Phylogeny and ecological radiation of New World thistles (Cirsium, Cardueae – Compositae) based on ITS and ETS rDNA sequence data

ETS primer design

Primers 18S-ETS (ACT TAC ACA TGC ATG GCT TAA) (Baldwin & Markos 1998) and ETS-Car-1 (TTC GTA TCG TTC GGT) were used for amplifying and sequencing the 3′ portion of the ETS examined in our study. Primer ETS-Car-1 was designed by sequencing upstream from the 18S gene into the IGS of Carduus nutans and Cirsium andrewsii and locating a region of conserved sequence in the two taxa, c. 660 bp from the ETS/18S gene border. The IGS of these taxa were amplified using primers 18S–IGS and 26S–IGS (Baldwin & Markos 1998). IGS polymerase chain reactions (PCR) comprised 25 µL of 50 mm Tris-HCL, 2.5 mm MgCl₂, 16 mm (NH4)₂SO₄, 150 µg/mL bovine serum albumin, 250 mm of each dNTP, 1 µm of each primer, 0.25 µL of Klentaq-LA by Clonetech and approximately 1–10 ng of template DNA. IGS PCR was conducted in thin-walled microfuge tubes using an MJ Research PTC-200 Peltier thermal cycler with the following parameters: initial denaturation at 94 °C for 3 min, followed by 30 cycles of 94 °C for 30 s and 68 °C for 7 min. Superior results were obtained by using the above machine rather than machines with slower temperature ramping rates.

ITS and ETS sequencing

Primers ITS4 (White et al. 1990) and ITS-I (Urbatsch et al. 2000) were used to amplify the ITS region (ITS-1, 5.8S gene and ITS-2) and primers 18S-ETS and ETS-Car-1 were used to amplify the 3′ end of the ETS via the PCR, following the protocol of Baldwin & Wessa (2000). For short-distance PCR of the ITS and the 3′ end of ETS, we used 25 µL reaction volumes including 1.25 µL glycerol, 50 mm KCl, 10 mm Tris-HCL, 2.5 mm MgCl₂, 250 µm of each dNTP, 0.5 µm of each primer, 0.5 units of AmpliTaq DNA Polymerase (Perkin-Elmer) and 1–10 ng of genomic DNA. PCR was conducted in thin-walled tubes in an MJ Research PTC-200 Peltier thermal cycler thermal cycler at 97 °C for 1 min (initial denaturation), followed by 40 cycles of 97 °C for 10 s, 48 °C (for ITS) or 55 °C (for ETS) for 30 s (annealing stage), and 72 °C for 1 min (extension phase), and concluding with 72 °C for 7 min (final extension stage).

Resulting products were cleaned using Millipore Ultrafree-MC tubes. Cycle sequencing reactions for the ITS PCR products were conducted using a Big Dye terminator cycle sequencing premix kit (Perkin-Elmer). Four µL of sequencing reagent premix, 0.4 µm of primer ITS-4 or ITS-5 (White et al. 1990) for ITS or primer 18S-ETS or ETS-Car-1 for ETS, 2 µL of DMSO and 20–40 ng of cleaned PCR products were added to thin-walled microfuge tubes. Cycle sequencing was performed in an MJ Research PTC-200 Peltier thermal cycler with the following settings: initial denaturation at 97 °C for 1 min, followed by 25 cycles of 96 °C for 10 s, 50 °C for 5 s and 60 °C for 4 min. Sequencing products were cleaned using isopropanol precipitation, then resolved on a 4.8% Long-Ranger polyacrylamide gel using an Applied Biosystems 377 automated sequencer (Perkin-Elmer).

Data analysis

Sequence data from the ITS and ETS were edited and easily aligned manually using Sequence Navigator, version 1.0.1. Gaps were coded as missing data. Approximately 1.4% of the matrix cells are represented as missing data. Copies of the aligned data set can be obtained from Treebase (SN1124). DNA sequences for all taxa in this study have been deposited in GenBank (see Table 1). Sequence data were compiled into a nexus file and analysed phylogenetically using maximum likelihood, as implemented in paup* test version 4.0d64 for Macintosh PPC (Swofford 2001). Phylogenetic analyses were conducted using the HKY85 model of sequence evolution and the heuristic algorithm of paup (with as-is stepwise addition and TBR branch swapping). Quartet-puzzling values were calculated to assess branch support in the resulting trees (Strimmer & von Haeseler 1996). All New World taxa were investigated for rate constancy of ETS and ITS sequence evolution using a likelihood ratio test for differences between clock-constrained and clock-unconstrained trees (Felsenstein 1988). The maximum likelihood analyses were conducted using the HKY85 model of sequence evolution, a gamma distribution of rate variation among sites (with the shape parameter set to four), and no assumed invariant sites. The difference in the likelihood values was examined for significance using a χ² test with the degrees of freedom equal to n-2 (n= number of terminal branches).

Average genetic divergence values were calculated for the endemic Californian Cirsium lineage, as well as for the Compositae lineages Baeriinae and Madiinae (Baldwin & Wessa 2000), Stephanomeria (Lee et al. 2002), Sidalcea (Andreasen & Baldwin 2001), Sanicula sect. Sanicoria (Vargas et al. 1999), the perennial lineage of Eriophyllum (Baldwin unpublished data) and Lessingia (Markos & Baldwin 2001). Clades originating in the California Floristic Province (CFP) were chosen using character optimization methods in MacClade (Maddison & Maddison 1999). When circumscription of CFP clades was equivocal, then a conservative circumscription was adopted by choosing the smaller of potential clades. Corrected pairwise distances were obtained using the HKY85 model. To estimate average divergence within each lineage, pairwise distance values were averaged for pairs of taxa spanning the ingroup root node. Taxa dispersed across a tree were chosen for pairwise comparisons. To compute an average divergence value, trees with highly asymmetrical levels of diversity on either side of the ingroup root-node required fewer pairwise comparisons than did trees with symmetrical levels of diversity across the root-node. Because, due to shared ancestry, pairwise comparisons of taxa are not independent, no standard measures of variance could be computed for the resulting mean values.

Ecological diversity within each of the Californian lineages examined was estimated by summing the number of plant communities in which members of a lineage naturally occur. Plant community definitions were based on those of Holland & Keil (1995) and were judged to represent a particular combination of environmental values, such as insolation, soil chemistry, temperature range, moisture availability and fire history. Occurrence of species within a particular community was based on our field observations, as well as statewide and local floras (Mason 1957; Munz 1973, 1974; Thomas 1961; Sharsmith 1982; McClintock et al. 1990; Hickman 1993; Twisselmann 1967; Best et al. 1996; Matthews 1997; Smith 1998).

ETS and ITS sequences in Cardueae

The ETS primers 18S-ETS and ETS-Car-1 are probably universal for the thistle tribe (Cardueae, Compositae) based on high yield of PCR products of uniform size (c. 650 bp) for taxa of Cardueae that span major lineages of the group. Absence of evident length variation in the 3′ region of the ETS (hereafter, ETS) of Cardueae contrasts with considerable length variation in the same region of another group of Compositae, Heliantheae sensu lato (Baldwin & Markos 1998). Use of the ETS for higher-level phylogenetic studies in Cardueae may be possible based on unambiguous alignment of the sequences across the disparate genera sampled. Length uniformity and ease of alignment of the ETS sequences may be indicative of an overall low level of variation in rDNA transcribed spacer sequences of the thistle tribe.

As in some other groups of angiosperms investigated to date [e.g. Sidalcea (Malvaceae), Andreasen & Baldwin 2001; Lasthenia (Compositae), Chan et al. 2001; Stephanomeria (Compositae), Lee et al. 2002)], the ETS and ITS-1 + ITS-2 appear to have evolved at similar rates in Cardueae. Average pairwise distances across the root node separating Onopordum acaulon from the other taxa are 0.10 for ETS and 0.15 for ITS-1 + ITS-2. Enhancement of resolution and support in rDNA trees by combined, rather than separate, analysis of ETS and ITS data accords with other studies indicating that the ETS often nicely complements phylogenetic data from the ITS region (e.g. Baldwin & Markos 1998; Clevinger & Panero 2000; Linder et al. 2000; Chan et al. 2001; Markos & Baldwin 2001; Lee et al. 2002).

Relationships among species

Use of rDNA transcribed spacers for resolving phylogeny within Cirsium is limited by low sequence variation among species but evidently not by the rate of concerted evolution. Conspecific populations of each of the three species represented by multiple samples (C. arvense, three samples; C. occidentale, two samples and C. vulgare, two samples) constitute lineages in the maximum-likelihood (ML) tree (Fig. 2). rDNA transcribed spacers in Cirsium appear to have undergone sufficiently rapid concerted evolution for use in phylogeographical study of at least one widespread species, C. arvense (Kelch, unpublished).

Strong support (quartet puzzling value = 86) for monophyly of Cirsium relative to the outgroup species sampled provides provisional evidence for the value of a pappus character for diagnosing the genus. Carduus is morphologically similar to Cirsium and differs most conspicuously from Cirsium by having nonplumose, rather than plumose, pappus bristles. Pappus evolution is known to have been highly dynamic in some groups of Compositae (e.g. in Heliantheae s.l., Baldwin et al. 2002), so reliance on a single pappus character for circumscribing genera of thistles might be regarded as questionable. Further sampling within Carduus is necessary to discern whether the genus and Cirsium are each truly monophyletic.

Putative monophyly of New World Cirsium (supported by a low quartet-puzzling value of 48; Fig. 2) leads us to suggest provisionally that Cirsium invaded North America only once before the advent of European peoples. If New World taxa of Cirsium indeed constitute a lineage, then indigenous North American members of the genus are each equally closely related to weedy thistles introduced into North America from the Old World. A single origin for New World Cirsium contradicts the traditional taxonomy, which treats New World species in multiple sections that each include Old World species (De Candolle 1837; Petrak 1917). For example, sect. Onotrophe includes the New World species C. crassicaule, C. edule and C. quercetorum, as well as the Old World species C. monspessulanum, C. palustre and C. spinosissimum. These taxa do not constitute a lineage in the ML tree (Fig. 2).

Within the New World thistles, one well-supported lineage (quartet puzzling score of 78) consists of species that are confined to the California Floristic Province (CFP; as delimited by Howell 1957; Raven & Axelrod 1978). This group contains rare endemics, including three of the four taxa of Cirsium that are listed as endangered by the US Fish and Wildlife Service. If a weakly supported lineage (quartet puzzling value of 16) consisting of the Californian endemic clade and a lineage of taxa from northwestern North America (some of which enter California in the Coast Ranges or Sierra Nevada) (Fig. 2) is upheld by future evidence, then all endemics of the CFP could be concluded to stem from the same western North American radiation (although probably from more than one ancestral source within the CFP).

The phylogeny presented here does not support the monotypic sections that Petrak (1917) erected for Californian taxa. The two monotypic sections sampled in our study, sect. Dermatolepis (C. fontinale) and sect. Mastigophyllum (C. rhothophilum), are both strongly nested within the lineage endemic to the CFP (Fig. 2). Petrak's taxonomy implies that C. fontinale and C. rhothophilum are isolated relicts; our results show that the two taxa are neoendemics, rather than palaeoendemics.

Ecological diversity in thistles of the California Floristic Province (CFP)

Like other regions on earth with a Mediterranean climate, the CFP has unusually high plant diversity and endemism for a temperate, continental area (Dallman 1998). Most plant groups restricted to the CFP are neoendemics that are believed to have evolved concurrently with or subsequent to the development of summer drought (Raven & Axelrod 1978; Axelrod 1992). Fossil and palaeoclimatic evidence indicate that the summer drying trend in western North America began at mid-Miocene (c. 15 Ma; e.g. Axelrod 1992; Flower & Kennett 1994).

Members of Cirsium that naturally occur within the CFP represent at least three separate lineages and cannot be reliably treated as representing a single example of radiation within the region (Fig. 2). Cirsium occidentale constitutes a highly variable lineage that occurs widely in western North America (including the CFP) and may be undergoing incipient diversification with zones of secondary introgression (Keil & Turner 1993; David Keil, pers. comm.). The Pacific Northwest lineage includes species that extend south into California along the coast (e.g. C. brevistylum) or along the higher mountain chains (e.g. C. andersonii). The California Endemic lineage includes various narrow endemics, most of which have a chromosome number of 2n = 32 (e.g. C. fontinale and C. rhothophilum; Ownbey et al. 1975). Only the California Endemic lineage was used in ecological comparisons. All taxa in the group are strongly associated with the CFP and therefore are likely to be products of a common radiation in the region that postdated the advent of the late Tertiary summer-drying trend (Axelrod 1992; Flower & Kennett 1994).

For purposes of comparison with Californian Cirsium, a conservative approach was used to identify other neoendemic angiosperm lineages of the CFP. For example, a clade corresponding to tribe Madieae (Compositae) sensu Baldwin comprises taxa (e.g. subtribe Baeriinae and subtribe Madiinae) that are mainly confined to the CFP (the Hawaiian silversword alliance being the primary exception) and others (e.g. Arnica) that are more widespread (Baldwin & Wessa 2000). Baeriinae, Madiinae and Arnica may represent a common CFP radiation or conceivably may represent separate colonizations of the CFP by different western North American ancestors. As with Cirsium, we chose the most cautious interpretation and treated Baeriinae and Madiinae as separate CFP radiations (Fig. 3). The number of species in each of these CFP endemic clades varied from seven in the California endemic clade of Cirsium to 89 in Madiinae, not including the Hawaiian taxa (Table 2). We chose mean genetic divergence to reflect divergence over time rather than taxon number, as the latter can be strongly skewed by artefacts of taxonomic judgement and extinction within lineages.

Ecological variation within the lineages compared in Fig. 3 is distributed mainly between, rather than within, taxa and appears to be largely the result of evolutionary shifts in ecology rather than phenotypic plasticity. In the California Endemic clade of Cirsium, taxa are narrowly restricted in habitat occurrence. In the other angiosperm groups, most taxa are confined to a limited number of plant communities, with notable exceptions (e.g. some perennial members of Eriophyllum and ruderal tarweeds).

Within the CFP, most angiosperm groups for which we have comparable rDNA and habitat data display a similar relationship between the level of rDNA divergence within a lineage and ecological breadth of the group (Fig. 3). The California Endemic lineage of Cirsium displays an extraordinarily high ecological diversity relative to the degree of rDNA divergence among taxa (Fig. 3). Unusually rapid ecological radiation or unusually low rates of ETS and ITS evolution could explain the results for Cirsium. Minimal chloroplast DNA divergence among the taxa of Cirsium in trnK intron/matK sequences (Kelch, unpublished) leads us to suspect that molecular divergence within the group is low in general and that rapid ecological radiation has occurred in the Californian thistles. Differences in rates of molecular evolution associated with differences in habit or life history (Gaut et al. 1996; Andreasen & Baldwin 2001) are unlikely to explain low rates of molecular evolution in the biennial or short-lived perennials in the California Endemic lineage of Cirsium; other angiosperm lineages included in our comparisons represent both annual and perennial groups.

High ecological diversity and low rDNA variation appear to extend throughout New World Cirsium and may reflect rapid ecological radiation on a continental scale. As noted above, other indigenous groups of Cardueae in North America are relatively depauperate and do not provide parallel examples of diversification in the New World. Columbines (Aquilegia, Ranunculaceae), on the other hand, constitute a particularly prominent and well-demonstrated example of extensive radiation of north temperate perennial herbs, with substantial diversity in North America (Hodges & Arnold 1994). In columbines, evolutionary radiation was mediated by changes in floral morphology and pollinator preference (Hodges 1997). In Cirsium, some morphological differentiation within and between species may be attributable to pollinator preference (e.g. flower colour and flower head length) but data on pollination and mating system are lacking. In general, differences in ecological setting or habitat among taxa of Cirsium are more pronounced than floral differences, as is evident in various examples of insular adaptive radiation in Compositae such as Hawaiian Madiinae (Baldwin 1997; Baldwin & Sanderson 1998), Macaronesian Sonchus (Kim et al. 1996; Kim et al. 1997) and Dendroseris (Sang et al. 1994; Esselman et al. 2000). In general, our data illustrates that Cirsium represents a continental parallel to insular composite lineages displaying low rDNA divergence and extensive ecological diversity.