The Californian Silene multinervia (Caryophyllaceae) and Eurasian members of section Conoimorpha in subgenus Behenantha are the only Silene species that have calyces with 15 or more prominent parallel, unbranched veins. We show that S. multinervia, which has been considered a recent introduction of the Asian S. coniflora (section Conoimorpha) to North America, is clearly not synonymous with the latter species based on morphological or molecular data. We present a chromosome count of S. multinervia (2n = 24), which is different from the base number x = 10, a putative synapomorphy for section Conoimorpha. Gene trees based on sequences from three different genomes fail to group S. multinervia with the European section Conoimorpha species. The S. multinervia sequences form a monophyletic group placed in an unresolved position within subgenus Behenantha.
Intercontinental disjunctions of plant species or species-pairs have received considerable interest from biogeographers (e.g. Raven 1972; Thorne 1972; Lee et al. 1996; Wen 1999; Milne 2006). Classical explanations often include vicariance or anthropogenic introduction. However, recent studies based on molecular data suggest that the most likely explanation for some Eurasia-North America disjunctions is pre-human dispersal events [e.g. Plantago ovata Forssk. (Meyers and Liston 2008), Oligomeris linifolia (Vahl) J. F. Macbr. (Martín-Bravo et al. 2009), and Senecio mohavensis A. Gray (Coleman et al. 2003)]. In other cases, species which previously have been regarded by some botanists as native to North America have been found to have been introduced by humans [e.g. Cakile edentula (Bigelow) Hook. (Raven and Axelrod 1978; Sauer 1988, p. 34) and Vulpia myuros (L.) C. C. Gmel. (Raven and Axelrod 1978)].
Silene L. (Caryophyllaceae) is a genus of approximately 700 species, most of which have their natural distribution in Eurasia (Oxelman et al. 2001). There are, however, also native Silene species in North and South America as well as in Africa, and species that have spread as weeds throughout the world. Silene is divided into the subgenera Silene and Behenantha (Otth) Endl. [syn. S. subgenus Behen (Dumort.) Rohrb.] (Popp and Oxelman 2004). Silene subgenus Silene includes the well-known species S. acaulis L. and S. gallica L., whereas major groups in Silene subgenus Behenantha include section Melandrium (Röhl.) R. K. Rabeler (containing the familiar S. latifolia Poir.), sections Physolychnis (Bentham) Bocquet and Conoimorpha Otth, the Silene vulgaris group, and S. noctiflora L. (with the closely related S. turkestanica Regel, Sloan et al. 2009). The Flora of North America lists 52 native and 18 introduced or naturalized North American Silene species (Morton 2005). Most of the North American species belong to subgenus Behenantha, either to the section Physolychnis s. l. (Popp et al. 2005; Popp and Oxelman 2007) or to the S. menziesii group (Popp and Oxelman 2007), while S. antirrhina L. and S. repens Patrin belong to Silene subgenus Silene (Eggens et al. 2007; Popp and Oxelman 2007).
Considerable attention has been given to the phylogenetic position of the section Melandrium to facilitate understanding of the evolution of dioecy (e.g. Atanassov et al. 2001; Filatov and Charlesworth 2002; Filatov 2005; Nicolas et al. 2005; Rautenberg et al. 2010), and in several cases section Conoimorpha has been suggested to be the sister group to these dioecious species (e.g. Desfeux and Lejeune 1996; Erixon and Oxelman 2008a).
Common morphological features for Silene, as circumscribed by Oxelman et al. (2001), are flowers with 10 stamens and three or five styles, five free petals, a synsepalous calyx, and a capsule that usually splits open into twice as many teeth as the number of styles. Two important characters in identification of Silene species are anthophore length and the coronal scales. The anthophore is a structure that separates the attachment of the calyx and corolla. The coronal scales are present as small appendages on the border between the petal limb and the petal claw (the part of the petal that is hidden in the calyx).
Silene multinervia S.Watson (Caryophyllaceae) is a Californian taxon (Hitchcock and Maguire 1947) that always has been placed into the otherwise Eurasian section Conoimorpha (e.g. Watson 1890; Hitchcock and Maguire 1947; Šourková 1971, as the separate genus Pleconax Raf.). Silene multinervia and section Conoimorpha share a distinctive morphological feature: all species have several (15–60) unbranched prominent parallel veins on the calyx (all other Silene species have 10 principal veins and/or branching nervature). Silene section Conoimorpha also has a base chromosome number of x = 10 (Greuter 1995), whereas all other Silene have x = 12 with one known exception (S. fortunei Vis., 2n = 30; Bari 1973). Members of section Conoimorpha have elevated nucleotide substitution rates in chloroplast (Erixon and Oxelman 2008b) and mitochondrial DNA (Sloan et al. 2009), compared to other members of the genus. The circumscription of the group (e.g., Rohrbach 1868; Chowdhuri 1957) has been uncontroversial since its first appearance in the taxonomic literature (Otth 1824). The species currently recognized in the group (Silene ammophila Boiss. & Heldr., S. conica L., S. coniflora Nees ex Otth, S. conoidea L., S. lydia Boiss., S. macrodonta Boiss., S. subconica Friv., S. grisebachi (Davidov) B. Pirker & Greuter, and S. sartorii Boiss. & Heldr.; Pirker and Greuter 1997) have their native distribution in Europe and southwest to central Asia (Table 1), although S. conica and S. conoidea are introduced as weeds around the world (e.g. Rozefelds et al. 1999; Morton 2005; Global Compendium of Weeds 2007). Silene multinervia has recently been put into synonymy with the southwest/central Asian species S. coniflora (Morton 2005; followed by Hartman and Rabeler 2008). On the other hand, Popp and Oxelman (2007) and Rautenberg et al. (2010) showed, based on cpDNA and nrDNA ITS sequence data, that S. multinervia does not form a monophyletic group with Eurasian samples from the section Conoimorpha. However, the sampling in either of these two studies was not focused on S. multinervia or section Conoimorpha.
TABLE 1.
Native distribution and number of calyx veins of Silene multinervia and the members of Silene section Conoimorpha. Silene grisebachii and S. sartorii were not included in the molecular analyses.
Using DNA sequences from samples of S. multinervia, S. coniflora, and six other taxa from Silene section Conoimorpha, a chromosome count of S. multinervia, and sequence data from several outgroup species with emphasis on potentially closely related species in Silene subgenus Behenantha, we address the following questions: Is there any morphological or molecular support for the synonymization of S. multinervia with S. coniflora? Is there any morphological or molecular support for the inclusion of S. multinervia in Silene section Conoimorpha? Does S. multinervia represent a recent introduction to the Californian flora? What is the phylogenetic position of section Conoimorpha?
MATERIALS AND METHODS
Study Species—The present study includes Silene multinervia and seven of the nine species from Silene section Conoimorpha (Table 1), as well as a large outgroup sampling, with special emphasis on potentially closely related species in Silene subgenus Behenantha.
The members of section Conoimorpha are briefly characterized in Table 1, but a few of them deserve mention here. Silene multinervia grows in California and Mexico on burnt open ground, after forest fires, and is recognized by 20 calyx veins and no coronal scales (Watson 1890; Jepson 1914; Hartman and Rabeler 2008). Silene coniflora grows from southwest to central Asia and has 20 calyx veins and oblong coronal scales (Schischkin 1970). Silene lydia is a species sharing some of the of features of section Conoimorpha (more than 10 unbranched parallel veins), but also having enough features to be placed in a section of its own (S. section Lydiae Greuter) by Greuter (1995). Silene lydia has a chromosome number of 2n = 20, or possibly 2n = 11 (preliminary data by B. Pirker, discussed in Greuter 1995), long eglandular hairs on the calyx, and no anthophore (Greuter 1995). It is distributed in the southeastern Balkans and western Anatolia (Greuter 1995). The Greek endemics S. griesebachi and S. sartorii were not included in the molecular analysis. They are similar to S. subconica, but differ in petal shape and venation, and the former has distinct seeds and longer anthophore (Pirker and Greuter 1997). Rautenberg et al. (2008) found some indications of a close relationship between S. noctiflora and section Conoimorpha. Previous molecular phylogenetic studies (e.g. Oxelman and Lidén 1995; Oxelman et al. 2001; Popp and Oxelman 2001, 2004, 2007; Rautenberg et al. 2010) have revealed that section Conoimorpha is confidently embedded in subgenus Behenantha, which has poorly resolved basal relationships, possibly due to a rapid radiation some six to seven million years ago (Erixon and Oxelman 2008a; Frajman et al. 2009). We therefore sampled outgroup taxa primarily to represent major lineages from subgenus Behenantha.
Chromosome Count—A chromosome count was determined for S. multinervia based on a plant grown from seeds collected in Napa County, California (Appendix 1). Prior to fixation in Carnoy I solution (3 volumes absolute alcohol and 1 volume glacial acetic acid), growing roots were pretreated with equal parts 0.1% colchicine and 0.002M 8-hydroxyquinoline for 2 hrs. After fixation and hydrolysis in 1N HCl at 60°C for 2 mins, root-tip meristems were prepared. Flower buds were fixed and hydrolyzed in Carnoy I solution. All tissues were stained with aceto-orcein on clean slides and squashed under a coverslip.
Morphology—Herbarium specimens from CAS, G, GB, LE, MW, S, UPS, and WU (abbreviations according to Holmgren and Holmgren 1998), and Arne Strid's private herbarium (in Ørbaek, Denmark), of S. multinervia, S. coniflora, and other representatives of Silene section Conoimorpha were studied as physical specimens or as images deposited in the Sileneae database ( http://www.sileneae.info). Specimens were compared to keys, descriptions, and illustrations in the literature (Otth 1824; Boissier 1867; Rohrbach 1868; Watson 1890; Williams 1896; Jepson 1914; Post 1932; Hitchcock and Maguire 1947; Blakelock 1957; Khoshoo and Bhatia 1963; Mouterde 1966; Zohary 1966; Bajtenov 1969; Schischkin 1970; Ghazanfar and Nasir 1986; Melzheimer 1988; El-Oqlah and Karim 1990; Hosny et al. 1992; Chater et al. 1993; Greuter et al. 1997; Boulos 1999; Morton 2005; Hartman and Rabeler 2008; Calflora 2009).
DNA Extraction, Amplification, and Sequencing—DNA was extracted from living or herbarium material using a modified Carlson/Yoon method (Oxelman and Lidén 1995). Voucher details and GenBank accession numbers are listed in Appendix 1. Three cpDNA regions (the matK gene, the rps16 intron, and the trnL gene and trnL-trnF intergenic spacer), three mitochondrial DNA (mtDNA) regions [the protein-encoding ATP synthase subunit 1 (atp1), cytochrome c oxidase subunit 3 (cox3), and NADH dehydrogenase subunit 9 (nad9)], ITS from nuclear ribosomal DNA, and four low-copy nuclear regions (parts of the RNA polymerase genes RPA2, RPB2, RPD2a, and RPD2b) were amplified. The PCR products were either purified using MilliPore multiscreen PCR plates in a vacuum manifold (Millipore, Billerica, Massachusetts) and sequenced by Macrogen Inc. in Seoul, South Korea or purified with Exonuclease I and shrimp alkaline phosphatase (USB Corporation, Cleveland, Ohio), cycle sequenced with BigDye v3.1 (Applied Biosystems, Foster City, California), and analyzed on an ABI 3130xl capillary sequencer. In addition to already published PCR and sequencing primers for matK (Fior et al. 2006; Mower et al. 2007; Sloan et al. 2009; Rautenberg et al. 2010), rps16 (Oxelman et al. 1997), tmLF (Oxelman et al. 2005), RPA2 (Popp and Oxelman 2004), RPB2 (Popp and Oxelman 2001), RPD2 (Popp and Oxelman 2004), cox3 (Duminil et al. 2002), nad9 (Duminil et al. 2002), and ITS (Popp and Oxelman 2001), the following primers were used for amplification and sequencing: atp1_Conoi_F (GCKGGAGAAATGGYKGAATTTG), atpl_Conoi_F2 (ATGCAAACYGGCTTAAAGGC), atp1_Conoi_F3 (ATTCTTGTAGCAGCCACTGC), atp1_Conoi_R (CCWACATTAATAGCWGGTCTA) atp1_Conoi_R2 (TCCAATCGCTACATAAACAC), atp1l_Conoi_R3 (CSGCTCTTTCTAAGAGACG), cox3_Conoi_F (GAATAACCAAACTACGTCCAC), cox3_Conoi_R (GGBGGTGAAATMCTGCTCAG), nad9_Conoi_F (ACCACNCGTTTTTCTGGATC), nad9_Conoi_R (CAAGAARTGGGTCAAAAGAATG). Eighty sequences were new to this study, and additional sequences were obtained from GenBank Appendix 1).
Sequence Alignment and Analysis—Sequence reads were assembled into contigs and edited using the Staden package version 1.6.0 for Mac OSX (Staden 1996) with phred version 0.020425.C (Ewing and Green 1998; Ewing et al. 1998) and phrap version 0.990319 ( http://www.phrap.org) or using Sequencher v4.5 (Gene Codes, Ann Arbor, Michigan). Base polymorphisms were coded using the NC-IUPAC ambiguity codes. Sequence alignment was performed manually in QuickAlign (Müller and Müller 2003), following the criteria of Popp and Oxelman (2004). The alignments of the three cpDNA regions were analyzed separately and checked for strongly supported conflicts (see definition below). As such conflicts were not found, the alignments were concatenated into a cpDNA data set. The mtDNA regions were analyzed both separately and concatenated into a single data set. The nuclear regions were analyzed separately. Simple indel coding (Simmons and Ochoterena 2000) was applied to the alignments using SeqState version 1.36, build 19.10.2007 (Müller 2005) for use in the PAUP* and MrBayes analyses.
Maximum parsimony analyses and maximum parsimony bootstrap support measures were performed with PAUP* v.4.0b10 (Swofford 2002). Maximum parsimony analyses were carried out using heuristic searches with TBR branch swapping, the multrees option on (but a limit of maxtrees set to 5,000), and 10 random addition sequences. For bootstrap support, 1,000 replicates were performed, with the multrees option off.
Bayesian phylogenetic analysis was performed using MrBayes 3.2 (Huelsenbeck and Ronquist 2001; Ronquist and Huelsenbeck 2003) with nucleotide models as proposed by MrModeltest version 2.2 (Nylander 2004), using the Akaike information criterion. Four MCMC chains were run for five million generations with trees and parameter values saved every 1,000th generation, in two parallel runs. Convergence of MrBayes analyses was checked using the split frequency diagnostic (runs with average standard deviations of < 0.01 were considered as converged), Tracer v1.5 (Rambaut and Drummond 2007), and AWTY (Wilgenbusch et al. 2004; Nylander et al. 2008). The first 25% of the trees were discarded as burn-in.
The *BEAST (starbeast) mode in BEAST v1.5.4 (Drummond and Rambaut 2007) infers gene trees and, at the same time, estimates a species tree that is compatible with the gene trees given a coalescent process. *BEAST was used to infer a species tree for the genera Lychnis and Silene based on the combined information from the cpDNA, RPA2, RPD2a, and RPD2b regions. Because of the strong incongruence between the gene trees from RPB2 and the other genes regarding the positions of Lychnis and the two subgenera in Silene, RPB2 was excluded from the *BEAST analysis. Input files for BEAST were created with BEAUti v1.5.4 (Drummond and Rambaut 2007) and with additional manual editing of the xml file, using a relaxed clock model (Drummond et al. 2006), with branch rates following a lognormal distribution, and the same substitution models as in the MrBayes analysis. We used a Yule prior for the species tree. Differences in effective population size will influence the coalescence times. The Ne of chloroplasts and mitochondria are generally considered to be ¼ of the Ne of the nuclear genes in a dioecious plant, assuming uniparental inheritance and an equal sex ratio [but see Lynch et al. (2006)]. In hermaphroditic plants however, the chloroplast Ne is ½ that of the nuclear genes. Therefore, the ploidy level of the cpDNA partition was adjusted manually in the xml file to accommodate this twofold difference. A prior on the age of the root of the species tree was set to 12.39 million years, with a normally distributed standard deviation of 2.1, based on the posterior age of the node containing Silene and Lychnis in a fossil calibrated matK tree of Caryophyllaceae (Frajman et al. 2009). One MCMC chain was run for 100 million generations with trees and parameter values saved every 1,000th generation. The tree files were summarized using TreeAnnotator v1.5.4 (Drummond and Rambaut 2007) into one maximum credibility tree with median node heights (discarding the first 10% of the trees as burn-in). To assess the effect of the priors on the posteriors, the run was compared to a run performed with the same settings but on an empty alignment.
TABLE 2.
Statistics for the data sets used in the maximum parsimony (PAUP*), MrBayes, and *BEAST analyses. The *BEAST data set included 38–54 sequences from four of the regions, representing 39 species from Silene and Lychnis, and had 22.9% missing data. The mitochondrial and chloroplast sequences were analyzed both separately and concatenated into one mtDNA data set (atp1, cox3, and nad9) and one cpDNA data set (matK, rps16, and trnLF). In order to make each species represented in each region in the *BEAST analysis, empty sequences were added to some data sets.
The resulting trees were visualized using FigTree version 1.2 (Rambaut and Drummond 2008). Posterior probabilities (PP) ≥ 0.95/bootstrap values (BS) ≥ 85% were considered as strong support, while values of 0.85–0.94 PP/75–84% BS were considered as moderate support, and values of 0.70–0.84 PP/50–74% BS as low support. We define incongruence as the presence of strongly supported conflicts between tree topologies. Data matrices and phylogenetic trees are available on TreeBASE (study number S11178).
RESULTS
Statistics for the alignments and phylogenetic analyses, as well as the model of evolution proposed by MrModeltest for the DNA regions are presented in Table 2.
Chloroplast Genes—In the concatenated chloroplast data set, Silene section Conoimorpha is a well-supported monophyletic group containing all species reported to belong to the section except S. multinervia (Fig. 1a). All S. multinervia accessions form a monophyletic group placed in an unresolved position in subgenus Behenantha, outside the rest of section Conoimorpha. Silene lydia is placed as sister to the rest of section Conoimorpha. All species relationships within section Conoimorpha are strongly supported. The pattern is congruent between all included cpDNA regions (data not shown), and between phylogenetic methods (Fig. 1a).
Mitochondrial DNA—As in the cpDNA tree, the European and Asian members of section Conoimorpha form a strongly supported monophyletic group in the mtDNA tree (Fig. 1b). In the concatenated mtDNA data set, section Conoimorpha groups with S. noctiflora + S. turkestanica with strong support and S. multinervia is weakly to moderately supported as sister to this clade (Fig. 1b). The branches are extremely long in section Conoimorpha, as well as in S. noctiflora + S. turkestanica (Fig. 1b). The different mtDNA gene trees show different patterns in terms of branch length variation (supplemental data S1). Silene multinervia occupies a branch that is some-what longer than the majority of other Silene branches, but still much shorter than the extreme lineages (Fig. 1b). The position of S. multinervia is more or less ambiguously resolved in all three mtDNA gene trees (supplemental data S1). The internal relationships within Eurasian Conoimorpha are strongly supported and agree with the cpDNA tree (Fig. 1b).
FIG. 1.
Phylogram obtained from MrBayes for the concatenated chloroplast DNA (cpDNA) sequences (a), and for the concatenated mitochondrial DNA (mtDNA) sequences (b). Values associated with nodes are Bayesian posterior probabilities/parsimony bootstrap values. Posterior probabilities/bootstrap values lower than 0.70/50 are not indicated. Branch lengths represent estimated number of changes per site. The gray boxes indicate the key groups C = section Conoimorpha, M = S. multinervia, and N = S. noctiflora + S. turkestanica. Numbers and letters after species names indicate different specimens (Appendix 1). Genera are represented as follows: Ag = Agrostemma, At = Atocion, B = Beta, E = Eudianthe, H = Heliosperma, L = Lychnis, P = Petrocoptis, S = Silene, and V = Viscaria.
FIG. 2.
Phylogram obtained from MrBayes for the low-copy nuclear RNA polymerase genes RPA2 (a), RPB2 (b), RPD2a (c), and RPD2b (d). Values associated with nodes are Bayesian posterior probabilities/parsimony bootstrap values. Posterior probabilities/bootstrap values lower than 0.70/50 are not indicated. Branch lengths represent estimated number of changes per site. Dashed lines represent parts of the tree where the maximum parsimony bootstrap consensus tree has a differing topology (bootstrap support lower than 60%). The gray boxes indicate the key groups C = section Conoimorpha, M = S. multinervia, and N = S. noctiflora + S. turkestanica. Numbers and letters after species names indicate different specimens (Appendix 1). GenBank accession numbers are used to identify different sequences from the same specimen. All species belong to Silene except At = Atocion, E = Eudianthe, L = Lychnis, and P = Petrocoptis.
Nuclear Genes—In all nuclear gene trees, the members of section Conoimorpha, with the exception of S. multinervia, form a strongly supported monophyletic group (Figs. 2–3). The relationships within Conoimorpha are generally well resolved, strongly supported, and congruent with other regions (Figs. 2–3). Generally, the topological relationships in Behenantha outside of section Conoimorpha are unresolved, or conflicting between different nuclear genes (Figs. 2–3). In the ITS tree the relationships between section Conoimorpha, S. multinervia, and S. noctiflora + S. turkestanica are unresolved (Fig. 3). In the RPD2a and RPD2b trees, S. multinervia is placed as a close relative of the Physolychnis group with moderate (RPD2a) or strong (RPD2b) support, while S. noctiflora + S. turkestanica form a moderately to strongly supported sister group to the members of section Conoimorpha (Fig. 2c–d). In RPB2, RPD2a, and RPD2b, S. multinervia and the rest of section Conoimorpha are separated by at least one moderately to strongly supported node (Fig. 2b–d).
*BEAST Analysis—In the species tree obtained by the *BEAST analysis based on cpDNA and data from the RNA polymerase genes RPA2, RPD2a, and RPD2b, the topological relationships between section Conoimorpha, S. multinervia, and S. noctiflora + S. turkestanica are poorly resolved (Fig. 4). There is no support for S. multinervia as the sister group to section Conoimorpha. In the RPD2a and RPD2b gene trees, S. noctiflora + S. turkestanica form a monophyletic group with section Conoimorpha (PP = 0.97 and 0.78, respectively; supplemental data Fig. S2), but in the species tree the PP for this grouping is 0.63 (Fig. 4).
Dating—The 95% HPD (highest posterior density) ages of the MRCA (most recent common ancestor) of the S. multinervia sequences vary in the different gene trees, between 0.0021 (RPD2a) and 0.64 million years (RPA2). In the combined species tree in the *BEAST analysis, the 95% HPD ages of the MRCA of section Conoimorpha are 1.6–5.7 million years (Fig. 4). The age of the MRCA of S. multinervia and its closest sister group (section Physolychnis) has a 95% HPD interval of 1.9–7.1 million years in the combined species tree, although this node has a posterior probability of only 0.60 (Fig. 4).
Chromosome Count—Twenty-four chromosomes could readily be counted from several metaphase plates prepared from root-tips of Silene multinervia, and also from mitotic metaphase plates prepared from flower buds.
Morphology—There are several phenotypic differences between the allegedly synonymous S. multinervia and S. coniflora: S. multinervia lacks coronal scales and has basal leaves that are oblanceolate and cauline leaves that are lanceolate-linear (Fig. 5A). Silene coniflora has coronal scales and grass-like linear leaves (Fig. 5B). The number of calyx veins is 20 in both S. multinervia and S. coniflora. Although the protologue by Otth, citing the original author Nees, states the number of calyx veins to be 30 (Otth 1824), the examined S. coniflora specimens have 20 calyx veins, a number that is also supported by previously published reports (Boissier 1867; Rohrbach 1868; Williams 1896; Post 1932; Blakelock 1957; Zohary 1966; Bajtenov 1969; Schischkin 1970; Hosny et al. 1992; Boulos 1999). Ghazanfar and Nasir (1986) and Melzheimer (1988) give a number of 15–20 calyx veins for S. coniflora.
FIG. 3.
Phylogram obtained from MrBayes for the ITS sequences of the nuclear ribosomal DNA. Values associated with nodes are Bayesian posterior probabilities/parsimony bootstrap values. Posterior probabilities/bootstrap values lower than 0.70/50 are not indicated. Branch lengths represent estimated number of changes per site. Dashed lines represent parts of the tree where the maximum parsimony bootstrap consensus tree has a differing topology (bootstrap support lower than 60%). The gray boxes indicate the key groups C = section Conoimorpha, M = S. multinervia, and N = S. noctiflora + S. turkestanica. Numbers and letters after species names indicate different specimens (Appendix 1). All species belong to Silene except At = Atocion, E = Eudianthe, L = Lychnis, and P = Petrocoptis.
FIG. 4.
Species tree obtained from *BEAST for Silene and Lychnis, based on cpDNA, RPA2, RPD2a, and RPD2b sequences. Values associated with nodes are Bayesian posterior probabilities (above branches) and median node ages in million years (under branches). The horizontal bars represent 95% HPD (highest posterior density) intervals of node ages. The gray boxes indicate the key groups C = section Conoimorpha, M = S. multinervia, and N = S. noctiflora + S. turkestanica. All species belong to Silene except L = Lychnis.
DISCUSSION
Is There any Morphological or Molecular Support for the Synonymization of S. multinervia to S. coniflora?—Silene coniflora is the representative of section Conoimorpha that most resembles the superficial appearance of S. multinervia, with the similarity mainly based on the number of calyx veins. Careful study of plant material, however, reveals that the North American and southwest/central Asian species are two distinct entities that easily can be distinguished morpho-logically based on leaf morphology and presence/absence of coronal scales. None of the gene phylogenies show any support for the synonymy of S. multinervia and S. coniflora.
Is There any Morphological Support for the Inclusion of S. multinervia in Silene Section Conoimorpha?—The common characteristic nervature of Silene section Conoimorpha and S. multinervia, with 15 or more densely packed, prominent parallel calyx veins, is not present in any other members of the genus. Other Silene species have calyces with 10 veins, or with a different distribution of the veins. Among the close relatives in Silene subgenus Behenantha, the dioecious members of section Melandrium have female flowers with 20 branching veins and male flowers with 10 veins, S. vulgaris and its close relatives have an anastomosing pattern on the calyx, whereas the members of section Physolychnis have 10 veins. We have not found any synapomorphies for S. multinervia and section Conoimorpha other than the nervature. Our chromosome count of S. multinervia (2n = 24) is the same as for most other diploid Silene species, but differs from what Morton (2005) reports for Asian S. coniflora material (2n = 20). We have, however, not been able to find any original chromosome counts of S. coniflora in Bari (1973), the IPCN database (Goldblatt and Johnson 1979), the S. coniflora literature listed in Material and Methods, or other literature on chromosome counts in Silene. Several reports show that other members of section Conoimorpha have 2n = 20 (e.g. Khoshoo 1960; Khoshoo and Bhatia 1963; Greuter 1995), or possibly 2n = 22 in S. lydia (Greuter 1995). Thus, cytological evidence do not support the inclusion of S. multinervia in section Conoimorpha, whereas the presence of many densely packed calyx veins is a potential synapomorphy.
FIG. 5.
Flowers of A. Silene multinervia (California, U. S. A., photo: Barry Breckling). B. S. coniflora (Israel, photo: Sara Gold, used with kind permission from http://www.wildflowers.co.il). C. S. conoidea (cultivated, Sweden, photo: Anja Rautenberg). D. S. conica (photo: Mikael Thollesson).
Is There any Molecular Support for the Inclusion of S. multinervia in Silene Section Conoimorpha?—The present study, based on a more thorough sampling of specimens and taxa, supports previous studies indicating that S. multinervia does not form a monophyletic group with the Eurasian species of section Conoimorpha (Popp and Oxelman 2007; Rautenberg et al. 2010). The relationships between the different groups from Silene subgenus Behenantha are largely unresolved, and hence it is difficult to pinpoint the phylogenetic position of S. multinervia. Although the *BEAST species tree and the gene phylogenies of RPA2, RPB2, RPD2a, and RPD2b are somewhat incongruent regarding the relationships within subgenus Behenantha, they all indicate that S. multinervia is not the closest relative of section Conoimorpha. Other molecular studies also have had problems resolving the positions of several groups in subgenus Behenantha (e.g. Popp and Oxelman 2007; Rautenberg et al. 2010), and Erixon and Oxelman (2008a) suggested that an ancient radiation is responsible for the pattern seen in the cpDNA data.
If S. multinervia and S. section Conoimorpha are sister lineages, the non-monophyly of the groups could potentially be explained by incomplete lineage sorting effects, which would be reasonable if the branching events leading to the radiation of subgenus Behenantha were separated by short time spans and/or large effective population sizes. If S. multinervia and section Conoimorpha are not each other's closest relatives, the apparent morphological synapomorphy (many densely packed unbranched calyx veins) could be caused by convergent evolution or by a deep coalescent event of the gene(s) responsible for this feature. The chronograms indicate that the split between S. multinervia and section Conoimorpha lineages must be several million years old, so even if S. multinervia and section Conoimorpha are sister groups, the hypothesis of human-mediated dispersal of S. multinervia from Eurasia to America can be safely rejected.
The species from section Conoimorpha included in our species tree analyses (S. ammophila, S. conica, S. coniflora, S. conoidea, S. lydia, S. macrodonta, and S. subconica) form a strongly supported monophyletic group. Silene grisebachii and S. sartorii could unfortunately not be sampled for the present study, but given their great morphological, ecological, and geographical resemblance (Pirker and Greuter 1997) to the rest of the species, it is sound to hypothesize that they also belong to the section Conoimorpha clade.
Silene Section Conoimorpha and S. noctiflora—In accordance with Sloan et al. (2009), the members of Silene section Conoimorpha and the monophyletic group S. noctiflora + S. turkestanica both have extremely high substitution rates in the mitochondrial genes atp1, cox3, and nad9. Silene multinervia has slightly elevated rates, as compared to the rest of the genus. Due to the extreme variations in substitution rates between the sampled taxa, it is difficult to use the mtDNA phylogeny to draw conclusions on the relationships between different lineages. In the RPD2a and RPD2b phylogenies, S. noctiflora + S. turkestanica form a monophyletic group with section Conoimorpha. This topology is partly supported by a recent study, where the 3′ part of the SlX1/SlY1 gene indicates monophyly of S. noctiflora and section Conoimorpha, although with low support (Rautenberg et al. 2008). If this sister group relationship reflects the species phylogeny, it would support a single origin of the elevated substitution rates in section Conoimorpha and S. noctiflora + S. turkestanica. However, the incongruence of the tree topologies inferred from other nuclear and chloroplast genes (Figs. 1a, 2a–b, 3′-′4; and the 5′ part of SlX1/SlY1 gene in Rautenberg et al. 2008) makes this relationship remain ambiguous.
Congruence Between Organellar Phylogenies—Recent studies in Silene vulgaris have found evidence of paternal transmission and recombination in organelle genomes, resulting in incongruence between cpDNA and mtDNA gene trees within the species (McCauley et al. 2005; Houliston and Olson 2006; McCauley et al. 2007; McCauley and Ellis 2008). These results raise the possibility of phylogenetic conflicts between cpDNA and mtDNA at the interspecific level. Plant mtDNA sequences are often uninformative at local phylogenetic scales, because substitution rates in plant mtDNA are generally low compared to those in plant chloroplast and nuclear genomes and compared to mtDNA of other organisms (Wolfe et al. 1987; Palmer and Herbon 1988). In our dataset, the extreme differences in branch lengths in the Silene mtDNA phylogenies preclude using mtDNA to infer relationships among the major lineages. On the other hand, the rate acceleration provides the rare opportunity to use plant mtDNA to resolve the relationships at a local phylogenetic scale within section Conoimorpha. Within section Conoimorpha, we found that the different mitochondrial regions are congruent with each other, with the cpDNA regions, and with the nuclear regions, except for a few weakly supported deviations. Therefore, if paternal leakage and recombination have occurred within section Conoimorpha, they do not appear to have generated significant phylogenetic conflicts between chloroplast and mitochondrial genomes.
ACKNOWLEDGMENTS.
We would like to thank Nahid Heidari for excellent assistance in the lab. Sunniva Aagaard and Per Erixon are thanked for fruitful discussions. Andreas Wallberg and Mats Töpel helped with computer cluster related questions. Magnus Lidén, Freek Bakker, journal editor Lynn Bohs, and two anonymous reviewers gave helpful comments on earlier versions of the manuscript. Božo Frajman, Erik Ljungstrand, Jake Ruygt, Peter Schönswetter, Douglas Stone, Arne Strid, and curators of the herbaria CAS, GB, LE, MW, S, UPS, and WU helped with material. Sara Gold ( http://www.wildflowers.co.il), Barry Breckling, and Mikael Thollesson kindly provided photographs of live specimens. The study was supported by grants from The Swedish Research Council (VR) and The Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS) (to BO), The Royal Physiographic Society in Lund, Helge Ax:son Johnssons stiftelse, The Royal Swedish Academy of Sciences, Liljewalchs resestipendium, Linné-stipendiestiftelsen, and Wallenbergstiftelsen (to AR), and NSF DEB-0808452 (to DBS). Part of this work was carried out using the resources of the Computational Biology Service Unit from Cornell University, which is partially funded by Microsoft Corporation. Authors' Contributions: AR was responsible for most of the extractions, sequencing, and assembly of the nuclear genes and some cpDNA genes. She also made the alignments, the analyses, and drafted the manuscript. DBS was responsible for sequencing and assembly of mitochondrial genes and matK, and helped to write the manuscript. VA was responsible for the chromosome count and some of the sequencing. BO conceived of the study, participated in its design and data analysis, and helped to write the manuscript.
LITERATURE CITED
Appendices
APPENDIX 1.
Voucher details (collector, number and herbarium), origin (for section Conoimorpha specimens), and GenBank accession numbers for the specimens analyzed in the present study. Sequences HQ334894–HQ334976 were produced for this study. Herbarium abbreviations are according to Holmgren and Holmgren (1998), except Strid = Arne Strid's private herbarium, Ørbaek, Denmark. Superscripts are used as index letters to match voucher specimens to sequences in the gene trees (Figs. 1–4) in those species where more than one specimen was used to produce sequences.
