HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Popp, M. Articles by Oxelman, B. Search for Related Content PubMed Articles by Popp, M. Articles by Oxelman, B. Agricola Articles by Popp, M. Articles by Oxelman, B.

Origin and evolution of North American polyploid Silene (Caryophyllaceae)¹

Uppsala University, Evolutionary Biology Centre, Department of Evolution, Genomics and Systematics, Norbyvägen 18D, SE-752 36 Uppsala, Sweden; Natural History Museum, University of Oslo, National Centre for Biosystematics, P.O. Box 1172 Blindern, NO-0318 Oslo, Norway

Received for publication April 25, 2006. Accepted for publication December 30, 2006.

ABSTRACT

Nuclear DNA sequences from introns of the low-copy nuclear gene family encoding the second largest subunit of RNA polymerases and the ribosomal internal transcribed spacer (ITS) regions, combined with the psbE-petL spacer and the rps16 intron from the chloroplast genome were used to infer origins and phylogenetic relationships of North American polyploid Silene species and their closest relatives. Although the vast majority of North American Silene species are polyploid, which contrasts to the diploid condition dominating in other parts of the world, the phylogenetic analyses rejected a single origin of the North American polyploids. One lineage consists of tetraploid Silene menziesii and its diploid allies. A second lineage, Physolychnis s.l., consists of Arctic, European, Asian, and South American taxa in addition to the majority of the North American polyploids. The hexaploid S. hookeri is derived from an allopolyploidization between these two lineages. The tetraploid S. nivea does not belong to any of these lineages, but is closely related to the European diploid S. baccifera. The poor resolution within Physolychnis s.l. may be attributed to rapid radiation, recombination among homoeologues, homoplasy, or any combination of these factors. No extant diploid donors could be identified in Physolychnis s.l.

Key Words: Caryophyllaceae • cpDNA • low-copy nuclear DNA • molecular phylogeny • North America • rapid radiation • RNA polymerase • Silene

At least one polyploidization, i.e., duplication of the entire nuclear genome, is estimated to have occurred in 30–80% of the extant plant lineages (Grant, 1981 ; Masterson, 1994 ; Wendel, 2000 ). It has also been inferred in many insects, fish, and amphibians (Otto and Whitton, 2000 ). Many economically important domesticated plants such as rice, wheat, potatoes, maize, soybean, and cotton are polyploids (Hilu, 1993 ). In addition, several plant lineages traditionally considered as diploids, for example, Arabidopsis thaliana and Zea mays, have turned out to bear evidence in their genomes of ancient polyploidy events (Arabidopsis Genome Initiative, 2000 ; Gaut and Doebley, 1997 ). Two extremes of polyploidy can be distinguished: genome duplication within an individual is referred to as an autopolyploidization, whereas a duplication of two or more divergent genomes within a hybrid is an allopolyploidization. It is common that polyploidization is followed by massive genome rearrangements including gene loss and silencing and/or subfunctionalization of the homoeologues (i.e., DNA regions duplicated by polyploidization) (Adams and Wendel, 2005 ). Thus, it is hard to overemphasize the importance of polyploidization in plant evolution. It is important to study not only the common model species and synthetic polyploids, but also natural auto- and allopolyploid plant lineages. By using several unlinked biparentally inherited DNA regions, we can test the mode of origin (i.e., auto- or allopolyploid) and hypotheses of the closest relatives of polyploids in a phylogenetic perspective (Popp and Oxelman, 2001 ; Tank and Sang, 2001 ; Cronn et al., 2002b ; Smedmark et al., 2003 ; Popp et al., 2005 ; Huber and Moulton, 2006 ; Huber et al., 2006 ).

In a revision of North American species of Silene L., Hitchcock and Maguire (1947) noted that the characters used to separate Melandrium Röhl., from Silene and Lychnis L. were too variable among American species to be of any taxonomical value at the generic level. They chose to include the majority of North American representatives of Melandrium in Silene, the revision thus encompassing 54 species in total. Later on, Chowdhuri (1957) too included Melandrium in Silene in his worldwide revision of Silene. The use of subgenera was dismissed, and he instead erected 44 sections, which he considered as natural groups. The native North American species were dispersed among several sections. Most of the species were included in the sections Occidentales Chowdhuri, Graminifoliae Chowdhuri, and Quadrilobatae Chowdhuri. Silene menziesii (= Anotites Greene) and its close allies were considered to belong to Silene section Rupifraga Otth in DC by Chowdhuri. A number of Melandrium species not considered by Hitchcock and Maguire (1947) were included in Silene section Physolychnis (Benth.) Bocquet [= S. section Gastrolychnis (Fenzl.) Chowdhuri] by Bocquet (1969) . Morton (Flora of North America, 2005 ) included both Lychnis and Viscaria Bernh. in Silene, and the genus thus comprises c. 70 species in North America. Recent phylogenetic studies based on molecular data from different genomic compartments have identified a well-supported group of representatives from S. section Physolychnis, section Occidentales, section Odontopetalae, the S. ajanensis (= Lychnis ajanensis) group, and a number of other, chiefly Asian species (Oxelman and Lidén, 1995 ; Oxelman et al., 1997 ; Oxelman et al., 2001 ; Popp and Oxelman, 2004 ; Popp et al., 2005 ). The group is informally named Physolychnis s.l. in the following text. Physolychnis s.l. is nested in a more inclusive clade corresponding to S. subgenus Behen (Dumort.) Rohrb., one of the two major clades in Silene. The second major clade, S. subgenus Silene, is represented by Silene antirrhina L., S. repens Patrin, and S. acaulis (L.) Jacq., which also are native to North America (Oxelman and Lidén, 1995 ; Oxelman et al., 1997 ; Oxelman et al., 2001 ; Popp and Oxelman, 2004 ; Eggens et al., 2007 ).

The basic haploid chromosome number in Silene is x = 12, and a majority of the species studied so far are diploids (2n = 24). However, Kruckeberg demonstrated extensive polyploidy during biosystematic studies of North American Silene (Kruckeberg, 1954 , 1960 , 1961 , 1962 , 1963 ). With few exceptions, the Silene species native to North America are polyploids, ranging from tetraploids (2n = 4x = 48) to octoploids (2n = 8x = 96). Both diploidy and polyploidy has been documented in S. section Physolychnis (Benth.) Bocquet, but information is lacking for the majority of taxa (Kruckeberg, 1962 ; V. V. Petrovsky, Dept. of Vegetation of the Far North, Komarov Botanical Institute, Russia, and R. Elven, National Centre of Biosystematics, Natural History Museum, University of Oslo, Norway, unpublished manuscript).

The predominantly maternal inheritance of the chloroplast among angiosperms (including Caryophyllacae; Corriveau and Coleman, 1988 ) makes chloroplast DNA (cpDNA) regions suitable for inferring the maternal lineages in polyploids. In combination with the internal transcribed spacers (ITS) of the nuclear ribosomal DNA (nrDNA), cpDNA has been used to infer relationships and origins of polyploid plants (Sang et al., 1995 ; Campbell et al., 1997 ; Ge et al., 1999 ; Popp and Oxelman, 2001 ; Popp et al., 2005 ). Several studies, however, have shown that the homoeologous ITS repeats in an allopolyploid may be homogenized, so that one of the parental lineages effectively gets lost (Wendel et al., 1995 ; Brochmann et al., 1996 ; Popp et al., 2005 ; but see Rauscher et al., 2002 ). If the homogenization is directed towards the paternal lineage, the allopolyploid may reveal its hybrid origin as incongruence between the plastid and ITS data sets (Smedmark and Eriksson, 2002 ; Popp et al., 2005 ). If, on the other hand, the homogenization is directed toward the maternal lineage, no phylogenetic signal revealing the hybridization will be left in the ITS region (Wendel et al., 1995 ; Brochmann et al., 1996 ). Moreover, it is also possible that the result will be a mosaic of the parental sequences (Mummenhoff et al., 1997 ). Thus, to infer relationships reliably in groups containing taxa of putative hybrid origin, it is therefore necessary to include several independent nuclear markers.

The RNA polymerase (RNAP) gene family in most eukaryotes consists of three large DNA-dependent RNA polymerase multisubunit enzymes, and a fourth member is unique to plants (Arabidopsis Genome Initiative, 2000 ). In Arabidopsis thaliana, the second largest subunits (RPA2, RPB2, RPC2, and RPD2) of each of the multisubunit enzymes are encoded on chromosomes 1, 4, 5, and 3, respectively (Arabidopsis Genome Initiative, 2000 ). Whereas RPD2 is duplicated, single-copy genes encode the RPA2, RPB2, and RPC2 subunits (Arabidopsis Genome Initiative, 2000 ). Oxelman et al. (2004) inferred a duplication of RPB2 early in the evolution of the core eudicots followed by several independent losses, and RPD2 has been hypothesized to be duplicated in the Caryophyllaceae (Popp and Oxelman, 2004 ; Popp et al., 2005 ). Intron regions of RPB2 have been used for phylogenetic studies among closely related species and polyploid complexes, whereas coding regions have been applied to infer deep phylogenetic relationships (Denton et al., 1998 ; Oxelman and Bremer, 2000 ; Popp and Oxelman, 2001 ; Pfeil et al., 2004 ; Oxelman et al., 2004 ; Goetsch et al., 2005 ; Thomas et al., 2006 ; Hajibabaei et al., 2006 ). The other subunits are not as well studied, but RPA2 and RPD2 were used with RPB2, ITS, and chloroplast regions to infer the phylogeny of the tribe Sileneae and an Arctic/subarctic polyploid species complex in Silene (Popp and Oxelman, 2004 ; Popp et al., 2005 ), and Grundt et al. (2004) used RPD2 in combination with ITS and random amplified polymorphic DNA (RAPD) to study a circumpolar species complex in Draba (Brassicaceae).

In this study, we used chloroplast and nuclear DNA sequences to infer phylogenetic relationships in Silene. Sequences from the rps16 intron (Oxelman et al., 1997 ) and the psbE-petL spacer (Popp et al., 2005 ) represent the chloroplast, and the ITS region (Baldwin et al., 1995 ), the putatively unlinked intron 23 from RPA2 and RPB2, and intron 7 from RPD2a and RPD2b (Popp and Oxelman, 2004 ) represent the nuclear genome.

In light of the very high frequency of polyploidy among the Silene species native to North America compared to Silene in general, we tested if the polyploid species of Silene native to North America form a monophyletic group. The mode of origin (auto- or allopolyploid origin) and the relationships of the polyploid taxa are also explored. Furthermore, the relationship of S. section Physolychnis (Benth.) Bocquet and the native North American Silene is explored. We also aimed to infer the diploid parental lineages of the polyploids. Diploids forming sister groups to polyploids represent putative parental lineages of polyploids.

MATERIALS AND METHODS

Plant material
Sixty-one taxa, of which eight are non-autonymous subspecies or varieties, are included in this study. Forty-three of the sampled taxa are native to North America. Although S. antirrhina, S. acaulis, S. williamsii, S. seelyi, and the Arctic/subarctic S. uralensis are diploids (2n = 24), the majority of the Silene species native to North America are tetraploids (2n = 48; Flora of North America, 2005 ; Kruckeberg, 1954 , 1960 , 1961 , 1962 , 1963 ). In addition, hexaploids (S. hookeri with 2n = 72) and octoploids (S. petersonii with 2n = 96) as well as species with more than one ploidy level (S. menziesii and S. repens with 2n = 24 and 48, S. laciniata with 2n = 48, 72, 96, and S. scouleri subsp. scouleri and S. parryi with 2n = 48, 96) have been reported (Flora of North America, 2005 ; Kruckeberg, 1961 , 1962 , 1963 ). As far as possible, we have used specimens with a known chromosome number for sequencing. The plant material in this study is presented with voucher data and GenBank/EMBL accession numbers in the Appendix.

PCR, cloning, and sequencing
Total genomic DNA was extracted from herbarium material as described in Oxelman et al. (1997) . For some extractions, a MiniBeadBeater with 2.5 mm zirconia/silica beads (BioSpec Products, Bartlesville, Oklahoma, USA) running for 40 s at 5000 rpm replaced the mortar and liquid nitrogen.

PCR was performed using several different commercial DNA polymerases according to the manufacturers' manuals. Usually either Taq polymerase (Advanced Biotechnologies, Epsom, UK) with 0.625 U enzyme/25 µL reaction, 1.5–2.5 mM Mg²⁺, and 0.01% bovine serum albumin or the TripleMaster enzyme (Eppendorf, Hamburg, Germany) with 1.25 U/25 µL reaction and HighFidelity buffer were used. Standard PCR conditions were 25 or 50 µL reactions with 200 µM of each dNTP, 0.5–1.0 µM of each primer, and 0.5–1.0 µL of total genomic DNA of unknown concentration. All reactions were run on an Eppendorf Mastercycler gradient with an initial denaturation at 95°C for 5 min followed by 35–45 cycles of 95°C for 30 s, 56–58°C for 1 min, and 72°C for 1–2.5 min. The cycling ended with 72°C for 10 min. Primers used for PCR and sequencing are described in Table 1.

View larger version (79K): [in this window] [in a new window]   Fig. 1. One of 430 most parsimonious trees found in the analysis of the combined psbE-petL and rps16 chloroplast data set. Heavy internal branches are retained in the strict consensus tree. Numbers associated with nodes indicate maximum parsimony bootstrap frequencies. Branch lengths are proportional to the number of changes. An asterisk or a plus sign following the taxon name indicates that the sequence is lacking from the psbE-petL data set or the rps16 data set, respectively. If known, ploidy level is given after the taxon names; ploidy level in italics indicates that the sequence is from a chromosome-determined specimen. The North American Silene menziesii group and Physolychnis s.l. are indicated by gray squares. The hexaploid S. hookeri is highlighted by a black arrow, and the tetraploid S. nivea and S. multinervia (unknown ploidy level) are highlighted by gray arrows. Geographic distribution is indicated in bold

View larger version (66K): [in this window] [in a new window]   Fig. 6. One of 9860 most parsimonious trees found in the analysis of the RPD2b data set. Heavy internal branches are retained in the strict consensus tree. Numbers associated with nodes indicate maximum parsimony bootstrap frequencies. Branch lengths are proportional to the number of changes. Numbers in parentheses indicate the number of clones used to construct consensus sequences. If known, ploidy level is indicated after the taxon names; ploidy level in italics indicates that the sequence is from a chromosome-determined specimen. The North American Silene menziesii group and Physolychnis s.l. are indicated by gray squares. The hexaploid S. hookeri and S. multinervia (unknown ploidy level) are highlighted by black and gray arrows, respectively. Geographic distribution is indicated in bold

Sequences were edited using Sequencher 3.1.1 (Gene Codes, Ann Arbor, Michigan, USA). To reduce the impact of PCR artifacts such as chimeric sequences and polymerase errors on the phylogenetic reconstruction (Hengen, 1995 ; Cline et al., 1996 ; Popp and Oxelman, 2001 ; Cronn et al., 2002a ; Kanagawa, 2003 ), consensus sequences from cloned taxa were constructed following Popp et al. (2005) : a maximum parsimony analysis (see Alignment and phylogenetic analysis) with only parsimony informative characters was performed including all sequences from a DNA region. The topology was inspected for obvious signs of chimeric sequences, revealed as long terminal branches due to the increased homoplasy. After visual inspection, the sequences suspected to be chimeras were excluded from further analyses. Monophyletic groups of clones obtained from a single accession differing by autapomorphic substitutions only were regarded as single sequence, i.e., autapomorphic substitutions were interpreted as polymerase errors. If a subset of two or more clones from a single accession had one or more parsimony informative substitution in common (thus distinguishing them from other clones from the same accession), they were regarded as a single, separate sequence. Majority rule consensus sequences were constructed for each group of clones forming such separate sequences within a single accession. Single clones that were not recovered within a monophyletic group of clones from the same accession (or species) were regarded as distinct sequences.

Alignment and phylogenetic analyses
Sequences were aligned manually using Se-Al (Rambaut, 1996 --2002). Gaps, i.e., inferred insertions/deletions were coded as absent/present using simple gapcoding (Simmons and Ochoterena, 2000 ) as implemented in SeqState (version 1.25; Müller, 2005 , 2006 ). Excluding taxa lacking from either the psbE-petL or rps16 data set, we used the incongruence length difference ILD test (Farris et al., 1994 ) as implemented in PAUP* (version 4.0b10; Swofford, 2002 ) with 1000 replicates, 10 random addition sequences, tree-bisection-reconnection (TBR) branch swapping, and saving 10 trees from each replicate to test for significant incongruence between the cpDNA data sets. No incongruence was detected (P = 0.57) and the cpDNA regions were combined for further analyses. Phylogenetic analyses were made using PAUP* (version 4.0b10) for Macintosh or UNIX (Swofford, 2002 ). Maximum parsimony (MP) analyses of the matrices were performed using a heuristic search strategy with 1000 random trees as starting points and TBR branch swapping, saving 10 trees from each replicate. Maximum parsimony bootstrap (MPB) analyses were carried out with full heuristics, 1000 replicates, TBR branch swapping, the MULTREES option off, and random addition of sequences with four replicates. Because initial analyses indicated several independent lineages in North America containing polyploids, the rps16, ITS, RPA2, RPB2, RPD2a, and RPD2b sequences from the North American S. menziesii (2n = 24, 48), S. williamsii (2n = 24), S. seelyi (2n = 24), S. nivea (2n = 48), and S. multinervia (2n = ?) were combined and analyzed together with a previously published combined matrix with representatives—including Physolychnis s.l.—from the tribe Sileneae (Popp and Oxelman, 2004 ).

Mode of origin of polyploidy
The mode of origin of polyploids is a mere hypothesis, and if we state H₀ = "the polyploid is an autopolyploid" and H₁ = "the polyploid is an allopolyploid," the hypothesis can be tested in a phylogenetic context. Autopolyploidy is the null hypothesis and as such cannot be "confirmed" even though the analysis may reveal a sister-group relationship of the homoeologues from a single accession. If the analysis reveals a nonsister-group relationship of the homoeologues, then autopolyploidy is rejected if at least one of the homoeologues (or clade of homoeologues) has a sequence from a plant with lower ploidy level as its sister group. Note that an autopolyploid may diversify ("speciate") so that the homoeologues will form parallel clades in the same way as after ordinary gene duplications. The descendant lineages may well in turn form allopolyploids, both with lineages stemming from the autopolyploidization event, as well as with other lineages. Given that polyploidy is an irreversible process, some gene phylogenies will always be inconsistent with polyploidy (i.e., a sequence from a diploid nested within sequences from polyploids) and as such reject polyploidy. Duplication of a DNA region, unlinked to polyploidy, followed by lineage sorting is a potential source of Type I error, i.e., H₀ (autopolyploidy) is wrongly rejected in favor of H₁ (allopolyploidy). Connected with—but not isolated to (see Shaked et al., 2001 )—greater age of a polyploidization event, the difficulties of discriminating between the effects of polyploidization and independent gene duplications combined with gene loss, i.e., orthology assessment, increase substantially (Wendel, 2000 ; Pfeil et al., 2005 ). Because it is unlikely that the same pattern of lineage sorting will be displayed in unlinked regions, the risk of Type I errors is minimized by analyzing several independent biparentally inherited DNA-regions. Type II errors, i.e., cases where H₀ (autopolyploidy) is not rejected when in fact H₁ (allopolyploidy) is true, is best addressed by carefully choosing DNA regions containing a phylogenetic signal strong enough to separate the lineages included in the analysis.

RESULTS

PCR and sequencing
EMBL/GenBank accession numbers for the sequences produced by direct sequencing of PCR products as well as consensus sequences produced from cloned PCR products are given in the Appendix, except for sequences that are only included in the combined data set, which are given in Popp and Oxelman (2004) . Direct sequencing of PCR products from the chloroplast regions rps16 and psbE-petL usually resulted in unambiguous sequences. The nuclear DNA regions, however, had to be cloned for the majority of the polyploid taxa, as well as some of the diploids. The number of sequenced clones used to construct each consensus sequence (as described in PCR, cloning, and sequencing) is given after the corresponding taxon name in the figures showing the inferred phylogenetic trees.

Phylogenetic analyses
Table 2 summarizes the characteristics of the matrices and the resulting trees.

ITS
Both the S. menziesii group and Physolychnis s.l. were supported as monophyletic (MPB 100 and 96%, respectively) in a large polytomy consisting of taxa from the subgenera Silene and Behen (Fig. 2). The diploid S. seelyi was weakly supported (MPB 53%, Fig. 2) as sister to the diploid accession of S. menziesii and the accession with unknown ploidy level, whereas the tetraploid accession occurred as unresolved in a basal polytomy in the S. menziesii group. In Physolychnis s.l., the Siberian diploid S. ajanensis and the Arctic/subarctic hexaploids S. sorensenis and S. ostenfeldii (MPB 79%) and tetraploid S. involucrata (MPB 99%) formed a previously recognized sister group to diploid S. keiskei from Japan (MPB 75%; Fig. 2 and Popp et al., 2005 ). The European diploid S. zawadskii was recovered as sister to that clade in the strict consensus tree but received a MPB < 50%. In contrast to the cpDNA analysis, all seven clones from hexaploid S. hookeri were recovered as a monophyletic group together with the majority of the North American polyploids in Physolychnis s.l. (Figs. 1 and 2). To obtain clean sequences, the North American S. oregana (2n = 48), S. laciniata subsp. laciniata (2n = 96), and the South American S. mandonii (2n = ?) were also cloned. All clones (8, 4, and 8, respectively) were recovered as monophyletic within their respective accession (Fig. 2). A South American group (MPB 99%) was formed by S. mandonii (2n = ?) and S. thysanodes (2n = ?) in a polytomy with the North American tetraploids S. verecunda and S. occidentalis (MPB 62%). Among the North American species, S. polypetala (2n = 48) was resolved as sister (MPB 68%) to a group (MPB 100%) of the tetraploids S. caroliniana, S. virginica, and the S. stellata accession from Burleigh and Holtsford (2003) . Of the eight clones from S. caroliniana subsp. pensylvanica, five formed together with a second accession of the same species a sister group (MPB 98%) to the remaining three clones and several accessions of S. caroliniana subsp. caroliniana, S. caroliniana subsp. wherryi, S. virginica, and the Burleigh and Holtsford S. stellata accession. Silene ovata (2n = 48) was recovered as sister (MPB 83%) to the tetraploids S. regia and S. subciliata, whereas S. laciniata subsp. californica (2n = 48, 72) was sister to S. campanulata subsp. glandulosa (2n = 48). As in the cpDNA analysis, the three subspecies of S. laciniata failed to form a monophyletic group (Figs. 1 and 2). The tetraploids S. bernardina, S. invisa, S. scaposa, S. douglasii var. oraria, and S. scouleri subsp. scouleri formed a weakly supported group (MPB 59%). The tetraploid North American S. nivea was recovered together with the diploid European S. baccifera (MPB 55%) in a large polytomy including also the S. menziesii group, Physolychnis s.l., the North American S. multinervia (2n = ?), several European diploids from subgenus Behen, and the taxa belonging to subgenus Silene (Fig. 2).

View larger version (83K): [in this window] [in a new window]   Fig. 2. One of 5840 most parsimonious trees found in the analysis of the ITS data set. Heavy internal branches are retained in the strict consensus tree. Numbers associated with nodes indicate maximum parsimony bootstrap frequencies. Branch lengths are proportional to the number of changes. Numbers in parentheses indicate the number of clones used to construct consensus sequences. If known, ploidy level is given after the taxon names; ploidy level in italics indicates that the sequence is from a chromosome-determined specimen. The North American Silene menziesii group and Physolychnis s.l. are indicated by gray squares. The hexaploid S. hookeri is highlighted by a black arrow, and the tetraploid S. nivea and S. multinervia (unknown ploidy level) are highlighted by gray arrows. Geographic distribution is indicated in bold. 1Sequences from Burleigh and Holtsford, (2003), 2sequences from Freeman et al. (2002)

View larger version (70K): [in this window] [in a new window]   Fig. 3. One of 8683 most parsimonious trees found in the analysis of the RPA2 data set. Heavy internal branches are retained in the strict consensus tree. Numbers associated with nodes indicate maximum parsimony bootstrap frequencies. Branch lengths are proportional to the number of changes. Numbers in parentheses indicate the number of clones used to construct consensus sequences. If known, ploidy level is given after the taxon names; ploidy level in italics indicates that the sequence is from a chromosome-determined specimen. The North American Silene menziesii group and Physolychnis s.l. are indicated by gray squares. The hexaploid S. hookeri is highlighted by black arrows, and the tetraploid S. nivea and S. multinervia (unknown ploidy level) are highlighted by gray arrows. Geographic distribution is indicated in bold

View larger version (84K): [in this window] [in a new window]   Fig. 4. One of 8330 most parsimonious trees found in the analysis of the RPB2 data set. Heavy internal branches are retained in the strict consensus tree. Numbers associated with nodes indicate maximum parsimony bootstrap frequencies. Branch lengths are proportional to the number of changes. Numbers in parentheses indicate the number of clones used to construct consensus sequences. If known, ploidy level is given after the taxon names; ploidy level in italics indicates that the sequence is from a chromosome-determined specimen. The North American Silene menziesii group and Physolychnis s.l. are indicated by gray squares. The hexaploid S. hookeri is highlighted by black arrows, and the tetraploid S. nivea and S. multinervia (unknown ploidy level) are highlighted by gray arrows. Geographic distribution is indicated in bold

View larger version (56K): [in this window] [in a new window]   Fig. 5. One of 41 most parsimonious trees found in the analysis of the RPD2a data set. Heavy internal branches are retained in the strict consensus tree. Numbers associated with nodes indicate maximum parsimony bootstrap frequencies. Branch lengths are proportional to the number of changes. Numbers in parentheses indicate the number of clones used to construct consensus sequences. If known, ploidy level is indicated after the taxon names; ploidy level in italics indicates that the sequence is from a chromosome-determined specimen. The North American Silene menziesii group and Physolychnis s.l. are indicated by gray squares. The hexaploid S. hookeri and S. multinervia (unknown ploidy level) are highlighted by black and gray arrows, respectively. Geographic distribution is indicated in bold

The combined data set
The majority of the results from the analysis of the combined data set has been presented in Popp and Oxelman (2004) , and we therefore only focus on the position of the newly added taxa. The North American tetraploid S. verecunda (erroneously named S. parishii in Popp and Oxelman, 2004 ), due to multiple nonmonophyletic sequences (i.e., homoeo- or paralogous copies) was excluded from the RPB2, RPD2a, and RPD2b parts of the combined data set. For the same reason, S. rotundifolia was excluded from the RPB2 part. Because of the possibility of that we may have combined non-orthologous homoeologues/paralogues from S. verecunda and S. rotundifolia, we caution for drawing any conclusions about the relationships among the many polyploids within Physolychnis s.l. These sequences should only be taken as representatives for the majority of North American polyploids in Physolychnis s.l. The analysis strongly supported Physolychnis s.l. and the S. menziesii group as monophyletic and only distantly related (Fig. 7). Physolychnis s.l., with the European diploid S. zawadskii as sister group, included S. verecunda and S. rotundifolia representing the majority of the North American polyploids. The North American tetraploid S. nivea formed together with the European diploid S. baccifera a sister group to the S. menziesii group, whereas S. multinervia (2n = ?) was moderately supported (MPB 65%) as sister to Physolychnis s.l. (including S. zawadskii).

View larger version (40K): [in this window] [in a new window]   Fig. 7. One of 12 most parsimonious (MP) trees found in the analysis of taxa from the Silene menziesii group, S. uralensis, S. nivea, S. multinervia, and the combined data set from Popp and Oxelman (2004) . Heavy internal branches are retained in the strict consensus tree. Numbers associated with nodes indicate maximum parsimony bootstrap frequencies. Branch lengths are proportional to the number of changes. The North American S. menziesii group and Physolychnis s.l. (including the North American tetraploid S. verecunda and S. rotundifolia) are indicated by gray squares, and the tetraploid S. nivea and S. multinervia (unknown ploidy level) are indicated by gray arrows

Monophyly of North American Silene
Our results clearly reject a hypothesis wherein the polyploid Silene native to North America form a single monophyletic group. The majority of the North American polyploids belongs to Physolychnis s.l., a group recognized in several previous studies (Oxelman and Lidén, 1995 ; Oxelman et al., 1997 , 2001 ; Popp and Oxelman, 2004 ; Popp et al., 2005 ). Physolychnis s.l. includes not only polyploid North American Silene, but also taxa from Japan (S. keiskei 2n = 24), Siberia (S. ajanensis 2n = 24), and Europe (S. zawadskii 2n = 24), as well as the representatives of S. section Physolychnis s.s. from South America (i.e., S. chubutensis, S. mandonii, and S. thysanodes), S. nigrescens from Central Asia (ploidy levels not known), and the circumpolar Arctic diploid S. uralensis. The relationships within Physolychnis s.l. are poorly resolved (Figs. 1--6), and the presence of homoeologues that are poorly resolved phylogenetically in addition to uneven taxon sampling between data sets makes the interpretation difficult. However, some interesting conclusions and suggestions can be made and are outlined later.

A second group consists of di- and tetraploid S. menziesii and the diploids S. williamsii and S. seelyi. The group is only distantly related to the rest of the North American species and clearly has a separate origin from Physolychnis s.l. Whereas each of the individual gene trees (Fig. 1--6) only weakly rejects a sister-group relationship between Physolychnis s.l. and the S. menziesii group, the combined analysis strongly supports that the North American taxa in these two clades do not share a most recent common ancestor (Fig. 7).

Although the two main lineages of North American Silene, Physolychnis s.l. and the S. menziesii group, are well separated, a strong connection was found in the form of the hexaploid S. hookeri. Our results are congruent with a hypothesis where the ancestor of the S. menziesii group contributed to the allopolyploid origin of S. hookeri as the chloroplast donor (Fig. 1). The ITS sequence of S. hookeri resides within Physolychnis s.l. with strong support (Fig. 2), which is consistent with the Physolychnis s.l. lineage as paternal donor. In all RNAP trees (Figs. 3, 4, 6), except in RPD2a (discussed later), we consistently recovered two S. hookeri lineages in Physolychnis s.l. and one in the S. menziesii group. Thus, an autopolyploid origin of S. hookeri can confidently be rejected, and the consistent pattern in RPA2, RPB2, and RPD2b together with the ITS and cpDNA data support an allopolyploid origin with a tetraploid paternal donor from Physolychnis s.l. and a diploid maternal donor from the S. menziesii group. Hitchcock and Maguire (1947) found S. menziesii similar to the tetraploids S. campanulata and S. suksdorfii in many characters, whereas Greene (1905) considered it more similar to Stellaria or Alsine. These similarities were not reflected in the phylogenies inferred from the chloroplast or the nuclear DNA data. The RPB2 region in S. menziesii has been subjected to extensive cloning, using several different primer combinations to amplify fragments that may link S. menziesii to S. campanulata and S. suksdorfii (no RPB2 sequences available for the latter) or Stellaria (RPB2 sequences too diverged to align with Silene). All attempts recovered only a single sequence from S. menziesii, and we find no support for the previously suggested hypotheses.

No RPD2a sequence was obtained from S. menziesii. Failure to amplify one or more "expected" paralogues or homoeologues have been explained by, for example, PCR selection, large inserts in one copy, or physical elimination of one of the redundant copies (Wagner et al., 1997 ; Shaked et al., 2001 ; Tank and Sang, 2001 ). In this particular case, it is noticeable that we failed to amplify not only the RPD2a region in S. menziesii but also the "expected" S. menziesii lineage homoeologue in S. hookeri (only a Physolychnis s.l. homoeologue was recovered from S. hookeri, Fig. 5). The phylogenetic analyses suggest that the hybridization between the two lineages occurred before the divergence of S. menziesii and its close diploid relatives (Figs. 3 and 6, unresolved in the other trees), but the apparent "loss" of the RPD2a paralogue from S. menziesii and the S. menziesii lineage homoeologue from S. hookeri requires that these losses were independent, thus contradicting that hypothesis. More data is needed to infer the order of hybridization and divergence of S. hookeri and the S. menziesii group. None of our analyses can reject an autotetraploid origin of S. menziesii itself (Figs. 1--4, 6; unpublished data). Unfortunately, the clade is poorly resolved (Figs. 1--6), and it is not possible to confidently discern from which of the diploid lineages, including diploid S. menziesii, the tetraploid may stem.

Neither the tetraploid S. nivea nor S. multinervia (ploidy level not known) belong to any of two main groups, including North American polyploids. To some extent, this is not unexpected for the latter; it is a weedy annual (Hitchcock and Maguire, 1947 ) and morphologically clearly belongs to the section Conoimorpha Otth, which is characterized by a base chromosome number of x = 10 and a ovate-conical calyx with 20–30 veins. Morton (Flora of North America, 2005 ) considered S. multinervia conspecific to S. coniflora Nees and introduced to the Pacific coast by early European settlers. It is striking that S. multinervia and S. conica, which is a close relative to S. coniflora, do not form a monophyletic group in any of the gene trees, despite their great morphological resemblance. The identity and historical origin of S. multinervia clearly warrants further study.

Chowdhuri (1957) placed S. nivea in section Siphonomorpha together with the tetraploid S. stellata among others. The bulk of section Siphonomorpha has been confidently shown to belong to subgenus Silene by previous molecular phylogenetic studies (Oxelman and Lidén, 1995 ; Desfeux et al., 1996 ; Oxelman et al., 1997 ; Popp and Oxelman, 2004 ). Silene stellata clearly is part of Physolychnis s.l. according to our results. The close relationship between S. nivea and S. stellata, supported by an analysis of chloroplast trnL intron sequences (Burleigh and Holtsford, 2003 ) wherein S. nivea was part of a more inclusive group of eastern North American taxa belonging to Physolychnis s.l. is contradictory to our results (Fig. 1). Analyses of chloroplast rps16 intron and psbE-petL spacer sequences with a restricted taxon sampling, similar to that of Burleigh and Holtsford's, do not group our S. nivea sequences with the rest of the North American polyploids in Physolychnis s.l. (results not shown). Our sequences from rps16 and psbE-petL were produced from two different DNA extractions from a single S. nivea specimen, both resulting in a similar position of S. nivea. The sequences obtained from S. nivea are not identical to any ITS or rps16 sequence published in the EMBL/GenBank databases, nor are they identical to any of the unpublished sequences in our database. The species is easily recognized by its long willow-like leaves and leafy inflorescence, and we are therefore confident with the plant identification. Our conclusion is that the phylogenetic positions of S. nivea can be explained in a biological context discussed later.

Mode of origin and phylogenetic relationships of American polyploids and Physolychnis s.l
Pointing out that the primary criterion for classifying a polyploid is its mode of origin, Ramsey and Schemske (1998) adopt a biological species concept and denote "autopolyploid" as polyploids arising from crosses within or between populations of a single species and "allopolyploid" to indicate polyploids derived from hybrids between species. This is much in line with the original definitions by Kihara and Ono (1926) . We agree fully with Ramsey and Schemske in as much as the primary criterion for classification of polyploids should be the mode of origin, but disagree with simply choosing one of many species concepts and then use an arbitrary taxonomic rank as the basis for classifying polyploids.

The North American polyploids usually had the number of homoeologues of the RNAP genes as expected according to their ploidy levels, that is, tetraploids had two different sequences (according to the principles outlined in Materials and Methods) and hexaploids three (Figs. 3--6). For those taxa with unknown ploidy levels, it might be suggested that S. nigrescens and S. multinervia are diploids, because they always had only one sequence (or a monophyletic group of sequences) per gene. The South American species S. chubutensis and S. mandonii would by the same logic be interpreted as tetraploids (two distinct, nonmonophyletic, homoeologues in RPA2, RPB2, and RPD2a). Silene laciniata subsp. californica, a taxon for which both tetraploids and hexaploids have been reported, could be interpreted as hexaploid, because three nonmonophyletic sequences were found in RPB2 (Fig. 4) and RPD2b (Fig. 6). However, strong caution should be taken, because alleles may form nonmonophyletic groups in the gene trees, and, conversely, homologues may be extinct or undetected.

Neither the RPA2 data set (Fig. 3) nor the RPD2a data set (Fig. 5) resolved the homoeologues from North American polyploids together with sequences from diploids. Thus, the autopolyploidy hypothesis cannot be rejected by these gene trees. In the RPB2 (Fig. 4) and RPD2b (Fig. 6) gene trees, there are groups (although with weak or moderate support only) of homoeologues that would reject the autopolyploidy hypothesis, because in each tree, one of the groups includes diploid accessions. However, for RPB2 there is an alternative explanation (outlined later) involving gene duplication unlinked to polyploidy. In conclusion, the support for a rejection of the autopolyploidy hypothesis is only as strong as the support for the two clades involving North American polyploids of Physolychnis s.l. displayed in Fig. 6. The number of tetraploidization events in the group is unresolved. The origin of the Arctic polyploids is separate, and the data presented here does not affect the hypotheses put forward by Popp et al. (2005) .

Among the diploids belonging to Physolychnis s.l., S. zawadskii is resolved as sister to the rest in the cpDNA, RPD2a, and RPD2b trees and is unresolved with respect to this in the RPA2 tree. The ITS sequence is most parsimoniously, but weakly supported, as nested within Physolychnis s.l. The RPB2 sequence of S. zawadskii, by contrast, is moderately supported as belonging to a clade consisting of North American and Arctic/subarctic taxa, which is nested within Physolychnis s.l. (Fig. 4). Both RPB2 sequences isolated from S. keiskei belong to a clade consisting of North American and South American taxa (Fig. 4). This clade was also identified by Popp et al. (2005) but with much smaller taxon sampling. One of the S. keiskei sequences was identified as belonging to a group of putative pseudogenes from the Arctic polyploids, unlinked to the polyploid past of these (Popp et al., 2005 ). A possible explanation to the RPB2 pattern might be a duplication of the RPB2 gene in a diploid ancestor to Physolychnis s.l. and subsequent differential gene loss or substitution patterns disabling PCR amplification with our primers. This would then indicate that the American tetraploids with RPB2 sequences resolved in the same clade as S. zawadskii have a separate origin from those that only have sequences resolved in the "S. keiskei" clade. However, the support for this clade is low, and there might be undetected homoeologues in some of the taxa.

The diploid Siberian S. ajanensis lineage, which has contributed repeatedly to polyploid formation in Arctic/subarctic taxa belonging to Physolychnis s.l. (the tetraploid S. involucrata and the hexaploids S. sorensenis and S. ostenfeldii, Popp et al., 2005 ) seems not to have been involved in the formation of any of the North American polyploid lineages included in this study. It always forms a well-supported group in the separate gene trees (Fig. 2--6). The second diploid contributor to the Arctic/subarctic complex, "S. uralensis" lineage (Popp et al., 2005 ) also seems unlikely as a contributor to the North American polyploids in Physolychnis s.l., because none of the RNAP homoeologues share a most recent common ancestor with this lineage (Figs. 3--6). In the ITS tree (Fig. 2), the moderately supported sister group relationship between S. uralensis and S. suksdorfii (2n = 48) may indicate a contribution to polyploidy in the latter species. Silene suksdorfii also share a unique cpDNA substitution with the S. uralensis lineage (Fig. 1). However, this substitution is also shared by S. parryi, S. scouleri subsp. scouleri, and S. drummondii, and the resolution among these sequences is poor. Thus, it is possible that the S. uralensis cpDNA lineage is separate from these North American polyploids. Further cpDNA sequencing might help resolve this. Unfortunately, we were not able to obtain sequences from any of the RNAP genes for S. suksdorfii. From our results, it is not possible to pinpoint any extant diploid lineages as strongly supported for being involved in the formation of the North American polyploids in Physolychnis s.l.

The tetraploid S. nivea does not appear to have any close relatives among the North American Silene species in our analyses (Figs. 1 and 2), but a close relationship with the European diploid S. baccifera is strongly supported by the RPA2 and RPB2 data sets (Figs. 3 and 4). This contradicts the analysis of trnL data by Burleigh and Holtsford (2003) , which indicated a close relationship to the tetraploid S. stellata. Such a pattern is only understandable in a biological sense if processes such as introgression or multiple allopolyploid origins (e.g., Soltis and Soltis, 1999 ) with different maternal lineages are invoked. Our results alone, however, do not reject an autopolyploid origin of S. nivea, where the ancestral lineage was only distantly related to either of the two other lineages that today are represented by the North American polyploids. Morphologically, S. nivea shows similarities to the S. menziesii group, especially in leaf and calyx morphology, whereas S. baccifera is dissimilar in having a berry-like fruit, a climbing habit, and different leaves and calyx. Thus, in addition to the deep polyploidization(s) in the ancestor(s) of the South and North American polyploids belonging to Physolychnis s.l., the relationships inferred here hypothesize at least three independent polyploid origins; S. menziesii (putative autotetraploid), S. hookeri (allohexaploid), and S. nivea (putative autotetraploid).

With the inclusion of South American representatives of S. section Physolychnis, i.e., S. mandonii, S. chubutensis, and S. thysanodes, a North and South American biogeographic connection is added to the previously known close relationship between North American Arctic and subarctic to Siberian and Central Asian species (discussed earlier, and Popp et al., 2005 ). A similar pattern was seen in Cerastium (Scheen et al., 2004 ) wherein some North American species are grouped with Arctic species and some with South American species. The two groups of North American taxa were only distantly related, and Scheen et al. (2004) inferred the pattern to correspond to two separate immigrations to North America. Although the North American Silene are separated in several distantly related lineages, both the South American and Arctic connections are recovered in the same clade, much in line with for example South and North American Astragalus (Wojciechowski et al., 1999 ).

At more shallow levels, it is hard to find simple and well-supported straightforward relationships among the American taxa in the Physolychnis s.l. clade. In agreement with Burleigh and Holtsford (2003) , we find a close relationship between the tetraploids S. caroliniana and S. virginica, the latter appearing to have originated within the former. However, our study does not indicate an obvious relationship of the tetraploid S. stellata with the former two as Burleigh and Holtsford (2003) suggested. It is also notable that S. andersonii, put in synonymy to S. verecunda (both are tetraploid) by Morton (Flora of North America, 2005 ), does not form monophyletic groups with the other S. verecunda sequences (Figs. 1 and 2). Silene scouleri subsp. pringlei (2n = 60) apparently does not share the same plastid ancestor as S. scouleri subsp. scouleri (2n = 48, 96), which is part of a group of Arctic and high-alpine taxa (Fig. 1). Perhaps most striking is that the different subspecies of S. laciniata not only differ in ploidy levels (2n = 48, 72, 96) but also appear to lack a common origin, thus indicating several independent origins of hummingbird pollination in addition to S. regia, S. virginica, S. rotundifolia, and S. subciliata.

The poor resolution in our phylogenetic analyses may be explained by rapid radiation following the southward migration from the Arctic. Rapid radiation, resulting in high morphological diversity and poorly resolved phylogenetic trees (Fishbein et al., 2001 ), has been suggested for other North American plants (e.g., Hershkovitz and Zimmer, 2000 ; Kelch and Baldwin, 2003 ). In addition, homogenization by recombination or gene conversion between homoeologous regions in allopolyploids may result in loss of resolution or inference of the "wrong" topology (Posada and Crandall, 2002 ). For example, interlocus concerted evolution of the ITS region in polyploids resulting in the loss of one of the parental ITS lineages has been reported in several studies (Wendel et al., 1995 ; Brochmann et al., 1996 ; Smedmark and Eriksson, 2002 ; Popp et al., 2005 ). The frequency of concerted evolution of low-copy nuclear DNA regions, however, is much less known but has been suggested to occur among PgiC loci in Clarkia (Gottlieb and Ford, 1996 ), Adh loci in Gossypium (Millar and Dennis, 1996 ), glutamine synthetase in Pisum (Walker et al., 1995 ), and RPD2 in Silene subgenus Silene (Popp and Oxelman, 2004 ). Whereas the homogenization of the ITS region in polyploid taxa included in this study is in an advanced stage (as indicated by the within-accession monophyly of all sequenced ITS clones), strong indications of recombination and/or concerted evolution is lacking among homoeologous/paralogous low-copy nuclear DNA regions. The phylogenetic position of the RPB2 sequences from S. keiskei and S. zawadskii, however, indicate that such processes cannot be ruled out.

Conclusions
Previous studies based on cpDNA, ITS, and RNA polymerase introns (Oxelman and Lidén, 1995 ; Oxelman et al., 1997 , 2001 ; Popp and Oxelman, 2005 ) have all identified the Physolychnis s.l. clade as the one to which most North American polyploid Silene species belong. However, there has been little resolution within this clade and no support for, nor rejection of, a single origin of polyploidy in North American Silene. In this study, we have increased the number of sequenced regions as well as the taxon sampling, and still, the phylogenetic resolution is poor. Taking into account that we have included data from seven DNA regions, including four putatively unlinked low-copy nuclear genes, an approach using regions with stronger phylogenetic signal (i.e., faster evolving or longer sequences) may prove necessary to resolve the complicated relationships within Physolychnis s.l. We find, however, support for a hypothesis including at least one deep polyploidization (auto- or allopolyploidization) in the lineage leading to the North American members of Physolychnis s.l., which is separate from the event forming the Arctic tetraploid S. involucrata lineage (Popp et al., 2005 ). In addition, a previously unrecognized North American lineage composed of the diploids S. seelyi and S. williamsii and the di- and tetraploid S. menziesii is identified outside the Physolychnis s.l. The hexaploid S. hookeri is the result of an allopolyploidization between a diploid from this lineage and tetraploids from Physolychnis s.l. The tetraploid S. nivea appear to be more closely related to the European diploid S. baccifera. Thus, polyploidy appears to have originated on at least four occasions in the history of North American polyploids. We agree with Hitchcock and Maguire (1947 , p. 3) who more than half a century ago, regarding the North American Silene, wrote, "While there are readily recognizable and natural alignments, many species and groups seem to be polyphyletic and their relationships to be reticulate."

APPENDIX

Voucher information and EMBL/GenBank accessions for specimens in this study. Para- and/or homoeologous sequences are separated by a comma. Missing data are indicated with an —. Chromosome number determined specimens are underlined.

Taxon—Voucher specimen and herbarium^a; Source; Determined by; Chromosome no.; EMBL/GenBank accession nos.: psbE-petL; rps16; ITS; RPA2; RPB2; RPD2a; RPD2b.

Atocion lerchenfeldiana (Baumg.) M.Popp—Strid et al. 24188 C; Greece, Florinis; Strid; 2n = 24^b; —; AJ409061; AJ409057; AJ629281; AJ634066; AJ634153; AJ634150. Silene ajanensis (Regel) Lidén—Popp 200206 UPS; Uppsala University Botanical Garden; Berkutenko; 2n = 24^c; AJ831754; AJ831763; AJ831780; AJ634167, AJ634165, AJ634166; AJ634196, AJ634197, AJ634198; AJ831181; AJ831180. S. andersonii Clokey—Clokey 7514 UPS; USA, Nevada, Clark Co.; Oxelman; 2n = 48^d; DQ908855; DQ908809; DQ908630