Waldsteinia within Geum s.l. (Rosaceae): Main Aspects of Phylogeny and Speciation History

: Waldsteinia is a small plant genus inhabiting the temperate regions of the Northern Hemi-sphere. According to the latest revisions, Waldsteinia is included in Geum . We have obtained a phylogenetic reconstruction based on the nuclear (ITS) and plastid ( trnL-trnF ) DNA to understand the phylogenetic structure of Waldsteinia and its relationships with other taxa of Geum s.l. Phylogenetic analysis based on the joint ITS + trnL-trnF dataset demonstrated Waldsteinia monophyly. The phylogenetic relationships of Waldsteinia species were better explained by their geographical distribution than their morphology. Hence, Euro-Siberian, Northeast Asian, and North American phylogeographic groups were distinguished, with East Asia having been suggested as the place of Waldsteinia origin. Considering the incongruence in W. geoides (a type species) position on the plastid and nuclear DNA trees, together with the discrepancy between the species morphology and its location on the plastid DNA tree, a hybrid origin was suggested for this species. Despite the fact that the position of W. maximowicziana is still not fully resolved, we support the point of view that claims it should be separated from the W. ternata aggregate (traditionally including W. trifolia , W. ternata s.str., and W. maximowicziana ) and considered a separate species. The American W. doniana , W. fragarioides , and W. lobata belong to a single maternal lineage, but the observed genetic differences are too small to serve as a convincing argument for species segregation, so their relationships still remain unresolved.


Introduction
Rapid integration of molecular biology approaches into classic botany makes it possible to review the existing phylogenetic relationships based on the differences in DNA sequences.In these taxonomic revisions, small groups of related organisms are often considered, as their position on the tree of life is often debatable and sometimes too subjective.One of the examples can be Waldsteinia Willd., a small genus of herbaceous plants containing only a few species [1], which was joined to the genus Geum L. as a result of the latest revision [2].Waldsteinia belongs to the Rosaceae family and inhabits the temperate zone of the Northern Hemisphere [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19].Before the recent taxonomical revision, it was divided into two subgenera: Waldsteinia and Comaropsis (Rich.ex.Nestl.)Teppner based on morphological differences [4,13], with all species except for W. geoides Willd.(the type species) belonging to the latter subgenus (Table 1).Waldsteinia geoides is a rather tall (15-25 cm) herb with an erect or shortly creeping rhizome and lobed basal leaves.The most Waldsteinia is of great interest in terms of historical biogeography.Waldsteinia presumably dates back to the Neogene [20,21], having been widely distributed across the Northern Hemisphere in former times and now representing a remnant of the tertiary flora [1,4,22,23].At least a few Waldsteinia species are considered true nemoral relicts with narrow and fragmented ranges [4,10,20,23].Presently, all Waldsteinia species inhabit mixed mesophytic nemoral and hemiboreal forests in lowlands and the piedmonts of mountainous areas within the Holarctic [4,6,19,24].The present geographical range of Waldsteinia is wide.However, it exhibits a discontinuous pattern, including the following clearly detached fragments: Central and Southeastern Europe (W.geoides and W. trifolia), South Siberia (W.tanzybeica and W. ternata s.str.),Eastern Asia (W.maximowicziana), eastern North America (W.doniana, W. fragarioides, and W. lobata), and the western part of North America bounded by Idaho and Montana states (W.idahoensis).Waldsteinia tanzibeica is the species with the narrowest range.It is considered a relict and the local endemic for the Western Sayan Mountains [16].It is believed that even W. fragarioides, one of the species with the most extensive range, was much more widespread in the past than it is now.For this reason, this species may also be discussed as a relict plant from a more northern area [7,25].This discontinuous range pattern of Waldsteinia correlates well with the long-known North American-Eastern Asian floristic relationship involving migration and interchange of Asian and American species via the region of the Bering Strait, followed by the disruption of the continuous ranges because of Land Bridge disappearance and the Pleistocene glaciations [23].At the same time, no pollen records or any other reliable evidence confirming the former distribution range and migration pathways of Waldsteinia species are present.Comparatively low levels of morphological divergence together with fuzzy ploidy patterns within Waldsteinia do not allow for any clear answer either.It appears that the historical dynamics of this group have never been thoroughly studied, and only speculations and hypotheses exist on this subject.
As it was mentioned above, according to the latest revision based on the sequences of the internal transcribed spacer (ITS) and intergenic spacers of plastid DNA (trnL-trnF) of Waldsteinia together with the closely related Coluria R.Br. and Taihangia T.T.Yu & C.L.Li, are nested in the Geum genus [2].Geum s.l. in this broad sense was previously considered the Geinae Schulze-Menz subtribe, including all herbaceous perennials.In its turn, Geinae (Geum s.l.), together with the woody genera Fallugia Endl.and Sieversia Willd., formed a clade corresponding to the Colurieae Rydb.tribe [26].DNA-based phylogenetic reconstruction of Geinae [27,28] partly confirmed the hypothesis of reticulate evolution through hybridization and allopolyploidization [1,3].Taking into account both the genetic interactions within species and the paraphyly of Geum (in the size suggested by W. Gajewski [1]), J.E.E.Smedmark [2] suggested that all herbaceous lineages belonging to the Geinae subtribe should be considered as a genus Geum.Despite the clear morphological segregation of Waldsteinia and Coluria by fruit types from other Geum s.l.[1,3], the broad generic concept was accepted by many taxonomists [18,26,29] and applied in global taxonomic databases, e.g., the Catalogue of Life ( [11], accessed on 15 March 2023).
In the aforementioned phylogenetic reconstructions, only the type species of Waldsteinia and Coluria (W.geoides and C. geoides (Pall.)Bunge, respectively) were considered [2,30].However, taking into account the low statistical support for some nodes, we speculate that the issue of the position of these groups on the tree and the relationships between them cannot be considered fully resolved.We suggest that a broader sampling of Waldsteinia and Coluria for the phylogenetic analysis can improve our understanding of evolutionary relationships within Geum s.l.
Our goals in this study were to: (1) specify the relationships between Waldsteinia and other taxa in Geum s.l.based on a broader sample; (2) estimate the inner phylogenetic structure of Waldsteinia; and (3) suggest the main historical pathways of species dispersal across the continents, which have never been investigated in detail.
To address these questions, we made a molecular phylogenetic reconstruction, adding more Waldsteinia species to those included in the previously reported phylogenies [2,30] based on the ITS region of nuclear DNA and the trnL-trnF intergenic spacer of plastid DNA.We have also used these markers to shed light on Waldsteinia history, mainly focusing on the place of origin and the presumed migration patterns of this group during the late Cenozoic.To assess the role of polyploidization in speciation in Waldsteinia and during its dispersal across the continents, we collected the known data on chromosome numbers for all the species in the genus.Since the information on chromosome numbers was in some cases provided only in the original language (mainly Russian), the data on the ploidy distribution of North Asian species has so far been quite hard to obtain.Thus, our study contains the most comprehensive review of Waldsteinia chromosome numbers known to date.Moreover, except for W. tanzybeica, which has an extremely narrow range, W. ternata s.str.remained, until recently, the least studied species in terms of ploidy polymorphism among all Eurasian Waldsteinia species.Therefore, an important point was to screen additional populations of W. ternata to identify the dominant chromosome race as well as other possible types of ploidies in the species, including unknown diploids.

Plant Material Collection
For DNA sequencing, eight of nine Waldsteinia and two Coluria species were sampled.Waldsteinia ternata was collected from a natural population on the Khamar-Daban Ridge and the Eastern Sayan Mountains; W. tanzybeica from the Western Sayan Mountains; W. maximowicziana from the Lower Amur region; and Coluria geoides from the Altai Mountains.Samples of W. geoides and W. trifolia were collected from living collections in the botanical gardens.The samples of North American Waldsteinia species and Coluria henryi were collected from herbariums.The detailed list of samples is presented in Table 2.We did not have access to the narrowly distributed North American species W. idahoensis, and it was the only species that was not covered by our research.At least six individuals were collected from each natural population, and single specimens per species were taken from botanical gardens and herbaria.Each sample was kept in an individual filter paper bag (23 g•m −2 ), dried, and stored in silica gel until DNA isolation.
For new chromosome counts, the root tips of plants from natural populations were sampled and then fixed.For that, fresh tips up to 0.5 cm in length were retrieved and washed in distilled water, briefly dried on filter paper, followed by their pre-treatment in a 0.2% colchicine water solution for 2-4 h.The roots were washed of colchicine, briefly dried on filter paper, and placed in Klark's fixative (3:1 mixture of 96% ethanol-glacial [absolute] acetic acid) for at least 24 h.Samples were then washed five times and stored in 70% ethanol.

Counting Chromosome Numbers
The ploidy level was determined by a direct count of chromosomes on the metaphase plates of the root meristem, as described by M.S. Navashin [31] and L.I. Abramova and I.N.Orlova [32].The root tips were macerated in 1 M hydrochloric acid for 10-15 s at 60 • C.Then, samples were washed three to five times in distilled water to eliminate residual hydrochloric acid and stained with 1% aceto-orcein for 8 h, followed by material squashing in 45% acetic acid.Metaphase plates were observed on an Axioscope 40 (Karl Zeiss, Oberkochen, Germany) under 100× magnification and captured by an AxiCam MRc 5 digital camera.

DNA Isolation and PCR
Total DNA was isolated from silica-dried leaf tissue following the cetyltrimethylammonium bromide (CTAB) method [33], with some authors' modifications [34].
For phylogenetic reconstruction, sequences of internal transcribed spacers (ITS1 and ITS2) of nuclear DNA (ncDNA) and trnL-trnF intergenic spacers of plastid DNA (ptDNA) were used as molecular markers.The ITS region was amplified using the forward ITS1-P2 [35] and the reverse ITS4 [36] primers, complementary to the flanking regions of the 18S and 26S rDNA genes.In order to reduce PCR-mediated recombination between ITS clones and to improve PCR accuracy, a proofreading polymerase was used together with lower initial template concentrations (not more than 5 ng per reaction) and PCR cycle numbers (not more than 30 cycles) as recommended by D.J.G.Lahr and L.A.Katz [37].In particular, the PCR was performed in a reaction mixture of 20 µL containing 1x Q5 Reaction Buffer and 0.4 units of Q5 High-Fidelity DNA Polymerase (New England BioLabs, Ipswich, MA, USA) with final concentrations of 2.0 mM of MgCl 2 , 250 µM of each dNTP, and 500 nM of each primer.The conditions of amplification were 98  C for 20 s, with a final elongation of 2 min at 72 • C. In the case of herbarium samples, insufficient amplification of the ITS region using Q5 polymerase was observed because of DNA degradation probably caused by long-term storage and treatment.GoTaq Flexi DNA Polymerase (Promega, Madison, WI, USA) was used to amplify the ITS region from herbarium samples and the trnL-trnF region from all the samples in the study.Using this polymerase, the ITS region could be successfully amplified using the primers indicated above.For the trnL-trnF region amplification, the combination of forward (e) and reverse primers (f) described in the study of P. Taberlet et al. [38] was used.The reaction mixture of 20 µL contained 1× Green GoTaq Flexi Buffer, 1 unit of GoTaq polymerase, and final concentrations of 2.5 mM of MgCl 2 , 250 µM of each dNTP, and 250 nM of each primer in the final volume of 20 µL.The conditions of amplification for both DNA regions and primer pairs were 95  Amplicons were either directly purified from PCR mixtures (ITS) using the GeneJET Purification Kit (Thermo Fisher Scientific, Vilnius, Lithuania) or visualized in 1% agarose gel stained by ethidium bromide after electrophoresis and then gel-purified (trnL-trnF) using the GeneJET Gel Extraction Kit (Thermo Fisher Scientific, Vilnius, Lithuania).

Cloning and Sequencing
Purified amplicons were either directly sequenced (ITS, trnL-trnF) or additionally cloned in Escherichia coli cells (ITS only).For molecular cloning, amplicons were ligated into plasmid vectors pMiniT 2.0 (New England Biolabs, Ipswich, MA, USA) in the case of blunt-end products or into pTZ57R/T (Thermo Fisher Scientific, Vilnius, Lithuania) in the case of products with single 3 -A overhangs.Ligation was carried out according to the manufacturer's protocols using the insert-to-vector molar ratio of 3:1 in 5 µL of reaction mixture containing 12.5 ng of pMiniT 2.0 or 27.5 ng of pTZ57R/T.Further, 50 µL of One Shot TOP10 E. coli chemically competent cells (Invitrogen, Waltham, MA, USA) were heat shock transformed at 42 • C for 35 s using 2.5 µL of the ligation mixture.After transformation, cells were incubated in SOC liquid medium at 37 • C for 1.5 h and plated onto LB agar containing 100 mg•L −1 ampicillin.In the case of the pTZ57R/T vector, 40 µL (20 mg•L −1 ) of X-Gal solution were surface-spread over agar plates to enable blue-white screening for identification of the colonies carrying the insert.In the case of the pMiniT 2.0 vector carrying a toxic minigene in the cloning site, all grown colonies were considered to contain the insert.Eight colonies from each plate were picked with a sterile pipette tip and inoculated into 5 mL of liquid SOC medium containing 100 mg•L −1 of ampicillin.In the case of low transformation efficiency, all colonies were used for further analysis.Cells were grown overnight at 37 • C. Plasmids were isolated from overnight cultures using the GeneJet Plasmid Miniprep Kit (Thermo Fisher Scientific, Vilnius, Lithuania).Isolated plasmids and amplicons were Sanger sequenced in both forward and reverse directions using the BigDye Terminator Cycle Sequencing Kit version 3.1 (Applied Biosystems, Waltham, MA, USA) and M13(-20) (pTZ57R/T vector-based plasmids) or region-specific primers mentioned above (pMiniT 2.0 vector-based plasmids and PCR products) in a 3500 Genetic Analyzer (Applied Biosystems and Hitachi, Tokyo, Japan).

Sequence Alignment and Phylogenetic Analysis
Raw sequencing data were edited using SnapGene Viewer software version 2.6.2 (GSL Biotech, San Diego, CA, USA) and deposited in GenBank of the National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm.nih.gov,accessed on 10 March 2023).The ITS region was analyzed as the ITS1 and ITS2 combined set, excluding sequences for the 18S, 26S, and 5.8S genes of rRNA.Phylogenetic analysis based on the ITS ribotypes (R) was carried out in two variants: (a) using only the main ITS variants, which were found in all specimens of each Waldsteinia species using molecular cloning and corresponding to the total signal of the PCR product, and (b) using the broad sample of the ITS variants revealed in Waldsteinia by molecular cloning.The trnL-trnF region was analyzed, excluding parts of the trnL and trnF genes, considering generally different evolutionary rates of coding and noncoding parts of DNA.The ITS1 + ITS2 and the trnL-trnF datasets were analyzed both separately and combined.For joint (ITS + ptDNA) analysis, the only main ITS ribotypes were used.For analysis, we aligned our original DNA sequences of Waldsteinia and Coluria with the sequences of other Colurieae published by J.E.E.Smedmark and T. Eriksson [30] and in some other research (Table 3).Notes: 1 The species names are mainly presented as cited in the study of J.E.E.Smedmark and T. Eriksson [30], the Waldsteinia names as given in Table 2, and the other taxon names as given in the studies where the sequences were originally mentioned. 2Locality names are abbreviated as given in Table 2. 3 The identified ribotypes and plastotypes were continuously numbered for each species separately. 4In the studies in which the sequences were mentioned, 'curr.' equals to the sequences obtained in the current study. 5The species was incorrectly mentioned as Coluria elegans Cardot in the original study of J.E.E.Smedmark and T. Eriksson [30], but later recognized as not belonging to Coluria at all [27].
For W. lobata and C. henryi, only the trnL-trnF region was used in the analysis because of insufficient amplification of the ITS1-ITS2 region from the available herbarium specimens.In total, 37 sequences belonging to 36 taxa were included in the analysis of the main ITS ribotypes (R), and 53 sequences belonging to 34 taxa of ITS ribotypes obtained by molecular cloning were also included in the analysis.In the case of the trnL-trnF region (plastotypes, P), 37 sequences of 37 taxa were included in the main analysis.Some species for which the ptDNA-and ncDNA-based phylogenies were not congruent were excluded from the joint ITS + trnL-trnF phylogeny because of disturbances in the clustering (see below).Therefore, in the joint ITS + trnL-trnF analysis, only 34 sequences from 33 taxa were included.
Phylogenetic reconstructions were obtained independently by the Bayesian inference method (BI) based on the matrices combining the nucleotide alignments and binary (gaps) datasets in MrBayes version 3.2.5 [45] and the maximum likelihood method (ML) based on multiple nucleotide sequence alignments in MEGA independently.The best-fit model of nucleotide substitutions based on the lowest Bayesian Information Criterion (BIC) calculated using the "find best DNA/protein models" tool in MEGA (Neighbor-Joining tree to use and ML as a statistical method were applied as the settings) was selected and then used to perform the analysis.Nucleotide frequencies calculated using the 'find best DNA/protein models' tool were also included to optimize the models implemented in MrBayes in the case of the Bayesian inference analysis.
A BI analysis of nucleotide datasets was performed using the models implemented in MrBayes with optimized parameters to better correspond with the models used in the ML analysis (see below).The analyses were performed by specifying the model and parameters for each partition of the DNA datasets using the 'applyto' option.In particular, for the ITS dataset, the HKY-like model [46] with fixed equal stationary state frequencies, gamma distribution, or no rate variation was applied to get the K80 or K80+G models depending on the dataset (see below).For the trnL-trnF sequences, we also used a HKY-like model with the base frequencies optimized for the T92 model and fixed on values 0.35, 0.15, 0.15, and 0.35 of A, C, G, and T, respectively, with no evolutionary rate variation among the sites.Analysis of the ITS + trnL-trnF datasets was performed by applying separated parameters for ITS and trnL-trnF partitions as described above using 'applyto' option.Binary data (indels + inversion) were analyzed using the F81-like model [47] implemented in MrBayes with equal stationary state frequencies to match the JC69 model [48].
For each dataset, two simultaneous and independent Markov chain Monte Carlo (MCMC) analyses were run with four parallel chains up to 10,000,000 generations, with sampling every 100 generations and diagnostic calculations every 1000 generations.The first 25% of samples from the cold chain were discarded.The standard deviation of split frequencies below 0.01 was regarded as sufficient convergence, and that value was considered chain stationarity being reached.The fluctuations of the cold chain likelihood in the stable range were also taken into account for the estimate of reaching stationarity.The sampled trees from both analyses were pooled, and 50% majority-rule consensus trees were constructed from 62,146 (joint dataset of ITS and trnL-trnF regions) to at least 139,000 (ITS, trnL-trnF) trees to estimate clade posterior probability values (PP).The final phylogenetic trees were edited in FigTree version 1.4.3 [49].
For ML analysis of the ITS region, the Kimura 2-parameter model (K80, [50]) was applied, and for the dataset of the main ITS ribotypes, the model was additionally optimized with gamma-distribution of substitution rate variation among sites (+G, 4 categories).For ML analysis of the trnL-trnF region, the Tamura 3-parameter model (T92, [51]) with no among-site rate variation was used.For ML analysis of the ITS + trnL-trnF joint dataset, the T92 +G (4 categories) was used.For all analyses, the initial tree for the heuristic search was obtained by applying the neighbor-joining method to a matrix of pairwise distances estimated using the maximum composite likelihood (MCL) approach.All aligned positions, including the indels, were used in the analysis.A bootstrap of 1000 replicates was used as a test of the phylogeny.In this study, ML-cladograms are presented as condensed trees computed in MEGA and based on the original tree with the highest loglikelihood and collapsing branches with bootstrap confidence levels (BS) lower than 50%.The bootstrapped 50% majority-rule consensus trees for each dataset were constructed to compare topology with the highest log likelihood tree.
To assess the matrilineal genealogical relationship between Waldsteinia species and closely related taxa, a network based on trnL-trnF plastotypes was constructed using the integer neighbor-joining method (IntNJ, with a reticulation tolerance parameter equal to '0') implemented in PopART software version 1.7 [52].The network was built based on the combined nucleotide alignment and binary (gaps) matrix of Waldsteinia taxa, C. geoides, C. henryi, and T. rupestris, with S. pusilla + F. paradoxa as outgroups.

Phylogenetic Analysis Based on Nuclear DNA
The evolutionary relationships between the nuclear genomes of the species under study were inferred based on the polymorphisms in the main ITS ribotypes (Figure 1, the left side).
The subtribe Geinae (Geum s.l.) was rather poorly structured and contained only a few clades, including but not limited to the clade formed by G. schofieldii and New Zealand Oncostylus spp.(node E, BS, 76; PP, 0.99) and the clade formed by G. andicola, G. bulgaricum, and N. glacialis (node H, BS, 88; PP, 1.00).Most taxa, including C. geoides and T. rupestris, remain unresolved.In what regards Waldsteinia, the main ITS ribotype (R1) sequence for W. geoides obtained in the present study was identical to the one published by J.E.E.Smedmark and T. Eriksson [30], which explains why only one sequence was used in the main phylogeny reconstruction (Table 3).The main W. ternata ribotype (R1) was shared by the specimens from different localities (Bz1, S1, and Z1, see Table 3).For this reason, only a single sequence for this species was used in the phylogeny reconstruction.On the phylogenetic tree, all studied Waldsteinia species were grouped together, although with only moderate support (node F, BS, 66).Within the clade F, W. geoides was a sister to the group including Comaropsis species (node L, BS, 90).This latter group contained a clade combining ribotypes from European (W.trifolia) and Siberian (W.ternata and W. tanzybeica) species (node K, PP, 0.96) and a North American group including W. fragarioides and W. doniana (node M, PP, 1.00).The unresolved branch corresponded to the Northeast Asian W. maximowicziana ribotype.
Figure 1.Phylogenetic trees based on the ML 50% bootstrap confidence level condensed tree tree with the highest log likelihood): tree based on the main ITS ribotypes (R), on the left side; trnF tree, on the right side.The BI analysis was carried out independently.Thick gray lines ind additional supported nodes (PP ≥ 0.95) on the BI 50% majority-rule consensus tree compared t ML tree.The branch leading to the clade combining Rosa persica and Sanguisorba officinalis was to root the trees.Bootstrap values are indicated above the branches, and posterior probabilit the corresponding clades on the BI tree (if relevant) are indicated below the branches.The ca letters at the nodes correspond to the groups discussed in the text.Species names are given as in Table 3 ('Taxon Name' column).The geographical pattern of Waldsteinia species distributi indicated by different colors: Europe is light-green, North America is light purple, Northeast is dark blue, and South Siberia is yellow.Note: * indicates the species incorrectly mentioned a luria elegans in the original study of J.E.E.Smedmark and T. Eriksson [30], but later recogniz not being Coluria [27].
The subtribe Geinae (Geum s.l.) was rather poorly structured and contained on few clades, including but not limited to the clade formed by G. schofieldii and New Zea Oncostylus spp.(node E, BS, 76; PP, 0.99) and the clade formed by G. andicola, G. bul cum, and N. glacialis (node H, BS, 88; PP, 1.00).Most taxa, including C. geoides an  3 ('Taxon Name' column).The geographical pattern of Waldsteinia species distribution is indicated by different colors: Europe is light-green, North America is light purple, Northeast Asia is dark blue, and South Siberia is yellow.Note: * indicates the species incorrectly mentioned as Coluria elegans in the original study of J.E.E.Smedmark and T. Eriksson [30], but later recognized as not being Coluria [27].
To detect the low-copy ITS variants that may bear evidence of the probable hybridization events between Eurasian Waldsteinia suggested based on the incongruent inheritance of ncDNA and ptDNA in W. geoides (see below), molecular cloning of this DNA region was additionally carried out for these species (Figure 2).
the group including Comaropsis species (node L, BS, 90).This latter group contained a clade combining ribotypes from European (W.trifolia) and Siberian (W.ternata and W. tanzybeica) species (node K, PP, 0.96) and a North American group including W. fragarioides and W. doniana (node M, PP, 1.00).The unresolved branch corresponded to the Northeast Asian W. maximowicziana ribotype.
To detect the low-copy ITS variants that may bear evidence of the probable hybridization events between Eurasian Waldsteinia suggested based on the incongruent inheritance of ncDNA and ptDNA in W. geoides (see below), molecular cloning of this DNA region was additionally carried out for these species (Figure 2).The scale bar indicates the number of expected changes (substitutions and/or indels) per site, corresponding to the unit of branch length.Note: 1 Since R2 and R6 of W. geoides are identical in their ITS1 and ITS2 partitions (they differ by only a single mismatch in the 5.8S part), the two ribotypes were combined for the phylogenetic analysis, which was based on the ITS dataset excluding the 5.8S part (see Section 2.5).  1.The color pattern also corresponds to that in Figure 1.The asterisk indicates the collapsed clade, which includes the following taxa: Acomastylis elata, A. rossii, A. sikkimensis, Erythrocoma triflorum, Geum aleppicum Jacq., G. canadense, G. geniculatum, G. heterocarpum, G. montanum, G. reptans, G. rivale, Geum sp., G. speciosum, G. urbanum, and G. vernum.The scale bar indicates the number of expected changes (substitutions and/or indels) per site, corresponding to the unit of branch length.Note: 1 Since R2 and R6 of W. geoides are identical in their ITS1 and ITS2 partitions (they differ by only a single mismatch in the 5.8S part), the two ribotypes were combined for the phylogenetic analysis, which was based on the ITS dataset excluding the 5.8S part (see Section 2.5).
The analysis revealed additional ribotypes in the ITS region.All ribotypes in W. geoides (R1-R6) showed high affinity (Figure 2, node R, BS, 97; PP, 1.00).The phylogenetic structure of Comaropsis (node L, BS, 79; PP, 0.95) reconstructed based on the extended ribotype sample in general corresponded with the results of the analysis of the main ITS ribotypes (Figure 1).In particular, the clade combining Euro-Siberian Comaropsis (node K, PP, 0.96) ribotypes and the North American species (node M, PP, 0.99) haplogroup were presented on the tree as they were in the case with the main ribotypes.The North American species showed extremely high affinity for each other, and not more than one or two mismatches were found in the ITS region between W. doniana (R1) and the ribotypes belonging to W. fragarioides (R2 and R1, respectively).Molecular cloning revealed an additional haplogroup combining the minor ribotypes of Euro-Siberian species and Northeast Asian W. maximowicziana (node P, PP, 0.96).The branch lengths in the phylogram suggested that the main ribotype of W. maximowicziana (R1) has the shortest distance to the node L, which corresponds to the most recent common ancestor of Comaropsis.
The results of the multiple alignment also showed that there was only one site (site # 395, dark green) that may be a synapomorphy for the Waldsteinia group.However, this site may also be considered a homoplasy or even plesiomorphy if we take the outgroups (Rosa and Sanguisorba) into account (Figure 3).PP, 0.96) ribotypes and the North American species (node M, PP, 0.99) haplogroup were presented on the tree as they were in the case with the main ribotypes.The North American species showed extremely high affinity for each other, and not more than one or two mismatches were found in the ITS region between W. doniana (R1) and the ribotypes belonging to W. fragarioides (R2 and R1, respectively).Molecular cloning revealed an additional haplogroup combining the minor ribotypes of Euro-Siberian species and Northeast Asian W. maximowicziana (node P, PP, 0.96).The branch lengths in the phylogram suggested that the main ribotype of W. maximowicziana (R1) has the shortest distance to the node L, which corresponds to the most recent common ancestor of Comaropsis.
The results of the multiple alignment also showed that there was only one site (site # 395, dark green) that may be a synapomorphy for the Waldsteinia group.However, this site may also be considered a homoplasy or even plesiomorphy if we take the outgroups (Rosa and Sanguisorba) into account (Figure 3).Eriksson [30], but later recognized as not being Coluria [27].Another position (site # 486, dark green) was identical in most W. geoides ribotypes and the two ribotypes from only one of the Comaropsis species.At the same time, there were at least nine sites by which W. geoides ribotypes differed from the sequences of other Waldsteinia species (Figure 3, light green).One site seems to be an autapomorphy for W. geoides (site # 297), and the others were shared by W. geoides and different sets of taxa.

Phylogenetic Analysis Based on Plastid DNA
The phylogenetic tree based on the trnL-trnF plastid DNA region (Figure 1, the right side) was better structured than the ITS tree and had similar topology to those presented by J.E.E.Smedmark and T. Eriksson [30].Colurieae received good support (node A, BS, 100; PP, 1.00) and included well-supported clades of the subtribe Geinae (Geum s.l.) (node B, BS, 100; PP, 1.00), monophyletic Sieversia (node C, BS, 100; PP, 1.00), and Fallugia paradoxa.However, the latter was included in the common clade with Sieversia on the ITS phylogenetic tree.In turn, herbaceous Geinae (node B) consisted of a big clade comprised of a majority of species (node D, PP, 1.00) and a small, well-supported clade (node E, BS, 92; PP, 1.00) of monophyletic Oncostylus spp.together with G. andicola and G. schofieldii.Except for G. andicola, the clade defined by node E had a structure similar to the corresponding clade on the ITS tree.The clade defined by the node D consisted of (1) the well-supported Waldsteinia clade (node F, BS, 90; PP, 1.00) in which C. geoides was now found as compared with the corresponding clade on the ITS tree, (2) the clade made up by C. henryi and T. rupestris (node O, BS, 82; PP, 1.00), and (3) the well-supported clade embracing the rest of Geum s.l.(node G, BS, 92; PP, 1.00) and characterized by additional well-supported structure.
In the Waldsteinia species, for which specimens from several populations were used in the analysis, we did not observe any intraspecies polymorphism in the trnL-trnF region.In particular, the plastotypes of W. ternata specimens from all three studied localities were identical (P1, see Table 3).Three North American species, including two populations of W. fragarioides, also had identical plastotypes.Moreover, the sequences of the trnL-trnF region in W. geoides and C. geoides obtained in the present study were identical to those reported by J.E.E.Smedmark and T. Eriksson for the mentioned species [30].Given this identity of plastotypes, we used only one sequence per each Waldsteinia taxon and C. geoides to perform phylogenetic reconstructions and designated them as P1 in Table 3 for each species individually.On the phylogenetic tree, all Siberian and European Waldsteinia species formed a separate subclade (node J, BS, 95; PP, 1.00), to which the European W. geoides was added in comparison with the corresponding clade on the ITS tree (the left side, node K).Inside the clade, W. ternata and W. tanzybeica combined on the BI (node N, PP, 0.95) but not on the ML tree.Three North American plastotypes being identical were grouped in a single clade (node M, PP, 0.99), similar to that on the ITS tree.The relationships between Northeast Asian W. maximowicziana, C. geoides, and the aforementioned Waldsteinia groups of species remained unresolved.However, we have found that these species tended to form a single clade together with the North American plastotypes (node M', BS, 73).

Combined Phylogenetic Analysis
The topology of the phylogenetic tree built based on the joint dataset (ITS + trnL-trnF) was well-structured and similar to the topology of the trnL-trnF tree.In particular, subtribe Geinae (Geum s.l.) consisted of a well-supported Waldsteinia clade (clade I, BS, 96; PP, 1.00), a clade containing most of the Geum s.l.species (clade IV, BS, 76; PP, 1.00), and a clade comprising of Oncostylus and G. schofieldii (clade V, BS, 98; PP, 1.00) (Figure 4).We excluded G. andicola from the analysis of the joint dataset because of the serious discrepancy between its position on the ITS and the trnL-trnF tree, as it showed affinity for both a subgroup within the clade IV and the group V (see Figure 1), which disturbed the normal clusterization of all other species.Coluria geoides (branch II) appeared to be sister to Waldsteinia, according to the BI analysis (PP, 0.95).The second analyzed Coluria species, C. henryi, was missing on the tree because of the insufficient amplification of the ITS1-ITS2 region.The relationships between T. rupestris (branch III) and any of the aforementioned taxa and clades remained unresolved because of the poor support observed for the I+II+III group.Fallugia (branch VI) and Sieversia (clade VII) were outgroup taxa for Geinae, and together with the latter they made the Colurieae.
The clade I, which contained all Waldsteinia species, was monophyletic and well supported by both ML and BI analyses.Within Waldsteinia, the group combining the European and Siberian taxa was well supported (BS, 78; PP, 1.00) and formed by a nested clade of three Comaropsis species (W.tanzybeica, W. ternata, and W. trifolia) and W. geoides as a sister.Within the latter clade, two Siberian species (W.ternata and W. tanzybeica) clustered together (BS, 75; PP, 0.99).The North American haplotypes H1 (R1 + P1) and H2 (R2 + P1) of W. fragarioides and the haplotype of W. doniana formed a common clade with high support (BS, 86; PP, 1.00).The relationships between Northeast Asian W. maximowicziana and the aforementioned groups within Waldsteinia remained unresolved.However, we found that this species tends to assemble with the North American haplotypes with high support (BS, 82) if the ML consensus tree instead of the highest log likelihood tree is built (Figure 4 Waldsteinia, according to the BI analysis (PP, 0.95).The second analyzed Coluria species, C. henryi, was missing on the tree because of the insufficient amplification of the ITS1-ITS2 region.The relationships between T. rupestris (branch III) and any of the aforementioned taxa and clades remained unresolved because of the poor support observed for the I+II+III group.Fallugia (branch VI) and Sieversia (clade VII) were outgroup taxa for Geinae, and together with the latter they made the Colurieae.[30]).The species names are given as they are presented in Figure 1.The color pattern also corresponds to that in Figure 1.Note: * indicates the species incorrectly mentioned as Coluria elegans in the original study of J.E.E.Smedmark and T. Eriksson [30], but later recognized as not being Coluria [27].
The clade I, which contained all Waldsteinia species, was monophyletic and well supported by both ML and BI analyses.Within Waldsteinia, the group combining the European and Siberian taxa was well supported (BS, 78; PP, 1.00) and formed by a nested clade of three Comaropsis species (W.tanzybeica, W. ternata, and W. trifolia) and W. geoides as a sister.Within the latter clade, two Siberian species (W.ternata and W. tanzybeica) clustered together (BS, 75; PP, 0.99).The North American haplotypes H1 (R1 + P1) and H2 (R2 + P1)  Phylogenetic tree based on the ML 50% bootstrap confidence level condensed tree constructed using the joint ITS + trnL-trnF dataset.The dashed line shows the extra node with high support on the ML bootstrap 50% majority-rule consensus tree.The BI analysis was carried out independently.The branch leading to the clade combining Rosa persica and Sanguisorba officinalis was used to root the trees.Bootstrap values are shown above the branches, and posterior probabilities of the corresponding clades on the BI tree (if relevant) are shown below the branches.The clades and branches discussed in the text are indicated by Roman numbers.Ploidy levels are presented in the column on the right side (including the literature data [30]).The species names are given as they are presented in Figure 1.The color pattern also corresponds to that in Figure 1.Note: * indicates the species incorrectly mentioned as Coluria elegans in the original study of J.E.E.Smedmark and T. Eriksson [30], but later recognized as not being Coluria [27].

Chromosome Numbers
The original data ('curr.')and the summary of the published information on chromosome numbers are presented in Table 4 as well as in Figure 4 and the next one.The basic chromosome number for all Waldsteinia species is a constant: x = 7.Most species investigated more than once exhibited the existence of different ploidy levels.Waldsteinia geoides and W. tanizbeica were exceptions that apparently were stable diploids (2x; 2n = 14).In addition to this species, diploids were revealed in two other species, i.e., W. doniana and W. lobata.However, because they were studied only once, their stability as diploids still needs to be confirmed.Diploids also occurred in the East Asian W. maximowicziana and in the North American W. fragarioides.
In addition to diploids, tetraploid (4x; 2n = 28) races were found to be one of the most common within the genus, especially for Eurasian species.This included W. ternata, W. trifolia (simultaneously with the pentaploid race), W. maximowicziana (simultaneously with the diploid race), and North American W. idahoensis.Hexaploids (6x; 2n = 42) occurred occasionally in different species and were reported mainly by H. Teppner [4,13,54].The repeated attempts to find hexaploids for W. ternata, once revealed in previous research on the Khamar-Daban Ridge [53], were unsuccessful.All new counts showed only tetraploids (Table 4).A similar situation was also observed for W. maximowicziana.Hexaploid variants of this species could not be consistently found in the Russian Far East [62].Chromosome races, including triploids (3x) and pentaploids (5x), were found to be codominant for W. fragarioides and W. trifolia, respectively.

Geographical Patterns of Plastotype Distribution
The phylogeographical analysis of Waldsteinia performed using the phylogenetic relationships based on the trnL-trnF plasotypes indicated maternal inheritance (Figure 5).
The plastotype network (Figure 5a) was prepared based on the well-supported clade F combining Waldsteinia species and C. geoides on the plastid DNA phylogram (Figure 5b, BS, 90; PP, 1.00), the sequences of closely related species (T.rupestris and C. henryi), and outgroup taxa (F.paradoxa and S. pusilla).The plastotype of W. maximowicziana appeared to be that belonging to the most recent common maternal ancestor of C. geoides and Waldsteinia species.The plastotypes of C. geoides, North American (W.doniana, W. fragarioides, and W. lobata), and European (W.geoides and W. trifolia) species of Waldsteinia were derivatives of that of W. maximowicziana.The youngest plastotypes within Waldsteinia belonged to the Siberian taxa (W.tanzybeica and W. ternata).The current geographical location of plastotypes is schematically shown in Figure 5c, where arrows indicate our hypothetical model of the most probable scenario of intercontinental species migrations (see Discussion).The repeated attempts to find hexaploids for W. ternata, once revealed in previous research on the Khamar-Daban Ridge [53], were unsuccessful.All new counts showed only tetra ploids (Table 4).A similar situation was also observed for W. maximowicziana.Hexaploid variants of this species could not be consistently found in the Russian Far East [62].Chro mosome races, including triploids (3x) and pentaploids (5x), were found to be codominan for W. fragarioides and W. trifolia, respectively.

Geographical Patterns of Plastotype Distribution
The phylogeographical analysis of Waldsteinia performed using the phylogenetic re lationships based on the trnL-trnF plasotypes indicated maternal inheritance (Figure 5).

The Updated Phylogenetic Reconstruction of Colurieae
Our reconstruction based on the joint ITS + trnL-trnF dataset (Figure 4) did not generally contradict the results obtained by J.E.E.Smedmark and T. Eriksson [30].However, there were some differences in Fallugia and Sieversia relationships.In the study of J.E.E.Smedmark and T. Eriksson [30], Fallugia was sister to the combined clade of Sieversia and Geinae (Geum s.l.).In our study, the relationships between Fallugia and Sieversia appeared to be unresolved in the final reconstruction.However, the ITS and trnL-trnF-based phylogenetic trees revealed different interactions between these two groups, but neither of them matched the result obtained in the work of J.E.E.Smedmark and T. Eriksson [30].We believe that the observed incongruence may be explained by the differences in taxon sampling between the two studies, considering the broad sampling of outgroup taxa in the study of J.E.E.Smedmark and T. Eriksson [30].Within Geinae, the clade V (Figure 4; bootstrap, 98; PP, 1.00) mainly formed by Oncostylus was sister to the well-supported clade (I+II+III+IV; BS, 73; PP, 1.00), including all the other representatives of the group.The relationship between the Oncostylus clade and the other Geinae was not fully resolved on the phylogenetic tree obtained by J.E.E.Smedmark and T. Eriksson ( [30]: Figure 5, node K).In both studies, nesting of North American G. schofieldii in the New Zealand Oncostylus clade (V) was observed.The Oncostylus clade additionally included G. andicola on the plastid DNA tree but not on the nuclear DNA tree (Figure 1).That incongruence between the ncDNA and ptDNA trees was first detected and discussed by J.E.E.Smedmark and T. Eriksson [30], who suggested that it might be evidence of the hybrid origin of G. andicola.We believe that using additional data for the aforementioned taxa as well as their relatives to avoid the misidentified species effect could clarify the situation within the Oncostylus lineage.Nevertheless, our results showed that the New Zealand Oncostylus and probably a few American Geum species linked by a common origin stood out from the other Geinae and were a sister group to the latter.
A sister group to Oncostylus consisted of two well-supported clades and two partially resolved taxa.The largest Geinae clade (Figure 4, Clade IV) combined the majority of Geum species.This clade is well supported in both our study and the studies of J.E.E.Smedmark and T. Eriksson [30].In fact, we suggest recognizing this clade as Geum s.str.

(see below).
In what regards Waldsteinia, Coluria, and Tahangia, in J.E.E.Smedmark and T. Eriksson's study [30], they formed a well-supported clade on the BI tree (PP, 0.99) but showed poor BS support (53) based on joint (ITS + trnL-trnF) datasets.Using the expanded sampling of Waldsteinia species did not allow us to confirm the relationship between Waldsteinia + C. geoides and Tahangia as sister groups due to the low support (Figure 4).Thus, the position of T. rupestris (branch III) within Geum s.l./Geinae remains unresolved.The relationship between Waldsteinia and Coluria is discussed in the next paragraph.

Relationships between Waldsteinia and Coluria
Previously, Waldsteinia and Coluria were declared to be closely related genera but separated from the rest of Geum [1].The common feature of Waldsteinia and Coluria that distinguished them from the other Geum s.l.species is a style jointed at the base that is entirely deciduous in fruits [1,3,30].At the same time, it should be recognized that there were attempts to describe Waldsteinia and Coluria as Geum even before the study of J.E.E.Smedmark [2].In particular, H. Baillon suggested that floral characteristics and fruit types are features not reliable enough to separate these two groups from Geum [76].The difference between Waldsteinia and Coluria lies mainly in the leaf shape, with leaves being lobed (W.geoides, W. idahoensis, and W. lobata) or 3-foliolate/deeply lobed (W.doniana, W. fragarioides, W. maximowicziana, W. tanzybeica, W. ternata, and W. trifolia) in Waldsteinia and pinnate in Coluria [1,2].The differences also manifest in the few carpels, quickly drying stamens, and a cone-or bowl-shaped receptacle in Waldsteinia, and the numerous carpels, persistent stamens, and a sacciform receptacle in Coluria [6,[77][78][79][80].
Our analysis based on the joint dataset (ITS + trnL-trnF) showed that C. geoides is likely to be a sister to Waldsteinia (Figure 4; PP, 0.95).At the same time, a dual position of Coluria on the plastid DNA tree was demonstrated, i.e., C. geoides nested in one clade together with Waldsteinia, whereas C. henryi formed a single clade with Taihangia (Figures 1 and 5).The observation that Coluria nested in two different clades on the plastid DNA tree was unexpected because, before the present study, the monophyly of this group was never questioned.Due to the lack of ITS1-ITS2 region sequences for C. henryi, the patrilineal lineage for this species and, therefore, the monophyly of Coluria currently cannot be established.
The position of C. geoides still remains incompletely resolved.In particular, the species was nested in the Waldsteinia clade on the plastid DNA tree, and it also tended to form a common clade with North American and Northeast Asian Waldsteinia species.At the same time, the nesting of C. geoides within Waldsteinia could not be confirmed yet based on the nuclear DNA tree (Figure 1, the left side; Figure 2).The C. geoides appeared to be a sister to Waldsteinia, according to the BI analysis of the joint tree (Figure 4).The inconsistent position of C. geoides on the nuclear and plastid DNA trees can be caused either by incomplete lineage sorting with Waldsteinia or introgression following a hybridization.
The possible polyphyly of Coluria and the incompletely resolved position of C. geoides complicate the understanding of the relationships between Coluria and other closely related groups.Although our data indicated close affinity between Waldsteinia and Coluria, the precise phylogenetic relationship between these groups can hardly be deciphered without including other species of Coluria and applying additional genetic methods.

Phylogenetic Structure of Waldsteinia
Waldsteinia geoides, which belongs to the Waldsteinia subgenus, significantly differs from the other species in the genus by its morphology.At the same time, according to the plastid DNA phylogenetic tree (Figure 1, the right side) and joint DNA analysis (Figure 4), W. geoides clustered together with the Euro-Siberian group of Comaropsis (W.tanzybeica, W. ternata, and W. trifolia).Nesting W. geoides in the aforementioned group was also confirmed by our previous results based on the trnH-psbA intergenic spacer of plastid DNA [81].In such a way, considering these data, Comaropsis does not appear to be a monophyletic group because of W. geoides interposition.On the other hand, according to the phylogenetic analysis based on the nuclear DNA, the position of W. geoides remained unresolved because of the low support levels, but it was outside the Comaropsis group anyway (Figure 1, the left side; Figure 2).No reliable synapomorphies in the ITS sequences that would justify nesting W. geoides in the clade of Euro-Siberian species were found.In particular, two sites in the multiple alignment may be claimed as synapomorphies for the two haplogroups of Euro-Siberian species (Figures 1 and 2, nodes K and P'), but none of them was shared with W. geoides (Figure 3, sites #78 and #181).Moreover, only one site was found (site #395, dark green) to be a possible synapomorphy for the Waldsteinia group as a whole.At the same time, at least nine positions found in the ITS ribotypes of W. geoides differed from the sequences of the rest of the Waldsteinia species but were shared by W. geoides and several Geum species.
In such a way, an incongruence was found between W. geoides morphology and its position on phylogenetic trees based on ptDNA and ncDNA.The morphological and nuclear DNA data showed that W. geoides is a sister to Comaropsis.On the other hand, plastid DNA and joint DNA analyses convincingly showed that W. geoides nested in one of the well-supported groups within Comaropsis, embracing Euro-Siberian species.This fact may suggest a possible introgression following a hybridization event between one of the Waldsteinia species from the Comaropsis clade and another, yet unknown, paternal ancestor from Colurieae.Our hypothesis mainly builds on the results of other studies, which demonstrated that similar mismatch patterns between nuclear and plastid DNA trees for the species appeared due to hybridization events [82][83][84][85][86][87].Another fact that may be related to the hybrid origin of W. geoides is the presence of apomictic embryonic sacs in addition to the meiotic ones [17,73].Most studies devoted to apomixis converge on the recognition that almost all apomictic plants are polyploids and/or hybrids [88].Being diploid, the W. geoides nuclear genome could have appeared as a result of homoploid hybridization without increasing the ploidy level [89][90][91] or chromosome number reductions during the 'diploidtetraploid-dihaploid cycle' following hybridization [92][93][94].Phylogenetic analysis based on plastid DNA polymorphism allows us to suggest a candidate for the maternal ancestor of W. geoides.It might be some ancestral form of European Comaropsis, which is now presented by W. trifolia and which has a common plastotype with W. geoides (Figure 5).
To clarify the possible intertaxa hybridization events within Waldsteinia, with an emphasis on W. geoides, and to define the possible paternal ancestors, we have conducted molecular cloning of the ITS region.We expected to get a picture similar to that described for the ITS2 region by M. Zarrei et al. [95].The study showed nesting of the ITS ribotypes of the Crataegus (Rosaceae) species with a hybrid origin in different clades belonging to presumed parents.Our attempts to find any clear patterns of hybridization proved to be unsuccessful.As the result of our study, the identified W. geoides ITS ribotypes showed high affinity to each other (Figure 2, node R), but none combined with the ribotypes of other Geum s.l. or nested in the Comaropsis clade.Only a single position in the alignment (Figure 3, site # 486, dark green) was identical in most W. geoides ribotypes and the two minor ribotypes of one Comaropsis species and might have possibly been introgressed due to hybridization.
Additionally, several substitutions in one of the ITS ribotypes belonging to W. geoides (R5) demonstrated a different ancestry compared to the other ribotypes (R1-R4, R6).Those substitutions may also turn out to be the remains of the former introgression.Nevertheless, we need to admit that the identified single nucleotide substitutions in the ITS region cannot be reliable evidence of the former hybridization and can hardly help in determining the probable ancestors.We are inclined to associate the absence of clear patterns of parental ribotypes in the ITS sequences of W. geoides with the concept of concerted evolution, where intra-individual variability in the multicopy DNA units is generally low or absent due to unequal crossing over, high-frequency gene conversion, and large deletions [96].Thus, if the ribosomal DNA of the hybrid was subsequently homogenized through concerted evolution in the direction of one of the ancestral genomes, the information concerning the other parent would be lost [30,96].In this case, the present genotype of W. geoides may be the result of chromosome rearrangements and DNA recombination, which have led to genetic homogenization [91,97] and the preferential retention of the paternal nuclear genome rather than the maternal one.Although we adhere to the interspecies hybrid hypothesis for W. geoides origin with a Comaropsis maternal ancestor and a non-Waldsteinia paternal ancestor, it cannot be ruled out that the genotype of the studied W. geoides specimens could have been formed by complete chloroplast capture from one of the Comaropsis species or through other types of introgressions based on backcrossing events with the paternal parent.Nonetheless, stable introgression might be complicated by different ploidy types in W. geoides and other modern Geinae (Geum s.l.) species, which are mostly polyploid (Figure 4).Furthermore, hybridization events between W. geoides and other modern Geum s.l.species have never been described, although this does not mean that this is not possible in principle [3].Moreover, incomplete lineage sorting is a known phenomenon that may lead to the observed mismatches in phylogenies.However, our previous results [81] based on the additional ptDNA marker (trnH-psbA) revealed a ptDNA phylogeny that is similar to that presented in this study.In particular, W. geoides sharing the common plastotype with W. trifolia nested in the clade of Euro-Siberian species.Hence, the question of the probable hybrid origin of W. geoides requires further special study, e.g., complete genome sequencing, or targeted sequencing using next-generation technologies.
Nevertheless, based on our data on plastotype diversity, the monophyly of the subgenus Comaropsis and the objectivity of dividing Waldsteinia into two subgenera [4,13] are becoming controversial.According to our reconstruction, the phylogenetic relationships of Waldsteinia taxa are better correlated with geographical patterns than with their morphology.Within Waldsteinia, the following geographically separated groups are also well-supported by genetic data: Euro-Siberian (W.geoides, W. tanzibeica, W. ternata, and W. trifolia), East Asian (W.maximowicziana), and North American (W.doniana, W. fragarioides, and W. lobata).The two latter groups could be assembled together on the ML consensus tree, and W. maximowicziana appears to be closer to the North American than the Euro-Siberian Comaropsis species.Moreover, our data have shown that W. ternata s.l. in the commonly accepted sense (i.e., including W. maximowicziana, W. ternata s.str., and W. trifolia) seems to be a paraphyletic or even polyphyletic group considering W. geoides nesting in this group and W. maximowicziana clustering with the North American species, according to the results of one of the performed analyses (Figure 4).The non-monophyletic status of the aforementioned group was also suggested by our previous results based on the trnH-psbA spacer [81].In such a way, despite the fact that the position of W. maximowicziana continues to be unresolved, we support the point of view that W. maximowicziana should be separated from the W. ternata aggregate and considered a separate species.The appropriate nomenclature combination for W. maximowicziana was validly published by N.S.Probatova [98].Although W. tanzibeica has never been considered an infraspecific taxon of W. ternata, its close relationship is evident from morphological similarity [10] and genetic analysis [81].Our present analysis also confirms the high affinity between these two species.This fact was expected because of the distribution of W. tanzibeica; its local endemic nature in the Western Sayan Mountains lies within the distribution range of W. ternata s.str.[10].
The genetic relationships among North American Waldsteinia are not clear.We have not detected any genetic evidence that W. doniana and W. lobata are more closely related to each other than to W. fragarioides [18] or that W. doniana is a subordinate taxon of W. fragarioides [17].Our present and previously obtained data [81] indicate that all three species have a common plastotype, i.e., a common matrilineal lineage, but the phylogenetic relationships between these species were not possible to establish from the available datasets.The observed differences between W. fragarioides and W. doniana in the ITS region are also too minor to suggest species segregation into different taxa.However, we believe that more complex research with additional genetic markers and data from North American populations (including W. idahoensis) is needed to make any taxonomic conclusions.

Possible Scenarios of Waldsteinia's History
The level of genetic distance based on the ITS phylogenetic tree indicated that the ribotype of W. maximowicziana (R1) was the nearest to the common ancestor of the Comaropsis group (Figure 2).According to the plastid DNA (Figure 5a,b, and [81]), W. maximowicziana carried the maternal ancestral plastotype for C. geoides and the entire Waldsteinia.Considering the current distribution range of W. maximowicziana (Table 1), we suggest an East Asian origin for the genus Waldsteinia and its subsequent speciation and distribution toward Europe and North America (Figure 5c).
Considering the estimated age of Colurieae, 38-49.6 million years old [39], and the tertiary age of some Waldsteinia species [4,10,[20][21][22], the probable time of Waldsteinia's origin as a genus can be suggested as the Late Oligocene or Miocene, when the climate, even at high latitudes, was warm and wet and mixed mesophytic forests were more or less continuously distributed throughout the Northern Hemisphere [99,100].These conditions should be favorable for warm-and moisture-dependent Waldsteinia [10,20,101].During that time, intercontinental migrations were possible through the Bering Land Bridge, which connected northeastern Asia and western North America throughout the late Cretaceous and the Neogene [99,102,103].The scenario of intercontinental dispersal events from the Old World to the New World, in particular, was assessed by Xiang and Soltis [104] and appeared to be the most common among temperate species with intercontinental disjunctions in the Northern Hemisphere [105][106][107][108][109].
The significant climate cooling approximately 15 Ma and the disappearance of the Bering Land Bridge about 5.5-5.3Ma [110,111] simulated the divergence of Asian and American Waldsteinia populations and their distribution through the continents.
In North America, further progression of climatic cooling and aridification culminated in the Pleistocene and led to phytogeographic barriers around the arid center of the continent [112].As a result, elements of temperate and subtropical mesophytic biotas survived in refugial regions of the south-eastern (Atlantic) and western (Pacific) parts of North America [111].Evidently, climate change was the reason for the split in the distribution range of Waldsteinia.Currently, W. idahoensis is endemic to Idaho and Montana and is the only representative of the genus in the Pacific refugium, whereas the Atlantic complex is represented by three species, W. fragarioides, W. doniana, and W. lobata, distributed from north to south from the Great Lakes to the southern piedmont of the Appalachian Mountains [19].
Concurrently, Eurasian Waldsteinia was spreading from East Asia to the west.The expansion of ancestral Waldsteinia to Europe might occur through continental Siberia following the outlines of the trans-Palearctic broad-leaved zone [20,21,101,113].Climate change and the increase in continentality were the reasons for the fragmentation of the broad-leaved forest zone into European and East Asian parts and the formation of a gap in continental Eastern Siberia and Central Asia in the Quaternary [101].After all, those events led to the disjunction of Waldsteinia into European (W.trifolia), South Siberian (W.tanzibeica, W. ternata), and East Asian (W.maximowicziana) fragments.Since then, the distant populations have developed independently, and their genotypes have accumulated nucleotide substitutions and DNA rearrangements at different rates.
The mountains of southern Siberia are of great importance because of the surviving mesophylous remnants of broad-leaved forests during the Pleistocene [101].In the case of Waldsteinia, such refugia are located on the piedmonts of the Khamar-Daban Ridge, the Eastern Sayan Mountains, and the Western Sayan Mountains, which are inhabited by W. ternata and W. tanzibeica, which are treated as tertiary (Neogene) nemoral relicts [9,10,20,21,114].A status of tertiary relict was also suggested for European W. trifolia, characterized by a highly fragmented range in the southeastern Alps and the Carpathian Arc [4,22].Considering that W. trifolia occurs mostly in the territory of modern Romania, the southern part of the Carpathians could be considered the main Pleistocene refugium for this species.Herwig Teppner [4] also defined some possible refugia in the southeastern Alps near Carinthia.
Although the European plastotype was evidently more ancient than the Siberian one, there was no other way to reach Europe than through continental Siberia.The youngest plastotype was found in the South Siberian populations, which may indicate the highest value of genome plasticity among Waldsteinia species or genetic drift.In particular, a dramatic reduction in the population size of Waldsteinia both in Europe and in southern Siberia during the Pleistocene cooling and glaciation could lead to accidental haplotype extinction or fixation resulting from the bottleneck and/or founder effects.By that, we may explain the fact that European species retained the plastotype that is ancestral to the Siberian species.
Despite the fact that populations of Comaropsis species have been isolated from each other since at least the Early Pleistocene, the modern plants morphologically [1,3,4,8,14,17,78] and genetically [81,115] are still very similar.The morphological similarity fits well with the fact that tertiary relict floras on separate continents have a shared history, leading to a slow morphological evolution ('stasis') in many taxonomic groups as compared to their genetic differentiation [112].However, further investigation is needed to understand whether a low genetic variance among Comaropsis species is due to a prevalence of vegetative propagation or apomixis over sexual one [4].We supposed that the observed moderate resolution in Comaropsis phylogeny (e.g., the unresolved position of W. maximowicziana) together with the clustering of ribotypes (Figure 2) or plastotypes [81] from a single species in non-monophyletic groups may be due to the progenitor/derivative situation where the ancestor (W.maximowicziana in our case) and in the short term, the derivative species do not show clearly monophyletic patterns.Moreover, ancestral populations may often show a lack of autapomorphies, which makes it problematic to designate them as separate species.Monophyly may be obtained over time, however, via the sorting and extinction of lineages [116].
Considering the incongruence in the position of W. geoides in plastid and nuclear DNA phylogenetic trees, we speculated that the species may have a hybrid origin with a matrilineal ancestor belonging to a European representative of the subgenus Comaropsis and an unknown patrilineal ancestor from Colurieae (see Section 4.3).European Comaropsis is currently represented only by W. trifolia, which has the same plastotype as W. geoides (Figure 5).The modern distribution ranges of both species overlap in the Carpathian Mountains, which could be the center of W. geoides origin.The age of W. geoides is hard to estimate based on available data; however, it had a chance to arise when Comaropsis relatives reached Europe.This assumption is also relevant because W. geoides has never been considered a tertiary relict in contrast to the European W. trifolia [4,22].These facts allow us to consider W. geoides as the youngest Waldsteinia species of the late Cenozoic age.
Thus, we suggest the scenario of a one-directional dispersal of Waldsteinia to the New World that was accompanied by a one-directional dispersal through Asia towards Europe (Figure 5c).This obvious dispersal pattern is known for many other East Asian angiosperms [104,108].The hypothesis is also confirmed by the distribution patterns of races with different ploidy levels.
Diploids in most cases represent the ancestral chromosome race in species with variable ploidy levels.Thus, the presence of diploids in W. maximowicziana fits well into the concept of the East Asian origin of the genus.At the same time, diploids frequently occur in all fragments of the disjunct Waldsteinia range.Furthermore, taking into account that polyploidization in plants happens much more often than the polyploid-diploid transition, we assume that the intra-and transcontinental dispersal of the group occurred predominantly with diploids (Figure 5c).In particular, we suppose that the transcontinental (trans-Pacific) dissemination of the ancient Waldsteinia from East Asia to North America may have been driven by diploids, with their subsequent and independent polyploidization and probable hybridization within continents.The appearance of the modern polyploid races in loco may also be indicated by the different dominant ploidy levels in the East Asian and North American populations.Specifically, the 4x race dominates in East Asia, while the 3x race dominates in North America.The same chromosome races, which are rarely found on either one (4x in North America) or both (6x) continents, could arise independently of each other.The possibility of such independent convergent evolution of different ploidy levels within the genus was first suggested by H. Teppner et al. [17].
The western (trans-Eurasian) pathway of Waldsteinia dispersal was also most likely related to diploids.The distribution by only one race is also confirmed by matrilineal plastotype evolution (Figure 5a,c), which definitely indicated that both tetraploid W. ternata and diploid W. tanzybeica had the same 'South Siberian' plastotype that arose from the common East Asian ancestry.However, the parallel distribution of the tetraploid race cannot be excluded (Figure 5c).Tetraploids are dominant in South Siberia (W.ternata) and codominant in Europe (W.trifolia), whereas diploids are very rare in Siberia (locally endemic W. tanzibeica) and represented in Europe by a single species, W. geoides, from a monotypic subgenus.If the homoploid origin of W. geoides took place, feasibly, the ancient diploid Comaropsis species could have disappeared after it was involved in hybridization.For another European species, W. trifolia, diploids were not known, and its present chromosome races (4x, 5x, and 6x) could have originated either from the common diploid ancestor with W. geoides or from the tetraploid Comaropsis ancestor that arrived from South Siberia.Noteworthy, hexaploids occur rarely in species from different geographical regions [4,13,53,54] and cannot be reliable criteria for relationship establishment between species and populations.
The relationships between chromosome races of the same species are unclear because of poor morphological differences [54].The distribution pattern of chromosome races was revealed only for W. trifolia [4,54].To obtain a complete picture of the phylogenetic relationships between Waldsteinia species, the study of DNA differences in clones with the same and different ploidy levels from matching and distant regions needs to be conducted.

Geum Sensu Lato vs. Geum Sensu Stricto
Despite the fact that the phylogenetic reconstruction of Colurieae proposed by J.E.E.Smedmark et al. [28,30] was well justified, in our view, the taxonomic decision to combine all herbaceous species into Geum [2,26] was too generalized.The challenges in Geum s.l.taxonomy are mainly associated with the complex interaction of processes such as hybridization, polyploidization, and gametophytic apomixis, which are widespread in the Rosaceae as a whole [84,[117][118][119].Some works have suggested a broad generic concept as the most appropriate solution to the convoluted nomenclature of groups with such kinds of relationships [29].Debates between 'lumpers' and 'splitters' are as old as taxonomy itself, and, apparently, it should be recognized that the taxonomic level of a particular taxon can be arbitrary and seems to be less important than the actual establishment of its monophyly.It should also be considered that clear monophyly patterns, at least at the species level, maybe occasionally obtained only with time via the sorting and extinction of lineages [116].The loss of generic status by small groups cannot always be justified since their unique 'identity' is ignored in these cases.It looks especially lamentably in the case of groups with well-defined morphology and evolutionary history, e.g., Waldsteinia.And although we generally support the allopolyploidy monophyly arguments in favor of Geinae (or Geum according to the new concept), we still believe that the suprageneric status of this clade better reflects the present genetic structure and evolutionary perspectives of internal groups, such as their divergence [116].In addition, gene flow between extant species has been limited for a long time.For example, the experimental attempts to perform intergeneric crosses between Waldsteinia, Coluria, and different Geum species were unsuccessful [1,3].Although the importance of reticular evolution and hybridization events throughout the early evolution of Geinae can hardly be ignored, we believe that the present relationships between the existing taxa are more important for making taxonomic decisions.
Despite the question of whether to consider Geinae as a suprageneric clade or as a genus Geum s.l. with a subgeneric structure being only a taxonomic issue and looking secondary, we would still suggest considering the genus Geum in a partly restricted sense combining most of the herbaceous species starting with G. reptans and other taxa that share more recent common ancestors with it than with Waldsteinia, Coluria, and Taihangia on the joint phylogenetic tree (Figure 4; clade IV).The designated clade was also well supported on the plastid DNA phylogenetic tree, both in our study (Figure 1, the right side, node G, BS, 92; PP, 1.00) and in the study of J.E.E.Smedmark and T. Eriksson [30] (BS, 99; PP, 1.00).The high affinity of species in the proposed sense of Geum was also shown by the phylogenetic reconstruction based on the nuclear low-copy granule-bound starch synthase (GBSSI) gene sequences [28].Thus, no evidence exists that at present any species belonging to the clade Geum s.str.have a part of the genome that is closer to Waldsteinia, Coluria, or Taihangia than to the other species of the mentioned clade.
If we follow the 'split' conception, New Zealand Oncostylus and probably several American Geum species (Figure 4, clade V, BS, 98, PP, 1.00) linked by the common origin should be separated from Geum s.str.The distinct status of this group is even more evident because of its sister position to the clade embracing the other Geinae (Geum s.l.) (Figure 4, clade IV, BS, 73; PP, 1.00).However, the confusing merging of Geum schofieldii and G. andicola (on the plastid DNA tree) with Oncostylus needs further investigation.In this view, G. andicola could probably be an intergeneric hybrid [30].
The relationships between the three remaining groups within Geinae (Coluria, Taihangia, and Waldsteinia) are still not entirely resolved.Coluria appears to be a polyphyletic group based on its matrilineal lineage and has common maternal ancestors with both Waldsteinia and Taihangia (Figure 1, the right side).In this case, we suggest a hybrid origin for some Coluria species or incomplete lineage sorting with other groups (see above).C. geoides clustering together with Waldsteinia on the plastid DNA phylogenetic tree and its predicted position as a sister to Waldsteinia according to the results obtained with the joint DNA dataset may mean that these groups may be more closely related to each other than to other groups of Geum.Taihangia rupestris, a generic endemic of China, still exhibits an unresolved position (in relation to Geum s.str., Coluria, and Waldsteinia) on the joint phylogenetic tree (see Figure 4).However, the position of T. rupestris on the plastid DNA phylogenetic tree suggests a certain affinity to Coluria (Figure 1, the right side, node O).Therefore, the discussed issues concerning the origin and the relationships between Coluria, Taihangia, and Waldsteinia could hardly be solved without including other representatives of Coluria in the analysis and applying additional genetic methods.We understand that our study has raised more questions than it has answered, and we admit that our present results are still preliminary and do not definitively demonstrate the taxonomic status of the groups included in the analysis.

Conclusions
Based on the present phylogenetic structure of Geum, Waldsteinia, and other closely related taxa, we believe that excluding Waldsteinia together with Coluria, Taihangia, and Oncostylus from Geum s.l. is more justified than 'lumping' them together into a single polymorphic genus.Even though, in our opinion, our phylogenetic reconstruction has resolved with greater precision the relationships between several groups, in fact, our taxonomic proposals are only based on a different 'chopping pattern' of the tree than the one previously suggested by J.E.E.Smedmark and T. Eriksson [30].The earlier division of Waldsteinia into two subgenera (Waldsteinia and Comaropsis) is now getting controversial since the subgenera appear to be non-monophyletic, as is indicated by the genetic data, mainly ptDNA.We tend to explain the morphological differences (on which the division into subgenera is based) between W. geoides (the only representative of the type subgenus) and Comaropsis species by the presumed hybrid origin of the former one.Our 'hybrid' hypothesis is based on the discrepancy observed between W. geoides positions on the plastid and nuclear DNA trees, along with a mismatch between the species morphology and its position on the plastid DNA tree.We also suggest that this species may have a matrilineal ancestor belonging to the European representatives of the Comaropsis subgenus and an unknown patrilineal ancestor from other Colurieae.Our data convincingly show that the phylogenetic relationships of Waldsteinia species are better explained by their geographical distribution than by the morphological differences between them.The Euro-Siberian (W.geoides, W. tanzibeica, W. ternata, and W. trifolia), Northeast Asian (W.maximowicziana), and North American (W.doniana, W. fragarioides, and W. lobata) phylogeographic groups within Waldsteinia were identified, and East Asia was proposed to be considered the place of the genus's origin.Three Waldsteinia species from eastern North America (W.doniana, W. fragarioides, and W. lobata) belong to a single maternal lineage, but the observed genetic differences are too small to serve as a convincing argument for the species segregation, so their relationships still remain unresolved.Despite the fact that the position of W. maximowicziana continues to be not entirely resolved, we support the point of view that W. maximowicziana should be separated from the W. ternata aggregate (the one including W. trifolia, W. ternata s.str., and W. maximowicziana) and considered a separate species.Our suggestion is based on the possible paraphyly or even polyphyly of the W. ternata aggregate considering W. geoides nesting in this group and on W. maximowicziana clustering with the North American Comaropsis species according to the results of one of the performed analyses.Our data has also shown that W. tanzibeica may belong to the aforementioned species aggregate because of its high affinity with W. ternata s.str.

Figure 1 .
Figure1.Phylogenetic trees based on the ML 50% bootstrap confidence level condensed tree (the tree with the highest log likelihood): tree based on the main ITS ribotypes (R), on the left side; trnL-trnF tree, on the right side.The BI analysis was carried out independently.Thick gray lines indicate additional supported nodes (PP ≥ 0.95) on the BI 50% majority-rule consensus tree compared to the ML tree.The branch leading to the clade combining Rosa persica and Sanguisorba officinalis was used to root the trees.Bootstrap values are indicated above the branches, and posterior probabilities of the corresponding clades on the BI tree (if relevant) are indicated below the branches.The capital letters at the nodes correspond to the groups discussed in the text.Species names are given as listed in Table3('Taxon Name' column).The geographical pattern of Waldsteinia species distribution is indicated by different colors: Europe is light-green, North America is light purple, Northeast Asia is dark blue, and South Siberia is yellow.Note: * indicates the species incorrectly mentioned as Coluria elegans in the original study of J.E.E.Smedmark and T. Eriksson[30], but later recognized as not being Coluria[27].

Figure 2 .
Figure 2. BI phylogram of multiple ITS ribotypes (R) of Waldsteinia and related species based on the 50% majority-rule consensus tree.The ITS ribotypes used in the associated phylogenetic analyses (Figures 1 and subsequent ones) are shown in bold.Posterior probabilities are indicated below the branches and bootstrap values of the respective clades on the independent ML tree (if relevant) above the branches.Capital letters at the nodes indicate the groups (including the collapsed clades) discussed in the text and correspond to those in Figure1.The color pattern also corresponds to that in Figure1.The asterisk indicates the collapsed clade, which includes the following taxa: Acomastylis elata, A. rossii, A. sikkimensis, Erythrocoma triflorum, Geum aleppicum Jacq., G. canadense, G. geniculatum, G. heterocarpum, G. montanum, G. reptans, G. rivale, Geum sp., G. speciosum, G. urbanum, and G. vernum.The scale bar indicates the number of expected changes (substitutions and/or indels) per site, corresponding to the unit of branch length.Note:1 Since R2 and R6 of W. geoides are identical in their ITS1 and ITS2 partitions (they differ by only a single mismatch in the 5.8S part), the two ribotypes were combined for the phylogenetic analysis, which was based on the ITS dataset excluding the 5.8S part (see Section 2.5).

Figure 2 .
Figure 2. BI phylogram of multiple ITS ribotypes (R) ofWaldsteinia and related species based on the 50% majority-rule consensus tree.The ITS ribotypes used in the associated phylogenetic analyses (Figure1and subsequent ones) are shown in bold.Posterior probabilities are indicated below the branches and bootstrap values of the respective clades on the independent ML tree (if relevant) above the branches.Capital letters at the nodes indicate the groups (including the collapsed clades) discussed in the text and correspond to those in Figure1.The color pattern also corresponds to that in Figure1.The asterisk indicates the collapsed clade, which includes the following taxa: Acomastylis elata, A. rossii, A. sikkimensis, Erythrocoma triflorum, Geum aleppicum Jacq., G. canadense, G. geniculatum, G. heterocarpum, G. montanum, G. reptans, G. rivale, Geum sp., G. speciosum, G. urbanum, and G. vernum.The scale bar indicates the number of expected changes (substitutions and/or indels) per site, corresponding to the unit of branch length.Note:1 Since R2 and R6 of W. geoides are identical in their ITS1 and ITS2 partitions (they differ by only a single mismatch in the 5.8S part), the two ribotypes were combined for the phylogenetic analysis, which was based on the ITS dataset excluding the 5.8S part (see Section 2.5).

Figure 3 .
Figure 3. Parts of the multiple alignment of the ITS ribotype (R) sequences showing the mismatches between W. geoides and other species.ITS region variants used for the main phylogenetic reconstruction (Figure 1) are given in bold.The number of ribotype copies found by molecular cloning is provided in brackets.Dots indicate the same base as in the reference sequence (W.geoides, R1); characters indicate the differences.Numbers indicate the positions in the multiple alignments, including the indel regions.The indels themselves and missing data are mostly not shown in the picture.Light green indicates the mismatch positions between the ribotypes belonging to W. geoides and other Waldsteinia species.The dark green indicates possible synapomorphies for W. geoides and other Waldsteinia species.The species names are provided as they are presented in Figure 1.Note: * indicates the species incorrectly mentioned as Coluria elegans in the original study of J.E.E.Smedmark and T.Eriksson[30], but later recognized as not being Coluria[27].

Figure 3 .
Figure 3. Parts of the multiple alignment of the ITS ribotype (R) sequences showing the mismatches between W. geoides and other species.ITS region variants used for the main phylogenetic reconstruction (Figure 1) are given in bold.The number of ribotype copies found by molecular cloning is provided in brackets.Dots indicate the same base as in the reference sequence (W.geoides, R1); characters indicate the differences.Numbers indicate the positions in the multiple alignments, including the indel regions.The indels themselves and missing data are mostly not shown in the picture.Light green indicates the mismatch positions between the ribotypes belonging to W. geoides and other Waldsteinia species.The dark green indicates possible synapomorphies for W. geoides and other Waldsteinia species.The species names are provided as they are presented in Figure 1.Note:* indicates the species incorrectly mentioned as Coluria elegans in the original study of J.E.E.Smedmark and T. Eriksson[30], but later recognized as not being Coluria[27].

Figure 4 .
Figure 4. Phylogenetic tree based on the ML 50% bootstrap confidence level condensed tree constructed using the joint ITS + trnL-trnF dataset.The dashed line shows the extra node with high support on the ML bootstrap 50% majority-rule consensus tree.The BI analysis was carried out independently.The branch leading to the clade combining Rosa persica and Sanguisorba officinalis was used to root the trees.Bootstrap values are shown above the branches, and posterior probabilities of the corresponding clades on the BI tree (if relevant) are shown below the branches.The clades and branches discussed in the text are indicated by Roman numbers.Ploidy levels are presented in the column on the right side (including the literature data[30]).The species names are given as they are presented in Figure1.The color pattern also corresponds to that in Figure1.Note: * indicates the species incorrectly mentioned as Coluria elegans in the original study of J.E.E.Smedmark and T. Eriksson[30], but later recognized as not being Coluria[27].

Figure 4 .
Figure 4. Phylogenetic tree based on the ML 50% bootstrap confidence level condensed tree constructed using the joint ITS + trnL-trnF dataset.The dashed line shows the extra node with high support on the ML bootstrap 50% majority-rule consensus tree.The BI analysis was carried out independently.The branch leading to the clade combining Rosa persica and Sanguisorba officinalis was used to root the trees.Bootstrap values are shown above the branches, and posterior probabilities of the corresponding clades on the BI tree (if relevant) are shown below the branches.The clades and branches discussed in the text are indicated by Roman numbers.Ploidy levels are presented in the column on the right side (including the literature data[30]).The species names are given as they are presented in Figure1.The color pattern also corresponds to that in Figure1.Note: * indicates the species incorrectly mentioned as Coluria elegans in the original study of J.E.E.Smedmark and T. Eriksson[30], but later recognized as not being Coluria[27].

Figure 5 .Figure 5 .
Figure 5. Plastotype network of Waldsteinia and several related species constructed using the IntN method (a); phylogram resulting from the BI analysis of the trnL-trnF region (b); and suggested scenarios of the historical pathways of Waldsteinia species dispersal across the continents based on their present phylogeographical structure (c).(a): Different haplotypes are presented as colored cir cles, with species abbreviations provided inside the circles, and are connected by lines, where hatch marks correspond to the number of evolutionary events (substitutions and/or indels).The color pat tern of Waldsteinia taxa corresponds to that in Figure 1, where unlabeled dark blue dots show net work vertices.Modern ploidy levels corresponding to each plastotype are indicated above the cir cles.(b): The BI phylogenetic tree; the posterior clade probabilities are shown above the branches and bootstrap values of the respective clades on the independent ML tree are below the branches Capital letters indicate the nodes (including the collapsed clades) discussed in the text and corre spond to those in Figure 1.The color pattern also corresponds to that in Figure 1.The scale bar indicates the number of expected changes (substitutions and/or indels) per site per unit of branch length.(c): The curved gray arrows schematically show the supposed distribution directions of spe cies carrying different plastotypes within Holarctic during the Cenozoic.The dashed arrow shows the hypothetical pathway that is not based on molecular genetic data.The dominant chromosome races participating in Waldsteinia dispersal are indicated by curved arrows.The straight lines sche matically indicate the branches on the phylogenetic tree corresponding to the closely related taxa (Coluria henryi and Taihangia rupestris).Notes: The species names were abbreviated by the following Figure 5. Plastotype network of Waldsteinia and several related species constructed using the IntNJ method (a); phylogram resulting from the BI analysis of the trnL-trnF region (b); and suggested scenarios of the historical pathways of Waldsteinia species dispersal across the continents based on their present phylogeographical structure (c).(a): Different haplotypes are presented as colored circles, with species abbreviations provided inside the circles, and are connected by lines, where hatch marks correspond to the number of evolutionary events (substitutions and/or indels).The color pattern of Waldsteinia taxa corresponds to that in Figure 1, where unlabeled dark blue dots show network vertices.Modern ploidy levels corresponding to each plastotype are indicated above the circles.(b): The BI phylogenetic tree; the posterior clade probabilities are shown above the branches, and bootstrap values of the respective clades on the independent ML tree are below the branches.Capital letters indicate the nodes (including the collapsed clades) discussed in the text and correspond to those in Figure 1.The color pattern also corresponds to that in Figure 1.The scale bar indicates the number of expected changes (substitutions and/or indels) per site per unit of branch length.(c): The curved gray arrows schematically show the supposed distribution directions of species carrying different plastotypes within Holarctic during the Cenozoic.The dashed arrow shows the hypothetical pathway that is not based on molecular genetic data.The dominant chromosome races participating in Waldsteinia dispersal are indicated by curved arrows.The straight lines schematically indicate the branches on the phylogenetic tree corresponding to the closely related taxa (Coluria henryi and Taihangia rupestris).Notes: The species names were abbreviated by the following: Cg, Coluria geoides; Ch, C. henryi; Fp, Fallugia paradoxa; Sp, Sieversia pusilla; Tr, Taihangia rupestris; Wb, Waldsteinia lobata; Wd, W. doniana; Wf, W. fragarioides; Wg, W. geoides; Wi, W. idahoensis; Wl, W. trifolia; Wm, W. maximowicziana; Wt, W. ternata; Wz, W. tanzybeica.

Table 2 .
The list of DNA samples used for the phylogenetic reconstructions.

Table 3 .
The taxa and DNA sequences used for the phylogenetic reconstructions.