The Biogeographic South-North Divide of Polygonatum (Asparagaceae Tribe Polygonateae) within Eastern Asia and Its Recent Dispersals in the Northern Hemisphere

Eastern Asia (EA) is a key region for the diversification of flowering plants in the Northern Hemisphere, but few studies have focused on the biogeographic history within EA in the context of the other northern continents. Polygonatum is an important medicinal genus widely distributed in the Northern Hemisphere with its highest species richness in EA, and it represents an excellent model for studying the evolution of biogeographic patterns in this region. Divergence time estimation was used to examine the biogeographic history of Polygonatum based on nuclear ITS and four plastid sequences (rbcL, matK, psbA–trnH and trnC–petN) from 30 Polygonatum species and 35 outgroup taxa. The ancestral area of Polygonatum and subsequent dispersal routes were inferred using Bayes-Lagrange. Polygonatum was estimated to have originated in southern EA during the middle Miocene (14.34–13.57 Ma) with subsequent south-to-north expansion in the late Miocene. Multiple intercontinental dispersal events were inferred between EA and Europe or North America, and all of them have occurred recently in the late Miocene to Pliocene. The separation of Polygonatum into the south and north lineages and their subsequent diversifications in the late Miocene supports the existence of a biogeographic divide between the northern and southern parts of EA that also coincides with the retreat and redevelopment of the arid zone in EA in the Neogene. Our results demonstrate the complexity of biogeographic history of Polygonatum in the Northern Hemisphere including early vicariance followed by frequent and recent dispersals in the Neogene.


Introduction
Eastern Asia (EA) is an important region for the biogeographic diversification of flowering plants in the Northern Hemisphere, showing one of the highest levels of species diversity and endemism [1].It has also been suggested to be the ancestral area for most EA-eastern North American disjunct lineages [2,3].The high level of species richness in EA can be attributed to secondary diversification due to habitat heterogeneity as well as a lower rate of extinctions in the Neogene, whereas the lower species diversity in North America and Europe is often explained by the hypothesis of more severe extinctions with global cooling beginning in the late Eocene to Oligocene (Paleogene) [4,5].Biogeographic studies on flowering plants in the Northern Hemisphere have largely focused on the classic EA and eastern North American disjunctions, but few studies have been devoted to exploring biogeographic patterns within EA [2,4,6].A better understanding of the biogeographic history of EA and its connections to the other continents, especially concerning the intercontinental disjunctions between EA and North America/Europe, requires the examination of diversification patterns in the Northern Hemisphere from a broader perspective.
Polygonatum (commonly known as Solomon's seal) provides an ideal model for studying the evolution of intercontinental disjunctions in the Northern Hemisphere, as well as the diversifications within EA.The genus contains approximately 60 species that are widely distributed in the warm-temperate to boreal zones of the Northern Hemisphere, and is the largest and most complex genus in the tribe Polygonateae of Asparagaceae [7][8][9].Most species of the genus are found in EA (c.50 species) with two diversification centers: one extending from the Himalayas to southeastern China and the other found in northeastern Asia [10][11][12].Apart from these areas, the other species of Polygonatum occur in moderate climate zones of North America and Europe, showing a continuous distribution between Europe and Asia and a disjunct distribution between Eurasia and North America.A better understanding of the biogeographic history of Polygonatum thus may shed light on the past floristic exchanges between different continents within the Northern Hemisphere, as well as the patterns of biogeographic migrations within EA.
Previous studies have provided insights into the evolutionary relationships of Polygonatum [12][13][14][15][16]. Particularly, a recent phylogenetic study based on four plastid markers suggested that three major lineages were well supported in Polygonatum [17].It has been suggested that the lineages showed correlations with geographic distributions, and the phylogenetic results supported the most recent classification at the sectional level.Meng et al. proposed a possible origin of Polygonatum in southern EA with subsequent dispersal into northern areas [17].
In this study, we further infer the historical biogeography of Polygonatum by employing a Bayesian uncorrelated lognormal and relaxed molecular clock approach based on our previous phylogenetic framework [17].The ancestral area of Polygonatum and subsequent dispersal routes are inferred and compared with the molecular dating results to identify the most likely explanation for the current distribution of the genus.Our study emphasizes elucidating patterns of biogeographic diversification of the genus within EA.

Ethics Statement
Polygonatum species are not included in any Eurasian and North American official list of threatened plants.No special permits were required for this study.The field studies involved no endangered or protected species and the locations of our study are provided in Table 1.Herbarium voucher specimens were deposited in the Kunming Institute of Botany, Chinese    Academy of Sciences (KUN) and National Museum of Nature and History (US).The sequences determined in this study are listed in Table 1.

Taxon sampling and sequence dataset
A total of 68 accessions representing 33 species of Polygonatum were included in the analyses (Table 1).Our sampling of the ingroup covers the geographic diversity of the genus (28 of over 50 recognized species in EA, three of the five species from Europe, and the two species from North America), and represents a wide range of morphological forms from all sections [12] and the eight series [9].The outgroup taxa were selected to include representatives from other genera in tribe Polygonateae (i.e., Heteropolygonatum, Disporopsis, and Maianthemum; 13 species) and the other members in Asparagaceae (21 species) based on previous broader analyses [18][19][20].Allium microdictyon from Amaryllidaceae was included in the outgroup to enable rooting of the tree topology and calibration of the molecular clock.
An existing dataset from two plastid coding regions (rbcL and matK) and two non-coding regions (psbA-trnH and trnC-petN) have been supplemented with new sequences for a few additional species (Table 1), using methodologies described in our earlier publications [17,21].Sequences for four plastid DNA regions were obtained from GenBank (with most of the ingroup sequences derived from our previous studies [17,21]).To provide better resolution with evidence from the nuclear genome, an internal transcribed spacer (ITS) dataset of 32 species (18 Polygonatum and 14 outgroup) was newly generated in our study.The entire ITS region was amplified and sequenced using the ITS1 (occasionally using ITS5) and ITS4 primers [22].When amplification of the ITS region was unsuccessful, two internal primers ITS2 and ITS3 [23] were used to obtain PCR products in two shorter fragment (ITS1-ITS2, ITS3-ITS4).The ITS sequences were amplified and sequenced following Meng et al. [21].Information on the voucher specimens used and the corresponding GenBank accession numbers are provided in Table 1.Sequence alignment was performed in MAFFT 6 using the default alignment parameters [24] followed by manual adjustment in PhyDE 1.0 [25].

Bayesian time estimations
Independent DNA substitution model and gamma rate heterogeneity were estimated on each molecular marker using the Akaike Information Criterion as determined by MrModelTest 2.3 (Akaike 1974; Nylander 2008).The HKY+G+I model was selected for rbcL and psbA-trnH; the GTR+G model was selected for matK and ITS; and the GTR+G+I model was selected for petN-trnC.For the molecular dating analyses, the strict molecular clock model was rejected (P <0.01) for our dataset based on a likelihood ratio test [26].We simultaneously estimated the tree topology and node ages of Polygonatum using a Bayesian relaxed clock model as implemented in BEAST 1.8.0 [27].
The molecular sequence fragments were partitioned using BEAUti 1.8.0 (as supplied within BEAST program) in order to incorporate different models of substitution rates for each partition separately.A Yule speciation tree prior was specified using an uncorrelated lognormal distributed (UCLD) relaxed clock model [28].After optimal operator adjustment as suggested by the output diagnostics from the preliminary BEAST runs, two independent Markov Chain Monte Carlos (MCMC) runs were performed for 50 million generations each and sampling every 1000 generations.Tracer 1.6 [29] was used to check for convergence between the runs.The results are considered reliable once the effective sampling size for all parameters exceeded 200, as suggested by the program manual [29].After the removal of the 10% burn-in, the sampled posterior trees were summarized to generate a maximum clade credibility tree using the program TreeAnnotator 1.8.0 [27] with a posterior probabilities (PP) limit of 0.5 and mean node heights.
To minimize the effects from the secondary-calibrations, an additional time estimation was also implemented based on a large dataset of rbcL and matK sequences which sampled 245 taxa representing all monocots to enable multiple fossil constraints (S1 Table ).We constrained the crown Pandanales, with a minimum age (89.3 Ma) corresponding to the oldest known fossil records from Mabelia and Nuhliantha [35].We also used fossil evidences to set the minimum stem age of Arecales at 70.6 Ma and the stem of Zingiberales at 72.5 Ma, as discussed in Hertweck et al. [36].A minimum age of 14.5 Ma was fixed at the stem age of Yucca based on the fossil records from Protoyucca shadishii [37,38].With the chloroplast phylogenetic framework and multiple fossil calibrations, the split between Acorus and the remaining monocots was estimated to be 131 (126-142) or 138 (132-143) Ma using the penalized likelihood or relaxed clock approaches, respectively [37].We thus constrained the root of our tree to 134.5 +/-5.0 and the maximum ages of all the fixed points as 134.5 Ma using uniform priors in our analyses.

Ancestral area reconstruction
The distribution range of Polygonatum and its sister taxa was divided into three areas, based on the geographic boundary and the presence of one or more endemic species.These areas are: A (eastern Asia, EA), B (North America), and C (Europe to central Asia).We also ran a second analysis with EA split into the southern (A) and the northern regions (D) in the biogeographic analysis, because Polygonatum has its highest species diversity and endemism in these two regions [7,[9][10][11].Eastern Asia can be subdivided into two distinct northern and southern regions [3,6,[39][40][41].Geological evidences suggest that an aridity barrier has existed from the western-most part of China to the eastern Asian coast from the Paleogene to the Miocene, and it has been thought to have acted as a climate barrier between the two regions [42].The arid belt around 35°N in China [43] was used as a boundary to define these two regions in this study.The southern region of EA (sEA) comprises southern and southeastern China, with extensions into the Himalayas, whereas the northern region of EA (nEA) comprises northeast China, Korea and Japan.
A likelihood method for biogeographic inference was applied in Lagrange 20130526 [44], which take into account genetic branch lengths and/or phylogenetic uncertainty and incorporates an explicit dispersal-extinction-cladogenesis (DEC) model of dispersal routes available at historical intervals correlating stochastic events with lineage persistence [44].We used a recently modified statistic DEC (S-DEC) analysis implemented in RASP 3.2 to reconstruct the possible ancestral ranges of Polygonatum on the phylogenetic trees [45].To account for uncertainties in phylogeny and DEC optimizations, we used 1000 randomly sampled trees from the post burn-in sampled trees derived from the first BEAST analysis.The 40 taxa in the Polygonatum-Disporopsis-Heteropolygonatum clade were included with the other taxa and pruned from the input trees.The Bayes-Lagrange analysis [45,46] was then performed in RASP on all of the trees to obtain the possible ancestral range at each node.

Results
The final datasets of rbcL, matK, psbA-trnH, trnC-petN, and ITS consist of 1447 (224 variable and 110 potentially parsimony-informative), 1864 (558 variable and 245 potentially parsimony-informative), 729 (56 variable and 21 potentially parsimony-informative), 985 (96 variable and 42 potentially parsimony-informative), and 741 (256 variable and 136 potentially parsimony-informative) nucleotide positions, respectively.The total number of characters included in the combined dataset is 5766 nucleotides.The BEAST analysis produced a wellresolved phylogeny of Polygonatum, which strongly supported the monophyly of Polygonatum and two major clades within the genus (Fig 1 1. The ancestral distributions inferred from S-DEC for internal nodes in the genus are shown in Fig 2 and Table 2.The first analysis revealed that Polygonatum originated in EA with several migrations or dispersals into other regions of the Northern Hemisphere.The second S-DEC analysis indicated two possible ancestral areas for node 1, A (sEA) or AD (both sEA and nEA), and the frequencies of occurrence of these ranges are 55% and 41%, respectively (Fig 2 ; Table 2).A similar result was found for the crown lineage of the genus (node 2, Fig 2 ), the north and the south clades (Fig 2 ; Table 2).A recent migration from sEA to central Asia was found in P. roseum with its sister taxon (P.zanlanscianense), with a divergence time of 0.61 to 0.45 Ma (node 5, Fig 2 ; Table 2).Dispersal into central Asia and Europe was also observed in P. multiflorum and its sister taxa (P.filipes and P. desoulavyi), which diverged at 2.25 Ma (node 9; Fig 2 ; Table 2).

Early divergence within EA
The BEAST analysis produced a robust phylogeny of Polygonum that is congruent with our previous phylogenetic results [17] and suggested an origin of the genus in the middle Miocene around 14.34 or 13.57Ma (Fig 2).Polygonatum together with its sister group (Disporopsis and Heteropolygonatum) are mainly restricted or endemic to EA [9,10,12] and our first S-DEC analysis also suggested EA as the ancestral area of the genus (Fig 2 ; Table 2).The results from the second S-DEC further revealed that sEA is more likely to be the ancestral range of the stem lineage of Polygonatum (A, 55%; node 1 in Fig 2 ), although EA (sEA + nEA) is also possible (AD, 41%).The favored ancestral range at the crown lineage of Polygonatum (node 2, Fig 2) is also sEA (A, 52%; Table 2), which suggested an early origin and differentiation of the genus in this area.   .However, to our knowledge, there have been very few biogeographic or phylogeographic studies to test the existence of a biogeographic divide between northern and southern EA during the Miocene and into the Pliocene, except for among Asian butternuts (Juglans section Cardiocaryon), which offer another case supporting this hypothesis [43].The genetic data from both chloroplast and nuclear regions consistently identified two clades (one northern and one southern) in the Asian butternuts, which diverged through climate-induced vicariance during the mid-Miocene [43].
Based on a sEA origin of the genus, a south to north expansion is suggested for the biogeographic origin of the northern clade within EA (AD, 62%, node 6 in Fig 2).However, plant migrations within EA have been proposed to be primarily in a north to south direction [51][52][53], which has been supported by molecular evidence from many angiosperm taxa.For example, Maianthemum is the second largest genus after Polygonatum in tribe Polygonateae and exhibits a similar distribution pattern to that of Polygonatum in the Northern Hemisphere.However, most Maianthemum species from the Himalayas to southwestern China form a well supported clade nested within several basal grades, including species from nEA and the New World [21], which may indicate a possible New World origin and a north to south migration within EA.Astilbe Buch.-Ham.ex D.Don is a well-known genus of Saxifragaceae with an early divergence in nEA, which then migrated southwards into southern and tropical Asia [54].It is also well supported that Phryma L. (Phrymaceae) diversified in nEA first and then migrated to sEA [55].In contrast, the biogeographic diversification of Polygonatum in EA represents one of the few examples showing a south to north migration scenario.Notably, some plant lineages have been reported to have southern or tropical origins with adaptations to northern and seasonal climates in EA [56], especially for north temperate taxa with tropical affinities such as Mitchella L. (Rubiaceae) [57] and Parthenocissus Planch.(Vitaceae) [58].

Recent dispersals in shaping modern distribution
Our biogeographic results revealed multiple recent migrations/exchanges among areas of EA, Europe (including central Asia), and North America from the late Miocene to the Pliocene (Fig 2).The divergence time between Eurasia (Polygonatum multiflorum, P. filipes and P. desoulavyi) and eastern North America (P.biflorum and P. pubescens) was estimated to be 2.79 to 3.31 Ma in the Pliocene (node 8, Fig 1; Table 2).The biogeographic inference supported a migration from EA into North America (Fig 2).The biogeographic disjunctions in EA and North America are common in plants and have been explained by either vicariance (via the Bering or the North Atlantic land bridges) or long-distance dispersal [2][3][4]59].The Bering land bridge was available for floristic exchanges until 3.5 to 5.0 Ma in the late Tertiary [60], and thus appears to be the most likely migration route for Polygonatum.Beringian migrations have been proposed for a number of flowering plant taxa, especially for herbs with recent disjunct ages, including Circaea L. (Onagraceae) [61], Symplocarpus R.A.Salisbury ex Nuttall (Araceae) [62], and Phryma (Phrymaceae) [55].The biogeographic disjunction between the North American and European and Asian alternate-leaved taxa suggest that there is a possibility that the North Atlantic land bridge might have been a viable pathway, given the absence of Polygonatum in most of western North America.However, the North Atlantic route has been discussed as a more likely route for thermophilic and woody taxa available before the Oligocene [63] and seems less likely in our case.
The other biogeographic connections are found between Europe to central Asia and EA, as indicated by many taxa from central Asia and Europe nested with taxa from EA (Fig 2 ).For example, Pliocene migrations or dispersals from EA into central Asia and Europe are found between P. roseum and its sister taxon (P.zanlanscianense) and between P. multiflorum and its sister taxon (P.filipes and P. desoulavyi) (Fig 2 ; Table 2).Since the late Eocene, the Turgai Strait retreated southward, coinciding with the global cooling of the Eocene-Oligocene boundary.This retreat allowed extensive biogeographic exchange between Asia and Europe [50].The wide spread of P. verticillatum, from Europe to EA, may be another example showing this biogeographic connection (Fig 2).
Recent exchanges between southern and northern regions within EA were also observed in the late Miocene and Pliocene.The first example is the northern species of P. stenophyllum that diverged from its relative P. kingianum from subtropical SW China to Indochina at 4.02 Ma (node 4, Fig 2; Table 2).The southern species of P. filipes from eastern China (node 10, Fig 2 ; Table 2) and P. arisanease from Taiwan (node 11, Fig 2 ; Table 2) show the opposite migration route (from north to south) at 1.54-1.84and 1.08 Ma with their sister species P. desoulavyi and P. involucratum from northeastern Asia, respectively.
In a word, all the disjunct dispersals or migrations in Polygonatum, either out of EA to Europe or North America, or within EA (e.g., from the southern to the northern part, or vice versa), are much recent and mostly restricted to the late Miocene to Pliocene (Fig 2 ).It seems that these recent dispersals finally led to a widespread distribution of the Polygonatum in the Northern Hemisphere.Similar pattern of recent dispersals have been found in various flowering plants which indicated the importance of the late Miocene-Pliocene in the assembly of the modern flora [64].

Conclusions
Polygonatum likely originated in sEA with a subsequent split into southern and northern lineages, supporting the existence of a biogeographic divide between the northern and the southern parts of EA during the Miocene.This biogeographic subdivision is probably explained by the retreat and redevelopment of the arid zone in EA in the Neogene.Plant migrations within EA have been proposed to have primarily occurred in a north to south direction; however, the biogeographic history of Polygonatum in EA documents primarily a south to north migration scenario.Our results also indicate that the modern distribution of Polygonatum in the Northern Hemisphere has resulted from recent regional dispersals from EA to North America, central Asia, and Europe during the late Miocene to Pliocene.
Information on taxa sampled and corresponding GenBank accession numbers.Accessions beginning with KX are new sequences published in this study; missing data are indicated by a dash.Herbarium acronyms follow http://sciweb.nybg.org..1371/journal.pone.0166134.t001

Fig 1 .
Fig 1. Maximum clade credibility tree of Polygonatum and the closely related taxa derived from a BEAST analysis.Posterior estimates of divergence times were inferred using a partitioned analysis of five combined DNA regions (ITS, psbA-trnH, trnC-petN, rbcL, and matK) and two fixes as normal age constraints (F1: 58.3 Ma; F2: 56.4 Ma).Nodes are posterior mean ages with blue node bars representing 95% highest posterior density intervals.Bayesian posterior probabilities <0.95 are indicated by grey and >0.95 by black circles.Geological epoch abbreviations: Pl, Pliocene; P, Pleistocene.Inset: Polygonatum pubescens.doi:10.1371/journal.pone.0166134.g001

Fig 2 .
Fig 2. Reconstruction of ancestral distributions in Polygonatum using Bayes-Lagrange optimizations.Pie charts at nodes represent relative frequencies of ancestral-area reconstructions.The letter and color-coding of regions for each tree is indicated on the map.Numbers 1-11 indicate nodes of interest (seeTable 2 for details).doi:10.1371/journal.pone.0166134.g002

Table 2 ,
). Polygonatum biflorum and P. pubescens from eastern North America are deeply nested within sect.Polygonatum, showing a close relationship to the Eurasian clade including P. multiflorum, P. filipes and P. desoulavyi.The Bayesian dating analyses based on two different datasets and different calibration schemes produced similar results (Table 2, see also S1Fig).An origin of the Polygonatum stem lineage was estimated to be 14.34 Ma with a 95% highest posterior density (HPD) range of 8.64-20.74Ma (node 1, Fig 1) or 13.57(9.13-18.42)Ma based on the two calibration schemes (S1 Fig).The initial diversification of the Polygonatum crown group was inferred to be 10.5 (6.21-15.54;node 2 in Fig 1) or 9.85 (6.3-13.65)Ma.The disjunction between Eurasian and North American species was estimated at 2.79 (1.46-4.38;Fig 1) or 3.31 (1.65-5.18)Ma.Details of ages for other nodes of interest are shown in Table

Table 2 . Posterior age distributions of major nodes of Polygonatum (Asparagaceae), with results of ancestral area reconstruction using the S-DEC.
Node numbers correlate with those in Fig 2. doi:10.1371/journal.pone.0166134.t002