Phylogenetic reassessment of tribe Anemoneae (Ranunculaceae): Non-monophyly of Anemone s.l. revealed by plastid datasets

Morphological and molecular evidence strongly supported the monophyly of tribe Anemoneae DC.; however, phylogenetic relationships among genera of this tribe have still not been fully resolved. In this study, we sampled 120 specimens representing 82 taxa of tribe Anemoneae. One nuclear ribosomal internal transcribed spacer (nrITS) and six plastid markers (atpB-rbcL, matK, psbA-trnQ, rpoB-trnC, rbcL and rps16) were amplified and sequenced. Both Maximum likelihood and Bayesian inference methods were used to reconstruct phylogenies for this tribe. Individual datasets supported all traditional genera as monophyletic, except Anemone and Clematis that were polyphyletic and paraphyletic, respectively, and revealed that the seven single-gene datasets can be split into two groups, i.e. nrITS + atpB-rbcL and the remaining five plastid markers. The combined nrITS + atpB-rbcL dataset recovered monophyly of subtribes Anemoninae (i.e. Anemone s.l.) and Clematidinae (including Anemoclema), respectively. However, the concatenated plastid dataset showed that one group of subtribes Anemoninae (Hepatica and Anemone spp. from subgenus Anemonidium) close to the clade Clematis s.l. + Anemoclema. Our results strongly supported a close relationship between Anemoclema and Clematis s.l., which included Archiclematis and Naravelia. Non-monophyly of Anemone s.l. using the plastid dataset indicates to revise as two genera, new Anemone s.l. (including Pulsatilla, Barneoudia, Oreithales and Knowltonia), Hepatica (corresponding to Anemone subgenus Anemonidium).

To date, phylogenetic analyses of Anemone s.l. are mainly based on nuclear ribosomal internal transcribed spacers (nrITS) and plastid atpB-rcbL intergenic spacer, because the two regions show high rates of variable and parsimony-informative sites, and they are powerful to resolve phylogenies at the infrageneric level [24-26, 40, 41]. Monophyly of Anemone s.l. was strongly supported in these studies. However, the monophyly of Anemone s.l. was not resolved in other studies using other regions, but these were with limited samples [4,39,42]. In addition, phylogenetic relationship between subtribes Anemoninae and Clematidinae is inferred just using nrITS and atpB-rcbL datasets [24,38,41]. In this study, we extensively sampled Hepatica and Pulsatilla in subtribe Anemoninae, as well as Anemoclema and Naravelia in subtribe Clematidinae, and we sequenced nrITS, atpB-rbcL, and five additional plastid regions (matK, rbcL, psbA-trnQ, rpoB-trnC and rps16). For the atpB-rbcL region, we only used the intergenic spacer, so there is no overlapping with the rbcL gene. Based on comprehensive phylogenetic analyses, we sought to: (1) infer the phylogenetic relationships among genera within the two subtribes; (2) reevaluate the monophyly of Anemone s.l.; and (3) resolve the phylogenetic placement of Anemoclema and Naravelia.

Phylogenetic analyses
New sequences were assembled, aligned, and adjusted using Geneious 7.0 [43]. Aligned matrices of the seven DNA regions were firstly analyzed separately, then plastid matrices were concatenated using SequenceMatrix 1.7 [44]. The DNA matrix of seven DNA regions was deposit at Figshare (DOI: 10.6084/m9.figshare.4774753). No nucleotide positions were excluded from analyses. According to the topologies of single marker datasets, monophyly of Anemone s.l. was recovered in nrITS and atpB-rbcL datasets. Previous studies using the nrITS + atpB-rbcL dataset well resolved the monophyly of Anemone s.l., therefore, the two datasets were combined in this study. To combine the plastid datasets, we did two treatments: one has all six plastid regions (i.e. six-plastid-gene dataset), and the second has five plastid regions without atpB-rbcL (i.e. five-plastid-gene dataset). Topological incongruence among nrITS, atpB-rbcL, nrITS + atpB-rbcL and five plastid datasets was investigated using the approximately unbiased (AU) test [45] and the Shimodaira-Hasegawa (SH) test [46]. Topologies were constrained using Mesquite 3.2 [47]. The SH and AU tests were performed using PAUP 4.0 [48].
Maximum likelihood (ML) analyses were conducted using RAxML [49]. These analyses used the GTR substitution model with gamma-distributed rate heterogeneity among sites and the proportion of invariable sites estimated from the dataset. The multiple-gene datasets were partitioned by genes. Support values for the node and clade were estimated from 1000 bootstrap replicates. ML bootstrap support (BS) values ! 70% were considered well supported, and BS < 50 were seen as an indication of nonsupport. Bayesian inference (BI) analyses was performed using MrBayes 3.2.6 [50], with DNA substitution models selected for each gene partition by the Bayesian information criterion (BIC) using jModeltest 2.0 [51]. Markov Chain Monte Carlo (MCMC) analyses were run in MrBayes for 10,000,000 generations for each dataset. The BI analyses were started with a random tree and sampled one tree every 1000 generations. The first 20% of the trees were discarded as burn-in, and the remaining trees were used to generate a majority-rule consensus tree. Internodes with posterior probability values (PP) ! 0.95 were considered as statistically significant. The best-fit model of nucleotide substitution for the seven DNA regions is listed in Table 1.

Phylogenetic analyses of single DNA marker
Phylogenetic relationships among genera resulting from of the seven DNA markers analyzed separately using ML and BI methods are presented in S1 Fig. As for Barneoudia, Knowltonia, and Oreithales only nrITS and atpB-rbcL sequences were available from GenBank, the three genera were not included in phylogenetic analyses of the other five plastid datasets. In addition, all samples of Hepatica failed to amplify for the rps16 region. Topologies of the seven datasets were divided into two types. The first type included nrITS and atpB-rbcL datasets, which supported the splitting of tribe Anemoneae into two clades, i.e. Clematis s.l. (including Archiclematis and Naravelia) + Anemoclema and Anemone s.l. (including Barneoudia, Hepatica, Knowltonia, Oreithales, and Pulsatilla). The clade Clematis s.l. + Anemoclema corresponds to a newly defined subtribe Clematidinae by Zhang et al. [38], and the clade Anemone s.l. corresponds to subtribe Anemoninae. The other type of dataset was the other five plastid regions. All five trees showed that Anemoclema was sister to Clematis s.l., while Anemone s.l. was paraphyletic. Overall, species of Anemone were divided into two clades in all seven trees, with one clade (Anemone I) close to Pulsatilla (not with atpB-rbcL), and another clade (Anemone II) close to Hepatica (but not with the nrITS and rps16 datasets). There is no species sharing between the two Anemone clades. In the clade Clematis s.l., six datasets of single marker, except matK dataset, strongly supported the monophyly of Naravelia.
Three traditional genera (Hepatica, Naravelia and Pulsatilla) were strongly supported as monophyletic, and all six samples of Anemoclema formed one clade. Clematis, including Naravelia, was paraphyletic; and Anemone was polyphyletic, separated into two subclades, Anemone

Phylogenetic analyses of the six-plastid-gene dataset
Topology of the six-plastid-gene dataset (Fig 3) recovered the same relationship of three major clades using five-plastid-gene dataset. However, two weakly incongruent clades between BI and ML trees were found in the clade Clade 1 and clade 2 were well supported as sister (BS/PP = 76/0.97). In clade 1, Anenoclema was sister to Clematis s.l. Then, C. alternata and C. aethusifolia were sister to remaining presented under branches, and posterior probability of BI above branches. Topological incongruence between ML and BI trees is indicated by colored nodes/branches, and topology of BI tree shows by dash lines with posterior probability in square bracket under branches.

Topological comparisons and dataset combinations
The SH and UA tests for constrained relationships using nrITS, atpB-rbcL, nrITS + atpB-rbcL and five-plastid-gene datasets are presented in Table 2. We only found that the unconstrained topology of the five-plastid dataset showed significant difference in both SH and AU tests when compared with the constraint nrITS topology, and in AU test when compared with the constrained atpB-rbcL topology. For combined analyses, the atpB-rbcL dataset was more suitable for concatenating with nrITS than the five-plastid-gene dataset, and nrITS dataset and the five-plastid-gene dataset should be analyzed separately.

Phylogenetic incongruence among datasets
Monophyly of tribe Anemoneae was strongly supported by seven single marker datasets (S1 Fig). Within tribe Anemoneae, five major groups were recognized in all seven datasets, six major groups in the six datasets (except rps16 dataset), and nine major groups in both nrITS and atpB-rbcL datasets. Species of Barneoudia, Knowltonia and Oreithales were absent from the psbA-trnQ, rbcL rpoB-trnC and rps16 datasets, and Hepatica from the rps16 dataset because we failed to generate sequences from the samples, or there was no sequence in GenBank. For the five datasets, the remaining major groups were well supported as monophyletic. Overall, phylogenetic resolution of the backbone was poor using the single marker datasets (S1 Fig), and relationships among groups were incongruent. Based on the similarity of topologies, and the SH and AU tests, the seven datasets tended to split in two groups: one group included nrITS and atpB-rbcL, and the other group included the remaining five plastid datasets. We confirmed that taxa sampling had no effect on backbone relationships obtained with either the nrITS or atpB-rbcL datasets, because clades Clematis + Anemoclema and Anemone s.l. were also supported when Barneoudia, Knowltonia and Oreithales were excluded (S3 Fig). Generally, the conflicting topologies in plants are found between nuclear and plastid datasets [52-the majority rule consensus of ML tree. Bootstrap values of ML are presented above branches, and posterior probability of BI under branches. Topological incongruence between ML and BI trees is indicated by colored nodes/branches, and topology of BI tree shows by dashed lines with posterior probability in square bracket under branches.
https://doi.org/10.1371/journal.pone.0174792.g002  56]. In tribe Anemoneae, the topologies based on the nrITS and atpB-rbcL datasets were consistent [26,41,57]. However, topological incongruence was found between the five-plastiddataset and atpB-rbcL suggested that plastid genes may be evolved independently in tribe Anenomeae. In a large-scale analysis, Zeng et al. [58] have documented that topologies showed differences between the single copy region genes and inverted repeat region genes, because genes in the inverted repeated region are more conservative than those in the single copy region. Meanwhile, the coding genes are more conservative than the non-coding genes. In this study, six plastid genes were not powerful enough to clarify this question. Based on published plastomes of Ranunculaceae, at least two large rearrangements (rps4 CDS and trnH tRNA-rps16 CDS) were found tribe Anenomeae, which has been detected using restriction enzymes [59]. As more and more chloroplast genomes are published [60], comparative analyses of whole chloroplast genomes may help to understand the evolutionary history of plastid genes.
Compared to the single marker datasets, phylogenetic resolution was significantly improved when the nrITS dataset was combined with the atpB-rbcL dataset, and five plastid datasets were concatenated. Meanwhile, phylogenetic conflicts between the two combined datasets became significant (AU test: P = 0.0588). In the topology, monophyly of subtribe Anemoninae was well supported by the nrITS + atpB-rbcL dataset; whereas subtribe Anemoninae was paraphyletic using the plastid dataset. In addition, support values for the clades 1 + 2 were not increased yet when the atpB-rbcL dataset was combined with the other five plastid datasets.
that of the majority rule consensus of ML tree. Bootstrap values of ML are presented above branches, and posterior probability of BI under branches. Topological incongruence between ML and BI trees is indicated by colored nodes/branches, and topology of BI tree shows by dash lines with posterior probability in square bracket under branches.
https://doi.org/10.1371/journal.pone.0174792.g003 Table 2. Summary of the Shimodaira-Hasegawa (SH) and the approximately unbiased (AU) tests. P values were less than 0.05 in boldface. Log likelihood scores for the unconstrained analysis are given, as well as the difference in log likelihood scores between the unconstrained and the constraint topologies (@).

Phylogeny of tribe Anemoneae
The AU test indicated that the atpB-rbcL and the five-plastid gene datasets were tended to analyze separately.

Phylogenetic placement of Anemoclema and Naravelia
Anemoclema is upgraded as an independent genus primarily based morphological characters [28]. The flowers of Anemoclema glaucifolium resemble to Anemone, and its persistent styles with hairs to Pulsatilla [28]. Therefore, Anemoclema should belong to Anemone s.l or subtribe Anemoninae. However, preliminary phylogenetic analyses show that Anemoclema is the sister to Clematis + Naravelia, while Anemone and Pulsatilla form another clade [4]. Due to the study of Wang et al. [4] focusing on resolving the relationships of Ranunculales, Anemoclema and the other three genera (Anemone, Clematis and Pulsatilla) only included one sample/species. Subsequently, Zhang et al. [38] sampled multiple species of Anemone, Clematis, and Pulsatilla, and three individuals of Anemoclema, and they sequenced the nrITS and atpB-rbcL regions. Their results strongly support the transfer of Anemoclema to subtribe Clematidinae.
In this study, we sampled six individuals of Anemoclema representing its whole distribution regions in southwestern China, and 18 taxa of Pulsatilla, and sequenced nrITS and six plastid regions. Phylogenetic analyses revealed that seven single marker datasets and three combined datasets all recovered the clade Anemoclema + Clematis s.l. Therefore, Anemoclema is clearly excluded from Anemone s.l. or subtribe Anemoninae as a distinctive genus that is sister to Clematis s.l. Morphological delimitation of the genus Clematis is very controversial, several small genera have been proposed [12]. Of these genera, Naravelia is widely accepted as an independent genus [2,3,18,21,29,61], although it is subsumed within Clematis s.l. by some taxonomists [14,22,62]. Naravelia is separated from Clematis as an independent genus by having narrow and long petals and leaflet tendrils. Traditionally, Clematis section Atragene (L.) DC. is supposed to have petals. However, floral development has shown that petals in Clematis macropetala are initiated from stamen primordia, and then antherless filaments expand to petal-like staminodia [63]. Therefore, we suggested that the "petals" of Naravelia may be the narrow and long staminodia.
Miikeda at al. [19] firstly revealed that Naravelia was nested with Clematis, then N. laurifolia and N. eichleri formed a clade. Subsequent studies [20,24,37,39] confirmed the result of Miikeda at al. [19] because they used same/similar dataset of Naravelia from GenBank, or sequenced the same species. Based on our extensive sampling of Naravelia, we recovered the monophyly of Naravelia (including N. eichleri), which should be treated as a subgenus or section. Naravelia eichleri was originally placed in Naravelia by Tamura [18] based on fruiting and imperfect specimens, then Tamura [64] himself transferred it to Clematis after he collected fertile specimens without petals and leaflet tendrils. However, the sequenced sample of N. eichleri was collected by Tamura from Thailand [19]. In the present study, we demonstrated that N. eichleri was included the Naravelia group. The nrITS + atpB-rbcL dataset strongly supported N. eichleri as sister to remaining species of Naravelia, indicating that species with petal-like staminodia and leaflet tendrils may be derived from an ancient without staminodia and leaflet tendrils only once.
There is no sample of Metanemone included in any phylogenetic analyses, so the systematic placement of this genus remains unclear.
Anemone s.l. has been suggested to include Barneoudia, Hepatica, Knowltonia, Oreithales, and Pulsatilla, because this group is strongly supported as monophyletic by the combined nrITS and atpB-rbcL dataset [25,26,41,65]. Our phylogenetic analyses also recovered the monophyly of Anemone s.l. using nrITS + atpB-rbcL dataset. Based on 26S rDNA and other three plastid markers (matK, rbcL, trnL-F), however, Wang et al. [4] revealed that the clade Pulsatilla + Anemone was nested with Clematis s.l., and that Hepatica was the sister to them. This conflicting result might be caused by limited sampling from tribe Anemoneae [26]. Nevertheless, the concatenated plastid dataset with extensive sampling of this tribe also revealed the paraphyly of Anenome s.l. in this study. Therefore, Barneoudia, Knowltonia, Oreithales, and Pulsatilla in clade 2 are strongly supported to subsume with Anemone s.l. [26], whereas Hepatica and Anemone II in clade 3 tends to be treated as an independent genus, i.e. Hepatica. The clade 3 corresponds to subgenus Anemonidium (Spach) Juz. [23,26], which is characterized by a chromosome number equal to 7; achenes are globose (usually wider than long) and nearly glabrous (or with short, straight hairs) with thick walls; and each head may yield no more than 50 achenes.

Recommendations for reclassification of tribe Anemoneae
Morphologically, two subtribes have been recognized in tribe Anemoneae [1,21]. Subtribe Anemoninae is characterized by erect herbs with basal leaves and imbricate sepals, and subtribe Clematidinae by lianas with opposite leaves (except Archiclematis alternata) and valvate sepals. However, Anemoclema, an Anemoninae-type genus, tends to transfer to subtribe Clematidinae [38]. When this treatment was adopted, diagnostic characters between subtribes Anemoninae and Clematidinae became confused. Moreover, the concatenated plastid datasets have demonstrated that subtribe Anemoninae is paraphyletic. Therefore, the subtribe rank in this tribe becomes inapplicable, and it should be abolished in future classifications.

Conclusions
Monophyly of tribe Anemoneae has been demonstrated by several studies [4,[8][9][10][11]. However, phylogenetic relationship among genera was not full resolved, due to limited DNA markers were used, and/or incomplete genera samplings were analyzed. In this study, we included nine of ten recognized genera in tribe Anemoneae (only Metanemone was not sampled) and used one nuclear and six plastid markers to reconstruct a comprehensive phylogeny of tribe Anemoneae. Based on evaluation of topological incongruence, seven DNA markers were classified as two groups, nrITS and atpB-rbcL, and the remaining five plastid genes. The combined datasets resolved tribe Anemoneae as three major clades: clade 1 included Anemoclema and Clematis s.l. (including Archiclematis and Naravelia), clades 2 and 3 corresponded to Anemone subgenus Anemone (including Barneoudia, Knowltonia, Oreithales, and Pulsatilla), and subgenus Anemonidium (including Hepatica), respectively. The nrITS + atpB-rbcL supported the monophyletic of Anomone s.l. (including clades 2 and 3). However, the five-plastid-gene dataset made subgenus Anemone (clade 2) sister to the clade Anemoclema + Clematis s.l. (clade 1). Our results strongly supported to subsume Archiclematis and Naravelia within Clematis s.l., and to retain Anemoclema as an independent genus. For the genus Anemone s.l., all analyses supported to include Barneoudia, Knowltonia, Oreithales, and Pulsatilla in this genus. However, the five-plastid-gene dataset tended to retain Hepatica as a separated genus, corresponding to Anemone subgenus Anemonidium. Therefore, the updated tribe Anemoneae consists of four revised genera, Anemoclema, Anemone s.l., Clematis s.l. and Hepatica, and an unresolved genus, Metanemone.
Supporting information S1  DNA samples; to Zhen-Shan He, Jing Yang, Ji-Xiong Yang, Wen-Bin Yuan, Chun-Xia Zeng and Zhi-Rong Zhang for their assistance in molecular experiments; to Lily Zeng for her English editing in early version; to Curator of herbaria of Kunming Institute of Botany, Chinese Academy of Sciences and Field Museum of Natural History for allowing us to access specimens; to Carl. S. Keener and Claude W. dePamphilis for their valuable discussions and suggestions; and to two anonymous reviewers for their valuable comments and suggestions.

Author Contributions
Conceptualization: NJ ZZ K-YG W-BY.