Molecular circumscription and major evolutionary lineages of the fern genus Dryopteris (Dryopteridaceae)

Background The fern genus Dryopteris (Dryopteridaceae) is among the most common and species rich fern genera in temperate forests in the northern hemisphere containing 225–300 species worldwide. The circumscription of Dryopteris has been controversial and various related genera have, over the time, been included in and excluded from Dryopteris. The infrageneric phylogeny has largely remained unclear, and the placement of the majority of the supraspecific taxa of Dryopteris has never been tested using molecular data. Results In this study, DNA sequences of four plastid loci (rbcL gene, rps4-trnS spacer, trnL intron, trnL-F spacer) were used to reconstruct the phylogeny of Dryopteris. A total of 122 accessions are sampled in our analysis and they represent 100 species of the expanded Dryopteris including Acrophorus, Acrorumohra, Diacalpe, Dryopsis, Nothoperanema, and Peranema. All four subgenera and 19 sections currently recognized in Dryopteris s.s. are included. One species each of Arachniodes, Leptorumohra, and Lithostegia of Dryopteridaceae are used as outgroups. Our study confirms the paraphyly of Dryopteris and provides the first strong molecular evidence on the monophyly of Acrophorus, Diacalpe, Dryopsis, Nothoperanema, and Peranema. However, all these monophyletic groups together with the paraphyletic Acrorumohra are suggested to be merged into Dryopteris based on both molecular and morphological evidence. Our analysis identified 13 well-supported monophyletic groups. Each of the 13 clades is additionally supported by morphological synapomophies and is inferred to represent a major evolutionary lineage in Dryopteris. In contrast, monophyly of the four subgenera and 15 out of 19 sections currently recognized in Dryopteris s.s is not supported by plastid data. Conclusions The genera, Acrophorus, Acrorumohra, Diacalpe, Dryopsis, Nothoperanema, and Peranema, should all be merged into Dryopteris. Most species of these genera share a short rhizome and catadromic arrangement of frond segments, unlike the sister genus of Dryopteris s.l., Arachniodes, which has anadromic arrangement of frond segments. The non-monophyly of the 19 out of the 21 supraspecific taxa (sections, subgenera) in Dryopteris strongly suggests that the current taxonomy of this genus is in need of revision. The disagreement between the previous taxonomy and molecular results in Dryopteris may be due partly to interspecific hybridization and polyplodization. More morphological studies and molecular data, especially from the nuclear genome, are needed to thoroughly elucidate the evolutionary history of Dryopteris. The 13 well-supported clades identified based on our data represent 13 major evolutionary lineages in Dryopteris that are also supported by morphological synapomophies.

The phylogenetic positions of most of these genera have recently been clarified. Stigmatopteris and Ctenitis are both rather isolated within the dryopteroid lineage [14], while Megalastrum Holttum, a relatively recent segregate of Ctenitis, forms a clade with Rumohra Raddi and the paraphyletic genus Lastreopsis [14,15]. None of these genera are in fact closely related to Dryopteris. Although the close affinity among some but not all of Acrophorus C. Presl, Acrorumohra (H. Itô) H. Itô, Diacalpe Blume, Dryopsis, Dryopteris, Nothoperanema, and Peranema D. Don has long been noticed (e.g., [16,17]), it has been unclear exactly how they are phylogenetically related. In studying the historical biogeography of the species of Hawaiian Dryopteris, Geiger & Ranker [18] sampled 63 species of Dryopteris and found Nothoperanema (represented by one species) to be embedded within a paraphyletic Dryopteris. Using rps4-trnS sequence data of 60 Chinese species of Dryopteris and several related genera, Li & Lu [19] reinforced Geiger & Ranker's [18] finding that Nothoperanema (three species sampled) should belong to Dryopteris and for the first time confirmed that Acrorumohra (one species sampled) belongs to Dryopteris as well. The inclusion of Acrophorus, Diacalpe, Dryopsis, and Peranema in Dryopteridaceae was strongly supported by Li & Lu's [20] and Liu et al.'s [21] works based on rbcL and rbcL + atpB data, respectively. With relatively small sampling, both of the works also found that Dryopteris is paraphyletic in relation to these genera plus Acrorumohra.
To date, Fraser-Jenkins [1] has carried out the most intensive taxonomic study on Dryopteris worldwide, partly on the basis of early such work carried out by Itô [22,23] on the Japanese species and by Ching [8] on species in China, the Himalaya, and neighboring areas. Fraser-Jenkins [1] recognized 225 species which he divided into four subgenera: D. subgen. Dryopteris, D. subgen. Erythrovariae (H. Itô) Fraser-Jenk., D. subgen. Nephrocystis (H. Itô) Fraser-Jenk., and D. subgen. Pycnopteris (T. Moore) Ching. He divided the former three subgenera further into 16 sections. In his series of studies of species of Dryopteris in Yunnan, China, Lu [2,24,25] proposed three new sections: D. sect. Caespitosae S. G. Lu, D. sect. Chrysocomae S. G. Lu, and D. sect. Indusiatae S. G. Lu, two of which were adopted by Wu & Lu [26] in their classification of the 127 Chinese species of Dryopteris. The nonmonophyly of D. subgen. Dryopteris and D. subgen. Pycnopteris has been detected by Geiger and Ranker [18] and Li and Lu [19,20], respectively, using DNA sequences of one or two loci and with relatively small species-level sampling (ca. 60 species in both studies). Based on analyses of seven plastid loci and 97 Dryopteris species, Sessa et al. [27] rejected monophyly of eleven of Fraser-Jenkins' [1] sections and three of the four subgenera. Several additional sections have never been tested for their monophyly using molecular data.
The goals of this study include: [1] vigorously testing the monophyly of Dryopteris by including representative species of every subgenus and every section of Dryopteris currently recognized and by including all controversially related genera; [2] resolving phylogenetic relationships between Dryopteris and Acrophorus, Acrorumohra, Diacalpe, Dryopsis, Dryopteris, Nothoperanema, and Peranema; [3] assessing the monophylies of supraspecific taxa at sectional and subgeneric ranks recognized in current classifications using relatively large sampling and DNA sequences of multiple loci; and [4] identifying major evolutionary lineages in Dryopteris.

Analyses of individual plastid regions
The characteristics and statistics of four individual plastid regions from MP and ML analyses are presented in Table 1. The four individual plastid regions as well as the combined trnL intron and trnL-F spacer yielded similar tree topologies in both MP and ML analyses (trees not shown). The most parsimonious, parsimony JK, and likelihood JK and BS trees for all analyses are available upon request from the first author. There were no well-supported (≥70% JK or BS support; [28]) clades that conflicted with one another in both the parsimony JK and likelihood BS trees. Therefore, the four plastid regions were combined.

Analyses of combined plastid data
The combined data matrix of four plastid regions consisted of 3,638 bases. A simultaneous analysis [29,30] of nucleotides from all plastid regions was conducted as the primary basis for phylogenetic inference in Dryopteris.
Unweighted MP simultaneous analysis generated 1,785,800 most parsimonious trees with a length of 1,831 steps, a consistency index (CI; [31]) of 0.5877, and a retention index (RI; [32]) of 0.8370. The MP simultaneous analysis terminated prematurely when it was out of memory. The ML simultaneous analysis generated one optimal tree which is shown in Figure 1. The tree topology of the MP simultaneous analysis was similar to that from the ML simultaneous analysis and there existed no well-supported conflicts between the two trees.

The monophyly and circumscription of Dryopteris
Our analyses showed that Dryopteris in its current circumscription is paraphyletic in relation to Acrophorus, Acrorumohra, Diacalpe, Dryopsis, Nothoperanema, and Peranema ( Figure 1). This resolution is consistent with that of Geiger & Ranker [18], in which only Dryopteris and one species of Nothoperanema were sampled. Our finding is also in accordance with that of Li & Lu [19] who sampled Acrorumohra, Dryopteris, and Nothoperanema. Sessa et al. [27] found Dryopteris to be monophyletic, but did not include representatives of any of these six genera in their sampling schemes. Our study provides the first strong molecular evidence that the traditionally defined Peranemataceae sensu Wu [33,34] and Wu & Ching [35] (Acrophorus, Diacalpe, and Peranema), and Acrorumohra, Dryopsis, and Nothoperanema, should all be merged into Dryopteris, though each is monophyletic except Acrorumohra. The expanded Dryopteris is supported as monophyletic with strong support (Figure 1; ML BS: 100%; MP JK: 100%). Interestingly, except for Acrorumohra and Nothoperanema, and despite the similarity among members of these genera, the study of Li & Lu [20] was the first to suggest that Dryopteris was paraphyletic with respect to Acrorumohra, Acrophorus, Diacalpe, Dryopsis (included in Ctenitis), and Peranamea. Such a close relationship among all of them had not previously been suggested in the literature, although the close affinities among Peranemataceae, Dryopsis (previously in Tectariaceae), and Dryopteridaceae have partially long been noticed (e.g., [8,13,16,17,33]). Morphologically, Peranemataceae, Dryopsis, Dryopteris, and Nothoperanema share short rhizome and catadromic arrangement of frond segments.

Resolution of Acrophorus
The genus Acrophorus has been recognized by numerous authors (e.g., [16,17,[33][34][35][37][38][39]41,42]), but was not recognized by Fraser-Jenkins [43], who synonymized it with Peranema. Acrophorus is characterized by having a cordate and often persistent scale at the bases of costae, membranaceous and semi-globose indusia, a few multicelled septate clavate paraphyses on the lower portion of the sporangiate stalk, and a few short multi-celled clavate appendages on the margins of scales at the stipe bases [33][34][35]. Acrophorus contains about 12 species occurring in Southeast Asia, westward reaching Papua New Guinea and Polynesia [44]. Six species are sampled in our study including the type, A. nodosus C. Presl. Acrophorus is strongly supported as monophyletic in our study ( Figure 1; ML BS: 100%; MP JK: 95%). In the ML tree it is sister to Diacalpe + Nothoperanema, but this resolution received low statistical support. In the ML BS tree, it formed an unresolved trichotomy with Diacalpe and Nothoperanema (tree not shown). Based on our study, Acrophorus belongs to Dryopteris, and represents a specialized group within Dryopteris with round indusia and cordate scales at costa base. Within Acrophorus, A. paleolatus Pic. Serm. is strongly supported as sister to the remaining species ( Figure 1).
Within Diacalpe, D. annamensis Tagawa is resolved as sister to the rest of species, with D. chinensis Ching & S. H. Wu then sister to D. aspidioides + D. christensenae Ching ( Figure 1).
(See figure on previous page.) Figure 1 Simultaneous-analysis maximum likelihood tree with parsimony jackknife values above each branch, and maximum likelihood bootstrap values below each branch. If a clade was resolved in one analysis but not the other, "/" is used to indicate which analysis that clade was not resolved in. Dashed branches indicate the disproportional branch lengths. The species in red color indicate those that are currently not in the genus Dryopteris but resolved as members of Dryopteris in this study. Major morphological and/or palynological synapomorphies are indicated in blue. Geographical provenances are indicated in green.
polita is sister to A. hasseltii (Blume) Ching plus A. subreflexipinna (Ogata) H. Itô, and together these three are sister to A. diffracta. Our results clearly show that Acrorumorha is a member of Dryopteris.
Recently, Acrorumohra subreflexipinna has been postulated to have arisen through recurrent hybridization between A. hasseltii and A. diffracta, with the former being its putative maternal parent and the latter its paternal progenitor [48]. Our resolution of these three taxa (Figure 1: Acrorumohra clade) supports A. hasseltii as the maternal progenitor of A. subreflexipinna.

Resolution of Nothoperanema
Tagawa [11] originally described this taxon as a subgenus of Dryopteris. Ching [10] elevated it to a genus. Nothoperanema has been accepted at the generic level by many pteridologists (e.g., [35,37,38,41,47,49]). In contrast, Itô [50], Copeland [51], and Ohwi [52] [20] resolution where Nothoperanema was resolved as paraphyletic in relation to Acrophorus, Diacalpe, and Peranema. Our study also confirmed Liu et al.'s [21] finding that Nothoperanema is embedded within a paraphyletic Peranemataceae. This resolution is accordant with the general morphological similarities except for differences in the morphology of the indusia between Nothoperanema and Diacalpe/Peranema [10]. Our study also reinforced Geiger & Ranker's [16; with N. rubiginosum A. R. Sm. & Palmer only sampled] finding that Nothoperanema is nested within a paraphyletic Dryopteris, and we conclude that Nothoperanema should be a member of Dryopteris. The two share similar reniform indusia, short rhizomes, and catadromic arrangement of frond segments.
Within Nothoperanema, N. diacalpioides Ching, N. rubiginosum, and N. squamisetum together are strongly supported as sister to the remaining members of the genus. The relationship between N. hendersonii (Bedd.) Ching and N. shikokianum (Makino) Ching needs further clarifications.
It is now widely recognized [17,39,53,54], though the relationships among Ctenitis, Dryopsis, and Dryopteris have been controversial. Morphologically, Dryopsis appears to be more distant from Ctenitis than from Dryopteris. Dryopsis has distinct venation on the abaxial surfaces, sori terminal on veinlets, and marginal, entire scales that are clathrate or not, but with long and dull areolae. Ctenitis has venation indistinct on both the adaxial and abaxial surfaces, sori middle on the veinlets, and scales ciliate on their margins, clathrate, and with nearly hexagonal and lustrous areolae [8,13,54]. The major difference between Dryopsis and Dryopteris is that the former has either shallow or deep rachis and costa grooves that are closed near their bases, as well as multi-celled hairs with a thickened base on the margins but not in the grooves of the rachis and costae. Dryopteris, in contrast, always has deep rachis and costa grooves that are connected near their bases, and normally has no hairs on the rachis or costae [8,13,54].
Dryopsis contains about 22 species [55] occurring in tropical and subtropical Asia and reaching southwestward to southern India and Sri Lanka, eastward to Japan and the Philippines, and southward to Malaysia and Indonesia. It is most diverse in the southern and southeastern Himalaya [13,55]. With two species sampled, Liu et al. [21] discovered that Dryopsis should be a member of Dryopteridaceae but Liu et al. [21] failed to resolve the relationships among Dryopsis, Dryopteris, and Peranemataceae sensu Ching [36,37], and Wu [33]. Liu et al. [21] also concluded that Dryopsis is not closely related to Ctenitis.
Six accessions of five species of Dryopsis, including the type, D. apiciflora (Wall. ex Mett.) Holttum & P. J. Edwards, are sampled in our analysis. Our results demonstrate that Dryopsis is monophyletic (ML BS: 76%; MP JK: 66%), in contrast to the resolution of Li & Lu [20], where three species of Dryopsis formed an unresolved trichotomy with two species of Dryopteris and one species of Acrorumohra. Our results also indicate that Dryopsis is nested within a paraphyletic Dryopteris (Figure 1), strongly suggesting that Dryopsis should be subsumed into Dryopteris. This resolution is not surprising given that the morphological difference between Dryopsis and Dryopteris is minute (see above).
Within Dryopsis, the species sampled were resolved into two clades.

Monophylies of supraspecific taxa in Dryopteris
Our 100-species sampling is still not dense enough to rigorously test the monophylies of all supraspecific taxa (sections or subgenera) recognized in recent classifications by Fraser-Jenkins [1], Lu [2,24,25], and Wu & Lu [26], given that Dryopteris s.s. contains between 225 [1] and 300 species [56], and in fact is even larger given that Dryopsis, Nothoperanema, and Peranemataceae should be included in Dryopteris following our current work. However, our sampling allowed us to reject the monophylies of some supraspecific taxa because all four subgenera and 17 out of all 19 sections sampled (except D. sect. Purpurascentes and the monotypic D. sect. Politae) were represented by two or more species in our study (Appendix I).
The non-monophyly of the 19 out of the 21 supraspecific taxa in Dryopteris strongly suggests that the current taxonomy of this genus is in need of revision. However, our data do not necessarily falsify the monophyly of these 19 sections. The disagreement between previous taxonomy and molecular results in Dryopteris may be due partly to interspecific hybridization and polyplodization [57,58].
There are four well-documented allopolyploids in Dryopteris that have evolved via inter-clade hybridization, based on plastid trnL-F sequences, nuclear PgiC sequences, and/ or biochemical evidence. D. guanchica, limited to Spain, Portugal, and the Canary Islands, has been postulated to be of hybrid origin between D. aemula (D. sect. Aemulae; our Aemulae clade) and possibly D. intermedia (D. sect. Lophodium) [57]. The Japanese endemic D. shibipedis Sa. Kurata, an obvious member of D. sect. Variae judging from the morphology [1,42], has possibly a hybrid origin between D. kinkiensis (D. sect. Erythrovariae; our Erythrovariae clade) and D. pacifica (Nakai) Tagawa (D. sect. Variae; our Variae clade) [59]. Using allozyme data Jiménez et al. [60] concluded that D. corleyi, an endemic of northern Spain, is of hybrid origin between D. aemula (D. sect. Aemulae) and D. oreades Fomin (D. sect. Dryopteris; not sampled in our study but would presumbly be in our Dryopteris clade). Our analyses based on plastid data and the resolution of D. corleyi as sister to D. aemula suggest that D. aemula is the maternal progenitor of D. corleyi. In addition, Sessa et al. [58] found evidence of extensive hybridization among the New World species of Dryopteris that has involved inter-continental long-distance dispersal as well as inter-clade hybridization. These examples of hybridization not only highlight the importance of reticulate evolution and thus the importance of nuclear data in understanding the evolutionary history of Dryopteris, but also strongly support the inclusion of these 13 lineages, including the small segregates, within Dryopteris, as opposed to breaking Dryopteris into many small genera.

Major evolutionary lineages in Dryopteris
Within the newly defined Dryopteris (incl. Acrorumohra, Dryopsis, Nothoperanema, and Peranemataceae; Figure 1), the 100 species included in the current study are resolved into the following 13 well-supported major clades based on our four-locus plastid data set (Figure 1). Most of these major clades are also defined by morphological synapomorphies.
Our data clearly show that the Dryopsis clade is sister to the Erythrovariae clade; these two together are sister to a clade containing the Acrorumohra clade and the Variae clade; these four clades together are sister to a clade containing the Aemulae clade and the Nothoperanema clade; and these six clades are strongly supported as monophyletic (ML BS: 99%; MP JK: 94%). The relationships among the remaining seven clades are resolved in the ML tree ( Figure 1) but with weak (<50%) branch support.

Conclusions
The genera, Acrophorus, Acrorumohra, Diacalpe, Dryopsis, Nothoperanema, and Peranema, should all be merged into Dryopteris. Most species of these genera share a short rhizome and catadromic arrangement of frond segments, unlike the sister genus of Dryopteris s.l., Arachniodes.
The non-monophyly of the 19 out of the 21 supraspecific taxa in Dryopteris strongly suggests that the current taxonomy of this genus is in need of revision. However, our data do not necessarily falsify the monophyly of these 19 sections. The disagreement between previous taxonomy and molecular results in Dryopteris may be due partly to interspecific hybridization and polyplodization.
The 13 well-supported clades identified with our data represent 13 major evolutionary lineages in Dryopteris that are supported by morphological synapomophies and may deserve circumscription as supraspecific entities within Dryopteris.
One species each of Arachniodes Blume, Leptorumohra H. Itô, and Lithostegia Ching of Dryopteridaceae are used as outgroups based on Liu et al. [21] where Arachniodes, Leptorumohra, Lithostegia, and Phanerophlebiopsis Ching together were resolved as sister to a clade consisting of Acrophorus, Acrorumohra, Diacalpe, Dryopsis, Dryopteris, Nothoperanema, and Peranema. All sequences used in this study together with their GenBank accession numbers and/or voucher information are listed in Appendix II.

Molecular phylogenetics
The alignment of the rbcL data was manually obtained using Microsoft Wordpad. Preliminary alignments of rps4-trnS and trnL-F (trnL intron + trnL-F spacer) data were obtained using the default alignment parameters in Clustal X [71] followed by manual adjustments. Gap characters were coded as missing data.
Equally weighted maximum parsimony (MP) tree searches were conducted for each data matrix using 1000 tree-bisection-reconnection (TBR) searches in PAUP* ver. 4.0b10 [72] with MAXTREES set to increase without limit. Parsimony jackknife (JK) analyses [73] were conducted using PAUP* with the removal probability set to approximately 37%, and "jac" resampling emulated. One thousand replicates were performed with 10 TBR searches per replicate and a maximum of 100 trees held per TBR search. In addition to the analyses of the four individual regions, MP and ML analyses of the combined trnL intron and trnL-F spacer were also conducted since these two linked regions are sometimes viewed as one locus.
MrModeltest 2.3 [74], a modified version of Modeltest 3.6 [75], was used to select the best fit likelihood model for maximum likelihood (ML; [76]) analyses. The Akaike Information Criterion [77] was used to select among models instead of the hierarchical likelihood ratio test, following Pol [78] and Posada and Buckley [79]. The models selected were GTR + G (trnL-F spacer), GTR + I (trnL intron), GTR + I + G (trnL intron & trnL-F spacer and the simultaneous analysis), HKY + G (rps4-trnS spacer), and SYM + I + G (rbcL gene). The selected models and parameters estimated ( Table 2) were then used for tree searches from the respective data partitions. One hundred jackknife replicates were performed with one TBR search per replicate and a maximum of 100 trees held per TBR search.
The simultaneous ML analyses of nucleotide characters and ML bootstrapping (BS) were performed using RAxML-HPC2 on TG ver. 7.2.8 ( [80,81]; available at http://www. phylo.org/) with 1000 rapid bootstrap analyses followed by a search for the best-scoring tree in a single run.

Monophylies of supraspecific taxa in Dryopteris
Our results show that none of the four subgenera, D. subgen. Dryopteris, D. subgen. Erythrovariae, D. subgen. Nephrocystis, and D. subgen. Pycnopteris (T. Moore) Ching, are monophyletic. Sessa et al. [27] also rejected monophyly of these subgenera, except for D. subgen Pycnopteris, for which they had insufficient sampling to test monophyly. In the current study, most of the members of D. subgen. Dryopteris are resolved in the Dryopteris clade, while others are placed in other major clades except the Acrorumohra, Dryopsis, Erythrovariae, Nothoperanema, and Variae clades. Members of D. subgen. Erythrovariae sensu Fraser-Jenkins [1] are resolved in the Acrorumohra, Erythrovariae, and Variae clades, but these clades are paraphyletic in relation to the Dryopsis clade as well as D. chinensis (Baker) Koidz. and D. gymnophylla Our results also demonstrated that 15 out of the 17 sections currently recognized [1,2,[24][25][26], for which two or more species are sampled, are not monophyletic (Figure 1). Only two sections, D. sect. Cinnamomeae Fraser-Jenk. and D. sect. Variae Fraser-Jenk., are resolved as monophyletic with our current sampling. This is at odds with Sessa et al. [27], which tested the monophly of eleven sections, including D. sect. Cinnamomeae and D. sect. Variae, and rejected the monophly of all. Although these two sections are found to be monophyletic in the current study, the sampling for both was larger in Sessa et al. [27], and they are thus not likely to be monophyletic either. The 15 polyphyletic sections discovered in our study are:  "G" = gamma distribution shape parameter [82]. "GTR" = general-time-reversible model [83]. "HKY" = Hasegawa-Kishino-Yano model [84]. "I" = proportion of invariable sites. "SYM" = symmetrical model [85]. "Ti/Tv" = transition/transversion ratio.