Using a multilocus phylogeny to test morphology-based classifications of Polystichum (Dryopteridaceae), one of the largest fern genera

Background Polystichum (Dryopteridaceae) is probably the third largest fern genus in the world and contains ca. 500 species. Species of Polystichum occur on all continents except Antarctica, but its highest diversity is found in East Asia, especially Southwest China and adjacent regions. Previous studies typically had sparse taxon sampling and used limited DNA sequence data. Consequently, the majority of morphological hypotheses/classifications have never been tested using molecular data. Results In this study, DNA sequences of five plastid loci of 177 accessions representing ca. 140 species of Polystichum and 13 species of the closely related genera were used to infer a phylogeny using maximum likelihood, Bayesian inference, and maximum parsimony. Our analyses show that (1) Polystichum is monophyletic, this being supported by not only molecular data but also morphological features and distribution information; (2) Polystichum is resolved into two strongly supported monophyletic clades, corresponding to the two subgenera, P. subg. Polystichum and P. subg. Haplopolystichum; (3) Accessions of P. subg. Polystichum are resolved into three major clades: clade K (P. sect. Xiphophyllum), clade L (P. sect. Polystichum), and the HYMASO superclade dominated by accessions of P. sect. Hypopeltis, P. sect. Macropolystichum, and P. sect. Sorolepidium, while those of P. subg. Haplopolystichum are resolved into eight major clades; and (4) The monophyly of the Afra clade (weakly supported), the Australasian clade (weakly supported), and the North American clade (strongly supported) is confirmed. Conclusions Of the 23 sections of Polystichum recognized in a recent classification of the genus, four (P. sect. Hypopeltis, P. sect. Neopolystichum, P. sect. Sorolepidium, P. sect. Sphaenopolystichum) are resolved as non-monophyletic, 16 are recovered as monophyletic, and three are monospecific. Of the 16 monophyletic sections, two (P. sect. Adenolepia, P. sect. Cyrtogonellum) are weakly supported and 14 are strongly supported as monophyletic. The relationships of 11 sections (five in P. subg. Haplopolystichum; six in P. subg. Polystichum) are well resolved. Electronic supplementary material The online version of this article (doi:10.1186/s12862-016-0626-z) contains supplementary material, which is available to authorized users.

A morphology-based infrageneric classification of a group is basically phylogenetic hypotheses based on morphology. An infrageneric treatment of a genus is important for floristic and monographic studies and this is particularly true for a large genus like Polystichum [18]. Although infrageneric classifications of Polystichum go back at least to Keyserling [35] who established P. sect. Parapolystichum Keyserling (= Parapolystichum (Keyserl.) Ching), the first relatively comprehensive attempt at subdividing the genus in a natural way was conducted by Tagawa [36]. Based on morphological characters such as pinnation, the aspect of scales and sori, Tagawa [36] divided the species of Korea, Japan, and Taiwan into eight sections: P. sect. Achroloma Tagawa, P. sect. Crucifilix Tagawa, P. sect. Cyrtomiopsis Tagawa, "P. sect. Eupolystichum" Diels (=P. sect. Polystichum), P. sect. Haplopolystichum, P. sect. Mastigopteris Tagawa, P. sect. Metapolystichum Tagawa, and P. sect. Sorolepidium (Christ) Tagawa (Table 1).
In revising the African species of Polystichum, Roux [39] classified the 24 species he recognized into nine sections including P. sect. Lasiopolystichum, P. sect. Metapolystichum, P. sect. Xiphopolystichum, and other six sections (nom. nud.) he proposed in his Ph.D. dissertation. Later when he published his work [29] he did not describe these six sections officially and recognized only 16 species and one hybrid for Africa. An extensive study of subdividing Polystichum was conducted by Kung et al. [4] where the then recognized 168 species of Polystichum in China were accommodated in 13 sections. Four of Tagawa's [36] eight sections and six of Daigobo's [37] 16 sections were adopted, albeit often with dramatically different circumscriptions. Two additional sections were introduced: P. sect. Neopolystichum Ching ex Li Bing Zhang & H.S.Kung and P. sect. Sphaenopolystichum Ching ex W.M.Zhu & Z.R.He (Table 1).
The most recent and comprehensive subdivision of Polystichum was performed by Zhang and Barrington [18] who arranged the 208 species recognized for Flora of China in two subgenera: P. subg. Polystichum and P. subg. Haplopolystichum (Tagawa) Li Bing Zhang, and the former further into 14 sections while the latter into nine sections ( Table 1). Nine of the 23 sections were newly proposed and most of the existing sections were circumscribed differently in comparison with their earlier delimitations by Tagawa [36], Daigobo [37], Fraser-Jerkins [2,38] and Kung et al. [4].
In the era of molecular phylogenetics, although substantial progress in understanding the phylogeny of Polystichum has been achieved using plastid rbcL, rps4-trnS, and trnL-F data [1,11,13,15,[40][41][42][43][44][45], the relationships among sections, species, and previously recognized genera, Cyrtogonellum and Cyrtomidictyum, as well as Cyrtomium subser. Balansana, have not yet been resolved and the majority of the Asian species not included in any molecular analyses. So far no monophyletic supraspecific taxa except Cyrtomidictyum (= P. sect. Cyrtomiopsis Tagawa) have been recovered using molecular data. Almost all morphological hypotheses about the relationships within Polystichum, especially in terms of subdivisions of the genus, have largely remained speculative.
The objectives of this study included: (1) to test the monophyly of Polystichum using the largest taxon and character sampling so far; (2) to resolve the major relationships within Polystichum worldwide with focus on the Old World taxa which represent ca. 80 % of the species diversity in the genus; (3) to evaluate the monophyly of the supraspecific taxa recognized in the most recent classification and to test other previous morphologybased hypotheses of relationships within Polystichum.

Results
This study generated 334 new sequences (Additional file 1). The dataset characteristics and tree statistics for the analyses are presented in Table 2. Comparisons of tree topologies from MPJK analyses of the individual markers did not identify any well-supported conflicts (MPJK ≥ 70 %; [46][47][48]). Thus, the five datasets were concatenated. The topology of the ML tree based on the concatenated dataset ( Fig. 1) is mostly identical to those based on each individual marker, but with generally increased support values.

Discussion
Monophyly of Polystichum and its relationships with Cyrtomium and Phanerophlebia While the monophyly of polystichoid ferns (i.e., Cyrtomium, Phanerophlebia and Polystichum) has been highly supported in previous studies [1,44], the relationships among these three genera remain ambiguous. With the largest sampling so far (about three times as large as the previous largest worldwide sampling by Driscoll and Barrington [44]), our study resolved Polystichum as monophyletic but only with weak support (MLBS: 57 %; MPJK: 58 %; BIPP: < 0.50). Although earlier studies [1,49] using a limited molecular sampling (only rbcL sequences) found Polystichum (sensu [18,20]) as paraphyletic in relation to Cyrtomium (weak support), the monophyly of Polystichum is further supported by several morphological features (i.e., lamina 1-3-pinnate, apex pinnatifid, without a clear apical pinna; venation mostly free, rarely anastomosing to form 1 or 2 rows of areoles). Our result is also consistent with some more recent studies based on multi-locus datasets [42,44,50]. The sister relationship between Cyrtomium and Phanerophlebia is highly supported by our phylogenetic reconstructions (MLBS: 86 %; MPJK: 90 %; BIPP: 1.00, Fig. 1a). As early as 1988 Yatskievych et al. already found that Cyrtomium and Phanerophlebia are convergent descendants from different progenitor groups based on chloroplast restriction site data [51]. However, a closer relationship of Polystichum with Cyrtomium than with Phanerophlebia was found by Li et al. [42] based on plastid trnL-F and rps4-trnS data and by Mc Keown et al. [50]. Our data do not support this resolution (Fig. 1).
Generally, the relationships among polystichoid ferns obtained here are in agreement with those found in most of the earlier phylogenetic studies but more studies are needed to fully resolve the relationships among these three genera.

Relationships within Polystichum
Within Polystichum, 164 accessions are resolved into two monophyletic clades corresponding to P. subg. Polystichum and P. subg. Haplopolystichum (Tagawa) Li Bing Zhang defined by [18], both with strong support. The sister relationship between these two subgenera agrees with the morphology and the findings with molecular data by Driscoll and Barrington [44] and Li et al. [42], but contrasts those by Little and Barrington [1] and Lu et al. [49] based on rbcL data alone which resolved P. subg. Polystichum as sister to Cyrtomium, and them together as sister to P. subg. Haplopolystichum (also see above). The bulbil-bearing species are resolved in five clades (A, C, D, N, T), suggesting that bulbils evolved at least five times in Polystichum, twice in P. subg. Polystichum and three times in P. subg. Haplopolystichum.

I Polystichum subg. Haplopolystichum (Tagawa) Li
Bing Zhang (Fig. 1a): Nine sections are recognized by Zhang and Barrington [18] in this subgenus. The monophyly of all but two sections is recovered and six sections are well supported as monophyletic. The relationships among all but four sections in the subgenus are well resolved. This section contains about five species, four of which are endemic to Southwest to central China [18]. We included three species in this study including the sexual tetraploid P. erosum Ching & K.S.Shing. Our data confirmed the monophyly of this section (MLBS: 100 %; MPJK: 100 %; BIPP: 1.00), consistent with our earlier findings based on more species and accessions sampled [10,15].
-This section contains about four species and is characterized by prolonged rachis apex with bulbils and broad-type microscales [18]. This section was often recognized as a genus, i.e., Cyrtomidictyum Ching (e.g., [52,53] (Fig. 1a: clade E). -This section in its new circumscription [18] contains about six species including four assigned to Cyrtomium in early classifications (e.g., [4]). We sampled four species in our study. Our analyses recovered the monophyly of the section but only with weak support in ML and MP analyses (MLBS: < 50 %; MPJK: < 50 %) but moderately support in BI analysis (BIPP: 0.94). Interestingly, the two former members of Cyrtomium, which have anastomosing venation [P. balansae Christ, P. hookerianum (C.Presl) C.Chr.], are paraphyletic in relation to two species with free venation, suggesting that the anastomosing venation in the section evolved at least twice or evolved once but reversed to free venation from anastomosing venation in the P. falcatilobum + P. formosanum clade. Polystichum sect. Adenolepia sensu Daigobo [37], which included P. obliquum (Don) T.Moore, a member of P. sect. Haplopolystichum, is apparently polyphyletic. 6 Polystichum sect. Crucifilix Tagawa (Fig. 1a: [53] sampled three species and recovered the monophyly of P. sect. Sphaenopolystichum but without any statistical support. Our data of eight accessions representing ca. eight species failed to recover the monophyly of the section in all three analyses; instead the eight accessions were resolved into three subclades: the P. alcicorne subclade containing ca. three species, the P. tonkinense subclade containing one species, and the P. auriculum subclade containing P. auriculum Ching, P. bifidum Ching, P. caruifolium (Baker) Diels, and P. christii Ching (Fig. 1a: clade H). In BI analysis, the P. alcicorne subclade is resolved as sister to P. sect. Cyrtogonellum with BIPP = 0.60. P. sect. Sphaenopolystichum together with P. sect. Cyrtogonellum is strongly supported as monophyletic (MLBS: 100 %; MPJK: 99 %; BIPP: 1.00). One member of the section, P. wattii (Bedd.) C.Chr., has never been included in any molecular studies and might not belong to this section. 9 Polystichum sect. Haplopolystichum Tagawa (Fig. 1a: clade J). -In its recent circumscription (i.e., sensu Zhang and Barrington [18]), P. sect. Haplopolystichum, is different from its original delimitation by Tagawa [36]. The latter contained also P. sect. Adenolepia, P. sect. Hecatoptera, and P. sect. Stenopolystichum in our definition [18].
Our results show that P. sect. Haplopolystichum sensu Tagawa [36] is highly polyphyletic and taxa included in its original definition are resolved in two subgenera (see above and our Fig. 1a, b). This section in our definition [18] is estimated to contain about 200 species [18] and almost all species recently described from southern China and Vietnam belong to this section [6-11, 15, 16, 19, 55]. We included seven accessions representing four species. An ongoing project focusing on this section will include many more species. Our current study shows that this section is strongly supported as monophyletic  [9,10] for the relationships of P. acutidens and P. obliquum), shows that P. sect. Polystichum sensu Fraser-Jenkins [38] is apparently polyphyletic as these members are resolved in three independent clades ( Fig. 1a and b: clades J, L, T). 12 Polystichum sect. Achroloma Tagawa (Fig. 1b: clade M). -When Tagawa [36] established this section, he included only the type P. nepalense (Spr.) C.Chr. (a diploid sexual species). Daigobo [37] added P. falcatipinnum Hayata (= P. manmeiense (Christ) Nakaike, tetraploid sexual) to this section. This delimitation was accepted by Zhang and Barrington [18] but rejected by Kung et al. [4] who included the two species in P. sect.
Polystichum. Fraser-Jenkins [38] placed the latter species in P. sect. Hypopeltis. Our study is the first to sample both of the species in a molecular analysis. Two species formed a clade with strong support (MLBS: 100 %; MPJK: 100 %; BIPP: 1.00) in our analysis. This section is resolved as sister to the P. sect. Macropolystichum clade (MLBS: 85 %; MPJK: 85 %; BIPP: 1.0), consistent with the resolution found by Driscoll and Barrington [44]. These two sections share evergreen leaves which are shiny adaxially. 13 Polystichum sect. Macropolystichum Daigobo (Fig. 1b: clade N). -As defined by Zhang and Barrington [18], this section contains species with or without proliferous bulbils but all members are of relatively large habit and laminae that are dark green and shiny adaxially. Ten accessions representing about 8 out of ca. 17 species of this section are sampled in our study (the largest sampling so far). Our analyses recovered the monophyly of P. sect. Macropolystichum sensu Zhang and Barrington [18] with strong support (MLBS: 78 %; MPJK: 73 %; BIPP: 1.00).
Prionolepis is a heterotypic synonym of P. sect. Macropolystichum. The same species was treated as a member of P. sect. Neopolystichum by Zhang and Kung [57], but this is not supported by our data. P. mucronifolium, resolved as a member of P. sect. Macropolystichum in our study, was placed in P. sect. Metapolystichum Tagawa, a heterotypic synonym of P. sect. Hypopeltis [18], by Fraser-Jenkins [38]. 14 Polystichum sect. Chingiarum Li Bing Zhang ( Fig. 1b: clade O). -This monospecific section contains P. chingiae Ching [18] and our study is the first to include it in a molecular analysis. Our study resolved the species in the MCSCHMANS clade, but its relationships are not well resolved. The isolated phylogenetic position is consistent with its special morphology. Morphologically, this species has lamina 1-pinnate, pinnae not cartilaginous at margins, and sori in 2 or 3 rows on each side of midrib and abaxial on veinlets [18]. Such a combination of morphological features is unique within the genus. 15 Polystichum sect. Micropolystichum Tagawa ( Fig. 1b: clade P). -This section contains only about six montane to alpine species [18]. We included three accessions representing two species. Our study resolved P. sect. Micropolystichum as a strongly supported clade (MLBS: 100 %; MPJK: 100 %; BIPP: 1.00) which is sister to P. grandifrons, but the sister relationship between those lineages is weakly supported statistically (MLBS: < 50 %; MPJK: < 50 %) and morphologically [18]. Fraser-Jenkins [38] also placed the diploid sexual Polystichum capillipes (Baker) Diels and P. wattii (Bedd.) C.Chr. in P. sect. Micropolystichum.
Neither of the species were sampled in our current study, but our earlier study [10] resolved the former species in P. sect. Basigemmifera (Fig. 1a: (Fig. 1b). Fraser-Jenkins [38] placed P. grandifrons C.Chr. in P. sect. Macropolystichum, which is not supported by our data. Given our limited phylogenetic sampling and the low support values, the taxonomic rearrangements in P. sect. Neopolystichum need further investigations. 17 Polystichum sect. Sorolepidium (Christ) Tagawa ( Fig. 1b: clade R). -Sorolepidium Christ was often recognized as a genus (e.g., [52,58] [18]. Our study clearly placed P. wilsonii as a member of clade X1 and the phylogenetic position of P. bakerianum is not resolved (Fig. 1b). 18 Polystichum sect. Hecatoptera (L.L.Xiang) Li Bing Zhang (Fig. 1b: clade S). -This monospecific section contains P. hecatopterum Diels only [18], a diploid sexual [61], and our study is the first to include this species in a molecular analysis. We could not amplify its rbcL gene. Our data from other four plastid loci show that this species is definitely a member of P. subg. Polystichum confirming our earlier hypothesis [18], in spite of its striking morphological similarities with members of P. subg. Haplopolystichum in once-pinnate lamina without bulbils on its rachis [18]. Xiang [62] established P. ser. Hecatoptera L.L.Xiang based on its long-spinulose pinna margins but placed it in P. sect. Haplopolystichum, as Tagawa [36] did. Interestingly, Daigobo [37] placed this species in P. sect. Stenophyllum, which is a section of P. subg. Polystichum (see our discussion below) although he did not recognize any subgenera in the genus. Our ML and BI analyses resolved P. hecatopterum as sister to a clade containing P. sect. Stenopolystichum and part of P. sect. Sorolepidium (R2) with weak support. 19 Polystichum sect. Stenopolystichum Daigobo ( Fig. 1b: clade T). -Tagawa [36] placed the type of the section, P. stenophyllum Christ, a diploid sexual species, in P. sect. Haplopolystichum based on its once-pinnate lamina and terminal sori on veinlets. Four accessions representing ca. three species of this section are included in our study. All species of this section have proliferous bulbils at the apex of lamina [18,37]. Our study is the first to confirm the monophyly of the section. This section is resolved as monophyletic (MLBS: 75 %; MPJK: 59 %; BIPP: 0.99) and sister to P. sect. Lasiopolystichum sensu Daigobo [37] with high support values (MLBS: 97 %; MPJK: 94 %; BIPP: 1.00). This sister relationship is unexpected given the huge differences between the two sections. Polystichum sect. Lasiopolystichum was merged with P. sect. Sorolepidium by Zhang and Barrington [18]. 20 Polystichum sect. Crinigera Li Bing Zhang ( Fig. 1b: clade U). -This monospecific section contains P. crinigerum (C.Chr.) Ching only [18] and our study is the first to include it in a molecular analysis. P. crinigerum, together with P. nepalense and P. chingiae Ching, was included in P. sect. Polystichum by Kung et al. [4]. Our study shows that P. crinigerum is not closely related to either of them suggesting that the similarity among them in once-pinnate lamina and asymmetrical pinna base is not a synapomorphy. However, the relationships of P. crinigerum are not well resolved with our data. Our ML and MP analyses resolved it as sister (MLBS: 53 %; MPJK: 51 %) to a species of P. sect. Hypopeltis and they together are sister (MLBS < 50 %) to P. longispinosum Ching ex Li Bing Zhang & H.S.Kung, a species assigned to P. sect. Neopolystichum [18,57].
Our BI analysis resolved it as sister to P. longispinosum (BIPP: 0.73). 21 Polystichum sect. Fimbriata Li Bing Zhang ( Fig. 1: clade V). -This monospecific section contains P. fimbriatum Christ [18] and our study is the first to include it in a molecular analysis. Polystichum fimbriatum is strongly (MLBS: 88 %; MPJK: 82 %) supported as sister to a clade containing portions of P. sect. Hypopeltis (Fig. 1b:  X1) in our sampling (see below). This sister relationship is unexpected given their dissimilarity in morphology of lamina shape and scales [18]. This resolution collapsed in BI analysis which resolved it as part of a trichotomy. 22 Polystichum sect. Subfimbriata Li Bing Zhang ( Fig. 1b: clade W). -This monospecific section contains P. subfimbriatum W.M.Chu & Z.R.He [18] and our study is the first to include it in a molecular analysis. When P. subfimbriatum was described, Chu and He [63] compared it with P. fimbriatum. Indeed, both species share once-pinnate and leathery lamina, but their scales on rachis and stipes are very different. A close relationship between these two species is not suggested with our analyses which resolved P. subfimbriatum as sister to portion of P. sect. Hypopeltis (Fig. 1b: X3; MLBS: < 50 %; BIPP: 0.57) but with low statistical support. 23 Polystichum sect. Hypolepis (Michx.) T.Moore (Fig. 1b: clade X). -Zhang and Barrington [18] re-defined this section and made it the most accommodating section in the genus. They noted that this section in their definition is possibly polyphyletic. We included 55 accessions in our study. Our results show that P. sect. Hypolepis is indeed polyphyletic. Accessions of this section are resolved in about nine subclades, which partially corresponds to the morphological heterogeneity noticed in this section. Although polyphyletic, the majority of species belonging to P. sect. Hypolepis are included in the HYSUFI clade which also contains P. fimbriatum and P. subfimbriatum. Within this clade, three relatively well-supported subclades can be identified: the P. ovatopaleaceum subclade (Fig. 1b: subclade X2), the P. polyblepharum subclade (Fig. 1b: subclade X3), and the P. sinensis subclade (Fig. 1b: [40], and members of the Afra clade [44]. But these issues do not affect our overall topology. The monophyly of P. sect. Achroloma, P. sect. Macropolystichum, P. sect. Polystichum, P. sect. Sorolepidium, P. sect. Stenopolystichum, and P. sect. Xiphopolystichum is strongly supported by our analyses (Fig. 1b).
Notably, the isolated positions of these species is in line with their peculiar morphology. Polystichum discretum (diploid) and P. weimingii were placed in P. ningshenense, and some species in our subclades X2 and X3 (Fig. 1b). Our current study did not recover the monophyly of P. ser. Brauniana. 24 The Afra, the North American, and the Australasian lineages (Fig. 1b: Afra

Conclusions
Our study based on the largest character sampling and most taxonomically comprehensive sampling so far successfully resolved the 164 accessions representing ca. 140 species of Polystichum into two well-supported major clades, corresponding to the two subgenera, P. subg. Polystichum and P. subg. Haplopolystichum. Although our study is still preliminary of many results, given that the taxon and character sampling still needs improvements and that some results are poorly supported, our current work is the first toward a new classification based on morphological and molecular evidence in the genus Polystichum. Of the 23 sections of Polystichum recognized in a recent classification of the genus, except three monospecific sections which are each represented by one accession, four sections (P. sect. Hypopeltis, P. sect. Neopolystichum, P. sect. Sorolepidium, P. sect. Sphaenopolystichum) are resolved as paraphyletic or polyphyletic, 16 are recovered as monophyletic. Of the 16 monophyletic sections, two (P. sect. Adenolepia, P. sect. Cyrtogonellum) are weakly supported and 14 are strongly supported. In addition, our study also recovered the monophyly of the Afra clade (moderately supported) and the North American clade (strongly supported). The relationships of 11 sections (5 in P. subg. Haplopolystichum; 6 in P. subg. Polystichum) are well resolved (MLBS ≥ 78 %; MPJK ≥ 76 %). However, several phylogenetic uncertainties remain, particularly in P. sect Hypopeltis. These issues probably linked to introgression and/or fast radiation highlight the fact that more data including nuclear data are needed to obtain a complete picture of the evolutionary relationships in polystichoid ferns and therefore draw a new taxonomic framework for one of the largest genera of ferns, Polystichum.

Taxonomic sampling
To test the monophyly of the two subgenera and 23 sections recognized in the most recent classification of Polystichum [18], we included 121 accessions representing about 106 species of P. subg. Polystichum and 43 accessions representing 34 species of P. subg. Haplopolystichum (see Table 1). Given that Cyrtomium (sensu [18]) and Phanerophlebia are both monophyletic [42,43,51,67] and each mainly distributed in only one area (i.e., East Asia and southwestern U.S.A. to Central America, respectively), we included six species of Cyrtomium and two of Phanerophlebia. Denser sampling of these two genera will be performed in a separate ongoing study (Le Péchon et al., unpubl. data). Based on previous molecular [1,41,42,44,49,59,[68][69][70] and morphological works [71], five species of Dryopteris Adans. and two of Arachniodes Blume were included as outgroups. In total, 177 accessions representing ca. 153 species in the subfamily Dryopterioideae (sensu [71]) were included in this study. Taxa included, their classification, voucher information, and GenBank accession numbers are given in Additional file 1.

DNA extraction, PCR amplification and sequencing
Genomic DNA was extracted from fresh, silica-gel dried, or herbarium leaf fragments using TIANGEN plant genomic DNA extraction kit (TIANGEN Biotech., Beijing, China) according to the manufacturer's protocols. We selected five chloroplast regions (the intergenic spacers psbA-trnH, trnS-rps4 and trnL-trnF, the trnL intron, and the protein-coding gene rbcL). The primers used to amplify these regions were based on previous studies or newly designed ( Table 3). The PCR protocols followed Zhang et al. [72] and Small et al. [73]. All regions were amplified in 25 μL volumes, with 15.85 μL deionized sterile water, 2.5 μL of 25 mol/L EasyTaq Buffer, 1.5 mL of 25 mmol/L MgCl 2 solution, 2 μL of a 2.5 mmol/L dNTP solution in equimolar ratio, 1 μL of each primer at 10 pmol/μL, 1 unit (0.2 μL) of EasyTaq DNA polymerase (TransGen Biotech, Beijing, China), and 1 μL of the template DNA. PCR products were purified and sequenced by Invitrogen (Shanghai, China).

Sequence alignment and phylogenetic reconstruction
The resulting sequences were edited and assembled with Sequencher V.4.14 (GeneCodes Corporation, Ann Arbor, Michigan, USA). We manually performed the sequence alignment using Bioedit [74]. Gaps (insertion/deletion events) were considered as missing data. Phylogenetic relationships were reconstructed using maximum parsimony (MP), maximum likelihood (ML), and Bayesian inference (BI). Maximum parsimony jackknife (MPJK) analyses [75] were conducted using PAUP* for each dataset with the removal probability set to approximately 37 %; and "jac" resampling emulated. One thousand replicates were performed, each from a different random addition sequence tree, with 100 TBR searches per replicate and a maximum of 100 trees held per TBR search. A final simultaneous MP analysis [76,77] was conducted based on the combined dataset including the five molecular markers.
Each DNA region of the concatenated molecular matrix was assigned a separate GTR+I+G substitution model. ML tree searches and 10,000 rapid bootstrapping (MLBS) were conducted using RAxML-HPC and default parameters, followed by a search for the best-scoring tree, in a single run [78,79]. jModelTest 2 [80] was used to select the best-fit likelihood model for Bayesian analyses. The Akaike information criterion [81] was used to select among models. The models selected were GTR+G (psb-trnH spacer), GTR+I+G (rps4-trnS spacer, the combined region trnL-trnF and rbcL gene). The selected models (Table 2) were then used for tree searches from the respective data partitions in combination. BI analyses were performed using MrBayes v3 [82]. For each DNA partition, we used the appropriate model selected by jModelTest 2, and each molecular region has independent parameters and the overall rate is allowed to be different across partitions. Four chains (i.e., three heated and one cold) of Metropolis-coupled Markov chain Monte Carlo were performed for 50 million generations, sampling every 1000th generation. After checking the convergence of parameter traces among generations using Tracer [83], we discarded the first 25 % of sampled trees as a "burnin phase". The remaining trees were then used to calculate Bayesian inference posterior probability (BIPP).
ML and BI analyses were run on the CIPRES cluster available at http://www.phylo.org/ [84].