Exploring phylogenetic relationships within the subgenera of Bambusa based on DNA barcodes and morphological characteristics

The genus Bambusa belongs to the subtribe Bambusinae and the subfamily Bambusoideae. The subgenera of Bambusa has not been satisfactorily circumscribed, and this remains a major taxonomic issue. Simultaneously, genera such as Dendrocalamus and Gigantochloa have not been confidently assigned to Bambusa. Here, the phylogenetic relationships among subgenera were investigated using five chloroplast DNA markers (rpl32-trnL, rpl16, matK, rbcL, and trnH-psbA) for a sample of 50 ingroup and 16 outgroup species. A total of 186 key morphological descriptors were studied for the 50 ingroup species. The results indicated that five chloroplast DNA markers were possible to distinguish Bambusa species from other species and divide them into several clusters. Phylogenetic analyses conducted using morphological descriptors and a combined marker (rpl32-trnL+rpl16) revealed three and five distinct lineages, respectively, among the currently recognized Bambusa species. The branching pattern of the dendrogram was not completely consistent with the classical taxonomic classification of Bambusa. In addition, not all varieties and cultivars were clustered with McClure classifications. As the maximum parsimony topology and morphological analyses were inconsistent, some clustering results overlapped. Overall, the results obtained here do not support the current classification of the Bambusa subgenera.

www.nature.com/scientificreports/ widely used methods applied in bamboo classification 7,8 , especially at the genus level 9,10 . Recently, substantial progress has been made towards understanding the evolutionary relationships of Bambusa and its allies (Bambusa, Dendrocalamus, Gigantochloa, and Melocalamus are classified as a close group, in particular based on their shared characteristic of a solid, thickened, and hairy ovary summit) using molecular data 11 . Yang 11 et al. used nuclear gene (GBSSI) and plastid DNA sequences (psbA-trnH, rpl32-trnL, and rps16), which allowed Bambusa and Dendrocalamopsis to be classified into one of two clades with reasonable support. Through this approach, 17 Bambusa samples were classified into three clades, and this result supported the present subgeneric classification of Bambusa. However, other studies have not supported this classification. The phylogeny of bamboo species has also been analyzed using only internal transcribed spacer (ITS) sequences. In this group, each branch was composed of several species of three subgenera (not including the subgenus Dendrocalamopsis), and the Bambusa and Dendrocalamus species formed a group with a bootstrap value of 100 12 . Goh 13,14 et al. used chloroplast DNA markers (rps16-trnQ, trnC-rpoB, and trnD-T) and a nuclear DNA marker (GBSSI) to classify Bambusa, Dendrocalamus, and Gigantochloa as distinct lineages. This approach identified four Bambusa subgenera, which differed from the subgeneric classification. Chloroplast DNA sequences have been extensively used to infer plant phylogeny for uniparental inheritance through comparison with nuclear DNA sequences 15 . Several DNA markers have been used as core plant barcodes, such as the plastid (chloroplast) markers rbcL, matK, and trnH-psbA. Nuclear ribosomal ITSs have also been used 16 . Statistical results revealed that these three plastid markers showed high levels of universality (87.1-92.7%) and that the combination of ITS and any of the plastid DNA markers was able to discriminate 69.9-79.1% of species 17 . In this study, many DNA barcoding primers (trnL-trnF, trnS-trnG, psbB-psbF, rpl16, rpl32-trnL, rbcL, matK, trnH-psbA, and ITS) were utilized with the aim of amplifying the DNA sequences of bamboo samples. Unfortunately, the plastid DNA markers trnL-trnF, trnS-trnG, and psbB-psbF failed to amplify most specimens, as did the nuclear marker ITS.
In addition to DNA barcoding, researchers have attempted to arrange morphological characteristics into a data matrix using cladistic analysis 18  In the present study, the phylogenetic relationships among the four subgenera of Bambusa were investigated (50 samples) using DNA sequence data and morphological characteristics, employing a much larger taxon sample than has been previously available. This included representatives from all subgenera of Bambusa that have previously been described. DNA sequence data were derived from the plastid markers.

Materials.
A total of 66 taxa from Bambusa and some other bamboo species (Table 1, all Latin names were obtained from the FOC) representing ten genera were sampled for molecular phylogenetic analysis. There were 50 species from Bambusa belonging to the four subgenera described in the FOC 2 , including D. oldhamii and Neosinocalamus affinis, which were not accepted as Bambusa in the FRPS, but were moved to Bambusa in 2007 2 and named B. oldhamii and B. emeiensis, respectively. The outgroup taxa included Dendrocalamus, Drepanostachyum, Indosasa, Melocanna, Neosinocalamus, Oligostachyum, Phyllostachys, Pseudosasa, Pleioblastus, Shibataea, and Sinobambusa. Fifty taxa from Bambusa were collected and analyzed for morphological phylogeny. DNA isolation, amplification, cloning, and sequencing. Leaves were collected from the Hua'an Bamboo Garden (Fujian Province, China) and Lin'an Taihu Lake Source Bamboo Garden (Zhejiang Province, China). Total DNA was extracted from silica-gel-dried young leaves, using a modification of the method described by Fulton 22 Table 2.
PCR was conducted using the TaKaRa Ex™ kit (Takara Biomedical Technology Co., Ltd., Beijing, China) with the following program settings: DNA sequence alignment and phylogenetic analyses. DNA  www.nature.com/scientificreports/ parsimony (MP) analysis was conducted based on the separate rpl32-trnL, rpl16, matK, rbcL, and trnH-psbA datasets and with a combined rpl32-trnL+rpl16 dataset. MP analysis was performed with MEGAX (https:// www. megas oftwa re. net/); all characteristics were equally weighted, and gaps were coded as missing data. Heuristic searches of 1,000 random addition replicates were conducted using subtree-pruning-regrafting (SPR) branch swapping. This was done to obtain the most parsimonious trees, and ten trees from each random sequence were saved. Estimates of clade robustness were obtained through bootstrap values (BV) calculated from 1000 replicate analyses, conducted using the heuristic search strategy and through a simple addition sequence of the taxa. The incongruence length difference (ILD) test of Farris 31 et al. was used to evaluate the statistical significance of character incongruence among the rpl32-trnL and rpl16 intron datasets before their combined analysis. Table 1. Sixty-six taxa from Bambusa and the outgroup. Marker name in the missing data column indicates that there was an amplification or sequencing failure; * indicates missing morphological characteristic data.  Table A1, which were assessed from each of the 50 OTUs (five replications per OTU) studied in the field. Mean values obtained from five independent replications were used as representative OTU data for each quantitative morphological descriptor. If the sample characteristics conformed to descriptors, they were marked as "0"; if not, they were marked as "1. " The scored qualitative and quantitative interval data were standardized to construct a dendrogram using neighbor-joining  (Table A2).

Results
Phylogenetic analyses. In this study, all five markers, rpl32-trnL, rpl16, matK, rbcL, and trnH-psbA, were independently detected by MP. Based on the five MP trees and supplemented by information on diversity acquired using DnaSP v5 32 , phylogenetic analyses of Bambusa were performed using combined DNA barcoding (rpl32-trnL+rpl16). The nucleotide (Pi) and haplotype (Hd) diversities of these five DNA barcodes indicated that rbcL and trnH-psbA were not suitable for identifying Bambusa as their Pi values were much lower than those of rpl 32-trnL, rpl 16, and matK at 0.00458 and 0.00406, respectively (Table 3). In contrast, rpl32-trnL, rp16, and matK appeared to be reasonable barcoding candidates for identifying Bambusa species according to the diversity information available. While matK, rbcL, and trnH-psbA could separate Bambusa from other genera, they grouped >70% of the sampled Bambusa into one cluster. In comparison, rpl32-trnL and rpl16 both divided Bambusa into several clusters.
A combined barcode (rpl 32-trnL+rpl 16) was also used to analyze the phylogeny of 66 taxa after the ILD was tested. The p-value of the ILD was 0.05 and the combined marker successfully divided bamboo into several clusters, as shown in Fig. 1 (left). The tree length, consistency index (CI), and retention index (RI) of the MP analyses for rpl32-trnL+rpl16 were 2392, 0.79, and 0.91, respectively. The BVs were mapped onto the MP topologies and shown as figures behind the branches. The analysis based on the rpl32-trnL+rpl16 combined dataset divided the entire group into three major clusters (A [100 BV], B [100 BV], C [61 BV]), with cluster A as an outgroup (Fig. 1,  left), constituting members of the Shibataea, Drepanostachyum. Phyllostachys, Pseudosasa, Oligostachyum, Sinobambusa, Indosasa, and Pleioblastus. However, three species, Dendrocalamus minor var. amoenus, Dendrocalamus membranaceus, and Melocanna baccifera (cluster C) were considered to form an outgroup and were therefore not included. Bambusa taxa formed two major clusters, B and C, and cluster C was further divided into four sub-clusters (C1 [54 BV], C2 [64 BV], C3 [66 BV], and C4 [52 BV]) and several monotypic and small clades.
The branching pattern of the dendrogram was not completely consistent with the classical taxonomic classification of Bambusa proposed by the FOC 2 , especially at the subgenus level. The subgenus Lingnania (blue strip in Fig. 1) contained the greatest number of species sampled in this study, while members the subgenus Bambusa (pink strip) were scattered among all the Bambusa clusters. The subgenus Leleba (green strip) did not appear  Analyses of morphological characteristics. In the absence of flower or fruit characteristics, culm sheaths and characteristics were treated as two taxonomic features for classifying Bambusa. According to 186 key morphological descriptors, the entire dendrogram (Fig. 1, right) was split into three clusters (H, I, and J). One main cluster (H) was divided into two sub-clusters (K and L) and four clusters (K, L, I, and J) did not completely conform to the existing classification. For instance, B. textilis and B. teres were totally isolated in a small cluster (J) belonging to the subgenus Leleba and species in the subgenus Lingnania were all in one subclade of cluster L. Critically, the subgenera Bambusa and Leleba were not separated from one another. Meanwhile, varieties and cultivars, such as B. vulgaris and B. multiplex, were more likely to stay with their McClure classifications. Bambusa vulgaris and its two cultivars formed a small clade in cluster I and B. chungii and B. chungii var. velutina were grouped into cluster L. The varieties and cultivars of B. multiplex were placed into cluster K; B. textilis, B. textilis cv. Purpurascens, and B. textilis var. gracilis were split into clusters J, L, and K, respectively, and B. tuldoides cv. Swolleninternode and B. tuldoides were split into clusters I and L.
Topological congruences. The MP topology analyses were largely inconsistent with the morphological analysis, but cluster C2 in the MP analysis was largely consistent with cluster K in the morphological analysis. Morphological characteristics analyses and subgeneric classification. According to the FOC, the genus Bambusa has four subgenera: Lingnania, Dendrocalamopsis, Bambusa, and Leleba. The subgenus Lingnania was found to share the following typical characteristics: a culm sheath with a narrow blade, a base only one-third of the width of the sheath apex; culm internodes that are usually longer than 30 cm, and thin walls (often < 8 mm). Three other subgenera shared the following characteristics: a culm sheath with a broad blade, a base 1/2-3/4 of the width of the sheath apex; culm internodes shorter than 30 cm, and thick walls (up to 2 cm). Meanwhile, the subgenus Dendrocalamopsis shared the following typical characteristics: culm sheath auricles and small, and rounded spikelets that are dense at maturity. The rest of the subgenera shared the following characteristics: culm sheath auricles that are large, rounded, irregular, or absent and spikelets that are loose at maturity, with broad florets on short rachilla segments. Otherwise, the characteristics of the subgenus Bambusa were found to be branchlets of lower branches specialized into tough or weak leafless thorns, and with culm sheaths with persistent blades. The subgenus Leleba had branchlets in their lower branches that were normal and leafy; and their culm sheath blade was deciduous.
To the best of our knowledge, this study represents the first attempt to distinguish Bambusa subgenera by using 186 morphological descriptors to sample more than 50 Bambusa taxa. As mentioned above, the traditional classification uses eight to 14 morphological characteristics to identify a subgenus, which are fewer than the number of morphological characters used in this study. Therefore, it is not surprising that the morphological phylogenetic tree generated here did not coincide exactly with the existing Bambusa subgenus classification. Establishing a phylogenetic tree based on morphological characteristics is a novel way to explore bamboo classification. According to the findings of this approach, we described more than 39 morphological features as 186 key morphological descriptors. Thus, the results were focused more on the overall characteristics of each species, rather than on one or several obvious or easily identifiable features. www.nature.com/scientificreports/ based on morphological characters and molecular analysis. Instead, they found that, from a morphological perspective, these four bamboo species (B. arundinacea, B. vulgaris, B. auriculata, and B. striata) differed from each other, and B. striata and B. vulgaris showed greater similarity to each other than the others in RAPD analysis.

Controversial bamboo species.
Here, B. auriculata and B. striata were not sampled, and the data of morphological characteristics and DNA sequence between B. arundinacea and B. vulgaris were different in this study. B. chungii var. velutina is a new variant of B. chungii that, to date, has only been found in the Fujian province of China. It was previously considered as a member of the genus Lingnania, but is now considered to be a sub-genus of Bambusa. Here, B. chungii and B. chungii var. velutina were found to be similar in both MP and morphological characteristic analyses.
Application of the codes. Compared with flowering plants, the classification of bamboos is more challenging for researchers and workers that are not engaged in examining phylogenetic relationships. Using DNA barcodes to classify or identify species will be more widely applied with the growth of molecular biology technology because of its easy operability, even though it may not align perfectly with traditional botanical classification. The codes rpl32-trnL and rpl16 are two loci on plastid DNA. Phylogenetic analyses that are based on whole chloroplast genomes have been used to resolve relationships within the subfamily Bambusoideae 5 . Wang 34 et al. suggested the use of a larger dataset, indicating that insufficient parsimony information characters were the main cause for poor resolution in temperate bamboos.
Based on morphological features, morphological codes were used as a classification method to evaluate whether they could be a match for traditional classification. However, following statistical analysis, results showed that it could not be appropriately explained in the context of morphological classification. A new operating model for morphological codes needs to be developed for the application of this technique in botanical classification.