The main morphological characters of I. sangzhiensis and three species of Chimonobambusa, viz., C. communis, C. opienensis and C. puberula, were compared based on protologues and descriptions from floras. Some detailed characters, such as indumentum of the culm leaf ligule, were observed with a stereomicroscope (Mshot MZ101). Measurements were taken using a ruler or micrometer.

Two different molecular regions including complete chloroplast genome (cpDNA) and nuclear ribosomal DNA (nrDNA) were utilized to reconstruct the phylogenetic relationships of I. sangzhiensis. A total of 23 species representing 7 genera from Arundinarieae were utilized to reconstruct the plastid tree, for which Bambusa emeiensis L.C.Chia & H.L.Fung and B. sinospinosa McClure were set as the outgroup taxa. For nrDNA, 16 species representing 5 genera from Arundinarieae were utilized to reconstruct the nrDNA tree, for which B. sinospinosa and B. multiplex (Lour.) Raeusch. ex Schult.f. were set as the outgroup taxa. The generic type of Indosasa McClure, I. crassiflora McClure, was added here to clarify the systematic position of I. sangzhiensis. All voucher information and accession number of cp genomes are listed in Table 1. Our sample information for nrDNA and its sequence matrix can be found in the supplementary material (Suppl. material 1: Table S1).

Total genomic DNA was isolated from silica-dried leaves following manufacturer specifications TIANGEN Genomic DNA Extraction Kit (TIANGEN, Beijing, China). DNA samples of concentration up to standard (≥1 μg) were randomly sheared into fragments using Covaris M220 (Covaris, Woburn, MA). Insert size of 350 bp fragments were enriched by PCR, and the paired-end (2 × 150 bp) libraries were constructed on NovaSeq 6000 platform. As a result, about 20 G genome skimming data were generated.

To improve assembly accuracy and efficiency, Trimmomatic v 0.39 were utilized to filter out unpaired and low-depth reads from clean data using default parameters (Bolger et al. 2014). The whole cpDNA and nrDNA were assembled using the software GetOrganelle v 1.7.4 pipeline (Jin et al. 2018), with five k-mer sets of 45, 65, 85, 105, 125 bp. The filtered reads were transferred to Bandage (Wick et al. 2015) for plastid and ribosomal DNA scaffolds connection. Two opposite plastid scaffolds exported from Bandage were aligned with the reference I. shibataeoides (MF066251), and the one matching the reference was annotated using the PGA software (Qu et al. 2019) based on the annotation of I. shibataeoides. The sequences were finally checked manually in Geneious 9.1.4 (Kearse et al. 2012). The assembled nrDNA sequences were annotated directly using Geneious. Illustration of the newly sequenced plastome of I. sangzhiensis was drawn by OGDRAW with default settings (Greiner et al. 2019).

To determine the systematic position of I. sangzhiensis, maximum likelihood (ML) and Bayesian inference (BI) analyses were conducted. A total of 22 complete cp and 16 nrDNA genomes were aligned with MAFFT v 7.450 (Katoh and Standley 2013). Maximum likelihood (ML) analysis was generated by RAxML v 8.0.0 (Stamatakis 2014). Rapid bootstrap analysis and GTRGAMMAI were set as the best-fit algorithm and model. The number 12345 was specified as the random seed of parsimony tree inference with 1000 replicates performed. Bayesian inference (BI) was conducted using MrBayes v 3.2.6 (Ronquist et al. 2012). The model of SYM+G was defined by MrModeltest v 2.4 (Nylander 2004). The rates of variations across sites were trimmed as gamma. At least 6,000,000 generations were run to ensure average standard deviation of split frequencies (ASDFs) < 0.01 with sampling frequency set as 100 generations. After rejecting the first 25% burn-in samples, the optimized topology was printed.

The plastid genome of I. sangzhiensis exhibited a typical quadripartite structure. Its genome size is 139,595 bp including a large single copy region (LSC) of 83,190 bp, a small single copy region (SSC) of 12,811 bp and a pair of inverted repeat regions (IRs) of 21,797 bp (Fig. 1A). The overall GC content is 38.9%. The whole plastid genome contains 108 unique genes, including 76 protein-coding genes, 28 transfer RNAs and 4 ribosomal RNAs. Among them, 43 genes are associated with photosynthesis and 59 genes are relevant to gene expression.

Figure 1. 

The plastome map A and nrDNA structure B for I. sangzhiensis.

For the tandemly repeated nrDNA, our de novo assembly obtained 5,799 bp sequences comprising 18S (1,811 bp), 5.8S (164 bp), and 26S (3,391 bp) ribosomal RNA gene along with two internal transcribed spacer I (ITS1) (215 bp) and ITS2 (217 bp) in the middle (Fig. 1B).

Indosasa sangzhiensis resembles C. communis, C. opienensis and C. puberula in having slightly 4-angled basal culm internodes, flat or slightly prominent culm nodes without root thorns, deciduous culm leaf sheaths, very small, erect and narrowly triangular or subulate culm leaf blades and well-developed foliage leaf oral setae; but can be easily distinguished from the latter three species by the morphological characters shown in Table 2. Specifically, it can be distinguished from C. communis in having culm internodes with dense white pubescence (Fig. 2B–E) (vs. glabrous internodes), culm leaf ligule margins with dense cilia (Fig. 3C) (vs. glabrous margin) and culm leaf blades broadly extending to join the sheath apex (Fig. 3B) (vs. slightly narrowed at the junction with the sheath apex). It differs from C. opienensis in having culm internodes with dense white pubescence (Fig. 2B–E) (vs. glabrous), 3 branches per mid-culm nodes (Fig. 2D, E) (vs. 2–3 branches per mid-culm node), culm leaf sheaths of glabrous adaxial surface (Fig. 3A) (vs. with sparse brown setose), culm leaf ligule margins with dense cilia (Fig. 3C) (vs. glabrous) and 1–3 foliage leaves per ultimate branch (Fig. 3D) (vs. 1(–2) foliage leaves). It differs from C. puberula in having glabrous rhizome internodes (Fig. 3E) with densely brown-setose parts below the nodes and glabrous rhizome internodes sheath scars (Fig. 2C) (vs. with densely brown setose); culm leaf sheaths with glabrous abaxial surface (Fig. 3A) (vs. with densely brown setose surface); culm leaf ligule margins with dense cilia (Fig. 3C) (vs. glabrous margin) and new shoots produced during April to June (vs. October).

Figure 2. 

The holotype of I. sangzhiensis B.M.Yang (HNNU).

Figure 3. 

Morphological characters of I. sangzhiensis A habit B culm leaf sheath on internodes covered with dense white pubescence C culm bud, node and sheath scar D two longitudinal ridges and three grooves above branches D, E branches at mid-culm nodes, young D and old E culm. All photos by Z.Y.Niu.

Figure 4. 

Morphological characters of I. sangzhiensis A culm leaf, detached B culm leaf blade C culm leaf ligule D ultimate leafy branches E foliage leaf oral setae F rhizome with partial young culms. All photos by Z.Y.Niu.

The topologies based on ML and BI methods did not indicate any conflict between the cpDNA and nrDNA phylogenetic analyses, thus only ML cladograms are shown here (Figs 5, 6), with BP and PP values noted at each node. The plastid and nrDNA phylogenetic trees both strongly supported the case that I. sangzhiensis is distantly related to I. crassiflora (the type of Indosasa) (BS = 100% & PP = 1.00), but forms a sister clade with members of Chimonobambusa (BS = 95% & PP = 1.00) (Fig. 5). Our results also indicate that all samples of Chimonobambusa formed a monophyletic clade while Indosasa was found to be polyphyletic, as the samples of Indosasa and those of two other genera, viz. Acidosasa and Oligostachym, were intermixed within a clade.

Figure 5. 

Phylogenetic tree reconstruction for I. sangzhiensis based on plastid genome dataset with maximum likelihood and Bayesian analyses. Only bootstrap values (BS) ≥ 70% and posterior probabilities (PP) ≥ 0.95 are indicated at each node.

Figure 6. 

Phylogenetic tree reconstruction for I. sangzhiensis based on nrDNA dataset by using maximum likelihood analysis and Bayesian analysis. Only bootstrap values (BS) ≥ 70% and posterior probabilities (PP) ≥ 0.95 are indicated at each node.