Straight From the Plastome: Molecular Phylogeny and Morphological Evolution of Fargesia (Bambusoideae: Poaceae)

Fargesia is ecologically and economically important in mountainous forests. Many Fargesia species are also important sources of food for some endangered animals such as the giant panda. Recent molecular phylogenetic analyses have revealed Fargesia as a polyphyletic group despite some unclear lineage affinities. In the present study, we reconstructed the phylogeny of Fargesia and its allies, including Thamnocalamus, Arundinaria (incl. Bashania), Yushania, Indocalamus, Ampelocalamus and Phyllostachys, from a plastome sequence matrix that contained 20 Fargesia and five Yushania species as ingroups, 16 species from nine other bamboo genera plus Oryza sativa and Zea mays as outgroups. Fargesia and its allies were broken into eight clades. Several Fargesia species were clustered into the Thamnocalamus clade and the Drepanostachyum + Himalayacalamus clade that rendered the polyphyly of Fargesia. The remaining six clades, including the Fargesia spathe clade, the Phyllostachys clade, Arundinaria fargesii, the Ampelocalamus clade, the Fargesia grossa clade, and the Fargesia macclureana clade, were identified. Molecular phylogenetic analyses supported that Yushania should be included in Fargesia (s.l.) which had synapomorphy of expanded leaf sheaths in varying degree at the basis of inflorescences, and further divided into the Fargesia spathe clade, the Fargesia grossa clade, and the Fargesia macclureana clade. All sampled species of Yushania were nested within the Fargesia grossa clade. The probable model of the origin of the species in the Fargesia spathe clade with spathe-like leaf sheath syndrome was proposed. Moreover, the formation of the spathe-like leaf sheath syndrome may be correlated with cold climatic conditions in Quaternary. Our results provide new sight into the phylogenetic relationship within Fargesia.

The infrageneric classification of Fargesia was established mainly based on vegetative characteristics including the shape of buds and culm sheath blades, and the number and thickness of branches (Yi, 1988a). However, high rates of morphological divergence may result in unstable generic/infrageneric classifications due to parallel evolution or morphological homoplasy (Wei et al., 2018), which prevail in vegetative traits of bamboos (Guo and Li, 2004;Attigala et al., 2016;Zhang et al., 2019). Therefore, it needs to be tested whether the current morphologically based infrageneric classification of Fargesia (Yi, 1988a(Yi, , 1996Yi and Jiang, 2010;Bamboo Phylogeny Group, 2012) is credible. Recent systematic studies revealed that the Fargesia + Yushania clade based on RAD-data (Wang et al., 2017) or plastid regions and nrITS (Zhang et al., 2019), including most Fargesia species and all sampled Yushania species, could be further divided into the Fargesia spathe clade and the non-spathe clade according to their morphological traits of inflorescences and distribution rather than the key vegetative characteristics listed by Yi (1988a) for the subdivision of Fargesia (Zhang et al., 2019). Moreover, the species in the Fargesia spathe clade shared a spathe-like leaf sheath syndrome, i.e., a series of spathe-like leaf sheaths that surround compressed inflorescences with the top spathe generally longer than the inflorescences and the spikelets exserted partially and uni-laterally, which was hypothesized to be an adaptation to cold environments in the north (Zhang et al., 2019). However, the phylogenetic relationships among major subclades of the non-spathe clade including the rest of Fargesia species and all sampled Yushania species are still poorly understood, with species from both genera being intermingled with other taxa such as Arundinaria, Indocalamus, and Chimonocalamus (Ma et al., 2014;Wang et al., 2017).
Sparse taxon sampling and insufficient informative characters had been the problems for investigating the phylogeny of Fargesia. Thus, a high-quality phylogeny, including more representative taxa and a large number of genetic markers, is necessary to understand the evolutionary history of Fargesia.
Full plastome sequencing has been shown to be useful in resolving complex evolutionary relationships in closely related species (Njuguna et al., 2013;Williams et al., 2016;Uribe-Convers et al., 2017;Fonseca and Lohmann, 2018). Despite the existence of previous phylogenetic analyses based on the plastomes of Arundinarieae (Ma et al., 2014;Attigala et al., 2016;Saarela et al., 2018), the phylogenetic relationships among Fargesia species are still poorly resolved. Our aims were to (1) test the monophyly of Fargesia; (2) establish a phylogenetic framework for Fargesia and closely related taxa; (3) evaluate the utility of whole chloroplast genome sequencing in providing a high-resolution estimation of the infrageneric phylogeny within Fargesia.

Taxon Sampling
Three to five fresh leaf samples from each individual (clumps) were collected in their native habitats and dried in silica gel and stored at room temperature for DNA extraction ( Table 1). Twenty-six species, including the type species for each section and series in Yi's (1988a) classification of Fargesia, were selected for Illumina sequencing. We also included the representatives of the major subclades in the Fargesia + Yushania clade based on the previous study by Zhang et al. (2019) and outgroups representing genera within the Phyllostachys clade of the Arundinarieae (since their phylogenetic relationship is unclear with Fargesia; see Ma et al., 2014) (Table 1). All vouchers were deposited in the Herbarium of Shaanxi Normal University (SANU, Xi'an, China). Seventeen previously published plastomes were downloaded from the NCBI GenBank database 1 ( Table 1).

Plastomes Sequencing, Assembly, and Annotation
Silica-dried leaf material was sent to the Center for Genetic and Genomic Analysis (Genesky Biotechnologies Inc., Shanghai) for library preparation and Illumina sequencing. Total genomic DNA was isolated using a DNeasy Plant Mini Kit (Qiagen, CA, United States). Short-insert (330 bp) libraries were constructed using the TruSeqTM DNA sample preparation kit (Illumina, United States) for Illumina HiSeq sequencing. We sequenced at least 2 GB of raw data for each bamboo species with an average read length of 150 bp. All of the sequencing reactions were conducted on the Illumina HiSeq 2 × 150 platform. The obtained raw reads were qualitytrimmed using CLC Genomics Workbench v.8.5.1 (CLC Bio, Aarhus, Denmark) 2 with default parameters. After trimming, high-quality paired-end reads were MIRA mapping assembly using MIRA v4.0.2 (Chevreux et al., 2004) to create initial reference sequences. These newly created reference sequences were baiting and iterative mapping with MITObim v1.8 (Hahn et al., 2013) to assemble the complete plastome sequences, and the previous published complete plastome sequences of Fargesia spathacea (GenBank no., JX513417), Yushania levigata (GenBank no., JX513426), Pleioblastus maculatus (GenBank no., JX513424), Ampelocalamus calcareus (GenBank no., KJ496369), and Arundinaria fargesii (GenBank no., JX513413) were used as seed reference. Raw reads were then remapped to newly assembled complete plastome sequences using Geneious v R9.0.2 (Kearse et al., 2012) under default parameters to check for misassemblies. Small gaps (46-193 bp) or ambiguous nucleotides in the remapped sequences were further confirmed by designed PCR amplifications (Supplementary Table S1) and sent to Sangon Biotech (Shanghai) Co., Ltd for Sanger sequencing. Gene annotations were assigned using DOGMA (Wyman et al., 2004). Inverted repeat (IR) boundaries were determined using the BLAST method (Altschul et al., 1990) 3 and verified using Geneious v R9.0.2. The genome map of Fargesia denudata was drawn using the online program OGDRAW v 1.1 (Lohse et al., 2013).

Data Partitioning and Alignment of Plastomes
We employed six datasets to reconstruct the phylogenetic trees: (1) the complete chloroplast genome sequences that excluded one copy of the IR region; (2) coding sequences including exons of protein-coding genes, tRNAs, and rRNAs (3) noncoding sequences including intergenic regions and introns; (4) the large single copy region (LSC), (5) the small single copy region (SSC) and (6) one inverted repeats region (IR). All 43 sequences from these six datasets were aligned using the MAFFT plugin (Katoh et al., 2002) in Geneious v.R9.0.2 with default parameters. As there is a negative effect of an increasing density of gap characters in multiple sequence alignments (Dwivedi and Gadagkar, 2009) and biases may be introduced when indels are included without precautions in Bayesian and Maximum Likelihood phylogenies (Warnow, 2012;Simmons, 2014), poorly aligned regions contained one or more gaps introduced by the alignments were excluded from the datasets (Attigala et al., 2016). Two different partitioning strategies were explored to access heterogeneity in the processes of molecular evolution for different sites in a sequence alignment. The first strategy included two partitions: coding sequences and non-coding sequences. The second strategy included three partitions: LSC, SSC, and one IR regions we then used PartitionFinder v 1.1.1 (Lanfear et al., 2012) to compare and evaluate partitioning schemes under 56 models (models = all) according to AICc values.

Phylogenetic Analysis
We used DAMBE v6.4.29 to assess substitution saturation of the matrices before phylogenetic analyses (Xia et al., 2003;Xia and Lemey, 2009). Six datasets mentioned above with unpartitioning strategies were analyzed using Maximum likelihood (ML), Maximum parsimony (MP), and Bayesian inference (BI). ML analysis with rapid bootstrapping and 1000 replications (RAxML; Stamatakis et al., 2008) was performed using the General Time Reversible + gamma model (GTR-GAMMA model) via the CIPRES Science Gateway server (Miller et al., 2010). MP analysis was implemented in PAUP * 4.0b10 (Swofford, 2002)  with 1000 random addition sequences and tree bisectionreconnection (TBR) branch swapping. A full heuristic bootstrap was conducted for MP with 1000 bootstrap replicates. The appropriate model for each dataset was selected according to AIC value (Posada and Crandall, 1998) implemented in jModelTest v2.1.4 (Darriba et al., 2012). GTR+I+G was found to be the most appropriate nucleotide substitution model for 43 complete chloroplast genome sequences, 41 complete chloroplast genome sequences, coding sequences, non-coding sequences, and LSC sequences; TVM+I G for SSC sequences; TVM+I for IR sequences; TIM1+I+G for 39 complete chloroplast genome sequences. BI analysis was performed using MrBayes v3.2.6 (Ronquist and Huelsenbeck, 2003). Two independent Markov chain Monte Carlo (MCMC) runs were executed, each with three heated and one cold chain for two million generations, sampling every 1000 generations, and a chain-heating temperature of 0.1. Twenty-five percent of trees were discarded as burn-in and the remaining samples were then summarized and a majority-rule consensus tree was constructed. Convergence of runs was also checked using Tracer v 1.6 (Rambaut et al., 2014) to assess the effective sample size (ESS) > 200 for all parameters.

Topology Hypothesis Testing
Taxon removal experiments can aid in the investigation of whether distantly related outgroups have a biased attraction to long branches within the ingroups (Bergsten, 2005;Attigala et al., 2016). To test if outgroup selection affected the topology of the ingroups, we recovered phylogenetic trees with two datasets: (1) removing Zea mays and Oryza sativa to include only species of bamboos; (2) further removing species of Bambuseae to leave only species of Fargesia and its allies. Assessments of competing hypotheses (e.g., the monophyly of the Fargesia spathe and non-spathe clades) of the phylogeny were conducted with the approximately unbiased (AU) test (Shimodaira, 2002) as implemented in CONSEL (Shimodaira and Hasegawa, 2001). The site-wise log likelihoods for the trees were calculated using TREE-PUZZLE v5.3 (Schmidt et al., 2002). Neighbor-Net algorithm based on uncorrected P-distances was performed with SplitsTree v4.14.4 (Huson et al., 2008). To better visualize and evaluate the phylogenetic signal conflicts between ingroup taxa, we excluded all outgroup taxa in the SplitsTree analysis.

Calibration and Estimation of the Divergence Time
A sequence matrix including 43 complete plastomes was created for divergence time dating. A molecular clock test was performed under the GTR model to assess the applicability of the strict molecular clock model. Our results showed that lnL = 429 (P < 0.05), so the relaxed log-normal molecular clock was used. Divergence time dating was analyzed using BEAST v2.4.2 (Bouckaert et al., 2014) on the Cipres web server (Miller et al., 2010) with a Yule tree prior. We selected a Bambusoideae cf. Chusquea fossil (ca. 35 Mya) to calibrate the crown Bambusoideae (Strömberg, 2005). Runs were performed for 60 million generations, and trees were sampled every 1000 generations. Tracer v 1.6 (Rambaut et al., 2014) was used to check that all ESS values were > 200. The maximum clade credibility tree was generated by TreeAnnotator with mean node heights and 95% highest posterior density intervals. The temporal dynamics of the diversification of the major clades within Fargesia was visualized by constructing lineages through time (LTT) plots implemented by the APE package in R v.3.2.4 (Paradis et al., 2004).

Plastome Phylogenomics
The multiple sequence alignment of all 43 species (including both newly generated and previously published plastomes) was 124,853 bp after excluding one of the IR regions. After removal of alignment columns with gaps, the alignment length was reduced to 108,515 bp, including 8841 (8.15%) variable sites and 2187 (2.02%) parsimony-informative sites ( Table 2  and Supplementary Table S3). The aligned sequences of six sequence datasets mentioned above were little saturated and were thus useful for phylogenetic analyses (Supplementary Table S4). The less partitioned model was found to be better than the highly partitioned model (Supplementary Table S5). The topologies estimated from the complete chloroplast sequences (Supplementary Figure S2F) were broadly similar to those estimated using the other datasets  Figure S2).

Topology Testing
Since the tree topologies of the major clades based on two taxon removal experiments showed no changes when outgroups were removed (Supplementary Figures S3, S4), we concluded that the tree topology was not affected by long branch attraction.
The incompletely resolved relationships in the network suggested substantial conflicting signals (Figure 2). All species were clearly sorted into respective clades, and the clades were almost distinct. However, the relationships among the Fargesia macclureana clade, the Drepanostachyum + Himalayacalamus clade, and the Thamnocalamus clade were less clear. We used the AU test to evaluate alternative tree topologies for the relationships among the major clades of Fargesia and its allies (Supplementary Figure S5). The results based on the complete plastomes sequences and non-coding sequences rejected the monophyly of the Fargesia spathe clade (Supplementary Figure S5A), the Fargesia grossa clade and the Fargesia macclureana clade and the monophyly of the Fargesia grossa clade and the Fargesia macclureana clade (Supplementary Figures S5A,C). However, the results based on coding sequences failed to  (Yi, 1996), Flora of China (Li et al., 2006;Zhang and Ren, 2016). Figure S5B) involving the Fargesia grossa clade, the Fargesia macclureana clade, and the Drepanostachyum + Himalayacalamus clade.

Morphological Characters Evolution
Most extant and ancestral taxa have cespitose culms and sympodium. Diffused culms and amphipodium (e.g., Ar. faberi and l. longiauritus) or monopodium (e.g., Phyllostachys species) have evolved independently in several species of different lineages in parallel (Figures 4A,B). Spathe-like leaf sheath syndrome is only restricted to the Fargesia spathe clade (Figure 4C), whereas species in the Fargesia grossa clade and the Fargesia macclureana clade have open inflorescences with expanded leaf sheaths in varying degree (Figures 1C,D,H). The semelauctant inflorescences seem to have evolved from iterauctant inflorescences, followed by a reversal in Phyllostachys  Figure 4D). The ancestral state of the number of stigmas in Arundinarieae was ambiguous due to some floral characters are unknown in outgroups ( Figure 4E). The selecting of different outgroups may affect the evolutionary trend of the number of stigmas. The situation is more complex within Fargesia (s.l.) due to the presence of both pistils with two stigmas and three stigmas taxa in the Fargesia spathe clade, the Fargesia grossa clade and the Fargesia macclureana clade. Thus, the pistils with two or three stigmas might arise independently in Fargesia (s.l.).

Polyphyly of Fargesia
Our phylogenetic analyses based on 43 complete plastome sequences, which represented a greater diversity known in Fargesia and its closely related species, confirmed that Fargesia (Yi, 1988a) is not monophyletic, which was consistent with previous analyses (Guo et al., 2002;Guo and Li, 2004;Triplett and Clark, 2010;Zeng et al., 2010;Wang et al., 2017;Zhang et al., 2019). In the present study, Fargesia damuniu and T. spathiflorus formed a monophyletic clade, which is in line with the previous recognition of a strongly supported the Thamnocalamus clade consisting of F. damuniu (Zhang et al., 2019) or F. crassinoda (Guo and Li, 2004;Zhang et al., 2019) and other Thamnocalamus species. The phylogenetic tree based on plastomes also supported that F. gyirongensis and F. collaris would be included in Drepanostachyum + Himalayacalamus clade. Phylogenetic relationships among the Drepanostachyum + Himalayacalamus clade, the Fargesia grossa clade, and the Fargesia macclureana clade varied depending on data partitioning. Differing rates, insufficient phylogenetic signals, and patterns of nucleotide substitution may be the cause of conflicts between partitioned data and unpartition data, as summarized by Ma et al. (2014). The isolated position of the Drepanostachyum + Himalayacalamus clade in the present study coincides well with the results from other studies that used nuclear genes (Hodkinson et al., 2010;Wang et al., 2017;Zhang et al., 2019) or plastid genes (Triplett and Clark, 2010;Zeng et al., 2010;Zhang et al., 2019). Furthermore, the morphology evidence (Keng, 1983;Ohrnberger, 1996;Li et al., 2006;Liu et al., 2017) supported F. damuniu, F. gyirongensis, and F. collaris should not be considered as Fargesia species.
Meanwhile, the phylogenetic relationships recovered in the present study corroborated Yushania species nested with Fargesia (Guo et al., 2001(Guo et al., , 2002Guo and Li, 2004;Wang et al., 2017;Zhang et al., 2019). Yushania was considered to be the most closely allied genus to Fargesia with similar vegetative characters, inflorescence types and habitats (Keng, 1957(Keng, , 1984(Keng, , 1987Wang andYe, 1980, 1981;Yi, 1982Yi, , 1985aYi, , 1988aYi, , 1996Keng and Wen, 1989;Stapleton, 1994aStapleton, , 1998Wang, 1997;Guo et al., 2001Guo et al., , 2002Guo and Li, 2004;Li et al., 2006). Although that opinion was supported by molecular systematic studies, the results were weakly supported and fluctuated depending on the gene and taxon sampling (Guo et al., 2001(Guo et al., , 2002Guo and Li, 2004;Triplett and Clark, 2010;Zhang et al., 2012;Ma et al., 2014). The recent studies based on a wider taxon sampling revealed that the sampled species of Yushania were nested within Fargesia (Wang et al., 2017;Zhang et al., 2019). Our backbone topology based on plastome data further revealed that all sampled Yushania species were nested within the Fargesia grossa clade. This result was congruent with the convergence of morphological traits. When Yushania was published, Keng (1957) stated that Yushania was different with Fargesia in aspects of rhizomes, simple inflorescences with only three spikelets and the acute lemma. Among those three traits, only the morphology of the rhizome was considered as the actual difference between Fargesia and Yushania (Yi, 1996;Li et al., 2006). Actually, the length and thickness of culm-necks in Fargesia and Yushania were considered as one of the highly variable traits (Wang et al., 2017) as the intermediate states were common (Yi, 1996;Guo and Li, 2004;Li et al., 2006). The pulvini at the point of the inflorescence branches were considered as an identifying characteristic of Yushania (Stapleton, 1994a), except for Y. lineolate which do not have pulvini, while Fargesia species have a synflorescence without pulvini except for F. yunnanensis with pulvini at the base of the inflorescence branches (Yi, 1996). A very small bract at the base of each of inflorescence branches in Yushania, which was considered as bract (Wang and Ye, 1981;Keng, 1982) or leaf sheath (Wang and Ye, 1980;Yi, 1982Yi, , 1985aYi, ,b, 1988aYi, , 1996Li et al., 2006;Yi et al., 2008;Yi and Jiang, 2010), is another identifying trait between these two genera (Stapleton, 1994a). However, a small bract exists at the base of each of inflorescence branches in the lower part of inflorescences in F. lincangensis and F. yunnanensis indicating the transition of this trait (Yi, 1996;Yi et al., 2008). Therefore, Fargesia and Yushania could not be distinguished by above-mentioned morphological traits. Yi (1988a) assumed that on the premise of the similar floral morphology and the same position of a series of leaf sheaths subtending at the base of inflorescences, the small or delicate leaf sheaths in Yushania might evolve from the small or large leaf sheaths in Fargesia. Fargesia species have a series of leaf sheaths which expand to varying degrees, or even to spathes subtending at the basis of inflorescences (Wang andYe, 1980, 1981;Keng, 1982;Yi, 1982Yi, , 1985aYi, ,b, 1988aYi, , 1996Wang, 1997;Li et al., 2006;Yi et al., 2008;Yi and Jiang, 2010) and spikelets are exserted partially and uni-laterally from leaf sheaths (Keng, 1982;Yi et al., 2008), while those of Yushania have small or delicate leaf sheaths which are not spathe-like (Wang andYe, 1980, 1981;Keng, 1982;Yi, 1982Yi, , 1985aYi, ,b, 1988aYi, , 1996Wang, 1997;Li et al., 2006;Yi et al., 2008;Yi and Jiang, 2010) and all spikelets are exserted from leaf sheaths (Keng, 1982). However, the transitional states of this trait exist in F. yunnanensis and Y. brevipaniculata. All spikelets of F. yunnanensis are exserted from the delicate leaf sheaths (Yi, 1988a(Yi, , 1996, while the spikelets of Y. brevipaniculata are exserted slightly from the non-expanded leaf sheaths (Yi, 1996). Based on basic morphological and positional similarities and the present phylogenetic analyses, we considered that a series of leaf sheaths, regardless of size, at the base of inflorescences might be a homological and synapomorphic trait in Fargesia and Yushania (Figures 1B,C, 4C). Combined with the results from our data and previous phylogenetic studies (Wang et al., 2017;Zhang et al., 2019), we suggested that Yushania might be included in Fargesia (s.l.).
In the present study, Arundinaria fargesii, Phyllostachys clade and Ampelocalamus clade are successive sisters to the Fargesia spathe clade, which was incongruent with nuclear regions (Zhang et al., 2012;Wang et al., 2017). Arundinaria fargesii was sister to the Fargesia spathe clade based on our plastomes analyses, in line with previous plastid markers analyses (Zhang et al., 2012) or plastome phylogenomics (Ma et al., 2014). However, Ar. fargesii may be a hybrid between F. decurvata and Bashania aristata based on GBSSI analyses or originated from hybridization between Chinese Pleioblastus and Indocalamus based on morphological features (Zhang et al., 2012). Thus, hybridization or plastid introgression may be a reasonable explanation for different phylogenetic positions of Ar. fargesii.
Phyllostachys species were clustered within the Fargesia (s.l.) in our plastomes analyses, while they grouped together with other species with leptomorph rhizomes such as Brachystachyum and Shibataea based on GBSSI gene tree (Zhang et al., 2012) or Pleioblastus, Acidosasa, Indosasa, and Shibataea based on RAD-seq data (Wang et al., 2017), which confirmed an isolated position from species of Fargesia (s.l.). Phyllostachys species have leptomorph rhizomes and iterauctant inflorescences which have few morphological similarities compared to species of Fargesia (s.l.). Therefore, plastid introgression or lineage sorting may be alternative causes for the incongruence.
The previous molecular studies revealed that Am. melicoideus and other Ampelocalamus species (excluding Am. calcareus and Am. actinotrichus) grouped together with Drepanostachyum and Himalayacalamus based on nuclear gene markers (Guo et al., 2002;Guo and Li, 2004;. The plastomes of Ampelocalamus, Fargesia, Yushania, and Arundinaria (incl. Bashania) might share the same origin. Plastid capture or incomplete lineage sorting could be an explanation for the incongruent phylogenetic position of Ampelocalamus species between plastid gene markers and nuclear gene markers .
The inflorescence traits also provided a clue for the differences among Fargesia (s.l.), Phyllostachys species, Arundinaria fargesii and Ampelocalamus species such as semelauctant synflorescences or iterauctant synflorescence, leafy or leafless flowering branches, and the position of inflorescence-affiliated leaf sheaths, prophylls, and bracts (Figures 1A,B,E,G, 4C). For example, the inflorescences of Am. melicoideus interrupted clustered on leafless flowering branches subtended by two or three bracts and one prophyll (Yi, 1996), which differ in leafy flowering branches subtended by a series of expanded leaf sheaths in varying degree of Fargesia (s.l.) species. However, morphological differences among these taxa might implicate intermediate forms. For example, inflorescence branches of Ar. fargesii subtended by very small remnants of sheaths or rings of hairs, pulvinate, which is similar to species in Fargesia (s.l.). Moreover, these intermediate forms may be the result of hybridization given the nested positions of complete plastomes phylogeny in the present study, which is similar to Ar. fargesii discussed above or what was found in the Thamnocalamus group (incl. Fargesia and Yushania)and Arundinaria groups (incl. Bashania) (Triplett and Clark, 2010).
Differing from the situation of Ar. fargesii, Phyllostachys and Ampelocalamus, Arundinaria faberi and Indocalamus longiauritus was nested within the Fargesia grossa clade and the Fargesia macclureana clade, respectively, in the present study. Arundinaria faberi (=Bashania fangiana) was considered to belong to Arundinaria subgenus Sarocalamus (Li et al., 2006) or alpine Bashania (Zhang et al., 2012), while Ar. fargesii was put into Arundinaria subgenus Bashania (Li et al., 2006). Even though they were put into Bashania in Floral Reipublicae Popularis Sinicae (Yi, 1996) due to similar morphology, molecular evidence suggested that no affiliation between these species was revealed (Li et al., 2006;Zeng et al., 2010;Zhang et al., 2012). Arundinaria faberi was clustered within the Fargesia macclureana clade in our plastome analyses, while Ar. faberi or other alpine Bashania species were sister to species in the Fargesia spathe clade rather than the species in the Fargesia grossa clade on the GBSSI gene tree  and RAD-seq analyses (Wang et al., 2017). In morphology, however, Ar. faberi have open inflorescences initially at the tips of leafy branches and later on leafless branches, and the inflorescence branches subtended by long hairs and delicate leaf sheaths (Yi, 1996;Li et al., 2006), which is similar to species in the Fargesia macclureana clade. Arundinaria faberi and species in Fargesia (s.l.) overlap in distribution which may provide a potential condition for hybridization. In addition, incomplete lineage sorting during speciation may be an alternative interpretation, as well as Ampelocalamus species.
The placement of I. longiauritus may stay incertae sedis due to a putative hybridization event between Pseudosasa guanxianensis and Bashania qingchengshanensis based on previous multiple plastid regions (Zeng et al., 2010) and GBSSI gene (Zhang et al., 2012). A lack of informative characters and taxa sampling may be responsible for the disparity between our results and previous plastid phylogenies (Zeng et al., 2010;Zhang et al., 2012;Ma et al., 2014). Indocalamus longiauritus has amphipodium rhizomes with spaced culms and open inflorescences with tiny bracts subtending at the basis of inflorescence branches ( Figure 1D; Li et al., 2006;Yi et al., 2008), which has some morphological differences compared with Fargesia (s.l.). Moreover, the geographic distribution of I. longiauritus overlaps less with that of species of Fargesia (s.l.) (Yi, 1996;Li et al., 2006;Yi et al., 2008). Thus, incomplete lineage sorting or homoplasy may be another cause for the present phylogenetic disparities among I. longiauritus species in Fargesia grossa clade based on plastomes and nuclear markers.
Our results also indicated the Arundinarieae began to diversify around 13.17 Mya and Fargesia (s.l.) along with the Phyllostachys clade, Ampelocalamus clade, Ar. fargesii, Ar. Faberi, and I. longiauritus undergo a rapid late Pleistocene radiation at 4.83 Mya (95%HPD: 2.19-7.30). This is close to the recent recognition of a relatively recent dispersal event in the Pliocene within Phyllostachys clade of the Arundinarieae ranging from an estimated 2.38 to 3.10 Mya . Arundinarieae may have experienced rapid radiation, especially within the Arundinaria clade, Phyllostachys clade and Shibataea clade of Arundinarieae, starting from ca. 7-8 Mya . A bottleneck to resolving the phylogenetic uncertainty among Fargesia (s.l.) and its allies might be the low information and different genetic background of molecular markers (nrDNA and plastid genes), like other clades in Arundinarieae (Triplett and Clark, 2010;Zeng et al., 2010;Zhang et al., 2012;Ma et al., 2014;Attigala et al., 2016).
Using a combination of phylogenetic estimation and morphological evolution, we revealed that Arundinaria fargesii, Phyllostachys species, and Ampelocalamus species might not be included in Fargesia (s.l.) due to recent radiation, plastid capture, incomplete lineage sorting, and hybridization/introgression among different clades within Fargesia (s.l.). We need more species Arundinaria (incl. Bashania) and Indocalamus with more nuclear markers and morphological analyses to amend Arundinaria (incl. Bashania) and Indocalamus and further disentangle the phylogenetic relationships among Ar. faberi, I. longiauritus and Fargesia (s.l.). Thus, we temporarily excluded them from Fargesia (s.l.) to discuss the infrageneric phylogenetic relationships within Fargesia (s.l.) below.
Our analyses highlight species with spathe-like leaf sheath syndrome formed the Fargesia spathe clade with high supported in line with previous nuclear molecular studies (Guo et al., 2001;Guo and Li, 2004;Wang et al., 2017;Zhang et al., 2019). Here, we speculate that the spathe-like leaf sheath syndrome may be a morphological synapomorphy of Fargesia (s. s.) species that can be used to distinguish the species of the Fargesia spathe clade from those of other Fargesia clades. On the other hand, species assigned to the non-spathe clade by an earlier study (see Zhang et al., 2019) fell into two strongly supported paraphyletic clades, the Fargesia grossa clade and the Fargesia macclureana clade. The result of the AU test also demonstrated the non-monophyly of the Fargesia grossa clade and the Fargesia macclureana clade. The use of whole plastid genome sequences has the potential to increase the resolution of phylogenetic analyses at low taxonomic levels or among recently diverged species (Njuguna et al., 2013;Williams et al., 2016;Sancho et al., 2017;Uribe-Convers et al., 2017;Fonseca and Lohmann, 2018). Taxon sampling (including ingroups and outgroups) is important for phylogenetic inference, particularly for the complex temperate woody bamboos. The non-spathe clade may be further broken, even more clades will be recovered as sampling more Fargesia and Yushania species in the future.
The Fargesia grossa clade in the present study included some species of Fargesia (incl. Borinda), all species of Yushania and Indocalamus longiauritus. Those species have open inflorescences subtended by expanded leaf sheaths and pachymorph rhizomes with a long (5-20 cm) and thick (0.25-7 cm) neck (Li et al., 2006;Yi et al., 2008), except for I. longiauritus. The Fargesia macclureana clade in the present study included F. macclureana, F. albocerea, F. hygrophila, F. communis, Fargesia sp. and Ar. faberi. In this clade, all species have lax inflorescences subtended by unexpanded leaf sheaths and pachymorph rhizomes with a short (3-11 cm) and thin (0.4-2.3 cm) neck (Li et al., 2006;Yi et al., 2008), expect for Ar. faberi. The range of the thickness and the length of the culm neck and the degree of the openness of the leaf sheaths at the basis of the inflorescences are overlapped in different species of two clades, so that these morphological traits may not be sufficient to distinguish between these two clades.
These differences may be explained, at least in part, by the absence of morphological autapomorphies, high levels of homoplasy in these two clades.
Temperate woody bamboos might flower frequently at the beginning of the rapid divergence, but climate fluctuation might influence their subsequent life cycles (Zhang et al., 2012). Thus, long generation times might mean that temperate bamboos have a relatively recent diversification compared to annual flowering plants or tropical bamboos (Stapleton et al., 2009). Basic knowledge on the genetics and inflorescence morphology of Fargesia is still lacking due in part to their unusual life cycle, with the vegetative phase ranging from a few to 120 or even 150 years (Janzen, 1976;Campbell, 1987). Our estimation of the divergence times for Fargesia (s.l.) and its allies implies that Fargesia (s.l.) might also undergo rapid radiation, which accounts for the difficulties of reconstructing the phylogeny of Fargesia to some degree. As plastomes only provide maternally inherited signals, further study with more nuclear DNA makers and an expanded sampling of taxa is necessary for a clearer elucidating involving the phylogenetic placement of major clades within Fargesia (s.l.) and complicated evolutionary history of Fargesia (s.l.).

A Hypothesis: Putative Hybrid Origin of the Fargesia Spathe Clade
Putative relictual lineages, transitional series, and hybrids are common in the Arundinarieae, which experienced rapid radiation (Triplett and Clark, 2010;Zhang et al., 2012. In particular, hybridization has been a key factor in bamboo diversification . Incongruence between trees based on nuclear and uniparentally inherited organellar DNA has been reported to be caused by hybridization or gene introgression (Hodkinson et al., 2010;. Arundinaria fargesii and Ar. faberi may have different phylogenetic relationships with species in the Fargesia spathe clade due to hybridization or gene introgression accounting for discrepancies between the findings of our plastomebased study and those of Ma et al. (2014) and nuclear phylogenies (Zhang et al., 2012;Wang et al., 2017). Additionally, species in Fargesia (incl. Yushania) are so widely distributed that substantial geographic overlap likely exists between the hybrids and putative parental species from Arundinaria (Li et al., 2006), thus providing the potential for hybridization or gene introgression.
Morphological evidence along with molecular phylogenies suggested that transitional state or hybrids might be common in the Yushania, and Arundinaria sp. nested within Fargesia (Zhang et al., 2012;Ma et al., 2014;Wang et al., 2017). Wang et al. (2017) revealed that Fargesia and Yushania may have an intermediate morphology regarding the length of culm necks and the open degree of spathe subtending; the resulting taxonomic uncertainty might be due to hybridization. The derived spathe-like leaf sheath syndrome might be interpreted as a significant indicator. Evolution of the morphology of the leaf sheaths at the basis of inflorescences or the bracts at the basis of branches has remained puzzling as comparable morphologies in related taxa were unknown until the present discovery of a relationship among Fargesia (s.l.) and Arundinaria. Arundinaria species have initially compressed inflorescences subtended by reduced leaf sheaths (Li et al., 2006;Kellogg, 2015). Species in the Fargesia grossa clade and the Fargesia macclureana clade have open inflorescences with a series of expanded leaf sheaths in varying degree but never spathe-like at the basis of inflorescence branches. In other words, the species in the Fargesia spathe clade retain partial morphological similarity to Arundinaria and species in Fargesia (s.l.) without spathe-like leaf sheath syndrome.
The spathe-like leaf sheath syndrome might be an adaptive trait facing decreasing temperature, as flower buds could be protected by the expanded spathes. Our divergence time estimation showed that the Fargesia spathe clade originated and diversified in ca. 2.28 Mya, in the Quaternary. Climatic oscillations in the Quaternary likely had tremendous effects on current plant distribution (Hewitt, 2000(Hewitt, , 2004. When temperature decreased in glacial times, species capable of enduring cold likely expanded northward, as in the case of the Fargesia spathe clade. The fact that this clade is currently distributed in the north compared to Fargesia species without spathe-like leaf sheath syndrome also supports the hypothesis. Our data thus present an interesting scenario concerning the evolution of the Fargesia spathe clade: hybridization or gene introgression between Arundinaria species and species that bore a preliminary spathe-like leaf sheath syndrome species in the Fargesia grossa clade or the Fargesia macclureana clade (i.e., the species have a series of leaf sheaths subtending at the basis of inflorescences or bracts at the basis of branches, which could be further expanded or elongated) produced an advanced spathe-like leaf sheath syndrome, which conferred the ability for the species in the Fargesia spathe clade to expand northward in the Quaternary. However, this hypothesis needs further validation from multiple nuclear marker-based phylogenies and ecological niche modeling.
We assembled and annotated 26 plastomes, which allowed us to infer a well-supported phylogenetic backbone for Fargesia (Yi, 1988a) and its allies. Our results revealed that all sampled species of Yushania were nested with most of Fargesia and we propose a new Fargesia (s.l.) and further divided into the Fargesia spathe clade, the Fargesia grossa clade and the Fargesia macclureana clade. Phyllostachys clade, Arundinaria fargesii, and Ampelocalamus clade are inserted in Fargesia (s.l.), sister to the Fargesia spathe clade. The derived spathelike leaf sheath syndrome was believed to be unique to the Fargesia spathe clade, which probably originated from hybridization or gene introgression between the Fargesia grossa clade or the Fargesia macclureana clade and Arundinaria species. Moreover, climatic oscillations in the Quaternary may also contribute to the formation of the derived spathe-like leaf sheath syndrome. Elucidating the complex evolutionary history of Fargesia and its allies remains difficult as multiple biological factors may play complex roles among lineages. Future work with multiple independent molecular regions will help to test different hypotheses of biological factors generated by the current study.

DATA AVAILABILITY
The datasets generated for this study can be found in GenBank (National Center for Biotechnology Information) with the accession numbers MH988715-MH988740.

AUTHOR CONTRIBUTIONS
Y-QZ and X-CX collected the materials and conducted the field investigations. YZ conducted the experiments and analyzed the data. YR, J-QZ, and YZ wrote the manuscript. All authors have read and approved the final manuscript.

FUNDING
This study was co-supported by the National Natural Science Foundation of China (No. 31570221) and the Fundamental Research Funds for the Central Universities (2017TS028 to YZ).

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.00981/ full#supplementary-material FIGURE S1 | Gene map of the chloroplast genome of Fargesia denudata. Individual genes are labeled and those belonging to different functional groups are color coded. Dashed area in the inner circle indicates the GC content of the genome.          (Lanfear et al., 2012).