Introduction

Dendrocalamus is a woody bamboo genus placed in the subtribe Bambusinae and tribe Bambuseae (Ohrnberger 1999). Species referred to the genus are characterized by their sympodial rhizomes and large-sized, dense clumps. The genus contains over 50 species, naturally distributed in tropical and subtropical regions of the Old World, many of which are economically exploited by communities in south and southeast Asia. The original description of the genus was based on the type species D. strictus. The description was subsequently expanded to include pericarp characters, which were used to distinguish between Dendrocalamus and Bambusa (Munro 1868; Bentham 1883; Gamble 1896). While at present it is taxonomically convenient for Dendrocalamus to be recognized in a broad sense, the limits between Bambusa and Dendrocalamus are not satisfactorily defined, thus creating confusion in their systematic classification. Lack of sound taxonomy is hindering scientific conservation and management of the woody bamboos belonging to this genus.

Since Dendrocalamus was separated from Bambusa by Nees von Esenbeck in 1834, over 70 species names have been assigned to the genus, although Ohrnberger (1999) retains only 51 of these. Most of the species that have not been maintained by Ohrnberger have been reduced to synonymy or to infraspecific rank. A few are transferred to, or sunk into, other genera: Ampelocalamus (subtribe Thamnocalaminae); Gigantochloa and Pseudoxytenanthera (Bambusinae).

Various infrageneric classifications of Dendrocalamus have been adopted by Chinese botanists. Hsueh and Li (1988) proposed the first infrageneric classification of Dendrocalamus by recognizing two subgenera and five sections, limiting the assignments to only those species reported from China. Ohrnberger (1999) assigned species only to sections Dendrocalamus, Bambusoidetes, Sinocalamus and Draconicalamus. Out of the 51 taxa recognized by Ohrnberger, 22 were assigned to particular sections, while 29 taxa were unplaced. A more recent taxonomic revision of Chinese Dendrocalamus (Li and Stapleton 2006) retains the subgenera proposed by Hsueh and Li (1988) but disregards sectional assignments, merging sections Dendrocalamus and Bambusoidetes as subgenus Dendrocalamus, and sections Sinocalamus and Draconicalamus as subgenus Sinocalamus. Li and Stapleton (2006) transferred 11 taxa previously referred to subgenus Sinocalamus to subgenus Dendrocalamus. The major problem faced in the infrageneric classification of Dendrocalamus is the paucity of published morphological character information for many of the species. Twenty-seven species do not appear to have been referred to any subgenus or section under any of the proposed schemes.

Bamboos have always been a taxonomically challenging group of plants because, while the classification of flowering plants depends largely on the characteristics of reproductive organs, flowering is rare in many bamboo species. Some bamboo species flower at intervals as long as 120 years, and for some there is no report of flowering to date. Because of apparent paucity of morphological characters in bamboos, taxonomists have long sought different sources of taxonomically informative data. The availability of molecular data in the final decade of the 20th century enabled taxonomists to review phylogenetic concepts of the Poaceae more objectively. Initially DNA products, viz. isozymes and secondary compounds such as phenolics, were used in exploring relationships among taxa (Chou and Hwang 1985), species identification (Alam et al. 1997) and assessment of infraspecific polymorphism (Biswas 1998). In a study involving five Dendrocalamus taxa, Arthrostylidium naibunensis W.C. Lin and Chimonobambusa quadrangularis Makino, a Dendrocalamus cluster could be differentiated from the other two genera using phenolic compounds and isozyme patterns of esterase and peroxidase. Within Dendrocalamus two sub-clusters were recognized; Dendrocalamus asper associated with D. giganteus, while D. latiflorus associated with its variety D. latiflorus var. mei-nung. However, Dendrocalamus strictus was distant from these two sub-clusters. Later on, variation in DNA itself was the subject of investigations. The more pertinent studies involving named Dendrocalamus taxa are those of Loh et al. (2000) and Sun et al. (2005), but these studies entailed only limited sampling of the genus. In the first, two Dendrocalamus taxa were sampled, with D. brandisii clustering with taxa from Bambusa, and D. giganteus appearing genetically distant from all other taxa included. In the second study three Dendrocalamus taxa were sampled. These three taxa did not form a separate Dendrocalamus cluster but were found in two distinct sub-clusters within Bambusa. D. membranaceus showed close affinity to D. strictus and both were placed within one Bambusa cluster, whereas D. latiflorus was associated with the other Bambusa cluster. The study reported wide genetic variation within Dendrocalamus and raised questions about its monophyly. These earlier molecular studies included limited samples from Dendrocalamus and did not throw much light on the infrageneric relationships within the genus. In the present investigation there was wider provision within Dendrocalamus with ten putative taxa, and five outgroup taxa from the subtribe Bambusinae. Amplified fragment length polymorphism (AFLP) markers were used with the objectives to (1) test the monophyly of Dendrocalamus, and (2) assess the molecular support for various infrageneric assignments proposed in Dendrocalamus.

Materials and methods

Site description

Genetic material was collected from the Bambuseta of five botanical gardens in India: Forest Research Institute (FRI), Dehra Dun; National Botanical Garden of Botanical Survey of India (NBG), Howrah; State Forest Research Institute (SFRI), Chessa; ICAR research complex for northeastern hill region (ICAR), Basar; and Rain Forest Research Institute (RFRI), Jorhat. The genetic material of the monotypic African bamboo Oxytenanthera abyssinica was available as the result of previous research work at Bangor University, UK (Inada 2004).

Genetic material

Unopened terminal leaves were collected from ten Dendrocalamus and five outgroup taxa from subtribe Bambusinae. The leaves were dried using silica gel as per the procedure of Chase and Hills (1991) and then transported from the field sites in India to the laboratory at CAZS Natural Resources, Bangor University, for DNA extraction and analysis. Six of the Dendrocalamus taxa (D. strictus, D. hamiltonii, D. membranaceus, D. brandisii, D. sikkimensis and D. asper) represented subgenus Dendrocalamus and two other taxa (D. giganteus and D. calostachyus) represented subgenus Sinocalamus, in the infrageneric classification of Dendrocalamus by Li and Stapleton (2006). No information was available regarding the infrageneric assignment of D. sahnii and D. somdevai. The outgroup taxa were from the genera Bambusa, Melocalamus, Oxytenanthera, Dinochloa and Thyrsostachys, all of which belong, like Dendrocalamus, to subtribe Bambusinae as recognized by Ohrnberger (1999). For each collected taxon, except Oxytenanthera abyssinica and Dendrocalamus brandisii, a voucher specimen was prepared and deposited at the herbarium of the Forest Research Institute, Dehra Dun, India (DD) (Table 1).

Table 1 Voucher specimens deposited at the Forest Research Institute (DD) herbarium, Dehra Dun, India
Full size table

DNA extraction

DNA was extracted from 50 mg of dried leaf tissue using a modified cetyltrimethylammonium bromide (CTAB) protocol of Doyle and Doyle (1990). Modifications of the protocol included an increased incubation period in the CTAB buffer (2 h), increased concentration of β-mercaptoethanol (1%) and two clean-up cycles with chloroform and isoamyl alcohol. The tissue was homogenized mechanically in a tissue lyser (Roche). About 700 μl pre-warmed CTAB buffer [2% CTAB, 1.4 M NaCl, 20 mM ethylenediamine tetraacetic acid (EDTA), 100 mM Tris, 1% polyvinylpyrrolidone (PVP) and 1% β-mercaptoethanol] was added to the homogenate and incubated at 65°C for 2 h. Immediately after incubation, about 700 μl chloroform:isoamyl alcohol (24:1) was added to the incubated sample and thoroughly mixed by inverting the sample tube. Following a centrifugation step, the supernatant was transferred to a fresh tube. DNA present in the supernatant was precipitated by adding ice-cold isopropanol (0.4 × volume) and 5 M NaCl (0.5 × volume). The DNA pellet was washed with 70% ethanol and resuspended in Tris-EDTA (TE). To degrade RNA, the enzyme RNAseA was added to the suspended DNA and incubated at 37°C for 1 h. A second cycle of clean-up and DNA precipitation, as described earlier, was repeated. In the final step, the DNA pellet was washed with 70% ethanol, dried and resuspended in TE. The quality of extraction was checked by running a 1% agarose mini-gel (run at 50 V for 30 min) in Tris–borate–EDTA (TBE buffer, 1×) containing 0.5 μg/ml ethidium bromide and visualizing under a ultraviolet light source. Genomic DNA concentration was determined in a Fluostar Galaxy Fluorometer using the fluorescent dye Pico green.

Generation of AFLP markers

The AFLP assay was performed following the protocol of Vos et al. (1995), adapted for the Beckman Coulter Sequencer. The process was carried out in four steps. In the first step, 250 ng genomic DNA was double-digested in a final volume of 40 μl at 37°C for 3 h with EcoRI (5 U) and MseI (Tru9I) (5 U). In the second step, a ligation reaction was set up by adding 10 μl of a master mix (containing EcoRI adapter, MseI adapter, T4 DNA ligase and ligase buffer) to the products of earlier digestion reaction. The mixture was incubated overnight at 16°C. In the third step, a pre-selective polymerase chain reaction (PCR) was performed with a final volume of 20 μl, containing 5 μl restriction-ligation reaction, 10 μl Reddy mix (Abgene) and 5 μl universal primer mix (E00—30 ng and M00—30 ng). The thermocycler conditions for the pre-selective PCR included 30 cycles consisting of 30 s denaturation at 94°C, 60 s annealing at 56°C, 60 s extension at 72°C and 600 s extension at 72°C. In the fourth step, a selective PCR was performed with a final volume of 20 μl, containing 5 μl pre-selective products, 10 μl Reddy mix (Abgene) and 5 μl selective primer mix (E00 + 3—25 ng and M00 + 3—30 ng). The selective primer E00 + 3 was end-labelled with a fluorescent dye instead of the radioactive labelling described in the original protocol of Vos et al. (1995). The thermocycler conditions for the selective PCR included 13 touchdown cycles to avoid amplifying non-specific sequences (30 s denaturation at 94°C, 30 s annealing at 65°C which was then reduced by 0.7°C per cycle, 60 s extension at 72°C), 23 normal cycles (30 s denaturation at 94°C, 30 s annealing at 56°C, 60 s extension at 72°C) and 420 s final extension at 72°C.

A primer screening experiment was set up with two samples and 20 primer pairs (four EcoRI primers and five MseI primers) to screen primer pairs having ability to generate most polymorphic loci. A locus was treated as polymorphic when a peak was detected in one sample but not in the other. Detection of peaks at the same locus in both samples made the locus monomorphic. Five primer pairs generating most polymorphic loci were short-listed and used in amplifying AFLP markers from all 15 operational taxonomic units (OTUs) used in the present investigation. The primer sequences are provided in Table 2. A negative control without template DNA was run in each batch of PCRs to confirm that no contamination had occurred. The reproducibility of AFLP peaks was checked by repeating the whole process a number of times.

Table 2 Primer sequences of selective primers used in the present investigation to generate AFLP markers in the 15 OTUs from subtribe Bambusinae
Full size table

Separation and scoring of AFLP markers

The selective PCR products were separated through capillary gel electrophoresis in the CEQ 8000 Genetic Analysis System (Beckman Coulter, Inc.) and analysed with the fragment analysis software. During fragment analysis the separated fragments were sized with the use of internal size standards (PA400). Following this, the sized fragments were subjected to an AFLP binning analysis that converted the AFLP peak profiles into binary matrix. The presence of a peak was scored as 1, and its absence was scored as 0. Peaks of size ranging from 60 to 400 bp were scored.

Data analysis

The AFLP dataset was analyzed both phenetically and cladistically. For phenetic analysis, the binary matrix (470 × 15) of multilocus peak patterns was converted to a matrix of pairwise distances between OTUs, expressed as Jaccard’s (1908) distance coefficient using the software package NTSYSpc 2.11X (Rohlf 2000). Cluster analysis was carried out using the neighbour-joining (NJ) algorithm of Saitou and Nei (Saitou and Nei 1987) in the same software package using Jaccard’s distance matrix as the input file. The generated phenogram was rooted using the outgroup option, taking Oxytenanthera abyssinica as the outgroup taxon. Statistical support for the internal branches was assessed by performing a bootstrap analysis with 1,000 replicates in the software package FREETREE (Pavlicek et al. 1999). A Mantel test was performed to test how well the phenogram represented the inter-OTU distances, following the procedure described by Koopman et al. (2001). Initially, the NJ tree was transformed to a tree-distance matrix using the COPH module in NTSYSpc 2.11X (Rohlf 2000), then the MXCOMP module was used to calculate the co-phenetic correlation coefficient (r), which indicated the goodness of fit of the cluster analysis to the distance matrix.

Cladistic analysis of the AFLP dataset was performed under maximum-parsimony criterion with PAUP 4.0b10 (Swofford 2002). The large number of included taxa (>12) ruled out an exhaustive search. So, heuristic search was used to identify the most parsimonious tree. Heuristic search was performed with the following criteria: 1,000 replicates, random additions of sequence, tree-bisection-reconnection (TBR) branch swapping, character optimizations using accelerated transformation (Perrie and Brownsey 2005). Output trees were rooted using the outgroup option with Oxytenanthera abyssinica as the outgroup taxon. Statistical support for internal branches was assessed using the bootstrap analysis (Felsenstein 1985) in PAUP 4.0b10 with the following criteria: 1,000 replicates, heuristic search and a 50% confidence level.

Results

The five primer pairs used in the present investigation generated 609 AFLP marker loci, out of which 99.2% (604) loci were polymorphic and only 0.8% (5) loci were monomorphic across the 15 OTUs. The dataset also contained 134 (22.0%) loci where only one peak was detected. These loci are uninformative as far as relationships among taxa are concerned and were excluded from subsequent analysis. The number of AFLP marker loci generated by the individual primer sets varied from 101 to 133 with a mean of 121.8 (Table 3).

Table 3 Summary statistics of AFLP markers obtained from 15 OTUs belonging to subtribe Bambusinae, with five selective primer pairs
Full size table

The genetic distance estimates based on Jaccard’s measure varied from 0.47 to 0.92 (Table 4). Although referred to the Bambusinae, the monotypic African bamboo Oxytenanthera abyssinica shared very few peaks with other taxa included in the study and was found genetically distant from them (distance ranged from 0.77 to 0.92). However, even within Dendrocalamus there was wide genetic variation. Dendrocalamus strictus appeared isolated with a minimum distance of 0.77 (from D. somdevai) and a maximum distance of 0.85 (from D. asper). Among the outgroups included in the present investigation, Bambusa balcooa was found to be the closest to Dendrocalamus sensu lato with a mean distance of 0.60.

Table 4 Half-matrix of pairwise Jaccard’s distance between 15 OTUs from subtribe Bambusinae
Full size table

Cluster analysis with the NJ algorithm resulted in a single tree (Fig. 1) with high co-phenetic correlation coefficient (r = 0.971). A well-supported (85% bootstrap support) major cluster was recovered, containing all Dendrocalamus taxa (except D. strictus) with Melocalamus compactiflorus as sister lineage. D. strictus was recovered near the root of the tree, distant from the cluster containing other Dendrocalamus. Three clusters that had varying degree of bootstrap support (≥50%) could be recognized within the major cluster: cluster 1: Dendrocalamus membranaceus and D. somdevai (93% bootstrap support); cluster 2: Dendrocalamus sikkimensis, Bambusa balcooa and D. hamiltonii (89% bootstrap support); cluster 3: D. giganteus and D. asper (50% bootstrap support). However, the relative relationship among these three clusters was not clear due to polytomy. Furthermore, the affinity of D. sahnii, D. calostachyus and Dinochloa macclellandii remained unresolved. Also, the affinity of D. brandisii to cluster 1 was unreliable owing to poor bootstrap support (46%).

Fig. 1
figure 1

Neighbour-joining phenogram depicting phenetic relationships among 15 OTUs sampled from subtribe Bambusinae. Numbers at the nodes indicate bootstrap (%) support for the respective clusters. OTU codes are indicated within parentheses. Scale bar indicates additive distance between pairs of taxa

Full size image

Maximum-parsimony analysis of the AFLP dataset (consisting of 470 characters, of which 460 were parsimony informative) with heuristic search yielded two equally parsimonious trees (Fig. 2a, b), giving two alternate but equally likely hypotheses of evolutionary relationships among the OTUs. Each tree was 1,468 steps (character state changes) long with consistency index (CI) = 0.320, retention index (RI) = 0.298 and rescaled consistency index (RC) = 0.095. These most parsimonious trees were congruent for most part of the trees but differed in the placement of Dendrocalamus calostachyus. Strict consensus of the most parsimonious trees resulted in a polytomy (Fig. 3). Three groups were recovered that were identical to the clusters in NJ tree. Because of polytomy, the relationship among these three groups was not clear. Bootstrap analysis with 1,000 bootstrap replicates showed varying degree of support for these groups.

Fig. 2
figure 2

Phylograms depicting two equally parsimonious trees resulting from maximum-parsimony analysis of a character matrix of 470 AFLP markers. Each tree has a length of 1,468 steps, CI = 0.320, RI = 0.298, RC = 0.095. Values above segments indicate character state changes (gains/losses of AFLP bands) supporting respective nodes. OTU codes are indicated within parentheses. Horizontal bars below trees represent 10 character state changes

Full size image
Fig. 3
figure 3

Cladogram depicting strict consensus of the two parsimonious trees obtained in maximum-parsimony analysis of a character matrix of 470 AFLP markers. Length = 1,480 steps (character state changes), CI = 0.317, RI = 0.289, RC = 0.092. Values above segments indicate bootstrap support for the respective nodes. Bootstrap support for nodes with <50% support and which collapse under the 50% majority rule tree is not shown. OTU codes are indicated within parentheses

Full size image

Discussion

The ability of AFLP markers to detect polymorphisms rapidly without prior knowledge of the target genome is an advantage that has been widely exploited in plants in a range of applications including genetic identification through fingerprinting, quantitative trait loci mapping, population genetics and systematics (Mueller and Wolfenbarger 1999). The use of AFLPs in the present investigation is justified on the grounds that woody bamboos have evolved recently and have undergone little molecular evolution, as a result of which sequencing studies (nuclear gene-internal transcribed spacer) in other genera have found very little sequence variation among closely related taxa (Hodkinson et al. 2000). Similarly, sequencing of plastid gene (ndhF) has not been successful below tribal level (Clark 1997; Zhang and Clark 2000). The AFLP assay was successful in detecting polymorphism across the 15 OTUs included in the present investigation. A mean number of 120.8 polymorphic loci were detected using fluorescence-labelled primers in combination with gel analysis in an automated DNA sequencer, a number comparable to finding in earlier study of bamboos (Hodkinson et al. 2000).

The two different approaches adopted for data analysis in the present investigation, viz. NJ cluster analysis and maximum-parsimony analysis, yielded partially resolved trees. Even though the basal lineages were well resolved, there was internal polytomy in both trees. Three identical monophyletic internal groups were recovered; however, the degree of statistical support for these groups differed in both analyses. Only one group containing Dendrocalamus membranaceus and D. somdevai had high bootstrap support (93% and 82% in NJ and parsimony analysis, respectively) in both analyses. The affinity of Dendrocalamus sahnii, D. calostachyus and D. brandisii were inconclusive in both analyses. Similarly, the affinity of Dendrocalamus sikkimensis to the group containing Bambusa balcooa and Dendrocalamus hamiltonii was not supported in the maximum-parsimony analysis, though it was well supported (89% bootstrap support) in the NJ analysis. Koopman et al. (2008) argue that, as the AFLP markers are randomly sampled from across different parts of the genome which may have different evolutionary history and hence may contain conflicting signals, phylogenetic analysis will result in low support values for those parts of the tree containing the species involved. Nevertheless, even the poorly supported parts of the phylogenetic tree showed congruence with that of the NJ phenogram. The three monophyletic groups did not agree with any of the sectional assignments within Dendrocalamus sensu lato circumscribed by Hsueh and Li (1988), Ohrnberger (1999) and Li and Stapleton (2006).

Neither the phenetic nor the phylogenetic analysis adopted in the present investigation supported monophyly of the genus Dendrocalamus as circumscribed by Hsueh and Li (1988), Ohrnberger (1999) and Li and Stapleton (2006), based on morphological data. The placement of Bambusa balcooa and Dendrocalamus strictus suggested otherwise. Bambusa balcooa, instead of forming a separate lineage, was recovered in cluster 2 shared with Dendrocalamus hamiltonii and D. sikkimensis (Fig. 1). This supports the findings of Stapleton (1994), who had reported the closeness of Bambusa balcooa to Dendrocalamus species on morphological grounds, stressing similarity in the profuse aerial roots at the culm nodes, the large rhizomatous branch bases and the culm wax. Further similarities between Bambusa balcooa and Dendrocalamus hamiltonii can be found in the reproductive parts, with both species having three stigmas each. The placement of Dendrocalamus strictus near the root of the tree away from rest of the Dendrocalamus sensu lato was not entirely unexpected considering the findings of Chou and Hwang (1985), who had reported the isolation of D. strictus from other Dendrocalamus taxa based on studies involving isozymes and phenolics.

Morphologically, the isolation of D. strictus could be explained by presence of inflorescence comprising fascicular pseudospikelets (2.2–2.5 cm in diameter) on each node of flowering branches, distinguishing them from other species included in the study (Yang et al. 2008). Ecologically also, D. strictus is very distinct from the other Dendrocalamus taxa included in the present investigation. Dendrocalamus strictus naturally occurs in dry deciduous open forests, receiving as little as 750 mm annual rainfall and exposed to low relative humidity. The other Dendrocalamus taxa included in the present investigation are confined to moister areas (moist deciduous to wet evergreen forests), with annual rainfall in excess of 1,500 mm.

Bambusinae as circumscribed by Ohrnberger (1999) is an Old World tropical subtribe with its centre of diversity in southeast Asia. It contains 17 genera, the relationships among which are not fully understood. In the present investigation Bambusa balcooa was placed within Dendrocalamus sensu lato, supporting the closeness, or even inseparability, of these two genera. Melocalamus and Thyrsostachys were recovered as sister lineages to Dendrocalamus and Bambusa. Watanabe et al. (1994), the first to study phylogenetic relationships among Asian bamboos using restriction fragment length polymorphism of chloroplast DNA, recovered a clade representing subtribe Bambusinae sensu Ohrnberger (1999), containing Bambusa, Dendrocalamus, Gigantochloa and Thyrsostachys. Internally, however, Watanabe’s clade was poorly resolved in terms of relationships among Bambusa, Gigantochloa and Dendrocalamus, suggesting close relationships among these genera. Thyrsostachys had emerged as a sister lineage to the other genera included in Watanabe’s study. The study of Loh et al. (2000) and Ramanayake et al. (2007), using AFLPs and random amplified polymorphic DNAs (RAPDs), respectively, also indicated a close relationship between Bambusa and Gigantochloa. The combined evidence from these earlier molecular studies and the present investigation suggests that taxa belonging to Bambusa, Dendrocalamus, Dinochloa and Gigantochloa form a close complex but are relatively distant from Melocalamus, Thyrsostachys and Oxytenanthera.

Conclusions

The phenetic and phylogenetic trees generated in the present study are plausible hypotheses pertaining to relationships within Dendrocalamus. The low statistical support for two of the three identified monophyletic groups might improve with inclusion of more informative characters, which could be generated by using more selective primer sets. The study confirms that the current taxonomic treatment of Dendrocalamus is unsatisfactory and needs revision. A broader study encompassing a wider selection of taxa from Bambusa, Dendrocalamus, Dinochloa and Gigantochloa, and inclusion of evidence from multiple data sources (including AFLP and sequencing of fast-evolving genes) might be expected to produce a robust phylogenetic tree for this suite of closely related taxa.