Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Application of DNA Barcodes in Asian Tropical Trees – A Case Study from Xishuangbanna Nature Reserve, Southwest China

  • Xiao-cui Huang ,

    Contributed equally to this work with: Xiao-cui Huang, Xiu-qin Ci

    Affiliations Laboratory of Plant Phylogenetics and Conservation, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Kunming, Yunnan, People’s Republic of China, Key Laboratory of Tropical Forest Ecology, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Menglun, Yunnan, People’s Republic of China, Key Laboratory of Tropical Plant Resources and Sustainable Use, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Menglun, Yunnan, People’s Republic of China, University of Chinese Academy of Sciences, Beijing, People’s Republic of China

  • Xiu-qin Ci ,

    Contributed equally to this work with: Xiao-cui Huang, Xiu-qin Ci

    Affiliations Laboratory of Plant Phylogenetics and Conservation, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Kunming, Yunnan, People’s Republic of China, Key Laboratory of Tropical Forest Ecology, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Menglun, Yunnan, People’s Republic of China, Key Laboratory of Tropical Plant Resources and Sustainable Use, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Menglun, Yunnan, People’s Republic of China, University of Chinese Academy of Sciences, Beijing, People’s Republic of China

  • John G. Conran,

    Affiliation Centre for Evolutionary Biology and Biodiversity & Sprigg Geobiology Centre, School of Biological Sciences, Benham Bldg DX, The University of Adelaide, Adelaide, South Australia, 5005, Australia

  • Jie Li

    Affiliations Laboratory of Plant Phylogenetics and Conservation, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Kunming, Yunnan, People’s Republic of China, Key Laboratory of Tropical Forest Ecology, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Menglun, Yunnan, People’s Republic of China, Key Laboratory of Tropical Plant Resources and Sustainable Use, Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Menglun, Yunnan, People’s Republic of China

Application of DNA Barcodes in Asian Tropical Trees – A Case Study from Xishuangbanna Nature Reserve, Southwest China

  • Xiao-cui Huang, 
  • Xiu-qin Ci, 
  • John G. Conran, 
  • Jie Li



Within a regional floristic context, DNA barcoding is more useful to manage plant diversity inventories on a large scale and develop valuable conservation strategies. However, there are no DNA barcode studies from tropical areas of China, which represents one of the biodiversity hotspots around the world.

Methodology and Principal Findings

A DNA barcoding database of an Asian tropical trees with high diversity was established at Xishuangbanna Nature Reserve, Yunnan, southwest China using rbcL and matK as standard barcodes, as well as trnH–psbA and ITS as supplementary barcodes. The performance of tree species identification success was assessed using 2,052 accessions from four plots belonging to two vegetation types in the region by three methods: Neighbor-Joining, Maximum-Likelihood and BLAST. We corrected morphological field identification errors (9.6%) for the three plots using rbcL and matK based on Neighbor-Joining tree. The best barcode region for PCR and sequencing was rbcL (97.6%, 90.8%), followed by trnH–psbA (93.6%, 85.6%), while matK and ITS obtained relative low PCR and sequencing success rates. However, ITS performed best for both species (44.6–58.1%) and genus (72.8–76.2%) identification. With trnH–psbA slightly less effective for species identification. The two standard barcode rbcL and matK gave poor results for species identification (24.7–28.5% and 31.6–35.3%). Compared with other studies from comparable tropical forests (e.g. Cameroon, the Amazon and India), the overall performance of the four barcodes for species identification was lower for the Xishuangbanna Nature Reserve, possibly because of species/genus ratios and species composition between these tropical areas.


Although the core barcodes rbcL and matK were not suitable for species identification of tropical trees from Xishuangbanna Nature Reserve, they could still help with identification at the family and genus level. Considering the relative sequence recovery and the species identification performance, we recommend the use of trnH–psbA and ITS in combination as the preferred barcodes for tropical tree species identification in China.


Species identification is of critical importance in conserving and utilizing biodiversity, but this is often hindered by a lack of professional knowledge of classification [1]. As one of the most vulnerable floras to the increasing threats from human activities [2], tropical plant species are badly in need of rapid identification methods to aid in the development of reasonable protection strategies [3]. Unfortunately, traditional morphological taxonomy is time-consuming and dependent on pre-determined classifications and expertise [4]. Furthermore, identification is a challenge for tropical trees, even for experts, due to the frequent absence of reproductive organs that are needed to distinguish among morphologically similar species, but are often unavailable during field surveys [5]. A wide range of molecular methods have been applied to overcome this, but Hebert et al. [6] presented an important tool of DNA barcoding which provides a fast and effective means for species assignment without the need for detailed taxonomic expertise.

An ideal barcode should meet the need for rapid enough evolution to distinguish between species, combined with conserved regions, which will function as universal primer binding sites for PCR [7]. However, because it has proved difficult to find a single barcoding locus for plants, a combination of two or more loci is normally proposed. Recently, a consensus has emerged for using the plastid genes rbcL and matK as standard markers to barcoding plants [8], as rbcL is the most effective locus for PCR amplification and sequencing [9], while matK performs well for species identification in some cases [10]. In addition, there are also reports suggesting the potential of the non-coding trnH–psbA [11,12] and nuclear ITS regions [13,14] as markers and these four loci have now been utilized in numerous plant of barcoding studies [1518].

Xishuangbanna Nature Reserve in southern Yunnan Province is located at an intermediate zone between tropical Southeast Asia and subtropical East Asia and, as such, represents the northern limit of tropical rain forest distribution in China. The region is of considerable interest to biologists for biodiversity conservation [19], containing 3,336 angiosperms from 1,140 genera in 197 families [20]. It also contains different vegetation types, of which tropical rain forest is the most common and the least threatened, whereas tropical seasonal moist forest associated with limestone habitat is considered vulnerable, mostly though habitat fragmentation due to land clearing.

In this study, we use a plot-based sampling strategy to establish a local DNA barcode database of tropical trees occurring two different vegetation types and evaluate the performance of DNA barcodes in the Xishuangbanna Nature Reserve. Specifically, we analyzed sequence recovery and species discrimination of the four barcodes rbcL, matK, trnH–psbA and ITS singly and in combination, particularly for the following:

  1. Assessment errors of morphological identification in ecological surveys based on core barcodes (rbcL + matK);
  2. Comparison of sequence recovery of the four selected markers between our study and other comparable studies;
  3. Evaluating species resolution for different methods with various barcodes combinations;
  4. Comparison of the ability of species identification between this study and others studies based on plant DNA barcoding with geographically bounded sampling;
  5. Evaluate the ability of DNA barcoding in this region.

Materials and Methods

Ethics statement

All the fieldwork was conducted at Xishuangbanna Nature Reserve under permit issued by the Forestry Department of Yunnan Province and Xishuangbanna National Nature Reserve Administration and collecting procedures were done with proper precautions for minimizing impacts to protected or endangered trees in these areas. We stated field studies did not involve any locations for which no specific permission was required.

Study site and sampling

Fieldwork in this study was conducted from 2008 to 2012. Four plots established by the Xishuangbanna Tropical Rainforest Ecology Station (XSTRES) were selected on basis of vegetation type and different level of ecological surveys (named the 20 ha Xishuangbanna tropical seasonal rainforest dynamics plot (BB), JJYL, GGYL and LSL) (Table 1). There are two vegetation types—tropical rain forest (BB, JJYL and GGYL) and tropical seasonal moist forest (LSL) in our study [19]. The 20 ha permanent dynamic plot was established in 2007 following the protocol for large forest dynamics plot of Center for Tropical Forest Science [21]and tree species were identified by ecologists and taxonomists. Other three plots were identified in the field by ecologists. We collected mature leaves from 1–6 individuals for each tree species ≥1cm of diameter at breast height and dried with silica gel. Cambium (bark) tissues were also used as an alternative for canopy trees if they were too tall for leaf collection. Voucher specimens were collected and deposited at the herbarium of Xishuangbanna Tropical Botanical Garden (XTBG), Chinese Academy of Science (CAS).

Table 1. Sampling information of four plots in the Xishuangbanna Nature Reserve.

DNA isolation, amplification and sequencing

Total genomic DNA was isolated from approximately 30 mg of dried leaf or cambial material using the Plant Genomic DNA Kit (Tiangen Biotech Co., China), either according to the manufacturer’s protocols, or modified as needed. For example, extraction with chloroform isoamyl alcohol (24:1) was repeated twice when the material was rich in secondary metabolites.

We amplified chloroplast regions rbcL, matK, trnH–psbA and the nuclear region ITS using multiple primers with broad taxonomic versatility. As standard barcodes, rbcL and matK are used widely and recommended due to high amplification levels in plants [22, 23], thus, these two barcodes are especially helpful for mass screening data. Similarly, ITS and trnH–psbA showed considerable utility for species identification [24]. For matK, four primer sets were tested, due to its generally poor performance of amplification and sequencing [25]. All PCR reactions had a total volume of 25 μL and DMSO and BSA were added to enhance the PCR performance for matK and ITS. In order to test the effects of different PCR procedures on sequence recovery, we conducted two sets of cycling conditions (general and Ramp procedures) applied to samples from the three small plots (JJYL, GGYL and LSL). For primer combinations, PCR thermal conditions and references see S1 Table. All the PCR products were sequenced at the Beijing Genomics Institute (BGI).

Sequence editing and alignment

We assembled consensus sequences using Sequencher 4.14 (GeneCodes Corp., Ann Arbor, MI, USA) and aligned them with different programs (i.e. ClustalW [26], MUSCLE 3.8.31 [27] and SATé [28]). For two core markers (rbcL, matK) and the nuclear marker ITS, a global multiple sequence alignment was used. The rbcL sequences were unambiguous due to the absence of insertion or deletion. Alignment of matK was more difficult due to the insertion of triplet codons, so we checked the alignment results visually. Both rbcL and matK were aligned several times by ClustalW and MUSCLE. Because the ITS sequences were more difficult to align, we used the Simultaneous Alignment and Tree Estimation (SATé) for global multiple alignment ( Similarly, the trnH–psbA sequences were highly variable and could not be handled with a global multiple sequence alignment. As a result, we conducted a family-based alignment using ClustalW and then created a supermatrix by concatenating them with the aligned sequences of the other markers [29].

Detecting errors in tropical tree identification

It is difficult to identify woody plants in the tropics because most of the trees encountered in the field are not reproductive at the time of sampling and must be identified using vegetative characters, but most species descriptions and keys rely on flower and fruit characters [30], often resulting in misidentification of sterile material. Here, we adopted a two-step procedure of reciprocal illumination, combining morphology and DNA sequence data to uncover and correct mistakes in species identification in three plots (JJYL, GGYL and LSL), in which the morphology-based identifications were undertaken in the field. Firstly, we detected potential errors through examination of the Neighbor-Joining trees using the core barcodes. Secondly, we reviewed the morphology of the species involved, comparing specimens with relevant herbarium vouchers from other studies, to confirm whether mistakes based on morphological identification had been made. DNA extraction and subsequent trials were also repeated when herbarium vouchers were absent, until all the samples were considered to be error free.

Data analysis

There are numerous methods used for the analysis of barcode data and species resolution, of which phylogenetic analysis [17,3133] and similarity approaches such as BLASTn [8,34] are the most commonly used for DNA barcode data analysis. The similarity-based BLASTn is an algorithm for comparing query sequences with reference database calculating pairwise alignments in the process. All sequences in our study served as both database and query and were queried individually to the database. Additionally, we also conducted stand-alone BLAST comparisons, only using the sequence database of the BB plot, where the most species occurred. All barcodes were tested singly and in combination. We considered an assignment to be correct when query sequences showed ≥95% identical sites to sequences of the same species in the database and all the sequences of the species showed higher identical sites compared with sequences for other taxa.

We also present results based on Neighbor-Joining and Maximum-Likelihood trees, because some studies have shown that different algorithms for reconstructing trees did not alter the performance of DNA barcodes significantly [35,36]. We tested to see if the individual species were retrieved as monophyletic groups for each barcode locus and their different combinations. The NJ tree reconstruction was constructed using Geneious 6.1.6, while the ML analysis was conducted using RAxML [37] via the CIPRES supercomputer cluster ( Bootstrap analyses were based on 1000 replicates for NJ trees and 100 for ML trees. For a given barcode locus or combination of loci, we used a cutoff of 50% to define support for “successful” resolution of monophyletic species [38].

Testing the barcoding accuracy at the regional scale

To determine if species identification success was lower between different plots than within them, we established a barcoding database from the BB plot. For the other three plots (JJYL, GGYL and LSL), only those individuals that belonged to taxa present in the BB plot were used for the regional scale analysis. To this end, we selected all the samples from the three plots that belonged to a species or genus represented by at least one individual in the database from BB plot, using the BLAST method to assign a species or genus to the specimens in the three smaller plots with the BB specimens as reference database.


PCR and sequencing success rates

In total, we obtained 5583 sequences from 2052 samples, representing 655 species, 259 genera and 76 families. These included 1654 sequences for rbcL, 1430 sequences for trnH–psbA, 1422 sequences for matK and 1077 sequences for ITS (Table 2). We recovered one sequence for at least one of the four markers in 1858 (90.5%) samples; however, 194 samples failed for all four regions. rbcL showed the highest PCR and sequencing success rates of 97.6% and 90.8%, respectively. The next best PCR and sequencing rates were exhibited by trnH–psbA (93.6%, 85.6%) and matK (89.5%, 79.5%), followed by ITS (86.2%, 71.0%). The two chloroplast genes (matK, trnH–psbA) gave poorer PCR results for the BB plot than the other three plots (JJYL, GGYL, LSL). In matK, the PCR success rate for the BB plot was only 80.7%, while for the other three plots were all over 90%. Similarly, the PCR success rate of trnH–psbA in the BB plot was 89.3%, but the other three plots gave rates over 95% (Table 2).

Table 2. PCR amplification and sequencing success of the four plots in Xishuangbanna.

Mistakes in taxonomic identification

Based on Neighbor-Joining analyses, 99 individuals (9.6%) were misidentified morphologically from the JJYL, GGYL and LSL plots (S1 and S2 Figs). Excluding the seven unknown individuals, we found that out of the remainder, 70 samples were misidentifications at the family level, while 17 and five samples were at the genus and species level, respectively. Comparing morphology-based identifications and corrected identification results derived from DNA sequences, we only observed seven cases in which all individuals of a certain species were mistaken for an another species. These were all cases of morphological convergence in vegetative characters e.g.: Chrysophyllum lanceolatum (Blume) A. DC. versus Ardisia scalarinervis E. Walker and Epiprinus siletianus (Baill.) Croizat versus Ixora amplexicaulis C.Y. Wu & W.C. Ko. However, most of the errors were found to be mistakes in individual sample identifications, resulting in some individuals of one species nesting with those of another.

Species resolution: single-region analysis

We conducted species identification analysis based on the corrected results derived from the reciprocal illumination procedure. The performances of the four markers using the three barcoding identification methods within our four plots in Xishuangbanna provided relatively similar results, although in this study, trnH–psbA could not be evaluated by analysis of reconstructing trees because we could not obtain good results from global multiple sequence alignments due to variations among such diverse taxonomic groups, mainly due to high numbers of insertions and/or deletions [31].

The highest success of species discrimination based on two tree building methods (NJ and ML) with single barcode were obtained with ITS (44.6% and 47.8%), followed by matK (34.1% and 35.3%) and then rbcL (28.5% and 27.8%). At the genus level, the results also performed best with ITS (72.8% and 77.2%), followed by matK (66.7% and 63.1%) and rbcL (64.3% and 59.1%) (see Figs 1 and 2).

Fig 1. Species resolution success at the family, genus and species levels for single regions and combinations, based on Neighbor-Joining Tree analysis of all the species (samples ≥ 2), collected from the four plots (BB, JJYL, GGYL and LSL) of Xishuangbanna Nature Reserve in southwest China.

(R, M, T, S represent rbcL, matK, trnH–psbA and ITS respectively.)

Fig 2. Species resolution success at the family, genus and species levels for single regions and combinations, based on Maximum Likelihood Tree analysis of all the species (samples ≥ 2), collected from the four plots (BB, JJYL, GGYL and LSL) in the Xishuangbanna Nature Reserve in southwest China.

(R, M, T, S represent rbcL, matK, trnH–psbA and ITS respectively.)

Of the three methods, BLASTn tended to show slightly higher discrimination success rates for the four genes. For all the 1858 samples for which we obtained at least one sequence, percent species-level resolution ranged from 58.1% (ITS) to 24.7% (rbcL) with trnH–psbA and matK having intermediate values of 43.4% and 31.6%, respectively. Similar patterns were observed for genus-level resolution; ITS again providing the highest genus discrimination success rate (76.2%) and rbcL the lowest (54.5%), while trnH–psbA (70.9%) and matK (64.4%) were intermediate. Thus, percentage species and genus resolution were higher with the two supplementary barcodes compared with the two core barcodes among the three analysis methods in our study. However, the four genes were all prone to higher species discrimination when we just used the samples of BB plot to manage a stand-alone BLAST. The species-level identification success rates for matK, rbcL, trnH–psbA and ITS were 60.0%, 61.3%, 79.7% and 84.7% respectively, with genus-level success rates for rbcL, matK, trnH–psbA and ITS of 79.8%, 84.4%, 94.6% and 95.3% (Fig 3).

Fig 3. Percent species resolution at the genus and species levels for single regions and combinations, based on BLAST analysis of all the samples, collected from the four plots (BB, JJYL, GGYL and LSL) of Xishuangbanna Nature Reserve in southwest China.

(R, M, T, S represent rbcL, matK, trnH–psbA and ITS respectively.)

Species resolution: multi-region analysis

We found little difference between the two methods of phylogenetic tree reconstruction (NJ and ML) for the different barcode combinations (Figs 1 and 2) as follows: rbcL + matK (41.3% and 42.9%), rbcL + matK + trnH–psbA (50.0% and 51.3%), rbcL + matK + ITS (58.9% and 60.8%) and rbcL + matK + trnH–psbA + ITS (68.6% and 60.7%).

The rbcL + matK barcodes identified 48.4% of the species we sampled in the four plots using BLASTn (Fig 3). The addition of a non-coding region to this combination increased resolution by 13.5% (to 61.9% for rbcL + matK + trnH–psbA) and a nuclear region by 16.7% (to 65.1% for rbcL + matK + ITS) and the addition of all four gene regions resulted in 70.6% species resolution. The genus-level resolutions across all four barcode combinations ranged from 70.2% (rbcL + matK) to 81.6% (rbcL + matK + trnH–psbA + ITS).

Barcoding accuracy at the regional scale

At the regional scale, we tested the effectiveness of BLAST for species and genus identification. Here, all samples from the dataset in the three small plots for species or genera identification should be present in our database of barcode sequences from the BB plot (Table 3). For the JJYL plot, identification was most successful with ITS at the species level (68.6%), followed by trnH–psbA (60.7%), matK (49.5%), then rbcL (44.0%). The genus-level identification success of the JJYL sequences reached 92.2% using ITS, 81.1% with rbcL, 73.0% with trnH–psbA and 71.2% with matK. The poorest performing barcode locus in the GGYL sequences was rbcL for both species- and genus-level discrimination, while the best was ITS. In contrast, in the LSL plot the performance of all four barcodes was different from the JJYL and GGYL plots, with the two core barcodes (rbcL, matK) showing higher discrimination success rates than the trnH–psbA and ITS sequences (Fig 4).

Table 3. Shared numbers of species and genera among the four plots (BB, JJYL, GGYL and LSL).

Fig 4. Barcoding success of Xishuangbanna tropical trees at a regional scale for species identification and genus identification.

(LSL-BB, GGYL-BB, JJYL-BB, BB-BB mean that the samples of BB serve as a database, while the samples of LSL, GGYL, JJYL and BB as queries respectively.)


Our study is the first attempt to barcode a local tropical tree flora from the Xishuangbanna Nature Reserve in Southwest China. The creation of our study provides a local platform for a broad range of applications that are reliant on large-scale species identification.

Sequence recovery

Our results showed higher sequence recovery for matK, compared with only 42% in one case [39] and around 70% in others [40,41,42], Fazekas et al. [43] reported a higher level of success (88%) for matK using 10 primer pairs, while sequencing success of 90% was obtained by the CBOL Plant Working Group using two primer pairs [24], which is similar with our results. This difference in relative performance may be explained either by the choice of primer combinations and/or the numbers of primer pairs used. Compared with other DNA barcode studies of tropical trees (e.g. 42% amplification success and 27% sequencing success for matK using three primer pairs in India [44]), our results for matK were much higher using only four primer pairs. This could have resulted from more sophisticated amplification, for example, Ford et al. (2009) obtained 85% success rate with matK using a combination of standard and nested multiplexed-tandem PCR (MT-PCR) [39]. In this study, we also conducted a modification of the cycling procedure named Ramp-PCR for each of the three chloroplast regions in the three smaller plots, resulting in higher PCR amplification and sequencing success for the three chloroplast genes and especially matK for these areas relative to the BB plot. Basic PCR programs were used during experiments of the BB plot, which were finished in the early 2009 at the beginning of DNA barcoding project in Xishuangbanna. Later, for the other plots, we attempted to use ramp-PCR. We consider that these observed differences in site-level barcoding success were due mainly to the combination of standard and Ramp PCR process used in the three small plots. As a result, we suggest that the additional cost of testing a large number of primer combinations of chloroplast genes might be less cost-effective than implementing more suitable, but non-standard PCR methods when conducting a DNA barcode study involving large numbers of phylogenetically diverse and genetically variable plant species.

The PCR amplification and sequencing success rates of the non-coding gene trnH–psbA region ranked only second to rbcL. This was in accordance with several other studies implying that the resolution of trnH–psbA was high enough to be considered as a barcode [7,8,36,45]. Although some studies have criticized the use of this marker because it is considered likely to develop errors during sequencing [46], we obtained high quality contigs with a success rate of 85.6%, which was nearly as good as rbcL, which is generally considered to be one of the most efficient barcode loci for plants [24,27]. Due to the high numbers of substitutions seen in trnH–psbA, it has the potential to be a suitable marker for discrimination between closely related species [47]. This barcode gene was recommended as one of the best performing locus for barcoding tropical tree species [35] and in terms of PCR amplification success, sequencing and species resolution, we would support this.

The only nuclear barcode gene in our study provided the highest species resolution ability among of the four tested regions using both tree-based and stand-alone BLAST methods in each of the four plots. ITS has been shown to discriminate species in many groups [14,16,48,49]; however, although we obtained an acceptable PCR amplification success rate of 86.2% for ITS, there was still a relatively low overall sequencing success rate of 71.0%. Some other studies have reported difficulties of sequencing ITS because of issues relating to secondary structure formations in this region [50,51]. Nevertheless, despite the lower sequence recovery by ITS compared to the other three markers at Xishuangbanna, we still obtained higher sequence recovery than some other DNA barcoding studies of tropical forests (41.0% for Amazonia [5] and 62.0% for India [44]). Therefore, we would support the use of ITS as a barcode marker for tropical tree species in Xishuangbanna.

Reducing mistakes in taxonomic identification

A tree-based approach using DNA barcoding in combination with morphology is very useful to revise mistakes of morphological identification [12]. We recommended the core barcodes-rbcL and matK to do this work for two reasons. First, the high rates of sequence recovery make it possible to find identification errors from as many samples as possible using the core barcodes. Second, data analysis is easy for the core barcodes because they are coding regions, for example, multiple sequence alignments. In contrast, it is difficult to assess error rates using ITS or trnH–psbA due in part to the high rate variation of sequences among such a large numbers of taxa, in contrast to work on the tropical tree Inga Mill. (Fabaceae), where the focus of that study was genus-specific [30]. From this perspective, it was less economical to detect error identifications or assign unknown samples to a certain taxon using the two complementary barcodes which were difficult to amplify, sequence and align among such a large numbers and variable taxa of tree species in tropics.


In this study, tree-based methods performed less well for identification than similarity-based methods using BLAST. This finding is also in agreement with the results of other studies comparing relative performance of DNA barcoding methods [13,52,53], probably because tree-based methods combine all sites and/or attempt to consider relationships among the species sampled, whereas the BLAST-based approaches use local comparisons among sequences, making them more sensitive to small differences among taxa [52].

Using the samples from all four plots in this study, we found that there was relatively low species identification success for the two core barcode regions, either alone (28.5% for rbcL and 35.3% for matK at the best), or in combination (48.4%) based on both tree-based and BLAST methods. These results are similar to previous findings for Indian tropical forests (rbcL 39.1%) [24] and suggest that the recommended standard barcode markers rbcL and matK may not always be suitable for tropical tree species discrimination at the species level. Before we assessed the performance of species identification, we corrected some apparent morphological identification errors only based on rbcL and matK which were amplified and sequenced more easily; detecting an overall 9.9% error rate in taxonomic identification using rbcL and matK: 77.1% (74/96) at the family level, 17.7% at the generic and 5.2% at the species level. This indicates that these two core barcode genes were still useful for detecting and correcting morphology-based identification mistakes in tropical Asian tree species at the family level, which makes them useful for tropical ecological research (e.g. for investigating phylogenetic communities).

The best results for species-level identification were gained by ITS (58.1%), followed by trnH–psbA (43.4%) in BLAST. The addition of each of the two barcodes to the combination (rbcL + matK) increased identification success from 48.4% to 61.9% by trnH–psbA and 65.1% by ITS. The three loci combinations of rbcL + matK + trnH–psbA and rbcL + matK + ITS provided slightly higher species resolution as that of the single loci of trnH–psbA and ITS or their combination, but the two barcode combination was more preferred here than three loci combinations in consideration of cost effectiveness. A number of studies relying upon trnH–psbA alone [16] or in combination with other regions [5456] have verified the utility and efficacy of this region for plant DNA barcoding. The high species resolution ability of ITS was tested and compared with other candidate regions in a barcoding context [18,57,58]. Though a suitable processing method of trnH–psbA and the low sequencing success of ITS needs to dealt with, we suggest both trnH–psbA and ITS as potential genes for tropical trees in the present study. This is in line with several previous studies [5,43,52,59].

One of the important observations was the relatively low species identification by all the four loci compared with earlier similar studies in tropical areas [5,24,29,60], prompting an investigation of this poor performance in our study. We considered the taxon proportion of different areas as one of the reasons for the discrimination performance. Species resolution in Panamanian forests reached up to 98% using trnH–psbA [29], while Xishuangbanna showed the poorest species discrimination (47.8%), with intermediate values for studies in Cameroon (84.3%) [60], Amazonia (64.0%) [5] and India (60.0%) [24]. The ratios of individual/genus were Panama: 5.7 (1035/296), Cameroon: 4.9 (772/159), Amazonia: 7.5 (1073/143), India: 3.66 (300/82) and Xishuangbanna: 7.9 (2052/259), while the matching ratios of species/genus for these regions was 1.63 (296/181), 1.71 (272/159), 1.77 (254/143), 1.82 (149/82) and 2.5 (655/259), respectively (see Table 4).

Table 4. Comparison of relationships between the ratio (individuals/genera or species/genera) and species identification success in different tropical areas.

A decreasing tendency in successful species identification was apparent when clade richness (species/genus) increased, yet the success rate was not affected by the number of samples per species. This result was in accordance with the research of Cameroon by Parmentier et al. [60]. This may also reflect the different generation times or mutation rates for the woody species in these areas, possibly contributing to the differences in species discrimination success rates [61,62].

This study showed barcode identification success with two data sets: one comprising samples with all sequences from taxa common to the four plots (BB, JJYL, GGYL and LSL) and the other, for all samples from the BB plot. Our species identification success rates for rbcL (61.3%) and matK (60.0%) for the BB plot alone were much closer to results from Amazonia (57.0% for rbcL and 61.0% for matK) using BLAST, while the results (79.7%,) of the supplementary barcode gene trnH–psbA performed at a level comparable to the Cameroonian rainforests of Africa (84.3%). The BB plot also displayed higher species resolution than the combined samples of the four plots in our study. In addition, we conducted a regional scale barcoding that involved the three small plots, JJYL, GGYL and LSL serving as query plots. The whole consequences using BLAST with the three small plots were lower than that with the BB plot itself and the results for JJYL and GGYL were much closer to each other than to LSL. This may have been a result of the similar vegetation tropical rainforest shared between JJYL and GGYL, whereas LSL consists of tropical seasonal moist forest.

These outcomes indicate an increase of genetic diversity from the local scale (single point of sampling) to the regional scale (multi-point of sampling) [63] and we did observe more intraspecific base substitutions when considering all samples in the four sites (Fig 5). Thus, the multi-point sampling strategy used here resulted in more variable intraspecific sequences, especially between different vegetation types, lowering species identification success.

Fig 5. Intraspecific single base difference among the four plots (BB = BB, JJYL = J, GGYL = G, LSL = L) in Xishuangbanna Nature Reserve.

(Gray = No Base Difference, Red = A, Green = T, Yellow = G, Blue = C)


This study is an initial assessment of barcoding tropical tree species within the Xishuangbanna Nature Reserve, southwest China. It demonstrated that in this area, there are ecological applications for identifying invasive species [33,64], construction of phylogenetic trees for community ecology [29,41] and evaluation of the effect of species identification errors on ecological theories [30].

Large-scale biodiversity inventories are based on accurate species identification. Unfortunately, errors are common for tropical trees, usually due to the lack of reproductive characters. DNA barcoding could quickly and effectively help to correct morphological identification errors [30]. Compared with the core DNA barcodes rbcL and matK, the species-level identification results for trnH–psbA and ITS in this study were more successful and we recommend using these two barcodes in combination as the preferred barcodes for tropical tree species in southwest China.

Supporting Information

S1 Fig. Neighbor-Joining (NJ) tree generated using rbcL sequences from the JJYL, GGYL and LSL plots.

Error identifications at the family, genus, species levels and unknown species were highlighted in blue, green, red and purple respectively.



S2 Fig. Neighbor-Joining (NJ) tree generated using matK sequences from the JJYL, GGYL and LSL plots.

Error identifications at the family, genus, species levels and unknown species were highlighted in blue, green, red and purple respectively.



S1 Table. The systems and reaction process of PCR amplification.




We are very grateful to David L. Erickson (National Museum of Natural History, Smithsonian Institute) and John W. Kress (National Museum of Natural History, Smithsonian Institute) for sharing a workflow of plant DNA barcoding in the plot and assistance with DNA barcoding analyses. We think those who assisted with sampling: Ling Zhang, Lang Li, Meng-meng Lu, Heng Li, Jie Yang, Jian-feng Huang and Hsi-wen Li (Kunming Institute of Botany, Chinese Academy of Sciences) and Ferry Slik for help with species identification. We thank the Xishuangbanna Tropical Rainforest Ecology Station (XSTRES), Chinese Academy of Sciences for providing logistical support and their help in the field.

Author Contributions

Conceived and designed the experiments: XQC JL. Performed the experiments: XCH XQC. Analyzed the data: XCH XQC. Contributed reagents/materials/analysis tools: XCH XQC JL. Wrote the paper: XCH XQC JGC JL.


  1. 1. Chase MW, Fay MF (2009) Barcoding of plants and fungi. Science 325: 682–683. doi: 10.1126/science.1176906. pmid:19644072
  2. 2. Janzen DH (1988) Tropical ecological and biocultural restoration. Science 239: 243–244. pmid:17769984
  3. 3. Brooks TM, Mittermeier RA, da Fonseca GA, Gerlach J, Hoffmann M, Lamoreux JF, et al. (2006) Global biodiversity conservation priorities. Science 313: 58–61. pmid:16825561
  4. 4. Costion C, Ford A, Cross H, Crayn D, Harrington M, Lowe A. (2011) Plant DNA barcodes can accurately estimate species richness in poorly known floras. PLoS ONE 6: e26841. doi: 10.1371/journal.pone.0026841. pmid:22096501
  5. 5. Gonzalez MA, Baraloto C, Engel J, Mori SA, Pétronelli P, Riéra B, et al. (2009) Identification of Amazonian trees with DNA barcodes. PLoS ONE 4: e7483. doi: 10.1371/journal.pone.0007483. pmid:19834612
  6. 6. Hebert PD, Ratnasingham S, de Waard JR (2003) Barcoding animal life: cytochrome c oxidase subunit 1 divergences among closely related species. Proceedings of the Royal Society of London Series B: Biological Sciences 270: S96–S99. pmid:12952648
  7. 7. Kress WJ, Wurdack KJ, Zimmer EA, Weigt LA, Janzen DH (2005) Use of DNA barcodes to identify flowering plants. Proceedings of the National Academy of Sciences of the United States of America 102: 8369–8374. pmid:15928076
  8. 8. Kress WJ, Erickson DL (2007) A two-locus global DNA barcode for land plants: the coding rbcL gene complements the non-coding trnH–psbA spacer region. PLoS ONE 2: e508. pmid:17551588
  9. 9. Roy S, Tyagi A, Shukla V, Kumar A, Singh UM, Chaudhary LB, et al. (2010) Universal plant DNA barcode loci may not work in complex groups: a case study with Indian Berberis species. PLoS ONE 5: e13674. doi: 10.1371/journal.pone.0013674. pmid:21060687
  10. 10. Gu J, Su JX, Lin RZ, Li RQ, Xiao PG (2011) Testing four proposed barcoding markers for the identification of species within Ligustrum L.(Oleaceae). Journal of Systematics and Evolution 49: 213–224.
  11. 11. Chase MW, Cowan RS, Hollingsworth PM, Van Den Berg C, Madriñán S, Petersen G, et al. (2007) A proposal for a standardised protocol to barcode all land plants. Taxon: 295–299.
  12. 12. Piredda R, Simeone MC, Attimonelli M, Bellarosa R, Schirone B (2011) Prospects of barcoding the Italian wild dendroflora: oaks reveal severe limitations to tracking species identity. Molecular Ecology Resources 11: 72–83. doi: 10.1111/j.1755-0998.2010.02900.x. pmid:21429102
  13. 13. Li DZ, Gao LM, Li HT, Wang H, Ge XJ, Liu JQ et al. (2011) Comparative analysis of a large dataset indicates that internal transcribed spacer (ITS) should be incorporated into the core barcode for seed plants. Proceedings of the National Academy of Sciences of the United States of America 108: 19641–19646. doi: 10.1073/pnas.1104551108. pmid:22100737
  14. 14. Muellner A, Schaefer H, Lahaye R (2011) Evaluation of candidate DNA barcoding loci for economically important timber species of the mahogany family (Meliaceae). Molecular Ecology Resources 11: 450–460. doi: 10.1111/j.1755-0998.2011.02984.x. pmid:21481203
  15. 15. Ebihara A, Nitta JH, Ito M (2010) Molecular species identification with rich floristic sampling: DNA barcoding the pteridophyte flora of Japan. PLoS ONE 5: e15136. doi: 10.1371/journal.pone.0015136. pmid:21170336
  16. 16. Gao T, Yao H, Song J, Liu C, Zhu Y, Ma X, et al. (2010) Identification of medicinal plants in the family Fabaceae using a potential DNA barcode ITS2. Journal of Ethnopharmacology 130: 116–121. doi: 10.1016/j.jep.2010.04.026. pmid:20435122
  17. 17. Starr JR, Naczi RF, Chouinard BN (2009) Plant DNA barcodes and species resolution in sedges (Carex, Cyperaceae). Molecular Ecology Resources 9: 151–163. doi: 10.1111/j.1755-0998.2009.02640.x. pmid:21564974
  18. 18. Wang W, Wu Y, Yan Y, Ermakova M, Kerstetter R, Messing J. (2010) DNA barcoding of the Lemnaceae, a family of aquatic monocots. BMC Plant Biology 10: 205. doi: 10.1186/1471-2229-10-205. pmid:20846439
  19. 19. Zhu H (2007) On the classification of forest vegetation in Xishuangbanna, southern Yunnan.
  20. 20. Zhu H, Cao M, Hu H (2006) Geological history, flora, and vegetation of Xishuangbanna, Southern Yunnan, China. Biotropica 38: 310–317.
  21. 21. Condit R (1998) Tropical forest census plots. Springer-Verlag and RG Landes Company, Berlin, Germany.
  22. 22. Cowan RS, Chase MW, Kress WJ, Savolainen V (2006) 300,000 species to identify: problems, progress, and prospects in DNA barcoding of land plants. Taxon 55: 611–616.
  23. 23. Burgess KS, Fazekas AJ, Kesanakurti PR, Graham SW, Husband BC, Newmaster SG, et al. (2011) Discriminating plant species in a local temperate flora using the rbcL + matK DNA barcode. Methods in Ecology and Evolution 2: 333–340.
  24. 24. Tripathi AM, Tyagi A, Kumar A, Singh A, Singh S, Chaudhary LB, et al. (2013) The internal transcribed spacer (ITS) region and trnH-psbA are suitable candidate loci for DNA barcoding of tropical tree species of India. PLoS ONE 8: e57934. doi: 10.1371/journal.pone.0057934. pmid:23460915
  25. 25. Fazekas AJ, Kesanakurti PR, Burgess KS, Percy DM, Graham SW, Barrett SC, et al. (2009) Are plant species inherently harder to discriminate than animal species using DNA barcoding markers? Molecular Ecology Resources 9: 130–139. doi: 10.1111/j.1755-0998.2009.02652.x. pmid:21564972
  26. 26. Thompson JD, Higgins DG, Gibson TJ (1994) CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position specific gap penalties and weight matrix choice. Nucleic Acids Research 22: 4673–4680. pmid:7984417
  27. 27. Edgar RC (2004) MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Research 32: 1792–1797. pmid:15034147
  28. 28. Liu K, Warnow TJ, Holder MT, Nelesen SM, Yu J, Stamatakis AP, et al. (2012) SATé-II: very fast and accurate simultaneous estimation of multiple sequence alignments and phylogenetic trees. Systematic Biology 61: 90–106. doi: 10.1093/sysbio/syr095. pmid:22139466
  29. 29. Kress WJ, Erickson DL, Jones FA, Swenson NG, Perez R, Sanjur O, et al. (2009) Plant DNA barcodes and a community phylogeny of a tropical forest dynamics plot in Panama. Proceedings of the National Academy of Sciences of the United States of America 106: 18621–18626. doi: 10.1073/pnas.0909820106. pmid:19841276
  30. 30. Dexter KG, Pennington TD, Cunningham CW (2010) Using DNA to assess errors in tropical tree identifications: How often are ecologists wrong and when does it matter? Ecological Monographs 80: 267–286.
  31. 31. Clerc Blain JL, Starr JR, Bull RD, Saarela JM (2010) A regional approach to plant DNA barcoding provides high species resolution of sedges (Carex and Kobresia, Cyperaceae) in the Canadian Arctic Archipelago. Molecular Ecology Resources 10: 69–91. doi: 10.1111/j.1755-0998.2009.02725.x. pmid:21564992
  32. 32. Mort ME, Crawford DJ, Archibald JK, O'Leary TR, Santos Guerra A (2010) Plant DNA barcoding: A test using Macaronesian taxa of Tolpis (Asteraceae). Taxon: 581–587.
  33. 33. Van de Wiel C, Van Der Schoot J, Van Valkenburg J, Duistermaat H, Smulders M (2009) DNA barcoding discriminates the noxious invasive plant species, floating pennywort (Hydrocotyle ranunculoides Lf), from non-invasive relatives. Molecular Ecology Resources 9: 1086–1091. doi: 10.1111/j.1755-0998.2009.02547.x. pmid:21564846
  34. 34. Blaxter M, Mann J, Chapman T, Thomas F, Whitton C, Floyd R, et al. (2005) Defining operational taxonomic units using DNA barcode data. Philosophical Transactions of the Royal Society B: Biological Sciences 360: 1935–1943.
  35. 35. Elias M, Hill RI, Willmott KR, Dasmahapatra KK, Brower AV, Mallet J, et al. (2007) Limited performance of DNA barcoding in a diverse community of tropical butterflies. Proceedings of the Royal Society B: Biological Sciences 274: 2881–2889. pmid:17785265
  36. 36. Lahaye R, Van der Bank M, Bogarin D, Warner J, Pupulin F, Gigot G, et al. (2008) DNA barcoding the floras of biodiversity hotspots. Proceedings of the National Academy of Sciences of the United States of America 105: 2923–2928. doi: 10.1073/pnas.0709936105. pmid:18258745
  37. 37. Stamatakis A, Hoover P, Rougemont J (2008) A rapid bootstrap algorithm for the RAxML web servers. Systematic Biology 57: 758–771. doi: 10.1080/10635150802429642. pmid:18853362
  38. 38. Liu J, Provan J, Gao LM, Li DZ (2012) Sampling strategy and potential utility of ndels for DNA barcoding of closely related plant species: A case study in Taxus. International Journal of Molecular Sciences 13: 8740–8751. doi: 10.3390/ijms13078740. pmid:22942731
  39. 39. Ford CS, Ayres KL, Toomey N, Haider N, Van Alphen SJ, Kelly LJ, et al. (2009) Selection of candidate coding DNA barcoding regions for use on land plants. Botanical Journal of the Linnean Society 159: 1–11.
  40. 40. Hollingsworth PM, Graham SW, Little DP (2011) Choosing and using a plant DNA barcode. PLoS ONE 6: e19254. doi: 10.1371/journal.pone.0019254. pmid:21637336
  41. 41. Kress WJ, Erickson DL, Swenson NG, Thompson J, Uriarte M, Zimmerman JK, et al. (2010) Advances in the use of DNA barcodes to build a community phylogeny for tropical trees in a Puerto Rican forest dynamics plot. PLoS ONE 5: e15409. doi: 10.1371/journal.pone.0015409. pmid:21085700
  42. 42. Fazekas AJ, Burgess KS, Kesanakurti PR, Graham SW, Newmaster SG, Husband BC, et al. (2008) Multiple multilocus DNA barcodes from the plastid genome discriminate plant species equally well. PLoS ONE 3: e2802. doi: 10.1371/journal.pone.0002802. pmid:18665273
  43. 43. CBOL Plant Working Group (2009) A DNA barcode for land plants. Proceedings of the National Academy of Sciences of the United States of America 106: 12794–12797. doi: 10.1073/pnas.0905845106. pmid:19666622
  44. 44. Newmaster S, Fazekas A, Steeves R, Janovec J (2008) Testing candidate plant barcode regions in the Myristicaceae. Molecular Ecology Resources 8: 480–490. doi: 10.1111/j.1471-8286.2007.02002.x. pmid:21585825
  45. 45. Devey DS, Chase MW, Clarkson JJ (2009) A stuttering start to plant DNA barcoding: microsatellites present a previously overlooked problem in non-coding plastid regions. Taxon 58: 7–15.
  46. 46. Chen S, Yao H, Han J, Liu C, Song J, Shi L, et al. (2010) Validation of the ITS2 region as a novel DNA barcode for identifying medicinal plant species. PLoS ONE 5: e8613. doi: 10.1371/journal.pone.0008613. pmid:20062805
  47. 47. Bruni I, De Mattia F, Martellos S, Galimberti A, Savadori P, Casiraghi M, et al. (2012) DNA barcoding as an effective tool in improving a digital plant identification system: A case study for the area of Mt. Valerio, Trieste (NE Italy). PLoS ONE 7: e43256. doi: 10.1371/journal.pone.0043256. pmid:22970123
  48. 48. Luo K, Chen S, Chen K, Song J, Yao H, Ma X, et al. (2010) Assessment of candidate plant DNA barcodes using the Rutaceae family. Science China Life Sciences 53: 701–708. doi: 10.1007/s11427-010-4009-1. pmid:20602273
  49. 49. Pang X, Song J, Zhu Y, Xu H, Huang L, Chen S. (2011) Applying plant DNA barcodes for Rosaceae species identification. Cladistics 27: 165–170.
  50. 50. DeSalle R (2007) Phenetic and DNA taxonomy; a comment on Waugh. Bioessays 29: 1289–1290. pmid:18022809
  51. 51. Waugh J (2007) DNA barcoding in animal species: progress, potential and pitfalls. Bioessays 29: 188–197. pmid:17226815
  52. 52. Arca M, Hinsinger DD, Cruaud C, Tillier A, Bousquet J, Frascaria LN. (2012) Deciduous trees and the application of universal DNA barcodes: a case study on the circumpolar Fraxinus. PLoS ONE 7: e34089. doi: 10.1371/journal.pone.0034089. pmid:22479532
  53. 53. Meier R, Shiyang K, Vaidya G, Ng PK (2006) DNA barcoding and taxonomy in Diptera: a tale of high intraspecific variability and low identification success. Systematic Biology 55: 715–728. pmid:17060194
  54. 54. Ragupathy S, Newmaster SG, Murugesan M, Balasubramaniam V (2009) DNA barcoding discriminates a new cryptic grass species revealed in an ethnobotany study by the hill tribes of the Western Ghats in southern India. Molecular Ecology Resources 9: 164–171. doi: 10.1111/j.1755-0998.2009.02641.x. pmid:21564975
  55. 55. Steven GN, Subramanyam R (2009) Testing plant barcoding in a sister species complex of pantropical Acacia (Mimosoideae, Fabaceae). Molecular Ecology Resources 9: 172–180. doi: 10.1111/j.1755-0998.2009.02642.x. pmid:21564976
  56. 56. Yi HW, XiaoYu T, Hai LL, Xiao MC, Ying XQ (2009) A two-locus chloroplast (cp) DNA barcode for identification of different species in Eucalyptus. Acta Horticulturae Sinica 36: 1651–1658.
  57. 57. Hollingsworth ML, Andra CA, Forrest LL, Richardson J, Pennington R, Long D, et al. (2009) Selecting barcoding loci for plants: evaluation of seven candidate loci with species-level sampling in three divergent groups of land plants. Molecular Ecology Resources 9: 439–457. doi: 10.1111/j.1755-0998.2008.02439.x. pmid:21564673
  58. 58. Sass C, Little DP, Stevenson DW, Specht CD (2007) DNA barcoding in the cycadales: testing the potential of proposed barcoding markers for species identification of cycads. PloS ONE 2: e1154. pmid:17987130
  59. 59. de Vere N, Rich TC, Ford CR, Trinder SA, Long C, Moore C, et al. (2012) DNA barcoding the native flowering plants and conifers of Wales. PLoS ONE 7: e37945. doi: 10.1371/journal.pone.0037945. pmid:22701588
  60. 60. Parmentier I, Duminil Jm, Kuzmina M, Philippe M, Thomas DW, Kenfack D, et al. (2013) How effective are DNA barcodes in the identification of African rainforest trees? PLoS ONE 8: e54921. doi: 10.1371/journal.pone.0054921. pmid:23565134
  61. 61. Lanfear R, Ho SY, Davies TJ, Moles AT, Aarssen L, Swenson NG, et al. (2013) Taller plants have lower rates of molecular evolution. Nature Communications 4: 1879. doi: 10.1038/ncomms2836. pmid:23695673
  62. 62. Smith SA, Donoghue MJ (2008) Rates of molecular evolution are linked to life history in flowering plants. Science 322: 86–89. doi: 10.1126/science.1163197. pmid:18832643
  63. 63. Dainou K, Bizoux JP, Doucet JL, Mahy G, Hardy OJ, Heuertz M. (2010) Forest refugia revisited: nSSRs and cpDNA sequences support historical isolation in a wide spread African tree with high colonization capacity, Milicia excelsa (Moraceae). Molecular Ecology 19: 4462–4477. doi: 10.1111/j.1365-294X.2010.04831.x. pmid:20854478
  64. 64. Bleeker W, Klausmeyer S, Peintinger M, Dienst M (2008) DNA sequences identify invasive alien Cardamine at Lake Constance. Biological Conservation 141: 692–698.