The phylogenetic analysis of Dalbergia (Fabaceae: Papilionaceae) based on different DNA barcodes

DNA extraction, amplification and sequencing of specimens:

The three Dalbergia samples were obtained from Guangdong and Hainan province (southern China), and tissue material for DNA extraction was obtained from living specimens. Genomic DNA was extracted from fresh tissue by means of the modified cetyltrimethyl ammonium bromide (CTAB) method (Zhou et al. 2016). Sample powder was transferred to a tube along with 5 ml CTAB buffer (4% CTAB, 0.1 M Tris-HCl, 1.4 M NaCl, 25 mM EDTA, 2% β-mercaptoethanol, 1% PVP, pH 8.0) (Dingguo, Beijing, China) and incubated at 65°C for 2 h in a water bath. Then the mixture was centrifuged at 12 000 g for 10 min. Clear supernatants were collected and extracted by mixing equal-volumes with a mixture of trichloromethane and isoamyl alcohol (24:1) three times. Clear supernatants were collected after centrifugation and precipitated by 90% ethanol for 30 min, then washed with 70% ethanol three times. The precipitated DNA was dissolved with 1×TE and stored at −20°C for further study. Three primer pairs corresponding to ITS, matK, rbcL and psbA-trnH were used for polymerase chain reaction (PCR) amplifications carried out on a S1000 Thermal Cycler (BIO-RAD, USA). Primer sequences are listed in the Supplementary Table S2. Using a reaction volume of 20 μl, PCR conditions were as follows: denaturation step at 94°C for 3 min, followed by 30 cycles at 94°C for 1 min, 56°C for 1 min and 72°C for 1 min and the final extension step at 72°C for 10 min. PCR products were purified and sequenced by Ruibiotech Inc. (Guangzhou, China), with the same primers as for amplification. All reagents were purchased in Beijing Dingguo Changsheng Biotechnology Co. Ltd., (Beijing, China).

Data analysis:

Dalbergia species sequences of ITS, matK, rbcL and psbA-trnH were received from NCBI. Multiple sequence alignment (including trimming, visual inspection and manual adjustments) was performed in MEGA version 6.0 (Tamura et al. 2013). Then four barcoding regions were concatenated to produce combinations of datasets according to the voucher number.

Distance-based analyses (TaxonDNA):

The intra- and interspecific distances calculation based on the method “uncorrected p-distances” was performed by means of the TaxonDNA program (Meier et al. 2006). With regards to the overlapping distribution, the covering range between intra- and interspecific distances, the error margin 5% was removed on both ends of the intra-/interspecific distances. However, overlapping distributions may be interpreted as a failure of DNA barcoding, and do not necessarily mean that barcodes perform poorly for identification (Collins and Cruickshank 2012). Dot plots of intra-/interspecific distances is a good way to illustrate the barcoding gap (Robinson et al. 2009). In this study, the maximum intraspecific distances and minimum interspecific distances for each sequence were calculated by the “extreme pairwise” function in TaxonDNA. The relationship between DNA barcoding and the barcode gap was plotted on several scatter graphs. Species, which fall above the 1:1 slope, exhibited a barcode gap. However, singletons could not generate scatter plots due to the lack of conspecific matching.

Specimen identification was performed using a Taxon DNA/Species Identifier 1.8 (Meier et al. 2006) based on three different criteria: BM, BCM and all species barcodes (ASB), where the methods uncorrected p-distances and a minimum sequence overlap of 300 bp were applied. In these criteria, each sequence from the dataset served as a query against other sequences from the same dataset. With BM, any query sequence was identified as a BM with the smallest distance to all conspecific matches. The BCM requires a threshold value (95% of all intraspecific distances) that defines an upper bound to identify queried sequences. As for ASB, the closest match for a query is a list of all known sequences for a single species (Meier et al. 2006).

Tree-based analyses:

Phylogenetic and molecular evolutionary analyses were performed using MEGA version 6.06 (Tamura et al. 2013). Neighbor-joining (NJ) trees were constructed in MEGA using the Kimura-2-Parameter (K2P) model, which is based on distance substitution (Kimura 1980). One thousand bootstrap replicates were run to assess the relative support for the branches, while uninformative characters (gaps and missing data) were completely deleted. However, unknown Dalbergia specimens were assigned by the NJ tree using with the ITS+matK+rbcL dataset including all known and unknown Dalbergia samples (n=191).

Sequences

The dataset totally included 1175 downloaded DNA barcoding sequences corresponding to 86 species (Table 1). The ITS region has the highest number of sequences and species, but it has 53 (15.3%) singletons. The matK, rbcL, as well as their combination, have an approximate singleton rate and species number less than 30%, because several researchers frequently combined matK and rbcL as one barcode for further research (Bhagwat et al. 2015; Hartvig et al. 2015). The cpDNA region psbA-trnH have fewer sequences and species. The sequences of three Dalbergia specimens were submitted to the NCBI GenBank database (KY489987 to KY489998).

Distance analysis and barcoding gap

Our results demonstrate that the mean of the intraspecific variation was significantly lower than the corresponding interspecific divergence (Table 2). The rbcL region had the lowest value of mean intra- and interspecific distances (0.08% and 0.71%, respectively), which was approximately half of matK. The combined barcode ITS+psbA-trnH shows higher values (0.96% and 11.7%, respectively) in the combined barcoding set, which are almost nine times as much as matK+rbcL (0.11% and 1.25%).

A suitable barcode for species identification should exhibit a barcode gap-maximum intraspecific distance that is smaller than the minimum interspecific distance within species (Meier et al. 2008; Hartvig et al. 2015). Dot plots of inter-/intraspecific distances that fall above the 1:1 line illustrate a barcode gap (Collins and Cruickshank 2012). As for single barcodes, psbA-trnH showed the most barcode gaps (56.6%), following the ITS region and matK (Figure 1, Supplementary Table S4). The chloroplast gene rbcL had the worst performance in barcode gap for Dalbergia specimens (only 11.6% of specimens can yield a gap). The combined barcoding ITS+psbA-trnH, ITS+rbcL and ITS+matK+rbcL have a better performance on barcode gaps (Figure 1).

Figure 1:

Distribution of barcode gaps.

Maximum intraspecific variation versus minimum interspecific divergence (%) for the single and combined barcodes. Singletons cannot generate scatter plots due to lack of conspecific matching. Each plot represents one or several species. (a) and (b) show the distribution barcode gaps for four single barcodes; (c)–(f) show the distribution of barcode gaps for combined barcodes.

Nucleotide variation and phylogenetic analysis

The conserved and variable sites and parsimony informative sites were comparable for both single and combined barcodes (Table 2). The ITS region shows 53.7% variable sites and 48.7% parsimony informative sites followed by ITS+psbA-trnH and other barcodes. For conserved sites, the highest was rbcL+psbA-trnH (94.2%) following rbcL and matK+rbcL+psbA-trnH.

The ITS+matK+rbcL dataset for phylogenetic analysis included 1636 positions, from which gaps and missing data were eliminated. The optimal tree, with the sum of branch length=0.35406786 is presented. Dalbergia ecastaphyllum and D. monetaria were 100% supported as sister species. However, a well-supported group, comprised of D. tamarindifolia, D. pinnata, D. rubiginosa, D. candenatensis and D. velutina was found in the tree. Two well-supported groups are visible: D. horrida and D. thorelii (97%), as well as D. latifolia, D. cochinchinensis and D. ovata (97%). The focus species D. odorifera was monophyletic and was nested within the species D. rimosa and D. entadoides.

Specimens identification through BM/BCM and ASB

The performance for specimens identification by DNA barcoding was tested by the methods BM, BCM and ASB as indicated above. The rate of correct identification, ambiguous and incorrect identification for each method and barcode are presented in Supplementary Table S3.

The combined barcoding ITS+psbA-trnH has the highest correct identification rates (97.7%) and rbcL the lowest (34.4%) obtained by the BM and BCM approaches. The correct identification rates of combined barcodes are higher than single barcodes except matK+rbcL. For the ASB method, the rates of correct identification decreased below 50% for four barcodes (ITS, matK, rbcL and matK+rbcL), but the other barcodes still have high correct identification rates.

The ambiguous and incorrect identification specimens in BCM for ITS+matK+rbcL are seen in the neighbor-joining tree (Figure 2, n=14). All singletons are marked on the tree as the dataset did not show a conspecific match. However, D. assamica and two D. rimosa specimens are marked on tree as their minimum intraspecific distance is larger than the incorrect identification rate in the BM/BCM evaluation.

Figure 2:

Neighbor-Joining tree of K2P distances for ITS+matK+rbcL barcode.

Some clades have been compressed to triangles due to more than four conspecific individuals. Specimens are colored regarding to misidentification of “best close match”.

Specimens identification ability of DNA barcoding

To test the ability of specimen identification by DNA barcoding, three distance based methods (BM, BCM and ASB) were used (Meier et al. 2006). BM and BCM generated comparable results except for the rbcL region and outperformed considerably the ASB approach (Table 3). The BM criterion is the assigned sequence with the smallest distance to query all conspecific data and the BCM request sequences within 95% of all intraspecific distances of the best match. The combined barcode ITS+psbA-trnH yielded the highest rate of correct identification, while rbcL yielded the lowest performance (Figure 3). The poor performance of rbcL is due to the low percentage of variable sites and parsimony informative sites (Albert et al. 1994) (Table 2), moreover, many specimens’ intra- and interspecific distance are zero in distance comparisons (Figure 1). The cpDNA region psbA-trnH and combined barcodes with psbA-trnH show satisfactory discrimination performances in BM and BCM, because of the low number of singletons. psbA-trnH was found to have high conservation within conspecific and own proper variation among the interspecific rather than the congeneric. Although psbA-trnH included inversions and indels, it still has a good discrimination ability in some closely related species (Liu et al. 2015; Yu et al. 2016).

Figure 3:

Species identification performance of DNA barcode in Dalbergia.

Resolution at species level for four DNA barcoding and combinations of two or three barcodes using BCM, “best close match” and ASB, “all species barcode” function of the program TaxonDNA. The letters on the y-axis represent different barcoding: I, ITS; M, matK; R, rbcL; P, psbA-trnH. Singleton were included as source of resolution failure.

Conversely, the low identification success of ASB is related to the more rigorous rules of this strategy (based on the “best close match” including all conspecific barcodes of the query as BM) (Meier et al. 2006; Virgilio et al. 2010). The correct identification rates of all barcodes in ASB were rapidly decreased comparing to BCM, except rbcL (increased from 34.4% to 43.9%). Singletons are also a source of incorrect identification due to the absence of other conspecific reference sequences in the dataset without matching (Ross et al. 2008).

Lahaye et al. (2008) showed that matK and rbcL barcoding cannot yield a barcode gap in the closely related plant species. Moreover, several studies also indicated the poor performance of the barcode gap for rbcL (Pettengill and Neel 2010; Bhagwat et al. 2015; Zhang et al. 2015). In the present study, 45.4% of the specimens can yield barcode gap in matK, while rbcL yields the lowest percentage in this context (Figure 1, Supplementary Table S4). However, the combined barcoding ITS+psbA-trnH yields the highest rates of barcode gap due to the presence of a high rate of parsimony informative sites.

Phylogenetic relationships within Dalbergia species

Vatanparast et al. (2013) performed a phylogeny analysis based on ITS barcoding and included 64 Dalbergia species and Hartvig et al. (2015) tested 31 Dalbergia species based on ITS, matK and rbcL barcoding. Our dataset included almost all the detected Dalbergia species sequences aimed at barcoding (ITS, matK, rbcL and psbA-trnH) accessed from GenBank (Supplementary Table S1).

The nuclear ribosomal DNA ITS can be amplified in two smaller fragments (ITS1 and ITS2) adjoining the 5.8S locus, which was especially useful for degraded samples (Kress et al. 2005). Here, it was found that ITS has a high proportion of variable sites, mainly contributed by the ITS1 and ITS2 region resulting from a high number of indels. Two chloroplast genes, matK and rbcL, with high variation and identification ability, were suggested as alternative barcodes (CBOL Plant Working Group 2009). The present study shows that matK and psbA-trnH provided appropriate rates of variable sites, and rbcL possessed a high proportion of conservation sites within Dalbergia species (Table 2).

With regards to the phylogenetic tree, there are some basal polytomies where the multiple species are still present. A few well-resolved groups, however, are seen in this study, which are consistent with the findings in previous studies. It was found that D. tamarindifolia and D. rubiginosa were closely relative to two sister species D. pinnata and D. candenatensis (Ibeiro et al. 2007; Hartvig et al. 2015). The sister relationship between D. ecastaphyllum and D. monetaria (members of the tropical section Ecastaphyllum (Carvalho 1997) was also observed in this study. Dalbergia assamica is not monophyletic, but has been placed together with D. balansae because D. assamica is a synonymy of D. balansae (Chen et al. 2010b; Hartvig et al. 2015). However, in the present study, D. latifolia voucher DlatP was not in the D. latifolia group, but was located together with D. melanoxlon. This is probably due to the misidentification of specimens.

Barcodes performance for unknown Dalbergia specimens

Three Dalbergia samples were assigned to D. odorifera in the “cluster” function and were nested with D. odorifera in the NJ tree (Figure 4). Tested with the three unidentified Dalbergia samples, the proposed identification of the specimens by these DNA barcoding was found to be the best discrimination. The combined barcode ITS+matK+rbcL has a good performance for specimen identification, which is consistent with data of a study of Hartvig et al. (2015).

Figure 4:

Neighbor-Joining tree for ITS+matK+rbcL barcode, including Dalbergia sp. specimens.

Some clades have been compressed to triangles due to more than four conspecific individuals.

Several studies stated the difficulty to discriminate all plants by means of a universal DNA barcode. Therefore, a multi-locus approach for species identification is necessary (Lahaye et al. 2008; Bourke et al. 2013; Tripathi et al. 2013; Bhagwat et al. 2015; Hartvig et al. 2015). Though the combinations ITS+matK+psbA-trnH and ITS+rbcL+psbA-trnH were not tested in this study, combined barcodes provide certainly a good species identification for Dalbergia species.

One of the main constraints of DNA barcoding is the incompletion of the reference dataset, which cannot comprehensively represent the taxonomic diversity of the group to be identified (Virgilio et al. 2012). Then the quality of the reference dataset needs to be taken into account. The dataset in this study included numbers of singletons which affected the correct identification rates in BM/BCM/ASB due to a lack of conspecific matching. Therefore, it is essential to establish a large DNA barcoding library representing the earth’s botanical diversity. A feasible method involves sampling several voucher specimens for each species, ideally from different geographical locations (Gonzalez et al. 2009; Bhagwat et al. 2015). The DNA barcoding has a high potential for further development.