The recently concluded Fourth International Conference on DNA barcode held at Adelaide, Australia emphasized on DNA barcoding of tree species. The main aim of barcoding tree species is to check the illegal trade of timbers as well as CITES-listed tree species. Forest trees are fast disappearing worldwide, especially in the developing countries. This is due to deforestation and urbanization and meeting various human needs. In a recent study, it has been predicted that over half of the estimated 11,000 Amazonian tree species may face a direct risk of extinction [1].Thus, there is an urgent need to carry out an inventory and manage diversity using innovative tools, like DNA barcoding. Significant progress has been made in mapping the Neotropical plants during the last decades [2]–[5], including studies on DNA barcoding in tree species [6] and other plant species [7]–[9]. However, such studies on tree species of India are lacking. Large-scale biodiversity inventories are critically needed in order to develop conservation strategies for the diverse ecosystems of India. It has been suggested that the application of new tools, like DNA barcoding, may help identify species with a high confidence, which would be useful in a wide array of applications, including discriminating forest species and large-scale biodiversity surveys [10]–[14].

The DNA barcoding technique utilizes a short fragment of DNA sequence to identify the species. Since its first report in 2003 [10], it has been successfully established in animals as a tool to identify species and taxonomic clarification [10]. A portion of the mitochondrial cytochrome c oxidase 1 (COI or cox1) gene sequence is currently being used as a universal barcode in certain animal groups [10], [11], [15], [16], fungi [17], diatoms [18], and red algae [19]. However, COI has proved to be an unsuitable barcoding marker in land plants [20]–[22], primarily because of the low nucleotide substitution rates of the plant mitochondrial genome [23]. The plastid DNA sequences have been the focus for DNA barcodes for plants. The success rate using plastid markers has not been so high, compared to animals. This is mainly because there are certain inherent difficulties for some groups of plants, which have been well discussed as far as the barcoding of these groups is concerned [24]–[27]. After evaluating the performance of seven leading candidate plastid DNA regions (atpF–atpH, psbK–psbI, trnH–psbA spacers and matK, rbcL, rpoB, rpoC1 genes), the Plant Working Group of the CBOL recommended the two-marker combination rbcL/matK as the standard DNA barcodes for plants [13] and, later on, added ITS and trnH-psbA as supplementary barcode loci. Yet, the screening for single or multiple regions from plastid and nucleus, appropriate for DNA barcoding in different plant groups, has been an important research focus around the globe.

The four loci, namely, matK, rbcL, ITS, and trnH-psbA have been studied in great detail in taxonomic and floristic contexts as the barcode loci for plants [6], [13], [25], [27]–[34]. Each region has its own strengths and weaknesses to be considered before establishing the universal barcode loci for plants. While rbcL remained as the most effective loci in terms of sequence quality and recovery, but not in species discrimination ability [6], [13], [25], matK was found to be effective in species discrimination in some cases [7], [13], [21], [29], [33], but not in others [25], [35], [36]. Moreover, in most cases, sequence recovery and the quality of matK was not as effective as that with rbcL [7], [25], [37]. Similarly, there are mixed reports for the consideration of trnH-psbA as barcode locus, some advocating in favor [14], [25], [36]–[39] and others disregarding it [13], [40]. The nuclear region that has so far been considered as barcode loci is the ITS. Though initially, this locus was not advocated by many due to its inherent difficulties–for example, the existence of paralogs, chances of fungal contamination, difficulties in sequence recovery and so on, as discussed [41]–yet, recently, ITS has been reported to be an efficient barcode locus by many [25], [33], [42], [43]. This is mainly due to the advantages gained in using ITS, which outweigh its limitations. More recently, ITS2–part of the ITS region–has been shown as the best barcode locus due to its higher species discrimination and sequence recovery ability across different plant groups [44]–[46]. Thus, these studies indicate that there is a need to further validate these loci among different plant groups in order to choose the most favorable one or their combination thereof. Though the effectiveness of these candidate barcode loci has been discussed, most of them were concerned with herbaceous or shrubs taxa, as mentioned above, and only a few studies are reported on tree and other long living taxa [36], [42], [47], [48]. As echoed by many, for example, Gonzalez et al. [6], we assume that DNA barcoding in tropical trees may be challenging as compared to temperate plants. Moreover, DNA extraction from tropical trees is sometime difficult due to the greater abundance of secondary metabolites [49]. Further, it has been shown that woody plant lineages show consistently lower rates of molecular evolution as compared to herbaceous plant lineages [50], suggesting that the application of DNA barcoding concept should be more difficult for tree floras than for non-woody floras [7], [27]. Applications of DNA barcoding in the tropical trees, especially from India, are still unexplored.

India is considered to be one of the 17 mega biodiverse countries of the world, with four biodiversity hotspots. There are around 19,294 species of flowering plants in India of which around 2,560 species are estimated to be tree species [51]. The country has been divided into 12 bio-geographical zones. The samples were collected from the National Botanic Garden (NBG), Lucknow, as well as other parts of the province of Uttar Pradesh (Figure 1). The NBG at Lucknow (Uttar Pradesh) lies at 26° 55′ N, 80°59′ E and at an altitude of 113 m above the mean sea level. The region falls under the range of the bio geographic zone of the Gangetic Plain, which has a typical moist tropical climate. The garden, spread over an area of 65 acres, is a repository of germplasms collected from the various parts of the country as well as some exotic collections. It harbors approximately 5,000 taxa, out of which about 345 species belong to trees. The repository represents almost all the tree species of Uttar Pradesh, which consists of approximately 400 tree species. Initially, we examined 236 accessions using all the five loci for PCR, sequencing, and species discrimination ability. ITS and trnH-psbA provided the best species resolution using this data. Then, we added another 64 accessions for ITS and trnH-psbA to determine the species discrimination ability in 300 accessions by these loci only. The representative tree species from the NBG as well as from the other parts of the province have been selected to examine the efficacy of standard and proposed barcode loci (ITS, matK, rbcL, trnH-psbA, and ITS2) in tropical tree species.

We examined the efficacy of five barcode loci in identifying the tropical tree species. This is the first report on the evaluation of barcode loci in tropical tree species from India, which is considered one of the mega bio-diverse countries in the world.

An ideal DNA barcode must have adequate conserved regions for universal primer design, high PCR amplification efficiency, and enough variability to be used for species identification [13]. The PCR success rate for all the loci was variable. The striking observation was the extremely low PCR success using matK primers. The three different primer pairs of matK were not able to provide more than 50 percent PCR success. There are mixed reports about PCR and sequencing success using matK primers, depending upon the use of particular primer/s and data sets [25]. This and other studies suggest that existing matK primers may not be suitable for plant barcode purposes in general, and tree species, in particular. We are not able to comment on the species identification ability of the matK in tree species based on our present study.

The PCR success rate using the trnH-psbA was lower than rbcL and ITS. This is in contrast to earlier findings, where PCR amplification (as well as species discrimination ability) of trnH-psbA was high enough to be considered as barcode loci [7], [8], [13], [21], [22], [37]. The higher rate of PCR amplification success in some of these reports may be attributed to the sampling strategies, which were restricted to particular taxon, or to sampling, which were very shallow in some cases. However, in the CBOL plant working group study involving large numbers of taxa, it was observed that trnH-psbA possess attributes that are highly desirable in a plant DNA barcoding system, such as universality, sequence quality and coverage, and discrimination ability [13]. The plant working group rejected this locus because of difficulties in bidirectional sequencing and also due to its higher length (>1000 bp) in some monocots [14] and conifers [52]. We did not increase our effort to obtain a higher PCR amplification rate of trnH-psbA using different PCR conditions. But, if efforts are exerted in this direction, like matK, the PCR amplification rate may be increased to suit the universality criterion. Finally, the use of the trnH-psbA locus has been much criticized because it is prone to errors during sequencing [53]. However, we were able to get high-quality bidirectional sequences with the success rate falling next to rbcL, which is considered the most efficient barcode locus as far as sequence quality is considered. The high quality bidirectional sequence was also obtained by Maia et al. in the family Bromeliaceae, where only 2 out of 101 samples failed in generating high quality trnH-psbA bidirectional sequences after the second trial [9]. Except for the low PCR amplification rate of trnH-psbA, our study agrees with the findings of the work reported by Gonzalez et al. [6], where they observed that trnH-psbA was the best performing locus for barcoding Amazonian trees. We emphasize here that, to the best of our knowledge, the study of Gonzalez et al. was the only study involving a large number of tropical tree species for tree barcoding. Our study also indicates that trnH-psbA can be considered as a potential tree barcode marker, especially for tropical tree species, with a success rate of species identification ranging from 57.4 percent to 67.7 percent, depending on the method and data set used.

The only nuclear-originated barcode locus that has gained popularity is the internal transcribed spacer region (ITS). The ITS provided the highest species resolution ability among all the five tested loci using tree-based and standalone BLAST approaches with data set 2. The ITS sequences have been proposed as a barcode locus for plants for some time [22]. It was recently suggested as an additional marker by CBOL [13]. ITS has been validated as an efficient barcode locus for identifying species in many groups [31], [42], [44], [54]–[56], including ashes [44] and other tree genera, such as Cedrela [42] and Quercus [39]. One of the most extensive latest barcoding studies also identified ITS as the best alternative to matK and rbcL [44]. In our earlier study, ITS provided the best species resolution in the tree Ficus and in the shrub Gossypium [25]. However, other studies described its inherent difficulties, for example, low PCR success [6], [14], [37], [44], problem of secondary structure formation resulting in poor quality sequence data [57], [58], and multiple copy numbers [59], [60] etc. In our study too, the sequencing success rate was low. In the study by Gonzalez et al. [6], ITS provided a low sequencing success rate (41.0%), but a higher species discrimination rate (80.3%) of tropical Amazonian forest trees. Another major apprehension that has been shown in using ITS as the plant barcode marker is the chance of contamination with endophytic fungus. As reported by Chen et al. [44], we also did not find any fungal sequences when we retrieved sequences after BLAST. In all cases, the best match retrieved the plant species, either as the same plant species sequence as query sequences or as the other plant species. Secondly, we looked for the presence of the characteristic conserved motif in the 5.8S rRNA gene of angiosperm plant ITS sequences [61]. The characteristic motif (5′-GAATTGCAGAATCC-3′) was found in all the ITS sequences whereas the variant of the motif generally found in fungi (5′-GAATTGCAGAATTC-3′) was not found in any of the sequences. Though we did not study the occurrence of paralogous ITS sequences, none of the PCR products showed multiple bands nor were the sequences unreadable, in a fashion easily attributable to the presence of multiple divergent copies. Because these copies generally evolve in a concerted fashion, leading to a single detectable sequence per plant. In order to consider ITS as the barcode locus, these potential problems and their remedies have been further discussed by Hollingsworth [41]. In the present study too, though we did not observe many of the difficulties associated with ITS as a barcode marker, the low sequencing success rate needs to be dealt with.

As an alternative to complete the ITS (ITS1-5.8S-ITS2) region, ITS2 has been suggested as a barcode locus [44], [46]. While one of these suggestions was based on in silico analysis [46], another study compared seven candidate DNA barcodes (psbA-trnH, matK, rbcL, rpoC1, ycf5, ITS2, and ITS) from medicinal plant species and found ITS2 to be the most suitable region (more than 6,600 plant samples belonging to 4,800 species from 753 distinct genera tested), with 92.7 percent discrimination ability at the species level [44]. Following these, other studies also reported about the usefulness of this locus as a barcode marker [31], [62], [63]. We observed a low PCR (59.0%) and sequencing success rate (60.0%) using these primers, whereas species resolution was found to be only 42.8 percent, 48.7 percent, and 9.0 percent using the barcoding gap principle, standalone BLAST, and BLASTn methods, respectively. In evaluating candidate DNA barcoding loci for the economically important timber species of the Mahogany family (Meliaceae), Muellner et al. observed that ITS2 is less variable and had fewer PICs as compared to ITS1. Similarly, Liu et al. did not find ITS2 to be a suitable marker for species identification in Taxus lineages, where only 5 of 11 Taxus lineages were discriminated by ITS2 [56]. More recently, Sun et al. [64] observed a low PCR success rate by using ITS and ITS2 in Dioscorea. Though ITS2 has some other advantages, for example, ease of routine sequencing (as observed in the present study) and conserved length, as compared to ITS1 [44], echoing Hollingsworth et al. [38], we assume that further sampling is needed, involving a wider range of plant species, to assess the potentiality of ITS2 as the barcode loci for land plants.

It is now well established that a universal DNA barcode for plants may be elusive. Therefore, the necessity for a multilocus approach has been suggested [13], [22], [42], [52], [65]. Though we did not try all possible combinations in this study, a two loci combination of ITS and trnH-psbA provided better species resolution in the barcode gap method as well as the tree-based approach than when used alone. A number of studies relying on trnH–psbA alone [66] or in combination with other regions [48], [67], [68] have confirmed the utility and efficacy of this region for plant barcoding [7]. Hollingsworth et al. [52] recently summarized the seven empirical studies published till date that have involved comparisons of multiple regions in a barcoding context [7], [8], [21], [22], [37], [40], [52]. In most of these studies, ITS was not considered as a component of the multilocus system, whereas trnH-psbA was one of the main loci in most cases. In our earlier work too, we found that the combination of these two loci worked better than other combinations in species resolution of the Indian Berberis [25]. More recently, Li et al. [33] showed that adding ITS to the multilocus system took the levels of species discrimination success from 50 percent to 62 percent for two or three marker plastid barcodes, to between 77 percent and 82 percent, when ITS was combined with two plastid markers. The higher success rate of species resolution that we observed here using the combination of ITS and trnH-psbA (80.0%) may be due to the lower number of species in this data (25 species). Though the three loci combination of ITS, trnH-psbA and rbcL provided slightly higher species resolution ability (may be due to less number of species than two loci combination) as that of two loci combination of ITS and trnH-psbA,for obvious reason of cost effectiveness, the two loci combination is preferred here than three loci combination.

One of the important observations was the very low species identification by all the loci in NCBI BLAST analysis, though other studies have reported higher species resolution in the BLAST approach [36], [44]. This prompted us to check the availability of a specific locus sequence of a particular species in the NCBI database. Our analysis suggests that 68.0 percent and 61.3 percent of the studied species were not represented by the trnH-psbA and ITS sequences, respectively, in the NCBI database (Table S5). This could be one of the reasons for the low species identification rate using NCBI BLAST. This further emphasized the need to enrich the NCBI database with nucleotide sequences of the different barcode loci of tree species to be effectively used as the standard barcode reference database.

Overall, although there were differences in species resolution, depending on the methods and data set used for a particular locus, considering the best possible representative data (data set 2, having higher species numbers with more than one accession), popular methods (barcode gap and tree-based), species discrimination ability, and results of our statistical analysis of inter- and intra-specific divergence, trnH-psbA and ITS, both can be considered as potential barcode loci for tropical tree species. This is in line with the standard success rate of plant species discrimination level at 60–72 percent, as reported in most of the earlier studies [6], [13], [29], [36].

Tropical forests play a key role in developing large-scale biodiversity inventories and conservation strategies. The DNA barcode of tropical tree species, as reflected here, may not be as difficult as presumed [6]. For example, DNA extraction is expected to be more difficult in tropical plants, due to the greater abundance of secondary metabolites [69], which may affect the overall performance of DNA barcoding. However, we did not find any difficulties in DNA extraction except in case of Shorea robusta and Delbergia sissoo. These two samples required slight modifications in the extraction protocol as described in materials and methods. Moreover, the high PCR and sequencing success rate by rbcL reflect that good quality DNA could be obtained from of tropical tree species for DNA barcoding. Thus, the low PCR and sequencing success rate by matK and ITS2 may not pertain to DNA extraction and quality of DNA, rather due to non specificity of primer sequences. Though the species discrimination rate here is not very high yet, the identification of ∼70 percent of tree species by ITS and trnH-psbA alone, following the tree-based approach, holds promise as a good candidate marker for tropical tree species identification. This is considering the present success rate of not more than 60–72 percent in a large number of plant species barcode studies reported so far. Further large-scale studies on tropical tree species are needed to establish that the combination of ITS and trnH-psbA is better for barcoding tropical trees. This is more important because the combination of a nuclear marker and a plastid marker has an added advantage: the nuclear marker, being inherited from both parents, would provide much more information than an organellar marker alone, which is inherited from only one parent. The matK and rbcL combination may not work as the ideal barcode loci for tropical tree species, as indicated in the present study and other studies involving tree species [42], [52].