The geographical distribution of different species of Berberis , Ficus and Gossipium . Circles indicate the regions from where species were collected. Exact GPS data are shown in Table S1. Region A (western Himalayas): B. aristata, B. asiatica, B. chitria, B. glaucocarpa, B. jaeschkeana, B. lycium, B. pachyacantha, B. umbellata; Region B (Eastern Himalayas): B. angulosa, B. griffithiana, B. insignis, B. macrosepala, B. replicata; Region C (Satpura - Central India): B. Hainesii; Region D (Nilgiri Hills-Southern India): B. tinctoria, B. wightiana; Region E (Gujrat): G. herbaceum, G. arboreum Region F (Assam): G. barbadense ; Region G (Karnataka): G. herbaceum, G. arboretum, G. hirsutum Region H (Andhra Pradesh): G. hirsutum; Region I (Tamil Nadu): G. barbadense ; Region J (Uttar Pradesh): Ficus species. doi:10.1371/journal.pone.0013674.g001 

Contexts in source publication

Context 1

... our findings in ITS sequences of Berberis indicate none of the tested three genera have paralogs of ITS. Our results, especially in Ficus and Gossypium suggest that ITS holds a good promise as a candidate barcode locus, as also reported earlier [51]. The low level of barcoding success observed in Berberis is not uncommon in plants. Similar difficulties have earlier been reported in Aspalathus [21], Crocus [52] Solanum sect. Petota [53] and Carex [44]. Even in the well studied taxonomic group, barley ( Hordeum L.), Seberg et al. (2009) reported recognition of less than 50% species using matK and rpoC1 [52]. In the morphologically distinct species of the Galapagos sun flower tree, Scalesia Arn, (Asteraceae), no variation was found in plastid loci and almost none in nuclear loci [52]. In our study, the morpho-taxonomic parameters of the selected species were considered for species delimitation. In Berberis the phenogram derived from this matrix showed that morphologically the species are well delineated whereas none of the four loci tested could distinguish species of Indian Berberis , except B. pachyacantha . This prompted us to check the species relationship in Berberis and evaluate genetic basis for the delimitation of species using amplified fragment length polymorphism (AFLP). Though the species recovery increased using AFLP method, most of the species remained unresolved. None of the Ahrendt’s (1961) sections or subsections in Berberis was recognized with the exception of B. pachyacantha , which belongs to section ‘vulgaris’. Only one species of section ‘vulgaris’ was included in the present study. It remains to be seen if the species would be resolved when other species of the section are included. Kim et al . (2004), in analyzing the ITS phylogeny of Berberis species including a few species from India ( B. edgeworthiana, B. insignis, B. coriaria, B. hookeri ) could not recognize the sections and subsections of Berberis proposed by Ahrendt (1961). That the species could not be fully resolved using AFLP has been noted before for complexes containing closely related or hybridizing taxa [54–57]. There are several reports on occurrence of hybridization in the genus Berberis [58–60]. The lack of resolution of most of the species e.g. B. asiatica, B. glaucocarpa, B. lycium of the section Asiaticae, B. chitria, B. aristata, B. tinctoria, and B. wightiana of the section Tinctoriae, B. angulosa of the section Angulosae and B. insignis of the section Wallichianae with AFLP data as well as ITS sequences indicates a probable hybridization in these species. This is further corrobo- rated by the fact that in AFLP analysis there was a tendency of the species of Berberis to cluster according to their geographic location rather than to species identity. This indicates that, species discrimination seems possible with morphological characters, but reproductive isolation appears to be weak in Berberis and in many cases probably only affected by geographical barriers. These findings are consistent with either non-monophyletic or reticulate evolution of these species. Kim et al (2004) [25] using ITS sequences of 79 taxa of Berberis representing four major groups including Septentrionales and 22 sections in the genus showed that these traditional geographical groups are monophyletic. Therefore, the latter hypothesis is preferred because of evidence of hybridization and relatively young age (5.33-0.01 Ma) of Indian Berberis (Pleistocene record of Kashmir, India) [61]. The taxonomic problems described here for Indian Berberis are not unique. Similar problems were reported by Spooner (2009) for Solanum sect. Petota [53]. Hawkes (1990) reported 232 species of sect. Petota [62] but Spooner and Salas (2006) reduced it to 190 [63] and more recently, Spooner has converged these to about 110 species [53]. Harlan and de Wet (1971) showed differences in the number of species recognized by different taxonomists in crops, e.g. 100 to 200 in wild relatives of potatoes, 2 to 24 in wheat and 1 to 31 in sorghum [64]. These are some examples where taxonomic disputes still remain unresolved. Plant DNA barcoding in these cases may be problematic and contribute to complexities in search of universal loci for plant DNA barcode. Barcoding in plant genera like Berberis with possible occurrence of natural hybridization and gene introgression may be quite challenging. The morphological, geographical and genomic diversity study in Indian species of Berberis indicates probable reticulate nature of the species. Our results with Ficus and Gossypium suggest that ITS and trnH-psbA are good candidates for plant DNA barcoding and the matK and rbcL , the standard barcode loci for plant barcoding do not work in all the tested species of these three genera. One hundred and sixty four accessions representing 16 species of Berberis were collected from four different geographical regions e.g. Eastern Himalayas, Western Himalayas, Central India and Sothern India (Figure 1). Out of these, 3–5 representative accessions from each species were evaluated by morpho- taxonomic analysis. All morphological characters were weighted equally. Multistate characters were unordered. The character matrix thus developed (Table S3) was used to generate the phenograms (Figure S1) with PAUP*4.0b [65]. The parsimony trees were generated using bootstrap analyses with 10,000 replicates. Bootstrap searches were heuristic with simple addition of taxa, TBR branch-swapping and MulTrees turned off. We sampled at least two species from each region except central India where only one species occurs. Thirty three accessions of 11 species of Ficus and 51 accessions of 4 species of cultivated Gossypium were collected from different parts of India. Multiple accessions were included for each species. Specimen vouchers were deposited in the Herbarium of National Botanical Research Institute, India (LWG). Accession numbers including specimen collection locations are given in Table S10. Genomic DNA was extracted from either fresh or silica gel dried leaf materials using DNeasy Plant Mini Kit (Qiagen, Germany) according to manufacturer’s instructions. PCR ampli- fication was performed in 50- m l reaction mixtures containing approximately 50–75 ng genomic DNA templates, 1.5 mM MgCl 2 , 0.2 mM of each dNTP, 1 m M of each primer, 0.1 mg BSA/ml and 1 unit Taq DNA polymerase. The thermocycler programme was 94 C for 1 min (1 cycle), 94 C for 40 sec, 48 C– 52 C (depending upon primer sets used) for 35 cycles, 72 C for 40 sec and 72 u C for 5 min (1 cycle). The primers, matK 2.1a and matK 3.2r reported by Plant Working Group failed to give PCR amplification in Berberis even after changing some PCR conditions including addition of DMSO. A modified matK forward primer, matK- NBRI (here after referred as matK ) was designed after aligning the matK sequences of genera closely related to Berberis e.g. Nandina , Ranzania , Mahonia . The matK along with the reported matK 3.2r primer was able to successfully amplify in Berberis , Ficus and Gossypium . For primer sequences and references see Table S11. The PCR products were cleaned by Qiaquick H PCR Purification kit (Qiagen, Germany). In a few cases, where multiple bands appeared, these were gel extracted and sequenced. Sequencing was carried out bidirectionally using automated capillary sequenc- er, ABI3730XL DNA analyzer (Applied Biosystems, UK). Pairwise alignments were made by using the sequences obtained from forward and reverse primers. Sequences which covered more than 70% overlap between forward and reverse sequences were considered (except a few sequence of matK where coverage was less than 50%). A minimum average QV of 30 was considered as quality sequences. DNA sequences were edited manually by visual inspection of the electropherograms of both end sequences using Sequencher 4.1.4. The GenBank accession numbers for the sequences are given in Table S10. Each nrITS sequence was searched in nucleotide data base using BLAST, to confirm its plant origin rather than from a possible fungal contamination of the sample. In all cases the best match retrieved the plant species, either as the same plant species sequence as query sequences or as the nearest plant species (e.g. in most cases sequences of Indian Berberis species were not available in data base and B. thumbergii showed the best match). Secondly, we looked for the presence of the characteristic conserved motif in the 5.8S rRNA gene of angiosperm plant ITS sequences [66]. The characteristic motif (5 9 - GAATTGCAGAAT C C-3 9 ) was found in all the ITS sequences where as the variant of the motif generally found in fungi (5 9 -GAATTGCAGAAT T C-3 9 ) was not found in any of the sequences. To establish the species diversity of Indian species of Berberis at the genome level, we employed amplified fragment length ploymorph- ism analysis in 55 accessions of 13 species of Berberis . We used the same DNA samples as used in barcode analysis except in few cases where DNA quality was poor for AFLP assays. AFLP protocol was followed as described in user manual (AFLP Plant Mapping Kit, Applied Biosystem, and USA) with minor modifications. Briefly, 0.5–1.0 m g genomic DNA was digested with 10 U EcoR I and 10 U Mse I in a 20 m L reaction and incubated at 37 u C for 5 h. Following 15 min heat inactivation of enzymes, 20 m L of ligation master mix containing 75 pmol each Mse I and EcoR I adapters with 20 U T4 DNA ligase in 1X T4 DNA ligase buffer was added and incubated overnight at 16 u C. The digestion- ligation mixture was diluted with 160 m L sterile water. Pre-selective amplification was performed by using a tri-selective nucleotide ( + 3) at the 3 9 . Ten primer combinations were employed to detect polymorphism among different genotypes: EcoR I + AAG/ Mse I + CAA, EcoR I + AAG/ Mse I + CAT, EcoR I + AAC/ Mse I + CTT, EcoR I + AAC/ Mse I + CTG, EcoR I + AGG/ Mse I + CAA EcoR I + AGG/ Mse I + CTG, EcoR I + ACC/ Mse I + ...

Context 2

... barcoding is the use of short DNA sequences for species identification. Since its inception as an approach for large scale species identification [1–3], several studies have reported the application of COI in a wide range of animal taxa [1,4–6]. However, the attempts to identify a single locus for barcoding in plants have largely been unsuccessful [7,8]. There are growing evidences which suggest the need of deploying more than one locus for barcoding in plants [9–13]. The following regions have been suggested for plant DNA barcoding: ITS (the internal transcribed spacer region of the nuclear ribosomal genes), rbcL and psbA-trnH [10]; ITS and psbA-trnH [14]; rbcL [15]; rpoC1 , rpoB and matK or rpoC1 , matK and psbA-trnH [9]; matK , atpF-atpH and psbA- trnH or matK and psbK-psbI [7], psbA-trnH and rbcL [11]; psbA-trnH [16]; trnLUAA [13]; matK and psbA-trnH [12], matK and rbcL [17] matK , rbcL and trnH-psbA [18] and ITS2 [19,20]. In most of the recent plant barcoding studies, the coding regions of matK and rbcL and the non-coding plastid intergenic spacer of trnH-psbA have been suggested as prime candidates for barcoding [17,18]. Following the first suggestion by Kress et al . (2005) [14], several subsequent reports projected trnH-psbA as a strong candidate for plant barcoding [9–12,16]. However, Consortium for the Barcoding of Life (CBOL) disregarded trnH-psbA as it does not consistently provide bidirectional unambiguous sequencing reads [17]. Erstwhile studies have focused predominantly on plastid regions for barcoding. Chase et al. (2005) [10] and Kress et al . (2005) [14] recovered highest mean percentage sequence divergence (2.81 and 5.7% respectively) for nrITS region for plant barcoding. However, the use of ITS region as barcode locus has often been considered unfavorable because of the presence of paralogs in several plant taxa. Yet, in other studies, ITS has been successfully used as barcode locus [10,14,21]. More recently, ITS2 has been projected as an important plant barcode locus [19,20]. We examined four DNA barcoding loci (one nuclear- nrITS and three plastid loci- trnH-psbA , rbcL and matK ) in 16 species of Berberis L. (Berberidaceae) from India. For validation of the techniques, we tested these loci in selected species of two other genera, Ficus L. (Moraceae), comprises keystone species in tropical rain forest ecosystems and Gossypium L. (Malvaceae), a pan tropical genus including the commercial cotton plants cultivated widely on the tropical and subtropical regions throughout the world. The genus Berberis comprises of about 500 species [22,23]. Based on phytogeographic distribution Schneider (1905) divided the species of Berberis into two groups, Septentrionales and Australes, [23]. The group Septentrionales (Old World) consists of ca 300 species occurring mainly in Eurasia but extending to North America (two species) and North Africa (four species). The group Australes (New World) contains about 200 species with most of them distributed in South America and a few in Middle America [22,24]. These geographical groups are supported by grouping based on morphological characters [23]. A recent molecular study based on the internal transcribed spacer (ITS) sequences, supports the treatment of these two groups within Berberis [25]. In subdividing these groups, Ahrendt (1961) accepted the schemes of Schneider (1905) and recognized 29 sections with some modifications. Ahrendt (1961) further subdivided these sections into numerous subsections. In India, Berberis is represented by 55 species [26], which according to Ahrendt (1961) belong to 8 sections and 7 subsections [22]. The majority of the species are centered in the Himalayan region extending from Pakistan to Western China and to Central and Southern China. The 16 species selected for the present study represent wide geographical distribution across India and belong to 5 sections and subsections (Table S1). The detailed morpho-taxonomic characters of these sections and subsections were described by Ahrendt (1961) [22]. The geographical locations of the selected species are indicated in Figure 1. Some species of Berberis are known for high medicinal value because of the presence of alkaloids, principally ‘berberine’ [27], which show activity against cholera, diarrhea, amoebiasis, malaria and leishmaniasis [28]. Some species of Berberis give a high value wood dye while some others provide edible berries. The taxonomy of Berberis is somewhat uncertain [27]. For example, Orisi (1984) considered 17 Patagonian species in Argentina [29], where as Landrum (1999) synonymized several of these species and recognized only nine [24]. Complexity in Berberis taxonomy has been attributed to hybridization and some degree of introgression in transitional zones which produce intermediate forms [27]. Out of 500 species of Berberis , Ahrendt (1961) recorded as many as 70 species and infraspecific taxa with a suspected hybrid origin [22]. Based on molecular phylogeny of Berberis species and previous taxonomic treatment of Landrum (1999) [24], Kim et al (2004) questioned the status of most sections and subsections in this genus [25]. Although taxonomic revision of Indian Berberis , based on morpho-taxonomic parameters has been proposed [26], no attempt has been made to establish the species delimitation in Indian Berberis using molecular approaches. In this study, we examined the application of standard plant DNA barcode loci in this challenging group considering Indian species of Berberis . In order to test the universality of the standard barcode loci we applied these loci to Ficus and Gossypium . Although, taxonomically the genus Ficus is considered to be quite difficult, it forms distinct natural groups in which many species are very common and conspicuous and can easily be identified even with sterile specimens [30]. Ficus consists of about 1000 species of woody trees, shrubs, vines, epiphytes and hemi- epiphytes [31], occurring in most tropical and subtropical forests thought the world. In India, the genus comprises about 100 species which are distributed throughout the country with maximum diversity in Western Ghats and North Eastern India [32,33]. The candidate species of Ficus selected in the present study belong to three subgenera and six sections (Table S2) [31]. Earlier phylogenetic study using ITS sequences indicates that three subgenera of Ficus studied here are monophyletic [34]. The four species of Gossypium considered here are well studied both taxonomically and at molecular diversity level elsewhere [35,36]. morphometric analysis, 3–5 representative accessions from each species of Berberis were studied, as described in material and methods. The character matrix thus developed is shown in Table S3. A consensus parsimony tree was developed using the matrix. The cladogram showed clear segregation of the accessions into distinct species clades (Figure S1). For DNA barcoding, we examined 129 DNA sequences for ITS (GenBank accession numbers GU934610-GU934738), 78 for matK (GenBank accession numbers GU934739-GU934816), 97 for rbcL (GenBank accession numbers GU934817-GU934913), and 83 for trnH-psbA (GenBank accession numbers GU934914-GU934996) representing 16 species of Indian Berberis . PCRs were generally successful with all the four loci. The maximum success in PCR was observed with rbcL and ITS (97%), followed by trnH-psbA (92%) and matK (76%). Sequencing success ranged from 95% for rbcL to 85% for matK (Table 1). The alignment of sequences was straight forward except in case of trnH-psbA , due to high variation in sequence length. The mean sequence lengths of ITS (ITS1 + 5.8- S + ITS2), matK, rbcL and trnH-psbA were 602.2, 488.1, 479.0 and 410.0 bp, respectively. The corresponding percentage frequencies of parsimony informative characters were 4.9, 1.8, 0.6 and 2.6 and the percentage variable sites were 6.1, 3.0, 0.8, and 3.2, respectively (Table 1). The genetic divergence within and between species was calculated. The highest mean intraspecific divergence was obtained in ITS and the lowest mean inter- and intraspecific divergence was obtained in case of rbcL (Table 1). ANOVA test showed ITS and trnH-psbA as the most divergent barcode loci at interspecific level followed by matK and rbcL (Table S4 A). At intraspecific level, rbcL was the least and ITS was the most divergent locus (Table S4 B). In multilocus analysis, ITS + trnH-psbA provided the highest divergence at interspecific level as compared to the two, three and four loci combinations (Table S5 A). At intra specific level, there was no significant difference in the sequence divergence between all combinations (Table S5 B). To evaluate the barcoding gap we looked at the minimum inter- and maximum intraspecific divergences for each locus. No distinct barcoding gap was noticed in any of the four loci (Figure 2, Table S6). To detect paralogs of ITS, if any, we cloned the PCR product from eight randomly selected species and sequenced at least eight clones from each species. None of the species showed the presence of multiple copies as ...