Use of ITS2 Region as the Universal DNA Barcode for Plants and Animals

1As one of the most important markers in molecular systematics and evolution –, ITS2 shows significant sequence variability at the species level or lower. The availability of its structural information permits analysis at higher taxonomic level,, –, which provides additional information for improving accuracy and robustness in the reconstruction of phylogenetic trees. Furthermore, ITS2 is potentially useful as a standard DNA barcode to identify medicinal plants – and as a barcode to identify animals –. ITS2 is regarded as one of the candidate DNA barcodes because of its valuable characteristics, including the availability of conserved regions for designing universal primers, the ease of its amplification, and enough variability to distinguish even closely related species.

2Since Hebert first proposed the use of the cytochrome c oxidase subunit 1 (CO1) as a barcode to identify animals, DNA barcoding has attracted worldwide attention,. Many loci have been proposed as plant barcodes, including ITS,, rbcL,, psbA-trnH,,, and matK –. Most recently, the Plant Working Group of the Consortium for the Barcode of Life recommended a two-locus combination of rbcL + matK as a plant barcode. However, some researchers have suggested that DNA barcodes based on uniparentally inherited markers can never reflect the complexity that exists in nature. In addition, nuclear genes can provide more information than barcoding based on organellar DNA, which is inherited from only one parent.

3Although ITS2 shows a great potential as a barcode to identify plants and animals, an extensive evaluation based on a comprehensive sample set is lacking. To validate the potential of using the ITS2 region to identify closely related species of plants and animals, we analyzed 50,790 plant and 12,221 animal ITS2 sequences (Table S1) available in a public database. The results support the conclusion that the ITS2 region can be used as an effective barcode for the identification of plant species and as a complementary locus to CO1 for identifying animals.

4For plants, the lengths of ITS2 sequences from dicotyledons and mosses were distributed between 100 and 700 bp, and the lengths of ITS2 sequences from monocotyledons, gymnosperms, and ferns were distributed between 100 and 480 bp. The average lengths of ITS2 sequences for dicotyledons, monocotyledons, gymnosperms, ferns, and mosses were 221, 236, 240, 224, and 260 bp, respectively. For animals, the ITS2 sequence lengths ranged from 100 to 1,209 bp (mainly dispersed between 195 and 510 bp), with an average of 306 bp. The GC contents of the ITS2 sequences of the dicotyledons, monocotyledons, gymnosperms, ferns, mosses, and animals were calculated, and the averages were 59.4%, 61.3%, 62.9%, 55.5%, 64.7%, and 48.3%, respectively. The average and distributions of ITS2 sequence lengths, as well as the GC contents of the six taxa, are shown in Figure 1 and Figure 2, respectively.

5Inter-specific divergence was assessed by three parameters: average inter-specific distance, average theta prime, and smallest inter-specific distance,,. In contrast, intra-specific variation was evaluated by three additional parameters: average intra-specific difference, theta (θ), and average coalescent depth,. The inter-specific genetic distances between congeneric species of plants and animals were greater than the intra-specific variations of the ITS2 regions of the different taxa (Table 1).

6BLAST1 method based on similarity was used to evaluate the identification capacity of the ITS2 region. At the genus level, the use of the ITS2 region had a >97% success rate for the identification of plants and animals (Table 2). At the species level, ITS2 sequences correctly identified 91.9% of 12,221 animal samples, whereas the success rates of using ITS2 sequences for the identification of 34,676 dicotyledons, 11,598 monocotyledons, 946 gymnosperms, 42 ferns, and 3,528 mosses were 76.1%, 74.2%, 67.1%, 88.1%, and 77.4% at the species levels, respectively (Table 2).

7In addition, we studied the possibility of using ITS2 sequences to identify closely related species in different families. First, we studied 34 dicotyledon families, each having more than 10 genera. For 13 families, the rates of successful identification were more than 80%; success rates for identification fell below 70% in only seven families (Fig. 3). Of the 14 monocotyledon families that each had more than 5 genera, identification success rates were lower than 70% in only two families (Fig. 3). The success rates for using the ITS2 region to identify species in families with more than 10 genera of mosses and gymnosperms and all families of ferns are also shown in Fig. 3. The success rates for using the ITS2 region to identify species in families with less than 10 genera of dicotyledons, mosses, gymnosperms, and with less than 5 genera of monocotyledons are listed in Table S2. Compared to the success rates when identifying species in plants, the success rates for identifying species in the nine phyla of animals studied were much higher (more than 90%), except for Cnidaria (77.1%) (Fig. 3).

8Second, we focused on the ability of ITS2 to discriminate amongst the lower taxa. Of the 35 dicotyledon genera that each had more than 80 species, identification success rates were more than 80% for 12 genera. The success rates for identification of species within the Draba and Rhododendron genera were the two lowest at 27.2% and 21.9%, respectively (Table 3). The success rates for the identification of species within the dicotyledon genera with less than 80 species can be found in Table S3. Of the 42 monocotyledon genera with more than 30 species, identification success rates were greater than 80% in 13 genera. The success rates for identification of species within the Kniphofia, Ophrys, and Diuris genera were the three lowest at 16.2%, 22.7%. and 31.1%, respectively (Table 4). The success rates for the identification of species within genera with less than 30 species of monocotyledons and of species from different genera of gynosperms, ferns, and mosses can be found in Table S3. All 28 animal genera with more than 20 species each had a species identification success rates greater than 80%, except for the genus Calligrapha and Dolichopus. The success rates for the identification of species within the genus Calligrapha and Dolichopus were the lowest, which were at 73.3% and 73.8%, respectively (Table 5). The success rates for the identification of genera with less than 20 species of animals are presented in Table S3.

9To identify the species, we focused not only on the divergence of primary sequences of ITS2, but also on the use of variations in the secondary structures of ITS2. The secondary structures and alignments of primary sequences of ITS2 were reconstructed in four different species from the same genus, four species from different genera of the same family, and four species from the different families of dicotyledons, monocotyledons, and animals. These are shown in Figures 4, S1, S2, S3, S4, and S5. All of the secondary structures in these species have four similar helices: Helix I, II, III, and IV (Figs. 4, S2 and S4),,. Helix III is relatively longer than the others. At the different taxa levels of dicotyledons, monocotyledons, and animals, the secondary structures show different levels of similarity, which result from the differences in the primary sequences of these species. Thus, the species of dicotyledons, monocotyledons, and animals could be identified by their secondary structure. And, the secondary structure of the ITS2 region could be considered as a molecular morphological characteristic.

10Although ITS2 sequences are advantageous for identification purposes, one of the concerns for accepting the ITS2 region as a barcode is the potential contamination of fungal sequences. We checked the studied ITS2 sequences of plants and animals using the Hidden Markov model (HMM) for fungal ITS2 annotation, in addition to conducting BLAST searches of the fungal nrITS database. For the plants, 139 and 136 ITS2 sequences may have been fungal sequences, as determined by BLAST and HMM, respectively. Less than 10 ITS2 sequences of gynosperms, ferns, and mosses may have been fungal sequences, as determined by the BLAST and HMM. There were 37 and 32 dicotyledon ITS2 sequences, as well as 30 and 27 animal ITS2 sequences that may have been fungal sequences as determined by the BLAST and HMM, respectively. There were 86 monocotyledon ITS2 sequences that may have been fungal sequences (Table S4).

11Finally, we developed a web application at http://its2-plantidit.dnsalias.org to allow researchers to further test the usefulness of ITS2 for species identification across plant and animal kingdoms. Four different modules have been implemented at the time of this writing. The first module, “View,” provides a gene-card like summary regarding the ITS2 reference sequence for a particular species. The users perform a query with a taxonomy ID used in NCBI's taxonomy browser. The module then displays all sequences associated with the taxonomy ID, as well as the reference barcode sequences for the ITS2 region of this species. The second module, “Retrieve,” allows the user to retrieve various segments of the ITS2 region, which can be divided into the 5.8S gene segment, the ITS2 core region, and the 28S gene segment. The sequences for these different regions can then be used to build various models, such as HMMs. The third module, “Annotate,” allows users to annotate the 5.8S gene segment, the ITS2 core region, and the 28S gene segment for their own sequences. The users need to provide the alignment of multiple sequences for the 5.8S gene and the 28S gene segments. The module then builds HMMs with these fragments, and uses HMM to query the input sequences to define the boundaries of the various fragments. The users can choose to export various segments individually or by batch. The last module, “Identify,” performs a BLAST search on a query sequence against our internal ITS2 reference barcode sequence database. Species identification is based on the assumption that the ITS2 sequence for this species is included in the reference database. In such a case, if the top hit represents a unique species, this species should represent the species to which the sample belongs. In contrast, if the top hit includes more than one unique species, the ITS2 sequence cannot be used to identify the sample, and additional DNA barcodes are needed to resolve the identity of the sample. If the reference database does not contain the ITS2 sequence of the species under investigation, the identification is more complicated, and has been stated elsewhere.

12In summary, a comprehensive reference database is critical for species identification, which is the reason this database was constructed.

13An ideal barcode should possess sufficient variation among the sequences to discriminate species; however, it also needs to be sufficiently conserved so that there is less variability within species than between species,. Chen et al. (2010) compared seven candidate DNA barcodes (psbA-trnH, matK, rbcL, rpoC1, ycf5, ITS2, and ITS) from medicinal plant species and proposed that ITS2 can be potentially used as a standard DNA barcode to identify medicinal plants. The ITS2 region has also been used as a barcode to identify spider mites, Sycophila, and Fasciola. In the present study, we extended this analysis across all plants and animals, and assessed the species discrimination capacity of ITS2 sequences for 50,790 plant and 12,221 animal sequences (Table S1). The success rates for identification of plants and animals were more than 97% and 74% at the genus and species level (Table 2), respectively, except for gymnosperms, which had a 67.1% success rate at the species level. In addition, the ITS2 region had a high success rate for discriminating between closely related species in plants and animals (Fig. 3, Tables 3, 4, 5, S2, and S3). The sequence length of ITS2 is short (Fig. 1), which satisfies the requirements for PCR amplification and sequencing. Finally, the secondary structures of ITS2 are conserved and can provide useful biological information for alignment,,; thus, it can be considered as molecular morphological characteristics for species identification.

14The ITS2 sequence lengths of plants and animals were mainly distributed in the 195–510 bp range. The identification of plant and animal voucher species and other collections using DNA barcoding techniques is one of the main tasks in natural museums and research institutes. The length of the ITS2 region is sufficiently short to allow amplification of even degraded DNA. In addition, the intra-specific variations in plants and animals are lower than the inter-specific divergences. But the overlap of genetic variation without barcoding gaps significantly increases when the number of closely related species is increased.

15Hebert et al. found that more than 98% of 13,320 congeneric species pairs, including representatives from 11 phyla, have sufficient sequence divergence to ensure easy identification. However, the sequence divergence of COI for some animal species, such as cnidarians and the West Palaearctic Pandasyopthalmus taxa, is relatively low, and even invariant. In addition, mtDNA is maternally inherited; other resources of data should be considered, such as nuclear DNA, morphology, or ecology. The success rate of using ITS2 for identification of animals is 91.7% at the species level based on testing of a comprehensive sample set, and the identification efficiency of ITS2 for sequences in cnidarians is more than 77%. ITS2 sequences have a relatively high divergence rate; thus, it can be used as a complementary locus to CO1 for identification of animal species.

16Recently, ITS2 region has been found to vary in primary sequences and secondary structures in a way that correlates highly with taxonomic classification. Several researchers have already demonstrated the potential for using ITS2 for taxonomic classification and phylogenetic reconstruction at both the genus and species levels for eukaryotes, including animals, plants, and fungi,,,,,. The ITS2 region of nuclear DNA provides a powerful tool because of sufficient variation in primary sequences and secondary structures. Analysis of the secondary structures formed by the RNA transcript as it folds back upon itself at transcription has been less commonly conducted; however, it has been proven extremely useful in aiding proper sequence alignment,. Schultz and Wolf described the utilization of ITS2′s primary sequence and secondary structure information, together with an ITS2-specific scoring matrix and an ITS2-specific substitution model, based on tools such as 4SALE, the CBCAnalyzer, and ProfDistS.

17Among of 50,790 ITS2 sequences of plants and 12,221 ITS2 sequences of animals,139 and 30 sequences, respectively, could be fungal sequences. Thus, the frequency is less than 0.3% in both plants and animals. This result is similar to that of Chen et al.. The frequency of suspected fungal sequences in monocotyledon ITS2 sequences is twice as high as in dicotyledons, which may be due to the presence of endophytic fungi in most monocotyledon species. Although the rate of fungal contamination is very low, we should pay more attention to the data from the public database.

18There are multiple copies of ITS (containing ITS1 and ITS2) in plants and animals. Although different copies of ITS exist, which may result in misleading phylogenetic inferences, there remain several advantages for its widespread use, such as the levels of variations and multicopy structure facilitating PCR amplification, even from herbarium specimens.

19In conclusion, we believe that the ITS2 locus can be used as a barcode for authenticating plant species, as well as a complementary locus to CO1 for identifying animal species. The sequences of the universal primers and the amplification conditions for obtaining the ITS2 sequences of plants and animals can be found in Table S5, as well as in the ITS2 application web. There were limited ITS2 sequences of ferns and vertebrates in the GenBank; therefore, the success rates for ITS2 to identify them need further investigation.

20All ITS2 sequences of dicotyledons, monocotyledons, gymnosperms, mosses, ferns and animals were downloaded from GenBank on June 28, 2010 by searching using the keywords “internal transcribed spacer 2,” which retrieved 160,295 sequences. These sequences were used to construct an analysis dataset. The raw data were annotated and trimmed using ITS2 annotation tools based on HMM. Two conserved regions of the 5.8S and 28S gene for plants and animals, respectively, were used to delimit the ITS2 region. A maximum E-value of 1.0 was used. The trimmed sequences were edited manually. The sequences with less than 100 bp length, or with ambiguous bases with more than two “Ns”, or with unnamed species (such as those with spp. and aff. in the species name) were excluded. The selected ITS2 sequences were filtered then with a HMM-based annotation and fungal nrITS database (http://www.emerencia.org/fungalitspipeline.html) using the BLAST tool. The ITS2 sequences belonging to a genus that contains only one species were excluded from the analysis. Finally, a reference database was constructed. The detailed sequences information can be found in Table S6. The workflow is shown in Figure 5.

21The GC content and sequence length were calculated for all of the ITS2 sequences of dicotyledons, monocotyledons, gymnosperms, ferns, mosses, and animals. The intra- and inter-specific divergences were calculated based on different taxa. Sequences were aligned using Clustal W, and Kimura 2-parameter (K2P) distances were calculated using PAUP4b10 (Florida State University, USA). The intra-specific variations and inter-specific divergences of congeneric species in the dicotyledons, monocotyledons, gymnosperms, ferns, mosses, and animals were calculated using a K2P distance matrix, as described previously,,.

22All ITS2 sequences of plants and animals were used as query sequences. Query sequences were divided into the following: dicotyledon, monocotyledon, gymnosperm, fern, moss, and animal. BLAST1, which was implemented using the BLAST program (Version 2.2.17), was used to search for the reference database for each query sequence.

23To identify the effect of primary sequence divergences on secondary structure, ITS2 sequences with different sequence divergence (∼1%, ∼5%, ∼10%) were subjected to the secondary structure prediction in a genus that had three other species and three other genera in the same family. Paphiopedilum (Orchidaceae) of monocotyledons, Acaena (Rosaceae) of dicotyledons, and Heterodera (Ceratopogonidae) of animals were used to construct secondary structures using tools from the ITS2 database.

24We developed a web application (http://its2-plantidit.dnsalias.org) to facilitate the utilization of the ITS2 sequence for various DNA barcoding studies. DNA sequences related to ITS2 regions were retrieved from GenBank, and were preprocessed to remove the flanking 5.8S and 28S rRNA gene sequences, as described in section Reference Database Construction. Sequences that belong to the same species, indicated by having the same taxonomy ID, were assembled using the program Phrap. The consensus sequence of the corresponding sequence clusters was considered as the average or reference sequence of the ITS2 region for the species, which can be retrieved from the application. The web application was built using the Catalyst web application framework (http://www.catalystframework.org/) for Perl language running in a Fedora 12 environment. This web application consists of four analytic modules at the time of the writing: View, Retrieve, Annotate, and Identify.