To design new ITS primers that amplify broad fungal taxa, we constructed databases of fungal ribosomal RNA gene sequences. Two local sequence databases were prepared by downloading sequences from the Nucleotide database of the National Center for Biotechnology Information (NCBI; http://www.ncbi.nlm.nih.gov/; accessed April 26, 2011). To construct a database of small subunit ribosomal RNA gene sequences, 134,928 sequences were downloaded using the taxonomic ID 4751 (fungi) and the query word ‘small subunit OR 18S’ and setting the range of sequence length from 200 to 5,000 bp. Likewise, to construct a database of 5.8S ribosomal RNA and large subunit RNA gene sequences, 26,616 sequences were downloaded using the taxonomic ID 4751 (fungi) and the query word ‘5.8S’ with the abovementioned range of sequence lengths. Note that we attempted to construct a specific database of large subunit ribosomal RNA gene sequences using the query word ‘large subunit OR 28S,’ but many downloaded sequences did not contain the 5′-end of the gene, which was necessary for designing reverse primers for the ITS2 region (Fig. 1). Since each database was composed of numerous sequences, which were difficult to align simultaneously, consensus sequences of closely related taxa were obtained using assembler programs. Assembly was first conducted by PCAP [47], which enabled parallelized rapid assembling, with a sequence identity setting of 92%. The contigs returned by PCAP were further assembled to generate consensus sequences of diverse fungal taxa using CAP3 [48], which facilitated accurate assembling, with a sequence identity setting of 80% and a max overhang percent length of 200 bp. Multiple alignments of the output consensus sequences were conducted using MAFFT v6.813b [49] with default options. High-coverage PCR primers targeting the ITS1 and ITS2 regions were designed at conserved nucleotide positions of the aligned consensus sequences.

The coverage of the designed ITS primers was evaluated and compared with those of existing fungal ITS primers developed in previous studies [40]–[43] based on in silico PCR using the ecoPCR v0.7.0 software [39]. Since the software requested sequence database files of focal taxonomic groups in the European Molecular Biology Laboratory (EMBL) database, fungal nucleotide sequence databases (STD, GSS, HTG, and WGS of release 107) were downloaded from the EMBL FTP site (ftp://ftp.ebi.ac.uk/pub/databases/embl/release/). Given that only a small proportion of sequences covered the entire part of the small subunit 5.8S and large subunit ribosomal RNA genes and the flanking ITS1 and ITS2 regions (Fig. 1), three subset databases covering either ITS1, ITS2, or the entire ITS region were constructed, as detailed in Bellemain et al. [39]. To create the first subset (subset 1), the ITS1 region was amplified in silico by extracting sequences matching both NS7 and ITS2 primers (Table 1, Fig. 1) from the initial fungal database. In this process, three mismatches to each primer were tolerated but sequences containing consecutive five or more ‘N’ were excluded, leaving a database containing 1,374 sequences. Likewise, subset 2 for the entire ITS region (ITS5-ITS4; 8,421 sequences) and subset 3 for the ITS2 region (ITS3-LR3; 3,217 sequences) were created.

Based on the three subset databases, the coverage of each ITS primer (Table 1) was evaluated in silico. In the in silico amplification, each target primer was paired with the forward or reverse primer that had been used in making the focal subset database, and from zero to three mismatches between each target primer and the template sequence (except mismatches at the two bases of the 3′-end of the primer) were allowed to simulate different PCR stringency conditions. Three mismatches between each nontarget primer and the template sequence were allowed in all cases to remove the effects of nontarget primers. Since each subset database included multiple sequences of the same fungal species, the coverage of PCR primers was calculated in terms of the proportion of “taxa” amplified in silico. In our subset databases, each sequence had its own taxonomic IDs in the NCBI taxonomy database, and the lowest taxonomic units (LTUs) of the subset/database sequences corresponded to species, subspecies, or varieties in the NCBI taxonomy. In total, 723 LTUs in subset 1, 3,474 in subset 2, and 1,869 in subset 3 were subjected to in silico PCR. Using the taxonomic information, the results of in silico PCR were evaluated independently for ascomycetes, basidiomycetes, and the remaining “non-Dikarya” fungi. Compared to ascomycete and basidiomycete fungi, “non-Dikarya” fungi represented only 7.6% (subset 1), 8.3% (subset 2) and 2.9% (subset 3) of the database sequences.

When selecting fungal PCR primers for DNA barcoding of environmental samples, both the coverage of primers across fungal taxa (species, subspecies, or varieties; see the above discussion of LTUs) but also the degree of “mismatches” to plant ITS sequences are important criteria for preventing the amplification of nontarget organismal sequences. Accordingly, we conducted in silico PCR of plant ITS sequences using fungal ITS primers (Table 1). From the EMBL FTP site (ftp://ftp.ebi.ac.uk/pub/databases/embl/release/), we downloaded plant nucleotide sequence databases (STD, GSS, HTG, and WGS of release 107) and used them to construct subsets 1–3, as detailed above. In silico PCR was conducted using the same settings as the fungal analysis; the results was are shown separately for Embryophyta (i.e., land plants) and the remaining “non-Embryophyta” plants.

On account of the coverage across fungal LTUs and the degree of mismatches to plant ITS sequences, candidate pairs of ITS primers for fungal DNA barcoding were selected for ITS1, ITS2, or the entire ITS region. The PCR fragment lengths amplified by these primer pairs were simulated using ecoPCR.

Optimal annealing temperatures for the selected fungal ITS primer pairs were explored in vitro. Three ascomycete and four basidiomycete fungal species collected on Yakushima Island, Kagoshima Prefecture, Japan, were subjected to the in vitro examination (A1–3 and B1–4: Table 2). DNA was extracted from fruiting body tissues of the fungal specimens using the CTAB method, as described elsewhere [32]. PCR was conducted using the buffer system of Ampdirect Plus (Shimadzu) with BIOTAQ HS DNA Polymerase (Bioline) under a temperature profile of 95°C for 10 min, followed by 35 cycles at 94°C for 20 s, 47°C, 50°C, 53°C or 56°C for 30 s, and 72°C for 20 s (40 s for the entire ITS region), followed by 72°C for 7 min. The concentration of MgCl₂, dNTPs, PCR primers and the template DNA in the reaction buffer were 1.5 mM, 200 µM, 0.5 µM and 1 ng/µl, respectively. Amplification of the DNA fragments was confirmed using the Flash Gel System for DNA (Lonza).

To evaluate the coverage of the selected fungal ITS primer pairs in vitro, an additional PCR assay was conducted under the optimal annealing temperature tested above using seven ascomycete and seven basidiomycete specimens (Table 2). All necessary permits of the sample collection were issued by Kyushu Regional Forest Office, Ministry of Agriculture, Forestry and Fisheries, Japan.

Within the first group of ITS primers for the small subunit ribosomal RNA gene (Table 1, Fig. 1), ITS9mun, ITS1-F_KYO2, and ITS5 realized more than 98% coverage across 723 fungal LTUs when one mismatch between a target primer and the template sequence was allowed. Commonly used primers, such as ITS1-F and ITS1, showed less than 90% coverage (Table 3, Fig. 2). Among the reverse primers for the 5.8S ribosomal RNA gene (Table 1, Fig. 1), ITS2_KYO1 and ITS2_KYO2 showed more than 98% coverage under one-mismatch conditions, while the coverage of commonly used ITS2 primers was relatively low (93.7%; Table 3, Fig. 2). Likewise, among the forward primers for the 5.8S ribosomal RNA gene (Table 1, Fig. 1), 58A2F, ITS3_KYO1, and ITS3_KYO2 showed more than 98% coverage under one-mismatch conditions, while the coverage of commonly used ITS3 primers was relatively low (95.1%; Table 3, Fig. 2). Among the reverse primers for the large subunit ribosomal RNA gene (Table 1, Fig. 1), the widely used ITS4 showed taxon coverage as high as the newly designed ITS4_KYO1, ITS4_KYO2, and ITS4_KYO3 under one-mismatch conditions, while other primers showed much lower coverage (Table 3, Fig. 2). Among primers for the large subunit ribosomal RNA gene, ITS4_KYO3 showed low coverage across “non-Dikarya” fungi (77.8%), while others covered more than 90% of “non-Dikarya” taxa. In silico PCR also indicated that primers with more than 97% coverage across fungal LTUs (i.e., ITS9mun, ITS1-F_KYO2, ITS5, ITS2_KYO1, ITS2_KYO2, 58A2F, ITS3_KYO1, ITS3_KYO2, ITS4, ITS4_KYO1, and ITS4_KYO2) are unlikely to exhibit a significant coverage bias toward ascomycetes or basidiomycetes (Table 3). Although a relatively small number of “non-Dikarya” sequences was available, a further inspection of sequences in the present data sets revealed that ITS1-F_KYO1, ITS1-F_KYO2, ITS2_KYO2, ITS3_KYO2, and ITS4 matched 99.2% (123/124), 96.0% (119/124), 98.8% (568/575), 99.0% (569/575), and 99.2% (387/390), respectively, of glomeromycetous sequences (one-mismatch conditions).

Among the above-mentioned high-coverage fungal primers, ITS1-F_KYO2 and ITS5 showed less than 10% coverage of plant LTUs under one-mismatch conditions, while other high-coverage primers matched the sequences of more than 96% of plant LTUs (Table 4). Thus, although ITS9mun showed the highest coverage across fungal LTUs (Table 3, Fig. 2), potential simultaneous amplification of plant sequences by this primer would be unfavorable in a variety of contexts, such as DNA barcoding of mycorrhizal fungi. Meanwhile, ITS1-F_KYO1, whose LTU coverage was relatively high (96.8%), showed a high specificity for fungi (Table 4). At the large subunit ribosomal RNA gene, ITS4_KYO3 matched the sequences of only 3.1% of land plant LTUs, while other primers were expected to amplify most land plant LTU sequences (Table 4). In terms of the 5.8S RNA gene, no primer with more than 95% fungal LTU coverage under one-mismatch conditions was expected to selectively amplify fungal sequences (Table 4). Accordingly, because of coverage across fungal LTUs and mismatches to plant ITS sequences, ITS1-F_KYO2–ITS2_KYO2, ITS3_KYO2–ITS4, and ITS1-F_KYO2–ITS4 primer pairs were tentatively selected as preferable for amplification of ITS1, ITS2, and the entire ITS regions, respectively, in the subsequent in vitro analyses (see below for further discussion). The average lengths of fragments amplified were 252.1 bp (SD = 53.2; N = 8,555 sequences) for the ITS1-F_KYO2–ITS2_KYO2 pair, 327.2 bp (SD = 40.1; N = 19,646) for the ITS3_KYO2–ITS4 pair, and 586.5 bp (SD = 73.9; N = 3,774) for the ITS1-F_KYO2–ITS4 pair (Fig. 3). Although the basidiomycete amplicons were the longest on average, considerable variation in amplicon lengths was observed within the ascomycete, basidiomycete, or “non-Dikarya” group (Fig. 3).

In vitro PCR analysis revealed that the optimal annealing temperature was 47°C for all the primer pairs examined (Fig. 4). The seven ascomycete and seven basidiomycete fungi examined were successfully PCR-amplified using the primer pairs (Fig. 5).

In fungal DNA barcoding founded on PCR-based next-generation sequencing of environmental samples such as soils, roots, and leaves, three important criteria exist regarding primer selection. First, as a general requirement for primers used in the exploration of biodiversity, fungal DNA barcoding primers should amplify sequences of broad fungal taxa to accurately describe the diversity and community structures of fungal species in environmental samples. As reported recently, several classic primers widely used in DNA barcoding of fungi (e.g., ITS1, ITS1-F, ITS2, and ITS3) do not cover a non-negligible proportion of fungal taxa (Table 3, Fig. 2), potentially causing taxonomic bias in the data [39]. Meanwhile, the newly designed primers ITS1-F_KYO2, ITS2_KYO2, and ITS3_KYO2 are expected to match the sequences of c. 99% of fungal LTUs, thereby enabling thorough and unbiased DNA barcoding of ascomycete and basidiomycete fungi. Results also showed that three classic primers, ITS9mun, ITS5, and ITS4, exhibited broad coverage of fungal LTUs (Table 3, Fig. 2). In particular, the coverage of ITS4 was as broad as that of any other primer, including our newly designed primers (Table 3, Fig. 2), suggesting that the classic primer is appropriate for fungal DNA barcoding as well as the newly-designed primers targeting the 5′-end of large subunit ribosomal RNA gene. Given the relatively small number of “non-Dikarya” sequences in the public databases, the coverage of the newly designed primers of fungal groups other than ascomycetes and basidiomycetes remains to be evaluated. Nonetheless, inspection of sequences in the present data sets revealed that concerning the “non-Dikarya” mycorrhizal fungal phylum Glomeromycota, ITS1-F_KYO1, ITS1-F_KYO2, ITS2_KYO2, ITS3_KYO2, and ITS4 matched with 96.0–99.2% of the sequences examined in silico. Thus, these primers are useful for not only DNA barcoding of ascomycetes and basidiomycetes, but also cataloging of glomeromycetes (and other “non-Dikarya” fungi) in the environment. In fact, pyrosequencing with the newly-designed primers successfully yielded glomeromycete ITS sequences in metagenomic analyses of root-tip and soil samples collected in Japanese temperate forests (Toju et al., unpubl.).

The second important criterion in selection of fungal barcoding primers is selective amplification of fungal sequences, i.e., the elimination of plant and other organismal sequences in PCR of mycorrhizal, leaf, or soil samples. Some fungi-specific primers such as ITS1-F and ITS4-B are widely used in DNA barcoding of various environmental samples (reviewed in [23]), although their low coverage of fungal taxa is expected to preclude comprehensive analyses of fungal diversity and community structures [39] (Table 3, Fig. 2). Basically, a trade-off exists between specificity for fungi (e.g., exclusion of plant taxa) and unbiased amplification within the fungal kingdom. Due to this trade-off, barcoding primer sets should be selected based on the purpose of the ecological/microbiological study. As substitutes for the classic fungi-specific primer ITS1-F, ITS1-F_KYO1 and ITS1-F_KYO2 were designed for selective amplification of fungal ITS sequences. In silico PCR showed that ITS1-F_KYO1 showed a high specificity for fungi that was comparable to that of the widely used primer ITS1-F; in contrast, ITS1-F_KYO2 might amplify sequences of diverse plant taxa under relaxed PCR conditions (e.g., low annealing temperatures). Thus, ITS1-F_KYO1 can be used as a highly specific primer for fungi at the slight expense of universality, while ITS1-F_KYO2 is a high-coverage primer, whose risk of amplifying plant sequences should be mitigated by adopting relatively stringent PCR conditions. Potential matches between these fungus-specific primers and other soil eukaryotic sequences (e.g., nematodes, springtails, and protozoa) warrants further extensive in silico and in vitro analyses.

The third criterion for the selection of barcoding primers is amplicon length. Next-generation sequencing is an invaluable technique for high-throughput DNA barcoding of fungi, being applied particularly to metagenomic analyses of mycorrhizal, phyllosphere, and soil fungi [9], [37], [50]. However, output read length is shorter in most next-generation sequencers than in Sanger sequencers (e.g., Applied Biosystems 3130 Genetic Analyzer), which has precluded DNA barcoding of fungi based on the entire ITS region. For example, since the average read length obtained from 454 pyrosequencing has until recently been 250 bp, pioneering studies on fungal metagenomics usually target the ITS1 region for DNA barcoding [9], [46], [51], [52]. The release of the GS FLX Titanium series (454 Life Sciences), which extended the length of output sequences to 400 bp, has allowed use of the potentially more informative ITS2 region for DNA barcoding [50], [53]. Given that the fungal ITS2 region can be highly variable at the species level [54], primers for amplification of this region should enable identification of fungal species and subsequent analyses of fungal communities [50], [53]. Therefore, DNA barcoding of the ITS2 region using ITS3_KYO2 and ITS4 with sequencers whose read length is less than 700 bp would be sufficiently informative in the context of ecological and microbiological studies (Fig. 3). Meanwhile, DNA barcoding targeting the entire ITS region using ITS1-F_KYO2 and ITS4 will increase the reliability of species identification if the upgraded FLX Titanium, whose read length is up to 1,000 bp, is used.

We herein developed high-coverage fungal ITS primers, which facilitate investigation of the remarkable diversity of fungi. Some of the newly designed primers enable selective amplification of fungal ITS sequences and are useful for investigating fungal diversity and community structures within mycorrhizal associations, leaf, soil, and other environmental samples that potentially contain plant tissues and detritus. There can be little doubt that DNA barcoding will dramatically enhance our knowledge of fungal diversity and communities during the next decade. To further fuel use of fungal DNA barcoding in ecological and microbiological studies, the potential bias caused by primers should be quantitatively evaluated based on next-generation sequencing [46].

Although the number of ITS sequences in public sequence databases is growing rapidly [22], [23], many studies use small or large subunit ribosomal RNA sequences as a DNA barcoding marker [34]–[38]. ITS sequences are hypervariable in most fungal clades and are used particularly in the context of identification at species or lower levels [25], while the considerable variation in the length of this region often impedes multiple sequence alignments and subsequent phylogenetic analyses. In contrast, small or large subunit sequences are more conserved across the fungal kingdom than ITS sequences, enabling, e.g., phylogenetic investigation of novel fungal clades. Thus, the simultaneous use of ITS and small/large subunit regions is desirable in future studies based on DNA barcoding after the read length of high-throughput sequencers is extended. To simultaneously conduct identification at species or lower levels and phylogenetic analyses, the ITS primers we report here should be paired with primers for amplification of small or large subunit ribosomal RNA sequences.