A constructive step towards selecting a DNA barcode for fungi

Abstract

Are organisms that cannot be assigned to species ‘DNA barcodable’? Hebert et al. (2003) define DNA barcoding as a mechanism for species identification based on a short standardized DNA sequence that corresponds to the same locus for all eukaryotes. When the concept of DNA barcoding was applied to eukaryotic organisms other than animals, doubts were voiced regarding whether the concept could be applied to organisms such as fungi where the alpha taxonomy is still largely unresolved. Arbuscular mycorrhizal (AM) fungi provide an even greater challenge than most fungi, as neither morphological nor biological or molecular phylogenetic species concepts can be applied easily. Ideally, for making the best possible selection, a DNA barcode region should be chosen based on predefined criteria, which are likely to include universality of primers, feasibility and species-level resolution. The selection should be based on screening numerous candidate genes, ideally entire genomes. For the final selection, the most promising candidate genes should then be tested on a wide range of taxa of the target group, including sister species (Consortium for the Barcode of Life, 2007). ‘… the poor man’s approach to data-driven selection of a DNA barcode.’ In this issue of New Phytologist, Stockinger and colleagues (pp. 461–474) present a pragmatic approach to the selection of a DNA barcode region for AM fungi. In their search for DNA barcode candidate regions, the authors’ primary criteria are indeed universality, feasibility and species resolution: the standard DNA barcode region, COI, is rejected because of the presence of extensive introns. As genomic data are not available for AM fungi (Martin et al., 2008), the candidate barcode regions are limited to the few accessible loci and the respective primer sets commonly used in AM studies. Molecular markers used in earlier studies (partial small subunit (SSU); internal transcribed spacer (ITS); partial large subunit (LSU)) are rejected because they either do not amplify all clades or are known not to discriminate among recognized taxa. Recently developed AM-specific universal primers now allow the amplification of a long continuous stretch of ribosomal DNA of c. 1500 bp, including part of the SSU, ITS and part of the LSU of all clades of AM fungi (Krüger et al., 2009). The taxonomic resolution of this long region was assessed using a set of sequences from six groups of closely related taxa representing more than one-third of the species available in culture. All taxa formed reciprocally monophyletic clades in neighbour-joining (NJ) trees with bootstrap support. The DNA extracts used for this analysis stemmed from starting material of unambiguous origin. Available larger data sets of partial sequences from the suggested barcode region, including environmental data, support the claim that it will be possible to assign sequences to named clades. However, the realities of undescribed and unculturable fungi in nature (and the current taxonomic impediment) require the inclusion of unnamed sequences in DNA barcoding reference databases of AM fungi and fungi in general, even though they do not meet the DNA barcoding standard of originating from vouchered specimens. Matching among environmental sequences may appear as a circular argument because taxon definition and taxon identification are achieved using the same diagnostic. For taxonomists working on enigmatic macroorganisms this may be an unusual perspective. But in microbiological practice, considering unnamed sequences is often the only way of getting a handle on otherwise unknown organisms. In summary, the paper of Stockinger et al. suggests that, yes, AM fungi can be DNA barcoded. If this is true for AM fungi, which seem to have the highest intraspecific variability among all fungal clades (Nilsson et al., 2008), then one might think that other fungi should also be ‘DNA barcodable’, given that species are normally easier to circumscribe in other fungal clades. With the relatively small number of DNA barcode candidate regions available for consideration, this paper presents the poor man’s approach to data-driven selection of a DNA barcode. A better example of how the selection of a DNA barcode locus could ideally look is given by the selection of the plant barcode regions (Hollingsworth et al., 2009), even though the result of the selection process is still somewhat controversial. What should be done for fungi in general? Standardization of the marker region is a mainstay of DNA barcoding and cannot be turned overboard without betraying the concept itself. In other words, we, the mycological community, should not want to escape the formal acceptance of one or more DNA barcoding loci. Following mixed initial results (summarized by Seifert, 2009), only a minority of mycologists is willing to test the applicability of COI for fungi. Fungal mitochondrial genomes tend to contain many introns. In spite of this, Santamaria et al. (2009) advocate the use of another mitochondrial gene, ND6, as a DNA barcode region for Ascomycota based on the analysis of available mitochondrial genomes. However, aligning the DNA sequence data of ND6 sequences retrievable from GenBank, it appears likely that this gene combines a disadvantage of COI (no universal primers) with another one of the ITS (alignment issues). Projects are underway searching for DNA barcode candidates through comparative genomics. A number of ‘new’ loci have already emerged from comparative genomics (i.e. Carbone & Kohn, 1999; Schmitt et al., 2009), which may have some potential as phylogenetic markers and possibly also as DNA barcode candidates. ‘Legacy barcode regions’ are the LSU D1/D2 for yeasts (Kurtzman & Robnett, 1998) and the ITS for ectomycorrhizal fungi (Kõljalg et al., 2005) and others. These loci have the advantage that they are easily amplifiable with universal primers and that the wealth of data available allows some predictions about their performance as barcodes. Depending on the perspective, the alignment, multicopy and intragenomic variation issues of the ITS, and to some degree also of the LSU, appear manageable (as demonstrated by Stockinger et al. for the AM fungi) or may constitute reasons for rejection. In addition, it is known that particularly in some groups of economically important fungi, such as molds, the ITS does not resolve all species (Seifert, 2009). For these fungi, LSU data are often not available, but as the gene is normally more conserved than the ITS, it appears unlikely that the LSU or a combined barcode will increase the degree of resolution. Even a data-driven approach, as adopted for plant DNA barcoding, cannot hide the fact that whatever part of the selection procedure is looked at, serving all possible purposes of DNA barcoding using the same criteria is unachievable. Depending on the selection criteria chosen for a fungal DNA barcode and the test panel, the rDNA regions referred to earlier may be selected or rejected. It is self-evident that different questions require different solutions. In a medical or phyto-sanitary context, for example, where the presence of a particular organism needs to be detected, DNA barcoding calls for markers and methods, leaving no margin for error. In such contexts, devising special DNA barcoding solutions may be dictated by necessity. A DNA barcoding concept with group-specific additional loci is likely to be the best possible compromise here. However, in environmental research, the foreseeable lack of a complete reference data set for some time to come will require a barcode with predictive power, even in the absence of a perfect match. High-level species resolution is desirable, but not of vital importance. In addition, having two loci instead of one does not necessarily improve the identification success for environmental sequences. Even in a world where environmental sequencing could easily be carried out for several loci, linking the resulting disparate data sets will require huge amounts of data and many processor hours, unless the reference database contains numerous matches with data for both loci. With a view to practicalities (the availability and universality of primers, ease of amplification) it appears very unlikely that a better candidate for a DNA barcode will be found in comparative genome searches. The contribution by Stockinger et al. is a big first step towards a formal suggestion of the ITS, and possibly the LSU, as a DNA barcode for fungi in general. A proposal to the Consortium for the Barcode of Life is in preparation, showing the performance of the ITS, or the ITS and the LSU (D1/D2), as a fungal barcode. Data from closely related taxa will be presented to show the species-level resolution of the data. As this will be obtained from published data it will not be possible to compare other loci across the entire fungal kingdom or to fully demonstrate the practical issues that distinguish these loci. In the absence of a formal decision on the DNA barcode for fungi, many mycologists have de facto adopted a system where the ITS or the LSU (i.e. in yeasts) are sequenced for any fungus entering their research. As pointed out before, a formal approval of the general practice of using nuclear ribosomal genes, particularly the ITS, would still be highly desirable for the fungal research community – or the finding of a spectacular new DNA barcode locus from genome comparisons. I thank Freek T. Bakker for useful discussions.