Using Barcode Similarity Groups to Organize Cortinarius Sequences

Transcription

1 Using s to Organize Cortinarius s by Emma Harrower A thesis submitted in conformity with the requirements for the degree of Master of Science Ecology and Evolutionary Biology University of Toronto Copyright by Emma Harrower 2010

2 Using s to Organize Cortinarius s Abstract Emma Harrower Master of Science Name of Graduate Ecology and Evolutionary Biology University of Toronto 2010 To improve fungal identification using a single DNA sequence, I introduce the Barcode Similarity (BSG) defined as a cluster of sequences that share greater than or equal to a threshold amount of genetic similarity with each other. As a test case, I created 393 BSGs from 2463 Cortinarius ITS sequences using a 94% similarity cut-off value in DOTUR. Some BSGs may contain multiple species. The BSG database was used to label environmental sequences, find misidentified or mislabeled sequences, and find potential cryptic species and novel species. Expert taxonomists will be needed to perform detailed morphological and phylogenetic studies to identify the individual species within each BSG. The main advantage of using BSGs is that it clusters together sequences using total genetic relatedness and does not rely on any taxonomy for identification. A website was created where the RDP Classifier is used to classify a query sequence into a BSG. ii

3 Acknowledgments I would like to thank Dr. Jean-Marc Moncalvo for his guidance, and Simona Margaritescu for her help in the lab. I would like to thank Dr. Nicholas Provart and Hardeep Nahal for their help with providing me with Perl scripts to use, for providing me an account on the BAR server and for helping set up a MySQL account for me. I would like to thank Bryn Dentinger for his understanding of the difficulties I faced and for his encouragement. I would like to thank my family and friends for their support and encouragement. The National Science and Engineering Research Council provided funding for this research. iii

4 Table of Contents ACKNOWLEDGMENTS...III TABLE OF CONTENTS... IV LIST OF ABBREVIATIONS... VI LIST OF FIGURES...VII LIST OF TABLES...VIII LIST OF APPENDICES... IX CHAPTER 1: INTRODUCTION...1 FUNGAL DIVERSITY AND ECOLOGICAL ROLES...1 SPECIES CONCEPTS IN FUNGI...2 DIFFICULTIES IN IDENTIFYING FUNGAL SPECIMENS...3 MOLECULAR IDENTIFICATION...4 GENBANK AND BLAST...4 THE DNA BARCODING INITIATIVE...5 WHAT IS A BARCODE SIMILARITY GROUP?...7 THE GENUS CORTINARIUS (AGARICALES, BASIDIOMYCOTINA)...8 MOLECULAR IDENTIFICATION OF CORTINARIUS...9 OBJECTIVES...10 CHAPTER 2: THE CREATION OF ITS-RDNA BARCODE SIMILARITY GROUPS FOR THE GENUS CORTINARIUS (AGARICALES, FUNGI)...12 INTRODUCTION...12 MATERIALS AND METHODS...15 Obtaining sequences...15 Creating s...15 Matching a query sequence to a BSG...16 Web access to the BSG database...17 RESULTS...18 ITS sequence diversity and taxonomic accuracy in public databases...18 Creating the BSG Database...19 Finding misidentified sequences...20 Finding novel and cryptic species...21 Collecting and labeling environmental sequences and matching them to voucher specimens...21 Metadata retrieved from BSGs...21 The BSG Classifier and website...22 iv

5 DISCUSSION...23 CHAPTER 3: GENERAL DISCUSSION AND PERSPECTIVES FOR FUTURE STUDIES...28 TAXONOMIC DELIMITATION AND CLASSIFICATION...28 Delimitation of Species...28 Classification tools...30 Fungal identification tools...32 BARCODE SIMILARITY GROUPS...33 How are BSGs an improvement on the current taxonomic database model?...33 How the 94% similarity value was chosen...33 Uses of BSGs...34 FUTURE DEVELOPMENTS...38 CONCLUSION...40 LITERATURE CITED...41 FIGURES...50 TABLES...58 APPENDICES...61 v

6 AFTOL: Assembling the Fungal Tree of Life BLAST: Basic Local Alignment Search Tool BOLD: Barcode of Life Database BSG: CBC: compensatory base pair change CO1: cytochrome c oxidase I DDBJ: DNA Databank of Japan List of Abbreviations DOTUR: Distance Based OTU and Richness Determination EMBL: European Molecular Biology Laboratory FESIN: Fungal Environmental Sampling and Informatics Network MEGAN: Metagenome Analyzer NCBI: National Centre for Biotechnology Information INSD: International Nucleotide Database ITS: Internal Transcribed Spacer MOR: a Perl script package developed by Hibbett et al. (2005) that performs automated phylogenetic taxonomy on mushroom-forming fungi. MOTU: molecular operational taxonomic unit RDP: Ribosomal Database Project SAP: Statistical Assignment Program UNITE: User-friendly Nordic ITS Ectomycorrhizal database group vi

7 List of Figures Figure 1: Steps taken to create the s. (p. 50) Figure 2: Accumulation of Cortinarius ITS sequences in over time. (p. 51) Figure 3: The number of s created by DOTUR at different similarity values. Values were calculated with the 5.8S removed. (p. 52) Figure 4: Rarefaction curve showing no saturation in the number of BSGs that were created. (p. 53) Figure 5. Maximum parsimony bootstrap consensus tree of a cluster of 168 Cortinarius sequences with 80% similarity. (p. 54) Figure 6: The proportion of s that contain one or more different specific epithets. The number of epithets found within a BSG ranged from zero to ten. A BSG that contains no specific epithets contains only environmental and/or unidentified voucher specimens. (p. 56) Figure 7: The percent of labeled sequences whose specific epithets are found in 1, 2, 3, 4, 5 or 6 different BSGs compared to the number of all the labeled sequences. (p. 57) vii

8 List of Tables Table 1: s that were removed from analysis because they BLASTed to genera other than Cortinarius. (p. 58) Table 2: Specific epithets that were found in more than one (BSG) and the total number of BSGs they were found in. (p. 61) viii

9 List of Appendices Appendix 1. Screenshot of the BLAST result from Accession GQ Misidentified sequences, environmental sequences and unidentified sequences are present and uncategorized. (p. 62) Appendix 2. List of sequences in the MySQL database each containing unique identification numbers, collection numbers, species names, BSG numbers, the continent of origin and the forest habitat. (p. 63) Appendix 3: List of all BSGs containing environmental sequences in the BSG database and the presence or absence of unidentified or identified vouchers in the same BSG. (p. 214) ix

10 1 Chapter 1: Introduction Fungal diversity and ecological roles The Kingdom Fungi encompasses a diverse group of eukaryotes ranging from unicellular to filamentous, all united by a cell wall made of chitin, and a heterotrophic, absorptive mode of nutrition. The Kingdom is divided up into eight different phyla: Microsporidia, Chyridiomycota, Neocallimastigomycota, Blastocladiomycota, Zygomycota, Glomeromycota, Ascomycota and Basidiomycota (James et al. 2006, McLaughlin et al. 2009). The number of extant species is unknown but the estimated number of species is 1.5 million (Hawksworth 2001). There are approximately 100,000 described species and this number is increasing by about 1.2% per year (Kirk et al. 2008, McLaughlin et al. 2009). Fungi play vital ecological roles as decomposers, pathogens, and mutualists (McLaughlin et al. 2009) and are found in every biome throughout the world from the arctic to the desert (Robinson 2001; Baptista-Rosas et al. 2007). The only organisms known to be capable of breaking down lignin are select fungi and bacteria (Tuomela et al. 2000). Thus fungi are integral to the nutrient cycle of forests. Fungi are pathogens of algae (Gromov et al. 2000), plants, and animals. They are responsible for crop diseases such as rusts, smut and powdery mildew. For example, Fusarium graminearum was the pathogen responsible for the loss of over 2.6 billion dollars to US agriculture in the 1990 s (McMullen et al. 1997). In animals, Geomyces destructans is associated with white-nosed bat syndrome and may be responsible for the deaths of thousands of North American bats (Gargas et al. 2009, Zimmerman 2009). The chytrid, Batrachochytrium dendrobatidis, is responsible for the global decline and extinction of >200 amphibian species (Fisher et al. 2009, Voyles et al. 2009). As endosymbionts in plant leaves, beneficial fungi may protect the plant against pathogens (Ownley et al. 2009). They form symbioses with cyanobacteria and algae to form lichens, helping to break down rock into soil (Brodo et al. 2001). They are cultivated in fungal gardens by fungivorous ants (Mehdiabadi and Shultz 2009). They are also present in the gut of certain insects and help break down plant metabolites (Gibson and Hunter 2010). Approximately 80% of all species of land plants form beneficial mutualistic mycorrhizal associations with fungi (Wang and Qiu 2006). However, there is a lack of knowledge about the majority of fungi and the wide range of ecological roles they

11 2 perform. To aid studies of fungal taxonomy and ecology, it is necessary to develop tools for rapid species identification. Species concepts in fungi The delimitation of one fungal species from another is often difficult and there is no one species concept that is widely applicable to all fungi. Mycologists often use some combination of the morphological, ecological, physiological, biological and phylogenetic species concepts when describing a new species. Historically, researchers have relied heavily on the morphological species concept to describe new fungi. It is widely used in the Agaricomycetes as only the fruiting body is generally observed and collected. In the Ascomycota, because of dual morphologies, the same organism may have two different names at the genus level: One name corresponding to the sexual state (teleomorph) and the other name pertaining to the asexual state (the anamorph). However, the morphological species concept does not hold up to scrutiny for the fungi. Morphological variation is highly plastic and the choice of defining characters is highly subjective. The range of natural variation within a species is difficult to assess. Armillaria mellea sensu lato was once thought to be a single highly variable species with a worldwide distribution (Pegler 2000). By testing for intersterility, ten different species of A. mellea were identified in North America and five species were identified in Europe, many of which are morphologically indistinguishable (Ullrich and Anderson 1978, Anderson and Ullrich 1979, Pegler 2000). An ecological species concept (ESC) has been employed to describe many different fungi based on their ecological niche. The ESC defines a species as a lineage (or a closely related set of lineages) which occupies an adaptive zone minimally different from that of any other lineage in its range and which evolves separately from all lineages outside its range (Van Valen 1976). This species concept makes the a priori assumption that most parasitic or symbiotic fungal species are host-specific (Moncalvo 2005). However, the extent of host specificity is unknown in most fungi (Moncalvo 2005), and thus this species concept cannot be used in most cases. The physiological species concept is closely related in that new species are described if they grow differently given the same substrate. For example, Pitt (1973) advocated for the identification of Penicillium species by culturing them on two different types of agar and measuring the water activity and growth rate in a range of temperatures. Frisvad (1981) suggested extracellular enzyme production would be useful in Penicillium taxonomy and described two new species

12 3 based on differing secondary metabolites (Frisvad et al. 1989). However, it is possible that different strains of the same species have different heat and water tolerances. For example, in contrast to Penicillium commune, different strains of Penicillium crustosum have different extrolite profiles and physiological data (Sonjak et al. 2009). It takes a significant amount of time and scientific expertise to acquire ecological and physiological data. Due to the difficulties in understanding the complete ecology and physiology of most fungal species, this species concept cannot be used ubiquitously. The biological species concept (BSC) is sometimes used in those fungi that can be cultured such as Armillaria (Anderson and Ullrich 1979), and Ganoderma (Adaskaveg and Gilbertson 1986). It defines species as interbreeding populations that are reproductively isolated from other populations (Mayr 1942). This method of defining species is ineffective for asexual species and for the majority of fungi that cannot be easily cultured or manipulated (Taylor et al. 2000). Also, populations may be genetically isolated from each other but still retain sexual compatibility when grown under laboratory conditions (Moncalvo 2005). Thus, while the biological species concept may hold in specific cases, it cannot be relied upon for the majority of fungi. Because of these difficulties, the phylogenetic species concept is now the favoured species concept in mycology. Its philosophy is that a species should represent a monophyletic group that shares at least one uniquely derived character from a common ancestor (Moncalvo 2005). Taylor et al. (2000) stressed that multi-gene phylogenies rather than single gene phylogenies be used to recognize fungal species. The success of the phylogenetic species concept has led to the use of DNA to help identify fungal species. Difficulties in identifying fungal specimens The correct identification of fungal specimens is fraught with difficulties at both the morphological and molecular level. Fungal taxonomy relies on the examination of asexual or sexual reproductive states because the vegetative state is usually in the form of a filament and provides few morphological characters. However, the fruiting of many fungi is irregular and unpredictable making it difficult to accurately sample the total fungal diversity of a given location at a particular time. The vast majority of fungi still await discovery and formal description (Hawksworth 2001, Blackwell et al. 2006). It is highly likely that the discovery of

13 4 many of the "missing fungi" will come through environmental sequence data rather than traditional cultural methods or fruiting body surveys (Schadt et al. 2003, Porter et al. 2008, Schmidt et al. 2008). It is therefore warranted to explore within species sequence variation in order to match environmental sequences to existing species, or to recognize the existence of novel, cryptic species based on molecular evidence alone. Molecular identification If the circumscription of species or higher taxonomic groups can be evaluated by certain DNA sequences, then surely they can be identified the same way. Carl R. Woese revolutionized systematic biology when he compared 16S (18S) ribosomal RNA sequence data between species of different Kingdoms to show that there were three separate domains of life: Bacteria, Archea and Eukarya (Woese and Fox 1977). Cletus P. Kurtzman was the first to use the D2 domain of the 5 end of the large subunit (26S) rrna to distinguish between different species of yeasts (Peterson and Kurtzman 1991). Today, the most popular locus for systematics and identification of fungi at lower taxonomic levels is the internal transcribed spacer (ITS) region of nuclear ribosomal DNA (Horton et al. 2001, Bruns 2004). From the very beginning, the ITS region has been used to identify both voucher specimens and environmental samples. Gardes et al. (1991) focused on selected Laccaria species to show that the ITS can identify different genera, species and strains of l fungi. Lee and Taylor (1992) showed that the ITS could be used to distinguish between different species of Phytophora by using phylogenetic reconstruction methods and by comparing the interspecific variation. They suggested that P. capsici and P. citropthora were the same species based on low interspecific variation. Today, the ITS has been used so much by mycologists that there are now over 450,000 ITS sequences present in the database. and BLAST NCBI created in 1979 in an effort to create a central national public computerized database of DNA and protein sequences (Brasser 2008). Since then, the International Nucleotide Database (INSD) has grown to include the, EMBL (European Molecular Biology Laboratory) and DDBJ (DNA Databank of Japan) databases (Benson et al. 2006). The Basic Local Alignment Search Tool (BLAST; Altschul et al. 1997) is a

14 5 program that is used to search for short stretches of homologous protein or DNA sequences in the INSD database. For the purpose of DNA-based identification, BLAST is widely used as a first step to retrieve published sequences that match a query sequence. Researchers frequently use the percent identity to evaluate the quality of their matches. However, this is a poor indicator of the overall similarity because it is dependant on the length of the sequence and it has a high false positive rate (Brenner et al. 1998). Crouch et al. (2009) report that misidentified 86% of their Colleototrichum isolates when using BLAST but only 10% were placed incorrectly when a robust phylogenetic method (maximum likelihood or Bayesian) was used. In addition, there are several systematic errors that erode the reliability of matches (Nilsson et al. 2006). First, is underpopulated with fungal sequences (Nilsson et al. 2005), thus confidence in the BLAST score may be overinflated (Munch et al. 2008, U Ren et al. 2009). Second, when regions of low species resolution are used, BLAST provides inaccurate results (Horton et al. 2008). Third, the entries in may not be labeled properly, the quality of the sequences may be poor, and voucher material for species annotation is not always submitted (Bridge et al. 2003, Vilgalys 2003, Bidartondo et al. 2008, Crouch et al. 2009). Bridge et al. (2003) estimated that up to 20% of the sequences in may be unreliable. Bidartondo et al. (2008) called for a community curated annotation process that would allow third parties to improve the annotation of sequences. The DNA barcoding initiative is an effort to improve the identification of specimens by using a database of well identified and well documented voucher specimens and by using a superior algorithm to find conspecific sequences. The DNA barcoding initiative The Barcode of Life initiative started in 2003 at the University of Guelph and is being spearheaded by Paul Hebert (Hebert et al. 2003). Since then, the international community has rallied around this idea. The goal of DNA barcoding is to use a single sequence of DNA to rapidly distinguish between different species. This initiative has also developed the Barcode of Life Database (BOLD; Ratnasingham and Hebert 2007), which differs from in that it has strict guidelines to ensure data quality and retrieval. For instance, DNA trace files (chromatograms) must be deposited together with sequences, and voucher specimens must have detailed documentation (Ratnasingham and Hebert 2007).

15 6 The ITS region has now been accepted by BOLD as the primary target region for a firstlevel approximation in identifying fungal species. The cytochrome c oxidase I (CO1) gene was originally proposed to be used as the universal identifier for all eukaryotes (Hebert et al. 2003) but it does not work well for plants and fungi (Kress and Erickson 2008, Seifert 2009, Vialle et al. 2009). The ITS region was selected as the primary barcode sequence for fungi because of its ease of amplification using universal primers and its ability to differentiate between different genera, species and varieties, depending on the group of interest (Nilsson et al. 2008). Its multicopy nature and its homogenization through concerted evolution are what make it easy to amplify (Ganley and Kobayashi 2007). Universal fungal primers can be used to detect almost all species of all the phyla in the kingdom (Gardes and Bruns 1993). This makes it easy to detect microfungi, such as chytrids, as well as spores or tissue from environmental samples (Gardes and Bruns 1993). The ITS region has three subloci of different rates of evolution, ITS1: highly variable, 5.8S: very conserved, and ITS2: variable to semi-conserved (Hillis and Dixon 1991, Hershkovitz and Lewis 1996). Its size varies between 450 and 800 bp with very few exceptions. It is flanked on both its 5' and 3' sides by very conserved regions (SSU [18S] and LSU [28S], respectively) which simplifies the design of primers targeted at the ITS region. Hundreds of studies in systematic mycology have used ITS sequences already as 963 papers were found in PubMed alone using the search words ITS and fungi (accessed May 21, 2010). However, the ITS region does not always work well for species discrimination. For instance, it cannot resolve well-characterized species in Heterobasidion, Armillaria, Fusarium, and Penicillium (Chillali et al. 1998, Skouboe et al. 1999, Bruns 2001, Seifert et al. 2007). In some cases, the ITS will not detect known cryptic species in species complexes (Seifert et al. 2007). In certain taxa, two or more different copies of the ITS may be present in the genome, creating complications for direct sequencing (Matheny et al. 2007) and ITS paralogues have been detected in Fusarium and Cantharellus (O Donnell et al. 1998a, Moncalvo personal communication). For the ITS to work perfectly as a barcode, there should be a barcode gap where the levels of intra- and interspecific variation do not overlap (Meyer and Paulay 2005). However, for most fungi, there is a range of genetic variability, both within species complexes and between species complexes, such that the levels of intra- and interspecific variation do overlap (Goldstein and DeSalle 2000). The ITS is not equally variable in all groups of fungi, thus there is no one single similarity cut-off value that can be used for automated species delimitation

16 7 (Nilsson et al. 2008). Given these limitations, the ITS should only be used for a first level species approximation. Because of its ease of amplification and resolution power at the species level, ITS sequences are also targeted in both the FESIN (Fungal Environmental Sampling and Informatics Network) (Horton et al. 2008) and UNITE (User-friendly Nordic ITS Ectomycorrhizal database group; Kõljalg et al. 2005) initiatives for the identification of fungal species from the environment, such as mycorrhizal root-tips and soil samples. These initiatives aim to match environmental sequences with sequences from well-identified vouchered specimens whose sequences have been deposited in public sequence databases such as, or in the specially curated UNITE or BOLD databases. There are several problems with using, BOLD and UNITE to identify novel specimens with the BLAST program. Taxonomic errors, unidentified and environmental sequences make it difficult to know which sequences are related to each other (Vilgalys 2003). Misidentified, unidentified and environmental sequences contribute to the diversity of fungi but may not ever be represented in curated vouchered databases such as BOLD or UNITE. To accommodate undescribed species, misidentified species, species complexes, cryptic species and environmental sequences, it is necessary to develop a method of grouping similar sequences based on genetic relatedness, rather than taxonomy. What is a? In this thesis, I introduce the concept of creating s to group together genetically similar fungal ITS sequences regardless of their taxonomic annotation. A (BSG) is a cluster of sequences that share greater than or equal to a threshold amount of genetic similarity with each other. It differs from a DNA barcode in that the taxonomic label does not refer to a species. In a BSG, the correlation between species and clades will not be initially known. Clades may reflect species boundaries well, they may reflect a group of closely related species or they may reflect a genetically divergent subpopulation within a species (Horton et al. 2008). Each BSG will have its own genus and number and be placed in a curated database where human annotations could be updated. The creation of BSGs would enable identification of genetically similar groups, which would facilitate the sharing of biogeographic and ecological data, regardless of how the sequence was originally labeled. Once

17 8 these BSGs have been created, it will be necessary to have trained taxonomists perform detailed morphological and systematic studies involving multi-gene phylogenies to determine the number of different species within a BSG. Unlike BLAST, which lists a continuous list of matches, the BSG classifier only lists those sequences that are the most likely to match the query sequence. In this thesis, I will explore the use of BSGs to circumvent ambiguous or improperly labeled sequences to organize these sequences better, leading to improvements in taxonomy and the collection of sequence data, using the mushroom genus Cortinarius as a test case. The genus Cortinarius (Agaricales, Basidiomycotina) The genus Cortinarius Fr. is the largest genus among the agaric fungi (Brandrud et al. 1990). There are 4638 published names of Cortinarius in Index Fungorum (CABI Databases, accessed Jan. 19, 2010). Elias Fries first used the term Cortinarius in his work Epicrisis ( ). Moser (1983) recognized six subgenera: Myxacium, Phlegmacium, Telamonia, Leprocybe, Seriocybe, and Cortinarius. The section Dermocybe was treated as a separate genus. Brandrud et al. (1990) only recognized four subgenera: Myxacium, Phlegmacium, Telamonia and Cortinarius. Bidaud et al. (1994) recognized Cortinarius, Dermocybe, Phlegmacium, Telamonia and Hydrocybe. Cortinarius is a monophyletic group that is well characterized morphologically by a rusty brown spore print, warty rough walled spores and a cobweb like partial veil called a cortina (Peintner et al. 2004). The habit ranges from mycenoid to tricholomatoid, the cap can be dry, silky, squamose, fibrillose or viscid, and the fruiting bodies can range from brightly coloured to dull brown (Peintner et al. 2004). However, there are exceptions to this morphology. Taxa previously classified in Rozites and Cuphocybe have membranous partial veils, Cortinarius renidens lacks a veil, and Raphacea mariae has olive coloured spores (Peintner et al. 2002). Molecular studies have revealed the genera Rozites, Cuphocybe, Raphacea, as well as the secotoid genera Thaxterogaster, Quadrispora, Protoglossum and Hymenogaster were nested within the Cortinarius clade (Peintner et al. 2001, Peintner et al. 2002). These names have now been synonymized with Cortinarius. The wide range of morphology in the group makes it difficult to identify species and categorize them into subgenera (Peintner et al. 2004). Peintner et al. (2004) showed that all the subgenera within the genus (sensu Moser in Singer 1986) are polyphyletic with the exception of subgenus Cortinarius. They recognized fourteen clades within

18 9 Cortinarius. Peintner et al. (unpublished) produced a phylogenetic tree from 5 genes (ITS, LSU, elongation factor 1-alpha, ATPase subunit 6 and NADH ubiquinone oxidoreductase chain 5), which showed short branch lengths in the early diversification of the genus. This study could not resolve deep phylogenetic relationships in Cortinarius, therefore, natural delimitation of subgenera and sections in the genus remain unclear. Cortinarius forms beneficial l associations with the roots of temperate trees worldwide. Their host trees primarily belong to the Fagales but they are also found with members of Pinaceae, Myrtales, Tilia, Ericales, Salicaceae, Cistaceae, Dipterocarpaceae, Eucalyptus and Dryas (Bougher et al. 1994, Peintner et al. 2003, Frøslev et al. 2005). Cortinarius e have also been reported associated with herbaceous species such as Kobresia myosuroides (Cyperaceae) (Muhlmann and Peintner 2008), Carex flacca and C. pilulifera (Cyperaceae) (Harrington and Mitchell 2002). They are the dominant l fungal group in terms of species diversity and above ground biomass in temperate forests (Brandrud et al ). Wright et al. (2009) found that the genus Cortinarius accounted for 32% of l fungal diversity in a single stand of Western Hemlock (Tsuga heterophylla) forest. They also found differences in Cortinarius species composition between different stands that varied in nitrogen and phosphorus levels. The abundance of the genus in temperate forests, the difficulty in identifying morphological species and its ecological importance make Cortinarius a good candidate for using molecular identification. Molecular identification of Cortinarius The main source of sequences for identifying species of Cortinarius is. There are 2891 sequences labeled as Cortinarius, Dermocybe, Thaxterogaster, Protoglossum, Quadrispora or Cuphocybe in (accessed March 12, 2010). In contrast, UNITE (accessed April 8, 2010) has only 986 Cortinarius sequences and 105 of these are unidentified. One unidentified sequence is labeled as Dermocybe. As seen from the number of different names of genera submitted to, changes in nomenclature are not kept up to date. There are 1546 unidentified Cortinarius sequences in according to Emerencia (accessed March 12, 2010; Nilsson et al. 2005). In addition, misidentified Cortinarius sequences are known to exist in. There is an abundance of Cortinarius sequences in, yet the

19 10 prevalence of unidentified sequences, and misidentified sequences makes it difficult to interpret the results of a BLAST search. When viewing the BLAST output of putative, yet unknown Cortinarius sequences, one cannot determine whether the top hit sequences belong to a single or to multiple putative species (Appendix 1). Every unidentified and environmental sequence submitted to is labeled differently. Cortinarius environmental sequences are labeled Ectomycorrhiza, Fungus, Ectomycorrhizal root-tip, and Fungal sp.. Each sequence may be given with its own collection number, yet the sequences may be 100% identical to each other. Appendix 1 is a screenshot of the BLAST results of accession number GQ Here we see sequences that are identical to each other, yet are labeled as different species. There are two from two different studies that are labeled using their collection numbers. A maximum parsimony analysis reveals that GQ159883, GQ159849, GQ159907, GQ159847, GQ159795, and GQ form a cohesive clade with little variation (data not shown). Thus, due to the abundance of Cortinarius sequences in and the large quantity of misidentified, unidentified and environmental sequences, Cortinarius would be a good test case for the creation and use of s. Objectives The focus of this thesis is to present and develop the idea of creating and maintaining a database of sequences that clusters and labels sequences based on their genetic similarity to each other, rather than relying on the taxonomy submitted by the authors. I will use the genus Cortinarius as a test case to create s. The objectives of this thesis are to: a) examine the taxonomic diversity and accuracy of ITS sequences labeled Cortinarius in and provide examples when using BLAST results to classify sequences gives inaccurate results. b) create a BSG database of ITS Cortinarius sequences c) use a classifier that is capable of automatically classifying a sequence into a BSG and gives a level of confidence in the classification d) create a custom BLAST tool that can be used for comparison and to calculate the percent similarity

20 11 e) show the usefulness in using the barcode similarity group database to: i) find misidentified sequences ii) find novel and cryptic species iii) collect and label environmental sequences iv) match environmental sequences to voucher specimens v) mine the database for metadata such as geographical and ecological data f) discuss the benefits of using BSGs compared to other barcode type databases In addition, a public website is created where one can submit a query sequence and have it classified into a BSG. Then, through access to the BSG Database, information about all the other sequences in that BSG is retrieved.

21 12 Chapter 2: The Creation of ITS-rDNA s for the Genus Cortinarius (Agaricales, Fungi) Introduction Global climate change and rapid habitat loss are threatening the world s biodiversity. A global effort is underway to provide a DNA barcode to every single living species on Earth. (DeSalle et al. 2005, Savolainen et al. 2005) This effort is mainly focused on facilitating species discovery and aiding biodiversity assessments (Savolainen et al. 2005). The identification of organisms by use of a DNA barcode requires the existence of a reference database of wellidentified vouchers and a suitable molecular marker that has a barcode gap showing higher inter- than intraspecific variation (Meyer and Paulay 2005). Less than 10% of the total estimated number of species has been described in the Kingdom Fungi (Hawksworth 2001) and of those that have been described, many are difficult to identify without taxonomic expertise. Thus there is great demand to identify fungi using a DNA barcode approach (Horton et al. 2008). While there is estimated to be 1.5 million (Hawksworth 2001) species of fungi, only approximately 100,000 species have been described (Kirk et al. 2008). The vast majority of fungi still await discovery and formal description (Hawksworth 2001; Blackwell et al. 2006). Because morphological characteristics are scant and reproductive structures used for the basis of identification may be rare, it is highly likely that the discovery of many of the "missing fungi" will come through environmental sequence data rather than traditional cultural methods or fruiting body surveys (Schmidt et al. 2008; Porter et al. 2008; Schadt et al. 2003). Thus, there is a need to match environmental sequences to existing species, or to recognize the existence of novel or cryptic species based on molecular evidence alone. For the purpose of DNA-based identification, BLAST (Altschul et al. 1997) is widely used as a first step to retrieve published sequences that match a query sequence in the database. However, taxonomic errors, unidentified and environmental sequences make it difficult to know which sequences are related to each other (Vilgalys 2003). The BOLD (Barcode of Life Data Systems; Ratnasingham and Hebert 2007) and UNITE (Kõljalg et al. 2005) databases were developed to improve the quality of annotation and have increased standards of documentation for submitted fungal sequences. However, these databases are still subject to the same human

22 13 subjectivity and error as that found in. Furthermore, misidentified, unidentified and environmental sequences contribute to the diversity of fungi but are not represented in the BOLD database. To accommodate this unseen diversity, it is necessary to develop a method of grouping similar sequences based on genetic relatedness, rather than taxonomy. The most popular locus for identifying unknown fungal sequences is the nuclear rdna ITS (internal transcribed spacer) region (Horton and Bruns 2001) because of its ease of amplification using universal primers and its ability to differentiate genera, species and varieties (Nilsson et al. 2008). The region has three subloci of different rates of evolution, ITS1: highly variable, 5.8S: very conserved, and ITS2: variable to semi-conserved (Hillis and Dixon 1991; Hershkovitz and Lewis 1996). Its size varies between 450 and 800 bp with very few exceptions. It is flanked on both its 5' and 3' sides by very conserved regions (SSU [18S] and LSU [28S], respectively) which simplifies the design of primers targeted at the ITS region. Its multi-copy nature and its homengenization through concerted evolution are what make it easy to amplify (Ganley and Kobayashi 2007). Universal fungal primers can be used to detect almost all species of all the phyla in the kingdom (Gardes and Bruns 1993). This makes it easy to detect microfungi, such as chytrids, as well as spores or tissue from environmental samples (Gardes and Bruns 1993). In this study, I introduce a method of grouping fungal sequences by genetic similarity rather than by taxonomy, and providing a label to each genetic group. I define a Barcode Similarity (BSG) as a cluster of sequences that share greater than or equal to a threshold amount of genetic similarity. BSGs are similar to molecular operational taxonomic units (MOTUs) used in many ecological studies of bacteria, algae, nematodes and fungi (e.g. Floyd et al. 2002, Case et al. 2007, Chen et al. 2009, Wright et al. 2009). Although a percent similarity threshold value is often used to delimit the MOTUs in these studies, there is no standard for the similarity value, or for the clustering algorithm that is used to group the sequences. The idea and desire to group sequences in based on a similarity cut-off value and to then re-name these sequences based on the cluster they are a member of, has been circulating for some time (e.g. Horton et al. 2009) but no one has completed this task to date. BSGs differ from DNA barcoding in that the taxonomic label does not refer to a species. In a BSG, the correlation between species and clades will not initially be known. Clades may

23 14 reflect species boundaries well, they may reflect a group of closely related species or they may reflect a genetically divergent subpopulation within a species (Horton et al. 2009). Each BSG will have its own genus and number and be placed in a curated database where annotations could be updated. ing together genetically similar groups would facilitate the sharing of geographic and ecological data despite a lack of annotation. This ability to instantly display all the accessions that are within a particular similarity of each other, regardless of the taxonomy provided by the author, is what fundamentally sets BSGs apart from other identification methods. This paper will explore the use of BSGs using the mushroom genus Cortinarius as a test case. Cortinarius Fr. is the largest genus in the gilled fungi (Brandrud et al. 1990). Its wide range of morphology makes it difficult to identify species in the group and to categorize them into subgenera (Peintner et al. 2004). There are 2891 sequences labeled as Cortinarius (or one of its synonyms: Dermocybe, Thaxterogaster, Protoglossum, Quadrispora and Cuphocybe) in (accessed March 12, 2010). This number includes unidentified sequences but does not include sequences that were not identified at the genus level. As seen from the number of different names of genera submitted to, changes in nomenclature are not kept up to date. There are 1546 unidentified Cortinarius sequences in according to Emerencia (accessed March 12, 2010; Nilsson et al. 2005). Misidentified Cortinarius sequences are known to exist in. Cortinarius is good genus to use to test the use of s because of the abundance of sequences in and the large quantity of misidentified, unidentified and environmental sequences. In this paper I use the genus Cortinarius as a test case to create and use Barcode Similarity s. I examine the taxonomic diversity and accuracy of ITS sequences labeled Cortinarius in and present a BSG database. I demonstrate the use of this database to find misidentified sequences, find novel and cryptic species, collect and label environmental sequences, match environmental sequences to voucher specimens, and search the database for metadata such as geographical and ecological data. I then present a website where unknown query sequences can be classified into one of the BSGs.

24 15 Materials and Methods Obtaining sequences Figure 1 illustrates the steps taken to create the BSGs. To create a database of Cortinarius ITS sequences, sequences were downloaded from on October 25, 2008 using the search words: Cortinarius ITS. Additional sequences were found by a BLAST search using a Cortinarius traganus sequence against all sequences containing Fungus,, Cortinarius, mycorrhiza, cf. Dermocybe, mycorrhizal basidiomycete, and Fungal sp. Fifty-eight sequences were downloaded from the Duke Forest Mycological Observatory Website ( One hundred and seven sequences were from specimens in the Royal Ontario Museum Fungarium (TRTC) and from environmental sequences obtained by Terry Porter. To obtain non-cortinarius reference sequences representative of all major fungal groups, on November 20, 2009, five hundred twenty six fungal ITS sequences were downloaded from using the search words AFTOL internal transcribed spacer. Creating s Before alignment, the DNA sequences were manually trimmed to the beginning of ITS1 (including the last 5 nucleotides of the 3-end of the 18S gene; CATTA) and to the end of ITS2 (including the first 5 nucleotides of the 5-end of the 25S gene; GACCT). s that did not include any nucleotides in the first and last 200 base pairs of this region were removed. The 5.8S region was removed from the sequence alignment to ensure the greatest phylogenetic signal was obtained from the data as recommended by Nilsson et al. (2008). s that did not belong to the genus Cortinarius (as confirmed by a BLAST search) were removed. Habitat and locality data were collected by extracting this information from each entry and by locating it in the publications cited. This data was entered into a MySQL 5.0 database. MAFFT (Katoh et al. 2005) was used to align the sequences using the FFT-NS-i method. This is a fast iterative refinement method that can handle a large dataset. It works by creating a distance matrix based on the number of shared 6-tuples. A guide tree is created and a progressive alignment is made. It employs a fast Fourier transform algorithm, which is just as accurate as the Needleman-Wunsch dynamic programming algorithm, but is much faster

25 16 (MAFFT Version 6 [internet]). The guide-tree is reconstructed and the sequences are re-aligned. The iterative refinement method is repeated until there is no more improvement in the weighted sum-of-pairs (WSP) score or the number of cycles reaches The default scoring matrix is derived from Kimura s two-parameter model. The ratio of transitions to transversions is set to 2 by default. Different clustering programs were tested for speed, ease of use and taxonomic accuracy. I decided upon DOTUR 1.53 (Distance Based OTU and Richness Determination) (Schloss and Handelsman 2005) to delimit the ITS sequences into BSGs. I tested different clustering algorithms and a range of threshold values between 90 and 100%. Before inputting the data into DOTUR, a corrected distance matrix was constructed for the alignment in PAUP* 4.0b10 (Swofford 2002) using a GTR + I + G model (selected by ModelTest 3.7 using Akaike weights; Posada and Crandall 1998). I used the furthest neighbour joining algorithm as recommended by Schloss and Handelsman (2005). The robustness of this particular approach is addressed in the Discussion. Each accession was assigned a Cortinarius BSG number from 1 to 393. These BSG numbers were imported into the MySQL database using the unique identifier. DOTUR was used to produce a rarefaction curve as well as Chao1 and ACE diversity estimates. To test the monophyly of the BSGs, I chose a random group of sequences with 80% similarity as determined by DOTUR. At this similarity value, ITS sequences can be aligned unambiguously across all taxa. s were aligned using MAFFT with the e-ins-i algorithm. The alignment was edited using Se-Al v2.0all ( PAUP* 4.0b10 (Swofford 2002) was used to create a maximum parsimony tree using heuristic searches consisting of 10 rounds of random addition sequences and TBR branch-swapping, keeping 100 trees per replicate. Branch support was evaluated from 100 bootstrap replicates of random addition sequences with maxtrees set to Matching a query sequence to a BSG The Ribosomal Database Project (RDP) Classifier (Wang 2007) was used to classify a query sequence into one of the BSGs that were created in the previous step. It is a naïve Bayesian classifier that rapidly and accurately classifies sequences into a given taxonomy (Wang 2007). The RDP Classifier works by breaking up the query sequence into 8-base words (Wang 2007). The prior likelihood of each word occurring in the entire database is calculated (Wang

26 ). The probability of observing a genus containing a given set of words is calculated (Wang 2007). Bayes theorem is used to calculate the probability of a given sequence belonging to a BSG (Wang 2007). The sequence is classified as a member of the BSG giving the highest probability score but the actual numerical probability estimate is ignored (Wang 2007). This process is repeated on the same sequence using different sets of 8-base words (Wang 2007). The number of times a BSG is selected out of 100 bootstrap trials is used as an estimate of the confidence in the assignment to that BSG (Wang 2007). This method was chosen because it is fast, accurate and easily customizable. The RDP classifier uses information averaged over the entire genus and thus is less influenced by individual misplaced training sequences. s, containing the entire ITS, were exported and formatted for entry into the RDP classifier and BLAST databases. The AFTOL sequences were included so that the classifier could identify sequences that were not Cortinarius sequences. The AFTOL sequences were exported for entry into the RDP classifier but were not assigned BSG numbers. The RDP classifier requires that a taxonomy is provided for each sequence and that each rank is known. The taxonomy supplied by the AFTOL group to was used. If in doubt about a particular level of hierarchy, the hierarchy in Index Fungorum (CABI Databases, was used. I created the ranks: Domain, Kingdom, Subkingdom, Phylum, Subphylum, Class, Subclass, Order, Family, Genus and, to enable expansion of the classifier to identify other groups of fungi. The FASTA file for the BLAST program includes the BSG number, the genus and the species. I used a cross-validation technique (Kohavi 1995) to test the ability of the classifier to correctly classify a query sequence into a BSG. All BSGs that consisted of a single sequence were removed, and the others were divided into two equal groups ( A and B). First A was used to train the classifier while sequences from B were submitted to the classifier. Then B was used to train the classifier while sequences from A were used to query the classifier. The results of the RDP classification were then compared to the original BSG classification made by DOTUR. Web access to the BSG database The URL for accessing the Classifier is located at: First the query is checked to see if its length is between 400

27 18 and 800 base pairs. If it is under 400 base pairs, this warning message is displayed: length is less than 400 base pairs. Please try again with a longer sequence. If it is over 800 base pairs, this warning message is displayed: length is greater than 800 base pairs. Please trim the sequence to 5'CATTA...3'GACCT. If the length requirement is met, the RDP classifier assigns the sequence to the genus Cortinarius or to a separate genus or family of fungi. If the classifier does not recognize the sequence as belonging to Cortinarius, it displays the message: This is NOT A CORTINARIUS sequence. The above classification is only based on 526 AFTOL sequences and is only a suggestion. DO NOT USE FOR CLASSIFICATION. If the classifier recognizes the sequence as belonging to the genus Cortinarius, a query will be sent to the MySQL database to retrieve all the sequences belonging to the closest BSG as well as the metadata associated with each sequence. The query sequence is then blasted against the customized BSG database to compare the BLAST output with the RDP classifier output. Only the BLAST results with an e-value of 0.0 are shown as these are the most statistically significant results. The BLAST result displays the accession number, BSG number, species annotation, bitscore, e-value and percent identity. Since the metadata retrieved from the classifier is unordered, the accession number can be used to match each specimen with a percent identity number. This is helpful when there is a large range of genetic variation within the BSG. Results ITS sequence diversity and taxonomic accuracy in public databases When viewing the BLAST output of putative, yet unknown Cortinarius sequences, one cannot determine whether the top hit sequences belong to a single or to multiple putative species (Appendix 1). Every unidentified and environmental sequence submitted to is labeled differently. Cortinarius environmental sequences were labeled Fungus,, Cortinarius, mycorrhiza, cf. Dermocybe, mycorrhizal basidiomycete, and Fungal sp.. Each sequence may be given with its own collection number, yet the sequences may be 100% identical to each other. Appendix 1 is a screenshot of the BLAST results of accession number GQ Here we see sequences that are identical to each other, yet are labeled as different species. There are two from two different studies that are labeled using their collection numbers. A maximum parsimony analysis reveals that GQ159883, GQ159849,

28 19 GQ159907, GQ159847, GQ159795, and GQ form a cohesive clade with little variation (data not shown). In this study, of the 2482 sequences retrieved as described above, nineteen sequences did not belong to the genus Cortinarius and were removed (Table 1). Four of these sequences were likely sequence submission mistakes as they came from fruiting bodies that one would not expect to get confused with Cortinarius. The rest of the mislabeled sequences came from e and why they were labeled Cortinarius remains unclear. In total, 2463 complete Cortinarius ITS sequences were obtained (Appendix 2). Seven different genera known to be synonymous with Cortinarius were included: Cortinarius, Cuphocybe, Dermocybe, Protoglossum, Hymenogaster, Quadrispora, and Thaxterogaster (Appendix 2). All sections of the genus Cortinarius are represented. Synonyms were recorded such as using the old genus name Dermocybe and the use of Cortinarius muscigenus Peck for Cortinarius collinitus (Pers.) Fr. (Appendix 2). Using the total number of recorded specific epithets (502 epithets including synonyms; Appendix 2), only about 25% of the total number of described species (~2000) are present in. Seventy-nine percent of the sequences were from basidiocarps, 17% were from root-tips, and 2% were from the soil (Appendix 2). The origin of the remaining 2% of sequences is unknown. Of the total number of basidiocarps (1926), 18% (340) were unidentified (Appendix 2). All the environmental sequences had no specific epithets except for seven mycorrhizae that were identified by the authors based on high sequence identity to identified specimens. The number of Cortinarius ITS sequences submitted to over time was plotted (Figure 2). The earliest date was April 22, 1996 and the oldest date was September 27, This graph shows that the entry of sequences obtained from fruiting bodies, root-tips and from the soil have been increasing exponentially since the late 1990 s and continues to increase. Creating the BSG Database I looked for a barcode gap where the level of intraspecific variation was less than that of the interspecific variation. I plotted the number of BSGs created by DOTUR against a range of threshold values from 90 to 100% (Figure 3). The number of BSGs created increased linearly with increasing threshold values, signifying continuous variation in the dataset. Thus there was no natural cut-off value that separates intra- and interspecific variation in Cortinarius. Since a

29 20 cut-off value could not be selected naturally, an arbitrary 94% similarity value was chosen to ensure all members of a species were included in a BSG. This similarity value was calculated with the 5.8S removed and may be similar to the 95 to 97% similarity values commonly used in mycological studies where the 5.8S has not been removed. Next, the robustness of using DOTUR to form the same BSGs given different datasets was evaluated. When half of the 2463 Cortinarius sequences, selected randomly, were realigned using the same MAFFT algorithm, the smaller dataset only had 345 BSGs instead of the 393 BSGs present when the full dataset was used. Thus, the contents of the BSGs were different depending on the size and variability of the input dataset. From the 2463 Cortinarius sequences included in our analysis, 393 Cortinarius BSGs were created using DOTUR (Appendix 2). A rarefaction curve (Figure 4) showed no saturation in the number of BSGs created. The Chao1 diversity estimate was BSGs with a 95% confidence interval of to BSGs. The ACE diversity estimate was BSGs with a 95% confidence interval of to BSGs. Each BSG contained anywhere from zero to thirteen different specific epithets (Figure 6). Seventeen percent of the BSGs did not contain any specific epithets and thus only contained sequences derived from unidentified or uncultured specimens (Figure 6). The same epithet was found in multiple different BSGs (Table 2). A bootstrap consensus tree, created from sequences that are 80% similar to each other (Figure 5), shows that, for this selected group, 14/15 BSGs are monophyletic with a bootstrap support greater than 70%. BSG 191 is split into four distinct clades each with a bootstrap value of 98 to 100%. The relationship between these four groups cannot be resolved and thus the monophyly of this group cannot be rejected. Finding misidentified sequences s that are misidentified at both the genus and species level were recognized. As stated above, Table 1 shows twelve sequences that were identified as Cortinarius but BLASTed to other genera. These sequences were removed from the dataset. Misidentified sequences within the genus Cortinarius may be found by looking at cases where the same specific epithet occurs in multiple BSGs. Ninety-six specific epithets were found in more than one BSG (Table

30 21 2). Of the total number of labeled sequences (1606), 41% of the sequences fall into this category (Figure 7). The majority of these possibly misidentified sequences (56%) are found in only 2 BSGs. Fifty-nine percent of the labeled sequences were not found in more than one BSG. Misidentifications may (but not always) be detected in BSGs that have more than one specific epithet in the same BSG. For example, phylogenetic analysis of BSG 121 shows that EU821682, originally identified as C. superbus, is likely C. elegantior (data not shown). Figure 7 shows 36% (142) of the BSGs contain two or more specific epithets. However, without seeing the specimens, we can only speculate if these are indeed misidentifications. Finding novel and cryptic species Novel and cryptic species may also be detected when one specific epithet is found in multiple BSGs. Cryptic species that have a low amount of genetic variation may not be detected. For example, C. comarostaphylii and its sister species C. leucophanes are both present in BSG 276. Novel species may be detected in BSGs that only contain unidentified vouchers. As stated above, 22% (86) of the BSGs contain only unidentified vouchers. However, detailed morphological and phylogenetic analyses are needed to confirm such instances. Collecting and labeling environmental sequences and matching them to voucher specimens In this study, 460 environmental sequences were placed into 110 different BSGs (Appendix 2). Of these BSGs, 66 (60%) contained an identified voucher specimen, 9 (8%) contained an unidentified voucher and 35 (32%) contained only environmental sequences (Appendix 3). The compilation of environmental sequences that do not have identified voucher specimens is useful. For example, BSG 297 contains unidentified vouchered sequences and environmental sequences collected from seven different coniferous forests located all across North America (Appendix 2). The label, BSG 297, denotes these sequences are at least 94% similar to each other and could now be discussed in a comparative context. Metadata retrieved from BSGs Of the sequences that gave continent data (2126; 86%), 54% were from North America, 37% were from Europe, 3% were from South America, 5% were from New Zealand, Australia and Tasmania, less than 1% were from Asia and none were from Africa (Appendix 2).

31 22 s came from 31 different countries (Appendix 2). A number of BSGs did contain both Northern Hemisphere and Southern Hemisphere species, but under phylogenetic analysis, these separated into different clades (data not shown). A probable instance of human-mediated migration can be documented by the northern temperate species C. fulvoconicus having been documented in Chile (Appendix 2). Habitat data included hardwood, coniferous and mixed forests. Unusual hosts were the monocots Carex flacca and C. pilulifera, Kobresia myosuroides, an orchid (Epipactis microphylla), the rosids Halimium halimifolium, and Helianthemum nummularium, an ericaceous plant (Comarostaphylis arbutoides), and various species in a tropical forest (under Hopea spp. (Dipterocarpaceae), Calophyllum sp. (Clusiaceae), Meiogyne sp. (Annonaceae) and Cansjera rheedii (Opiliaceae) (Appendix 2)). The BSG Classifier and website The RDP Classifier was used to classify a query sequence into a BSG and to perform a BLAST search. When a Cortinarius traganus (GQ159890) sequence that contained the 18S and/or the 25S regions was submitted, the top BLAST matches correctly matched the input sequence, but the classifier assigned the query to a different BSG with low confidence. When a partial sequence was submitted, neither BLAST nor the classifier was able to give any indication of classification. When a sequence that contained the last five nucleotides of the 18S, ITS1, 5.8S, ITS2 and first five nucleotides of 25S was submitted, both the classifier and BLAST were able to identify the sequence. Thus only sequences that have been trimmed should be used to classify an unknown query sequence. When sequences from B were used as queries against the A training sequences in a cross-validation test, 98.9% of the sequences were correctly classified. When sequences from A were used queries against the B training sequences, 98.6% were correctly identified. On average 98.8% of the sequences were correctly classified. The majority of the misidentified sequences had a bootstrap value of less than 70%. They were found to be unique sequences that were dissimilar to other sequences in the same original BSG. Thus the reliability of the classification is high but depends somewhat on the full range of variability being present in the training dataset.

32 23 Discussion Barcode similarity groups can improve sequence identification when compared to a simple BLAST search. Classifications based on BLAST matches in are unreliable. Sequencing errors, misidentification or mislabeling, contamination of cultures, PCR-based errors, and differences in taxonomic opinion between specialists are all common problems one is faced with when performing a BLAST search (Vilgalys 2003). BSGs allow for sequencing errors, can detect misidentification or mislabeling, can detect novel or cryptic species and allow for differences in taxonomic opinion. They also provide a common nomenclature for environmental sequences and can serve as a repository for ecological and geographical data. However, before BSGs become widely used, one must first consider the limitations of the method used to create them. This study suggests that there is no robust method to create BSGs that is fully satisfactory. DOTUR was chosen because it is fast and can handle a large number of sequences. In constructing BSGs, the particular parameters input into DOTUR (the multiple sequence alignment, distance correction, clustering algorithm and distance threshold) greatly affect the number of OTUs created as well as the composition of the OTUs (White et al. 2010). White et al. (2010) tested different sets of parameters on a 16S rdna bacterial dataset and found that the distance threshold (95% to 99%) and the choice of multiple sequence alignment algorithm (ClustalW, NAST and MUSCLE) made the greatest difference to the composition of OTUs when using DOTUR (White et al. 2010). The threshold that best represented the true species composition was 0.05, which is well below the 97-99% similarity threshold typically used for microbial biodiversity studies (White et al. 2010). The distance correction measures did not change the variability significantly (White et al. 2010). Given the limitations of the dataset, and limited methods publicly available to create OTUs, I strove to optimize the parameters entered into DOTUR. Deciding which threshold value to use to both delimit species and identify sequences is a widely debated issue. Percent similarities used to date, to identify both vouchered and environmental fungal sequences, vary between 95% and 99% (Tedersoo et al. 2006, Bidartondo and Read 2008, Ryberg et al. 2008, Wright et al. 2009). The ITS region is not equally variable in all groups of fungi and thus there is no one single value that works for all fungi (Nilsson et al.

33 ). An arbitrary 94% threshold value was selected for delimiting BSGs to ensure that all members of a species were included in the same group even if there were sequencing errors, the alignment was poor or if there was a large amount if intraspecific variability. However, this low similarity value also means that multiple species may be present within a BSG. Unfortunately, the ITS is unable to differentiate between all Cortinarius morphospecies (Frøslev et al. 2007, Niskanen et al. 2008) and thus further molecular analysis of the specimens within each BSG should be undertaken. The creation of BSGs should not be seen as an end-point but be seen as a starting point for taxonomic investigation. In this study, I found that when the number of Cortinarius sequences was reduced to a half, the number of BSGs was reduced and the composition of the BSGs changed subtly. I argue that although the contents of a BSG will vary from dataset to dataset, this is of little consequence. It is understood that there may be multiple species within a BSG and that detailed taxonomic work, involving morphological work and multi-gene phylogenies, will be needed to differentiate between species. Since the ITS cannot always be relied upon to resolve interspecific relationships, it is more important that all members of a species be present in a BSG than the precise composition of the BSG. Maximum parsimony analyses of a subset containing 168 sequences sharing 80% similarity and representing 15 BSGs was conducted. The strict consensus tree indicated that 14 of them were monophyletic with > 70% bootstrap support, whereas one BSG was split into 3 unresolved clades (Figure 5). This tree also suggests that the majority of BSGs are likely to contain two to three species (as shown by distinct monophyletic groups within a single BSG with bootstrap support greater than 70%). Figure 4 shows that there is no saturation in the number of BSGs that could be created. Both the Chao1 and ACE estimators predict approximately 750 BSGs would be needed to reflect the entire species diversity of Cortinarius. However, these estimators have been shown to vary with sample size, particularly by severely underestimating the total diversity given low sample size (Roesch et al. 2007). Since the sequence rarefaction curve from this study is still unsaturated (Figure 4), one can therefore expect a higher BSG diversity to exist in Cortinarius. This would be in agreement with the observation that less than twenty-five percent of the total number of described Cortinarius species were included in the present study. The main improvement the Database brings to the current taxonomic database model (such as, UNITE and BOLD) is that BSGs cluster together

34 25 sequences based on total genetic relatedness and does not rely on any taxonomy for classification. I used the RDP classifier as an alternative to using BLAST to classify query sequences into a BSG. BLAST matches are largely unreliable because the BLAST algorithm was designed to find large local regions of similarity and not designed to identify species. Unfortunately, the BLAST algorithm is often used to identify species (Koski and Golding 2001; e.g. Gharizadeh et al. 2003, Geml et al. 2005: Kõljalg et al. 2005, Öpik et al. 2009). U Ren et al. (2009) write that using the top match almost always misidentified their environmental sequences when compared to a Bayesian analysis. Both Munch et al. (2008) and U Ren et al. (2009) warn that certainty regarding BLAST identification can be over-inflated because only sequences present in are considered. This is a problem since is underpopulated with fungal sequences (Vilgalys 2003). My cross-validation test indicates that the Cortinarius BSG Classifier was effective at assigning the query sequence to the correct BSG 98.8% of the time. However, it did find it difficult to correctly classify sequences of variable length and sequences that do not have a close match in the database. The RDP classifier provides a much higher level of confidence in the classification and has a higher accuracy rate than BLAST (Wang 2007), and it can be customized to fit any taxonomic structure. Being able to customize the taxonomic structure is a vital component of a classifier for BSGs. MEGAN (Metagenome Analyzer) (Huson et al. 2007) is a java computer program that analyzes large environmental DNA datasets and relies on the NCBI taxonomy to summarize and order the results. MEGAN would be unsuitable for identifying unknown Cortinarius sequences as it would not recognize, for example, the genus Dermocybe as a synonym of Cortinarius or that some sequences are simply misidentified. The BSG classifier relies on a statistical probability rather than a percent similarity value for classification. However, it does depend on the sequences being the same length. The BSG classifier developed in this study was designed to be able to accommodate other groups of fungi in the future. Also, it has the potential to be used to recognize different species within the same BSG. Further improvements to the BLAST method of sequence identification were the exclusion of the 5.8S to delimit species, as recommended by Nilsson et al. (2008), as well as the definition of the ITS to between 5 -CATTA and 3 -GACCT. This ensured the greatest possible mean variability was used to delimit species and ensured that all BLAST and RDP Classifier matches were equally comparable. It also ensured that only regions of high species resolution

35 26 were used. BLAST has been known to give faulty results when a region of low species resolution is used (Horton et al. 2008). The 94% Cortinarius BSGs could be helpful in revealing cases of misidentification and mislabeling, in finding novel and cryptic species, in naming environmental sequences and in compiling ecological and geographic data. Misidentification or mislabeling could be detected by investigating the presence of the same specific epithet occurring in multiple BSGs (Table 2). This type of error affects 41% of the named Cortinarius sequences (Figure 6). Fifty-nine percent of the labeled sequences were not found in more than one BSG. However, this does not mean that these sequences are not misidentified as synonyms are present and not all the vouchers are derived from detailed taxonomic studies. Misidentification could also take the form of multiple specific epithets occurring in the same BSG. Cryptic species could be detected by finding the same specific epithet occurring in more than one BSG, or by finding multiple clades within a BSG, when a phylogenetic analysis is performed on the BSG. Four hundred sixty environmental sequences were placed into 110 different BSGs (Appendix 2) and thus were named according to genetic distance. However, as there may be multiple species per BSG, the presence of a voucher specimen does not provide a species level identification. Further phylogenetic analysis may be required. There is great interest in grouping and naming environmental sequences based on similarity to facilitate comparison across studies (Horton et al. 2008). However, as there may be multiple species per BSG, the presence of a voucher specimen does not provide a species level identification. Despite the availability of metadata on environmental sequences, few studies to date have attempted to compile it. However, in order to document the world s biodiversity, it is important to identify and label these genetically distinct clusters of DNA sequences. For example, Soil Clone I represents a novel subphylum of Ascomycota but is only known from environmental sequence data (Porter 2008). BSGs have the potential to make it easier to assimilate data on the geographical range of a species or species complex. However, this is still a laborious task because there is no requirement to submit this kind of data to (Horton et al. 2008). Ryberg et al. (2008) used geographic and ecological metadata of environmental sequences to show that Inocybe is widely distributed and can be found in many climatic regions. Similarly, this study shows that

36 27 Cortinarius is widely distributed and that sampling is biased towards Europe and North America. The host species were predominantly angiosperm and coniferous trees but members of the rosids, Ericaceae and Cyperaceae were also present. The BSGs could be used for studies relating to host specificity and speciation. However, most clades have few representatives and have been collected from few sites. As the number of ITS sequences submitted to continues to climb (Figure 2), whether the sequences be from fruiting bodies or environmental samples, the need to organize this large quantity of sequences will only increase. In the future, BSGs could be used to group and classify all fungi. The technology to do this is already present. s could be classified into BSGs and BSGs could be further investigated based on a phylogenetic approach such as that used by MOR (Hibbett et al. 2005) and UNITE (Kõljalg et al. 2005). New sequences could be automatically retrieved from NCBI (as performed by MOR (Hibbett et al. 2005) and Emerencia (Nilsson et al. 2005)) and then automatically classified into an existing BSG or into a newly created BSG. The use and acceptance of BSGs could lead to great improvements in fungal taxonomy (by the detection of novel species and cryptic species, and by annotating misidentified specimens) and ecology (by grouping similar sequences from different studies together and providing nomenclature to their sequences). In addition we can look forward to improvements in the method of delimiting sequences into groups of genetic similarity (White et al. 2010).

37 28 Chapter 3: General Discussion and Perspectives for Future Studies Taxonomic Delimitation and Classification Mycologists, bacteriologists and entomologists all deal with hyperdiverse groups of organisms, in which only a small fraction of the diversity is known. They frequently encounter difficult taxonomic situations such as having few defining morphological features, cryptic species, and differing morphologies in different life stages. Thus, the use of DNA data is essential in these fields. These researchers all have a vested interest in rapid delimitation and classification using large molecular datasets and have developed a number of different solutions. Coalescent theory, the alignment program Sequencher (Gene Codes, Ann Arbor, Mich.), DOTUR (Schloss and Handelsman 2005), Cd-Hit (Li and Godzik 2006) and XplorSeq (Frank 2008) are all methods used to create operational taxonomic units. MEGAN (Huson et al. 2007), Trichokey (Druzhinina et al. 2005), SAP (Munch et al. 2008), CBCAnalyzer (Wolf et al. 2005), and the RDP Classifier (Wang 2007) are programs that have been developed to classify sequences once a taxonomy and a set of voucher sequences has been established. UNITE (Kõljalg et al. 2005), Emerencia (Nilsson et al. 2005), and MOR (Hibbett et al. 2005) contain programs that can assist in classifying a fungal sequence, but human inspection is still needed. In choosing a method to delimit and classify sequences into BSGs each program was critiqued. Delimitation of Species The nature of the Cortinarius dataset places certain limitations as to what method of delimitation can be employed. First, the method must not be computationally intensive as a large dataset is used. Second, it cannot rely on a method that requires good support of interspecific relationships since the ITS region cannot be relied upon to provide this type of resolution. DOTUR 1.53 (Distance Based OTU and Richness Determination) (Schloss and Handelsman 2005) was used to delimit ITS sequences into BSGs. The program operates by grouping a set of aligned sequences based on a furthest neighbour-joining algorithm. It is currently used in fungal soil ecology studies (Tedersoo et al. 2006, Porras-Alfaro et al. 2007, Allison et al. 2008) as a way of grouping sequences into molecular operational taxonomic units (motus).

38 29 XplorSeq (Frank 2008) is a software environment for the analysis of rdna studies. Its multiple functions include performing BLAST searches of NCBI and local databases, multiple sequence alignments, creation of phylogenetic trees, creation of operational taxonomic units (OTUs) and estimate biodiversity indices. The OTUs are created using a furthest neighbourjoining algorithm and assume a Jules-Cantor model of evolution. Thus, the program is much like DOTUR but with limited parameters that can be optimized. U Ren et al. (2008) used Sequencher (Gene Codes, Ann Arbor, Mich.) to create OTUs for fungal endophytes. In a comparison of methods, they found that DOTUR yielded a higher estimate of species richness than did Sequencher. In a pilot study, I found that Sequencher did not create monophyletic groups and was sensitive to differences in rates of evolution between clades (data not shown). Clades that showed low variability in the ITS, such as the subgenus Dermocybe, formed monophyletic groups at high percentage values, but where split into paraphyletic groups at lower percentage values (data not shown). The composition of clusters changed dramatically across different threshold values. In addition, the algorithm for clustering sequences is proprietary information and is not given publicly. The composition of the clusters changed depending on the size and variability of the dataset. Sequencher could work well to delimit sequences if the dataset is small and if there are large differences between sequences, such as that between genera. However Sequencher cannot be relied upon for a dataset as large and as variable as Cortinarius. The secondary structure of the ITS2 has been proposed as an easy and reliable way to distinguish species (Müller et al. 2007; Coleman 2009, Shultz and Wolf 2009). The presence or absence of compensatory base pair changes (CBCs) has been shown to correlate well with sexual incompatibility across a wide range of eukaryotes (Coleman 2009). While CBCs are not the direct cause of the incompatibility, they seem to indicate that enough time has passed for a reproductive barrier to develop (Müller et al. 2007). Shultz and Wolf (2009) outline a protocol by which the secondary structure can be retrieved in the ITS 2 database, or homologous structures can be found (Shultz et al. 2006, Selig et al. 2008), alignments can be made using 4SALE (Seibel et al. 2006) and phylogenetic trees can be created using ProfDistS (Wolf et al. 2008) and CBCAnalyzer. This method requires a number of different programs, and yet still relies on a good sequence alignment. This method was deemed too time consuming to be employed for the Cortinarius study.

39 30 A number of people have applied coalescent theory to delimit species. Pons et al. (2006) developed a likelihood method to delimit species that predicts the transition between speciation (and extinction) events and population-level (coalescence) processes. This method would be ineffective for the genus Cortinarius as the dataset is too large to perform a maximum likelihood analysis. As well, the ITS region does not provide enough resolution to elucidate interspecific relationships. Abdo and Golding (2007) also used coalescent theory to classify butterfly specimens. Again, their method was more accurate than distance based methods, but was computationally expensive and would not work for the Cortinarius dataset. Cd-Hit (Li and Godzik 2006) is a protein or DNA/RNA clustering program that uses short word filtering, which determines the similarity between two sequences without performing a sequence alignment. It is extremely fast and can handle large databases (Li and Godzik 2006). The cd-hit-est algorithm can be applied to non-intron containing sequences. In this algorithm, the sequences are first sorted in order of decreasing length. The longest sequence forms the first cluster, and then each remaining sequence is compared to the representatives of the existing clusters. If the similarity value is above a certain threshold, it is grouped into the most similar cluster; otherwise the sequence becomes the representative of a new cluster. At 94% similarity, it only took 22 seconds to create 255 clusters. This is much faster than DOTUR which takes over an hour to run. The use of this program to create BSGs should be investigated in the future because it is fast and creates clusters without doing a sequence alignment. In addition, incremental clustering can be used to add new sequences to the clusters or create new clusters without changing the composition of the cluster. This aspect of the program would make it easy to add new sequences to the BSG database. Classification tools When selecting a classification tool, I was looking for particular traits. First, I was looking to customize the taxonomy and not rely on the taxonomy present in. Second, the program had to be fast and accurate. Also, it had to be able to produce a statistic showing the confidence in the classification. MEGAN (Metagenome Analyzer) is a java computer program that analyzes large environmental DNA datasets (Huson et al. 2007). s are compared against a database of known sequences by doing a BLAST search. The program then uses the NCBI taxonomy to

40 31 summarize and order the results. The problem with this method for Cortinarius is that the NCBI taxonomy cannot be trusted. It will not, for instance, recognize that Dermocybe is a synonym of Cortinarius or that some sequences are simply misidentified. The program assigns each read to the lowest common ancestor of the set of taxa that it hit in the comparison. This program works well in that it uses a number of different thresholds to place the taxon in a taxon level. The first threshold is the bit score threshold. Any hit that is below the threshold is discarded. The second threshold is the percent similarity. The program will discard any hits that do not fall below a certain threshold. The third threshold states that there must be a minimum quorum of two sequences to be placed in a taxon level. Another problem with MEGAN is that it relies on a BLAST comparison to come up with the initial stats and the NCBI taxonomy cannot be updated to reflect recent changes. Overall, MEGAN would work if the taxonomy in could be relied upon and if BLAST were accurate enough to classify sequences. However, it is not and therefore did not meet my criteria. Trichokey is an online tool used to identify Trichoderma and Hypocrea spp. (Druzhinina et al. 2005). Trichokey searches a database for diagnostic regions of the sequence that can identify the genus, the section and then the particular species. The problem with applying this approach to Cortinarius is the sheer number of species present in the genus. This method relies on having voucher sequences for all species and it depends on a good alignment to identify species. However, as there is no clear species concept in Cortinarius, and due to the large number of sequences, it would be too much effort to identify species-specific domains. The addition of new Cortinarius sequences could change the size of the species-specific domains and thus this method was deemed to be too labor intensive. The Statistical Assignment Program (SAP) BLASTs the query sequence to the database (or other custom database) to retrieve a number of sequences with high homology, aligns those sequences to each other (and to the unknown sequence), then determines the posterior probability of membership to a particular group using a Markov chain Monte Carlo approach similar to the one commonly used in phylogenetic inference. This method is very effective for small datasets but is impractical for large datasets. I personally found this method to be very time intensive compared to other methods. An aspect about this program that I like is the use of posterior probabilities to provide a level of confidence in the assignment. However, the program is reliant on the sequences in the database and can lead to wrong inferences if the

41 32 database is not representative. This method is more accurate than using BLAST (Munch et al. 2008) and should be considered as a good candidate to classify BSGs. Finally, the Ribosomal Database Project (RDP) Classifier (Wang 2007) fit all the criteria. It is fast, accurate and highly customizable. A review of this classification method can be found in Chapter 2. Fungal identification tools These tools have been developed to correct some of the taxonomic problems in. Emerencia (Nilsson et al. 2005) has a number of features that are attractive to the Cortinarius identification problem. It BLASTs an insufficiently identified sequence against all the identified sequences in. However, it assumes that because a specific epithet is given, it is a good identification, when this may not be the case. Also, it doesn t give any confidence in the BLAST match. It can search for everything with a particular name, such as Cortinarius, and retrieves all the BLAST matches that BLAST to that name. However, it won t be able to exclude a sequence labeled as Cortinarius when it is not. Although Emerencia has features that aid in the identification of unidentified sequences, it lacks the robustness to be confident in the matches Emerencia is able to produce. MOR (Hibbett et al. 2005) automatically downloads identified homobasidiomycete LSU rrna sequences in and aligns them. It uses RAxML to produce a maximum likelihood tree. MOR is ineffective for Cortinarius since the LSU cannot differentiate between species and because MOR screens out environmental sequences. However, automatically importing sequences from and incorporating them into a gigantic phylogenetic tree are favourable aspects of this project that could be used with BSGs. UNITE (User-friendly Nordic ITS Ectomycorrhizal database group; Kõljalg et al. 2005) BLASTs query sequences against and against its own well-curated database. So far, the majority of Cortinarius sequences in UNITE represent European species and do not contain North American or Southern Hemisphere endemics. It also uses a CGI script called galaxie (Nilsson et al. 2004) that places the query sequence within a phylogenetic tree of the BLAST results. As galaxie relies on BLAST, it will not be able to retrieve all the related sequences, only select sequences, depending on the length of the input sequence. However, if galaxie were used to create a phylogeny from sequences in the customized BSG database, where sequences are

42 33 approximately the same size, it would be useful in determining the relationships between different species within each BSG. s How are BSGs an improvement on the current taxonomic database model? The main improvement the Database brings to the current taxonomic database model (such as, UNITE and BOLD) is that BSGs cluster together sequences based on total genetic relatedness and does not rely on any taxonomy for classification. Human taxonomies have the potential to be debated and changed in the future, but the BSGs will stay constant. BOLD will not accept sequences without vouchers but the BSG database is able to accommodate the total fungal diversity, including environmental sequences. Unlike BLAST, the program used to match a query sequence to a voucher in UNITE and, the RDP Classifier is accurate and provides confidence in the classification. The BSG classifier allows one to quickly find all the sequences that are at least 94% similar to the sequence of interest. In contrast, has no one easy way of determining conspecific sequences. If one uses the related sequences link on, phylogenetic analysis needs to be done to determine the relationships between sequences. If one uses a node in the distance tree provided by, one must examine each accession separately. In addition, the BSG classifier provides the metadata associated with each specimen, in a manner that facilitates comparison within and between conspecific sequences. only provides this information within each entry, making it difficult to compare across entries. Thus if one is interested in species level questions, the BSG database makes it easier to classify the query sequence and retrieve conspecific sequences and their associated metadata. How the 94% similarity value was chosen After choosing to use DOTUR to delimit the sequences into BSGs, it was necessary to choose a threshold value. Originally, the criteria for the threshold value was that it should best differentiate between different species. However, after understanding the effect that changing the multiple sequence alignment has on the composition of the OTUs, this perfect boundary between inter and intraspecific variation could not be found (Figure 3) and the criteria was changed to the threshold value that ensures that all members of species are included.

43 34 I attempted to choose a threshold value at which the number of new OTUs dramatically increased, signifying greater variation within species than between species. However, Figure 3 shows that for this alignment, the increase in variation across different threshold values was continuous and thus a value could not be chosen. It was determined there is no one single percentage that could be used to delimit taxa in the genus Cortinarius without either splitting or lumping some species together. The approach I eventually used to choose a delimitation value was to weigh the costs and benefits of performing a type 1 error (splitting a species up into many BSGs) and performing a type 2 error (lumping many species into a single BSG). As the global alignment is of a large number of species, of differing sequence lengths and of differing sequence quality, the likelihood of performing a type 1 error is increased. And, as the contents of each BSG will have to be inspected more comprehensively by a taxonomist anyway, using detailed morphological and multi-gene phylogenies, it was deemed more appropriate to ignore type 2 errors and reduce type 1 errors. Using 391 sequences representing 102 exemplar species, an alignment using MAFFT and the FFT-NS-i method was made (data not shown). These exemplar sequences contained the whole ITS with no ambiguities and were chosen based on the reputation of the authors and the monophyly of the groups under neighbour-joining analysis. When these sequences were input into DOTUR, 94% was the similarity value at which no putative single species were split into multiple BSGs. Seventy-five percent of the species were placed in a single BSG. The other twenty-five percent of the species represented type 2 errors and were lumped with at least one other species in a BSG. Uses of BSGs Finding misidentifications and areas that need taxonomic work Despite the majority of Cortinarius species entered into originating from published taxonomic papers, the presence of the same specific epithet occurring in multiple BSGs (Table 2) reveals that misidentification or mislabeling is highly prevalent. In addition, twelve sequences, identified as Cortinarius, were excluded from this study because they BLASTed to fungal genera other than Cortinarius (Table 1). The BSG database allows one to recognize where these errors occur (by examining the contents of BSGs that have more than one specific epithet or by examining the contents of BSGs where the specific epithet is found in more than one BSG, and by examining the contents of each BSG) and can be used to update the taxonomy of submitted sequences. For example, the name C. elegantior is found in two BSGs.

44 35 On retrieving all the C. elegantior sequences, the accession EU was found to be in BSG 121 while eight other C. elegantior sequences were found to be in BSG 234. A phylogenetic tree of BSG 121 shows that EU is closely related to C. superbus, another member of subgenus Phlegmacium that is morphologically similar to C. elegantior (data not shown). Thus, EU is likely to be misidentified as C. superbus. Phylogenetic methods and inspection of vouchers will be needed to confirm cases of misidentification, but BSGs can be used to find putative misidentifications. Novel species and cryptic species may be detected when one specific epithet is found in multiple BSGs. This may indicate the same species definition being applied to genetically dissimilar entities. For example, the specific epithet vibratilis occurs in four different BSGs. Novel species may also be found in clades that only contain unidentified vouchers. BSG 148 contains 3 unidentified vouchers from North America. However detailed taxonomic and phylogenetic expertise is needed to discern if these are misidentifications or novel species. Other novel species may not be easily detectable due to the low amount of genetic divergence between species. For example, C. comarostaphylii and its sister species C. leucophanes are both in BSG 276. These sequences are 99% similar to each other and only differ by one base pair. Ammirati et al. (2009) interpreted these two morphologically similar species as cryptic species and differentiate between them on the basis of their different hosts and geographical locations. C. leucophanes is found with Pinus sylvestris and Larix sp. in Europe, while C. comarostaphylii is found with Comarostaphylis arbutoides in Costa Rica. Because they are so similar to each other, this BSG database would not be able to differentiate between the species. Classifying environmental and unidentified sequences There is great interest in grouping and naming environmental sequences based on similarity to facilitate comparison across studies (Horton et al. 2008). In this study, 460 environmental sequences were placed into 110 different BSGs (Appendix 2). Of these BSGs, 66 (60%) contained an identified voucher specimen, 9 (8%) contained an unidentified voucher and 35 (32%) contained only environmental sequences (Appendix 3). However, as there may be multiple species per BSG, the presence of a voucher specimen does not provide a species level identification. Further phylogenetic analysis may be required. The compilation of environmental

45 36 sequences that do not have identified voucher specimens is useful. For example, BSG 297 contains unidentified vouchered sequences and environmental sequences collected from seven different coniferous forests located all across North America (Appendix 2). The label, BSG 297, denotes these sequences are at least 94% similar to each other and could now be discussed in a comparative context. Despite the availability of metadata on environmental sequences, few people have attempted to compile it. Revealing cryptic species s could be used to detect cryptic species. This could be diagnosed from the presence of the same specific epithet occurring in more than one BSG, or it could take the form of multiple distinct clades within a BSG following phylogenetic analyses. The contents of some BSGs were compared with some recently studied cryptic species to determine whether BSGs could detect cryptic species. Garnica et al. (2009) studied Cortinarius section Calochroi using a maximum likelihood analysis of the ITS-5.8S + D1/D2 rrna sequences. They found C. arcuatorum clustered into four closely related monophyletic groups (each with 99% or more bootstrap support). There was little phenotypic variation between the specimens but there was a range of variability in spore size between collections from different parts of North America and Europe. Since the different groups differed in their habitat, Garnica et al. (2009) hypothesized that some populations had evolved specialized habitat requirements with allopatric diversification. Garnica et al. (2009) stopped short of describing these groups as different species as these were small collections from specific areas. In my study, C. arcuatorum is split into two different BSGs: BSG 238 and BSG 239, which respectively include sequences from 3 and 1 of the four monophyletic groups detected by Garnica et al. (2009). This shows that my method of creating BSGs is able to detect putative cryptic species. However, the BSGs will not be able to detect cryptic species in all cases. Garnica et al. (2009) also found there was a high level of genetic diversity in C. elegantior as DNA sequences contained several unique substitutions and indels. Garnica et al. (2009) considered C. elegantior var. americanus to represent a separate species from C. elegantior, which differs in the colour of the veil. All these sequences clustered into BSG 234, thus in this case, the BSGs would not be able to reveal these cryptic species.

46 37 Testing hypotheses about speciation The fundamental question of what drives speciation in mycorrhizal fungi has yet to be solved. Two hypotheses are used to explain current species distributions, but neither of these hypotheses have been thoroughly tested. The first hypothesis is that since spore dispersal is poor and since fungi have limited rates of growth, vicariance (e.g. O Donnell et al. 1998b, Hibbett 2001, Nilsson et al. 2003) and/or allopatry (e.g. Anderson et al. 1980, O Donnell et al. 2004) better predict current species distributions. The second hypothesis is that species evolve in sympatry by becoming specially adapted to a particular niche (e.g. Peever 2007). Cortinarius can be used to answer these questions since it is mycorrhizal, has a large number of species and has a worldwide distribution. Many organisms, including Cortinarius, have a disjunct distribution where species from the southern hemisphere do not occur in the northern hemisphere and vice versa (Peintner et al. 2001, 2002). The separation of Gondwana and Laurasia that took place approximately 120 mya is often used to explain this distribution (Raven and Axelrod 1974). However, this Southern Hemisphere endemism typically only occurs at the species level, not the genus level, which provides doubt to this hypothesis (Moncalvo and Buchanan 2008). By dating the divergence of Northern Hemisphere and Southern Hemisphere endemics (found within a BSG) and by using a molecular clock, the genus Cortinarius could be used to test the tectonic plate hypothesis. For example, Moncalvo and Buchanan (2008) used a molecular clock to show that the saprophytic Ganoderma applanatum-australe species complex diversified before the break-up between Gondwana and Laurasia. They provide evidence for episodic events of long-distance dispersal within the southern hemisphere. However, this method has not been used to test the Gondwana/Laurasia hypothesis in an l species. Ecology The advent of rapid DNA sequencing has spurned a growing interest in studying the ecology of fungi in the field (Kõljalg et al. 2005). Projects are underway to create microarrays that use the ITS region to assay different species of fungi across different habitats (Bruns et al. 2008). However, very seldom is this data collected in such a way that could be compared across studies. Compiling data from multiple studies could assist in our understanding of biological processes within a genus. For example, a widely used hypothesis is that ecology and host-

47 38 specificity drive speciation (den Bakker et al. 2004). New species could evolve through switching to a new host (den Bakker et al. 2004) or through adapting to new environmental conditions (Giraud et al. 2008). For example, Chapela and Garbelotto (2004) used the ITS region to show that, in North America, matsutake mushrooms switched from strictly angiosperm hosts to strictly coniferous hosts as the climate changed during the Eocene. Similarly, den Bakker et al. (2004) used the ITS2 and glyceraldehyde 3-phosphate dehydrogenase (Gapdh) in the l genus Leccinum to study host-specificity and evolution. The evolution of a generalist from a specialist was noted. They showed episodes of rapid speciation coincided with or immediately followed host-switches. The relationship between ecology and host-specificity has never been investigated in the genus Cortinarius. Using molecular phylogenetics and a molecular clock, Peintner et al. (unpublished) estimate that there was a single rapid radiation event in Cortinarius that coincides with the radiation of the angiosperms and especially with its main symbiotic partner, the Fagales. BSGs could be used to study rates of speciation between host-specific clades and clades with multiple hosts. However, a more complete sampling of the genus and more complete data on the host species are needed. Future Developments The Classifier is a starting point for grouping together sequences based on their genetic similarity. However, several improvements could be made, particularly in the ability to query the BSG database. Fortunately, the technology to make these improvements has already been developed and only needs to be implemented. The first improvement would be to create a phylogenetic tree of the sequences within each BSG because a single BSG could contain any number of different species with a varying amount of genetic variation. Nilsson et al. (2004) developed a set of CGI scripts, called galaxie, that aligns the query sequence to related species in a local database using Clustal W and presents either a neighbour joining tree (NJ) or a maximum parsimony (MP) tree using Phylip. The output is a phylogenetic tree displayed in the user s browser and a Newick tree file that can be downloaded for further analysis. Another desirable feature would be a CGI script that would display a phylogenetic tree of individual BSGs even in the absence of a query sequence.

48 39 Taxonomists could use this tool to search through BSGs to find misidentified sequences, cryptic species and novel species. A severe limitation of the BSG database is that it does not respond to changes made to entries and it does not automatically update when new sequences are added to. However, this feature could be added. Once a week, MOR (Hibbett et al. 2005) uses BioPerl to connect to, and retrieves new sequences using a Boolean search. The same script could be used to retrieve new Cortinarius sequences. There are two additional challenges that would have to be overcome to conserve accuracy in the BSG database. The first challenge is to filter out the sequences that are misidentified as Cortinarius but really belong to a different genus. The second challenge is to filter out ITS sequences that are too short and to trim the 18S and 25S portions when they are too long. The third, and most difficult challenge is to retrieve those environmental sequences that are not labeled as Cortinarius. When sequences are downloaded from, they are often of differing lengths. The RDP classifier requires sequences to be of roughly equal length. As well, the BLAST output is more accurate when sequences are of equal length. Nilsson et al. (unpublished) produced a Perl script that uses hidden Markov Models (HMM) trained on detecting the ends of the ITS region in a variety of fungal lineages. Once this program enters the public domain, it could be used to trim the entries to the desired length. After retrieving sequences from, sequences should be classified into an existing BSG or a new BSG should be made. As the original BSG creation process depended on an alignment, this process cannot be repeated. It would be possible to have the RDP classifier classify sequences into existing BSGs. However, as the BSG classifier is sensitive to the length of sequence, a poor bootstrap value is not necessarily a sign that a new BSG should be created. Solid criteria need to be developed to define the rules of when a new BSG should be created. As well, a script for creating new BSGs in the MySQL database still needs to be written. If the program Cd-Hit (Li and Godzik 2006) were used to create BSGs, incremental clustering could be used to add new sequences to existing BSGs or create new BSGs if needed. Similar to the problem of automatically adding new sequences from, is the problem of updating all three databases. Presently the MySQL database serves as the hub to create the personal RDP and BLAST databases. However, the subservient databases have

49 40 different FASTA formats. A script would need to be created to update the subservient databases after the MySQL database is updated. Currently, the submission of ecological data to is not required and often is not given. It would be desirable to create an accession entry page where users could upload geographical and ecological data (Horton et al. 2008). Conclusion s, as produced by DOTUR, are effective at clustering sequences of greater or equal to a threshold value of genetic similarity. They can find misidentification or mislabeling, label environmental sequences and match them to voucher sequences, have the potential to find novel and cryptic species, and are an effective way of compiling geographic and ecological data. However, this method is not without its faults. Expert taxonomic work on the contents of each BSG is needed to identify the species within each BSG and to correct misidentifications and mislabeling. This method does not identify species. The small genetic differences within a BSG may represent different species with large ecological differences. Thus BSGs should not be treated as an endpoint species label for environmental sequences, but rather should be treated as a proxy. In the future, BSGs could be used to group and classify all fungi. The technology to do this is already present. s could be classified into BSGs and BSGs could be further investigated based on a phylogenetic approach such as that used by MOR (Hibbett et al. 2005) and UNITE (Kõljalg et al. 2005). New sequences could be automatically retrieved from NCBI (as performed by MOR (Hibbett et al. 2005) and Emerencia (Nilsson et al. 2005) and then automatically classified into an existing BSG or into a newly created BSG. The use and acceptance of BSGs could lead to great improvements in fungal taxonomy (by the detection of novel species and cryptic species, and by annotating misidentified specimens) and ecology (by grouping similar sequences from different studies together and classifying their sequences). In addition we can look forward to improvements in the method of delimiting sequences into groups of genetic similarity (White et al. 2010). The method of using Cd-Hit (Li and Godzik 2006) to create BSGs should be investigated in the future.

50 41 Literature Cited Abdo, Z. and G.B Golding A step towards barcoding life: a model-based, decisiontheoretic method to assign genes to preexisting species groups. Systematic Biology. 56: Adaskaveg, J.E. and R.L. Gilbertson Cultural studies and genetics of sexuality of Ganoderma lucidum and G. tsugae in relation to the taxonomy of the G. lucidum complex. Mycologia. 78: Allison, S.D., Czimczik, C.I. and K.K. Treseder Microbial activity and soil respiration under nitrogen addition in Alaskan boreal forest. Global Change Biology. 14(5): Altschul, S.F., Madden, T.L., Schaffer, A.A., Zhang, J., Zhang, Z., Miller, W., and D.J. Lipman Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Research. 25: Anderson, J.B. and R.C. Ullrich Biological species of Armillaria mellea in North America. Mycologia. 71 (2): Anderson, J.B., Korhonen, K., and R.C. Ullrich Relationships between European and North American biological species of Armillaria mellea. Experimental Mycology. 4: Baptista-Rosas, R.C., Hinojosa, A., and M. Riquelme Ecological Niche Modeling of Coccidioides spp. in Western North American Deserts. Annals of the New York Academy of Sciences : Benson, D.A., Karsch-Mizrachi, I., Lipman, D.J., Ostell, J., and D.L. Wheeler Nucleic Acids Research. 35: D25-D30. Bidartondo, M.I. et al Preserving accuracy in. Science. 319: Bidartondo, M. I., and D. J. Read Fungal Specificity Bottlenecks During Orchid Germination and Development. Molecular Ecology. 17: Bidaud, A., Möenne-Loccoz, P. and P. Reumaux Atlas des Cortinaires. VI. Fédération Mycologique Dauphiné-Savoie. Blackwell, M., Hibbett, D. S., Taylor, J.W. and J.W. Spatafora Research coordination networks: a phylogeny for kingdom Fungi (Deep Hypha). Mycologia. 98: Bougher, N.L., Fuhrer, B.A., and Horak, E Taxonomy and biogeography of Australian Rozites species mycorrhizal with Nothofagus and Myrtaceae. Australian Systematic Biology. 7: Brandrud, T.E., Lindström, H., Marklund, H., Melot, Jacques and S. Muskos Cortinarius: Flora Photographica I (English version). Cortinarius HB, Matfors. Brasser, B.J Natural History in the 21 st Century? Science. 322 (5901):

51 42 Brenner, S.E., Chothia, C. and T.J.P. Hubbard Assessing sequence comparison with reliable structurally identified distant evolutionary relationships. Proceedings of the National Academy of Sciences. 95: Bridge, P.D., Roberts, P.J., Spooner, B.M., and G. Panchal On the unreliability of published DNA sequences.. New Phytologist. 160: Brodo, I.M., Sharnoff, S. D., and S. Sharnoff Lichens of North America. Yale University Press, New Haven. Bruns, T.D ITS reality. Inoculum. 52: 2. Bruns T.D. and R.P. Shefferson Evolutionary studies of l fungi: recent advances and future directions. Canadian Journal of Botany. 82: Case, R.J., Boucher, Y., Dahloff, I., Holstrom, C., Doolittle, W.F. and S. Kjelleberg Use of 16S rrna and rpob genes as molecular markers for microbial ecology studies. Applied and Environmental Microbiology. 73: Chapela, I.O and M. Garbelotto Phylogeography and evolution in matsutake and close allies inferred by analyses of ITS sequences and AFLPs. Mycologia. 96 (4): Chen, M.J., Chen, F.Z., Zhao, B.Y., L. W.Q.L. and F.X. Kong Spatio-temporal variability of microbial eukaryotic community composition in a large shallow subtropical lake, assessed by 18S rdna sequences. Acta Protozoologica. 48: Chillali, M., Idder-Ighili, H., Guillaumin, J.J., Mohammed, C., Escarmant, B.L. and B. Botton Variation in the ITS and IGS regions of ribosomal DNA among the biological species of European Armillaria. Mycological Research. 102: Coleman, A.W Is there a molecular key to the level of biological species in eukaryotes? A DNA guide. Molecular Phylogenetics and Evolution. 50: Crouch, J.A., Clarke, B.B., and B.I. Hillman What is the value of ITS sequence data in Colletotrichum systematics and species diagnosis? A case study using the falcate-spored graminicolous Colletotrichum group. Mycologia. 101: den Bakker, H.C., Zuccarello, G.C, Kuyper, T.W., and M.E. Nooredeloos Evolution and host-specificity in the l genus Leccinum. New Phytologist 163: DeSalle, R., Egan, M.G. and M. Sidall The unholy trinity: taxonomy, species delimitation, and DNA barcoding. Philosophical Transactions of the Royal Society B. 360: Druzhinina, I.S., Kopchinskiy, A.G., Komón, M., Bissett, J., Szakacs, G., and C.P. Kubicek An oligonucleotide barcode for species identification in Trichoderma and Hypocrea. Fungal Genetics and Biology. 42: Fisher, M.C., Garner, T.W., and S.F. Walker. Global emergence of Batrachochytrium dendrobatidis and amphibian chytridiomycosis in space, time, and host Annual Review of Microbiology. 63: Floyd, R., Abebe, E., Papert, A., and M. Blaxter Molecular barcodes for soil nematode identification. Molecular Ecology. 11: Frank, D.N XplorSeq: A software environment for integrated management and phylogenetic analysis of metagenomic sequence data. BMC Bioinformatics. 9: 420.

52 43 Frisvad, J.C Physiological criteria and myco-toxin production as aids in identification of common asymmetric penicillia. Applied and Environmental Microbiology. 43: Frisvad, J.C., Filtenborg, O., Samson, R.A. and A.C. Stolk Chemotaxonomy of the genus Talaromyces. Antonie van Leeunhoek. 57: Frøslev, T.G., Jeppesen, T.S., Læssøe, T. and R.K. Frøslev Molecular phylogenetics and delimitation of species in Cortinarius section Calochroi (Basidiomycota, Agaricales) in Europe. Molecular Phylogenetics and Evolution. 44(1): Frøslev, T.G., Matheny, P.B., and D.S. Hibbett Lower level relationships in the mushroom genus Cortinarius (Basidiomycota, Agaricales): A comparison of RPB1, RPB2, and ITS phylogenies. Molecular Phylogenetics and Evolution. 37(2): Ganley, A.R.D., and T. Kobayashi Highly efficient concerted evolution in the ribosomal DNA repeats: Total rdna repeat variation revealed by whole-genome shotgun sequence data. Genome Research. 17: Gardes, M., White, T.J., Fortin, J.A., Bruns, T.D. and J.W. Taylor Identification of indigenous and introduced symbiotic fungi in e by amplification of nuclear and ribosomal DNA. Canadian Journal of Botany. 69: Gardes, M. and T.D. Bruns ITS primers with enhanced specificity for Basidiomycetes application to the identification of mycorrhizae and rusts. Molecular Ecology. 2: Gargas, A., Trest, M.T., Christensen, M., Volk, T.J. and D.S. Blehert Geomyces destructans sp. nov. associated with bat white-nose syndrome. Mycotaxon. 108: Garnica, S., Weiß, M., Oertel, B., Ammirati, J. and F. Oberwinkler Phylogenetic relationships in Cortinarius, section Calochroi, inferred from nuclear DNA sequences. BMC Evolutionary Biology. 9: Geml, J., Via, Z., Long, J. Huston, S., Marr, T., and D.L. Taylor The fungal metagenomics project. < Gharizadeh, B., Norberg, E., Löffler, J., Jalal, S., Tollemar, J., Einsele, H., Klingspor, L. and P. Nyrén Identification of medically important fungi by the pyrosequencing technology. Mycoses. 47: Gibson, C. M. and M.S. Hunter Extraordinarily widespread and fantastically complex: comparative biology of endosymbiotic bacterial and fungal mutualists of insects. Ecology Letters. 13 (2): Giraud, T., Refrégier, G., Gac, M.L., de Vienne, D.M. and M.E. Hood Speciation in Fungi. Fungal Genetics and Biology. 45: Goldstein, P.Z. and R. DeSalle Phylogenetic species, nested hierarchies, and character fixation. Cladistics. 16: Gromov, B.V., Mamkaeva, K.A., and A.V. Pjusch Cultures of chytrids Parasites of algae in Biological Institute of Petersburg University. Phycology. 36: 26. Harrington, T.J. and D.T. Mitchell Colonization of root systems of Carex flacca and C. pilulifera by Cortinarius (Dermocybe) cinnamomeus. Mycological Research. 106:

53 44 Hawksworth, D.L The magnitude of fungal diversity: the 1.5 million species estimate revisited. Mycological Research. 105: Hebert, P.D.N., Cywinska, A., Ball, S.L. and J.R. dewaard Biological identifications through DNA barcodes. Proceedings of the Royal Society B. 270: Hershkovitz, M.A., and L.A. Lewis Deep-level diagnostic value of the rdna-its region. Molecular Biology and Evolution. 13: Hibbett, D.S Shiitake mushrooms and molecular clocks: historical biogeography of Lentinula. Journal of Biogeography. 28: Hibbett, D.S., Nilsson, R.K., Snyder, M., Fonseca, M., Costanzo, J., and M. Shonfeld Automated Phylogenetic Taxonomy: An Example in the Homobasidiomycetes (Mushroom-Forming Fungi). Systematic Biology. 54: Hillis, D.M. and M.T. Dixon Ribosomal DNA: Molecular evolution and phylogenetic inference. Quarterly Review of Biology. 66: Horton T. R., and T.D. Bruns The molecular revolution in l ecology: peeking into the black-box. Molecular Ecology. 10: Horton, T.R., Arnold, E. and T.D. Bruns FESIN workshops at ESA the mycelial network grows. Mycorrhiza. 19(4): Huson, D.H., Auch, A.F., Qi, J., and S.C. Schuster MEGAN analysis of metagenomic data. Genome Research. 17: James, T.Y., Kauff, F., Schoch, C.L., Matheny, P.B., Hofstetter, V., Cox, C.J., Celio, G., Gueidan, C., Fraker, E., Miadlikowska, J., Lumbsch, H.T., Rauhut, A., Reeb, V., Arnold, A.E., Amtoft, A., Stajich, J.E., Hosaka, K., Sung, G.H., Johnson, D., O'Rourke, B., Crockett, M., Binder, M., Curtis, J.M., Slot, J.C., Wang, Z., Wilson, A.W., Schuszler, A., Longcore, J.E., O'Donnell, K., Mozley-Standridge, S., Porter, D., Letcher, P.M., Powell, M.J., Taylor, J.W., White, M.M., Griffith, G.W., Davies, D.R., Humber, R.A., Morton, J.B., Sugiyama, J., Rossman, A.Y., Rogers, J.D., Pfister, D.H., Hewitt, D., Hansen, K., Hambleton, S., Shoemaker, R.A., Kohlmeyer, J., Volkmann-Kohlmeyer, B., Spotts, R.A., Serdani, M., Crous, P.W., Hughes, K.W., Matsuura, K., Langer, E., Langer, G., Untereiner, W.A., Lucking, R., Budel, B., Geiser, D.M., Aptroot, A., Diederich, P., Schmitt, I., Schultz, M., Yahr, R., Hibbett, D.S., Lutzoni, F., McLaughlin, D.J., Spatafora, J.W., and R. Vilgalys Reconstructing the early evolution of Fungi using a six-gene phylogeny. Nature. 443: Katoh, K., Kuma, K., Toh, H., and T. Miyata MAFFT version 5: improvement in accuracy of multiple sequence alignment. Nucleic Acids Research. 33: Kirk, P.M., Cannon, P.F., Minter, D.W., and J.A. Stalpers Dictionary of the Fungi 10th edition. CAB International, Wallingford. Kress, W.J. and D.L. Erickson DNA Barcodes: Genes, genomics and bioinformatics. Proceedings of the National Academy of Sciences USA. 105: Kohavi, R A study of cross-validation and bootstrap for accuracy estimation and model selection. Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence. 2:

54 45 Kõljalg, U., Larsson, K.H., Abarenkov, K., Nilsson, R.H., Alexander, I.J., Eberhardt, U., Erland, S., Hoiland, K., Kjoller, R., Larsson, E., Pennanen, T., Sen, R., Taylor, A.F.S., Tedersoo, L., Vralstad, T., and B.M. Ursing UNITE: a database providing web-based methods for the molecular identification of l fungi. New Phytologist. 166: Koski, L.B. and G.B. Golding The closest BLAST bit is often not the nearest neighbour. Journal of Molecular Evolution. 52: Lee, S.B. and J.W. Taylor Phylogeny of five fungus-like protoctistan Phytophthora species, inferred from the internal transcribed spacers of ribosomal DNA. Molecular Biology and Evolution. 9: Li, W. and A. Godzik Cd-Hit: A fast program for clustering sequences and comparing large sets of protein or nucleotide sequences. Bioinformatics. 22: MAFFT Version 6: Algorithms and parameters. Accessed March 1, Matheny, P.B., Wang, Z., Binder, M., Curtis, J.M., Lim, Y.W., Nilsson, R.K., Hughes, K.W., Hofstetter, V., Ammirati, J.F., Schoch, K.L, Langer, E., Langer, G., McLaughlin, D.J., Wilson, A.W., Froslev, T., Ge, Z.W., Kerrigan, R.W., Slot, J.C., Yang, Z.L. Baroni, T.J., Fischer, M., Hosaka, K., Matsuura, K., Seidl, M.T., Vauras, J., and D.S. Hibbett Contributions of rpb2 and tef1 to the phylogeny of mushrooms and allies (Basidiomycota, Fungi). Molecular Phylogenetics and Evolution. 43: Mayr, E Systematics and the Species from the Viewpoint of a Zoologist. Columbia University Press, New York. McLaughlin, D.J., Hibbett, D.S., Lutzoni, F., Spatafora, J.W. and R. Vilgalys The search for the fungal tree of life. Trends in Microbiology. 17(11): McMullen, M., Jones, R. and D. Gallenberg Scab of wheat and barley: A re-emerging disease of devastating impact. Plant Disease. 81 (12): Meyer, C.P., and G. Paulay DNA barcoding: Error rates based on comprehensive sampling. PloS Biology. 3(12): e422. Mehdiabadi, N.J. and T.R. Shultz Natural history and phylogeny of fungus-farming ants (Hymenoptera: Formicidae: Myrmicine: Attini). Myrmecological News. 13: Moncalvo, J.M Molecular systematics major fungal phylogenetic groups and fungal species concepts. Pages 1-33 in Evolutionary Genetics of Fungi (J.P. Xu, editors). Horizon Scientific Press, Norfolk, U.K. Moncalvo, J.M. and P.K. Buchanan Molecular evidence for long distance dispersal across the Southern Hemisphere in the Ganoderma applanatum-australe species complex (Basidiomycota). Mycological Research. 112: Moser, M.M Die Röhrlinge und Blätterpilze. In Gams, H. (Ed.). Kleine Kryprogamenflora, 2b/2 (5 th ed., pp. 533). Stuttgart: Gustav Fischer Verlag. Moser, M.M., and E. Horak Cortinarius und nahe verwandte Gattungen in Südamerica. Beihefte zur Nova Hedwigia. 52:

55 46 Muhlmann, O. and U. Peintner Ectomycorrhiza of Kobresia myosuroides at a primary successional glacier front. Mycorrhiza. 18: Müller, T., Philippi, N., Dandekar, T., Shultz, J. and Wolf, M Distinguishing species. RNA. 13: Munch, K., Boomsa, W., Huelsenbeck, J.P., Willerslev, E., and R. Nielsen Statistical assignment of DNA sequences using Bayesian Phylogenetics. Systematic Biology. 57: Nilsson, R.H. et al. ITS1/ITS2 extractor for fungal ITS sequences. Unpublished. Nilsson, R.H., Hallenburg, N., Norden, B., Maekawa, N., and S.H. Wu Phylogeography of Hyphoderma setigerum (Basidiomycota) in the northern hemisphere. Mycologia. 107: Nilsson, R.H., Larsson, K.H., and B.M. Ursing galaxie---cgi scripts for sequence identification through automated phylogenetic analysis. Bioinformatics. 20 (9): Nilsson, R.H., Kristiansson, E., Ryberg, M., and K.H. Larsson Approaching the taxonomic affiliation of unidentified sequences in public databases an example from the mycorrhizal fungi. BMC Bioinformatics. 6: 178. Nilsson, R.H., Rhyberg, M., Kristiansson, E., Abarenkov, K., Larsson, K.H. and U. Koljalg Taxonomic reliability of DNA sequences in public sequence databases: a fungal perspective. PLoS ONE. 1: e59 doi: /journal.pone Nilsson, R.H., Kristiansson, E., Ryberg, M., Hallenberg, N., and K.H. Larsson Intraspecific ITS Variability in the Kingdom Fungi as Expressed in the International Databases and its Implications for Molecular Species Identification. Evolutionary Bioinformatics. 4: Niskanen, T., Liimatainen, K. and I. Kytövuori Cortinarius sect. Brunnei (Basidiomycota, Agaricales) in North Europe. Mycological Research. 113: O Donnell, K., Cigelnik, E., and H.I. Nirenberg. 1998a. Molecular systematics and phylogeography of the Gibberella fujikori species complex. Mycologia. 90: O Donnell, K., Kistler, H.C., Cigelnik, E., and R.C. Ploetz. 1998b. Multiple evolutionary origins of the fungus causing Panama disease of banana: concordant evidence from nuclear and mitochondrial gene genealogies. Proceedings of the National Academy of Sciences. 95: O Donnell, K., Ward, T.J., Geiser, D.M., Kistler, H.C., and T. Aoki Genealogical concordance between the mating type locus and seven other nuclear genes supports formal recognition of nine phylogenetically distinct species within the Fusarium graminearum clade. Fungal Genetics and Biology. 41: Öpik, M., Metsis, M., Daniell, T.J., Zobel, M., and M. Moora Large-scale parallel 454 sequencing reveals host ecological group specificity of arbuscular mycorrhizal fungi in a boreonemoral forest. New Phytologist. 184: Ownley, B., Gwinn, K. and F. Vega Endophytic fungal entomopathogens with activity against plant pathogens: ecology and evolution. Biocontrol. 55(1):

56 47 Peever, T Role of host specificity in the speciation of Ascochyta pathogens of cool season legumes. European Journal of Plant Pathology. 119: Pegler, D. N Taxonomy, nomenclature and description of Armillaria. Pages in Armillaria Root Rot : biology and control of honey fungus (R. T. V. Fox, editor). Intercept, Andover. Peintner, U., Bougher, N.L., Castellano, M.A., Moncalvo, J.M., Moser, M.M., Trappe, J.M., and R. Vilgalys Multiple Origins of Sequestrate Fungi related to Cortinarius (Cortinariaceae). American Journal of Botany. 88: Peintner, U., Horak, E., Moser, M.M., and R. Vilgalys Phylogeny of Rozites, Cuphocybe and Rapacea inferred from ITS and LSU rdna sequences. Mycologia. 94: Peintner, U., Moncalvo, J. M. and R. Vilgalys Toward a better understanding of the infrageneric relationships in Cortinarius (Agaricales, Basidiomycota). Mycologia. 96: Peintner, U., Moncalvo, J. M., Moser, M. and R. Vilgalys. Rapid and Ancient Radiation of the Ectomycorrhizal Mushroom Genus Cortinarius (Agaricales, Basidiomycota) as indicated by Multiple-Gene Phylogenies. Unpublished. Peterson, S.W. and C.P. Kurtzman Ribosomal RNA sequence divergence among sibling species of yeasts. Systematic and Applied Microbiology. 14: Pitt, J.I An appraisal of identification methods for Penicillium species: Novel taxonomic criteria based on temperature and water relations. Mycologia. 65: Pons, J., Barraclough, T.G., Gomez-Zurita, J., Cardoso, A., Duran, D.P., Hazell, S., Kamoun, S., Sumlin, W.D., and A.P. Vogler based delimitation for the DNA taxonomy of undescribed insects. Systematic Biology. 55: Porras-Alfaro, A., Herrera, J., Natvig, D. O and R.L. Sinsabaugh Effect of long-term nitrogen fertilization on mycorrhizal fungi associated with a dominant grass in a semiarid grassland. Plant and Soil. 296: Porter, T.M., Schadt, C.W, Rizvi, L., Martin, A.P., Schmidt, S.K., Scott-Denton, L., Vilgalys, R., and J.M. Moncalvo Widespread occurrence and phylogenetic placement of a soil clone group add a prominent new branch to the fungal tree of life. Molecular Phylogenetics and Evolution. 46: Posada, D. and K. A. Crandall Modeltest: testing the model of DNA substitution. Bioinformatics. 14 (9): Ratnasingham, S. and P.D.N. Hebert BOLD: The Barcode of Life Datasystem ( Molecular Ecology Notes. 7 (3): Robinson, C.H Cold adaptation in arctic and antarctic fungi. New Phytologist. 151(2): Roesch, L.F.W., Fulthorpe, R.R., Riva, A., Casella, G., Hadwin, A.K.M., Kent, A.D., Daroub, S.H., Camargo, F.A.O., Farmerie, W.G. and E.W. Triplett Pyrosequencing enumerates and contrasts soil microbial diversity. The ISME Journal. 1:

57 48 Ryberg, M., Nilsson, R.H. Kristiansson, E., Töpel, M., Jacobsson, S., and E. Larsson Mining metadata from unidentified ITS sequences in : A case study in Inocybe (Basidiomycota). BMC Evolutionary Biology. 8: 50. Savolainen, V., Cowan, R.S., Vogler, A.P., Roderick, G.K. and R. Lane Towards writing the encyclopedia of life: and introduction to DNA barcoding. Philosophical Transactions of the Royal Society B. 360: Seifert, K.A., Samson, R.A., dewaard, J.R., Houbraken, J., Levesque, C.A., Moncalvo, J.M., Louis-Seize, G. and P.D.N. Hebert Prospects for fungus identification using CO1 DNA barcodes, with Penicillium as a test case. Proceedings of the National Academy of Sciences USA. 104: Seifert, K.A Progress towards DNA barcoding of fungi. Molecular Ecology Resources. 9: Selig, C., Wolf, M., Müller, T., Dandekar, T., and J. Schultz The ITS2 Database II: homology modelling RNA structure for molecular systematics. Nucleic Acids Research. 36: D377 D380. Schadt, C.W., Martin, A.P., Lipson, D.A. and S.K. Schmidt Seasonal dynamics of previously unknown fungal lineages in tundra soils. Science. 301: Schloss, P.D., and J. Handelsman Introducing DOTUR, a computer program for defining operational taxonomic units and estimating species richness. Applied and Environmental Microbiology. 71: Schmidt, S.K., Wilson, K.L., Meyer, A.F., Porter, T.M., Schadt, C.W., and J.M. Moncalvo The missing fungi New insights from culture-independent molecular studies of soil. Pages in Accessing Uncultivated Microorganisms (K. Zengler, editor). American Society for Microbiology Press, Washington, DC. Seibel, P.N., Müller, T., Dandekar, T., Shultz, J., and M. Wolf SALE-a tool for synchronous RNA sequence and secondary structure alignment and editing. BMC Bioinformatics. 7: 498. Singer R The Agaricales in modern taxonomy. 4th ed. Koeltz Scientific Books, Koenigstein, Germany. Skouboe, P., Frisvad, J.C., Taylor, J.W., Lauritsen, D., Boysen, M., and L. Rossen Phylogenetic analysis of nucleotide sequences from the ITS region of terverticillate Penicillium species. Mycological Research. 103: Sonjak, S., Frisvad, J.C. and N. Gunde-Cimerman Fingerprinting using extrolite profiles and physiological data shows sub-specific groupings of Penicillium crustosum strains. Mycological Research. 113: Swofford, D.L PAUP*. Phylogenetic Analysis Using Parsimony (*and Other Methods). Version 4.0b10. Sinauer Associates, Sunderland, Massachusetts. Taylor, J.W., Jacobson, D.J., Kroken, S., Kasuga, T., Geiser, D.M., Hibbett, D.S., and M.C. Fisher Phylogenetic species recognition and species concepts in fungi. Fungal Genetics and Biology 31:

58 49 Tedersoo, L., Suvi, T., Larsson, E., and U. Kõljalg Diversity and community structure of l fungi in a wooded meadow. Mycological Research. 110(6): Tuomela, M., Vikman, M., Hatakka, A., and M. Itavaara Biodegradation of lignin in a compost environment: a review. Bioresource Technology. 72 (2): Ullrich, R. C., and J. B. Anderson Sex and diploidy in Armillaria mellea. Experimental Mycology. 2: U Ren, J. M., Dalling, J. W., Gallery, R. E., Maddison, D. R., Davis, E. C., Gibson, C. M. and A. E. Arnold Diversity and Evolutionary Origins of Fungi Associated with Seeds of a Neotropical Pioneer Tree: a Case Study for Analysing Fungal Environmental Samples. Mycological Research. 113: Van Valen, L Ecological species, multispecies, and oaks. Taxon 25, Vialle, A., Feau, N., Allaire, M., Didukh, M., Martin, F., Moncalvo, J.M. and R.C. Hamelin Molecular Ecology Resources. 9: Vilgalys, R Taxonomic Misidentification in Public DNA Databases. New Phytologist. 160 (1): 4-5. Voyles, J., Young, S, Berger, L., Campbell, C., Voyles, W.F., Dinudom, A., Cook, D., Webb, R., Alford, R.A., Skerratt, L.F., and R. Speare Pathogenesis of chytridiomycosis, a cause of catastrophic amphibian declines. Science. 326: Wang, B. and Y.L. Qiu Phylogenetic distribution and evolution of mycorrhizas in land plants. Mycorrhiza. 16 (5): Wang, Q Naïve Bayesian Classifier for Rapid Assignment of rrna s into the New Bacterial Taxonomy. Applied and Environmental Microbiology. 73(16): White, J.R., Navlakha, S., Nagarajan, N., Ghodsi, M.R., Kingsford, C. and M. Pop Alignment and clustering of phylogenetic markers implications for microbial diversity studies. BMC Bioinformatics. 11: 152. Wolf, M., Friedrich, J., Dandekar, T., and T. Müller CBCAnalyzer: Inferring phylogenies based on compensatory base changes in RNA secondary structures. In Silico Biology. 5: Wolf, M., Ruderisch, B., Dandekar, T. and T. Müller ProfdistS: (Profile-) Distance based phylogeny on sequence-structure alignments. Bioinformatics. 24: Woese, C.R. and G.E. Fox Phylogenetic Structure of the Prokaryotic Domain: The Primary Kingdoms. Proceedings of the National Academy of Sciences. 74 (11): Wright, S.H.A., Berch, S.M., and M.L. Berbee The effect of fertilization on the belowground diversity and community composition of l fungi associated with western hemlock (Tsuga heterophylla). Mycorrhiza. 19: Zimmerman, R Biologists struggle to solve bat deaths. Nature. 324 (5931):

59 50 Figures Figure 1: Steps taken to create the BSGs.

60 Figure 2: Accumulation of Cortinarius ITS sequences in over time. 51

61 52 Figure 3: The number of s created by DOTUR at different similarity values. Values were calculated with the 5.8S removed.

62 Figure 4: Rarefaction curve showing no saturation in the number of BSGs that were created. 53

63 54 Figure 5. Maximum parsimony bootstrap consensus tree of a cluster of 168 Cortinarius sequences with 80% similarity.

64 Figure 5. (continued) 55

65 56 Figure 6: The proportion of s that contain one or more different specific epithets. The number of epithets found within a BSG ranged from zero to ten. A BSG that contains no specific epithets contains only environmental and/or unidentified voucher specimens.

66 57 Figure 7: The percent of labeled sequences whose specific epithets are found in 1, 2, 3, 4, 5 or 6 different BSGs compared to the number of all the labeled sequences.

67 58 Tables Table 1: s that were removed from analysis because they BLASTed to genera other than Cortinarius. taxonomy Best BLAST match Percent Similarity Notes AB mycorrhizal basidiomycete AB Laccaria laccata 99% AB Cortinarius FJ Atheliaceae sp. 90% AF Hymenogaster alnicola EU Hymenogaster tener 96% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius AF Hymenogaster tener AY Naucoria escharoides 91% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius AF Hymenogaster glacialis AY Hymenogaster rubyensis 100% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius AF Hymenogaster sp. AF Hymenogaster subalpinus 99% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius AF Hymenogaster griseus EU Hymenogaster griseus 97% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius AF Hymenogaster populetorum EU Hymenogaster griseus 97% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius AF Hymenogaster parksii AF Hymenogaster subalpinus 99% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius AF Hymenogaster gardneri AF Hymenogaster subalpinus 97% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius

68 59 taxonomy Best BLAST match Percent Similarity Notes AF Hymenogaster subalpinus AF Hymenogaster parksii 99% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius AF Hymenogaster bulliardii AF Hymenogaster olivaceus 99% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius AF Hymenogaster olivaceus AF Hymenogaster bulliardii 99% More related to Hebeloma, Alnicola, Naucoria than to Cortinarius AF Ectomycorrhizal root tip EF Meliniomyces variabilis 99% AJ Cortinarius caerulescens AY Bolbitius lacteus 99% AY (Cortinarius) FJ Atheliaceae sp. 90% AY (Cortinarius) AY Inocybe sp. 98% sequence too short to know AY mycorrhizal fungus CORTICI1 EF Atheliaceae sp. 88% unique sequence close to Piloderma and Tylospora AY (Cortinarius) EU Inocybe cf. soriora 83% dfmo1803 Hygrocybe dfmo1808 Hygrocybe dfmo1813 Hygrocybe dfmo1814 Hygrocybe dfmo2238 Russula

69 60 taxonomy Best BLAST match Percent Similarity Notes dfmo4704 Bolete DQ Cortinarius callisteus GQ Laccaria proxima 99% DQ Cortinarius vibratilis DQ Hygrophorus eburneus 100% DQ soil fungus clone GQ Laccaria proxima 98% DQ soil fungus clone GQ Laccaria proxima 99% DQ soil fungus clone GQ Laccaria proxima 99% DQ soil fungus clone GQ Laccaria proxima 99% DQ soil fungus clone GQ Laccaria proxima 99% DQ soil fungus clone GQ Laccaria proxima 99% DQ (Cortinarius) AY Cenococcum geophilum 97% EU Fungal sp. MR36 AF Laccaria laccata 98% EU (Cortinarius) EF Cadophora finlandica 97% EU Cortinarius sp. AM Inocybe ochroalba 99% EU (Cortinarius) EU Inocybe geophylla 100% FJ (Cortinarius) FJ Atheliaceae sp. 90%

70 61 Table 2: Specific epithets that were found in more than one (BSG) and the total number of BSGs they were found in. of Barcode Similarity s Specific Epithets 2 albovestitus, aprinus, arcuatorum, armeniacus, armillatus, arquatus, aurora, badiovinaceus, balaustinus, balteatoalbus, balteatus, barbarorum, bigelowii, bovinus, cacaocolor, caesiocanescens, californicus, camphoratus, catharinae, cinnamomeus, clandestinus, corrugatus, cotoneus, cupreorufus, delibutus, duracinus, elegantior, elegantissimus, evernius, flavobulbus, gentilis, glaucopus, hemitrichus, humidicola, humolens, idahoensis, junghuhnii, laetus, langei, laniger, limonius, muscigenus, ochrophyllus, odorifer, olearioides, parasuaveolens, phoeniceus, platypus, raphanoides, renidens, rufo-olivaceus, sanguineus, saturninus, scandens, spectabilis, spilomeus, stillatitius, subarquatus, subcastanella, subfoetidus, sulphurinus, talus, torvus, traganus, venetus, verrucisporus 3 anomalus, argutus, azureus, cedretorum, claroflavus, corrosus, croceocaeruleus, fulmineus, magellanicus, mucosus, multiformis, mutabilis, nanceiensis, percomis, rapaceus, salor, sodagnitus, violaceus 4 alboviolaceus, allutus, barlowensis, flexipes, infractus, magnivelatus, vibratilis 5 brunneus, calochrous, obtusus, semisanguineus 6 acutus

71 62 Appendices Appendix 1: A BLAST search showing the problems associated one encounters. There are unidentified, misidentified and uncultured sequences. It is not apparent from the BLAST search that the first 9 sequences all belong to the same species.