How to know the fungi: combining field inventories and DNA‐barcoding to document fungal diversity

The fungi kingdom is among the most diverse eukaryotic lineages on Earth with estimates of several million extant species (O'Brien et al., 2005; Blackwell, 2011; Taylor et al., 2014). Fungi play critical roles in carbon and nutrient cycling of terrestrial and aquatic ecosystems, and they are important pathogens and mutualists (Read & Perez-Moreno, 2003; Taylor et al., 2012; Grossart et al., 2016). More than 80% of plant species form symbioses with fungi and these symbioses have been crucial to the colonization of terrestrial ecosystems (Field et al., 2015a; Selosse et al., 2015). Despite their impacts on primary ecosystem functions, assessments of fungal biodiversity estimate that only c. 10% of fungal species have been described (Bass & Richards, 2011; Hibbett et al., 2011). Traditionally, specimen-based taxonomic studies have been the only way to discover new species. Because most fungi have microscopic life-stages and convergent morphological features (Rivas-Plata & Lumbsch, 2011; Wynns, 2015), many fungal groups remain severely undersampled. DNA-barcoding and high-throughput sequencing methods have provided a new framework for studying fungal biodiversity (Fierer et al., 2012; Schoch et al., 2012; Myrold et al., 2014), and diversity estimates based on environmental sequences have increased exponentially. Although these 'sequence-based classification and identification' methods are a powerful means to rapidly detect hidden diversity, careful interpretation of these data is needed to make accurate inferences (Kõljalg et al., 2013; Lindahl et al., 2013; Nguyen et al., 2015; Hibbett et al., 2016). In particular, many environmental sequences cannot be associated with a known fungal species or lineage. This remains a major challenge to decipher fungal community composition and understand ecological roles of fungi in leaf litter, soil, or inside plants (Yahr et al., 2016). In some cases, these fungi are truly undescribed and their ecological roles are unknown but in other cases they represent described taxa for which no sequence is available (Nagy et al., 2011; Nilsson et al., 2016). DNA barcoding of herbarium specimens and culture collections is extremely valuable to link unidentified sequences to known taxa (e.g. Brock et al., 2009; Nagy et al., 2011; Osmundson et al., 2013; Garnica et al., 2016). DNA sequences have been generated from fungal type specimens > 200 years old (Larsson & Jacobsson, 2004), but in many cases obtaining sequences from historical material is challenging (Dentinger et al., 2010). Today's threats to biodiversity from habitat loss and climate change are occurring at an unprecedented scale, and it is possible that many species may become extinct before they have been discovered (Costello et al., 2013; Monastersky, 2014). In the need to describe and protect as many species as possible we addressed the following questions: what are the best methods to rapidly document fungal biodiversity? Are traditional, specimen-based approaches still useful? The Southern Hemisphere harbors many unique fungal lineages that are absent from the Northern Hemisphere (Tedersoo & Smith, 2013; Tedersoo et al., 2014). In southern South America, recent studies based on environmental sequences have detected several previously unknown fungal lineages, thereby demonstrating that fungal diversity is probably much higher than presently known (Nouhra et al., 2013; Geml et al., 2014; Roy et al., 2017). As part of a project investigating ectomycorrhizal (ECM) fungi of southern South America, our team collected 1430 fungal fruiting bodies during four collecting expeditions, equaling c. 50 d with 3–4 collectors per day (Supporting Information Methods S1). We primarily collected ECM fungi in temperate forests dominated by Nothofagaceae but also opportunistically collected nonECM fungi. Vouchered specimens were photographed and dried for future study. Internal transcribed spacer rDNA sequences (ITS, e.g. ITS1-5.8S-ITS2) were obtained from a representative selection of 957 specimens using the Extract-N-Amp Plant kit (Sigma-Aldrich) for rapid DNA extraction and amplification. ITS sequences were clustered into operational taxonomic units (OTUs) at 97–99% similarity against the UNITE 'species hypothesis' dynamic database (Abarenkov et al., 2010a; Kõljalg et al., 2013) using Qiime 1.9.1 (Caporaso et al., 2010). Sequences that did not correspond to an existing 'species hypothesis' in the reference database were subsequently clustered de novo at 97% similarity (Methods S1). One representative sequence per OTU was subsequently compared to UNITE+INSD (UNITE and the International Nucleotide Sequence Databases) using megaBlast on the PlutoF workbench (Abarenkov et al., 2010b). We generated 439 OTUs (Table S1), of which 308 (c. 70%) did not match the UNITE dynamic database at 97–99% similarity, and thus did not correspond to any of the 'species hypothesis' currently in UNITE. The efforts of our research group were modest when compared to the high volume and novelty of the data. Although most of these ITS sequences were generated by our research team over a two-year period, they correspond to c. 1.5% of the total diversity in UNITE (19 698 representative sequences in v.7.1, August 2016). Comparisons between our dataset and the full UNITE database (which also includes singleton sequences) did not alter the number of new OTUs we detected. For comparison, in the one-year period before this study (between August 2015 and August 2016) the global efforts of the scientific community contributed only 360 unique 'species hypotheses' to the UNITE database. Working with fresh specimens was extremely efficient. The Extract-N-Amp method rapidly generated ITS sequences with a success rate of c. 80% (including extraction, amplification and sequencing). The success rate improved to c. 90% when partial ITS sequences (e.g. ITS1 + 5.8S only) were included. Success rate of DNA sequencing varies among fungal groups, according to the age of specimens, and based on how they were preserved. Sequencing from type specimens and important historical collections remains the gold standard to link DNA sequences to species names (e.g. Liimatainen et al., 2014; Sánchez-García et al., 2014; Schoch et al., 2014). However, generating sequence data from historical vouchers may be challenging in some groups (25–50% success rate for specimens > 10 years old according to Dentinger et al., 2010). Working with old herbarium specimens is also more time-consuming and sensitive to contamination. It is thus more expensive because it requires more rigorous DNA extraction and purification procedures as well as PCR troubleshooting (Brock et al., 2009; Osmundson et al., 2013). For generating high throughput data, fresh or recently collected specimens are clearly advantageous when compared to preserved herbarium materials. Morphological examination of specimens combined with Blast searches helped identify most vouchers to genus level (Table S1). Thirty-two OTUs could only be identified to family (14), order (14), or class (4), mostly because they belong to groups that require extensive systematic revision (e.g. Helotiales). Due to our focus on Nothofagaceae-dominated forests, we collected mostly ECM species (66%), but also many saprobes (31%), and fungi for which the trophic mode is unknown (3%). Agaricales (Basidiomycota) were diverse, abundant, easily sampled and therefore conspicuous in our dataset (Fig. 1c). We also found a large number of Pezizales and Helotiales (Ascomycota), many of which had low similarity to any known sequences. Within Agaricales, Cortinarius was the most diverse genus in both species richness and abundance, constituting c. 33% of all OTUs (Fig. 1c). The high diversity of Cortinarius lineages corroborates previous studies carried out in other southern temperate forests with Nothofagaceae (Tedersoo et al., 2008; Dickie et al., 2009; Nouhra et al., 2013; Fernandez et al., 2015; Horton et al., 2017). In addition, our dataset revealed surprisingly high species diversity in some fungal lineages where only a handful of species have been described from South America (e.g. /austropaxillus, /descolea – Peintner et al., 2001; Skrede et al., 2011). Similar to previous results based on ECM root tips (Nouhra et al., 2013), we detected relatively low diversity in several ECM lineages that are hyperdiverse in other regions of the world (e.g. /amanita, /boletus, /russula-lactarius). We also found several Northern Hemisphere exotic species in South American Nothofagaceae forests, including Inocybe ochroalba (MES1236), Hebeloma mesophaeum (MES1358) and Amanita muscaria (MES1647). All three of these taxa putatively fall into the category of 'introduced species that spread to local hosts' as outlined by Vellinga et al. (2009). Evidence of a native South American Inocybe species (MES1895) fruiting in pure Pseudotsuga plantations suggests that this species may be a potential invasive species in the Northern Hemisphere. We are actively engaged in taxonomic work to compare our collections with described species and formally describe novel taxa (Kneal & Smith, 2015; Trierveiler-Pereira et al., 2015; Kumar et al., 2017). Approximately one quarter of our OTUs (Fig. 1a, 23–32% according to percent Blast similarity) matched environmental sequences in UNITE+INSD with no available voucher specimen (Table S1; note that in some cases the closest Blast match was a GenBank sequence with no corresponding UNITE 'species hypothesis'). In addition, 43 OTUs were < 90% similar to any ITS sequence in UNITE+INSD. It is likely that our assessments may still underestimate the total number of species in some groups for which ITS is very similar among divergent species, for example many Cortinarius species (Ryberg, 2015; Garnica et al., 2016). Most OTUs (46%) matched sequences originating from South America, but c. 20% had a closest match to a sequence from Australasia (Fig. 1b), highlighting both the shortage of sequences from South America and the historical biogeographic connection of these two regions. This biogeographic pattern was particularly striking within some fungal lineages. For example, we found the first evidence of truffle-like species of Ruhlandiella, Amylascus and Gymnohydnotrya (Pezizales) in South America, despite the fact that described members of these genera are so far known only from Australasia (Table S1). Among the 309 unique OTUs detected, we identified two examples of distinctive plant-associated fungi that illustrate the exciting data generated from our collections. In the first case, we collected two small coralloid fungal specimens whose ITS sequences (KY462416, KY462417) were c. 80% similar to each other. The closest ITS Blast matches for these two OTUs were sequences from ECM root tips of Nothofagaceae that corresponded to an anomalous ECM lineage identified by Tedersoo & Smith (2013) as /agaricomycetes1. We subsequently sequenced 18S and 28S rDNA (Methods S1) and found that these specimens are phylogenetically affiliated with Tremellodendropsis tuberosa (Fig. 2a). Berbee et al. (2016) recently showed that T. tuberosa belongs to a unique Agaricomycete lineage in the order Tremellodendropsidales. Our rDNA phylogenies suggest that Tremellodendropsidales includes diverse ECM fungi that associate with a wide range of angiosperms (including Fagus, Eucalyptus, Lithocarpus, and Nothofagaceae) across the globe (Figs 2a, S1; Table S2). Together with T. tuberosa, our vouchered specimens will provide new insights into the morphology and ecology of this group. They will also allow the description of new species and provide fresh material for the phylogenomic placement of Tremellodendropsidales within the Agaricomycetes. A second case of unique plant-associated fungi was an Endogone-like specimen (KY462475) whose closest ITS Blast matches are fungal symbionts of nonvascular plants (liverworts and hornworts – Bidartondo et al., 2011; Desiro et al., 2013; Yamamoto et al., 2015) with no corresponding fungal specimens. We generated 18S and 28S rDNA sequences and placed this OTU in the Mucoromycotina (Fig. 2b; Table S2). Our specimen is nested in a large clade composed of multiple lineages, including specimens of Densospora solicarpa from Australia and Sphaerocreas pubescens from Japan. The Sphaerocreas-Densospora clade is sister to the Endogonales, which comprises Endogone species and additional sequences of early-diverging plant symbionts. Some members of the Endogonales and the Sphaerocreas-Densospora clade are associated with early-diverging plants (Bidartondo et al., 2011; Desiro et al., 2013), whereas others are ectomycorrhizal (Yamamoto et al., 2015, 2016) or 'arbuscular-like' symbionts of vascular plants (Orchard et al., 2016). These fungi have recently been documented from many hosts, habitats, and geographic locations, suggesting that the diversity of species and trophic modes of Mucoromycotina is much higher than previously understood. Because of their inconspicuous habit and the difficulties with culturing and DNA sequencing (Berch & Fortin, 1983; Tedersoo et al., 2016; Yamamoto et al., 2016), these fungi remain poorly represented in sequence databases, culture collections and herbaria, despite the growing evidence that species of Mucoromycotina may have played critical roles in the early colonization of terrestrial habitats (Strullu-Derrien et al., 2014; Field et al., 2015b, 2016; Rimington et al., 2015). The examples mentioned earlier are among the most illustrative in our dataset but are by no means the only discoveries. They demonstrate that systematic collecting, documenting, and sequencing from fresh specimens in undersampled regions are efficient and viable methods to capture unknown fungal diversity and provide substantial improvements to public DNA repositories. This approach is particularly relevant to 'fill the gap' of knowledge from geographic regions where comparatively fewer collections exist (e.g. South America – Roy et al., 2017) and this remains an efficient approach to obtain new fungal data at any site. Although environmental sequencing can rapidly detect diversity and elucidate ecological patterns, these approaches depend on informative sequence databases for fungal identification (Costello et al., 2013). Due to our current lack of knowledge, a large portion of environmental sequences cannot be identified at a meaningful taxonomic level (Nilsson et al., 2016). There is currently a movement to identify and classify fungi known only from sequences (Hibbett et al., 2016). Although we agree that it is critical to compile and validate high-quality environmental sequences, we nonetheless argue that 'traditional' methods should be considered irreplaceable and complementary to 'next-generation' approaches. Unfortunately, in the quest for cutting-edge science it is sometimes the case that 'traditional' methods are viewed negatively by funding agencies. We argue that, along with barcoding herbarium collections, high-throughput collect-and-sequence inventories are highly effective to document fungal diversity and are instrumental for future studies of plant–fungal symbioses. Herbarium vouchers provide much more than just DNA barcodes. Fresh well-documented specimens remain critical to reconstruct robust phylogenies, link sequence data to morphology, and supply ecological data on hosts and substrate associations (Peay, 2014). Specimens can also be used for stable isotopic analyses (Hobbie et al., 2001) and genomic studies (Tedersoo et al., 2016) in a way that environmental samples cannot. Given the increasing threats to biodiversity from habitat loss and climate change, responsibly collecting vouchered specimens with associated data and openly sharing these resources are more necessary today than ever before (Costello et al., 2013; Rocha et al., 2014). This work was supported by NSF grants DEB 1354802 (M.E.S. and P.B.M.) and DEB 1441677 (M.E.S.), an Advanced Postdoc Mobility Fellowship from the Swiss NSF (C.T.), and the David Rockefeller Center for Latin American Studies at Harvard University (D.P.). Collecting permits were issued by the Administración de Parques Nacionales of Argentina (Projecto 2016/720, E.N.), Secretaría de Desarollo Sustentable y Ambiente of Tierra del Fuego (no. 0218/2015, C.T.), Chilean Corporación Nacional Forestal (no. 014/2014, M.E.S.) and the Wildlife Conservation Society Chile in Parque Karukinka (C.T.). Vouchered specimens are deposited at the Museo Botánico of Córdoba (CORD), Museo Nacional de Historia Natural of Santiago (SGO), Farlow Herbarium (FH), the University of Tennessee Herbarium (TENN), and Florida Museum of Natural History (FLAS). Marc-André Selosse and three anonymous referees provided valuable comments to improve the manuscript before publication. C.T., A.B.M. and R.H. generated the DNA sequences. C.T. and A.B.M. carried out the analyses. C.T. and M.E.S. wrote the manuscript and constructed the figures and tables. M.E.S. provided expertise at all stages of research. All co-authors (C.T., A.B.M., R.H., F.K., G.F., D.T., T.N., P.A.S-L., N.F., J.M.E., A.M., G.P., D.P., E.N., R.S., M.S-G., P.B.M., and M.E.S.) participated in the field collections. Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office. Fig. S1 Maximum likelihood phylogeny of Tremellodendropsidales fungi based on ITS ribosomal DNA. Table S1 List of operational taxonomic units (OTUs) generated from this study Table S2 List of species and samples included in the phylogenetic analyses Methods S1 The full methods for specimen collection, DNA extraction and sequencing, and data analysis. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.