Mutualistic interactions between fungi and plant roots have been a fundamental prerequisite for evolution and biodiversity of land plants. More than 80% of known species of land plants are associated with mutualistic fungi in their roots, facilitating mineral nutrient uptake of the plants [1]. These associations are known as mycorrhizae. There is an amazing morphological and physiological diversity among numerous different types of these plant-fungal interactions. No other fungal group shows a diversity of mycorrhizal types comparable to that found in the Sebacinales, a basidiomycetous order described only recently [2], [3], [4], [5], [6], [7], [8], [9]. DNA sequence analyses have demonstrated a high phylogenetic diversity in this group [2], [7], which is divided into two distinct subgroups, informally designated group A and group B [2]. Though it is known from molecular phylogenetic analyses that Sebacinales belong to the mushroom-forming basidiomycetes (Agaricomycotina) [10], only a few sebacinalean morphospecies producing basidiomes have been described, all of them belonging to group A. Morphological data on group B Sebacinales is very sparse.

Endophytes, as opposed to mycorrhizal or endoparasitic fungi, are commonly defined as fungi colonizing tissues of living plants without formation of detectable interaction structures such as interaction apparatus or arbuscules and without causing disease symptoms on the hosts [11]. Endophytes were shown to occur in all organs of plants [12]. Classic examples are some ascomycetes in grasses which might have gone unnoticed if they did not produce alkaloids having a strong deleterious effect on cattle feeding on the respective grasses [13]. In practice, the separation of mycorrhizal, pathogenic and endophytic habits often seems to be problematic and a better understanding of symbiont interactions is required to refine these definitions. The recent discovery of the beneficial effects of ascomycetous fungal endophytes of the genus Curvularia on host plants in geothermal environments [14] has raised considerable interest for endophytes and the mechanisms of their interaction.

Some Sebacinales strains commonly considered as endophytes, particularly the Piriformospora indica model strain belonging to group B, have recently been studied intensively, because they significantly enhance plant growth and seed yield, and induce systemic resistance of their host plants against abiotic stress and fungal pathogens [15], [16], [17], [18]. Experimental studies suggest that the fungus improves the nutritional status of its host plants [19], and that programmed death (apoptosis) of cortical cells, which are subsequently densely colonized by Piriformospora hyphae, plays an important role in this endophytic interaction [15], [20]. Generally, since some Sebacinales can easily be maintained and propagated without their plant hosts, these strains may be ideal models for the study of beneficial fungus-plant interactions and have a promising perspective for application in sustainable horticulture and agriculture [15], [17], [18], [21], [22].

Sequences of Sebacinales have been sporadically detected by PCR/cloning approaches from herbaceous plants and soil [18], [23]. Sebacinalean fungi other than the strains used in experimental studies were recently detected in a preliminary study based on a few environmental plant samples [24]; these authors raised the question whether Sebacinales may occur as endophytes in the field more frequently than previously thought. Here we address this question and show that these fungi are indeed found as endophytes within plants and that they are ubiquitous.

We analyzed 128 root samples from phylogenetically and ecologically diverse plants from 27 families from four continents and studied the phylogenetic distribution of their sebacinalean endophytes. We used diagnostic group-specific PCR primers for nuclear-encoded rDNA regions designed for this study to detect members of Sebacinales groups A and B and molecular phylogenetics to assess their evolutionary relationships. In addition, we analyzed selected specimens by transmission electron microscopy for the presence of fungal hyphae in the roots with septal pore structures typical for the Sebacinales. This combination of methods is crucial; diagnostic PCR and sequencing alone cannot definitely exclude sebacinalean soil fungi only present on the root surface, and ultrastructural analysis alone cannot discriminate Sebacinales from certain other groups of Agaricomycotina that exhibit a similar septal pore structure. We classify sebacinalean sequences from plants known to be non-mycorrhizal or hosts of arbuscular mycorrhiza, which is known to exclusively involve fungi from the phylum Glomeromycota, as endophytic (Fig. 1).

Molecular analysis yielded 135 sebacinalean nuclear LSU sequences from 128 root samples. We found sebacinalean endophytes in all examined plant families, which span a broad phylogenetic range, from liverworts to Asteraceae (Table S1). Our PCR variants 2 and 3, which involve Sebacinales-specific primers designed for the present study, allow to contiguously amplify and analyze two key rDNA regions (ITS+5.8S and the D1-D2 regions of the nuc LSU; see Fig. S1) presently used for molecular identification and phylogenetic reconstruction employed in the fungi. We even succeeded to amplify Sebacinales sequences from herbarium specimens collected in the 1830-1840s by G.W. Schimper and T. Kotschy on their pioneering botanical excursions to North Africa (FJ556825-30, FJ556857; Fig. S2: 15a,b; 16-20), and were able to sequence the type material of Sebacina vermifera from the 1960s (Fig. S2: 21). DNA sequences obtained with these PCR variants can be analyzed both in the context of fungal barcoding (using the ITS portion) and for constructing phylogenetic trees that span both Sebacinales group A and group B or even Fungi as a whole (using the LSU portion). Such data is therefore particularly applicable for analyses of Sebacinales communities using recently developed methods that combine phylogenetic and ecological approaches [25], [26].

Evidence that Sebacinales were in fact present within the plant roots and not just in the rhizosphere was obtained by electron microscopy (Fig. 2). We found that fungal hyphae colonized few dead cortical cells of their host's fine roots, nearly completely filling the host cells. This is well compatible with results obtained from in vitro experiments with P. indica and the model plants barley and Arabidopsis [15], [20]. The colonizing hyphae had the septal pore structure typical for the Sebacinales (Fig. 2).

An overview of the phylogenetic tree of all Sebacinales nuc LSU sequences currently available in GenBank together with our new endophytic sequences is shown in Fig. 1 (for the full tree see Fig. S2). The endophyte sequences were placed in both Sebacinales subgroups A and B, however they were not evenly distributed across these two groups, as most endophytic sequences were placed in group B (Fig. 1). Within group B some clades were dominated by endophyte sequences (Fig. 1: e.g., 3, 5); on the other hand, there are other clades that at this time nearly completely lack known endophytic sequences (Fig. 1: 1). The most basal clades in groups A and B (Fig. 1: 8–10, 12–14) were endophytic; we therefore hypothesize that the endophytic habit may be ancestral in the Sebacinales and the starting point for the development towards specialized mycorrhizal symbioses. However, the close relationship of endophytes, in particular with orchid mycorrhizal strains, could also be indicative of the capability of Sebacinales strains to switch between symbioses, or to fall back on the endophytic habit if no appropriate mycorrhizal partner is present.

Our analyses demonstrate that Sebacinales with closely related or even identical LSU sequences can be found in geographically distant areas (Fig. 1: 7, 11; Fig. S2: 22). The absence of any obvious geographical patterns suggests efficient dispersal. Diverse sequences could be found in the same field site, even in the root system of the same plant host specimen (Fig. S2: 23a-d; 15a,b; 24-26a-b). Host specificity seems to be low as identical sequences were often found across different hosts (Fig. S2: 7, 11). According to present knowledge, sebacinalean fungi involved in ectomycorrhiza (ECM), ectendomycorrhiza (arbutoid mycorrhiza; EEM), and mycorrhizas with heterotrophic or mixotrophic orchids (ORM) have only been found in group A, while ericoid (ERM) and cavendishioid (CAV) mycorrhizas are only known from group B [2], [7]. This is widely confirmed by the present study. Since links between ORM and ECM as well as between EEM and ECM via the same fungus seem to be common [5], [7], it has been hypothesized that all group A Sebacinales are ectomycorrhizal. It was thus surprising to detect endophytic group A Sebacinales (Fig. S2: 12, 27) in meadows where known ECM hosts were missing. Thus, ectomycorrhizal activity is probably not an obligate feature for group A Sebacinales.

In addition to linking ECM to ORM and to EEM, it is likely that Sebacinales are able to connect other mycorrhizal types. We detected the same nuclear LSU sequence in roots of Sherardia (Rubiaceae) from Italy and in an Australian orchid (Fig. S2: 6). Other identical sequences were found in an ericad (Vaccinium) and a liverwort (Riccardia; Fig. S2: 28) or in another ericad (Cavendishia), Melittis (Lamiaceae) and Calamagrostis (Poaceae; Fig. S2: 29). Considering also the numerous cases in which the same nuclear LSU sequence of sebacinalean endophytes was detected in distinct plant species (e.g., Fig. 1: 7, 11), we hypothesize that Sebacinales play a crucial role in connecting individual plants in terrestrial ecosystems across mycorrhizal types. However, nutrient transfer studies are necessary to elucidate this issue.

On the basis of our molecular phylogenetic analysis we can link some of the endophyte sequences to morphospecies (Fig. 1: 6, 13). However, S. vermifera strains mostly originating from roots of Australian terrestrial orchids [27] are scattered all over group B, demonstrating the limited usefulness of morphospecies in this context. In fact, from the high genetic distances between specimens that have all been assigned to S. vermifera we conclude that much of the huge biodiversity in Sebacinales group B is covered by cryptic species that lack macroscopic fruiting bodies.

Most experimental studies in the Sebacinales have been conducted using the asexual model strain Piriformospora indica, which was originally isolated from soil of the Indian Thar desert [28]. Our study reveals that P. indica belongs to a well-supported group of closely related endophytic species, in our study represented by sequences from Western European and Namibian Fabaceae, Poaceae, or Araceae (Fig. 1: 2). In view of the biodiversity of endophytic Sebacinales still to be expected (Fig. S3) there is a huge resource of strains that are potentially useful for plant cultivation. Recent studies have shown that the available strains of S. vermifera and P. indica differ quantitatively in the plant-beneficial effects they induce in different hosts [15], thus the vast genetic diversity of plant-associated Sebacinales offers interesting perspectives for future experimental research and inoculum development.

We show that Arabidopsis thaliana is associated with Sebacinales under natural conditions (Fig. S2: 30). This model species belongs to the Brassicaceae, which have widely been believed to lack mycorrhizal interactions. Our finding gives new practical relevance to experimental studies on endophytic interactions between Piriformospora and the model plant A. thaliana [29], [30], since Brassicaceae contain many economically important plants such as cabbage and rape. The presence of sebacinalean endophytes in Triticum (Fig. S2: 25a,b; Fig. 2) as well as in Zea mays (Fig. S2: 31–33) collected in the field is equally important. Given the positive effects that sebacinalean fungi had on growth, yield and resistance against abiotic stress and fungal pathogens of their plant hosts in experiments under controlled conditions [15], [17], these results underline the feasibility of applying Sebacinales as biological fertilizers and biocontrol agents for arable crops in the future. At the same time, however, our findings imply that inoculated fungi have to be sufficiently competitive against diverse local Sebacinales communities already present in roots and soil.

Since the database of Sebacinales sequences on a worldwide scope is still too sparse, we are currently unable to answer detailed questions on biogeography of this group. Also, since for the present study we merged sequences obtained with different PCR approaches (see the Materials section) and did not rigorously clone all PCR products that could not be sequenced directly, we cannot address the abundance of Sebacinales in more details here. These issues will be dealt with in subsequent, geographically more restricted studies using extended sample numbers.

The ubiquity and diversity of sebacinalean endophytes shown in this study emphasize a previously unrecognized aspect of the plant interactions of this fungal group. Sebacinalean endophytes are not isolated phenomena, but extremely common and potentially important hidden players in plant ecosystems, which may have given rise to the large diversity of mycorrhizal symbioses the Sebacinales participate in. These findings are both relevant for applied research as well as for basic research on the role of Sebacinales in ecosystem functioning and possible shaping of plant communities.