The samples were obtained from an experimental site established in 2013 (Unterseher et al. 2016), consisting of two plots of 4 years-old Fagus sylvatica trees from a different origin. The tree seeds originated from the Lustian lowland area (central eastern Germany, HKG 81005) and were grown in a nursery in northern Germany (near Hamburg) for two years. (Unterseher et al. 2016). The plots are located at the same slope of the mountain massif “Untersberg” in Bavaria, Germany at 517 m asl. (above sea level, respectively) (Valley site; Lat: 47.712946 Long: 13.040101) and at 975 m asl. (Mountain site; Lat: 47.683158 Long: 13.002102) (Siddique and Unterseher 2016). It is humus-rich with a well developed topsoil layer (A horizon) at the valley site, while the A horizon is only weakly developed at the mountain site. The ground and understory vegetation of the mountain site was mainly composed of Acer pseudoplatanus, Picea abies, Daphne mezereum, Cardamine (= Dentaria) enneaphyllos, Helleborus niger and Hepatica nobilis (Siddique and Unterseher 2016). At the valley site, ground vegetation was different with dominance of Acer pseudoplatanus, Mentha spp., Petasites hybridus, Equisetum sylvaticum and Rubus sp (Unterseher et al. 2016). Five trees each from each site were selected randomly in October 2014, exactly one year after planting. Ten symptomless green leaves per tree were removed and processed as described in Unterseher et al. (2016).

Isolation of endophytes followed the dilution-to-extinction cultivation (Solis et al. 2016). In brief, samples were blended into tiny particles and filtered. Smallest particles (ø < 0.2 mm) were resuspended and strongly diluted before plating onto malt extract agar (MEA, 1.5%) containing 48-well plates (Carl Roth, Karlsruhe, Germany). The plates were inspected regularly for four weeks. Emerging colonies were transferred into Petri dishes containing the same growth medium.

Instead of classifying the fungal cultures according to macroscopic and microscopic characters (Khoyratty et al. 2015) and selecting only a few representative strains per morphotype for downstream processing, genomic DNA was extracted from all isolates using a traditional, chloroform-based protocol (e.g. Solis et al. 2016). The ITS region was amplified with the primer pair V9G - ITS4 (de Hoog and Gerrits van den Ende 1998) using approved amplification kits (MangoTaq; Bioline, Germany) and cycling conditions (Solis et al. 2016). Unpurified PCR products were shipped to Beckman Coulter Genomics (Takeley, UK) for sequencing. Sequences were discarded if their corresponding chromatograms showed pronounced signs of ambiguous base calling after end trimming.

Total genomic DNA was extracted with the Charge Switch® gDNA Plant Kit (Invitrogen, Germany) from the same fresh leaf particle mass that was used for cultivation. Library preparation consisted of two consecutive amplification steps in order to add sample-specific tag combination for multiplexing. Please refer to Siddique and Unterseher (2016) and Unterseher et al. (2016) for the detailed description of this procedure. Amplicons were sequenced in pair-end mode on an Illumina MiSeq platform (Illumina Inc.) at the Genetics Section, Biocentre of the LMU Munich, Germany.

Demultiplexing and quality filtering of Illumina reads relied on QIIME (Navas-Molina et al. 2013; see Suppl. material 1) and is also detailed in Unterseher et al. (2016) and Eusemann et al. (2016). Extraction of ITS1 (forward R1 Illumina reads) was done with ITSx (Bengtsson-Palme et al. 2013) followed by reference-based chimera checking (Nilsson et al. 2015), open-reference OTU picking (complete-linkage clustering at 97% similarity; Rideout et al. 2014), selection of representative sequences and taxon assignment (Kõljalg et al. 2013). These last steps were also performed with all high-quality Sanger sequences to guarantee best comparability. Final quality filtering of HTS OTUs consisted of the removal of unique (occurring in only one sample) and rare OTUs (having less than five reads, cf. Brown et al. 2015). In contrast, all OTUs from cultivation data were retained, knowing that they belonged to true fungi. The reasons for using only the ITS1 region for subsequent analyses is comprehensively explained in recent papers (e.g. Unterseher et al. 2016, Eusemann et al. 2016).

The analysis of fungal biodiversity comprised the assessment of OTU richness and further indicators of diversity (Fisher’s Alpha, Shannon index and three numbers of Hill’s series of diversity, the latter considering different levels of rarity) (Hill 1973). Given the strong positive correlation of OTU richness and read numbers (data not shown, but compare Siddique and Unterseher (2016)) as well as the strongly different sequencing depth between the two approaches, read counts were standardised (divided) by sample totals.

Community composition was assessed with PCoA (principal coordinate analysis) and NMDS (non-metric multidimensional scaling) and tested with PERMANOVA (permutational multivariate analysis of variance). Functional guild analysis was performed according to Nguyen et al. (2016). The entire biodiversity analysis was performed with R (freely available at www.r-project.org, last accessed 11/2016). The corresponding command script and necessary input files are available as “Suppl. material 2”. Curated Sanger sequences were taxonomically annotated as far as possible and made available in ENA/GenBank through accession numbers LT604837–LT604881. High-throughput data are available under SRA accession number SRX1211311.

Data volumes differed strongly between Illumina sequencing and cultivation. Illumina sequencing resulted in 597 OTUs from 170480 curated ITS1 reads and cultivation revealed 70 OTUs from 438 culture-based Sanger sequences with the same settings for OTU clustering. The combined data set comprised 630 OTUs (+ 33 OTUs compared with Illumina data). Thirty-seven OTUs were detected with both methods (see Table S1 on Suppl. material 3 and Suppl. material 4).

An insignificant trend of lower fungal diversity at the mountain site across all indexes was observed for HTS data (Fig. 1A). Richness analysis of cultivation data was partly contradictory with significant differences of Fisher’s Alpha between valley and mountain samples (p <0.01, see Table S2 and S3 “Suppl. material 3”) but nearly identical accumulation curves (Fig. 1B). In addition and contrary to HTS data, Hill numbers N2 and N3 were significantly different for cultivation data (Fig. 1B, for statistical details see Table S2 and S3; “Suppl. material 3”). The accumulation curves for HTS data (Fig. 1A) revealed a clearly lower fungal richness for mountain than for valley samples, whereas the cultivation data failed to show such differences (Fig. 1B).

Figure 1. 

Diversity indexes and accumulation curves for a Illumina and b cultivation data of fungal leaf-inhabiting endophytes of beech. Except of the accumulation curves of cultivation data, both methods revealed a clear and partly significant trend of higher fungal diversity at the valley site.

The analysis of community composition with non-metric multidimensional scaling (NMDS), principal coordinate analysis (PCoA) and PERMANOVA discovered a significant influence of the isolation method (df = 1, F = 7.58, R² = 0.30, p= 0.001) as well as of locality (df = 1, F = 2.87, R² = 0.14, p= 0.014) (Fig. 2). The differences between valley and mountain site were more pronounced for HTS (F = 3.94, R² = 0.33, p= 0.007) than for cultivation data (F = 1.96, R² = 0.20, p= 0.027) (Fig. 2).

Figure 2 

. Principal coordinate analysis (PCoA) of fungal leaf-inhabiting endophytes of beech display strongly differing assemblages obtained with Illumina sequencing and cultivation. Both methods revealed differing mycobiomes from valley and from mountain leaves, although these differences were less pronounced for cultivation data. Abbreviations: IM = Illumina data from mountain samples, IV = Illumina data from valley samples, CM = cultivation data from mountain samples, CV = cultivation data from valley samples

Three of the five most abundant orders from Illumina data were also most abundant in cultivation data (Capnodiales – both methods, Helotiales – both methods, Saccharomycetales – Illumina only, Pleosporales – both methods; all Ascomycota) (Fig. 3). In addition, HTS revealed further and abundant orders from both Asco- and Basidiomycota, which were not recovered during cultivation (Fig. 3), such as the yeast fungi Saccharomycetales (Ascomycota) and Malasseziales (Basidiomycota). The Xylariales (Ascomycota, present with one isolate) was the only order from the cultivation data that was not detected with HTS.

Figure 3. 

Abundance distribution of the 20 most abundant orders of fungal leaf-inhabiting endophytes of beech on a logarithmic scale. Three of the five most abundant orders from high-throughput sequencing were also most abundant in cultivation data.

The five main guilds (pathotrophs, patho-saprotrophs, patho-symbiotrophs, saprotrophs and symbiotrophs) were all detected by HTS. Cultivation largely failed to detect pathotrophs (including patho-saprotrophs and -symbiotrophs). The relative abundance of saprotrophs was clearly higher in cultivation than in HTS data (Fig. 4B, C).

Figure 4. 

Relative abundance of fungal leaf-inhabiting endophytes of beech among the five main trophic guilds as revealed by analysis with FUNGuild (Nguyen et al. 2016). A compares the two methods for each trophic guild and unassigned data. B displays the trophic guilds and unassigned taxa for Illumina data, C for cultivation data. Abbreviations in [B and C]: U = Unassigned, P = Pathotrophs, PSa = Patho-Saprotrophs, PSy = Patho-Symbiotrophs, Sa = Saprotrophs, Sy = Symbiotrophs

When analysing the influence of locality for the occurrence of different ecological guilds, it turned out that the abundance of pathotrophs was significantly higher in leaves of mountain trees than of valley trees (Fig. 5A). Saprotrophs were also more abundant in leaves of mountain trees, whereas symbiotrophs were more abundant in valley than in mountain trees (Fig. 5A, B).

Results from HTS and cultivation data were congruent in as much as saprotrophs and symbiotrophs revealed similar abundance patterns for both methods.

Figure 5. 

Relative abundance distribution of fungal leaf-inhabiting endophytes of beech among the five main trophic guilds as revealed by analysis with FUNGuild (Nguyen et al. 2016). A compares the two localities for each trophic guild on the basis of Illumina data B compares the two localities for each trophic guild on the basis of cultivation data.

The most abundant orders were the same for both cultivation and HTS, namely Capnodiales, Helotiales and Pleosporales (Fig. 3). Ascomycota clearly dominated the cultivation data thus confirming results of many earlier cultivation studies (e.g. U’ Ren et al. 2012, Scholtysik et al. 2013, Unterseher et al. 2013b). Many of own isolates, such as those with highest sequence similarities to the genus Mycosphaerella and its anamorphs, are often described in literature as frequent asymptomatic inhabitants of living leaves. In this study, taxa generally known as saprotrophs were the dominant trophic guild of the cultivable part (see below) and, in terms of relative abundance, surpassed this guild as detected with HTS by far (Fig. 4B, C).

The compositional difference in the two mycobiomes also corresponded to the presence of parasitic taxa (Taphrinales, Erysiphales) and yeast-like fungi (Saccharomycetales and Tremellales) in the HTS data, whereas the cultivation data were devoid of fungi with obligate parasitic, biotrophic or pathogenic lifestyle. The latter guilds usually cannot be cultivated, and yeasts are often detected only during cultivation studies when growth of filamentous fungi is slowed down with low-temperature incubation. In this study, HTS retained a wide range of taxa (compare Al-Sadi et al. 2015) and all guilds available in the FUNGuild reference data base (Nguyen et al. 2016) as expected (Figs 4, 5).

A poor comparability of cultivation and HTS data, as it is presented here, was recently reported for a microbiome study (Eevers et al. 2016). It was caused by the fundamentally different sample coverage with the most abundant OTU counting 28621 reads for HTS (Mycosphaerellaceae, Sphaerulina) and 102 isolates for cultivation (Hyaloscyphaceae, Lachnum) (Suppl. material 4).

On the one hand side our results clearly demonstrate the limitations and biases of cultivation approaches for comprehensive biodiversity assessments. On the other hand, the results did not meet our expectations (see above), because significant correlations to environmental parameters (here, it was the difference between valley and mountain samples) were still recognized. The present cultivation data are in concordance with similar studies (Unterseher et al. 2013b) and confirm general knowledge about the community ecology of leaf endophytes (Cordier et al. 2012, Zimmermann and Vitousek 2012, Meng et al. 2013, Glynou et al. 2015, Rojas-Jimenez et al. 2016). Moreover, similar results were observed in the present comparative study on the basis of HTS data (see also Siddique and Unterseher 2016, Unterseher et al. 2016).