Beyond detection: unveiling microbial dynamics in oak seedlings using fungal isolation and amplicon sequencing

Abstract

Background

Forest nurseries are critical for producing resilient nursery stock for reforestation and planting of ornamental trees, yet the microbial communities associated with nursery grown plants remain poorly characterized. Quercus robur L. seedlings from seven Czech forest nurseries were analyzed to assess microbial diversity, co-occurrence patterns, and environmental drivers.

Results

Microbial communities were characterized using fungal ITS2 and bacterial 16S rRNA gene amplicon sequencing, fungal isolation, and comprehensive soil chemistry. HTAS revealed broader taxonomic and functional profiles compared to isolation, which selectively enriched for fast growing pathogens. Both methods were found to be complementary, emphasizing the value of methodological integration. Fungal communities exhibited pronounced site specific beta diversity and responded to inorganic soil parameters, particularly calcium, phosphorus, and pH/CaCl₂. Bacterial communities were more spatially cohesive and primarily associated with humification related factors. Trends in fungal alpha diversity were observed in relation to organic matter fractions (e.g., C_ox, D_H). SparCC genus level association analyses revealed high magnitude compositional association patterns, but no network edges remained significant after FDR correction; these patterns are therefore interpreted as exploratory and hypothesis generating rather than as evidence of direct microbial interactions.

Conclusions

Mineral soil properties were associated with microbial community structure and fungal trophic composition. Despite standardized nursery conditions, edaphic variability exerted strong filtering effects on fungal and bacterial communities. These findings provide ecological insight into seedling microbe interactions and offer a basis for microbiome informed nursery management strategies.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s40793-026-00929-0.

Introduction

Pedunculate oak (Quercus robur L.) is a foundation species of European temperate forests, crucial for biodiversity maintenance, carbon storage, and timber production [1–4]. Climate change, frequent droughts, and pest outbreaks have intensified efforts to restore resilient broadleaf forests, increasingly favoring species such as Q. robur and Fagus sylvatica in European restoration programs [5–11]. Millions of oak seedlings, marketed as Q. robur, are produced annually in Central European nurseries to support reforestation programs, making seedling health and quality a priority for sustainable forest regeneration and a relevant system for investigating plant-associated microbiomes under forest production conditions.

Plant associated fungi and bacteria form a key interface between Q. robur seedlings and their nursery environment. They may support nutrient acquisition, stress tolerance, and pathogen suppression, but can also include latent or opportunistic pathogens. Studying these communities in symptomatic nursery seedlings therefore helps place pathogen detection into a broader microbial and environmental context. While diverse ectomycorrhizal fungal communities have been documented in forest nurseries [9, 12], broader bacterial and fungal microbiomes of Q. robur seedlings remain largely unexplored.

Forest nurseries therefore represent critical phytosanitary checkpoints, as planting material may serve as a vector for fungal and bacterial pathogens, including latent or cryptic infections that manifest only under environmental stress. Documented cases of pathogen dissemination via international seedling trade highlight the importance of early and accurate detection [13, 14].

Among these, trunk disease pathogens (TDPs) have been associated with substantial ecological and economic losses in both forest and agricultural systems [14–16]. Fungal families such as Botryosphaeriaceae, Nectriaceae, and Diaporthaceae include aggressive taxa capable of colonizing vascular tissues, disrupting water transport, and causing dieback and mortality across a wide range of woody hosts, including oaks, beeches, maples, firs, pines, grapevines, and walnuts [17–20]. However, pathogen presence alone does not necessarily reflect active disease development, as plant health is shaped not only by pathogenic load but also by the structure and balance of the associated microbial community [21–25].

The plant microbiome, comprising both fungal and bacterial communities, plays a key role in host defense, stress tolerance, and nutrient acquisition [26–28]. Interactions within these communities may suppress or facilitate pathogen activity depending on local ecological conditions [29, 30]. Therefore, assessing microbial diversity and functional structure, including trophic strategies and co-occurrence patterns, is essential for understanding plant health beyond simple pathogen presence/absence diagnostics.

Amplicon sequencing of fungal ITS regions and bacterial 16S rRNA genes provides a powerful approach for profiling microbial communities, capturing both dominant and rare taxa, including those not readily culturable [31–33]. In contrast, classical isolation methods are biased toward fast growing organisms but remain indispensable for functional validation, including pathogenicity and virulence testing. These approaches should therefore be considered complementary tools for comprehensive microbial diversity assessment.

In this study, we investigated fungal and bacterial communities associated with two year old Q. robur seedlings collected from seven forest nurseries across the Czech Republic. All sampled seedlings exhibited decline symptoms, including bark necrosis and shoot dieback. Rather than screening asymptomatic individuals, the study focused on symptomatic seedlings to characterize the microbial and environmental context associated with early decline syndromes in forest nursery material. The objective was to identify fungal groups potentially associated with seedling decline and to determine how microbial community structure, trophic composition, and soil properties vary across nurseries and plant compartments.

To address this objective, we combined culture dependent fungal isolation, high-throughput amplicon sequencing (HTAS) of root and trunk tissues, and soil physicochemical profiling. This combined design allowed us to distinguish between broad community patterns captured by sequencing and viable culturable fungi recovered from symptomatic tissues. Specifically, we tested the following hypotheses: (i) symptomatic Q. robur seedlings harbor a complex fungal community in which culture dependent isolation and HTAS recover partly overlapping but complementary subsets of dominant, potentially pathogenic, and functionally diverse fungal taxa; (ii) fungal and bacterial communities exhibit structured, non random co-occurrence patterns that reflect shared habitat preferences and potential ecological associations rather than direct biotic interactions; (iii) soil physicochemical properties are associated with variation in microbial alpha and beta diversity across nursery sites; (iv) root and trunk tissues differ in microbial community composition due to compartment specific habitat filtering; and (v) highly connected taxa within microbial association networks may indicate environmentally responsive or structurally important community members, but should be interpreted as exploratory indicators rather than direct proxies of ecological stability.

Materials and methods

Experimental design and sampling

During the autumn of 2020 and spring of 2021, a total of 70 bare-rooted seedlings and 7 soil samples were collected from seven forest nurseries across the Czech Republic (10 seedlings and 1 kg soil sample per nursery) (Table 1, Table S1, Figs. 1, 2, and S1). Soil samples were collected from the rhizosphere (10–20 cm depth) near the root zones of multiple seedlings within each site using a shovel. The subsamples were pooled per site to generate a 1 kg of homogenized composite soil sample representing each nursery. Soil was placed into plastic bags and stored at 5 °C until further processing. The sampled seedlings were two years old and exhibited visible decline symptoms, including bark necrosis, bud and branch dieback, and leaf chlorosis. Sampling sites, GPS coordinates, sample codes, and plant compartments are summarized in Table 1. Sample codes combine the nursery identifier (N1-N7) with the sampled plant compartment, where B denotes base/root tissue and S denotes stem/trunk tissue.

Table 1.

Overview of forest nursery sites, GPS coordinates, sample codes, and corresponding plant compartments used in this study

Forest nursery	GPS coordinates	Sample code	Plant compartment
N1	49.8159317N, 17.2012250E	N1B	Base/root
N1		N1S	Stem/trunk
N2	49.5368700N, 17.3117400E	N2B	Base/root
N2		N2S	Stem/trunk
N3	48.8175681N, 14.9459272E	N3B	Base/root
N3		N3S	Stem/trunk
N4	48.9143172N, 15.9296406E	N4B	Base/root
N4		N4S	Stem/trunk
N5	49.6077514N, 16.7167122E	N5B	Base/root
N5		N5S	Stem/trunk
N6	49.3218225N, 16.2383442E	N6B	Base/root
N6		N6S	Stem/trunk
N7	48.9749733N, 16.1258161E	N7B	Base/root
N7		N7S	Stem/trunk

The code consists of the nursery identifier (N1-N7) and plant compartment abbreviation: B = base/root tissue, S = stem/trunk tissue

Fig. 1.

Map of the sampled forest nurseries (N1-N7), generated using Folium (Python library)

Fig. 2.

Workflow for Microbiome Analysis of 70 Oak Seedlings from 7 Nurseries Using Isolation and HTAS In total yielding 140 plant tissue samples (root and trunk). DNA was extracted and used to amplify the fungal ITS region and bacterial 16S rRNA gene

All nurseries followed a standardized low-input plant protection protocol. A herbicide application was performed in early spring before budbreak, using a glyphosate-based formulation (e.g., Roundup®, 360 g L⁻¹ glyphosate acid) at 1–3 L ha⁻¹, applied to weed zones around the root collars. Manual and mechanical weed control (including hand weeding and inter-row hoeing) was conducted throughout both the first and second growing seasons as part of routine agrotechnical nursery management.

In the previous growing season, powdery mildew was managed by a single foliar application of elemental sulfur (e.g., Kumulus® DF, 80% w/w S) at 6–8 kg ha⁻¹ in August. No additional fungicides or any insecticides were applied directly to the oak seedlings. Where aphid pressure occurred, control measures were restricted to inter-row vegetation and did not involve treatment of the seedlings themselves. Seedlings were undercut in the autumn of their first growing season to promote lateral root development.

Each sampled plant was rinsed with tap water to remove soil residues. Trunk and roots were cut into 10 cm fragments using scissors sterilized with ethanol and flame between each cut. The cut fragments were then surface-sterilized by immersion in 70% ethanol and briefly flamed. Sterile scalpels were used to excise internal woody tissue separately from the trunk and root portions (see Fig. 2). In total, 140 wooden samples were obtained for subsequent total DNA extraction and fungal isolation.

Soil analysis

Soil samples (1 kg per site) were collected from the rhizosphere and refrigerated at 5 °C prior to analysis. Soil organic matter was characterized through a combination of chemical analysis and spectrophotometric techniques. The oxidizable carbon (C_ox) [34] was determined using an oxidimetric titration method according to Nelson et al. [35], with results expressed as percentages of dry soil. Humic substances (HS) were extracted from the fine soil by a solution of sodium pyrophosphate in an alkaline mixture + NaOH using short fractionation [36, 37] and fractionated into humic acids (HAs) and fulvic acids (FAs), and their relative proportions were used to evaluate the degree and quality of humification [38, 39]. The structural properties of humic matter were further assessed using UV–VIS spectrophotometry. Absorbance values were used to calculate an optical index reflecting the dominance of either labile aliphatic (465 nm) or stable aromatic compounds (665 nm) [37, 39]. The colour quotient (Q_4/6) was measured using a spectrophotometer SPECORD PLUS 200 (Analytik Jena AG, Jena, Germany) [37]. The shape and intensity of the absorbance curves were interpreted to characterize the quality and maturity of humic substances across sites. The degree of humification (D_H) of soil organic matter (%) was calculated according to the relation: D_H = C_ox FA/C_ox × 100 [37]. Macronutrients (P, K, Ca, Mg) were extracted using the Mehlich III method [40]. The content of available P was determined colorimetrically using the SPECORD PLUS 200 spectrometer (Analytik Jena AG, Jena, Germany). The contents of available K, Ca and Mg were determined by atomic absorption spectrometry with the ContrAA 700 instrument (Analytik Jena AG, Jena, Germany). Soil pH/CaCl₂ was determined in 0.01 M CaCl₂ solution using a pH meter InoLab pH Level 2 (WTW, Weilheim, Germany). The full list of measured soil parameters included: C_ox (%), HSs (%) (humic substances), HAs (%) (humic acids), FAs (%) (fulvic acids), HA/FA ratio, D_H (%) (degree of humification), Q_4/6 (colour quotient), soil pH/CaCl₂, and available P, K, Ca, and Mg (mg kg⁻1).

Fungal isolation and morphotyping

Sterile wood fragments were cut into segments 0.5–1.5 cm in length and placed onto 90 mm Petri dishes containing potato dextrose agar (PDA; HiMedia Laboratories LLC, Kennett Square, PA, USA) supplemented with 0.5 g L⁻¹ of streptomycin sulphate to inhibit bacterial growth. For each seedling, four Petri dishes were prepared, with 9–10 wood segments per dish, resulting in a total of 280 plates (Fig. S1).

Plates were incubated in a growth chamber under controlled conditions (24 °C, 30% relative humidity, and darkness). Fungal colonies typically began to emerge within three days. Once visible, actively growing colonies were transferred daily onto fresh PDA plates without antibiotics to obtain pure cultures. All isolates were subsequently cultivated on 60 mm PDA plates at 20 °C in darkness for morphological observation.

In total, 1007 fungal colonies were subcultured. Cultures were grouped by forest nursery and sampling site into 14 groups. Within each group, colonies were sorted based on macroscopic morphological characteristics, including colony texture, pigmentation (surface and reverse), and radial growth rate after 7 days of incubation on PDA at 22 °C. Morphological traits were assessed at two time points after 1 week and after 3 weeks of cultivation (Fig. S1). From each morphotype group, representative isolates were selected for molecular identification.

Molecular identification of fungal isolates

A total of 364 representative fungal colonies were selected for molecular identification. Approximately 100 mg of fresh mycelium from each isolate was transferred into a 2 mL homogenization tube containing ceramic beads. Homogenization was performed using a FastPrep instrument (MP Biomedicals) at 4.5 m/s for 20 s, in two consecutive rounds. Total DNA was extracted using the NucleoSpin® Tissue Kit (Macherey–Nagel, Düren, Germany) according to the manufacturer’s protocol for tissue. DNA quantity and purity were assessed spectrophotometrically using a SpectrostarNano (BMG Labtech, Offenburg, Germany).

Fungal ITS rDNA regions were amplified using the primer pair ITS1 (CTTGGTCATTTAGAGGAAGTAA) [41] and ITS4 (TCCTCCGCTTATTGATATGC) [42]. PCR reactions (22 µL total volume) were prepared using the GoTaq® Flexi DNA Polymerase system (Promega Corporation, Madison, WI, USA), and included the following components: 1 µL of each primer (10 µM), 0.2 µL of dNTP mix (10 mM each), 6 µL of 5 × Colorless GoTaq® Reaction Buffer, 2 µL of MgCl₂ (25 mM), 2.6 µL of 5 × Green GoTaq® Loading Dye, 0.2 µL of GoTaq® polymerase, and nuclease-free water up to 22 µL. Thermal cycling was performed as follows: initial denaturation at 95 °C for 2 min; followed by 35 cycles of denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 60 s; with a final extension step at 72 °C for 5 min.

Amplification success was verified by electrophoresis on 1.3% agarose gels (Serva, Heidelberg, Germany). Successfully amplified PCR products were purified using the NucleoSpin® Tissue Kit (Macherey–Nagel) and submitted for Sanger sequencing at Eurofins Genomics (Ebersberg, Germany).

Sample processing and DNA extraction

Wooden tissues were sampled into sterile mortars and frozen at -80 °C for 2 h. Subsequently, samples were homogenized in pre-cooled mortars maintained at -80 °C. The homogenized material was transferred into 2 mL sterile tubes and stored again at -80 °C until extraction.

Total DNA was extracted using the Macherey–Nagel NucleoSpin® Tissue Kit, following the manufacturer’s protocol for DNA isolation from tissue. DNA was extracted from 100 mg of homogenized wood tissue, with an overnight incubation using Proteinase K and Buffer T1 as a pre-lysis step.

To maximize DNA yield and concentration, elution was performed in two steps using pre-heated elution buffer (70 °C), with a 3-min incubation per step.

In total, DNA was extracted from 140 samples: 70 trunk and 70 root segments. DNA concentration and purity were assessed spectrophotometrically using the SpectrostarNano (BMG Labtech), including measurement of 260/280 absorbance ratio to evaluate sample purity.

Amplicon library preparation

Fungal ITS regions were amplified from total DNA using primers gITS7 (5′-GTGARTCATCGARTCTTTG-3′; [43]) and ITS4 (5′-TCCTCCGCTTATTGATATGC-3′; [42]). The V3–V4 region of the bacterial 16S rRNA gene was amplified using primers S-D-Bact-0008-a-S-16 (5′-AGAGTTTGATCCTGGCTCAG-3′) and S-D-Bact-0785-a-A-21 (5′-GACTACHVGGGTATCTAATCC-3′) [44]. Amplifications were performed using the GoTaq® Flexi DNA Polymerase Kit (Promega) in 25 µL reactions, prepared according to the manufacturer’s protocol. Thermal cycling conditions for fungal ITS amplification consisted of an initial denaturation at 95 °C for 2 min, followed by 35 cycles of 95 °C for 30 s, 58 °C for 30 s, and 72 °C for 60 s, and a final extension at 72 °C for 5 min. For bacterial 16S amplification, the annealing temperature was set to 55 °C, with all other cycling steps identical to the ITS protocol.

PCR products were verified on a 1.3% agarose gel (Serva, Heidelberg, Germany) and purified using the NucleoSpin® Tissue Kit (Macherey–Nagel). Amplicon concentrations were determined fluorometrically, and each sample was diluted to 10 ng·µL⁻¹ using nuclease-free water.

Although DNA was extracted and PCR amplification was performed separately for 140 individual samples (70 seedlings × 2 tissues), amplicons were subsequently pooled by nursery site and plant compartment (base and stem), resulting in 14 composite samples (7 sites × 2 compartments). These pooled samples were used as the effective sample size for all downstream analyses.

Genomic libraries were prepared using the Nextera XT DNA Library Prep Kit (Illumina, San Diego, USA), and Nextera XT v2 indexes were added following the standard Illumina protocol.

Library quantification and size distribution were assessed using a Bioanalyzer 2100 (Agilent Technologies) with the High Sensitivity DNA Kit. Sequencing was performed on an Illumina MiSeq platform using the MiSeq Reagent Kit v3 (2 × 150 bp paired-end reads, 300 cycles) at Mendeleum—Department of Molecular Genetics, Mendel University in Brno.

Negative controls (extraction blanks and PCR no-template controls) were included throughout the entire workflow to monitor for potential contamination during extraction, amplification, and sequencing.

Bioinformatic data evaluation

Sanger sequencing results from fungal isolates were quality-checked and visualized using Geneious Prime (v2024.0.5). Taxonomic identification was conducted using NCBI’s BLASTn algorithm (megablast), with an e-value threshold of 0.05 and a minimum identity of 98% [45].

HTAS data quality was assessed using FastQC v0.10.1 [46]. Read preprocessing and quality filtering were performed in SEED v2.1.2 [47], with paired-end reads merged using fastq-join [48]. Primer sequences and adapters were removed, and reads shorter than 70 bp or containing ambiguous bases were discarded. Only sequences with an average Phred quality score ≥ Q30 were retained. The ITS2 region was extracted from full-length amplicons using ITSx v1.1.2 [49].

OTU clustering was performed using USEARCH v8.1.1861 at a 97% sequence similarity threshold [50]. Singleton reads and chimeric sequences were removed using UPARSE implemented in USEARCH v8.1.1861 [51]. Representative OTU sequences were taxonomically assigned using BLASTn against the UNITE fungal database (v8.2, released in 2021) for fungal sequences and the SILVA 16S rRNA gene database (SILVA v138, released in 2021) for bacterial sequences, using a minimum sequence identity threshold of 98% and a maximum e-value of 1 × 10⁻5⁰ [52, 53]. Sequences were excluded if they lacked a fungal hit, matched non-fungal organisms, or fell below the similarity/e-value thresholds [54, 55].

Data analysis

For descriptive taxonomic visualizations, OTU tables were transformed to relative abundance using total sum scaling. Compositional patterns were visualized using stacked bar plots and heatmaps generated with the matplotlib and seaborn Python libraries [56]. The compositional nature of amplicon sequencing data was considered in downstream association analyses by applying centered log-ratio transformation or SparCC where appropriate [57, 58].

Alpha diversity was assessed to describe microbial richness and evenness. Shannon diversity and Simpson index were calculated from relative abundance profiles, whereas Chao1 richness was calculated from count data. All alpha diversity metrics were calculated using the skbio.diversity.alpha module in scikit-bio [59, 60]. Diversity indices were calculated for the 14 pooled sequencing libraries representing seven nursery sites and two plant compartments. Where group comparisons were performed, statistical differences were tested using the nonparametric Mann–Whitney U test from the scipy.stats module, and results were visualized using boxplots generated with seaborn.

Beta diversity was evaluated using Bray–Curtis dissimilarity and the Jaccard index to compare community composition among the 14 pooled sequencing libraries. Bray–Curtis dissimilarities were calculated from relative abundance profiles, whereas the Jaccard index was used to compare presence absence based community patterns. Dissimilarity matrices were computed using the pdist and squareform functions from the scipy.spatial.distance module and visualized as heatmaps using seaborn.heatmap. Differences in community composition among predefined groups were tested using permutational multivariate analysis of variance and analysis of similarities, implemented in the skbio.stats.distance module, with 999 permutations. Principal Coordinates Analysis was conducted on Bray–Curtis and Jaccard distance matrices using skbio.stats.ordination.pcoa, and the first two principal coordinates were visualized using matplotlib.pyplot and seaborn.scatterplot. Each point in the ordination represents one pooled sequencing library.

To identify ecological roles of fungal taxa, FUNGuild was applied to taxonomically assigned fungal OTUs [61]. Functional assignments were collapsed into broader trophic categories, including Pathotroph, Saprotroph, Symbiotroph, and mixed trophic modes. Redundant or overlapping FUNGuild labels were harmonized before visualization. Functional structure across sites was visualized using bar plots.

Associations between microbial diversity or functional profiles and soil chemical parameters were evaluated at the nursery site level. Because soil parameters were measured once per nursery site, microbial data were aggregated to the same site level before correlation analyses, resulting in seven independent observations. Soil variables included pH/CaCl₂, available Ca, Mg, P, K, and organic soil fractions. Spearman correlations were calculated between Shannon diversity indices and individual soil variables separately for fungi and bacteria. Because these analyses were based on seven nursery sites, empirical p-values were obtained using exact permutation testing across all possible permutations. Multiple testing was controlled separately for fungal and bacterial Shannon correlations using the Benjamini and Hochberg (BH) false discovery rate procedure [62].

For fungal trophic group profiles, redundant FUNGuild categories were first collapsed, and Spearman correlations with soil parameters were calculated using site level values. Multiple testing was controlled using the BH false discovery rate procedure. Soil related correlations were used to describe exploratory trends rather than causal relationships.

To explore associations between soil parameters and ordination patterns, a correlation-based vector fitting approach was applied to PCoA ordinations representing fungal community composition, bacterial community composition, and fungal trophic group composition. Each soil parameter was projected into the ordination space as a vector, with its direction defined by correlations with the first two PCoA axes and its length proportional to the combined correlation strength. Spearman correlation was used for total sum scaled taxonomic data, whereas Pearson correlation was applied to Hellinger transformed trophic group data [63, 64]. This approach, conceptually similar to the envfit function in R [55, 65], was used to visualize the strength and direction of associations between soil variables and community compositional shifts.

Microbial association patterns were analysed at the genus level using SparCC, a correlation inference method developed for compositional genomic survey data [57]. Prior to analysis, unassigned genera were removed, and genera were retained if they occurred in at least three samples and reached a mean relative abundance of at least 0.1%. To reduce sparsity, the analysis was restricted to the 50 most abundant retained genera per kingdom. Fungal and bacterial intrakingdom associations were inferred separately from genus level count tables using a pseudocount of 0.5. Empirical p-values were estimated using 999 permutations, and multiple testing was controlled using the BH false discovery rate procedure [62]. Strong associations were defined as |SparCC r|≥ 0.80 and were considered statistically supported only if they remained significant after correction.

For visualization, exploratory networks were reduced using a predefined ranking rule. Only strong associations were considered, and visual subsets were initialized with the top ten positive and top ten negative associations ranked primarily by empirical p value and secondarily by absolute SparCC correlation. If fewer than 20 unique genera were represented, additional top ranked associations introducing new genera were added until at least 20 genera were included or a maximum of 35 edges was reached. This reduction was used only for figure readability and did not define statistical significance. Cross kingdom fungal bacterial associations were analysed using the same SparCC framework and reported as supplementary exploratory results.

Soil taxon association maps were generated separately from microbial association networks. Because soil variables were measured at the nursery site level, microbial count data from base and stem samples were first aggregated by location. Genus level abundances were then centered log-ratio transformed to account for compositional constraints of amplicon sequencing data [58], and Spearman correlations were calculated between transformed genus abundances and soil variables across the seven nursery sites. Empirical p-values were obtained using exact permutation testing across all possible permutations, followed by BH correction [62]. Soil taxon maps were used to visualize exploratory soil related compositional trends and were not used to infer microbial interactions.

All networks and association maps were visualized in Cytoscape v3.10.0 [66]. Edge color indicated association direction, edge width reflected absolute association strength, node size represented mean relative abundance, and node color indicated taxonomic order. In soil taxon maps, soil variables were displayed separately from microbial genera.

All analyses were conducted in Python 3.9 using pandas v1.5.3, numpy v1.24.2, scipy v1.10.1, scikit bio v0.5.8, seaborn v0.12.2, and matplotlib v3.7.1. Detailed outcomes of the ordination analyses, including the proportion of variance explained by principal coordinates and the statistical significance of group separations, are reported in the Results section. Scripts used for statistical analyses and visualization are available in the GitHub repository linked in the Data availability statement.

Results

Soil physicochemical and organic parameters

Soil conditions varied substantially across nursery sites (Table S1, Fig. S2–S3). Most soils were slightly acidic (pH/CaCl₂ < 5.5), with N5 and N4 showing the highest pH/CaCl₂, and N7 and N2 the lowest. Phosphorus was generally suboptimal, especially in N1 and N2, while N3 and N4 had the highest levels.

Organic matter content (C_ox) peaked in N5 and N3, whereas N7 and N6 had the lowest values. Sites rich in carbon also exhibited more advanced humification (high D_H, HA/FA ratio). In contrast, aromaticity (Q_4/6) was higher in N1, N2, and N6, reflecting more mature but less reactive organic pools.

Spearman correlation analysis (Fig. S4) revealed strong internal structure among soil parameters. C_ox was positively correlated with D_H (ρ = 0.68) and HA/FA (ρ = 0.83), while Q_4/6 was negatively correlated with all organic fractions (ρ = -0.75 to -0.82), indicating an inverse relationship between aromaticity and humification. Among nutrients Ca, Mg, and K were highly collinear (ρ = 0.75–0.89).

Complementarity of culture based isolation and HTAS in fungal community profiling

To assess how culture dependent isolation and HTAS resolve fungal communities associated with symptomatic Q. robur seedlings, we compared both approaches in terms of diversity, dominant taxonomic composition, and predicted trophic mode structure. Sequencing data are available in the NCBI SRA under accession numbers PV495278-PV495641, and fungal HTAS data are deposited under BioProject PRJNA1240605.

HTAS recovered higher Shannon diversity than culture-based isolation across the full dataset and within both plant compartments (Fig. S5). Because the data were not normally distributed, differences were evaluated using the Mann–Whitney U test. Shannon diversity was significantly higher in HTAS than in isolates for all samples combined (p = 0.0005), root samples (p = 0.0175), and trunk samples (p = 0.0111), with mean Shannon indices of 3.02 for HTAS and 2.14 for isolates (Table S2). These differences reflect the distinct biological fractions targeted by the two approaches rather than methodological superiority. HTAS captures sequence diversity from both culturable and non culturable fungi and may distinguish closely related OTUs, whereas cultivation selectively recovers viable fungi capable of growth under the applied isolation conditions.

Accordingly, the method comparison was interpreted primarily in terms of overlap and complementarity among dominant fungal groups. At the family level, HTAS produced a broader and more even taxonomic profile, with Botryosphaeriaceae, Erythrogloeaceae, Psathyrellaceae, Diaporthaceae, Pleosporaceae, and Nectriaceae among the most abundant families (Fig. 3). In contrast, cultivation based profiling was concentrated in fewer recovered families, including Nectriaceae, Aspergillaceae, Mucoraceae, Diaporthaceae, Hypocreaceae, and Erythrogloeaceae. The repeated detection of Diaporthaceae, Nectriaceae, and Erythrogloeaceae by both approaches indicates partial agreement in the dominant fungal component associated with symptomatic seedlings. However, the different relative representation of these families also shows method dependent recovery patterns, with HTAS capturing a broader community background and cultivation preferentially representing viable, readily culturable taxa.

Fig. 3.

Family level composition of fungal communities (HTAS) and cultivation based isolation. Treemaps show the ten most abundant fungal families in each dataset, with all remaining families grouped as “Others”. Rectangle size represents relative abundance within each method. Shared dominant families, including Diaporthaceae, Nectriaceae, and Erythrogloeaceae, indicate partial overlap between HTAS and cultivation-based profiling, whereas differences in family representation reflect method-dependent recovery of the seedling-associated fungal community

Predicted trophic mode composition provided additional support for this complementary interpretation (Fig. 4). In the HTAS dataset, the dominant category was Pathotroph + Saprotroph + Symbiotroph (37.5%), followed by Saprotroph (21.3%) and Pathotroph + Saprotroph (13.5%). In the cultivation based dataset, Pathotroph + Saprotroph + Symbiotroph was also the most abundant category (45.5%), suggesting that both approaches recovered fungi with potentially flexible ecological strategies. However, cultivation based profiles contained a higher proportion of unassigned taxa (22.7%) and were concentrated in fewer trophic categories, whereas HTAS provided a broader representation of predicted trophic assignments. Overall, these results indicate that HTAS and cultivation are not interchangeable diversity estimators, but complementary approaches: HTAS characterizes the wider seedling associated fungal community, while cultivation provides access to viable isolates relevant for downstream taxonomic verification, functional characterization, and pathogenicity testing.

Fig. 4.

Stacked barplots showing the normalized relative abundance of fungal trophic modes across seven nursery sites (a) HTAS based community profiles derived from root and trunk DNA for each site. (b) Culture-based community profiles derived from fungal isolates obtained from the same sites. Colors indicate trophic modes assigned using FUNGuild.

Microbial communities in oak roots and trunks

Alpha diversity of fungal and bacterial communities across tissues and forest sites

Bacterial HTAS data are available under BioProject no. PRJNA1240598. Community composition was dominated by Pseudomonas, followed by Rhizobium across all sites (Fig. S9).

Alpha diversity was assessed using Shannon, Simpson, and Chao1 indices to capture complementary aspects of microbial diversity, including evenness, dominance, and estimated richness. All diversity metrics were calculated on pooled samples representing each nursery and plant compartment.

Fungal alpha diversity varied among sites (Fig. 5, Tab. S4). N2, N3, and N5 exhibited the highest Shannon and Simpson values, suggesting relatively diverse and even fungal communities at the site level. In contrast, N4 exhibited reduced diversity across all metrics. Chao1 estimates indicated the highest species richness in N2 and N1, whereas N4 contained fewer rare taxa.

Fig. 5.

Boxplots Shannon, Simpson, and Chao1 alpha diversity indices for fungal (a) and bacterial (b) communities obtained via HTAS across seven nurseries. Due to the limited number of samples per locality, alpha diversity is presented descriptively, and no statistically significant differences were observed

Fungal diversity was slightly higher in stems than in bases across all indices (Fig. S10b, Tab. S4), but differences were not statistically significant (Shannon: p = 0.318; Simpson: p = 0.902; Chao1: p = 0.128).

In contrast, bacterial diversity showed clearer site and tissue associated patterns (Fig. 5, S10a, Tab. S3). N6 and N7 exhibited higher diversity across all metrics, while N5, N3, and N4 were consistently lower. Bacterial diversity was significantly higher in stems than in bases (Shannon: p = 0.011; Simpson: p = 0.018), with Chao1 showing a similar trend (p = 0.405, non-significant). This pattern may reflect differences in microhabitat heterogeneity between plant compartments.

Beta diversity of fungal and bacterial communities

Beta diversity analysis revealed significant compositional differences in fungal communities across forest nurseries. Both Bray–Curtis and Jaccard dissimilarity metrics indicated clear between-site variation, with Bray–Curtis values ranging from 0.21 (N4) to 0.96 (between N1 and N4), and Jaccard from 0.27 to 0.80 (Tab. S5, Fig. 6, Fig. S11). PERMANOVA indicated a significant effect of site for both metrics (Bray–Curtis: F = 1.95, p = 0.002; Jaccard: F = 1.90, p = 0.001), supported by ANOSIM (R = 0.585 and R = 0.782, respectively; both p < 0.005).

Fig. 6.

Beta diversity of fungal (a–b) and bacterial (c–d) communities derived from HTAS data of trunk and root samples. PCoA ordinations based on Bray–Curtis dissimilarity (a, c) demonstrate site specific clustering, with significant differences among localities confirmed by PERMANOVA (Fungi: F = 1.95, p = 0.002; Bacteria: F = 1.79, p = 0.015). Heatmaps (b, d) present pairwise Bray–Curtis dissimilarities among nursery sites, highlighting compositional differences across microbial communities

PCoA ordination explained 37.2% of the total variance and showed site-related structuring of fungal communities. Some sites (e.g., N4 and N1) appeared more closely grouped, whereas others were more separated in ordination space. Separation was more pronounced when using Jaccard distances.

Bacterial communities also exhibited significant site-associated structuring. PERMANOVA indicated a significant effect of site (Bray–Curtis: F = 1.79, p = 0.013; Jaccard: F = 1.73, p = 0.014) (Tab. S5), and ANOSIM yielded moderate but significant R values (R = 0.327 and 0.367, respectively; p < 0.01). PCoA ordination (Tab. S5, Fig. 6, Fig. S12) explained 56.5% of total variance and showed site related patterns, with some sites (e.g., N2 and N4) appearing more closely grouped.

Compared to fungi, bacterial communities showed less distinct separation based on Jaccard distances, suggesting a higher proportion of shared taxa across sites. Some sites (e.g., N3, N7) appeared more distinct.

Functional structure and environmental gradients

Internal functional structure: trophic group associations

Pairwise Spearman correlations among fungal trophic modes were assessed using nursery level raw abundances aggregated per site (Fig. S14, Table S6). After BH correction, two associations remained significant: a strong negative correlation between Pathotroph and Pathotroph + Saprotroph (ρ = -0.93, adjusted p = 0.0378), and a strong positive correlation between Saprotroph and Unassigned taxa (ρ = 0.93, adjusted p = 0.0378). All remaining pairwise associations were not significant after multiple-testing correction and are therefore interpreted as exploratory patterns rather than robust evidence of ecological interactions.

Relationships between trophic groups and soil parameters

Spearman correlation analysis of nursery level averages revealed limited associations between fungal trophic groups and soil properties (Fig. S13). After BH correction for multiple testing, only the negative correlation between Pathotroph + Symbiotroph abundance and soil Ca concentration remained significant (ρ = -0.96, adjusted p = 0.0472). Several additional nominal trends were detected before correction, including negative associations of Pathotroph + Symbiotroph with pH/CaCl₂ and Mg, a negative association of Pathotroph + Saprotroph + Symbiotroph with P, and positive associations of Saprotroph with pH/CaCl₂ and P. As these relationships were not retained after multiple-testing correction, they are interpreted cautiously as exploratory site-level patterns. No evident relationships were observed for organic matter-related variables.

Envfit analysis showed stronger associations of trophic composition with inorganic than with organic soil parameters (Fig. 7). Among inorganic variables, P, Mg, and K showed the highest vector strengths (‖r‖ = 0.74, 0.73, and 0.73, respectively), indicating that nutrient-related soil chemistry was most strongly aligned with variation along the first two ordination axes. In contrast, organic parameters showed overall weaker associations. Among these, DH exhibited a moderate relationship with trophic composition (‖r‖ = 0.42), whereas HS and the remaining organic variables showed only weak associations.

Fig. 7.

Principal Coordinates Analysis (PCoA) of fungal trophic group composition based on Bray–Curtis dissimilarity, overlaid with Pearson correlation vectors of soil parameters. Panel (a) shows the ordination with organic soil variables (DH, HS, HA, FA, HA/FA, Q4/6), whereas panel (b) shows the ordination with inorganic soil variables (pH/CaCl₂, P, K, Ca, Mg). The first two axes explained 74.6% of the total variation (PC1 = 55.76%, PC2 = 18.81%). Organic variables showed generally weak associations with trophic composition, with DH displaying the strongest relationship (|r|= 0.421). In contrast, inorganic parameters were more strongly associated with the ordination pattern, particularly P (|r|= 0.738), K (|r|= 0.733), and Mg (|r|= 0.730)

Relationship between fungal alpha diversity and soil parameters

Permutation based Spearman correlations between fungal Shannon diversity and soil parameters were evaluated at the nursery level (n = 7), using site level soil data and Shannon diversity averaged across the two plant compartments. Exact permutation testing was applied, corresponding to all 5040 possible permutations for seven observations, followed by BH correction. No nominally significant or FDR supported associations were detected (Fig. S15). The strongest positive trends were observed for HA (ρ = 0.64, p = 0.139, adjusted p = 0.707), Cox (ρ = 0.56, p = 0.206, adjusted p = 0.707), and HS and FA (both ρ = 0.54, p = 0.236, adjusted p = 0.707). Inorganic soil parameters, including pH/CaCl₂ and available P, K, Ca, and Mg, showed weak and non significant associations with fungal Shannon diversity.

Soil-driven structuring of fungal diversity

To assess the potential effects of soil parameters on community structure, envfit vector fitting was applied to the Bray–Curtis PCoA ordination. The first two axes explained 39.8% of the variation. Among the inorganic factors (Fig. 8), P (‖ρ‖ = 0.71) and pH/CaCl₂ (‖ρ‖ = 0.54) exhibited the strongest associations with fungal beta diversity. Organic variables (Fig. 8) showed weaker associations, with the HA/FA ratio (‖ρ‖ = 0.30) being the most notable.

Fig. 8.

PCoA of fungal community composition based on Bray–Curtis dissimilarity, overlaid with inorganic (a) and organic (b) soil properties and Shannon diversity index (point colour). Point shape indicates nursery locality (N1–N7). Environmental vectors represent Spearman correlations between soil variables and fungal community composition; arrow length reflects the magnitude of the correlation (‖ρ‖), and arrow direction indicates gradient orientation in the ordination space. Among inorganic variables, the strongest correlations were observed for phosphorus (P), pH/CaCl₂, and magnesium (Mg). Among organic variables, the strongest correlations were found for the HA/FA ratio and HS, followed by FA and Cox

Relationship between bacterial alpha diversity and soil parameters

Permutation based Spearman correlations between bacterial Shannon diversity and soil parameters were evaluated at the nursery level (n = 7), using site level soil data and Shannon diversity averaged across the two plant compartments. Exact permutation testing was applied, corresponding to all 5040 possible permutations for seven observations, followed by BH correction. No soil parameter showed an FDR supported association with bacterial Shannon diversity (Fig. S16). Calcium showed the strongest negative nominal trend (ρ = -0.79, p = 0.048), but this association did not remain significant after correction (adjusted p = 0.576). Other variables showed weaker non significant trends, including pH/CaCl₂ (ρ = -0.57, p = 0.200, adjusted p = 0.606), Mg (ρ = -0.54, p = 0.236, adjusted p = 0.606), and HS and FA (both ρ = -0.46, p = 0.302, adjusted p = 0.606).

Soil driven structuring of bacterial beta diversity

The first two PCoA axes explained 39.3% (PC1) and 16.3% (PC2) of the variation in bacterial community composition, indicating a moderate compositional gradient across samples (Fig. 9). To explore potential associations with environmental variables, vector fitting based on Spearman correlations between soil parameters and the first two PCoA axes was applied.

Fig. 9.

PCoA of bacterial community composition based on Bray–Curtis dissimilarity, overlaid with inorganic (a) and organic (b) soil properties and Shannon diversity index (point colour). Point shape indicates nursery locality (N1–N7). Environmental vectors represent Spearman correlations between soil variables and bacterial community composition; arrow length reflects the magnitude of the correlation (‖ρ‖), while arrow direction indicates the orientation of the environmental gradient in the ordination space. The strongest associations among inorganic variables were observed for Ca, pH/CaCl₂, and Mg, whereas Cox, HA, Q4/6, and HA/FA showed the strongest associations among organic variables

Among inorganic soil parameters (Fig. 9), available Ca, P, and pH/CaCl₂ showed the strongest associations with community composition (‖ρ‖ = 0.56, 0.32, and 0.52, respectively). The corresponding vectors were oriented toward samples with higher bacterial Shannon diversity (e.g., N7B, N7S), suggesting an association between soil nutrient status and community composition. Organic soil fractions (Fig. 9) showed comparable or slightly stronger associations. Cox, the HA/FA ratio, and HA exhibited relatively high vector magnitudes (‖ρ‖ = 0.49–0.54), indicating their contribution to the observed compositional gradients. Samples such as N6S and N3B were positioned along gradients defined by these variables.

Microbial association networks and soil taxon interactions

Soil related taxon association maps

Exploratory soil taxon association maps were generated to visualize relationships between site level soil parameters and CLR-transformed genus level fungal and bacterial abundances aggregated at the nursery location level (Tab. S8, S9, Fig. 10, S17). These analyses were not intended to infer microbial interactions, but to summarize potential associations between environmental gradients and microbial composition. To reduce visual complexity, only a restricted subset of the strongest associations was displayed, retaining up to 10 positive and 10 negative links per map based on |Spearman’s ρ|≥ 0.70 and ranking by empirical permutation-based p-values and absolute correlation strength. Displayed associations showed high absolute correlation values (|ρ|= 0.70–1.00), but none remained significant after Benjamini-Hochberg (BH) correction.

Fig. 10.

Exploratory CLR-Spearman soil taxon association maps for inorganic soil variables. Selected associations between inorganic soil parameters and fungal or bacterial genera are shown. Soil variables are displayed as grey rounded rectangles, fungal genera as circles, and bacterial genera as squares. Node color denotes taxonomic order, node size reflects mean relative abundance, and edge color indicates the direction of the association, with green edges representing positive and purple edges representing negative associations. Edge width corresponds to association strength. These maps are exploratory and do not imply direct ecological interactions or causality

The strongest visual patterns were mainly linked to inorganic soil parameters. In fungi, available P and pH/CaCl₂ contributed most prominently, with negative trends between P and Penicillium, Sticta, Diplodia, Dothiorella, and Trichoderma, and positive trends between pH/CaCl₂ and Filobasidium, Cortinarius, Coprinopsis, and Cladosporium. Organic fractions (Fig. S17) were more apparent in the fungal association map, particularly through relationships with Cadophora, Peniophora, Rhodosporidiobolus, and Podila. In bacteria, inorganic parameters, especially Ca, Mg, P, K, and pH/CaCl₂, dominated the visual structure, with positive trends involving Brenneria, Beijerinckia, Sinorhizobium, Nitrosospira, Pseudomonas, and Rahnella, and negative trends involving Psychrobacter, Paraburkholderia, Paracraurococcus, and Novosphingobium. Overall, these patterns suggest possible soil related compositional trends, but they do not provide evidence for direct microbial interactions or causal soil effects.

Microbial association patterns

Genus level intrakingdom association patterns were evaluated separately for fungi and bacteria using a SparCC based compositional framework (Tables S9–S12, Figs. 11 and 12). For each kingdom, 50 retained genera were tested, corresponding to 1,225 genus pairs per dataset. Strong associations were defined as |SparCC r|≥ 0.80, and empirical p-values were estimated using 999 permutations followed by BH correction across all tested pairs.

Fig. 11.

Exploratory SparCC based fungal association network. High magnitude compositional associations among fungal genera are shown as a reduced visualization. Edge color indicates association direction, edge width reflects absolute SparCC correlation strength, node size represents mean relative abundance, and node color denotes fungal order. No displayed association remained significant after BH correction; therefore, the network should be interpreted as exploratory

Fig. 12.

Exploratory SparCC based bacterial association network. High magnitude compositional associations among bacterial genera are shown as a reduced visualization. Edge color indicates association direction, edge width reflects absolute SparCC correlation strength, node size represents mean relative abundance, and node color denotes bacterial order. No displayed association remained significant after BH correction; therefore, the network should be interpreted as exploratory

High magnitude associations were detected in both datasets, with 117 fungal and 178 bacterial associations passing the |SparCC r|≥ 0.80 threshold. However, none remained significant after FDR correction. Given the limited number of pooled samples and the sparse compositional structure of the data, these high absolute correlation values were interpreted cautiously. The networks were therefore used only as exploratory visualizations of compositional association patterns, not as statistically supported microbial interaction networks.

For visualization, the networks were reduced to exploratory subsets of the strongest high magnitude associations. This resulted in a fungal subset of 20 edges connecting 22 genera and a bacterial subset of 23 edges connecting 20 genera. In the fungal subset, positive and negative associations were equally represented, with Psathyrella showing the highest connectivity. In the bacterial subset, negative associations were slightly more frequent, and the most connected genera were Phyllobacterium, Methylosinus, and Yersinia.

Cross kingdom fungal bacterial associations were also explored and are provided in the supplementary material (Tab. S12, Fig. S18). The curated interkingdom network contained 48 high magnitude associations among 41 genera, but no edge was BH FDR supported after correction. These networks are therefore presented as hypothesis generating summaries of possible association structure among dominant genera and should not be interpreted as evidence of direct ecological interactions.

Discussion

HTAS recovered a broader fraction of the fungal community than cultivation based isolation across most nurseries, consistent with the tendency of culture dependent workflows to favor fast growing and medium adapted taxa [67–69]. However, the site specific exception where cultivation recovered comparable or higher diversity shows that sequencing does not fully represent the viable culturable fraction. Rather than ranking the two approaches hierarchically, our results support their complementarity: HTAS provides broad community profiling, whereas cultivation yields living material for taxonomic verification and downstream ecological or pathogenicity assays [68, 69].

The divergence in trophic composition between HTAS and cultivation likely reflects both ecological structure and methodological selectivity. HTAS captured a larger share of symbiotrophic and slow growing saprotrophic lineages, whereas cultivation enriched opportunistic pathotrophs and generalist saprotrophs able to establish on artificial media. Functional assignments based on FUNGuild were useful for broad ecological interpretation, but they remain dependent on taxonomic resolution, database completeness, and the context dependence of fungal lifestyles [61, 70]. Trait databases such as FungalTraits may further broaden or modify ecological assignments relative to FUNGuild, particularly for endophytes, saprotrophs, and other plant associated fungi [70, 71].

At the nursery scale, fungal trophic structure appeared to covary more consistently with inorganic soil gradients than with humus related descriptors. In contrast, univariate relationships between fungal alpha diversity and individual soil variables were weak, indicating that richness and compositional turnover did not respond identically to edaphic variation. This distinction is ecologically relevant because broader soil microbiome studies identify pH and mineral nutrient status as major correlates of fungal community composition [33, 72, 73], while recent work in forest mineral soils suggests that site level physicochemical context may outweigh local mineral identity in structuring fungal assemblages [74]. Thus, our results are best interpreted as evidence for nursery level edaphic filtering of community composition, not as proof of direct causal effects of individual soil properties.

Fungal and bacterial communities also differed in their apparent sensitivity to nursery conditions. Fungal beta diversity showed clearer site dependence and stronger turnover than bacterial beta diversity, whereas bacterial alpha diversity showed only weak or nominal associations with soil chemistry. Such decoupling between alpha and beta diversity is plausible because environmental filtering may alter community composition without producing strong changes in local richness [72, 75]. Large scale studies similarly show that bacteria often respond strongly to broad physicochemical gradients such as pH, while fungal communities can retain stronger spatial, host associated, and forest context signals [33, 72, 76]. In European forests, fungal, but not bacterial, composition and diversity have also been linked to variation in tree growth and carbon storage, supporting the interpretation that fungal communities may carry stronger habitat specific information in woody systems [77, 78].

The association networks should be interpreted conservatively. Although most inferred intra kingdom and inter kingdom links were positive, correlation based networks from amplicon data cannot distinguish direct biotic interactions from shared responses to the same host or environment [79, 80]. This limitation is particularly relevant for compositional datasets, where relative abundance structure, sparsity, indirect dependencies, and normalization choices can distort network topology [57, 58, 81, 82]. Therefore, the SparCC networks presented here should be viewed as exploratory, hypothesis generating structures that highlight candidate taxa and soil variables for follow up testing rather than as evidence of facilitation, competition, or trophic dependency. The absence of FDR supported network edges further reinforces this cautious interpretation.

Despite these constraints, the results fit better within a pathobiome framework than within a simple one pathogen, one disease model. Pathobiome theory emphasizes that disease expression emerges from interactions among host condition, multiple microorganisms, and abiotic context [22, 83, 84]. Tree decline studies support this view, including acute oak decline as a polymicrobial disease syndrome [85] and recent pine decline studies showing distinct microbiota and metabolomic profiles in symptomatic and asymptomatic trees [86, 87]. However, because our sampling focused on symptomatic seedlings only, the present data characterize microbial communities associated with declining oak seedlings across nurseries but cannot identify decline specific dysbiosis or distinguish disease linked taxa from taxa reflecting site conditions or host compartment.

These limitations define the scope of interpretation rather than the value of the dataset. Pooling samples and analysing composite soil data improved comparability among nurseries and emphasized site level ecological patterns, but reduced the ability to resolve within site heterogeneity and individual seedling trajectories. A further methodological limitation concerns the use of 97% OTU clustering for fungal ITS data. Although ASV based workflows provide high sequence resolution, fungal ITS markers show substantial length variation, uneven interspecific divergence, intraspecific variation, and occasional intragenomic variation. For this reason, OTU based clustering remains a defensible approach in fungal ITS metabarcoding and may reduce over interpretation of fine scale sequence variants as biologically independent taxa [32, 88, 89]. Future work should combine symptomatic and asymptomatic seedlings, repeated temporal sampling, compartment resolved profiling, absolute quantification, sequence resolved analyses, and isolate based validation. Integrating cultivation with amplicon sequencing, metabolomics, transcriptomics, and targeted inoculation experiments would provide a stronger mechanistic bridge from descriptive microbiome patterns to functional inference and microbiome informed nursery management.

Conclusion

Symptomatic Quercus robur seedlings from forest nurseries harbored complex fungal and bacterial communities whose composition was shaped by nursery site, plant compartment, and soil chemical context. Amplicon sequencing and cultivation recovered partly overlapping but method dependent fungal fractions: sequencing captured a broader community background, whereas cultivation provided viable fungal isolates, including several families of phytopathological relevance.

Second, fungal and bacterial communities showed structured compositional variation across nursery sites and plant compartments, but these patterns should be interpreted at the level of pooled site by compartment libraries rather than individual seedlings. Third, soil chemistry, particularly inorganic parameters such as pH/CaCl₂, phosphorus, calcium, and magnesium, was associated with microbial community composition and fungal trophic structure, whereas alpha diversity showed only weak or non-significant relationships with individual soil variables.

Fourth, SparCC based genus level association analyses revealed high magnitude compositional association patterns, but no intrakingdom or interkingdom network edges remained significant after FDR correction. These networks therefore provide exploratory, hypothesis generating information rather than evidence of direct microbial interactions. Finally, because only symptomatic seedlings were analyzed, the study does not identify disease specific microbial signatures, but instead provides a community level and environmental context for microbial assemblages associated with declining oak nursery stock.

Overall, the main finding is that pathogen detection in symptomatic oak nursery seedlings should be interpreted within a broader microbiome and soil context, and future studies should combine asymptomatic controls, individual level replication, temporal sampling, and isolate based functional validation to test these exploratory patterns mechanistically.

Abbreviations

16S rRNA

16S ribosomal ribonucleic acid

BLASTn

Basic Local Alignment Search Tool (nucleotide)

Chao1

Chao1 richness estimator

CLR

Centered log-ratio transformation

COX

Oxidizable carbon content

D_H

Degree of humification

envfit

Environmental fitting

FA

Fulvic acids

FA/HA

Ratio of fulvic acids to humic acids

FastQC

Quality control of sequencing data

HA

Humic acids

HS

Humic substances

HTAS

High-throughput amplicon sequencing

ITS

Internal transcribed spacer

ITSx

Bioinformatics tool for extracting ITS regions

N1–N7

Codes for the studied forest nurseries (Nursery 1 to Nursery 7); suffix Stem, B – root (base)

OTU

Operational taxonomic unit

PC1

Principal coordinate 1

PC2

Principal coordinate 2

PCoA

Principal coordinates analysis

PERMANOVA

Permutational multivariate analysis of variance

p_perm

Permutation-based p-value

Q₄/6

Colour quotient

TSS

Total sum scaling

UNITE

Database of fungal ITS sequences

UPARSE

Algorithm for clustering sequences into operational taxonomic units

USEARCH

Software for sequence processing, including clustering and homology search

V3–V4

Hypervariable regions of the bacterial 16S rRNA gene

Author contributions

LF designed the study, performed laboratory experiments, conducted bioinformatic processing, analyzed the data, and wrote the manuscript. MMG assisted with laboratory experiments and revised the manuscript. PŠ conducted soil analyses and revised the manuscript. MT contributed to experimental design, provided consultation, and revised the manuscript. AE supervised the study, contributed to experimental design, revised the manuscript, and performed Illumina sequencing together with LF.

Funding

This outcome was supported by the Internal Grant Schemes of Mendel University in Brno, Internal Grant Agency of the Faculty of Forestry and Wood Technology, Project no. LDF_VP_2021028. The experiment was also supported by the OP VVV project supporting young researchers no. CZ.02.1.01/0.0/0.0/16_017/0002334.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.