The study was conducted at the Mukhrino field station and the adjacent Mukhrino Bog in the middle taiga zone of Western Siberia, near Khanty-Mansiysk (60.89, 68.68) (Lapshina et al. 2015). The bog is an ombrotrophic peatland complex of approximately 10 × 15 km, situated on the left terrace of the Irtysh River close to its confluence with the Ob. The vegetation comprises typical ombrotrophic to oligo-mesotrophic communities, with two dominant types: pine–dwarf shrub–Sphagnum treed bogs (dominated by Pinus sylvestris, Chamaedaphne calyculata and Sphagnum fuscum) and open graminoid–Sphagnum lawns (dominated by Scheuchzeria palustris, Carex limosa and Sphagnum balticum).

To compare fungal communities across substrates, we conducted a systematic eDNA sampling campaign from 2021 to 2023, collecting 266 samples from four main substrates: peat (82 samples), plant litter (44 direct litter samples + 40 from experimental decomposition litter bags), wood (60 samples), and mycorrhizal roots (40 samples) (Fig. 1). The molecular survey was designed building upon 11 years of sporocarp-based mycological surveys (sporocarp monitoring, opportunistic collections and morphological identification) that provided a reference baseline for the visible component of the fungal community.

To study the fungal community of peat, six plots were located in two habitat types: treed bogs and sphagnum lawns. Each plot was sampled regularly in June, July, August and September. Several shoots of dead Sphagnum were collected from 10 points 5 m apart within each plot to create a composite sample of approximately 5 g field weight. The general sampling depth was 2–30 cm below the surface. Additional samples were collected at deeper depths (1 down to 3 m) in August and September to cover the anoxic peat layer. In total, 46 samples from the upper aerobic peat and 36 samples from deeper anoxic peat were collected (Table 1). Samples were flash-frozen in the field and stored at –20 °C until processing.

Sporocarp-based observations of terrestrial macrofungi (including peat saprotrophs and mycorrhizal species) were obtained through plot-based sporocarp monitoring and opportunistic surveys over 11 years (Filippova and Lapshina 2019; Filippova et al. 2023).

The fungal community of plant litter was studied by collecting leaf litter from six dominant bog plant species: Rhododendron groenlandicum, Chamaedaphne calyculata, Rubus chamaemorus (in treed bog habitats), Andromeda polifolia, Eriophorum vaginatum and Scheuchzeria palustris (in sphagnum lawn habitats). Litter was picked randomly from the surface of Sphagnum over an area of approximately 10 m² in each plot. Sampling was performed in the same plots and on the same dates as for peat (June, July, September), totalling 44 samples of freshly senesced litter (Table 1). Samples were dried at 40 °C and stored dry until processing.

In addition, 40 samples of decomposed litter from experimental litter bags (deployed for up to six years with different plant species) were included to assess successional changes during decomposition (Table 1). These samples were processed following the same eDNA protocol as the fresh litter.

To study the fungal community of early-decaying wood, we used a standardised wooden dowel of three tree species: pine (Pinus sylvestris), larch (Larix sibirica) and birch (Betula pubescens) following Shumskaya et al. (2025). Sterilised wooden dowels were buried in the upper peat surface in treed bog habitats and retrieved at two-week intervals throughout the growing season. A total of 60 wood samples were collected (Table 1). Dowels were dried at 40 °C and stored dry until processing.

To study the mycorrhizal community of bog trees, we collected ectomycorrhizal root tips of two common bog-dwelling tree species: Pinus sylvestris and Pinus sibirica. Roots were collected in two localities (Mukhrino and Shapsha, approximately 30 km apart) to cover spatial variability. In each locality, 5–10 trees growing about 10 m apart were selected. Approximately 30 g of fine roots were extracted from several locations around each tree trunk, cleaned from debris and collected in Eppendorf tubes. Sampling was performed twice a year (June and September), totalling 40 samples (Table 1). Samples were flash-frozen in the field and stored at –20 °C until processing.

All eDNA samples were processed following standardised protocols. Samples were lyophilised (except wood dowels, which were already dried) and homogenised. Total genomic DNA was extracted from 0.05 g of homogenised material per sample using a commercial soil DNA kit (SileksMagNA). The ITS2 fungal barcode region was amplified using primers fITS7 and ITS4 (White et al. 1990; Ihrmark et al. 2012). Libraries were prepared following the Illumina 16S Metagenomic protocol and sequenced on an Illumina MiSeq platform (v.3 chemistry, 2 × 300 bp), yielding an average of 150,000–200,000 raw paired-end reads per sample.

Raw paired-end reads were processed in QIIME 2 (v.2025.4.0) (Bolyen et al. 2019) following our standardised pipeline (Ishmanov 2025). Adapters were trimmed with cutadapt. Reads were merged using VSEARCH (min merged length 150 bp, max ambiguous bases 5, max expected errors 3, max mismatches 20, min overlap 8 bp). Quality filtering was performed with the quality-filter plugin (min Q-score ≥ 20, max ambiguous bases ≤ 2). Identical sequences were dereplicated and representative sequences were clustered de novo into operational taxonomic units (OTUs) at 99% similarity using VSEARCH; this threshold was chosen to minimise sequencing errors while preserving biologically relevant intraspecies variation. Chimeras were detected and removed with VSEARCH uchime_ref against the UNITE database (v.10.0 dynamic).

OTU taxonomy was assigned via a two-step approach. First, representative sequences were classified with the naïve Bayesian classifier (scikit-learn) trained on UNITE, yielding probabilistic predictions across all taxonomic levels. Second, all representatives were validated using BLAST+ (v.2.16.0) against the same UNITE reference, retaining up to 50 top hits per query in tabular format. For each OTU, the best hit with the highest percent identity was selected as the final assignment, linking the sequence to the corresponding UNITE SH.

The final taxonomy was determined by integrating BLAST assignments, based on BLAST percentage identity: for hits with ≥ 98% identity and a valid species name, the species-level assignment was accepted; for hits below 98%, the species assignment was discarded.

Community analyses were conducted in Python using pandas, numpy and scikit-bio (code available on GitHub: Filippova (2025)). Alpha-diversity indices (observed SHs, Shannon) and beta-diversity (Bray-Curtis dissimilarity) were calculated. To test the influence of substrate type and environmental parameters (Table 1) on community structure, we employed PERMANOVA and indicator species analysis. Statistical significance was assessed at p < 0.05.

All primary metabarcoding datasets are publicly available: raw sequence reads are deposited in NCBI SRA (PRJNA1007262), processed occurrence data on GBIF (Filippova et al. 2024; Filippova et al. 2026) and the complete set of input tables (including metadata, taxonomy, checklist, read counts, sporocarp records) together with the full analysis code on Zenodo (Filippova 2026, https://doi.org/10.5281/zenodo.21492477).

Sequencing of the ITS2 region from 266 substrate samples produced 22.7 million high-quality reads after bioinformatics processing. Clustering at 99% similarity yielded 295,050 OTUs, of which 95.9% were singletons. Assignment against UNITE (v.10.0 dynamic) identified 8,932 fungal species hypotheses (SH). Amongst these, 39.8% were singletons (3,554 SH), while 46.2% (4,128 SH) occurred in ≥ 3 samples and comprised 99.8% of all reads. All SH were retained to capture rare taxa.

This molecular dataset spanned 14 fungal phyla, 862 genera and 1,223 species, with Ascomycota (66% of sequences) and Basidiomycota (33%) being dominant. The taxonomic richness recovered varied substantially across substrate types and sampling events. For example, the 84 litter samples collectively yielded 7,297 SHs, while the 40 mycorrhizal root samples yielded 1,091 SHs. Taxonomic resolution was limited, with only 13.7% of SH assigned to named species, while the majority remained resolved at higher taxonomic levels (genus: 17.2%, family: 14.5%, order or above: 54.6%).

In comparison, the cumulative output of 11 years of sporocarp-based observations documented 208 fungal species across four ecological groups (barcoding macrofungi, lichens, plant litter fungi and wood decay fungi). Thus, the metabarcoding approach detected approximately 40-fold more species-level units than sporocarp-based observation.

Jaccard similarity analysis revealed a clear ecological gradient in fungal community composition across substrate types (Fig. 2, upper panel). At the species level, the highest similarity was observed between litter and litter bag (0.42), followed by peatupper with litter (0.31) and litter bag (0.30). Standardised wood (dowels) showed moderate similarity with the anoxic peat layer (0.32) and upper layer (0.27), while mycorrhizal samples exhibited the lowest similarity with all other substrates (range: 0.19–0.31). This pattern was consistent across genus and family levels, with mean Jaccard similarity increasing from 0.25 (species) to 0.40 (genus) and 0.56 (family), while the relative ranking of substrate pairs remained unchanged.

In contrast to the structured similarity observed within the metabarcoding data, comparison of the sporocarp-based checklists revealed extremely low overlap (Fig. 2, lower panel). Across all four groups (barcoding macrofungi, lichen survey, plant litter survey, wood decay survey), Jaccard similarity coefficients was below 0.05 at all taxonomical levels.

Species richness estimation (based on Chao1 index), calculated from metabarcoding data, suggested a total species pool of approximately 11,524 SHs across the entire ecosystem, indicating that our sequencing effort captured about 77.5% of the estimated total fungal diversity at the site. The overall Shannon diversity index for the ecosystem was 5.27, reflecting a highly diverse community.

At the level of individual samples, the mean observed richness was 251 SHs per sample (± 254 SD; range: 9–1,350) and a mean Shannon index of 2.3 (± 0.87 SD; Table 2). Alpha diversity metrics varied significantly across substrate types. Peat samples exhibited the highest mean sample-level diversity (Shannon index = 2.8), followed by litter (2.5) and mycopins (1.8). In contrast, mycorrhizal roots supported the least diverse communities (mean Shannon index of 1.4).

Fungal community composition, analysed via PCoA, based on Bray-Curtis dissimilarity, was strongly structured by substrate type (Fig. 4). PERMANOVA confirmed that substrate type significantly structured fungal communities (Pseudo-F = 10.82, p = 0.001). The ordination plot revealed a distinct separation of mycorrhizal samples from all saprotrophic substrates, while litter samples occupied a central position, partially overlapping with upper oxic peat layer. The compact confidence ellipses for most substrate groups indicate high within-group community homogeneity.

To further explore the drivers of community composition within each substrate type, we conducted a series of PERMANOVA tests (Suppl. material 1). The results are summarised below:

The magnitude of the effects, as indicated by PERMANOVA F-statistics, varied considerably across factors (Suppl. material 1). The strongest effect was observed for substrate type (F = 10.82), followed by habitat within peatupper (F = 8.66), plant host in mycorrhizal roots (F = 5.86) and plant host in litter (F = 4.11). Weaker, but significant effects were detected for season (mycorrhiza, F = 2.94), series (litter bag, F = 2.71; mycopins, F = 2.40) and depth within peatupper (F = 1.70). This hierarchy of factors highlights the primary importance of substrate type and habitat in structuring fungal communities, followed by host-specific and temporal effects that operate within particular substrate contexts.

Indicator species analysis identified 268 fungal species hypotheses significantly associated with specific substrate types. The highest number of indicators was found in mycorrhizal roots (85 SH, mean point-biserial correlation (Rpb) = 0.26), followed by peatstrat (82 SH, Rpb = 0.26) and peatupper (70 SH, Rpb = 0.27). In contrast, standardised wood yielded only 23 indicators and litter had the fewest (8 indicators, Rpb = -0.045), suggesting that litter-associated communities are largely composed of species shared with other substrates (Table 3).

The taxonomic composition of indicator species varied markedly across substrates at all taxonomic levels (Suppl. material 2). At the phylum level, Ascomycota dominated all substrates, ranging from 67% of indicators in litter to 39% in mycorrhiza. Notably, Rozellomycota (4 SH) was uniquely associated with deep peat, suggesting a specialised anaerobic community. The most striking differences emerged at the family and genus levels. Deep peat was characterised by Myxotrichaceae (6 SH) and Mollisiaceae (5 SH), with the genus Oidiodendron (6 SH) as the most abundant indicator. In contrast, the upper peat layer was characterised by ectomycorrhizal families Thelephoraceae (4 SH) and Russulaceae (3 SH), with corresponding genera Thelephora and Russula (2 SH each). Mycorrhizal roots were dominated by Piskurozymaceae (3 SH), while litter had only two identifiable indicator families (Aspergillaceae and Clavulinaceae) and a single genus (Penicillium).

Within the litter layer, partitioning samples by plant host revealed the high number of host-specific indicator taxa: 40–123 indicator SH per host species (S. palustris: 123; R. groenlandicum: 118; E. vaginatum: 92; V. oxycoccus: 89; C. calyculata: 81; R. chamaemorus: 52; A. polifolia: 40; Suppl. material 3). All seven plant hosts had significant indicator SH with high average point-biserial correlations (Rpb = 0.55–0.59), indicating strong and consistent associations between indicator fungi and their respective host plants. At the higher taxonomic level, Leotiomycetes (particularly the order Helotiales) dominated the indicator communities of all hosts (10–39 SH per host). Several highly specific plant–fungus associations were identified: Exobasidium (6 SH) was exclusively associated with C. calyculata, consistent with its known role as a gall-forming pathogen on ericaceous shrubs; Nagrajchalara (5 SH) and Venturia (5 SH) were associated with R. groenlandicum; Rhytismataceae (3 SH) were characteristic of E. vaginatum; and Coleophoma (3 SH) and Diplocarpon (2 SH) were indicators of V. oxycoccus and R. chamaemorus, respectively.

Indicator species analysis from the upper peat layer by the factor of habitat identified 22 significant indicator SH at the genus level, with 12 associated with Treed bog and 10 with Sphagnum lawn. The Treed bog indicators showed a positive mean correlation (Rpb = 0.19), while Sphagnum lawn indicators had a negative mean correlation (Rpb = -0.117). Notably, ectomycorrhizal genera commonly observed as sporocarps in Treed bog, such as Cortinarius, Suillus, Hebeloma and Russula, did not emerge as significant indicators at the genus level in the metabarcoding data.

To complement the metabarcoding data, we analysed 11 years of macrofungi sporocarp observations from the same peatland plots. At the species level, indicator analysis identified 45 indicator species for Treed bog and 23 for Sphagnum lawn. Treed bog was characterised by a diverse suite of ectomycorrhizal indicators, including multiple Cortinarius species, Hebeloma incarnatulum, Suillus spp., Lactarius and Russula spp., alongside saprotrophic litter fungi (Gymnopus, Mycena). The Sphagnum lawn was characterised by a smaller, but ecologically distinct set of indicators. Four species emerged as positive indicators: Galerina sphagnicola, Arrhenia bigelowii, Galerina tibiicystis and Hypholoma elongatum.

The contrasting results between the two approaches prompted a direct comparison of genus presence in both datasets. Of the 37 genera detected in fruiting body surveys, 32 were also detected in metabarcoding data, with only five genera completely absent (Fig. 5). In metabarcoding data, the mycelium of ectomycorrhizal taxa Cortinarius, Suillus and Hebeloma was widely distributed in both habitats: Suillus was detected in 100% of Treed bog and 69.6% of Sphagnum lawn samples; Cortinarius in 95.7% and 82.6%, respectively; Hebeloma in 60.9% and 39.1%. Overall, metabarcoding showed consistently higher detection rates (average 22.8% in Treed bog, 15.5% in Sphagnum lawn) compared to fruiting body records (20.8% and 9.8%, respectively) at the plot level, reflecting the persistent presence of mycelium versus the ephemeral nature of fruit bodies.

Indicator species analysis within mycorrhizal samples identified only four significant SH associated with plant host. Two SH (belonging to the genera Suillus and Leptodontidium) were positive indicators of Pinus sibirica (Rpb = 0.604) and the same two SH were negative indicators for Pinus sylvestris. No other SH showed significant host specificity.

To further investigate whether host-specific taxa exist below the indicator threshold, we examined the distribution of SH within ectomycorrhizal genera traditionally known to associate with pines (Cortinarius, Suillus, Hebeloma, Inocybe, Lactarius). This revealed that, while most SH were shared between the two hosts, some unique SH were present: for Cortinarius, 11 SH were unique to P. sylvestris and one to P. sibirica (out of 136 total CortinariusSH); for Suillus, five SH were unique to P. sylvestris and two to P. sibirica (out of 28 total SuillusSH). However, these unique SH occurred in fewer than three samples and had very low relative abundances (< 0.05% in most cases), preventing them from reaching statistical significance as indicators after FDR correction.

Functional annotation of fungal community revealed by metabarcoding using the FunGuild database (Nguyen et al. 2016) resulted in functional assignments for 44.4% (3,969 out of 8,932) of SH, while the remaining 55.6% remained unassigned. This moderate assignment rate reflects the limited taxonomic resolution of the checklist, as FunGuild assignments rely primarily on genus- or family-level identification. Amongst assigned taxa, the most common trophic mode was Saprotroph (28.0% of assigned), followed by mixed trophic strategies including Pathotroph-saprotroph-symbiotroph (9.1%) and Pathotroph-Saprotroph (6.7%). Pure Symbiotrophs accounted for 6.6% of all SH, while pure Pathotrophs were the least abundant (2.8%).

Analysis of trophic mode distribution across substrate types revealed clear differences in functional composition (Fig. 6). Mycorrhizal roots were dominated by Symbiotrophs (80.9% of relative abundance), reflecting the obligate biotrophic nature of mycorrhizal fungi. In contrast, all saprotrophic substrates – litter (37.3% Saprotrophs, 30.3% Pathotrophs, 32.4% Symbiotrophs), litter bag (43.2%, 26.7%, 30.1%), mycopins (45.0%, 25.2%, 29.8%), peatstrat (55.1%, 22.6%, 22.3%) and peatupper (39.5%, 26.2%, 34.4%) – showed a higher proportion of saprotrophic and pathotrophic modes, consistent with their roles as decomposer substrates.

The estimated fungal species richness (~ 11,500 SHs) in the Mukhrino Bog underscores peatlands as reservoirs of fungal diversity. This number substantially exceeds earlier global checklist-based estimates of ~ 1,500 peatland-associated species (Filippova and Rudykina 2024) and highlights the limitations of knowledge derived solely from sporocarp-based studies. For comparison, metabarcoding studies in other boreal peatlands have reported comparable richness levels: for instance, Sun et al. (2016) detected 1,000–2,000 OTUs per site in Canadian bogs, while Asemaninejad et al. (2017a, 2017b) found 2,000–4,000 OTUs across boreal peatlands, though methodological differences (primer choice, sequencing depth, filtering thresholds) complicate direct comparisons. Elliott et al. (2015) reported relatively low fungal OTU richness (300 OTUs at 97% similarity) in degraded and restored blanket bogs in the UK, with Ascomycota (65%) and Basidiomycota (29%) dominating, comparable to our phylum-level distribution.

The high diversity and distinct taxonomic profile confirm our hypothesis that northern West Siberian peatlands, representing one of the world’s largest continuous peatland areas, host a unique and under-explored fungal microbiome, filling a critical gap on the global map of fungal eDNA data.

Our analysis identified substrate type as the principal factor structuring the fungal community, explaining the strongest dissimilarities in beta-diversity and community composition (PERMANOVA F = 10.82, p = 0.001). The Jaccard similarity analysis revealed a clear ecological gradient across substrates: the high similarity between litter and litter bag confirms that litter bags serve as a reliable experimental proxy for natural litter decomposition.

The moderate similarity between litter and upper peat layer (0.313) reflects the vertical continuum of organic matter incorporation into the peat profile, while the similarity between standard wood dowels and lower anoxic peat (0.319) suggests shared wood- and peat-associated decomposer communities. The consistently low similarity of mycorrhizal samples with all other substrates (Jaccard < 0.31) confirms the ecological separation between biotrophic mycorrhizal fungi and saprotrophic communities. The stability of these patterns across species, genus and family levels indicates that the observed ecological structure is robust and not an artefact of taxonomic resolution.

This strong substrate effect is consistent with the findings of Vítovcová et al. (2024), who demonstrated that fungal assemblages in bogs are predominantly structured by substrate identity rather than site, with only a few OTUs behaving as generalists and the most specific fungi associated with particular plant litter types (e.g. Calluna vulgaris, Trichophorum cespitosum, Drosera rotundifolia). In their study, the effect of substrate accounted for 89.2% of explained variation, while site explained only 0.2%, mirroring our observation that substrate type overrides spatial variation in structuring fungal communities.

Within specific substrates, other ecological filters became apparent. The homogeneity of fungal communities in the anaerobic catotelm (100–300 cm depth) contrasts with the clear influence of microtopography (treed bog vs. lawn) in the upper peat layer. This suggests that, while oxygen availability is a fundamental barrier structuring the deep peat community, finer-scale environmental gradients and temporal dynamics shape the active surface community. This pattern is consistent with the findings of Lin et al. (2012), who documented a distinct vertical stratification of fungal communities in a bog–fen complex, with Ascomycota and Basidiomycota prevalent near the surface and Zygomycota dominant in the subsurface. They further reported that fungal rRNA gene abundance declined with depth at both sites.

The significant change in wood-associated communities over time, but not amongst wood species, suggests that successional dynamics override substrate chemistry on the standardised wooden dowels used in this experiment. However, this result should be interpreted with caution: the dowels represent a homogenised, processed substrate lacking bark and likely other natural features that typically differentiate woody species in nature. In contrast, Shumskaya et al. (2025), using a similar standardised wood substrate, observed differences in fungal communities between hardwood and softwood, suggesting that processed wood may still retain some host-specific signals under certain conditions. Nevertheless, both studies agree on the strong temporal successional dynamics of wood-associated fungal communities, confirming that time is a primary driver of community change on these substrates.

The indicator species analysis revealed a clear hierarchy of substrate specialisation in fungal communities. Mycorrhizal roots harboured the highest number of indicator species (85 SH), confirming that this substrate represents a highly specialised ecological niche with a distinct fungal community. This is consistent with the obligate biotrophic nature of mycorrhizal fungi, which form intimate symbioses with plant roots and are expected to show strong substrate specificity.

Peat substrates also harboured a large number of indicator species (oxygenic peat: 82, anoxic peat: 70), reflecting the distinct environmental conditions of these layers. Notably, the deep anoxic layer had a unique set of indicators, including Rozellomycota (4 SH) and the genus Oidiodendron (6 SH). Rozellomycota are known as early-diverging fungi often associated with anaerobic or low-oxygen environments and their presence as indicators of deep peat aligns with the anoxic conditions of this layer. Oidiodendron is a cold-adapted genus commonly found in peat soils, consistent with the permanently cold conditions of deep peat layers. The upper peat layer was characterised by ectomycorrhizal families Thelephoraceae and Russulaceae, with corresponding genera Thelephora and Russula. The presence of these ectomycorrhizal taxa in the upper peat layer is consistent with the occurrence of tree roots (Pinus sylvestris, P. sibirica) in this horizon.

In contrast, litter had the fewest indicator species when combined with litter bag (8 SH), but, when separated, litter yielded 38 indicators and litter bag 23. This indicates that fresh and decomposing litter support distinct fungal communities and the low number of indicators for combined litter reflects the ecological heterogeneity of the litter layer.

Indicator species analysis revealed that, while litter as a whole harboured only eight indicator SH, the same samples analysed at the plant-host level yielded 40–123 indicators per host species. This is explained by the high ecological heterogeneity of the litter layer, which is not a uniform substrate, but rather a mosaic of microhabitats defined by different plant species. Aggregating litter samples across all plant hosts dilutes host-specific signals: a fungus that is consistently associated with a particular plant species may not occur on other litter types and, thus, fails to appear as a general indicator of litter when compared against other substrate types. Vítovcová et al. (2024) similarly found that sequencing conventional soil core samples (peat and mixed litter) captured only 9–36% of the total OTU pool detected when multiple microhabitats were sampled and species accumulation curves did not reach an asymptote even after analysing 13 different substrates. Their observation that most OTUs occurred in only 1–2 substrate types parallels our finding that host-specific fungi in the litter layer are detectable only when samples are partitioned by plant identity rather than aggregated.

Each of the seven plant hosts supported a distinct set of indicator taxa, with S. palustris and R. groenlandicum showing the highest number of indicators (123 and 118, respectively). The taxonomic composition of these indicators further supports host specificity: Exobasidium was exclusively associated with C. calyculata, Nagrajchalara and Venturia with R. groenlandicum, Rhytismataceae with E. vaginatum and Coleophoma and Diplocarpon with V. oxycoccus and R. chamaemorus, respectively. These associations suggest that both pathogenic and saprotrophic fungi contribute to host-specific community differentiation in the litter layer. From a conservation perspective, the high number of indicators associated with S. palustris and R. groenlandicum suggests that these plants are particularly important hosts for fungal diversity in the litter layer. This parallels the findings of Vítovcová et al. (2024), who showed that co-occurring plant species – particularly peatland specialists such as V. oxycoccus, E. vaginatum and specific Sphagnum species – can serve as reliable indicators of suitable conditions for peatland fungal indicators, even when fruit-bodies are absent. As Vítovcová et al. (2024) emphasised, the presence of peatland specialists is strongly linked to the availability of specific host plants and restoration success depends on restoring both the physical environment (water table, peat layer) and the dominant plant species composition that supports associated fungal diversity. Our results from an undisturbed reference bog thus provide a baseline for understanding the natural levels of trophic specialisation that restoration efforts should aim to achieve.

In contrast to the strong host-specific signal observed in the litter layer, the mycorrhizal communities associated with the two pine species showed remarkably weak differentiation. The majority of SH in these genera were shared between the two hosts, indicating that the PERMANOVA signal reflects a gradual turnover of many taxa rather than strong host-specific indicators. The diagnostic analysis of Cortinarius and Suillus revealed that host-specific SH do exist (11 unique to P. sylvestris and one to P. sibirica for Cortinarius; five and two for Suillus). However, these unique SH are rare components of the community: most occur in fewer than three samples and have very low relative abundances (< 0.05%). This pattern of weak host specificity between closely-related pine species is consistent with the findings of Aučina et al. (2019), who studied ECM fungal communities of Scots pine along a peatland–forest gradient in Lithuania. They found that the majority of ECM taxa were shared across habitats, with community composition shifting gradually rather than showing discrete habitat-specific assemblages. Interestingly, they also documented an increase in ECM fungal taxa richness from peatland (23 taxa) to forest (31 taxa), with the peatland site dominated by medium-distance fringe and long-distance exploration types – mostly Cortinarius and Suillus – which are adapted to waterlogged conditions due to their hydrophobic mycelia. In contrast, forest sites hosted more hydrophilic taxa of the contact and short-distance types, represented by Russula and Lactarius.

The comparison between metabarcoding and sporocarp-based surveys revealed that mycelium of ectomycorrhizal genera (Cortinarius, Suillus, Hebeloma) was detected by metabarcoding in both Treed bog and Sphagnum lawn, whereas sporocarps of these taxa were recorded almost exclusively in Treed bog. The widespread mycelial presence in Sphagnum lawn – often exceeding 80% of samples for Cortinarius and Suillus – can be explained by two non-exclusive mechanisms. First, extensive mycelial networks of ectomycorrhizal fungi may extend beyond the immediate root zone of trees, possibly through connections with ericaceous shrubs or as saprotrophic mycelium. Second, the detection may reflect the presence of vital propagules (spores or mycelial fragments) in the peat matrix that remain viable, but are incapable of fruiting under water-saturated conditions or in the absence of a suitable host. This discrepancy is particularly striking given the large spatial scale of the system – the Sphagnum lawns extend up to 500 m from the nearest trees, making it unlikely that the detected DNA originates from active mycelium connected to root systems. Whether ectomycorrhizal fungi can maintain a non-fruiting, vegetative mycelium in the complete absence of a host remains an open question. Distinguishing between these alternatives will require complementary approaches such as RNA-based metatranscriptomics to assess metabolic activity or stable isotope probing to trace carbon flow, which could help resolve whether ectomycorrhizal mycelium in Sphagnum lawn represents an active functional component of the community or a legacy of inputs from adjacent forested microhabitats.

The functional group analysis revealed a clear ecological dichotomy between mycorrhizal roots and other saprotrophic substrates. Mycorrhizal roots were dominated by symbiotrophs (80.9%), reflecting the obligate biotrophic nature of mycorrhizal fungi. This finding is consistent with the strong indicator status of ectomycorrhizal genera in Treed bog habitats and confirms that the mycorrhizal root samples captured a functionally distinct community.

In contrast, all saprotrophic substrates – litter, litter bag, mycopins and both peat layers – were dominated by saprotrophs and pathotrophs, with saprotrophs accounting for 37–55% of relative abundance. The highest proportion of saprotrophs was observed in lower anoxic peat layer (55.1%), reflecting the dominance of decomposer fungi in the anoxic peat layer.

The presence of symbiotrophs in saprotrophic substrates (30–34% in litter, litter bag and upper peat) is noteworthy. This likely reflects the contribution of ectomycorrhizal extraradical mycelium and mycorrhizal root fragments to these substrates, particularly in upper peat layer, where tree roots are abundant.