Introduction

Old-growth coniferous forests in Scandinavia are usually composed of trees of different species and ages, but today most forest areas are managed into even-aged stands, representing stages from clear cuts to mature forests. This has major effects on biodiversity and potential impacts on carbon (C) storage (Jandl et al., 2007). In uniform, young stands, undergrowth vegetation is markedly different from that in old stands (Hart and Chen, 2006), as is nutrient and water availability, soil microclimate, and litter quantity and quality (Jurgensen et al., 1997). These differences are likely to be paralleled by shifts in fungal community composition with potential feedbacks on C sequestration, nutrient cycling and plant production (Clemmensen et al., 2015). Since fungi dominate production of extracellular enzymes involved in organic matter decomposition in forest soils (Schneider et al., 2012; Eichlerová et al., 2015; Hesse et al., 2015), shifts in fungal communities induced by management are, supposedly, reflected in the quantity and composition of enzymes.

In boreal forest soils the organic horizon becomes stratified, as fresh litter constantly is deposited on top of more decomposed organic matter. Different functional guilds of fungi have been shown to occupy spatially distinct vertical niches within the organic horizon (Baldrian et al., 2012; Clemmensen et al., 2015). In the litter layer, saprotrophic fungi are the principal decomposers of plant-derived litter, producing extracellular enzymes to degrade biopolymers and release metabolic resources. During decomposition, saprotrophic fungi retain and re-allocate nutrients within their mycelium and are, by this, able to overcome local nutrient limitation (Boberg et al., 2014). In the deeper, more decomposed humus layers, biotrophic mycorrhizal fungi dominate. Receiving carbohydrates from their plant host, ectomycorrhizal fungi generally have a reduced set of genes coding for plant cell wall-degrading enzymes (Kohler et al., 2015) and limited capacity for decomposition, as compared with free-lining saprotrophs. It has been proposed that saprotrophic and mycorrhizal fungi have overlapping fundamental niches, but that antagonistic mycorrhizal fungi exclude more efficient saprotrophic decomposers from deeper organic layers (Bödeker et al., 2016), and that competition between these fungal guilds might reduce decomposition rates (Gadgil and Gadgil, 1975; Averill and Hawkes, 2016; Fernandez and Kennedy, 2016).

Although mycorrhizal fungi in general are thought to be less efficient decomposers than saprotrophs, recent studies suggest that ectomycorrhizal fungi may still play an important role in organic matter transformation. Under strong nutrient limitation, with low mineralization and nutrients largely bound in complex organic forms, certain mycorrhizal fungi may use energy from host-derived sugars to power nutrient mobilisation from complex organic matter (Read and Perez-Moreno, 2003; Lindahl and Tunlid, 2015). Cortinarius species have been highlighted as particularly efficient in acquiring nutrients from organic pools, producing Mn-peroxidases, which are basidiomycetes-specific enzymes that facilitate oxidative decomposition of recalcitrant humus (Bödeker et al., 2014).

Clear-cutting has negative short-term effects on ectomycorrhizal fungal biomass and diversity (Jones et al., 2003; Grebenc et al., 2009) and leads to changes in fungal species composition (Hartman et al., 2012). Elimination of ectomycorrhizal fungi may open niches for saprotrophic fungi, proliferating in response to relaxed competition and increased resource availability (Lindahl et al., 2010; Averill and Hawkes, 2016; Bödeker et al., 2016). Furthermore, clear-cutting results in a flush of litter and dead fine roots and create favourable conditions for grasses and herbs, which together with increased variation in soil temperature, moisture and soil evaporation facilitate high rates of organic matter decomposition (Chatterjee et al., 2008). Such enrichment disturbance in combination with decreased competition for space and resources after harvesting—the ‘Gadgil effect’ (Gadgil and Gadgil, 1975; Fernandez and Kennedy, 2016)—may favour opportunistic saprotrophic species associated with ample production of hydrolytic enzymes involved in holocellulose decomposition (Moorhead and Sinsabaugh, 2006). In the absence of mycorrhizal fungi, saprotrophs may proliferate downward into relatively nitrogen (N) rich, well decomposed organic matter, leading to increased N mineralization and contributing to the generally observed flush of inorganic N (Kreutzweiser et al., 2008) and C loss (Magnani et al., 2007) after clear-cutting.

On seedlings establishing after the clear-cut phase, a new generation of ectomycorrhizal fungi establishes, initially as a low-diversity community of nitrophilic, disturbance-adapted ‘nursery species’ (Dahlberg and Stenström, 1991; Wallander et al., 2010). Within the ectomycorrhizal fungal community, species within the Atheliaceae family have been found to dominate in younger stands (Twieg et al., 2007; Wallander et al., 2010; Sun et al., 2015), possibly adapted to the relatively high N availability (Taylor et al., 2000). As stands develop, re-establishment of the mycorrhizal fungal community may restore mycorrhizal competition with saprotrophs, reducing organic matter turnover. Nitrogen immobilisation in proliferating mycorrhizal mycelium may further intensify N limitation (Wallander et al., 2010; Näsholm et al., 2013) with a potential feedback on fungal community composition (Sterkenburg et al., 2015). Specifically, species that are efficient in organic nutrient acquisition, for example Cortinarius, may obtain a competitive advantage over less efficient Atheliaceae species, as N availability declines (Hobbie and Agerer, 2010; Deslippe et al., 2011; Bödeker et al., 2014).

In this study, we address the theoretical framework outlined above of how soil fungi and their enzyme production may respond to stand development in managed forests, and discuss the results in the context of C sequestration and nutrient cycling.

We hypothesize that:

  1. i)

    Ectomycorrhizal fungi will increase in abundance with increasing time since clear-cutting, whereas saprotrophs will decline and become increasingly restricted to the surface litter. Activities of cellulose degrading enzymes will correlate with saprotrophic fungal abundance.

  2. ii)

    Atheliaceae species will gradually be replaced by a more diverse ectomycorrhizal fungal community, increasingly dominated by Cortinarius species, in correlation with activities of enzymes involved in mobilization of organic nutrients.

To test these hypotheses, we selected 10 even-aged Pinus sylvestris stands with tree ages ranging from 1 to 158 years. We assessed community composition in different organic layers by high throughput sequencing of fungal ITS2 amplicons (Ihrmark et al., 2012; Lindahl et al., 2013) and investigated correlations between community shifts along the age gradient and activity patterns of extracellular enzymes (Saiya-Cork et al., 2002).

Materials and methods

Study sites

In early October 2012, samples were collected from 10 forest stands situated in Uppsala County, Sweden. The forest stands were all subjected to forestry, unfertilized and differed in stand age from recent clear-cuts to mature forests (Supplementary Table S1). The younger stands (<50 years old) had been re-planted within 1–2 years after clear-cutting, while even aged structure of older stands (>80 years old) was obtained by selective thinning (the history of the 59-years-old stand is uncertain). The stands were dominated by Pinus sylvestris L. mixed with Picea abies L. with an understory of ericaceous dwarf shrubs (Vaccinium vitis-idaea L., Vaccinium uliginosum L. and Calluna vulgaris (L.) Hull), mosses (mainly Pleurozium schreberi (B.) Mitt.), and on young stand also grasses (primarily Deschampsia flexuosa L. and Festuca ovina L.). The stands were selected to be distant enough (at least 0.2 km apart) to be treated as independent replicates, but close enough (within 28 km distance) to avoid major differences in climate and geology.

Soil sampling and analyses

At each stand, nine soil cores (3 cm in diameter) were collected at random locations at least 5 m apart (Lilleskov et al., 2004), to avoid resampling of single individuals. Upon collection, all green plant parts were removed, and each core was split into three layers: litter (Oi; slightly decomposed plant material), fragmented litter (Oe; moderately decomposed plant material) and humus (Oa; highly decomposed plant material). Materials from respective layers were pooled within each stand, resulting in three pooled samples per stand. In total, 30 composite samples were analysed. After storage at −20 °C, samples were finely ground in liquid N with mortar and pestle. Roots with a diameter of more than 2 mm were removed, whereas fine roots were retained. Subsamples were kept at −20 °C for enzyme analyses and the rest of the material was freeze-dried before DNA and ergosterol extractions. Extractable pools of inorganic N, soil pH and organic content were determined on humus samples (a detailed description is provided in Supplementary Information, Section 1.1). All samples were subjected to ergosterol extraction (Nylund and Wallander, 1992; Supplementary Information, Section 1.2) and assayed for activities of selected hydrolytic enzymes: cellobiohydrolase (CBH), β-1.4-glucosidase (BG), β-1.4-xylosidase (BXD), leucine aminopeptidase (LAP), chitobiosidase (CHiB), β-1.4-N-acetylglucosaminidase (NAG), acid phosphatase (aP) and oxidative enzymes: Mn-peroxidase and Laccase (Saiya-Cork et al., 2002; Supplementary Table S2; Supplementary Information, Section 1.3).

Fungal community analysis

DNA was extracted from 50 mg of litter or 250–500 mg of fragmented litter or humus following the extraction protocol by Clemmensen et al. (2016) (Supplementary Information, Section 1.4). ITS2 amplicons were produced using the forward primer gITS7 (Ihrmark et al., 2012) and the two mixed reverse primers ITS4 (75%; White et al., 1990) and ITS4arch (25%; 5′-CACACGCTGTCCTCGCCTTATTGATATGC-3′) elongated with unique identification tags (Clemmensen et al., 2016). Products were mixed in equal concentrations and cleaned with the E.Z.N.A. Cycle Pure Kit (Omega Bio-Tek Inc., Nocross, GA, USA). Adaptor ligation and Pacific Biosciences PSII sequencing were performed by SciLifeLab (NGI, Uppsala, Sweden) using eight SMRT Cells.

Sequence analysis

Sequences were analysed using the bioinformatics pipeline SCATA (https://scata.mykopat.slu.se). After removal of sequences with mean quality <20 or containing bases with quality <3, sequences (complementary reversed, if needed) were searched for primers (requiring 90% identity) and identification tags. Only sequences containing matching tags at both ends were retained. Sequences passing quality control were randomly down-sampled to equal sequencing depth (1238 sequences per sample). USEARCH (Edgar, 2010), as implemented in SCATA, was used to cluster sequences into species hypotheses (SHs) (Kõljalg et al., 2013) based on single-linkage clustering with a 1.5% threshold distance for sequences to enter a SH and equal penalties for mismatch and gap extension. The UNITE reference sequence database (http://unite.ut.ee) was included in the clustering, in order to obtain identification of SHs. Sequence data are stored at NCBI (www.ncbi.nlm.nih.gov/sra) under the accession number SRP093592 (Supplementary Table S1). SHs were subjected to taxonomic and functional identification in order of decreasing global relative abundance, until at least 85% of the sequences in each sample were covered. Representative sequences from SHs were aligned and further analysed by neighbour joining with pairwise gap deletion in MEGA 6 (Tamura et al., 2013). Relative abundances were recalculated after removal of non-fungal sequences. SHs were further identified using the BLASTn algorithm against the UNITE database with neighbour joining trees as guidance and verification. Local SHs were assigned to UNITE SHs (if at least 98% similar), or identified to genera or orders based on neighbour joining trees (if justified by boot strap support). Further, SHs were assigned to ecological functions based on taxonomic information or sequence similarity to available sequences (NCBI) obtained from well-defined substrates, such as cleaned roots or plant leaves (Clemmensen et al., 2013). The following functional groups were included: ectomycorrhizal, ericoid mycorrhizal, root ascomycetes (excluding known mycorrhizal taxa), litter ascomycetes, litter basidiomycetes, moulds (Morteriellales, Eurotiales, Capnodiales, Mucorales, Hypocreales), yeasts (Saccharomycetales, Cystofilobasidiales, Leucosporidiales, Sporidiobolales, Tremellales) and unknown (Supplementary Tables S3 and S4; Supplementary Figure S1).

Statistical analysis

The total fungal community data set was analysed using ordination methods in CANOCO 5 (Microcomputer Power, Ithaca, NY, USA). For graphical representation of variation in fungal community composition between forest stands, detrended correspondence analysis was implemented. Correlation between forest age and fungal community composition was evaluated for significance using canonical correspondence analysis (CCA) with Monte Carlo permutations (without permutations within plots to account for dependency between layers from the same plot). CCAs were conducted at the levels of SHs or genera/orders (genera and orders were combined in a single analysis due to taxonomic ambiguities at the finer phylogenetic levels of the same clades). In order to eliminate the expected effect of recent clear-cutting and focus on community changes in established stands, all CCAs were repeated with the two youngest stands (1 and 5 years old) excluded from the data set. Separate post-hoc CCAs were conducted for each organic layer as well as for each functional group.

Effects of forest age, organic layer and their interaction on relative abundances of functional groups, total amount of fungal biomass (μg ergosterol per m2) and biomass concentration (μg ergosterol per gram organic matter) were evaluated using mixed effect models within the nlme package in R version 3.1.3 (R Core Team, 2015). Age and layer were specified as fixed effects, whereas forest stand was included as a random factor. Tests were repeated with the two youngest stands excluded.

To analyse the effect of forest age on enzyme activities, a CCA with Monte Carlo permutations (without permutations within plots) was performed. In order to test correlations between variation in enzyme activities and variation in fungal community composition (genera/orders), simple Mantel tests were performed separately for each layer, using the R package ecodist (Goslee and Urban, 2007) and the Bray–Curtis index as dissimilarity metric. For each fungal SH, a set of specific enzyme activities ES,X were calculated as abundance weighted averages across samples:

where EXi is the activity of enzyme X in sample i, Pi is the relative amplicon abundance of SHs in sample i and n is the total number of samples (Bödeker et al., 2014). Similarity in enzyme profiles between fungal genera/orders was visualised by principal component analysis of these specific enzyme activities (Equation 1). Only genera accounting for more than 1% of total amplicons in two or more samples were included. Post-hoc correlations between specific combinations of genera/orders (amplicon abundances) and enzyme activities were established using the R package Hmisc. To be comparable with fungal community data, which were expressed as relative abundances, enzyme data were expressed in relation to ergosterol concentrations and thereafter log transformed. Species data were arcsine transformed before all analyses (for calculation of specific enzyme activities, non-transformed relative abundance data were used).

Results

Soil characteristics

Soil pH ranged from 4.0 to 4.8 and decreased with increasing forest age (r2=0.54, P=0.01; Supplementary Table S1). Ammonium levels were highest in the two youngest stands and declined to low levels in stands above 12 years (Supplementary Table S1).

Sequencing output

In total, 101 971 PacBio sequences passed quality control. After random subsampling and removal of plant sequences (10%), the remaining 34 048 sequences clustered into 1305 SHs, with each sample represented by on average 1135 sequences (range 868–1221). Out of this, 92% of the sequences (458 SHs) were subjected to phylogenetic and functional identification.

Community composition

Fungal community composition was significantly affected by forest age both at the SHs and genera/orders levels, with higher explanatory power at genera/orders level (Figure 1; Table 1). Significance remained after removal of the two youngest stands. Different layers were colonised by distinct fungal communities (Figure 1), and post-hoc CCAs on separate layers indicated that the relationship between community composition and forest age was strongest in the deepest humus layer and marginally insignificant in the litter layer (Table 1).

Figure 1
figure 1

Variation in fungal community composition in different organic layers of 10 Pinus sylvestris forest stands of different ages (1–158 years), as visualised by a sample plot of a detrended correspondence analysis based on PacBio sequencing of amplified ITS2 markers. Numbers correspond to forest age; circles are shaded according to layers. Axes 1 and 2 explained 8 and 5.9%, respectively, of the total inertia of 6.9.

Table 1 Effect of forest age on fungal community composition in organic layers of 10 Pinus sylvestris forest stands

Functional groups segregated significantly between layers, confirming a preference of fungi a priori classified as litter-associated to surface litter, whereas root-associated taxa were predominantly found in fragmented litter and humus (Figures 2 and 3; Supplementary Figure S2; Table 2). Litter basidiomycetes declined in relative abundance with increasing forest age and litter ascomycetes were more strongly confined to the litter layer in older forests. Moulds and yeasts also declined in the humus layer with increasing forest age, but their overall relative abundance was low. Balancing the general decline in saprotrophs, ectomycorrhizal fungi increased in relative abundance with increasing stand age. The difference in functional guild dominance was most obvious between the two youngest stands (1 and 5 years) and the rest (12–158 years). However, the decline in litter basidiomycetes with increasing forest age remained significant after removal of the two youngest stands (Figure 2; Table 2).

Figure 2
figure 2

Distribution of fungal functional groups in different organic layers: (a) litter (b) fragmented litter (c) humus, of 10 Pinus sylvestris forest stands of different ages (1–158 years), as estimated by PacBio sequencing of amplified ITS2 markers. Abundances are given as percent of the identified amplicon sequences (accounting for 92% of total sequences).

Figure 3
figure 3

Distribution of fungal genera/orders in organic layers of 10 Pinus sylvestris forest stands of different ages (1–158 years), as visualised by a species plot of a canonical correspondence analysis (CCA) based on PacBio sequencing of amplified ITS2 markers. The CCA included 66 fungal genera/orders, but only the 30 most abundant genera/orders are shown. Circles are colour-coded according to functional groups with area indicating relative abundance. Triangles and the red vector represent constraining environmental variables, while grey vectors represent additional supplementary variables plotted in ordination space. Axes 1 and 2 explained 19.1 and 8.9%, respectively, of the total inertia of 1.3.

Table 2 Results of linear mixed models

Within litter ascomycetes, the most dominant orders were Venturiales and Rhytismatales (Lophodermium species), and litter basidiomycetes were strongly dominated by the genus Mycena. The effect of forest age on community composition within litter-associated fungi was marginally insignificant (P=0.06; Supplementary Table S5). Within the ectomycorrhizal fungal community, a significant effect of stand age was observed, both at the SHs and genera/orders levels and after removal of the two youngest stands (Supplementary Table S5). Piloderma and Tylospora (Atheliaceae) species dominated in younger stands (12–34 years) and Cortinarius and Russula species increased in relative abundance and became dominant in the oldest stands (over 100 years) (Figures 3 and 4; Supplementary Figure S1). The root ascomycete community was dominated by Archaeorhizomyces and Helotiales and changed significantly with increasing forest age. Mould composition was also significantly affected by forest age, but only at SHs level and when all stands were included (Figure 3; Supplementary Table S5).

Figure 4
figure 4

Distribution of the most dominant ectomycorrhizal fungal families in the humus layer of 10 Pinus sylvestris forest stands of different ages (1–158 years), as indicated by PacBio sequencing of amplified ITS2 markers.

Overall species richness and evenness declined significantly with increasing forest age in the humus layer, but only when tested across all forest stands (Figure 5a and c; Supplementary Table S6a). In contrast, ectomycorrhizal fungal richness and evenness in the humus and fragmented litter layers increased significantly with increasing forest age (Figure 5b; Supplementary Table S6b). For richness, the relationship was significant also with the two young stands excluded (Figure 5d).

Figure 5
figure 5

Richness and evenness of total (a, c) and ectomycorrhizal (b, d) fungal communities in different organic layers of 10 Pinus sylvestris forest stands of different ages (1–158 years), as analysed by PacBio sequencing of amplified ITS2 markers. The indices were calculated on all fungal species hypotheses (SHs) (n=1560) or on identified ectomycorrhial SHs (n=78).

Total amounts of fungal biomass (μg ergosterol per m2) increased significantly with increasing forest age in the humus layer (Table 2; Supplementary Figure S3), whereas there was no significant effect on biomass concentration (μg ergosterol per gram organic matter).

Enzymes

Enzyme activity profiles were not significantly correlated with forest age (P=0.5). However, enzyme profiles correlated with fungal community composition (at genera/orders level) in the litter and fragmented litter, as indicated by significant Mantel statistics (r=0.46, P=0.02 and r=0.46, P=0.03, respectively). In the humus layer, Mantel statistics indicated no significant correlation between enzymes and community matrices (r=−0.4, P=0.9) and data were not further explored.

In litter, Cantharellales correlated positively with hydrolytic enzymes acting on polysaccharides (CBH, BG and BXD) and N-containing substrates (LAP and CiBH). The saprotrophic genera Pleosporales, Hypocreales and Trechisporales also correlated positively with CBH and BXD, CBH and Laccase, respectively. Venturiales, on the other hand, correlated negatively with CBH, BG, BXD and LAP (Figure 6a; Supplementary Table S7a).

Figure 6
figure 6

Similarity in enzyme profiles among fungal genera/orders, based on principal component analysis of their specific enzyme activities, calculated as abundance weighted averages (according to formula 1). The figures are based on enzyme activity analyses and PacBio sequencing of amplified ITS2 markers from (a) litter and (b) fragmented litter of 10 Pinus sylvestris forest stands of different ages (1–158 years). Only genera accounting for more than 1% of total amplicons in two or more samples are shown. Genera/orders and enzymes with significant positive correlations (Supplementary Table S7) are highlighted by bold text and black vectors, respectively. aP, acid phosphatase; BG, β-1.4-glucosidase; BXD, β-1.4-xylosidase; CBH, β-D-cellobiohydrolase; CiBH, chitobiohydrolase; LAP, leucine aminopeptidase; MnP, Mn-peroxidase; NAG, β-1.4-N-acetylglucosaminidase.

In fragmented litter, enzymes acting on polysaccharides were positively correlated with saprotrophs: Mycena (CBH, BG, BXD), Helotiales (CBH, BXD) and Pleosporales (BXD). Nutrient-acquiring enzymes were positively correlated with ectomycorrhizal fungi: Cenococcum (aP), Piloderma (aP), Cortinarius (aP) and Russula (NAG), but negatively correlated with saprotrophs: Mycena (aP), Helotiales (aP), Pleosporales (aP, NAG) and Sordariales (NAG). Mn-peroxidase was positively correlated with Cortinarius but negatively with saprotrophic Rhytismatales (Figure 6b; Supplementary Table S7b).

Discussion

Ecosystem development may be described as driven by interactions between environmental conditions, composition of communities and their activities. These activities, in turn, have a major effect on the environment through feedback loops that may either stabilize ecosystems or lead to directional development. Fungal species are central mediators of these interactions (Lindahl and Clemmensen, 2016). In species, response traits (that determine population responses to environmental changes) are connected to effect traits (that determine how communities influence their environment; Koide et al., 2014). By combining molecular methods with enzyme analyses, we here demonstrate significant correlation between environmental drivers, fungal species and their enzyme expression leading towards a better mechanistic understanding of ecosystem responses to forestry-related disturbances. Clear-cutting radically altered fungal communities in pine forest soils, after which composition of species and functional groups changed progressively with increasing stand age (Figure 2). Data correlations supported that these compositional shifts were associated with changed effect traits, in terms of divergent enzyme activity profiles with presumed relevance for organic matter turnover (Figure 6).

In line with our first hypothesis (i), relative abundance and diversity of ectomycorrhizal fungi were negatively affected by clear-cutting but thereafter increased with increasing forest age (Figure 3; Table 2). In contrast, saprotrophs displayed high relative abundance right after harvest but subsequently declined, particularly in the deeper organic layers (Figure 2; Table 2). Potentially, the observed decline of saprotrophic fungi could be relative rather than absolute (as a result of increasing abundance of ectomycorrhizal fungi). However, more or less stable ergosterol concentrations across the age gradient (Table 2) support that shifts in community composition can be interpreted in absolute terms. Moreover, the decline in overall fungal diversity in the humus layer with increasing forest age (Figure 5) further supports the picture that ectomycorrhizal fungi proliferate during stand development on the expense of other functional groups.

Elimination of ectomycorrhizal fungi by clear-cutting opened a new niche for saprotrophic fungi (Figure 2), in line with the ‘Gadgil effect’ (Gadgil and Gadgil, 1975; Fernandez and Kennedy, 2016), which involves stimulated litter decomposition after severing of root connections and loss of ectomycorrhizal fungi. Presumed saprotrophs within Pleosporales and Hypocreales in the upper litter layer and Mycena in more decomposed litter correlated positively with enzyme activities involved in holocellulose and protein degradation (Figure 6a). These groups occurred more abundantly in younger stands (Figure 3), indirectly supporting that clear-cutting stimulated saprotrophic assimilation of carbohydrates and N from organic matter. Thus, ectomycorrhizal fungi and saprotrophs seem to have overlapping fundamental niches, but ectomycorrhizal fungi may restrict the realized niche of saprotrophs, repressing their decomposer activity under undisturbed conditions (Averill and Hawkes, 2016; Bödeker et al., 2016). Activated organic matter mineralization by saprotrophic fungi due to loss of mycorrhizal competition (that is, a ‘Gadgil effect’) in combination with improved plant litter quality might explain the high net C losses in forest ecosystems during the first decade after clear-cutting (Peltoniemi et al., 2004; Magnani et al., 2007).

After an almost complete loss of ectomycorrhizal fungal species as a result of clear-cutting, richness in the humus layer increased gradually for at least 60 years of stand development (Figure 5). These results are consistent with previous research (Wallander et al., 2010) and emphasize that ectomycorrhizal fungi are particularly sensitive to forestry. During the last 60 years, 95% of the Swedish forest area has been subjected to forestry practices involving clear-cutting, and it is difficult to predict long-term effects on populations of ectomycorrhizal fungi in relation to preindustrial forest fire dynamics (Dahlberg, 2002). For example, species within the genus Cortinarius, frequently recorded as fruit-bodies in old-growth forests (Dahlberg, 2001), usually form long-lived and large individuals, which could make them particularly vulnerable to clear-cutting practices.

In agreement with our second hypothesis (ii), ectomycorrhizal fungal species composition (based on abundance of ITS markers) changed gradually throughout stand development from dominance by species within the Atheliaceae family (Piloderma and Tylospora species) at middle developmental stages to dominance by Cortinarius and Russula species in the oldest stands (Figures 3 and 4). Although not statistically testable, the fungal community in the 59-years-old stand still showed some similarity to younger stands (Figure 1). Swedish production forests are commonly subjected to repeated thinning during stand development, which could favour disturbance-adapted fungal communities also in older stands. In contrast, the oldest stands (>80 years) in this study were never clear-cut, and the longer-term effects of clear-cutting on fungal communities are not yet possible to evaluate.

Atheliaceae species arrived early (about 10 years after re-planting) and dominated the ectomycorrhizal fungal community in younger stands. The highest relative abundance of Atheliaceae was found in the 34-years-old stand, corresponding to timing of canopy closure (Croft et al., 2014), the highest standing biomass of fine roots (Konôpka et al., 2015) and maximum ecosystem C sink (Magnani et al., 2007). Athelioid species remained a significant component of the ectomycorrhizal fungal community also in older stands, but at lower relative abundance. Accelerated decomposition after clear-cutting in combination with increased access of saprotrophs to low C:N ratio substrates in deeper layers may have contributed to increased N mineralisation and elevated soil pH after clear-cutting (Smolander et al., 1998) compared with old forests (Supplementary Table S1). Furthermore, in the same age gradient, Hagenbo et al. (2016) observed higher rates of mycorrhizal mycelial production in the younger stands. Presumably, the combination of high resource availability, both in terms of host-derived C and inorganic N, together with relaxed acidity stress favoured establishment of rapidly growing Atheliaceae species. Tylospora (Atheliaceae) species have previously been highlighted as favoured by high soil fertility (Parrent et al., 2006; Sterkenburg et al., 2015) and atmospheric N deposition (Taylor et al., 2000). Ammonium pools decreased during the first 12 years after stand re-establishment (Supplementary Table S1), but gross N mineralization could have remained high for longer time. Soil pH was elevated for about 50 years (Supplementary Table S1), and potentially contributed to the long-term effect of clear-cutting on fungal community composition and ectomycorrhizal fungal diversity. Further, priority effects (Kennedy et al., 2009) may have favoured persistence of successful early colonizers and prevented establishment of other species. Dispersal limitation has been shown to play an important role in short-term assembly of ectomycorrhizal fungal communities (Peay and Bruns, 2014), but it seems unlikely that low dispersal could have delayed re-establishment of certain ectomycorrhizal genera for decades within a patchy forest landscape.

High production of ectomycorrhizal mycelium in regenerating pine forests (Hagenbo et al., 2016) has been suggested to be associated with rapid N immobilization in fungal mycelium (Wallander et al., 2010), resulting in overall intensified N limitation of the ecosystem (Näsholm et al., 2013). Progressively increased competition for N combined with restored competitive suppression of saprotrophs by mycorrhizal fungi (Gadgil and Gadgil, 1975) may have hampered decomposition (Averill and Hawkes, 2016) and favoured accumulation of organic matter in the soil (Averill et al., 2014). In the long run, retention of nutrients in accumulating organic soil stocks leads to decreasing ecosystem productivity in the absence of disturbance (Magnani et al., 2007; Clemmensen et al., 2013), but mycorrhizal symbioses may enable plants to ameliorate N limitation, providing access to organic nutrient pools (Read, 1991).

Ectomycorrhizal taxa, such as species within the genera Cortinarius, Cenococcum and Russula, were positively correlated with activities of enzymes mobilising N and/or phosphorus from organic matter (Figure 6b). The correlation of the genus Cortinarius with lignin-degrading Mn-peroxidase in the fragmented litter confirms previous observations (Bödeker et al., 2014) and highlights these fungi as particularly important decomposers, facilitating mobilization of nutrients from complex organic matter (Lindahl and Tunlid, 2015). We postulate that successional establishment of ectomycorrhizal fungi with capacity to access more complex nutrient pools sustains plant productivity and delays ecosystem retrogression (Clemmensen et al., 2015; Baskaran et al., 2016).

Direct effects of forest age on enzyme activities did not receive statistical support. Further, a prokaryotic contribution to enzyme activities cannot be dismissed. Nevertheless, the significant correlations between stand age and fungal community composition, on the one hand, and fungal community composition and enzyme activities, on the other hand, suggest that enzyme activities shift during stand development in concert with fungal succession. Supposedly, our sampling was insufficient to capture this indirect effect of stand age on enzyme activity estimates, which are notoriously noisy due to spatial patchiness (Baldrian, 2014). The more direct causal links between environmental drivers and fungal communities, and between fungi and enzymes, may have provided more certain statistical support for these relationships in our study.

Conclusion

Here, we present relationships between fungal community dynamics and associated enzyme activities along an age gradient of managed pine forest stands. We propose that compositional shifts in fungal species and enzymes play an important role in regulating plant productivity and C storage during forestry rotation cycles. Clearly, clear-cutting has a major and long-lasting influence on fungal diversity and its role in ecosystem processes. Potentially, alternative management regimes, such as continuous cover forestry, could limit harvest-associated losses of nutrient and organic matter stores by maintaining the competitive balance between saprotrophic and mycorrhizal fungi (that is, limit the ‘Gadgil effect’ associated with clear-cutting). Further, maintenance of functional diversity in the ectomycorrhizal fungal community (particularly preserving Cortinarius and Russula species) may sustain long-term forest production by retaining a capacity for symbiosis-driven recycling of organic nutrient pools throughout the forest rotation.