Bark coverage shifts assembly processes of microbial decomposer communities in dead wood

Abstract

Bark protects living trees against environmental influences but may promote wood decomposition by fungi and bacteria after tree death. However, the mechanisms by which bark determines the assembly process and biodiversity of decomposers remain unknown. Therefore, we partially or completely removed bark from experimentally felled trees and tested with null modelling whether assembly processes were determined by bark coverage and if biodiversity of molecularly sampled fungi and bacteria generally benefited from increasing bark cover. The community composition of fungi, wood-decaying fungi (subset of all fungi) and bacteria clearly separated between completely debarked, partly debarked and control trees. Bacterial species richness was higher on control trees than on either partly or completely debarked trees, whereas the species richness of all fungi did not differ. However, the species richness of wood-decaying fungi was higher on partially and completely debarked trees than on control trees. Deterministic assembly processes were most important in completely debarked trees, a pattern consistent for fungi and bacteria. Our findings suggest that human disturbances in forests shift the dominant assembly mechanism from stochastic to deterministic processes and thus alter the diversity of wood-inhabiting microorganisms.

1. Introduction

Tree characteristics, such as wood density and chemistry, shape the microbial communities that inhabit dead or dying trees [1–3]. In turn, these characteristics determine the decomposition rates of wood [4]. Bark is the outermost part of woody plants and consists of the periderm, cortex and phloem, each of which has specific physiological functions, morphological properties and chemical properties [5]. Bark components transport and store resources and plant products, protect living trees against physical (e.g. fire) and biological (e.g. bark beetles and microbial pathogens) influences, and form a buffer against climatic oscillations of temperature and moisture [5–8]. By mediating the moisture content of wood, the bark of most tree species also promotes dead wood decomposition [9–11]. Natural forces, such as wind breakage, desiccation cracks and browsing, change the bark cover of dead or dying trees, as may anthropogenic forces such as logging activities, which in the last 2 decades have damaged the bark of upwards of 28% of the trees in some commercial forests [12]. Furthermore, in Norway spruce (Picea abies (L.) H. Karst.), the partial or complete removal of bark is an active management strategy aimed at decreasing populations of pest bark beetle species [13].

The microclimate of dead wood depends on the cover of the tree layer; thus, with increasing canopy openness, the moisture content decreases and the temperature increases [14,15]. Microclimate as mediated by canopy openness is a major driver of the diversity and community composition of different saproxylic species groups [15,16]. However, empirical evidence for both the independent and interactive effects of bark coverage and canopy openness on the community composition of microbial decomposers is lacking. For example, under the harsh physical condition of sun-exposed sites, the buffering function provided by the physical properties of the bark layer may determine the microbial assemblage.

Prokaryotic and eukaryotic microorganisms are ubiquitous, occur in high abundances and play key roles in many biological processes [17]. They include saprotrophic microorganisms, which are important drivers of wood decomposition [18]. Bark, and particularly its phloem layer, is a structurally complex microhabitat characterized by proportions of sugars and nutrients that are higher than in heartwood and sapwood, as well as by distinct chemical properties [10,11]. Consequently, the bark layer of dead wood hosts a diverse fungal community [19]. Thus, dead or dying trees covered by bark should have a high heterogeneity of microhabitats that results in a high fungal species richness [16]. Indeed, based on fruit-body inventories, Thorn et al. [20] showed that the species richness of wood-inhabiting fungi is lower on fallen trees without than with bark. However, fruit-body inventories provide only a glimpse of the fungal community, as only a minority of fungal species in dead wood form fruit bodies [21]. Nevertheless, fruit-body and molecular inventories often yield comparable fungal diversity patterns, and the most abundant species identified by molecular surveys commonly correspond to those identified by fruit-body inventories [21,22]. Both sets of observations suggest that bark coverage promotes the species richness of wood-inhabiting fungi. In addition, based on the nutritional richness of bark, Deschamps [23] proposed that a high diversity of bacteria is responsible for primary degradation in the bark layer, but empirical evidence is lacking. However, the protective properties of bark for living trees may also hamper the access of microorganisms to the tree after its death, such that species richness will increase only when the bark layer is removed [6,24].

Processes that assemble species into communities are simultaneously influenced by deterministic and stochastic processes [25,26]. Deterministic processes include the selection of species by the abiotic environment (i.e. environmental filtering) and the biotic interactions within a community, and are thus predictable. By contrast, stochastic processes, such as the ecological drift imposed by factors that include dispersal and random birth–death dynamics, are unpredictable [27]. A main challenge in ecology is to understand the habitat properties that determine the relative influences of deterministic and stochastic processes [28]. Recent methodological advances in the field of microbial ecology have enabled deeper investigations of the drivers of deterministic versus stochastic processes and the change in their relative importance [25,29,30]. Yet, our knowledge on how bark affects the assembly processes and biodiversity of different wood-inhabiting microorganisms during decomposition is thus far limited.

We investigated the function of tree bark, as a characteristic of woody plants, in driving assembly processes and diversity patterns of wood-inhabiting fungi and bacteria. Specifically, we removed the bark of experimentally felled Norway spruce trees either partially (approx. 20%) or completely, or left the bark intact (control), and 1.5 years later sampled the wood-inhabiting microbial communities by high-throughput sequencing. We expected that: (i) tree bark coverage would have an effect on assembly processes and, together with microclimate (as mediated by canopy openness), alter the community composition of wood-inhabiting microbial communities; (ii) the moisture-retaining properties of tree bark and the increase in microhabitat diversity would demonstrate the general positive effect of bark coverage on bacterial and fungal diversity; and (iii) for wood-inhabiting microorganisms, deterministic processes would become more relevant under the extreme condition represented by completely debarked trees.

2. Material and methods

(a). Study area and experimental design

This study was conducted in the Bavarian Forest National Park in southeastern Germany. To focus on the natural microbial community involved in decomposition, we selected the naturally dominant tree species in this area, Norway spruce (Picea abies (L.) H. Karst), whose decaying bark layer hosts a rich microbial fauna [19].

We established 12 study sites, selecting at each site three vital mature spruce trees without any bark injury. All trees had similar physical attributes, a mean height of 21.6 ± 4.5 m and a mean diameter at breast height (1.3 m) of 0.40 ± 0.08 m. To test for possible effects of canopy openness (sun-exposed or shaded), six of the sites were established in the shade and six in sunny salvage-logged areas or small forest gaps. Each of the six pairs of sunny/shaded sites was separated by at least 200 m.

The selected 36 spruce trees were pulled down with steel cables and winches in April 2013. At each site, one tree (control) was left with its bark completely intact, another was completely debarked, by fully removing the bark and phloem (figure 1) using a debarking device mounted on conventional chainsaws, and another was partially debarked using a bark-scratching device, mounted on a conventional chainsaw, that disrupted the bark and phloem approximately every 3 cm and thus removed approximately 20% of the bark (figure 1). This set-up was part of an experimental approach to investigate the effects of mechanical bark removal on biodiversity (for details of the experimental design, see [20]).

Figure 1.

Photographs of the bark coverage of the experimentally felled Norway spruce (Picea abies (L.) H. Karst) in the Bavarian Forest National Park, Germany. Bark was left intact (control), or partly or completely mechanically removed. (Online version in colour.)

(b). Sampling protocol

The microbial biodiversity in a particular environment is frequently analysed using molecular-based approaches (see below) that identify molecular operational taxonomic units (OTUs) [2,31,32]. The felled trees were sampled in October 2014, 1.5 years after their death, by full vertical cross-section drilling using a 0.8 cm × 30 cm long auger bit and including the bark layers of partly debarked and control trees [2,33]. On each felled tree, five cores regularly distributed over the length of the tree were drilled, pooling the sawdust of each tree in one sample. If the diameter of the tree was greater than 30 cm, opposite sides of the felled tree were sampled, so that each total drilling core had the same length as the tree diameter. To minimize the effect of microorganisms occurring on the surface of the tree, we mechanically cleaned the point of drilling with ethanol. To avoid cross-contamination between different trees, the auger bit was flamed after each mechanical cleaning. Sawdust samples were stored in clean plastic bags and directly transferred to a freezer at −40°C.

(c). Laboratory protocol

Each wood sample was homogenized and ground to a fine powder using liquid nitrogen and a swing mill (Retsch, Haan, Germany). Total community DNA was isolated from 0.25 g of each homogenized wood sample using the ZR Soil Microbe DNA MiniPrep kit (Zymo Research, Irvine, CA, USA) according to the manufacturer's protocol. Fungal ITS2 regions were amplified using a mixture of primers P7-3N-fITS7 and P7-4N-fITS7 (forward) and P5-5N-ITS4 and P5-6N-ITS4 (reverse), which were modified from primers fITS7 and ITS4 as described in Ihrmark et al. [34]. Bacterial 16S rRNA genes were amplified using a mixture of the primer pair P5-8N-515F and P5-7N-515F (forward) and P7-2N-806r and P7-1N-806r (reverse), as modified by Caporaso et al. [35]. Primer sequences are given in electronic supplementary material, appendix S1.

The resulting PCR products (for the detailed PCR protocol, see electronic supplementary material, appendix S1) were sequenced on an Illumina MiSeq system at the Deep Sequencing Group of the Technische Universität Dresden. Briefly, the purified PCR products with universal 5′ tails were subjected to a second round of PCR, consisting of six to eight cycles and conducted using Phusion HF (NEB) and the indexing primers P5 and P7 (for the primer sequences, see electronic supplementary material, appendix S1). After indexing PCR, the final libraries were purified (1 × XP Beads, Agencourt), pooled in equimolar amounts and used for 2 × 300 bp paired-end sequencing on a MiSeq System.

(d). Bioinformatics

Raw data (FASTQ files) were processed using Geneious R9 software [36]. The 5′-ends of all sequences were trimmed, and the adapters removed. Forward and reverse reads were paired and then separated into fungal and bacterial sequences based on the specific primer sequence. Paired reads were quality trimmed using BBDUK (trim low quality, minimum quality = 13) and then merged using BBMerge (very high merge rate). Sequence fragments less than 200 bp and greater than 450 bp were discarded. The Mothur software (v. 1.36.1) [37] function chimera.uchime was used to filter out chimeras, resulting in the exclusion of 5530 fungal and 69 330 bacterial sequences. The remaining data were clustered into OTUs using CD-HIT-EST and based on a consensus of 97% [38]. Note that the number of sequences does not conclusively reflect the abundance of OTUs in the sample [39]. All processed and merged OTU sequences were submitted to the short-read archive (SRA, https://www.ncbi.nlm.nih.gov/sra/) under accession number SRP136590.

For the taxonomic annotation of the fungal and bacterial sequence data, representative sequences of the OTUs were blasted in SEED 2.1 [40] using Megablast against the NCBI nt database. Taxonomic classifications were checked using UNITE [41] and the NCBI GenBank. To build a subset of fungal OTUs known to be involved in wood decomposition, taxonomically identified fungal OTUs were classified based on the information from FUNGuild [42]. Thus, fungal species were designated as wood-decaying fungi if FUNGuild listed the species as a ‘brown rot,’ ‘white rot’ or ‘soft rot’ fungus (FUNGuild—trait) or if the species was explicitly listed as a ‘wood saprotroph’ (FUNGuild—guild) with a likelihood of ‘highly probable’ or ‘probable’ (FUNGuild—confidence).

(e). Statistics

All statistical analyses were performed in R [43]. Prior to the analyses, all OTUs represented by just one sequence read (singletons) were excluded from the dataset.

First, the effects of bark coverage and canopy openness on species richness (number of OTUs) and the potential effects of their interaction were tested. Quasi-Poisson linear mixed-effect models were applied for the three OTU datasets (i.e. those of fungi, the subset of wood-decaying fungi and bacteria), with bark coverage (control, partly debarked, completely debarked) and canopy openness (sun-exposed, shaded) as fixed terms and site as a random term. Simultaneous inference procedures with adjustments of p-values were used [44] for multiple testing by means to compare the three classes of bark coverage (glht function of R package multcomp [45]). Quasi-Poisson linear mixed-effect models were also applied for the species richness of each taxonomic subgroup of fungi (i.e. Ascomycota, Basidiomycota) and bacteria (i.e. Proteobacteria, Bacteroidetes, Actinobacteria, Firmicutes, Planctomycetes, Acidobacteria, Verrucomicrobia; see electronic supplementary material, appendix S2), with bark coverage and canopy openness as explanatory variables and site as a random term.

Second, the effect of bark coverage and canopy openness on the community composition of fungi and bacteria was tested using non-parametric multivariate analysis of variance [46] for the entire dataset and for the three subsets of pairwise comparisons (control–partly debarked, control–completely debarked, partly debarked–completely debarked) based on the Bray–Curtis dissimilarity distances and including site as a grouping factor, constraining the permutations within the group (adonis function of R package vegan [47]). The obtained p-values were adjusted [44]. Non-metric multidimensional scaling (NMDS) based on the Bray–Curtis dissimilarity distances for the OTU presence–absence data was used for visualization (metaMDS function of R package vegan [47]).

Third, the ecological processes that influence the assembly of fungi, the subset of wood-decaying fungi and bacteria within dead wood were explored by discriminating between stochastic processes (e.g. ecological drift) and deterministic processes (e.g. selection [48]). This was accomplished by calculating the phylogenetic turnover as the standardized effect size (SES) of the β mean nearest taxon distance (βMNTD). To retrieve the SES, βMNTD was first calculated for the original community matrix (observed βMNTD). Then tip labels of the phylogeny were randomized 1000 times and the βMNTD was repeatedly calculated (expected βMNTD). The SES βMNTD was then calculated by subtracting the expected from the observed βMNTD, divided by the standard deviation of the expected βMNTD. To calculate the βMNTD, phylogenies for fungi and bacteria were inferred (for details, see electronic supplementary material, appendix S3). Following Stegen et al. [25,29], |SES βMNTD| > 2 was interpreted as deterministic and |SES βMNTD| < 2 as stochastic community turnover. To retrieve an intuitive measure of the influence of deterministic processes among treatments, we calculated the proportion of pairwise |SES βMNTD| > 2 values divided by the total number of pairwise comparisons, which we call ‘proportion of deterministic processes'. Note that the absolute values of the assembly processes can be interpreted only within a sampling group/taxon (in this study: fungi, the subset of wood-decaying fungi, and bacteria) and not between taxa [49].

3. Results

(a). Sequence data analysis

Among the 2207 fungal OTUs based on 795 780 fungal sequences, 1460 OTUs belonged to Ascomycota (represented by 38% of all sequences) and 621 OTUs to Basidiomycota (represented by 62% of all sequences; figure 2a). From all 2207 fungal OTUs, 82% could be classified to the genus or species level (electronic supplementary material, appendix S4). A subset of 231 fungal OTUs (54% of all fungal sequences) was annotated as known wood-decaying fungi, of which Basidiomycota dominated (99% of these sequences; figure 2b).

Figure 2.

Diversity analyses of fungi, the subset of wood-decaying fungi and bacteria on felled trees with bark (control), and felled trees from which the bark was partly or completely removed. The proportions of (a) fungal phyla, (b) fungal phyla within the subset of wood-decaying fungi and (c) bacterial phyla are shown, together with the OTU numbers of (d) fungi, (e) wood-decaying fungi and (f) bacteria. (Online version in colour.)

The 5265 bacterial OTUs based on 1 090 251 bacterial sequences belonged to seven bacterial phyla: 1424 to Proteobacteria (represented by 59% of all sequences), 363 to Bacteroidetes (7%), 317 to Actinobacteria (14%), 244 to Firmicutes (2%), 217 to Planctomycetes (1%), 164 to Acidobacteria (11%) and 160 to Verrucomicrobia (2%; only phyla with greater than 1% of all sequences are listed; figure 2c). Of these bacterial OTUs, 61% could be classified to the genus or species level (see electronic supplementary material, appendix S4).

The frequency distribution of the sequence data showed a typical right-skewed distribution of rank abundance curves, with a few taxa dominating the community. Thus, 75% of all fungal sequences consisted of the 36 most abundant fungal OTUs and 75% of all bacterial sequences of the 168 most abundant bacterial OTUs (electronic supplementary material, appendix S5).

(b). Microbial species richness

An average of 325 fungal OTUs per felled tree was determined, but there was no difference in the number of fungal OTUs in felled trees differing in their bark coverage (figure 2d; electronic supplementary material, table S6.1). Diversity analyses for different q parameters also showed no differences for fungal diversity as a function of bark coverage (electronic supplementary material, appendix S7). Separate analyses of Ascomycota and Basidiomycota showed a similar species richness of each phylum on felled trees with different bark coverage (electronic supplementary material, appendix S2.1). For the subset of wood-decaying fungi, the number of OTUs ranged from 20 to 53 per felled tree and was lower on control trees than on either partly or completely debarked trees (figure 2e). The same pattern was found for the diversity of this group for the different q parameters (electronic supplementary material, appendix S7). Canopy openness had no effect on the species richness of fungi nor on the subset of wood-decaying fungi. For both, there was also no significant interaction between bark coverage and canopy openness (electronic supplementary material, table S6.1).

An average of 979 OTUs per felled tree could be assigned to bacteria. The number of bacterial OTUs per tree was higher on control trees than on partly or completely debarked trees (figure 2f; electronic supplementary material, table S6.1). The diversity of bacteria for the different q parameters was highest on control trees and higher for partly debarked than for completely debarked trees (electronic supplementary material, appendix S7). There was no evidence of a relationship between bacterial species richness and canopy openness, nor was there a significant interaction between bark coverage and canopy openness (electronic supplementary material, table S6.1). In separate analyses of the species richness of the different bacterial phyla, control trees had the highest number (Proteobacteria, Bacteroidetes, Actinobacteria, Planctomycetes and Verrucomicrobia) or the same number (Firmicutes and Acidobacteria) of OTUs as partly and completely debarked trees (electronic supplementary material, figure S2.2). The species richness of Proteobacteria and Verrucomicrobia was higher on control trees than on partly debarked trees, which in turn had a higher species richness for these groups than completely debarked trees (electronic supplementary material, figure S2.2). The number of colonized trees (OTU incidence) and the number of sequences per OTU were strongly correlated for both fungi and bacteria (electronic supplementary material, appendix S8).

(c). Microbial community structure

The composition of the fungal communities differed between trees with different bark coverage and between sun-exposed versus shaded sites. These differences explained 14% and 6% of the fungal community dissimilarity, respectively (ADONIS all adjusted p-values less than 0.05; electronic supplementary material, table S9.1). The community composition diverged from control trees to partly debarked trees to debarked trees along the second NMDS axis (figure 3a; electronic supplementary material, table S9.1). Similarly, bark coverage and canopy openness in the subset of wood-decaying fungi explained 15% and 4% of the community dissimilarity, respectively (figure 3b; electronic supplementary material, table S9.1). Rank abundance curves showed that the seven most abundant wood-decaying fungal species overall were also the most abundant species on control and partly debarked trees; however, among these species, Fomitopsis pinicola and Heterobasidion abietinum had a lower abundance on completely debarked trees (electronic supplementary material, figure S10.1).

Figure 3.

NMDS of the community composition of (a) fungi, (b) the subset of wood-decaying fungi and (c) bacteria on felled Norway spruce trees (12 per bark coverage level). Ellipses represent the interquartile range of the NMDS scores for trees of each bark coverage. Centroids of sun-exposed and shaded communities are indicated in bold in the ordination. (Online version in colour.)

Bacterial community composition differed between trees of different bark coverage and between sun-exposed versus shaded sites, which explained 15% and 5% of the bacterial community dissimilarity, respectively (ADONIS all adjusted p < 0.05; electronic supplementary material, table S9.1). The community composition diverged from control trees to partly debarked trees to debarked trees along the first NMDS axis (figure 3c; electronic supplementary material, table S9.1). Spatial analyses showed no significant effects of the distance between trees on the community patterns of fungi and bacteria (electronic supplementary material, appendix S11).

(d). Assembly processes

The proportion of deterministic processes was highest in completely debarked trees, which was a consistent pattern for fungi, the subset of wood-decaying fungi and bacteria (figure 4). For fungi, deterministic processes played a larger role in completely debarked trees than in control trees, whereas in partly debarked trees, they played no role at all (figure 4). For the subset of wood-decaying fungi, the proportion of deterministic processes increased from control to partly debarked to completely debarked trees (figure 4). For bacteria, the proportion of deterministic processes decreased from control to partly debarked trees; by contrast, in completely debarked trees, they played a larger role (figure 4). Within all three treatments and for all three taxa, stochastic processes predominated over deterministic processes (figure 4).

Figure 4.

Proportion of deterministic processes within felled trees with bark (control) and felled trees from which the bark was partly or completely removed for the community of (a) fungi, (b) the subset of wood-decaying fungi and (c) bacteria. (Online version in colour.)

4. Discussion

Our analyses showed that bark coverage determines the assembly processes and the diversity of microorganisms inhabiting dead wood. Deterministic assembly processes were more relevant on debarked than on either partly debarked or control trees, consistent with an important contribution of bark to microbial colonization. Microbial community composition differed between trees that differed in their bark coverage and between sun-exposed and shaded sites. In contrast with our expectation, a higher biodiversity on control trees was obtained only for bacteria, whereas fungal biodiversity did not differ with respect to bark coverage and the biodiversity of the subset of wood-decaying fungi was lower on barked trees than on partly or completely debarked trees. This change in species number due to bark coverage should translate to changes in the rates of wood decomposition and CO₂ emissions [50,51].

(a). Relevance of stochastic and deterministic assembly processes

Our results indicate the prevalence of stochastic processes relative to deterministic processes in the community assembly of microorganisms inhabiting dead wood. Recent studies of other habitats (i.e. deserts and rock pools) also reported that stochasticity is a major determinant of microbial assemblages [52,53]. Despite the ongoing and dichotomous debate about the dominance of stochastic or deterministic processes in structuring microbial assemblages [28], increasingly, the contributions of both have been acknowledged and the focus has shifted to the factors determining their relative importance [28,29]. Our null-model approach revealed the bark coverage of dead trees as both an important ecological control variable and a deciding factor in the relative importance of deterministic versus stochastic processes. The proportion of deterministic processes was highest in completely debarked trees for fungi, the subset of wood-decaying fungi and for bacteria. Thus, natural events and anthropogenic pressures which result in bark removal from dead trees [12,13] change the assembly of microbial communities to increase the importance of deterministic processes [54]. Therefore, microbial community composition is more predictable for completely debarked trees than for control trees. In the latter, stochasticity exerts a greater influence on species assembly. This finding is supported by other assembly studies showing that factors such as anthropogenic disturbances, which shift habitats away from their predominant status, increase the importance of deterministic processes in community assembly [54–56]. In other words, if a habitat is changed such that evolutionarily established niches and biotic interactions are changed as well, then deterministic processes will increase in relevance.

(b). Wood-inhabiting fungi

Community analyses of wood-inhabiting fungi and the subset of wood-decaying fungi showed that bark coverage explained community dissimilarity, with a gradual shift from control to partly debarked to completely debarked trees. The establishment of different saproxylic communities in relation to bark coverage has also been shown for epixylic bryophytes and saproxylic beetles [20,57,58]. Thus, our results are consistent with the idea that bark coverage alters the physical properties (e.g. moisture and temperature) of dead wood and therefore the competitive interactions of wood-inhabiting fungi. These interactions ultimately determine species persistence.

Species richness of all sampled fungi, and in particular for the subset of fungi known as wood-decaying fungi, our results contradicted our general expectation that bark coverage would promote microbial species richness. The species richness of fungi did not differ with bark coverage, but that of wood-decaying fungi was higher on partially and completely debarked trees than on control trees. Because our molecular samples included the bark layer, we expected that species numbers would be lower in partly and especially completely debarked trees than in control trees, as bark serves as an additional habitat for microorganisms. Potential explanations for our findings are that (i) the properties of bark directly or indirectly influence the endophytic fungal species pool [59], and (ii) the protective properties of bark hamper the access of microorganisms [6,24]. Consequently, the partial or complete removal of bark could promote colonization by other microbial species. Further, biotic interactions affect the establishment of species within a habitat and are crucial for assembly processes [60]. Thus, a change in environmental factors (e.g. bark coverage) could alter the biotic interactions within a habitat, and therefore species establishment [60]. In trees with intact bark, a few abundant and highly competitive wood-rotting species, such as the red-belted bracket (F. pinicola (Sw.) P. Karst), are typical early colonizers and dominate the community (as also shown by our data; see electronic supplementary material, figure S10.1) while possibly also inhibiting co-occurring species [61]. This view is supported by experimental manipulations of fungal communities on wood discs, in which competitive interaction were shown to increase with an increasing number of co-occurring species, leading to 30% less wood decomposition [51,62]. Another study showed that the dominance of F. pinicola in spruce reduced the species richness of other wood-inhabiting organisms [63,64]. The experimental finding of less wood decomposition for trees with a lower abundance of the dominant species is also in line with previous observations in our study area of a very slow decomposition of debarked spruce trees over decades [61]. The lower species richness of wood-decaying fungi on trees with bark together with the above-cited studies suggests that bark removal promotes the co-occurrence of wood-decaying fungal species by relaxing the competitive pressure exerted by strongly competitive wood-rotting species [62].

Our molecularly based results of the species richness of wood-inhabiting fungi contradict previous fruit-body inventories of the same felled trees, which showed a lower fungal species richness on debarked trees [20]. A positive effect of bark coverage on species richness as determined in fruit-body inventories was also reported in studies covering 13 tree species [4,65]. The differences in fungal species richness due to bark coverage as evaluated by molecular means (this study) versus fruit-body inventories [4,20,65] indicate that bark coverage shapes both processes: first the colonization of dead trees, which determines the community of wood-inhabiting fungi, and second the number of species able to produce fruit bodies for sexual reproduction. A key factor in the latter process is the physiological properties of the substrate, such as moisture content and temperature, whereas in sporophore primordium production, the surface of the substrate is critical [66]. Because both factors are influenced by bark [11], the link between fructification and bark coverage may determine the diversity of fruiting fungal species.

(c). Wood-inhabiting bacteria

Similar to the results obtained for fungi, community analyses of bacteria showed that bark coverage explained the dissimilarity of the bacterial community. The significant changes in the community of wood-inhabiting bacteria from control to partly debarked to completely debarked trees suggested that tree bark is a characteristic of dead wood that shapes bacterial community composition. Both Hoppe et al. [1] and Moll et al. [67] found significant variations in the bacterial communities of dead wood with different physico-chemical properties (e.g. pH, C and N concentration, relative wood moisture and density). Because tree bark mediates many of the physico-chemical properties of wood [8], our results are consistent with those earlier findings.

In line with our prediction, the species richness of bacteria was higher on control trees than on partly or completely debarked trees. Bark is a structurally complex microhabitat with a diverse set of niches available to bacteria. Among all wood components, bark, and in particular phloem, has the highest proportion of sugars and nutrients [10]. These features could facilitate bacterial richness in the bark layer and account for its distinct niches that favour bacterial colonization, but which are not found in heartwood. This point may explain higher bacterial species richness in control tress than in partly or completely debarked trees. Moreover, bacteria are important agents of tannin degradation [23,68]. As a component of tree bark, tannin could facilitate bacterial species richness in trees with bark. Although our study samples were taken from bark-bearing dead trees left intact during the sampling process, it would also be of interest to analyse the individual components of dead trees to determine their relative contributions to microbial diversity [67]. A further possible explanation for the higher bacterial species richness on control trees is the buffering function of bark [69], especially its ability to retain moisture [69], which would be a major reason for the higher species richness of bacteria on bark-covered dead trees [70].

(d). Microclimate mediated by canopy openness

Microclimate, as mediated by canopy openness (i.e. shaded versus sun-exposed sites), changed the community composition of fungi, wood-inhabiting fungi and bacteria. This result is in line with earlier studies of the community composition of saproxylic species (e.g. bacteria, beetles and fungi) [1,15,16], and can be attributed to the effects of canopy openness on the moisture content and temperature of a dead wood object [14,15]. However, we found that canopy openness only explained around 5% of the variance in the community composition, whereas bark coverage had a greater effect of around 15%. We therefore argue that bark coverage had more direct effects on the physical properties of dead wood than did canopy openness. For the species richness of fungi, wood-decaying fungi and bacteria, we found no direct effect of canopy openness nor an interactive effect with bark coverage. The results of fruit-body surveys of wood-inhabiting fungi are in line with those of our study, in that no differences in species number between shaded and sun-exposed dead trees were found [16]. Rather, host tree identity and decay stage were of greater importance [16].

5. Conclusion

Our study demonstrated that bark coverage is among the factors controlling the relative influence of stochastic versus deterministic processes in shaping the assemblages of wood-inhabiting fungi and bacteria. Factors that reduce bark coverage shifted assembly processes from a high relative importance of stochasticity to a higher relative importance of deterministic processes. This shift might have been a general reaction of species assemblies following a change in the relative dominance of different ecological variables (i.e. bark coverage, e.g. this study) [54–56]. Further, we showed that the effect of bark on biodiversity patterns differed between bacteria and fungi and between the processes of fungal colonization (this study) and fructification [20]. This complexity confounds a general positive effect of bark cover for biodiversity of wood-inhabiting microorganisms. A deeper understanding of the factors controlling the biodiversity of wood-inhabiting microorganisms provides insights into the linked process of nutrient cycling, which is relevant for carbon cycle–climate feedback and for conservation strategies [50].

Data accessibility

All OTU sequence data were submitted to the Sequence Read Archive (SRA, https://www.ncbi.nlm.nih.gov/sra/) under accession number SRP136590. A list of fungal and bacterial OTUs with their taxonomic classification is given in electronic supplementary material, appendix S4.

Authors' contributions

J.H., C.B., J.M. and S.T. conceived the study; B.H., H.K. and E.S. carried out the laboratory work and assembled the data; J.H. performed the data analysis; J.H. wrote the manuscript with support from C.B., A.G., B.H., H.K., F.-S.K., J.M., S.S., E.S. and S.T.; S.T. supervised the project.

Competing interests

We declare we have no competing interests.

Funding

S.T. was funded by the scholarship program of the German Environmental Foundation.