The deadwood samples from Norway spruce (P. abies (L.) Karst) were collected in a montane forest at an altitude of 1995 m above sea level (a.s.l.) in Val di Rabbi (Trentino, Italy). A detailed description of the study site is given in Petrillo et al. (2015). The samples were classified in situ into DCls I, III, and V based on their visual, geometric and tactile features (Hunter, 1990). Samples from DCl I correspond to hard wood penetrable with a knife for only a few mm. Wood of DCl III is distinctly softer and can be penetrated with a knife to around 1–4 cm. In DCl V, wood is very soft and has a crumbly, instable texture which is easily penetrable with a knife for >10 cm; the shape of the original log is barely recognizable. For each DCl, three samples (biological replicates) were collected from different logs, pre-treated by cut-milling and stored at −20 °C for further analysis according to Gómez-Brandón et al. (2017a).

The total DNA pool was extracted from 0.1 g cut-milled sample using MP Biomedicals DNA Extraction Kit according to the manufacturer’s protocol. To assure a complete disruption of the wood tissue, one 1/400 ceramic sphere (MPI biochemical cat# 6540-424) was added to the lysing matrix. The extra- and intracellular DNA fractions were sequentially extracted by applying the procedure by Gómez-Brandón et al. (2017a). Briefly, 0.1 g cut-milled sample was treated with 500 μl 0.12 M Na₂HPO₄ (pH 8) and gently shaken on a horizontal shaker for 30 min; thereafter, the liquid phase containing the recovered extracellular DNA fraction was separated from the solid phase (pellet) containing intact cells by low-speed centrifugation (4 °C, 30 min, 7,500 × g). The procedure was repeated twice, for a total of three washing and centrifugation cycles, and the supernatants were pooled. The remaining pellet was processed as described for the total DNA extraction, yielding the intracellular DNA fraction.

All DNA extracts were purified using the GeneClean Kit (components of the MP Biomedical DNA Extraction Kit) and their quality and quantity was checked via agarose gel electrophoresis and fluorimetric measurement (Qubit, LifeTechnologies), respectively.

The amplicon library for Illumina Miseq sequencing was prepared in a two-step procedure according to the standard protocol. First, the bacterial 16S rRNA genes of the DNA extracts were amplified using Illumina TruSeq chemistry adapters fused to bacterial 16S primers 515f and 806r (Caporaso et al., 2011). In a second step, samples were barcoded using Illumina TruSeq chemistry and flow cell adapters were attached to the amplicons. The library was sequenced in an Illumina MiSeq PE250 nano run at VBC Vienna. Sequencing data were deposited in SRA under BioProject Number PRJNA 682981.

From demultiplexed fastq files, an amplicon sequence variant (ASV) table was obtained using Dada2 pipeline (version 1.16, Callahan et al., 2016) following the standard protocol provided by the developers on github (version 1.16). Briefly, adapters and primers were trimmed and sequences with ambiguous base pairs were removed from the dataset. Error models were predicted from randomly selected samples from the run and both forward and reverse reads were filtered according to the error profiles. Sequences were merged and filtered for chimeric sequences. Sequences outside a range of 250–254 bp were discarded. A frequency ASV table was generated from the dataset. Taxonomy was assigned to ASVs using SILVA reference database v.138. An ASV-approach was chosen over an approach based on operational taxonomic units (OTUs) for reasons of precision and reproducibility (Callahan et al., 2017).

The resulting ASV table was filtered prior to further analysis. Only ASVs detected in at least two out of the three samples per sample group were considered, that is the presence of ASVs in the same DNA type extracted from different logs classified into the same DCl. In addition, ASVs with <5 reads per sample group were removed from the dataset. Applying these criteria, the initial number of 1,946 ASVs was reduced to 623 ASVs in the final dataset. Although the removal of ASVs <5 reads reduced the total richness calculated for the samples, it did not affect the comparisons of α-diversity among sample groups. The removal of filtered ASVs also did not affect the compositional distances between samples as inferred from comparing the ordination resulting from both principal component analysis (PCA) of the ilr-transformed compositional data and Non-metric Multidimensional Scaling (NMDS) on Bray-Curtis distances using Procrustes analysis (PCA: m²_Procrustes = 0.0985, correlation of symmetric rotation = 0.950, p = 0.001; NMDS: m²_Procrustes = 0.1568, correlation of symmetric rotation = 0.918, p = 0.001). Both ordination methods showed comparable results in terms of clustering by sample groups. For illustration purposes (Figure 1B), NMDS was chosen here.

Sequencing depth differed between sampling groups (p_Anova = 7e-5) (Supplementary Figure 1A). In addition to smaller differences, especially the extracellular DNA fraction of DCl V had a significantly higher number of sequencing reads (prior to filtering and thereafter) compared to all other sample groups (min_{extra; DCl V} = 14,382, median_{extra; DCl V} = 14,593, max_{extra; DCl V} = 22,789 reads after filtering). No differences were found among the other sample groups (min = 4,608; median = 8,007; max = 11,715 reads after filtering). Moreover, there were no significant differences in sequencing depth comparing the groups of main effects, which are DCl (I, III, and V) and DNA type (extracellular, intracellular, and total) (p_{Anova; DCl} = 0.124, p_{Anova; DNA type} = 0.058).

All data analyses were performed in R 4.0.2 (R Core Team, 2020). Package vegan (Oksanen et al., 2019) was used for ecological multivariate analysis, including NMDS and Procrustes analysis. For PCA, data were ilr-transformed (isometric log ratio) using the R package compositions (van den Boogaart et al., 2020). ASVs with differential abundance among DCls were identified by calculating a generalized linear model using ALDEx2 (Fernandes et al., 2013). Breakaway (Willis et al., 2020) was used to estimate the α-diversity/ASV richness. Data were visualized using lattice (Sarkar, 2008), ggplot (Wickham, 2016), ggtern (Hamilton and Ferry, 2018), and Vennerable (Swinton, 2013). Colors were chosen from RColorBrewer (Neuwirth, 2014).

Assuming that more niches become available as deadwood decomposition progresses, we would expect that the number of ASVs detected increases with higher degree of decay that is from DCl I to V (Hunter, 1990). However, we did not observe an increase in ASV richness as a function of progressing deadwood decomposition. This was in line with the study on P. abies logs by Hoppe et al. (2015), who also applied sequencing methodology (454 pyrosequencing). Rinta-Kanto et al. (2016), in contrast, reported a higher bacterial richness in P. abies logs observed via microbiome sequencing data (454 pyrosequencing) and in line with this, a higher bacterial abundance was observed via quantitative PCR of 16S rRNA gene copy numbers as decomposition progressed. The increased richness with more progressed decay has also been confirmed by Kielak et al. (2016b), who researched decaying Pinus sylvestris logs. Tláskal et al. (2017), however, did not confirm an increased bacterial richness with progressing decay of different tree species, including P. abies. As the majority of α-diversity measures (richness), including the number of OTUs/ASVs is largely influenced by the number of samples, sequencing depth and data treatment, such as rarefying and filtering (Willis, 2019; Willis et al., 2019), an explanation for these controversial observations might be methodological rather than ecological. The sample sizes in the studies of Rinta-Kanto et al. (2016) and Tláskal et al. (2017) exceed the sample size of our study (n = 27), which is more comparable to the one from Hoppe et al. (2015). A higher number of samples per DCl, potentially even higher than that studied in Rinta-Kanto et al. (2016) and Tláskal et al. (2017) might further optimize the microbiome representability of the sample group and contribute to solving the riddle of decaying deadwood α-diversity. In addition to sample size, we hypothesize that richness estimations in microbial ecology suffer from an upper threshold of microbial units that can be observed in a sample. The reasons for this are a combination of the vast bacterial diversity known today, the patchiness of most ecosystems that are studied by pooled samples across physical space and unavoidable methodological limitations and restrictions, such as the amount of environmental sample used for DNA extraction, the template volume applied to PCR and the limited total number of amplicons subjected to sequencing. In addition, limited sample sizes, by means of low biological replications restrict the prediction of more reliable richness estimators. The increased depth of analysis obtained by the sequentially extracted DNA fractions of the total DNA pool clearly evidences the potential of the applied fine-tuning DNA approach, pinpointing that the statements on bacterial richness in natural ecosystems are problematic. Further investigations optimizing experimental design, sampling strategy, richness estimation, data analysis, and interpretation are necessary to make studies more comparable and a better representation of the ecosystem.

Significant differences with regard to the bacterial community composition were expected among DCls according to their varying physico-chemical properties (Gómez-Brandón et al., 2017a). While pH, electrical conductivity and cellulose concentration decreased with increasing DCl, the moisture content as well as the concentrations of micro- and macronutrients such as N, Fe, and P increased (Gómez-Brandón et al., 2017a).

The vast differences in bacteriome compositions (Figure 1) reflected the vast ecological differences among deadwood at different decay stages. Despite the ability of some bacteria to degrade complex polymers, fungi are recognized as the main deadwood decomposers (Eastwood et al., 2011). The bacterial and fungal compositions are, therefore, likely interconnected within the deadwood habitat. As bacteria usually feed on the metabolites released by fungal activity, fungi may inhibit bacterial growth by the acidification of their environment and/or by the production of antibiotics (Folman et al., 2008; Johnston et al., 2016). With this in mind, acidotolerant, antibiotic resistant bacteria and those beneficial for the wood-inhabiting fungi are more likely to appear in higher quantities (Figures 1, 2). A high number of reads and a high number of ASVs were annotated as Acetobacteraceae (Alphaproteobacteria). Members of this taxon usually have an optimum pH ranging between 5 and 6.5 or even lower (Reis and dos Santos Teixeira, 2015), which might be advantageous for the fungi-dominated deadwood environment. Moreover, some members of this taxon are even able to fixate nitrogen (Reis and dos Santos Teixeira, 2015). Genes coding for enzymes involved in N-cycling, including N-fixation, have been found wide-spread across the bacterial phylogenetic tree (Nelson et al., 2016). Also some members of the taxa Burkholderiales and Rhizobiales, which were also found in high read proportions and numbers of ASVs (Figure 1), have been associated to this phylogenetically unconserved trait (Nelson et al., 2016; Shamseldin et al., 2017). In fact, members of these taxa have also been associated to methylotrophy (Chistoserdova et al., 2009), potentially giving them another selective advantage, especially with regards to co-inhabiting the deadwood together with fungi. Recently, Gómez-Brandón et al. (2020) showed that OTUs annotated as both Rhizobiales and Burkholderiales (Gammaproteobacteria) were associated to deadwood colonizing fungi in the very early stages of P. abies decay (2 years of decomposition). In their study, associations between mycorrhizal fungi and Burkholderia ASVs, which were also differentially abundant across DCls here (Figure 2), were reported. The changes in the fungal composition as a function of progressing decay (Arnstadt et al., 2016; Gómez-Brandón et al., 2020) are therefore likely to influence the bacteriome of the deadwood. In fact, this is reflected by the differences in relative abundance between DCls with regard to Rhizobiales, Burkholderiales, and Acetobacteraceae (Figure 2). The high selective pressure on bacteria in deadwood might result in a high turnover and a strong fluctuation of low-abundance populations, which is reflected by the high numbers and read proportions of ASVs that were detected in solely one DNA type (extracellular vs. intracellular fraction vs. total DNA pool) (Figure 4).

In addition to Proteobacteria, Actinobacteriota (peak abundance in DCl I), Acidobacteriota (peak abundance in DCl III), and Verrucomicrobiota (peak abundance in DCl V) were detected in our P. abies logs. These three phyla have been commonly found in bacterial communities colonizing deadwood from P. abies and other tree species (Sun et al., 2014; Hoppe et al., 2015; Kielak et al., 2016a; Rinta-Kanto et al., 2016; Tláskal et al., 2017; Johnston et al., 2019). Within Actinobacteriota, Solirubrobacterales accounted for the highest relative abundance in our study. Some members of Solirubrobacterales can act as lignin-degraders (Wilhelm et al., 2019), which might explain both their presence in all DCls and their peak abundance in DCl I.

In the case of Acidobacteriota, they peaked in relative abundance in DCl III. While Acidobacteriota ASVs were frequently detected in intra and in/ex (Figures 3B, 4C) and, therefore, likely reflect potentially living bacterial populations (intact cells), Actinobacteriota ASVs were frequently found in extra and tot/ex (Figure 4D) and are, therefore, potentially not viable. This latter phylum has been reported to occur on tree stems, including those from P. abies (Cregger et al., 2018; Ren et al., 2019). In our study, Actinobacteriota had a high ASV richness and to a certain extent read proportion in DCl I, suggesting that they might originate from populations that colonized the living tree, even though their metabolic role in P. abies deadwood degradation was probably minor owing to their pronounced presence in the extracellular DNA fraction. In fact, the capacity of some members of the Acidobacteriota to grow on cellobiose and N-acetyl-glucosamine and to produce enzymes required for chitin and cellobiose degradation has previously been shown for Acidobacteriota from forest soil (Lladó et al., 2016) and their adaptation to acidic conditions (Sait et al., 2006; Kielak et al., 2016a) might make some Acidobacteriota equally fit for DCl I and potentially more resistant to fungal competition than Actinobacteriota. The roles of these two phyla, their dynamics and their interactions with fungal deadwood degraders will help in understanding deadwood ecology and warrants deeper investigation.

In DCl V, a high number of ASVs and high read proportions were annotated to Verrucomicrobiota, exceeding both Actinobacteriota and Acidobacteriota (Figures 1C, 4B). Although Verrucomicrobiota are wide-spread across a range of environments, their vast majority remains uncultivated and is consequently, metabolically undescribed (Wagner and Horn, 2006). However, some few members isolated from high-temperatures, acidic environments have been shown to be methylotrophic (Wagner and Horn, 2006). As methylotrophy could provide an advantage in the deadwood environment due to fungal lignolytic activities that may result in the release of methanol, this trait might be spread across deadwood Verrucomicrobiota, thereby explaining their peak abundance in DCl V and both their high ASV richness and read proportions in the set of intra and in/ex ASVs (Figures 3B, 4C).

Despite their low proportions, Planctomycetota, mainly WD2101 soil group, were increasingly detected from DCl I to V (Figure 2). Several members of this taxon are known for degrading complex polymers, including cellulose (Wang et al., 2015), suggesting their potential role in deadwood degradation, which still remains underexplored (Tláskal et al., 2017; Probst et al., 2018). Planctomycetota have unique cell-wall and outer membrane structures as well as the ability to form biofilms and extracellular polysaccharides. Extracellular DNA has been shown to stabilize biofilms, to contribute to cell-cell-interaction within microbial populations and to serve as nutrient source (as reviewed by Pietramellara et al., 2009). In addition, extracellular DNA has been shown to increase Planctomycetota (WD2101) in soil (Morrissey et al., 2015). We hypothesize that the high concentration of extracellular DNA (Table 1) and the deadwood’s environmental approximation to soil with increasing decay might have benefited Planctomycetota abundance (Figure 4D) and might also contribute to the higher read proportions of Planctomycetota in DCl V compared to the other DCls and the higher numbers of unique extra ASVs observed in this DCl (Figure 4A).

It seems that multiple factors shape the microbiome of deadwood and determine its trajectory. Especially in terms of functional roles, more research involving the characterization, identification and description of isolates is necessary to shed light into the complex degradation process. In addition, co-cultivation experiments with wood inhabiting fungi should be performed in the future to study the interaction of deadwood inhabiting fungi and bacteria and to understand to what extent fungi shape the deadwood bacteriome.

The information about the bacteriomes derived from the three DNA types (extracellular vs. intracellular fraction vs. total DNA pool) followed the same trajectory (Figure 1). However, probably due to both methodological constraints and ecological aspects, they appear to differ in substantial details (Figure 4). The crucial impact of the rare biosphere on complex ecosystems is widely recognized (Jousset et al., 2017). The sequential extraction of the extra- and intracellular DNA fractions increased the ecological depth of analysis of the P. abies wood bacteriome, as compared to the total DNA pool (Figure 3). However, it often lowers the total amount of DNA extracted from the environmental material, due to repeated washing, ultimately resulting in the sum of intra- and extracellular DNA being lower compared to the total DNA pool (Table 1). Although repeated rounds of extraction from the very same sample can be applied in total DNA extraction procedures and may increase the microbial diversity observed by microbiome sequencing, the methodological differences between the total DNA pool and its extra- and intracellular DNA fractions (Wagner et al., 2015) suggest that analyzing these two fractions inherently results in greater depth (Figure 3). This observation was also reported for soil even with low-resolution molecular methods (Ascher et al., 2009; Chroňáková et al., 2013; Gómez-Brandón et al., 2017b), and for biogas slurries by using next-generation sequencing technologies (Nagler et al., 2018a; Nagler et al., 2020). From an ecological viewpoint, the extracellular DNA can be considered a natural product of microbial growth (Nagler et al., 2018a), including former ecosystem states. This latter aspect addresses the persistence of extracellular DNA in the environment as its degradation can be prevented by binding to various organic (e.g., cell debris) and inorganic, mineral colloids (soil components) or bacterial extracellular polymeric substances (Agnelli et al., 2007; Pietramellara et al., 2009; Nagler et al., 2018a). In agreement with this, we found a higher ASV richness in the extracellular DNA fraction compared to the intracellular one (Figure 1). Moreover, in line with the recent study by Nagler et al. (2020) on slurry bioreactors, we showed that DNA of certain bacterial populations remained detectable in the extracellular fraction, while it was not detectable in the respective intracellular DNA fraction, suggesting that these populations were no longer occurring in the monitored biological assemblage at the timepoint of sampling (Table 2). The conceptual differences between the extra- and intracellular DNA fractions imply that persisting DNA (extracellular DNA occurring in the extracellular environment) might partially mask intracellular DNA (methodologically extracted from intact cells via induced cell disruption) and thus obscure the snapshot of occurring microbes. Like the differential copy numbers of marker genes observed between different taxa, species-specific differences in DNA release might bias the composition observed in the extracellular DNA fraction and the total DNA pool. This potential masking effect of the extracellular DNA fraction over the intracellular one affects the assessment of the microbiome when directly analyzing the total DNA pool (made up by both DNA fractions), depending on its proportion with regard to the intracellular DNA fraction (Nagler et al., 2018a; Carini et al., 2020). For example, in soil and the marine environments up to 60% (Agnelli et al., 2004) and 90% (Torti et al., 2015) of the total DNA pool account for extracellular DNA, respectively. While the proportions of both DNA fractions are unknown for the majority of ecosystems, they were comparable in our P. abies deadwood samples (Table 1). They might even be patchy within ecosystems. Especially in terms of smaller operational units, single populations and rare biosphere, the proportions of DNA fractions and the potential masking effect of extracellular DNA should be considered. More studies on the masking effect of extracellular DNA in different ecosystems are necessary to confirm its generalizability (Nagler et al., 2020).

In our study, we found that the frequencies of several bacterial classes differed among DNA types (Supplementary Table 2). Our results suggest that the ecological role of Acidobacteriota in the decomposition of P. abies deadwood might be underestimated. Among the intra ASVs (Figure 3B), the diversity and abundance of Acidobacteriota was significantly higher than in the other two DNA types, especially in DCl I and also in DCl III (Figure 4). In addition, the abundance of indicator ASVs among DNA types fine-tuned their ecological interpretation (Table 2). In studies based on the total DNA pool, the composition of the extracellular DNA can mask the arrival, establishment and extinction of the ASV corresponding organisms despite “sufficient” sequencing/sampling depth, usually analyzed by rarefaction curves and sequencing coverage. In fact, the variations in bacterial abundances or even detection introduced by the potential masking effect of the extracellular DNA could complicate the assessment of microbiome-environment interactions. Consequently, the ecological role of organisms might be misinterpreted. Given the differential frequencies of bacterial classes in our study, this seems to hold true not only for species, but also for higher levels of taxa, which might change the perspective on a broader ecological scale. Due to the absence of indicator ASVs in the total DNA pool in some DCls (Table 2, Figure 4E), we might conclude that we do have statistical tools that can handle microbiome data, while we do not feed them with high-resolution microbiome data sets, which is not necessarily achieved solely by sequencing depth. Our findings provide further evidence that this can be accomplished with relatively little effort by the proposed fine-tuning DNA approach. Future studies can greatly benefit from studying different DNA types or at least considering their concept for a more accurate interpretation of molecular environmental data.

Despite the concerns raised regarding their conceptual differences, the DNA types studied here broadly resembled a comparable bacteriome and our results indicate that the differences in terms of composition with regard to their ecological interpretation might benefit scientific research. The analysis of the total DNA pool resulted in a robust, meaningful picture of the deadwood ecosystem. However, its resolution – independent of sequencing depth – is not as deep as the one provided by intracellular DNA, which is in agreement with Nagler et al. (2020) who made the same observation for slurry bioreactors. Especially if no exact time-depending compositional data are needed, the total DNA pool appears as a good choice, as it reflects the picture of the ecosystem including its former states. Studies focusing on microbial diversity or microbial populations with low-abundance might benefit from the increased sequencing depth provided by the intracellular DNA fraction. In these studies, the more complex extraction of intracellular DNA might be worth the effort. A general disadvantage of intracellular DNA analysis – equally to total DNA – by using commercial extraction kits, is the usually small amount of starting material. Studies on large and complex ecosystems – such as soil – can suffer from this, especially if they are heterogeneous and patchy. Analysis of the extracellular DNA can easily be up-scaled to larger amounts of input material and provide an efficient way of DNA extraction. However, ecosystems with high rates of microbial turnover and ecosystems underlying big disturbances or short-term successional changes might lack accuracy in terms of microbiome composition. Further studies are necessary to understand general concepts of DNA release into the extracellular environment. For evolutionary studies, it might be meaningful to analyze both DNA fractions of the environmental DNA pool. Extracellular DNA might be incorporated by living cells (natural transformation) thereby contributing to genome diversification and consequently, bacterial evolution/adaptation. Intracellular DNA can give a clear indication regarding which organisms diversify. In this context, metagenomic studies could prove especially valuable.