Environmental patterns of brown moss- and Sphagnum-associated microbial communities

Northern peatlands typically develop through succession from fens dominated by the moss family Amblystegiaceae to bogs dominated by the moss genus Sphagnum. How the different plants and abiotic environmental conditions provided in Amblystegiaceae and Sphagnum peat shape the respective moss associated microbial communities is unknown. Through a large-scale molecular and biogeochemical study spanning Arctic, sub-Arctic and temperate regions we assessed how the endo- and epiphytic microbial communities of natural northern peatland mosses relate to peatland type (Sphagnum and Amblystegiaceae), location, moss taxa and abiotic environmental variables. Microbial diversity and community structure were distinctly different between Amblystegiaceae and Sphagnum peatlands, and within each of these two peatland types moss taxon explained the largest part of microbial community variation. Sphagnum and Amblystegiaceae shared few (< 1% of all operational taxonomic units (OTUs)) but strikingly abundant (up to 65% of relative abundance) OTUs. This core community overlapped by one third with the Sphagnum-specific core-community. Thus, the most abundant microorganisms in Sphagnum that are also found in all the Sphagnum plants studied, are the same OTUs as those few shared with Amblystegiaceae. Finally, we could confirm that these highly abundant OTUs were endophytes in Sphagnum, but epiphytes on Amblystegiaceae. We conclude that moss taxa and abiotic environmental variables associate with particular microbial communities. While moss taxon was the most influential parameter, hydrology, pH and temperature also had significant effects on the microbial communities. A small though highly abundant core community is shared between Sphagnum and Amblystegiaceae.

The majority of global wetlands are peatlands (70%) 1 , of which 80% are natural and pristine 2 . Due to the high water content, low temperatures 3 , anoxia and the production of recalcitrant organic matter and microbial inhibitors (e.g., polyphenols) 4 , dead organic matter in northern peatlands decomposes slowly. Almost 20% of the global soil organic carbon and 10% of the global freshwater reservoirs are stored in peatlands 5 . Thus, these ecosystems fulfil crucial functions like maintaining regional water balance and sequestering carbon 6,7 . The vegetation of natural peatlands is dominated by bryophytes of the Amblystegiaceae family (brown mosses) and of the genus Sphagnum. Both moss types play an important role in succession and peat formation. Amblystegiaceae initialize peat formation by colonizing open water bodies ('terrestrialisation'), which leads to the formation of minerotrophic fens. Over time, Sphagnum species establish and contribute to peat accumulation ('paludification'), transforming those fens into ombrotrophic bogs [8][9][10][11][12] . The transition from a fen to an ombrotrophic bog profoundly alters the ecosystem carbon budget due to doubled net primary productivity and a fourfold decrease in the decomposition rate, causing a several fold increase in the peat accumulation rate 13 . Such shifts especially in vegetation have been observed to coincide with drops in pH from 7 to around 4 [14][15][16] .
Microorganisms play a crucial role in the C and N cycles of peatlands [17][18][19] . Moss-associated microbial communities are suggested to maintain host plant health and pathogen defence [20][21][22][23][24] . In addition, moss shoots, specifically their cell walls, provide a matrix where the microorganisms can attach; additionally, cell anatomy and cell wall chemistry are expected to influence microbial community structure. Sphagnum leaves differ from Amblystegiaceae and other mosses mainly by their unique anatomy, as most of the leaf volume is formed by

Materials and methods
Study sites and sampling. We studied two main ecosystems (Amblystegiaceae-and Sphagnum-dominated peatlands), represented by four sites, which are analogous to different stages in the transition from fens to incipient ombrotrophic bogs: 1) High-Arctic lakes with colonies of Amblystegiaceae on Svalbard (SV), Norway; 2) Arctic polygonal tundra with densely growing Amblystegiaceae on Samoylov Island (SA), Lena Delta, Russia; 3) Sub-arctic Sphagnum palsa peatlands in Neiden (NEI), Northern Norway, and 4) temperate Sphagnum kettle bogs in the Mueritz National Park (MUE), Northern Germany. An overview about the study sites and a simplified sampling scheme is given in Fig. 1, and a detailed overview of the 157 samples and the corresponding sampling sites, biotic (plant species) variables, environmental variables and geographical coordinates is provided in Tables S1A and S1B.
On SV (78.9° N, 11.9° E), we sampled a mixture of submerged Amblystegiaceae from three subsites in the vicinity of Ny-Ålesund: Bryum pseudotriquetrum in Twin Water (TW), Drepanocladus trichophyllus and Scorpidium turgescens in Knudsenheia (KNU) and Drepanocladus revolvens and S. turgescens in Gludneset (GLU), each in duplicates (sample type 1; all sample types are depicted in Fig. 1). On SA (72.4° N, 126.5° E), we sampled a mixture of submerged Scorpidium scorpioides and Meesia triquetra from an interpolygonal crack (PC; sample type 2a) and S. scorpioides from a polygonal pond (three replicate of plants subsumed to PP; sample type 2b). At both locations, SV and SA, we collected sediment underneath the mosses as references. In NEI (69.7° N; 29.4° E), different successional palsa stages were selected: thermokarst ponds with Sphagnum riparium adjacent to degrading palsas (one plant within each of the subsites NEI1, NEI2; sample type 3a), thermokarst ponds with S. riparium as remnants of collapsed palsas (one plant within each of the subsites NEI3, NEI4; sample type 3c) and hollows with Sphagnum lindbergii, representing old successional stages of previously collapsed palsas (one plant within each of the subsites NEI5, NEI6, NEI7; sample type 3b). From MUE (53.3° N, 13.2° E), three subsites were chosen: Heidbergmoor, a hummock-hollow complex (sample type 4a) with emerged Sphagnum fallax (three replicate plants within the subsite called HEI2) and submerged S. fallax (one plant within the subsite called HEI1); Klockenbruch, a kettle bog with an oligotrophic, elevated center (sample type 4c) with Sphagnum magellanicum (three replicate plants within KLO1) and a meso-oligotrophic, lower margin (sample type 4b) with S. fallax (three replicate plants within KLO2); Kiebitzmoor, a formerly drained and rewetted kettle bog (sample type 4c) with S. magellanicum (three replicate plants within KIE). In NEI we collected the sedges Eriophorum sp. and Carex sp. (NEI1, NEI5, NEI6, NEI7) and sediment underneath the mosses (NEI1, NEI2, NEI3, NEI4) as references, with duplicates for each site and reference type. In MUE, Eriophorum vaginatum (HEI2, KLO1 and KLO2) and Carex sp. (KIE) were collected as references with duplicates for each site and plant type. Sampling in all sites was done between June and September 2013.
Peat or moss batches were sampled using gloves and sterile knifes or spoons. Leafs, stem and upper root material of vascular plants were manually extracted from the peat body, washed with sterilized tap water to get rid of microorganisms that came from the surrounding environment and were not attached to the plants, cut and Pore water analysis (pH, temperature, methane, DOC and O 2 ). Values of pH were measured in the field using a multi parameter probe Multi 350i from WTW (Laboratory and Field Products, Nova Analytics). Air and peat temperatures were measured with a hand-held digital thermometer 2000T (Thermocouple Thermometer, Digitron Instrumentation Ltd, England) equipped with a 50 cm long probe. Headspace methane concentrations were measured by gas chromatography shortly after pore water sampling as described elsewhere 37 . For the determination of DOC values, 20-mL glass vials (Agilent) were flushed with ultrapure water, baked at 550 °C for 2 h, closed with aluminium-sealed PTFE/butyl septa and acidified with 3% HCl Suprapur (VWR). 15 mL of the pore water was filtered with 0.7 µm GF/L filter (Whatman). The samples were sent to 'Potsdamer Wasser-und Umweltlabor GmbH' (PWU) for DOC analysis. Pore water O 2 contents were measured in the field at different depths (above, within and below moss layer, where possible), using an optical oxygen meter (FireStingO 2, Pyro-Science).
Cell wall analysis (CEC, HC, LLP). Cation exchange capacity (CEC). Up to 45.0 mg of dry moss samples were sealed into labelled polyamide mesh bags. The bags were submerged in 2 L of 20 mM HCl to soak the moss up and to convert all carboxylic cation-exchange sites to un-dissociated form; free protons were then replaced by repeated thorough wash with distilled water. All the bags were then transferred to 2 L of 0.5 M ammonium acetate and after pH equilibration the ammonium acetate solution was renewed and adjusted to pH 7.0 using NH 4 OH. The bags were repeatedly washed with large amount of distilled water to replace free NH 4 + and dried. The bags were individually immersed to 50 mL of 20 mM HCl and shaken for 15 min to elute cell-wall bound NH 4 + ions. The eluate was sampled and NH 4 + analyzed colorimetrically using Flow Injection Analysis (Foss Tecator AB, Sweden).  Separation of loosely (epiphytic) and closely associated (putative endophytic) microorganisms. Between 2.2 and 5.3 g of the moss material pre-treated as described above was thawed and amended with extraction buffer containing ultrapure DEPC water (AppliChem), 0.85% NaCl (Merck), and 0.01% Tween20 (AppliChem) in a ratio 2:1 (weight percent), a modification of a previously published protocol 38 . The mixture was shaken horizontally for 1 h at 4 °C prior to ultrasonication (Bandelin Sonoplus HD3100) with pulsation for 2 min (1 s off, 2 s on) at 0.45 W/mL 39 . Extraction buffer containing the epiphytes was filtered through a 0.2 µm cellulose filter (Sartorius Stedium). The remaining moss was surface-sterilized with 0.15% NaOCl (Roth) for 1 min, and rinsed 7 times with DEPC water according to a modified protocol 40 . Filters and sterilized mosses were ground to powder under sterile conditions with liquid nitrogen, transferred to lysis tubes and stored at − 20 °C until DNA extraction. For each moss sample, one filter with wash-off (epiphytes) and two technical replicates of the surface-sterilized moss (putative endophytes) were used for DNA extraction and sequencing (below).
DNA extraction and sequencing. Genomic DNA was extracted from 0.4-0.8 g of each of the surface treated mosses, the filters containing the wash-off as described above, and from about 0.4-0.8 g untreated sedges and sediment samples as references for the moss samples following the CTAB/phenol-chloroform-based method of 41  . The library preparation and sequencing was performed on an Illumina MiSeq sequencer by the company GATC (Konstanz, Germany) according to their standard protocols. In short, the library was prepared with the MiSeq Reagent Kit V3 for 2 × 300 bp paired-end reads. To account for the low-diversity amplicon sampling, we used 15% PhiX control v3 library.
Sequence analyses and bioinformatics. Raw data was demultiplexed using CutAdapt 44 ; e 0.1; -trim-n; no error in barcodes allowed. Paired-reads were merged using PEAR 45 (Q25; p 10^-4; v20) and sequence orientation was standardized using own scripts. Low quality sequences were filtered and trimmed using Trimmomatic 46  www.nature.com/scientificreports/ pick_open_reference.py script with a cutoff value of 97% 47 . Representative sequences of the clusters were annotated with usearch utilizing the curated Greengenes 13.8 taxonomy database 48 . OTUs assigned to chloroplasts, bacterial OTUs within archaeal samples and vice versa, and OTUs with a small, sample-wise relative abundance (< 0.01%) were filtered before further exploration.
Statistical analyses. Differences in microbial community composition between the sites were obtained by calculating the inverse Simpson index and counting the number of OTUs, as measures of the OTU diversity and richness, respectively. Bubbleplots for Figs. 5 and 6 were generated using the package 'ggplot2' (version 2.2.0) within the statistical software R (version 3.2.2) 49 . Correlation matrices of samples (16S rRNA gene datasets of either bacteria or archaea) were generated using the R function 'cor' , specifying the Spearman rank correlation coefficient. Hierarchical clustering of the samples based on the correlation matrices to generate dendrograms were calculated using the method 'agnes' within the R package 'cluster' , with default settings. All heatmaps were created using the R package 'heatmap3' (version 0.3.3). The inverse Simpson index diversity estimates for bacteria were calculated using the R package 'asbio' (version 1.6-5). Pairwise t-tests were used for environmental variables and carried out using the R function 'pairwise.t.test' . Pairwise Mann-Whitney-Wilcoxon tests were used for diversity indices and carried out using the R function 'pairwise.wilcox.test' . Canonical correspondence analysis (CCA) was carried out to quantify the explanatory power of biotic and environmental variables with respect to the microbial ecology of the peatlands (package: vegan (version 2.2.1)). Correspondence analysis (CA) was carried out as described in 50 and plotted using 'ggplot2' . Eight variables; cation exchange capacity, ligninlike polymers, hemicellulose, total nitrogen, total carbon and C:N ratio, DOC, oxygen and water content were removed from the initial full model due to lacking observations for between 23 and 45% of the samples. We constrained the variation in the microbial communities to the remaining variables; (1) sites (SV, SA, MUE and NEI), (2) subsite (e.g., KIE1), (3) plant species or reference sediment, (4) location above or below water table, (5) washed and surface-sterilized moss plant (putative endophytes) or wash-off (epiphytes) (6) pH, (7) methane concentration in pore water, and 8) temperature. In order to estimate and account for the spatial autocorrelation that the sites (1) and subsite (2) variables represent, we introduced partial CCA. Running the model without (1) and (2) we observed that the constrained inertia was reduced from 72% of total inertia to 40%. Subsequent analysis of variance inflation factors showed that no remaining variables were redundant. Core communities were calculated with a restrictive 66% threshold, meaning that an OTU has to be present in 80 out of 122 samples and in both system types (Amblystegiaceae and Sphagnum) for it to be considered part of the core microbiota.
Moss system core communities (Amblystegiaceae or Sphagnum) were calculated with the same threshold, 66%. Moss species communities were calculated with a more restrictive threshold of 75% 30 .   (Fig. S1). Interestingly, the microbial communities in the Amblystegiaceae sites displayed the same level of diversity, independent of the geographical location. This also applied to the Sphagnum sites. Unlike the bacterial diversity, the archaeal diversity was similar between the Amblystegiaceae and Sphagnum peatlands with overall little differences between mosses, sediments and vascular plants (Fig. S2). An exception was that the sediments in the Amblystegiaceae sites displayed slightly higher archaeal richness than the other sites.

Results
To identify the association between moss taxa, abiotic environmental variables and the bacterial moss microbiota, we performed a canonical correspondence analysis (CCA). Using variance partitioning we quantified the contribution of the variables to the explanation of total inertia in the following order from most to least important: (1) plant species and reference sediment: 19.7% (p. value < 0.001), (2) temperature: 4.8% (p. value < 0.001), (3) putative endophytes or epiphytes: 4.6% (p. value < 0.001), (4) methane concentration in pore water: 4.1% (p. value < 0.001), (5) pH: 3.3% (p. value < 0.001) (6) location above or below water table: 3.2% (p. value < 0.001). Repeating the procedure with Hellinger transformed data to control for large effects of low abundant OTUs we observed the same patterns at highly similar total and constrained inertia, suggesting that the impact of rare OTUs on CCA ordination was minor. Initially, our model included sites and subsites in addition to the six above Scientific Reports | (2020) 10:22412 | https://doi.org/10.1038/s41598-020-79773-2 www.nature.com/scientificreports/ mentioned variables, together accounting for 32% of the differences between the microbial communities (see materials and methods). As site effects and plant species effects correlate, it is likely that by removing site effects we have substantially underestimated the plant species influence. The removed fraction of the inertia contained in the site variables we consider the 'environment' , a mix of abiotic and biotic variables that cannot be studied in isolation with our dataset. Due to its complexity, we were unable to visualize the major gradients in the dataset using a single CCA plot. Thus, we plotted the constraints of the final model above separately (Fig. 3). The plots show that substantial parts of the bacterial communities correlate with the moss or vascular plant species (Fig. 3A) and by being submerged in water or above (Fig. 3F). At one instance we saw that the water table had a stronger impact on the microbiota than the plant species; the S. fallax samples closest to the S. riparium in Fig. 3A were submerged, as were all the S. riparium samples. There were also consistent differences between endophytic and epiphytic communities (Fig. 3C). Furthermore, in line with the differences in pH and temperature, we observed clear differences between the Amblystegiaceae and Sphagnum bacterial communities (Fig. 3B,E), while the effects of altered methane concentrations on the microbial communities were similar in Amblystegiaceae and Sphagnum ecosystems (Fig. 3D). The CCA explained approximately 40% of the variance in the dataset. The small explained variance was to a large extent due to the removal of area and subsite variables which was not considered explanatory variables.
To allow an evaluation of this area and site-dependent structure of the microbial communities we constructed a Spearman correlation based dendrogram of the OTU profiles, along with some of the categorical variables. The analysis revealed a very high level of agglomerative clustering in the dataset (0.86), particularly considering the large size of the dataset. The resulting dendrogram confirmed some of the previously observed patterns, such as the differences associated with dominating moss vegetation (Amblystegiaceae or Sphagnum) and hydrology (Fig. 4). However, it also revealed additional data structures. Starting from the top of Fig. 4, we can see the Amblystegiaceae-and Sphagnum-associated microbial taxa. To identify which bacterial and archaeal groups accounted for the majority of microbial community variation we studied the communities at family level. Within bacteria, we observed an evenly high abundance of families in the Amblystegiaceae moss microbiota; Acidimicrobiales_C111, Pseudoanabaenaceae, Hyphomicrobiaceae, Sphingomonadaceae and Comamonadaceae (Fig. 5).
In contrast, only two bacterial families dominated the Sphagnum moss microbiota; Acetobacteraceae and Acidobacteriaceae. Sphingomonadaceae was the only family present at similar relative abundances in the Amblystegiaceae and Sphagnum systems. To identify the nature of these large differences we studied the OTU composition of Acetobacteraceae in more detail (Fig. S3). The relative abundance of Acetobacteraceae was higher in www.nature.com/scientificreports/ Sphagnum than in Amblystegiaceae systems and the majority of Acetobacteraceae OTUs were present only in Sphagnum. However, some OTUs were only present in Amblystegiaceae, while a handful of OTUs were present in both the Amblystegiaceae and Sphagnum peatlands. To identify whether this distinctly different OTU composition with few overlaps was a general pattern, we repeated the analysis for other major bacterial taxa. The same pattern was observed for Acidobacteria (Fig. S4), Acidimicrobiales (Fig. S5), and Cyanobacteria (Fig. S6). This points towards distinct bacterial communities of Sphagnum-and Amblystegiaceae-dominated peatlands, while only individual OTUs occurred in both peatland types. Among the community of methane oxidizing bacteria (MOB), Methylocystis was most abundant (Fig. S7). Unlike most bacterial taxa, Methylocystis occurred in almost all of the sites but its relative abundance varied and correlated positively with the amount of methane in the pore water. In addition to Methylocystis, the MOB community contained members of within Methylomonas. In contrast to Methylocystis, OTUs affiliated with Methylomonas occurred primarily in Sphagnum sites, preferentially under submerged conditions. Finally, Methyloferula-associated OTUs were also detected among MOB but at low relative abundances. These OTUs were restricted to emerged Sphagnum sites. The fasta files of methanotrophic OTUs in Fig. S7 is provided as additional supplement (S_methanotrophs_fasta). The complete OTU table for bacteria is provided as supplementary information (Supplement_Bacteria_OTU_table). Clustering of the archaeal communities did not reveal any hierarchical clustering patterns related to sample origin (Fig. S8), as was observed for bacteria. The OTUs within the phylum Euryarchaeota, the majority of which belonged to families/genera of methanogenic archaea, dominated the archaeal communities (Fig. 6).
The most abundant OTU belonged to the hydrogenotrophic methanogenic family Methanobacteriaceae. This OTU was present in almost all the samples of both Amblystegiaceae and Sphagnum ecosystems (Fig. S9). Furthermore, Methanomassiliicoccaceae, Methanocellales, and Methanosarcinaceae were widespread while Methanosaetaceae occurred mainly in the Amblystegiaceae sites. Outside the phylum of Euryarchaeota, the Bathyarchaeota were abundant throughout most of the sites while Woesearchaeota mainly occurred in the Amblystegiaceae sites. The complete OTU table for archaea is provided as supplementary information (Supplement_Archaea_OTU_table).
The core and moss-associated microbiota. Only about 0.4% of all bacterial OTUs identified (49 out of 13,799) were observed in both Sphagnum and Amblystegiaceae ecosystems and were designated as the core microbiome (see materials and methods for definition of core microbiome). The majority of these OTUs belonged to Acetobacteraceae and Acidobacteriaceae, the dominating bacterial families of the Sphagnum micro-  (2), shows that the majority of samples from the same sites cluster together, but that exceptions related to hydrology and reference samples occur. Level (3) shows that bacterial communities from the same subsite almost exclusively cluster together. Level (4) show that the putative endophyte amplicon libraries from the washed moss plant almost always cluster with the wash-off library containing the putative epiphytic library from the same moss plant. Level (5) show a clear microbial community separation by the hydrology of the site. Level (6) show some clustering based on whether the community is from a vascular plant, sediment or moss.  (Table S2A). Further, we addressed whether the core microbiome of all mosses was similar in size to the individual bacterial core microbiome of Amblystegiaceae and Sphagnum mosses (moss system core communities), respectively. We found that by applying the same threshold as for the total core microbiome (TCM), the Amblystegiaceae core microbiome (ACM) was 348 OTUs, while the Sphagnum core microbiome (SCM) was 142 OTUs (Table S2A). Out of these, 20 were shared between TCM and ACM, while 46 were shared between TCM and SCM. To identify whether specific moss types shared larger core microbiomes, we calculated the moss species communities of the Amblystegiaceae mosses from Svalbard, the Amblystegiaceae mosses of Samoylov (only Scorpidium), Sphagnum riparium, S. fallax, S. lindbergii and S. magellanicum, respectively (Table S2B). This showed that the individual core microbiomes are in a similar size range as for the broader core microbiomes at 295, 548, 126, 132, 252 and 154 OTUs, respectively. Calculating the intersects of these core microbiomes we found that the Sphagnum mosses share a larger proportion of their core microbiomes with each other than with the Amblystegiaceae mosses (Table S2C). Interestingly, Scorpidium mosses shared more OTUs with the Sphagnum species than Amblystegiaceae mosses from Svalbard shared with Sphagnum mosses. The Svalbard and Samoylov mosses shared the highest number of OTUs, reflecting the larger overall number of OTUs associated to these mosses and their larger core microbiomes. Consistently for all core microbiomes calculated, the relatively few OTUs compared to the total number of OTUs identified accounted for a large proportion of the relative abundance in the microbial communities.
Next we sought to identify the dominant endophytic communities of Amblystegiaceae and Sphagnum mosses. Thus, we plotted the most abundant OTUs of significantly higher abundance in endophytic than epiphytic communities. This showed that almost none of the most abundant putative endophytes (washed and surface sterilized moss samples) of Amblystegiaceae were shared with Sphagnum (Fig. 7).
The taxonomic assignment and list of fasta files for the putative endophytes is provided as supplementary material (S_endophytes_taxonomy; S_endophytes_fasta). The Sphagnum endophytes belonged to several families www.nature.com/scientificreports/ within Proteobacteria and Acidobacteria, while the Amblystegiaceae endophytes belonged to Actinobacteria, Proteobacteria, Chloroflexi, Firmicutes and Gemmatimonadetes. Interestingly, of the 24 most abundant Sphagnum endophyte OTUs, 19 were observed in the total core microbiome. Thus these are primarily epiphytes of Amblystegiaceae.

Discussion
We studied the relationship between the moss-associated microbial communities and their respective moss taxa and physico-chemical environments in four sites, in which two were analogous to early (mosses of the family Amblystegiaceae) and two to later (mosses of the genus Sphagnum) successional stages of natural northern peatlands. We identified both biotic and abiotic controls on the microbial community structure and distinctly different communities in the Amblystegiaceae and Sphagnum mosses. The core community consisted of few but highly abundant members of the microbial community, of which many were endophytes of Sphagnum.
Biotic and environmental drivers of moss-associated bacterial communities. Our study approach allowed ranking the influence of individual variables on microbial community structure relative to the combined dataset. Given that the sum of individual variables reduced the explanatory power by almost 50% compared to the combined dataset that included also area and subsites, and also considering the hierarchical clustering in the bacterial dendrogram, we conclude that a large set of biotic and environmental variables correlate with the ecosystem type, Sphagnum-or Amblystegiaceae-dominated peatlands, and that the sum of these exert a major influence on the bacterial community structure. This is in accordance with other studies where soil bacterial communities were reported to structure according to ecosystems 51 and characteristic microbial communities which evolved in contrasting peatland ecosystems that differed mainly in vegetation, water chemistry and hydrology 52,53 . Sphagnum versus brown moss cover was also suggested as the major factor determining diversity and assemblages of testate amoebae 54 . Not only did the peatland ecosystem shape bacterial community structure, but also species richness and diversity, which was significantly higher in sub-neutral Amblystegiaceae compared to acidic Sphagnum peatlands, corresponding to findings from neutral compared to acidic www.nature.com/scientificreports/ soil environments 51,55 . Presumably, most bacteria lack substantial mechanisms to regulate their intracellular pH close to neutral when exposed to low extracellular pH and thus are not able to survive under acidic conditions 56 .
Contrarily, most bacteria adapted to low pH can survive at neutral pH, possibly explaining the occurrence of Acidobacteriaceae and Acetobacteraceae in all samples studied, but in much higher abundances in sites with low pH. Acetobacteraceae and Acidobacteriaceae were previously reported to be among the dominant Sphagnumassociated microbial groups 31,57,58 and generally thrive in low pH habitats 59,60 .
Among the individual variables, the specific plant taxa explained four to five times more of the observed inertia than any of the other controlling variables such as temperature, hydrology, and pH. This essentially shows that the different plants harbor their own microbiome which is similar to that of closely related plants. Differences in bacterial communities between different Sphagnum species were shown previously 27,31,32 . Our findings extend this knowledge on the influence of the host plant on bacterial community structure to other Sphagnum species such as S. riparium and S. lindbergii, to Scorpidium scorpioides, a member of the Amblystegiaceae, and also to some vascular peatland plants like Carex and Eriophorum. It further supports that although the sediment www.nature.com/scientificreports/ microbiome does correspond with the different dominant species, it is distinct from that associated with plants as indicated before 61 . The relationship between temperature and microbial community structure accounted for only 4.8% of inertia in the partial CCA although the investigated mosses were located in different climatic zones. Here, temperature is based on a point measurement and thus not necessarily expected to be representative of growing season average temperatures. Still, a poor influence of temperature on bacterial assemblages both on short and long time-scales is in line with other studies 62,63 . Hydrology was identified as another variable that affects moss-associated bacterial communities in natural peatlands. In some instances hydrology was more important than the influence of the host plant; emerged and submerged S. fallax sampled from the same subsite (HEI) carried different microbial communities while the microbiota of submerged S. fallax (Mueritz, Germany) was more similar to submerged S. riparium (Neiden, Norway) than S. fallax growing above the water table. This confirms a strong influence of hydrology on host moss morphology and physiology [64][65][66] and the associated microbial communities. The influence of the water table on moss-associated microbial community structure and activity was already reported for Sphagnum 20,24,67 but here we extend it to a larger spatial and environmental scale including Amblystegiaceae. Sphagnum and Amblystegiaceae-associated archaea were dominated by methanogenic Euryarchaeota, Bathyarchaeota and Woesearchaeota. This supports the hypothesis that Woesearchaeota occur in methanogenic environments, possibly as syntrophic partners with methanogens 68 . The role of Bathyarchaeota is unclear, but its presence corresponds to its former observations in peatlands 69 . There was some site-dependent clustering of the archaeal communities, e.g. Methanosaeta was only found in association with the neutral Amblystegiaceae mosses and associated sediments which is consistent with the biogeography of Methanosaetaceae 70 . The most abundant methanogen, belonging to Methanobacterium, was present in all sites and samples, which is also in accordance with the biogeography of Methanobacterium being among the most abundant methanogenic taxa both in pH neutral and low pH soils 70 . Similar to the community of methanogens, also the most abundant methane oxidizing OTU belonging to Methylocystis was present throughout all sites corresponding to its prevalence in wetlands 71 . Its abundance correlated with pore water methane concentrations (Fig. S7) which corresponds to related studies 72,73 . It was proposed that historical contingencies (i.e., random events), rather than evolutionary acquired fitness underlie the variations in Methylocystis communities 74,75 . Based on this, we suggest that Methylocystis adapted to the changes associated to succession, including pH. The observation that the communities of methane oxidizers in this study are dominated by Methylocystis, with high abundances of Methylomonas and Methyloferula in the Sphagnum-sites, extends on related studies which showed that Methylobacter typically thrives in Arctic pHneutral peatland soils 18,76 . Together with a number of studies focusing on the cultivation of methane oxidizers in Sphagnum peatlands [77][78][79][80] , this study further shows that the vegetation-associated methanotrophic community of Arctic peatlands is different from that of the peat soil (similar to what was discussed earlier for the overall microbiota) and that the low pH Sphagnum peatlands favor specific methanotrophic taxa such as Methyloferula.
Distinct patterns of endophytic bacteria. We identified distinct patterns of putative endophytic bacteria for both Sphagnum and Amblystegiaceae. The different bacterial endophytes in Amblystegiaceae compared to Sphagnum likely reflect a direct influence of the moss taxa on the microbiota. It was suggested earlier that Sphagnum plants select for beneficial bacteria through secondary metabolites 32 . The cell wall of Sphagnum contains polysaccharides and lignin-like polymers 81 , and our data show that the cell wall composition of Sphagnum and Amblystegiaceae is basically the same (Table S1A). Further, pectin-like polymers represent a minor fraction of cell wall polysaccharides, providing the mosses with substantial cation exchange capacity (CEC) 81,82 which is similar in both moss groups (Table S1A), corresponding to previous findings 13 . However, Amblystegiaceae have a greater proportion of lignin-like phenolic polymers compared to Sphagnum. Also, in Sphagnum-dominated sites the cation exchange maintains low pH. By contrast, the cation exchange does not reduce and control pH in Amblystegiaceae-dominated fens due to the substantial neutralization capacity of the mineral-rich groundwater. Besides a direct pH control, the cation exchange sites of Sphagnum pectin-like polymers inhibit microbial activity when these polymers hydrolyze and are released to the surroundings as so-called sphagnan 81 . Apart from the selection of beneficial microorganisms, Sphagnum protects itself against pathogens, e.g. via close association with antagonistic and antifungal bacteria or by release of antimicrobial substances [81][82][83][84] .
We further found that the putative endophytic microbiota was distinct from the epiphytic microbiota. It is known that bacterial endophytes can be host plant-specific promoting growth or health of their hosts 85,86 . Several OTUs, shown here to represent putative endophytes of Sphagnum or Amblystegiaceae, were previously reported as host plant-specific, e.g., Kineosporiaceae, Hyphomicrobiaceae, Intrasporangiaceae and Acidimicrobiales [87][88][89][90] . Endophytic microorganisms can be transferred from one generation to another, as shown for Sphagnum, or colonize the host from the surroundings after attraction by moss secreted chemical signals 31,40,91 . The inheritance and selection of potentially beneficial endophytes may provide an explanation for the distinct endophytic communities not only of Sphagnum but also of Amblystegiaceae. The core microbiota and its possible role for peatland succession. The total core microbiome of Amblystegiaceae and Sphagnum ecosystem samples was small compared to the total number of OTUs identified (49 vs. 13,799). However, the Sphagnum core microbiome in our study (142 OTUs) spanning Subarctic and temperate regions was similar in size to the alpine Sphagnum bog core microbiome (260 OTUs) reported elsewhere 30 . The sum of relative abundances of the OTUs in the total core microbiome was very high. Thus, this core community is a collection of few but highly abundant and thus, presumably, important members of the moss-associated microbial communities. On the other hand, the large numbers of low abundant OTUs that are not part of the core microbiomes identified in this study are possibly a result of local assemblies of microorganisms from the immediate surroundings. Notably, a major part of the total core microbiome were epiphytes on Scientific Reports | (2020) 10:22412 | https://doi.org/10.1038/s41598-020-79773-2 www.nature.com/scientificreports/ Amblystegiaceae, while dominant endophytes in Sphagnum. It needs to be emphasized that the patterns are dominated by the moss samples which reflects the sample design at least to some extent since we mainly collected mosses. Nevertheless, those patterns appear robust since the reference samples associate well with the moss samples (Fig. 4). Considering that these observed patterns of epi-and endophytes are consistent across large distances, many subsites, moss species and conditions, we hypothesize that Sphagnum recruited parts of the Amblystegiaceae microbiome during establishment in Amblystegiaceae peatlands, after which parts of the recruited microbiome adapted and became dominant endophytic members of the Sphagnum core microbiome that can be vertically transferred to the next generation as shown previously 31 . This way, over time, a Sphagnum core microbiome that originated in part from Amblystegiaceae, may have established. Alternatively, recruitment of microorganisms by Sphagnum is independent from peatland succession. If so, the presence of abundant OTUs that are dominant endophytes of Sphagnum and epiphytes of Amblystegiaceae is a coincidence and these bacteria dominate both systems because they are able to survive in both types of environments. However, considering that Amblystegiaceae and Sphagnum co-exist during peatland succession, a frequently occurring event through history 9,11,16 , it is possible that parts of the shared microbiome have transferred during times of co-existence. However, as the existence of an abundant core microbiome was first identified during the analysis of these data, the question of its role during succession is outside the scope of the current study.

Conclusion
We have uncovered the identity of moss-associated microorganisms in both Sphagnum and Amblystegiaceae peatlands and show that the dominant part of the moss microbiota in northern peatlands is influenced by the plant host. Amblystegiaceae and Sphagnum mosses share a small but highly abundant bacterial core microbiome with members that are dominant endophytes in Sphagnum and epiphytes in Amblystegiaceae. Future work is needed to test whether this core microbiome is a result of transfers of epiphytic bacteria from Amblystegiaceae to Sphagnum during natural peatland succession from fens to bogs.

Data availability
Demultiplexed read sequence data has been deposited at NCBI/Genbank database under the BioProject PRJNA356121 with accession numbers SRR6442387-SRR6442509 for bacteria and SRR6442615-SRR6442637 for archaea.