Metabarcoding of Bacteria Associated with the Acute Oak Decline Syndrome in England

Outbreaks of acute oak decline (AOD) have been documented in England from 2006. Both species of native oaks (Quercus robur and Quercus petraea) are affected. To complement isolation efforts for identification of putative causative biotic agents and increase our understanding of bacteria associated with oak tissue, five sites in England were chosen for this study. Samples of outer bark, inner bark, sapwood and heartwood were taken from healthy oak and trees with symptoms at varying stages of the syndrome. Furthermore, larval galleries attributed to infestation with Agrilus biguttatus were included. After DNA extraction and amplification of the V3–V5 fragment of the bacterial 16S rRNA genes by pyrosequencing, the dataset was analyzed to identify patterns in bacterial communities in oak tissue samples with and without AOD symptoms at each site. The composition of bacterial communities differed greatly according to the site from which the samples were obtained. Within each site, the composition of the bacteria associated with symptomatic tissue varied between advanced stages of the syndrome and healthy tissue. Key players in healthy and symptomatic tissue were identified and included members of the Gammaproteobacteria related to Pseudomonas sp. or Brenneria goodwinii and members of the Firmicutes.


Introduction
Acute oak decline (AOD) appears to be a complex, secondary disease of native oak (Quercus robur and Q. petraea), and is increasingly found on turkey oak (Q.cerris) in the UK.It is widespread across England and the Welsh borders, with northerly limits in the counties of Cheshire, Derbyshire, Lincolnshire and Nottinghamshire (around the 53 ˝N latitude) and westerly limit along the Welsh borders [1,2].Quantitative studies on the number of trees affected have not been carried out, but estimates from the sector suggest many thousands of oak trees are affected.
The decline is characterised by two distinctive symptoms (stem bleeds and internal bark decay) and two signs (exit holes on tree stems and larval galleries in the inner bark tissues).It is considered a distinctive condition within the wider oak decline complex [1].To date research indicates a complex cause of the lesions and bark decay in the inner bark tissues.Using destructive sampling and isolation studies an insect and bacterial component were revealed [3].As a large number of microbes are not tractable to culture, additional methods are required to identify as many putative necrogenic organisms Forests 2016, 7, 95 2 of 20 as possible.Here, we present data derived from a metabarcoding technique to gain insights into the community structure of affected and healthy oak trees from five sites in England.
Advances in high throughput sequencing are helping to unravel the composition and identity of the vast diversity of microorganisms in various habitats.Such habitats include eukaryotic hosts for whose microbiomes strong differentiation into microniches can be shown [4][5][6][7].
It is generally thought that the plant microbiome serves several functions for its host ranging from growth stimulation and providing a competitive advantage for resource acquisition to other mechanisms such as suppressing damaging microbes, and enhancing stress resistance, and nutrient transport ( [8] and references therein).The composition of the microbiomes of phyllospheres and rhizospheres have been widely studied for several plant species [9].In general, a particularly large bacterial diversity has been observed in the rhizosphere of a wide range of plants with 10 4 bacterial species detected [10][11][12].Interestingly, rhizosphere associated microorganisms were shown to influence the plant's immune response and inversely, plants were capable of influencing their rhizosphere microbiome ( [10] and references therein) implicating elaborate relationships.Studies analyzing factors shaping the phyllosphere microbiome were carried out in various scales, ranging from the landscape level (covering a range of different plant species) to the individual level (where the different plant parts of various genotypes of a single species were investigated).There is consensus that the phyllosphere microbiome is less taxa rich than the rhizosphere microbiome.The microbiome can vary between different plant species, for example trees.There were less than 100 taxa associated with Ginkgo trees [13]; but several hundred bacterial species detected on a number of tree species including Metrosiderospolymorpha [6], Tamarix trees [14], and various trees sampled in forests of Brazil [15] and Panama [16].Strong differences in the phyllosphere were attributed to plants' functional traits [16], although variation was observed in relation to season and shorter timescales ( [17] and references therein).These findings imply that the phyllosphere responds to species specific characteristics as well as plant genotype [18,19].Furthermore, physiological changes of the hosts as well as environmental factors, including the distance of the organ to soil [7], can influence the phyllosphere community structure ( [17] and references therein).Spatial structuring at the plant part level [6,7,13] suggests the adaptation of specific taxa to available microniches within a plant's phyllosphere, whereas on a larger scale, geographical distance [14] shapes phyllosphere communities as shown for Tamarix trees.
Many studies have focused on the ectosphere (outer surfaces, [8]), studying the epi-and endophytic microbiome of plant organs as in the case of the phyllosphere [8], partly due to its accessibility for sampling.The current study focused on the oak's stem endosphere and its associated microbiome.In general, endophytic microbes residing within plant tissues (the endosphere), whether in leaves, roots or stems, can establish beneficial, neutral or detrimental associations of varying intimacy with their host plants.Since woody tissue is such a specific niche, adaptation of microorganisms to the compounds present is required for growth and survival.For example the type and composition of carbohydrates [20], phenolic compounds [21], and other extractables [22] in woody stems can vary.Furthermore, studies have shown differences even between different trees in dry extract (dry wood), ellagitanins and ellagic acid, as well as whisky-lactone extracted from pedunculate and sessile oaks in France [22].Variation was also discovered in the aromatic compounds present in woody stems [23,24], for which geographical origin can be partly a determining factor ( [21] and references therein) as it varies in environmental factors like soil type, sun exposure and rainfall [21].
Despite the specific niche woody tissue provides, some organisms are adapted to utilize this environment as a food source.It can therefore be the place of microbial attack [25][26][27][28][29].Primary plant pathogens are usually studied using classic microbiology, including isolation of microbes from an environment on nutrient media to obtain cultures of the pathogens involved, which often misses important microbial diversity.New technologies like metabarcoding can complement traditional investigative methods and since the plant microbiome is a key determinant of plant health and productivity [10], it has received substantial attention in recent years [9,30].Previous studies focusing on identifying unknown bacterial taxa from oak trees suffering from AOD have used classical Forests 2016, 7, 95 3 of 20 microbiology.Various gram-negative bacteria of the family Enterobacteriaceae have been associated with AOD symptoms including Gibbsiella quercinecans [31], Lonsdalea quercina ssp.britannica [32], Brenneria goodwinii [33], Gibbsiella greigii [34], a range of species of the genus Rahnella [35], and Brenneria roseae ssp.roseae [36].This metabarcoding study aimed to supplement those efforts and helps to further understand the plant:microbial interactions taking place within oak trees affected or not affected by the AOD syndrome.Specifically, high-throughput sequencing technologies were used to identify thousands of sequences per sample and thus attempt to reveal the various taxa and their relative abundance interacting in healthy and diseased trees, including any rare or uncultivable microbial species.Thus, the 16S rRNA gene, the most comprehensive taxonomic marker available to date [37,38] was selected for metabarcoding.
To our knowledge this is the first study of the bacterial microbiome associated with oak bark and internal tissue.We hypothesized that we would find spatial structuring of the microbial community at different levels from landscape scale to within tree tissue types.Furthermore, we expected that the microbiomes would differ according to the health status of trees.

Sampling
Five sites, representative of the known distribution of AOD in England, were selected for this study based on the availability of representative material and the willingness of the landowners to donate and fell affected trees for research.The sites chosen were: Attingham Wood (AW, OS Eastings 356033, OS Northings 310372), Bisham Wood (BW, OS Eastings 485362, OS Northings 184100), Great Monks Wood (GMW, OS Eastings 582100, OS Northings 225300); Stratfield Brake (SB, OS Eastings 449400, OS Northings 211900) and Runs Wood (RW, OS Eastings 563207, OS Northings 310858) (Table 1).At BW and SB only two trees at each site were sampled, a healthy and a diseased tree with advanced symptoms.At GMW, an early stage symptomatic tree was included bringing the number of trees sampled there to three; and at the two remaining sites, AW and RW, an additional category of tree was included, those showing moderate symptoms, bringing the total to four trees sampled on both these sites.Trees were sampled as described by Denman et al. [1] and summarized below.The location of each tree was recorded using GPS (Garmin GPSmap 60CSx-Global Positioning Systems, UK) for mapping and recording purposes.Diameter at breast height (DBH, i.e., 1.3 m) was measured using a Forestry Suppliers metric diameter tape to get an indication of age/maturity of the tree.Severity of AOD attack was assessed by categorizing the number of bleeding points on stems where more than 20 stem bleeds per tree trunk scored advanced/severe; between 10 and 20 bleeds was moderate and less than 10 bleeds was considered a light (early) infection; healthy trees showed no stem bleeds at all.A sharp, surface disinfected chisel was used to remove a fully barked panel taken to a 5-7 cm depth in the heartwood.Panels were approximately 20 cm ˆ15 cm (LxB).The sampling equipment was thoroughly surface disinfected by dousing with industrial methylated spirits for 2 min and wiping dry with sterile paper towels, between samples.Each symptomatic panel was selected to include one or more bleed points, and both bark and wood tissues.On each of the symptomatic trees a non-symptomatic panel was cut some distance from the panel with symptoms.Healthy tree panels and non-symptomatic panels in a symptomatic tree did not have stems bleeds.Once cut, it was placed in a clean polythene bag, kept cool and taken to the laboratory for processing.Material to be processed was kept in a cold room at 4 ˝C overnight and was sorted, packaged and posted within 24 h of the field sampling.During the sorting process each panel was manually separated into outer bark, inner bark, sapwood and heartwood tissues by clamping the panel in a vice grip and prising the tissues apart using a disinfected chisel.Where galleries were present (i.e., in symptomatic tissues only), small blocks of tissue were chiseled out containing the galleries.Blocks of material were placed in clean plastic zip-lock bags, placed in a cooler with ice blocks and shipped overnight to the laboratory where they were processed as described below.

Bacterial Communities Associated with Oak Tissue
Genomic DNA was extracted from different tissue samples using a CTAB DNA extraction protocol as follows: From each sample, between 600 and 700 mg of material were scraped into a weigh boat using disposable scalpels.After weighing, the material was transferred into a clean bore mill canister (Kleco, Visalia, CA, USA) containing a 2.5 cm ball bearing.7.5 mL of prepared CTAB buffer (100 mM Tris-HCl, pH 8.0, 2% CTAB (Cetyltrimethylammoniumbromide), 20 mM EDTA, 1.4 M NaCl) including freshly added 1.0% sodium sulphite, 2.0% PVP-40 and 1% antifoam agent (A5757, Sigma, Gillingham, Dorset, UK) were then added to each canister which was subsequently sealed with a lid and o-ring.Canisters were loaded onto a ball mill (Kleco) and the samples were ground for 2 min.After grinding, 1 mL of extract was transferred with a pipette into a 2 mL tube for incubation at 65 ˝C for 10-15 min. 1 mL chloroform:isoamyl alcohol (24:1) were added and mixed by inverting followed by centrifugation at 11,000ˆg for 10 min.The aqueous (top) layer was transferred to a new tube.This process was repeated once.An equal volume of ice cold isopropanol and 0.5 volumes 5 M NaCl were added to the crude DNA extract, which was incubated at ´20 ˝C for 30 min.After centrifugation at 13,000ˆg for 10 min the supernatant was discarded and the pellet was washed by addition of 400 µL of 70% ethanol followed by centrifugation at 13,000ˆg for 4 min.The ethanol was discarded and the pellet dried for 5 min.The pellet was resuspended in 100 µL molecular grade water (Severn Biotech, Kidderminster, UK).Since DNA extracts from specific tissues contained large amounts of phenolic substances, DNA extracts were further purified by gel electrophoresis.For this, 20 to 25 µL of DNA extract were run on a 0.8% agarose gel and the band containing high molecular weight DNA was excised.It was then purified using the UltraClean ® GelSpin ® DNA Extraction Kit (MoBio, Carlsbad, CA, USA) following the manufacturer's instructions followed by storage at ´20 ˝C.
Bacterial community structure was assessed based on the V3-V5 region of the 16S rRNA gene [39].PCR primers flanking this region were adapted to 454 amplicon sequencing.For the forward primer a 25-mer Lib-L specific sequence adapter (underlined) was followed by a 454 amplicon sequencing specific 4-mer amplification key (italic) which was accompanied by a 10-mer barcode sequence to aid multiplexing (NNNN) and the template specific primer sequence (bold), specifically primer 341f (5 1 -CCATCTCATCCCTGCGTGTCTCCGACTCAGNNNNCCTACGGGNGGCWGCAG-3 1 ).For the reverse primer a 25-mer Lib-L specific sequence adapter (underlined) was followed by the target primer (bold) specifically 805r (5 1 -CCTATCCCCTGTGTGCCTTGGCAGTCTCAGGACTACHVGGGTATCT AATCC-3 1 ).An overview of the barcode sequences can be found in Table 1.Amplification reactions of 20 µL were carried out using the Phusion High-Fidelity DNA Polymerase (New England Biolabs, Hitchin, UK) containing 4 µL of HF buffer, 0.3 µM of forward and reverse primers respectively, 0.3 mM of dNTPs, 0.4 U Phusion DNA polymerase and 1 µL of template DNA.The final reaction volume was made up with nuclease-free water (Severn Biotech).Amplification was carried out on a C1000 thermal cycler (BioRad, Hercules, CA, USA) starting with an initial single denaturing step at 98 ˝C for 2 min, followed by 25 cycles of denaturation at 98 ˝C for 10 s, annealing at 55 ˝C for 30 s and extension at 72 ˝C for 30 s, followed by a final extension at 72 ˝C for 10 min.PCR products containing unique barcodes were pooled aiming at equimolar concentrations.Gel purification was used to remove small DNA fragments from pooled amplicons prior to emulsion PCR, which was carried out according to Roche's instructions.Sequencing of all amplicons was completed using the GS FLX+ System (Roche, Burgess Hill, UK).

Data Analysis
Bioinformatics analysis was carried out using QIIME v1.90 [40] based on denoised data processed via FlowClus [41].A minimum read length of 100 bp and a maximum read length of 500 bp were used.Quality filtering was applied removing reads with a Phred quality score (Q) less than 30 on average and those with less than Q25 across a 50 bp sliding window.A 1 bp mismatch in the primers and barcode sequences was allowed.Reads obtained per sample were highly variable ranging from 633 to 19,265 with an average of 5748 sequences.The QIIME protocol "de novo OTU (Operational Taxonomic Unit) picking" [42] was used for clustering of the sequences followed by taxonomy assignment using the greengenes database (reference version gg_13_8_otus, [43]).In detail, OTUs were picked using uclust [44] utilizing the first seed as representative sequence.Pynast was subsequently utilized to align sequences [45] followed by taxonomy assignment via uclust [44].After filtering of the alignment via pynast applying default values, a phylogenetic tree was constructed by fasttree [46].ChimeraSlayer [47] was applied for detection of chimeric sequences.No chimeric sequences were found.The procedure was carried out on the 4 June 2015.Removal of OTUs assigned to chloroplasts, chlorophyta, mitochondria, archaea and sequences unassigned at kingdom level were removed from the dataset as described previously [6].We observed a high variation in sequence number obtained for different samples as described previously [48].To allow comparison of disparate libraries, normalization of sequence number is achieved by random selection of sequences referred to as rarefaction [49][50][51][52][53][54][55].The effect of this process on beta diversity has been demonstrated [56][57][58].However, other studies could not find an effect [59] or showed the importance of rarefaction to reduce bias in pyrosequencing datasets [49].Therefore, libraries were rarefied to the same sequencing depth (1000 sequences).This resulted in loss of data for samples A5, A10, A22, A26, A31, R1, R2, R3, R4, R5, R6, R7, R8, R9, and R10.Singletons detected in the rarefied dataset were kept [59][60][61][62].This approach was chosen since rare taxa can play an important role in the environment [63][64][65][66], whilst their role in health or disease is unclear [67].Thus, it is important to study such taxa instead of dismissing them [68].
Differences in alpha (within sample) diversity were assessed by comparing the number of OTUs, chao1 richness estimator [69] and phylogenetic diversity (PD, [70]) obtained for each sample based on the rarefied dataset.Relative abundances of OTUs clustered at 97% sequences similarity were used to assess differences between bacterial communities.Multivariate analysis of square root transformed community data was performed using the subroutines nonmetric multidimensional scaling (nMDS) and analysis of similarities (ANOSIM, [71]) based on Bray-Curtis similarities of the PRIMER software package version 7 (Primer-E Ltd., Lutton, UK).Clustering was applied using complete linkage mode.Phylogenetic analysis was performed by MEGA 6 [72,73] using alignment by ClustalW followed by the Maximum Likelihood method based on the Jukes-Cantor model.

Nucleotide Sequence Accession Numbers
The 16S rRNA gene sequences derived from pyrosequencing have been deposited in the European Nucleotide Archive under study accession number PRJEB12457 (sample accession numbers ERS1037713-ERS1037800).

Results and Discussion
We studied the composition of bacterial communities associated with oak tissues from trees with and without disease symptoms at five sites in England.In total, 15 trees were sampled, five without disease symptoms (called healthy) and ten with disease symptoms.In addition, panels without symptoms were tested from diseased trees (non-symptomatic samples) (Table 1).In total 88 tissue samples were processed.We first present results of the overall community structure and relative composition at the family and genus level and then consider results according to tree health status.
It needs to be noted that species level identification is hindered in this dataset due to the limited taxonomic resolution of the 16S rRNA gene for specific groups like the Enterobacteriaceae, for which other markers may be more suitable [74,75].It is thus thought that classical pathogen detection could be complemented with metabarcoding approaches if relevant targets like the gyrase B gene [74] are chosen.However, the lack of sequence information for a wide range of taxa currently hinders the broad use of such markers in microbial ecology.Since this is the first molecular report on the composition of the oak microbiome at a range of sites we selected a broad taxonomic marker to gain an initial overview.Furthermore, sequencing errors introduced by pyrosequencing were eliminated by using a stringent quality control approach that included the latest algorithm for denoising of the raw data via FlowClus [41].
Species diversity (number of OTUs and Shannon diversity) and estimated richness (chao1) varied greatly with highest values found in samples from GMW followed by RW.Lowest diversity and PD (phylogenetic diversity) were observed in samples from BW (Table 2).Community structure varied most between samples from different locations as illustrated in Figure 2. GMW showed the strongest difference from other locations followed by samples from RW.However, some similarities in communities were observed for samples from BW, SB and AW.The overall separation was confirmed by ANOSIM (Global R = 0.816, p = 0.001).Pairwise comparisons indicated significant differences among sites apart from BW and SB, which were very similar in community composition.Both sites had considerably lower OTU numbers and the phylogenetic diversity was lower than in the other three sites (Table 2).The number of taxa detected at all sites was generally comparable to the phyllosphere microbiome consisting of several hundred or less taxa [6,[13][14][15][16].
By comparison, alpha diversity values were far greater at the sites GMW and RW for which the chao1 richness estimator indicated a taxon diversity with thousands of different taxa but still not reaching the richness observed in the rhizosphere, where tens of thousands of taxa have been commonly recorded [10][11][12]76].For these samples, we did not achieve the coverage required to describe the full diversity of these samples.Nevertheless, we are convinced that the present analysis describes differences for the most common taxa found in our samples as suggested previously [77].
The different oak tissue types (sapwood, heartwood, etc.) did not display a consistent separation across all sites as indicated by ANOSIM (Global R = ´0.041,p = 0.894).This implicates, that spatial structuring at the level of tissues did not override site specific differences.However, when the gallery associated bacteria from all sites were compared with all of those associated with heartwood, significant differences were found in pairwise comparison (R = 0.151, p = 0.037).By comparison, alpha diversity values were far greater at the sites GMW and RW for which the chao1 richness estimator indicated a taxon diversity with thousands of different taxa but still not reaching the richness observed in the rhizosphere, where tens of thousands of taxa have been commonly recorded [10][11][12]76].For these samples, we did not achieve the coverage required to describe the full diversity of these samples.Nevertheless, we are convinced that the present analysis describes differences for the most common taxa found in our samples as suggested previously [77].
The different oak tissue types (sapwood, heartwood, etc.) did not display a consistent separation across all sites as indicated by ANOSIM (Global R = −0.041,p = 0.894).This implicates, that spatial structuring at the level of tissues did not override site specific differences.However, when the gallery associated bacteria from all sites were compared with all of those associated with heartwood, significant differences were found in pairwise comparison (R = 0.151, p = 0.037).
To identify taxa representing the oak microbiome we used the core microbiome algorithm available via QIIME.We identified that a phylotype belonging to the genus Pseudomonas, Gammaproteobacteria (OTU 1542) was found in 95% of samples regardless of health status or tissue type.This phylotype was accompanied by a member of the genus Halomonas (Gammaproteobacteria) in up to 80% of samples (OTU 21506).
Similar to studies of other plant organs, Proteobacteria dominated the communities associated with oak tissue and included the presence of Bacteroidetes and Firmicutes [6,13,15,19,[78][79][80][81].Interestingly, Juncker et al. [6] identified a member of the Gammaproteobacteria (Enterobacteriaceae) as dominant in stamina and styles whereas a member of the Pseudomonadaceae was identified as dominant in the oak microbiome.Our finding is in line with findings from Ottesen et al. [7], who also detected dominance by members of the Pseudomonadaceae and Micrococcineae on tomato fruit samples.
Geographical distance appeared to shape the oak microbiome implicating strong spatial structuring and adaptation to the local environment which was also shown previously on Tamarix trees [14].On a smaller scale, spatial structuring was not apparent for the various oak tissues examined within the trees, although we assumed that they would differ considerably in the microniches provided to bacteria.However, we detected a significant difference between galleries and heartwood supporting the hypothesis for the two most discerned tissues.To identify taxa representing the oak microbiome we used the core microbiome algorithm available via QIIME.We identified that a phylotype belonging to the genus Pseudomonas, Gammaproteobacteria (OTU 1542) was found in 95% of samples regardless of health status or tissue type.This phylotype was accompanied by a member of the genus Halomonas (Gammaproteobacteria) in up to 80% of samples (OTU 21506).
Similar to studies of other plant organs, Proteobacteria dominated the communities associated with oak tissue and included the presence of Bacteroidetes and Firmicutes [6,13,15,19,[78][79][80][81].Interestingly, Juncker et al. [6] identified a member of the Gammaproteobacteria (Enterobacteriaceae) as dominant in stamina and styles whereas a member of the Pseudomonadaceae was identified as dominant in the oak microbiome.Our finding is in line with findings from Ottesen et al. [7], who also detected dominance by members of the Pseudomonadaceae and Micrococcineae on tomato fruit samples.
Geographical distance appeared to shape the oak microbiome implicating strong spatial structuring and adaptation to the local environment which was also shown previously on Tamarix trees [14].On a smaller scale, spatial structuring was not apparent for the various oak tissues examined within the trees, although we assumed that they would differ considerably in the microniches provided to bacteria.However, we detected a significant difference between galleries and heartwood supporting the hypothesis for the two most discerned tissues.

Bacterial Community Structure in Relation to the Tree's Health Status
The link between the health status of a tree or specific tissue and its microbiome was tested using ANOSIM.There was no significant difference between the total microbial communities from healthy and diseased trees (Global R = 0.106, p = 0.090).However, we could detect a relationship between the oaks' microbiomes and the health status of the tissue (Global R = 0.091, p = 0.001), which was strengthened when the trees' health status was also taken into account (Global R = 0.151, p = 0.001).For comparison of samples with varying sequence depth we applied random sampling of sequences, which can cause overestimation of β-diversity [57].Thus, we limited our analyses to major shifts and differences detected between samples.
Healthy Trees: When identifying the subset of OTUs indicative of a healthy oak microbiome, the phylotypes 1542 and 21506 as well as a member of the genus Shewanella (Gammaproteobacteria, OTU 12625) were identified for 80% of healthy samples.Two members of the Proteobacteria, namely OTUs 15035 and 8364, (Achromobacter (Betaproteobacteria) and Stenotrophomonas (Gammaproteobacteria) respectively) were detected in seven of nine healthy samples.
Diseased trees: Advanced symptom tissues only: The oak microbiome of trees displaying advanced stages of the AOD syndrome was indicated by the phylotype 5035, a member of the family Enterobacteriaceae, Gammaproteobacteria, in 80% of samples.Phylogenetic analysis indicated a close relationship to Brenneria goodwinii previously isolated from AOD tissue [33] supporting its important role in the syndrome.Although it was also detected in healthy tissue, its levels were increased drastically, where symptoms were observed.Additionally, other enteric taxa detected in this study showed a close relationship to plant pathogenic bacteria like Erwinia [82] implying a role in the acute oak decline syndrome.Additionally, members of the Tenericutes [83,84] were detected at one site, but these occurred also in healthy tissue contradicting a unique contribution to the symptoms seen in symptomatic tissue.
Since OTU 1542 was also present in 80% of samples from symptomatic tissue, it is assumed that this phylotype is generally associated with oak tissue regardless of health status.Pseudomonas species were described as plant endophytes before ( [85] and references therein) with P. stutzeri being described as denitrifying organism [86] with antifungal activity [87] suggesting a beneficial role in its interaction with the plant.In contrast, Firmicutes like relatives of Robinsoniella peoriensis are involved in the degradation of cellobiose [88], a metabolite in cellulose degradation indicating a saprophytic lifestyle.
Due to the high site-specific variability observed, we analyzed the datasets per site with regards to the health status of the tissues (Figure 3).We observed a strong link between microbiomes and tissue health supported by ANOSIM at the sites AW (Global R = 0.299, p = 0.001), RW (Global R = 0.395, p = 0.001) and SB (Global R = 1, p = 0.029).In contrast, associated bacteria at the sites BW and GMW were not significantly correlated with the health status of the tissue.Hence, metabarcoding was unable to detect a strong signature for an AOD associated microbiome.
In summary, the overall health status of the trees had only a very small effect on the general oak microbiome, which indicates that there was no common shift towards a disease indicating community across the sites studied.This implies, that the whole bacterial community does not shift drastically towards a disease indicator microbiome independent of site.Instead, rather small global effects are observed by AOD in symptomatic tissue, presumably masked by location specific signatures.This is supported by the finding, that specific sites did display distinct shifts in oak associated communities for symptomatic tissues, especially in advanced stages when signatures from other sites were removed.This finding implies that the observed shifts are dependent on geographical conditions highlighting that spatial structuring occurs regardless of the studied AOD syndrome being apparent.In future studies, tree age or specific composition of the plant's metabolites and phenolic compounds could be included to characterize the tissues analyzed.Furthermore, the height of the samples' origin will be of interest as highlighted previously [7].Such information would describe the microniche, the bacteria reside in, allowing a correlation between such characteristics and community structure.The latter requires the development of taxonomic targets that are suitable for high throughput sequencing and complement 16S rRNA gene sequencing by providing higher taxonomic resolution for families like the Enterobactericeae.In addition, the influence of other microorganisms present should be studied including fungi, microeukaryotes and viruses.For the identification of novel putative disease causing organisms site independent characteristics like metabolites would also be beneficial, to allow the identification of bacteria solely changing in relation to the health status of the trees studied.Furthermore, an assessment of tree associated microbiomes covering the present variety of tree species in a given location should be taken into account.Although we could not detect a distinct shift in the microbiome due to AOD, we could confirm the recurring presence of previously isolated Enterobacteriaceae, namely Brenneria goodwinii in symptomatic tissue.
identification of novel putative disease causing organisms site independent characteristics like metabolites would also be beneficial, to allow the identification of bacteria solely changing in relation to the health status of the trees studied.Furthermore, an assessment of tree associated microbiomes covering the present variety of tree species in a given location should be taken into account.Although we could not detect a distinct shift in the microbiome due to AOD, we could confirm the recurring presence of previously isolated Enterobacteriaceae, namely Brenneria goodwinii in symptomatic tissue.

Conclusions
We demonstrated that strong site-specific signatures exist for oak associated bacteria studied in various tissues.The number of phylotypes ranged from several hundreds to below hundred.Apart from this variability we could identify two taxa mainly associated with tissue from healthy trees belonging to the Gammaproteobacteria.Due to the high variability between sites we could not detect a distinct shift towards an acute oak decline syndrome associated microbiome.However, one phylotype (OTU 5035) belonging to the Enterobacteriaceae was consistently found at elevated levels in symptomatic tissues at all sites suggesting a role of this bacterium in the AOD syndrome.

Conclusions
We demonstrated that strong site-specific signatures exist for oak associated bacteria studied in various tissues.The number of phylotypes ranged from several hundreds to below hundred.Apart from this variability we could identify two taxa mainly associated with tissue from healthy trees belonging to the Gammaproteobacteria.Due to the high variability between sites we could not detect a distinct shift towards an acute oak decline syndrome associated microbiome.However, one phylotype (OTU 5035) belonging to the Enterobacteriaceae was consistently found at elevated levels in symptomatic tissues at all sites suggesting a role of this bacterium in the AOD syndrome.Furthermore, members of the Firmicutes were abundant at most sites highlighting their importance in oak-prokaryote interactions.This study offers a so far unknown description of oak associated microbiomes using a metabarcoding approach on a wide range of sites.

ForestsFigure 1 .
Figure 1.Phylogenetic tree of the 20 most abundant oak associated bacteria, irrespective of health status of the tissue.Bootstrap values above 50% are displayed.Sites where the OTUs were detected are indicated in brackets (Attingham Wood (AW), Runs Wood (RW), Great Monks Wood (GMW), Stratfield Brake (SB), Bisham (BW)).

Figure 1 .
Figure 1.Phylogenetic tree of the 20 most abundant oak associated bacteria, irrespective of health status of the tissue.Bootstrap values above 50% are displayed.Sites where the OTUs were detected are indicated in brackets (Attingham Wood (AW), Runs Wood (RW), Great Monks Wood (GMW), Stratfield Brake (SB), Bisham (BW)).

Figure 3 .
Figure3.NMDS plot based on Bray-Curtis similarities of relative bacterial abundances after square root transformation in relation to the health status of the tissue sampled including all samples.Tissue health status was categorized as non-symptomatic tissues from trees with AOD (), symptomatic tissues from trees with AOD () and healthy tissue ().Displayed are the communities from the following sites: Attingham Wood (A); Runs Wood (B); Great Monks Wood (C); Stratfield Brake (D); Bisham Wood (E).

DFigure 3 .
Figure3.NMDS plot based on Bray-Curtis similarities of relative bacterial abundances after square root transformation in relation to the health status of the tissue sampled including all samples.Tissue health status was categorized as non-symptomatic tissues from trees with AOD (), symptomatic tissues from trees with AOD ( ) and healthy tissue ( ).Displayed are the communities from the following sites: Attingham Wood (A); Runs Wood (B); Great Monks Wood (C); Stratfield Brake (D); Bisham Wood (E).

Table 1 .
Bacteria 16S rRNA gene sequences after denoising of pyrosequencing data.

Table 2 .
Overview of alpha diversity of bacterial communities.
* Tree displayed advanced stages of the syndrome; † Phylogenetic Diversity.