2 MATERIAL AND METHODS

2.1 Sampling and DNA extraction

Microhabitat samples were collected in October 2017 in cooperation with the Canopy Crane Facility in the floodplain forest in Leipzig, Germany (51.3657 N, 12.3094 E). We sampled three different specimens of three autochthonous tree species: The small-leaved lime (Tilia cordata ), the European ash (Fraxinus excelsior ) and the pedunculate oak (Quercus robur ). The samples can be classified into two strata: (i) canopy samples and (ii) ground samples. Canopy samples were taken at 20 to 30 m height with replicates at all four cardinal directions of each tree. At each of the four replicate sites on every tree, following seven microbial microhabitat compartments related to tree surface were sampled: Fresh leaves, dead wood, bark, arboreal soil and three cryptogam epiphytes (lichen and two moss genera,Hypnum and Orthotrichum ). In addition, two ground samples (soil and leaf litter with four replicates per tree) at 2 m distance from each trunk were sampled. The soils were collected at the surface layer (~10 cm depth after removal of leaf litter and stones) throughout each station. All 324 samples were stored at -22°C until further processing. For DNA extraction, all canopy and litter samples were decorticated and/or chopped with a sterile razor blade and cut into small, regular pieces. DNA extraction was done according to the manufacturer’s instruction with the DNeasy PowerSoil kit (QIAGEN, Hilden, Germany). DNA concentration and quality were checked using a NanoDrop Spectrophotometer (NanoDrop Technologies, Wilmington, USA). For following PCR amplification, all four replicates of each microhabitat per tree were pooled.

2.2 PCR amplification, barcoding and sequencing

PCRs with taxon specific primers were conducted in two steps. The hypervariable V4 region of the 18S ribosomal RNA gene (SSU rDNA) was used for cercozoan community profiling with specific primers (Fiore-Donno, Richter-Heitmann, & Bonkowski, 2020). For the first PCR the forward primers S616F_Cerco and S616F_Phyt were mixed in the proportions of 50% and 50%, and used with the reverse primer S963R_Phyt. For a following semi‐nested PCR a mixture of the reverse primers S947_Phyt and S947_Vamp in an equal proportion has been used. The thermal program consisted of an initial denaturation step at 95°C for 2 min, 24 cycles at 95°C for 30 s, 52°C for 30 s, 72°C for 30 s; and a final elongation step at 72°C for 5 min. For amplifying the ITS 1 of the oomycete communities we used the specific primer pair ITS_177F and 58SR_Oom (Fiore-Donno & Bonkowski, 2020). Amplicons of the first PCR were again used as template for a semi-nested PCR with the primer pair I1786F_Stra and 58SR_Oom. The thermal program started with a denaturation step at 95°C for 2 min, followed by 24 cycles at 95°C for 30 s, 58°C for 30 s, 72°C for 30 s; and a final extension step at 72°C for 5 min.
We used 1 μl of DNA template for the first PCR amplification and 1 μl of the obtained amplicons as a template for a second semi‐nested PCR which was conducted with tagged primers. Tags were designed as described in Fiore-Donno, Richter-Heitmann, & Bonkowski (2020). The used primers and tag combinations are provided in Supplementary Tables 3 and 4.
We applied the following final concentrations: DreamTaq polymerase (Thermo Fisher Scientific, Dreieich, Germany) 0.01 units, Thermo Scientific DreamTaq Green Buffer, dNTPs 0.2 mM and primers 1 µM. To reduce the artificial dominance of few amplicons by PCR competition, all PCRs were carried out twice. PCR products were pooled, then purified and normalized using SequalPrep Normalization Plate Kit (Invitrogen GmbH, Karlsruhe, Germany). Sequencing was performed with a MiSeq v2 Reagent kit of 500 cycles for the shorter ITS amplicons (c. 250 bp) of Oomycota and a MiSeq v3 Reagent kit of 600 cycles for the amplified V4 Region fragments (c. 350 bp) of Cercozoa. Sequencing was conducted by a MiSeq Desktop Sequencer (Illumina Inc., San Diego, CA, USA) at the Cologne Center for Genomics (Germany).

2.3 Sequence processing

All bioinformatic and statistical methods were applied to both Oomycota and Cercozoa datasets independently if not stated otherwise. Raw reads were merged using vsearch v2.10.3 (Rognes et al. , 2016) at default settings. Merged contigs were demultiplexed with cutadapt v1.18 (Martin, 2011) allowing no mismatches in neither primer nor tag sequence. Cutadapt was also used to trim primer and tag sequences after demultiplexing. Sequences were then clustered into operational taxonomic units (OTUs) using swarm v2.2.2 (Mahé et al. , 2015) with d = 1 and fastidious option on. Chimeras were de novo detected using vsearch. OTUs were removed from the final OTU table if they were flagged as chimeric, showed a quality value of less than 0.0002, were shorter than 150bp (Oomycota) or 300bp (Cercozoa), or were represented by less than 0.005% of all reads (i.e. 141 reads for Oomycota or 269 reads for Cercozoa).
For taxonomic assignment, OTUs were first tentatively assigned by using BLAST+ v2.9.0 (Camacho et al. , 2009) with default parameters against the non-redundant NCBI Nucleotide database (as of June 2019). OTUs were removed if the best hit in terms of bitscore was a non-oomycete or non-cercozoan sequence, respectively. For a finer taxonomic assignment, two databases were used: The PR2 database (v4.12.0, Guillou et al. , 2012) served as a taxonomic reference set for cercozoan V4 sequences, while for the Oomycota all available oomycete sequences were downloaded from NCBI Nucleotide (as of July 2019). Both databases were used as a template for an in-silicoPCR with cutadapt, with the same primer sequences used in this study. The resulting virtual amplicons served as a database with the same length and genetic origin as our sequenced amplicons, which offers the advantage of penalising terminal gaps during the taxonomic annotation - which was performed with vsearch. The annotation was refined by assigning the species name of the best vsearch hit to the corresponding OTU if the pairwise identity was over 95%. OTUs with lower percentages were assigned higher taxonomic levels.
To account for random effects due to low sequencing depth, the final OTU table was loaded into Qiime2 v2018.11 (Bolyen et al. , 2019) to explore the sequencing depth by sample metadata. The minimum sequencing depth was determined depending on how many samples per metadata would be excluded. It was set as high as possible while retaining at least five samples per microhabitat and 15 samples per tree species.

2.4 Statistical analyses

All statistical analyses were conducted in R v3.5.3 (R Core Team, 2019). Rarefaction curves were carried out with the iNEXT package (Chaoet al. , 2014; Hsieh, Ma, & Chao, 2019) to determine if a higher sequencing depth would have revealed more OTUs. Alpha diversity indices were calculated for each sample using the diversity function in the vegan package (Oksanen et al. , 2019). Both former methods were applied on the OTU table with absolute abundances. To explore differences in the community composition across the samples, the following beta diversity-based methods were conducted on relative abundances. Non-metric multidimensional scaling was performed on the Bray-Curtis dissimilarity matrix of the log transformed table (functionsvegdist and metaMDS in the vegan package, respectively). The same method was used in a permutational multivariate analysis of variance (permANOVA, function adonis ), to test if oomycete and cercozoan OTU diversity differed across the strata, habitats and tree species. To analyse the effects of environmental factors to the variance of the community composition, a redundancy analysis was carried out on the Hellinger-transformed table (function rda in the vegan package). The function nestedtemp was used to test if the community of a microhabitat might be a subset of a larger one. Species accumulation curves were calculated using the specaccum function and the number of shared OTUs between different combinations of microhabitats was visualised using the UpSetR package (Lex et al. , 2014; Gehlenborg, 2019). All figures were plotted with the ggplot2 package (Wickham, 2016).