Assembly processes of bacterial endophytic and epiphytic communities in the canopy of neotropical trees

Bacterial phyllosphere communities play a key role in the fitness of plants. Understanding the factors and mechanisms shaping them is necessary to predict how trees would respond to a variety of stresses or disturbances. In tropical forests, the steep vertical gradients in light, rainfall or temperature within the canopy are likely to drive the assembly of bacterial communities in leaves.We explored the assembly processes driving epiphytic and endophytic bacterial communities within the canopy of three Neotropical tree species in French Guiana. We used 16S high-throughput sequencing to characterize leaf bacterial communities and measured foliar traits of host trees to describe variations within the canopy. We found that selection by foliar traits was more important in endophytic communities with a particular effect of traits related to space and available nutrients. Epiphytic communities were also subjected to selection by foliar traits but were more prone to dispersal limitation on top of the canopy. This study provides evidence that the assembly bacterial communities in leaves of tropical trees is jointly driven by determinism and stochasticity. Moreover, the clear differences found between endophytic and epiphytic communities emphasize the major role of the plant host, and provide new perspectives to further investigating the phyllosphere-bacteria interactions. Leaf sampling has been done on tree individuals belonging to three neotropical species : <em>Eperua</em> <em>falcata</em> Aubl. (Caesalpiniaceae), <em>Macrolobium</em> <em>bifolium</em> Aubl. (Caesalpiniaceae) and <em>Tetrasgastris</em> Gaertn. sp (Burseraceae). Five individuals of each tree species were selected. On each individual, we sampled leaves at 3 heights: on top of the crown (Top Canopy, TC), in the middle of the crown (Middle Canopy, MC) and on the lower branch of the crown (Bottom Canopy, BC). Three leaves were collected at each height for DNA analysis resulting in 135 leaf samples (3 tree species x 5 individuals x 3 heights x 3 leaves). In addition, we collected 2 leaves for each height and each individual, that were further used for leaf traits measurements. We collect epiphytic and endophytic DNA on each leaf, resulting in 270 DNA samples (3 tree species x 5 individuals x 3 heights x 3 leaves x 2 DNA types (epiphytic or endophytic)). To collect epiphytic DNA, the whole upper and lower surfaces of each leaf were carefully wiped with a piece (2x2 cm) of Whatman paper sterilized by autoclaving (120°C, 20 min) and soaked in sterile CTAB buffer (2% cetyl trimethylammonium bromide, 1% polyvinyl pyrrolidone, 100 mM Tris-HCl, 1.4 M NaCl, 20 mM EDTA), within 4 hours after leaves harvesting. The Whatman paper was stored in a 2ml Eppendorf tube filled with CTAB for downstream DNA extraction. Surface-sterilization protocol was performed as recommended by Compant <em>et al.</em>, (2021) to collect endophytic communities. Each leaf was surface-sterilized in 0.525% sodium hypochlorite (2 min) and 70% ethanol (2 min) and two segments of 2 cm² area from the lamina were cut and stored in a 2ml Eppendorf tube filled with CTAB for downstream DNA extractions close to a field Bunsen burner to create a sterilized environment. This resulted in 270 DNA samples (3 tree species x 5 individuals 3 heights x 3 leaves x 2 DNA types (epiphytic or endophytic)). DNA was extracted using a CTAB extraction method (Carrell &amp; Frank, 2014). We added 800 µl of CTAB solution to 0.6 g of tissue, incubated for 2 h at 60°C, and homogenized with glass beads for 3 min. Proteins were removed with the addition of an equal volume of chloroform, centrifuged for 10 min at 16 000 g, and placed the top aqueous phase in a sterile tube. Nucleic acids were precipitated with the addition of 1/10 volume of cold 3 M sodium acetate and 1/2 volume cold isopropanol, froze them at -20°C for 12 h, and centrifuged for 30 min at 16 000 g. The supernatant was discarded, 700 µl of 70% ethanol was added to the solution, and centrifuged for 10 min. The air-dried pellet was resuspended with 30 µl of DNA resuspension fluid (1.0 M Tris-HCL and 0.1 M EDTA) and stored it at -20°C. The V5-V6 region of the bacterial 16S rRNA gene was amplified using the chloroplast-excluding forward primer 799f (Chelius &amp; Triplett, 2001) and the reverse primer 1115R (Reysenbach &amp; Pace, 1995). Forward and reverse primers were tagged in 5’ with a combination of two different 8-nucleotide labels. The PCR amplification was done in 25 μl with 1x buffer, 0.22 mM dNTP each, 0.45 μM each tagged primer, 2.84 mM MgCl2, 0.11 mg/ml bovine serum albumin and 0.04 U/μl Taq polymerase (Solis Biodyne). The thermocycling conditions were as follows: 5 min at 94°C, 30 cycles of 30 sec at 94°C, 30 sec at 58°C and 40 sec at 78°C. The PCR reactions were done for each sample separately and amplicons were quantified with a fluorescence-based method (Qubit 3.0, Invitrogen Life Tech) and pooled in equimolar conditions. The library was built using the Fasteris MetaFast protocol (FASTERIS SA, Plan-les-Ouates, Switzerland) and sequenced on one run of a MiSeq Illumina platform (FASTERIS SA,) using the paired-end sequencing technology. To control for potential contaminants (Salter <em>et al.</em>, 2014) and false positive sequences caused by tag-switching events (Esling <em>et al.</em>, 2015) the sequenced multiplexes comprised extractions/PCR negative controls and unused tag combinations. A total of 3.181.610 sequencing reads were obtained and curated using the OBITools3 package (Boyer <em>et al.</em>, 2016) and the R software version 3.5.3 (R Core Team 2019) following the procedure described in (Zinger <em>et al.</em>, 2019). Paired-end reads were assembled and assigned to their respective samples. After dereplication, low quality sequences (i.e. containing Ns, being shorter than 80 pb or singletons) were excluded. PCR/sequencing errors were removed from the dataset and remaining reads were clustered into operational taxonomic units (OTUs) at 97% similarity using the Sumaclust algorithm (Kopylova <em>et al. </em>2016). The most abundant read was considered representative of the OTU and taxonomically assigned using the SILVAngs r138.1 (Quast <em>et al.</em>, 2013). Finally, we checked the taxonomic assignment and kept only sequences assigned to the Bacteria kingdom. Curation procedure yielded to 649,212 reads corresponding to 10152 bacterial OTUs in 203 samples. Finally, because the sequencing depths were uneven across samples (Figure S1), the sequencing depth of each sample was standardized by randomly resampling a number of reads equal to the first quartile of read number across samples (n=1864) which removed 51 samples and kept 152 samples for subsequent analyses. Data provided consist of metadata.txt ( table gathering the complete list of samples with the 10 leaf traits measured on each leaf, the species identity of the tree individual, its geolocalisation, its position within the canopy and the type of DNA (epiphytic or endophytic)), taxo.txt. (contains the taxonomy assigned to each OTUs using the SILVA database), R_script.txt (contains the main R scripts used for the statistical analysis), OTU_table.txt (row representing each OTU, columns the features of the OTU namely their id code, the number of read counts in the analysed dataset, the length of the sequence and the sequence itself).