The type of mycorrhizal associations (i.e. ecto- or arbuscular mycorrhizal) formed by trees is of fundamental importance for a range of soil properties and processes in forest ecosystems, yet their importance for the distribution of other important soil biota such as bacteria is still largely unknown. This is especially so in diverse tropical forests where trees of different mycorrhizal types are intermingled in a highly heterogeneous biotic and abiotic environment. Here we used an experimental common garden that helped us to assess how abiotic and biotic variation differentially influenced bacterial communities associated with trees planted in a secondary tropical forest of Borneo. We used high-throughput amplicon sequencing to characterize rhizosphere bacterial communities of 13 climax tree species (8 ectomycorrhizal and 5 arbuscular mycorrhizal) in relation to plant traits, plant neighborhood, and abiotic environment. Rhizosphere bacterial (RB) communities differed significantly between EM and AM trees but not among EM species and only marginally among AM species. Foliar nutrients, especially potassium, showed relationships with RB community composition. Rhizosphere bacterial communities were related to the density and size of neighboring ectomycorrhizal but not arbuscular mycorrhizal trees. Diversity of RB on AM trees responded positively to AM neighbors and negatively to EM neighbors but RB diversity associated with EM trees was unaffected by neighborhood. Rhizosphere bacterial communities of AM trees were more responsive to environmental variation such as light availability and position on a slope. Plant-growth-promoting taxa of RB assorted similarly to total RB but more strongly.

Synthesis: Our results suggest that the distribution of rhizosphere bacterial communities is linked to plant functional group and plant neighborhood. Because rhizosphere bacteria play important roles in nutrient cycling that influence plant species composition, it is likely that their distributional patterns are important for understanding ecosystem processes and plant demographics.

Methods for generation of microbial abundance table

Rhizosphere soil sample collection: In October through November 2016 when trees were eight years old, we collected fine roots from trees of 13 tree species belonging to six families (4-5 trees per species; Table 1). We selected 13 species from the total pool of 34 species focusing on an even sampling of trees that form EM and AM and also focusing on tree species with high enough survival to allow replicate sampling. Unfortunately, a high proportion of AM trees planted initially in the common garden did not survive resulting in the inclusion of only 5 AM species. Individuals were selected randomly from the 20 replicates per species. To characterize the RB community of each tree, rhizosphere soil was collected by following lateral roots out from the trunk until the fine roots were reached, ensuring that samples belonged to the focal tree. Fine roots were collected opportunistically and were found anywhere from several centimeters to 10+ meters from the trunk. Trees were sampled at three different directions from the tree trunk, with a minimum collection of 15 cm of fine roots per direction. Because soils were wet and adhered to roots when removed from the ground, we were able to exclusively sample soil located within 1 cm of fine roots (“rhizosphere soil”). All soil that was further than 1 cm from a fine root was discarded. Bacterial communities were sampled only from rhizosphere soil to exclude endophytes. Soil samples were kept frozen before being transported on ice to a -60 °C freezer. We homogenized rhizosphere soil by stirring with a rod and subsampled 25-30 g.

Molecular analysis: We extracted genomic DNA from rhizosphere soil using a DNeasy PowerSoil kit (Qiagen Biotechnology). The concentration and purity of the DNA were determined with a NanoDrop 2000 Spectrophotometer (GE Healthcare) and standardized to 25 ng/μL (soil). We produced rDNA PCR amplicons as follows: we amplified for the V4-V5 region of the 16S rRNA gene using the primers 515FB (Parada et al., 2016) and 926R (Quince et al., 2011). Reaction and cycling conditions for PCR are found in the Supplemental Material. Library preparation was carried out with a Nextera DNA library prep kit (Illumina, Inc.). Sequencing was carried out at Biotechnology Research Institute, Universiti Malaysia Sabah on a MiSeq Desktop Sequencer (Illumina Inc.) running in paired end 2×300 mode. All molecular analyses, including sample preparation, PCR, library preparation, and sequencing were performed at Biotechnology Research Institute, Universiti Malaysia Sabah. We homogenized rhizosphere soil, subsampled 25-30 g and extracted gDNA using a DNeasy PowerSoil kit (Qiagen Biotechnology). The concentration and purity of the DNA were determined with a NanoDrop 2000 Spectrophotometer (GE Healthcare) and standardized 25 ng/μL. Details of the extraction protocol are provided in the Supplemental Material. We produced rDNA PCR amplicons as follows: We amplified the taxonomically informative V4-V5 region of the 16S rRNA gene using the primers 515FB (Parada et al., 2016) and 926R (Quince et al., 2011). Samples were amplified on a C1000 Touch Thermal Cycler (Bio-Rad) with 40 PCR cycles using the GoTaq Flexi PCR Kit (Promega Corp.) with recommended protocol to prepare amplicon pools for sequencing (details of PCR reactions and cycling conditions in supplementary materials). The locus-specific PCR products were purified using KAPA Pure Beads kit (Kapa Biosystems) with recommended protocol and standardized to a concentration of 10 ng/mL using the elution buffer. Library preparation was carried out using a Nextera DNA Sample Preparation Index Kit (96 Indices, 384 samples; Illumina, Inc.) with the following alterations to the recommended protocol. Indexing PCR reactions were set up using the KAPA HiFi HotStart ReadyMix PCR Kit (Kapa Biosystems) with 5 μL of the purified PCR products as template and 5 μL each of index primers 1 and 2 (i7 & i5, respectively) and amplified for 12 cycles (details in supplementary materials). The indexed PCR product was purified as above. The concentration of the indexed PCR product was measured spectrophotometrically and standardized to a concentration of 10 ng/mL using the elution buffer as diluent. To allow inclusion of more samples, pools of 16S amplicons were indexed together with 18S and ITS samples for each sample (results for 18S and ITS analysis not presented here, details of pooling protocol in supplementary material). Sequencing was carried out using MiSeq Reagent Kit v3 (600 cycle) on a MiSeq benchtop sequencer (Illumina) running in paired end 2 x 300 mode. The loading concentration of the pooled DNA library was 7 pM and spiked with 30% PhiX.

Bioinformatics: Loci of interest (16S) were separated from the other pooled loci (18S and ITS) prior to demuliplexing using fqgrep  (Das, 2011). Contaminating PhiX sequence was removed using the akutils phix_filtering command in akutils v1.2 (Andrews, 2018). Primers were trimmed from the amplicons with the strip_primers command. 16S paired-end reads were merged using the join_paired_reads command. Merged 16S and forward read 18S sequences were assessed for quality with FastQC (Andrews, 2010). Low-quality 3’ base calls were trimmed using FastX Toolkit (http://hannonlab.cshl.edu/fastx_toolkit/). OTU picking and taxonomy assignments were performed using the pick_otus workflow command to process samples through QIIME 1.9.1 (Caporaso et al., 2010). Samples were demultiplexed and quality-filtered to retain sequences with a minimum quality of q20, allowing no ambiguous base calls and discarding any reads less than 95% of the original length after filtering. Chimeras were filtered from sequences with VSEARCH 1.1.1 (Rognes et al., 2016). OTU-picking was performed using SWARM (Mahé et al., 2014), allowing a distance of 3 (approximately 99% similarity). Taxonomy assignment was done with BLAST (Altschul et al., 1990) in QIIME (max e-value 1 x 10-20, assign best match) against SILVA version 132 database (Quast et al., 2013). Reference databases were filtered to include only the regions of our amplicons (V4-V5), avoiding false assignments. Rare OTUs comprising less than 0.005% of the tables were filtered out (Bokulich et al., 2013).

Tree neighborhood data

We established 1.5 m radial plots around all focal trees and identified and measured the diameter at breast height (DBH, 1.3 m above soil surface) of all trees with DBH of 10 cm or greater. We also established 3.0 m radial plots and identified all tree species with DBH of 20 cm or greater.

Plant traits & abiotic variables

Traits and abiotic data were taken from the dataset used for Gustafsson et al., (2016).

*See manuscript for references

Microsoft Office or Open Office.