Combining whole genome shotgun sequencing and rDNA amplicon analyses to improve detection of microbe-microbe interaction networks in plant leaves

Abstract

Microorganisms from all domains of life establish associations with plants. Although some harm the plant, others antagonize pathogens or prime the plant immune system, acquire nutrients, tune plant hormone levels, or perform additional services. Most culture-independent plant microbiome research has focused on amplicon sequencing of 16S rDNA and/or the internal transcribed spacer (ITS) of rDNA loci, but the decreasing cost of high-throughput sequencing has made shotgun metagenome sequencing increasingly accessible. Here, we describe shotgun sequencing of 275 wild Arabidopsis thaliana leaf microbiomes from southwest Germany, with additional bacterial 16S rDNA and eukaryotic ITS1 amplicon data from 176 of these samples.The shotgun data were dominated by bacterial sequences, with eukaryotes contributing only a minority of reads. For shotgun and amplicon data, microbial membership showed weak associations with both site of origin and plant genotype, both of which were highly confounded in this dataset. There was large variation among microbiomes, with one extreme comprising samples of low complexity and a high load of microorganisms typical of infected plants, and the other extreme being samples of high complexity and a low microbial load. We use the 1 Regalado et al.Combining sequencing methods in plant metagenomics metagenome data, which captures the ratio of bacterial to plant DNA in leaves of wild plants, to scale the 16S rDNA amplicon data such that they reflect absolute bacterial abundance. We show that this cost-effective hybrid strategy overcomes compositionality problems in amplicon data and leads to fundamentally different conclusions about microbiome community assembly.

The relationship between microbial population size and disease in the Arabidopsis thaliana phyllosphere

“…Lastly, there is a possibility that unknown biases in our extraction and sequencing methods could potentially skew the relative abundance of microbiota. While this is true of every extraction and sequencing method to date, metagenomic results are concordant with those of the 16S rDNA data 17 , and the prominence of Pseudomonadaceae, Peronosporaceae, Sphingomonadaceae, and Albuginaceae in wild phyllospheres has been observed repeatedly across experiments and methodologies 2,3,44,46 .…”

“…When we regressed the abundance of microbial family on the presence of Pseudomonadaceae, the presence of Pseudomonadaceae positively and significantly affected four of the ten most abundant families (p < 0.005). The family Sphingomonadaceae was positively correlated with the abundance of Pseudomonadaceae, a result that contrasts with previously observed antagonistic relationships between some strains in these families 1 , but replicates in an expanded dataset the findings of Regalado and colleagues 17 .…”

“…We mapped the shotgun metagenomic data to the A. thaliana Col-0 reference genome 19 , and classified unmapped reads using the kmer metagenomic mapping tool centrifuge 20 . As shown before, the coverage estimates are robust and largely insensitive to the A. thaliana reference genome used 17 . We then estimated the ratio of average coverage over microbial genomes to plant nuclear genomes.…”

“…We previously demonstrated that one can obtain quantitative measures of microbial colonization in wild plants by performing shotgun metagenomic sequencing of whole plant tissue, followed by determining the ratio of reads assigned as having either a microbial or plant origin 2 . We have further validated our approach by showing that compositional inferences agree well between this method and more conventional amplicon sequencing methods 17 . In the current study, to maximize reproducibility across samples, all samples were processed in the same manner.…”

“…In order to mask host derived sequences, reads were mapped against the A. thaliana TAIR10 Col-0 reference genome 64 with bwa mem 19 using standard parameters. We have shown that metagenomic load can be reliably assessed with as few as 30,000 reads 17 . After removing samples that had fewer than 30,000 reads, all read pairs flagged as not mapping to A. thaliana were extracted with samtools 65 as the putatively "metagenomic" fraction.…”