The Leaf Microbiome of Arabidopsis Displays Reproducible Dynamics and Patterns throughout the Growing Season

Abstract

Leaves are primarily responsible for the plant's photosynthetic activity. Thus, changes in the leaf microbiota, which includes deleterious and beneficial microbes, can have far-reaching effects on plant fitness and productivity. Identifying the processes and microorganisms that drive these changes over a plant's lifetime is, therefore, crucial. In this study, we analyzed the temporal dynamics in the leaf microbiome of Arabidopsis thaliana, integrating changes in both composition and microbe-microbe interactions via the study of microbial networks. Field-grown Arabidopsis were used to monitor leaf bacterial, fungal and oomycete communities throughout the plant's natural growing season (extending from November to March) over three consecutive years. Our results revealed the existence of conserved temporal patterns, with microbial communities and networks going through a stabilization phase of decreased diversity and variability at the beginning of the plant's growing season. Despite a high turnover in these communities, we identified 19 "core" taxa persisting on Arabidopsis leaves across time and plant generations. With the hypothesis these microbes could be playing key roles in the structuring of leaf microbial communities, we conducted a time-informed microbial network analysis which showed core taxa are not necessarily highly connected network "hubs," and "hubs" alternate with time. Our study shows that leaf microbial communities exhibit reproducible dynamics and patterns, suggesting the potential of using our understanding of temporal trajectories in microbial community composition to design experiments aimed at driving these communities toward desired states. IMPORTANCE Utilizing plant microbiota to promote plant growth and plant health is key to more environmentally friendly agriculture. A major bottleneck in the engineering of plant-beneficial microbial communities is the low persistence of applied microbes under filed conditions, especially considering plant leaves. Indeed, although many leaf-associated microorganisms have the potential to promote plant growth and protect plants from pathogens, few of them are able to survive and thrive over time. In our study, we could show that leaf microbial communities are very variable at the beginning of the plant growing season but become more and more similar and less variable as the season progresses. We further identify a cohort of 19 "core" microbes, systematically present on plant leaves that would make these microbes exceptional candidates for future agricultural applications.

Keywords: community dynamics; core microbial community; hub microbes; leaf microbiome; microbial communities; microbial hubs; microbial networks; persistence; plant-microbe interactions; time dynamics.

FIG 1

Monitoring leaf microbiome dynamics throughout the natural growing season of A. thaliana. (A) Experimental setup. The four global Arabidopsis accessions Ws-0, Col-0, Ksk-1, and Sf-2 were planted in a common garden (Max Planck Institute, Cologne, Germany). Every month from November to March, three individual plants per ecotype were collected, and leaf samples were taken for microbiome analysis (destructive sampling). The experiment was repeated three times over the years 2014 to 2015 (experiment 1), 2015 to 2016 (experiment 2), and 2016 to 2017 (experiment 3), with a total number of 206 plant leaf samples analyzed (see Table S1). Average temperature and rainfall during the sampling season are shown. (B) Composition of the leaf microbiome. Microbiome analysis was conducted via Illumina-based amplicon sequencing (Miseq 2 × 300 bases). Taxonomic markers included the bacterial 16S rRNA v5-v7 region, fungal ITS1, and the oomycete ITS1 region. Bar charts show the average relative abundance of the main microbial groups (order level) by months, across three experiments. Arrowheads indicate taxa exhibiting marked seasonal patterns (see Fig. S2).

FIG 2

Persistent core members of the Arabidopsis leaf microbiome. (A) Core taxa were identified as OTUs showing high occurrence results (≥95% for fungi and oomycete, ≥98% for bacteria) in each of the three experiments. Bubbles depict the average relative abundance of each core OTU, per sample. The dendrogram depicts taxonomical distances between OTUs (hierarchical clustering on Gower distances from OTU taxonomy). (B) Changes in the relative abundance of core taxa over time (month averages; n > 38 samples per month).

FIG 3

Changes in alpha diversity and variability in leaf microbial communities over time. The alpha diversity (Shannon’s H index), within-month variability (distance to the group centroid; beta-dispersion), and between-month variability (Bray-Curtis distances between samples from consecutive months) in bacterial, fungal, and oomycete communities are shown. Each plot shows combined data from the three experiments, with n > 38 samples per month. Dots represent individual samples, whiskers depict the dispersion of the data (1.5 × interquartile range), and different letters indicate significant differences between groups (Dunn test, P < 0.05). Single BC distances between samples are not shown because of the high number of comparisons (>700).

FIG 4

Changes in phyllosphere microbial interaction networks throughout A. thaliana’s growing season. (A) Data from the three experiments were aggregated to reconstruct co-abundance networks for each time point (month) using the SparCC algorithm. Nodes (dots) represent OTUs; edges (colored lines) depict potential positive and negative interactions between OTUs (connections). Nodes from core microbes are indicated. Gray lines connecting networks show nodes conserved in networks from 1 month to the next (inherited nodes). (B) Number of nodes and edges in each month network. (C) Percentage of nodes and edges in a given month network which are inherited from (shared with) the previous month network. (D) Percentage of edges inherited for a given inherited node. (E) Node degree, i.e., number of edges per node in each month network. (F) Node-rewiring score (Dn-score) calculated in DyNet. For each node, its connected neighbors are compared between two networks (consecutive months) and the changes (rewiring) are quantified. Points represent rewiring scores from single nodes, high values indicate important changes in the node’s connections between the compared networks. Different letters indicate significant differences between conditions (Dunn test, P < 0.05).

FIG 5

Identification of microbial hubs within A. thaliana’s core leaf microbiome. The correlation networks calculated with SparCC (Fig. 3) were used to identify microbial hubs as nodes with high betweenness centrality (i.e., the fraction of shortest paths passing through the given node) and high closeness-centrality (i.e., the average shortest distance from the given node to other nodes). (A) Values for single taxa, with dotted lines indicating the top 5% values. Circles are colored based on microbial phyla. Circle sizes depict the node’s degree. Closed circles indicate taxa identified as part of the core leaf microbiome. Two core OTUs (12 and 4) are annotated. (B) Changes in the connectivity of core taxa. The product of “betweenness centrality × closeness centrality” was used to depict monthly changes in the connectivity of core OTUs. Hub taxa are indicated.