Capacity of soil bacteria to reach the phyllosphere and convergence of floral communities despite soil microbiota variation

Abstract

Leaves and flowers are colonized by diverse bacteria that impact plant fitness and evolution. Although the structure of these microbial communities is becoming well-characterized, various aspects of their environmental origin and selection by plants remain uncertain, such as the relative proportion of soilborne bacteria in phyllosphere communities. Here, to address this issue and to provide experimental support for bacteria being filtered by flowers, we conducted common-garden experiments outside and under gnotobiotic conditions. We grew Arabidopsis thaliana in a soil substitute and added two microbial communities from natural soils. We estimated that at least 25% of the phyllosphere bacteria collected from the plants grown in the open environment were also detected in the controlled conditions, in which bacteria could reach leaves and flowers only from the soil. These taxa represented more than 40% of the communities based on amplicon sequencing. Unsupervised hierarchical clustering approaches supported the convergence of all floral microbiota, and 24 of the 28 bacteria responsible for this pattern belonged to the Burkholderiaceae family, which includes known plant pathogens and plant growth-promoting members. We anticipate that our study will foster future investigations regarding the routes used by soil microbes to reach leaves and flowers, the ubiquity of the environmental filtering of Burkholderiaceae across plant species and environments, and the potential functional effects of the accumulation of these bacteria in the reproductive organs of flowering plants.

Keywords: bacterial communities; flower; phyllosphere; soil.

Fig. 1.

Experimental design.

Fig. 2.

Soil community influence on the phyllosphere. (A) Hexagonal heatmap of the proportion of ASVs detected in the phyllosphere at the outside location also detected in phyllosphere communities in the growth chamber. All samples collected outside were combined in one pool, and those collected inside were combined in another before comparisons. (B) Unsupervised hierarchical clustering of phyllosphere communities (identified by their sample number) collected at outside locations, using the Bray–Curtis metric as a measure of dissimilarity. The data were all rarefied according to the number of reads from the smallest plant sample collected outside. Each panel consists of a dendrogram and a heatmap generated from rarefied data and two majority-rule consensuses: the first one summarizes the hierarchical clustering analyses based on 1,000 bootstrapped datasets, and the second one summarizes the hierarchical clustering analyses based on 1,000 rarefied datasets. The nodes are labeled with the percentage of recovery of each cluster across these iterative analyses. The tips are colored according to the community used to inoculate the soil in which the plants were grown (gray, AG community; blue, FO community). prop., proportion.

Fig. 3.

Bacterial diversity. (A) Relative abundances of bacterial lineages in the different types of samples. (B) Community richness in the different types of samples and different location and soil treatment combinations (data were rarefied; numbers of samples are indicated above each boxplot; centerline, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range; points, outliers).

Fig. 4.

Convergence of phyllosphere communities. Unsupervised hierarchical clustering analyses of bacterial communities. The communities are identified by their sample number. (A) Communities of all the samples collected outside at the flowering stage. The first dendrogram and heatmap were generated from one rarefied dataset; the second and third dendrograms are majority-rule consensuses of hierarchical clustering analyses conducted with 1,000 bootstrapped and 1,000 rarefied datasets, respectively. (B) Communities of the plant samples collected during the flowering stage at outside locations plus communities of the plant samples grown in the growth chamber in soil communities. The layout is the same as that in A. (C) Communities of all the samples collected outside during the flowering stage after the removal of the candidate ASVs responsible for the convergence of the floral cluster. The first dendrogram and heatmap were generated with a rarefied dataset, and the majority-rule consensus was obtained from hierarchical clustering analyses of 1,000 rarefied datasets. (D) Majority-rule consensus of hierarchical clustering analyses after 1,000 subsequent removals of 28 ASVs (not the ASV candidates responsible for the clustering of floral communities observed in A). The numbers presented at the nodes of the majority-rule consensus dendrograms are the percentage of clusters recovered across analyses conducted with 1,000 bootstrapped datasets and 1,000 rarefied datasets. The numbered clusters (clusters 1 and 2) are not supported with bootstrapped and 1,000 rarefied datasets. The clusters supported with bootstrap and 1,000 rarefied datasets are labeled as described in the text to allow comparisons across panels of the figure. Asterisks indicate that the relative positions of tips give the misleading impression that cluster D exists in the first dendrogram of C.