Arabidopsis thaliana and Pseudomonas Pathogens Exhibit Stable Associations over Evolutionary Timescales

Abstract

Crop disease outbreaks are often associated with clonal expansions of single pathogenic lineages. To determine whether similar boom-and-bust scenarios hold for wild pathosystems, we carried out a multi-year, multi-site survey of Pseudomonas in its natural host Arabidopsis thaliana. The most common Pseudomonas lineage corresponded to a ubiquitous pathogenic clade. Sequencing of 1,524 genomes revealed this lineage to have diversified approximately 300,000 years ago, containing dozens of genetically identifiable pathogenic sublineages. There is differentiation at the level of both gene content and disease phenotype, although the differentiation may not provide fitness advantages to specific sublineages. The coexistence of sublineages indicates that in contrast to crop systems, no single strain has been able to overtake the studied A. thaliana populations in the recent past. Our results suggest that selective pressures acting on a plant pathogen in wild hosts are likely to be much more complex than those in agricultural systems.

Keywords: A. thaliana; clonal expansion; microbial population genomics; pathogenicity; pseudomonas.

Graphical abstract

Figure 1

Natural Pseudomonas Populations in A. thaliana Leaves Are Dominated by the OTU5 Lineage (A) Overview of 16S rDNA survey of epi- and endophytic compartments of A. thaliana plants (dots indicate sampled plants). Red numbers indicate individuals from which Pseudomonas isolates were cultured and metagenome analysis was performed in parallel. (B) Heatmap of relative abundance of 56 Pseudomonas OTUs in the 16S rDNA survey. Color key to samples on top according to (A). Pseudomonas species assignments on the right. P. veronii, P. fragi, and P. umsongensis belong to the P. fluorescence complex; P. nitroreducens and P. alcaligenes to the P. aeruginosa complex. (C) Correlation between occurrence across all samples and average relative abundance within samples of the 56 Pseudomonas OTUs in the endo- and epiphytic compartments. (D) Pseudomonas abundance (gray bars) and percentage of Pseudomonas reads belonging to OTU5 (red bars), in the endo- and epiphytic compartments. (E) OTU5 is significantly more abundant in the endophytic compartment (Wilcoxon test, p = 10⁻⁴). (F) ML phylogenetic tree illustrating the similarity between amplicon sequencing-derived and isolation-derived Pseudomonas OTUs defined by distance clustering at 99% sequence identity of the v3-v4 regions of the 16S rDNA. For isolate OTUs, exact 16S rDNA sequences were used; for amplicon sequencing OTUs, the most common representative sequence was used. Gray dots on branches indicate bootstrap values >0.7. Colored bars represent the relative abundance or the number of isolates. The most abundant Pseudomonas OTU in both the endophytic and epiphytic compartments, OTU5, was identical in sequence to the most abundant sequence observed among isolates and to a P. viridiflava reference genome (NCBI AY597278.1/AY597280.1). See also Figures S1, S2, and S5.

Figure 2

The Most Abundant OTU, which Encompasses OTU5 (OTU5-enc), Is Correlated with Microbial Load (A) Bacterial and plant fraction of metagenome shotgun sequencing reads in 176 plants. (B) Correlation between fraction of bacterial reads in metagenome data and relative abundance of OTU with 100% identity to OTU5 in the v4 region of 16S rDNA amplicons from the same 176 samples. (C) Distribution of Pearson correlation coefficients between microbial loads as inferred from fraction of bacterial reads and OTU abundances (as shown for OTU5 in B). The correlation coefficient for the OTU5-associated sequence abundance is the highest among any of the 3,647 OTUs detected across all samples. See also Figure S2.

Figure 3

OTU5 Is Composed of Multiple Expanding Lineages that Are Pathogenic (A) ML whole-genome phylogeny and abundance of strains in Eyach, Germany, in December 2015 and March 2016. Diameters of circles on the right indicate relative abundance across all isolates from that season. Purple circles at nodes relevant for OTU5 classification and gray numbers indicate support with 100 bootstrap trials. (B) Examples of OTU5 strains that can reduce growth and even cause obvious disease symptoms in gnotobiotic hosts. (C) Quantification of effect of drip infection on growth of plants. Pst DC3000 was used as positive control. The negative control did not contain bacteria. Values represent mean ± SEM. See also Figures S2 and S5.

Figure 4

Genome-wide Divergence and Dating of OTU5 Strains (A) Pairwise nucleotide diversity in 1,000 bp sliding windows. One randomly chosen OTU5 reference strain was separately compared with four different other OTU5 strains. (B) Genome-wide distribution of segregating sites (S_n) in OTU5, calculated in 1,000 bp sliding windows. Putative recombination tracts were removed from the core genome alignment to calculate the coalescence of OTU5. This removal reduced the fraction of segregating sites by half (0.14 versus 0.07). (C) The TMRCA of 107 isolates representing the genetic diversity of OTU5 strains as calculated using a substitution rate estimated in McCann et al. (2017). See also Figure S3.

Figure 5

Different OTU5 Strains Expand Clonally within Different Plants (A) Distribution of relative OTU5 and non-OTU5 strain abundances in single plants. (B) Phylogenetic trees of isolates collected from individual plants. (C) Strain diversity as function of season. (D) Pseudomonas load as function of season. For both (C) and (D), seasons are significantly different (Student’s t test, p = 1.32 × 10⁻¹⁵). Boxplots show median and first and third quartiles. Related to Figure S4.

Figure 6

OTU5 Strains Vary in Gene Content but Share an Effector (A) ML tree topology of 1,524 isolates (for all panels) and presence (dark purple) or absence (light purple) of the 30,000 most common orthologs as inferred with panX (Ding et al., 2018). OTU5 strains share 622 ortholog groups that are found at less than 10% frequency outside OTU5. (B) Presence/absence of 30 effector homologs. Only avrE homologs are present in more than 50 isolates. (C) Genes for toxins and phytohormones. Only a few genomes contain the full set of genes required for synthesis of syringomycin and syringopeptin. (D) Similarity of avrE and hrp-hrc genes to Pto DC3000 and the most similar P. viridiflava genome from NCBI. Higher values indicate greater similarity. Related to Table S2.