Convergent gain and loss of genomic islands drive lifestyle changes in plant-associated Pseudomonas

Abstract

Host-associated bacteria can have both beneficial and detrimental effects on host health. While some of the molecular mechanisms that determine these outcomes are known, little is known about the evolutionary histories of pathogenic or mutualistic lifestyles. Using the model plant Arabidopsis, we found that closely related strains within the Pseudomonas fluorescens species complex promote plant growth and occasionally cause disease. To elucidate the genetic basis of the transition between commensalism and pathogenesis, we developed a computational pipeline and identified genomic islands that correlate with outcomes for plant health. One island containing genes for lipopeptide biosynthesis and quorum-sensing is required for pathogenesis. Conservation of the quorum-sensing machinery in this island allows pathogenic strains to eavesdrop on quorum signals in the environment and coordinate pathogenic behavior. We found that genomic loci associated with both pathogenic and commensal lifestyles were convergently gained and lost in multiple lineages through homologous recombination, possibly constituting an early step in the differentiation of pathogenic and commensal lifestyles. Collectively this work provides novel insights into the evolution of commensal and pathogenic lifestyles within a single clade of host-associated bacteria.

Fig. 1

A pathogen within a clade of beneficial Pseudomonas spp. a A species tree of characterized Pseudomonas strains was generated using 217 conserved housekeeping genes. Strain names are color-coded to reflect plant-associated lifestyles based on their effects on Arabidopsis. Pie charts at each node reflect the proportion of the genome that encodes “core” genes that are present in each member of the clade, showing that there can be large differences in gene content over short evolutionary distances. b Gnotobiotic Arabidopsis seedlings inoculated with N2C3, N2E2, or a buffer control. c Fresh weight of seedlings treated with 10 mM MgSO₄, N2C3, N2E2, NFM421, and WCS365. Lateral root density of seedlings treated with 10 mM MgSO₄, N2E2, NFM421, and WCS365. Lower-case letters indicate statistically significantly different (p < 0.01, 26 > n > 12) means as determined by a two-sided Student’s T-test for unpaired samples. Boxplots are quartile plots of the observed data. Data shown are from one of two independent experiments performed with similar results

Fig. 2

A genomic island necessary for pathogenesis. a A large genomic island encoding lipopeptide biosynthesis and quorum-sensing genes (the LPQ island) is present in N2C3 but not in beneficial strains such as N2E2. b Diagram of the genomic island showing putative functions and locations of quorum-sensing genes (luxI_LPQ—acyl-homoserine lactone (AHL) synthase, luxR_LPQ—AHL-binding transcriptional regulator) and the lipopeptide biosynthesis clusters (syringomycin and syringopeptin). c Fresh weight of Arabidopsis treated with 10 mM MgSO₄, wild-type N2C3, wild-type N2E2, N2C3 with a syringomycin cluster deletion (∆SYR), a syringopeptin cluster deletion (∆SYP), and a syringomycin and syringopeptin double deletion (∆SYR∆SYP). Lowercase letters indicate statistically significant (p < 0.01, 20 > n > 15) differences as determined by a two-sided unpaired Student’s T-test. Boxplots are quartile plots of the observed data. Data shown are from one of two independent experiments performed with similar results. d Fresh weight of Arabidopsis treated with 10 mM MgSO₄, wild-type N2C3, N2C3 with an AHL synthase deletion, (∆luxI_LPQ), and N2C3 with an AHL-binding transcriptional regulator deletion (∆luxR_LPQ). Lowercase letters indicate statistically significant (p < 0.01, 11 > n > 5) differences as determined by a two-sided unpaired Student’s T-test. Data shown are from one of two independent experiments performed with similar results. Boxplots are quartile plots of the observed data. e Lateral root density of seedlings treated with 10 mM MgSO₄, wild-type N2C3, a syringomycin and syringopeptin double deletion (∆SYR∆SYP), and wild-type N2E2. Lowercase letters indicate statistically significant (p < 0.01, 19 > n > 29) differences as determined by a two-sided unpaired Student’s T-test. Boxplots are quartile plots of the observed data. f Species tree showing the distribution of the island among divergent members of the Pseudomonas clade. Color coding of taxon labels summarizes the phenotypic data from g. Colored squares indicate that an individual gene (as classified by PyParanoid) is present in a given taxon. Only 15 of the 28 genes in b (shown in green and teal in a, Data S1 for details) that are unique to the pathogenic strains are shown in e. The other 13 genes (shown in purple in b) are part of larger lipopeptide biosynthesis homolog groups that are not specific to the LPQ island. g Fresh weight of Arabidopsis seedlings treated with one of the 18 wild-type strains assayed in this study. Shown here are means of data from two independent experiments normalized to a 10 mM MgSO₄ control (10 < n < 35). Error bars represent a 95% confidence interval for each mean using a non-parametric bootstrap of 1000 resamples

Fig. 3

Polyphyletic distribution of pathogenic and commensal genomic islands within the bcm clade. A species tree was built for the bcm clade using a concatenated alignment of 2030 conserved genes. Color-coded squares represent presence and absence of individual genes associated with each locus based on PyParanoid presence-absence data. In the case of both the LPQ and T3SS, only a subset of genes to the island are shown because both islands contain genes that are part of larger homolog groups not specific to either island. Bold taxon names highlight strains experimentally tested for commensal (green) or pathogenic (red) behavior (Fig. 2g). For strains that we did not test experimentally, we annotated rhizosphere lifestyle based on previous reports in the literature with the same color scheme. All six loci exhibit a highly polyphyletic distribution in extant strains suggesting that the pathogenic and commensal lifestyles have a complex evolutionary history in this clade

Fig. 4

Phylogenomic evidence for independent island gain and loss through homologous recombination of flanking regions. a Insertion sites for each locus are conserved throughout the bcm species complex. The LPQ and type III secretion system (T3SS) island insertion sites are shown here for the seven bcm clade strains that we characterized experimentally. All insertion sites for all six loci in the 85 strains can be found in Figures S5-S10. The tree shows the evolutionary relationship of the bcm strains. The gray background highlights that the conserved regions flanking both islands are adjacent in strains lacking the island. The location of the trx-like gene is shown adjacent to the T3SS, which is used as a marker for recombination in c. b Circular genome plots of two representative complete genomes from the pathogenic strain N2C3 and the beneficial strain NFM421. Darker gray regions of the genome represent the core genome of the bcm clade, while light gray denotes the variable genome. c A species phylogeny for the bcm clade adapted from Fig. 3 to show evidence of homologous recombination of the T3SS island inferred through phylogenetic incongruencies between the species tree and the phylogeny of the trx-like gene (see Figure S13). Dark green rectangles denote taxa that have the T3SS island, while white rectangles represent taxa missing the island. Based on the phylogenetic evidence, we propose a parsimonious model for evolution of the T3SS island where gain of the T3SS island led to divergence of an ancestral lineage (purple arrow/rectangle) into a lineage with the T3SS present (light green arrow/rectangle) and a lineage with the T3SS absent (light red arrow/rectangle). However, subsequent gain (green arrows) or loss (red arrows) of the T3SS island has led to phylogenetic incongruence between the adjacent trx-like gene and the conserved loci used to construct the species tree (i.e. the bcm clade clonal phylogeny). d A diagram of the proposed homologous recombination mechanism leading to island gain (top) and loss (bottom). Both events lead to insertion of a trx-like allele with a distinct history from the surrounding conserved chromosomal loci

Fig. 5

A lipopeptide-associated quorum-sensing in divergent Pseudomonas spp. a Gene tree showing the phylogenetic relationship of the six acyl-homoserine lactone (AHL) synthase homology groups identified by PyParanoid. Light gray squares indicate the absence of a gene and dark gray or purple squares show the presence of a gene. Group 22008 contains the LuxI_LPQ protein sequence from N2C3 as well as protein sequences from other LPQ+ strains (PdfI, PcoI, and PmeI) [41, 42, 45]. b Correlation coefficients between each AHL synthase homology group and the 14 other LPQ island homology groups across the entire Pseudomonas clade showing that the presence of LuxI_LPQ is correlated with the capacity for lipopeptide biosynthesis across the entire Pseudomonas genus. c Fresh weight of seedlings inoculated with 10 mM MgSO₄, wild-type N2C3, wild-type DF41, the AHL synthase deletion mutant N2C3∆luxI_LPQ alone or co-inoculated with DF41, and the AHL-binding transcriptional regulator mutant N2C3∆luxR_LPQ alone or co-inoculated with DF41. Lowercase letters indicate statistically significant (p < 0.01, 23 > n > 12) differences as determined by a two-sided unpaired Student’s T-test. Data shown are from one of two independent experiments performed with similar results. Boxplots are quartile plots of the observed data. d Species tree based on 217 housekeeping genes showing the distribution of six AHL synthase homology groups identified using PyParanoid. All strains with an asterisk were tested in a visual screen for induction of pigment production in the C. violaceum CV026 AHL biosensor strain, but only strains in the bcm clade with the LPQ island resulted in visible pigment production. Within the bcm clade, purple and brown taxon names show strains that were confirmed to induce or have no effect, respectively, on violacein production in a quantitative assay. e Barplot showing quantitative violacein production by C. violaceum CV026 by specified strains in the bcm clade. Data are the average of three biological replicates and error bars indicate the standard deviation (n = 3). Violacein production was normalized to total bacterial growth by calculating the ratio of OD₅₈₅/OD₇₆₄