Introduction

Plants and animals are colonized by symbiotic bacteria. Symbionts include mutualists that help with nutrient uptake or pathogen protection, commensals that live in association with hosts without causing harm or benefit, and pathogens that cause disease. Transitions between symbiotic lifestyles [1, 2] and even hosts [3] can occur over relatively short evolutionary distances. While mechanisms underlying bacterial virulence and mutualism are widely studied, relatively little research has focused on the mechanisms by which bacteria initiate symbiotic host association and if these mechanisms are conserved across closely related pathogenic and commensal bacteria.

Both plants and animals have functionally similar receptors and signaling pathways that allow them to sense and respond to the presence of bacteria [4]. Accordingly, Pseudomonas aeruginosa makes use of the same virulence factors to infect both plants and animals [5, 6], suggesting that P. aeruginosa targets similar cellular processes across hosts. Consistently, the nitrogen-fixing plant mutualist Sinorhizobium meliloti has genes required for mutualism that are homologous to virulence factors in the animal pathogen Brucella abortus and the plant pathogen Agrobacterium tumefaciens [3]. Regardless of symbiotic lifestyle, host-associated bacteria must be able to colonize their host. As a result, we hypothesized that genes required for host colonization may be shared by pathogens, commensals, and mutualists.

The bacterial genus Pseudomonas includes plant commensals and opportunistic animal and plant pathogens [7,8,9] and thus provides a model genus to study conserved symbiotic mechanisms. We hypothesize that the ability to be symbiotic is ancestral within the genus Pseudomonas, and thus should be encoded by components of the core genome, with further evolution of species or strains leading to distinct lifestyles. Indeed, the Pseudomonas accessory genome (genes not conserved across the majority of Pseudomonas strains) has previously been shown to encode virulence factors [1, 10] while colonization factors may include components of the core genome [11].

Two-component systems are common signal-transduction cascades in bacteria that sense environmental cues, including the presence of a host, and cause altered gene expression. Orthologous two-component systems in Brucella abortus (BvrR/S), Sinorhizobium meliloti (ExoS/ChvI) and Agrobacterium tumefaciens (VirA/VirG), are required for virulence in animals and mutualism or virulence in plants [12,13,14,15]. The ColR/S two-component system is required for colonization of the tomato rhizosphere by Pseudomonas fluorescens WCS365 [16, 17], and ColS is required for Pseudomonas aeruginosa TBCF10839 virulence during infection of the nematode Caenorhabditis elegans [18]. As different species spanning the Pseudomonas genus can be host associated [19], we hypothesized that distantly related Pseudomonas species may use conserved two-component signaling pathways to sense and adapt to diverse host environments.

Because ColR/S has been identified in phylogenetically diverse Pseudomonas spp. as required for rhizosphere colonization [16, 20], or virulence in C. elegans [18], we tested whether ColR/S, and its regulon, are broadly conserved host-association factors across diverse Pseudomonas spp. and diverse hosts. The genes encoding for ColR/S are components of the core Pseudomonas genome, and previously identified components of the ColR regulon are broadly conserved across the genus [11]. ColS has been proposed to bind iron or zinc through a conserved ExxE motif [21]; as bacteria are known to colonize host environments with low pH [22, 23], and metals have higher solubility at low pH, this suggests ColS has the potential to sense diverse host environments. Collectively this indicates that identifying the genes and mechanisms by which ColR/S regulates host association may reveal conserved mechanisms that predispose members of the genus Pseudomonas for symbiosis.

To test the role of ColR/S in association across hosts, we used P. fluorescens WCS365, P. aeruginosa PAO1, and P. aeruginosa LESB58 as model strains to study mechanisms of plant and animal host association across the genus Pseudomonas. P. fluorescens WCS365 is a beneficial plant growth-promoting strain [24]. P. aeruginosa PAO1 is a human and plant opportunistic pathogen [25]. It can efficiently colonize the plant rhizosphere, but causes disease in plants including Arabidopsis, sweet basil, canola, and poplar trees [26,27,28]. P. aeruginosa PAO1 also quickly disseminates into the tissue of mice to cause infection, causing death of most mice within three days, making it an non-ideal strain for studying prolonged association with a murine host [29]. P. aeruginosa LESB58 was isolated from a cystic fibrosis patient and causes chronic infection in a subcutaneous murine infection model, making it an ideal strain for studying host association over many days [29]. We used growth of P. aeruginosa PAO1 and P. fluorescens WCS365 in an Arabidopsis rhizosphere model [30] to identify conserved or divergent ColR-dependent genes required for rhizosphere colonization. We then used both P. aeruginosa PAO1 and LESB58 in a murine subcutaneous abscess model to test if the genes required for rhizosphere colonization are also required for virulence in a murine model [29]. Collectively, use of these strains and hosts allows us to test whether ColR/S is broadly required for host association.

We generated deletions of colR in P. aeruginosa and P. fluorescens and found that ColR is required for rhizosphere colonization in both P. fluorescens and P. aeruginosa, and abscess formation by P. aeruginosa, indicating that ColR is required for host association across diverse strains and hosts. Using comparative transcriptomics, we found that ColR-dependent genes have largely diverged between Pseudomonas strains. We propose that ColR/S allows Pseudomonas to sense and respond to the presence of a host, but the ColR-regulon has diverged between Pseudomonas strains with distinct lifestyles.

Results

The ColR response regulator is conserved and required for rhizosphere colonization across Pseudomonas species

To assess the degree of functional conservation of the ColR response regulator across the genus Pseudomonas, we used the model strains P. fluorescens WCS365, P. aeruginosa PAO1, and P. aeruginosa LESB58 to study common mechanisms required for host association. An alignment of the ColR and ColS sequences from more than 500 orthologs across the genus Pseudomonas shows a high degree of similarity across the genus with conserved predicted DNA- and zinc-binding domains, respectively (Supplementary Fig. 1). P. aeruginosa PAO1 and P. fluorescens WCS365 ColR and ColS are 88% and 61% identical at the amino acid level, respectively, suggesting that ColR and ColS are highly conserved between these two species (Supplementary Fig. 1).

To test the conservation of the requirement of colR in rhizosphere colonization, we deleted colR in P. fluorescens WCS365 and in P. aeruginosa PAO1. We found that consistent with disruption of colS being required for potato root tip colonization [16], deletion of colR leads to a seven-fold decrease in growth of WCS365 (Fig. 1A) and a ten-fold decrease in rhizosphere growth of PAO1 (Fig. 1B) when inoculated into the rhizosphere of hydroponically grown Arabidopsis seedlings. In contrast, colR mutants grew at similar rates as wild-type bacteria when grown in M9 minimal media or MS media with succinate as a sole carbon source (Supplementary Fig. S2) indicating that the ∆colR mutants do not have a general growth defect. Microscopy with P. aeruginosa PAO1 expressing GFP revealed lower abundance of the ∆colR mutant in the rhizosphere (Fig. 1C). Complementation of colR deletion mutants with the copies of the colR from their respective strains expressed under the colR native promoter restored rhizosphere colonization of the colR deletion mutants to wild-type levels or higher (Fig. 1A, B). The complemented WCS365 and PAO1 colR mutants had significant growth defects in minimal media (Supplementary Fig. S2) suggesting that expression of colR from its native promoter on a plasmid might result in overexpression and/or altered phosphorylation, which might be detrimental to growth in vitro but not in vivo. Because of the high degree of conservation of ColR across species, we tested if P. aeruginosa PAO1 colR could complement a P. fluorescens WCS365 ∆colR mutant in vivo. We found that PAO1 colR expression from a plasmid is able to complement the WCS365 ∆colR deletion and rescue rhizosphere growth (Fig. 1A). Collectively these data indicate that colR is a conserved rhizophore colonization factor across diverse Pseudomonas.

Fig. 1: ColR is a conserved host association factor across Pseudomonas spp. in plant and murine hosts.
figure 1

A To test if colR is necessary for rhizosphere colonization, seedlings growing hydroponically in 48-well plates were inoculated with GFP-expressing P. fluorescens WCS365, WCS365 ∆colR, WCS365 ∆colR containing an empty vector (EV), a vector expressing WCS365 colR under its native promoter (WcolR), a vector expressing PAO1 colR under its native promoter (PcolR) or (B) P. aeruginosa PAO1, PAO1 ΔcolR, or PAO1 ΔcolR expressing colR under its native promoter (PcolR). A, B GFP fluorescence values in each plant-containing well were measured 5 days post-inoculation and converted to OD600 using a standard curve. Each data point represents the estimated CFU values from a single well containing a single plant. Experiments were repeated three independent times with 10-30 plants per replicate (n = 30–90). Mean +/− standard error is shown and letters indicate significant (p < 0.05) differences as determined by a one-way ANOVA follow by a post-hoc Tukey HSD test. C Light microscopy of P. aeruginosa PAO1 expressing GFP (green) growing for two days on Arabidopsis roots shows that deletion of colR leads to visibly less PAO1 colonization of the root. D, E Wild-type P. aeruginosa LESB58 and a LESB58 ∆colR mutant were injected (~5 ×107 CFU inoculum) into the subcutaneous thin skeletal muscle on the dorsum of mice. CFU counts (E) and abscess lesion size (E) were determined three days post infection. Means +/− standard error are shown; each dot represents the results from one animal. Different letters indicate significant differences at p < 0.05 determined using a one-way ANOVA followed by a Tukey’s HSD test, or using a t-test if only one comparison was made. F Representative images of abscess formation by wild-type LESB58 and the ∆colR mutant.

Because previous research has shown that virulence mechanisms are conserved across Pseudomonas aeruginosa infecting plants and animals [6, 31], we also tested if colR was required for virulence in a murine subcutaneous abscess model [29]. We deleted colR in the Liverpool Epidemic Strain P. aeruginosa LESB58 (Methods; Supplementary Dataset S1) and found that the colR mutant formed a significantly smaller abscess than wild-type LESB58 (54% reduction in size; 90 ± 9 mm for wild-type versus 48.6 ± 3 mM for the ∆colR mutant), indicating that colR was required for virulence (Fig. 1E, F). In contrast to ∆colR mutants in the rhizosphere, the LESB58 ∆colR mutant grew to similar levels as wild-type bacteria in an abscess (Fig. 1D) suggesting that colR in LESB58 may primarily contribute to virulence rather than growth in the abscess model. We also tested the more acutely infectious PAO1 ∆colR mutant for abscess formation and found that the ∆colR mutant did not have a pronounced virulence defect (Supplementary Fig. S3). Collectively these data indicate that ColR plays a role in host association through colonization or establishment of chronic infection in both P. aeruginosa and P. fluorescens.

The ColR regulons are highly divergent across Pseudomonas strains

The requirement of colR in symbiosis (commensalism or virulence) in P. fluorescens and P. aeruginosa led us to hypothesize that colR regulates an overlapping set of genes in these two species. As these two strains have distinct host-association lifestyles (beneficial rhizosphere bacterium vs. opportunistic human and plant pathogen), we hypothesized that colR may regulate genes required for host association rather than mutualism or virulence. To identify ColR-dependent genes in both P. fluorescens WCS365 and in P. aeruginosa PAO1, we performed RNA-Seq on wild-type bacteria and ∆colR mutants in the rhizosphere. We grew both PAO1 and WCS365 in the Arabidopsis rhizosphere and in M9 minimal media with 20 mM L-glutamine (L-gln) (P. aeruginosa PAO1) or 30 mM sodium succinate (P. fluorescens WCS365) and isolated RNA 6 h post-inoculation to identify rhizosphere-specific colR-dependent and independent responses. The sequencing reads from three biological replicates were mapped to the PAO1 or WCS365 transcriptomes. As previously only a draft genome of WCS365 was available, a complete WCS365 genome assembly was generated for use in this study (Methods).

To determine how divergent the transcriptomes were between wild-type bacteria and the colR mutants of the same strain, and between the rhizosphere and M9 minimal media, we performed principal component analysis (PCA) with the RNA-Seq data. We found that the majority of the differences in gene expression within a single strain (94–97%) were due to growth condition (media vs. rhizosphere) and only a small (1–2%) amount of variance in gene expression was due to bacterial genotype (wild-type vs. ∆colR) (Fig. 2A). This suggests that Pseudomonas undergo major transcriptional reprogramming in the rhizosphere, but that only a subset of these changes are ColR-dependent.

Fig. 2: ColR regulates a limited number of genes in the rhizosphere that are distinct between Pseudomonas strains.
figure 2

A PCA plots of transcript count matrices of RNA-Seq analysis comparing gene expression in wild-type P. aeruginosa PAO1 and the PAO1 ΔcolR mutant (left) or P. fluorescens WCS365 and the WCS365 ΔcolR mutant (right). Heat maps and Venn diagrams showing the number of significantly differentially expressed genes (greater than 0.585-fold change, padj <0.01) between wild-type and the colR mutant in the rhizosphere or in minimal media in PAO1 (B) and WCS365 (C). Gene counts higher than row average are shown in yellow, and gene counts lower than row average are shown in blue. D A phylogenetic tree showing orthologous genes across Pseudomonas strains. Genes fell into 5 distinct clusters: Cluster 1. Genes with ColR-dependent expression in both WCS365 and PAO1; Cluster 2. Genes with ColR-dependent expression in WCS365 with ColR-independent orthologs in PAO1; Cluster 3. Genes with ColR-dependent expression in PAO1 with ColR-independent orthologs in WCS365; Cluster 4. Genes with ColR-dependent in WCS365 without orthologs in PAO1; Cluster 5. ColR-dependent expression in PAO1 without orthologs in WCS365. Genes tested for rhizosphere colonization (Fig. 3) are marked with an asterisk.

Because only a small portion of differences in transcription in the rhizosphere were ColR-dependent, we hypothesized that failure to induce expression of these genes may explain the requirement of ColR in rhizosphere colonization. We identified genes that were specifically upregulated in the rhizosphere in wild-type relative to a ∆colR mutant in both WCS365 and in PAO1 (Dataset S2; Methods). We identified 50 and 42 positively ColR-regulated genes with higher expression in the rhizosphere in wild-type than in the colR mutant in WCS365 and in PAO1, respectively (Fig. 2B, C). These data indicate that ColR may directly or indirectly upregulate a limited set of genes during rhizosphere colonization.

To determine if the ColR regulon is conserved between Pseudomonas spp. WCS365 and PAO1, we used the RNA-Seq data described above to perform comparative transcriptomics to identify orthologous genes present in the ColR regulons of PAO1 and WCS365 [1]. We identified only four genes that were ColR regulated in both PAO1 and WCS365 (Fig. 2B–D; Supplementary Dataset S2). These include eptA and eptC (which encode phosphoethanolamine transferases), tpbA (which encodes a tyrosine phosphatase), and warA (which encodes a methyltransferase) (Fig. 2B–D; Cluster 1). Of these, expression of warA was previously shown to be ColR-regulated in P. putida [32]. We found that 26 ColR-dependent genes in P. fluorescens WCS365 and 21 ColR-dependent genes in P. aeruginosa PAO1 had a ColR-independent ortholog in the other strain (Fig. 2B, C; Cluster 2 and 3, and Supplementary Fig. S4), indicating these genes have divergent regulation. The remaining genes in the ColR regulon encoded genes unique to PAO1 or WCS365; 24 ColR-dependent genes in WCS365 and 16 genes in PAO1 did not have any orthologs in the other strain (Fig. 2B–D; Cluster 4 and 5) and are not highly conserved across the genus Pseudomonas (Fig. 2D). Despite ColR serving a similar function in rhizosphere colonization by P. aeruginosa PAO1 and P. fluorescens WCS365, these results suggest the ColR regulon has largely diverged between PAO1 and WCS365.

Deleting components of the ColR regulon revealed strain-specific, but not host-specific, requirements for ColR-dependent genes

We reasoned that orthologous ColR-regulated genes shared in both P. fluorescens WCS365 and P. aeruginosa PAO1 might be part of a conserved ColR regulon required for host association. We tested if eptA, tpbA, and warA, which are ColR-dependent in both P. fluorescens WCS365 and P. aeruginosa PAO1, were required for rhizosphere colonization in both strains. We used either transposon insertion mutants (PAO1 tpbA::Tn5 and PAO1 warA::Tn5 [33]), or generated clean deletions (PAO1 ∆eptA and WCS365 ∆eptA, ∆tpbA, and ∆warA). We found that the tpbA::Tn5 mutant had reduced rhizosphere colonization in PAO1 but deletion of the WCS365 tpbA ortholog had similar rhizosphere colonization as wild-type WCS365 (Fig. 3A, B). While the WCS365 ∆eptA and ∆warA mutants had reduced rhizosphere colonization, the PAO1 ∆eptA and warA::Tn5 mutants did not (Fig. 3A, B). None of the mutants had growth defects in vitro indicating that they are specifically required for host association (Supplementary Fig. S5). Collectively these data indicate that while ColR is a conserved determinant of rhizosphere colonization, ColR-dependent genes required for rhizosphere colonization have diverged between P. fluorescens WCS365 and P. aeruginosa PAO1.

Fig. 3: Strain-specific ColR-dependent genes are required for both rhizosphere colonization and virulence in a mouse abscess model.
figure 3

A Wild-type P. aeruginosa PAO1 or PAO1 mutants [∆colR, warA::Tn5 (PA4379), tpbA::Tn5 (PA3885); ∆eptA (PA1927), dgkA::Tn5 (PA3603), pap2::Tn5 (PA4382), or PA4518::Tn5] and B Wild-type P. fluorescens WCS365 or WCS365 mutants [∆colR, ∆warA (LRP86_00786), ∆tpbA (LRP86_02280), or ∆eptA (LRP86_03626)] containing GFP-expressing plasmids were inoculated into the Arabidopsis rhizosphere of hydroponically grown seedlings. Each data point represents the estimated cfu/mL for a single plant. Three independent experiments, each with a minimum of 10 plants, were performed (n > 30). Letters indicate significant differences at p < 0.05 using an ANOVA and Tukey’s HSD. Colors correspond to the gene clusters shown in Fig. 2D. C Genomic arrangement of orthologous genes in P. fluorescens WCS365 and P. aeruginosa PAO1. Orthologous genes are shown in the same color. Red arrows and striped, red boxes indicate putative ColR-binding sites and ColR-dependent transcription. Black arrows indicate ColR- independent transcription. D and E Wild-type P. aeruginosa LESB58 or LESB58 mutants [ΔdgkA (PALES_14321), ΔtpbA (PALES_10921), and Δpap2 (PALES_47610)] were injected into the subcutaneous thin skeletal muscle on the dorsum of mice. Abscess lesion size (D) and CFU counts (E) were determined three days post-infection. Error bars represent standard deviation. Letters indicate significant differences at p < 0.05 as determined by a Mann-Whitney U test; each dot represents the results from one animal.

We investigated if colR-dependent genes required for P. aeruginosa PAO1 rhizosphere colonization were also required for virulence in the murine subcutaneous abscess model. To find additional ColR-dependent genes required for rhizosphere colonization and abscess formation in P. aeruginosa, we tested rhizosphere colonization by 25 transposon insertion mutants in PAO1 genes with ColR-dependent expression (Supplementary Dataset S2; Methods). From this screen we identified two additional mutants, dgkA (encoding a diacyl glycerol kinase) and pap2 (with similarity to type 2 family phosphatidic acid phosphatases), that had significantly decreased colonization in the rhizosphere (Fig. 3A) but did not have growth defects in vitro (Supplementary Fig. S5).We found that deletion of the P. aeruginosa LESB58 orthologs of the genes with rhizosphere colonization defects in PAO1 (dgkA, pap2, and tpbA), resulted in decreased virulence in the abscess model (Fig. 3D). In addition, deletion of tpbA and pap2 led to a small but significant decrease in bacterial recovery (Fig. 3E), indicating that virulence of these mutants was impaired. Collectively these data indicate that an overlapping set of ColR-dependent genes is required for P. aeruginosa to associate with diverse hosts.

Transcriptional profiling suggests Pseudomonas strains experience diverse stresses in the rhizosphere

We found that the largest differences in gene expression were due to differences in bacterial growth condition (minimal media vs. rhizosphere, Fig. 2A) rather than differences in genotype. To identify rhizosphere stressors encountered by Pseudomonas spp. that might reveal why ColR is necessary for rhizosphere colonization, we queried categories of genes that are induced in the rhizosphere in wild-type bacteria (Fig. 4A, B) using gene ontology (GO) analysis. We found that in both P. fluorescens WCS365 and in P. aeruginosa PAO1, the global changes in gene regulation were largely similar between wild-type bacteria and a colR mutant in the rhizosphere when compared to minimal media (Fig. 4A, B; Supplementary Dataset S3). In both Pseudomonas species, we observed enrichment of genes involved in detoxification and survival in a harsh environment. These include metabolism of antibiotics and organic compounds, and genes involved in detoxifying the environment and responding to a new environment (Fig. 4C, D; Supplementary Dataset S3). The upregulation of genes involved in stress tolerance in both WCS365 and in PAO1 suggested that to grow in the rhizosphere, bacteria must protect themselves against harmful rhizosphere components.

Fig. 4: The P. aeruginosa PAO1 ∆colR mutant is sensitive to low pH and high iron.
figure 4

Venn diagrams of the number of total genes in PAO1 (A) and WCS365 (B) that are significantly upregulated in wild-type or a colR mutant in the rhizosphere vs. M9 minimal media (greater than 0.585-fold change, padj < 0.01). C, D GO categories of significantly differentially expressed genes between the rhizosphere and M9 minimal media in wild-type PAO1 (C) and wild-type WCS365 (D). Significantly enriched GO terms were identified using GOfuncR. EH Growth curves of wild-type PAO1 and PAO1 mutants (∆colR, tpbA::Tn5, dgkA::Tn5 or pap2::Tn5) in LB at pH 5 or 7 with and without the addition of 0.2 mM FeSO4. Growth curves were performed at least three times with similar results. Single experimental replicates are shown with averages and standard deviations of 6 technical replicates per time point.

Because we found upregulation of GO terms and processes related to protection against environmental stressors and membrane modifications, we tested if known rhizosphere stressors, including the presence of cationic peptides [34], could explain the growth inhibition of a colR mutant. We found that the minimal inhibitory concentration (MIC) of the cationic peptide polymyxin B was similar between wild-type PAO1 and the colR mutant (Supplementary Fig. S6). We also tested biofilm formation in a colR mutant, as biofilm formation was previously shown to be increased in a colR mutant, and is important for virulence [35] and rhizosphere colonization [36]. However, the colR mutant formed slightly lower levels of biofilm as wild-type bacteria in PAO1 and did not form lower levels of biofilm in LESB58 (Supplementary Fig. S7). This could be due to the use of different P. aeruginosa isolates (PA14 vs. PAO1 and LESB58), or the different conditions under which biofilm formation was tested (endotracheal tube vs. 96-well plate). These data indicate that the defect in rhizosphere growth of a colR mutant is unlikely to be due to the presence of cationic peptides or a change in biofilm formation.

ColR is necessary for growth at low pH and resistance to transition metals

We found that genes induced in the rhizosphere are enriched for GO terms including metal ion binding and electron transfer activity (Fig. 4), suggesting that bacteria experience metal stress in the rhizosphere. Previous studies showed that a P. putida colR mutant is sensitive to high levels of zinc and iron, and that this metal sensitivity is dependent on the loss of multiple ColR-dependent membrane modifying genes [21, 32]. We therefore confirmed that a P. aeruginosa PAO1 ∆colR mutant is similarly sensitive to metals. We determined the minimum inhibitory concentration (MIC) of PAO1 and the ∆colR mutant. We found that the P. aeruginosa PAO1 ∆colR mutant has lower tolerance to ZnSO4 (WT MIC = 10 mM; ∆colR MIC = 5 mM) and FeSO4 (WT MIC = 5 mM; ∆colR MIC = 2.5 mM) and similar tolerance to MnSO4 (WT MIC = 10 mM; ∆colR MIC = 10 mM) and CuSO4 (WT MIC = 10 mM; ∆colR MIC = 10 mM) than wild-type PAO1. Using metal concentrations below the MIC of P. aeruginosa PAO1 ∆colR, we performed growth curves with PAO1 and the ∆colR mutant and found that a PAO1 colR mutant has reduced growth relative to wild-type PAO1 in the presence of zinc, iron, and manganese (Supplementary Fig. S8).

The concentrations of metals in 1/2 X MS media (15 µM Zn, 0.1 mM Fe, and 0.05 mM Mn), the media in which the plants are grown, is below the levels that inhibit the ∆colR mutant (3 mM Zn, 1 mM Fe, and 3 mM Mn); however, metal solubility increases at low pH, and the rhizosphere is acidic [22] so we hypothesized that a synergistic effect of pH and metal stress might be inhibitory [37]. We measured bacterial growth at pH 5 in the presence of 0.2 mM FeSO4 and found that colR is required for tolerance to this combination of metal and pH stress (Fig. 4H). We tested if P. aeruginosa tpbA::Tn5, dgkA::Tn5 and pap2::Tn5 mutants also had increased sensitivity to FeSO4 at low pH and found that tpbA::Tn5 but not the other mutants partially phenocopied the growth defect of the ∆colR mutant (Fig. 4H). We also measured bacterial growth at pH 5 in ½ X MS plant growth media with succinate as the sole carbon source and varying concentrations of FeSO4.We found that both wild-type PAO1 and a ∆colR mutant have shorter doubling times in 1/2X MS media containing an additional 0.25 mM FeSO4 at pH 5 and that a ∆colR mutant has a consistent delay in growth relative to wild-type bacteria under this condition (Supplementary Fig. S9F, H). These data indicate that there is a complex interaction between media composition, pH, and metal concentration. These data also indicate that either the loss of multiple ColR-dependent genes is required for sensitivity to the synergetic effects of 0.2 mM FeSO4 and low pH, or that the loss of a ColR-dependent gene which we did not test is required for this sensitivity.

A mechanism of defence present in both plants and animals is the generation of reactive oxygen species (ROS) using iron through Fenton chemistry [38, 39] suggesting a potential shared environment across hosts that might require ColR. We tested whether the ColR mutant has increased sensitivity to ROS using H2O2 and found that while a katA::Tn5 mutant unable to make catalase cannot grow in 20 mM H2O2, growth of the P. aeruginosa PAO1 ∆colR mutant is indistinguishable from wild-type bacteria (Supplementary Fig. S10A). Additionally, survival of a ∆colR mutant after 30 min of treatment with 1 or 20 mM H2O2 was similar to wild-type P. aeruginosa (Supplementary Fig. S10B). As zinc and manganese do not participate in Fenton chemistry, and the PAO1 ∆colR mutants had dysregulation of genes involved in membrane modification (eptA, pap2, and dgkA) and ion transporters (mntP, fecA, and copA1), this suggests colR mutants might have increased uptake and decreased efflux of metals leading to toxicity.

Discussion

Bacteria in the genus Pseudomonas can form symbiotic associations with a diverse range of hosts with outcomes ranging from mutualism to pathogenesis. We showed that ColR is a conserved symbiosis factor, required for host-association not only in diverse Pseudomonas species, but also in diverse hosts. We found that the ColR regulon, and the ColR-dependent genes required for host association in distinct Pseudomonas spp., are largely divergent. Only four orthologous genes, eptA, eptC, warA, and tpbA, are ColR-dependent in both P. fluorescens WCS365 and P. aeruginosa PAO1. Of the remaining ColR-dependent genes, some were conserved between P. fluorescens WCS365 and P. aeruginosa PAO1 with divergent regulation, while others were not conserved between strains. This indicates that the divergence of the ColR regulon between strains may be due to both gain and loss of accessory (non-conserved) genome components, or promoter evolution, for example through the loss or gain of a ColR binding site in the promoter. Due to the conservation of ColR DNA binding residues, and ability to complement a WCS365 ∆colR mutant with PAO1 colR, it seems unlikely that changes to the ColR protein itself underly differences in the ColR regulon between the two strains. We found that none of the four conserved ColR-dependent genes are required for rhizosphere colonization in both species. This indicates that ColR/S may play a conserved role in sensing the host and responding through changes in transcription; however, the specific transcriptional changes depend on the bacterial genetic background.

We performed transcriptional profiling and GO analysis of genes upregulated in the rhizosphere in comparison to M9 minimal media, to identify genes that are likely involved in adaptation to growth in the rhizosphere environment (Fig. 4). By identifying stressors that bacteria experienced in the rhizosphere, we found that a ∆colR mutant is sensitive to a combination of low pH and high iron levels, suggesting a mechanism by which the rhizosphere or an abscess might inhibit growth or virulence, respectively, of a colR mutant. Metals are more soluble at low pH [37] and have diverse mechanisms of toxicity including mismetallation, inhibition of electron transport, and antagonizing uptake of other essential metals [40]. We found that the PAO1 ∆colR mutants had dysregulation of genes involved in membrane modification (eptA, pap2, and dgkA) and ion transport (mntP, fecA, and copA1; Fig. 2D and Supplementary Dataset S2), suggesting that colR mutants might have increased membrane permeability resulting in higher cytosolic metal levels. Coupled with decreased expression of efflux transporters this may lead to metal toxicity in a colR mutant.

As P. aeruginosa pap2 and dgkA mutants were impaired in rhizosphere colonization and virulence, but were not sensitive to low pH and high metals, this indicates that there must be an additional stress or process that ColR is required to adapt to in both the plant rhizosphere and in the murine abscess. This could be an inhibitory antimicrobial peptide, lack of an essential nutrient, or a heightened immune response in the plant rhizosphere or mouse tissue. DgkA and Pap2 convert diacylglycerol (DAG) to phosphatidic acid (PA), and PA to DAG, respectively [41, 42]. Both PA and DAG are key intermediates and precursors to the synthesis of many major membrane phospholipids [42]. Additionally, DAG is generated as a result of phosphoethanolamine (pEtN) addition to Lipid A [43]. We identified three pEtN transferase enzymes that were ColR-dependent in PAO1 [eptA (PA1972), eptA2 (PA3310) and eptC (PA4517)] and two in WCS365 [eptA (LRP86_03626) and eptC (LRP86_05619)]. This suggests that dgkA may be required for turnover of the resulting DAG and that loss of colR may result in decreased membrane structural integrity through disruption of phospholipid synthesis and turnover.

This study has found that colR is required for rhizosphere colonization in P. fluorescens and in P. aeruginosa and abscess formation in P. aeruginosa but not through the regulation of conserved genes. This suggests that ColS may be a conserved host-sensing protein that activates ColR to adapt to a host environment; however, the ColR regulon may have diverged according to bacterial lifestyle. Divergence of two-component system regulons has been previously shown to occur in members of the family Enterobacteriaceae, in which the PhoP response regulator targets were largely divergent among species [44]. Similarly, virulence through type III secretion systems relies on a conserved secretion apparatus coupled with host-specific diversification of effectors. This suggests that conservation of two-component systems, coupled with lifestyle dependent diversification of the regulon, may play a role in host association and lifestyle transitions.

Materials and methods

Bacterial strains, media, and growth conditions

Bacteria strains and plasmids are listed in Supplementary Dataset S1. All transposon mutants were obtained from the PAO1 two-allele library [33]. Routine bacterial culturing was performed using LB broth or agar at 37 °C (for P. aeruginosa PAO1 and LESB58 as well as E. coli) or at 28 °C (for P. fluorescens WCS365). When appropriate, LB was supplemented with sucrose, 25 µg/ml kanamycin (Km), 15 µg/ml gentamicin (Gm), 200 µg/ml carbenicillin (Cb), 10 µg/ml nalidixic acid, and/or 25 µg/ml irgasan.

Strain construction

E. coli DH5α was used for construction and maintenance of plasmids, and E. coli SM10 λpir was used for biparental conjugation with P. aeruginosa and P. fluorescens. Deletion mutants were generated using the two-step allelic exchange method [17, 45,46,47]. Deletion constructs for PAO1 ∆colR (PA4318), ∆eptA (PA1972), LESB58 ∆colR (PALES_47611), ∆tpbA (PALES_14321), ∆dgkA (PALES_47610), ∆pap2 (PALES_10931), WCS365 ∆colR (LRP86_00788), ∆eptA (LRP86_03626), ∆warA (LRP86_00786), and ∆tpbA (LRP86_02280) were generated using suitable primers (Supplementary Dataset S1). As the flanking regions have 100% sequence identity, the same deletion construct for ∆colR was used to delete colR from both PAO1 and LESB58.

Briefly, approximately 700–800 bp of sequence flanking the target genes were amplified using PCR. The flanking regions were joined using overlap extension PCR [48]. The PCR products were digested using the appropriate restriction enzymes (Supplementary Dataset S1), and ligated into the multiple cloning site of pEXG2 suicide vector containing sacB [49]. After confirming the insert by Sanger sequencing, the vector was transformed into E. coli SM10 λpir and transformants containing the plasmid were selected for on LB agar plates supplemented with 15 µg/ml Gm. The plasmid was then transferred to Pseudomonas via conjugation into PAO1, LESB58, or WCS365, and integrated into the genome through homologous recombination. Merodiploids were selected using Gm (50 µg/ml for PAO1, and 500 µg/ml for LESB58) and irgasan (25 µg/ml) for P. aeruginosa, or Gm (15 µg/ml) and nalidixic acid (10 µg/ml) for P. fluorescens. Counter-selection on LB plates with 10% sucrose was performed to select for loss of the plasmid. PCR and Sanger sequencing was used to screen for and validate deletion mutants.

The colR complementation plasmids were created using the pBBR1MCS-5 plasmid as a backbone for PAO1 and pBBR1MCS-2 for WCS365. PAO1 colR, or WCS365 colR, and approximately 200 bp upstream of each gene, respectively, were amplified using corresponding primers (Supplementary Dataset S1). Amplicons and plasmids were digested with the appropriate enzymes and then ligated together, and the resulting insert was confirmed by Sanger sequencing. The colR complementation strains were generated by electroporating pBBR1MCS-5-PAO1colR (PcolR) into the PAO1 ∆colR mutant, and pBBR1MCS-2-WCS365colR (WcolR) into the WCS365 ∆colR mutant. Gm (PAO1) or Km (WCS365) were used for selection and maintenance of the plasmid. Wild-type and ∆colR with empty vector (EV) were generated by electroporation with pBBR1MCS-5 for PAO1 or pBBR1MCS-2 for WCS365. GFP-expressing P. aeruginosa and P. fluorescens were generated by electroporating strains with pSMC21 (Ptac-GFP) [50, 51]. Electrocompetent cells were prepared by pelleting, washing and resuspending cultures in 300 mM sucrose. Transformants were selected on and maintained using Cb (300 µg/ml) for P. aeruginosa, and Km (25 µg/ml) for P. fluorescens.

Plant growth conditions

Axenic plants were generated by surface sterilizing Arabidopsis (wild-type accession Col-0) seeds in 70% ethanol for approximately 2 min, 10% bleach for approximately 2 min, and washed three times with sterile water (H2O). Seeds were resuspended in H2O or 0.1% phytoagar and stored at 4 °C in the dark for at least 48 h before sowing on Murashige and Skoog (MS) liquid media [52], as described below. Plants were grown at 16 h light/8 h dark under 100 µM cool white fluorescent lights at 22 °C.

Rhizosphere colonization in 48-well plates

Plants were grown hydroponically in flat-bottom 48-well plates as described [17, 30]. Briefly, one autoclave-sterilized Teflon mesh disk was placed in each well of a 48-well plate containing 250–300 µl MS plant growth media supplemented with 2% sucrose at pH 5.8. Sterilized Arabidopsis seeds were placed individually in the center of each disk. The plant media was changed to 270 µl of ½X MS media at pH 5.8 without sucrose after 10 days. At 12 days, overnight cultures of GFP-expressing bacteria were diluted to an OD600 of 0.0002 in 10 mM MgSO4, and 30 µl was inoculated into each well. GFP fluorescence (485 nm/535 nm excitation/emission) was read from the bottom of the wells at 5 days post-inoculation, using a SpectraMax i3x fluorescent plate reader (Molecular Devices). Bacterial growth in a minimum of 10 plant root-containing wells was measured for each strain per experiment.

Bacterial abundance in the rhizosphere was estimated from the GFP fluorescence readings of the entire well containing bacteria free-living in the well and those attached to the root and reported in CFU per ml by first converting GFP fluorescence readings to OD600 using standard curves specific to each strain or mutant, and then converted to CFU per ml using known values of number of bacteria per OD600 for both P. fluorescens WCS365 and P. aeruginosa PAO1. Standard curves of GFP fluorescence vs. OD600 were constructed for P. aeruginosa PAO1 and P. fluorescens WCS365. Separate standard curves were constructed for PAO1 and WCS365 strains that harbor two plasmids (pSMC21 and pBBR1MCS-5 or pBBR1MCS-2), because of possible differences in plasmid copy number due to the presence an additional expression vector. These plasmids contain a stability fragment that has been shown to confer a high degree of plasmid stability in the absence of selection [50].

Screening of the Manoil lab PAO1 two-allele library transposon insertion mutants (Supplementary Dataset S2) for rhizosphere colonization defects were performed similarly, but OD600 of the liquid media from each well was measured instead of GFP fluorescence. Non-GFP-expressing bacteria were inoculated into wells containing 12-day-old plants to a final OD600 of 0.0002. At 5 days post inoculation, 80 µl of media were transferred from each plant well into a 96-well plate. PAO1 and PAO1 ∆colR were included in each experiment as controls. OD600 readings were taken from the 96-well plate using SpectraMax i3x. A minimum of 8 wells were measured per strain. The ratio of mutant to wild-type colonization was estimated by dividing mutant OD600 over PAO1 OD600 after subtraction of the blank reading. Mutants with a OD600 ratio of below 0.7 relative to PAO1 were considered candidates for rhizosphere colonization defects and were retested using the GFP method above.

Study approval and animals

Animal experiments were performed in accordance with the Canadian Council on Animal Care (CCAC) guidelines following approval by the University of British Columbia Animal Care Committee (A14-0253). Mice used in this study were outbred CD-1 mice (female, 7–8 weeks). All animals were purchased from Charles River Laboratories, Inc. (Wilmington, MA). Mice weighed 25 ± 5 g at the experimental endpoint. Animals were group housed in cohorts of 4–5 littermates exposed to the same pathogen. Littermates were randomly assigned to experimental groups and standard animal husbandry protocols were employed.

Murine subcutaneous abscess infection model

The role of ColR in virulence during high density infection was examined using P. aeruginosa LESB58 [46] in a murine subcutaneous abscess model [29]. Briefly, the LESB58 wild-type and mutant strains (∆colR, ∆dgkA, ∆pap2, ∆tpbA) were sub-cultured at 37 °C with shaking (250 rpm) to an OD600 = 1.0 in LB. Cells were washed twice with sterile phosphate buffered saline (PBS) and resuspended to a final OD600 = 1.0. Bacteria were injected (50 μl) subcutaneously into the right or left dorsum of mice for an inoculum density of ~5.0 ×107 CFU. Abscesses were formed for 72 h with daily clinical grading. At the experimental endpoint, mice were euthanized by CO2 and cervical dislocation. Visible dermonecrosis was measured using a caliper, and abscesses were harvested in PBS. Then abscess tissues were homogenized using a Mini-Beadbeater (BioSpec Products, Bartlesville, OK) for bacterial enumeration on LB following serial dilution. Two or three independent experiments containing 2–4 biological replicates each were performed (n = 6–12).

Bacterial RNA extraction and RNA-Seq

RNA from rhizosphere-growing and minimal media-growing P. aeruginosa PAO1 and P. fluorescens WCS365 was extracted for RNA sequencing and analysis. M9 minimal media supplemented with 24 mM L-glutamine (L-gln) (P. aeruginosa PAO1), or supplemented with 30 mM sodium succinate (P. fluorescens WCS365) was used as the minimal media control. Plants were grown hydroponically as described in the 48-well plate rhizosphere colonization assay. Wells containing 12-day-old plants were inoculated with the wild-type and colR deletion mutants of P. aeruginosa PAO1 and P. fluorescens WCS365. To obtain sufficient RNA for sequencing, 6 wells containing plants or minimal media were inoculated with a final OD600 of 0.2 for PAO1 and 0.1 for WCS365. Both plant and minimal media plates were incubated in the plant growth conditions described above for 6 h. After 6 h, 40 µl of media from each of 6 plant wells or media from 3 minimal media wells were pooled and stabilized in RNAprotect Bacteria Reagent (QIAGEN) before performing RNA extraction using RNeasy Mini Kit (QIAGEN). When necessary, RNA was concentrated using ethanol precipitation.

RNA from three biological replicates was used with RNA integrity numbers (RIN) > 9.9. cDNA library preparation and RNA sequencing, using paired-end 150 bp reads, were performed by GENEWIZ using HiSeq (Illumina). The sequencing yielded an average of 16,897,599 (standard deviation = 429,191) reads for each sample for all PAO1 and PAO1 ∆colR samples, and an average of 16,055,232 (standard deviation = 760,587) reads for each sample for all WCS365 and WCS365 ∆colR samples. The quality of the reads was assessed using FASTQC v.0.11.8 [53]. Salmon v.1.1.0 [54] was used to align reads to the PAO1 transcriptome and to obtain the count files for each sample. Reads were mapped to the PAO1 transcriptome with an average mapping rate of 70.8% (standard deviation = 1.3%) for each sample for PAO1 and were mapped to the WCS365 transcriptome with an average mapping rate of 75.5% (standard deviation = 3.5%) for each sample for WCS365. DESeq2 v1.26.0 was used for differential expression analysis in R [55]. Log2 fold change values were corrected using the apeglm package in R [56]. Genes with a log2 fold change of ≥1.5 (±0.585 fold change) and adjusted p value ≤ 0.01 were considered significantly differentially expressed. Normalized reads for each gene were generated using the DESeq2 package in R. The dplyr package was used for data manipulation (sorting and merging lists). The heatmap function in R was used to generate a heatmap using reads normalized for length of each gene.

Comparative genomics

Ortholog groups between the ColR-dependent genes in WCS365 and PAO1 were identified using the PyParanoid pipeline described previously [1]. Briefly, the PyParanoid pipeline was previously used to create a database using 3886 Pseudomonas genomes, identifying 24,066 homologous protein families (“gene groups”) that covered 94.2% of the generated Pseudomonas pangenome. Groups were identified based on the PAO1 or WCS365 amino acid sequences and mapped onto a phylogenetic tree (Fig. 2). Phylogenetic trees were generated using the PyParanoid pipeline as described previously [1, 57].

Functional analysis of RNA-seq results

Gene ontology (GO) terms assigned to WCS365 were assigned using Blast2GO [58]. Lists of genes significantly differentially expressed in each condition or genotype were created using DEseq2, as described above. Lists of genes with assigned GO terms were inputted into R. GOfuncR [59] was used to identify GO terms significantly enriched in the conditions inputted (wild-type bacteria or ∆colR in the rhizosphere or minimal media). The go_enrich command was used to identify significantly enriched GO terms and GO terms with a FWER (family-wise error rate) ≤0.1 were refined to include only those terms with a refined p value ≤ 0.05. Code is available on the Haney Lab GitHub page: https://github.com/haneylab/ColR_paper_RNAseq_analysis.

In vitro metal, pH and H2O2 sensitivity assays

For all growth curves, bacterial cultures were pipetted into clear, flat-bottom 96-well plates to a final OD600 of 0.02 and a final volume of 100 or 200 µl. Growth was monitored using a shaking plate reader (Molecular Devices VersaMax or SpectraMax microplate reader), at a fixed temperature of 37 °C for PAO1 or 28–30 °C for WCS365. OD600 readings were taken once every 15 min for 24 h. At least four technical replicates were used for each strain and condition. The growth curves for each combination of strain and condition were repeated three times.

Growth in metals or H2O2 were assessed by addition of the appropriate concentration of metals or H2O2 to LB (FeSO4, ZnSO4, MnSO4, and CuSO4) adjusted to pH 5 or 7 with HCl or KOH. Metal concentrations used were below the MIC for both PAO1 and PAO1 ∆colR. To determine the MIC of the transition metals tested, LB media supplemented with 2X the highest concentration of metals (ZnSO4, MnSO4, FeSO4, and CuSO4) to be used for the MIC assay were prepared. The solutions of LB + metal were then serially diluted by 2X down the rows of 96-well plates. The final volume in each well was 50 µl in LB or 200 µL in 1/2X MS. Overnight bacterial cultures in LB were resuspended in fresh LB and diluted to an OD600 of 0.04 or to an OD600 0.1 in 1/2X MS. 50 µl (LB) or 200 µL (1/2X MS) of diluted bacterial cultures were added to each well. The 96-well plates were placed in an incubator at 37 °C for 24 h (LB) or 12 h (1/2X MS) before assessment of MIC. The MIC values were determined by eye (LB) or by dilution plating (1/2X MS) and scored as the lowest concentration of metals in which there was at least a 50% reduction in growth. Percent survival in H2O2 was performed by diluting overnight cultures 1:100 to an OD600 of 0.04–0.08 and adding H2O2 to a final concentration of 1 or 20 mM. After a 30 min incubation, bacteria were plated on LB plates and % survival was calculated relative to time 0 for each genotype and replicate.

Growth curves in ½X MS or M9 were performed as described above. Succinate at 20 mM was used as the carbon source in M9 minimal media and 1/2X MS media. For growth curves at varying pH, media was adjusted to pH 5, 6, or 7 using KOH. Overnight bacterial cultures in LB were washed and resuspended in 10 mM MgSO4 and diluted to an OD600 of 0.2. 90 µl of sterile media were aliquoted into each well in a 96-well plate before inoculation with 10 µl of the diluted bacterial suspension.

Polymyxin B MIC assays

The MIC assay for PAO1 was performed according to a modified MIC method for antimicrobial peptides (http://cmdr.ubc.ca/bobh/method/modified-mic-method-for-cationic-antimicrobial-peptides/). Briefly, individual bacterial colonies grown on MH agar media were used to start overnight cultures in MH broth. Serial dilutions of polymyxin B were performed by making 20X solutions of the antibiotic in 0.01% acetic acid, 0.2% BSA and serially diluting the antibiotic two- fold into sterile 96-well polypropylene microtitre plates. Cultures were then diluted to 5 ×106 cfu/mL (OD600 = 0.0073) and 100 µL of this diluted culture was added to each well. The MIC values were determined by eye and scored as the lowest concentration of polymyxin B in which there was at least a 50% reduction in growth.

Biofilm assay

P. aeruginosa PAO1 strains were tested in 24 h biofilm assays quantifying total biofilm biomass [60]. Briefly, bacteria from overnight cultures were diluted to an OD600 of 0.05 in Synthetic Cystic Fibrosis Media [61] without ammonium chloride (PAO1 strains) or LB (LESB58 strains) and 100 µl were inoculated into 96 well plates. After static growth for 24 h at 37 °C, the planktonic bacteria were rinsed off with deionized water. Biofilm biomass was stained with 105 µl 0.1% crystal violet and destained with 110 µl 70% ethanol. Lastly, absorbance was measured at 595 nm. Absorbance of each sample was blanked by subtracting the average absorbance of the sterility controls, and then normalized to the average absorbance of the wild-type and expressed as %wild-type biofilm. Statistical significance was determined using an unpaired t-test if comparing two values, or a one-way ANOVA followed by a post-hoc Tukey’s HSD test of multiple comparisons to correct for multiple comparisons if more than one value was being compared. p values <0.05 were considered significant.