Systemic co-infections of commensal Pseudomonas with an individual pathogen
To examine the ability of commensal Pseudomonas strains to protect host plants from members of a pathogenic Pseudomonas lineage, we made use of a local isolate collection [16]. We henceforth refer to an operational taxonomic unit (OTU) as reported in that study as “ATUE” (isolates from Around TUEbingen), and following previous findings [16, 17], we refer to the lineage ATUE5 as pathogenic, and to all non-ATUE5 lineages as commensals.
We grew plants on MS agar and monitored plant growth and health by extracting the number of green pixels from images over time (illustration in Fig. 1A). Green pixel count and rosette fresh weight were strongly correlated (Supplementary Fig. S1; R2 = 0.92, p value < 2.2e-16), validating the use of green pixels as a proxy for the plant biomass.
A Illustration of image processing to enumerate plant green pixels, approximating plant biomass. B Phylogenetic tree of 127 representatives, putatively commensal non-ATUE5 strains sampled from southwest Germany [16]. All other non-ATUE5 isolates in this collection are represented in this core collection by a strain with which they share ≥ 99% of genes. Colors indicate the ATUE group [16], and asterisks the 99 strains used here. C Daily median of plant green pixels among the different treatments. Control treatments: Bacteria-free buffer, pathogen only, and co-infections of the pathogen with the protective strain and the heat-killed protective strain. Plants were imaged daily. n = 8 replicates per treatment.
We use a conservative definition of plant protection, with protective strains leading to near-normal plant growth (comparable with uninfected plants) in the presence of a pathogen. To estimate how common the ability of non-ATUE5 strains to protect against the impact of pathogenic Pseudomonas is, we infected plants with a panel of non-ATUE5 isolates in the presence of a common ATUE5 representative, strain p4.C9, which hereafter will be referred to as “the pathogen” or “the ATUE5 pathogen”. This isolate was chosen because it is dominant over other ATUE5 strains but at the same time highly susceptible to the presence of non-ATUE5 strains, correlating with plant protection [17]. We excluded highly similar isolates, with Jaccard distances ≥ 0.99 in gene content. From a total of 151 non-ATUE5 strains in our local collection [16], we initially chose a subset of 127 isolates, from which we were able to revive 99 (Fig. 1B; Supplementary Table S1). We included three isolates from the pathogenic ATUE5 clade as control (Supplementary Table S1), resulting in 102 strains that were tested in co-infections with the pathogen p4.C9.
One of the 102 strains, p5.F2, was known to suppress ATUE5 strains inside plants [17] and was used as a “protective control”. To confirm that protection was due to bacterial activity and not merely a host response to the inoculum, e.g., PAMP-triggered immunity, we carried out infections with a heat-killed preparation of our protective control. As expected, co-infection with the protective strain resulted in normal plant growth, while treatment with the pathogen by itself or co-inoculation with the heat-killed strain impaired growth (Fig. 1C; Supplementary Fig. S2; mean growth [Δ7dpi-0dpi green pixels]: control 29,227 [15,721, 40,662], pathogen 3328 [−10,350, 14,759], pathogen + heat-killed protective −5222 [−16,410, 6211] and protective 29,585 [16,323, 40,812], at 95% confidence interval).
Plant protection is phylogenetically widespread, but also lineage-specific
Co-inoculation of non-ATUE5 strains with the ATUE5 pathogen led to a range of outcomes for plant health, from reduced to enhanced growth in comparison to uninfected plants, with protection being common (Fig. 2A).
A Mean difference in plant growth to control, after co-infection with different Pseudomonas strains and the focal p4.C9 pathogen. Growth was measured as the change in green pixels between days 0 and 7 after infection. Vertical lines indicate 95% Bayesian credible intervals and dots indicate the median estimate. The dashed vertical line signifies the positive control, that is, the pathogen-only treatment and the dashed horizontal line signifies the negative control, that is, the baseline average growth in the absence of bacteria. n = 8 replicates each. B Plant growth after co-infections, binned by ATUE group. In each ATUE group, raw data for individual replicates are shown with dots; the overlain shades of blue indicate the posterior predictive intervals, as presented on the bottom right. The mean growth for each ATUE group is also shown; the dot indicates the median, the thin horizontal line the 95% credible interval, and the thick line the 67% credible interval. Green shade indicates the 95% credible interval of the negative control, growth in the absence of bacteria. See Supplementary Table S1 for the number of strains in each ATUE group. C Median plant growth after co-infections, ordered by phylogeny. Colors indicate ATUE groups [16]. Medians of plant growth in mono-association with pathogen p4.C9 or without infection indicated by dashed vertical lines.
Protection was unevenly distributed among the three highly sampled ATUE groups—ATUE2, ATUE3, and ATUE4—with ATUE2 being the most protective mean growth (Δ[7 dpi]-[0 dpi] green pixels, 95% CI): ATUE2 22,231 [19,522, 24,908], ATUE3 14,120 [10,780, 17,389], ATUE4 10,868 [6844, 14,937], and bacteria-free 27,069 [19,882, 34,264]) (Fig. 2B).
Within each ATUE group, there was considerable variation in protective ability, even among strains with highly similar genomes (Fig. 2C). In some cases, closely related strains had contrasting activities, with one providing robust protection and the other having no effect, e.g., ATUE2 strains p11.F1 and p12.H7: mean growth after coinfection with p11.F1 being 25,554 pixels [8303, 42,567], but with p12.H7 being −28,914 pixels [−45,789, −11,837]).
Even within ATUE5, some isolates had the protective ability. We had chosen the three ATUE5 strains because in a previous set of experiments [17], they had appeared to be less competitive than our focal ATUE5 pathogen p4.C9, with lower abundance in the context of A. thaliana infections with synthetic communities of ATUE5 and non-ATUE5 strains. Surprisingly, two of them also mitigated the pathogen effect, causing normal plant growth. This further exemplifies the differential functions within ATUE groups (Fig. 2C).
ATUE2 protective genes are lineage-specific
That functional variation could not be explained by phylogeny, considering both topology or branch length, suggested that variation in gene content, possibly due to horizontal gene transfer, was causal for protection. We, therefore, were curious whether the presence/absence (P/A) of gene orthology groups could explain the propensity for plant protection.
There were 32,753 gene orthology groups, that were found in more than one strain, but not shared by all ATUE strains. We examined the correlation between plant growth and the presence of each of the individual orthology groups using treeWAS [19], a tool for genome-wide association studies (GWAS) in bacteria. Because treeWAS accounts for population structure, it will remove true positives that are highly correlated with population structure [20]. To address this issue, we used Spearman’s rank correlation coefficient (SRCC) to search for genes that were correlated with protection. SRCC should also be useful for differentiating a global from a lineage-specific signal. To this end, we calculated SRCCs separately among the highly sampled ATUE2, ATUE3, and ATUE4 groups, considering the difference in median green pixels between 0 and 7 dpi. For treeWAS, we used three additional plant growth metrics: the median change in green pixels between the last day of the experiment (7 days post-infection) to the day of infection (‘‘median growth’’), a Bayesian-derived approximation of the ‘median growth’’ (‘‘cdl50’’, i.e., the posterior distribution median of the ‘median growth’’), a binary categorization based on cdl50 and the area under the curve while accounting for all sampled time points (Supplementary Table S2; Methods). The extent of agreement between the four phenotypes was used as another indicator for the robustness of each association (Supplementary Table S2). We overlapped the results from all four analyses and removed genes with a negative SRCC (SRCC < 0, which implies negative effects on plant health in addition to those of the pathogen rather than protection from the pathogen), leaving us with 14 strong candidates for plant protection (Supplementary Table S2). The nine genes with the highest SRCC values (Rho 0.37-0.46) were unique to ATUE2 (Fig. 3A).
A Presence/absence variation of nine genes that are most strongly associated with protection. Strains are ordered by their phylogeny. Strains belonging to the ATUE2 group are indicated in magenta. Gene presence is indicated in grey, absence in white. Growth was measured as the change in green pixels between days 0 and 7 after infection, with the scale indicated on the right. B Daily median of plant green pixels among all strains (top panel) or the ATUE2 subset (bottom panel). Shades of magenta indicate the number of protective genes out of 9 present in each strain, as indicated by the scale on the bottom. Eight replicates per treatment. C Plant growth after co-infections with the pathogen p4.C9, binned by (i) the presence of at least one gene from the set of the nine protective genes (presented in panel A) in a given commensal strain, and (ii) membership in the ATUE2 group. In each group, raw data for individual replicates are shown as dots; the overlain shades of blue indicate the posterior predictive intervals, as indicated on the bottom right. The mean growth for each group is also shown; the dot indicates the median, the thin horizontal line the 95% credible interval, and the thick line the 67% credible interval. n = 8 replicates per strain, with the number of strains in the different categories as follows: non-ATUE2 with 0/9 genes = 67; ATUE2 with 0/9 genes = 20, and ATUE2 with at least 1/9 genes = 16.
The existence of a gene set unique to one lineage implies that plant protection by Pseudomonas is not driven by a universal mechanism in this genus, but rather by lineage-specific processes. To test this assumption, we ran the same GWAS analysis on each of the main ATUE groups separately, and independently analyzed the subsets of strains related to ATUE2, ATUE3, and ATUE4. In total, 95 genes with positive SRCCs were significantly associated with plant protection: 14 genes from the full set of strains, 46 from ATUE4, 35 from ATUE3, and none from ATUE2. Except for a single gene, we found no overlap among hits from the different ATUE groups, providing further evidence for lineage specificity (Supplementary Fig. S3).
Half of all protection-associated genes, 47 out of 95, were annotated as encoding ‘uncharacterized protein’’ or had no-hit in TrEMBL and Swiss-Prot databases (Supplementary Table S2). Out of the well-annotated genes, a few seemed to be likely to be directly related to microbe-microbe interactions: three iron-uptake-related genes unique to ATUE2 (GC00000450_7: ‘TonB_C domain-containing protein’’, GC00000032_87: ‘Putative iron(III) dicitrate sensor protein FecR’’ and GC00000050_54: ‘Probable RNA polymerase sigma factor FecI’’), a gene related to resistance to antimicrobial peptides in ATUE3 (GC00000089_55: ‘UDP-4-amino-4-deoxy-L-arabinose–oxoglutarate aminotransferase’’ [arnB] [21]), and an antitoxin gene in ATUE4 (GC00007392_r1_1: ‘Antitoxin FitA’’) alongside mobility-related genes (GC00002204_5: ‘Twitching motility protein PilT’’, GC00000715_r1_r1_2: ‘Pilus assembly protein PilW’’). These putative functions suggest not only a diverse set of genes but also diverse protective mechanisms among commensal Pseudomonas.
Several of these genes had identical SRCC values in a specific ATUE group and turned out to be physically linked (Supplementary Table S2; details in Methods). There was also evidence for genetic elements associated with horizontal gene transfer, such as phage genes (clusters 3 and 4 in Supplementary Table S2), consistent with these clusters being genomic islands. Cumulatively, these results suggest a scenario in which plant protection by commensal Pseudomonas is driven by multiple, clade-specific mechanisms that were horizontally transferred.
As described above, of the 14 protection-associated genes found among all strains, nine were unique to ATUE2 (Fig. 3A). Although there were no significant hits when we analyzed the ATUE2 subset on its own, strains carrying these nine genes were more protective not only when compared among all strains, but also within ATUE2 (Fig. 3B, C). For ATUE2 strains with at least one of these genes, plant growth was 29,299 pixels (22,567, 35,983, 95% confidence interval). In contrast, plant mean growth for ATUE2 strains without any of these genes was only 12,780 pixels (6891, 18,821), and for non-ATUE2 strains without any of these genes, it was only 10,791 pixels (7541, 14,078). Moreover, we found that the effect of these nine genes was additive within ATUE2, and plant growth was correlated with the number of genes present in a given Pseudomonas strain (Supplementary Fig. S4; R2 = 0.55, p value = 0.0004). These results highlight the importance of these nine genes in plant protection both among all Pseudomonas strains in our collection, and within ATUE2, despite the lack of statistical signal in the ATUE2-subset alone.
Plant protection by ATUE2 linked to iron acquisition and biofilm formation
We selected the nine ATUE2-lineage specific candidates for gene deletion in the protective strain p5.F2 as a representative for the ATUE2 clade, to validate the role of these genes in plant protection. Since two genes were found next to each other (GC00000032_87 and GC00000050_54), we included them in one deletion, resulting in eight mutations. Because our knockout mutants were marked with gentamicin resistance, we generated a wild type strain that also expresses gentamicin (as well as the lux operon), ensuring a similar metabolic burden (hereafter “wild type” or “WT”; see Methods).
Out of the eight knockout mutants, three lost their ability to mitigate the effect of the pathogen p4.C9 following co-infection, and we call them henceforth “non-protective mutants” (Fig. 4A; Supplementary Fig. S5A; results of an additional experiment with the three non-protective mutants in Supplementary Fig. S6A, B). These three non-protective mutants did not affect plant weight when tested individually, similarly to the wild type (Supplementary Fig. S7A, B), implying that the weight reduction after co-infection was due to the pathogen or to the interaction with the pathogen, rather than the knockout mutants themselves.
A Daily median of plant green pixels after treatment with control, p4.C9 pathogen, the protective strain p5.F2, a mixture of the pathogen and p5.F2, and mixtures of the pathogen and each of eight p5.F2 knockout mutants, indicated as p5.F2 K.O. Dashed lines indicate three treatments with p5.F2 knockout mutants that lost their ability to protect the plant (Δ GC00000032_87 // GC00000050_54, Δ GC00000450_7, and Δ GC00003884_12), as analyzed in Supplementary Fig. S5A. n = 20 replicates. A second experiment gave similar results, as detailed in Supplementary Fig. S6A, B. B In vitro growth of the pathogen, p5.F2 and two knockout mutants (Δ GC00000032_87 // GC00000050_54 and Δ GC00000450_7) as a function of 2,2’-bipyridine concentration, analyzed in six 50 nM increments from 0 to 300 nM. OD600 was monitored for 10 h (see also Supplementary Fig. S5 and Methods). The logistic area under the growth curve was extracted as a proxy for bacterial growth. The shaded area indicates 95% confidence intervals of the regression curve. n = 4 for each bacterial strain grown in the corresponding 2,2’-bipyridine concentration. C Daily median of plant green pixels after treatment with control, Pseudomonas syringae pv. tomato DC3000 (Pst), mixture of Pst and p5.F2, and mixture of Pst and each of the three tested p5.F2 knockout mutants, indicated as p5.F2 K.O. Dashed lines indicate two treatments with p5.F2 knockout mutants that lost their ability to protect the plant (Δ GC00000032_87 // GC00000050_54 and Δ GC00000450_7), as analyzed in Supplementary Fig. S5B. n = 20 replicates. A second experiment gave similar results, as detailed in Supplementary Fig. S6C, D.
Out of the four genes in the three mutants that had lost their ability to protect from pathogens, three adjacent genes are annotated as iron-related: one encodes a TonB_C domain-containing protein (GC00000450_7) [22, 23], one a putative iron (III) dicitrate sensor related to FecR (GC00000032_87) [24], and one a probable RNA polymerase sigma factor related to FecI (GC00000050_54) [25] (SupplementaryTable S2).
The fourth gene, GC00003884_12, is annotated as encoding an ‘uncharacterized protein’’ (Supplementary Table S2). When ΔGC00003884_12 was grown in tubes while shaking, clumps formed, suggesting the aberrant formation of cellular aggregates. To investigate this further, we grew the wild type and ΔGC00003884_12 overnight, and then let the cell suspension settle on the benchtop. After 1 h, cell sediment had formed in the ΔGC00003884_12 tube, with a clear liquid forming at the top and aggregates at the bottom (Supplementary Fig. S8), while the wild-type cells had stayed in suspension and the culture still had retained its homogenous opaque appearance. This phenotype implies that at least in the tested conditions, GC00003884_12 suppresses cell-cell aggregation and biofilm formation.
To validate the role of GC00000032_87 // GC00000050_54 and GC00000450_7 in iron uptake, we performed an in vitro iron-deficiency growth assay. The wild-type strain, the ΔGC00000032_87 // GC00000050_54 and ΔGC00000450_7 mutants and the pathogen were grown in LB with increasing levels of the iron chelator 2,2’-dipyridyl. We also tested ΔGC00003884_12 in the same assay, but noted an unusual growth curve shape for this strain, regardless of chelator levels (Supplementary Fig. S9), in agreement with the formation of cell aggregates we had observed (Supplementary Fig. S8). Consequently, we excluded the mutant ΔGC00003884_12 from further analysis.
Both ΔGC00000032_87 // GC00000050_54 and ΔGC00000450_7, as well as the pathogen p4.C9, grew more slowly than the wild-type commensal strain in LB without chelator (Fig. 4B; Supplementary Figs. S9, S10A; mean difference to control: pathogen −4424 AUC [−5299, −3551, 95% confidence interval], ΔGC00000450_7 −3526 AUC [−4412, −2634] and ΔGC00000032_87 // GC00000050_54 -2,469 AUC [−3337, −1587]). This confirms that both loci have a role in growth that is independent of iron availability. Increasing chelator levels led to a reduction in the growth of the mutants, the pathogen and the wild-type commensals, with the pathogen and ΔGC00000450_7 being more sensitive to the iron deficiency than the other three strains (Fig. 4B; Supplementary Figs. S9, S10B; mean slope difference to control: pathogen −5.6 AUC [−10.5, −0.7, 95% confidence interval], ΔGC00000450_7 −10.1 AUC [−15.1, −5.3] and ΔGC00000032_87 // GC00000050_54 -0.4 AUC [−5.3, 4.3]). These findings provide evidence for the involvement of iron acquisition in plant protection by ATUE2 strains and specifically imply that ATUE2 strains protect the plant by outcompeting the pathogen over iron.
Taken together, these results suggest that ATUE2 members can protect A. thaliana from coexisting pathogenic ATUE5 isolates via mechanisms related to iron uptake, and apparently also to biofilm formation. To test whether these mechanisms act specifically against the local ATUE5 strains or whether they are also effective against other Pseudomonas pathogens, we co-infected the protective wild-type commensal and the three non-protective mutants with the model pathogen Pseudomonas syringae pv. tomato DC3000 (Pst) [26]. Pst-like strains were rare in the local A. thaliana populations from which ATUE2 was sampled [16], hence it is unknown whether ATUE2-like strains and Pst-like strains co-exist. The wild-type commensal protected against Pst infection, greatly reducing the effects of Pst on plant growth (Fig. 4C; Supplementary Fig. S5B). While this suggests that ATUE2 protective ability is not restricted to the pathogenic lineage ATUE5, the exact modes of protection against the two pathogens seem to differ, since our mutants that had lost protection against the ATUE5 pathogen were still able to protect against Pst infection (Fig. 4C; Supplementary Fig. S5B; replicated in an additional experiment, shown in Supplementary Fig. S6C, D).
Source: Ecology - nature.com
