Phage biocontrol success of bacterial wilt depends on synergistic interactions with resident rhizosphere microbiota

Abstract

Phages can successfully be used in vitro and in planta to biocontrol the phytopathogenic Ralstonia solanacearum bacterium-the causal agent of bacterial wilt disease. However, phage biocontrol outcomes are still variable, and it is unclear what causes this. In this study, we assessed the efficiency of four phages in controlled in vitro and in planta experiments in all one- and two-phage combinations. We found that using phages in combination did not improve the phage biocontrol efficiency relative to single phage treatments, while certain phages and their combinations were more effective than the others. High intra-treatment variability in phage efficiency was observed across all phage treatments, which was associated with clear shifts in microbiome composition, a reduction in R. solanacearum and an increase in phage densities. We further identified the bacterial taxa that were associated with these 'shifted' microbiomes and conducted additional plant growth experiments, demonstrating that some of the enriched bacterial species could protect plants from R. solanacearum infections-a pattern which was also observed using partial least squares path modelling (PLS-PM). Together, these results suggest that phages could open niche space for beneficial bacteria by reducing pathogen densities and that variability in phage biocontrol outcomes is rhizosphere microbiome-dependent, which can introduce between-replicate variation, even in controlled greenhouse conditions.

FIGURE 1

Testing phage biocontrol efficiency in a greenhouse experiment. (A) Bacterial wilt disease development over time (days post‐inoculation (DPI)). (B) Area under the diseased progression curve showing one‐phage (orange) and two‐phage (blue) treatments, positive control with the R. solanacearum only (purple) and negative control (grey) plants (same colour coding also used in panels C and D). (C) Pathogen relative abundances in the rhizosphere averaged over one‐ and two‐phage treatments relative to control treatments. (D) Phage abundances (log(PFU/mL)) in the rhizosphere averaged over one‐ and two‐phage treatments relative to control treatments. (E) Bacterial wilt disease development over time in terms of area under the diseased progression curve for all 10 phage treatments (colours), positive control with the R. solanacearum UW551 only (black) and negative control (grey; N = 12 plants per treatment). (F) Pathogen relative abundances in the rhizosphere for all phage and control treatments. (G) Phage abundances (log(PFU/mL)) in the rhizosphere for all phage and control treatments. (H) Pearson correlation between relative Ralstonia abundances (log of the counts) and the area under the disease progression curve for the mean of phage and control treatments. (I) Pearson correlation between phage abundances (log(PFU/mL)) and the area under the disease progression curve for the mean of phage treatments. (J) Pathogen relative abundances in the rhizosphere between healthy and diseased plants. (K) Phage abundances (log(PFU/mL)) in the rhizosphere between healthy and diseased plants.

FIGURE 2

Phage application shifts bacterial community composition and relative taxa abundances in a subset of phage treatment replicates. Panels (A and B) show bacterial alpha diversities in terms of Shannon index for control and one‐ and two‐phage treatments (A) and healthy and diseased plants (B). (C, D) show correlations between R. solanacearum (log counts) (C) and phage (log(PFU/mL)) (D) densities along with Non‐metric multidimensional scaling (NMDS) axis 1 scores. (E) NMDS plot comparing differences in bacterial community composition in terms of beta diversity between one‐ and two‐phage treatments and control plants for ‘centred’ and ‘shifted’ plant replicates. (F) Relative bacterial phyla abundances in ‘shifted’ and ‘centred’ plant microbiome replicate samples for control and one‐ and two‐phage treatments.

FIGURE 3

Identification of key bacterial taxa associated with shifted plant rhizosphere microbiome plant replicates. (A) Multidimensional scaling plot showing the distribution of shifted and centred rhizosphere microbiome replicate samples based on the random forest analysis. (B) The relative importance of different bacterial genera for grouping samples to ‘shifted’ and ‘centred’ microbiomes based on Gini values (in descending order). (C–G) Comparison of relative abundances of five bacterial taxa with high Gini values between control and ‘shifted’ and ‘centred’ phage treatment replicate samples (based on read counts).

FIGURE 4

Direct experiments testing the in vitro and in planta biocontrol effect of selected rhizosphere taxa. (A) Inhibitory potential of the nine selected bacteria against R. solanacearum based on inhibition halos (mm) observed in spotting assays. (B) Area under the growth curve during 48 h in liquid culture measuring GFP (emission at 509 nm) when growing individually or when growing with GFP‐tagged R. solanacearum. (C) Number of healthy and diseased plants in plants treated Pseudomonas P19, Rhodanobacter R55, Burkholderia B12 and Burkholderia PB18 and R. solanacearum. (D) Comparison of relative abundance of ASV1852, showing the highest similarity to Burkholderia B12 between control and ‘shifted’ and ‘centred’ phage treatment samples (based on read counts).

FIGURE 5

Partial least squares path models (PLS‐PM) comparing significant associations between different variables on pathogen densities and bacterial wilt disease progression. (A) Path coefficients for the PLS‐PM in the shifted microbiome sample replicates. (B) Path coefficients for the PLS‐PM in the centred microbiome sample replicates. In both models, ‘Phages’ refer to phage densities (log(PFU/mL)), ‘Microbiome’ to NMDS1 score, ‘Ralstonia’ to pathogen densities (read counts) and ‘Disease’ to disease score at 21 dpi and the area under the disease progression curve. Positive and negative associations between variables are shown in blue and red, respectively, and the arrows show the direction of effects. The R ² values show the percentage of variation explained by the models.