A model to explain plant growth promotion traits: a multivariate analysis of 2,211 bacterial isolates

Abstract

Plant growth-promoting bacteria can greatly assist sustainable farming by improving plant health and biomass while reducing fertilizer use. The plant-microorganism-environment interaction is an open and complex system, and despite the active research in the area, patterns in root ecology are elusive. Here, we simultaneously analyzed the plant growth-promoting bacteria datasets from seven independent studies that shared a methodology for bioprospection and phenotype screening. The soil richness of the isolate's origin was classified by a Principal Component Analysis. A Categorical Principal Component Analysis was used to classify the soil richness according to isolate's indolic compound production, siderophores production and phosphate solubilization abilities, and bacterial genera composition. Multiple patterns and relationships were found and verified with nonparametric hypothesis testing. Including niche colonization in the analysis, we proposed a model to explain the expression of bacterial plant growth-promoting traits according to the soil nutritional status. Our model shows that plants favor interaction with growth hormone producers under rich nutrient conditions but favor nutrient solubilizers under poor conditions. We also performed several comparisons among the different genera, highlighting interesting ecological interactions and limitations. Our model could be used to direct plant growth-promoting bacteria bioprospection and metagenomic sampling.

Figure 1. PCA analysis of the soil characteristics from the 40 soils samples (numbered black circles) that were used for bacterial isolation.

The percentages show how much variation is explained by each principal component. The soils with higher pH, organic matter (OM), potassium (K), phosphorus (P), and clay (Clay) contents are plotted to the right. There are three clusters along the first principal component (PC1) that grouped the soils by overall richness. Based on these clusters, all 40 of the soil samples were classified according to their overall soil richness: poor, average or rich. The appropriate soil richness was attributed to each bacterial isolate (according to its origin) before further analysis. Supervised statistics of these data on S1 Fig.

Figure 2. CatPCA analysis of 2,211 bacterial isolates.

The indolic compounds production, TCP solubilization, siderophores production and soil richness are shown as colored vectors, with arrows indicating the vector's direction in the plot. The black numbers show the average position of each bacterial genus. In the right column are shown the bacterial genera, the number they represent in the plot (Plot code), and their frequency in the dataset (Freq). Cronbach's alpha value was 0.774.

Figure 3. Indolic compound production ability of the isolates (average rank ±1 SE) according to the soil nutrient conditions and TCP solubilization and siderophores production abilities.

The phosphate solubilization and siderophores production scores are 1 =  no halo, 2 =  small or average halo, and 3 =  large halo. The soil richness score is according to the PCA analysis (Fig. 1). Different letters show significant differences. Sample sizes and p values are presented on S3 Table.

Figure 4. Heat map associations of the TCP solubilization (left) and siderophores production (middle) abilities of bacterial isolates with soil conditions and with each other (right).

Phos  =  TCP solubilization, and Sid  =  siderophores production. 1 =  no halo, 2 =  small or average halo, and 3 =  large halo. The red cells  =  less isolates than expected under those conditions, the green cells  =  excessive number of isolates under those conditions, and the yellow cells  =  no significant differences between the observed and expected values. Percentages and residuals are shown in S2 Fig. Sample sizes and p values are presented on S3 Table.

Figure 5. Niche effect on ICs production (average ±1 SE) between endophytic (root) and rhizospheric (soil) isolates under each soil condition.

The best ICs producers shift their colonization site according to soil richness. Sample sizes and p values are presented on S3 Table.

Figure 6. Heat map associations of the TCP solubilization and siderophores production abilities of endophytic (root) and rhizospheric (soil) isolates under each individual soil condition.

Phos  =  TCP solubilization, and Sid  =  siderophores production. 1 =  no halo, 2 =  small or average halo, and 3 =  large halo. The red cells  =  less isolates than expected under those conditions, the green cells  =  excessive number of isolates under those conditions, and the yellow cells  =  no significant differences between the observed and expected values. Percentages and residuals are shown in S3 Fig. Sample sizes and p values are presented on S3 Table.

Figure 7. Heat map associations of bacterial genera and PGP traits (left), soil richness (middle), and occurrence of putative endophytic (Root) and rhizospheric (Soil) bacteria under each soil richness condition (right).

Phos  =  TCP solubilization, Sid  =  siderophores production, with 1 =  no halo, 2 =  small or average halo, and 3 =  large halo. ICs  =  Indolic compounds production, with 1 =  low (0–10 µg of ICs ml⁻¹), 2 =  average (11–80 µg of ICs ml⁻¹) and 3 =  high (80 or> µg of ICs ml⁻¹). The red cells  =  less isolates than expected under those conditions, the green cells  =  excessive number of isolates under those conditions, and the yellow cells  =  no significant differences between the observed and expected values. “–”  =  an association could not be calculated due to the lack of cases (no expected total marginal values). Percentages and residuals are shown in S4 Fig. Sample sizes and p values are presented on S3 Table.

Figure 8. PGP traits of some bacterial strains shifted due to the soil richness.

Only those bacterial genera that significantly changed their PGP traits are shown. Each box is a separate chi-square test, with non-significant tests shown entirely in yellow. Phos  =  TCP solubilization, and Sid  =  siderophores production, with 1 =  no halo, 2 =  small or average halo, and 3 =  large halo. ICs  =  Indolic compounds production, with 1 =  low (0–10 µg of ICs ml⁻¹), 2 =  average (11–80 µg of ICs ml⁻¹) and 3 =  high (80 or> µg of ICs ml⁻¹). The red cells  =  less isolates than expected under those conditions, the green cells  =  excessive number of isolates under those conditions, and the yellow cells  =  no significant differences between the observed and expected values. “–”  =  an association could not be calculated due to a lack of cases (no expected total marginal values). Percentages and residuals are shown in S5 Fig. Sample sizes and p values are presented on S3 Table.

Figure 9. CatPCA analysis of 2,211 bacterial isolates (the legend and interpretation are similar to those of Fig. 2).

The genera Burkholderia, Enterobacter, Klebsiella, Pseudomonas, Stenotrophomonas, Herbaspirilum, Rhizobium, and Grimontella are represented one at a time. Each black dot represents an isolate, but isolates with the same characteristics are stacked on the same dot.

Figure 10. A model to explain the distribution of bacteria displaying different plant growth promotion traits.

In soils with fewer nutrients, plants leave the best growth hormone producers in the rhizosphere, while both endophytic and rhizospheric bacteria are good nutrient solubilizers. In soils with more nutrients, the best growth hormone producers are found inside plant roots, but the endophytic bacteria are poor nutrient solubilizers, with the best solubilizers found in the rhizosphere. In addition, genera diversity and growth hormone producers are more abundant in soils with more nutrients, while phosphate solubilizers and siderophores producers are more abundant in soils with fewer nutrients. Siderophores producers and phosphate solubilizers seem to co-occur, while indolic compound producers are clearly opposed to phosphate solubilizers. Plants seem to select bacterial PGP traits according to their nutritional needs: nutrient solubilizers under poor conditions and growth hormone producers under rich conditions.