Fig. 1. Isolation and identification of pathogenic strain RPI 2201. A, Rice nursery tray with seedling blight. B, Rice plants with and without seedling blight. The position within the circle is the area that has been locally magnified. C, Colony morphology of RPI2201 strain. D, Photograph of RPI2201’s spores. E, Rice seedlings inoculated with RPI2201 strain. F, Phylogenetic analysis of RPI2201 strain based on the sequencing of PCR products (primers ITS1 and ITS2 are listed in Table S1).

Fig. 2. Combination of two different Bacillus strains better suppresses rice seedling blight than applying them individually. A, Phenotype of rice seedlings in the presence of Fusarium oxysporum with applying T40 and T208 strains individually or in combination. B‒E, Disease incidence of seedling blight (B), seedling emergence rate (C), plant height (D), and rice seedling biomass (E) under Bacillus strain treatments. Different lowercase letters above the bars indicate significant differences at the 0.05 probability level according to the Duncan’s test (n = 8).

Fig. 3. Microbial community compositions in rhizosphere soil under different treatments. A, Relative abundance of Fusarium spp. treated with the T40 and T208 strains individually or in combination. Different lowercase letters above the bars indicate significant differences among the treatments according to the Duncan’s test (P < 0.05, n = 8). B and C, Stacked bar charts showing the relative abundance of various bacterial (B) and fungal (C) phyla communities. D and E, Ternary plots of bacterial and fungal amplicon sequence variants (ASVs). Each circle represents one ASV, and the size indicates its relative abundance. Circles marked NS indicate ASVs that did not significantly enrich in any of the three treatments.

Fig. 4. Combination of T40 and T208 strains did not significantly change alpha diversities of microbial communities but significantly altered microbial community structures. A and B, Alpha-diversity of bacterial communities in rhizosphere soil of rice, expressed as richness (Chao1 index, A) and evenness (Shannon index, B). C and D, Alpha-diversity of fungal communities in rhizosphere soil of rice, expressed as richness (Chao1 index, C) and evenness (Shannon index, D). E and F, Beta-diversities of bacterial (E) and fungal (F) communities between different treatments. Different lowercase letters above the bars indicate significant differences among the treatments according to the Duncan’s test (P < 0.05, n = 8).

Fig. 5. Combining different Bacillus strains enhances microbial network stability. Co-occurrence networks of microbial communities, including both bacterial and fungal species in the samples. A connection represents a significant (P < 0.01) correlation between two amplicon sequence variants. The size of each node indicates the number of connections (i.e., degree), and the thickness of each connection (i.e., edge) between two nodes represents the value of the Spearman correlation coefficient. A blue edge indicates a positive correlation, while a red edge indicates a negative correlation.

Fig. 6. Spearman’s correlation analysis between rice seedling biomass and major factors that explored by a random forest classification approach. 1, Rice seedling biomass; 2, Disease incidence; 3, β-nearest taxon index; 4, Fungal Chao1 index; 5, Fungal Shannon index; 6, Average degree; 7, Density; 8, Modularity; 9, Average clustering coefficient; 10, Positive correlation within the co-occurrence networks; 11, The ratio of positive to negative correlations in the co-occurrence networks; 12, ITS32; 13, ITS193. Asterisks (*) represent significant differences (Student’s t-test, *, P < 0.05; **, P < 0.01; ***, P < 0.001).