Greenhouse experiment
In the experiment conducted under greenhouse conditions with 10 treatments and six replicates, the results showed that the shoot dry mass varied between 5.02 and 15.03 grams and that the treatments did not differ from each other (p > 0.05) according to the 5% Duncan test (Fig. 1A). The root dry mass ranged between 0.46 and 2.00 grams, and the values also did not differ from each other (p > 0.05) (Fig. 1B). The highest mean values (p < 0.05) of nitrogen levels in the SDM were found in treatments T5 (T. harzianum furrow application) and T6 (T. harzianum foliar application), followed by treatments T1 (control), T3 (P. lilacinum via foliar) and T7 (T. harzianum via furrow + foliar). The lowest nitrogen contents were found in treatments T2 (P. lilacinum via furrow), T8 (P. lilacinum + T. harzianum via furrow), T9 (P. lilacinum + T. harzianum via foliar) and T10 (P. lilacinum + T. harzianum furrow + foliar) (Fig. 2A).
Regarding the average of phosphorus content in the SDM, the highest values were found in treatment T10 (P. lilacinum + T. harzianum via furrow and foliar), followed by treatments T1 (control) and T5 (T. harzianum via furrow), T6 (T. harzianum via foliar) and T7 (T. harzianum via furrow and foliar). Treatments T4 (P. lilacinum via foliar) and T8 (P. lilacinum + T. harzianum via furrow) had lower average phosphorus contents (Fig. 2B).
The highest mean values of nitrogen from the root were found in the T8 treatment (P. lilacinum + T. harzianum via furrow) compared to the T2 treatment. (P. lilacinum via furrow). The other treatments did not differ (p > 0.05) from each other (Fig. 3A). There was no significant difference (p > 0.05) between the treatments and the control treatment (Fig. 3B).
Regarding the diversity of endophytic bacteria from the roots of soybean plants, inoculations of the fungi P. lilacinum and T. harzianum performed in the experiment under greenhouse conditions showed no difference between the treatments. The bacterial genera found most frequently in the treatments were Pseudomonas sp., Salmonella sp., Rhizobium, Rhizobiaceae, Agrobacterium and Stenotrophomonas. The bacterial genera and species least abundant in the treatments were Achromobacter, Pantoa dispersa and bentonitic Stenotrophomonas (Fig. 4).
There was no significant difference (p > 0.05) in the diversity of endophytic fungi isolated from the roots in the pot experiment under greenhouse conditions. The fungal genera and species present in greatest quantities were Fusarium, F. oxysporum, Plectosphaerella and Curvularia. The genera and species that were found in smaller numbers were Mythecium inundatum, Chaetomium globosum, Fusarium suflorianum and Saccharomyces (Fig. 5).
The mean values found in the split analysis with two factors, factor 1 (fungal species, T. harzianum; P. lilacinum and T. harzianum + P. lilacinum) and factor 2 (application methods, furrow, foliar and furrow + foliar), under potted conditions showed no difference (p > 0.05) for the parameters of SDM (Fig. 6), nitrogen content from SDM (Fig. 7) or phosphorus content from SDM (Fig. 8).
Experiment under field conditions
The highest values for SDM under field conditions were found in treatments T2 (T. harzianum), T3 (P. lilacinum) and T4 (T. harzianum + P. lilacinum) compared with the control treatment T1, which had the lowest value (p < 0.05) (Fig. 9).
There was no significant difference (p < 0.05) in nitrogen content between the treatments based on SDM (Fig. 10A). Nevertheless, the phosphorus content was the highest in the T3 treatment, which utilized P. lilacinum (Fig. 10B).
The most abundant phylum was Pseudomonas, and the percentages in the T1 (control), T2 (T. harzianum), T3 (P. lilacinum) and T4 (T. harzianum + P. lilacinum) treatments were 44.93%, 29.18%, 20.01% and 53.88%, respectively. The second most abundant genus was Bradyrhizobium at 31.24%, 33.482%, 44.72%, and 17.43%, and the third most abundant genus was Enterobacter at 16.77%, 24.50%, 25.552%, and 24.377%, respectively (Fig. 11). Interestingly, there was a significant difference in the number of species between treatments T1 (control), T2 (T. harzianum), T3 (P. lilacinum) and T4 (T. harzianum + P. lilacinum), with 70, 50, 13 and 29 species, respectively.
Comparative analysis between treated samples (T2 to T4 treatments) and control samples (T1 treatment) revealed a total of 77 differentially abundant taxa (DA), categorized into 2 phyla, 4 classes, 5 orders, 19 families, and 26 genera. Among these, 58 taxa were predominantly more abundant in the T1 treatment (control), while 6, 4 and 9 taxa were more abundant in the T2 (T. harzianum), T3 (P. lilacinum) and T4 (T. harzianum) treatments. + P. lilacinum), respectively. The DA taxa at the genus level are visualized in Fig. 13. Notably, with the exception of Kosakonia, all other affected genera exhibited a relative abundance lower than 1%. Genera such as Oligoflexus, Achromobacter, Shinella and Sphingobacterium were differentially abundant in the T1 treatment (control) in multiple comparisons. Notably, the genus Luteibacter was also differentially abundant in treatments T2 (T. harzianum) and T4 (T. harzianum + P. lilacinum). The greatest differences (log2-fold changes) were observed in the absence of genus in the opposite treatment. However, the greatest variations in terms of relative abundance were recorded for Kosakonia, which was significantly more abundant in T3 (P. lilacinum) (3.6%) than in T1 (control) (0.08%), and Siccibacter, which was more abundant in T1 (control) (0.81%) than in T2 (T. harzianum) (0.003%).
Principal coordinate analysis (PCoA) based on Bray‒Curtis distances was applied to investigate the microbial composition of the samples under different treatments. The results indicate a trend toward a significant separation of the samples as a function of the treatments, although the p value is marginally above the significance threshold (p value = 0.051). The dimensionality reduction performed by PCoA was able to explain 76.07% of the total variability observed in the samples based on the first three main axes (Fig. 14). Although considerable variability was observed within the groups and some overlap between them, post hoc analyses indicated trends of significant compositional differences between treatments, particularly between T3 (P. lilacinum) and T4 (T. harzianum + P. lilacinum) (p value = 0.028), as well as between T1 (control) and T3 (P. lilacinum) (p value = 0.058). These results suggest the existence of distinct patterns in the microbial composition associated with each treatment, reflecting the specific influence of each intervention on the microbial community from the leaves of soybean plants.
Venn diagram analysis (Fig. 15) revealed significant sharing of taxa at the higher taxonomic levels, with 12 phyla and 98 genera identified as common among all treatments. However, the comparison at the level of ASVs showed a distinct pattern, with only 59 ASVs out of a total of 2,207 being shared by all treatments. This result suggested conservation of the main taxonomic groups among the treatments, while the differences in the ASV indicated significant variations in the population composition. Regarding the presence of exclusive taxa, there was a trend in which the control treatment (T1) had the greatest number of unique taxa, followed by treatments T3 (P. lilacinum), T2 (T. harzianum) and, finally, T4 (T. harzianum + P. lilacinum), which were consistent at all the taxonomic levels analyzed.
The structuring of the microbiomes of the samples was inferred from the co-occurrence networks of the different treatments. There were changes in the relationships between the identified genera (Fig. 16). In this sense, treatment T1 (control) had a greater number of correlated genera ("N. of nodes") and correlations ("N. of edges"). Although this treatment had more than twice the amount of bonding compared to the other treatments, treatments T2 (T. harzianum) and T3 (P. lilacinum) presented a greater number of negative bonds ("negative edges"), with 57 and 71, respectively, against only 25 in T1 (control). The characteristics of T1 (control) are mainly due to the presence of a cluster with a large number of correlations, reflecting higher means of connections ("Mean degree") and average and maximum potential of betweenness/Max. betweenness"). This is further reinforced by the high number of key taxa ("main hubs") found for this treatment. In general, treatments with fungal inoculants showed reduced binding, increased negative binding (except for T4 (T. harzianum + P. lilacinum) and a reduction in the number of key taxa, which in turn differed between the treatments. These observations suggest that treatments differentially influence the structure and dynamics of microbial communities, altering the patterns of co-occurrence and the relative importance of certain taxa within the networks.
The highest mean yield value (p < 0.05) was found for treatment T3 (P. lilacinum) compared to treatment T1 (control). The T2 (T. harzianum), T3 (P. lilacinum) and T4 (T. harzianum + P. lilacinum) treatments did not differ (p > 0.05) from each other (Fig. 17).
The correlation between the taxonomic groups showed that some genera had a positive and significant correlation with some plant growth parameters. Specifically, for productivity, the genera Erwinia and Bacillus followed the genus Blautia (Fig. 18).
Treatment T3 (P. lilacinum) was the only treatment that promoted increased productivity compared to the other treatments (Figs. 17 and 19).
The parameters that were significantly influenced by the T3 treatment (P. lilacinum) were the phosphorus concentration in SDM, productivity and nitrogen content (Fig. 20).