Sorghum yields
Table 1 shows the total dry biomass and grain yields of sorghum. Total biomass and grain yields obtained in the NPK and P-Comp-Soil treatments were not significantly different. They gave significantly higher yields than the other treatments, except that there was no significant difference between P-Comp-soil and the Control for biomass yields and between P-Comp-Soil, P-Comp, and Control for grain yields. In general, Sorghum Straw, Comp, P-Comp, and BPR treatments influenced less the sorghum yields. The percentage of good ears (ears well-filled with grains) were significantly higher in the NPK, P-Comp-Soil, P-Comp, and Comp treatments, with slightly higher values in NPK and P-Comp-Soil. The NPK treatment gave 4.43 t ha-1 of total dry biomass, 0.86 t ha-1 of dry grain, while 72.51% of ears were filled with grains. For the P-Comp-Soil treatment, these values were 3.87 t ha-1, 0.82 t ha-1, and 72.23%, respectively.
Chemical properties of the rhizosphere soil
The two-way ANOVA applied to the sampling days and treatments showed no significant interaction between these two variables for soil TN, TC, and the sum of exchangeable cations (data not shown). Therefore, the mean values of these three parameters across the three sampling days were considered and are shown in Table 2. Significant differences among treatments were observed in TN and the sum of exchangeable cations. Compared to the Control, Sorghum Straw, and BPR treatments, the soil contained a significantly higher total nitrogen when fertilized with the P-Comp-Soil, P-Comp, and Comp treatments. However, the nitrogen content in the soil amended with Comp and P-Comp was similar to the NPK treatment. Also, the soil fertilized with P-Comp-Soil contained significantly higher exchangeable cations (1.08 cmolc kg-1 dry soil), while the Control and BPR treatments led to the lowest contents with 0.88 soil and 0.87 cmolc kg-1 dry soil, respectively. Table 2 also shows that TN and TC were significantly higher at the early sorghum growth stage (52nd day) and decreased with the cultivation’s progress.
A significant interaction existed between the sampling days and the treatments for the pH, C/N, and available P (data not shown). Thus, these parameters were analyzed separately for each sampling day following one-way ANOVA (Table 3). For the pH, significant differences were found between treatments at the 52nd and 93rd days and not at harvest (115th day). Although the soil was, in general, slightly acidic, the acidity was significantly less under P-Comp-Soil, P-Comp, Comp, and Sorghum Straw treatments compared to Control and NPK on the 52nd day. At this period, the pH of the BPR treatment (pH = 5.43) was significantly lower than that of P-Comp-Soil (pH = 5.60) but was the same as that of P-Comp, Comp, and Sorghum Straw. In contrast, the soil amended with the BPR treatment was significantly less acidic (pH = 5.90) than the remaining six treatments on the 93rd day. At this second soil sampling day, the pH of the soil fertilized with NPK was significantly lower than that of Sorghum Straw and BPR treatments. The difference among treatments for the C/N ratio was only significant on the 93rd day, where C/N (9.52) was significantly higher in the Sorghum Straw treatment than all remaining six treatments.
The soil available P (Bray2-P) content followed contrasting patterns on the three sampling periods (Table 3). On the 52nd day, it was significantly higher in NPK (32.81 mg kg-1 dry soil) and Sorghum Straw (28.49 mg kg-1 dry soil) than the other treatments. However, it was not significantly different between the Sorghum Straw and the BPR (26.98 mg kg-1 dry soil) treatment. The Control and the P-Comp treatments recovered fewer amounts of available P at this period with 9.39 and 12.06 mg kg-1 dry soil, respectively. In contrast, the pattern changed on the 93rd day when Comp, Sorghum Straw, and BPR-treated soil contained significantly higher P with 34.20, 31.89, and 30.32 mg kg-1 dry soil. Sorghum Straw and BPR-treated soil showed similar levels of available P to the NPK treatment with 26.27 mg kg-1 dry soil, while the Control still contained significantly less available P. On the 115th day, the soils treated with Sorghum Straw, Comp, and BPR showed significantly higher P levels corresponding to 35.50, 34.03, and 30.23 mg kg-1 dry soil, almost like the 93rd day. There was no significant difference in available P between BPR and NPK and NPK, P-Comp-Soil, and P-Comp. We observed an increasing trend in soil P throughout the cultivation period in the Sorghum Straw, Comp, P-Comp, BPR, and a decrease in NPK and P-Comp-Soil treatments, although we did not perform statistical analysis.
Abundance of microbial genes in the rhizosphere soil
The two-way ANOVA showed significant interactions (data not shown) between sampling days and treatments for the abundance of the soil microbial genes gcd, AMF, and pstS. Therefore, we described these genes on each sampling day following one-way ANOVA, and the results are reported in Table 4. A significant difference among treatments for the gcd gene was observed only on the 52nd day. At this period, the soil amended with P-Comp-Soil contained a significantly higher abundance of the gcd gene than the other treatments. The soil of the remaining treatments hosted a similar abundance of gcd except that NPK had a significantly lower abundance of the gene than the control.
On the 52nd day, the soil that received P-Comp contained higher copy numbers of AMF than the group formed by the Control, Sorghum Straw, BPR, and NPK treatments, which showed the least AMF abundance. However, there was no significant difference in soil AMF among P-Comp, P-Comp-Soil, and Comp. In contrast, Sorghum Straw-amended soil recovered significantly higher copy numbers of AMF on the 115th day (harvest) than all other treatments. It was followed by Comp and P-Comp, while AMF was significantly less abundant under the Control, P-Comp-Soil, BPR, and NPK treatments. We also noticed an increase (about 3-fold) in AMF abundance in the soil at harvest compared to the 52nd day.
The pstS gene was also significantly different among treatments on the 52nd and 115th days, similarly to AMF. It was significantly higher in soils treated with P-Comp-Soil and BPR on the 52nd day than the other five treatments that did not show differences. At harvest, the soil amendment with P-Comp led to a significantly higher abundance of the pstS gene than the remaining six treatments that showed no significant difference.
The two-way ANOVA was not significant between sampling days and treatments for phnX, phoD, ITS, 16S rRNA, entA, pqqE, and aphA genes (data not shown). Thus, the mean values of the three sampling periods were considered for each treatment (Table 5). The treatments showed a significant (p < 0.05) effect for the abundance of phnX, ITS, 16S rRNA, and aphA genes in soil. It was highly significant (p < 0.01) for the pqqE gene. The three compost treatments (Comp. P-Comp, P-Comp-Soil) promoted the proliferation of microorganisms harboring the phosphonatase phnX gene more significantly than the other treatments. The abundance of phnX gene was significantly lower in the NPK treatment. Soil fungi (ITS) were significantly more abundant in the NPK treatment (4.85 107 copies g-1 dry soil) than in the other treatments. This treatment was followed by the P-Comp-Soil/Comp/BPR group. The Control soil contained the lowest abundance of fungi (3.37 107 copies g-1 dry soil), which was not significantly different from the Sorghum Straw treatment. In addition, the P-Comp-Soil treatment led to a significantly higher population of soil bacteria (5.17 108 copies of 16S rRNA g-1 dry soil) than the Control, BPR, and the NPK treatments. However, it did not show significant differences with the Sorghum Straw, Comp, and P-Comp treatments. The soil of these three treatments contained similar numbers of total bacteria to the Control and the NPK treatments. Pyrroloquinoline quinone gene (pqqE), a co-factor of the gcd gene involved in phosphate solubilization, was significantly more abundant in soils treated with Comp, P-Comp-Soil, and NPK. The BPR treatment generated the lowest population of soil microbes with the pqqE gene. Compared to the remaining six treatments, the P-Comp-Soil treatment significantly enhanced the copy numbers of the acid phosphatase gene (aphA) (4.98 105 copies g-1 dry soil). It was followed by the P-Comp treatment with 2.91 105 copies g-1 dry soil.
Table 5 showed a highly significant effect (p < 0.01) of the sampling period for the abundances of phnX, phoD, and entA, and a significant effect (p < 0.05) for that of the pqqE gene in soil. The overall soil’s phosphonatase phnX gene increased with the cultivation progress to be significantly higher at harvest (5.18 105 copies g-1 of dry soil). A similar trend was observed for the siderophore enterobactin gene (entA), although the overall abundance on the third sampling period (115th day) was not significantly different from that on the first sampling period (52nd day). In contrast, the alkaline phosphatase gene (phoD) and the pqqE genes were significantly more abundant in soil on the 52nd day than on the 93rd and 115th days. They decreased in number with the progress of the cultivation.
Multivariate analysis: interactions of soil chemical and microbiological properties with sorghum yields
Principal Component Analysis (PCA) was performed to understand the relationships between soil chemical and microbiological properties obtained during the three sampling periods and sorghum yields (dry biomass and dry grains) using the correlation matrix of Pearson. The factors loadings of different variables on PC1, PC2, PC3, PC4, PC5, and Eigenvalues (standard deviation, proportion variation, cumulative variation) were shown in Table S1. These first five components explained 81% of the observed variation. However, PC1, PC2, and PC3, which accounted for 65.8% of the variation, were discussed. The cumulative variation of PC1 (29.5%) and PC2 (22.7%) was 52.2% (Fig. 1a). PC1 and PC2 separated the variables into two groups. On the negative side of PC1, sorghum yields positively strongly correlated with total carbon (TC), acid phosphatase (aphA), alkaline phosphatase (phoD), total nitrogen (TN); and moderately with inorganic phosphorus-solubilizing microbes harboring the gcd and pqqE genes, phosphate specific transporter pstS, exchangeable cations. These variables were strongly influenced by the first sampling period (52nd day), and more by the P-Comp-Soil treatment during this period. The remaining variables clustered on the positive side of PC1 were highly influenced by the third sampling period (115th day). Bray-2 P strongly positively correlated with total fungi and moderately with soil pH, arbuscular mycorrhizal fungi (AMF), phosphonatase (phnX). The P-Comp-Soil treatment strongly influenced the population soil AMF and phnX at harvest. When PC1 and PC2 are plotted, the second sampling (93rd day) does not influence the distribution of variables. The highest factor loading in PC1 was obtained for the variable pH (0.343). It was followed by AMF (0.283), phnX (0.276), and total fungi (0.218). The lowest factor loadings for negative correlation were -0.391 (TN), -0.335 (phoD), -0.327 (TC), -0.313 (pqqE), -0.236 (aphA), while that of dry biomass yield and dry grain yield were -0.153, -0.141, respectively. The factor loadings on PC2 were all negative, with the lowest obtained for variabale exchangeable cations (-0.421), pstS (-0.355), 16S rRNA (-0.351).
PC3 accounted for 13.6% of the variation, and when plotted with PC1 (29.5%), both components explained 43.1% of the observed variration (Fig. 1b). On the PC1/PC3 plan, the three sampling periods were separated. The second sampling, which was not expressive on PC1/PC2, strongly influenced the dry biomass yield, dry grain yield, exchangeable cations, Bray-2 P, entA, and C/N, showing strong positive relationship of yield variables with exchangeable cations, with a strong positive influence of the P-Comp-Soil and NPK treatments, and moderate of P-Comp. Bray-2 P, C/N, and entA negatively correlated with yield variables and were positively correlated among them with the influence of BPR, sorghum straw, and Comp. The third sampling (115th day) did not influence the yield variables. However, total bacteria and fungi were positively related to P-Comp-Soil, while P-Comp, mainly influenced the phnX, AMF, and the pH. Dry biomass yield and dry grain yield of sorghum showed loading factors of 0.538, 0.533, respectively, on PC3. They were followed by total fungi (0.390) and total bacteria (0.195).
The Spearman’s correlation (Table S2) confirmed that sorghum biomass was moderately negatively correlated with entA, AMF, pH, and Bray2_P, and moderately positively correlated with total nitrogen, total carbon, and exchangeable cations. The grain yield was moderately negatively correlated with AMF, pH, and Bray2_P and positively correlated with total bacteria (16S rRNA), total nitrogen, total carbon, and exchangeable cations. Among the other existing interactions, we noticed that the abundance of soil bacteria was moderately positively correlated with AMF, pH, and exchangeable cations but weakly with C/N. The soil fungi moderately negatively correlated with pqqE and weakly negatively correlated with nitrogen and carbon. It correlated moderately positively with AMF, weakly with pH, and strongly with bacteria. The phosphonatase phnX gene strongly positively correlated with AMF and pH, moderately positively with total fungi and total bacteria, but moderately negatively correlated with phoD, and pqqE.