Physicochemical properties of rhizosphere soils and tobacco leaves
A total of 17 physicochemical parameters at ten different tobacco fields were measured (Table 1). The value of pH ranged from 5.23±0.16 (YX3) to 7.28±0.45 (QJ15) and all samples, except QJ15, were weakly acidic. The heatmap of the rhizosphere soil physicochemical parameters showed that the 17 parameters could be divided into four groups (Fig. 1A). pH, Mg, TN, and OM were in one group and their value (pH) and contents were higher in QJ15. Zn, Cl, and Ca were in a group and their contents were relatively high in QJ15, QJ13, and WS18. Fe, Cu, TP, Mn, and S were grouped together and their contents were high in the YX2 sample. TK, P, AP, and K made the last group and all the four parameters were high in KM1.
For tobacco leaves, seven parameters were measured and the results are shown in Table 2. The heatmap based on the seven parameters revealed three groups of the seven parameters (Fig. 1B). ST, TS, and RS formed a group, while NT and SC formed another group, and TN and PO formed the last one. ST, TS, and RS had higher contents in KM14. NT and SC had relatively higher content in YX2, while TN and PO had higher contents in BS21. For WS18 and YX3, the three repeat samples in each site were not clustered together.
Illumina sequencing and bacterial community structure
The Illumina Hiseq sequencing generated a total of 692,836 clean reads with an average length of 414.6 bp. The reads number of each sample was resampled to 19,548. The OTU numbers of the 30 samples ranged from 1365 (QJ13-3) to 2034 (DL22-1) (Table S1). The rarefaction curve based on the observed species and reads showed that the numbers of observed species for each sample were almost saturated and were sufficient for microbial community analysis (Fig. S1). The mean relative abundances of the top 20 phyla in each sample site are shown in Fig. 2A. Proteobacteria possessed the highest relative abundance in all samples, ranging from 32.50% (DL21-1) to 46.81% (YX3-2) (Table S2). The three dominate phyla were Proteobacteria, Acidobacteria, and Actinobacteria, which accounted almost 80% of the relative abundance. At the genus level, about 40% of the reads could not assigned to a specific genus (Fig. 2B). The top five genera were Gp6 (6.92%–26.04%), Gemmatimonas (2.69%–13.79%), Gp4 (1.91%–8.11%), Gp3 (1.95%–7.96%), and Sphingomonas (1.54%–7.42%) (Table S3). However, the most abundant genus in each sample was different.
The OTUs in each sample was compared by Venn diagram (Fig. 3). A total of 170 core OTUs were shared by all samples. CX8-1 possessed the most specific OTUs at 15 while YX3-3 possessed no specific OTUs. The specific OTUs of each sample were very few and, for most of them, were lower than 10. The core OTUs and the specific OTUs together with the OTU sequences are listed in Table S4.
The alpha diversity of all samples are listed in Table S5. The highest mean observed species was found in DL22 at 1935 ± 33.79, while the lowest was QJ13 at 1351 ± 19.61 (Fig. S2A). Similarly, DL22 had the highest PD whole tree at 100.04 ± 2.54, while QJ13 had the lowest at 82.44 ± 1.74 (Fig. S2B). All samples had a good coverage higher than 0.97, indicating that the sequence numbers were enough for each sample to perform microbial community analysis.
To illustrate the differences of the microbial communities of the ten sample sites, principal component analysis (PCA) based on the OTUs demonstrated that samples from the same site could have different microbial community profiles (Fig. 4). Samples from Yuxi city YX3 had a more similar microbial community with KM11 than YX2 while KM14 had a more similar microbial community with BS21 than KM11. YX2, QJ15, CX8, and QJ13 had a relative specific microbial community. At the OTU level, PC1 explained 21.32% and PC2, 16.11% of the total variation (Fig. 4).
Relationship among soil properties, rhizosphere microbial communities and tobacco leaf properties
The correlation analysis based on the Spearman method between the physicochemical parameters of tobacco leaves with the soil physicochemical parameters and rhizosphere microbial community at the class level is shown in Fig. 5. Acidobacteria_Gp3 and Flavobacteria had a significant (P < 0.05) positive correlation with RS and ST. K showed a significant positive correlation with TS and an extreme significant (P < 0.01) positive relationship with ST. Meanwhile, K also possessed a significant negative correlation with PO and an extreme significant negative correlation with NT. Cl also showed an extreme significant negative correlation with ST.
RDA analysis indicated that pH, OM, SN, TN, TP, AP, Cu, Fe, K, and Mg were significantly associated with microbial community diversity (forward selection with a Monte Carlo test, P < 0.05). The first two axes explained 25.42% of the microbial community diversity information (Fig. 6A). RDA was also used to assess the relationship between the physicochemical parameters of tobacco leaves with the soil physicochemical parameters and rhizosphere bacterial microbial communities of the top 20 classes (Fig. 6B). Bacteria of Actinobacteria, Deltaproteobacteria, Acidobacteria_Gp4, Acidobacteria_Gp3, Acidobacteria_Gp1, and the soil physicochemical parameters of K, Fe, and Ca were significantly associated with tobacco leaf physicochemical parameters, based on the Monte Carlo test (P < 0.05). The first two axes explained 54.95% of the total variance.
The rhizosphere microbial community similarity of different cultivars, different landforms, different soil types, different altitudes, and different rotation crops were analyzed using the ANOSIM method. The results showed that the landform, altitude, and rotation crop had a significant influence in the rhizosphere microbial community, while different cultivars and different soil types showed no significant difference in the rhizosphere microbial community (Table 3).
Variation partitioning of tobacco leaves physicochemical parameters
To investigate the contribution of the rhizosphere microbial community and soil physicochemical characteristics to tobacco leaf property variation, variance partitioning analysis was conducted based on the RDA model and the results are shown using a variation partitioning diagram (Fig. 7). Fifteen classes of bacteria (Bacteroidetes_incertae_sedis, Thermoplasmata, Verrucomicrobiae, Armatimonadetes_gp5, BRC1_genera_incertae_sedis, Candidatus Hydrogenedens, Latescibacteria_genera_incertae_sedis, WPS-1_genera_incertae_sedis, Nitrososphaerales, and six unidentified class) and six soil physicochemical characteristics (pH, SN, Zn, TP, TK, and Ca) were selected as explanatory variables through a forward selection procedure. The rhizosphere microbial community and the soil physicochemical characteristics together could explain 82.68% of the variation in the tobacco leaf properties. The pure effect of the rhizosphere microbial community explained 56.99% of the variation while the pure effect of soil physicochemical characteristics explained 14.77% of the variation. We found that 10.98% of the variation could be explained by the rhizosphere microbial community and soil physicochemical characteristics simultaneously.