3.1 Soil microbial community carbon metabolism
The AWCD of all samples increased with incubation time (Fig. 1A) and of the rhizosphere microorganisms of poisonous plant species was higher than the bulk (non-rhizosphere) soil in the four levels of degraded grassland. This indicated that rhizosphere microorganisms of poisonous plants had higher metabolic activity and could make better use of the carbon source in the Biolog-Eco plates. The AWCD reached maximum rate within 48–96 h, and the 72 h measurements were used for PCA (Fig. 1B). The values of the first principal component in the four degraded grasslands of S, H, M, and L were 54.2%, 50.5%, 48.5%, and 47.0%, respectively, and indicated that the microorganisms of different plant rhizospheres and bulk soil used carbon sources differently (Fig. 1B).
There were 12, 13, 11, and 13 carbon sources that were significantly (P < 0.05) related to principal component 1 in S, H, L, and M, respectively (Additional file 1: Table S3). Except for D-xylose, which was correlated negatively with L (P < 0.05), all were correlated positively. The microorganisms in S had greater metabolic activity for amino acids than other carbon sources. The rhizosphere microorganisms in each degraded grassland used six more types of carbon sources than bulk soils, and the utilization efficiency of amino acids by soil microbes was greater than for other carbon sources (P < 0.05) (Table 1). In L grassland, bulk soil had a higher utilization efficiency for carbohydrates, while rhizosphere soil had a higher utilization efficiency for phenolic acids (P < 0.05) (Table 1).
Table 1
Soil utilization of different carbon sources by microorganisms in 72 h cultivation
Samples | Amines | Phenolic acids | Carboxylic acids | Carbohydrates | Amino acids | Polymers | P value |
S-Nr | 0.226 B | 0.135 Cb | 0.116 Cc | 0.249 Bc | 0.340 Ab | 0.258 B | P < 0.001 |
S-Sg | 0.352 B | 0.518 ABa | 0.401 Ba | 0.517 ABab | 0.641 Aa | 0.390 B | P = 0.040 |
S-Oo | 0.364 B | 0.363 Bab | 0.229 Bbc | 0.656 Aa | 0.562 Aa | 0.349 B | P < 0.001 |
S-Pk | 0.325B | 0.498 ABa | 0.330 Bab | 0.543 ABab | 0.639 Aa | 0.459 AB | P = 0.095 |
S-Lv | 0.338 BC | 0.248 Cb | 0.271 Cab | 0.409 Bbc | 0.592 Aa | 0.273 C | P < 0.001 |
| P = 0.232 | P = 0.015 | P = 0.010 | P = 0.003 | P < 0.01 | P = 0.153 | |
L-Nr | 0.306 BC | 0.274 C | 0.251 C | 0.632 A | 0.487 AB | 0.314 BC | P = 0.008 |
L-Mk | 0.558 BC | 0.841 A | 0.430 C | 0.587 BC | 0.769 AB | 0.500 C | P = 0.009 |
| P = 0.049 | P = 0.006 | P = 0.148 | P = 0.762 | P = 0.009 | P = 0.017 | |
M-Nr | 0.207 B | 0.126 B | 0.145 Bc | 0.281 ABb | 0.399 Ac | 0.269 ABd | P = 0.020 |
M-Al | 0.421 B | 0.487 B | 0.202 Cbc | 0.402 Bab | 0.654 Aa | 0.358 Bcd | P = 0.001 |
M-Ef | 0.382 | 0.468 | 0.253 abc | 0.443 ab | 0.443 bc | 0.366 cd | P = 0.616 |
M-Mk | 0.379 B | 0.357 BC | 0.272 Cabc | 0.523 Aab | 0.550 Ab | 0.493 Aab | P < 0.001 |
M-Ap | 0.419 BC | 0.514 ABC | 0.338 Cab | 0.658 ABa | 0.666 Aa | 0.592 ABa | P = 0.041 |
M-Lv | 0.484 | 0.460 | 0.409 a | 0.647 a | 0.629 a | 0.412 bc | P = 0.119 |
| P = 0.202 | P = 0.062 | P = 0.024 | P = 0.040 | P = 0.002 | P = 0.001 | |
H-Nr | 0.229 c | 0.157 c | 0.177 c | 0.311 c | 0.344 d | 0.281 d | P = 0.088 |
H-Al | 0.329 Babc | 0.229 Cbc | 0.275 BCabc | 0.304 BCc | 0.505 Abc | 0.354 Bcd | P < 0.001 |
H-Ef | 0.315 Cabc | 0.256 Cbc | 0.298 Cabc | 0.627 Aa | 0.508 ABbc | 0.383 BCbcd | P = 0.002 |
H-Sg | 0.287 Cbc | 0.399 Bab | 0.235 Cbc | 0.529 Aabc | 0.526 Abc | 0.376 Bbcd | P < 0.001 |
H-Mk | 0.354 abc | 0.442 ab | 0.338 abc | 0.534 abc | 0.615 ab | 0.440 abc | P = 0.345 |
H-An | 0.262 ABbc | 0.217 Bbc | 0.179 Bbc | 0.353 Abc | 0.384 Abc | 0.287 ABd | P = 0.021 |
H-Ad | 0.486 a | 0.416 ab | 0.272 abc | 0.553 ab | 0.523 cd | 0.444 abc | P = 0.057 |
H-Ap | 0.440 ab | 0.540 a | 0.348 ab | 0.699 a | 0.740 a | 0.536 a | P = 0.074 |
H-Lv | 0.427 Bab | 0.403 Bab | 0.429 Ba | 0.716 Aa | 0.662 Aab | 0.491 Bab | P = 0.001 |
P value | P = 0.04 | P = 0.024 | P = 0.044 | P = 0.003 | P = 0.001 | P = 0.003 | |
Note: Values with different lowercase letters in the same column indicate significant differences between them (P < 0.05). Values with different capital letters in the same row indicate significant differences between them (P < 0.05). L, lightly degraded grassland; M, moderately degraded grassland; H, heavily degraded grassland; S, degraded sown grassland. Nr, non-rhizosphere soil. Al, Ajuga lupulina; Ef, Euphorbia fischeriana; Sg, Sphallerocarpus gracilis; Oo, Oxytropis ochrocephala; Mk, Morina kokonorica; Pk, Pedicularis kansuensis; An, Artemisia nanschanica; Ad, Artemisia dubia; Ap, Aconitum pendulum; Lv, Ligularia virgaurea. |
3.2 Convergence and microbial composition in bulk soil and rhizosphere
The Illumina MiSeq 16S sequence determined the bacterial diversity in the rhizosphere soil of poisonous plant species and bulk soil. A total of 12,282 operational taxonomic units (OTUs) were identified in 2017 (5923, 8031, 6877, and 3110 in S, H, L and M, respectively) and 5447 OTUs were identified in 2018 (2730, 2525, 1992 and 2898 in S, L, M and H, respectively). OTUs were classified into 44 microbial phyla in 2017 and 41 microbial phyla in 2018 with Proteobacteria, Actinobacteria, Acidobacteria and Planctomycetes the four most abundant in all soil groups (Fig. 2). The bulk soil of L grassland had the lowest α-diversity indices of bacterial and fungal communities across all samples (Kruskal-Wallis with Kruskal.test, P < 0.05) (Fig. 3a).
If roots establish stable associations with microbial communities across different habitats, a strong host-filtering effect on habitats is expected. Analysis of microbial community structure based on average Bray-Curtis distances across the 4 degraded grasslands revealed that bacterial and fungal communities in bulk soil and in rhizosphere clustered by grassland type (Figs. 3a and 3c). This pattern was corroborated by permutational multivariate analysis (PERMANOVA) of variance with Bray-Curtis distances (Adonis function in R library vegan), which indicated that species × habitat explained more of the variation in microbial community (Adonis: degrees of freedom (d.f.) = 21; bacterial: coefficient of determination (R2) = 0.53; P < 0.001; fungal: R2 = 0.32; P = 0.59) than compartment (d.f. = 1; bacterial: R2 = 0.04; P < 0.001; fungal: R2 = 0.01; P = 0.62), habitat (d.f. = 3; bacterial: R2 = 0.16; P < 0.001; fungal: R2 = 0.05; P = 0.9 1) or species (d.f. = 10; bacterial: R2 = 0.27; P < 0.001; fungal: R2 = 0.15; P = 0.29) (Fig. 3b). The bacterial community was affected to a greater extent than the fungal community by habitat and species.
PCoA ordinates based on Bray-Curtis distances revealed marked differences in soil bacterial and fungal communities in the 4 habitats. The bacterial community in M and the fungal community in S, in particular, were separated clearly from the others. These differences appeared in the same plant species in different habitats, suggesting divergence in the microbial community composition in plant rhizosphere (Fig. 3c).
Nitrate nitrogen (NO3-N) in L grassland was lower than the other three habitats (P < 0.05), while ammonium nitrogen (NH4-N) exhibited an opposite trend (Fig. 4a). The concentrations of TN, AP and TOC in bulk soil of the four habitats were lower than in the rhizosphere soil (Additional file 1: Table S4). Redundancy analysis showed that soil physico-chemical properties explained 28.3% of the variation in rhizosphere bacterial communities (Fig. 4b) (P < 0.05), but had no effect on fungal communities. The bacterial communities were correlated positively with TP, pH and AP, but negatively with TN, TOC, NO3, NH4 and WC (Fig. 4b).
3.5 Relationship of modules of root exudates and rhizosphere bacterial community
In total, 3613 effective OTUs and 1092 root exudates were obtained for network construction after screening. After the mad function removed unimportant values, the sample clustering results were consistent. There was similarity in results between OTU and root exudates, in which L. virgaurea was significantly different (P < 0.05) from the other poisonous plant species (Figs. 7 and S4). The rhizosphere microorganisms and root exudates of E. nutans and P. kansuensis were similar. Fifteen modules were combined in the clusterings in 16S OTUs (Additional file 1: Table S5; Additional file 1: Figure S4). A high average connection degree emerged inside the turquoise module (39.25), which contained 1769 OTUs (Additional file 1: Table S5). This module was correlated with all soil groups and had a significant negative correlation with L. virgaurea. The strong correlation between the module eigengenes from the WGCNA network analysis and each plant species indicated that the plant rhizosphere accumulated microorganisms/root exudation in the module. After removing the turquoise module, the OTU correlation network filtering with threshold (0.2) showed that the connectivity within the module was higher, especially in the blue, midnight blue and red modules (Fig. 7A and C). The tan, salmon and cyan modules had few OTUs (Fig. 7C). P. kansuensis correlated with the red and tan modules (P < 0.05), S-Lv-MD with midnight blue, S-Lv-HD with green, H-Lv with black, blue and yellow, L-Lv with magenta, green yellow and pink, and M-Lv with cyan, brown, and purple (Figs. 7A, 7C, and S4).
Most root exudates contained benzene ring structures, including mainly esters, ketones, aromatic acids, alkenes, amino acids, phenols and derivatives. Poisonous plant species correlated with corresponding modules; P. kansuensis with pink (P < 0.05), M. kokonorica with green (P < 0.05), and A. pendulum with blue (P < 0.05). M-Lv correlated with black, magenta, and yellow modules (P < 0.05), L-Lv with red (P < 0.05), S-Lv-HD with brown (P < 0.05), and S-Lv-MD and S-Lv-LD with turquoise (P < 0.05) (Additional file 1: Figure S5). There was no module related to E. nutans and K. pygmaea, indicating that the content of compounds in the root system of poisonous plant species were higher than non-poisonous grasses. It was evident that root exudates were plant species specific and each species had a unique module (Figs. 7B and 7D).
The correlation analysis heat map of the root exudate module and the bacterial module are presented in Additional file 1: Figure S6. In the OTU module corresponding to poisonous plant species, more than 85% had positive correlations with roots exudates. The pink (exudate) and red (OTU) modules correlated with P. kansuensis and L. virgaurea; magenta (exudate) with brown and purple (OTU) in M-Lv (P < 0.05); yellow (exudate) with cyan (OTU) in M-Lv (P < 0.05); red (exudate) with green yellow (OTU) in L-Lv (P < 0.05); and turquoise (exudate) with brown and green (OTU) in S-Lv (P < 0.05) (Figs. 8A and S6). It demonstrated that microorganisms had preferences for plant species exudates and, consequently, plant species could manipulate microorganisms through exudates from roots.
Figure 8A outlines the network between OTU and exudate modules. There were two networks in M-Lv. One network was cyan (OTU) and yellow (exudate), with 7 phyla (n = 28) in the cyan module (OTU) (Fig. 8Aa), and Actinobacteria (25%) and Proteobacteria (43%) being the main phyla. Twelve genera in the module were related to exudates (P < 0.05) (Additional file 1: Table S6), including Acidibacter, Pseudonocardia, Bauldia, Stenotrophomonas and Nocardioides. Approximately 60% of the compounds in the yellow module were alkaloids, lupinic acid, sesquiterpene, artemisinin and other secondary metabolites (Additional file 1: Table S6), and 70% of the metabolites in the magenta module were alkaloids, coumarins and derivatives, especially alkaloids. The second network was OTU modules (brown and purple) and magenta (exudate) (Fig. 8Aa). The brown module had 14 main phyla (n = 52) (P < 0.05), of which Proteobacteria, Planctomycetes, Bacteroidetes and Chloroflexi were 26%, 16%, 12% and 10%, respectively. It contained a large number of bacteria with nitrification and denitrification abilities, such as Flavobacterium, Luteimonas, Arenimonas and Planctomyces. There were 6 phyla (n = 21) in the purple module, mainly Chloroflexi (19%), Plantomytomytes (24%), and Proteobacteria (24%) (Figs. 8Aa and 8C; Additional file 1: Table S6).
There were 17 OTUs and 34 exudates which were correlated with P. kansuensis (Fig. 8Ab; Additional file 1: Table S6). The 17 OTUs in the red module were divided into 5 phyla, including mainly Proteobacteria (41%) and Bacteroidetes (41%), while Alphaproteobacteria and Sphingobacteriia were abundant classes. There were 12 genera, including Aureimonas, Segetibacter, Methylobacterium, and Pleomorphomonas that related significantly to rhizosphere exudates (P < 0.05). Most of the root exudates were insecticides with insecticidal effects and some soil microorganism inhibitors (about 60%), such as carbofuran, cycloheximide and flusilazole.
For L-Lv, 11 phyla (n = 44) were included in the green yellow module, and Planctomycetes and Firmicutes accounted for 29% and 27%, respectively (Figs. 8Ac and 8C; Additional file 1: Table S6). Nineteen genera were sequenced out that promoted plant growth (Clostridium, Lactobacillus, Acidibacter) and also some pathogenic bacteria (Methylocella, Aquicella, Veillonella, Enterococcus) were included. About 80% of the exudation in the red module that related significantly to OTUs were intermediate metabolites, such as terpenes, alkaloids, sugars, and fatty acid conjugates.
The blue (OTU) module was related significantly to the grey module (exudates) (Fig. 8Ad; Additional file 1: Table S6). For H-Lv, 99% of the compounds in the grey module was nitrogen-containing heterocyclic compounds, mainly amino acids. The blue module contained 94 OTUs, mainly Proteobacteria (38%), Plantomytomytes (19%) and Acidobacteria (10%). The module contained bacteria with nitrogen fixation, nitrification and denitrification abilities, such as Mesorhizobium, Steroidobacter, Pirellula and Rhizobacter (Figs. 8Ad and 8C; Additional file 1: Table S6).
For S-Lv, 8 and 9 phyla in the green module (OTU) correlated with the turquoise (exudate) and brown (exudate) modules, respectively, among which Proteobacteria was the main phylum and Alphaproteobacteria the main class. It contained Mycobacterium, Inquilinus, Nitrospira, Pirellula, and Singulisphaera (Figs. 8Ae and 8C; Additional file 1: Table S6). Of the compounds in the turquoise module, 80% included alkaloid, sesquiterpene, coumarin and phenolic compounds, and some compounds that inhibit the growth of mold.
Overall, S-Pk, S-Lv, L-Lv, M-Lv, H-Lv had 5, 10, 11, 14 and 10 phyla, respectively, in the bacterial communities correlated with root exudates. The relative abundance of Actinobacteria (0.50%, 0.71% and 0.69%), Bacteroidetes (0.17%, 0.47% and 0.22%) and Proteobacteria (0.99%, 3.52% and 2.53%) were contained in L-Lv, M-Lv and H-Lv, respectively, and the relative abundance was highest in M-Lv. Some unique dominant phyla were identified in different habitats. For relative abundance of phyla > 0.1%, there were Cyanobacteria (1.15%) and Firmicutes (2.82%) in L-Lv, Chloroflexi (0.11%) in M-Lv; Gemmatimonadetes (0.26%) in H-Lv; Bacteroidetes (0.16%) in S-Pk; Acidobacteria (0.6%), Chloroflexi (0.12%) and Proreobacteria (0.69%) in S-Lv, and Acidobacteria (0.88% and 0.55%) and Planctomycetes (0.23% and 0.37%) in M-Lv and H-Lv (Fig. 8B).