Vegetation and soil information of study sites
One grazing exclusion area (GE), and three free grazing areas (FGs) having different vegetation biomass per unit area and representative vegetation composition, were selected in this study. Based on the evaluation criteria of grassland degradation [18] and the vegetation diversity and biomass of four study sites (Table 1), GE was considered as the non-degraded area, and FG1, FG2, and FG3 were initially identified as the slightly degraded area, moderately degraded area, and severely degraded area, respectively. To verify the determination of degradation degree of each area, the correlations between degradation degree and diverse factors about vegetation biomass and diversity were analyzed (Table 1). The results suggested that total plant species richness (SR) and aboveground biomass (AB), Shannon-Wiener diversity index (H), Pielou evenness index (E), litter mass (LM), and the species richness of different functional types (effective species, gramineous species, forb species, and shrub species), and the aboveground biomass of different functional types (effective species, gramineous species, and annual and biennial species) were all significantly related with the determination of degradation degree, further proving the rationality of degradation degree evaluation of the study sites. Because the sampling sites were set in the same steppe, the soil properties and enzyme activities should not be too different, while due to the different grassland management of grazing exclusion and free grazing, there were indeed certain impacts on soil nutrients, especially on nitrogen and phosphorus contents (Table 2). Referring to the relationship between soil and degradation degree, only soil organic matter content (SOM) had significantly negative response (Table 2).
Effect of grazing exclusion on soil bacterial abundance
Long-term grazing exclusion resulted in the obvious increase of absolute abundance of soil bacterial community based on the high-throughput 16S rRNA absolute quantitative sequencing result (Fig. 1). Acidobacteria, and Actinobacteria, and Proteobacteria were the predominant phyla in all soil samples, which accounting for at least 12.8% of the bacterial community. Except for these three phyla, the proportions of Bacteroidetes, Planctomycetes, and Chloroflexi were relatively high in most soil samples, accounting for more than 5% of the bacterial community. Gemmatimonadetes, Thaumarchaeota, candidate division WPS-1, Verrucomicrobia, Nitrospirae, Firmicutes, Armatimonadetes, and Cyanobacteria were present in relatively low proportions, accounting for more than 1%. Ignavibacteriae, Latescibacteria, Candidatus Saccharibacteria, Euryarchaeota, and Elusimicrobia showed lower proportions, with relative abundances of more than 0.1%.
Based on the Kruskal-Wallis test, a total of 3 and 5 phyla exhibited significant differences (P<0.05) in absolute and relative abundance of soil bacterial community, respectively, among different study sites (Figs. 2, 3). Among them, two phyla, Euryarchaeota and Cyanobacteria/Chloroplast both showed significant differences in the absolute and relative quantitative analysis of soil bacterial community among the four study sites. Euryachaeota, Microgenomates, and Verrucomicrobia exhibited high abundance in GE, and low abundance in FGs; Chloroflexi and Firmicutes exhibited relatively low abundance in GE, but high abundance in FGs; Cyanobacteria/Chloroplast exhibited very low abundance in GE and severely degraded area (FG3), relatively higher abundance in FG1 and FG2.
Effect of grazing exclusion on soil bacterial composition
As shown in the Venn map, most of the soil bacterial species in different study sites (accounting for 63.10%-64.98%) were overlapping (Fig. 4). However, considering the different important orders within and between groups, the partial least squares discriminant analysis (PLS-DA) results showed that four study sites could be completely separated from each other, although the FG2 and FG3 were relatively close (Fig. 5). This indicated that grazing exclusion could not only increase the absolute abundance of soil bacterial community, but also influence the composition of soil bacterial community.
Effect of grazing exclusion on the ecological functions of bacterial community
Using PICRUSt software, the COG functional annotations and KEGG pathways were used to predict the ecological functions of bacterial community. A total of 24 COG categories were annotated for the functions; those with relatively high abundance were related to transcription, amino acid transport and metabolism, signal transduction mechanisms, cell wall/membrane/envelope biogenesis, energy production and conversion, carbohydrate transport and metabolism, replication, and recombination and repair, etc. The two-way analysis of variance (ANOVA) results showed that 5 functional annotations were significantly different (P<0.05) among the four study sites, including general functions predicted only, carbohydrate transport and metabolism, signal transduction mechanisms, replication, recombination and repair, and cell motility (Additional file 1). Among these functions, general functions predicted only, and carbohydrate transport and metabolism were more active in FG3 than in GE, while the other three functional annotations with significant differences were the opposite and, being more active in GE than in FG3.
Based on the PICRUSt results, a total of 312 KEGG pathways were involved in this study. The ANOVA results showed that 63 KEGG pathways were significantly different (P<0.05) among the four study sites (Additional files 2, 3). Among them, 3 KEGG pathways with significant differences overlapped with the COG categories with significant differences among the four study sites, which were general functions predicted only, replication, recombination and repair, and cell motility.
According to the comprehensive analysis of the predicted bacterial function, it indicated that long-term grazing exclusion could effectively improve cell motility by increasing the contents of bacterial motility proteins and improving flagellar assembly, thus significantly enhancing bacterial chemotaxis (Additional file 3). In addition, the biosynthesis of isoflavonoids which participated in plant growth, development and stress resistance [19, 20], was significantly higher in GE than in the other three free-grazing grasslands (Additional file 2), thus, it was not difficult to recognize that GE possessed higher glycan biosynthesis and metabolism, and higher activity of glycosyltransferases than FG (Additional file 3). This was because the isoflavonoid biosynthesis required the participation and catalysis of a large number of glycan and glycosyltransferases [21]. This result suggested that long-term grazing exclusion could significantly enhance isoflavonoid biosynthesis by increasing glycan biosynthesis and metabolism and glycosyltransferase activities and thus favor plant growth, development and stress resistance.
Correlations between bacterial community and environmental factors
To understand the effects of environmental factors on soil bacterial community, we examined multiple indices related to vegetation diversity, biomass, soil properties, and enzyme activities, and estimate their correlations with the bacterial community. The Pearson correlation results indicated that no matter the relative abundance or absolute abundance of the bacterial community, these environmental factors all showed significant correlations with them (P<0.05, Table 3). According to the redundancy analysis (RDA) between soil bacterial community and vegetation biomass indexes, the soil bacterial community showed the closest correlations with shrub species aboveground biomass (SAB, Fig. 6a), SR, H, (Fig. 6B), and SOM (Fig. 6c).
Due to the accurate discrimination of grassland degradation degree for four investigated study sites based on vegetation biomass and diversity (Table 1), we introduced the grassland degradation degree (DG) as an important factor, and used it to further analyze the correlation of the soil bacterial phylum with environmental factors. As for vegetation biomass, annual and biennial species aboveground biomass (AAB) and DG were divided into one category, the other factors were grouped as one category (Fig. 6a). Compared with the vegetation diversity factors, DG formed one category by itself, and the annual and biennial species diversity (ASR) was had the worst consistency with other factors (Fig. 6b). As for soil properties, the soil pH value (PH) and soil bulk density (SBD) were grouped with DG, and the other factors were grouped together (Fig. 6c). As for soil enzyme activities, DG was separated from any one of three soil enzyme activities (Fig. 6d).
Through the correlation analysis with diverse environmental factors and soil bacterial community in the phylum level, we found vegetation biomass and vegetation diversity showed similar results of that all bacteria were divided into three big groups showing different responses to various factors: one group which were positively related to most vegetation biomass and diversity factors, but negatively related to DG and AAB, included Euryarchaeota, Microgenomates, candidate division WPS-1, Verrucomicrobia, Proteobacteria, Acidobacteria, Ignavibacteriae, Omnitrophica, Elusimicrobia, and Parcubacteria; one group which were negatively related to most factors, but positively related to DG and AAB, included Thaumarchaeota and Firmicutes; the other group had no significant relations with most factors, such as BRC1, Armatimonadetes, Latescibacteria, Bacterioidetes, Hydrogenedentes, Actinobacteria, Nitrospirae, and Pacearchaeota, and so on. Compared with the correlations with soil properties and soil bacterial community in the phylum level, microorganisms could also be divided into three categories according to their response patterns: one group which were positively related to most soil property factors except for PH, SBD, and DG, included Bacteroidetes, Hydrogenedentes, Armatimonadetes, BRC1, candidate division WPS-1, and Omnitrophica; one group which were negatively related to most soil property factors, included Pacearchaeota, Thaumarchaota, and Firmicutes; the other group did not show significant relations with most factors, or only showed significant relations with very few factors, such as Parcubacteria, Ignavibacteriae, Elusimicrobia, Euryarchaeota, Microgenomates, Verrucomicrobia, Acidobacteria, Proteobacteria, and so on. Based on the correlation analysis between soil enzyme activities and soil bacterial community in the phylum level, the phyla which responded significantly to them were greatly reduced. All microorganisms could be divided into four categories: one group were positively related to the soil enzyme activities, including Omnitrophica, candidate division WPS-1, Armatimonadetes, Bacteroidetes, Hydrogenedentes, and BRC1; one group were negatively related to them, including Thaumarchaeota, Firmicutes, Latescibacteria, Actinobacteria, and Nitrospirae; one group were only negatively related to DG and urease activity, including Euryarchaeota, and Microgenomates; the other group had no significant relations with the soil enzyme activities, such as Parcubacteria, Ignavibacteriae, Elusimicrobia, Proteobacteria, Verrucomicrobia, Acidobacteria, and so on.
It was worth nothing that Bacteroidetes, Hydrogenedentes, Armatinomadetes, and BRC1 and Pacearchaeota did not respond to vegetation factors, but they had remarkable responses to soil properties and enzyme activities. However, the bacterial phyla, such as Euryarchaeota, Microgenomates, Verrucomicrobia, Acidobacteria, Proteobacteria, Parcubacteria, Ignavibacteriae, and Elusimicrobia, which had significantly positive responses to vegetation factors, had no or few responses to soil properties and enzyme activities. To be noted, Latescibacteria, Actinobacteria, and Nitrospirae did not significantly respond to vegetation factors and most soil properties, but significant-negatively to available potassium content (AK), and soil enzyme activities.