The Loess Plateau of China is located in the upper middle stretches of the Yellow River and includes the 624,000 km² loess deposition area west of the Taihang Mountains, east of the Riyue Mountain, north of the Qin Mountains, and south of the Yin Mountains. The plateau is the most severely eroded area in the world; over 60% of the land is affected by erosion due to natural (topography, climate, soil, and vegetation) as well as anthropogenic (unreasonable production and construction) causes, and the average annual soil loss is about 2000–2500 t/km² (Shi and Shao, 2000). Soil erosion, which is mainly caused by grassland degradation, has severely damaged the land resources and the ecosystem of the Loess Plateau.

Grassland soil degradation is a reverse evolutionary ecosystem process instigated by human activities as well as natural factors. One of the most important causes is overgrazing (Lusigi and Glaser, 1984; Sinclair and Fryxell, 1985; Hammond, 1992). Livestock behaviors such as feeding, trampling, and fouling lead to changes in the plant community and the physical and chemical properties of soil. Generally, the pasture recovery ability and total biomass decrease with increasing stocking rate (Christiansen and Srejcor, 1988). However, some studies have suggested that moderate grazing can actually improve grassland, yielding higher plant diversity and productivity (Mcintyre et al., 2003; Ruifrok et al., 2014). In addition, grazing compacts the soil (Weigel et al., 1990), reduces the soil water-holding capacity and water potential (Kobayashi et al., 1997), and affects soil humus and nitrogen accumulation (Severson and Debano, 1991).

Soil microbes are an important component of grassland ecosystems; they regulate matter cycling and promote energy flow in the ecosystem. Owing to their wide biochemical activities, they act as stores and sources of active soil nutrients (Smith and Paul, 1990). As a result, microbes are very sensitive to environmental changes (Nielsen and Winding, 2002), and they adapt to different environments. Grazing activities affect the soil microbial community structure primarily by changing the vegetation and soil physical and chemical properties. Moreover, interactions exist between plants and microbes; plants produce a large variety of root exudates and secretions, including organic substances, enzymes, and polysaccharides, which can benefit or inhibit microbes. In addition, plants can distinguish compounds from various microbial origins and adjust their defense and growth responses on the basis thereof. Conversely, microbes can initiate their colonization strategies in the rhizosphere and synthesize canonical plant growth regulators such as auxins and cytokinins when appropriate host plants are detected (Ortiz-Castro et al., 2009).

For the purpose of clear and reasonable grazing management, an increasing number of studies have focused on the effects of grazing on soil microbial communities. Some studies have found that soil microbial biomass and diversity increase along a gradient of increasing grazing intensity (Ruess and McNaughton, 1987; Bardgett et al., 1997; Qi et al., 2011; Gou et al., 2015); Jangid et al. (2008) and Zhou et al. (2010) suggested that soil microbe diversity is maximal at an intermediate grazing intensity. In addition, Patra et al. (2005) found no obvious differences in fungal and bacterial abundances between light grazing and intensive grazing in seminatural grassland. In studies of salt marsh and sand-dune grassland, Ford et al. (2013) discovered that grazing behavior has a limited impact on soil microbial community structure, which was instead primarily affected by the type of grassland. Together, these studies indicate that results tend to vary between regions with different environments. The relationship between biodiversity and disturbance is affected by geographic location (Death and Barquín, 2012), and different soil types (Arenic Endoaqualfs and Arenic Glossaqualfs) are exposed to different grazing pressures, which cause significant differences in bulk density, organic matter, and carbon mineralization rate in Arenic Endoaqualfs as compared to that in Arenic Glossaqualfs (Wang et al., 2006). For the easily eroded grassland regions of the Loess Plateau, which are highly susceptible to the effects of wind and water, the development and enforcement of reasonable grazing management systems are even more important. A 16S rRNA gene clone library was previously generated to analyze the bacterial community structure of soils with different grazing intensities in the plateau, and it was found that the biomass and diversity increased along a gradient of increasing grazing intensity (Gou et al., 2015). However, this library method detects only a very limited group of microbes, which is a serious limitation with respect to gaining an overall scientific understanding of the soil microbial diversity (McCaig et al., 2001; Nacke et al., 2011).

This study adopted high-throughput sequencing methods to sequence partial bacterial 16S rRNA and fungal 18S rRNA genes to compare the bacterial and fungal community structures in four types of Loess Plateau semiarid grasslands with different grazing intensities and to elucidate their interactions with vegetation and soil physical and chemical properties. A better understanding of the biotic and abiotic components of grasslands with different grazing intensities provides a theoretical basis for the management of Loess Plateau grazed grassland.

In this study, four sample plots of grassland with different grazing intensities were set up in the Loess Plateau to analyze the relationships between soil bacterial and fungal communities and grazing intensity, and the interaction between plant communities and soil physical and chemical properties resulting from grazing intensity and bacterial and fungal communities. The predominant flora in the bacterial and fungal communities of Loess Plateau grassland soil was generally consistent between the four grazing regimens, but there were differences in relative abundance, and additionally, each regimen had its own unique microbial populations. Bacterial and fungal abundance and diversity were highest in S2.67 and S0, respectively. Similarly, Lienhard et al. (2013) found that maximum bacterial and fungal diversity occur with different utilization schemes. Microbial biomass and/or diversity can increase, decrease, or remain the same depending on the type of grassland, geographical location, and the grazing system and intensity (Ruess and McNaughton, 1987; Bardgett et al., 1997; Wang et al., 2006; Jangid et al., 2008; Zhou et al., 2010; Qi et al., 2011; Death and Barquín, 2012; Ford et al., 2013; Gou et al., 2015).

Although bacterial and fungal communities respond to grazing intensity in different ways with regard to relative abundances and diversity, our study showed that the soil microbial community structure changed significantly along a gradient of grazing intensity, which was consistent with changes in soil and plant characteristics. Although similar findings have been reported, these previous studies often relied on culture plate counting (Clegg, 2006; Wang and Sheng, 2012), phospholipid fatty acid analysis (Bardgett et al., 2001; Ingram et al., 2008), or denaturing gradient gel electrophoresis (Clegg, 2006) to determine the effects of grazing on microbes. With respect to microbe identification and classification, these methods may not be sufficiently accurate and/or are able to analyze only a limited set of microbial species in samples from complex environments and thus, do not allow a comprehensive understanding of the effects of grazing intensity on grassland microbial communities. As a consequence, inferences about regulators of the ecosystem and energy flow are limited when such methods are used (Chapin et al., 2000). Therefore, this study aimed to provide a more complete analysis of the microbial community in order to achieve a better understanding of the contributions made by the microbial communities.

We found that the environmental changes that occur with changes in grazing intensity contribute differently to different microbial groups in the community (Figures 3A,B). While shifts in bacterial communities showed a significant relationship with plant biomass, no such relationship was found in fungal communities. Zhou et al. (2010) found that livestock feeding affects aboveground vegetation biomass and community structure, and indirectly changes soil physical and chemical properties. This is a result of the interaction of microorganisms with plants (Ortiz-Castro et al., 2009). The fungal community structure in our study area was not as sensitive as the bacterial community structure to vegetation biomass changes. This may be because fungi are more likely to degrade lignocellulose from different plants than bacteria, allowing them to first obtain resources from many of the relevant available substances (Chapin et al., 2011). In addition, we found significant direct relationships of bacterial and fungal community structures with soil total nitrogen and NO₃-N. With increasing grazing intensity, soil carbon and nitrogen storage decreases (Cui et al., 2005; He et al., 2008). In addition to soil total nitrogen and NO₃-N, organic carbon also decreased, albeit not significantly, with the increase in grazing intensity in this study. Nitrogen is one of the most important nutrients for life; therefore, plant and microbial activities may gradually reduce the content of nitrogen in soil. However, correlations between microbial communities and environmental factors must be carefully explained because it is often very difficult to firmly establish the relationship between microbial communities and soil nutrient cycling (Bardgett, 2005). In addition, this study employed 16S rRNA and 18S rRNA sequencing to analyze microbial communities; the gene abundances indicate genetic potential and not necessarily microbial activity. Despite these limitations, which were considered in this study, we found that the environmental factors evaluated explained over 50% of the shift in the microbial communities, which suggests that they are the primary factors influencing microbial community structure. Multiple studies have shown that the effect of mild or moderate grazing on grassland soil is relatively small, and these grazing intensities are beneficial to dry matter production, nutrient cycling, and carbon and nitrogen storage (Han et al., 2008; Li et al., 2008; Chen et al., 2010). In this study, grazing in S2.67 had relatively little effect on grassland soil and plant characteristics, and with respect to bacterial communities, 2.67 sheep/ha can be considered a moderate stocking rate in the study region. According to the intermediate disturbance hypothesis (Connell, 1978), 2.67 sheep/ha may be close to the moderate-disturbance stocking rate in the Loess Plateau grassland.

Compared with the other regimens, ungrazed soil was the most enriched and showed the largest variety in indicator bacteria and fungi. One of the bacteria enriched in S0 was Chlorobi, also known as green sulfur bacteria, which are obligate anaerobic bacteria. However, some members of this family are facultative anaerobic and can degrade cellulose (Iino et al., 2010; Podosokorskaya et al., 2013). Presence of Chlorobi was significantly positively correlated with water-holding capacity and aboveground biomass; these parameters were significantly greater in S0 than in the other regimens (Supplementary Table S1). This could be owing to the air in the soil being saturated with water vapor, thus increasing anaerobicity, and/or plant litter, which increases the soil cellulose content. Proteobacteria was one of the most abundant phyla in all four grazing regimens, with the exception of Desulfobacterales enrichment in S8.67; the other significant clades were primarily enriched in S0. Most of the significantly enriched microbial clades in S0 showed a significant positive correlation with plant biomass (Supplementary Table S1). These microbial clades (Betaproteobacteria, Caulobacterales, Sphingomonadales, Xanthomonadales, Leotiomycetes, and Zygomycota) all have a close relationship with lignocellulosic degradation (Yoon et al., 2010; Štursová et al., 2012). Enrichment for Caulobacterales (Verastegui et al., 2014), Sphingomonadales (Alphaproteobacteria) (Abrusci et al., 2009), Xanthomonadales (Gammaproteobacteria) (Eichorst and Kuske, 2012), and Zygomycota (Chatzinotas et al., 2013) may be caused by a higher plant biomass in S0 than in the three other grasslands. In addition, Sphingomonadales, Haliangiaceae, and Pucciniomycetes also showed a significant positive correlation with OC, suggesting that increases in plant litter decomposition products or root secretions affect these microbial communities. Kindaichi et al. (2013) found that Caldilineae are effective phosphorus-removing bacteria that can absorb dissolved phosphorus from the environment and synthesize polyphosphates intracellularly. The soil nutrients in S5.33 may have created the most appropriate growth environment for Caldilineae, which might have accumulated the soil phosphorus in their cells, thereby significantly reducing the soil phosphorus content (Supplementary Table S1). S8.67 was primarily enriched for Actinobacteria and Desulfobacterales, and the fungal Pezizales_norank. No significant correlation was detected between Actinobacteria and the main environmental factors measured. This may be because correlations between higher taxonomic ranks and nutrient indicators may conceal relations with environmental parameters at the subfamily level (Cruz-Martínez et al., 2012). Total N was significantly increased in S8.67, which is easily explained by the fact that with increasing occupancy, the amount of excrement and consequently, soil nitrogen content, increases. A significant positive correlation was observed between Desulfobacterales and total N, which might be related to the nitrogen cycle; Widdel (1987) found that Desulfobacter hydrogenophilus fixates nitrogen, and Nitrospinaceae (also significantly enriched and a member of Desulfobacterales) show ammonia oxidation activity (Ionescu et al., 2012). The only significantly enriched fungi in S8.67 soil were Pezizales, which generally grow easily on disturbed soils (Petersen, 1985). Although some microbial taxa (such as Tremellales_incertae_sedis, Lineage_IIb, Glycomycetales, Subgroup_3, AKYH767, F0723, Aquamicrobium, Oceanobacillus, Acidobacteriales, Ktedonobacteria_norank) showed significant enrichment in different grazing intensities of the grassland, their contribution to their environment might be limited due to their low abundance (<0.1% of the sequences were from these taxa).

In summary, the response of soil microbial communities to grazing intensity was studied by partial 16S/18S rRNA gene analysis. MiSeq results showed that grazing changed the microbial community structure and increased bacterial community diversity and relative abundance in S2.67. Grazing intensity impacted the soil and vegetation characteristics to varying degrees, with relatively minimal impact in S2.67. In addition, changes in microbial communities were primarily attributable to plant biomass, and soil total nitrogen and NO₃-N. Considering microbial community structure and environmental factors, a stocking rate of about 2.67 sheep/ha is appropriate for the semiarid Loess Plateau grassland evaluated in this study. Owing to the limited sampling and restricted sampling area, these conclusions cannot be generalized to other Loess Plateau grasslands; however, our findings represent an important step toward understanding how grazing has changed the soil microbial communities in the Loess Plateau grasslands. We are currently planning a more comprehensive study of the interaction between grazing activities and soil microbial communities in the Loess Plateau grassland system to formulate even more accurate recommendations for the conservation of the fragile grasslands of the Loess Plateau.

H performed the sampling, contributed to the experimental design, data analysis, and writing of the manuscript. XC contributed to the sampling and data analysis. FH provided to the study site. YW contributed to the data analysis and writing. YC contributed to the sampling, experimental design, data analysis, and writing of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.