Drivers of Soil Bacterial Diversity in Sandy Grasslands in China

Bacteria constitute great abundances and groups on Earth and control many important processes in terrestrial ecosystems. However, our understanding of the interactions between soil bacteria and environmental factors remains limited, especially in sensitive and fragile ecosystems. In this study, geographic patterns of bacterial diversity across the four sandy grasslands along a 1600 km north-south transect in northern China were characterized by high-throughput 16S rRNA gene sequencing. Then, we analyzed the driving factors behind the patterns in bacterial diversity. The results showed that of the 21 phyla detected, the most abundant were Proteobacteria, Actinobacteria, Acidobacteria and Firmicutes (average relative abundance > 5%). Soil bacterial α diversity, calculated as the bacterial phylotype richness and Faith’s phylogenetic diversity, was highest in the Otingdag Sandy Land and lowest in the Mu Us Sandy Land. Soil EC was the most inuential factor driving bacterial α diversity. The bacterial communities differed signicantly among the four sandy grasslands, and the bacterial community structure was signicantly affected by environmental factors and geographic distance. Of the environmental variables examined, climatic factors (MAT and MAP) and edaphic properties (pH and EC) explained the highest proportion of the variation in bacterial community structure. Biotic factors such as plant species richness and aboveground biomass exhibited weak but signicant associations with bacterial α diversity. Our ndings revealed the important role of climate and salinity factors in controlling bacterial diversity; understanding these roles is critical for predicting the impacts of climate change and promoting sustainable management strategies for ecosystem services in these sandy lands.


Introduction
Soil microbes have great abundances and groups on Earth, and control many important processes in terrestrial ecosystems 1 . Microbe spatial patterns and their driving forces have received much attention in recent years and are important subjects in community ecology. Numerous studies have reported patterns in the diversity and community structure of soil microbes at different scales in different ecosystems [2][3][4][5] .
Several studies have demonstrated large-scale patterns of soil microbes in China, for instance forest 5 , alpine grassland 6 and agricultural ecosystems 7 . However, little is known about the geographic patterns of soil microbial diversity and factors controlling that diversity in sandy land ecosystems. There are four sandy lands in China, i.e., Hulun Buir, Horqin, Otindag and Mu Us, and they account for approximately 1.7% of the land area in China. Investigating the driving factors of soil microbial diversity in these sensitive and fragile ecosystems is essential due to their importance to nutrient cycling, climate change resistance and human activities.
The patterns of microbial diversity are in uenced by a wide range of biotic and abiotic factors. At small scales, the patterns of the microbial community are in uenced by the nutrient status 8 , texture 9 , moisture 4, and pH of soil 2,10 . Microbial communities have also been found to be controlled by climatic features, geographic distance, and vegetation and soil properties at large spatial scales 5,11,12 . Generally, these driving factors vary with the scales and ecosystems studied. However, no study has investigated how soil microbial communities are organized in the soils of the four sandy grasslands in China until now. In summary, soil microbial communities are closely correlated with abiotic variables, such as climate and edaphic properties 6,13 , and biotic variables, namely plant diversity 14 . Studying microbial diversity and community structure and their relationships with environmental factors would expand our knowledge about the factors that regulate Earth's biodiversity and biogeochemistry.
Additionally, most studies in sandy lands have been performed at the site scale and in limited areas 15,16 .
A comprehensive analysis of the effects of environmental variation on microbial diversity and community structure across the four sandy grasslands is still lacking. In the present work, bacterial diversity and community structure were estimated based on high-throughput sequencing data, and their relationships with environmental variables were analyzed. We aimed to (i) explore and compare the bacterial diversity and community structure patterns among the four sandy grasslands, and (ii) identify the main environmental factors controlling the patterns in the bacterial community.

Relationships between bacterial α diversity and climate, vegetation and edaphic factors
The bacterial α diversity (number of observed OTUs and Faith's phylogenetic diversity) changed signi cantly across the four sandy grasslands. The observed bacterial phylotype richness and Faith's phylogenetic diversity were the highest in the Otindag Sandy Land, followed by the Horqin Sandy Land and the Hulun Buir Sandy Land, and were the lowest in the Mu Us Sandy Land (Fig. 2).
The bacterial phylotype richness and Faith's phylogenetic diversity in the four sandy grassland soils were both signi cantly correlated with soil EC and PR (Table S1). The bacterial phylotype richness was also signi cantly correlated with soil pH and AB, while Faith's phylogenetic diversity was signi cantly correlated with MAP. Among all 10 environmental variables examined, EC was the most in uential factor in uencing the bacterial phylotype richness and Faith's phylogenetic diversity. The results of the stepwise regression further con rmed the best prediction of EC for both phylotype richness and Faith's phylogenetic diversity. The variation in EC explained 45.6% and 48.4% of the variation in phylotype richness and Faith's phylogenetic diversity, respectively ( Table 1).
Correlations of bacterial β diversity with geographical distance and environmental factors NMDS showed that the soil samples from the four sandy grasslands were clearly separated from each other, revealing that the structure of bacterial communities differed signi cantly across the four sandy grasslands (Fig. 3). This nding was further con rmed by the results of ANOSIM, MRPP and PERMANOVA ( Table 2). The structure of soil bacterial communities was signi cantly related tospatial factors (geographic distance) and climate factors (MAT, MAP) ( Fig. 4), suggesting that spatial isolation and climate characteristics affected the degree of similarity between bacterial communities. Signi cant correlations were also observed between the structure of bacterial communities and salinity factors (pH and EC) ( Table S2). The results of DistLM showed that climate factor (MAT, 6.37%, p < 0.001) explained the largest proportion of the variation in the bacterial community structure, followed by SOM (5.50%, p < 0.001) and salinity factors (EC, 4.17%, p < 0.001; pH, 3.01%, p < 0.001). Climate factor, namely MAP (p < 0.01; Table 3), also played an important role in explaining the variation in the bacterial community structure.

Discussion
The results of this study demonstrate that a group of biotic and abiotic variables can explain the variation in soil bacterial diversity across the four sandy grasslands in China. Speci cally, our results show that the soil bacterial communities differed signi cantly across the sandy grasslands and were mainly controlled by a suite of variables, including edaphic variables, for instance soil EC and pH; climate variables, such as MAT and MAP; and spatial variables, namely geographic distance. Surprisingly, vegetation-related variables played a minor role in explaining the variation in bacterial community diversity.
Soil properties played an important role in explaining the variation in the diversity of bacterial communities. Soil EC was the most important factor regulating diversity patterns across a large scale. This is surprising because saline soils are often controlled by EC. In this study, both bacterial phylotype richness and phylogenetic diversity decreased with increasing EC ( Figure S2); however, EC was the third most important variable for β diversity. These results are in agreement with other results obtained in saline environments [27][28][29] . The role of soil EC in determining microbial diversity can be explained by examining the variations in the relative abundances of speci c taxonomic groups, especially the dominant groups. For example, in this study, the relative abundance of Actinobacteria decreased with increased EC, but the relative abundance of Proteobacteria increased (Table S3). Proteobacteria and Actinobacteria accounted for 66.56% of the obtained bacterial sequences. This indicates that salinity is highly in uential not only in saline soils but also in nonsaline soils.
Soil EC is not the only stressor for soil biota in desert environments and is often accompanied by high pH 30 . Our results indicate that soil pH was another important factor in uencing the structure of soil bacterial communities at a large scale. The contribution of soil pH in regulating the structure of soil bacterial communities has also been widely estimated by previous studies 5, 31−33 . The strong relationship between soil pH and the structure of microbial communities can be attributed to the interaction between pH and other environmental variables. The intracellular pH of most microorganisms is usually within 1 pH unit of neutral 34 . Any signi cant variation in environmental pH imposes stress on single-celled organisms, which has signi cant effects on microbial community composition in a variety of environments [35][36][37] .
In addition, our ndings demonstrate that soil nutrient contents, such as C,N and P, were not strong predictors of soil bacterial α and β diversity, which is consistent with the ndings of a study conducted in the Loess Plateau by Zeng et al. 11 . However, this is inconsistent with the global soil bacterial diversity patterns identi ed in a meta-analysis of 600 soil samples 13 . This inconsistency arose because in the current study, there were no obvious differences in soil C, N and P levels across the four sandy grasslands. Thus, no signi cant correlations were observed between soil bacterial α and β diversity and soil C, N and P contents. However, multicollinearities were excluded from the step wise regression, and only variables that were closely related, such as soil EC, were selected, as they contributed to most of the variance.
Vegetation properties (PR and AB) exhibited weak but signi cant correlations with bacterial phylotype richness and phylogenetic diversity but not with bacterial community structure (Table 3; Table S1). Many studies have examined the relationships between vegetation characteristics and soil bacterial community properties; however, the results have been inconsistent. Some studies in grassland ecosystems have reported the signi cant effect of plant diversity on the diversity of soil microbial communities 38,39 . Others have shown that vegetation properties do not signi cantly affect the soil microbial community 40 . Plant diversity and productivity control the contents of limiting resources and thereby control the soil biota 41,42 .
The results of this study con rm the importance of resource availability in regulating the soil microbial community in these sandy grasslands and indicate that the in uences of vegetation properties on soil biodiversity in sandy grassland ecosystems should also be considered in future studies.
Climate also played an important role in explaining the variation in the structure of soil bacterial communities. Our ndings reveal that MAT and MAP signi cantly affected the structure of soil bacterial communities. This nding is consistent with studies demonstrating the importance of MAT and MAP in in uential the structure of bacterial communities at different scales 2,43 , especially in arid and semiarid regions such as temperate grasslands in northern China 44 . Any change in MAT or MAP directly affects the processes of microbial metabolism 45 . Furthermore, MAT and MAP are the main factors affecting plant community composition and AB, which determine the quantity and quality of resources available to soil bacterial communities 5,46,47 . In this sense, MAT and MAP control the structure of soil bacterial communities through their impacts on vegetation properties and soil nutrient levels.
The strong relationship between the degree of dissimilarity among bacterial communities and geographical distance indicates that the probability of dispersion is the main constrainton the spatial patterns of bacterial communities across the four sandy grasslands. The impacts of dispersal limitation in soil bacterial communities have also been extensively described in previous studies 31,48,49 . However, the results of this study are inconsistent with the results of Fierer and Jackson 2 , who quantitatively compared the diversity and structure of bacterial communities across North and South America, and found that there was no signi cant relationship between the degree of dissimilarity among soil bacterial communities and geographic distance. This discrepancy may be attributed to the different scales of that study and our study 50,51 . Our results are in agreement with those of Martiny et al. 52 , who suggest that the effects of dispersal limitation on the biogeography of soil bacterial communities are most likely to be observed at intermediate scales (10-3,000 km).
In summary, this work is the rst attempt to comprehensively detect the geographic patterns of soil bacteria across the four sandy grasslands in northern China. A combination of biotic and abiotic factors controlling soil bacterial α and β diversity were identi ed. Our study provides evidence for the in uence of soil EC on the diversity of bacterial communities, which is often neglected in nonsaline soils. Soil pH was another important predictor of the structure of soil microbial communities. Geographic distance and climate factors (MAT and MAP) explained the largest proportion of the variance in the structure of soil microbial communities. To improve our knowledge about the dynamics of these sensitive and fragile ecosystems, future efforts should be put into exploring the general patterns that microbes respond to climate change and human activities. The results of this study strengthen our knowledge about the processes that generate and maintain microbial biogeographic patterns.

Study area and eld sampling
This study was performed at 19 sites in the four sandy grasslands in China (107.5°E to 120.7°E, 37.2°N to 49.2°N; Figure S1). There were 6 sites in the Hulun Buir Sandy Land, 5 sites in the Horqin Sandy Land, 5 sites in the Otindag Sandy Land and 3 sites in the Mu Us Sandy Land. The mean annual temperature (MAT) ranged from − 0.85 to 8.25 °C, and the mean annual precipitation (MAP) ranged from 290 to 439 mm.
Field sampling was conducted in August 2018 during the period with the highest vegetation productivity.
Fifty-six soil samples (0-10 cm; 3 samples per site except at one ofthe HulunBuir Sandy Land) were collected. At each site, three plots (1×1 m 2 ) were established at least 100 m apart. Within each plot, all plant species were identi ed to the species level. The aboveground biomass of each species was determined by clipping the plants at ground level, drying them at 65 °C for 72 h, and then weighing them 17 . In addition, 5 soil corsamples (3.5 cm in diameter) were collected within each plot. The samples were combined and then divided into two portions in the eld. One portion was stored in polyethylene bags, placed in a portable refrigerator and then refrigerated at − 20 °C in the laboratory until processed for molecular analysis; the other was air-dried for the determination of its edaphic physicochemical properties.

Climate, Vegetation And Edaphic Properties
Climate factors, such as MAT and MAP, were collected from the WorldClim database 18 ( www.worldclim.org).
Soil organic matter (SOM) was determined by the K 2 Cr 2 O 7 -H 2 SO 4 oxidation method. Total nitrogen (TN) was measured using an elemental analyzer (VarioMacro Cube elementar, Germany). Total phosphorus (TP) was examined using the Mo-Sb colorimetric method after H 2 SO 4 -H 2 O 2 digestion. Available phosphorus (AP) was measured by the Mo-Sb colorimetric method after extraction with 0.5 M NaHCO 3 .
Plant species richness (PR) was expressed as the total number of species 18 . The total aboveground biomass (AB) was estimated as the sum of each species' biomass 18 .
Soil Dna Extraction, Pcr And 16s Rrna Sequencing DNA from each sample was extracted from 0.5 g fresh soil using a magnetic bead DNA extraction kit (Tiangen Biochemical Technology Co. Ltd., Beijing, China). The primers 341F (CCTAYGGGRBGCASCAG) and 806R (GGACTACNNGGGTATCTAAT) were used to amplify the V3-V4 hypervariable regions ofthe 16S rRNA genes in bacteria. PCRs were conducted in a 30 µl mixture containing 15 µL of Phusion® High-Fidelity PCR Master Mix (New England Biolabs), 1 µl of each primer (10 µM), and ~ 10 ng of template DNA (gDNA). Thermal cycling was performed at 95 °C for 5 min, followed by 34 cycles at 91 °C for 1 min, 57 °C for 45 s, 72 °C for 1 min and nally at 72 °C for 10 min. Sequencing was carried out on an Ion S5 XL platform (Thermo Fisher Scienti c) with single-end reads at Novogene Bioinformatics Technology Co. Ltd., Beijing, China.

Statistical analysis
The bacterial phylotype richness was calculated as the number of OTUs, while phylogenetic diversity was determined with Faith's index 20 . Pearson correlation was performed to assess the associations between bacterial phylotype richness and phylogenetic diversity and the measured environmental variables using DPS software 21 . To avoid collinearities among the environmental variables, the most discriminating variables for bacterial phylotype richness and diversity were selected by stepwise multiple regression analysis. A Mantel test was also performed with PCORD 5.0to relate the structure of bacterial communities based on Bray-Curtis distance to the environmental variables. The importance of each environmental variable to bacterial community structure was estimated with a distance-based multivariate linear model (DistLM) with the 'forward3' procedure in Primer 7 22 .
Nonmetric multidimensional scaling (NMDS) was applied to test the variation in the bacterial community using the VEGAN package in R 23 . Analysis of similarity (ANOSIM) 24 , multiresponse permutation procedure (MRPP) 25 , and permutational multivariate analysis of variance (PERMANOVA) 26 were also performed to con rm the signi cance of differences among the bacterial communities.

Declarations
Availability of data and materials All raw sequences have been deposited at the NCBI Sequence Read Archive (SRA) the accession number PRJNA667483.

Declaration of competing interest
The authors declare that they have no competing interests.    Soil bacterial phylotype richness and Faith's phylogenetic diversity across the four sandy lands in northern China. Means (±SE) with different letters indicate signi cant differences.