Nitrogen Application Increases Soil Organic Carbon And Crop Productivity By Altering Soil Autotrophic Bacterial Community On the Semiarid Loess Plateau

Background and aims Soil autotrophic microorganisms play crucial roles in carbon xation and crop productivity. However, the information is remains limited to whether and how soil autotrophic microbe mediate SOC pool dynamics and crop productivity. Methods Here, we investigated the effects of the structure and co-occurrence network of soil autotrophic bacterial community on SOC storage and maize yield. A long-term eld experiment was conducted with four chemical nitrogen (N) application rates on the semi-arid Loess Plateau, including N application at 0 kg ha –1 (N0), 100 kg ha −1 (N1), 200 kg ha −1 (N2), and 300 kg ha −1 (N3), respectively. Results Our results showed that the N application signicantly impacted the structure and co-occurrence network of soil cbbL-carrying bacterial community via changing soil pH, nitrate nitrogen (NO 3 -N), and soil water content (SWC). The diversity of soil autotrophic bacterial community decreased with the increasing rate of N application. We detected a high abundance of the autotrophic bacterial dominant genera Xanthobacter, Bradyrhizobium, Aminobacter, and Nitrosospira. The co-occurrence network of autotrophic bacteria contained four distinct modules that consisted of closely interconnected microorganisms. Structural equation modeling further indicated that the diversity, composition, and network of autotrophic bacterial community had signicant relationships with SOC storage and maize yield. Conclusion Taken together, our results highlight that the soil autotrophic bacterial community may drive carbon xation SOC accumulation and contribute positively to crop productivity. The present study indicated the N treatments strongly impacted the abundance, diversity, composition, and co-occurrence network of soil autotrophic bacterial community via changing pH, NO 3 -N, and SWC. We provided evidence that the abundance, diversity, and network soil autotrophic bacterial community contributed to RubisCO activity, SOC storage, and maize productivity. Taken together, understanding the biological mechanisms of soil autotrophic bacteria mediating carbon xation may provide crucial implications for enabling SOC sequestration and crop productivity. Future research by the microcosm incubation and stable isotope-based evidence could help verify the casual relationships between soil autotrophic bacterial community and carbon xation activity.


Introduction
Soil organic carbon (SOC) is a key predictor for soil quality and crop production (Wiesmeier et al. 2019).
Given that soil organic carbon is closely related to crop yield, agricultural system productivity can be enhanced by increasing soil organic carbon (Yuan et al. 2021a). Agricultural management regimes can drive soil microbial community structure to improve SOC storage through a serious of biochemical To date, four RubisCO forms (forms I-IV) have been found to be different in structure, catalytic property, and O 2 sensitivity, and form I of RubisCO is the most abundant among the four forms. The cbbL gene, which encodes a large subunit of RubisCO I, has been often used as a phylogenetic marker to investigate the autotrophic bacterial community (Kovaleva et al. 2011). A growing body of evidence has supported that fertilization regimes, tillage, and mulching practices have strong impacts on the soil autotrophic bacterial community and enzyme activities through altering SOC, pH, bulk density, and available phosphorus (Yuan et  Here, we evaluated the importance of soil autotrophic bacterial community on SOC and crop productivity under eld conditions. We performed an 8-year eld experiment with four N application rate on the semiarid Loess Plateau. We asked the following two questions: (1) How do the composition and cooccurrence network of soil autotrophic bacterial community response to the N treatments? (2) How does the soil autotrophic bacterial community drive SOC dynamics and crop productivity? We hypothesized that the N treatments signi cantly changed the composition and co-occurrence network of soil autotrophic bacterial community. Furthermore, we expected that soil properties and the autotrophic bacterial community jointly mediated SOC dynamics and crop productivity.

Field experiment description
The long-term fertilization experiment was conducted at the Rainfed Agricultural Experimental Station of Gansu Agricultural University (35°28'N, 104°44'E) in Gansu province, China. The experiment site located in a warm temperate zone with a continental monsoon climate, with mean annual temperature of 6.4℃ and mean annual precipitation of 390 mm. The soil is classi ed as Calcaric Cambisol according to the Food and Agricultural Organization (FAO) classi cation system. The long-term eld experiment followed a completely randomized design with four nitrogen application rates (three replicates) at 0 kg ha −1 (N0), 100 kg ha −1 (N1), 200 kg ha −1 (N2), and 300 kg ha −1 (N3), respectively. The experiment was started in 2012, which consisted of 12 plots. Nitrogen fertilizer was applied as urea (46% N) in two splits: one-third was broadcast on the soil surface and incorporated by moldboard plowing to soil and the remaining two-thirds was applied at jointing stage of maize. Triple superphosphate (P 2 O 5 16%) was applied at 150 kg P 2 O 5 ha −1 , and was evenly broadcast on the soil surface of all plots.
The experimental plots were 18.7 m 2 (4.25 m length and 4.4 m width) and consisted of narrow ridges (15 cm height and 40 cm width) alternated with wide ridges (10 cm height and 70 cm width). All ridges were covered by the plastic lm (0.01 mm thickness and 140 cm width). Maize monoculture (cv. Pioneer 335) was planted annually from April to October, with a density of 52,500 plants ha −1 . No management regimes were adopted, expect for weeding by hand.
Soil sampling and soil properties assays Soil samples from each plot were collected at the owering stage in early August 2019. In each plot, 10 soil cores were collected from the surface layer (0−20 cm) using a Dutch auger (5 cm diameter), and mixed to form a composite sample. After eld collection, fresh samples were placed on ice and immediately transported to the laboratory, and then were sieved (2 mm) to remove visible residues. Soil samples were homogenized to analyze soil chemical properties, microbial biomass carbon, soil autotrophic bacterial community, and RubisCO activity.
Soil pH was measured by a pH meter (Mettler Toledo FE20, Shanghai, China) with water: soil ratio of 2.5: 1 (v/w). SOC was determined by a modi ed Walkley−Black wet oxidation method (Nelson and Sommers 1983). TN was determined using the micro−Kjeldahl method (Sparks et al. 1996). NO 3 −N and NH 4 −N were extracted with 2 M KCl and measured using a continuous ow analyzer (Skalar, Breda, Netherlands). Available phosphorus (AP) was extracted with sodium bicarbonate and measured using the

Crop yield and RubisCO activity
All maize plants were manually harvested at the maturity. After harvest, the air-dried grain in each plot was weighted to calculate grain yield. The aboveground biomass was determined on a dry-weight basis by oven drying the crop samples at 105°C for 45 min and subsequently to constant weight at 85°C (Xie et al. 2020).
Soil RubisCO activity was determined by the method described by Yuan et al. (2012b). Brie y, 2 g of freeze-dried soil sample was placed in centrifuge tubes and suspended in protein extractant containing Tris-HCl buffer and dithiothreitol. The soil suspension was disrupted by ultrasonication in ice bath. After centrifuged, the supernatant was amended with solid ammonium sulfate to reach 80% saturation, and then stirred for 30 min and centrifuged at 4℃. The precipitate was dissolved to determine RubisCO activity. The absorbance of reaction mix was measured at 340 nm wavelength using a spectrophotometer (UV−2450, Shimadzu, Japan). All reactions were carried out in triplicate, and negative controls were set up. The RubisCO activity was expressed as nmol CO 2 g −1 soil min −1 .
The cbbL gene copy number and Illumina sequencing Total genomic DNA was extracted from soil samples using the EZNA. Soil DNA Kit (Omega Bio-tek, Norcross, GA, USA) according to manufacturer's protocols. The quality of DNA was determined by 1% agarose gel electrophoresis, and the concentration and purity were detected using a NanoDrop−2000 spectrophotometer (Thermo Fisher Scienti c, Wilmington, DE, USA). DNA samples were stored at -80℃ for subsequent analysis.
The copy number of cbbL gene was determined by quantitative PCR using the primers K2f PCR parameters were as follows: a pre-denaturation 95°C for 2 min, followed by 25 cycles at 95°C for 30 s, 55°C for 30 s, and 72°C for 30 s, and a nal extension at 72°C for 5 min. The standard curve was obtained using a 102 to 108 dilution of plasmid DNA carrying the cbbL gene fragment. Melting curve analysis was performed at the end of the PCR ampli cation to check the speci city of ampli cation. The cbbL copy number was calculated according to the standard curve.
Puri ed PCR products were quanti ed by Qubit3.0 (Life Invitrogen) and mixed equally. The amplicon library was paired-end sequenced (2 × 300) on an Illumina platform. Raw data were extracted, trimmed, and quality screened using the Quantitative Insights Into Microbial Ecology (QIIME, v1.8.0) (Caporaso et al. 2010). Brie y, the low-quality sequencing reads with length <150, with an average Phred scores <20, with ambiguous bases in barcodes and with mononucleotide repeats greater than 8 bp were ltered out.
The paired reads from the original bacterial DNA fragments were merged using FLASH (version 1.2.7). After chimera detection, sequence analyses were performed with the UPARSE (version 7.1) software package, and the operational taxonomic unit (OTU) was clustered at 97% sequence identity by UCLUST

Statistical analyses
The analysis of variance was analyzed by Tukey's honest signi cance test at P ≤ 0.05 using SPSS 22.0 (IBM SPSS, USA). Pearson's correlation coe cients were used to assess linear relationships among soil properties, the abundance and composition of soil autotrophic bacterial community, RubisCO activity, DOC, MBC, and crop yield across N treatments. Principal coordinate analysis (PCoA) was used to evaluate the Bray-Curtis distances of soil autotrophic bacterial community compositions using the 'vegan' package in R software.
To describe the potential co-occurrence patterns, the autotrophic bacterial network was constructed using the Spearman's correlation and Kullback-Leibler dissimilarity (Faust et al. 2012). The OTUs more than ve-sixths of soil samples were kept for network construction. A valid co-occurrence was considered a statistically robust correlation between taxa when the correlation coe cient (r) was > 0.6 or < −0.6 and the P value was < 0.05. The co-occurrence network visualization was conducted using Gephi software (Bastian et al. 2009), and the modules were de ned as clusters of closely associated nodes (i.e., groups of co-occurring microbes) (Layeghifard et al. 2017).
Random forest tool was used to assess the important predictors of RubisCO activity and maize yield.
Random Forest modeling was conducted using the 'RandomForest' package (Bento et al. 2002), and the 'rfPermute' package was used to determine the model signi cance and predictor importance (Archer, 2016

Soil properties, MBC, grain yield, and Rubisco activity
One-way analysis of variance showed that the N treatments have signi cant effects on soil properties (P < 0.05, Table S1). TN and NO 3 −N were signi cantly (P < 0.05) increased with the increasing rates of N application, while soil pH and SWC exhibited an opposite tread (P < 0.05). However, no signi cant difference was found in SOC (P > 0.05), NH 4 −N (P > 0.05), and AP (P > 0.05) among three N treatments (N1, N2, and N3). DOC and MBC were enhanced by the N application (Fig. S1), with the signi cantly (P < 0.05) higher values under the N3 treatment than under the N1 and N0 treatments. The N2 and N3 treatments were characterized by signi cantly (P < 0.05) higher grain yield and aboveground biomass than the N1 and N0 treatments (Fig. S1), as well as RubisCO activity (P < 0.05, Fig. 1).
Abundance and structure of soil autotrophic bacterial community The abundance of autotrophic bacteria indicated by the copy number of cbbL gene ranged from 0.82 × 10 6 to 2.78 × 10 6 copies g −1 soil. The clear differences were observed among treatments (P < 0.05), with the highest abundance of cbbL gene under the N3 treatment (Fig. 1). A total of 289, 656 sequences of soil autotrophic bacterial community were obtained using Illumina sequencing after quality control. The diversity of autotrophic bacteria indicated by Chao 1 richness was signi cantly higher under the N0 and N1 treatments than under the N2 treatment, with intermediary value under the N3 treatment (P < 0.05, Fig.   1). However, there was no signi cant difference in Shannon index among the four treatments (P = 0.10, Fig. 1).
The autotrophic bacterial co-occurrence networks Co-occurrence networks were constructed to examine the different co-occurrence patterns of the soil autotrophic bacterial community under the four treatments (Fig. 3). In total, there were 367 nodes, 1956 links, and four distinct modules (modules I, II, III, and VI) in the autotrophic bacterial network. In the autotrophic bacterial network, the number of positive correlations (1953 edges) were greater than that of the negative correlations (3 edges). The modules I, II, III, and VI in the bacterial networks consisted of 117, 72, 99, and 79 nodes, respectively. At the phylum level, the relative abundance of Alphaproteobacteria was signi cantly (P < 0.05) higher in modules II and VI than in modules I and III, while those of Betaproteobacteria, Actinobacteria, and Chloro exi followed the opposite trend (Fig. 4). At the genus level, modules I and II showed the signi cantly (P < 0.05) greater abundances of Aminobacter and Bradyrhizobium, but lower abundances of Mycobacterium, Nitrosospira, and Saccharomonospora than module III. In contrast, Xanthobacter was signi cantly greater in module VI than in modules I, II, and III. Module VI was positively correlated with TN (r = 0.77, P < 0.01) and NO 3 −N (r = 0.89, P < 0.001), but negatively correlated with pH (r = −0.71, P < 0.01) (Fig. 5). Furthermore, module VI had positive correlations with RubisCO activity (r = 0.77, P < 0.01), SOC (r = 0.63, P < 0.05), DOC (r = 0.59, P < 0.05), MBC (r = 0.85, P < 0.001), grain yield (r = 0.72, P < 0.01), and aboveground biomass (r = 0.71, P < 0.05) (Fig. 5).

Discussion
Nitrogen application affected soil autotrophic bacterial community Our results showed that N fertilizer application led to great changes in soil properties, and consequently strongly impacted the abundance, diversity, composition community, and co-occurrence network of soil autotrophic bacterial community. The abundance of cbbL gene was signi cantly increased with the increasing application rate of N fertilization, with the highest value under the N3 treatment. Increasing cbbL-containing bacterial abundance has been found to be positively correlated with microbial In the present study, the impact of N application on soil autotrophic bacterial community composition was signi cantly associated with soil properties. The majority of the autotrophic bacterial community . We suggest that the reduced autotrophic bacterial diversity was bene t for enhancing RubisCO activity and CO 2 xation rate of bacterial populations, and improved SOC sequestration and maize productivity. The soil autotrophic bacterial community was dominated by facultative autotrophic bacteria, including Xanthobacter, Bradyrhizobium, Aminobacter, and Nitrosospira. In fact, these facultative autotrophs exhibit metabolic exibility with both heterotrophic and autotrophic metabolic pathways, allowing them to grow on alternative C and energy sources (Esparza et al. 2010;Xiao et al. 2019). Additionally, the genus Xanthobacter is the main implementer of xing CO 2 mainly via the ribulose-biphosphate pathway (Wiegel 2006). Notably, the genus Xanthobacter was the dominant taxa in module IV of co-occurrence network, and module IV had positive correlations with SOC storage, which con rmed the potentials of soil autotrophic bacterial community for carbon xation. Our results highlight that soil autotrophic bacteria have signi cant potentials for carbon xation and contribute to carbon pool and crop productivity.

Conclusions
The present study indicated the N treatments strongly impacted the abundance, diversity, composition, and co-occurrence network of soil autotrophic bacterial community via changing pH, NO 3 -N, and SWC.
We provided evidence that the abundance, diversity, and network soil autotrophic bacterial community contributed to RubisCO activity, SOC storage, and maize productivity. Taken together, understanding the biological mechanisms of soil autotrophic bacteria mediating carbon xation may provide crucial implications for enabling SOC sequestration and crop productivity. Future research by the microcosm incubation and stable isotope-based evidence could help verify the casual relationships between soil autotrophic bacterial community and carbon xation activity.    Relative abundance of different dominant modules in soil autotrophic bacterial networks at phyla (a) and genera (b)-level. Different small letters indicate the signi cant difference among modules at P < 0.05.

Figure 6
Mean contribution (% of increased mean square error, MSE) of soil properties and soil autotropic bacterial community (cbbL) to RubisCO activity (a) and maize yield (b) based on random forest modelling. Random forest modelling was performed based on 12 samples (4 fertilization treatments × 3 replicates).
Soil properties include pH, total nitrogen (TN), nitrate nitrogen (NO3−N), ammonia nitrogen (NH4−N), available phosphorus (AP), and soil water content (SWC). The autotrophic bacterial community include abundance of cbbL gene, diversity (Shannon index), composition ( rst principal coordinates, PCoA1), and four module eigengenes in trophic co-occurrence network. c, Impacts of soil properties and soil autotrophic bacterial community on RubisCO activity, SOC storage, and maize productivity (grain yield) as estimated using structural equation modeling (SEM) analysis. Based on random forest analyses, the signi cant predictors were chosen to perform the SEM analysis. The latent (soil properties and the autotrophic bacterial community) inside the boxes were used to integrate the effcts of multiple multiple conceptually related observed variables into a single-composite effect. Soil properties are represented by pH, nitrate nitrogen (NO3-N), and soil water content (SWC). Soil autotrophic bacterial community (cbbL) is represented by the composition [ rst principal coordinates (PCoA1)], diversity (Shannon index), and module VI of soil autotrophic bacterial networks. SOC storage is represented by soil oraganic carbon (SOC), microbial biomass carbon (MBC), and dissolved organic carbon (DOC). *** P < 0.001; ** P < 0.01; * P < 0.05.