Prolonged Forest Nitrogen Addition Shifts Bacterial Assembly While Network Complexity Persists

A five-year field experiment conducted in a subtropical forest demonstrates that prolonged nitrogen addition systematically alters the mechanisms governing soil bacterial community assembly over time while leaving the overall topological complexity of bacterial co-occurrence networks surprisingly intact. Atmospheric nitrogen deposition, which has intensified across developing regions over recent decades, continuously introduces reactive nitrogen into forest soils, altering physicochemical conditions and substrate chemistry. By following a continuous five-year fertilization regime in a subtropical stand of Pinus taiwanensis, researchers discovered that stochastic ecological processes initially governed bacterial community construction following nutrient addition, but environmental filtering and deterministic forces progressively expanded in influence as the soil environment stabilized in later years. Despite these succession dynamics, bacterial interaction networks maintained constant connectivity and complexity across all experimental treatments and years.

The field manipulation was established in 2018 within the Daiyun Mountain National Nature Reserve, located in Quanzhou, Fujian Province, China. The experimental site experiences a typical subtropical maritime monsoon climate, characterized by annual mean temperatures between 18.1 and 19.5 degrees Celsius, annual rainfall ranging from 1268 to 1964 millimeters, and a frost-free period spanning approximately 260 days, with the prevailing soil classified as mountainous red soil. To simulate elevated atmospheric nitrogen deposition, the research team implemented a randomized block design within a pure stand of Pinus taiwanensis, establishing twelve 10-meter by 10-meter plots separated by 5-meter buffer strips. Treatments comprised three distinct levels: an ambient control without nitrogen input, a low nitrogen addition treatment delivering 40 kilograms of nitrogen per hectare annually, and a high nitrogen addition treatment delivering 80 kilograms of nitrogen per hectare annually. Aqueous urea solutions were sprayed evenly using backpack sprayers across seven monthly applications from March to September throughout the growing season, while control plots received equivalent volumes of deionized water.

Using a stainless-steel auger, five soil cores were taken from the zero to ten centimeter mineral layer in each plot following a five-point sampling protocol, cleared of stones and plant roots, and homogenized into composite samples. In the laboratory, samples were partitioned for microbiological, chemical, and spectroscopic analyses. Soil dissolved organic matter was extracted by mixing soil with deionized water at a one-to-four ratio and filtering through sterile 0.45-micrometer syringe filters. The researchers quantified dissolved organic carbon using an organic carbon analyzer and measured dissolved organic nitrogen with a continuous flow analyzer. Spectral absorbance at 254 and 260 nanometers was determined via spectrophotometry to evaluate aromaticity and hydrophobicity, while three-dimensional excitation-emission matrix fluorescence spectroscopy coupled with parallel factor analysis identified three humic-like fluorescent components and established the humification index of the dissolved organic matter pool.

Microbial profiling was performed on genomic deoxyribonucleic acid extracted from the sixty collected soil samples, targeting the hypervariable V3 to V4 regions of the bacterial 16S ribosomal RNA gene amplified with primers 338F and 806R prior to paired-end sequencing on an Illumina MiSeq platform. Principal component analysis revealed that continuous nitrogen addition did not dramatically alter the overarching bacterial community structure across the five-year period. Bacterial alpha diversity, community life-history strategies, and ecological niche breadth remained statistically indistinguishable across nitrogen treatments and control plots throughout the experiment. However, linear discriminant analysis effect size uncovered distinct taxonomic responses among specific bacterial lineages. High nitrogen addition significantly enriched several taxa, including Micromonosporaceae and Micromonosporales within the phylum Actinobacteriota, Rhodanobacter and Nitrospirillum within the phylum Proteobacteria, and Bacillales within Firmicutes, each exhibiting linear discriminant analysis scores exceeding two. In low nitrogen plots, only the genus Skermanella, a diazotrophic member of the Proteobacteria capable of flexible carbon metabolism, showed significant enrichment with a linear discriminant analysis score exceeding 3.5.

To evaluate the ecological mechanisms directing community succession, the researchers compared three analytical frameworks: the neutral community model, quantifying assembly processes based on entire-community null model analysis (QPEN), and inferring community assembly mechanisms by phylogenetic-bin-based null model analysis (iCAMP). All three modeling approaches converged on the finding that stochastic processes overwhelmingly dominated the assembly of soil bacterial communities across the five-year study period. When evaluated across the pooled five-year dataset, the neutral community model demonstrated relatively strong fits, with explained variance values ranging from 0.654 to 0.677 across nitrogen treatments, although annual model fits proved markedly lower. The QPEN framework confirmed this predominance, yielding five-year mean beta-nearest taxon index values between 1.37 and 1.80, which fall within the threshold indicating stochasticity. The iCAMP framework further revealed that stochastic processes accounted for 85.1 to 90.6 percent of total community assembly, with ecological drift representing the single largest mechanism by explaining 49.6 to 54.4 percent of the turnover across treatments.

Beyond demonstrating persistent stochastic dominance across the broader timeline, both the QPEN and iCAMP frameworks revealed an important temporal transition in the relative balance of ecological assembly forces between early and late experimental stages. During the initial two years of nitrogen enrichment, bacterial communities in treated plots shifted toward significantly higher degrees of stochasticity compared to control soils. This early surge in stochastic assembly was fueled by immediate resource pulses, characterized by a rapid doubling of soil nitrate concentrations and a 29 percent decline in the humification index of dissolved organic matter, which reduced environmental filtering and enabled opportunistic taxa to undergo boom-and-bust dynamics. In contrast, during the fourth and fifth years of continuous nitrogen application, deterministic processes gained substantial prominence as soil environmental conditions stabilized.

Measured soil physicochemical variables and dissolved organic matter properties explained approximately 45 percent of the overall variation in bacterial community structure in redundancy analysis, with the remaining 55 percent unaccounted for by measured environmental parameters. Nitrogen addition significantly depressed soil dissolved organic carbon content in years one, two, and four, and lowered the humification index during the first year, whereas dissolved organic nitrogen, aromaticity, and hydrophobicity indices showed no significant overarching response across treatments. Nitrate nitrogen concentrations increased sharply under nitrogen addition during the first two years, while ammonium nitrogen, total carbon, and total phosphorus only exhibited significant increases in year three under high nitrogen inputs. Multiple linear regression indicated that environmental variables explained 21 percent of the variation in the beta-nearest taxon index and 10 percent of the variation in deterministic processes. Total nitrogen and nitrate nitrogen exerted significant negative effects on phylogenetic turnover, while total carbon, soil pH, and the humification index were positive predictors of deterministic selection.

To assess how microbial interactions responded to nutrient enrichment, the authors constructed co-occurrence networks based on pairwise Spearman correlation coefficients among bacterial amplicon sequence variants, setting a correlation threshold of absolute value 0.6 or higher and a significance level below 0.05. Across all treatments, the reconstructed ecological networks were predominantly populated by Proteobacteria, accounting for 30 to 31 percent of network nodes, and Acidobacteria, representing 26 to 28 percent, followed by Actinobacteria and Verrucomicrobia. Contrary to the initial study hypothesis that network connectivity would change under chronic nutrient inputs, topological parameters remained stable throughout the five-year duration. The total number of nodes, number of edges, average degree, and average path length did not differ significantly between the nitrogen addition treatments and the untreated control plots, indicating that the overall architecture of interspecific microbial interactions resisted destabilization despite five years of continuous exogenous urea applications.

Although nitrogen addition treatments did not directly alter topological network metrics, statistical modeling revealed that the intrinsic substrate quality of dissolved organic matter acts as the dominant environmental driver of bacterial network complexity. Random forest regression models incorporating dissolved organic matter parameters, soil physicochemical factors, and bacterial community traits explained 58.6 percent of the total variation in network complexity. Dissolved organic matter characteristics alone accounted for 33.7 percent of this variation, whereas removing dissolved organic matter predictors dropped the model explanatory capacity to 49.0 percent. Linear regression analyses established that the humification index, the ratio of dissolved organic carbon to dissolved organic nitrogen, parallel factor analysis component C3, and soil cation exchange capacity were significantly and negatively associated with bacterial network complexity. Conversely, dissolved organic nitrogen concentration and bacterial taxonomic richness showed significant positive correlations with network complexity, demonstrating that more bioavailable, less humified organic substrates promote richer microbial interaction networks.

The mechanistic link between organic substrate chemistry and network structure reflects fundamental metabolic thresholds governing microbial interactions in forest soils. A lower humification index indicates a smaller proportion of condensed, aromatic macromolecules and a higher abundance of low-molecular-weight, labile compounds, which reduces the metabolic energetic barriers required for microbial decomposition and lowers the necessity for extensive metabolic cross-feeding among distinct taxa. Concurrently, a reduced ratio of dissolved organic carbon to dissolved organic nitrogen improves carbon-nitrogen stoichiometry in the soil solution, allowing functional bacterial consortia to establish dense, coordinated interaction networks through cooperative resource partitioning. The minimal responsiveness of the overall dissolved organic matter pool to nitrogen fertilization in this subtropical stand explains why bacterial co-occurrence network complexity persisted without structural collapse. These findings indicate that bacterial interaction resilience is tightly anchored to dissolved organic matter substrate quality, offering a mechanistic basis for designing adaptive nutrient management and forest conservation strategies under ongoing global atmospheric nitrogen deposition.

Subject of Research: Temporal dynamics of soil bacterial community assembly and co-occurrence networks under nitrogen addition in a subtropical forest

Article Title: Temporal dynamics of bacterial community assembly and network complexity under five years of nitrogen addition in a subtropical forest

Article References: Yuan, X., Zeng, Q., Fu, X., Zhang, X., Zhang, Q., Ren, M., Bai, X., Zhang, Y., Tian, Y., Sun, H., Chen, L., Zhou, J., Cui, J., Gao, X., Cai, M., Zheng, Y., Guo, W., Wang, J., & Chen, Y. (2026). Temporal dynamics of bacterial community assembly and network complexity under five years of nitrogen addition in a subtropical forest. Carbon Research, 5(1), Article 25. https://doi.org/10.1007/s44246-026-00274-4

Image Credits: AI Generated

DOI: 10.1007/s44246-026-00274-4

Keywords: nitrogen deposition, bacterial community assembly, network complexity, subtropical forest, Pinus taiwanensis, dissolved organic matter, humification index, co-occurrence networks, stochastic processes, deterministic selection, ecological drift, iCAMP null model, QPEN framework, neutral community model, soil microbial ecology, dissolved organic carbon, nitrate nitrogen