Root-Driven Filtering Overrides Biochar and Microbial Inoculants in Structuring Bacterial Assemblages of Seawater Rice Cultivation Ecosystem in a Saline–Alkali Soil

TL;DR

This study shows that seawater rice plants have a stronger influence on their root bacteria than soil amendments like biochar or beneficial microbes.

Contribution

The study reveals that host plant selection overrides biochar and microbial inoculants in shaping root-associated bacterial communities in saline–alkali soils.

Findings

Biochar altered soil properties and bacterial composition, but its effects were reduced when combined with PGPR.

Bacillota and Bacteroidota increased in abundance from bulk soil to the root endosphere across all treatments.

Root compartment selection, not soil amendments, dominated bacterial community structure in the rhizoplane and endosphere.

Abstract

Saline–alkali soils significantly hinder agricultural productivity in China’s coastal areas. Although both plant growth-promoting rhizobacteria (PGPR) and biochar have individually demonstrated the capacity to boost crop yield and soil fertility, their synergistic effects on seawater rice and soil ecosystems remain uncertain. In this study, we examined the individual and interactive influences of lychee biochar (2.5% and 5% w/w) and PGPR inoculation on soil physicochemical properties and bacterial community assembly along a soil–root continuum, encompassing bulk soil, rhizosphere soil, rhizoplane, and root endosphere, in a controlled pot experiment with seawater rice. The application of biochar significantly altered soil pH, electrical conductivity, and nutrient availability in both bulk and rhizosphere soils, resulting in pronounced changes in bacterial community composition. The effects generated by biochar were partially mitigated when PGPR was co-applied. The relative abundances of Bacillota and Bacteroidota grew progressively from bulk soil to the root endosphere across all treatments, indicating a significant compartment-dependent selection. Co-occurrence network analysis and FAPROTAX-based functional predictions revealed several taxa and functions that were progressively enriched toward the root, including the halotolerant genera Exiguobacterium and Chryseobacterium, highlighting a significant host-mediated filtration process that functioned independently of the inoculated strains. Multivariate analyses further demonstrated that soil pH was the primary driver of bacterial community structure in bulk and rhizosphere soils, whereas plant-root selection dominated in the rhizoplane and endosphere. Overall, our results demonstrate that, within a seawater-rice and soil ecosystem, the selective influence of the host plant on root-associated microbiomes exceeds that of either biochar amendment or PGPR inoculation. This work improves our understanding of biochar–PGPR–plant interactions in saline–alkali soils and provides insight into sustainable strategies for enhancing rice production under salinity stress.

Figures

Full text

1. Introduction

Saline–alkali soils, which are characterized by high concentrations of soluble salts and exchangeable sodium, constitute one of the most severe abiotic constraints on global rice production. Excessive salinity imposes multiple interacting stresses on plants, including osmotic stress, ion toxicity, and nutrient imbalance, which together restrict root development, disrupt microbial colonization, and ultimately reduce plant productivity [1]. These constraints are particularly pronounced in coastal reclamation regions, such as eastern China, where saltwater intrusion and poor soil structure further hinder the establishment of stable plant–soil–microbe interactions [2]. Although large-scale engineering interventions and agronomic practices have expanded land availability in these areas, the biological limitations governing rice growth and rhizosphere functioning in saline–alkali soils remain insufficiently resolved.

Biochar has been widely proposed as a soil amendment to mitigate salinity stress through improvements in soil structure, cation exchange capacity, and nutrient retention [3]. Numerous studies have reported reductions in soil electrical conductivity and concurrent increases in organic carbon and microbial biomass following biochar application [4,5,6]. However, the effects of biochar are highly variable and depend strongly on feedstock type, pyrolysis conditions, and application rate [7]. Importantly, biochar-induced alterations in soil chemistry can also restructure microbial niches, in some cases suppressing sensitive taxa or amplifying abiotic stress when applied at high dosages [8,9]. Thus, while biochar can substantially modify the soil environment, its influence on the assembly of root-associated microbial communities remains context dependent and is still poorly understood.

Plant growth–promoting rhizobacteria (PGPR) provide another promising strategy for enhancing rice performance under salinity stress. These microorganisms can improve plant tolerance through multiple mechanisms, including the production of phytohormones, facilitating nutrient acquisition, and modulation of stress-related signaling pathways [10,11,12]. However, the effectiveness of PGPR in saline soils is inconsistent. Successful inoculation depends not only on the functional traits of the introduced strains but also on their ability to colonize roots, compete with native microbial communities, and persist under host-mediated selection pressures [13]. Increasing evidence indicates that many inoculants fail to establish beyond early growth stages, particularly when soil physicochemical conditions are strongly altered [14]. To address these limitations, the combined application of biochar and PGPR has recently been proposed as a potentially synergistic approach to improve soil quality and crop productivity [15,16]. While some studies report enhanced nutrient availability and microbial activity following co-application, others observe neutral or even antagonistic effects [17]. These contrasting results highlight a critical knowledge gap: the mechanisms through which biochar-mediated niche modification interacts with host-driven microbial filtering and PGPR establishment remain poorly understood, particularly in saline rice systems.

In rice paddies, microbial communities are organized along a soil–root continuum that includes bulk soil, rhizosphere soil, the rhizoplane, and the root endosphere, with each compartment representing a distinct ecological niche shaped by progressively stronger plant selection. This spatial framework provides a powerful basis for disentangling the relative contributions of soil amendments and host-driven processes to microbial assembly. However, how biochar amendment and PGPR inoculation interact across these compartments to influence microbial community structure and function remains largely unknown. Here, we conducted a controlled pot experiment using seawater rice grown in coastal saline–alkali soil to test the hypothesis that biochar primarily modifies soil chemical constraints in outer soil compartments, whereas PGPR effects become increasingly constrained or reshaped in root-associated niches by host filtering. Specifically, we investigated: (i) whether biochar and PGPR exert synergistic or compartment-specific effects on bacterial community composition, (ii) the relative importance of soil chemistry versus host-mediated selection in structuring root-associated microbiomes, and (iii) how these processes shape microbial network organization and predicted functional potential. By integrating multi-compartment sampling with high-resolution bioinformatic analyses, this study aims to clarify the mechanisms underlying biochar–PGPR–plant interactions in saline–alkali rice systems.

2. Materials and Methods

2.1. Soil Collection, Lychee Biochar and Bacterial Agent Preparation

Soil samples were collected from a coastal barren mudflat of the Jujue reclamation area (32°30′16″ N, 121°10′52″ E), which is part of the North Jiangsu radial sand ridges. The soil primarily consists of fine sand and silt, with median grain sizes ranging from 8 to 63 μm [18]. The provenance of the soil is estimated to be approximately 70–73% of the sediment originates from the abandoned Yellow River, while 20–29% is derived from the Yangtze River and associated sand ridges [19]. The soil is characterized by high salinity and alkalinity, together with low availability of phosphorus, nitrogen, and organic carbon [20,21]. Prior to experimental treatment, the soil exhibited an initial pH of 9.35 and an electrical conductivity (EC) of 4.68 mS cm^−1^.

Lychee biochar, derived from the wood of Litchi chinensis, was obtained from a commercial supplier in China. According to the manufacturer, the biochar was produced using traditional kiln carbonization, in which dried lychee wood is carbonized under oxygen-limited conditions by staged sealing of the kiln, rather than by precise control or continuous monitoring of temperature. During production, key pyrolysis parameters, including in situ temperature, heating rate, and residence time were not recorded; therefore, detailed pyrolysis conditions cannot be specified. After production, the biochar was ground and sieved to a particle size of <3 mm prior to use. The biochar was characterized before the pot experiment, exhibiting a carbon content exceeding 95%, a BET surface area of 1530–2450 m^2^ g^−1^, 95% microporosity (pore diameter <2 nm), and a pH of 7.9. These measured properties provide an integrated description of the biochar used in this study and support its consistency across treatments.

The bacterial inoculant used in this study was strain SHZ02, which belongs to the genus Stenotrophomonas. Based on whole-genome sequencing and genome-level comparative analyses, including average nucleotide identity and digital DNA–DNA hybridization, strain SHZ02 was identified as a novel species and designated Stenotrophomonas shiheziensis SHZ02. The strain has been deposited in the Guangdong Microbial Culture Collection Center (GDMCC No. 66422), and its whole-genome sequence is available in the National Genomics Data Center under accession number CRA028892. Strain SHZ02 has been characterized as a plant growth-promoting rhizobacterium (PGPR) with multiple functional traits related to nutrient mobilization and plant growth promotion. Detailed physiological and functional characterization of this strain will be reported elsewhere.

2.2. Pot Experiment Design and Samples Collection

This pot experiment was conducted to evaluate the effects of biochar amendment and bacterial inoculation on soil properties and plant–soil interactions. Lychee biochar was applied at two rates (2.5% and 5.0%, w/w). The pot experiment was established in 200 L high-density polyethylene flanged drums (95 cm × 58 cm) made of, each filled with 300 kg of soil. Biochar was incorporated only into the surface soil layer (0–20 cm) at three levels (0%, 2.5%, and 5.0%, w/w). These application rates were selected to represent moderate and high biochar inputs commonly used in controlled pot and microcosm experiments (typically 1–10% w/w), which are intended to induce measurable physicochemical and microbial responses within a short experimental period rather than to simulate field-scale amendment practices [22].

Six treatments were established: (1) CK, control without bacterial inoculation; (2) BC25, CK supplemented with 2.5% lychee biochar; (3) BC50, CK supplemented with 5.0% lychee biochar; (4) STCK, CK with Stenotrophomonas shiheziensis SHZ02 inoculation before transplantation; (5) STBC25, STCK supplemented with 2.5% lychee biochar; and (6) STBC50, STCK supplemented with 5.0% lychee biochar. Rice (Oryza sativa L.) seeds were surface-sterilized with 6% H_2_O_2_ solution, rinsed with distilled water, and incubated at 25 °C for germination. Before transplantation on 2 June 2024, rice seedlings were soaked for 2 h in a biochar–bacterial suspension containing Stenotrophomonas shiheziensis SHZ02 (10^6^ CFU mL^−1^), in which biochar and the bacterial suspension were mixed at a ratio of 1:1000 (w/w). Biochar was used as a carrier to facilitate bacterial attachment to the root surface. Two weeks post-transplantation, only one seedling was retained per pot.

Rice plants and soil samples were collected on 4 November 2024. Bulk soil samples were obtained from soil not adhering to the roots, while rhizosphere soil was collected by gently shaking soil that remained attached to the root surface. Rhizoplane samples were prepared by pre-washing roots with sterile phosphate-buffered saline (PBS) solution, followed by sonication (40 Hz, 30 s) in 50 mL PBS in a 100 mL flask. The sonication process was repeated 4–6 times until the root color changed from yellow to light yellow or white. The resulting suspension was filtered through a 0.22 μm membrane, which was subsequently used for DNA extraction. After sonication, the roots were surface-sterilized with 2.5% sodium hypochlorite for 5 min, rinsed with 70% ethanol for 30 s, and washed five times with sterile distilled water. The final rinse water was plated using the spread-plate method to verify surface sterility. All samples were stored at −80 °C until further analysis.

2.3. Soil Physicochemical Properties Analysis

Soil physicochemical properties, including total carbon (TC), total nitrogen (TN), pH, electrical conductivity (EC), dissolved organic carbon (DOC), available phosphorus (AP, Olsen-P), soil organic matter (SOM), soil organic carbon (SOC), ammonium (NH_4_^+^-N), and nitrate (NO_3_^−^-N), were determined. TC and TN were quantified using a Vario EL Cube Elemental Analyzer (Elementar, Langenselbold, Germany). SOM was determined by the loss-on-ignition (LOI) method. Briefly, approximately 3 g of soil was placed in a ceramic crucible, dried at 105 °C for 2 h to remove moisture, and weighed. The dried sample was then combusted in a programmable muffle furnace (XD433-12, Shanghai Mafu Furnace Technology Instrument Co., Ltd., Shanghai, China) at a controlled heating rate of 8 °C/min to 550 °C, held for 4 h, cooled in a desiccator, and reweighed. Soil moisture content (MS) was identified by the weight loss method under 105 °C for 12 h. SOM content was calculated as the percentage of weight loss relative to the initial dry soil weight. The same sample was subsequently subjected to LOI at 950 °C to decompose carbonates, with SIC calculated based on weight loss between 550 °C and 950 °C, representing CO_2_ release from carbonate decomposition, and converted to carbon content using a molecular weight ratio of CO_2_ to C (3.67). SOC was derived as the difference between TC and SIC.

Soil pH and EC were measured in a 1:5 (w/v) soil-to-water suspension using a portable pH/EC meter (Mettler Toledo, Shanghai, China). DOC was extracted by shaking 5 g of air-dried soil with 25 mL of deionized water at 150 rpm for 30 min, followed by filtration through filter paper and a 0.45 μm membrane. The DOC concentration was then determined using a Multi N/C 3100 Analyzer (Analytik Jena AG, Jena, Germany). AP was extracted using 0.5 M NaHCO_3_ (pH 8.5) and quantified via molybdenum-antimony colorimetry method with a UV spectrophotometer. Soil ammonium (NH_4_^+^-N) was determined using a 1 M KCl extract and a modified indophenol blue method, employing sodium nitroprusside, sodium salicylate, sodium hydroxide, and sodium dichloroisocyanurate as reagents. Nitrate (NO_3_^−^-N) was quantified using a modified reduction-based colorimetric method with copper sulfate, zinc sulfate, sodium hydroxide, hydrazine sulfate, and naphthylethylenediamine dichloride.

2.4. DNA Extraction and Sequencing

Genomic DNA from soil microbial communities was extracted using the FastDNA Spin Kit for Soil (Bio 101, Carlsbad, CA, USA), following the manufacturer’s protocol. A 0.5 g (wet weight) soil sample was used for the extraction, and the obtained DNA was dissolved in 50 μL of nuclease-free deionized water. DNA concentration and purity were assessed using a Thermo Scientific NanoDrop-2000 UV-Vis Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Additionally, a 1% agarose gel electrophoresis was conducted to verify DNA integrity.

To profile the bacterial community, the V3–V4 region of the 16S rRNA gene was amplified using the primer pair 338F/806R. Unique barcode sequences were incorporated into the reverse primer to differentiate between samples. The PCR reaction was carried out in a 50 μL system containing 25 μL of 2X GoTaq^®^ Green Master Mix (Promega, Madison, WI, USA), 10 μL of a 10-fold diluted DNA template, and 1.5 μL of each primer (10 μM). PCR amplification was conducted using a T960c Thermal Cycler (Heal Force, Shanghai, China) under the following conditions: initial denaturation at 95 °C for 5 min, followed by 35 cycles of denaturation at 95 °C for 30 s, annealing at 53 °C for 45 s, and extension at 72 °C for 45 s, with a final elongation step at 72 °C for 5 min. A negative control (without a DNA template) was included to ensure no contamination in the PCR reaction. Each sample was amplified in triplicate, and the resulting PCR products were pooled. The amplicons were purified, pooled in equimolar concentrations, and used to construct a sequencing library with a MiSeq Reagent Kit V3. The library was sequenced on an Illumina MiSeq PE300 platform (Illumina, San Diego, CA, USA) by Majorbio (Shanghai, China). The raw sequencing data of the soil microbial community have been deposited in the Genome Sequence Archive via the BIG submission portal (CRA024037, CRA024038).

2.5. Bioinformatic Data Processing

Raw Illumina reads underwent a standardized processing pathway that included demultiplexing, quality filtering, read merging, and amplicon sequence inference. Quality control and trimming were performed with fastp v0.23.2 [23], paired-end overlaps were merged using FLASH v1.2.7 [24], and DADA2 in QIIME 2 was used to resolve amplicon sequence variants (ASVs). Taxonomic classification of the resulting ASVs was assigned via the naïve Bayes classifier against the SILVA 16S rRNA v138 reference database. To evaluate soil bacterial α-diversity, we calculated the following indices: Chao1 and Sobs (species richness), Shannon diversity, Pielou’s evenness, and Faith’s phylogenetic diversity (Faith-PD). Co-occurrence networks were constructed in Wekemo Bioincloud (https://www.bioincloud.tech) [25], selecting the top 200 ASVs per network based on Spearman’s r > 0.6 and p < 0.05, and visualized using Gephi v0.9.2. Functional profiles of the bacterial community were inferred with FAPROTAX [26], while PICRUSt2 was used to predict KEGG Ortholog (KO) abundances, focusing on key genes in the C, N, P, and S cycles. Differential bacterial communities among different treatments were identified via DESeq2, and results were displayed as Sankey and volcano plots in Wekemo Bioincloud. Finally, a random forest analysis (R package randomForest, R v4.3.2) was performed to determine the relative importance of soil physicochemical variables in shaping community structure.

2.6. Statistical Analysis

Univariate analysis of variance (ANOVA) was applied to detect treatment-related differences in soil physicochemical parameters at a significance threshold of p < 0.05, with mean separations evaluated by Duncan’s multiple range test. Microbial diversity indices were compared across treatments using the nonparametric Kruskal–Wallis test [27]. To exmine how environmental variables shape bacterial community structure at the genus level, we conducted canonical correspondence analysis in Brodgar v2.7.5. Twelve soil factors: pH, EC, TC, TN, C:N, AP, DOC, SOC, SOM, MS, NH_4_^+^-N, and NO_3_^−^-N, were included in the ordination. Explanatory variables were standardized using a weighted scaling method, and all parameters were centered and scaled based on inter-Y correlations. Significance of environmental drivers was determined by forward selection guided by the Akaike information criterion (AIC), with 1000 permutations used to assess statistical support.

3. Results

3.1. Soil Physicochemical Properties

Biochar amendment substantially altered soil physicochemical conditions, and these effects varied clearly between soil compartments (bulk soil and rhizosphere) and between inoculated and non-inoculated treatments (Table 1). Across all treatments, the rhizosphere responded more strongly than bulk soil, indicating an important interaction among plant roots, biochar, and microbial activity. In bulk soil, pH remained relatively stable following inoculation with Stenotrophomonas inoculation, whereas a pronounced pH decline was observed in the rhizosphere under high biochar input (BC50). This localized acidification likely associated with intensified root–microbe interactions and the production of organic acid in the rhizosphere, rather than with a direct chemical effect of biochar alone. By contrast, biochar addition without bacterial inoculation resulted in smaller and more variable changes in pH, suggesting that microbial processes contributed to the regulation of biochar-induced buffering effects. EC increased markedly under BC50, particularly in the rhizosphere of inoculated soils, indicating enhanced ion retention or mobilization in root-influenced zones. This increase was much weaker in non-inoculated soils, implying that Stenotrophomonas inoculation amplified biochar-driven salinity-related changes at the soil–root interface. Carbon and nutrient pools exhibited clear compartment-specific responses. High-dose biochar significantly increased SOC, SOM, and C:N ratios, especially in the rhizosphere under bacterial inoculation. The elevated C:N ratios under these conditions indicate greater carbon availability relative to nitrogen, which may favor copiotrophic and biochar-adapted microbial taxa. DOC showed contrasting patterns: co-application of biochar and bacterial inoculation tended to stabilize DOC concentrations in the rhizosphere, whereas biochar applied alone resulted in a greater variability. AP showed the most pronounced compartmental divergence. Under bacterial inoculation, AP preferentially accumulated in the rhizosphere at the high biochar rate, and exceeded the corresponding concentrations in bulk soil, while biochar alone produced inconsistent responses of AP across compartments. This pattern indicates that microbial inoculation enhanced biochar-mediated phosphorus mobilization in the root-associated zone. Collectively, these results demonstrate that the effects of biochar on soil chemistry are not uniform across the soil-root system. Instead, they are strongly modified by rhizosphere processes and bacterial inoculation, leading to the formation of distinct physicochemical niches that are likely to influence subsequent microbial community assembly.

3.2. Microbial Diversity and Community Structure

After quality filtering and rarefaction, sequencing depth was sufficient to capture bacterial diversity across all samples, as indicated by the saturation of rarefaction curves (Figure S1). In total, 28,943 ASVs were retained for subsequent analyses. Clear spatial and treatment-related patterns were observed across the four compartments, including bulk soil (BS), rhizosphere soil (RS), root rhizoplane (RP), and root endosphere (ES) (Figure S2). Compartmental differentiation was the primary determinant of bacterial α-diversity. Across all treatments, richness and diversity indices (Sobs, Chao1, Shannon, and Pielou’s evenness) declined progressively from BS to ES, and the endosphere consistently exhibited the lowest richness and evenness. This pattern reflects strong plant-mediated filtering during the recruitment of microorganisms from the surrounding soil into root associated niches. The effects of biochar addition on bacterial diversity differed markedly among compartments and application rates. Low-dose biochar (BC25) significantly reduced bacterial richness and diversity in BS and RS, whereas these reductions in RS were partially alleviated when bacterial inoculation was applied. In contrast, both BC25 and BC50 consistently reduced richness and evenness in RP, regardless of inoculation, indicating that the rhizoplane community was particularly sensitive to biochar amendment. Distinct responses were observed in the endosphere. Specifically, BC25 increased bacterial richness (Sobs and Chao1), while BC50 increased phylogenetic diversity (Faith’s PD). These results suggest that moderate biochar inputs may partially relax host filtering and allow colonization by a broader phylogenetic spectrum. Bacterial inoculation alone increased endosphere richness under control conditions and altered biochar responses in root-associated compartments. Notably, the combined application of Stenotrophomonas and BC25 mitigated diversity losses in the rhizosphere, whereas the combination of bacterial inoculation with high-dose biochar (BC50) resulted in pronounced diversity reductions in the rhizoplane. Together, these results indicate that the effects of biochar on bacterial diversity are strongly context dependent and are jointly shaped by microbial inoculation and plant compartment.

3.3. Shift of Microbial Community Under Bacterial Inoculation and Biochar Addition

Across all samples, bacterial communities were dominated by four major phyla—Bacillota, Bacteroidota, Pseudomonadota, and Chloroflexota—which together accounted for most of the relative abundance (Figure 1). Pronounced and consistent compositional shifts were observed along the soil–root continuum, irrespective of treatment. The relative abundances of Bacillota and Bacteroidota increased progressively from bulk soil (BS) to the root endosphere (ES), whereas Chloroflexota declined sharply along the same gradient, indicating strong habitat filtering toward root-associated niches. These patterns were maintained at finer taxonomic resolution. Taxa affiliated with Bacteroidota, including Bacteroidia, Flavobacteriales, and Chryseobacterium, were consistently enriched in the endosphere, while Chloroflexota-related groups, such as Anaerolineae and Caldilineales, decreased from soil to root compartments. Members of Bacillales and their associated families reached their highest relative abundances in the rhizosphere, indicating preferential colonization of root-proximal soil.

Biochar amendment and bacterial inoculation further modified these baseline compartmental patterns. The combined application of biochar and bacterial inoculation increased the relative abundance of Pseudomonadota in the rhizoplane (RP), suggesting a partial relaxation of rhizoplane filtering under combined treatments. Within Bacillota, several genera—most notably Exiguobacterium—were consistently enriched in both RP and ES, with the highest relative abundances observed in biochar-amended and inoculated treatments. LEfSe analysis identified distinct compartment-specific biomarkers among treatments (Figure 2 and Figure S3). The rhizosphere and rhizoplane compartments contained the greatest numbers of differentially enriched taxa, whereas bulk soil and the endosphere harbored fewer but more distinct biomarkers. Bacterial inoculation increased the number of enriched phyla relative to non-inoculated treatments, with Actinomycetota, Bacillota, Bacteroidota, and Pseudomonadota being consistently associated with inoculated systems. Differential abundance analysis based on DESeq2 further supported these patterns, identifying 25–30 significantly regulated taxa per compartment in response to biochar addition and bacterial inoculation (Figures S4–S7). Low-dose biochar generally resulted in a greater number of upregulated than downregulated taxa, while high-dose biochar produced stronger treatment × compartment interactions. Several genera, including Sporacetigenium, Pseudazoarcus, Anaeromyxobacter, and Pedobacter, showed consistent responses across multiple compartments, whereas members of the family Sphingomonadaceae exhibited widespread downregulation. Taken together, these results demonstrate that biochar amendment and bacterial inoculation reshape microbial community composition in a compartment-specific manner, superimposed on strong background of plant-driven selection along the soil-root continuum.

3.4. Relationship Between Soil Properties and Bacterial Composition

Canonical correspondence analysis (CCA) revealed significant associations between soil physicochemical properties and bacterial community composition (Figure 3a). The first two canonical axes explained 24.5% of the total variance, with samples primarily separated according to biochar amendment and bacterial inoculation. Multiple environmental variables significantly contributed to community structuring, including pH, DOC, SOC, SOM, TC, TN, NH_4_^+^-N, and NO_3_^−^-N. Biochar-amended soils without bacterial inoculation formed distinct clusters along the CCA axes, whereas bacterial inoculation reduced this separation, indicating that inoculation moderated the compositional shifts induced by biochar. Random forest analysis produced a consistent pattern and identified pH as the most important predictor of bacterial community structure, followed by DOC, SOC, the C:N ratio, and TN (Figure 3b). Together, these variables reflect changes in carbon availability and nutrient stoichiometry, suggesting that the influence of biochar on microbial assembly is largely mediated through modification of soil resource conditions. The consistency between the CCA and random forest results further indicates that bacterial inoculation may reduce the magnitude of biochar-associated community differentiation, potentially by buffering community responses to underlying physicochemical gradients.

3.5. Network Topology and Functional Predictions of Microbial Communities

To evaluate how biochar amendment and bacterial inoculation influenced microbial interactions beyond taxonomic composition, co-occurrence networks were conducted for each compartment (Figure 4 and Figure S8). All networks showed significantly higher connectivity, clustering coefficients, and modularity compared with randomized networks, indicating non-random community organization (Table S1). Network structure differed markedly among compartments. The bulk soil network contained the largest number of nodes and exhibited a wide range of inter-phyla associations, whereas networks in the rhizoplane and root endosphere showed higher clustering coefficients, suggesting more cohesive and tightly connected microbial assemblages. Positive correlations were dominant in bulk soil and the rhizoplane, while the rhizosphere and endosphere displayed comparatively simpler interaction patterns. Keystone taxa also varied among compartments. In bulk soil, network hubs were primarily affiliated with Chloroflexota and Pseudomonadota. In contrast, Bacillota-related hubs were prevalent in rhizosphere and rhizoplane, whereas endosphere networks were characterized by a limited number of hubs belonging mainly to Bacillota, Pseudomonadota, and Bacteroidota. At the genus level, Paenisporosarcina and Exiguobacterium were identified as keystone taxa in the rhzosphere and rhizoplane, respectively, whereas endosphere networks were distinguished by hubs such as Chryseobacterium, Comamonas, Exiguobacterium, and Pseudomonas.

Functional predictions derived from FAPROTAX and PICRUSt2 revealed clear compartment-dependent patterns in metabolic potential (Figure 5). The relative abundances of functions related to chemoheterotrophy and aerobic chemoheterotrophy increased progressively from bulk soil to the endosphere, indicating enhanced carbon utilization in root-associated habitats. Nitrate reduction was most prominent in the rhizoplane, whereas phototrophic functions were largely confined to bulk soil. Predicted functional gene profiles showed similar trends. Genes associated with central carbon metabolism, including glycolysis and the tricarboxylic acid cycle, were enriched in bulk soil and the rhizosphere, whereas genes related to aldehyde metabolism increased in root-associated compartments. Nitrogen cycling genes exhibited a shift from denitrification-related pathways in soil compartments toward dissimilatory nitrate reduction in the endosphere. In addition, sulfur cycling genes were predominantly enriched in soil compartments and declined toward the plant interior. Overall, these network and functional predictions indicate that microbial communities undergo coordinated changes in interaction structure and inferred metabolic potential along the soil–root continuum, and that both biochar amendment and bacterial inoculation modify these patterns in a compartment-dependent manner.

4. Discussion

4.1. Biochar and Bacterial Inoculation Modulate Soil Physicochemical Properties

The alkaline ash components and cation exchange capability of biochar are widely recognized as key drivers of soil pH regulation, often increasing pH in acidic soils and stabilizing pH alkaline systems [28]. In the present study, bulk soil pH in Stenotrophomonas-inoculated treatments remained relatively stable (8.01–8.21) across biochar levels, whereas a clear pH decline was observed in the rhizosphere under the high biochar application rate (BC50; 8.14 ± 0.25). This localized decrease in the rhizosphere pH is more likely attributable to intensified root-microbe interactions, including organic acid exudation and proton release associated with nutrient acquisition, rather than to a direct chemical effect of biochar itself. In contrast, pH changes in uninoculated bulk soil were small and inconsistent (e.g., 7.95 ± 0.02 under BC25), suggesting that microbial activity modulated biochar–soil buffering processes in the rhizosphere.

Increased application rates of biochar can release larger quantities of base cations (K^+^, Ca^2+^ and Mg^2+^) and ash-derived salts, thereby elevating soil electrical conductivity [29]. Consistent with this mechanism, EC increased markedly under BC50, particularly in Stenotrophomonas-inoculated soils (194.7 ± 4.4 µS cm^−1^ in bulk soil and 267.6 ± 25.4 µS cm^−1^ in the rhizosphere), whereas EC values in the uninoculated rhizosphere remained substantially lower (145.3–161.8 µS cm^−1^). These results indicate that the combined effects of biochar and microbial activity intensified ion mobilization or retention in the root-influenced zone.

The porous structure of biochar and its oxygen-containing surface functional groups can promote NH_4_^+^-N adsorption and reduce nitrogen losses through leaching [30]. In the present study, NH_4_^+^-N concentrations in Stenotrophomonas-inoculated soils remained relatively stable (0.52–0.56 mg kg^−1^) across biochar treatments, whereas greater variability was observed in uninoculated bulk soil, which reached a maximum of 0.59 mg kg^−1^ under BC25. This pattern suggests that microbial inoculation may contribute to a more buffered ammonium pool under biochar amendment.

Biochar is also known to enhance phosphorus availability through liming effects, desorption of phosphate from mineral surfaces, and stimulation of microbial P mineralization [31]. In the present study, available phosphorus in the inoculated rhizosphere increased substantially under BC50 (7.94 ± 0.38 mg kg^−1^), exceeding the corresponding bulk soil concentration (7.00 ± 0.29 mg kg^−1^). By contrast, available phosphorus in uninoculated bulk soil declined under BC50 (0.87 ± 0.07 mg kg^−1^), while the rhizosphere still exhibited an increase (7.45 ± 0.64 mg kg^−1^). These contrasting responses highlight the strong context dependency of biochar-mediated phosphorus dynamics and emphasize the role of rhizosphere processes and microbial activity in regulating phosphorus availability.

High-carbon biochar inputs are known to increase soil C:N ratios and may influence microbial nitrogen immobilization when the overall C:N ratio exceeds critical thresholds (approximately 32:1) [32]. In this study, the C:N ratio in the inoculated rhizosphere reached 63.77 ± 3.50 under BC50, compared with 59.03–60.63 in the uninoculated rhizosphere, indicating substantial carbon enrichment under combined biochar and microbial treatments. In addition, biochar can release labile dissolved organic carbon (DOC) through surface desorption and leaching of low-molecular-weight compounds [33]. Consistent with this mechanism, DOC in inoculated bulk soil reached a maximum of 94.50 ± 2.34 mg kg^−1^ under BC25. In contrast, DOC in the rhizosphere decreased under BC50 (53.33 ± 4.33 mg kg^−1^), which likely reflects rapid microbial assimilation and/or adsorption of dissolved organic compounds onto biochar surfaces. In uninoculated soils, DOC remained comparatively stable under BC50 (85.23 ± 6.23 mg kg^−1^), further indicating that microbial activity strongly modified DOC dynamics in the rhizosphere.

Meta-analyses have reported that biochar application can increase soil organic carbon (SOC) by 30–100% in agricultural and grassland systems, thereby promoting carbon sequestration and improving soil aggregate stability [34]. In agreement with these findings, SOC and SOM in inoculated bulk soil increased markedly under BC50 (SOC, 1.84 ± 0.31%; SOM, 7.89 ± 1.47%), compared with the corresponding control values (SOC, 0.07 ± 0.01%; SOM, 1.30 ± 0.19%). The nitrogen content of biochar and its indirect effects on nitrogen cycling may also contribute to moderate increases in total nitrogen, particularly when biochar is combined with microbial inoculants [35]. In the present study, total nitrogen in inoculated bulk soil reached 0.046 ± 0.007% under BC50. In parallel, total carbon generally increased with higher biochar inputs, consistent with previous meta-analyses reporting approximately 64% increases in total carbon at elevated application rates [36].

Notably, carbon dynamics differed between compartments. In uninoculated bulk soil, total carbon increased to 4.66 ± 0.48% under BC50, whereas in the inoculated rhizosphere, total carbon declined from 2.29 ± 0.05% under BC50 to 1.89 ± 0.14% under BC25, indicating strong compartment-specific regulation of carbon pools and possible enhancement of microbial turnover in the rhizosphere. Overall, the observed improvements in nutrient retention, carbon storage, and microhabitat conditions under the combined application of biochar and Stenotrophomonas suggest complementary benefits for soil fertility and resilience. At the same time, the pronounced decrease in rhizosphere pH under high biochar input indicates that biochar–microbe interactions can generate localized acidification, highlighting the need for further investigation of their longer-term consequences for plant–microbe interactions in saline–alkali soils.

4.2. Shift of Bacterial Community Under Biochar and Bacterial Inoculation Condition

A clear decline in bacterial richness and diversity was observed from bulk soil to the rhizosphere, rhizoplane, and root endosphere. The consistently lower α-diversity in the endosphere provides strong evidence that rice roots exert a pronounced filtering effect on the surrounding soil microbiome. This pattern is consistent with the widely accepted two-step assembly model [37], in which microorganisms are first enriched in the rhizosphere and subsequently a subset of taxa colonizes the internal root tissues. Such hierarchical filtering across physicochemical and plant-imposed barriers has been well documented in rice systems [38]. The pronounced decline in Chloroflexota and the concomitant enrichment of Bacillota and Bacteroidota in the endosphere further support niche specialization driven by root exudates and oxygen gradients [39], highlighting the predominance of host-mediated selection over bulk soil conditions in shaping endophytic bacterial communities.

Changes at both the phylum and genus levels further illustrate this filtering process. The relative abundance of Bacillota increased steadily from bulk soil to the endosphere, consistent with its classification as a root-associated bacterial group [40]. In particular, the enrichment of Bacillota-affiliated genera such as Exiguobacterium in the rhizoplane and endosphere can be attributed to the halotolerant and stress-resistant traits commonly reported for Exiguobacterium species, which are frequently detected within plant tissues [41]. Similarly, the increased abundance of Chryseobacterium (Bacteroidota) in the endosphere agrees with previous studies showing that several Chryseobacterium strains are beneficial to plants [42]. In contrast, soil-adapted groups such as Chloroflexota tended to decline toward the root interior. These taxonomic shifts indicate that only bacteria possessing appropriate functional traits, such as motility, biofilm formation and stress tolerance, are able to penetrate and persist in the endosphere. This interpretation is consistent with genomic evidence from plant-associated bacteria, including Stenotrophomonas, which harbor genes related to motility and biofilm formation that facilitate root colonization [43].

Biochar amendment altered bacterial community composition in a dose-dependent and compartment-specific manner. The reduction in bacterial diversity observed in bulk soil, rhizosphere and rhizoplane under BC25 (Figure S2) may be associated with biochar-induced changes in soil pore structure and carbon availability that selectively favor oligotrophic taxa [22]. The alleviation of these negative effects in the rhizosphere under Stenotrophomonas inoculation suggests that strain SHZ02 may enhance nutrient solubilization or buffer pH fluctuations, thereby maintaining greater niche heterogeneity, a property commonly associated with plant growth-promoting rhizobacteria [44]. In contrast, the persistent sensitivity of the rhizoplane community to biochar, regardless of inoculation, indicates that the root surface represents a highly selective interface. At this boundary, biochar-induced changes in attachment sites or chemotactic signals may override the influence of introduced inoculants [45].

Notably, BC25 increased bacterial richness in the endosphere, in contrast to its inhibitory effects in the outer compartments. This contrasting response may reflect the selective recruitment of beneficial taxa, such as Chryseobacterium, potentially facilitated by enhanced root exudation under moderate carbon supplementation [46]. By comparison, the increase in Faith’s phylogenetic diversity observed under BC50 suggests that higher biochar inputs allow phylogenetically more diverse taxa to establish in the endosphere, possibly through the adsorption of inhibitory compounds (e.g., phenolics) or modification of local redox conditions [47]. Consistent with these trends, DESeq2 analysis identified several genera that were repeatedly enriched by biochar across compartments, including Sporacetigenium, Pseudazoarcus, Anaeromyxobacter and Pedobacter. These taxa include anaerobic fermenters and decomposers that may exploit recalcitrant carbon fractions or anoxic microsites created within biochar-amended soils (e.g., Anaeromyxobacter and Pedobacter are capable of using a wide range of substrates for respiration) [48,49]. In contrast, members of the family Sphingomonadaceae were generally suppressed following biochar amendment. Many Sphingomonas species are considered oligotrophic and may therefore be disadvantaged under the altered chemical and resource conditions created by biochar addition [50].

Inoculation with Stenotrophomonas further modified spatial differentiation of bacterial communities, particularly in the endosphere under control conditions (STCKES vs. CKES). This observation is consistent with the proposed role of certain PGPR as keystone taxa that may influence root immune responses or cross-kingdom signaling processes involved in microbiome recruitment [51]. The higher number of indicator taxa detected by LEfSe in inoculated treatments relative to non-inoculated controls suggests that Stenotrophomonas introduced additional selection pressures or microhabitats that promoted community differentiation. Moreover, the enrichment of Paenisporosarcina in STBC50-treated compartments implies that inoculation may enhance the ability of biochar-amended soils to recruit stress-tolerant taxa, potentially through co-metabolism of recalcitrant carbon substrates [52]. However, the inability of Stenotrophomonas to mitigate BC50-induced diversity losses in the rhizoplane highlights the limits of microbial inoculation under strong abiotic constraints, where imbalances in resource stoichiometry (e.g., C:N:P) may weaken mutualistic interactions and network stability [53].

The dominance of Bacillota in the rhizoplane and endosphere further reflects their adaptation to nutrient-rich and microaerophilic conditions, as their facultative anaerobic metabolism is well suited to root-derived exudates [37]. The prevalence of Bacteroidota, including Chryseobacterium, likely indicates functional specialization for the degradation of plant-derived polysaccharides [54]. In addition, the persistence of Pseudomonadota in the rhizoplane under combined inoculation and biochar amendment (Figure 1) suggests enhanced capacities for biofilm formation and chemotactic responses. Finally, LEfSe results showed that Stenotrophomonas inoculation promoted taxa affiliated with Actinomycetota and Thermodesulfobacteriota, which are commonly associated with secondary metabolite production and sulfur cycling, respectively, indicating that functional reorganization accompanied the observed taxonomic restructuring.

4.3. The Assembly of Bacterial Community and Function Under Biochar and Bacterial Inoculation Condition

Network topology provided additional evidence for strong compartmental specialization of bacterial communities. All compartment-specific networks (bulk soil, rhizosphere, rhizoplane and endosphere) exhibited clear modular structures (Louvain Q  >  0.3), indicating the presence of discrete clusters of co-occurring taxa. Networks associated with the rhizoplane and endosphere showed pronounced clustering (average clustering coefficient > 0.5), whereas the bulk-soil network was larger in size and comparatively less clustered. In line with previous observations [55], the bulk-soil network in our study contained the highest numbers of nodes and edges, reflecting a more taxonomically complex and loosely connected assemblage.

By contrast, the rhizoplane and endosphere networks were organized into more compact modules and were dominated by positive correlations, resulting in higher clustering coefficients. Similar patterns have been reported in other plant-associated systems. For example, highly clustered fungal networks were previously observed in the rhizoplane [56], and bacterial networks within the endosphere of grafted apple trees were found to be dominated by co-occurrence relationships, whereas rhizosphere networks showed a mixture of positive and negative associations [57]. Consistent with these studies, our results suggest that microbial communities within root tissues tend to form tightly connected and predominantly cooperative consortia, whereas bulk soil supports more extensive and weakly connected interaction networks.

Soil chemistry properties closely tracked these patterns in community organization. Both canonical correspondence analysis and random forest analyses identified indices of C and N availability as major drivers of community composition, with DOC, TN, SOC, pH, and NH_4_^+^-N ranking among the most influential environmental predictors. Similar relationships between soil nutrient status and microbial community structure have been reported previously. For instance, soil TN, NO_3_^−^-N, pH, EC and organic matter together explained approximately 26% of the variation in bacterial communities in the rhizosphere of Lycium barbarum L. [58]. In the present study, variation in pH and C:N conditions appears to regulate how root exudates and soil amendments influence microbiome assembly. Importantly, Stenotrophomonas inoculation modified these relationships. In several cases, inoculated treatments exhibited reduced community divergence across biochar levels, suggesting that the introduced PGPR dampened the sensitivity of bacterial communities to changes in soil chemistry. This apparent buffering effect may arise from stabilization of nutrient cycling processes or competitive interactions with taxa that would otherwise respond strongly to the altered resource regime [13,14]. Collectively, the combined CCA and random forest results indicate that the effects of both plants and soil amendments on microbial assembly are largely mediated through changes in root-zone chemistry, particularly carbon availability, nitrogen status and pH, underscoring the complexity of host–soil–microbe interactions.

Functional predictions based on FAPROTAX and PICRUSt2 further revealed a clear compartmental gradient in metabolic potential. The relative abundance of functions associated with aerobic chemoheterotrophy and chemoheterotrophic metabolism increased progressively along the soil–root continuum (bulk soil < rhizosphere < rhizoplane < endosphere), indicating strong selection for heterotrophic lifestyles in the carbon-rich root environment. This functional shift is consistent with increasing availability of plant-derived carbon substrates as microorganisms approach and colonize root tissues [59]. In contrast, many nutrient-cycling functions declined as microbial habitat shifted from soil into the plant interior. Predicted genes related to carbon metabolism exhibited greater variability among compartments than those associated with nitrogen, phosphorus, or sulfur cycling, Suggesting a progressive narrowing of metabolic breadth and increasing functional specialization in the endosphere.

Notably, phototrophic functions were enriched in bulk soil under Stenotrophomonas-only treatments (STCKBS), indicating that biochar-derived microhabitats may favor the establishment of cyanobacteria, consistent with the light-absorbing and surface properties of biochar [44]. In addition, the predominance of nitrate reduction functions in the rhizoplane, including genes such as norB and nosZ, points to enhanced denitrification activity at the root surface, likely driven by root exudation and localized oxygen limitation. Denitrification represents an important pathway for nitrogen loss in rice production systems, and similar compartment-specific functional shifts have been reported previously [37,60], where changes in microbial community structure during rice growth were closely coupled with alterations in metabolic capacity.

5. Conclusions

Under the specific conditions of this pot experiment—using a single rice cultivar, one bacterial inoculant, one saline–alkali soil, and two biochar application rates—our results demonstrate that host-associated compartmentalization is the primary factor structuring bacterial communities along the soil–root continuum. Across all treatments, microbial diversity, community composition, and interaction networks were governed mainly by progressive host filtering from bulk soil to the root endosphere, whereas biochar amendment and PGPR inoculation exerted secondary and compartment-dependent influences.

The effects of biochar and bacterial inoculation on microbial assembly were expressed largely through indirect modification of soil physicochemical conditions. These effects were most pronounced in bulk soil and the rhizosphere and became increasingly constrained toward the plant interior. Co-occurrence network analyses further revealed tighter clustering of root-associated communities, indicating enhanced microbial co-selection under host control rather than increased ecosystem stability. In addition, functional profiles derived from predictive approaches suggested clear spatial differentiation in metabolic potential across compartments; however, these patterns represent inferred functional capacities rather than directly measured process rates.

By elucidating how plant-mediated selection constrains microbial responses to soil amendments in saline–alkali soils, this study provides mechanistic insight that is particularly relevant for regions where soil salinization continues to limit crop production and threaten food security. Future studies incorporating multiple host genotypes, agronomic performance indicators, and field-scale experiments will be essential to determine how these microbiome assembly principles translate into sustainable productivity outcomes under realistic agricultural conditions.