Pepper (Capsicum annuum L.) seeds with uniform size and viability were obtained from the Agricultural Science Research Institute of Loudi City, Hunan Province, China. The seeds were surface-sterilized by immersion in a 2% (v/v) sodium hypochlorite (NaClO) solution for 30 min, followed by five rinses with sterile deionized water. After 25 days of soil germination under controlled conditions (14 h photoperiod, 25/20°C day/night temperature, and 80% relative humidity), morphologically homogeneous seedlings were selected and transplanted into experimental pots (20 cm diameter × 18 cm height) containing prepared growth substrate.

The Sb-tolerant PGPR strain Cupriavidus sp. S-8-2 was isolated from the rhizosphere of ferns collected from the XKS mine, according to our previously published protocol (Zheng et al., 2023). The bacterium was cultured in lysogeny broth (LB) medium (10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl) at 30°C with shaking at 120 rpm for 48 h. Cells were harvested during mid-exponential growth phase (OD₆₀₀ ≈ 1.0) by centrifugation at 5,000 g for 10 min, washed twice with sterile physiological saline (8.5 g/L NaCl), and resuspended to a final concentration of 1 × 10⁹ CFU/mL as determined by OD₆₀₀ calibration.

Surface soil (0–20 cm depth) was sampled from uncontaminated agricultural fields at the Jiuer Experimental Station, located in Loudi City (27°44’N, 111°59′E), with the background Sb concentration of 1.96 mg/kg. The soil exhibited a yellow-brown color with the following physicochemical properties: organic matter content of 18.2 g/kg, cation exchange capacity of 15.3 cmol/kg, total phosphorus content of 1.97 g/kg, total nitrogen content of 3.22 g/kg, and pH of 5.41. Following air-drying, homogenization, and sieving through a 2 mm mesh to remove debris, soils were amended with potassium antimony tartrate (C₈H₄K₂O₁₂Sb₂, analytical purity, Sigma) to achieve target Sb concentrations of 500 and 1,000 mg/kg Sb (dry weight basis), based on previously reported low-to-moderate contaminations levels in agricultural soils near Sb mining areas (Kong et al., 2024). The contaminated soils were then equilibrated for 28 days at 25 ± 1°C, with moisture maintained at 60% water-holding capacity, to facilitate Sb aging.

A pot experiment was carried out from April to July 2024 to evaluate the alleviation of Sb stress in pepper plants through inoculation with Cupriavidus sp. S-8-2. Morphologically uniform seedlings (25-day-old) were transplanted into pots (23.5 cm diameter × 29.0 cm height; 4 kg soil/pot). Six treatments were implemented: uninoculated control (0 mg/kg Sb; UCK), inoculated control (0 mg/kg Sb + Cupriavidus sp. S-8-2; ICK), uninoculated low Sb stress (500 mg/kg Sb; ULT), inoculated low Sb stress (500 mg/kg Sb + Cupriavidus sp. S-8-2; ILT), uninoculated high Sb stress (1,000 mg/kg Sb; UHT), and inoculated high Sb stress (1,000 mg/kg Sb + Cupriavidus sp. S-8-2; IHT). The experiment was conducted using a completely randomized block design, incorporating four biological replicates for each treatment group (4 biological replicates × 6 treatments = 24 pots in total). The bacterial suspension, containing 1 × 10⁹ CFU/mL, was prepared by resuspending pelletized cells in sterile distilled water following centrifugation (8,000 × g, 6 min) to remove residual growth media. It was then applied to the rhizosphere at 0, 30, and 45 days post-transplantation. Control treatments received equivalent volumes of sterile distilled water to ensure comparability and eliminate potential confounding effects associated with nutrient addition. The pots were maintained under natural photoperiod conditions, with daily irrigation to maintain 60–70% of the water holding capacity, at ambient temperature ranging from 19°C (night) to 28°C (day). Measured Sb concentrations in spiked soils showed 4.6% mean deviation from nominal values (<5.2% at 500–1000 mg/kg; Supplementary Figure S1).

Plant samples were harvested at the fruit-bearing stage (120 days after post-sowing). Four biological replicates per treatment were randomly collected following a stratified random sampling protocol. Rhizosphere soil (approximately 200 g), adhering to roots within 1–2 mm, was carefully collected using gentle brushing and shaking for subsequent analysis. Roots were systematically separated, and subsamples were processed as follows: one portion was air-dried and sieved (0.074 mm) for the determination of pH, TN, TP, and SOM; a second portion was analyzed for total Sb concentration; a third portion was immediately placed in sterile cryovials, transported under refrigeration at 4°C, and stored at −80°C for enzyme activity assays and microbial community analysis.

Root systems were meticulously excavated while maintaining their structural integrity. Plant organs (roots, stems, leaves, and fruits) were separated and sequentially washed with tap water followed by deionized water to remove surface particulates. Morphometric parameters (total root length, stem basal diameter measured with digital caliper ±0.01 mm, and fresh biomass recorded via analytical balance ±0.01 g) were quantified immediately after processing. For the quantification of Sb concentration, tissues were freeze-dried at −50°C, ground into particles smaller than 0.5 mm in size, and subsequently digested using microwave-assisted digestion. The Sb content in each tissue was determined by inductively coupled plasma mass spectrometry (ICP-MS, model 7500c, Agilent Technologies, United States), following established methodology (Zheng et al., 2023).

Each fresh leaf sample (0.2 g fresh weight) was collected at harvest, with midribs carefully excised. The samples were subsequently homogenized in 20 mL of ice-cold 95% ethanol (HPLC grade, Sigma-Aldrich) using a mortar and pestle. Extraction was proceeded in amber glass vials at 4°C for 18 h until complete tissue depigmentation was achieved. The extracts were then adjusted to a final volume of 50 mL with 95% ethanol, followed by centrifugation at 8,000 × g for 10 min at 4°C. The concentrations of Chlorophyll a and chlorophyll b were determined spectrophotometrically (Shimadzu UV-1800) based on absorbance measurements at 665 nm and 649 nm, respectively, and calculated using Lichtenthaler’s equations (1987).

For carotenoids quantification, fresh leaf tissues (0.2 g) were thoroughly pulverized in liquid nitrogen and extracted with an acetone: petroleum ether (1:1 v/v) mixture under dim light conditions. Following vortexing for 2 min, the samples were centrifugated at 8,000 × g for 10 min at 4°C. The resulting supernatants were collected, pooled, and subsequently evaporated under N₂ stream. The residues were then reconstituted in 5 mL of acetone, and absorbance was measured at 450 nm. The carotenoid content was determined according to the protocol outlined in the reference material (Zhang et al., 2020).

Soil pH was determined in a 1:2 (w/v) soil-water suspension using a calibrated pH meter. TN content was quantified via semi-micro-Kjeldahl digestion and subsequently analyzed using a TOC-TN analyzer (Vario EL III, Elementar, Germany). SOM was assessed through potassium dichromate (K₂Cr₂O₇) wet oxidation followed by colorimetric determination. TP was quantified after HF-HClO₄ digestion using the molybdenum blue method.

Rhizosphere soil enzyme activities were analyzed post-harvest. Urease activity was determined by quantifying ammonia release following urea substrate incubation. Acid phosphatase activity was assayed by measuring p-nitrophenol (PNP) liberation from p-nitrophenyl phosphate disodium (pH 6.0, 115 mM) after 1 h incubation at 37°C. Saccharase activity was evaluated using sucrose as the substrate with 3,5-dinitrosalicylic acid reagent, with reducing sugars measured at 508 nm. All enzyme activities were quantified spectrophotometrically (UV-1800, Shimadzu), following the protocol provided.

Fresh root tips were harvested and promptly snap-frozen in liquid nitrogen to halt metabolic activity, followed by storage at-80°C. Subsequently, samples (100 μg) were homogenized in ice-cold phosphate-buffered saline (PBS, pH 7.4) and centrifuged at 8,000 g for 20 min at 4°C. The resulting supernatants were filtered through double-layers cheesecloth to remove insoluble materials. Activities of SOD, POD, and CAT activities were quantified using commercial kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) according to the manufacturer’s protocols. Enzyme activities were expressed as units per milligram of protein (U/mg protein), with protein concentrations determined using detection kits from the same supplier. Malondialdehyde (MDA) content, a widely recognized biomarker of membrane lipid peroxidation, was assessed using lipid peroxidation assay kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) following the provided protocol.

For the determination of total flavonoids, 0.5 g dried root tissue was finely pulverized in liquid nitrogen. The resulting powder was extracted with 10 mL 80% methanol in a 50-mL polypropylene tube, followed by vortex mixing for 1 min and ultrasonic treatment at 45°C for 45 min. Subsequently, the extracts were filtered through Whatman No. 1 filter paper to remove insoluble residues. The total flavonoids content was quantified using the aluminum chloride colorimetric method as described by Hussain et al. (2023) (details in Supplementary Text S1).

The rhizosphere microbiome was characterized through 16S rRNA gene sequencing. Genomic DNA was extracted from samples using the CTAB/SDS method (Andriyanto et al., 2022). The V3–V4 hypervariable region of the 16S rRNA gene, which offers superior taxonomic resolution across bacterial phyla while maintaining high sequencing accuracy, was selected for analysis (Klindworth et al., 2013). The region was amplified via PCR with a thermocycler under conditions specified by Li H. et al. (2020) and Li X. et al. (2020). PCR products were purified with the GeneJET™ Gel Extraction kit (Thermo Fisher Scientific, United States) to remove non-target DNA fragments. Sequencing libraries were prepared using the TruSeq^® DNA PCR-Free Sample Preparation kit (Thermo Fisher Scientific) in strict accordance with the manufacturer’s protocols. Rigorous quality control included: removal of PCR duplicates through unique molecular identifier filtering, excision of non-target fragments via GeneJET™ Gel Extraction (Thermo Fisher), and elimination of chimeric sequences during bioinformatic processing using DADA2 within QIIME2. Purified amplicons underwent library preparation with TruSeq^® DNA PCR-Free kits and were sequenced on Illumina HiSeq X Ten/NovaSeq 6,000 platforms (Thermo Fisher Scientific), yielding single-end reads of 400 bp or 600 bp. Full bioinformatic workflows including ASV clustering, alpha/beta diversity calculations, and PICRUSt2 functional predictions are detailed in Supplementary Text S2.

IBM SPSS (v26.0.0, IBM Corp., United States) was employed for statistical analysis, while data visualization was conducted with Origin 9.1 (Origin Lab, United States). To assess the effects of inoculation treatments on soil properties, a one-way analysis of variance (ANOVA) coupled with Fisher’s least significant difference (LSD) post hoc test was performed at a 95% confidence interval (α = 0.05). The significance of differences between inoculated and non-inoculated groups was categorized as follows: p < 0.05 (*, moderate), p < 0.01 (**, high), and p < 0.001 (***, extreme). In graphical representations, distinct lowercase letters above bars indicate statistically significant differences (p < 0.05) derived from post hoc pairwise comparisons. Microbial co-occurrence networks were constructed using Spearman correlations (|r| > 0.6, p < 0.05) and visualized in Gephi. Redundancy analysis (RDA) and Mantel tests linked soil parameters to microbial composition. Structural Equation Model (SEM) analysis was conducted to explore the effects of soil chemical properties (pH, SOM, TP, and TN), enzyme activity and microbial community on plant Sb extraction efficiency using the “plspm” package in R. Detailed information is available in the Supplementary Text S3. All experiments were conducted with four biological replicates, and results are expressed as mean values ± standard error (SE).

Inoculation with Cupriavidus sp. S-8-2 significantly enhanced pepper growth across all treatments (Table 1). Root length and fresh weight, leaf fresh weight, and fresh fruit biomass all exhibited significant increases following inoculation, under both Sb-free and Sb-stressed environments. Specifically, root length increased by 6.86% (p < 0.05) in the absence of Sb stress, by 14.35% (p < 0.001) at 500 mg/kg Sb, and by 24.88% (p < 0.001) at 1000 mg/kg Sb. Root fresh weight increased by 29.10% (p < 0.001) without Sb stress, by 36.97% (p < 0.01) at 500 mg/kg Sb, and by 19.38% (p < 0.01) at 1000 mg/kg Sb. Leaf fresh weight increased by 33.33% (p < 0.01) under both Sb concentrations. Furthermore, inoculation significantly enhanced fresh fruit biomass by 18.48% (p < 0.05) at the highest Sb level, suggesting that Cupriavidus sp. S-8-2 contributes to improved reproductive development under Sb stress.

Under Sb stress, the accumulation of Sb in roots, stems, and leaves was observed in both inoculated and non-inoculated pepper plants (Figure 1). In non-inoculated plants, the Sb content followed the order: roots > leaves > fruits, with the highest total Sb concentration recorded under Sb-only treatment conditions. Under 500 mg/kg Sb stress, no significant difference in Sb content was observed between inoculated and non-inoculated plants in leaf tissues. However, inoculation with Cupriavidus sp. S-8-2 significantly reduced Sb accumulation in fruit tissues (p < 0.01). Notably, under 1,000 mg/kg Sb stress, compared to non-inoculated treatment, the Sb content decreased by 61.31% in roots (p < 0.01), 55.02% in leaves (p < 0.001), and 54.75% in fruits (p < 0.001).

Inoculation with Cupriavidus sp. S-8-2 significantly increased chlorophyll a content in pepper leaves by 22.58% (p < 0.01) under Sb-free conditions and 12.25% (p < 0.05) under 1,000 mg/kg Sb stress (Figure 2a1). Chlorophyll b content also showed marked improvements across all Sb treatments, with increases of 25.42% (p < 0.001) at 0 mg/kg Sb, 11.11% (p < 0.05) at 500 mg/kg Sb, and 20.00% (p < 0.001) at 1000 mg/kg Sb compared to non-inoculated treatments (Figure 2a2). Carotenoid levels remained stable at 0.133 ~ 0.147 mg/g. FW across treatments except under 500 mg/kg Sb stress, where inoculated plants exhibited a 10.52% increase relative to non-inoculated treatment (p < 0.05) (Figure 2a3).

Cupriavidus sp. S-8-2 inoculation significantly altered root antioxidant enzyme activities relative to non-inoculated controls (Figure 2B). Specifically, SOD activity showed a significant decrease of 45.86% in the inoculated controls (p < 0.01) and a reduction of 21.50% under low Sb stress conditions (p < 0.001). In contrast, the decrease in SOD activity was not statistically significant under high Sb stress conditions (12.11%) (Figure 2b1). POD activity was reduced by 21.94% (p < 0.001), 38.48% (p < 0.001), and 16.67% (p < 0.001) under 0, 500, and 1,000 mg/kg Sb treatments, respectively (Figure 2b2). Similarly, CAT activity decreased by 1.45-fold (p < 0.01), 2.29-fold (p < 0.001), and 0.91-fold (p < 0.01) in corresponding treatments (Figure 2b3). MDA content, indicating lipid peroxidation, was consistently lower in inoculated roots, decreasing by 26.69% (p < 0.05) at 500 mg/kg and 27.52% (p < 0.05) at 1000 mg/kg Sb (Figure 2b4). Total flavonoids, contributing to stress tolerance, was elevated its contents by 9.24% (p < 0.01) and 8.77% (p < 0.05) at 500 mg and 1,000 mg Sb/kg, respectively (Figure 2b5). These results suggest that inoculation with Cupriavidus sp. S-8-2 mitigates oxidative damage through coordinated modulation of antioxidant defenses in Sb-stressed pepper roots.

The inoculation treatment significantly improved rhizosphere physicochemical properties under Sb stress (Table 2). SOM increased by 6.30, 4.50, and 8.40 g/kg under Sb concentrations of 0, 500, and 1,000 mg/kg, respectively (p < 0.05). TP levels increased by 26.42, 35.42, and 36.54% under the same Sb levels (p < 0.05). Acid phosphatase activity markedly decreased under Sb stress in the absence of inoculation, but increased by 48.33–81.35% when inoculated (p < 0.05). Urease activity was significantly enhanced by inoculation, with notably higher levels observed across all tested Sb concentrations, showing a 2.03-fold increase even under highest Sb concentration (1,000 mg/kg). CAT activity exhibited the most pronounced improvement following inoculation under non-Sb stress conditions. Although sucrase activity showed relatively weaker response to inoculation, increases were still evident under Sb stress compared to the uninoculated control treatments.

The observed growth promotion and biochemical alterations underscored the role of Cupriavidus sp. S-8-2 in rhizosphere remodeling under Sb stress. To elucidate the mechanisms underlying these phenotypic enhancements, we performed Illumina sequencing of the rhizosphere microbiota. After stringent quality control, 1,158,364 high-quality sequences were clustered into 10,280–10,524 OTUs at a 97% similarity threshold (Supplementary Table S1), revealing 878 conserved OTUs across all treatments and 1,493–2,063 group-specific OTUs (Supplementary Figure S2A). Under 1,000 mg/kg Sb stress, inoculated treatments showed significantly higher OTU richness from 1771 to 1914 compared to the non-inoculated treatment (Supplementary Figure S2A). Alpha diversity exhibited distinct response to different treatments. Inoculated treatments significantly influenced all alpha-diversity indices (Shannon, Chao1, Simpson, Good’s coverage, Pielou’s evenness; p < 0.05), confirming microbial communities restructuring (Supplementary Figure S2B). The Chao1 index increased by 19.87% in the inoculated control under Sb-free conditions. Under low Sb stress (500 mg/kg), the Chao1 decreased by 6.16% following inoculation. Under high Sb stress (1,000 mg/kg), inoculation resulted in a 10.2% reduction in the Chao1 index. Notably, inoculation consistently reduced Shannon indices across all Sb stress levels relative to uninoculated counterparts.

Beta diversity analysis via PCoA demonstrated clear stratification among treatments under 1,000 mg/kg Sb stress (Supplementary Figure S3A), revealing three distinct groupings: (1) Sb-free groups (0 mg/kg with or without Cupriavidus sp. S-8-2) formed a cohesive cluster; (2) moderate Sb stress (500 mg/kg, with or without Cupriavidus sp. S-8-2) exhibited transitional grouping; and (3) high Sb treatments (1,000 mg/kg, with or without Cupriavidus sp. S-8-2) showed unique segregation between each other. These results suggest threshold-dependent restructuring of microbial communities under varying Sb stress levels. The outcomes of ANOSIM also revealed inoculated communities clustered distinctly from non-inoculated counterparts (R = 1, p = 0.001, Supplementary Figure S3B).

The shifts in microbial community composition were visualized using heatmap analysis of the top 10 phyla and 30 genera across Sb concentration gradients (Figures 3A,B). Seven dominant phyla with relative abundances exceeding 1% were consistently identified (Figure 3A; Supplementary Table S2). The abundance of Proteobacteria generally increased with elevated Sb concentrations, peaking at 52.10% under 1,000 mg/kg Sb stress in inoculated treatments. In contrast, Acidobacteria exhibited a significant decline upon Sb exposure, with further suppression observed following inoculation compared to non-inoculated controls. Chloroflexi displayed Sb-dependent depletion, showing a 41.2% greater reduction in inoculated compared relative to non-inoculated ones across all Sb levels. Low-abundance phyla (<0.1%) displayed distinct ecological niches responses to Sb levels and inoculation (Supplementary Table S2). For instance, Chlamydia was detected exclusively in non-inoculated treatments and was undetectable in inoculated soils exposed to Sb concentrations exceeding 500 mg/kg. Elusimicrobia demonstrated Sb-sensitive colonization, being detected exclusively in Sb-free soils and inoculated soils treated with 500 mg/kg Sb. Notably, Cyanophyta showed an increased abundance with rising Sb concentration and exhibited a 1.9-fold higher abundance in inoculated treatments compared to non-inoculated ones under 1,000 mg/kg Sb stress. These findings underscore the distinct responses of microbial taxa to Sb exposure and emphasize the impact of Cupriavidus sp. S-8-2 inoculation on the structure of microbial community in the pepper rhizosphere.

The heatmap analysis of the Top 30 genera within the rhizosphere microbial community also revealed compositional differences between inoculated and non-inoculated treatments (Figure 3B; Supplementary Table S3). Dominant genera, defined as those with relative abundances exceeding 1%, included Rhodanobacter, JG30-KF-AS9, Pseudolabrys, Chujaibacter, Nocardioides, and Jatrophihabitans. Inoculated treatments exhibited reduced relative abundances of Rhodanobacter and JG30-KF-AS9 compared to non-inoculated controls. Conversely, the relative abundances of Pseudolabrys, Jatrophihabitans, and Saccharimonadales were elevated in the inoculated treatments. The relative abundance of Lysobacter was significantly greater in inoculated treatments under low Sb stress conditions (0 and 500 mg/kg Sb; p < 0.05). Luteimonas and the Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium group also showed increased abundance under 500 mg/kg Sb stress in inoculated treatments (p < 0.05). These genera continued to exhibit elevated abundances under 1,000 mg/kg Sb stress in inoculated treatments (p < 0.05). The abundance of Nocardioides increased significantly in inoculated treatments under 500 mg/kg Sb stress (p < 0.05). In contrast, Conexibacter decreased at 0 and 500 mg/kg Sb but increased at 1000 mg/kg in inoculated treatments (p < 0.05). Flavobacterium was detected only under Sb stress and showed higher abundance in inoculated treatments. Some rare genera with abundances below 0.1%, such as Dyella, Gaiella, and Amycolatopsis, exhibited significantly increased abundance in inoculated treatments under Sb stress (p < 0.05, Figure 3D).

The LEfSe analysis (LDA score > 4.0) identified distinct microbial biomarkers across treatments and Sb stress conditions (Figure 3C). Under Sb-free conditions, non-inoculated treatments were characterized by phylum-level biomarkers Actinobacteria and Chloroflexi. Conversely, inoculated treatments under the same conditions showed specific enrichment of Gemmatimonadetes at the phylum level. At a moderate Sb stress level (500 mg/kg), the genus JG30-KF-AS9 was identified as a biomarker in non-inoculated treatments, whereas inoculated treatments were distinguished by the family Rhizobiaceae. Under elevated Sb stress (1,000 mg/kg Sb), two family-level biomarkers, Enterobacteriaceae and Sphingomonadaceae were observed in inoculated treatments. Notably, no significant biomarkers were detected in non-inoculated treatments exposed to 1,000 mg/kg Sb stress.

The functional prediction of rhizosphere microbial communities was performed using PICRUSt2 across KEGG metabolic hierarchies (Supplementary Figure S4). Inoculation with Cupriavidus sp. S-8-2 significantly upregulated multiple metabolic pathways associated with detoxification and stress resistance. Notably, under conditions of 1,000 mg/kg Sb, a marked enhancement was observed in pathways related to ABC transporter, including cofactor, prosthetic group, electron carrier degradation, and fatty acid and lipid biosynthesis. Interestingly, the activity of the tricarboxylic acid (TCA) cycle remained consistently higher in inoculated treatments compared to non-inoculated treatments across all tested Sb concentrations. Furthermore, analysis revealed a substantial upregulation of antibiotic resistance pathways in the inoculated treatments, particularly under 1,000 mg/kg Sb stress conditions. Concurrently, significant activation of secondary metabolite biosynthesis pathways was observed in the inoculated treatments.

The interspecific interactions were analyzed using SparCC-based co-occurrence networks (|r| > 0.60, p < 0.05). The network topology exhibited distinct treatment-specific variations (Figure 4; Table 3). The non-inoculated treatments under severe Sb stress (1,000 mg/kg) exhibited the least connected network structure, whereas the inoculated treatments under the same stress level demonstrated significantly higher connectivity. Notably, the average node degree was markedly increased in inoculated treatments under 1,000 mg/kg Sb stress compared to their non-inoculated counterparts. Additionally, a higher proportion of negative correlations was detected in inoculated treatments under 0 and 500 mg/kg Sb stress conditions. The analysis identified specific members of Actinobacteria, Proteobacteria, Acidobacteria, and Planctomycetes as occupying key topological positions, suggesting their roles as keystone species (Figure 4). Proteobacteria consistently displayed the highest level of network connectivity across all treatments, engaging in both positive and negative correlations. Under inoculation treatments at 1000 mg/kg Sb, significant antagonistic interactions were observed between Proteobacteria and Actinobacteria. Furthermore, inoculation treatments exhibited significantly enhanced intra-phylum connectivity within Proteobacteria compared to non-inoculation treatments. Inoculation notably enhanced intra-phylum connectivity within Proteobacteria, with this effect becoming more pronounced as Sb concentrations increased. Actinobacteria exhibited strong correlations with Planctomycetes in inoculated samples, forming a dense subnetwork that was most evident under high Sb stress. In the absence of Sb stress, Bacillus (phylum Firmicutes) acted as a central hub genus, showing correlations with PGPR-associated genera such as Rhodanobacter and Luteibacter; however, this specific regulatory network structure collapsed under moderate Sb stress (500 mg/kg). Under severe Sb stress condition (1,000 mg/kg), genera such as Sericytochromatia (phylum Cyanobacteria) and Acidipila (phylum Acidobacteria) partially compensated for the topological collapse by assuming hub-like functions. In contrast, non-inoculation treatments lacked such compensatory hubs, resulting in a marked reduction in overall network connectivity compared to their inoculated counterparts. Among the inoculated treatments under Sb stress, Rhodanobacter and Luteimonas emerged as core genera within two separate subnetworks, likely due to their divergent metabolic patterns; notably, Luteimonas displayed correlation patterns that were largely opposite to those of Rhodanobacter.

Redundancy analysis (RDA) was employed to characterize the environmental drivers influencing the phyla-level structure of the microbial community, explaining 97.77% of the total variance (Figure 5A). Among the analyzed factors, rhizosphere enzyme sucrase activity was identified as the most significant chemical factor affecting the distribution of rhizosphere bacterial communities across all treatments (R² = 0.20, Figure 5A). Soi Sb concentration exhibited a significantly negative correlation with microbial community structure (p < 0.01). Urease activity, TN, and SOM were also recognized as key drivers shaping community structure. Chloroflexi, Acidobacteria, and Gemmatimonadetes correlated positively with urease and SOM, while Proteobacteria and Planctomycetes associated with sucrase, TN, and TP (p < 0.05). In contrast, sucrase activity, TN, and TP showed positively associated with the abundances of Proteobacteria and Planctomycetes (p < 0.05). At 1000 mg/kg Sb, the correlations between microbial communities and acid phosphatase, sucrase activity, SOM, and TN were significantly enhanced in inoculated treatments compared to non-inoculated treatments.

Pearson correlation analysis revealed strong associations among rhizosphere enzyme activities, physicochemical properties, and microbial abundance at both the phylum and genus levels across different treatments (Figures 5B,C). Total Sb exhibited significant negative effects on most bacterial phyla, especially Chloroflexi, Firmicutes, Nitrospirae, and Elusimicrobia, and multiple genera such as SC-I-84, Nitrolancea, Subgroup_6, Acidibacter, Aeromicrobium, and Marmoricola. In inoculated treatments, Chloroflexi (phylum) and JG30-KF-AS9 (genus) exhibited significant negative correlations with total Sb (p < 0.01), while Firmicutes showed more variable trends. Total Sb was positively correlated with several phyla, including Proteobacteria, Bacteroidetes, and Euryarchaeota, as well as specific genera such as Pseudolabrys, Sphingomonas, Massilia, Luteimonas, and Burkholderia-Caballeronia-Paraburkholderia. Soil nutrients (TP and TN) exerted limited influence on microbial phyla regardless of treatment conditions. However, SOM and enzyme activities (urease, sucrase, and acid phosphatase) were positively influence specific phyla (Chloroflexi, Nitrospirae, and Fibrobacteres) and genera (SC-I-84, JG30-KF-AS9, Aeromicrobium, and Marmoricola), but negatively impacted other genera such as Massilia.

Mantel correlation analysis was performed to investigate the potential associations between various environmental factors and rhizosphere bacterial communities across different treatments (Figure 5D). Root Sb content exhibited a strong negative correlation with bacterial community structure following inoculation with Cupriavidus sp. S-8-2 (|r| > 0.6, p < 0.01), as reflected by in alpha-diversity metrics, including Observed OTUs, Chao1, and Shannon indices. In contrast, under non-inoculated conditions, inverse relationships were observed, suggesting that root Sb accumulation may disrupt the integrity and stability of the bacterial community. Moreover, SOM Additionally, SOM showed significant positive correlations with the Chao1 and Shannon indices in inoculated systems. Enzyme activity and soil pH were positively correlated with bacterial diversity under inoculated conditions (|r| > 0.6, p < 0.05). However, no significant correlations were detected between soil nutrients, including TP and TN, and microbial community indices across all treatments. Carotenoids content and chlorophyll levels were correlated with bacterial communities exclusively under non-inoculated conditions (|r| > 0.6, p < 0.05). while total flavonoids showed a negative correlation with tissue Sb content only under inoculated conditions. Furthermore, chlorophyll a was positively correlated with SOM, urease activity, and acid phosphatase (|r| > 0.6, p < 0.05) in inoculated treatments, while it exhibited a negative correlation with tissue Sb content under the same treatments.

The SEM results revealed that inoculation with Cupriavidus sp. S-8-2 exerted both direct and indirect influence on the structure and diversity of the microbial community, which in turn affected microbial functions, plant growth, and Sb uptake through alterations in soil nutrient profiles (Figure 6). Among these factors, the correlation coefficient between Cupriavidus sp. S-8-2 inoculation and plant traits, including fresh weight, root height, and chlorophyll content was determined to be 0.484. In contrast, a significant negative correlation was observed between Cupriavidus sp. S-8-2 inoculation and Sb uptake, with a correlation coefficient of −0.536. Notably, the correlation coefficient between Cupriavidus sp. S-8-2 inoculation and bacteria abundance and diversity was 0.171, which was not statistically significant. This lack of significance may be attributed to the fact that inoculation with Cupriavidus sp. S-8-2 primarily alters microbial community structure and functions rather than directly influencing bacterial abundance or diversity. These findings collectively highlight the complex interactions between PGPR inoculation, plant performance, and nutrient dynamics in the soil environment under Sb stress.

Inoculation with Cupriavidus sp. S-8-2 significantly alleviated the inhibitory effects of Sb on the growth of pepper plants, highlighting its potential application as a bioremediation agent in soils contaminated with low-to-medium level of Sb (Sb ≤ 1,000 mg/kg). In this study, inoculation with strain S-8-2 markedly enhanced root length, biomass, and chlorophyll content in pepper plants exposed to Sb stress. These enhancements can be attributed to the plant growth-promoting (PGP) properties of strain S-8-2, including its capability to regulate phytohormone levels, such as IAA, as our previously reported (Zheng et al., 2023). Such hormonal regulation may facilitate root elongation and enhance overall plant development. Notably, inoculated plants exhibited higher chlorophyll content compared to non-inoculated controls, particularly under exposure to 1,000 mg/kg Sb (Figure 2). Research has shown that HMs can accumulate in chlorophyll molecules, triggering their degradation, a process is often associated with ROS-mediated interference (Karimi et al., 2025). Conversely, several studies have shown that inoculation with Sb-tolerant PGPR strains can increase chlorophyll content in plants grown under HM stress (Aziz et al., 2025). Additionally, carotenoids are known to function as a photo-protective pigment. A reduction in carotenoids content has been previously observed in various plants under HM stress due to ROS-mediated inhibition of carotenogenesis (Espinola et al., 2025). However, PGPR inoculation enhanced carotenoid levels in Sesbania sesban under stress conditions (Zainab et al., 2021), a result consistent with our observations. The elevated chlorophyll and carotenoid content can be attributed to the ability of PGPR to upregulate carotenogenic genes, maintain chloroplast structural integrity, enhance the efficiency of photosynthetic photon capture, and suppress ROS-dependent pigment degradation (Bhardwaj et al., 2024; Li X. et al., 2022; Li Z. et al., 2022).

Exposure to HM stress increases ROS accumulation in plants, which activates the antioxidant defense system and enhances antioxidant enzyme activity (Mansoor et al., 2023). Among these antioxidant enzymes, SOD, POD, and CAT play crucial roles in ROS scavenging (Song et al., 2025). Studies have demonstrated that synergistic action of SOD and CAT effectively reduces ROS levels, thereby creating favorable conditions for photosynthesis (Moustaka et al., 2025; Rahim et al., 2025). In the present study, inoculation with Cupriavidus sp. S-8-2 significantly attenuated the activities of POD, CAT, POD, and SOD in peppers across all tested Sb concentration (Figure 2). These attenuations may be attributed to Cupriavidus sp. S-8-2 substantially reducing Sb accumulation in root tissues, thereby mitigating Sb-induced toxicity to pepper roots. This reduction in Sb uptake was associated with decreased ROS levels and modulation of antioxidant enzyme activities. Similar observations were reported by Park et al. (2017), who found that inoculation of soybean plants with PGPR Bacillus aryabhattai resulted in decreased CAT and POD activities. Likewise, Cd-tolerant PGPR strains (Burkholderia gladioli and Pseudomonas aeruginosa) have been shown to reduce Cd uptake and subsequently lower antioxidant enzyme levels in tomato plants (Khanna et al., 2019a). Additionally, MDA, a key biomarker of lipid peroxidation and oxidative damage, was significantly reduced in pepper plants inoculated with Cupriavidus sp. S-8-2 compared to non-inoculated controls. Consistent with this, treatment with Bacillus subtilis strain IU31 resulted in decreased MDA levels in rice plants exposed to As stress (Ullah et al., 2024). Furthermore, previous studies have demonstrated that flavonoids can suppress ROS biosynthesis and reduce MDA content, thereby alleviating plant stress under adverse conditions. Supporting this, Wu et al. (2025) reported that inoculation with Pseudomonas koreensis in Cd-stressed rice significantly increased total flavonoids content, which was accompanied by enhanced antioxidant capacity, chlorophyll content, and biomass accumulation. Taken together, these observations are consistent with the present findings, indicating that inoculation with Sb-tolerant PGPR effectively alleviates Sb-induced stress, reduces oxidative damage, and enhances the survival and growth of pepper plants under high Sb conditions.

Plant roots critically regulate the dynamics at the soil-root interface by modifying rhizosphere physicochemical properties. In the present study, inoculation with Cupriavidus sp. S-8-2 under high Sb stress (1,000 mg/kg) significantly decreased rhizosphere pH from 5.13 to 4.88 (Table 2). This acidification likely resulted from enhanced secretion of low-molecular-weight organic acids (e.g., malate, citrate, and oxalate), which facilitate Sb immobilization through precipitation mechanisms under elevated Sb stress (Zheng et al., 2023). Furthermore, HMs can form stable complexes with SOM, including humic and fulvic acids, as evidenced by the observed relationship between SOM and pH in the inoculated treatments under 1,000 mg/kg Sb stress. Conversely, under moderate Sb stress (500 mg/kg), inoculated treatments exhibited increased rhizosphere pH compared to non-inoculated controls. This observation suggests that organic acid secretion may be limited under lower stress conditions, thereby exerting minimal influence on the pH of the soil solution. Instead, alternative mechanisms that result in a net increase in rhizosphere pH may predominate under such conditions (Kangi, 2024). Further investigation into the underlying processes governing these differential pH responses could enhance our understanding of the adaptive strategies employed by plant–microbe systems under varying levels of Sb stress.

Concurrently, HM contamination triggered alterations in soil chemical processes while potentially suppressing biological activities linked to nutrient cycling. This shift appears to align with changes in specific soil physicochemical properties, such as decreased soil enzyme activities. These observations are consistent with the findings obtained under elevated Sb exposure. However, inoculation with Cupriavidus sp. S-8-2 significantly increased soil TN, TP, and SOM contents compared to the uninoculated controls (Table 2). These results suggest that PGPR inoculation enhances soil quality through augmenting nutrient pools. Our observations are in agreement with Ju et al. (2019), who reported that PGPR inoculation significantly improved soil properties and fertility in copper-contaminated system by elevating tissue nutrient (N, P, and K) content and promoting plant growth. Soil enzymes may indirectly influence carbon dynamics through nutrient leaching or microbial activity modulation, while also directly participating in the conversion of organic carbon, thereby contributing to improved soil fertility. SOM, as the primary reservoir of organic carbon in soil, serves as the fundamental substrate for these enzymatic activities. In this study, the activities of urease, sucrase, CAT, and acid phosphatase were observed to decrease progressively with increasing Sb concentration in the soil. In contrast, inoculation with Cupriavidus sp. S-8-2 significantly enhanced enzyme activities compared to non-inoculation treatments (Table 2), which is consistent with previous findings demonstrating the capacity of PGPR to stimulate soil enzyme activities (Li X. et al., 2020). This enhancement may be attributable to PGPR-mediated modifications in rhizosphere microbial activity, improvements in soil physicochemical properties (e.g., increased TN content), and the mitigation of soil toxicity (Ju et al., 2019; Zhou et al., 2022). Soil urease, a key enzyme in the nitrogen cycle, showed significant positive correlations with TN and SOM content (Figure 5), supporting its role in the conversion of organic nitrogen into plant-available ammonium-N (Cheng et al., 2025). This aligns with documented positive association between urease activity and soil nutrient content levels (Fu et al., 2025). Similarly, acid phosphatase facilitates phosphorus mineralization, thereby directly influencing soil-available phosphorus and alleviating phosphorus limitation under stressful environments conditions (Zhu et al., 2018), as evidenced by the present results. PGPR have been found to enhance CAT activity, as demonstrated in Gossypium hirsutum L. (Qureshi et al., 2019), as well as sucrase and urease activities in HM-contaminated soils through strains like Bacillus sp. ZC3-2-1 (Liu et al., 2022). CAT enzyme contributes to HM detoxification by catalyzing the decomposition of toxic H₂O₂ into less harmful products and is also involved in soil carbon cycling (Das and Sen, 2024). Therefore, inoculation with Cupriavidus sp. S-8-2 enhances nitrogen and phosphorus cycling through the activation of specific enzymes, leading to improved soil properties and enhanced plant growth under Sb stress. Furthermore, the pronounced negative correlation observed between urease activity and Sb concentration underscores the urease activity could serve as a potential indicator for evaluating Sb toxicity in soil ecosystems. Further investigation into this correlation may elucidate the mechanisms underlying Sb-induced enzymatic inhibition and its broader ecological implications.

Rhizosphere biochemical properties can influence microbial community structure, which in turn regulate plant growth and HM uptake through diverse mechanisms (Solomon et al., 2024). PGPR treatments have been shown to critically shape bacterial diversity under HM exposure conditions (Hu et al., 2021; Muratova et al., 2023). A significant decrease in alpha diversity (Shannon index) was observed following inoculation under 500 mg/kg Sb stress (p < 0.05, Supplementary Figure S2B). These results suggest that inoculation with Cupriavidus sp. S-8-2 promoted a rapid response of rhizosphere microorganisms to Sb stress by competitive exclusion of taxa with lower Sb tolerance, while concurrently enriching HM-tolerant families such as Rhizobiaceae and Sphingomonadaceae. This restructuring of microbial community may contribute to enhanced plant survival under adverse conditions. Supporting this hypothesis, PCoA analysis (Supplementary Figure S3) reveals potential structural changes in the rhizosphere microbial community that could be associated with altered rhizosphere toxicity dynamics. Crucially, inoculation significantly modulated the relative abundance of several bacterial taxa involved in HM resistance and plant growth promotion. These included the phyla Proteobacteria and Firmicutes, as well as genera such as Ramlibacter, Brevundimonas, Pseudomonas, Microvirga, Dyella, and Nitrospira. An increase in the relative abundance of Proteobacteria was observed under elevated Sb stress when inoculation, which is consistent with earlier reports in Cd-stressed wheat (Zhao et al., 2024). This phylum is known for its intrinsic HM tolerance mediated by detoxification enzymes and regulatory proteins that enhance its competitive advantage in contaminated environments (Bai et al., 2022). Moreover, Proteobacteria include genera key genera, such as Rhodanobacter and Luteimonas, which have been linked to oxidative stress mitigation through enzymatic ROS detoxification (Caldeira et al., 2021). These observations are consistent with our findings showing reduced MDA levels in inoculated roots (Figure 2). Despite their recognized role in SOM decomposition and nutrient cycling, the abundance of Proteobacteria showed a significant negative correlation with SOM content (Figure 5B). Many members of Proteobacteria also exhibit PGP traits, including symbiotic nitrogen fixation (Li P. et al., 2024; Shi et al., 2023), which may improve nutrient availability for plants. Other enriched genera, such as Ramlibacter, Pseudomonas, and Dyella have previously been implicated in HM resistance and PGP capabilities (Zhao et al., 2024; Shi et al., 2023; Cai et al., 2023). Similarly, Microvirga has been associated with HM resistance and nitrogen cycling processes (Zhang N. et al., 2025). The inoculation-induced enhancement of nutrient availability under Sb stress may have facilitated the proliferation of copiotrophic microbial groups, such as Proteobacteria and Firmicutes, potentially leading to a decrease in the relative abundance of oligotrophic taxa, including Chloroflexi and Acidobacteria (Figure 3A). This trend was further corroborated by LefSe analysis, which identified Proteobacteria as significant discriminative biomarkers in inoculated treatments subjected to high Sb stress, while Chloroflexi were more prominently associated with non-contaminated controls (Figure 3C). Within the rhizosphere of pepper inoculated with Cupriavidus sp. S-8-2, Rhizobiaceae and Sphingomonadaceae were identified as key biomarkers under 500 mg/kg and 1,000 mg/kg Sb stress, respectively. Rhizobiaceae harbors essential genes involved in Sb(III) oxidation, N₂ fixation, and carbon fixation (Li Y. et al., 2022), whereas Sphingomonadaceae has been reported to exhibit increased abundance in response to Cd exposure in previous studies (Zhou et al., 2019). Collectively, these findings suggest that inoculation with Cupriavidus sp. S-8-2 may help shape the rhizosphere microbiome under Sb contamination by enriching bacterial taxa associated with HM tolerance and nutrient cycling, potentially contributing to enhanced plant growth.

The network analysis of microbial co-occurrence provides a powerful framework for elucidating complex microbial interactions beyond conventional sample-level comparisons, particularly in extreme environments (Guo et al., 2022; Ma et al., 2020; Mercado et al., 2022). In this study, inoculation with Cupriavidus sp. S-8-2 was found to significantly enhance node connectivity and mean degree in rhizosphere bacterial networks, especially under severe Sb stress (1,000 mg/kg). This increase in structural complexity, as reported by Wang J. et al. (2023), may indicate an improved capacity for adaptation to environmental stress (Xing et al., 2024). Interestingly, the microbial network, whether under individual Sb Stress or PGPR inoculation, showed an increased proportion of negative correlations (Table 2). Such shifts in correlation patterns may reflect dynamic adjustments in both symbiotic and competitive relationships among microbial taxa, as previously observed under nutrient-limited stress conditions (Anas et al., 2025). These competitive interactions warrant further investigation under Sb-induced stress conditions.

Key taxa within microbial networks under PGPR treatment were identified as functional groups involved in nutrient cycling (Liu et al., 2024; Chi et al., 2025), potentially contributing to the structuring microbial co-occurrence patterns (Ding et al., 2023). Several keystone taxa, including Proteobacteria, Rhodanobacter, Ramlibacter, and Luteibacter, were identified, all of which have been previously associated with HM removal and detoxification (Peng et al., 2022). Notably, Ramlibacter has been shown to enhance phosphorus solubilization and promote ryegrass growth (Zhao et al., 2024). The emergence of Rhodanobacter as a keystone hub under high Sb stress (Figure 4) is consistent with its previously reported capacity for Sb(III) oxidation in rhizosphere systems (Kataoka et al., 2018). This oxidation mechanism contributes to the alleviation of Sb-induced ROS generation, a process that has been mechanistically elucidated by Zhang Y. et al. (2025). Inoculation with Cupriavidus sp. S-8-2 was observed to modify the rhizosphere co-occurrence network structure (Table 2), thereby enhancing functional stability, as indicated by increased modularity and reduced the number of positive correlations (Coyte et al., 2015).

Although inoculation with Cupriavidus sp. S-8-2 did not significantly alter the overall structure of the rhizosphere microbiota, functional predictions generated using PICRUSt2 revealed notable modifications in specific secondary metabolic pathways (Supplementary Figure S4). In particular, TCA cycle, a central metabolic pathway that supplies energy and anabolic precursors essential for cellular proliferation and survival (MacLean et al., 2023), exhibited mitigated inhibition following inoculation. Given the reported synergistic relationship between glycolysis and TCA cycle under Al exposure (Guan et al., 2024), these findings suggest that Cupriavidus sp. S-8-2 may help alleviate energy supply deficits under Sb stress (Supplementary Figure S4). Inoculated treatments under high Sb stress (1,000 mg/kg) also showed increased activity in pathways associated with aromatic amino acid biosynthesis. Aromatic amino acids play key roles in regulating auxin signaling and serve as precursors for the biosynthesis of antioxidant phenolic compounds (Samsami and Maali-Amiri, 2024) which is consistent with the observed increase in total flavonoid content. Specifically, tyrosine-derived flavonoids have been shown to enhance plant tolerance to abiotic stress through stress-induced accumulation (Sanches Silva et al., 2020). Additionally, inoculated systems exhibited significantly improved aldehyde degradation, linked to IAA production, and enhanced carboxylate metabolism, associated with ACC deaminase synthesis. Aldehyde degradation is biochemically driven by aldehyde dehydrogenase, which converts indole-3-acetaldehyde into IAA—the final step in auxin biosynthesis (Zhang et al., 2023). Carboxylate metabolism produces α-ketobutyrate, a key co-substrate for ACC deaminase synthesis, which helps regulate ethylene-mediated stress responses (Gupta et al., 2022). Both processes have been shown to improve root growth and stress tolerance, as confirmed by our earlier research (Zheng et al., 2023). Pathways related to glutathione metabolism, known for their roles in pollutant sequestration and antioxidant defense mechanisms (Dorion et al., 2021), were also significantly upregulated under Sb stress following inoculation with Cupriavidus sp. S-8-2. The metabolic changes observed after inoculation with Cupriavidus sp. S-8-2 were correlated with reduced Sb accumulation in plants and enhanced growth under stress conditions. This functional reconfiguration of metabolic processes likely enhanced Sb tolerance and supported host plant acclimation. Future studies should monitor the persistence of Cupriavidus sp. S-8-2 in soil environments to validate its ecological role.

PGPR-mediated modulation of HM uptake involves a series of interconnected processes encompassing plant physiology, soil HM bioavailability, and intraplant transport mechanisms (Sultana et al., 2024; Qin et al., 2024). Integrative analyses (RDA, Pearson correlation, Mantel tests, and SEM; Figures 5, 6) revealed that soil physicochemical properties directly influenced pepper biomass and the distribution of Sb in plant tissues, thereby governing Sb accumulation patterns under Sb stress conditions.

In this study, Cupriavidus sp. S-8-2 inoculation directly enhanced plant growth by improving the availability of soil nutrients (TN, TP, SOM). The resulting changes in N and P availability, along with shifts in soil pH, rapidly modulated soil enzyme activities, thereby altering the structure of the soil microbial community. The inoculation treatments enriched HM-tolerant genera, such as Gaiella, Nitrospira, JG30-KF-AS9, Haliangium, which are adapted to oligotrophic conditions and exhibited positive correlations with plant biomass under Sb stress. Genus Gaiella has been previously reported to exhibit notable HM tolerance (Wu et al., 2024) and to promote plant growth in contaminated soils (Li Y. et al., 2022; Li Z. et al., 2022; Wu et al., 2024). While JG30-KF-AS9 is associated with the mineralization of soil organic carbon (Cao et al., 2025). Nocardioides is known to produce phytohormones and perform nitrogen fixation under HM stress (Meena et al., 2020), whereas members of the Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium complex are capable of degrading xenobiotics and simultaneously fixing nitrogen (Talwar and Chatli, 2020). Additionally, rare genus like Pseudarthrobacter encodes genes involved in auxin biosynthesis and atmospheric N₂ fixation (Li X. et al., 2020), further supporting the role of inoculation in recruiting functionally diverse bacterial taxa to improve plant fitness under Sb stress. Significant positive correlations were observed between microbial community composition and soil enzyme activities. Notably, the abundance of Proteobacteria was positively correlated with sucrase activity and SOM content, which aligns with the metabolic versatility of this phylum in organically enriched soils.

PGPR application has been shown to reduce the entry of HMs into food chains, with plant uptake primarily governed by bioavailability, which is modulated by soil pH and SOM content (Li J. et al., 2024; Li P. et al., 2024). Although tissue Sb concentrations increased with elevated Sb exposure levels, inoculation with PGPR significantly reduced root Sb accumulation (p < 0.01), leading to reduced concentrations in leaves and fruits. This reduction may help mitigate the risk of secondary pollution and improve food safety in Sb-contaminated environments. These findings are consistent with previous observations of PGPR-induced decreases in HM bioavailability in ryegrass (Ke et al., 2021). However, some PGPR consortia, such as Pseudomonas aeruginosa have been reported to enhance HM uptake (Shi et al., 2022), highlighting the functional variability among these PGPR strains. The observed reduction in Sb content in pepper plants may be due to the superior phosphate solubilizing ability and the secretion of low-molecular-weight acids by Cupriavidus sp. S-8-2 under Sb stress, which could potentially immobilize Sb through precipitation reactions (Zheng et al., 2023). Additionally, rhizosphere SOM has been shown to facilitate HM binding, thereby reducing Sb bioavailability (Wang et al., 2024), with the efficiency of PGPR being closely related to SOM content (Bai et al., 2024). Moreover, elevated TP levels can compete for soil binding sites, thus decreasing Sb mobilization (Yan et al., 2017; Lisac et al., 2019). These mechanisms were concurrently observed under high Sb stress (1,000 mg/kg), as inoculation not only increased TP and SOM levels but also reduced root Sb accumulation. Moreover, enhanced root growth (Table 1) contributed to enhanced tolerance to Sb and decreased its translocation to above-ground tissues.

Furthermore, inoculation with Cupriavidus sp. S-8-2 was found to enhance the abundance of certain microbial taxa that exhibited negative correlations with Sb concentrations (Figure 5). For instance, the genus Gemmatimonas showed an inverse relationship with Sb bioavailability, suggesting its involvement in Sb immobilization (Li et al., 2021). Conversely, Flavobacterium showed a strong positive correlation with Sb accumulation in plant roots, which may be attributed to its HM resistance mechanisms involving oxidation and methylation processes (Sun et al., 2023). Additionally, Dyella has been reported to participate in the cycling of Al and iron ions under oligotrophic conditions (Chen et al., 2023). Notably, inoculation led to a decrease in the relative abundance of Patescibacteria, a phylum associated with HM activation and migration (Tian et al., 2022). These observations are consistent with previous studies indicating that Proteobacteria, Actinobacteria, Bacteroidetes, Firmicutes, and Gemmatimonadetes are frequently associated with variations in HM accumulation patterns in plants (Wang X. et al., 2023). Halim et al. (2025) demonstrated that PGPR employ multiple mechanisms, including biosorption, bioaccumulation, plant growth promotion, organic acids secretion, siderophore enhancement, and extracellular polymer synthesis to immobilize Cd at the soil–plant interface, thereby restricting Cd translocation to above-ground plant tissues. Collectively, Cupriavidus sp. S-8-2 contributes to the enrichment of functionally diverse microbial communities in the rhizosphere of pepper plant, including phosphorus-solubilizing bacteria, HM-immobilizing microorganisms, plant growth-promoting rhizobacteria, and nitrogen-cycling bacteria. These functionally distinct microbial groups may work in concert to decrease Sb bioavailability through immobilization processes, while simultaneously supporting plant growth under Sb stress conditions.

Beyond providing preliminary mechanistic insights, this study highlights the practical applications of sustainable remediation strategies for Sb-contaminated environments. Cupriavidus sp. S-8-2 demonstrates potential as a bioaugmentation agent or biofertilizer in agricultural areas affected by Sb contamination, supported by a 54.75% reduction in Sb accumulation in fruits and 18.48% increase in fruits biomass, even under a high Sb stress level of 1,000 mg/kg in pot experiments. However, further targeted investigations are required to validate the underlying mechanisms. The transition from laboratory to field application necessitates confirmation of the strain’s persistence in non-sterile soil environments and field validation of its bioaugmentation efficacy, both of which are essential for the development of precision bioformulations tailored for Sb-affected agricultural systems.