Nitrogen cycle induced by plant growth-promoting rhizobacteria drives “microbial partners” to enhance cadmium phytoremediation

Abstract

Background

Using plant growth-promoting rhizobacteria (PGPR) combined with hyperaccumulator is an ecologically viable way to remediate cadmium (Cd) pollution in agricultural soil. Despite recent advances in elucidating PGPR-enhanced phytoremediation, the response of plant-associated microbiota to PGPR remains unclear.

Results

Here, we found that the effective colonization of PGPR reshaped the rhizosphere nutrient microenvironment, especially driving the nitrogen cycle, primarily mediated by soil nitrate reductase (S-NR). Elevated S-NR activity mobilized amino acid metabolism and synthesis pathways in the rhizosphere, subsequently driving a shift in life history strategies of the rhizosphere microbiota, and enriching specific rare taxa. The reconstructed synthetic community (SynCom3) confirmed that the inclusion of two crucial collaborators (Lysobacter and Microbacterium) could efficiently foster the colonization of PGPR and aid PGPR in executing phytoremediation enhancement. Finally, the multi-omics analysis highlighted the critical roles of phenylpropanoid biosynthesis and tryptophan metabolism pathways in inducing SynCom3 reorganization and PGPR-enhanced phytoremediation.

Conclusions

Our results underscore the significance of the rhizosphere microenvironment modification by PGPR for its colonization and efficacy, and highlight the collaborative role of rare microbiota in the context of PGPR-enhanced phytoremediation.

Video Abstract

Introduction

Soil contamination by heavy metal is a pervasive and severe environmental issue that threatens agricultural productivity and food safety [1, 2]. Cadmium (Cd) is the most prevalent heavy metal in agricultural soil with high bioavailability, leading to concerns about the presence of Cd-contaminated rice [3]. Despite stricter environmental regulations and numerous efforts to remediate Cd-contaminated agricultural soil, Cd pollution continues to accumulate, rendering millions of hectares of agricultural land unproductive [3, 4]. Consequently, there is a growing interest in developing sustainable and efficient remediation strategies.

Phytoremediation, an ecological remediation approach, employs plants to separate pollutants from the environment, demonstrating significant potential in addressing Cd pollution [5]. Among all plants, hyperaccumulators are frequently used as phytoremediation vectors due to their exceptional ability to concentrate and tolerate specific heavy metals [6, 7]. Solanum nigrum L., a type of hyperaccumulator, is well-documented for its ability to hyperaccumulate Cd [8, 9]. However, the efficacy of phytoremediation remains constrained by the sluggish growth rate and diminutive biomass of hyperaccumulators.

Microbiota associated with their hosts provides crucial support for plant growth and stress resistance [10]. Plant growth-promoting rhizobacteria (PGPR) perform a series of crucial biological functions, including facilitating nutrient acquisition and alleviating stress [11]. Certain effective PGPR, e.g., Bacillus and Pseudomonas, were identified and successfully utilized in the enhancement of Cd phytoremediation [12,13,14]. Nevertheless, the substantial benefits and precise mechanisms of plant-PGPR interactions in Cd-contaminated soil remain overlooked, potentially offering new opportunities for sustained improvements in phytoremediation efficiency [15].

Rhizosphere nutrient availability is clearly a key factor in mediating plant-microorganism interactions, especially soil nitrogen availability. Nitrogen can restrict the plant growth and microbial communities across diverse ecosystems; organic nitrogen, nitrate nitrogen, and ammonium nitrogen are the three forms of nitrogen utilized by organisms, and they play a crucial role in plant-microorganism interactions through nitrogen allocation among species [16, 17]. Some studies suggest that nitrogen forms and availability may also affect plant uptake of Cd [18, 19]. Soil microorganisms can metabolize nitrogen in various forms [20]. For example, microorganisms with the ability to assimilate nitrate can trap nitrate into organic nitrogen. Organic nitrogen will be decomposed into ammonia under microbial mineralization and rapidly converted to ammonium, a form of nitrogen that may be more conducive to plant growth and Cd uptake [18, 21].

Despite these characteristics, the inefficient colonization of microbial inoculants in the rhizosphere limits their functional performance. This is especially true for single exogenous microbial inoculants, which often struggle to proliferate due to spatial niche competition with indigenous microbiome [22, 23]. In addition, the significant heterogeneity of microbial inoculants performance also indicates that the functionality and longevity of inoculants rely on the interaction with the environment and other microorganisms within the microbiome [10]. Consequently, the compositional alterations in the microbiome triggered by the proliferation of inoculants can offer invaluable insights into the interactions between beneficial bacterial inoculants and the plant-associated microbiome [24].

Synthetic communities (SynComs) of varying complexity were constructed using different strategies to enhance inoculant colonization and demonstrate more robust and stable ecological functions compared to a single microorganism [25, 26]. One key to this process is mining crucial microorganisms. Successfully recruiting and maintaining specific microbial members in sufficient abundance determines the outcome of plant-microbiota interactions [27]. Another important aspect is the deconstruction of community member assembly patterns, as optimizing the functional redundancy and stability of the microbiota further benefits rhizosphere nutrient acquisition and plant health [25, 28].

In this study, Bacillus megaterium NCT- 2 was employed as a representative of PGPR to elucidate the plant-PGPR interactions in the context of PGPR-enhanced Cd phytoremediation. B. megaterium NCT- 2 has been demonstrated to regulate environmental nitrogen cycling (including nitrate assimilation and organic nitrogen mineralization) and superior plant growth promotion [21, 29]. We hypothesize that modifying the rhizosphere environment with PGPR will enrich beneficial partners, thereby enhancing the effectiveness of PGPR in Cd phytoremediation. Hence, we first characterized rhizosphere environmental parameters under PGPR inoculation and mined beneficial units in the plant-related microbiome mediated by PGPR through high-throughput sequencing. Further adopt a top-down approach to construct and simplify the functional microbiota with NCT- 2 strain as the core. Finally, the potential molecular mechanisms underlying microbiota interactions were elucidated through the integration of multi-omics analyses.

Methods

Materials preparation

The PGPR B. megaterium NCT- 2 was screened by our laboratory and stored in China General Microbiological Culture Collection Center (CGMCC, No. 4698). The strain grew in Luria–Bertani liquid medium and was harvested after cultured in a shaker at 32 ℃ and 200 rpm for 24 h. The pellets were washed three times with sterile deionized water, and then suspended to approximately 10⁹ CFU/mL in sterile deionized water as stock inoculum.

The experimental soil was taken from Cd-contaminated agricultural soil (Zhuzhou, Hunan Province, China) with a Cd concentration of 0.88 mg/kg. The basic physiochemical properties of the soil, including nitrogen, phosphorus and potassium contents, pH value, etc., were characterized and listed in Supplementary Table S1. After homogenization, the soil was packed into polyethylene plastic pots (15.7 cm diameter × 16.5c m height), each containing 1.5 kg of soil.

S. nigrum seeds were purchased from Shouguang Youhe Agricultural Technology Co., Ltd. (Shandong Province, China). The seeds were disinfected with 70% ethanol and sodium hypochlorite solution (2.5% active chlorine), washed with sterile water four times, and vernalized in 1‰ agar solution at 4 ℃ for 2 days, then sown in seedling pots and cultured in a plant growth chamber.

Pot experiments

S. nigrum seedlings with 5–7 true leaves and uniform growth were selected, and their roots were washed with sterile water and soaked in bacterial inoculum (10⁸ CFU/mL) or sterile water for 2 h, and then transplanted into pots (Fig. 1a) [7]. Each group was provided with 10 pots of S. nigrum. The pot experiment lasted for 2 months, during which the soil was sprayed with 150 mL bacterial stock inoculum (to maintain the concentration of soil bacterial inoculum at 10⁸ CFU/g soil) or sterile water (negative control group) every 2 weeks. The experiment was conducted in a greenhouse at Shanghai Jiaotong University, and the temperature can be maintained at 15 ~ 30 ℃ during the experiment. Plant and soil samples were taken separately at the end of the experiment. Roots were first carefully removed from the pot, and soil away from the roots (> 1 cm) was defined as bulk soil. Gently shaken the roots to remove the soil particles with weak adhesion, and then used a sterile clean brush to remove the residual soil attached to the roots and collected rhizosphere soil sample [7]. Plant samples were washed with sterile deionized water, followed by soaking the roots in 20 mM EDTA-Na₂ for 30 min to remove metal ions adhering to the root surface [30]. Part of the plant samples were oven-dried at 105 ℃ for 1 h, then dried at 80 ℃ to constant weight, and the dry weight and Cd content of each part were detected (Supplementary text: Determination of Cd and other metal elements concentration in samples). To obtain the root endosphere and leaf endosphere samples of S. nigrum, the surface of the sample was thoroughly disinfected according to the previously reported method, as described in the (Supplementary text: Surface sterilization of plant samples).

Fig. 1

Overview of the experimental design for this study. a Pot experiment to investigate the enhanced phytoremediation effect of PGPR and explore feature vectors. b Biomarkers mining based on community analysis and metabolic pathway exploration based on metabolomics. c Screening and identification of biomarkers. d Construction and simplification of synthetic communities

DNA extraction, amplification, and sequencing

DNA was extracted from bulk soil, rhizosphere soil, root endosphere, and leaf endosphere samples (Fig. 1b). The concentration and purity of DNA was assessed using Nanodrop (NC2000, Thermo Scientific, USA) and agarose electrophoresis. The hypervariable regions V5 and V7 of the bacterial 16S rRNA gene were amplified, employing the forward primer 799 F (5′-AACMGGATTAGATACCCKG- 3′) and the reverse primer 1193 R (5′-ACGTCATCCCCACCTTCC- 3′). The sequencing of the amplicons was performed using the Illumina Miseq platform, provided by Shanghai Personal Biotechnology Co., Ltd., China. Sequencing data processing is visible in the Supplementary text (Sequencing data processing).

Rhizosphere soil non-targeted metabolomics analysis

For metabolite extraction, an extraction solution (methanol: water = 4:1 (v:v)) containing 0.02 mg/mL of internal standard (L- 2-chlorophenylalanine) was used. The sample solution was ground for 6 min using a frozen tissue grinder (− 10 ℃, 50 Hz), followed by a low-temperature ultrasonic extraction for 30 min (5 ℃, 40 kHz). The sample was then allowed to stand at – 20 ℃ for 30 min and centrifuged for 15 min (4 ℃, 13,000 × g). The supernatant was transferred to a sample vial with an insert and analyzed using the instrumentation. The metabolic composition of S. nigrum rhizosphere soil samples was determined by non-targeted metabolic profiling using ultra-high performance liquid chromatography tandem Fourier transform mass spectrometry (UHPLC-MS/MS, Thermo Scientific, USA) supported by Majorbio Bio-Pharm Technology Co. Ltd. (China). Detailed parameters of instrument operation are described in Supplementary text (UHPLC-MS/MS analysis).

The LC–MS raw data were imported into the metabolomics processing software Progenesis QI (Waters Corporation, Milford, USA) for pre-treatment, resulting in the export of a three-dimensional data matrix in CSV format. This matrix encompassed comprehensive information, including sample details, metabolite names, and their corresponding mass spectral response intensities. Internal standard peaks and known false positive peaks (arising from noise, column bleed, and derivatized reagent peaks) were meticulously removed from the data matrix. Concurrently, metabolite identification was conducted by searching against multiple databases, including the HMDB (http://www.hmdb.ca/), Metlin (https://metlin.scripps.edu/), and the Majorbio Database. Further data analysis was presented in the Supplementary text (Soil metabolism data analysis).

Subsequently, the R package “ropls” (version 1.6.2) was employed to conduct principal component analysis (PCA) and orthogonal least partial squares discriminant analysis (OPLS-DA). To assess the stability of the model, a 7-cycle interactive validation was performed. Metabolites were deemed significantly different if they exhibited a Variable Importance in the Projection (VIP) score greater than 1 and a P value derived from two-sided unpaired Student’s t-test less than 0.05 after adjustment by Benjamini-Hochberg (BH) method.

The differentially expressed metabolites (DEMs) were further annotated through the metabolic pathways in KEGG (Kyoto Encyclopedia of Genes and Genomes) database (http://www.genome.jp/kegg/) to obtain the pathways in which the DEMs were involved. The python package “scipy.stats” (https://docs.scipy.org/doc/scipy/) was used for pathway enrichment analysis, and Fisher’s exact test was used to obtain the biological pathway most relevant to the experimental treatment. The P value was adjusted by the Benjamini-Hochberg (BH) method to control the false positive of enrichment results.

Isolation, purification, and identification of culturable bacteria

To maximize the diversity of bacterial species, we conducted screenings across all experimental samples, encompassing bulk soil, rhizosphere soil, root endosphere, and leaf endosphere samples (Fig. 1c). Furthermore, in addition to commonly used bacterial media (0.1 × and 0.5 × Tryptone soy agar (TSA), Reasoner’s 2 A (R2 A), and Luria–Bertani (LB)), we employed the Komodo website (http://komodo.modelseed.org) [31] to predict the specific media components suitable for each core bacterial genus (Supplementary Table S2). The bacterial isolation method was adapted from previous studies with appropriate modifications [26]. In brief, an appropriate amount of sample was added to sterile 10 mM MgCl₂ solution (1:10, w/v; plant samples need to be homogenized first) and incubated in a shaking incubator at 180 rpm and 28 °C for 1 h. Subsequently, the suspension was serially diluted (10⁻⁴ ~ 10⁻⁷) and plated on solid agar plates containing the aforementioned media. All plates were incubated upside down in a constant temperature incubator at 28 °C for 3–7 days. Subsequently, colonies exhibiting distinct morphologies on each medium were individually selected for further isolation and purification. Single colonies were picked and resuspended in 20 µL of sterile deionized water to create a bacterial suspension. Ten microliters of the suspension were aspirated into a PCR tube for further use. Additionally, 10 µL of 40% glycerol (v/v) was added to the remaining bacterial suspension, which was then stored at − 80 °C for preservation.

Colony PCR was performed using universal primers specific for the bacterial 16S rRNA gene (27 F: 5′-AGAGTTTGATCCTGGCTCAG- 3′; 1492R: 5′-CTACGGCTACCTTGTTACGA- 3′) and Taq DNA polymerase (Sangon Biotech, Shanghai, China) for amplification. The PCR reaction mixture consisted of 25 µL of Taq PCR Master Mix, 2 µL of each forward and reverse primer, 10 µL of bacterial suspension, and 11 µL of sterile deionized water. The PCR cycling parameters were as follows: pre-denaturation at 94 °C for 4 min, followed by 30 cycles of denaturation at 94 °C for 30 s, annealing at 58 °C for 30 s, and extension at 72 °C for 100 s, with a final extension at 72 °C for 10 min. The PCR products were verified through agarose gel electrophoresis and subsequently sequenced by Sangon Biotech Co., Ltd. (Shanghai, China). Sequence alignment was performed on the NCBI website (https://blast.ncbi.nlm.nih.gov/Blast.cgi).

Construction and simplification of synthetic communities

By aligning the sequences of isolated bacteria with those of ASVs belonging to key genera, those with a similarity greater than 97% are considered as successfully screened [15]. Multiple synthetic communities were further constructed: a synthetic community containing 8 key strains and NCT- 2 strain (1 community, including 9 species), and synthetic communities with one strain knocked out (9 communities, each containing 8 species) (Fig. 1d). According to the previously described protocol for preparing bacterial inoculum (section Materials preparation), each strain was first washed and resuspended, and the bacterial suspension was adjusted to a concentration of 10⁹ CFU/mL. Subsequently, the strains within the community were mixed in equal proportions. A microenvironment for cultivation was established on sterile soil, and the final inoculation concentration was set to 10⁸ CFU/g soil [25].

The source of the sterile soil is consistent with the aforementioned soil, achieved through two consecutive days of high-pressure steam sterilization (121 °C, 30 min). The treated soil was verified as sterile through liquid and plate culture methods. Sterile S. nigrum seedlings were obtained by disinfecting the seed surfaces and then germinating them on 1/2 Murashige and Skoog (MS) solid medium (Caisson, USA) for approximately 10 days. The experiments were conducted in tissue culture bottles (75 mm diameter × 110 mm height), with each bottle containing 80 g of sterile soil and three S. nigrum seedlings. Each experimental group had 12 replicate tissue culture bottles to ensure reproducibility. The tissue culture bottles were situated within a plant growth chamber, maintained at a temperature of 25 °C, a day/night photoperiod of 16/8 h, and a relative humidity level of 70%.

Two weeks later, the plants were gently extracted from the tissue culture bottles, and their biomass and Cd concentration were assayed as described earlier. Given the small size of the samples, the whole plant of S. nigrum was treated as a single unit for the detection of these parameters. And a biological repeated fresh weight was obtained by calculating the average value of three S. nigrum plants in each bottle, while a biological repeated Cd concentration was detected by mixing three S. nigrum plants in each bottle together. There were 9 biological replicates for each parameter.

To further simplify the community, three strains that led to a loss of enhancing effects during the one-strain knockout validation were combined to create a synthetic community designated as SynCom3, and the subsequent validation process followed the same methods as described above with a slight modification. The experiment period was extended to 4 weeks to obtain higher plant biomass sufficient for subsequent transcriptomic and metabolomic studies of root samples. Similarly, the roots of three S. nigrum plants in each bottle were mixed together and treated as a single sample for analysis. Furthermore, rhizosphere soil was collected, and the nitrate reductase activity of the rhizosphere soil samples was assayed according to the aforementioned methods.

RNA sequencing and data analysis

The root samples of S. nigrum were obtained as described above and subjected to RNA-seq analysis. The extraction of RNA and the construction of the sequencing library followed the Supplementary text (RNA extraction and library construction). The raw paired-end reads underwent trimming and quality control using fastp [32] with default parameters. The resulting clean data were then utilized for de-novo assembly with Trinity [33]. To enhance the assembly quality, we employed CD-HIT [34] and TransRate [35] for filtering the assembled sequences. Additionally, we assessed the quality of the assembly using BUSCO (Benchmarking Universal Single-Copy Orthologs) [36] analysis. The assembled transcripts were queried against the NCBI protein nonredundant (NR), Clusters of Orthologous Groups of proteins (COG), and KEGG database. To discern differentially expressed genes (DEGs) across the two groups, the expression level of each transcript was quantified using the transcripts per million reads (TPM) approach. RSEM [37] served as the tool for quantify gene abundances. Further, differential expression analysis was conducted utilizing the DESeq2 tool [38], and unigenes with a |Log₂fold change|≥ 1 and a P-value < 0.05 (after BH adjustment) were deemed to DEGs. Integrating DEGs into a gene set, further KEGG enrichment analysis was conducted using Python’s scipy software. Through Fisher’s exact test, a P-value less than 0.05 after BH adjustment was considered indicative of significant enrichment of a KEGG pathway within the gene set.

Root non-targeted metabonomics analysis

Extraction, detection, and analysis of root metabolites refer to the aforementioned non-targeted metabolomics analysis procedures of rhizosphere soil samples.

Real-time quantitative PCR

The abundance of SynCom3 members in the rhizosphere soil of S. nigrum was determined using qPCR. DNA extraction and detection were consistent with the above methods. Specific primers were set for the V5-V7 regions of the 16S rRNA genes of Sn319-Lysobacter and Sn486-Microbacterium, respectively. For the NCT- 2 strain, we used previously proven specific primer pairs, and all primers were listed in the Supplementary Table S3. We constructed the specific fragments of each strain onto the PUC-SP vector and established respective standard curves through a series of tenfold serial dilutions. The gene copy number was calculated using the following formula [7]:

$$\text{Gene copy number (copy) = }\frac{\text{6.02} \times {10}^{23}\text{(copy/mol)}\times\,\text{plasmid amount (g/}\upmu\text{L)}}{\text{plasmid length (bp)} \times 6{5}\text{0 (g/mol/bp)}}$$

in which 6.02 × 10²³ is Avogadro constant. The qPCR was performed on the QuantStudio 3 system (Thermofisher, USA) using TB Green® Premix Ex Taq™ II reagents (Takara Bio), with the reaction system and instrument operating parameters detailed in the Supplementary text (qPCR reaction system and operating parameters). Furthermore, the expression of nitrate reductase functional genes, nasB and nasC, in the rhizosphere soil of S. nigrum was also characterized using the aforementioned method.

Statistical analysis

Microsoft Excel 2010, SPSS 25.0, and R (v4.3.2) software were used for statistical analysis. Two-sided unpaired t-test, analyses of variance (ANOVA), and Duncan’s post hoc multiple test were used for significance analysis when the dependent variables met the normality requirement, and Mann–Whitney U test, Kruskal–Wallis test, and Dunn’s post hoc test were used for statistical analysis when the dependent variables did not meet the normality requirement. The R package, Origin 2023b, and a range of commercial cloud platforms (https://www.genescloud.cn/home, https://cloud.majorbio.com/page/tools/, https://www.chiplot.online/) were used for data presentation. Bioconcentration factor (BCF), translocation factor (TF), and phytoextraction ratio were calculated as previously described [29]. Least absolute shrinkage and selection operator (LASSO) regression analysis was implemented with the “glmnet” package. Fast expectation–maximisation microbial source tracking (FEAST) was calculated through the “FEAST” package [39] and visualized in the Origin 2023b. Co-occurrence network analysis was performed using the “WGCNA” package and visualized in Gephi (v0.10.1) or Cytoscape (v3.8.0). Structural equation modeling (SEM) and partial least squares path modeling (PLS-PM) were completed through “lavaan” and “plspm” packages, respectively. Phylogenetic tree of screened strains was constructed in MEGA7 based on maximum likelihood method (bootstrap 1000) and visualized through “ggtree” package. Linear regressions between variables were assessed using ordinary least squares (OLS) regression models.

Results

PGPR-mediated changes in physicochemical properties of rhizosphere soil induced enhanced phytoremediation effects

Here, we found that strain NCT- 2 successfully improved the Cd remediation efficiency of S. nigrum (Fig. 2a). Compared to the control group, the whole plant dry weight, Cd concentration, and Cd accumulation significantly increased by 16.83%, 16.92%, and 36.85% under strain NCT- 2 inoculation, respectively (Supplementary Fig. S1a–c, P < 0.05). We further studied the regulation effect of strain NCT- 2 inoculation on soil physiochemical properties. Compared to the control group, the electrical conductivity (EC) value of S. nigrum rhizosphere soil in strain NCT- 2 inoculation group was significantly increased (Supplementary Fig. 2b, P < 0.05), while the pH and cation exchange capacity (CEC) values of S. nigrum rhizosphere soil were significantly decreased (Supplementary Fig. S2a, c, P < 0.05). In addition, the available Cd content in the rhizosphere soil of the control group was significantly lower than that of the bulk soil (Supplementary Fig. S2 d, P < 0.05), while the strain NCT- 2 inoculation effectively increased Cd bioavailability for plant, so there was no significant change compared with the bulk soil (Supplementary Fig. S2 d, P > 0.05).

Fig. 2

Effects of strain NCT- 2 on Cd phytoremediation, and the relationships between soil physicochemical indexes and phytoremediation indexes that showed significant changes under strain NCT- 2 inoculation. a Effects of strain NCT- 2 on BCF (bioconcentration factor), TF (translocation factor), and PR (phytoextraction ratio) of S. nigrum (n = 4). b Pearson correlation analysis between soil physicochemical indexes and phytoremediation indexes that showed significant changes under strain NCT- 2 inoculation (n = 8). CEC, cation exchange capacity; EC, electrical conductivity; S-NiR, soil nitrite reductase; S-NR, soil nitrate reductase; S-UE, soil urease; S-SC, soil sucrase; S-ACP, soil acid phosphatase; ACd, available cadmium; TFe, total iron; TCa, total calcium; TN, total nitrogen; SOM, soil organic matter; SOC, soil organic carbon; AP, available phosphorus; AN, available nitrogen; NH₄ ⁺-N, ammonium nitrogen; NO₃ ⁻-N, nitrate nitrogen. c LASSO regression model for plant Cd accumulation and soil physicochemical indexes showing significant changes under strain NCT- 2 inoculation. Asterisks indicate significance: *P < 0.05, **P < 0.01, ***P < 0.001. The two-sided unpaired t-test was used for statistical significance testing. Error bars represent standard deviations

Similarly, the changes of macronutrients and micronutrients contents in soil were investigated. Overall, the effect of strain NCT- 2 inoculation on nutrient changes of bulk soil was negligible. For rhizosphere soil, the inoculation of strain NCT- 2 mainly changed the contents of soil total nitrogen (TN), calcium (TCa), iron (TFe), nitrate nitrogen (NO₃ ⁻-N), ammonium nitrogen (NH₄ ⁺-N), available nitrogen (AN), available phosphorus (AP), and organic carbon (SOC), which increased by 8.89%, − 24.55%, 25.22%, − 14.04%, 30.92%, 18.14%, 27.07%, and 16.90% respectively compared with the control group (Supplementary Fig. S3a, b and Table S4, P < 0.05). Therefore, we proceeded to assess the activity of soil enzymes involved in the soil C-N-P cycle. The results presented a high match with nutrient changes, in which the five selected soil enzymes: acid phosphatase (S-ACP, P-cycle), sucrose (S-SC, C-cycle), urease (S-UE), nitrate reductase (S-NiR), and nitrite reductase (S-NR, N-cycle), all showed higher enzyme activity in the rhizosphere soil under strain NCT- 2 treatment than that in the control group rhizosphere soil (Supplementary Fig. S3c–g, P < 0.05).

To discover the correlation between the above significant difference indicators and the enhanced phytoremediation effect by strain NCT- 2 inoculation, Pearson correlation analysis was used. We found that S-NR, SOC, and NH₄ ⁺-N were significantly positively correlated with plant dry weight, Cd concentration, and Cd accumulation (Fig. 2b, P < 0.05). We then used the least absolute shrinkage and selection operator (LASSO) regression model to quantify the contribution of various environmental factors mediated by strain to the enhanced remediation effect. We found that variables S-NR and AP were the key indicators that had a critical impact on the target variable (Cd accumulation), among which S-NR was the dominant vector given by the model, and its influence on Cd accumulation accounted for 97.02% of the total influence (Fig. 2c, R² = 0.652). In summary, our results indicated that inoculation of strain NCT- 2 could effectively improve the remediation efficiency of Cd in S. nigrum, and this strengthening effect may be mainly mediated by the improvement of rhizosphere S-NR activity.

PGPR altered the microbiome composition in the hyperaccumulator-soil interaction system

We first studied the α diversity of bacterial community of S. nigrum-soil interaction system after strain NCT- 2 inoculation. As shown in Fig. 3a, there was no significant difference in the α diversity and richness (Shannon and Chao1 index, P > 0.05) of bacterial community in each compartment between the strain NCT- 2 inoculation group and the control group. The principal coordinate analysis (PCoA) based on Bray–Curtis showed differences in the overall bacterial community composition, and significant differences were observed between the control group and the strain NCT- 2 inoculation group in bulk soil (BS), rhizosphere soil (RS), and root endosphere (RE) bacterial community composition, respectively, while insignificant differences were observed in leaf endosphere (LE) between the two groups (Fig. 3b, Adonis test; BS, RS, RE: P < 0.05; LE: P > 0.05). Additionally, we found that there was a close relationship between S-NR and bulk or rhizosphere soil bacterial communities both in strain NCT- 2 inoculation group and the control group (Supplementary Fig. S4, Mantel test, P < 0.05).

Fig. 3

Variation in bacterial communities after strain NCT- 2 inoculation. a Alpha diversity of bacterial communities in different compartments (n = 4). The P value was obtained through Kruskal–Wallis test and statistically significance was marked using Dunn’s multiple comparison test (*P < 0.05, **P < 0.01). b Principal coordinates analysis (PCoA) of bacterial communities based on Bray–Curtis distance (PerMANOVA by Adonis). Relative abundances of bacterial communities at the phyla (c) and genera (d) level in different compartments (n = 4). e Heat map of the top- 30 abundant bacterial genera across samples (n = 4). Similar samples were clustered horizontally, and vertical patterns illustrate the phylogenetic relationships. CBS, control group bulk soil; CRS, control group rhizosphere soil; BBS, treatment group bulk soil; BRS, treatment group rhizosphere soil; CRE, control group root endosphere; BRE, treatment group root endosphere; CLE, control group leaf endosphere; BLE, treatment group leaf endosphere

According to the results of ASV classification, the dominant phyla in each compartment were Proteobacteria (47.65% ~ 90.36%), Actinobacteria (3.25% ~ 35.50%), and Firmicutes (1.09% ~ 34.48%), accounting for 86.82% ~ 97.68% of the total reads (Fig. 3c). Compared with the control group, the relative abundance of Firmicutes in rhizosphere soil and root endosphere was significantly increased under the strain NCT- 2 inoculation, while the relative abundance of Chloroflexi was decreased (Supplementary Fig. S5b, c, P < 0.05). Besides, the relative abundance of Acidobacteria, Cyanobacteria, and Planctomycetes in rhizosphere soil were also significantly decreased under the strain NCT- 2 inoculation (Supplementary Fig. S5b, P < 0.05). For bulk soil, strain NCT- 2 inoculation only significantly reduced the relative abundance of Actinobacteria, while significantly increased the relative abundance of Acidobacteria, Cyanobacteria, Bacteroidetes, and Chloroflexi (Supplementary Fig. S5a, P < 0.05). Similarly, strain NCT- 2 inoculation did not significantly affect the composition of the bacterial communities in the leaf endosphere at the phylum level (Supplementary Fig. S5 d, P > 0.05).

At the genus level, the dominant genera in soil were different from those in plant endosphere, in which Sphingomonas, Bacillus, and Massilia were the dominant genera in soil, while Halomonas, Candidatus_Portiera, and Rickettsia were the dominant genera in plant endosphere (Fig. 3d). Interestingly, inoculation of strain NCT- 2 only significantly increased the relative abundance of Bacillus in rhizosphere soil and root endosphere, but significantly increased the relative abundance of Massilia, Ramlibacter, Arthrobacter, etc., in bulk soil compared with the control group (Supplementary Fig. S6, P < 0.05). In general, strain NCT- 2 inoculation mainly affected the bacterial community composition in S. nigrum bulk soil, rhizosphere soil, and root endosphere. Moreover, the heat maps of the top 30 genera and phyla showed that the treatment group bulk soil-control group rhizosphere soil (BBS-CRS) and treatment group rhizosphere soil-control group bulk soil (BRS-CBS) were clustered together, respectively (Fig. 3e). Combined with PCoA and community composition analysis, it seems that at the dominant phyla and genera level, strain NCT- 2 inoculation mainly enriched Bacillus (Firmicutes) in S. nigrum rhizosphere soil and root endosphere, and drove some taxa bacteria into bulk soil at the same time (e.g., Acidobacteria, Chloroflexi), resulting in certain similarities in bacterial community composition between BBS and CRS.

PGPR converged the bulk soil microbiome with the uninoculated rhizosphere soil

We then employed fast expectation–maximization microbial source tracking (FEAST) to trace the potential sources of microbiome in each compartment (Supplementary Fig. S7). We found that the microbiome in BBS was mainly derived from CRS, accounting for 34.19%. UpSet Venn diagram also presented that the number of ASVs shared in BBS-CRS was the highest (Supplementary Fig. S8).

In addition, we performed co-occurrence network analysis for ASVs with relative abundances greater than 0.1% (Supplementary Fig. S9 and Table S5). We found that although strain NCT- 2 inoculation resulted in a decrease in the modularity of the rhizosphere soil bacterial community network compared with the control group, it improved the complexity and stability of the rhizosphere soil bacterial community network, according to the edge, average degree, density, average clustering coefficient, and average path length indexes of the network diagram. Interestingly, in this case, the rhizosphere soil bacteria community did not show higher competitive intensity, but tended to cooperate with each other, as reflected by the positive correlation proportion (CRS 77.45%; BRS 91.99%). On the contrary, for bulk soil, strain NCT- 2 inoculation improved the modularity of the bacterial community network, while reducing the complexity and stability of the bacterial community network, as evidenced by the edge, average degree, density, average clustering coefficient, and average path length indexes. What’s more, on the whole, the bacterial community networks of the BBS and CRS showed higher similarity.

Dynamic changes of microbiome and S-NR were associated with PGPR colonization in rhizosphere soil

Furthermore, we carried out a detailed analysis of the genus Bacillus to which the strain NCT- 2 belongs at the ASV level. By sequence alignment, we observed that only the sequence of ASV_20544 was 100% similar to the sequence of strain NCT- 2, and the ASV_20544 was largely colonized in S. nigrum rhizosphere soil and root endosphere of strain NCT- 2 inoculation group (Supplementary Fig. S10). We previously noted that strain NCT- 2 inoculation mainly significantly altered the relative abundance of genus Bacillus in S. nigrum rhizosphere soil and root endosphere, so we then performed ordinary least squares (OLS) regression analysis on the variation of Bacillus and ASV_20544 abundance in the rhizosphere soil and the root endosphere. We found that there was a highly significant correlation between them, both in the rhizosphere and root endosphere (Supplementary Fig. S11, RS: R ² = 0.963, P < 0.001; RE: R ² = 0.996, P < 0.001). However, Mantel test showed that only the alterations of bacterial community in rhizosphere soil were significantly correlated with the changes of ASV_20544 abundance, while the strain NCT- 2 colonization did not significantly affect the alterations of the bacterial community in root endosphere (Fig. 4a, Mantel test, RS: P < 0.05; RE: P > 0.05). Also, we observed that the abundance of ASV_20544 in rhizosphere soil was also significantly positively correlated with S-NR (Fig. 4b, R² = 0.778, P < 0.01).

Fig. 4

Association of strain NCT- 2 colonization with soil bacterial communities and nitrate reductase and effects on enriched biomarkers and metabolic profiles in rhizosphere soil. a Correlations between bacterial community and ASV_20544 abundance (n = 8). The statistical significance of comparisons is assessed using Mantel test based on Pearson’s product moment correlation using 9999 permutations. RS, rhizosphere soil; RE, root endosphere. b Ordinary least squares (OLS) linear regression between the soil nitrate reductase (S-NR) and the abundance of ASV_20544 (n = 8). The fitted lines are regression lines from OLS regression, and the shaded areas indicate 95% confidence interval of the fit. The statistical test used is F-test. c Based on the random forest model prediction biomarkers of rhizosphere soil bacterial community. d Linear discriminant analysis (LDA) effect size cladogram of the rhizosphere soil bacterial community. The taxon from the inner circle to the outer circle is phylum to genus, and the node size corresponds to the average relative abundance of the taxon, in which hollow and solid nodes represent insignificant and significant differences between groups, respectively. CRS, control group rhizosphere soil; BRS, treatment group rhizosphere soil. e KEGG pathway enrichment analysis of differentially expressed metabolites (DEMs) in rhizosphere soil under strain NCT- 2 inoculation. DA Score reflects the overall changes of all metabolites in the metabolic pathway, with a score of 1 indicating an upregulation of the expression trend of all annotated DEMs in the pathway, and − 1 indicating a downregulation of the expression trend of all annotated DEMs in the pathway. The size of the dots indicates the number of DEMs annotated in the pathway. The dots are distributed on the right side of the central axis, and the longer the line segment, the more inclined the overall expression of the pathway is to be upregulated; On the contrary, it indicates that the overall expression of the pathway tends to be downregulated. Enrichment analysis was performed using Fisher’s exact test, and P values were corrected by the BH method (**P < 0.01, ***P < 0.001)

Biomarkers with low-abundance were enriched by PGPR colonization

We sought to illustrate whether strain NCT- 2 inoculation would recruit beneficial and collaborative partners within S. nigrum rhizosphere soil at the overall genus level, including some genera with low abundance ranking, hence LDA effect size (LEfSe) analysis and random forest prediction were conducted to retrieve key genera. LEfSe analysis demonstrated that at the genus level, compared with the control group, genera Granulicella, Microbacterium, 37_13, Bacillus, Clostridium_sensu_stricto_11, Sphingopyxis, Blfdi19, Pseudoduganella, Ralstonia, Vogesella, Rhodanobacter, Lysobacter, Pseudoxanthomonas, and Thermomonas were significantly enriched in the rhizosphere soil under strain NCT- 2 inoculation (Fig. 4d). Moreover, in addition to Sphingobacterium, the other biomarkers predicted by random forest analysis were also the genera that were significantly enriched in LEfSe analysis (Fig. 4c). Further analysis suggested that except genus Bacillus, the other enriched genera in the control group rhizosphere soil were all low-abundance bacteria (relative abundance was lower than 0.1%), and the relative abundance of half of them was higher than 0.1% after strain NCT- 2 inoculation (Supplementary Table S6). We also found that the abundance of most of the enriched biomarkers was significantly positively correlated with the colonization abundance of strain NCT- 2 (Supplementary Fig. S12).

Potential metabolites associated with key microbial taxa

Soil non-targeted metabolomics was used to investigate the variation of rhizosphere soil metabolic profile under strain NCT- 2 inoculation. Partial least squares discriminant analysis (PLS-DA) showed that strain NCT- 2 inoculation significantly interfered with the metabolic profile of rhizosphere soil (Supplementary Fig. S13a), with 153 metabolites significantly adjusted, 114 of which were significantly upregulated (Supplementary Fig. S13b, P _BH < 0.05). Then, we conducted enrichment analysis of the KEGG pathway for differentially expressed metabolites (DEMs), and found that there were six pathways with significant enrichment. According to the differential abundance score map, we found that five of them were significantly upregulated under strain NCT- 2 inoculation, including tryptophan metabolism, arachidonic acid metabolism, phenylpropanoid biosynthesis, etc. (Fig. 4e, P _BH < 0.01).

Subsequently, we analyzed the link between DEMs in the six pathways, key genera of bacterial community, as well as the abundance of ASV_20544 and S-NR (Supplementary Fig. S14). We found strong associations between key bacterial genera and DEMs, with Lysobacter in key genera showing the largest degree of connectivity and being significantly associated with most of the DEMs. Additionally, more positive and close co-occurrences were observed between ASV_20544 and key bacterial genera, while more positive and close co-occurrences were observed between S-NR and DEMs. The structural equation model (SEM) further explained the relationship between the four factors: rhizosphere colonization of the strain stimulated an increase in S-NR activity (path coefficient 0.89, P < 0.001), which further altered the rhizosphere metabolic profile (path coefficient 0.57, P < 0.001), driving the enrichment of key taxa (path coefficient 1.49, P < 0.001) (Supplementary Fig. S15). Interestingly, although the path coefficients from S-NR to DEMs and strain colonization abundance to DEMs were both significant, the coefficient of the former was larger, suggesting a stronger direct relationship between S-NR and DEMs, which also supports the hypothesis of the above network diagram.

Construction and simplification of synthetic communities

To acquire the biomarkers significantly enriched in the rhizosphere soil of S. nigrum under strain NCT- 2 inoculation, we used the universal medium and specific screening medium to screen the key species. A total of 542 bacterial isolates were obtained from five phyla belonging to 56 genera (Supplementary Fig. S16). Sequence alignment suggested that a total of 8 key genera were successfully screened (Supplementary Table S7). Based on the abundance of corresponding ASVs, sequence similarity, and significant differences between groups (Supplementary Fig. S17a, Supplementary Table S7), we selected 8 strains of the isolated bacteria, which were Sn486-ASV_8196, Sn70-ASV_18570, Sn492-ASV_68734, Sn504-ASV_8575, Sn319-ASV_61176, Sn328-ASV_1219, Sn501-ASV_67244, Sn17-ASV_320, and they can well represent the corresponding genus according to OLS linear regression analysis (Supplementary Fig. S17b, c, R ² > 0.9, P < 0.001).

Based on the key strains obtained by screening, we further constructed multiple synthetic communities, including the synthetic community of 8 key genera and strain NCT- 2 (SynCom), and the synthetic community of knocking out one of the strains (SynCom-Lys, SynCom-Bac, SynCom-Sph, SynCom-Pse, SynCom-Mic, SynCom-Ral,SynCom-The, SynCom-Rho, SynCom-NCT- 2), and investigated their effects on Cd remediation by S. nigrum in a sterile soil system (Fig. 5a, b; Supplementary Fig. S18a–c). We found that under sterile conditions, single strain NCT- 2 inoculation had limited effect on rhizosphere S-NR activity, and could not effectively promote the plant growth and improve the Cd accumulation in S. nigrum (P > 0.05). However, SynCom inoculation not only significantly increased the rhizosphere S-NR activity, but also effectively improved the Cd accumulation in S. nigrum (P < 0.05). The one-strain knock out test showed that when Lysobacter, Microbacterium, or strain NCT- 2 was knocked out respectively, the synthetic community could not effectively improve the S-NR activity, as well as the Cd accumulation in S. nigrum (P > 0.05). Nevertheless, the remaining one-strain knocked out synthetic communities still significantly increased the rhizosphere S-NR activity and the Cd accumulation in S. nigrum (P < 0.05).

Fig. 5

Effects of different synthetic communities on Cd phytoremediation by S. nigrum. a Results of one-strain knockout experiment. b Cd accumulation results of one-strain knockout experiment. Biomass (c), Cd concentration (d), Cd accumulation (e), and soil nitrate reductase (f) performance in simplified synthetic community experiments (n = 9). g Colonization abundance of each SynCom3 member in rhizosphere soil (n = 9). h Expression of soil nitrate reductase encoding genes in rhizosphere soil (n = 9). i OLS linear regression between the strain NCT- 2 colonization abundance and the Sn319 or Sn486 colonization abundance in the rhizosphere soil (n = 18). The fitted lines are regression lines from OLS regression, and the shaded areas indicate 95% confidence interval of the fit. The statistical test used is F-test. In b-d and f, the two-sided unpaired t-test was used for statistical significance testing (**P < 0.01, ***P < 0.001). In g, different letters above bars indicate significant differences (P < 0.05) according to Duncan’s multiple comparison. In e and h, Mann–Whitney U test was used for statistical significance testing (***P < 0.001). Error bars represent standard deviations

We then simplified the synthetic community, constructed a microbiota composed of Lysobacter (Sn319), Microbacterium (Sn486), and strain NCT- 2 (SynCom3), and verified its reinforcement Cd remediation effect compared to single strain NCT- 2 inoculation. We were surprised to find that SynCom3 inoculation significantly increased the whole plant biomass, Cd concentration, Cd accumulation, and rhizosphere S-NR activity by 94.83%, 32.17%, 153.04%, and 65.99%, respectively, compared with single strain NCT- 2 inoculation (Fig. 5c–f, P < 0.01). Furthermore, the OLS linear regression analysis of S-NR and Cd accumulation in S. nigrum again verified that there was a significant positive correlation between them (Supplementary Fig. S19a, R ² = 0.893, P < 0.001). In short, our results indicate that the simplified SynCom3 consisting of Lysobacter, Microbacterium, and strain NCT- 2 is effective and efficient.

We further examined the colonization levels of community members in the rhizosphere soil of S. nigrum (Supplementary Fig. S20). qPCR results indicated that the abundance of strain NCT- 2 in the rhizosphere soil was relatively low when inoculated alone, while its colonization abundance was significantly increased by 335.70% under the SynCom3 inoculation (Fig. 5g, P < 0.05). In addition, colonization of Lysobacter (Sn319) and Microbacterium (Sn486) was not detected in the single NCT- 2 strain inoculation group, but their colonization was successfully detected in the SynCom3 inoculation group (Fig. 5g), and their colonization abundance was significantly positively correlated with the colonization abundance of NCT- 2 strain (Fig. 5i, R² > 0.8, P < 0.001). Interestingly, although the colonization abundance of all three strains was significantly positively correlated with S-NR, it is evident that there was a higher correlation between colonization abundance of strain NCT- 2 and S-NR (Supplementary Fig. S19b, NCT- 2: R ² = 0.745, P < 0.0001; Sn319: R ² = 0.549, P < 0.001; Sn486: R ² = 0.579, P < 0.001). Moreover, qPCR results also showed that the key coding gene for S-NR, nasC, showed higher expression levels under SynCom3 inoculation (Fig. 5h, P < 0.001), and its expression level was significantly positively correlated with S-NR activity, while nasB existed no significant difference and was not related to changes in S-NR activity (Supplementary Fig. 19c, nasC: R ² = 0.750, P < 0.001; nasB: R ² = 0.157, P > 0.05).

SynCom3 induced enrichment of phenylpropanoid biosynthesis and tryptophan metabolism pathways

We further used the transcriptome sequencing to investigate the differentially expressed genes (DEGs) transcription profile in the roots of S. nigrum inoculated with SynCom3. A total of 171,923 transcripts were obtained, of which 94,761 were unigenes. PCA analysis of the samples revealed a separation between the SynCom3-inoculated group and the single strain NCT- 2-inoculated group (Supplementary Fig. S21a). Analysis of differential gene expression between the groups showed that, compared to inoculation with strain NCT- 2 alone, inoculation with SynCom3 significantly upregulated 5943 genes and downregulated 6388 genes in the roots of S. nigrum (Supplementary Fig. S21b, P _BH < 0.05). Functional annotation analysis revealed that the DEGs were primarily clustered in the metabolism category of carbohydrate metabolism and amino acid metabolism (Fig. 6a). KEGG enrichment analysis further demonstrated significant enrichment in pathways phenylpropanoid biosynthesis, ubiquinone and other terpenoid-quinone biosynthesis, phenylalanine, tyrosine and tryptophan biosynthesis, etc., under SynCom3 inoculation (Fig. 6b, P _BH < 0.05). Notably, both phenylpropanoid biosynthesis and tryptophan metabolism pathways were significantly enriched in the previously identified soil DEMs.

Fig. 6

Transcription and metabolic profiles of S. nigrum roots inoculated with SynCom3 or single NCT- 2 strain. a Histogram for KEGG pathway annotation of the differentially expressed genes (DEGs). b Bubble diagram for KEGG enrichment analysis in transcriptomics. The size of bubbles is directly proportional to the number of DEGs enriched in the KEGG pathway. c Bubble diagram for KEGG topology analysis in metabolomics. The size of bubbles is directly proportional to the importance of the KEGG pathway. The relationships between the DEMs in pathways phenylpropanoid biosynthesis and tryptophan metabolism and the colonization abundance of strain NCT- 2 (d), Sn319 (e), and Sn486 (f) in rhizosphere soil (n = 12). The fitted lines are regression lines from OLS regression, and the shaded areas indicate 95% confidence interval of the fit. The statistical test used is F-test. In b and c, Fisher’s exact test was used, a P-value less than 0.05 after BH adjustment was considered indicative of significant enrichment of a KEGG pathway

We next endeavored to ascertain the alterations in the root transcriptomic induced by SynCom3 inoculation, particularly the heightened expression of genes associated with the phenylpropanoid biosynthesis and tryptophan metabolism, would result in corresponding metabolites variations within the root, and employed non-targeted metabolomics on S. nigrum roots by using UHPLC-MS/MS. We found that compared with single strain NCT- 2 inoculation, 118 metabolites in roots exhibited significant differences under SynCom3 inoculation (Supplementary Fig. S22a, P _BH < 0.05). Moreover, we conducted a KEGG pathway enrichment analysis for the DEMs. Interestingly, pathways phenylpropanoid biosynthesis and tryptophan metabolism were again significantly differentially enriched under SynCom3 inoculation (Fig. 6c, P _BH < 0.05), and the related metabolites in these pathways were all significantly upregulated compared to the single strain NCT- 2 inoculation (Supplementary Fig. S22b, P _BH < 0.05), further suggesting that SynCom3 inoculation could stimulate these two pathways in S. nigrum roots during its enhancement of Cd remediation. The correlation analysis between DEMs in these two pathways and the colonization abundance of each member of the SynCom3 showed that the abundance of Lysobacter and Microbacterium were significantly positively correlated with DEMs, while the direct relationship between strain NCT- 2 and DEMs was relatively weak (Fig. 6d–f).

Further, the analysis results based on the partial least squares path modeling (PLS-PM) showed that, strain NCT- 2 affected the Cd accumulation by S. nigrum through multiple direct and indirect pathways (Supplementary Fig. S23). The direct effect of strain NCT- 2 on S-NR was the most significant (path coefficient 0.86, P < 0.001), and S-NR significantly promoted the Cd accumulation (path coefficient 0.50, P < 0.05). These results indicated that S-NR was the key mediating factor of the effect of strain NCT- 2 on Cd accumulation by S. nigrum. In addition, S-NR had significant positive regulatory effects on DEMs (path coefficient 0.80, P < 0.05), and DEMs also had a significant positive regulatory effect on the recruitment of partners within the community (Lysobacter (Sn319) and Microbacterium (Sn486)). However, S-NR had no significant effect on the enrichment of Sn319/Sn486 (path coefficient 0.08, P > 0.05). This again suggests that the enrichment of these two partners under strain NCT- 2 inoculation may be directly driven by DEMs rather than S-NR. Moreover, although the effect of Sn319/Sn486 on Cd accumulation was negative (path coefficient − 0.08), it did not reach a significant level and the path coefficient value was low. Overall, the R ² value of the model was 0.966 with high explanatory power, and the overall fit of the model was high (GoF = 0.87), indicating the reliability of the results. These results reveal the complex regulatory mechanism of strain NCT- 2 in the accumulation of Cd in S. nigrum, emphasizing the core role played by S-NR in this process.

Discussion

Beneficial microbial inoculation can effectively stimulate plant growth and development under heavy metal pollution and enhance phytoremediation in contaminated soil. Here, we elucidate how exogenous PGPR strengthen Cd-phytoremediation by S. nigrum. Initially, we found that the PGPR strain NCT- 2 primarily enhanced rhizosphere nutrient acquisition and stimulated Cd-phytoremediation by increasing the activity of soil nitrate reductase, a key enzyme involved in soil nitrogen cycle. Further, we observed a close relationship between S-NR activity and the colonization abundance of the strain NCT- 2. Additionally, colonization by the strain may occur through reshaping the bacterial community’s microecology and enriching certain low-abundance bacterial genera in the rhizosphere soil, which also affected the soil metabolic profile simultaneously. Finally, we constructed and simplified a synthetic community with the strain NCT- 2 as the core and two low-abundance bacterial genera as collaborators, and validated its effectiveness. Particularly, by utilizing transcriptomics and non-targeted metabolomics, we reaffirmed the critical role of phenylpropanoid biosynthesis and tryptophan metabolism during the strain enhancement process.

S-NR is the key factor of PGPR mediated enhanced phytoremediation

Heavy metal pollution, including Cd contamination, can adversely affect soil fertility and hinder nutrient uptake by plants, thereby impeding plant growth [40]. Here, our results suggested that the inoculation of strain NCT- 2 effectively increased the available nutrients content (e.g., NH₄ ⁺-N, AP, SOC) in the rhizosphere soil of S. nigrum, which could be attributed to the increase in soil enzyme activity related to nutrients cycling following strain inoculation. It is noteworthy that S-NR, defined as the primary predictor of the enhanced Cd remediation effect of the strain NCT- 2 on S. nigrum, exhibited a significant increase in enzyme activity following strain inoculation. The activity of S-NR serves as a limiting factor in the nitrogen assimilation pathway, which in turn is the initial process of soil nitrogen priming effect and can promote soil organic nitrogen mineralization [41, 42]. We previously confirmed that the strain NCT- 2 primarily utilizes nitrate efficiently through the assimilation pathway and maintains its stable growth [43]. Here, the high-density of strain colonization in rhizosphere soil was indeed accompanied by a more active nitrate assimilation process, representing more nitrate being utilized by the strain to convert into microbial biomass nitrogen required for its own growth. Although the assimilated organic nitrogen was stored within cells and cannot be utilized by other organisms, the rapid turnover of the microbial community resulted in the subsequent release of a substantial amount of organic nitrogen [44, 45]. However, plants are generally unable to compete with soil microorganisms for organic nitrogen [16]. Coincidentally, strain NCT- 2 also possesses the ability to mineralize organic nitrogen into ammonia, thereby increasing the environment ammonium concentration [21]. Spatial and temporal variations in environmental nutrients drive metabolic heterogeneity in microbial population [46]. The nascent strain NCT- 2 population may swiftly convert the organic nitrogen compounds released by the prior population into ammonia via mineralization, further enabling other organisms to utilize them. Additionally, under Cd stress or acidic soil conditions, plants primarily utilize ammonium, while the uptake system for nitrate may be inhibited [19, 47]. Therefore, the strain may promote the utilization of nitrogen in S. nigrum by regulating the nitrogen cycle in rhizosphere soil, thus stimulating the plant growth. On the other hand, higher contents of ammonium can also increase the uptake and transport of Cd by S. nigrum, thereby enhancing Cd phytoextraction [18]. The reason may be related to the regulation of soil pH by ammonium nitrogen. In the soil environment, ammonium nitrogen may acidify the soil by releasing H⁺, thereby improving the solubility of Cd in the soil and making Cd more easily absorbed and utilized by plants [48]. Our results also confirmed that PGPR inoculation can reduce soil pH.

PGPR colonization reshapes rhizosphere microbiome

The effectiveness of microbial inoculants is closely related to their colonization abundance [25, 49]. Here, we found that the strain NCT- 2 predominantly colonized in rhizosphere soil and root endosphere of S. nigrum. Moreover, the strain’s colonization directly led to a significant increase in the relative abundance of genus Bacillus in its residing environment, making it the dominant genus, especially in the rhizosphere soil as the highest abundance genus. Interestingly, the changes in bacterial community within the rhizosphere soil were significantly correlated with the abundance of the strain NCT- 2. Similarly, they were significantly correlated with the activity of S-NR within the rhizosphere soil, rather than the bacterial community within the root endosphere. Therefore, we are more concerned about the ecological effects brought about by the colonization of the strain NCT- 2 in the rhizosphere soil.

The destiny of inoculants hinges on biological interactions, potentially encompassing complex networks involving competition, cooperation, or interactions with other microorganisms [24, 50]. Based on co-occurrence network analysis, we found that strain inoculation led to more positive interactions within the bacterial community network in the rhizosphere soil. This increased cooperation among bacteria in the community while enhancing the complexity and stability of the bacterial community network. Increased cooperation may imply the displacement of competitors. The reduced relative abundance of Chloroflexi, Acidobacteria, and Cyanobacteria in the rhizosphere soil following strain inoculation precisely corresponds to the increased relative abundance of these phyla in the bulk soil. Furthermore, the bacterial community network topology of BBS also tends to be similar to that of CRS, rather than CBS or BRS. Additionally, heatmap analysis of community composition similarly demonstrate a higher similarity between BBS and CRS. These suggested that while strain NCT- 2 dominated the primary ecological niche in the rhizosphere soil, it drove some indigenous microorganisms to the bulk soil. This led to a convergence of community composition and structure between BBS and CRS. And it was further confirmed through tracing the sources of communities in BBS and discussing the shared number of ASVs among different compartments.

We then wondered whether the strain colonization strategy would recruit partners in addition to driving out competitors. Apparently, no genera in the high-abundance genus (> 0.1%) were enriched, except for the Bacillus genus to which the strain belongs. Further screening of biomarkers revealed 15 genera significantly enriched in the rhizosphere soil by the inoculant. Without exception, these genera were all of low-abundance in the control group (except for the genus Bacillus), and half of them became high-abundance genera after inoculation. Notably, most of the biomarkers are Proteobacteria, which are widely recognized as copiotrophs that rapidly respond to nutrient inputs and grow at high rates [51]. Either the expulsion of competitors or the recruitment of collaborators may be based on the modification of rhizosphere environment by inoculants. As we described earlier, strain NCT- 2 inoculation led to an increase in available nutrients in the rhizosphere soil, which may allow the copiotrophs (Firmicutes, Proteobacteria) to proliferate rapidly to dominate the rhizosphere niche, to the disadvantage of oligotrophs (Chloroflexi, Acidobacteria, and Cyanobacteria) [52,53,54].

We then aimed to understand the role of rhizosphere metabolites in strain’s colonization strategies. We found that several pathways related to the metabolism and synthesis of amino acids and fatty acids were significantly enriched under the strain inoculation. And there was a strong correlation between the DEMs in these pathways and the enriched key genera. Fatty acids and amino acids can be used as biomarkers of carbon and nitrogen respectively. Their metabolism and synthesis directly affect the nutrients cycle and are closely related to the growth and activity of microorganisms [55, 56]. Additionally, we noticed that the relationship between S-NR and DEMs in rhizosphere soil was closer than that between inoculum abundance and DEMs. We hypothesized that the alteration of rhizosphere metabolic profile was induced by inoculant driven increase of S-NR activity rather than directly induced by inoculant colonization, and the alteration of rhizosphere metabolic profile further enriched key genera.

Syncom3-mediated cascade effects of plant-microbial interactions

Using selected biomarkers, we established and further streamlined the microbiota using a single-strain knockout strategy in a sterile environment, resulting in the optimized community SynCom3 centered around strain NCT- 2 with Lysobacter and Microbacterium serving as collaborators. SynCom3 demonstrated superior reinforcement Cd-phytoremediation, which was achieved based on the more efficient colonization of strain NCT- 2. The community assembled by bacteria with diverse ecological functions may exhibit greater functional redundancy and stability. For instance, with one key bacterium as the core species, while the remaining members can effectively strengthen its ecological functions by promoting its growth or stimulating the expression of its functional genes [25, 26, 57]. We observed that higher colonization abundance of strain NCT- 2 in rhizosphere soil was accompanied by higher expression of NR coding gene nasC and higher S-NR activity. The improvement of S-NR activity promoted nitrate assimilation, made nitrogen cycling more active in plant rhizosphere and directly affect the metabolism and synthesis of root amino acids.

Similar to the aforementioned changes in soil metabolic profiles, SynCom3 inoculation also drove significant enrichment of phenylpropanoid biosynthesis and tryptophan metabolism pathways in S. nigrum roots. As a precursor substance for auxin synthesis, the more active tryptophan metabolism here did drive the accumulation of 3-indoleacetic acid in S. nigrum roots, which has been widely reported to promote the growth and development of plant roots [58, 59]. Additionally, it was found that the content of quinolinic acid in the tryptophan metabolic pathway was significantly increased under the SynCom3 inoculation, which may be conducive to the SynCom3 assisting in the detoxification of Cd in S. nigrum. Quinolinic acid is an essential nitrogenous aromatic heterocyclic compound, which has been shown to enhance plant tolerance to stress by removing reactive oxygen [60].

Moreover, the phenylpropanoid biosynthesis pathway results in the production of numerous phenolic acids, including 4-hydroxycinnamic acid and trans-cinnamic acid, that were significantly enriched under SynCom3 inoculation in this study [61]. These compounds are extensively involved in plant stress resistance and environmental adaptation [62, 63]. As the initial product of phenylpropanoid metabolic pathway, trans-cinnamic acid is catalyzed by phenylalanine through phenylalanine ammonia lyase, and then it can be converted into a variety of downstream metabolites, such as lignin and other secondary metabolites [64, 65]. Similarly, 4-hydroxycinnamic acid, as one of the key intermediates in the metabolic pathway of phenylpropane, is also involved in the synthesis of lignin [27]. These results suggest that SynCom3 inoculation may enhance the mechanical strength and resistance of plant cell walls by stimulating lignin synthesis, and help plants maintain structural integrity and chelate and detoxify Cd under heavy metal stress. In addition, 4-hydroxycinnamic acid is also a precursor of secondary metabolites such as flavonoids and coumarins, which has been widely proven to have antioxidant and chelating effects and can effectively relieve oxidative stress under heavy metal stress [66]. It is noteworthy that 4-hydroxycinnamic acid has recently been reported to drive the enrichment of beneficial bacteria, thereby maintaining the homeostasis of the rice phyllosphere [27]. Metabolite-driven enrichment and assembly of specific functional bacteria or expulsion of harmful bacteria within the environment are important ways to maintain ecological niche homeostasis and microbiota stability as well as plant health [67,68,69]. Hence, the enrichment of 4-hydroxycinnamic acid under SynCom3 inoculation may also facilitate the assembly and colonization of the SynCom3 community, thus enhancing the efficacy of Cd phytoremediation.

Again, we observed that the correlation between most DEMs and the colonization abundance of Lysobacter and Microbacterium surpasses that between NR and these same parameters. Consequently, the enrichment growth of Lysobacter and Microbacterium, directly fueled by DEMs, may be more compelling than that driven by S-NR. However, the molecular details of bacterial interactions mediated by metabolites within three bacterial species at the root-soil interface may be highly intricate, and will be the focus of future research.

Our study revealed that effective colonization of the beneficial strain B. megaterium NCT- 2 in S. nigrum rhizosphere is the prerequisite for its enhancement of Cd phytoremediation, and the key enzyme of nitrogen cycle, S-NR, primarily mediates this enhancement. The colonization of the strain and the modification of the rhizosphere environment by S-NR altered the metabolic profile at the root-soil interface, particularly in terms of amino acid metabolism and biosynthesis, thereby shaping a more favorable microbiota for strain survival. Further simplified community experiment analysis confirmed the positive feedback of the two auxiliary strains on the colonization and strengthening capabilities of strain NCT- 2. Collectively, our findings advance our understanding of how beneficial bacteria regulate phytoremediation of Cd-contaminated soil by modifying root-soil interface environmental parameters (Fig. 7), and will help guide future microbiome operations aimed at promoting plant health and phytoremediation efficiency and sustainable production under heavy metal stress.

Fig. 7

A schematic model summarizing the hypothesized mechanisms of how PGPR enhances its effect on Cd phytoremediation by modifying the rhizosphere microenvironment. PGPR, plant growth-promoting rhizobacteria