Sampling of the rhizosphere soil and various plant tissues, including leaf, stem, and root, was conducted in December 2022 using the methods we adopted in a previous study (Fallah et al., 2023a). In summary, random sampling of plant tissues of three healthy sugarcane plants was carried out from each plot as a biological replicate. The plant tissues were washed using phosphate buffer solution and cleaned using 75% alcohol cotton. We wrapped the samples in tin foil, placed them in liquid nitrogen for 5 min, stored in dry ice, transported to the laboratory, and stored at −80°C. The sample of soil taken from the plant root zones was classified as rhizosphere soil, with each group of the soil sample containing three replicates. Finally, we generated a total of 36 samples. Later, we air-dried a portion of the soil sample, which was ground and sieved using a 2-mm mesh to investigate soil chemical properties.

Extech Portable Sucrose Brix Refractometer (Mid-State Instruments, CA, United States) was used to estimate sucrose content and measured using the formula: sucrose (%) = Brix (%) × 1.0825–7.703 (Yuan et al., 2022). A random sampling of 25 sugarcanes was conducted in each row to estimate the diameter of the plants using a vernier. Sugarcane heights were measured in centimeters (cm) from the soil’s surface to the sugarcane’s top by randomly sampling 25 plants in each row using a meter rod. We added the average of three replicates to determine the mean of sugarcane heights.

An element analyzer (vario MAX cube, Germany) was used to estimate the TN. Soil potential of hydrogen (pH) was tested using a glass electrode pH meter. We used fresh soil samples to extract soil NO₃ ⁻-N and NH₄ ⁺-N with 2.0 M KCl and calculated using a continuous-flow analyzer (San++, Skalar, Holland) (Zhao et al., 2019).

The soil samples were incubated using buffer sodium carboxymethylcellulose solution and cellulose activity (glucose, mg/g 24 h, 37°C) to assess colorimetrically by measuring a decrease in 3,5-dinitrosalicylic acid from reducing sugar. We estimated soil urea activity (NH₃-N, mg/g 24 h, 37°C) using sodium hypochlorite colorimetry and improved sodium phenolate. Acid phosphatase activity was tested using a nitrophenyl phosphate disodium substrate (phenol, μg/g, 1 h, 37°C). Soil β-glucosidase activity was measured with a colorimetric p-nitrophenol assay after buffering the soil with p-nitrophenyl-β-glucopyranoside (p-nitrophenyl, μg/g, 1 h, 37°C). The measurement of soil urease activity involved incubating the samples with 20 mL of citric acid buffer (pH 6.7) and 10 mL of 10% urea solution at 37°C (24 h) to assess the release of NH₄ ⁺-N (Sun et al., 2014).

We conducted metabolite extractions following the methods documented in previous works (Chen et al., 2013; Chenkun et al., 2021). The various samples were ground into powder at 30 Hz using a mixer mill (MM400, Retsch) for 1.5 min after freeze-drying. We used 100 mg powder to conduct metabolite extraction. The extraction was performed at night at 4°C using 0.8 mL aqueous methanol (methanol: H₂O₂, 70:30, v/v) and pure methanol and subsequently centrifuged for 10 min at 10,000 g. We collected, homogenized, and filtered the supernatants (SCAA-104, 0.22 mm pore size; ANPEL Shanghai, China, www.anpel.com.cn/). Detailed information regarding further processing and annotation of the samples was documented in recent studies (Fallah et al., 2022).

The extraction of genomic DNA was performed by employing the Fast DNA™ Spin Kit (MP Biomedicals, LLC, Santa Ana, USA) using 0.5 g fresh soil as recommended by the manufacturer. DNA absorbance (A260 and 280 nm) was computed using BioTek Synergy H1 Hybrid Multi-Mode Microplate Reader (BioTek, USA) to investigate the quantity and quality of DNA. The bacterial 16S rRNA gene hypervariable V3–V4 region was amplified using 341F and 805R primers. PCRs were conducted in a 50-µL combination using 1 mM dNTPs (deoxynucleoside triphosphate), 1 × PCR buffer, 1 U of Platinum Taq, DNA template (10 ng), and each primer at 5 µM. PCR amplifications with an initial denaturation at 94°C (3 min), denaturation (five cycles at 94°C, 30 s), annealing at 45°C (20 s), and extension at 65°C (30 s) were performed. We subsequently carried out denaturation (20 cycles at 94°C) (20 s), annealing at 55°C (20 s), extension at 72°C (30 s), and the last extension at 72°C (5 min). Finally, an Illumina HiSeq 2500 platform (2 × 250 paired ends) at Biomarker Technologies Corporation, Beijing, China, was employed to conduct high-throughput sequencing. The raw data were later deposited on the NCBI Sequence Read Archive platform (accession no. PRJNA929962).

FLASH was adopted to merge the paired-end reads of the original DNA fragments (Tan et al., 2017). The paired-end reads appropriated to each sample were aligned to a sample-specific barcode. Sequence clusterings were conducted with operational taxonomic units (OTUs) based on 97% similarity. Thereafter, taxonomic information was annotated for each sequence using the Ribosomal Database Project (RDP) (Wang et al., 2007). Sequences that contained low quality and did not correspond with the barcode or primer with a high average quality score (Q ≥ 20) or without ambiguous base pairs were excluded. We then clustered all the sequences at 97% nucleotide similarity. Taxonomic classification was carried out using the SILVA database (SILVA Release 138, Bacteria), and biomarker biocloud platform (www.biocloud.net) was used to conduct the bioinformatics analysis.

We employed Bioconductor (http://www.bioconductor.org/) package “Mfuzz” and R software (http://www.r-project.org/) to investigate the expression patterns of metabolites based on fuzzy c-means. We adjusted the fuzzification parameter to m = 2 and the number of clusters to c = 12 to retain the soft clustering of the entire metabolite. Ternary plot analysis was conducted using R language-based packages, namely, grid and ggtern, an extension of the package ggplot2 to identify the upregulated and downregulated metabolites in the different compartments. Interactive networks of plant–soil systems and metabolites were conducted to establish the association between metabolites in the different compartments and under both treatments (Toju et al., 2016). The correlations between metabolites and plant growth-promoting bacteria were assessed using a correlation matrix. The potential pairwise Spearman’s ranks were determined and displayed using Cytoscape (version 3.6.1). ANOVA was used to evaluate the test data and visualized with GraphPad Prism (version 10.0.0). Tukey’s HSD test (p < 0.05) was used to compare the difference between mean values. The Biomarker biocloud platform (www.biocloud.net) was used to generate the rest of the used in this work.

Sugarcane traits responded strongly to the BC-supplemented soil. For instance, the crop brix, sucrose content, and chlorophyll content significantly increased (p < 0.05) under the BC-supplemented soil compared with the CK treatment ( Figures 1A–C ). Additionally, Figure 1D showed that the crop weight was promoted under the BC-supplemented soil, but no significant difference was observed between both treatments.

Meanwhile, we also noticed that soil chemical properties, including phosphatase, β-glucosidase, and cellulase, peaked considerably (p < 0.05) under the BC-supplemented soil than the CK treatment ( Figures 1E–G ). In addition, soil urease activity under the BC-supplemented soil outperformed those in the CK treatment but revealed no significant difference ( Figure 1H ). Figures 1I–K demonstrates that soil NH₄ ⁺-N, followed by TN and NO₃ ⁻N, were significantly increased (p < 0.05) under the BC-supplemented soil relative to the CK treatment. In addition, soil pH exhibited a 55.01% increase under the BC-supplemented soil compared with a 44.99% increase under the CK treatment ( Figure 1L ).

We conducted principal coordinate analysis (PCoA) to assess metabolite composition in the various compartments under both treatments. The metabolite composition distribution trend was largely compartment-driven ( Figure 2A ). The relative abundance of metabolite taxa in the different compartments, including prenol lipids (31.42%), fatty acyls (21.12%), organooxygen compounds (10.78%), steroids and steroid derivatives (8.73%), and benzene and substituted derivatives (3.53%), was dominant. Furthermore, a number of taxa, including carboxylic acids and derivatives (1.61%), flavonoids (1.89%), isoflavonoids (0.83%), phenols (0.52%), pyridines and derivatives (0.14%), and other (18.44%) were more prevalent ( Figure 2B ). Figure 2C reveals that these aforementioned taxa were more pronounced in the rhizosphere soil of the BC-supplemented soil, followed by the rhizosphere soil of the KC treatment.

Figure S1 further indicates that the distribution pattern of metabolite composition was primarily compartment-driven, suggesting that metabolite composition is more sensitive to different plant compartments rather than fertilization.

Mfuzz package was later used to explore the expression pattern of metabolite composition in each compartment under the BC-supplemented soil ( Figure 3A ) and the CK treatment ( Figure 3B ). The results showed that a number of essential metabolites peaked in the different clusters. In the BC-supplemented soil, jasmonic acid was among many metabolites that peaked in the leaf tissue in cluster 1. Moreover, benzaldehyde and benzene were some of the metabolites demonstrating a similar trend in cluster 2 in the root and leaf tissues. Similarly, phenylacetic acid and fluridone in cluster 3 peaked considerably in the stem tissue, whereas tyramine, homoveratric acid, and sebacic acid were considerably high in the stem and root tissues of clusters 4 and 5, respectively. In addition, mesaconate, biocytin, and abscisic acid were some of the dominant metabolites that peaked in the leaf tissue in clusters 6 and 7, respectively. Abscisic aldehyde, acetoin, and traumatic acid were among the dominant metabolites detected in rhizosphere soil in clusters 8, 9, and 10, respectively. Apigenin exhibited a similar pattern in the root and leaf tissues in cluster 11, whereas paromomycin increased significantly in cluster 12, especially in the root tissue ( Figure 3A ; Table S1 ).

In the CK treatment, apigenin and jasmonic acid were among the metabolites that peaked in both rhizosphere soil and leaf tissue in cluster 1. Ononin and pogostone revealed a similar trend in the stem and rhizosphere soil in clusters 2 and 3, respectively. Tyramine and daidzein were some of the metabolites that were considerably expressed in the root tissue in clusters 4 and 6, whereas pisatin, succinic acid semialdehyde, and retinal exhibited a similar pattern in the leaf tissue in clusters 5, 8, and 9, respectively. Likewise, coumestrol was considerably expressed in the rhizosphere soil in cluster 7, whereas sucrose and betanin in clusters 10 and 12 peaked in stem and root tissues, respectively. Abscisic acid demonstrated a similar trend in both root and leaf tissues ( Figure 3B ; Table S2 ).

Ternary plot analysis was performed to identify the specific metabolites that were upregulated or downregulated in the different compartments in each treatment ( Figures 4A, B ; Tables S3, 4 ). The analysis revealed that a number of key metabolites were significantly upregulated (p < 0.05) in the different compartments of the BC-supplemented soil, including sucrose and melatonin in the stem and root tissues, respectively. Similarly, phenyllactic acid was considerably upregulated (p < 0.05) in the stem and root tissues whereas zeatin exhibited a similar trend in the leaf, stem, and root tissues ( Figure 4A ; Table S3 ). On the other hand, biotin sulfone revealed the opposite in the leaf tissue ( Figure 4B ; Table S4 ).

In the CK treatment, bacampicillin was significantly upregulated (p < 0.05) in the stem tissue, whereas apiin revealed a similar pattern in the leaf, stem, and root tissues ( Figure 4C ; Table S5 ). However, some essential metabolites, namely, methyl jasmonate and salicylic acid, were downregulated in the leaf, stem, and root tissues of the CK treatment. Abscisic aldehyde and sucrose demonstrated the same trend in leaf and root tissues ( Figure 4D ; Table S6 ).

The differential abundance of metabolites showed that steroid biosynthesis, anthocyanin biosynthesis, and monoterpenoid biosynthesis performed better in the BC rhizosphere soil compared with the CK rhizosphere soil ( Figure S2A ). Similarly, the BC rhizosphere soil promoted sphingolipid metabolism, betalain biosynthesis, and glycerophospholipid metabolism compared with the CK rhizosphere soil ( Figure S2B ). Moreover, benzoxazinoid biosynthesis was significantly higher in the leaf tissue of the BC amendment compared with the leaf tissue of the CK treatment ( Figure S2C ). At the same time, sesquiterpenoid and triterpenoid biosynthesis and cutin suberin and wax biosynthesis marked a significant increase in the stem tissue of the BC amendment compared with the stem tissue of the CK treatment ( Figure S2D ).

The enriched KEGG pathway revealed that carotenoid biosynthesis, biosynthesis of unsaturated fatty acids, and alpha-linolenic acid metabolism were enriched considerably in the BC rhizosphere soil compared with the CK rhizosphere soil, followed by arginine biosynthesis, limonene and pinene degradation, and histidine metabolism ( Figure S2E ). In the root tissue of the BC amendment, a number of enriched KEGG pathways, including histidine metabolism; cysteine and methionine metabolism; glycine, serine, and threonine metabolism; betalain biosynthesis; glycerophospholipid metabolism; pantothenate and CoA biosynthesis; aminoacyl-tRNA biosynthesis; and beta-alanine metabolism outperformed those detected in the root tissue of CK treatment ( Figure S2F ). Additionally, the stem tissue of the BC amendment exhibited the advantage of significantly enriching biosynthesis of unsaturated fatty acids, phenylpropanoid biosynthesis, tyrosine metabolism, and carotenoid biosynthesis relative to those detected in the stem tissue of the CK treatment ( Figure 5A ). The leaf tissue of the BC amendment had a similar effect on biosynthesis of unsaturated fatty acids, phenylpropanoid biosynthesis, tyrosine metabolism, and carotenoid biosynthesis relative to those detected in the leaf tissue of the CK treatment ( Figure 5B ).

The pathways of the annotated metabolites of the Human Metabolome Database (HMDB) showed that lipids and lipid-like molecules (e.g., fatty acyls and prenol lipids), organic acids and derivatives (e.g., carboxylic acids and derivatives), and organic oxygen compounds (e.g., organooxygen compounds) were enriched considerably ( Figure 5C ). In addition, LIPID MAPS of metabolite annotation demonstrated that fatty acyls (e.g., fatty acids and conjugates, eicosanoids, and fatty esters), polyketides (e.g., flavonoids), and prenol lipids (e.g., isoprenoids) were considerably high ( Figure 5D ), whereas the KEGG pathway of metabolite annotation revealed the opposite trend ( Figure 5E ).

Meanwhile, we also identified a number of bacteria in different plant compartments. The various plant tissues were dominantly occupied by Proteobacteria, Firmicutes, Bacteroidota, Myxococcota, Actinobacteriota, Acidobacteriota, Gemmatimonadota, Nitrospirota, and Patescibacteria ( Figure S3A ). Figure S3B reveals that the community composition of these bacteria was region-specific. For example, bacteria identified in the leaf and stem tissues (aboveground compartments) were separated from those in the root tissue and rhizosphere soil (belowground compartments). Later, we employed BugBase functional analysis to confirm the suppressive effect of BC on pathogenic bacteria in the various compartments. We observed that pathogenic bacteria belonging to Proteobacteria were suppressed under the BC-supplemented soil relative to those in the CK treatment, especially in the rhizosphere soil and root tissue. Moreover, pathogenic bacteria belonging to Acidobacteria and Nitrospirae in the rhizosphere soil of the BC-supplemented soil were suppressed relative to those detected in the CK rhizosphere soil. Disease-causing bacteria belonging to Bacteroidetes exhibited similar behavior in the leaf tissue of the plant, whereas those belonging to Firmicutes were also identified in the leaf and stem tissues of the CK plants, implying that the BC-supplemented soil was effective in eradicating pathogenic bacteria belonging to Firmicutes ( Figure S3C ).

Metabolite associations with a number of classified plant growth-promoting bacteria were performed, and the result indicated that these targeted bacteria exhibited distinct relationships with a number of vital plant metabolites belonging to different taxa. Noticeably, Bacillus had a strong positive correlation with melatonin, whereas Bradyrhizobium, Mesorhizobium, and Serratia were strongly and positively correlated with abscisic aldehyde, whereas Flavobacterium displayed a significant positive association with fexofenadine in the aboveground compartments (stem and leaf tissues) ( Figure 6A ; Table S7 ).

In the belowground compartments (root tissue and rhizosphere soil), a significant number of metabolites exhibited strong positive associations with plant growth-promoting bacteria, especially Bradyrhizobium. For instance, some important metabolites, including melatonin and abscisic acid, responded strongly and positively to Bradyrhizobium, whereas Enterobacter showed a significant positive relationship with phenyllactic acid and Serratia had a similar pattern with zeatin. Likewise, Enterobacter and Frankia were significantly and positively correlated with phenyllactic acid and melatonin, respectively ( Figure 6B ; Table S8 ).

We also evaluated metabolite association with soil chemical properties. The results showed that some key metabolites exhibited a significant positive association with soil chemical properties. For example, sucrose, apigenin, and melatonin tended to favor soil TN. Moreover, phenyllactic acid and zeatin exhibited a significant positive association with soil β-glucosidase. Abscisic acid and jasmonic acid were significantly and positively correlated with soil phosphatase. Zeatin favored soil NH₄ ⁺-N and β-glucosidase, whereas apigenin and melatonin responded strongly and positively to soil TN ( Table 2 ).

Meanwhile, we also established the association between metabolites identified in the various plant tissues with the crop traits ( Tables 3 – 6 ). We observed that a number of plant growth and stress-regulating metabolites detected in the leaf tissue demonstrated a significant positive correlation with various crop traits. For example, metabolites, including melatonin, abscisate, and sucrose, showed a significant positive association with the crop brix. Likewise, abscisic acid and melatonin exhibited the same pattern with the crop height and chlorophyll content, respectively ( Table 3 ). Metabolites identified in the stem tissues, including abscisic aldehyde, biotin sulfone, and jasmonic acid, showed a significant positive correlation with the crop brix. Additionally, abscisic acid and jasmonic acid were significantly and positively associated with the crop weight whereas biotin sulfone revealed a similar pattern with the crop chlorophyll content ( Table 4 ). In the root tissue, metabolites such as abscisate and biotin sulfone revealed a strong and positive association with the crop brix. In addition, abscisate, apigenin, and biotin sulfone demonstrated the same trend with the crop sucrose content ( Table 5 ). However, metabolites identified in the rhizosphere soil of the crop revealed no significant association with the traits ( Table 6 ).