The experimental site was located in Wuding County, Chuxiong Yi Autonomous Prefecture, Yunnan Province (101°55′-102°29′E, 25°20′-26°11′N). The proven minerals in terms of geological structure consist of metals like iron, titanium, copper, lead, and zinc, along with non-metallic minerals like phosphorus and gypsum. During April 2024, we systematically collected three crops that are widely planted in the abandoned iron ore area, Musa basjoo Siebold, Amygdalus persica, and Triticum aestivum. The three crops in the mining area were named Mbs, Ape, and Tae. In a non-mining area adjacent to the mining area with similar environmental conditions, soil and root tissues of the same three crops were collected as a control group. The three crops in the non-mining area were named Mbs-CK, Ape-CK, and Tae-CK, respectively. Adhering to the strict 5-point sampling method (Data collection was carried out by selecting five equally spaced sample points within a given area.), three independent sampling points were placed in each designated area to ensure the samples’ representativeness. While conducting the sampling, the soil layer within 0 to 5 cm depth was extracted to obtain rhizosphere soil and plant roots (The root systems of most plants are usually in the 0–5 cm depth range, and these root systems are the primary interface for plant–soil microbial interactions.). Following the removal of loose soil around the roots, the soil remaining on the roots was collected using a sterile brush and identified as rhizosphere soil. Roots, leaf litter and stones have been removed from inter-root soil samples where they could affect the results of the analyzes (Zhao et al., 2024). Afterwards, the soil was filtered through a 2 mm sieve and air-dried for a series of physicochemical property analyzes to comprehensively understand the soil quality. The samples were stored at −20°C.

The pH of the soil was measured precisely using glass electrode technology (NY/T 1121.2-2006). The organic carbon content in the soil was quantitatively analyzed through the oxidation of potassium dichromate, combined with external heating (Shen et al., 2024). Soil total nitrogen was determined through the pretreatment step of combined digestion with sulfuric acid and an accelerator, followed by the Kjeldahl method (LY/T 1228-2015). The determination of soil total phosphorus was performed through NaOH alkali fusion treatment, followed by quantitative analysis using molybdenum-antimony spectrophotometry. The total potassium content was determined through NaOH melt flame spectrometry (Wang et al., 2024). The measurement of soil alkaline nitrogen was based on the principle of alkaline hydrolysis diffusion (LY/T 1229-1999). Soil effective phosphorus is determined by combined leaching of ammonium fluoride-hydrochloric acid solution and sodium bicarbonate solution, followed by molybdenum antimony antimony colourimetric method to complete the determination of its content (NY/T 1121.7-2014). The multi-element contents of the soil were determined via four-acid digestion, inductively coupled plasma-mass spectrometry (DZT 0279.2-2016). The measurement of each soil sample was conducted at least three times, and consequently, the mean value was derived.

In order to obtain pure samples of plant root microorganisms, the surface soil was first washed with sterile distilled water, soaked in 70% ethanol, and then washed several times with sterile distilled water until the washings were free of colony growth after incubation on culture medium. Genomic DNA was extracted using the SDS method. The bacterial V5–V7 highly variable region of the 16S rRNA gene was amplified using genomic DNA diluted to 1 ng/μl with sterile water as a template, with primers 799F (5′-AACMGGATTAGATACCCKG-3′) and 1193R (5′-ACGTCATCCCCACCTTCC-3′). The library was constructed using NEBNext^® Ultra^™ IIDNA Library Prep Kit, the constructed library was subjected to Qubit and Q-PCR quantification, and after the library was qualified, NovaSeq6000 was used for on-line sequencing.

The sample data were truncated to remove the Barcode and primer sequences and then the reads of each sample were spliced using FLASH to obtain Raw Tags (Magoč and Salzberg, 2011). The spliced Raw Tags were strictly filtered using fastp software to obtain Clean Tags (Bokulich et al., 2013; Deng et al., 2024). The above processed Tags were compared with the species annotation database and the chimeric sequences were removed to obtain the final Effective Tags (Edgar et al., 2011). For the Effective Tags, the DADA2 module of QIIME2 (Version QIIME2-202006) software was used to reduce noise and filter out the sequences with abundance less than 5 to obtain the final ASVs. The ASVs obtained were then compared to a database using the classify-sklearn module in the QIIME2 software in order to identify the species corresponding to each ASV (Wang et al., 2021). Tables of species abundance at different taxonomic levels were obtained based on the results of the ASVs annotation, and statistical tests of differential abundance were performed (Gloor et al., 2017).

To analyze the microbial community diversity within the samples, alpha diversity indices (observed_otus, shannon, simpson, chao1, goods_coverage, and pielou_e) were calculated for the different samples using QIIME2 software and visualized using R v2.15.3. For the comparison of microbial community composition among different samples, Beta Diversity Index intergroup difference analysis was used. This method detects differences between different samples (groups) using Principal Coordinates Analysis (PCoA). For the functional prediction of different microorganisms, metagenomic functional prediction was carried out in the COG, KO, and pathway databases based on marker genes (16S rRNA) using PICRUSt2 software. Correlation heatmaps were used to show the correlations between inter-root soil physicochemical properties and alpha diversity indices, CCA and RDA was used for correlation analysis between soil physicochemical properties and root endophytic bacterial microbiota. The RDA analysis was performed by the genescloud tools.

Excel 2019 was used to calculate the diversity indices, such as each soil physicochemical index. Analysis of variance (ANOVA) was performed using SPSS 21.0 statistical software. Correlation analysis of significantly different bacteria based on 16S rRNA analysis was performed according to the Pearson correlation coefficient. The data satisfies the normal distribution based on the Anderson-Darling test and meets the homogeneity of variances test when analyzed using SPSS. The significance of differences between samples was analyzed: a t-test was used for two groups of samples, and Tukey’s test was used for more than two groups of samples. By employing Duncan’s multiple range test, distinctions among different means were established at a significance level of p < 0.05.

The values of the soil pH, OC content and trace element contents of Fe, Zn, Cu, Ti and Pb in the mining area test group and non-mining control group are shown in Table 1. The pH values of the Ape and Tae experimental groups both decreased significantly (p < 0.05). This change was not significant between the acidic soil control Mbs-CK and the Mbs experimental groups. The pH value of the Mbs group was lower than that in all the experiments, and it was the lowest in the group. Concerning the OC content, each of the experimental groups demonstrated a progressive trend when compared to their corresponding control groups, with the Mbs experimental group showing the most substantial increase of 212%. The mining of iron ore has resulted in a generally higher Fe content in the soil of the experimental group compared to the control group, while the Ti levels were lower in the experimental group compared to the control group. Additionally, the Ape and Tae experimental groups exhibited significantly lower levels of Zn and Cu compared to the control groups (p < 0.05). On the other hand, the levels of Pb in the Mbs and Ape experimental groups were observed to be elevated compared to the control group.

In different groups, we also measured the total nitrogen (TN), alkaline nitrogen (AN), total kilocalories (TK), available kilocalories (AK), total phosphorus (TP), and available phosphorus (AP) of the soil in different environments, as well as other indicators of soil physicochemical properties (Figure 1). The soil physicochemical properties (p < 0.05) showed significant differences between the experimental group and the control group across the three plants. The TN and AN contents in the mining area under the Mbs system were significantly higher than those in the non-mining control group in terms of nitrogen. Despite this, the nitrogen content in the mining soil where the Ape and Tae crops were planted was considerably lower than in the control group, particularly in the case of the Ape.

The phosphorus levels showed contrasting patterns of change. While the TP content in the soil of the mining area under Mbs was high, its AP content was significantly lower than that of the control group. In contrast, the AP content in the soil of the mining area under Ape and Tae was notably higher than that in the control group, especially under Tae. The AP content increased by 289%, particularly under Tae. Notably, the TP content in the Ape soils significantly decreased to one-fourth of that in the control group. The soil in the experimental groups across all mining areas showed a significantly higher TK content compared to the control group, indicating a richness of potassium in the mining area or an increase in potassium input due to mining activities. However, for AK, the soil content in the mining area under Mbs and Tae outperformed the control group, whereas Ape fell short compared to the control group.

In general, the mining environment makes the soil nutrient content complex and variable. Different crops have different growth characteristics under mining area conditions. Crops maintain a relatively stable internal environment by increasing or decreasing the uptake of elements. The pH value of the soil in the mining area was generally lower and weakly acidic. The contents of OC, Fe, and TK were higher than those in the non-mining area, while the Ti content was lower. The contents of other elements varied due to the adaptability of different plants. Mbs had an advantage in promoting the accumulation of nitrogen and potassium. Ape faced the challenge of decreasing nitrogen, phosphorus, and available potassium contents. Tae significantly activated phosphorus, giving it an advantage.

The ASV data were obtained with noise reduction. The results of the Venn diagram (Supplementary Figure S1) revealed that 3,316 endophytic bacterial ASVs were detected in the roots of the three crops in the mining area test group. Of these, the Mbs, Ape, and Tae values were 1765, 1,011, and 1,120, respectively, and 11 species were common to all three crops. In contrast, 1,332 endophytic bacteria, namely ASV, Mbs-CK, Ape-CK, and Tae-CK, were detected in the roots of the three crops in the control group in the non-mining area (365, 505, and 584), respectively. This number was lower than that in the experimental group, but 101 endophytic bacteria were shared by the three species. There were 66 types of Meb and Mbs-CK, 41 types of Ape and Ape-CK, and 94 types of Tae and Tae-CK. To further assess the richness of bacterial communities under different crops and treatments, we used the sparse curve of OTUs (Supplementary Figure S2) as an analysis tool. Sparse curve analysis revealed that all of the soil samples had sufficient sequencing depth. When the sequencing read length exceeded 6,000 reads, the dilution curve gradually stabilized, the amount of sequencing data progressed reasonably, and more data volume did not significantly affect the alpha diversity index, ensuring the accuracy and reliability of microbial diversity analysis. Therefore, the abundance of sequences can be used to infer the diversity of bacteria in the samples.

The results showed that in all root samples, 23 phyla, 55 classes, 131 orders, 231 families, and 427 genera of endophytic bacteria were identified. We compared the changes in the top 10 most abundant phyla and genera (Figure 2). The main endophytic bacterial taxa at the phylum level were Proteobacteria, Actinobacteria, Firmicutes, Acidobacteriota, and Bacteroidota, which accounted for more than 90% of the total bacterial abundance. Proteobacteria was the most abundant, accounting for 55.88% of all bacteria. All experimental groups exhibited a significantly lower abundance of Proteobacteria and a significantly higher abundance of Actinobacteria when compared to the control group (p < 0.05). At the genus level, Stenotrophomonas was the main bacterial group, accounting for 35.2% of all bacteria. In comparison to the control group, the Mbs and Tae groups showed significantly lower abundance of Stenotrophomonas. Moreover, the Ape group exhibited significantly lower abundance of Streptomyces, while the Mbs group had significantly lower abundance of Bradyrhizobium. The abundance of Nocardia in the Ape group increased by 59.69%, and the abundance of Promicromonospora in the Tae group increased by 60.72%.

We calculated the α-diversity indices of the three crops in the experimental group and the non-mining control group (Figure 3). Compared with the control group in the non-mining area, all the experimental groups showed increased Chao1 and Shannon indices (Supplementary Figure S3). Compared with Mbs-CK, Mbs also significantly increased the Simpson, Pielou_e, and observed OTUs indices (p < 0.05), indicating that the community diversity was greater and the species distribution was more even. The Simpson diversity of Tae significantly increased by 71.3% (p < 0.05) compared to Tae-CK, while the Simpson diversity of the Ape test group did not significantly change.

In order to delve deeper into the beta diversity among the flora in various regions and distinguish the variations in microbial community structure, we conducted PCoA in conjunction with OTU abundance using the weighted UniFrac distance (Supplementary Figure S4). The PC1 and PC2 axes explained 55.36 and 25.45% of the results, respectively. The distance separating Mbs-CK and Tae-CK was relatively near, unlike the other samples, while the distance between Ape and Ape-CK was also close, indicating a greater overlap in bacterial communities.

For our experiments, we utilized PICRUSt2 to forecast bacterial functions. The top 35 functions were selected based on their abundance and functional annotation information in the COG, KO, and pathway databases. These functions were then depicted in a heat map along with their abundance information for each sample. They were then clustered at different functional levels (Figure 4). According to the COG database, the experimental group promoted the functions of flavin-dependent oxidoreductase (COG2141) and DNA-binding transcriptional regulator (COG1846). Compared to the control group, the Mbs experimental group showed increased functions of nucleoside-diphosphate-sugar epimerase (COG0451) and NAD(P)-dependent dehydrogenase (COG1028), and decreased functions of DNA-binding transcriptional regulator (COG0789). In comparison to Ape-CK, Ape reduced the function of signal transduction histidine kinase (COG0642). In contrast, the Tae group showed increased functions of signal transduction histidine kinase ComP (COG4585), pyridoxal reductase PdxI or related oxidoreductase (COG0667), ADP-ribose pyrophosphatase YjhB (COG1051), and DNA-binding transcriptional regulator, MerR family (COG0789) genes when compared to the control group.

In the KO database, the Myb_DNA-binding (K00249) function of Mbs was upregulated compared to that of the control. Compared to the control, Ape increased the ability of light to harvest chlorophyll a/b-binding protein (K02067). Tae increased the expression levels of raffinose transport system permease protein (K10119) and beta-galactosidase (K12308) compared to those in the control group.

In the pathway database, the function of pyruvate fermentation to isobutanol (PWY-7111) was upregulated in the experimental group. Compared to the control group, the function of L-isoleucine biosynthesis I (ILEUSYN-PWY) in Mbs was upregulated, while the function of peptidoglycan maturation (PWY0-1586) was downregulated. Additionally, the function of toluene degradation II (PWY-5182) was upregulated, while the function of urate biosynthesis (PWY-5695) was downregulated. Furthermore, Tae decreased the function of fatty acid β-oxidation I (FAO-PWY) compared to the control group.

To investigate the associations between the soil physicochemical parameters and the α diversity indices of the root endophytic bacteria, Pearson correlation analysis (PC, p < 0.05) was conducted (Figure 5). Results indicated a significant positive correlation between the soil TK content, as well as the Fe and Pb metal contents, and the α diversity index. Conversely, the α diversity index exhibited a significant negative correlation with the Zn and Ti contents. Additionally, the Shannon and Simpson indices were negatively correlated with the Cu content (p < 0.05).

In order to better understand how environmental factors impact the microflora of root soil and endophytic bacteria in root systems, we conducted redundancy analysis. Our focus was on the top 10 most prevalent microbial species at the genus level and how they relate to soil environmental factors (Figure 6). RDA 1 and RDA 2 explained 62.23 and 20.31%, respectively, of the total variation, revealing the complex connection between environmental factors and the composition of microflora. Among them, Promicromonospora was significantly positively correlated with the soil AK content, Nocardia was significantly positively correlated with the soil AN and OC contents, and Streptomyces was significantly positively correlated with the soil TK, TP, and AP contents.