Structure and variation of root-associated bacterial communities of Cyperus rotundus L. in the contaminated soils around Pb/Zn mine sites

Soil contamination due to mining activities is a great concern in China. Although the effects of mining pollution resulting in changes of soil characteristics and the microbiome have been documented, studies on the responses of plant root-associated microbial assemblages remain scarce. In this work, we collected bulk soil, rhizosphere soil, and root endosphere samples of Cyperus rotundus L (Cyp) plants from two Pb/Zn mines, of which, one was abandoned (SL) and the other was active (GD), to investigate the bacterial community responses across different site contamination levels and Cyp plant compartments. For comparison, one unpolluted site (SD) was included. Results revealed that soils from the SL and GD sites were seriously contaminated by metal(loid)s, including Pb, Zn, As, and Sb. Bacterial richness and diversity depended on the sampling site and plant compartment. All sample types from the SL site had the lowest bacterial diversities and their bacterial communities also exhibited distinct patterns compared to GD and SD samples. As for the specific sampling site, bacterial communities from the root endosphere exhibited different patterns from those in bulk and rhizosphere soil. Compared to the GD and SD sites, the root endosphere and the rhizosphere soil from the SL site shared core microbes, including Halomonas, Pelagibacterium, and Chelativorans, suggesting that they play key roles in Cyp plant survival in such harsh environments.


Introduction
Soil contamination due to mining activities is a cause of great concern worldwide, especially the mining of heavy metals, which diminish the availability of arable lands (Yang et al. 2018). In China, this problem in abandoned mining sites is unprecedented Zhao et al. 2016). Long-term exposure to heavy metal contamination may pose serious threats to ecological safety and human health (Gao et al. 2015;Wang et al. 2017a, b;Zhong et al. 2020). Previous studies reported distinct patterns of soil characteristics and microbiome in such harsh environments, which negatively affect soil ecosystems Yun et al. 2018).
Soil microbes comprise the largest group of biodiversity in terrestrial ecosystems and play important roles in plant growth (Voges et al. 2019). Plant-microbe interactions allow microbes to adapt to various soil conditions, enabling them to resist or tolerate chemical toxicity and pathogen disease in contaminated environments (Berendsen et al. 2012;Mendes et al. 2011). However, long-term selective pressure may change soil or plant microbiomes, resulting in great differences in microbial structures and compositions between the root endosphere and external environments, including the rhizosphere and bulk soils (Gottel et al. 2011;Sun et al. 2020). The rhizosphere is a major hotspot for root-microbe interactions and consists of nutrient availability, growth promotion, and microbial enrichment (Henneron et al. 2020;Liu et al. 2020). Various factors, such as geographical site, plant species, developmental stage, soil type, and land management, affect the colonization pattern of rhizosphere microbes (Beckers et al. 2017;Edwards et al. 2015). Several studies have found that the root microbiome might be quite distinct from the external microbes in the soil (Gkarmiri et al. 2017;Hartman and Tringe 2019). Changes in the soil microbial community due to external disturbances (i.e., mining activities) greatly influence the root-associated microbiome. Distinct bacterial and fungal communities in the rhizosphere and root endosphere of Populus deltoids have been observed (Gottel et al. 2011). Peiffer et al. (2013) reported that geographical location had a significant effect on variations in the maize rhizosphere microbiome. Our previous study found that rhizosphere bacterial communities in the adjacent areas of Sb and Pb/Zn mines were crop-specific and strongly linked to concentrations of Cr and V (Sun et al. 2018a). Similarly, plant species and mine tailing pH were found to drive root-associated microbial community patterns in boreal trees and shrubs (Gagnon et al. 2020). A recent study by Li et al. (2020) found that plant growth changed the recruitment of soil microbes colonizing rhizosphere compartments in response to acid mine drainage pollution. However, the relative contributions of the soil characteristics, contamination due to mining activities, and variations in root-associated bacterial assemblages have not been well documented.
Cyperus rotundus L. (Cyp) is a popular perennial species distributing worldwide in tropical and subtropical regions. Cyp plants can grow well in soils contaminated by mining activities, suggesting that they are able to adapt to harsh environments. Previous studies reported that Cyp plants were suitable for use in diesel phytoremediation and heavy metalcontaminated soils (Ashraf et al. 2012;Bordoloi and Basumatary 2016;Hou et al. 2016;Lum and Chikoye 2018;Subhashini and Swamy 2014). It has been confirmed that Cyp plants have significant bioaccumulation factors for various heavy metals, including Zn, Pb, As, Cd, Cu, Cr, and Ni (Ashraf et al. 2012;Lum et al. 2014). Thus, Cyp plants play an important role in the phytoremediation of heavy metal contamination. However, the responses of microbial communities to pollutant elimination have not been investigated and the contribution of root-associated microbes remains unknown.
In this study, we collected bulk soil, rhizosphere soil, and root endosphere samples of Cyp plants from two Pb/Zn mines, of which, one was abandoned (SL) and the other was actively mined (GD). Additionally, an unpolluted site (SD) was selected as a control for comparison. Geochemical factors, including soil pH, content of soil organic carbon (SOC), nitrate, and sulfate, as well as concentrations of metal(loid)s, including Cr, Cu, Cd, Pb, Zn, As, and Sb, were analyzed. The bacterial communities of plant compartments across different sampling sites were characterized. The aims of this work are to (1) explore root-associated bacterial community responses to mining activities; (2) compare bacterial distribution patterns in bulk soil, rhizosphere soil, and root endosphere of Cyp plants; and (3) determine the core bacterial communities in Cyp plant compartments that are tolerant to heavy mining contamination.
A total of 45 bulk and rhizosphere soils, as well as the root endosphere samples with five replicates for each component were collected from the GD, SL, and SD sites. Bulk soil was taken adjacent to each Cyp plant and was kept on ice as soon as being transported back to laboratory for analysis. The Cyp roots were collected and the rhizosphere soil attached on the roots were washed with phosphate buffer saline (PBS) solution after the loose soils were shaken off. The remaining roots were cleaned with PBS solution for three times and were considered as endosphere samples. Bulk soil, rhizosphere soil, and root endosphere samples were stored at −20°C prior to DNA extraction.

Soil physicochemical parameters analysis
Soil samples were freeze-dried, thoroughly grounded, and passed through a 200-mesh sieve prior to physicochemical characterization. Parameters of pH, sulfate, and nitrate were measured according to the standard methods, which were narrated previously . SOC was directly determined with a vario MACRO cube elemental analyzer (Elementar, Hanau, Germany). For measuring the concentrations of metal(loid)s including Cr, Cu, Cd, Pb, Zn, As, and Sb and their acid-soluble fractions, the dried soil samples were completely digested with HNO 3 /HF at a volume ratio of 5:1 (Sun et al. 2018a) and 0.11 M acetic acid (Zhai et al. 2018), respectively. The resulting digestion solutions were then determined by an Agilent 7700x inductively coupled plasma mass spectrometer (ICP-MS) (Agilent, Santa Clara, USA) equipped with a concentric nebulizer and a baffled cyclonic spray chamber using high-purity argon (99.996%) as carrier gas.

DNA extraction and 16S rRNA amplicon sequencing analysis
Approximately 0.5 g each of bulk soil, rhizosphere soil, and root endosphere samples (frozen with liquid nitrogen and pestled) was individually prepared for extraction of the total bacterial DNA, using the Power Soil DNA Isolation kit (MoBio Laboratories, Inc., Carlsbad, CA, USA) according to the protocol provided by the manufacturer. The quantity of the extracted DNA was determined with the NanoDrop ND-2000 spectrophotometer (ThermoFisher Scientific, Waltham, MA, USA), and the quality of DNA was determined with 1% (w/v) agarose gel electrophoresis.
Based on our previous optimization experiments, the V4-V5 hypervariable region of the 16S rRNA gene was amplified using the universal primers 515F (5′-GTG YCA GCM GCC GCG GTA A-3′) and 907R (5′-CCY CAA TTC MTT TRA GTT T-3′). The obtained amplicons were barcoded uniquely in each sample and sent out for high-throughput sequencing on an Illumina MiSeq platform (Tianjin Novogene Bioinformatic Technology Co., Ltd, China). After sequencing, the raw sequence data were merged and filtered in accordance with the previously reported criteria ). The remaining high-quality sequences within 97% similarity threshold were clustered into an operational taxonomic unit (OTU) and annotated against the Greengenes Database following our established pipelines (DeSantis et al. 2006;Sun et al. 2019a, b).

Bioinformatics and statistical analysis
OTU counts were normalized in R package (v3.5.2) and OTU-level alpha diversity indices (i.e., observed species number, Chao1 richness estimator, abundance-based coverage estimator (ACE), Simpson index, and Shannon diversity index) were calculated (McMurdie and Holmes 2013; Sun et al. 2021). To compare the richness and evenness of OTUs among samples, the rarefaction curves were generated by using Mothur program. Principal coordinate analysis (PCoA) was performed using UniFrac distance metrics to investigate the structural variation of bacterial communities across samples and visualized by origin from either sampling sites or Cyp compartments (Li et al. 2021). Taxa relative abundances at different levels were statistically compared among samples and visualized as stack plots. Linear discriminant analysis effect size (LEfSe) analysis was performed with the default parameters to determine differentially abundant taxa across Cyp compartments (Segata et al. 2011). Statistical analysis (i.e., two-sided t-test) was performed using IBM SPSS (v19.0) to evaluate the differences across samples, and it was considered statistically significant when P values less than 0.05.

Soil characteristics
A total of 15 bulk soil samples (5 each from the GD, SL, and SD sites) were used for pH, SOC, nitrate, and sulfate measurements ( Fig. 1). SD bulk soil had a lower pH (6.40 ± 0.04) and sulfate content (9.15 ± 1.70 mg/kg), but higher SOC (18.89 ± 4.10 g/kg) and nitrate content (20.83 ± 4.17 mg/kg) when compared to the GD and SL bulk soil. There were no significant differences in these characteristics between the GD and SL bulk soils. Metal(loid)s exhibited a wide variation of concentrations in the bulk soil samples (Fig. 2). The contents of all metal(loid)s, except Cr in the SD bulk soil, were significantly (P < 0.05) lower than those of the GD and SL bulk soil. In particular, the concentrations of Pb, Zn, and As in GD (908.8 ± 554.4, 5230 ± 2746, and 146.3 ± 85.1 mg/kg, respectively) and SL bulk soil (1842 ± 939, 6096 ± 2744, and 335.2 ± 146.7 mg/kg, respectively) were one and two orders of magnitude higher compared to the SD bulk soil (19.9 ± 5.7, 75.4 ± 23.1, and 10.7 ± 1.5 mg/kg, respectively). Surprisingly, the content of Sb in the SL bulk soil was significantly higher Fig. 1 Boxplots of the physicochemical parameters of bulk soil collected from the active Pb/Zn mine (GD), the abandoned Pb/Zn mine (SL), and an unpolluted site (SD) (1857 ± 1337 mg/kg) than that of the GD (19.8 ± 11.7 mg/kg) (P = 0.04) and SD bulk soil (1.1 ± 0.2 mg/kg) (P = 0.04). In contrast to the total metal(loid) contents, the concentrations of their acetic acid-soluble fractions were much lower (Fig. S2).

Bacterial diversity
Rarefaction curves were constructed using Mothur with 97% sequence similarity for bacterial samples (Fig. S3), which evaluated the species richness of each sample that approached saturation. Good's coverage scores were highly comparable across samples, ranging from 98.4-100%, indicating an adequate sequencing depth for analyzing the bacterial microbiome. Results revealed that the bacterial communities were heavily dependent on the sampling site and Cyp compartment. All SL sample types had a significantly (P < 0.05) lower richness (i.e., Chao1, ACE, and observed species number) and diversity (i.e., Simpson and Shannon indices) than the GD and SL samples (Fig. 3).
In the SD samples, the rhizosphere soil had the highest richness and diversity indices, while the root endosphere had the lowest values. Moreover, the richness and diversity indices between the bulk and rhizosphere soils were comparable, while the rhizosphere soil and the root endosphere were significantly different (P < 0.05). In the GD samples, the bulk soil had the highest richness and diversity indices, and were comparable to the rhizosphere soil. The root endosphere had the lowest values, but was not significantly different from the rhizosphere soil. The SL samples had the highest richness indices of the rhizosphere soil and diversity indices of the root endosphere, while the lowest richness and diversity indices were observed in the bulk and rhizosphere soils, respectively. Significant (P = 0.045) differences were detected in the richness indices between the bulk and rhizosphere soils. The bacterial richness indices revealed a decreasing gradient from the rhizosphere soil to the root endosphere ( Fig. 3a-c), while the diversity indices showed a decreasing trend from the rhizosphere soil to the root endosphere in the GD and SD samples, but exhibiting an increasing trend in the SL samples (Fig.  3d, e).
Similarities in the bacterial communities across various samples are visualized with a PCoA analysis (Fig. 4). PCo1 and PCo2 explained 28.4% and 20.4% of the total variation, respectively. All samples exhibited distinct clustering based on the sampling site (Fig. 4a). Within each cluster, the bacterial communities of bulk and rhizosphere soils overlapped considerably (Fig. 4b), but no consistent pattern was observed between rhizosphere soil and root endosphere. These findings were consistent with the alpha diversity results of sitedependent bacterial communities, which indicated that there were no significant differences in the bacterial richness or diversity indices between the bulk and rhizosphere soil samples within each site.

Bacterial community composition
The main bacterial phyla in all samples are presented in Fig.  5a and Fig. 6. Proteobacteria was the most abundant phylum with a relative abundance of 35.9% in the GD bulk soil and 91.7% in the SL rhizosphere soil, followed by Acidobacteria (3.2% and 31.7%, respectively) and Actinobacteria (1.7% and 7.9%, respectively). Bacterial phyla exhibited site-dependent effects, as the relative abundance of Proteobacteria in the SL samples (42.8-91.7%) was much higher when compared to GD (35.9-47.6%) and SD (36.9-57.8%) samples. Specifically, Proteobacteria in the SL rhizosphere (91.7%) Fig. 2 Total concentrations (tot) of heavy metal(loid)s in the bulk soil collected from the active Pb/ Zn mine (GD), the abandoned Pb/ Zn mine (SL), and an unpolluted site (SD) and bulk soils (83.9%) were two times more abundant than in the GD (40.8% and 35.9%, respectively) and SD samples (39.4% and 36.9%, respectively). Similar effects were observed for Acidobacteria and Actinobacteria, which were much lower in the SL samples (3.2-5.3% and 1.7-3.3%, respectively) than in the GD (14.3-31.7% and 6.8-8.7%, respectively) and SD samples (9.6-29.1% and 5.9-7.9%, respectively).
There were no significant differences in the dominant phyla (i.e., Acidobacteria, Proteobacteria, and Actinobacteria) relative abundance patterns between the rhizosphere and bulk soil samples collected from the same site (Fig. 5a). Notably, overall distinct patterns in the relative abundances of the dominant phyla were observed between the rhizosphere soil and the root endosphere bacterial communities, including Proteobacteria and Acidobacteria. In the SD rhizosphere soil, a significant (P = 0.001) enrichment of Acidobacteria was observed, while Proteobacteria was significantly (P = 0.021) depleted when compared to the root endosphere. Proteobacteria were significantly (P = 0.002) enriched in Samples marked as Bul, Rhi, and End represent bulk soil, rhizosphere soil, and root endosphere, respectively Fig. 4 PCoA of the bacterial communities depicting the samples obtained from a the GD (red), SL (blue), and SD (green) sampling sites, and b the bulk soil (red), root endosphere (green), and rhizosphere soil (blue) sample types SL rhizosphere soil, while Acidobacteria were significantly (P = 0.02) depleted in SL rhizosphere soil when compared to their respective root endosphere compartments. Moreover, other bacterial phyla, including Deinococcus Thermus, and Tenericutes, exhibited higher relative abundances in SL than in GD and SD, especially phylum Tenericutes of the Mollicutes subclass, which had a higher relative abundance in the SL root endosphere (14.1%) than other samples.
At the genus level, the core bacterial microbiome was identified, including Gp6, Halomonas, Bradyrhizobium, Pelagibacterium, Gp16, Chelativorans, Gaiella, and Spartobacteria genera incertae sedis (Fig. 7). We detected site effects on these core bacterial communities. In the SD samples, core communities in the rhizosphere soil exhibited Fig. 6 LefSe program showing the differences in relative abundances of the dominant phyla in bulk soil, rhizosphere soil, and root endosphere Fig. 7 Core bacterial microbiome at the genus level across all samples significant (P < 0.05) differences when compared to root endosphere communities, of which, Halomonas, Bradyrhizobium, Pelagibacterium, and Chelativorans were enriched and Gp6, Gp16, Gaiella, and Spartobacteria genera incertae sedis were depleted. In the SL samples, significant (P < 0.05) enrichment of Halomonas, Pelagibacterium, and Gaiella were observed in the rhizosphere soil, while only Chelativorans was significantly (P = 0.05) enriched when compared to the root endosphere. In contrast, only Bradyrhizobium was significantly (P = 0.035) enriched and Gp16 was significantly (P = 0.006) depleted in the rhizosphere soil of GD samples. No significant differences in the relative abundances of the core bacterial communities were observed between the rhizosphere and bulk soils, regardless of the sampling site, except Bradyrhizobium in the SD samples (P = 0.01).

Discussion
Heavy metal pollution caused by mining activities are severe environmental threat in China. Bioremediation mediated by plants and microorganisms is an effective method for alleviating mining contamination. The roots connecting soils and plants are hotspots for interactions between microorganisms and the environments. Although microbial communities in mining-contaminated sites have been extensively characterized, less attention has been focused on the roots, especially the root endosphere. In this study, microbial communities in plant compartments, including the bulk soil, rhizosphere soil, and the root endosphere, were characterized and compared. Their interactions with geochemical conditions were also analyzed and discussed.

Bacterial diversity was influenced by soil characteristics and contamination level
Endosphere species richness was lower than the bulk and rhizosphere soil richness, indicating the non-uniform and selective colonization of plant roots (Beckers et al. 2017). It is well known that the root microbiome is primarily assembled from external soil microbes (Edwards et al. 2015), which may migrate to the rhizosphere due to the attraction of rich nutrients like root exudates (Zhalnina et al. 2018), resulting in the rhizodeposition of various microbes. However, rootcolonizing bacteria are highly competitive in order to ensure successful colonization. Therefore, root-colonizing bacteria possess traits that favor resistance to harsh environments, such as chemical toxicity and pathogen disease (Mendes et al. 2011;Berendsen et al. 2012). Thus, bacteria inside the root endosphere are much less rich and diverse than in the rhizosphere soil, as confirmed in this study, with the exception of the SL root endosphere samples (Fig. 3), which had the highest Shannon and Simpson diversity indices when compared to the bulk and rhizosphere soils (Fig. 3d, e). These findings may be attributed to the long-term impact pollution on Cyp roots in the surrounding SL tailing, which may affect Cyp development, shape root-associated microbes, and change acquisition patterns in the root endosphere (Edwards et al. 2015;Chaparro et al. 2014).
Based on the bacterial richness and diversity indices, we found that the alpha diversity was highly dependent on the sample site. The SL samples had the lowest richness and diversity, regardless of sample type, followed by the GD samples, while the SD samples possessed the highest alpha diversity (Fig. 3). These results implied that the contamination level of the Pb/Zn mines greatly affected and shaped plantassociated microbes. The concentrations of metal(loid)s, including Cu, Cd, Pb, Zn, As, and Sb, in the SL bulk soil were significantly higher than those in SD bulk soil, which exerted persistent selection pressure on the assemblage of soil microbes (Fig. 2b). Additionally, soil characteristics, including pH and the nitrate and sulfate contents in the SL bulk soil, were significantly higher compared to those of the SD bulk soil (Fig. 2a), which have been reported to closely correlate with soil bacterial communities (Gagnon et al. 2020;Xiao et al. 2016;Chen et al. 2013). In contrast, the GD site was being actively mined and its contamination of the surrounding soils was less severe. The concentrations of metal(loid)s in the GD bulk soil were lower compared to the SL bulk soil, but were still significantly higher than in the SD bulk soil (Fig.  2b). Moreover, a significantly higher pH and lower SOC were also observed (Fig. 2a). These factors likely modulate the population and composition of plant-associated microbes, resulting in the loss of bacterial richness and diversity. Additionally, the PCoA analysis showed site-dependent clustering of the bulk soil, rhizosphere soil, and root endosphere communities (Fig. 4a). This result is consistent with the alpha diversity results, confirming that contamination conditions greatly affected microbial assemblages. As for specific sites (i.e., the GD, SL, and SD), the majority of bulk and rhizosphere soil samples clustered together (Fig. 4b), suggesting that bacterial diversity was comparable between the bulk and rhizosphere soil samples. In contrast, the root endosphere samples were sporadically distributed, indicating distinct patterns of bacterial communities.

Structural distribution of the dominant bacterial communities
Bacterial structures and compositions were dependent on the sampling site and Cyp compartment. However, phylum Proteobacteria (mostly Alphaproteobacteria and Gammaproteobacteria) consistently dominated bulk soil, rhizosphere soil, and root endosphere samples, followed by Acidobacteria and Actinobacteria. These phylotypes have been commonly identified in the soil microbiome surrounding mining areas (Gao et al. 2019;Sun et al. 2018b). It was previously reported that the ratio of Proteobacteria and Acidobacteria could be an indicator of soil trophic levels (Castro et al. 2010;Smit et al. 2001;Gottel et al. 2011;Beckers et al. 2017), in which, Acidobacteria and Proteobacteria were associated with nutrient-rich and nutrient-poor soils, respectively. The relative abundance of Proteobacteria (mostly Alphaproteobacteria) increased from rhizosphere soil to the root endosphere, while Acidobacteria decreased in the GD and SD samples, indicating nutrient-rich conditions of the root endosphere compared to rhizosphere and bulk soils. Similar results were observed in other plants, including poplar (Beckers et al. 2017;Gottel et al. 2011;Shakya et al. 2013) and rice (Edwards et al. 2015). In the SL samples, however, contrasting results were observed, where the relative abundance of Proteobacteria (mostly Alphaproteobacteria and Gammaproteobacter) decreased remarkably from rhizosphere soil to the root endosphere, while Acidobacteria increased slightly, which was similar to previous reports on grasslands (Singh et al. 2007) and soybeans (Xu et al. 2009).
These findings suggested that heavy contamination (i.e., high contents of metal(loid)s) in the SL site likely changed the soil nutrient status and thereby shaped bacterial assemblages. Similar results were previously reported by Sun et al. (2018a) who found that plant rhizosphere communities strongly correlated with Cr and V in Pb/Zn mining sites. Moreover, Acidobacteria and Proteobacteria in the bulk and rhizosphere soil samples from the GD and SD sites were comparable, indicating an intermediate nutrient level. This result is consistent with the alpha diversity results (Fig. 3). As for the SL samples, Proteobacteria were enriched in the rhizosphere soil when compared to the bulk soil, which may be attributed to the nutrient-rich conditions of the rhizosphere soil. This result is likely due to the production of root exudates under the selection of metal(loid)s in polluted soils (Kozdrój and van Elsas 2000;Qin et al. 2007), as well as the increased pollution resistance and/or tolerant Proteobacteria, such as Alphaproteobacteria, in this study (Sandaa et al. 1999). In addition to metal(loid)s, soil pH was previously found to correlate with overall bacterial community composition and phylogenetic diversity, which affect bacterial relative abundances, such as Acidobacteria (Lauber et al. 2009;Qi et al. 2018). Unfortunately, the pH values of the rhizosphere soil and the root endospheres were not directly measured in this study due to a lack of samples. Rhizosphere pH is mediated by plant roots (i.e., exudates of organic acids) in response to environmental constraints (Hinsinger et al. 2003;Sasse et al. 2018). As a result, plant-associated microbial communities in contaminated mining sites are largely shaped by various factors, including plant genotype and soil conditions. Additionally, phylum Tenericutes was exclusively found in the SL root endosphere samples with a relative abundance of 14.1%, which was exclusively consisted of genus Acholeplasma. Previous studies found that Acholeplasma belongs to wall-less, saprophytic, and free-living bacteria (Zhao et al. 2015;Mitter et al. 2017), which is transmitted to plants and colonize in the vascular tissues as a sap-sucking endophyte (Blain et al. 2017). Phylum Deinococcus Thermus (mostly Meiothermus) was remarkably richer in SL samples than in GD and SD samples. This result can be ascribed to the heavy contamination conditions of SL sites, as Meiothermus species are always present in extreme environments (Thokchom et al. 2017). Consistently, the dominance of Meiothermus has been observed in mining tailings (Sun et al. 2018b). Thus, it was proposed that Meiothermus may have the potential to oxidize and reduce As.

Core members of the root bacterial microbiome
Within the core bacterial microbiome, similar site-specific distributions were observed. At the genus level, the GD and SD bulk and rhizosphere soil communities were dominated by Gp6, Spartobacteria genera incertae sedis, Bradyrhizobium, Gp16, and Gaiella, of which, Gp6 and Bradyrhizobium also dominated root endosphere assemblages. This finding suggested that the root endosphere communities consisted of a subset of specific microbes from the corresponding rhizosphere soil, which was consistent with the findings of previous reports (Hallmann et al. 1997;Miliute et al. 2015). Similar results were observed in SL rhizosphere soil and root endospheres, which were both primarily dominated by Halomonas, Pelagibacterium, and Chelativorans, and their rhizosphere relative abundances were almost one order of magnitude higher. Bradyrhizobium is a common endophyte with higher abundance in the root endosphere than in the bulk and rhizosphere soils, which has been reported to improve plant growth in the heavy metal-contaminated environment (Wani et al. 2007;Seneviratne et al. 2016). In contrast, overall distinct compositions of root core microbes between SD and SL samples, which could be ascribed to the selection pressures of the surrounding soil conditions (described above). Halomonas of class Gammaproteobacteria is capable of indole acetic acid production and phosphate solubilization in the presence of high concentrations of heavy metals and highsalinity, which thereby promotes plant growth (Desale et al. 2014). Additionally, Halomonas species isolated from mangrove Avicennia marina rhizosphere soil increased exopolysaccharide production under arsenic stress, thereby enhancing rice growth (Mukherjee et al. 2019). Pelagibacterium of class Alphaproteobacteria has been found in Tamarix ramosissima root-associated communities and is tolerant to salinity (Taniguchi et al. 2015). Some species have been found to grow well in high-salinity environments, such as seawater Xu et al. 2011;Wang et al. 2017a, b), lake water , and deserts (Yang and Sun 2 0 1 6 ) . S e v e r a l C h e l a t i v o r a n s s p e c i e s o f c l a s s Alphaproteobacteria have been isolated from EDTAenriched soils (Bohuslavek et al. 2001). These species grew with EDTA as carbon, nitrogen, and energy sources (Doronina et al. 2010). This microbe was relatively abundant in SL rhizosphere soil, which may be due to EDTA pollution from mining. Chelativorans species have also been found in seawater (Evans et al. 2018). A recent study reported that Chelativorans species were abundant in marine spongeassociated microbes in the Palk Bay of India, which has suffered from sewage wastewater, salt pan, and heavy metal pollution (Meenatchi et al. 2020). In this study, it was interesting to find that the core bacterial microbes in SL rhizosphere soil (i.e., Halomonas, Pelagibacterium, and Chelativorans) were salinity-and metal-tolerant communities, which could be ascribed to long-term selection under heavy soil pollution in the SL site.

Conclusions
In this study, we performed physicochemical analyses to assess the contamination level of bulk soils in unpolluted and polluted Pb/Zn mining sites, and 16S rRNA amplicon highthroughput sequencing to characterize and compare bacterial community variations in Cyp plants across different sampling sites and plant compartments. Both site-and compartmentdependent patterns of bacterial communities were detected based on the alpha and beta diversity results. Soil nutrient status and contamination level affected the core bacterial microbiome in the rhizosphere and root endosphere compartments, though many abundant microbes were shared among compartments. There was little variation in the dominant bacterial phyla in the bulk and rhizosphere soils between unpolluted and active mining sites. Overall, these findings improve our understanding of soil-plant-microbe interactions under pollution selection and provide insight into microbial associations for potential phytoremediation applications.