Effects of long-term discharge of acid mine drainage from abandoned coal mines on soil microorganisms: microbial community structure, interaction patterns, and metabolic functions

More than twenty abandoned coal mines in the Yudong River Basin of Guizhou Province have discharged acid mine drainage (AMD) for a long time. The revelation of microbial community composition, interaction patterns, and metabolic functions can contribute to a better understanding of such ecosystems, which in its turn can be helpful in the development of strategies aiming at the ecological remediation of AMD pollution. In this study, reference and contaminated soil samples were collected along the AMD flow path for high-throughput sequencing. Results showed that the long-term AMD pollution promoted the evolution of γ-Proteobacteria, and the acidophilic iron-oxidizing bacteria Ferrovum (relative abundance of 15.50%) and iron-reducing bacteria Metallibacterium (9.87%) belonging to this class became the dominant genera. Co-occurrence analysis revealed that the proportion of positive correlations among bacteria increased from 51.02 (reference soil) to 75.16% (contaminated soil), suggesting that acidic pollution promotes the formation of mutualistic interaction networks of microorganisms. Metabolic function prediction (Tax4Fun) revealed that AMD contamination enhanced microbial functions such as translation, repair, and biosynthesis of peptidoglycan and lipopolysaccharide, etc., which may be an adaptive mechanism for microbial survival in extremely acidic environment. In addition, acidic pollution promoted the high expression of nitrogen-fixing genes in soil, and the discovery of autotrophic nitrogen-fixing bacteria such as Ferrovum highlights the possibility of using this taxon for bioremediation of AMD pollution.


Introduction
The total number of coal mines in China has been decreased from more than 80,000 Xu et al. 2016) to around 4700 (by the end of 2020) under the guidance of a series of national policies. Among all coal mines in China, 90% belong to underground coal mines. Large amounts of groundwater rebound into the mining area after the abandonment of underground coal mines, then, a series of physical, chemical, and biological reactions will occur inside the mine site, making the abandoned mining area a collection of contaminants (Chen et al. 2019). For high-sulfur abandoned coal mines, associated FeS 2 in coal seams and gangue was oxidized by microbial catalysis, producing large amounts of acid mine drainage (AMD) (Sharma et al. 2020). AMD is considered to be one of the major environmental issues in the mining industry, typically characterized by low pH (usually < 3.0) and high levels of iron, sulfate, and other toxic heavy metals (Akcil and Koldas 2005). Therefore, if not treated properly, AMD can contaminate soil, groundwater, and surface water, etc., posing a serious threat to humans and the environment (Jin et al. 2020;Rambabu et al. 2020). Owing to the lack of effective management measures, AMD from abandoned coal mines usually discharged directly into the environment without any treatment, posing a higher ecological risk compared to active mines. China currently has large numbers of abandoned coal mines, and the associated safety and environmental problems need more attention.
Microorganisms are an important part of the soil system and play a significant role in the transformation and cycling of elements. Soil microbial community structure and diversity are highly influenced by organic matter, temperature, salinity, pH, and heavy metal, etc. (Zhao et al. 2014). AMD pollution can alter the structure of ecosystems by changing the physicochemical properties of the soil and the composition of the microbial community (Sun et al. 2016a). AMD is typically characterized by excessive dissolved solids, high mineralization, strong acidity, and extremely low pH, which can be a harsh environment for most living organisms. As a result, only acidophilic and metal-tolerant microorganisms can survive, including Proteobacteria, Firmicutes, Actinobacteria, Nitrospirae, and Acidobacteria (Baker and Banfield 2003). High enrichment of such bacteria was also discovered in the water and sediment of the AMD dam in Lancaster, South Africa (Lukhele et al. 2019). Acidophilic genera Metallibacterium, Aciditerrimonas, Halomonas, Shewanella, Ferrovum, Alicyclobacillus, and Syntrophobacter were found in the AMD-contaminated river sediment of Aha watershed, southwest China (Sun et al. 2015). Microorganisms typically found in AMD environments have novel cellular mechanisms and specific elemental metabolism to better adapt to harsh conditions (Johnson and Hallberg 2005). For example, ironand sulfur-metabolizing bacteria showed high abundance in samples contaminated by abandoned coal mine drainage (Sun et al. 2015). These microorganisms with special metabolic functions in the AMD environment can be candidates for bioremediation of contaminated sites (Johnson and Hallberg 2005). Therefore, an in-depth understanding of microbial communities, co-occurrence interaction patterns, and metabolic functions in AMD-contaminated environments is beneficial for the screening of microorganisms with specific metabolic functions and provides a basis for better bioremediation of contaminated environments. AMD pollution has been reported worldwide. However, due to the wide variation in geochemical conditions of different AMD, it is essential to select microorganisms suitable for bioremediation in AMD-contaminated environments. Multiple AMD overflow sites from abandoned coal mines were identified during our preliminary investigation. A large amount of yellow sediment has been deposited on the surface soil in AMD runoff areas from abandoned coal mines, which severely limited the function of the soil and the development of biological communities. Microbial remediation provides a better way to carry out the restoration of ecological functions of contaminated soil, which requires a clear understanding of microorganisms in AMD-contaminated environments first. Therefore, we collected soil from contaminated areas of four abandoned coal mines, along with reference soil as controls, to analyze the microbial communities, co-occurrence patterns, and metabolic functions of the soil samples, respectively. The objectives of this study were to (1) reveal the typical microbial communities in long-term AMD-contaminated soils from abandoned coal mines; (2) reveal the effects of AMD contamination on the co-occurrence patterns and interaction relationships of soil microorganisms; and (3) explore the special metabolic functions of microorganisms in AMDcontaminated environments.

Site description and sample collection
The study area is located in the Yudong River Basin, Kaili City, southwestern Guizhou Province. Local coal mining started in the 1980s, and most of the mines have been closed now. Its coal-bearing rock system is the Middle Permian Liangshan Group, which is a high-sulfur, medium-high-ash fatty coal. The sulfur (S t, ad ) content of the coal ranges from 0.90 to 16.65% (generally > 3%), with sulfide (S p, ad ) content ranging from 0.80 to 4.60%, mostly in spherical particles, stellately distributed in the coal seam. For abandoned coal mines, AMD is formed when atmospheric precipitation seeps into the shaft or groundwater bounces into the mine, and sulfide minerals in the coal seams and surrounding rocks are oxidized. When the groundwater level exceeds the surface, AMD will overflow into the surface environment along the coal mining channel, causing serious environmental pollution .
We collected 13 topsoil samples, of which eight contaminated soils (Group M, M1-M8) were collected from the AMD runoff areas of four abandoned coal mines and five reference soils (Group S, S1-S5) were collected from areas more than 500 m away from the AMD runoff areas (Fig. 1). To avoid DNA contamination of samples, the sampling apparatus was first sterilized (with 75% ethanol) before sampling. Soil samples (sampling depth 2-5 cm) were collected after removal of the surface soil, homogenized, and later placed in sterile centrifuge tubes. Each sample was a mixture of four subsamples collected from adjacent points, homogenized, and divided into two parts. Soil samples for physicochemical parameter testing were collected and stored at 4°C in the sampling box and transported to the laboratory for processing within 24h. Samples for DNA extraction were stored at 4°C after field collection and transported to a −20°C refrigerator for freezing within 2h, and then placed on ice for transport to the laboratory for storage at −80°C within 24h until analysis.

Physicochemical parameter measurements
Soil samples were freeze-dried using a freeze-dryer (FD-1A-50, BioCool, Beijing, China) for 48 h. A total of 10.0 g of dry samples sieved through a 2-mm mesh was mixed with 25 mL of distilled water (1:2.5 soil-water ratio) for testing soil pH (HJ 962-2018) (Sun et al. 2016b). After shaking for 2 min, the mixture was left to stand for 30 min, and then measured using a calibrated pH meter (PXSJ-216F, Shanghai, China). For samples with pH < 7, KCl solution (1 mol/L) was used instead of distilled water for testing (ISO 10390:2005) (Asta et al. 2014). Total iron in the soil was measured according to the method reported (Sun et al. 2018a). Extractions were performed using 1.0 M HCl for 30 min at room temperature, and then, the total Fe was measured spectrophotometrically at 510 nm with 1,10 phenanthroline (722, Sunny Hengping, Shanghai, China). A total of 0.2 g dry sample sieved through a 0.15-mm mesh was digested in a microwave digestion bath, and then, heavy metals Zn, Mn, Cu, Cr, Pb, and Cd were determined by inductively coupled plasma mass spectrometry (ICP-MS, TFS, USA) (Lu et al. 2012). After sample digestion, the samples were analyzed for Hg and As by atomic fluorescence spectrometry (AFS8130, Titan Instruments, Beijing, China) (Jiang et al. 2020). All measurements were performed in triplicate.

Illumina MiSeq sequencing and processing
Total soil DNA was extracted from 1.0 g sample using E.Z.N.A.® Soil DNA Kit (Omega Bio-tek, Norcross, GA, USA). The concentration and quality of the extracted DNA were assessed using a Nanodrop®ND-2000 UV-Vis Spectrophotometer (NanoDrop Technologies, USA) for evaluation. Bacterial universal primers 338F and 806R for the V3-V4 region of the 16S rRNA gene were used to generate an amplification library (Lee et al. 2012). PCR products were extracted using 2% agarose gel, further purified by AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA), and quantified using QuantiFluor™-ST (Promega, USA) ). Sequencing of the purified amplicons (300 bp paired) was performed on the Illumina Sequences from the Illumina MiSeq platform were processed using the QIIME (v 1.9.1) package (Campbell et al. 2010). Raw fastq files were demultiplexed, quality controlled with Trimmomatic filtering, and merged by FLASH (v 1.2.11) (Magoč and Salzberg 2011). Clustering of operational taxonomic units (OTUs) was performed according to the RDP Classifier algorithm (v 2.11) against the SILVA database with a 97% similarity control (Elmar et al. 2007).

Statistical analyses
The diversity of bacterial communities was assessed using the alpha diversity indices OTUs (Sobs), Chao (1984), Shannon (Shannon and Weaver 1950), and Simpson. The interactions between microbial populations in the samples were studied by the co-occurrence network analysis with Networkx software. Interactions between "microbe-microbe" were calculated using the Spearman correlation coefficients . Positive and negative correlations may indicate mutualistic and competition relationship between microorganisms, respectively. Phylogenetic tree was constructed based on the maximum likelihood (ML) method (Wim and Olivier 2005) and IQ-tree (v 1.6.8) for the top OTUs. The above data was analyzed on the free online platform of Majorbio Cloud Platform (www.majorbio.com).
The potential functions of the soil microbial community were profiled using Tax4Fun (v 0.3.1) (Kathrin et al. 2015). Differences in the level 2 predicted functions of contaminated and reference soils were compared and analyzed by the Wilcoxon rank-sum test. The main level 3 predictive functions of contaminated and reference soils were compared and clustered using an R-language-based heat map. Bubble plot and heat map were plotted in an online platform for data analysis and visualization (http://www.bioinformatics.com.cn).

Physicochemical properties of soil samples
The physicochemical parameters of the soil samples are shown in Table 1. The local soils, mainly from long-term weathering of limestone, are weakly alkaline with a pH range of 7.04 to 8.22. The pH of the soils decreased significantly after AMD contamination, changing from weakly alkaline to acidic with a pH range of 2.53 to 6.79. According to the classification criteria of soil pH (Xiong and Li 1990), most of the samples (6 out of 8) in the contaminated group were extremely acidic soils (pH < 4.5). This indicates that longterm scouring and erosion of AMD from abandoned coal mines has led to severe acidification and contamination of soils in the runoff area (Zhang et al. 2013a). Statistical analysis showed that the pH of the contaminated soil was significantly lower (p < 0.001) compared to that of the reference group, while the total iron content of the contaminated group (average 321.80 g/kg) was significantly higher than that of the reference soil (average 31.28 g/kg), showing the typical physicochemical characteristics of AMD contamination . The high total iron content in the contaminated soil may be due to the high concentration of iron ions in AMD, which generates secondary minerals such as jarosite, schwertmannite, and goethite deposited on the surface soil under the combined action of oxygen and microorganisms (Bao 2018).

Sequencing results and microbial diversity
A total of 547,649 effective reads were identified in 13 sequencing libraries with an average length of 430 bp. To make the samples comparable and avoid bias due to differences in sequencing depth, we drew the sequences flat to the same sequencing depth based on the sample with the minimum sequence reads (18,528). The number of OTUs obtained for each sample and the alpha diversity indices are shown in Table 2. Based on the 97% similarity threshold, all sequences were finally clustered into 4224 operational taxonomic units (OTUs), ranging from 153 to 1961. The abundance of soil microorganisms decreased significantly (p < 0.05) after AMD contamination, with the mean number of OTUs obtained for groups M and S being 781 and 1708, respectively. As shown in Table 2, the alpha diversity indices Chao1, Shannon, and Simpson showed similar trends; i.e., the richness and diversity of bacteria in contaminated soils were significantly reduced. Statistical analysis showed that the alpha diversity indices of the contaminated soil were significantly lower than those of the reference group (p < 0.05). This may be due to the fact that AMD contamination altered the physicochemical properties of the soil (Table 1), where the survival of most microorganisms was inhibited in extremely harsh environments and only some acid-tolerant microorganisms were able to grow, showing an overall reduction in bacterial richness and diversity (Zhang et al. 2019). Figure 2 shows the relative abundance of the main bacterial phyla in the samples. The dominant phylum in all samples was Proteobacteria, which accounted for 54.22% of the total bacterial reads in Group M (31.91 to 69.66% per sample), while in Group S, it was only 29.26%. The percentage of Proteobacteria in the soil increased significantly after AMD contamination, which is consistent with the findings of other studies . Proteobacteria were the most abundant phylum in soil samples along the AMD flow path of Dapo Coal Mine, accounting for an average of 43.5% of the total sequence and up to 64.5% in some samples ). In addition, we found that although Proteobacteria were the most abundant phylum in both Groups M and S, there were significant differences in composition at the lower taxonomic levels. At the class level, γ -Proteobacteria were predominant in Group M with an average of 47.56%, while α-Proteobacteria dominated in Group S (19.55%). Actinobacteriota ranked as the second abundant phylum, accounting for 10.54% and 23.24% in Group M and Group S samples, respectively. Actinobacteriota are suitable for survival in neutral or slightly alkaline environments (Yang 2017), which may explain their low abundance in AMD-contaminated soil.

Comparison of microbial community composition
Chloroflexi and Acidobacteriota showed the same characteristics as Actinobacteriota, with an average of 20.29% and 13.69% in Group S compared to only 6.25% and  4.58% in Group M, suggesting that AMD pollution may inhibit the growth of bacteria belonging to Chloroflexi, Acidobacteriota, and Actinobacteriota. Figure 3 shows detailed information on the relative abundance of the main genera (at least 2% in one sample). It is evident that Ferrovum spp. and Metallibacterium spp. dominate in Group M, especially in soil samples with extremely low pH value (Table 1). Ferrovum can use the energy generated by ferrous iron oxidation to fix carbon through the Calvin-Benson-Bassham cycle (Hua et al. 2015), and is often considered the main carbon sequestrator within AMD microbial communities (Moya et al. 2014). Metallibacterium is   (Sylvan et al. 2012), and some certain species, such as Metallibacterium scheffleri, have the ability to reduce Fe (III) but not oxidize Fe (II) (Ziegler et al. 2013). In addition, studies have shown that some strains of Metallibacterium have carbon fixation and sulfur oxidation genes, suggesting their potential for chemoautotrophic growth (Sibylle et al. 2017). The other genera with relatively high abundance in contaminated soil samples also include A c i d i t h i o b ac i l l u s , A c i di b a c t e r, F e r r i t h r i x , and Leptospirillum. Among them, Acidithiobacillus is the most studied chemoautotrophic microorganism in acidic environments, and one of the species, Acidithiobacillus ferrooxidans, can rapidly catalyze the oxidation of Fe 2+ to Fe 3+ and the dissolution of sulfide minerals, etc. (Williams and Kelly 2013). Acidibacter is considered acidophilic, exclusively heterotrophic bacteria with the ability to reduce Fe (III) (Falagán and Johnson 2014). Ferrithrix is a Gram-negative, thermotolerant heterotrophic acidophilic bacterium capable of oxidizing ferrous iron (Johnson et al. 2009). Leptospirillum has been reported to be responsible for AMD production (Schrenk 1998), and this group can be involved in metal and sulfur metabolism (Goltsman 2009). The species Leptospirillum ferriphilum is an obligate aerobic microorganism that can live in extremely acidic environment (pH 1.5-1.8), and can only obtain energy through ferrous oxide, and some studies have shown that it is highly tolerant to arsenic (Tuffin et al. 2006). In summary, it is evident that the dominant genera in AMD-contaminated soil are acidophilic bacteria, which can metabolize iron and sulfur, especially the oxidation of ferrous iron.
AMD contamination can alter the microbial community structure of the soil by promoting the development of some acidophilic bacteria as well as inhibiting the growth of other bacteria. As shown in Fig. 3, the relative abundance of Sphingomonas, Gaiella, Nocardioides, Ramlibacter, and Pedomicrobium was significantly higher in the uncontaminated reference soil than in the contaminated soil. These bacteria are suitable to live in neutral environments; for example, Gaiella grows at an optimum pH between 6.5 and 7.5 (Albuquerque et al. 2011), and Nocardioides is an aerobic, pH-neutral actinomycetes in soil (Du et al. 2012). Sphingomonas is strictly aerobic heterotrophic bacteria widely distributed in various habitats such as aqueous and terrestrial, and plant roots (Chen et al. 2018). Pedomicrobium belongs to heterotrophic and aerobic bacteria with Mn oxidation ability (Larsen et al. 1999).
Ramlibacter (Chaudhary and Kim 2017), Arthrobacter (Jones and Keddie 2006), and Lysobacter (Weon et al. 2007) are all aerobic, mesophilic, and neutrophilic bacteria in soil. These neutrophilic aerobic bacteria are an important part of the soil microbial system and play an important role in soil material cycling and energy metabolism.

Phylogenetic tree analysis of the main bacteria in contaminated soil
To clarify the metabolic capacity of the major bacteria in contaminated soil, we selected the top OTUs (relative abundance > 0.5% in all samples) and constructed phylogenetic trees between the known bacteria and top OTUs by comparison with the NCBI database (Fig. 4). Clearly, the top OTUs in contaminated soil are similar with several known ironoxidizing and iron-reducing bacteria.
Most of the top OTUs (17 out of 31) are more closely related to iron-oxidizing bacteria, such as Ferrovum myxofaciens, Acidithiobacillus ferrooxidans, Gallionella spp., Sideroxydans lithotrophicus, Leptospirillum spp., and Ferrithrix thermotolerans. The relative abundance of these top OTUs with iron oxidation capacity totaled accounted for 29.71% of the total library of contaminated soil. Ferrovum myxofaciens is an extremely acidophilic, obligate autotroph that can use the energy of ferrous iron oxidation for endogenous nitrogen fixation as well as CO 2 fixation via the Calvin-Benson-Bassham pathway. In addition, it can produce large amounts of extracellular polymers that lead to intercellular attachment ). In our study, four top OTUs were related to Ferrovum myxofaciens, namely OTU833 (9.76%), OTU839 (4.27%), OTU796 (0.76%), and OTU450 (0.71%), with a total relative abundance of up to 15.50%, indicating the presence of considerable amounts of autotrophic bacteria with iron oxidation and nitrogen fixation abilities in contaminated soil. This can be attributed to a survival strategy of bacteria in a strongly acidic and oligonitrogenous environment, which provides carbon and nitrogen sources for cell survival through autotrophic carbon and nitrogen fixation, contributing for their sequestration (Sun et al. 2020b). OTU835 and OTU322 were identified to be similar with Acidithiobacillus ferrooxidans, an acidophilic chemolithoautotrophic bacterium that can grow by oxidizing Fe (II) and reducing sulfur compounds (Jin et al. 2020). Phylogenetic evolutionary tree analysis revealed that five OTUs, namely OTU479, OTU342, OTU542, OTU475, and OTU829, are more similar with Gallionella spp. and Sideroxydans lithotrophicus, both of which are ironoxidizing bacteria (Kipry et al. 2013;Anke et al. 2019). The other OTUs that can be identified to the known iron-oxidizing bacteria included OTU287 (Leptospirillum spp.), OTU522 and OTU546 (Thiobacillus thioparus), and OTU760, OTU284, and OTU141 (Ferrithrix thermotolerans). Among them, Leptospirillum spp. is considered to be an ironoxidizing bacterium and carries genes for dissimilatory reduction of sulfate (Goltsman 2009); Ferrithrix thermotolerans is an extremely acidophilic, heterotrophic, and iron-oxidizing Actinobacteria (Johnson et al. 2009).
Six of the 31 top OTUs were more similar with ironreducing bacteria, with a total relative abundance of 15.23%.
Most of these OTUs with iron-reducing ability were more similar with Metallibacterium Scheffleri, including OTU834 (6.26%) and OTU143 (3.61%), with a total relative abundance of up to 9.87%. Metallibacterium scheffleri is regarded as a facultatively anaerobic and acid-tolerant microorganism with the ability to reduce Fe (III) (Ziegler et al. 2013). Metallibacterium scheffleri can thrive on casitone as an electron donor and carbon source, possibly due to the production of NH 3 during the cellular metabolism of amino acids, which may be the survival strategy for Metallibacterium scheffleri to be able to tolerate the extremely acidic environments (Ziegler et al. 2013). Additionally, genes for sulfur and hydrogen oxidation were found in Metallibacterium scheffleri, suggesting a possible potential for sulfur oxidation (Sibylle et al. 2017). The other three OTUs are thought to be more similar with Acidibacter ferrireducens, an acidophilic, obligate heterotroph that can reduce Fe (III) (Falagán and Johnson 2014).

Changes of microbial interactions after AMD pollution
In addition to soil microbes being influenced by environmental conditions, microbial interactions, such as mutualism, Fig. 4 Phylogenetic tree of the top OTUs (relative abundance > 0.5% in all samples) in contaminated soil. Number in the box indicates the relative abundance of each OTU in all contaminated samples. Bold and underlined font indicates the closest known bacteria in the NCBI database. Fonts in yellow and green refer to iron-oxidizing bacteria and iron-reducing bacteria, respectively. Black fonts indicate OTUs that could not be clearly classified commensalism, and parasitism, may also play a role in the structure and composition of microbial communities (Nemergut et al. 2013). Co-occurrence networks among microorganisms can provide considerable information and have been successfully used to identify microbial interactions (Sun et al. 2016a). In this study, co-occurrence networks between microorganisms were constructed separately for the main microorganisms in contaminated soil (Fig. 5a) and reference soil (Fig. 5b).
The number of links in the co-occurrence network can be used to assess the inter-microbial interactions, and the higher the number of links, the stronger the microbial interactions (Banerjee et al. 2018). The numbers of strong links in the reference soil and contaminated soil were 833 and 318, respectively (Fig. 5d), indicating that the strength of microbial interactions in the reference soil was stronger than that in the contaminated soil. AMD contamination decreased the richness and diversity of soil microorganisms, which in turn leads to a weakening of the overall interactions between microorganisms. The interaction between microorganisms can be divided into positive and negative effects, which represent mutualistic and competition interactions, respectively. In this study, the proportion of positive and negative links in the reference soil was 51.02% and 48.98%, respectively (Fig.  5d), indicating that the positive and negative interactions between microorganisms were basically in balance. However, we observed a significant increase in the proportion of positive links in contaminated soils, although the overall intensity of the interaction between microorganisms decreased, with 75.16% and 24.84% of positive and negative links in contaminated soils, respectively (Fig. 5d). Positive correlations among microorganisms may indicate symbiosis, parasitic relationships, or collective synergistic degradation of organic matter, while negative correlations may be attributed to Fig. 5 Network of co-occurrence based on top OTUs (relative abundance > 0.2% in the corresponding group samples) in the contaminated soil (a) and reference soil (b). A connection stands for the significant Spearman correlation with 0.8 < |r| < 1 (p < 0.05). The size of each key node in the figure is proportional to the relative abundance of that OTU. The color of a node represents the assignment at phylum level. The red and green edges stand for positive and negative correlations, respectively. Number of keystone OTUs within each network was counted at phylum level and presented as percentages (c). Number of the strong links (0.8 < |r| < 1), including positive and negative links, was counted in each group (d) competition among microorganisms or different living environments . The predominance of positive correlations in contaminated soils indicates an enhanced reciprocal symbiosis among microorganisms, which may be a survival strategy for microorganisms in extreme environments by forming mutually beneficial interaction networks for better survival in the harsh environment of strong acidity.
There was a significant difference in the assignment of key nodes in contaminated and reference soils when considering the taxonomic attribution of each node (Fig. 5c). The key nodes in the reference soil samples were mainly attributed to the phyla Proteobacteria, Actinobacteriota, Acidobacteriota, and Chloroflexi, accounting for 38.38%, 28.28%, 17.17%, and 12.12% of the total number of nodes, respectively. Most of the key nodes in the contaminated soil also belonged to these four phyla; however, the proportion changed significantly. The percentage of key nodes belonging to Proteobacteria increased from 38.38 to 46.75%, while the proportion of the other three phyla all decreased, accounting for 12.98% (Actinobacteriota), 9.09% (Acidobacteriota), and 3.89% (Chloroflexi), respectively. In addition, some new nodes were displayed in the contaminated soil compared to the reference soil, mainly attributed to Firmicutes, Nitrospirota, and Patescibacteria, accounting for 5.19%, 3.90%, and 3.90% of the total number of nodes, respectively. Bacteria of Firmicutes usually show a higher resistance to harsh environments, which is attributed to the high peptidoglycan content of the bacterial cell wall that can protect the cells from the damage of external environment; in addition, most bacteria of this phylum can produce spores to preserve cellular activity (Dworkin et al. 2006). The presence of Leptospirillum makes Nitrospirota become the key nodes in the microbial network of contaminated soil. Leptospirillum was almost absent in the reference soil, but it showed highly abundance in AMD-contaminated soil. Leptospirillum has been reported as an acidophilic ironoxidizing bacterium with carbon fixation genes, as well as with multiple pathways for osmotic protection that allow it to thrive in the oligotrophic and extremely acidic environments (Goltsman 2009;Tuffin et al. 2006). Patescibacteria is commonly found in groundwater systems (Herrmann et al. 2019;Tian et al. 2020), and its emergence as a new critical node in contaminated soils may be related to the source of AMD in abandoned coal mines. After the abandonment of coal mines, groundwater rebounded into the mining area to form AMD, and some of the microorganisms in the groundwater will be carried into the surface environment along with the overflowing of AMD.

Co-occurrence pattern of microorganisms in contaminated soil
The previous analysis showed that the dominant bacteria in contaminated soil were mainly iron-metabolizing bacteria, and positive interactions between these microorganisms were significantly enhanced after AMD contamination. To determine the specific interactions between these microorganisms, we constructed a co-occurrence network of the main ironmetabolizing bacteria in contaminated soil at the genus level (Fig. 6).
As illustrated in Fig. 6, the main bacteria in the contaminated soil have different interactions with each other. The main iron-metabolizing bacteria in contaminated soil can be divided into three groups according to the interactions between microorganisms. The first group included Acidicapsa, Metallibacterium, Acidibacter, and Acidiphilium, and these iron-reducing bacteria have a positive correlation with each other. Among them, Metallibacterium is concurrently autotrophic bacteria, and the other three genera are all heterotrophic bacteria. In addition, we also found iron-oxidizing bacteria Ferrithrix, Acidithiobacillus, Acidibacillus, and Ferrovum are positively correlated with each other. These four bacteria all can oxidize Fe (II), but their nutritional patterns are quite different. Ferrovum and Acidithiobacillus belong to autotrophic bacteria (Holanda et al. 2015;Schopf et al. 2017), while Ferrithrix and Acidibacillus are heterotrophic. Autotrophic and heterotrophic bacteria have obvious synergistic relationship in carbon source utilization, mutual elimination of growth, and inhibition of metabolism (Baker and Banfield 2003), which may account for the positive correlation between these bacteria. Gallionella, Thiobacillus, Sideroxydans, Fig. 6 Co-occurrence network showing correlations among the main iron-metabolizing genera in contaminated soil. A connection indicates a strong (Spearman's 0.7 < |r| < 1) and significant (p < 0.05) correlation. The size of a node is proportional to the number of connections (degree). The nodes colored in yellow and blue represent for iron-oxidizing and iron-reducing bacteria, respectively. The width of an edge is proportional to the Spearman correlation coefficient. The red and green edges stand for positive and negative correlations, respectively Curvibacter, Ferritrophicum, Geothrix, and Geobacter formed another group of bacterial networks that are positively correlated with each other. Among them, Geothrix and Geobacter belong to iron-reducing bacteria, and the other bacteria were all iron-oxidizing bacteria.
The distribution of the above three groups of bacteria in the contaminated soil samples also showed significant differences (Fig. 3). The bacteria in the first and second groups showed higher relative abundance in the contaminated soil, and these bacteria are mainly distributed in samples M1, M3, M4, M5, M7, and M8 with low pH value (pH < 4). However, bacteria in the third group such as Sideroxydans, Thiobacillus, Geothrix, and Geobacter were mainly distributed in samples M2 and M6 with pH values of 6.79 and 5.88, respectively. This indicates that microorganisms formed different interaction networks in moderately (pH range 5.9 to 6.8) and severely (pH < 4) acidic contaminated soils. Iron-oxidizing and ironreducing bacteria formed separate networks of independent interactions in soils with pH < 4. In contrast, iron-oxidizing bacteria and iron-reducing bacteria suitable for living in a moderately acidic environment (pH value 6~7) together formed a collaborative network with a positive correlation with each other.

Predictive functional profiling of microbial community
The microbial community composition of environmental samples is closely related to environmental conditions. However, studies have demonstrated that environmental conditions are more associated with microbial community function; i.e., microbial communities in similar environments may differ significantly, while their community functions may be similar (Gibbons 2017). Therefore, in addition to revealing the composition and interaction of microbial communities in soils, it is particularly important to reveal the differences in metabolic functions of microbial communities in contaminated and reference soils. The application of metagenome is limited due to the inappropriate pretreatment of samples or prohibitive costs, and the functional prediction based on 16S rRNA sequence has become a beneficial complement (Kathrin et al. 2015), despite being indirectly obtained through analysis of available genomic and functional information and not directly from the organisms in the samples under study. As one of the best functional prediction methods, Tax4Fun has shown high correlation with metagenomic sequencing on functional profiles (Kathrin et al. 2015). It has been successfully applied in various habitats such as the pond sediment (Silveira et al. 2020), soil (Sun et al. 2020a), and hot spring (Mehetre et al. 2018). In this study, we used Tax4Fun, a metabolic prediction method with SILVA, as reference database (Kathrin et al. 2015), to explore the changes of metabolic potential in soils after AMD contamination. In this method, metabolic pathways based on Kyoto Encyclopedia of Genes and Genomes (KEGG) predictions were compared between contaminated and reference soil samples.
The prediction results (Table 3) showed that the predicted KEGG pathways at level 1 for both contaminated and reference soil samples were mostly dominated by metabolism (61.12% and 61.80%, respectively), followed by environmental information processing (18.08% and 19.88%), and genetic information processing (13.88% and 10.64%). Results of the Wilcoxon rank-sum test showed that most of the predicted KEGG pathways at level 2 (32 out of 40 functions) were significantly different in contaminated and reference soil samples (p < 0.05, Fig. 7). The proportions of amino acid metabolism, membrane transport, xenobiotic biodegradation, and metabolism all decreased significantly after AMD contamination, from 12.74 to 11.01%, 12.72 to 9.24%, and 4.70 to 3.28%, respectively. However, we also discovered that some metabolic functions were enhanced after contamination, such as some metabolism functions (energy metabolism, glycan biosynthesis and metabolism, nucleotide metabolism, etc.), and most of the genetic information processing functions (translation, folding, sorting and degradation, replication and repair) and part of the environmental information processing functions (signal transduction) (Fig. 7). The high expression of these functions may be related to the survival mechanism and adaptations of microorganisms in harsh environment. Studies have demonstrated that microorganisms in AMD environment have novel cellular mechanisms for adaptation to harsh conditions (Johnson and Hallberg 2003). Genes associated with adaptation are overexpressed in AMD systems compared to normal metabolism (Ram et al. 2005).
The main functions (proportion > 1.0% in all predicted functions) of the KEGG pathway at level 3 and their hierarchical clustering results are shown in Fig. 8. The contaminated and reference soil were marked in yellow and blue, respectively. The results of the clustering analysis showed that the soil samples could be divided into two groups according to the community function. Studies have confirmed that pH has an important effect on the distribution of microbial community structure and function (Sharma et al. 2020;Plazas et al. 2019). In our study, the microbial community functions of M2 and M6 with a pH value of 6.79 and 5.88, respectively, were close to those of the reference soil, while the community functions of other contaminated soil samples with lower pH value were similar. As shown in Fig. 8, most of the main microbial functions of level 3 (20 out 27 functions) are higher in cluster II than those in cluster I, especially in severely acid contaminated samples M1, M7, and M8. Metabolic functions are overexpressed in cluster II samples translation (ribosome, and aminoacyl-tRNA biosynthesis), replication and repair (homologous recombination, nucleotide excision repair, and mismatch repair), nucleotide metabolism (purine metabolism, and pyrimidine metabolism), RNA degradation, twocomponent system, and bacterial secretion system, which were regarded as the essential factors for the survival of microorganisms in harsh environmental conditions. For example, the bacterial secretion system is thought to regulate the secretion of protective molecules that allow bacteria to withstand the harsh conditions (Lukhele et al. 2019). Moreover, since the extreme conditions of AMD systems can cause significant damage to biomolecules (Lukhele et al. 2019), microorganisms in AMD have evolved an efficient repair system usually exhibiting the overexpression of genes involved in replication, repair, recombination, and translation. In our study, in addition to the high expression of genes related to heredity and repair functions, we also observed that the synthetic functions of peptidoglycan and lipopolysaccharide were overexpressed in the samples of cluster II. Peptidoglycan is the main component of the cell wall of Gram-positive bacteria, and the solid cell wall plays an important role in maintaining the bacterial cell morphology and resisting osmotic damage (Reith and Mayer 2011). Lipopolysaccharide is a unique component of the cell wall of Gram-negative bacteria, which enables the outer membrane of bacteria to act as a tight barrier to Fig. 7 Statistical differences of the predicted KEGG pathways (level 2) of significant variations between contaminated soil and reference soil based on all OTUs obtained from samples. Corrected p values were calculated using the Wilcoxon rank-sum test (p < 0.05) most substances harmful to bacteria (Zhang et al. 2013b). Therefore, we hypothesize that the high expression of the synthetic functions of peptidoglycan and lipopolysaccharide in the AMD environment is also an adaptation mechanism of microorganisms to the acidic environment. Nitrogen is an essential element for biological life activities, and since AMD environments are usually oligotrophic, this limits the possibility of bioremediation using microorganisms or plants. In this study, we found that the nitrogen metabolism of bacteria was significantly enhanced after AMD pollution, and the average abundance of nitrogen metabolism function in cluster II (2.33%) was higher than that in cluster I (2.10%). The distribution of some KEGG orthologs (KOs) associated with nitrogen fixation in soil samples is shown in Fig. 9. Results showed that the total relative abundance of nitrogenfixing genes in cluster II (0.30% on average) was significantly higher than that in cluster I (0.09%), especially in the extremely acidic M1, M7, and M8 samples with total relative abundance of 0.34%, 0.37%, and 0.37%, respectively, while their total abundances in the reference soil samples were all below 0.1%. The relative abundance of nitrogen fixation genes can reveal their potential nitrogen fixation capacity, and we found that the nitrogen fixation capacity of the microorganisms in AMD-contaminated soils was significantly enhanced, especially in extremely acidic soils. Studies have shown that genes for carbon fixation, nitrogen fixation, and metal resistance are widely present in a variety of microorganisms in AMD environments, which are survival strategies for microorganisms to survive in extremely acidic, oligotrophic environments (Sun et al. 2018b). These autotrophic microorganisms can grow autotrophically using inorganic compounds to provide essential carbon sources to Fig. 8 Tax4Fun-derived heatmap and hierarchical clustering comparing the main predicted functions on level 3 of contaminated soil and reference soil the environment through carbon fixation processes. In addition, microorganisms with nitrogen fixation capabilities can provide a nitrogen source for the environment by converting atmospheric nitrogen into bioavailable nitrogen through nitrogen fixation processes (Sun et al. 2020b). In this study, Ferrovum, as the most abundant genus in AMD-contaminated soil, was considered to endogenously fix nitrogen and CO 2 using the energy from ferrous iron oxidation ). Moreover, since species belonging to Ferrovum can secrete large amounts of extracellular polymers (EPS) that form strong biofilms on solid surfaces, they can be used in the treatment of iron-rich AMD to enhance the oxidation of Fe 2+ to Fe 3+ and ultimately remove iron ions in AMD (Hedrich and Johnson 2012). Meanwhile, the organic nitrogen and carbon source produced by Ferrovum in the iron-rich AMD environment can be further utilized by heterotrophic microorganisms. For example, heterotrophic sulfatereducing bacteria (SRB) may use these carbon sources to reduce sulfate and generate OH − during this progress, which in turn can increase the pH of the AMD system as well as remove some heavy metals through precipitation (Burns et al. 2012). At the same time, nitrogen-fixing bacteria can also provide nitrogen sources for the growth of hyper-accumulator (e.g., Bidens bipinnata) and promote their accumulation of large amounts of heavy metals during growth to facilitate phytoremediation of contaminated sites (Sun et al. 2018b). The above analysis showed that the autotrophic nitrogen-fixing bacterium Ferrovum may provide a possible way for the bioremediation of AMD environments, such as the removal of iron ions from iron-rich acidic water, or the provision of organic nitrogen and carbon sources for the growth of heterotrophic bacteria and plants in the initial stages of remediation. However, the restoration mechanism of Ferrovum and the synergistic process with heterotrophic bacteria or plants during the ecological restoration of AMD environments need further study.

Conclusions
In this study, we found that the long-term AMD pollution significantly affected the microbial community structure, interaction patterns, and metabolic functions of soil microorganisms, mainly included:

The AMD contamination inhibited the survival of
Chloroflexi, Acidobacteriota, Actinobacteriota, etc. of soil, significantly reducing the richness and diversity of soil microorganisms while promoting the evolution of γ-Proteobacteria. Acidophilic iron-metabolizing bacteria became the dominant microbial community in the longterm acidic contaminated soil, including iron-oxidizing bacteria such as Ferrovum, Acidithiobacillus, Gallionella, Sideroxydans, and Leptospirillum, and ironreducing bacteria, Metallibacterium and Acidibacter, etc. 2. Co-occurrence network analysis revealed that the percentage of positive correlations among bacteria in soil increased from 51.02 (reference soil) to 75.16% (contaminated soil), which may be a survival strategy for microorganisms in extreme environments by forming mutually beneficial interaction networks for better survival in the harsh environment of strong acidity. Some new nodes were displayed in the co-occurrence network of soil microorganisms after AMD contamination, mainly attributed to Firmicutes, Nitrospirota, and Patescibacteria. 3. The microbial interaction patterns varied in soils with different levels of AMD contamination. Iron-oxidizing and iron-reducing bacteria formed separate networks of independent interactions in soils with pH < 4. In contrast, ironoxidizing bacteria and iron-reducing bacteria suitable for living in a moderately acidic environment (pH value 5.9~6.8) together formed a collaborative network with a positive correlation with each other. 4. Metabolic function prediction (Tax4Fun) was used to explore the changes of metabolic potential in soils after AMD contamination. Results showed that most of the Fig. 9 Distribution of predicted genes related to nitrogen fixation in soil samples main microbial functions of level 3 (20 out of 27 functions) are higher in contaminated soil. AMD contamination enhanced the microbial functions such as translation, remediation, and biosynthesis of peptidoglycan and lipopolysaccharide, etc., which may be an adaptive mechanism for microbial survival in extremely acidic environment. 5. In addition, we found that the AMD contamination of soil promoted the enhancement of nitrogen metabolic functions and the high expression of nitrogen fixation genes, which may be responsible for the high development of autotrophic nitrogen-fixing bacteria such as Ferrovum. Therefore, we suggest that Ferrovum can be used in the bioremediation of AMD environments to achieve the provision of organic nitrogen and carbon sources in the initial stage of remediation.