A total of 24 AMD water samples with different mineralogy were collected from four mining areas; the detailed information of sampling locations is provided in Supplementary Table 1. Nine water samples named ZJA11, ZJA12, ZJA13, ZJA21, ZJA22, ZJA23, ZJA31, ZJA32, and ZJA33 were collected from AMD collection pond A1, A2, and A3 of the heap bioleaching area of ZJS copper mine in Shanghang, Fujian Province, in October 2018. The three acid mine water wells had a continuously high-temperature inflow (about 40°C). Three extremely acidic water samples (approximate pH 0.7), MYW1, MYW2, and MYW3, were collected from the MYW copper mine in Myanmar. Six AMD samples, named BD1, BD2, and BD3, and CD1, CD2, and CD3, were collected from two naturally formed acid mine water pits in the Bainiuchang polymetallic mining area (MZ) in Mengzi City, Yunnan Province, in September 2017. Six water samples, C1W, C3W, C4W, C5W, C6W, and C9W, were collected from water pits C1, C3, C4, C5, C6, and C9 in the DBS polymetallic mine in Guangdong Province in October 2018. The temperature, pH, and conductivity of each sampling point were measured by a portable water quality multi-parameter meter (Star A329, Thermo Orion, United States). Iron (Fe²⁺, Fe³⁺) and SO₄²⁻ concentration were determined using a portable colorimeter (DR/890, HACH, United States) based on the o-phenanthroline spectrophotometry (Environmental Protection Industry Standard of China, HJ/T 345-2007) and turbidimetric method, respectively. Approximately 10ml water samples were digested by solution composed of nitric acid and hydrochloric acid (8:4v/v) using a microwave digestion instrument (TOPwave, Analytik Jena, Germany) and heated on an electric hot plate to remove the acid. After diluting to an appropriate concentration, the solution was loaded onto an inductively coupled plasma-optical emission spectrometry (ICP-OES) analyzer (Optima 5300 DV, PerkinElmer, United States) to determine element contents. Hierarchical clustering was used to determine the similarity/dissimilarity of the physicochemical parameters of the samples, and Principal Component Analysis (PCA) was performed to retain the most important environmental factors using R 3.6.3 software package vegan (Dixon, 2003), FactoMineR (Lê et al., 2008), and factoextra.

For DNA extraction, water samples were filtered through a sterile microporous filter membrane (PES, Φ47–50mm, 0.22μm, PALL, United States) using a solvent filter and a vacuum pump. Genomic DNA was extracted from the filters by following the manual of the PowerMax Soil DNA Isolation Kit (Qiagen, United States). The DNA was dissolved in tris hydrochloric acid buffer (pH 8.0) and frozen at −20°C. The V4–V5 region of the 16S rRNA gene of almost all Bacteria and Archaea in the samples was amplified using the primer set 515F (5'-GTGYCAGCMGCCGCGGTAA-3') and 926R (5'-CCGYCAATTYMTTTRAGTTT-3'). PCR reactions were performed using KAPA HiFi HotStart ReadyMix (KAPA, Woburn, MA, United States) with amplification conditions comprising a pre-denaturation step at 95°C for 3min, then 25 cycles of denaturation at 98°C for 20s, annealing at 55°C for 15s, and extension at 72°C for 15s, followed by a final extension at 72°C for 1min. The resulting PCR products were purified by gel electrophoresis, the DNA concentration was determined by Qubit fluorometer (Qubit4, ThermoFisher Scientific, United States), and the sequencing adapter was connected to complete the library construction. Sequencing of 16S rRNA gene amplicons was carried out on an Illumina MiSeq PE250 platform by Beijing Boaohuijiu Biotechnology Co., Ltd.

Clean sequencing reads of each sample were obtained by dividing datasets into subsets according to the sample tags. Single-ended sequences were merged using USEARCH v11, and primers were trimmed using Cutadapt v2.3 (Edgar, 2010; Martin, 2011). After removing low-quality sequences and chimeric and filtering singletons, high-quality reads were denoised to obtain zero-radius operational taxonomic units (zOTUs) using USEARCH v11. In Qiime 2, the primer set 515F/926R was used to train the RDP species classifier (Bolyen et al., 2019). Taxonomic classification was performed against the Silva 138 16S rRNA gene database (December 2019 release). Data transformation and manipulation were performed using R packages phyloseq v1.28.0, microbiome v1.6.0¹ and microeco v0.3.4 (McMurdie and Holmes, 2013; Liu et al., 2021a). According to the annotation results, mitochondrial and chloroplast sequences were removed. To decipher the metabolic and ecologically relevant functions of the AMD microbial communities, taxon abundance profiles were converted and mapped into putative functional abundance profiles using FAPROTAX (Louca et al., 2016).

Community compositions at every taxonomic level were counted, and alpha diversity indexes of each sample were calculated. zOTUs with a relative abundance ≥1% were defined as abundant zOTUs and those with a relative abundance <0.01% were defined as rare zOTUs (Xue et al., 2020). Similarities and differences in communities were determined by principal coordinate analysis (PCoA) of Bray–Curtis distances obtained using vegan v2.5.6 (Dixon, 2003). The Mantel test was used to quantify the relationship between environmental factors and microbial assembly. Distance-based redundancy analysis (db-RDA) was conducted on core zOTUs and selected environmental parameters (SO₄²⁻, temperature, Fe²⁺, Mg²⁺, and Ca²⁺) to identify possible correlations between environmental gradients and community changes. Values of p for multiple comparisons were adjusted using the false discovery rate of the Benjamini–Hochberg method. All statistical analyses were performed using the R v3.6.3 software package.

Microbial co-occurrence networks were constructed to characterize the microbial niches with different mineralogy. The correlation was calculated using the OTU that appeared in at least five samples and had more than 10 average sequences in the sample. SParse InversE Covariance estimation for Ecological Association and Statistical Inference (SPIEC-EASI) was performed and correlations with coefficients <0.3 were filtered (Kurtz et al., 2015). Positive correlations were exclusively focused on based on their mathematical interpretation and biological meaning. Global network properties were calculated to describe the topology using igraph packages in R (Csardi and Nepusz, 2006), and the networks were visualized using Gephi.

The AMD samples collected from DBS, MZ, ZJS, and MYW were characterized by extremely low pH (0.7–2.59), high ferrous and ferric ion concentrations (ranging from 1.575 to 6,482mgL⁻¹ and from 409.18 to 36,504mgL⁻¹, respectively), and high sulfate concentration (1,721–161,220mgL⁻¹; Supplementary Table 1). Dissolved metals including Al³⁺, Mg²⁺, Cu²⁺, and Mn²⁺ were also abundant in the samples with concentrations ranging between 40.58–12,346, 44.64–3,576, 45.66–1,923.2, and 15.24–1,229.6 mgL⁻¹, respectively. Other metal species present in substantial amounts included Zn, As, B, and Cr. Ba, Cd, Li, Mo, Ni, Pb, and Sr were measured as trace elements (<5 mgL⁻¹) in most of the AMD samples except for the high concentration of Ni (~33.45 mgL⁻¹) in MYW1-3. All chemical parameters were analyzed and included in PCA analysis and Hierarchical Clustering (Figures 1A,B). Hierarchical clustering analysis showed that AMD samples were divided into two main groups according to physicochemical parameters. One cluster included all samples from the ZJS and MYW sites, which were both copper mines. Another cluster comprised most polymetallic mines samples from the MZ and DBS sites. Heatmap for the environmental factors of AMD confirmed that the cluster of all samples from the ZJS and MYW copper mines did differentiate from the cluster of MZ and DBS polymetallic mines (Figure 1B). Cluster 1 had higher concentrations of Fe³⁺, SO₄²⁻, Cu²⁺, and Zn²⁺ than cluster 2. PCA showed that the two clusters were significantly different from each other (Anosim, R=0.751, p=0.001; Figures 1A,C). PCA axes 1 explained 55.2% variance of samples, which was mainly contributed by SO₄²⁻ and metals including V, Cr, Cu²⁺, and Fe³⁺ (Figure 1D). PCA axes 2 explained 19.2% variance, which was mainly contributed by temperature, Mg²⁺, Mo²⁺, and Fe²⁺ (Figure 1E). These results demonstrated that SO₄²⁻, metal ions, and temperature were the factors that contributed most to the variation of samples from the different mining areas.

Sequencing and data manipulation generated 611,447 quality sequences, with an average of 26,184 sequences and a range of 5,894–49,612 sequences per sample (samples with <5,000 sequences were filtered). A total of 184 zOTUs were obtained after denoising and filtering. Sample clustering based on microbial community composition also demonstrated that the AMD samples clustered together according to the mining areas (Figure 2). Comparative analysis of the alpha diversity index including Shannon, Simpson, Chao1, and Observed OTUs (Richness) revealed that samples of MZ, DBS, and ZJS had significantly higher microbial diversity than samples of MYW (Figure 2A). All of the alpha diversity indexes (with an average Shannon index score of 0.31 and Observed OTUs number of 7) indicated that microbial diversity was extremely low in AMDs of MYW. Although the Shannon and Simpson indexes were similar for MZ, DBS, and ZJS AMDs, MZ AMD had the highest observed OTUs and Chao1 indices, followed by DBS AMD, which means that microbial community richness was higher in MZ and DBS AMDs. Compositional similarities and dissimilarities of microbial communities were further explored using PCoA of Bray–Curtis distances. The total interpretation of the first two axes, PCo1 and PCo2, was 65.37%. There were significant differences between the microbial communities across mining areas (Figure 2B). Prevalence analysis of zOTUs showed that four shared zOTU were detected in all AMD samples (accounting for 34.1% of the total sequences), while there were 54, 15, 9, and 1 unique zOTUs in samples of MZ, DBS, ZJS, and MYW, respectively (Figure 2C).

From the total sequences generated, over 97.8% of reads (the average among samples) could be classified under the Kingdoms of Bacteria and Archaea, but approximately 2.2% remained unclassified at the domain level. Among the classified sequences, 13 phyla were identified, with Thermoplasmatota, Nitrospirota, and Proteobacteria (average relative abundance of 41, 28.2, and 24.3%, respectively) representing the most abundant phyla (Figure 2D). At the genus level, more than 52.5% of the total sequences were distributed in the two most dominant genera in all AMD samples; these genera were Leptospirillum and Ferroplasma (Figure 2E). ZJS contained the highest abundance of Leptospirillum, while MYW was dominated by Ferroplasma. Acidithiobacillus were particularly richer in MZ and DBS (with an average relative abundance of 27.1%) than in ZJS and MYW (with an average relative abundance of 0.07%). The relative abundance of other phylotypes varied across the mining areas. Acidiphilium, Ferrovum, Acidicapsa, and Acidibacter were more abundant in MZ and DBS (average relative abundance of 12.5, 0.3, 0.8, and 0.9%, respectively) than in ZJS and MYW. ZJS samples had the highest relative abundances of A-plasma (9.3%), Sulfobacillus (2.3%), Cuniculiplasma (2.4%), and a genus affiliated with Babeliaceae (2.9%) which all occurred in lower proportions in other AMD (0.3, 0.17, 0.81, and 0%, respectively). Ferrimicrobium, Acidisphaera, Metallibacterium, Acidocella, and Acidiferrobacter were exclusively detected in DBS AMD.

A significant correlation between physicochemical variation and microbial community dissimilarity in AMD was confirmed by the Mantel test (r=0.521, p=0.001). Furthermore, db-RDA was performed to identify possible correlations between physicochemical characteristics and microbial community patterns of the four mining areas (Figure 3A). Environmental factors with multicollinearity were filtered based on the value of Variance Inflation Factors (VIF). The step model based on the lowest Akaike Information Criterion (AIC) was performed to determine the fewest influencing factors but with the best interpretation. In this db-RDA analysis, variations of five optimal physicochemical parameters—SO₄²⁻, Mg²⁺, Ca²⁺, temperature, and Fe²⁺—collectively explained 64.2% of the microbial community dissimilarity (Figure 3B). Among them, Mg²⁺ and SO₄²⁻ were the environmental factors with the highest explanation values of 18.4% (r²=0.44, p=0.001) and 17.6% (r²=0.49, p=0.001), respectively, followed by Ca²⁺ with 11.3% (r²=0.35, p=0.005), temperature with 11.1% (r²=0.41, p=0.006), and Fe²⁺ with 5.7% (r²=0.11, p=0.307; Figure 3B).

At the level of individual dominant taxa, the genus Leptospirillum was positively correlated with temperature and Ca²⁺ but negatively correlated with Mg²⁺ (Figure 3C). Positive correlations of Ferroplasma with Fe²⁺, Mg²⁺, and SO₄²⁻ were observed, while Acidithiobacillus showed negative correlations with temperature, Ca²⁺, and SO₄²⁻ (Figure 3C). A genus affiliated with Thermoplasmatota, A-plasma, and a genus affiliated with Babeliales were also positively correlated with temperature and Ca²⁺ but negatively correlated with Mg²⁺ like the genus Leptospirillum. Acidiphilium, Ferrimicrobium, Acidibacillus, Acidisphaera, a genus affiliated with uncultured Acidobacteriaceae Subgroup1, E-plasma, Ferrovum, Acidicapsa, Acidibacter, and a genus affiliated with uncultured Acidobacteriaceae were all negatively correlated with temperature, Ca²⁺, and SO₄²⁻. Ferrovum, Acidicapsa, and a genus affiliated with uncultured Acidobacteriaceae were negatively correlated with Fe²⁺, while Sva0485 was positively correlated with Fe²⁺. In summary, the microbial communities in AMD were mainly influenced by the environmental factors of Mg²⁺, SO₄²⁻, Ca²⁺, temperature, and Fe²⁺.

To decipher the co-occurrence pattern of the AMD microbial community, correlations between microorganisms at the zOTU level of samples from the four mining areas were calculated. Strong correlations (r>0.3) were retained to build co-occurrence networks (Figure 4). A total of 121 nodes and 92 edges were generated. Based on within-module connectivity (Zi) and among-module connectivity (Pi), nodes were classified as network hubs (Zi>2.5, Pi>0.6), module hubs (Zi>2.5, Pi<0.6), connectors (Zi<2.5, Pi>0.6), and peripherals (Zi<2.5, Pi<0.6; Fan et al., 2018). The results implied a sparse and loose network because 116 peripherals, five connectors, no network hub, and no module hub were identified. Co-occurred zOTUs varied from the mining sites. There were 40, 12, 4, and 1 node(s) exclusively detected in MZ, DBS, ZJS, and MYW, respectively, and unique correlations existed among these nodes. In MZ AMD, the sulfur-oxidizing bacterial genus Sulfobacillus was correlated with iron- and sulfur-oxidizing bacteria Acidithiobacillus, iron-reducing bacteria Acidibacter, and uncultured Rickettsiaceae. Rare zOTUs that were only detected in MZ were found to be related to each other. For example, uncultured CPla-3 termite group was closely related to unclassified Babeliales, uncultured Acidobacteriaceae Subgroup1, and uncultured Obscuribacter, while Turribacter was correlated with Clostridium sensu stricto 1, Romboutsia, and “Candidatus Captivus.” For zOTUs that occurred in at least two sites, correlations were observed between archaea and archaea (such as unclassified Thermoplasmatota and unclassified archaea), archaea and bacteria (such as uncultured Thermoplasmataceae and Leptospirillum, and Cuniculiplasma and Acidithiobacillus), sulfur- and/or iron-metabolizing and heterotrophic bacteria (such as sva0485 and Acidibacillus, and Ferrovum and unclassified Acetobacteraceae), and heterotrophic and heterotrophic bacteria (such as Acidisphaera and Acidiphilium).

To investigate the potential biogeochemical cycles mediated by microorganisms of different mining areas, microbial community metabolic functions were predicted using FAPROTAX. Across the four mining areas, the metabolisms associated with anaerobic and aerobic chemoheterotrophy and dark iron oxidation were dominant in the AMD environment (Supplementary Table 2 and Figure 3D). Sulfur metabolisms including oxidation of sulfur compounds, sulfur and sulfur compound respiration, and dark sulfur oxidation were predicted in the MZ, DBS, and ZJS sites but not in the MYW area. Acidithiobacillus, Leptospirillum, Ferroplasma, Acidiferrobacter, Ferrimicrobium, and Ferrithrix were responsible for iron oxidation in MZ and DBS, while Ferroplasma was the only taxa predicted as an iron-oxidizing microorganism in MYW. In MZ, Leptospirillum and Clostridium sensu stricto 1 were involved in nitrogen metabolism, including nitrogen fixation, nitrite ammonification, and nitrite respiration. Nitrogen fixation was mainly driven by Leptospirillum in DBS and ZJS.

Correlations between potential metabolism functions and environmental factors were also calculated (Figure 3D). SO₄²⁻ and Mg²⁺ were significantly positively correlated with aerobic chemoheterotrophy and dark iron oxidation but negatively correlated with dark oxidation of sulfur compounds and nitrogen fixation. Anaerobic chemoheterotrophy and sulfur respiration were positively correlated with temperature and Ca²⁺. Fermentation and nitrogen metabolisms (including nitrogen respiration, nitrite ammonification, and nitrite respiration) were significantly positively related to Fe²⁺ but negatively related to temperature.

Being mined and abandoned mines that contained varying amounts and types of metal sulfides minerals have the potential to generate acidic mine discharge water that contain high concentrations of sulfate and metals and metalloids. The physicochemical characteristics of AMD are subject to considerable variation (Johnson, 2003). In this study, the physicochemical characteristics of the AMD generated from four mining areas with different types of mines were distinct but were all extremely acidic and with high concentrations of SO₄²⁻, iron, and other toxic metal ions. Similar to previous reports on other AMD systems (Lukhele et al., 2019), Al³⁺, Mg²⁺, Cu²⁺, Mn²⁺, Zn²⁺, As, B, and Cr were detected as abundant metal species, and Ba²⁺, Cd²⁺, Li⁺, Mo²⁺, Ni²⁺, Pb²⁺, and Sr²⁺ were detected as trace elements in all AMD samples of the current study. However, AMD from copper (ZJS and MYW) and polymetallic (MZ and DBS) mining areas presented different metal concentration profiles; AMD of the copper sulfide mines contained a higher concentration of metal ions (except for Li⁺ and Sr²⁺) compared with polymetallic mines. Lower pH and higher SO₄²⁻ concentration were also observed in copper sulfide AMD. MYW and ZJS are both low-grade copper sulfide deposits; the dominant mineral of them was chalcocite, with acid-insoluble pyrite as associated minerals. AMD from them was generated during the bioheap-leaching process. In the case of acid-insoluble sulfide minerals oxidation, more ferric iron and protons are produced (Johnson, 2003). In contrast, MZ and DBS are polymetallic mines containing more complex ores, AMD from them were formed by multimineral dissolution. That might be the reason why AMD of MYW and ZJS contained higher concentrations of metal ions with lower pH compared to DBS and MZ. In practice, SO₄²⁻ concentration is an important tracer to distinguish AMD-impacted waterbodies from unimpacted waterbodies (Havig et al., 2017). Statistical analysis in the current study also indicated that SO₄²⁻ was the most important contributing factor determining the environmental divergence of AMD. Congruent with previous research (Liu et al., 2014), results from the present study demonstrated that acidity and concentrations of SO₄²⁻ and metals of AMD were largely dependent on the composition of the exposed minerals; therefore, AMD from copper sulfide mines show more extreme characteristics than that of polymetallic mines.

Community dissimilarity was further explored along environmental gradients. Mg²⁺, SO₄²⁻, Ca²⁺, temperature, and Fe²⁺ were the most contributing factors that affected the microbial community in AMD environments. In many previous studies, pH was the key determinant relating to and structuring AMD microbial communities (Kuang et al., 2013; Teng et al., 2017). pH can reflect the stages of the mineral oxidation process (Chen et al., 2013, 2014) and define the ecological niches of closely associated bacteria and archaea (Belnap et al., 2011). However, these previous studies captured a wide range of pH gradients from very low (below 2) to neutral (about 7) and thus may have underestimated and veiled the influences of other environmental factors on microbial communities (Aguinaga et al., 2018). The current study focused on AMD with a narrow range of pH at a low level (pH 0.7–2.59) and detailed the factors regulating extremely acidic AMD microbial populations. It has been known that pH has a strong influence on the dissolution of metals from ores, especially under low pH (<3.0). Accordingly, in our study, AMDs (such as MYW) contained higher concentrations of metal ions with lower pH, which might successively affect the microbial diversity. Nevertheless, microorganisms living in AMD appeared to have high resistance to highly toxic metals, which is why highly toxic metals such as copper, zinc, cadmium, and so on did not show a significant relationship with the microbial community. The research of Walton and Johnson (1992) indicated that the influence of pH change on the abundances of microorganisms was probably less significant than the change of substrate concentrations, such as Magnesium.

Mg²⁺ and Ca²⁺ were previously suggested to exert essential roles on aquatic microbial community composition in terrestrial aquifers (Fillinger et al., 2019) like hypersaline lakes (Podell et al., 2014), high-altitude Andean lakes (Núñez Salazar et al., 2020), and hot springs (Huang et al., 2013). Calcium ions participate in a variety of microbial functions, including heat shock, chemotaxis, and biodegradation. Furthermore, Ca²⁺ concentration in the aquatic environment is inextricably linked to the pH value and is related to inorganic carbon metabolism driven by microorganisms (Zavarzin, 2002). Indeed, almost all core taxa were affected by Ca²⁺ concentration in the current study. SO₄²⁻ and Fe²⁺ were also identified as important factors that contributed to the variance of microbial communities in the present study. SO₄²⁻ and Fe²⁺ were the main ions produced during the sulfide mineral oxidation process and likely reflected the degree of minerals dissolved (Teng et al., 2017). Most heterotrophic microorganisms were negatively related to SO₄²⁻concentration, and this might be why the abundance of these groups decreased significantly during the later stage of mineral oxidation. In addition, SO₄²⁻ was directly correlated with the metabolism of sulfur-oxidizing microbes (Dopson and Johnson, 2012) and Fe²⁺ concentration largely affects iron-oxidizing microorganisms (Walton and Johnson, 1992). The current study demonstrated that SO₄²⁻ and Fe²⁺ may have correlations with the nitrogen and carbon cycles driven by microorganisms in the AMD environment. This may be a direct influence since these essential geochemical processes were mainly attributed to key iron- and sulfur-oxidizing microorganisms like Leptospirillum or/and Acidithiobacillus, which are strongly affected by SO₄²⁻ and Fe²⁺. The hazardous metals such as arsenic were generally believed to have a significant influence on the microbial community structure. Nevertheless, our results showed only a few microorganisms that had strong negative correlations with dissolved arsenic (Supplementary Figure 1). Microorganisms living in AMD appeared to have high resistance with highly toxic arsenic (Laroche et al., 2018). The temperature was observed as another important contributing factor that influenced the community structure and prevalence of the dominant microbes. Temperature is an essential condition for the growth of microorganisms. Different species were reported to dominate the microbial community on ore surfaces at different temperatures during the bioleaching process (Wang et al., 2018). Furthermore, temperature-regulated microbial iron and arsenic oxidation in AMD (Tardy et al., 2018) was correlated with the microbial population and acid generation (Jia et al., 2018). Individual microbes likely have distinct functional responses to variations in temperature. Elevated temperature significantly upregulated the carbon fixation and amino-acid metabolism proteins of Leptospirillum in AMD biofilms collected from the Richmond Mine (Mosier et al., 2015). This was consistent with the findings of the current study that Leptospirillum was positively correlated with temperature. Archaea, mainly affiliated with phylum Thermoplasmatota, were previously found to be dominant in AMD with low-to-moderate temperatures and were considered significant contributors to carbon- and iron-cycling (Korzhenkov et al., 2019). In the present study, Thermoplasmatota-related taxa, including Ferroplasma and E-plasma, negatively correlated with temperature, demonstrating the potential association between carbon and iron metabolism with temperature, and that the niche of dominant taxa is divided by temperature.

The role of mineral type on AMD microbial communities was explored in the current study. AMD of polymetallic mines had higher microbial diversity than copper mines, and many rare taxa were exclusively detected in polymetallic AMD, suggested a more diverse community. Heterotrophic iron-metabolizing bacteria Acidiphilum, reported to can reduce soluble- and solid-phase ferric iron, was significantly more abundant in polymetallic AMD (12% relative abundance) than in copper AMD (relative abundance of 0.00086%). Other heterotrophic microorganisms, including Acidisphaera, an unclassified genus within Acetobacteraceae, Acidibacter, Acidibacillus, and Acidicapsa were also more prevalent in the polymetallic AMD environment. Furthermore, functional redundancies were revealed in polymetallic AMD. Low-abundance taxa might serve as the functional pool whose members are recruited during the specific acidification process to generate AMD. For example, the acidophilic iron-oxidizing microorganisms Ferrimicrobium, Ferrithrix, and Ferrovum, as well as Acidiferrobacter, were exclusively observed in polymetallic AMD (Supplementary Table 3). In addition to Leptospirillum, Clostridium detected as rare taxa in polymetallic AMD were also identified as nitrogen-metabolizing bacteria. Together, these results demonstrated that the prevalence of rare taxa in AMD was mainly influenced by the AMD mineralogy type and highlighted the contributing roles of rare taxa groups to AMD (Liu et al., 2014).

Leptospirillum, Ferroplasma, and Acidithiobacillus were identified as the core genera in AMD as they are detected globally in AMD sites and play important roles in the dissolution of sulfide minerals and the generation of AMD (Huang et al., 2016; Xiao et al., 2016). In the current study, the dominance of Leptospirillum and Ferroplasma in AMD of all four mining sites was reported, but Acidithiobacillus was only abundant in polymetallic AMD. Leptospirillum and Ferroplasma have been frequently discovered in extremely acidic environmental conditions, like biofilm floating on acidic water or growing on the surface of an ore body (Bond et al., 2000a,b; Lo et al., 2007). Leptospirillum and Ferroplasma were always found to coexist in environments where the pH was below 1 (Tyson et al., 2004; Demergasso et al., 2005). Leptospirillum is extremely acidophilic, chemolithotrophic bacteria, with the capabilities of dark iron oxidation, growth by carbon fixation via the reductive tricarboxylic acid (rTCA) cycle (Levican et al., 2008), and nitrogen fixation (Goltsman et al., 2013). The paucity of the iron uptake system may explain the dominance of these bacterial taxa in extremely low pH environments with soluble iron concentrate more than 14gL⁻¹ (Battaglia et al., 1994; Osorio et al., 2008). Ferroplasma was described as a chemomixotrophic, cell-wall-lacking, iron-oxidizing archaea and was predicted to assimilate carbon and nitrogen sources from outside the cell (Dopson et al., 2005). With a tetraether lipid membrane monolayer structure, Ferroplasma was profoundly acid-tolerant and heavy metal resistant (Golyshina and Timmis, 2005), especially it could endure more than 36gL⁻¹ iron in AMD of MYW. Acidithiobacillus is psychrophilic to mesophilic (Nordstrom et al., 2015) and thus may dominate the AMD microbial communities at lower temperatures as indicated by the results of the current study. A negative correlation with SO₄²⁻ suggested the essential role of Acidithiobacillus in the sulfur oxidation of AMD (Wang et al., 2019).

Network analysis is an efficient tool to decipher the sophisticated interactions of microbial communities and provides constructive information for experimental verification of interactions among microbes. Bioinformatic tools have been developed to calculate the relationships within a microbial community according to the variations in microbial abundance profiles. Taxa count data of the AMD microbial communities in the current study were overdispersed and had excess zeros. To fit the count data with overdispersion and excess zeros and minimize false positives of the correlations, SPIEC-EASI was applied to infer the AMD microbial ecological networks. Sparse but robust interactions were present among microbes across mining areas. There were distinct mineral-type-related correlations and specific relationships among various heterotrophic bacteria in polymetallic AMD. Rare taxa linked to mammalian diseases, including Legionella, Mycobacterium, Turicibacter, Terrisporobacter, and Romboustia, were correlated with typical AMD groups such as Metallibacterium and Sulfobacillus. It remains unclear that why these neutrophilic heterotrophs could appear in an extremely acidic environment like AMD. Previous researches highlighted the importance of Horizontal Gene Transfer events (HGT) in the environmental adaptions of some eukaryotes like Galdieria sulphuraria and prokaryotes like Acidiphilium and Alicyclobacillus to the AMD environment (Schonknecht et al., 2013; Li et al., 2020; Liu et al., 2021b). Genes acquired horizontally are involved in ecologically important processes including heat tolerance, toxic metals resistance, and metabolites transportation and metabolism, etc. Our results suggested possible HGT and/or metabolic coupling among these taxa. HGT may be a common strategy for microorganisms to adapt to the harsh and acidic environment. Distinct interactions between sulfur- and/or iron-metabolizing and heterotrophic microorganisms were also found, which implied indispensable roles of heterotrophic microorganisms in the geochemical cycles for key elements in AMD.