Metagenomic insights into sulfate-reducing bacteria in a revegetated acidic mine wasteland

The widespread occurrence of sulfate-reducing microorganisms (SRMs, which are typically considered anaerobic organisms) in temporarily oxic/hypoxic aquatic environments indicates an intriguing possibility that SRMs can prevail in continuously oxic/hypoxic terrestrial environments rich in sulfate. However, little attention has been paid to such a possibility, leading to an incomplete understanding of microorganisms driving terrestrial part of the global sulphur cycle. from in

Due to the di culty in isolating pure cultures of SRMs from the environment (especially acidic habitats) [8], as evidenced by the limited number of cultivated SRMs (approximately 240 species) [9], the functional genes aprBA and reductive dsrAB have been widely employed to explore SRMs diversity in the environment [9][10][11][12][13]. A striking example of such research was the work reported by Vigneron et al. [9]. The authors found 167,397 different species-level dsrB OTUs a liated with 47 different families through highthroughput sequencing of dsrB genes in 200 environmental samples from 14 different ecosystems. Among these OTUs, approximately 99% were previously not detected, greatly improving our knowledge of the global species-level biodiversity of SRMs.
Recently, a new trend in SRMs research is the application of genome-centric metagenomics [14][15][16][17][18]. Compared to gene-centric approaches, the most important advantage of genome-centric metagenomics lies in the recovery of near-complete genomes representing species-level microorganisms of interest from complex environments [19]. Such an advantage renders not only discovery of previously uncultured (possibly uncultivable) microorganisms of interest but also insights into principal metabolic potentials of individual microorganisms of interest and putative interspeci c interactions of microorganisms involved in a particular biogeochemical process [20][21][22]. Indeed, by retrieving and analyzing near-complete genomes containing reductive dsrAB from various environments, recent studies have revealed that: (1) eight prokaryotic phyla that were not previously reported to have the ability to reduce sulfate harbor the canonical functional genes required for dissimilatory sulfate reduction [14]; (2) two Acidobacteria genomes encoding reductive DsrAB but not Sat and AprBA contain additional genes for sul te-producing enzymes that render the relevant SRMs to use organosulfonates as a growth substrate [15]; and (3) viruses can infect SRMs in wetland sediments and thus likely affect functions of their hosts in presently unknown ways [16].
Despite the aforementioned research advances, little is known about SRMs in terrestrial environments, which can be attributed largely to the fact that currently available studies on SRMs have focused almost exclusively on aquatic environments [2]. This situation represents an incomplete understanding of microorganisms responsible for terrestrial part of the global S cycle, as the geographic distribution of soils rich in sulfate (the most important growth substrate for SRMs) is by no means restricted to aquatic environments [23]. In fact, many natural processes (e.g. prolonged droughts and sea level declines) and anthropogenic interventions (including mining operations and draining wetlands for agricultural purposes) can lead to the distribution of sulfate-rich soils in terrestrial environments [23]. A major distinction between terrestrial and aquatic sulfate-rich soils lies in better aeration and continuously oxic/hypoxic conditions associated with the former, although oxic/hypoxic conditions can temporarily exist in the latter [24]. Such discrepancy, however, should not preclude the occurrence and activity of SRMs in terrestrial sulfate-rich soils.
The functioning of SRMs in temporarily oxic/hypoxic aquatic environments had long been recognized [25,26], although SRMs are anaerobic organisms in nature. Moreover, there is considerable evidence that several Deltaproteobacteria-related model cultivated SRMs possess a variety of genes encoding enzymes conferring them a high tolerance to oxidative stress (see [24] for a review). A recent study even reported the rst SRM pure culture (i.e. Desulfovibrio vulgaris strains) that can grow using energy derived from oxygen reduction [27]. In contrast, existing literature provided only a few lines of evidence demonstrating the occurrence and activity of SRMs in terrestrial sulfate-rich soils under continuously oxic/hypoxic conditions (as indicated by the associated Eh values ranging from approximately 100 to 700 mV), which are related to sulfate-generating mine tailings deposited onto the surface soils of several Canadian mining areas (see [28] for a review). An attempt was made to characterize the community composition of SRMs in such mine tailings using Bacteria domain-, Deltaproteobacteria class-and Desulfotomaculumspeci c primers, showing that the majority of potential SRMs were a liated to Firmicutes and Deltaproteobacteria [29]. Additionally, two Acidobacteria genomes retrieved from an acidic sul de mine waste rock site in Canada were found to encode canonical enzymes required for sulfate reduction [14]. However, no information regarding the geochemical properties of the site was available in the literature [14], which greatly limits our ability to exploit them to get a better understanding of the principal metabolic potentials (e.g. oxygen defense mechanisms) of SRBs in continuously oxic/hypoxic terrestrial environments. Nonetheless, more work is urgently needed, considering the importance of the topic.
In this study, we employed genome-centric metagenomics to characterize SRMs in a revegetated acidic mine wasteland rich in sulfate. Before revegetation, the wasteland (i.e. a mine tailings pond) was abandoned and drained for approximately eight years. This mine site was selected, as it consisted of different habitats with soil Eh values varying from approximately 180 to 680 mV [30], a representative Eh range of various terrestrial environments under continuously oxic/hypoxic conditions [31]. We recovered 12 Acidobacteria and four Deltaproteobacteria genomes encoding reductive DsrAB, of which ve are members of three genera without previously known SRMs. Comparative genomic analysis revealed that the major metabolic potentials of Acidobacteria-related SRMs were considerably different from those of Deltaproteobacteria-related SRMs. Most importantly, we found for the rst time that SRM-infecting viruses harbor auxiliary metabolic genes (AMGs) encoding proteins dedicated to glycoside hydrolysis, chemotaxis and antioxidation of their hosts.

Study site and soil sampling
We selected a revegetated acidic mine wasteland located in southern China (29°40′52″N, 115°49′21″E) as our study site. Brie y, this site was revegetated in spring of 2013 and thereby consisted of three different habitats: amended layer of the revegetated tailings (0-10 cm, ALRT), unamended layer of the revegetated tailings (11-20 cm, ULRT) and unrevegetated tailings (UT). These habitats were rich in sulfate (1.80-25.9 g SO 4 2− kg − 1 dry soil) and were under continuously oxic/hypoxic conditions (as indicated by a soil Eh range of approximately 180-680 mV). Three independent soil samples were collected from each of these habitats in July 2016 and 2017, respectively. More details on the study site, soil sampling and soil physic-chemical properties were presented elsewhere [30].
DNA extraction and sequence processing DNA extraction, metagenomic sequencing and data processing (including assembly, binning, re nement, genome completion estimates, gene prediction, etc.) were described in detail elsewhere [30]. Metagenomic data analyzed in this study were deposited at EMBL under accession number PRJEB31441 [30] and the genome bins reported in this study have been deposited in GenBank under accession numbers SAMN15699825 and SAMN15808056-70.

Retrieval of key genes involved in dissimilatory sulfur metabolism
Genome-speci c metabolic potential for sulfate/sul te reduction was determined as follows. All predicted ORFs in a genome bin were searched against eggNOG [32] and KEGG [33] databases using Diamond [34] and against HMM pro les using InterProScan [35]. Then, the key sulfate reduction/sulfur oxidation genes (dsrAB, dsrD, dsrT, dsrMKJOP, aprAB, sat and dsrEFH) in genome bins were identi ed based on conserved domain hits elaborated in Anantharaman et al. [14].

Phylogenetic analysis of DsrAB proteins
A total of 214 DsrAB sequences, including those from previous studies [14,36,37], were used for phylogenetic analysis. The DsrAB sequences were aligned using MUSCLE [38] with default parameters. The alignments were then ltered by TrimAL [39] with the parameters -gt = 0.95 and -cons = 50. The concatenated DsrAB tree was constructed using RAxML [40] with the parameters set as -f a -m PROTGAMMAIJTT -p 12345 -x 12345 -# 100. The Newick les with the best tree topology were uploaded to the Interactive Tree of Life (iTOL) online interface [41] for visualization and formatting.
Taxonomic classi cation of reductive dsrAB-containing genome bins 16 genome bins retrieved in our study harbored reductive dsrAB sequences (Table S1). The direction of dissimilatory sulfur metabolism for each genomic bin was determined, according to the rules elaborated in Anantharaman et al. [14]. Taxonomic classi cation of the 16 SRM genome bins was inferred from two phylogenetic trees constructed with the reference genomes using GTDB-Tk [42] and PhyloPhlAn [43]. For Acidobacteria subdivision-level classi cation, 12 genome bins from Acidobacteria recovered in our study were used for phylogenetic analysis with published reference genomes spanned subdivisions 1, 3, 4, 6, 8 and 23 [15,44]. One genome bin from Syntrophobacteraceae family without genus-level classi cation was used for phylogenetic tree construction with public reference genomes from Syntrophobacteraceae downloaded from NCBI. The maximum-likelihood phylogenetic trees were constructed based on a concatenated dataset of 400 universally conserved marker proteins using PhyloPhlAn and visualized using iTOL.

Calculation of relative abundances of genomic bins
Relative abundances of the 16 genome bins were calculated as previously described [30]. Brie y, the highquality reads from each genomic dataset were mapped to all of the dereplicated genome bins using BBMap with the parameters k = 14, minid = 0.97, and build = 1. The coverage of the genome bin was calculated as the average scaffold coverage, and each scaffold was weighed by its length in base pairs.
Then, the coverage of each genome bin divided by the total coverage of all genome bins in each sample was considered as its relative abundance.

Selection of genome bins for metabolic potential analysis
Genome bins with a completeness > 90% and contamination < 10% were chosen for further metabolic potential annotation, including six Acidobacteria genomes and one Deltaproteobacteria genome.  [49]. The reference sequences were aligned using MUSCLE with default parameters, then the alignment was converted into Stockholm format and databases were built using hmmscan [50]. The noise cutoffs for individual HMM pro le were determined by manual inspection. Protein sequences that showed best hit with the HMM pro les with 1) bit-score greater than the calibrated threshold, and 2) over 90% sequence coverage, were retained.

Identi cation of proteins involved in respiration
All predicted ORFs in the selected genomes were searched for proteins involved in four respiratory complexes based on eggNOG annotation results, including NADH dehydrogenase, succinate dehydrogenase, quinol-cytochrome-c reductase, terminal oxidase and ATP synthase [15].

Identi cation of proteins involved in chemotaxis and oxidative stress
Besides methyl-accepting chemotaxis proteins (MCPs), the central components of bacterial chemotaxis system include CheA, CheB, CheR CheW and CheY [51]. MCPs were identi ed by Pfam annotation hits to PF00015, while the other central protein sequences were identi ed by KEGG annotation hits and were further con rmed based on eggNOG annotations. Classi cation of MCPs was according to Ud-Din & Roujeinikova [52]. Proteins involved in the core agellum [53] and type IV pilus [54] systems were identi ed by KEGG annotation hits and were further con rmed according to eggNOG annotation results.
The antioxidative enzymes analyzed in this study were selected based on two previous reviews [24,55] and were identi ed by KEGG, eggNOG and InterPro annotations.

Recovering and annotating viral scaffolds
VirSorter [56] was used to recover viral scaffolds from the 16 SRM genome bins as well as the two SRM reference genomes. The scaffolds from VirSorter categories 1, 2, 4 and 5 were retained. For scaffolds with predicted proviruses, only predicted proviral regions were retained. To taxonomically classify the viral scaffolds, a gene content-based network analysis was performed to cluster viral scaffolds into viral clusters at approximately the genus level, using vConTACT2 with ProkaryoticViralRefSeq94 database [57]. The ORFs in viral scaffolds were predicted with MetaProdigal. Virus signature proteins like terminase, integrase, capsid and tail were identi ed by Pfam hits. Viral sequences that encoded tail genes could be tentatively assigned to the order Caudovirales.

Estimation of viral-host abundances
Host (Acidobacteria-speci c SRM) abundance and the abundance of virus for that host were both calculated as the normalized mean coverage depth as described elsewhere [58]. Brie y, the high-quality reads from each metagenomics dataset were mapped to all of the dereplicated viral scaffolds or dereplicated genome bins using BBMap with the parameters k = 14, minid = 0.97, and build = 1. The viral or host abundances were pulled from the BBMap mapping coverage output, normalized by the number of metagenomic reads in each sample, respectively. Pearson correlation analysis was used to correlate the viral and host abundances via the vegan package within the R statistical computing environment.

Identi cation of viral AMGs
To examine the potential roles of SRM-infecting viruses in S biogeochemistry, we assessed whether they contained AMGs. The predicted viral proteins were searched against dbCAN2 meta server and with HMM pro les using InterProScan as described above. For GHs, MCPs and nickel-containing superoxide dismutase (Ni-SOD) encoded by AMGs identi ed in this study, their protein sequences were structurally modeled using PHYRE2 in normal modeling mode (http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi? id=index) to con rm and further resolve functional predictions.
The 12 Acidobacteria genome bins were all a liated to subdivision 1 of Acidobacteria ( Figure S2). Among them, three (UT3_2.bins.71, UT3_3.bins.87 and UT4_3.bins.137) formed a monophyletic clade and had average 63% AAI to their closet relative Granulicella tundricola MP5ACTX9. Similarly, ULRT4_2.bins.48 formed a monophyletic clade and had 56%-62% AAIs to its closest relatives. These four genomes represented two genera not previously reported to contain SRMs, given that no SRMs from Acidobacteria have been successfully cultivated and that currently known Acidobacteria-related genome bins containing dsrAB [15] were a liated to other genera ( Figure S2).
Three out of the four genome bins from Deltaproteobacteria were a liated to the well-known SRM genus Desulfovibrio (Table S1). In contrast, another one (i.e. ULRT4_2.bins.61) can be only assigned to Syntrophobacteraceae and formed a monophyletic clade ( Figure S3). It had 54-60% AAIs to its closest relatives. Therefore, we inferred that it belonged to a new genus.
The Dsr operon structures of Acidobacteria were different from those of Deltaproteobacteria (Fig. 1). Multiple alignments of DsrD and DsrT sequences with published references con rmed highly conserved residues ( Figures S4 & S5), indicating that these proteins are likely active. According to the rules for determination of direction of dissimilatory sulfur metabolism for uncultivated microorganisms [14], 11 genome bins (eight from Acidobacteria and three from Deltaproteobacteria) recovered in our study encoded the complete pathway for reduction of sulfate to sul de ( Fig. 1 & Table S2). Notably, seven genome bins (six Acidobacteria and one Deltaproteobacteria, Table S1) had a completeness > 90% and contamination < 10%.
Glycoside hydrolysis of SRMs 71 GH families were encoded by the nine SRM genomes considered in this study (Table S3). Among them, only one (i.e. GH 50) was not found in the six near-complete Acidobacteria-related genomes. This result is reasonable, as GH50 currently consists of β-agarase (EC 3.2.1.81), which is responsible for hydrolysis of (1→4)-β-D-galactosidic linkages in agarose (a polysaccharide produced by some aquatic red algae) [59]. In contrast, the three Deltaproteobacteria-related genomes encoded only 10 GH families. The numerical predominance of GH families observed for Acidobacteria-related genomes is remarkable even as compared to the results of Eichorst et al. [44] who identi ed 131 GH families in 24 non-SRB Acidobacteria-related genomes. According to the average number of GH genes (per genome), GH3, GH13, GH23, GH2, GH31, GH29, GH28, GH27, GH92 and GH35 were the top 10 most abundant ones across all investigated genomes (Fig. 2). Except for GH23, these abundant GH families were largely represented by Acidobacteria-related genomes. A striking example was GH3, which was encoded by 9-13 genes in each Acidobacteria-related genome but by only one gene in per Deltaproteobacteria-related genome (Fig. 2).

Hydrogen metabolism of SRMs
Genes encoding eight groups of hydrogenases (including Group A1, A2 of [FeFe]-hydrogenase and Groups 1a, 1b, 1d, 3d, 4c, 4e of [NiFe]-hydrogenase) were identi ed in this study (Table S4). Among them, [FeFe]-hydrogenase was encoded only by D. vulgaris. These results seem to agree with those of Hausmann et al. [15] who showed that genome bins of Acidobacteria-related SRMs harbored genes encoding Groups 1 (excluding 1 h), 3, and 4 of [NiFe] hydrogenase. When individual genomes were taken into account, they differed considerably in the total number of hydrogenase genes (  [49] were identi ed in the remaining four Acidobacteria genomes (Fig. 2).

Respiratory chain of SRMs
All investigated genomes encoded the major components of respiratory chain (Table S5) Figure S6) that contained two experimentally validated redox and oxygen sensors (i.e. DcrA and DcrH) [52], acidobacterial MCPs were mainly from classes IVa and IVb. More speci cally, three out of the ve Acidobacteria genomes lacked genes encoding class Ia MCPs (Table S7).

Flagellum system of SRMs
The entire set of 24 core agellar genes [53]

Viruses of SRMs
Six viruses (prophages) were identi ed across the 12 acidobacterial genome bins recovered in this study ( Fig. 3 & Table S9), while no virus sequences were detected in the four deltaproteobacterial genome bins. Seven and three viruses were found in D. vulgaris and D. multivorans respectively, which seemed to be not in complete agreement with previous ndings that eight and no viruses were identi ed in the two model SRMs correspondingly [47,61]. This discrepancy likely resulted from the utilization of different viral prediction methods. Notably, PFAM annotations revealed that 11 out of the 16 viruses identi ed in this study harbored at least one virion-associated gene (Table S9), suggesting that these viruses still have the genetic potential to complete a lytic cycle [44].
Most of the identi ed 16 viruses could not be clustered with isolated viruses or those identi ed in publicly available microbial genomes or metagenomes using a gene-content based classi cation (genus-level grouping) [57], although half of them could be tentatively assigned to the order Caudovirales (Table S9). Speci cally, four acidobacterial viruses formed an exclusive cluster, while the remaining two acidobacterial viruses were not closely related at the nucleotide level to any previously sequenced bacteriophages (i.e. singletons [62]; Table S9 & Fig. 3). Similarly, among the viruses of D. vulgaris and D. multivorans, one was a liated to Myoviridae family, three were clustered exclusively, and the remaining six were singletons (Table S9 & Fig. 3). Additionally, the abundances of the viruses targeting acidobacterial SRMs were positively correlated with their host abundances ( Figure S8).

Discussion
To our knowledge, this work represents the rst genome-centric metagenomic study that was speci cally aimed at characterizing SRMs in a continuously oxic/hypoxic terrestrial environment. It seems somewhat surprising to nd that SRMs a liated to Acidobacteria prevailed in the mine wasteland ( Fig. 1), since the limited evidence currently available indicates the dominance of Firmicutesand Deltaproteobacteriarelated SRMs in two other similar mine wastelands [29]. Nonetheless, the recovery of seven near-complete SRM genomes (six Acidobacteria and one Deltaproteobacteria) in this study provided us a particular opportunity to explore the principal metabolic potentials of these SRMs and their viruses, with emphasis on a comparison between Acidobacteria and Deltaproteobacteria. To this end, genomes of two Deltaproteobacteria-related model cultured SRMs with contrasting tolerance to oxygen stress (i.e. oxygentolerant D. vulgaris and oxygen-sensitive D. multivorans) were included [45].

The revegetated mine wasteland fostered novel SRM genera
The occurrence of SRMs in Acidobacteria was unveiled very recently [14]. So far, only 12 genomes of SRMs in Acidobacteria (with a completeness varying from 29.7-98.0%) have been reported [14][15][16]. In this context, this study not only doubled the number of such genomes but also detected the genus-level taxonomic novelty of Acidobacteria-related SRMs (Fig. 1). In addition, an unexpected nding of this study is that one Deltaproteobacteria-related SRM genome (i.e. ULRT4_2.bins.61) represents a genus without previously known SRMs, given that SRMs in Deltaproteobacteria were investigated extensively [2]. It thus becomes obvious that more work is needed to achieve a 'complete understanding' of the taxonomic diversity of SRMs, although recent studies have greatly expanded the diversity of SRMs at the OTU and phylum levels [9,14].

Acidobacteria-related SRMs encoded more GHs and oxygen-tolerant hydrogenases than Deltaproteobacteriarelated SRMs
Many members of Acidobacteria are thought to have the ability to use a wide range of carbohydrates, as they dedicate a large portion of their genomes to carbohydrate metabolism [44]. Furthermore, there is evidence that three DsrAB-encoding Acidobacteria genomes recovered from peatland sediments harbor more GH genes (on average approximately 105 genes per genome) than not only the other known SRMs but also the majority of non-SRM Acidobacteria (approximately 60 genes per genome) [15]. Coincidently, the average number of GH genes in the six near-complete DsrAB-encoding Acidobacteria genomes of this study was up to 120 per genome (Table S3), being much greater than that of the three Deltaproteobacteria-related SRMs. Our results indicate apparent adaptation of these Acidobacteria-related SRMs to the oligotrophic conditions of this mine wasteland [30]. Note that the most prevalent GH family across these Acidobacteria genomes is GH3 (Fig. 2 . Two near-complete Acidobacteria genomes seemed to lack genes encoding such hydrogenases (Fig. 2), suggesting their inability to couple sulfate reduction to oxidation of H 2 . This characteristic likely puts the SRMs at a disadvantage in competition for growth substrates with co-existing microorganisms [49] and thus provides a possible explanation for the fact that they occurred in the wasteland at a lower relative abundance as compared to most of the other ve SRMs possessing such genes (Table S1). Indeed, the hydrogenases encoded by the other four near-complete Acidobacteria genomes are oxygen-tolerant (i.e. 1d and/or 3d of [NiFe]hydrogenase, Fig. 2) [49]. This seems to be in consistence with a previous nding that seven Acidobacteria genome bins, which can be putatively assigned as facultative anaerobes or aerobes (as they have low-a nity terminal oxidase genes) [68], all encoded oxygen-tolerant hydrogenases (i.e. 3b and/or 3d of [NiFe]-hydrogenase) [15]. Note, however, that D. multivorans rather than the other two Deltaproteobacteria-related SRMs also encoded an oxygen-tolerant hydrogenase (Fig. 2), which is different from our expectation. It is currently unclear what causes this discrepancy. Nonetheless, the speci c functions of oxygen-tolerant hydrogenases in individual SRMs belonging to Acidobacteria and Deltaproteobacteria deserve further research.
Acidobacteria-related SRMs differed from Deltaproteobacteria-related SRMs in oxygen defense The presence of genes encoding high-a nity terminal oxidases in all seven near-complete SRM genomes of this study (Table S5) suggests that these SRMs have the potential for microaerobic metabolism [68].
However, such genes are also present in the genome of the oxygen-sensitive model cultured SRM D. multivorans (Table S5), highlighting that elaborate antioxidant strategies are additionally required for the persistence of these seven SRMs in the mine wasteland.
According to the current knowledge of antioxidant mechanisms in SRMs, which are based largely on data from several Deltaproteobacteria-related cultured species, peculiar behaviors (including aggregation and aerotaxis) and antioxidant enzymes (e.g. SOD and SOR) are the two major antioxidant strategies [24]. MCPs and agellum are indispensable components of the behavioral strategy [24]. The incompleteness of MCP and agellum systems in D. multivorans (Fig. 2) is thus consistent with its sensitivity to oxygen, while the opposite is true for D. vulgaris. As expected, the near-complete Deltaproteobacteria-related SRM genome of this study encodes complete MCP and agellum systems (Fig. 2). Moreover, the number (up to 34) and diversity of MCP genes on it seem to be higher than those of D. vulgaris (Table S7). On the other hand, it is actually somewhat surprising to nd that the six near-complete Acidobacteria-related SRM genomes of this study harbor very few or even no MCP genes (Fig. 2). This surprising nding raises two possibilities: (1) the behavioral antioxidant strategy is insigni cant for these Acidobacteria-related SRMs; and (2) there exist currently unknown proteins that have similar functions to MCPs. Note that the former possibility received little support from our results, as the Acidobacteria-related genomes encoding an incomplete MCP or agellum system (i.e. ULRT3_2.bins.110, UT4_3.bins.137 and UT3_2.bins.71; Fig. 2) occurred in the wasteland at a lower relative abundance than their relatives with the ability to encode complete MCP and agellum systems (Table S1).
SOD and SOR are thought to play an important role in scavenging superoxide ions in periplasm and cytoplasm of SRMs, respectively [24]. In agreement with this traditional wisdom, we found that genes encoding SOD are widespread across all SRM genomes considered in this study (Fig. 2). However, our results showed that genes encoding SOR are absent in the six near-complete Acidobacteria-related SRM genomes, despite their presence in those of Deltaproteobacteria-related SRMs. Similarly, most of the Acidobacteria-related SRM genomes lack genes encoding catalase (one major enzyme responsible for hydrogen peroxide elimination), although they all encode NPX (the other major enzyme responsible for hydrogen peroxide elimination; Fig. 2) [24]. These ndings indicate a remarkable difference between the Acidobacteriaand Deltaproteobacteria-related SRMs in enzyme-based antioxidant strategy. It is likely that the Acidobacteria-related SRMs prefer to reduce oxygen to water before the powerful oxidant generates various types of damages in cytoplasm, as they tend to have more genes encoding CCO (i.e. a type of high-a nity terminal oxidases that are also involved in antioxidative defense) [24] than the Deltaproteobacteria-related SRMs (Fig. 2).

Virus infection in SRMs was widespread
Virus infection in SRMs was rstly recognized by Heidelberg et al. [46] who performed a whole-genome sequence analysis of D. vulgaris. Martins et al. [16] extended this work by documenting that Acidobacteria-, Candidatus Aminicenantes-, Chloro exi-, Deltaproteobacteria-, Nitrospiraeand Planctomycetes-related SRM genomes retrieved from wetland sediments are hosts of viruses. In consistence with these previous studies, we found that ve Acidobacteria-related SRM genomes recovered from the wasteland and the genome of D. multivorans were infected at least by one virus (Fig. 3). Taken together, the abovementioned ndings indicate that virus infection in SRMs is more widespread than previously thought. Additionally, the Acidobacteria-speci c virus/host abundance ratio recorded in this study ( Figure S8) appeared to be greater than that for the phylum Acidobacteria in soils collected worldwide [58], which may be attributed to a scenario that the incidence of lysogeny in oligotrophic environments (e.g. mine wastelands) is high [69]. On the other hand, most of the viruses infecting SRMs in this study cannot be taxonomically assigned or even are not closely related at the nucleotide level to any known sequenced viruses (Fig. 3), supporting a notion that the diversity of environmental viruses is largely unexplored [16].

SRM-infecting viruses contributed to glycoside hydrolysis of their hosts
Viruses are widely thought to modulate S biogeochemical cycle in aquatic environments, as aquatic viruses can harbor several AMGs (i.e. rdsrA, rdsrC or dsrC) encoding enzymes directly involved in dissimilatory S oxidation or reduction [70,71]. In contrast, little is known about the potential roles of viruses in S biogeochemical cycle in terrestrial environments. In this study, we failed to nd viral AMGs encoding enzymes directly responsible for dissimilatory S oxidation or reduction. However, we identi ed two viral AMGs encoding enzymes dedicated to the oxidation of organic compounds (i.e. glycoside hydrolysis; Fig. 4 & Table S10). These results are remarkable, given that no previous studies have documented such metabolic potentials of viruses hosted by SRMs and that the oxidation of organic compounds is coupled to the reduction of sulfate in SRMs [1,2]. In a wider context, a recent analysis of viral community composition and metabolic potential in mangrove sediments has speculated that viral carbohydrate AMGs may facilitate hosts to obtain energy for growth by decomposing complex carbohydrates in soil ecosystems [72]. Our nding on the viral AMG encoding D-4,5-unsaturated βglucuronyl hydrolase allows us to provide a novel mechanistic explanation for such an interesting notion, as this enzyme can degrade plant cell-wall-derived oligosaccharides into rhamnose (Fig. 4) [63], which is accessible directly to SRMs for dissimilatory sulfate reduction [66]. As to the endochitinase encoded by viruses in D. vulgaris, its products seem uncoupled with sulfate reduction in the host (Fig. 4) [66]. However, some of its products, such as chitobiose, can be metabolized further by the host for use in cell wall biogenesis [73]. Such a kind of synergism should be advantageous for viral hosts in oligotrophic environments, shedding some light on the 'black box' of soil virus-host interactions [69].

SRM-infecting viruses participated in chemotaxis and antioxidation of their hosts
Going beyond reporting viral GH genes, we found that three viruses infecting the two model cultured SRMs a liated to Deltaproteobacteria are able to encode enzymes involved in chemotaxis or antioxidation (Fig. 4). There is only one previous study that documented viral AMGs encoding MCP [74]. We extended this work by identifying a speci c host of such AMGs for the rst time (Fig. 4). Intriguingly, our in-depth analysis of the viral MCPs recorded in this study revealed that they possessed an extracellular LBD for C2/C3 carboxylates or alanine/lactate (Table S11). These characteristics of the viral MCPs may endow the host with a survival advantage in oligotrophic environments, considering that most of their targeted ligands are growth substrates for SRMs (Fig. 4)

Conclusions
Our results showed that members of Acidobacteria and Deltaproteobacteria were major microorganisms responsible for dissimilatory sulfate reduction in the revegetated acidic mine wasteland under continuously oxic/hypoxic conditions. Among them, ve are novel SRMs at the genus level. Acidobacteria-related SRMs tend to encode more GHs, oxygen-tolerant hydrogenases and cytochrome c oxidases than Deltaproteobacteria-related SRMs, while an opposite trend was found for MCPs. More interestingly, a viral AMG associated with an Acidobacteria-related SRM is dedicated to liberation of rhamnose from plant cell-wall-derived oligosaccharides. Additionally, those viruses infecting Deltaproteobacteria-related SRMs can contribute to chemotaxis and antioxidation of their hosts by encoding MCPs and an antioxidative enzyme. Therefore, this study does not only improve our understanding of microorganisms driving dissimilatory sulfate reduction in terrestrial environments under continuously oxic/hypoxic conditions but also provides the rst evidence for roles of viruses in S cycle in terrestrial ecosystems. Looking more widely, this study sheds some light on the 'black box' of soil virushost interactions as well. Analysis of dissimilatory sulfate reduction genes on microbial genomes (bins) retrieved from the revegetated mine wasteland. Bins with a completeness > 50% and contamination < 10% are shown. Functions of the bins in sulfate reduction were predicted according to the presence and/or absence of those key genes for the pathway. Representative organization of dissimilatory sulfate reduction genes on the bins belonging to Acidobacteria and Deltaproteobacteria is displayed, respectively. More details were provided in Table S1 and Table S2. Analysis of dissimilatory sulfate reduction genes on microbial genomes (bins) retrieved from the revegetated mine wasteland. Bins with a completeness > 50% and contamination < 10% are shown.
Functions of the bins in sulfate reduction were predicted according to the presence and/or absence of those key genes for the pathway. Representative organization of dissimilatory sulfate reduction genes on the bins belonging to Acidobacteria and Deltaproteobacteria is displayed, respectively. More details were provided in Table S1 and Table S2.

Figure 2
Comparative analysis of selected metabolic potentials of sulfate-reducing microorganisms (SRMs). The wasteland-borne genomes (whose names are in blue) with a completeness > 90% and contamination <  Comparative analysis of selected metabolic potentials of sulfate-reducing microorganisms (SRMs). The wasteland-borne genomes (whose names are in blue) with a completeness > 90% and contamination < 10% were included in the analysis. For comparison, the complete genomes of Desuifovibrio vulgaris Hildenborough (an oxygen-tolerant model cultured SRM) and Desulfococcus multivorans DSM 2059 (an oxygen-sensitive model cultured SRM) were chosen. The top 10 most abundant glycoside hydrolase families across the genomes are shown. Those hydrogenase subgroups that are known to be involved in sulfate reduction or to be oxygen-tolerant are listed. MCP, methyl-accepting chemotaxis protein; BCP, bacterioferritin comigratory protein; Cbo, cytochrome bd oxygen reductase; Cco, cytochrome c oxidase; CcPx, cytochrome c peroxidase; SOD, superoxide dismutase; CysK, cysteine synthase; GPX, glutathione peroxidase; NPX, NADH peroxidases; ROO, rubredoxin-oxygen oxireductase; SOR, superoxide reductase; TPX, thioredoxin peroxidase. More details were presented in Tables S3-S8. to the lack of abundance information, the two model cultured SRMs and their viruses are shown in blue and yellow, respectively. *, calculated as the normalized mean coverage depth. More details were presented in Table S9. proportional to the predicted viral genome sizes, using a size scale smaller than that for their hosts.
Those hexagons lled with the same number were a liated to the same viral cluster. Singletons represent novel viruses. Increasing abundances of the acidobacterial viruses are indicated by darker red colors. Due to the lack of abundance information, the two model cultured SRMs and their viruses are shown in blue and yellow, respectively. *, calculated as the normalized mean coverage depth. More details were presented in Table S9.