Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Capability for arsenic mobilization in groundwater is distributed across broad phylogenetic lineages

  • Robert E. Danczak,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Department of Microbiology, Ohio State University, Columbus, OH, United States of America

  • Michael D. Johnston,

    Roles Investigation, Methodology

    Affiliation School of Earth Sciences, Ohio State University, Columbus, OH, United States of America

  • Chris Kenah,

    Roles Investigation, Methodology

    Affiliation Ohio Environmental Protection Agency, Columbus, OH, United States of America

  • Michael Slattery,

    Roles Investigation, Methodology

    Affiliation Ohio Environmental Protection Agency, Columbus, OH, United States of America

  • Michael J. Wilkins

    Roles Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Department of Soil and Crop Sciences, Colorado State University, Fort Collins, CO, United States of America


Despite the importance of microbial activity in mobilizing arsenic in groundwater aquifers, the phylogenetic distribution of contributing microbial metabolisms is understudied. Groundwater samples from Ohio aquifers were analyzed using metagenomic sequencing to identify functional potential that could drive arsenic cycling, and revealed mechanisms for direct (i.e., Ars system) and indirect (i.e., iron reduction) arsenic mobilization in all samples, despite differing geochemical conditions. Analyses of 194 metagenome-assembled genomes (MAGs) revealed widespread functionality related to arsenic mobilization throughout the bacterial tree of life. While arsB and arsC genes (components of an arsenic resistance system) were found in diverse lineages with no apparent phylogenetic bias, putative aioA genes (aerobic arsenite oxidase) were predominantly identified in Methylocystaceae MAGs. Both previously described and undescribed respiratory arsenate reduction potential via arrA was detected in Betaproteobacteria, Deltaproteobacteria, and Nitrospirae MAGs, whereas sulfate reduction potential was primarily limited to members of the Deltaproteobacteria and Nitrospirae. Lastly, iron reduction potential was detected in the Ignavibacteria, Deltaproteobacteria, and Nitrospirae. These results expand the phylogenetic distribution of taxa that may play roles in arsenic mobilization in subsurface systems. Specifically, the Nitrospirae are a much more functionally diverse group than previously assumed and may play key biogeochemical roles in arsenic-contaminated ecosystems.


Arsenic contamination of groundwater is a pressing health issue throughout the world [1]. For example, elevated arsenic concentrations pose a threat to over 40 million people in Bangladesh and eastern India alone, particular those living in Bangladesh [2,3]. Meanwhile in the US, a recent study revealed that 20 of 37 principal aquifers contained arsenic concentrations above the maximum contaminant level of 10 μg L-1, affecting approximately 2.1 million [4,5]. Arsenic can occur in a range of redox states (-3, 0, +3, and +5), with mobile arsenite (As3+) and arsenate (As5+) the most common [6] in groundwater systems [7]. In this context, arsenite is more soluble and affected by sorption than arsenate [6] and presents a greater risk for human consumption.

The mobility of arsenic in these aquifer systems is at least partly mediated by microbial metabolism, via both direct and indirect mechanisms. Bacteria from diverse phylogenetic lineages feature resistance mechanisms that reduce and transport arsenic, such as the one encoded by the ars operon, where arsB is a membrane bound efflux pump and arsC is a cytoplasmic arsenate reductase [811]. Alternatively, a narrower phylogenetic distribution of microorganisms can participate in dissimilatory arsenic reduction via the arr system, utilizing arsenate as a terminal electron acceptor and converting it to arsenite [811]. Bacteria can also immobilize arsenic using arsenite oxidation mechanisms, such as the anaerobic (arx) and aerobic (aio) oxidase systems. Arsenic is further affected by the biogeochemical cycles of other inorganic groundwater constituents. The reductive dissolution of iron oxides through either direct microbial activity [1214] or indirect biogenic sulfide production [15] can lead to release of adsorbed arsenic species. Although some previous studies have suggested that microbial sulfate reduction could immobilize arsenic through co-precipitation of sulfide-arsenic-iron species [1618], other research has indicated increased arsenic mobilization associated with the formation of thioarsenic species that are more soluble and less likely to adsorb to iron minerals [19].

Despite our knowledge of microbial mechanisms that contribute to arsenic cycling, the phylogenetic distribution of this functional potential is less well understood. The recent application of metagenomic tools to shallow subsurface microbial populations revealed greater phylogenetic distribution of taxa involved in both nitrogen [20] and sulfur [21] cycling than was previously appreciated. Here, we applied similar tools to investigate the microbial potential for catalyzing arsenic transformations in shallow aquifers across three counties in central and southern Ohio (Athens, Greene, Licking). Geochemical and mineralogical analyses had previously suggested that groundwater in each county was at risk for elevated arsenic concentrations [7,22,23]. Leveraging metagenomic datasets, we investigated the relationship between the functional potential across diverse bacterial lineages, and elevated arsenic concentrations. Our data suggests that arsenic mobilization potential is found in groundwater ecosystems regardless of current geochemical conditions. Moreover, our results demonstrate that the potential for arsenic and iron mobilization is broadly distributed across the bacterial tree of life, with current analyses potentially missing many microbial groups capable of these transformations.


Sample collection

Groundwater samples were collected from three groundwater wells operated by the Ohio Department of Natural Resources (ODNR) in three different counties. These three wells are located within separate buried valley aquifers, consisting mostly of glacial sands and gravels, and some till. The observation wells were sampled on a quarterly basis over a two-year period from July 2014 to July 2016. One private drinking water well, located in a sand and gravel aquifer within a thick till sequence in western Licking County, was also sampled once in June 2016 [24].

Groundwater wells were sampled as previously reported [24,25]. Briefly, they were purged of more than 250 L of water to ensure that aquifer-derived water was being sampled (dedicated pumps were placed at the screened interval for the ODNR wells). Approximately 38L of post-purge groundwater was pumped sequentially through a 0.2 μm then 0.1 μm Supor PES Membrane Filter (Pall Corporation, NY, USA). Filters were then immediately flash frozen in an ethanol-dry ice bath, and kept on dry ice before being stored at -80°C at Ohio State University.

DNA extraction, sequencing and processing

DNA was extracted from roughly a quarter of each 0.2 μm Supor PES membrane filter by using the Powersoil DNA Isolation Kit (MoBio Laboratories, Inc., Carlsbad, CA, USA). Final DNA concentrations were determined by using a Qubit Fluorometer (Invitrogen, Carlsbad, CA, USA).

Metagenomic data for 8 samples (filters from July 2014, Oct. 2014 and April 2016 for Greene and Athens, Oct 2014 for Licking, and the Licking–Private sample) was collected by shotgun sequencing on an Illumina HiSeq 2500 at the Genomics Shared Resource at the Ohio State University. Raw reads were trimmed and filtered based upon read quality using sickle pe with default parameters [26]. Resulting reads were subsequently assembled into larger contigs and then scaffolds using idba_ud with default parameters [27]. Assembly statistics are listed in S1 Table.

Whole metagenome analysis

Arsenic and sulfur gene analysis.

Each assembly was searched for a series of functional marker genes related to arsenic mobilization. All metagenomes were gene-called and translated using Prodigal with default parameters [28]. For genes encoding proteins which directly act upon arsenic (i.e. arsB, arsC, aioA, and arrA) [11], previously sequenced genes obtained from NCBI (S1 File) were used to create a BLAST database against which metagenomes could be searched [29]. Any gene that hit a previously identified sequence with an e-value of 1x10-40 or lower was considered a match. However, due to the interrelated nature of molybdopterin proteins, sequences identified as arrA or aioA underwent further analysis. Sequences potentially matching these two genes were subsequently aligned to other molybdopterin proteins (including NapA, DMSO reductase, etc.; S1 File) using MUSCLE with default parameters [30]. Alignments were then trimmed using Geneious v9.1.5 [31] to remove regions of at least 95% gaps and used to generate a RAxML tree with 100 bootstraps and an evolutionary model determined by ProtTest [3234] (raxmlHPC-PTHREADS -f a -m ‘ProtTest-Output’ -n output -N 100 -p 1234 -s file.phy -x 1234 -T 20; ‘ProtTest-Output’ for arrA and aioA was WAG while all other tress were LG). Trees were visualized with R using ggtree [35] and only those sequences which fell into the respective arrA or aioA clades were analyzed further. The marker gene for sulfate reduction, dsrD, was identified in metagenomes using an HMM with trusted cutoff values generated by Anantharaman et al. (hmmsearch—tbloout result.hres—noali—cut_tc -o result_hmm.txt dsrD.hmm input.faa) [21,36].

Multiheme c-type cytochrome analysis.

Multiheme c-type cytochromes (MHCs) required separate analyses for identification due to their numerous biochemical roles [37]. Potential MHC sequences were identified if a given amino acid sequence contained at least 3 CXXCH (Cys-X-X-Cys-His) motifs [37]. Potentially non-metal active MHCs were then removed through three primary steps. Firstly, the MHCs were annotated by comparing sequences to the KEGG, UniRef90, and InterproScan [3842] databases using USEARCH [43] to scan for single and reverse best hit (RBH) results [24]. These annotations helped identify which cytochromes have a known function unassociated with metal cycling (i.e. NapC/NirT, NrfA, HAO, etc.). These sequences, along with seeded MHC sequences from Geobacter spp. and Shewanella oneidiensis, were clustered into a network by alignment score using EFI-EST [44,45] with a score cutoff of 67. These networks were then visualized using Cytoscape v3.4.0 [46] and parsed by hand in order remove hypothetical or misidentified proteins clustering with non-metal active proteins. Lastly, sequences were analyzed using PSORTb v3.0.2 [47] to localize encoded proteins within the cell, removing any protein destined for the cytoplasm only. Resulting protein and motif counts were plotted in R using the ggplot2 package (v2.2.1) [48,49].

Gene abundance calculation.

Functional gene abundances were determined by mapping trimmed reads from a metagenome to corresponding functional sequences using bowtie2 (bowtie—fast) [50]. These mapped reads were then normalized to the length of the gene and the number of mapped reads in the assembly in order to obtain a “reads per kilobase million” or “RPKM” metric. This measurement allows for cross sample comparisons in gene abundances. These calculations were performed in R and then plotted using the ggplot2 package (v2.2.1) [48].

Metagenomic binning and analyses

Scaffolds were binned using a combined binning approach called “DAS Tool” [51], which allows for the dereplication of bins generated through different bin generation strategies. Using the read mapping information obtained during the whole metagenome analyses, assembled scaffolds >2500 bp were binned using MetaBAT (metabat—superspecific) [52]. Assembled scaffolds >1000 bp were also binned using Maxbin [53]. Bins obtained from both algorithms were then processed using DAS Tool with default parameters [51]. Diversity of these bins was assessed by identifying ribosomal protein S3 (rps3) sequences using AMPHORA 2 [54]. Obtained S3 sequences were aligned to an rps3 database from Hug et al. [55] and used in tree generation according to the protocol outlined in the “Arsenic and sulfur gene analysis” section (LG determined with the added utilization of Gblocks to mask phylogenetically uninformative regions) [56]. The maximum-likelihood tree was visualized using ggtree in R [35], with major phylogenetic groups highlighted. Genome completeness and contamination were measured using CheckM [57]. Bin completion measurements can be found in S2 Table.

Only those bins which were of medium quality or greater based upon newly updated standards (i.e., > = 50% complete, <10% contaminated; 306 total) [58] were analyzed further. Each bin was annotated using the pipeline described above (i.e., USEARCH comparing sequences against KEGG, UniRef90, and InterproScan databases); resulting general metabolic characteristics were summarized. Bins were blasted against the specific genes of interest found in the whole metagenomes (arsB, arsC, arrA, aioA, dsrD, and MHCs) to find organisms potentially capable of arsenic mobilization. If a MAG of interest contained a sequence encoding a potential RuBisCO, an alignment followed by tree generation as described with arrA and aioA was performed with RuBisCO sequences from NCBI to assess whether it was type-I, -II, -III, or -IV.

Genome-based replication rates were approximated by performing an “iRep” analysis on genomes at least 75% complete and less than 3% contaminated [59]. Given that iRep requires genomes to be less than 2% contaminated, we have noted where these divergences occurred in the supplemental materials (S3 Table). The reason for this difference was to increase the representation of arsenic mobilizing functions in our analyses. Trimmed reads from each metagenome were mapped to corresponding bins using bowtie2 (bowtie2—fast) [50] with unmapped reads removed using shrinksam with default parameters [60]. The iRep command was then run with defaults and results were plotted in R using the ggplot2 package (v2.2.1).

Approximate taxonomy was assigned by placement in the rps3 maximum-likelihood tree as well through an NCBI BLAST search [61]. Those genomes belonging to the Nitrospirae, Ignavibacteria, and Methylocystaceae were analyzed more deeply using 43-protein concatenated trees. Single copy genes (SCGs) utilized during the phylogenetic placement step in CheckM [57] were identified in all genomes from this study and those published on NCBI using HMMs obtained from PFAM (hmmsearch -Z 26740544 -E 1e-20—tblout result.hres phylo.hmm input.faa) [36,62]. These SCGs were then individually aligned, trimmed, and used in tree generation according to the method established in the “Arsenic and sulfur gene analysis” section, with an additional alignment concatenation step performed in Geneious v9.1.5 [31]. The trees were then visualized in FigTree v1.4.3 ( An exhaustive literature search was performed to ascertain where throughout Bangladesh and eastern India these microorganisms were previously described. Studies were selected if they fell within the specified geographic region and whether they described, at some point, members of the Methylocystaceae or Nitrospirae (or sub-classifications).

Metagenomic read files, assemblies, bins, and protein files can be found on Cyverse ( To access the files, users must create an account and log in. The following files are in the folder pathway “/iplant/home/danczakre/Ohio Groundwater Metagenomes”. Files are also available at the NCBI associated with bioproject number PRJNA512237.

Map generation

R-compatible shapefiles for Bangladesh and India were downloaded from the Database of Global Administrative Areas (GADM) [63]. The shapefiles were imported in R using the sp package v1.3.1 [64], processed using the “fortify” command (rgeos package v0.3.28) [65], and plotted using the ggplot2 package (v2.2.1) [48]. Points on the map were placed by hand based upon GPS coordinates.

Results and discussion

Groundwater microbiomes host functional potential for arsenic cycling, regardless of geochemical conditions

The geochemical conditions within three of these aquifers have been reported previously [25]. Briefly, the aquifer in Athens County was characterized by reducing conditions, and contained high concentrations of sulfate and dissolved iron. In contrast, samples from the Greene County aquifer were the most oxidizing, while the Licking aquifer featured elevated arsenic concentrations [25]. The fourth location, Licking–Private, had sulfate concentrations comparable to the Athens aquifer (151 mg/L), iron concentrations higher than the other Licking location (3.1 mg/L) and arsenic three times higher than the maximum contaminant limit (31 μg/L) (S4 Table).

To investigate the metabolic potential for iron and arsenic cycling within these systems, normalized abundances of marker genes associated with arsenic (arrA, arsBC, aioA) and iron (multiheme c-type cytochromes; MHCs, dsrD) mobilization were compared across locations (Fig 1). The Licking–Private Well featured the greatest functional potential for both direct iron reduction and arsenic reduction, mirroring the surrounding geochemical conditions (Fig 1). Similarly, the genomic potential for arsenic oxidation was highest in the most oxidizing Greene County samples (Fig 1). However, the absence of correlations between genomic functional potential and geochemical conditions in the Athens and Licking aquifers suggest that measures of activity may provide additional insights into links between microbiomes and biogeochemical transformations. Similar inferences regarding the disconnect between functional potential and microbial activity have been made in other ecosystems [6668]. This additionally supports previous research which noted that arsenic metabolism genes were found in rice paddies containing low concentrations of arsenic [69].

Fig 1. RPKM values for genes involved in potential direct or indirect arsenic mobilization.

RPKM values for (A) arsenic mobilizing and sulfate reducing genes, and (B) c-type cytochromes across the 4 wells. Different colors represent different samples in both panels. In panel B, different shades of colors group sequences based upon similar CxxCH-motif counts. Licking–Private Well has the highest arsenic concentrations while Green has the lowest.

To obtain a proxy for microbial activity, average replications rates for metagenome-assembled genomes (MAGs) were calculated utilizing iRep (Fig 2) [59]. In brief, iRep measures variations in read coverage along a genome from the origin to the terminus to approximate replication rates [59]. While this metric does not directly relate to metabolic activity, it may help determine variations in replication rates between groups of microorganisms inferred to perform different functional roles in the environment. MAGs encoding the potential for iron, sulfate, and arsenate reduction featured similar, but variable, iRep values suggesting comparable overall rates of replication (Fig 2). In contrast, MAGs encoding the potential for arsenic oxidation (aioA) were all phylogenetically related to methanotrophs, and had potentially the slowest rates of replication, as inferred from the lowest iRep values (Fig 2). Given the absence of putative methanogens in this dataset, and the high oxidation-reduction potential measured in groundwater, slower inferred growth rates for methanotrophs may be related to electron donor availability.

Fig 2. iRep values for genomes separated by functional potential.

Higher log(iRep) values for the genomes containing the specified functional genes suggest faster growth. For example, those genomes which encoded putative AioA sequences had lower iRep values and replicated at a slower rate than those genomes encoding the other functions.

The metabolic potential for direct arsenic transformations exists throughout the bacterial tree of life

To expand our understanding of metal-cycling microorganisms in shallow subsurface systems, MAGs were investigated for specific functions related to iron and arsenic transformations (Fig 3; S2 File). In total, 306 MAGs of medium or greater quality were recovered from 8 groundwater metagenomes (S2 Table). Of these MAGs, 194 were subsequently analyzed because they featured a ribosomal protein S3 sequence and at least one function of interest. Reflecting the conservation of the ars operon throughout archaea, bacteria, and eukaryotes, both arsB and arsC–arsenic detoxification genes–were broadly distributed throughout the bacterial tree of life and exhibited little phylogenetic bias within our dataset [9,11] (Fig 3).

Fig 3. A maximum-likelihood tree generated from rps3 (rpsC) sequences.

Each major lineage detected in this study is highlighted in a different color. Colored circles below each name indicate which of the examined functions were observed in MAGs belonging to each lineage. Gray circles at the end of branches indicate lineages observed in this study. In the lower portion of the figure, panels depict the frequency with which the putatively encoded functions occurred across MAGs belonging to the detected linages. For example, AioA is found exclusively in MAGs belonging to the Alphaproteobacteria.

Contrastingly, the phylogenetic distribution of aioA was significantly more constrained, occurring only in genomes related to the Methylocystaceae, Rhodosprillaceae, and Rhodobacteraceeae families within the Alphaproteobacteria. Out of nine higher quality MAGs containing aioA, six genomes belonged to the Methylocystaceae and were most closely related to Methylocystis rosea SV97 and associated isolates, suggesting that these might constitute a new species within the genus. While aioA has been previously detected within Methylocystis sp. SC2, potential functionality has not been explored in significant detail. Based upon the analyses presented here, it appears that our MAGs feature metabolisms of previously characterized Methylocystaceae genomes, including methanotrophy via an encoded serine pathway as determined by the presence of methanol dehydrogenase and methane monooxygenase. Five MAGs appear to be missing a glycine hydroxymethyltransferase, however, and all are missing an alanine-glyoxylate transaminase. The aioA sequences within the Methylocystaceae appear to be most closely related to those found in the Order Rhizobiales (S3 File, S1A Fig), including the aioA found in Rhizobium sp. NT-26 which can obtain energy from arsenite oxidation [70]. Additionally, each genome appears to encode a putative pseudoazurin, a potential electron acceptor for the arsenite oxidase [70]. While these results suggest that the Methylocystaceae MAGs might be capable of conserving energy by oxidizing arsenite to arsenate, a role for these genes in canonical arsenic detoxification cannot be discounted. Regardless of energy conservation, members of this family have been found in other arsenic contaminated sites throughout the world, including numerous locations throughout Bangladesh [71,72] (Fig 4). Given their distribution throughout As-impacted sediments, members of the Methylocystaceae may play an under-appreciated role in arsenic cycling.

Fig 4. Biogeography throughout the India and Bangladesh.

Map of eastern India and Bangladesh–areas historically affected by arsenic contamination–indicating the biogeographical distribution of bacteria related to the Methylocystaceae and Nitrospirae and the studies in which they were observed.

Nine medium or higher quality MAGs within the Betaproteobacteria, Deltaproteobacteria, and Nitrospirae contained putative arrA sequences (Fig 3). All five MAGs belonging to the Betaproteobacteria were members of the family Rhodocyclaceae, a functionally diverse and broadly distributed bacterial group [73]. The only Deltaproteobacteria MAG was a member of the Geobacteraceae. Lastly, the remaining three MAGs were members of the phylum Nitrospirae and were most closely related to genomes obtained from other subsurface environments [20,74,75]. Our results indicate that the organisms potentially capable of arsenic reduction might be more diverse than existing databases suggest; out of nine MAGs, only one belonged to a traditional arsenic reducing group (Geobacteraceae) (Fig 3). Particularly, both the Rhodocyclaceae and Nitrospirae represent groups that might be overlooked when considering only characterized arsenate reducers, and may represent a significant underestimation of arsenic mobilization potential throughout shallow aquifer systems due to their broad geographical distribution (Fig 4).

Lineages throughout the tree of life have previously undescribed sulfate and iron reduction potential

Twenty-eight MAGs recovered in this study featured the potential to reduce sulfate to hydrogen sulfide. The majority (15) of genomes encoding this functional potential belonged to the Deltaproteobacteria (Fig 3), including members of the Desulfobulbaceae, Desulfovibrionaceae, Desulfobacteraceae, and Syntrophaceae. Additionally, a member of the family Polyangiaceae also displayed genomic features for this function. Although this group is not traditionally associated with sulfate reduction, this MAG encoded a potentially reductive dsrABD (S4 File, S1B Fig) and featured the potential to use a broad range of smaller carbon substrates as electron donors, including acetate, alcohols, and lactate, providing added genomic support for this function. Beyond the Deltaproteobacteria, the next abundant group of MAGs (10) inferred to perform sulfate reduction belonged to the phylum Nitrospirae. While traditionally considered to be nitrite oxidizers [76], increasing recent evidence has greatly expanded the functional potential of members within this phylum [21,77,78]. Again, these data suggest that the potential for sulfate reduction is present in phylogenetically broader groups than might be assumed by simply considering well-characterized isolates. The metabolic activity of these organisms could potentially drive indirect metal mobilization across diverse subsurface ecosystems.

The reductive dissolution of iron oxides drives metal mobilization, with iron-reducing microorganisms using multi-heme c-type cytochromes (MHCs) as terminal reductases for the final step of electron transfer. Using the presence of putative metal-active MHCs as a screen for iron-reducing microorganisms, thirty-seven MAGs were selected for further analyses. (Figs 3 and 5). Nine of these MAGs were affiliated with the Deltaproteobacteria, including members of the Desulfobacteraceae, Geobacteraceae, and Desulfobulbaceae that are commonly inferred to catalyze iron reduction in the environment [7981]. Moreover, many of these organisms encoded the capability to utilize shorter carbon compounds as electron donors via acetate kinases and alcohol dehydrogenases.

Fig 5. CxxCH-motif content for each MAG featuring over 50 motifs per genome.

Each bar consists of small boxes which represent individual multiheme c-type cytochromes (MHCs) and are scaled based upon the number of CxxCH motifs within each sequence. Relative direct electron transfer potential of an organism can be approximated by the height of each bar. The phylogenetic relationship for each MAG is given in the parentheses along the x-axis and reference genomes of well-characterized iron reducing microorganisms (e.g., G. bemidjiensis Bem and S. oneidensis MR1) have bolded names and gray bars.

Beyond these traditional iron-reducing taxa, many MHC-encoding MAGs (10) belonged to the phylum Ignavibacteria within the FCB Superphylum. Some members of the Ignavibacteria, namely Melioribacter roseus, have previously been characterized as iron reducers [82,83], although their role in environmentally relevant metal cycling is less well understood [84]. Genome-level phylogeny of these Ignavibacteria indicated that eight of the ten were most closely related to Melioribacter (S5 File, S1C Fig). A closer examination of the metabolic potential in these organisms revealed many similarities with other Ignavibacteria and canonical iron reducers, such as the inferred ability to utilize organic acids (acetate, formate, and lactate). Moreover, each genome encoded a complete electron transport chain with all but 2 organisms exhibiting the potential to utilize diverse terminal electron acceptors including nitrite (reduced to ammonia via the NrfAH system) and nitrous oxide (reduced to nitrogen via NosZ). MHCs within the Ignavibacteria MAGs featured an average of 10 heme-binding motifs, placing them within the range of MHCs detected in other iron cycling microorganisms (Fig 5). In the absence of traditional autotrophic pathways associated with iron oxidizers, and the phylogenetic relationship to the known iron reducer Melioribacter rosesus, we infer that these Ignavibacteria MAGs are potentially capable of iron reduction. Although members of the Ignavibacteria appear to be broadly distributed in numerous groundwater ecosystems, relatively few seem to have been specifically detected in arsenic contaminated systems. We hypothesize that this may at least partly be due to mis-identification of these taxa as deeply branching members of the Chlorobi prior to their recent organization into a new phylum (Ignavibacteriae) [82]. Overall, these results suggest that members of the Ignavibacteria–beyond the characterized isolate Melioribacter rosesus–likely catalyze reductive dissolution of iron oxides, contributing to the mobilization of adsorbed metals in reducing environments.

Additional MAGs containing a high number of putative metal-active MHCs were distributed across the Betaproteobacteria and Nitrospirae. Within the Betaproteobacteria, four out of the five MAGs were members of the Gallionellaceae. Each of these MAGs featured chemolithoautotrophic metabolic potential, including an encoded RuBisCO, and were most closely related to other groundwater Gallionellaceae genomes, although BLAST results suggest that the nearest related isolates were the iron-oxidizing organisms Siderooxydans lithotrophicus and Ferriphaselus amnicola [85]. Together, these results indicate that the Gallionellaceae MAGs recovered here likely catalyze iron oxidation reactions. The functional potential encoded within the Nitrospirae MAGs is discussed in more detail below.

The phylum Nitrospirae could play significant, previously undescribed roles in arsenic cycling

The phylum Nitrospirae contains the distantly-related genera Leptospirillum, Thermodesulfovibirio, and Nitrospira, and features a broad diversity of metabolic functionality [76,77,8689]. Isolated members of the Leptospirillum have been characterized as chemolithoautotrophic iron-oxidizers, in contrast to the sulfate reducing capacity of the Thermodesulfovibrio, or the nitrite-oxidizing capabilities of the Nitrospira [76,77,86,90]. Additionally, a clade related to the Thermodesulfovibrio contains genes necessary for magnetotaxis via intracellular magnetite formation [91]. We were able to assemble fourteen medium quality or greater MAGs spanning the phylogenetic and functional diversity within the phylum. Moreover, we were able to uncover a novel potential function in arsenic cycling for the Nitrospirae, in addition to assigning a potentially greater role in iron cycling.

A 43-protein concatenated tree (S6 File, S1D Fig) revealed that the 13 Nitrospirae MAGs generated here are more closely related to genomes obtained from other subsurface environments than to isolated species, while one is related to Candidatus Nitrospira nitrificans. Metabolically, these organisms encoded respiratory functionality based upon the presence of oxidative phosphorylation genes, and contained a broad range of functions typical within the Nitrospirae. Firstly, two MAGs contained genes (e.g., nxrA/B) for nitrite-oxidizing capabilities traditionally characteristic of this phylum. Three MAGs encoded nitrate reductases via napA, three MAGs featured sequences for copper-containing nitrite reductases (e.g., nirK), and three MAGs encoded putative nitric oxide reductases (norBC). Evidence for nitrite reduction to ammonium was present in one MAG featuring the Nrf-system, and others encoded a distantly related ammonia forming nitrite reductase. Additionally, ten MAGs encoded the inferred functional potential for sulfate reduction via the presence of reductive-type dsrABD genes [21].

MHC profiles similar to known iron-reducers, Geobacter bemidjiensis and Shewanella oneidensis, were found in six MAGs, suggesting that members outside of the genus Leptospirllum may be capable of iron cycling (Fig 5). The presence of MHCs with a broad range of CxxCH motifs (up to 49 within one protein) suggests that members of the Nitrospirae may be capable of direct election transfer to or from iron oxides. Due to similar enzymatic machinery and associated genomic features, disentangling iron reduction from iron oxidation is challenging in the absence of microbial isolates. Some evidence suggests that at least two of these Nitrospirae MAGs may perform chemolithoautotrophy due to the presence of the large subunit of a type-II RubisCO (S7 File, S1E Fig). While this mechanism is found in the genus Leptospirillum, these MAGs appear to be unrelated, suggesting that this functionality might be widespread throughout the phylum. In contrast, the absence of RuBisCO-based autotrophic machinery suggests that four Nitrospirae MAGs may perform iron reduction. These MAGs are potentially capable of utilizing ethanol or formate as electron donors via an alcohol dehydrogenase and pyruvate formate lyase / Wood-Ljungdahl pathway, respectively. Additionally, two MAGs may be capable of acetate utilization via an ADP-forming acetyl-CoA synthetase; propionate and butyrate are unlikely substrates, however. While these observations are based upon genomic inferences, results suggest that members of the Nitrospirae may play additional, previously unrecognized roles in iron cycling.

Lastly, the capacity for respiratory arsenate reduction via a putatively encoded arrA was found in three of the high quality Nitrospirae MAGs representing a possible metabolic expansion for this phylum. Broadening our analyses to 185 other published genomes, arrA sequences were found in only eleven other Nitrospirae beyond the three described here. All 14 arrA sequences appear to be most closely related to those sequences found in Deltaproteobacteria genomes, suggesting a potentially shared evolutionary history or horizontal gene transfer (S8 File, S1F Fig). Geographically, members of the Nitrospirae appear to be constituents of microbial communities in many ecosystems, including those featuring elevated arsenic concentrations [9295] (Fig 4). While much of this evidence is based upon 16S rRNA gene data, numerous Nitrospirae are members of groundwater bacterial communities obtained from Bangladesh and might participate in direct arsenic transformations. These results further suggest that the capability for arsenate reduction is broadly distributed and likely exists in other under-sampled lineages. Moreover, the Nitrospirae might represent a previously unidentified group of organisms capable of widespread arsenic mobilization.


Arsenic-contaminated groundwater is a pressing issue throughout the world, although the relationship between contributing geochemical conditions and microbial communities is underexplored. Here we present evidence that the functional potential for arsenic mobilization is both geographically and phylogenetically widespread (Figs 1 and 3). In the groundwater samples studied here, moderate changes in environmental conditions could potentially stimulate the activity of arsenic-cycling microorganisms. For example, the onset of more reducing conditions in the Greene County location studied here could lead to increasing concentrations of mobile arsenic species through the activity of arsC and arrA genes that catalyze arsenate reduction.

Furthermore, our research demonstrates the importance of investigating microbial community members beyond “canonical” iron or arsenic cycling organisms. The Methylocystaceae are a family consisting of methanotrophs but some members appear to encode an aioA, allowing them to potentially oxidize arsenite (Fig 3). The phylum Ignavibacteria appears to contain a greater diversity of members encoding potential genes necessary for iron reduction, a function previously described only within the genus Melioribacter. Lastly, members of the Nitrospirae are much more functionally diverse than previously thought, with new groups potentially capable of iron reduction, iron oxidation, and arsenate reduction. Moreover, members of the Methylocystaceae and Nitrospirae are detected in locations affected by elevated groundwater arsenic concentrations and may help explain the mobilization potential missed by tracking standard organisms. Although these functions are inferred from genomic data alone, these results highlight the importance of looking beyond exclusively isolation-based and phylogenetic assumptions. By obtaining a more complete understanding of the diverse functional potential encoded within subsurface microbiomes, we can begin to combine isolation-based techniques with arsenic mobilization studies in an attempt to better address this ongoing crisis.

Supporting information

S1 Fig.

PDF containing all trees generated within this study: A) AioA tree, B) DsrAB tree, C) Ignavibacteria concatenated tree, D) Nitrospirae concatenated tree, E) RuBisCO tree, and F) ArrA tree. Red labels indicate sequences from this study, while blue are reference sequences.


S1 File. FASTA file containing all arsenic-related sequences from NCBI used in this study.


S2 File. Raw maximum-likelihood rpsC tree run with 100 bootstraps.


S3 File. Maximum-likelihood AioA and other molybdopterin-protein tree run with 100 bootstraps.


S4 File. Concatenated dsrAB maximum-likelihood tree run with 100 bootstraps.


S5 File. Concatenated maximum-likelihood tree generated from 16 ribosomal proteins for all published Ignavibacteria genomes from NCBI.


S6 File. Concatenated maximum-likelihood tree generated from 16 ribosomal proteins for all published Nitrospirae genomes from NCBI.


S7 File. Maximum-likelihood tree for RuBisCO sequences run with 100 bootstraps.


S8 File. Maximum-likelihood tree for ArrA sequences run with 100 bootstraps.


S1 Table. Table of assembly statistics for each of the 8 metagenomes.


S2 Table. Table of bin information, including phylogeny, completion, contamination, and other relevant information.


S4 Table. Geochemistry results for historical samples and those used in this study (highlighted in green).



  1. 1. Ravenscroft P, Brammer H, Richards K. Arsenic Pollution [Internet]. Oxford, UK: Wiley-Blackwell; 2009.
  2. 2. Karim MM. Arsenic in groundwater and health problems in Bangladesh. Water Res. 2000;34: 304–310.
  3. 3. Charlet L, Polya DA. Arsenic in shallow, reducing groundwaters in Southern Asia: An environmental health disaster. Elements. 2006;2: 91–96.
  4. 4. DeSimone LA, McMahon PB, Rosen MR. The quality of our Nation’s waters—Water quality in Principal Aquifers of the United States, 1991–2010. U.S. Geological Survey Circular. 2014.
  5. 5. Ayotte JD, Medalie L, Qi SL, Backer LC, Nolan BT. Estimating the High-Arsenic Domestic-Well Population in the Conterminous United States. Environ Sci Technol. 2017;51: 12443–12454. pmid:29043784
  6. 6. Lièvremont D, Bertin PN, Lett MC. Arsenic in contaminated waters: Biogeochemical cycle, microbial metabolism and biotreatment processes. Biochimie. Elsevier Masson SAS; 2009;91: 1229–1237. pmid:19567262
  7. 7. Thomas MA. Arsenic in groundwater of Licking County, Ohio, 2012—Occurrence and relation to hydrogeology: U.S. Geological Survey Scientific Investigations Report 2015–5148. 2016.
  8. 8. Yamamura S, Amachi S. Microbiology of inorganic arsenic: From metabolism to bioremediation. J Biosci Bioeng. Elsevier Ltd; 2014;118: 1–9. pmid:24507904
  9. 9. Andres J, Bertin PN. The microbial genomics of arsenic. FEMS Microbiol Rev. 2016;40: 299–322. pmid:26790947
  10. 10. Cavalca L, Corsini A, Zaccheo P, Andreoni V, Muyzer G. Removal of Arsenic From Water. 2013; 753–768.
  11. 11. Amend JP, Saltikov C, Lu G-S, Hernandez J. Microbial Arsenic Metabolism and Reaction Energetics. Rev Mineral Geochemistry. 2014;79: 391–433.
  12. 12. Islam FS, Gault AG, Boothman C, Polya DA, Chamok JM, Chatterjee D, et al. Role of metal-reducing bacteria in arsenic release from Bengal delta sediments. Nature. 2004;430: 68–71. pmid:15229598
  13. 13. Xie X, Wang Y, Su C, Liu H, Duan M, Xie Z. Arsenic mobilization in shallow aquifers of Datong Basin: Hydrochemical and mineralogical evidences. J Geochemical Explor. 2008;98: 107–115.
  14. 14. Burton ED, Bush RT, Sullivan LA, Johnston SG, Hocking RK. Mobility of arsenic and selected metals during re-flooding of iron- and organic-rich acid-sulfate soil. Chem Geol. 2008;253: 64–73.
  15. 15. Hansel CM, Lentini CJ, Tang Y, Johnston DT, Wankel SD, Jardine PM. Dominance of sulfur-fueled iron oxide reduction in low-sulfate freshwater sediments. ISME J. Nature Publishing Group; 2015;9: 2400–2412. pmid:25871933
  16. 16. Onstott TC, Chan E, Polizzotto ML, Lanzon J, DeFlaun MF. Precipitation of arsenic under sulfate reducing conditions and subsequent leaching under aerobic conditions. Appl Geochemistry. Elsevier Ltd; 2011;26: 269–285.
  17. 17. Buschmann J, Berg M. Impact of sulfate reduction on the scale of arsenic contamination in groundwater of the Mekong, Bengal and Red River deltas. Appl Geochemistry. Elsevier Ltd; 2009;24: 1278–1286.
  18. 18. Kirk MF, Holm TR, Park J, Jin Q, Sanford RA, Fouke BW, et al. Bacterial sulfate reduction limits natural arsenic contamination in groundwater. Geology. 2004;32: 953–956.
  19. 19. Stucker VK, Silverman DR, Williams KH, Sharp JO, Ranville JF. Thioarsenic Species Associated with Increased Arsenic Release during Biostimulated Subsurface Sulfate Reduction. Environ Sci Technol. 2014;48: 13367–13375. pmid:25329793
  20. 20. Anantharaman K, Brown CT, Hug LA, Sharon I, Castelle CJ, Probst AJ, et al. Thousands of microbial genomes shed light on interconnected biogeochemical processes in an aquifer system. Nat Commun. Nature Publishing Group; 2016;7: 13219. pmid:27774985
  21. 21. Anantharaman K, Hausmann B, Jungbluth SP, Kantor RS, Lavy A, Warren LA, et al. Expanded diversity of microbial groups that shape the dissimilatory sulfur cycle. ISME J. 2018; pmid:29467397
  22. 22. Thomas MA. Arsenic in Midwestern Glacial Deposits: Occurrence and Relation to Selected Hydrogeologic and Geochemical Factors: Water-Resources Investigations Report 03–4228 [Internet]. 2003. Available:
  23. 23. Thomas MA, Schumann TL, Pletsch BA. Arsenic in ground water in selected Pparts of Southwestern Ohio, 2002–03: U.S. Geological Survey Scientific Investigations Report 2005–5138. 2005.
  24. 24. Danczak RE, Johnston MD, Kenah C, Slattery M, Wrighton KC, Wilkins MJ. Members of the Candidate Phyla Radiation are functionally differentiated by carbon- and nitrogen-cycling capabilities. Microbiome. Microbiome; 2017;5: 112. pmid:28865481
  25. 25. Danczak RE, Johnston MD, Kenah C, Slattery M, Wilkins MJ. Microbial Community Cohesion Mediates Community Turnover in Unperturbed Aquifers. Mason O, editor. mSystems. 2018;3. pmid:29984314
  26. 26. Joshi N, Fass J. Sickle: A sliding-window, adaptive, quality-based trimming tool for FastQ files [Internet]. 2011. Available:
  27. 27. Peng YY, Leung HCM, Yiu SM, Chin FYL. IDBA-UD: a de novo assembler for single-cell and metagenomic sequencing data with highly uneven depth. Bioinformatics. 2012;28: 1420–1428. pmid:22495754
  28. 28. Hyatt D, Chen G-L, LoCascio PF, Land ML, Larimer FW, Hauser LJ. Prodigal: prokaryotic gene recognition and translation initiation site identification. BMC Bioinformatics. 2010;11: 119. pmid:20211023
  29. 29. Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 1997;25: 3389–402. Available: pmid:9254694
  30. 30. Edgar RC. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 2004;32: 1792–1797. pmid:15034147
  31. 31. Kearse M, Moir R, Wilson A, Stones-Havas S, Cheung M, Sturrock S, et al. Geneious Basic: An integrated and extendable desktop software platform for the organization and analysis of sequence data. Bioinformatics. 2012;28: 1647–1649. pmid:22543367
  32. 32. Stamatakis A. RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics. 2014;30: 1312–1313. pmid:24451623
  33. 33. Darriba D, Taboada GL, Doallo R, Posada D. ProtTest 3: fast selection of best-fit models of protein evolution. Bioinformatics. 2011;27: 1164–1165. pmid:21335321
  34. 34. Guindon S, Gascuel O. A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst Biol. 2003;52: 696–704. Available: pmid:14530136
  35. 35. Yu G, Smith DK, Zhu H, Guan Y, Lam TT-YY. ggtree: an R package for visualization and annotation of phylogenetic trees with their covariates and other associated data. Methods Ecol Evol. 2017;8: 28–36.
  36. 36. Eddy S, Wheeler T. HMMER 3.1 [Internet]. 2013. Available:
  37. 37. Sharma S, Cavallaro G, Rosato A. A systematic investigation of multiheme c-type cytochromes in prokaryotes. J Biol Inorg Chem. 2010;15: 559–571. pmid:20084531
  38. 38. Kanehisa M, Sato Y, Kawashima M, Furumichi M, Tanabe M. KEGG as a reference resource for gene and protein annotation. Nucleic Acids Res. 2016;44: D457–D462. pmid:26476454
  39. 39. Kanehisa M, Goto S. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 2000;28: 27–30. pmid:10592173
  40. 40. Kanehisa M, Furumichi M, Tanabe M, Sato Y, Morishima K. KEGG: new perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res. 2017;45: D353–D361. pmid:27899662
  41. 41. Suzek BE, Wang Y, Huang H, McGarvey PB, Wu CH. UniRef clusters: a comprehensive and scalable alternative for improving sequence similarity searches. Bioinformatics. 2015;31: 926–932. pmid:25398609
  42. 42. Quevillon E, Silventoinen V, Pillai S, Harte N, Mulder N, Apweiler R, et al. InterProScan: protein domains identifier. Nucleic Acids Res. 2005;33: W116–W120. pmid:15980438
  43. 43. Edgar RC. Search and clustering orders of magnitude faster than BLAST. Bioinformatics. 2010;26: 2460–2461. pmid:20709691
  44. 44. Gerlt JA, Bouvier JT, Davidson DB, Imker HJ, Sadkhin B, Slater DR, et al. Enzyme Function Initiative-Enzyme Similarity Tool (EFI-EST): A web tool for generating protein sequence similarity networks. Biochim Biophys Acta—Proteins Proteomics. 2015;1854: 1019–1037. pmid:25900361
  45. 45. Gerlt JA. Genomic Enzymology: Web Tools for Leveraging Protein Family Sequence-Function Space and Genome Context to Discover Novel Functions. Biochemistry. 2017;56: 4293–4308. pmid:28826221
  46. 46. Smoot ME, Ono K, Ruscheinski J, Wang PL, Ideker T. Cytoscape 2.8: New features for data integration and network visualization. Bioinformatics. 2011;27: 431–432. pmid:21149340
  47. 47. Yu NY, Wagner JR, Laird MR, Melli G, Rey S, Lo R, et al. PSORTb 3.0: improved protein subcellular localization prediction with refined localization subcategories and predictive capabilities for all prokaryotes. Bioinformatics. 2010;26: 1608–1615. pmid:20472543
  48. 48. Wickham H. ggplot2: Elegant Graphics for Data Analysis [Internet]. Springer-Verlag New York; 2009. Available:
  49. 49. R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2016.
  50. 50. Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat Methods. 2012;9: 357–359. pmid:22388286
  51. 51. Sieber CMK, Probst AJ, Sharrar A, Thomas BC, Hess M, Tringe SG, et al. Recovery of genomes from metagenomes via a dereplication, aggregation and scoring strategy. Nat Microbiol. 2018;3: 836–843. pmid:29807988
  52. 52. Kang DD, Froula J, Egan R, Wang Z. MetaBAT, an efficient tool for accurately reconstructing single genomes from complex microbial communities. PeerJ. 2015;3: e1165. pmid:26336640
  53. 53. Wu Y-W, Tang Y-H, Tringe SG, Simmons BA, Singer SW. MaxBin: an automated binning method to recover individual genomes from metagenomes using an expectation-maximization algorithm. Microbiome. 2014;2: 26. pmid:25136443
  54. 54. Wu M, Scott AJ. Phylogenomic analysis of bacterial and archaeal sequences with AMPHORA2. Bioinformatics. 2012;28: 1033–1034. pmid:22332237
  55. 55. Hug LA, Baker BJ, Anantharaman K, Brown CT, Probst AJ, Castelle CJ, et al. A new view of the tree of life. Nat Microbiol. 2016;1: 16048. pmid:27572647
  56. 56. Castresana J. Selection of Conserved Blocks from Multiple Alignments for Their Use in Phylogenetic Analysis. Mol Biol Evol. 2000;17: 540–552. pmid:10742046
  57. 57. Parks DH, Imelfort M, Skennerton CT, Hugenholtz P, Tyson GW. CheckM: assessing the quality of microbial genomes recovered from isolates, single cells, and metagenomes. Genome Res. 2015;25: 1043–1055. pmid:25977477
  58. 58. Bowers RM, Kyrpides NC, Stepanauskas R, Harmon-Smith M, Doud D, Reddy TBK, et al. Minimum information about a single amplified genome (MISAG) and a metagenome-assembled genome (MIMAG) of bacteria and archaea. Nat Biotechnol. 2017;35: 725–731. pmid:28787424
  59. 59. Brown CT, Olm MR, Thomas BC, Banfield JF. Measurement of bacterial replication rates in microbial communities. Nat Biotechnol. 2016;34: 1256–1263. pmid:27819664
  60. 60. Thomas BC. shrinksam—shrinks a SAM file while maintaining mate pair information. 2013.
  61. 61. Boratyn GM, Camacho C, Cooper PS, Coulouris G, Fong A, Ma N, et al. BLAST: a more efficient report with usability improvements. Nucleic Acids Res. 2013;41: W29–W33. pmid:23609542
  62. 62. Finn RD, Coggill P, Eberhardt RY, Eddy SR, Mistry J, Mitchell AL, et al. The Pfam protein families database: towards a more sustainable future. Nucleic Acids Res. 2016;44: D279–D285. pmid:26673716
  63. 63. Database of Global Administrative Areas (GADM) [Internet]. 2012. Available:
  64. 64. Pebesma EJ, Bivand RS. Classes and methods for spatial data in R. R News. 2005;5.
  65. 65. Bivand R, Rundel C. rgeos: Interface to Geometry Engine—Open Source ('GEOS’). 2018.
  66. 66. Jones SE, Lennon JT. Dormancy contributes to the maintenance of microbial diversity. Proc Natl Acad Sci. 2010;107: 5881–5886. pmid:20231463
  67. 67. Campbell BJ, Yu L, Heidelberg JF, Kirchman DL. Activity of abundant and rare bacteria in a coastal ocean. Proc Natl Acad Sci. 2011;108: 12776–12781. pmid:21768380
  68. 68. Hunt DE, Lin Y, Church MJ, Karl DM, Tringe SG, Izzo LK, et al. Relationship between abundance and specific activity of bacterioplankton in open ocean surface waters. Appl Environ Microbiol. 2013;79: 177–184. pmid:23087033
  69. 69. Xiao K-Q, Li L-G, Ma L-P, Zhang S-Y, Bao P, Zhang T, et al. Metagenomic analysis revealed highly diverse microbial arsenic metabolism genes in paddy soils with low-arsenic contents. Environ Pollut. 2016;211: 1–8. pmid:26736050
  70. 70. Santini JM, Kappler U, Ward SA, Honeychurch MJ, vanden Hoven RN, Bernhardt P V. The NT-26 cytochrome c552 and its role in arsenite oxidation. Biochim Biophys Acta—Bioenerg. 2007;1767: 189–196. pmid:17306216
  71. 71. Sultana M, Härtig C, Planer-Friedrich B, Seifert J, Schlömann M. Bacterial communities in Bangladesh aquifers differing in aqueous arsenic concentration. Geomicrobiol J. 2011;28: 198–211.
  72. 72. Das S, Bora SS, Yadav RNS, Barooah M. A metagenomic approach to decipher the indigenous microbial communities of arsenic contaminated groundwater of Assam. Genomics Data. Elsevier Inc.; 2017;12: 89–96. pmid:28409115
  73. 73. Oren A. The Family Rhodocyclaceae. The Prokaryotes. Berlin, Heidelberg: Springer Berlin Heidelberg; 2014. pp. 975–998.
  74. 74. Probst AJ, Castelle CJ, Singh A, Brown CT, Anantharaman K, Sharon I, et al. Genomic resolution of a cold subsurface aquifer community provides metabolic insights for novel microbes adapted to high CO2 concentrations. Environ Microbiol. 2017;19: 459–474. pmid:27112493
  75. 75. Probst AJ, Ladd B, Jarett JK, Geller-McGrath DE, Sieber CMK, Emerson JB, et al. Differential depth distribution of microbial function and putative symbionts through sediment-hosted aquifers in the deep terrestrial subsurface. Nat Microbiol. Springer US; 2018;3: 328–336. pmid:29379208
  76. 76. Daims H, Lücker S, Wagner M. A New Perspective on Microbes Formerly Known as Nitrite-Oxidizing Bacteria. Trends Microbiol. Elsevier Ltd; 2016;24: 699–712. pmid:27283264
  77. 77. Zecchin S, Mueller RC, Seifert J, Stingl U, Anantharaman K, von Bergen M, et al. Rice paddy Nitrospirae encode and express genes related to sulfate respiration: proposal of the new genus Candidatus Sulfobium. Appl Environ Microbiol. 2017;84: AEM.02224-17. pmid:29247059
  78. 78. Dalcin Martins P, Danczak RE, Roux S, Frank J, Borton MA, Wolfe RA, et al. Viral and metabolic controls on high rates of microbial sulfur and carbon cycling in wetland ecosystems. Microbiome. 2018;6: 138. pmid:30086797
  79. 79. Holmes DE, Bond DR, Lovley DR. Electron Transfer by Desulfobulbus propionicus to Fe(III) and Graphite Electrodes. Appl Environ Microbiol. 2004;70: 1234–1237. pmid:14766612
  80. 80. Lovley DR, Roden EEEE, Phillips EJP, Woodward JC. Enzymatic iron and uranium reduction by sulfate-reducing bacteria. Mar Geol. 1993;113: 41–53.
  81. 81. Lovley DR, Giovannoni SJ, White DC, Champine JE, Phillips EJP, Gorby YA, et al. Geobacter metallireducens gen. nov. sp. nov., a microorganism capable of coupling the complete oxidation of organic compounds to the reduction of iron and other metals. Arch Microbiol. 1993;159: 336–344. pmid:8387263
  82. 82. Podosokorskaya OA, Kadnikov V V., Gavrilov SN, Mardanov A V., Merkel AY, Karnachuk O V., et al. Characterization of Melioribacter roseus gen. nov., sp. nov., a novel facultatively anaerobic thermophilic cellulolytic bacterium from the class Ignavibacteria, and a proposal of a novel bacterial phylum Ignavibacteriae. Environ Microbiol. 2013;15: 1759–1771. pmid:23297868
  83. 83. Kadnikov V V., Mardanov A V., Podosokorskaya OA, Gavrilov SN, Kublanov I V., Beletsky A V., et al. Genomic Analysis of Melioribacter roseus, Facultatively Anaerobic Organotrophic Bacterium Representing a Novel Deep Lineage within Bacteriodetes/Chlorobi Group. PLoS One. 2013;8. pmid:23301019
  84. 84. Converse BJ, McKinley JP, Resch CT, Roden EE. Microbial mineral colonization across a subsurface redox transition zone. Front Microbiol. 2015;6: 1–14.
  85. 85. Hallbeck L, Pedersen K. The Family Gallionellaceae Lotta. 2013; 545–577.
  86. 86. Bhatnagar S, Badger JH, Madupu R, Khouri HM, Connor EMO, Robb FT, et al. Genome Sequence of the Sulfate-Reducing Thermophilic Bacterium Thermodesulfovibrio yellowstonii Strain DSM 11347 T (Phylum. Genome Announc. 2015;3: 1994–1995.
  87. 87. Fujimura R, Sato Y, Nishizawa T, Oshima K, Kim SW, Hattori M, et al. Complete genome sequence of Leptospirillum ferrooxidans strain C2-3, isolated from a fresh volcanic ash deposit on the Island of Miyake, Japan. J Bacteriol. 2012;194: 4122–4123. pmid:22815442
  88. 88. Lucker S, Wagner M, Maixner F, Pelletier E, Koch H, Vacherie B, et al. A Nitrospira metagenome illuminates the physiology and evolution of globally important nitrite-oxidizing bacteria. Proc Natl Acad Sci. 2010;107: 13479–13484. pmid:20624973
  89. 89. Marietou A. Nitrate reduction in sulfate-reducing bacteria. FEMS Microbiol Lett. 2016;363: 2016–2019. pmid:27364687
  90. 90. Schrenk MO, Edwards KJ, Goodman RM, Hamers RJ, Banfield JF. Distribution of Thiobacillus ferrooxidansand Leptospirillum ferrooxidans: Implications for generation of acid mine drainage. Science (80-). 1998;279: 1519–1523. pmid:9488647
  91. 91. Lin W, Zhang W, Zhao X, Roberts AP, Paterson GA, Bazylinski DA, et al. Genomic expansion of magnetotactic bacteria reveals an early common origin of magnetotaxis with lineage-specific evolution. ISME J. Springer US; 2018;12: 1508–1519. pmid:29581530
  92. 92. Gnanaprakasam ET, Lloyd JR, Boothman C, Ahmed KM, Choudhury I, Bostick BC, et al. Microbial community structure and arsenic biogeochemistry in two arsenic-impacted aquifers in Bangladesh. MBio. 2017;8: 1–18. pmid:29184025
  93. 93. Ghosh D, Bhadury P, Routh J. Diversity of arsenite oxidizing bacterial communities in arsenic-rich deltaic aquifers in West Bengal, India. Front Microbiol. 2014;5: 1–14.
  94. 94. Gorra R, Webster G, Martin M, Celi L, Mapelli F, Weightman AJ. Dynamic Microbial Community Associated with Iron-Arsenic Co-Precipitation Products from a Groundwater Storage System in Bangladesh. Microb Ecol. 2012;64: 171–186. pmid:22349905
  95. 95. Das S, Jean J-S, Kar S, Liu C-C. Changes in Bacterial Community Structure and Abundance in Agricultural Soils under Varying Levels of Arsenic Contamination. Geomicrobiol J. 2013;30: 635–644.