|Home | About | Journals | Submit | Contact Us | Français|
Acid mine drainage (AMD) is a major environmental problem affecting tens of thousands of kilometers of waterways worldwide. Passive bioremediation of AMD relies on microbial communities to oxidize and remove iron from the system; however, iron oxidation rates in AMD environments are highly variable among sites. At Scalp Level Run (Cambria County, PA), first-order iron oxidation rates are 10 times greater than at other coal-associated iron mounds in the Appalachians. We examined the bacterial community at Scalp Level Run to determine whether a unique community is responsible for the rapid iron oxidation rate. Despite strong geochemical gradients, including a >10-fold change in the concentration of ferrous iron from 57.3 mg/liter at the emergence to 2.5 mg/liter at the base of the coal tailings pile, the bacterial community composition was nearly constant with distance from the spring outflow. Scalp Level Run contains many of the same taxa present in other AMD sites, but the community is dominated by two strains of Ferrovum myxofaciens, a species that is associated with high rates of Fe(II) oxidation in laboratory studies.
IMPORTANCE Acid mine drainage pollutes more than 19,300 km of rivers and streams and 72,000 ha of lakes worldwide. Remediation is frequently ineffective and costly, upwards of $100 billion globally and nearly $5 billion in Pennsylvania alone. Microbial Fe(II) oxidation is more efficient than abiotic Fe(II) oxidation at low pH (P. C. Singer and W. Stumm, Science 167:1121–1123, 1970, https://doi.org/10.1126/science.167.3921.1121). Therefore, AMD bioremediation could harness microbial Fe(II) oxidation to fuel more-cost-effective treatments. Advances will require a deeper understanding of the ecology of Fe(II)-oxidizing microbial communities and the factors that control their distribution and rates of Fe(II) oxidation. We investigated bacterial communities that inhabit an AMD site with rapid Fe(II) oxidation and found that they were dominated by two operational taxonomic units (OTUs) of Ferrovum myxofaciens, a taxon associated with high laboratory rates of iron oxidation. This research represents a step forward in identifying taxa that can be used to enhance cost-effective AMD bioremediation.
Acid mine drainage (AMD) is a devastating and widespread industrial pollution problem (1,–6). The effects of AMD are felt especially hard in the Appalachian Plateau in the eastern United States, where over 8,000 km of streams are polluted by AMD. Many of these streams are located in Pennsylvania, where a long history of coal mining has set the stage for a major AMD problem (7). Acid mine drainage occurs when pyrite or other sulfide minerals exposed by mining are exposed to water and oxygen and react to produce ferrous iron and sulfuric acid via the following reactions [where “(s)” means “solid”]:
Equation 1 describes a frequent event in which oxygen is consumed in groundwater and Fe(II) is mobilized, resulting in AMD springs that emerge anoxic and mildly acidic (for examples, see references 8 to 11). Iron oxidation and iron oxide precipitation (equations 2 and 3) occur immediately downstream from the springs, spatially distant from pyrite-containing coal deposits. Due to the legacy of coal mining in the Appalachians, AMD spring discharges are extremely common (2).
Acid mine drainage springs and associated iron oxide mounds are chemically and physically heterogeneous. They differ in iron oxidation rates (for examples, see references 8, 9, 12, and 13), geochemistry, and microbial community composition. Previous work suggests that the AMD bacterial community composition is strongly influenced by pH and the concentration of ferrous iron (10, 14). Differences in bacterial community composition can account for some of the observed variation in iron oxidation rates (11, 15). However, it is clear that we require more detailed knowledge of the physical and geochemical variables that control bacterial population distributions (ecological niches), especially at sites where Fe(II) oxidation rates are high.
Iron oxidation rates at Scalp Level Run (Cambria County, PA) (Fig. 1) are 10 times greater than those at other sites in Pennsylvania and the Iberian Pyrite Belt (13). We hypothesized that a unique bacterial community is responsible for the rapid oxidation of Fe(II) at Scalp Level Run. We used fluorescence in situ hybridization (FISH) and high-throughput 16S rRNA gene sequencing to identify and quantify bacterial populations at Scalp Level Run and field geochemical methods to investigate in situ the environmental parameters that control their distribution.
The average concentration of ferrous iron in Scalp Level Run decreased from 57.3 mg/liter at the emergence to 2.5 mg/liter at the base of the coal tailings pile (SL 82; Fig. 2A; see also Table S1 in the supplemental material). Similarly, the concentration of total iron decreased from 70 mg/liter at the emergence to 49.7 mg/liter at SL 8.5. Total iron then increased to 63.5 mg/liter at SL 25 before continuing to decrease to 10.6 mg/liter at the base of the coal tailings pile (Fig. 2A; Table S1). The pH varied from 2.77 at SL 82 to 2.96 at SL 8.5 (Fig. 2A; Table S1). Dissolved oxygen increased rapidly from 0.5 mg/liter at the emergence to 9.1 mg/liter at SL 25 and then decreased to 8.4 mg/liter at SL 47 and SL 82 (Fig. 2B; Table S1). Temperature and oxidation-reduction potential (ORP) increased from the emergence, 13.3°C and 386.2 mV, respectively, to the base of the coal tailings pile (20.7°C and 491.7 mV) (Fig. 2B; Table S1).
MiSeq Illumina bacterial sequence tags were obtained for SL 8.5, SL 25, and SL 82 samples (Fig. 3A). Between 67% and 92% of the sequences at each site fell within the Proteobacteria. Within the Proteobacteria, the Betaproteobacteria order Ferrovales was the most common (49 to 63% of sequences). There were two abundant operational taxonomic units (OTUs) within the Ferrovales. Both were closely related to strains of the autotrophic Fe(II) oxidizer Ferrovum myxofaciens. The most abundant Ferrovum OTU was closely affiliated with isolate EHS8 (KC155322.1; 100% similarity), and the second OTU was most closely related to Ferrovum sp. PN-J185 (KC677657.1; 97% similarity). The Gammaproteobacteria order Xanthomonadales was also abundant and composed 9 to 24% of the sequences (Fig. 3A). All Xanthomonadales OTUs were classified as Xanthomonadaceae. The most abundant Xanthomonadales OTU was most closely related to Metallibacterium scheffleri strain DKE6, an anaerobic Fe(III) reducer (16) (NR_118103.1; 100% similarity). The Alphaproteobacteria Rhodospiralles were also common (1 to 18% of sequences). The most abundant Rhodospiralles OTU was closely related to the heterotrophic acidophile Acidiphilium rubrum.
The phylum Acidobacteria was the second most abundant phylum after Proteobacteria and was more abundant in SL 25 (7% of bacterial sequences) than in SL 8.5 and 82 (3 and 1% of bacterial sequences, respectively). Chloroflexi and Planctomycetes were present in all libraries. Rare sequences related to Cyanobacteria were present at all sites (<1%). The most abundant cyanobacterial OTU was affiliated with the Melainabacteria, a basal group of nonphotosynthetic Cyanobacteria (17). Chloroplast sequences were found at all sites and increased in abundance downstream from SL 8.5 (23% of total sequences) to SL 85 (39% of total sequences; see Fig. S1 in the supplemental material). Chloroplast sequences were identified as Stramenophiles, Streptophyta, Euglenozoa, or Chlorophyta. Stramenophiles were the most abundant. The abundance of chloroplasts from Euglenozoa decreased from SL emergence to SL 82. Chloroplast sequences from Chlorophyta were present only in SL 25 and SL 82. Chloroplast sequences were removed from the data set for bacterial community composition analyses presented above.
Fluorescence in situ hybridization was used to quantify the relative abundance of common AMD bacterial groups, including Ferrovum spp. (Ferri643 probe), acidophilic Gallionella (Gal177 probe), Acidithiobacillus ferrooxidans, and an AMD clade of Xanthomonadales (Fig. 4; Table 1). The Betaproteobacteria were the most abundant group detected by FISH and accounted for 41 to 53% of total bacterial cells. The Betaproteobacteria taxon Gallionella was not detected at any sample location. The Ferritrophicum probe targets a portion of the ribosome that Fuchs et al. (18) classify as a class V region, indicating that a probe in that region is only 6% to 20% as bright as a reference probe. In fact, the probe was too dim to accurately count cells using standard FISH protocols, and we therefore excluded Ferritrophicum from FISH-based analyses. Ferrovum spp. composed between 3% and 35% of Eubmix-stained cells. This taxon steadily decreased from the emergence to the base of the pile, except at SL 8.5, where the relative abundance sharply decreased (Fig. 5A). Unidentified betaproteobacterial sequences increased from 5% of Eubmix-stained cells at the emergence to 49% at the base of the tailings pile. The Xanthomonadales were the second most abundant population detected by FISH and made up between 6% and 29% of the community. The relative abundance of the Xanthomonadales was between 6% and 13% across the site, except at SL 8.5, where it composed 29% of the community (Fig. 3B). The Gammaproteobacteria genus Acidithiobacillus was not detected by FISH at any sampling site.
The bacterial community at Scalp Level Run was primarily composed of taxa common to other AMD sites in Pennsylvania and across the globe, including those in the orders Rhodospiralles (Alphaproteobacteria), Xanthomonadales (Gammaproteobacteria), and Ferrovales (Betaproteobacteria) (for examples, see references 10, 11, 19, 20, and 21). The Xanthomonadales at Scalp Level Run were closely related to other sequences from AMD sites, including in the Rio Tinto, Spain (22), in South Korea (23), and in the United States (9, 10). Acid mine drainage-associated Xanthomonadales sequences made up a monophyletic clade nested within Xanthomonadales (Fig. 4). The AMD clade was most closely related to cultivated strain WJ2, a moderately acidophilic Fe(II) oxidizer isolated from an AMD wetland remediation site near Cornwall, England (24). This taxon was also abundant (23% to >50%) in the surface and subsurface of the Mushroom Farm AMD mound in Ohio, USA (25). Nearly all the sequences identified as Rhodospirillales fell within the genus Acidiphilium, a heterotrophic iron-reducing Alphaproteobacterium. This taxon is commonly found in AMD sites, including the Dabaoshan Mine, China (26), and the Rio Tinto watershed (27).
The most abundant order at Scalp Level Run was the Ferrovales. There were two abundant OTUs from this order, and these OTUs fell within different clusters of the Ferrovales as defined by Tischer and others (28). The first OTU was contained within cluster 1 of the Ferrovales, and the second OTU fell within cluster 3 (Fig. 6). The cluster 1 sequence had no mismatches to the Ferri643 probe, whereas the cluster 3 sequence had a one-base-pair mismatch to the probe and therefore would not have been visualized in our FISH experiments. Two distinct Ferrovales sequences were found in the clone libraries, but neither of these sequences is affiliated with cluster 3 and both have no mismatches to the Ferri643 probe (Fig. 6).
We identified two abundant species of Betaproteobacteria in the amplicon data sets. Both were members of the genus Ferrovum. We expected that the relative abundance of Ferrovum spp. in the amplicon data set would equal the relative abundance of Ferrovum identified by FISH. However, they were not equivalent; the relative abundance of Ferrovum as calculated using FISH was lower than the abundance in the amplicon data set. The discrepancy between the FISH and amplicon data sets likely results from a one-base-pair mismatch between the Ferrovum probe and the corresponding 16S rRNA gene sequence in the cluster 3 Ferrovum OTU.
The relative abundance of Betaproteobacteria (Bet42a) identified by FISH was approximately equal to the combined abundance of the two Ferrovum OTUs retrieved from the amplicon data set (Fig. 5A). This is consistent with the result that no other abundant Betaproteobacteria were detected in the sequence tag or clone libraries. Thus, the betaproteobacterial community was composed almost exclusively of Ferrovum phylotypes. We therefore used the relative abundance of Betaproteobacteria as a proxy for the total abundance of Ferrovum spp. in the FISH data set and calculated the relative abundance of the cluster 3 Ferrovum isolates by difference (see Table S2 in the supplemental material). Using this approach, we found that the relative abundances of the cluster 1 and cluster 3 Ferrovum phylotypes are inversely correlated in FISH analyses (R2 = 0.99; P < 0.01), with cluster 1 more abundant near the spring emergence and cluster 3 more abundant at the base of the iron mound. Both temperature and pH vary across the iron mound, and both of these parameters are found to influence the growth rate of F. myxofaciens PG3, the type strain of Ferrovum (29). These factors may be controlling the distribution of Ferrovum in separate niches across the Scalp Level Run iron mound.
The only non-Ferrovum betaproteobacterial sequence in our clone library fell within the order Ferritrophicales, and this order was not abundant in the sequence tag libraries (<0.1%). The Ferritrophicales are Betaproteobacteria related to the Gallionellales, Ferrovales, and Rhodocyclales (29). The sequence that we retrieved from Scalp Level Run was closely related to the Fe(II) oxidizer Ferritrophicum radicicola (Fig. 7) and has not been detected at other Pennsylvania AMD sites. Ferritrophicum radicicola is an Fe(II) oxidizer isolated from plant roots near an AMD stream (30). Other sequences closely related to Ferritrophicum radicicola have been retrieved from the Yunfu sulfide mine (31) and Yinshan lead-zinc mine, China (32), the Carnoulès acid mine drainage, France (33), and La Zarza, Spain (34). Ferritrophicum species are not abundant in AMD sites studied to date, with few exceptions (35).
Like most acid mine drainage sites, Scalp Level Run hosts both Fe(II)-oxidizing and Fe(III)-reducing populations. Previous research has shown that there are strong oxygen gradients in the first few centimeters of AMD sediments (e.g., reference 25); therefore, it is not surprising to find both iron-reducing and Fe(II)-oxidizing populations in AMD surface sediments. In these sediments, factors that favor the activity of Fe(II)-oxidizing bacteria over iron-reducing populations could increase the net iron oxidation rate, including light or nutrient availability that enhances oxygenic photosynthesis or a lack of organic matter inputs from the surrounding ecosystem or from algal photoautotrophy at the iron mound surface.
Scalp Level Run surface sediments were colonized by eukaryotic phototrophs, including Stramenophiles, Streptophyta, Euglenozoa, and Chlorophyta. Oxygenic phototrophs, including these taxa, could enhance iron oxidation at AMD sites by increasing the concentration of dissolved oxygen at the sediment-water interface, thereby altering the depth at which iron reduction occurs in the subsurface of the mound. In this scenario, the net Fe(II) oxidation rate would increase due to both an increase in the rate of iron oxidation and a decrease in the rate of iron reduction. This mechanism could explain the higher Fe(II) oxidation rates in alga-dominated than in alga-poor sites (11). The same mechanism could explain why the field iron oxidation rates at Scalp Level Run were 20 times greater than the rates in dark laboratory chemostats inoculated with sediment from Scalp Level Run (36). However, the phototrophs are also producing organic carbon, which could be used in dissimilatory iron reduction. The influence of phototrophy on the system is likely dependent on the turnover over time of both the carbon and oxygen and the activity of the heterotrophs. We do not have sufficient data to quantify the effect of phototrophy on our system.
The results that we obtained from Scalp Level Run place new constraints on emerging ideas about physical and chemical niches of AMD phylotypes. In the ecological niche model for Fe(II)-oxidizing bacteria described by Brown et al. (14) and Senko et al. (10), pH and the concentration of ferrous iron were strongly correlated with the abundance of Fe(II)-oxidizing bacterial populations in Appalachian AMD. Both pH and the concentration of ferrous iron at Scalp Level Run were lower than at the sites used to construct the niche model. Scalp Level Run also provided a unique opportunity to isolate the effects of ferrous iron concentrations on bacterial communities independent of major changes in pH.
Both FISH and 16S rRNA gene sequence tag data show that the bacterial community composition was largely constant across the Scalp Level Run site (Fig. 3A). Of 42 bacterial orders that were detected at Scalp Level Run, 30 were found at all three sampling locations. These 30 orders made up between 98.9% and 99.9% of the community at each sampling location. Moreover, only eight orders were restricted to a single sampling location and all of these “endemic” orders represented rare populations (less than 0.01% of the community). The constant community composition at Scalp Level Run despite large changes in the concentration of ferrous iron suggests that ferrous iron availability is not an important driver of community composition at Scalp Level Run.
Unlike most other AMD sites in the Appalachian region, pH varies by only 0.2 units across the Scalp Level Run iron mound, whereas the concentration of ferrous iron changes over 1 order of magnitude. This situation presented a unique opportunity to test relationships proposed in earlier work at Appalachian coal mine drainages. Whereas Brown et al. (14) and Senko et al. (10) suggested that AMD populations respond to both pH and [Fe(II)], the Scalp Level Run bacterial community composition does not appear to respond to the large changes in ferrous iron concentrations across the site. However, the majority of the sites used to construct the niche model proposed by Brown et al. (14), Senko et al. (10), and others were above pH 3. The pH at Scalp Level Run is low enough to keep most ferric iron in solution. Under these conditions, the precipitation of iron oxides and production of H+ (equation 3) is kinetically inhibited. Therefore, iron oxidation does not immediately lead to a decrease in pH as it does at less acidic AMD springs where ferric iron is insoluble.
Overall, our results indicate that pH is the main controller of the composition of bacterial communities in coal-associated AMD and that Ferrovum, acidophilic Gallionella-like spp., and Acidithiobacillus populations inhabit niches defined by pH, where acidophilic Gallionella-like spp. dominate above pH 3.5 and A. ferrooxidans is most abundant below pH 3 (Fig. 8). Similarly, Kuang and others found that Ferrovum spp. and other Betaproteobacteria were predominant in bacterial assemblages in moderate AMD systems whereas the Alpha- and Gammaproteobacteria and Nitrospira were found in more-acidic environments (37).
Ferrovum myxofaciens appears to be most abundant between pH 3 and 4 but has a broader distribution than either acidophilic Gallionella-like spp. or A. ferrooxidans. This broad distribution may indicate that F. myxofaciens is a more generalist taxon with respect to pH than acidophilic Gallionella-like spp. or Acidithiobacillus spp. Alternately, an as-yet-undiscovered geochemical variable may be a more important driver of Ferrovum abundance than pH. The two Ferrovum populations that we detected at Scalp Level Run were inversely correlated with each other and correlated with the concentration of ferrous iron. These correlations lead us to speculate that the concentration of ferrous iron partially controls within-genus niche specificity among Ferrovum populations, whereas the pH is a primary driver of abundance among higher-level taxonomic units such as Ferrovum, Gallionella, and Acidithiobacillus. This hypothesis remains to be tested in future work.
Scalp Level Run (40°14′43.72″N, 78°51′33.18″W) is an acidic coal mine discharge associated with a former bituminous coal mine in Scalp Level Run, Cambria Co., PA. Polluted water discharges from an artesian spring fed by an abandoned mine and flows down a steep hillside and over a coal tailings pile before joining Little Paint Creek (Fig. 1). Five sites were selected along the flow path from the emergence to the base of the coal tailings pile (Fig. 1). Sediment samples for community characterization and water samples for geochemical characterization were collected aseptically at each of the five sites (SL 0, SL 8.5, SL 25, SL 47, and SL 82). Sediment samples were preserved in RNAlater (Qiagen) in the field and stored at −20°C in the lab until analysis.
Conductivity, oxidation-reduction potential (ORP), pH, dissolved oxygen, and iron concentrations were measured in the field. Dissolved oxygen and ORP were measured using a Hach HQ40d portable meter with an LDO101 dissolved oxygen probe and an MTC101 ORP probe (Hach, Loveland, CO, USA). Temperature, pH, and conductivity were measured using a WTW Multiline pH Conductivity 340 m and associated probes. Ferrous and total iron (method 8008) were measured in the field with a Hach DR 2700 portable spectrophotometer (Hach, Loveland, CO, USA) using the 1,10-phenanthroline method (Hach methods 8146 and 8008, respectively).
DNA from environmental samples was obtained using a MoBio Power Biofilm DNA extraction kit (MoBio, Carlsbad, CA, USA). To minimize DNA extraction bias, each sample was extracted twice, with either a 2- or 4-min vortex bead-beating time. The products from the two extractions were pooled. Clone libraries were constructed to obtain nearly full-length 16S rRNA gene sequences. These sequences were used to clarify the phylogenetic placements of the abundant taxa from the Scalp Level Run community. Small-subunit (16S) rRNA genes were amplified and cloned from the SL emergence sample as described previously (38). Briefly, 16S rRNA genes were amplified using PCR with bacterium-specific primers 27f and 1492r, cloned into chemically competent Escherichia coli OneShot Top10 cells. Cells were plated on and grown in agar containing 0.1% kanamycin to select for cells containing the 16S rRNA gene sequences. Inserts were amplified by PCR using T3 and T7 primers and sequenced at the Penn State University Nucleic Acid Facility using plasmid-specific primers. DNA sequences were assembled and manually checked and trimmed in CodonCode Aligner v. 2 (CodonCode Corporation, Dedham, MA, USA) and identified using NCBI Nucleotide BLAST (39). We picked and sequenced 45 clones. From these, we obtained 43 nearly full-length sequences (see below for accession numbers).
Nearly full-length 16S sequences were aligned with ClustalW (40). Phylogenetic trees were constructed using the maximum likelihood method, with the Tamura-Nei nucleotide substitution model (41). Evolutionary analyses were conducted in MEGA7 (42).
High-throughput sequencing was performed on samples SL 8.5, SL 25, and SL 82 using an Illumina MiSeq as described previously (36, 43). Briefly, the V4 region of the 16S rRNA gene was amplified using primers 515f and 806r with attached sequencing barcodes and adaptors and pooled in equimolar quantities for sequencing. We chose to use bacterium-specific primers because Archaea composed less than 5% of 4′,6-diamidino-2-phenylindole (DAPI)-stained cells (data not shown). Initial quality control was done within the MacQIIME (www.wernerlab.org/software/macqiime) implementation of QIIME 1.7.0 (44). Sequences were split using sample-specific barcodes and trimmed to a minimum q score of 20. Sequences were then processed in mothur (45) as described previously (46). Briefly, sequences shorter than 200 bp and longer than 300 bp, sequences with ambiguous bases, and those with homopolymer sequences greater than 8 bp were removed. Sequences were aligned to the Silva nonredundant database (v119) (47). Chimeras were removed using chimera.uchime (48). Chloroplast and mitochondrion sequences were also removed. From each sample, 10,000 sequences were randomly chosen for downstream analysis to allow even comparisons. Operational taxonomic units (OTUs) were picked using a 97% similarity threshold and the nearest-neighbor method. We picked and sequenced 45 clones. From these, we obtained 43 nearly full-length sequences (see below for availability of the demultiplexed and quality-filtered sequences).
FISH was performed as described previously (11) on all sediment samples. Briefly, environmental samples were fixed in 4% (wt/vol) paraformaldehyde and stored in a 1:1 ethanol-phosphate-buffered saline (ethanol-PBS) solution at −20°C. Samples were stained with taxon-specific probes (Table 1), counterstained with DAPI (Sigma-Aldrich, Darmstadt, Germany), and viewed using epifluorescence microscopy. A minimum of 1,000 DAPI-stained cells were counted for each probe combination. New FISH probes Xan940 and Ferri836 were designed with sequences from clone libraries from Scalp Level Run and other sites according to the protocol described by Hugenholtz et al. (49) using the PROBE_DESIGN tool in ARB (50). The optimum formamide concentration for Xan940 was determined using Scalp Level Run sample SL 8.5 as a positive control. This sample contains 29% Xanthomonadales. E. coli was used as a negative control (5-bp mismatches). The Xan940 probe targets a monophyletic clade of Xanthomonadales sequences retrieved from acidic or mining-related environments. The optimum formamide concentration for Ferri836 was determined using a culture of Ferritrophicum radicicola. The Ferri836 probe targets sequences isolated from acidic environments that are closely related to the Ferritrophicum radicicola. The Ferri836 probe has a 1-bp mismatch to a single nontarget sequence and a 2-bp mismatch to sequences in Thiobacillus and Gallionellales clades (Betaproteobacteria). Unidentified betaproteobacterial sequences were quantified by counting the cells stained by the Betaproteobacterium-specific probe Bet42a and subtracting the relative abundance of Gallionella and Ferrovum.
Linear regressions were performed in R (51) to test the significance of the relationships between the average relative abundances of taxa.
The FISH counts for three AMD taxa (an acidophilic Gallionella sp., Acidithiobacillus ferrooxidans, and Ferrovum myxofaciens) were retrieved from Jones et al. (13) and Brown et al. (14) and combined with the counts from Scalp Level Run. A local polynomial regression (Loess) curve was fitted to the relationship of pH and relative abundance for each taxon using a span value of 1 in the “loess” command in R (51).
We thank I. Schaperdoth for laboratory assistance and A. Vikram for technical assistance with sequencing. We thank Melissa Reckner from the Paint Creek Regional Watershed Association for introducing us to the Scalp Level Run site.
Supplemental material for this article may be found at https://doi.org/10.1128/AEM.00015-17.