Sulfur Isotope Fractionation as an Indicator of Biogeochemical Processes in an AMD Passive Bioremediation System

Sulfate, the main dissolved contaminant in acid mine drainage (AMD), is ubiquitous in watersheds affected by coal and metal mining operations worldwide. Engineered passive bioremediation systems (PBS) are low-cost technologies that remediate sulfate contamination by promoting (1) precipitation of sulfate-bearing compounds, such as schwertmannite and gypsum; and (2) microbially-mediated sulfate reduction (BSR) to sulfide with subsequent precipitation of sulfide minerals. In this study, chemical and sulfur isotopic data are used to infer multiple pathways for sulfate sequestration in the Tab-Simco PBS. By simultaneously monitoring sulfate concentrations and δSSO4 values at four sampling points across the PBS, we (1) identified that the organic layer within the bioreactor was the primary site of BSR processes contributing to sulfate sequestration; (2) observed seasonal variations of BSR processes; (3) estimated that initially the BSR processes contributed up to 30% to sulfate sequestration in the Tab-Simco bioreactor; and (4) determined that BSR contribution to sulfate sequestration continuously declined over the PBS operational lifetime. Together, our results highlight the utility of combining geochemical and microbial fingerprinting techniques to decipher complementary processes involved in sulfur cycling in a PBS as well as the value of adding the sulfur isotope approach as an essential tool to help understand, predict, prevent and mitigate sulfate contamination in AMD-impacted systems.


Introduction
Sulfate (SO 4 2− ) is a major contaminant in many watersheds impacted by coal and metal mining activities and is difficult to remediate [1][2][3][4][5][6].The increased SO 4 2− presence in ecosystems affected by mining has negative environmental and human health effects [7] since SO 4 2− can (1) promote secondary water quality impacts related to sulfur (S) redox processes and biological hydrogen sulfide (H 2 S) production; (2) stimulate methylation of mercury to methylmercury, the most toxic and bioaccumulative form of mercury; (3) enhance biodegradation of organic matter in soils; and (4) promote release of nutrients and potentially toxic compounds from sediments during biologically-mediated sulfate reduction processes [8][9][10][11].Consequently, decreasing SO 4 2− discharge from mine drainages associated with coal and metal mining operations is a priority worldwide [1][2][3][4].
distribution of biotic and abiotic pathways of S cycling and sequestration in a PBS could have major benefits on designing more effective bioremediation systems.This contribution presents geochemical and S isotope data of S-bearing phases in a field-scale PBS treating coal mine AMD.We have employed stable S isotope analyses because this approach is commonly used to discriminate between biologically-mediated sulfate reduction [19][20][21][22][23] and abiological SO 4 2− immobilization in SO 4 2− bearing nNP in natural and engineered systems [24,25] as BSR produces some of the largest sulfur isotope fractionations observed in nature [20].Our previous field experiments showed that S isotopic patterns could provide constraints on the occurrence of BSR in field experiments simulating a PBS [18].In this study, we illustrate that S isotopes, together with additional supporting microbial and geochemical data, can fingerprint BSR and provide insights into the S cycling within a field-scale PBS.Importantly, sulfur isotope patterns can become an essential tool to help understand, predict, prevent and mitigate SO 4 2− contamination in AMD-impacted systems.However, for a quantitative assessment of BSR, additional experimental work is necessary to further resolve the relationships between isotopic fractionation factors between SO 4 2− and microbially-produced H 2 S and relevant environmental parameters in highly metalliferous AMD-impacted environments.

Description of the Tab-Simco Site
Tab-Simco is an abandoned mine land site located ~6 km SE of Carbondale, IL, USA [5].The Tab-Simco coal mine was abandoned in the early 1970s prior to national coal mining environmental regulation and is one of the most highly AMD contaminated sites in the U.S. mid-continent region [26].The site is located on a 12-hectare highly-dissected plateau with underlying underground mine workings into two Pennsylvanian-age coal seams, the Murphysboro and Mt.Rorah Members of the Spoon Formation [27].Weathering of coal in the underground workings resulted in the formation of AMD mine pools.Subsequent contour-type surface mining removed the outcrop barrier that The sulfate-reducing bioreactor was initially built in 2007 with ~5887 m 3 of limestone-amended organic substrate (53% wood chips, 27% straw, 11% seasoned municipal yard waste) and 9% agricultural ground limestone by volume [28].In 2013, after six years of operation, due to decreasing contaminate removal rates, the bioreactor was rehabilitated by replacement of the organic substrate with higher amount of limestone.Two monitoring wells were also completed: B-1 into the underground mine pool and B-2 within mine waste saturated with AMD from the mine pool.Details regarding the site, bioremediation system, and geochemical monitoring are presented in our previous reports [5,6,17,18].

Biogeochemical Sulfur Cycling in the Tab-Simco PBS
Figure 2 presents a simplified conceptual model of S cycling in a typical PBS, such as the Tab-Simco PBS.In the open limestone channel, S sequestration occurs through precipitation of amorphous and crystalline SO4 2− -bearing nNP such as schwertmannite [Fe8O8(OH)6SO4] and jarosite [KFe3(SO4)2(OH)6] and by adsorption and incorporation into the Fe(III)-rich nNP such as goethite (α-FeOOH) and lepidocrocite (γ-FeOOH) [6] (Figure 2a).The presence in AMD of the detrital nanoand micro-scale particles (dNP), originated in the coal mine waste, promotes heterogeneous nucleation and growth of SO4 2− -bearing nNP [6] and thus contributes to increased SO4 2− sequestration.Furthermore, alkalinity generated by limestone dissolution favors the precipitation of nNP.However, such precipitates would also form coatings on limestone, thus inhibiting alkalinity The sulfate-reducing bioreactor was initially built in 2007 with ~5887 m 3 of limestone-amended organic substrate (53% wood chips, 27% straw, 11% seasoned municipal yard waste) and 9% agricultural ground limestone by volume [28].In 2013, after six years of operation, due to decreasing contaminate removal rates, the bioreactor was rehabilitated by replacement of the organic substrate with higher amount of limestone.Two monitoring wells were also completed: B-1 into the underground mine pool and B-2 within mine waste saturated with AMD from the mine pool.Details regarding the site, bioremediation system, and geochemical monitoring are presented in our previous reports [5,6,17,18].(SO 4 ) 2 (OH) 6 ] and by adsorption and incorporation into the Fe(III)-rich nNP such as goethite (α-FeOOH) and lepidocrocite (γ-FeOOH) [6] (Figure 2a).The presence in AMD of the detrital nano-and micro-scale particles (dNP), originated in the coal mine waste, promotes heterogeneous nucleation and growth of SO 4 2− -bearing nNP [6] and thus contributes to increased SO 4 2− sequestration.

Biogeochemical Sulfur Cycling in the Tab-Simco PBS
Furthermore, alkalinity generated by limestone dissolution favors the precipitation of nNP.However, such precipitates would also form coatings on limestone, thus inhibiting alkalinity production [18].
Although the high capacity of nNP to sequester SO 4 2− has been well documented, recent studies have shown that subsequent abiotic and biologically-mediated processes promote recrystallization and/or dissolution of nNP, which results in SO 4 2− remobilization [6,29].
Minerals 2017, 7, 41 4 of 20 production [18].Although the high capacity of nNP to sequester SO4 2− has been well documented, recent studies have shown that subsequent abiotic and biologically-mediated processes promote recrystallization and/or dissolution of nNP, which results in SO4 2− remobilization [6,29].In the bioreactor and the wetland, SO4 2− is sequestered as SO4 2− -bearing nNP and by microbially-mediated reduction of SO4 2− to H2S and precipitation of low-solubility metal-sulfide minerals.Optimal conditions for sulfate reducing microorganisms comprise anaerobic, alkaline settings with moderate salt and metal content as well as the availability of suitable electron donors, which may be either simple organic compounds or molecular hydrogen (H2) [30][31][32].Due to high spatial and temporal variability of environmental conditions in the bioreactor and wetland, zones of optimal BSR are therefore created, with their location and extent controlled by the distribution and nature of the microbial communities, availability of organic carbon, biochemical gradients, and overall local environmental conditions [15,16].BSR-produced sulfide can then react with aqueous Fe(II) and/or with Fe(III)-bearing minerals and precipitate iron sulfides [5].Transport of unreacted sulfide out of the bioremediation system commonly occurs in PBS (Figure 2b).Both dissolved sulfide and sulfide minerals are prone to oxidation either abiotically or by S oxidizing microorganisms, and this S fraction can be remobilized as SO4 2− back into solution.The residual SO4 2− that has not been reduced to H2S or sequestered in nNP will be transported with the treated AMD out of the bioreactor cell.Finally, the continuous retention of the S-bearing phases as precipitates within the bioreactor cell results in mineral accumulations either as coating on reactive surfaces or as individual layers that can cause modification of the environmental conditions and a lowering of permeability in the PBS [6].

Sulfur Isotope Systematics in the Tab-Simco PBS
Sulfur isotope fractionation ( 34 S/ 32 S) has been used to differentiate biological from abiological S pathways, as tracers of sources, mixing processes and transformations of S compounds in a variety In the bioreactor and the wetland, SO 4 2− is sequestered as SO 4 2− -bearing nNP and by microbially-mediated reduction of SO 4 2− to H 2 S and precipitation of low-solubility metal-sulfide minerals.Optimal conditions for sulfate reducing microorganisms comprise anaerobic, alkaline settings with moderate salt and metal content as well as the availability of suitable electron donors, which may be either simple organic compounds or molecular hydrogen (H 2 ) [30][31][32].Due to high spatial and temporal variability of environmental conditions in the bioreactor and wetland, zones of optimal BSR are therefore created, with their location and extent controlled by the distribution and nature of the microbial communities, availability of organic carbon, biochemical gradients, and overall local environmental conditions [15,16].BSR-produced sulfide can then react with aqueous Fe(II) and/or with Fe(III)-bearing minerals and precipitate iron sulfides [5].Transport of unreacted sulfide out of the bioremediation system commonly occurs in PBS (Figure 2b).Both dissolved sulfide and sulfide minerals are prone to oxidation either abiotically or by S oxidizing microorganisms, and this S fraction can be remobilized as SO 4 2− back into solution.The residual SO 4 2− that has not been reduced to H 2 S or sequestered in nNP will be transported with the treated AMD out of the bioreactor cell.Finally, the continuous retention of the S-bearing phases as precipitates within the bioreactor cell results in mineral accumulations either as coating on reactive surfaces or as individual layers that can cause modification of the environmental conditions and a lowering of permeability in the PBS [6].

Sulfur Isotope Systematics in the Tab-Simco PBS
Sulfur isotope fractionation ( 34 S/ 32 S) has been used to differentiate biological from abiological S pathways, as tracers of sources, mixing processes and transformations of S compounds in a variety of natural and engineered systems [18,20].Both biological and abiotic processes can fractionate S isotopes, but microbial activity has been shown to produce larger fractionation in low-temperature, sedimentary environments [20,33].
In the Tab-Simco PBS, S biogeochemical cycle starts with the weathering of sulfide minerals in the coal mine waste.The sulfide minerals oxidation, both through abiogenic and biogenic pathways, produces dissolved SO 4 2− with δ 34 S reflecting the isotopic composition of sulfide minerals in the coal mine waste [18].Within the Tab-Simco PBS, the dissolved SO 4 2− can be removed from the untreated and treated mine drainage through precipitation, co-precipitation, and sorption.Processes such as precipitation/dissolution of a wide range of SO 4 2− -rich minerals and adsorption/desorption of SO 4 2− on surfaces, known to dominate the S cycle in AMD-impacted environments [6], produce relatively small S isotope fractionation between reactants and products [24,25,33].
In the Tab-Simco bioreactor, biologically-mediated S cycling is expected to produce a diverse range of reactive S-species with different oxidation states (Figure 2b) and also generate distinct and often large S isotope fractionations between various reactants and products [15,16,20].For example, the microbially-mediated dissimilatory SO  [20].Residual H 2 S that does not react with metals to form sulfide minerals may be re-oxidized under anoxic and oxic conditions abiotically as well as by a range of microorganisms [17,34].The multi-step oxidation process of H 2 S back to SO 4 2− releases eight electrons and may include the production of intermediate S species, such as sulfite, thiosulfate, and elemental sulfur [33].Fe(III)-nNP, abundantly present in the bioreactor, can mediate the heterolytic dissociation of H 2 S and, in coupled redox reactions, the formation of elemental sulfur (S 0 ) with no isotopic fractionation during complete consumption [20].S 0 , which is relatively stable under the low pH conditions typical of AMD, can be microbially disproportionated to SO  [20].Repeated cycles of SO 4 2− reduction followed by subsequent reoxidation of H 2 S back to SO 4 2− mediated by the abundant Fe(III)-nNP typical for AMD can contribute to a 34 S-depleted S reservoir and large ε 34 S SO 4 -H 2 S kinetic isotope fractionation.Overall, recycling S in its most reduced valence state back to more oxidized forms could play a key, yet unexplored role in the overall S cycling in AMD-impacted systems.Sulfur isotopes, along with relevant physical, geochemical, and microbial data, can potentially be used to differentiate among the multiple and complex S sequestration and mobilization pathways in a PBS.

Microbial Communities in the Tab-Simco PBS
Microorganisms play key roles in the metal and nutrient cycles in AMD-impacted environments due to their ability to adapt to a remarkable range of extreme conditions.With AMD characterized by low pH and high SO 4 2− and Fe concentrations, the major metabolites are the oxidation and reduction of Fe and S for energy, and CO 2 as a carbon source [34][35][36][37][38].
A snapshot of the microbial community present at the Tab-Simco site in August 2008 was obtained through the generation of 16S rRNA gene clone libraries [17].This analysis indicated that bacteria related to Fe-oxidizing Gallionella species dominated the monitoring well B-1 (Figure 3).Recent detection of Gallionella-like sequences in AMD environments suggests that acidophilic members of the genus are important players in the oxidation of sulfide minerals in the coal mine waste and the continuous generation of AMD [38].As the AMD moved from the underground mine system to the bioreactor (Figure 1), the community remained dominated by Fe-oxidizing bacteria (Figure 3).However, the predominant Fe oxidizer shifted to one related to a Ferrovum, a genus that has also been detected in a wide range of AMD sites [39][40][41][42][43].This community analysis also suggested limited biological S cycling in the pretreated AMD drainage as only 2% of the analyzed sequences were related to a potential S oxidizing Thiomonas species and no sequences related to S or SO4 2− reducers were detected (Figure 3).As the treated mine drainage was released from the bioreactor, the community strongly shifted away from one dominated by Fe oxidizers to one supporting biological S cycling.Organisms related to species capable of S and SO4 2− reduction, including the anaerobic Desulfuromonas (S reducer), Desulfotomaculum, and Desulfobacteraceae (family) were detected in the Bioreactor Outlet sample (Figure 3).In conjunction with the presence of reduced S species in the bioreactor, sequences related to the S oxidizer Sulfuricurvum were detected in abundance.Some species of Sulfuricurvum have been shown to completely oxidize S 0 , sulfide, or sulfite to SO4 2− and thrive in high-sulfide, low oxygen microenvironments [44].As the AMD from the bioreactor travelled through the oxidation pond/wetland, the community in the effluent shifted back to one that was similar to the Bioreactor Inlet, in that it was dominated by Fe oxidizing bacteria and chloroplast-like sequences.However, the predominant sequence type detected was more closely related to the Fe-oxidizing genus Sideroxydans.While some acidophilic Sideroxydans species have been isolated, others are limited to circumneutral pH [45].Thus, the Fe oxidizer phylotype shifts are most likely linked to the pH of the sample sites.It should be noted that ~4.5% of the sequences detected in the System Outlet sample were still related to the S oxidizer Sulfuricurvum species, thus suggesting the remaining presence of some reduced S species as an electron source.

Sampling and Geochemical Analyses
Field parameters were measured and water samples collected at the Tab-Simco site over a six-year period at the following sampling locations: (1) Main Seep, the largest AMD seep feeding the bioreactor; (2) Bioreactor Inlet, the point where the AMD collection ditch discharges into the bioreactor cell; (3) Bioreactor Outlet, the point where the treated AMD emerges from the bioreactor cell; and (4) System Outlet, the point where the treated water exits the PBS and enters the receiving As the AMD moved from the underground mine system to the bioreactor (Figure 1), the community remained dominated by Fe-oxidizing bacteria (Figure 3).However, the predominant Fe oxidizer shifted to one related to a Ferrovum, a genus that has also been detected in a wide range of AMD sites [39][40][41][42][43].This community analysis also suggested limited biological S cycling in the pretreated AMD drainage as only 2% of the analyzed sequences were related to a potential S oxidizing Thiomonas species and no sequences related to S or SO 4 2− reducers were detected (Figure 3).As the treated mine drainage was released from the bioreactor, the community strongly shifted away from one dominated by Fe oxidizers to one supporting biological S cycling.Organisms related to species capable of S and SO 4 2− reduction, including the anaerobic Desulfuromonas (S reducer), Desulfotomaculum, and Desulfobacteraceae (family) were detected in the Bioreactor Outlet sample (Figure 3).In conjunction with the presence of reduced S species in the bioreactor, sequences related to the S oxidizer Sulfuricurvum were detected in abundance.Some species of Sulfuricurvum have been shown to completely oxidize S 0 , sulfide, or sulfite to SO 4 2− and thrive in high-sulfide, low oxygen microenvironments [44].As the AMD from the bioreactor travelled through the oxidation pond/wetland, the community in the effluent shifted back to one that was similar to the Bioreactor Inlet, in that it was dominated by Fe oxidizing bacteria and chloroplast-like sequences.However, the predominant sequence type detected was more closely related to the Fe-oxidizing genus Sideroxydans.While some acidophilic Sideroxydans species have been isolated, others are limited to circumneutral pH [45].Thus, the Fe oxidizer phylotype shifts are most likely linked to the pH of the sample sites.It should be noted that ~4.5% of the sequences detected in the System Outlet sample were still related to the S oxidizer Sulfuricurvum species, thus suggesting the remaining presence of some reduced S species as an electron source.

Sampling and Geochemical Analyses
Field parameters were measured and water samples collected at the Tab-Simco site over a six-year period at the following sampling locations: (1) Main Seep, the largest AMD seep feeding the bioreactor; (2) Bioreactor Inlet, the point where the AMD collection ditch discharges into the bioreactor cell; (3) Bioreactor Outlet, the point where the treated AMD emerges from the bioreactor cell; and (4) System Outlet, the point where the treated water exits the PBS and enters the receiving stream [5] (Figure 1).Sample sites 1 and 2 are connected by a limestone-lined ditch that also serves as a collection ditch for a series of smaller AMD seeps.Additional samples were periodically collected from the two monitoring wells: B-1 and B-2.Field parameters, including pH and temperature were measured on unfiltered samples immediately following sample collection using a HI9829 Hanna ® multi-sensor probe (Hanna Instruments, Woonsocket, RI, USA).The pH electrode Hanna HI769828-1 field probe (pH/ORP) was calibrated with Orion pH 1.68, 4.01, and 7.00 buffers and then checked against a pH 10 buffer.Water samples selected for laboratory analyses were filtered through 0.45 µm cellulose acetate filter papers (Millpore ® HAW, Billerica, MA, USA) and stored at 4 • C if not analyzed immediately after collection.The samples were subsequently analyzed for dissolved anions (i.e., SO 4 2− ) by means of ion chromatography (Dionex ® ICS 2000, LabX, Midland, ON, Canada) using an IonPac ® AS18 anion-exchange column (ThermoFisher Scientific, Waltham, MA, USA), [5,18].The standard deviation for standard and duplicate sample analyses was consistently <5% for SO 4 2− analysis.Standard deviation bars are not displayed on figures as the symbol size for individual data points are larger than the standard deviation of the measurement.

Stable Isotope Measurements
The sulfur isotopic composition of dissolved SO 4 2− in untreated and treated mine drainages, pore water SO 4 2− in AMD sediments, and sulfide minerals was measured in selected samples collected at Tab-Simco PBS.Dissolved SO 4 2− was recovered from filtered, acidified, water samples (50 mL) by addition of 0.2 M BaCl 2 solution and precipitation of BaSO 4 .Recovered BaSO 4 was dried, weighed, and retained for isotope analysis [24,25].Sulfide minerals were handpicked from Murphysboro and Mount Rorah coal samples and analyzed directly as mineral powder.In August 2013, when the limestone-amended organic substrate layer in the bioreactor was replaced during system maintenance, we collected additional samples from the bioreactor cell, which included AMD sediments deposited on top of the organic substrate [6] and a composite of the organic substrate materials.Porewater SO 4 2− measurements were obtained by first collecting duplicate samples of sediment segments immediately after the cores were sectioned in the laboratory.Subsequently, these segments were transferred to 50-mL centrifuge tubes and centrifuged for 30 min at 8000 relative centrifugal force (RCF) to isolate the porewater from the sediment solids.Sulfate levels in this supernatant were determined as described for dissolved SO 4 2− .
Additionally, samples of sulfide minerals were collected from the bioreactor organic substrate.These sulfide minerals were separated from the organic matrix, rinsed with DI water, dried in the oven at 50 • C and, finally, analyzed directly as mineral powder.For sulfur stable isotope measurements, aliquots of BaSO 4 , FeS 2 , or FeS were loaded into tin cups, mixed with V 2 O 5 , and then combusted on-line in an EA 1110 elemental analyzer at 1010 • C and analyzed with a Finnigan MAT 252 mass spectrometer.The analytical precision (±1σ) of δ 34 S values was <±0.05 of reference materials, whereas sample reproducibility was typically ±0.2 .All isotopic data are expressed in customary δ 34 S in parts per thousand ( ) relative to Vienna Cañon Diablo Troilite (V-CDT).The following international standards were used for calibration: IAEA-S1 = −0.3, IAEA-S2 = +21.6 , IAEA-S3 = −31.3, NBS-127 = +20.3 .Sulfur isotope results are presented in Table S1.

Coal Composition and Sulfur Compounds in Coal and Coal Mining Waste
In coal, S takes on several forms including organic, elemental, and inorganic [46,47].The inorganic S forms include sulfide and sulfate minerals, with pyrite being the main component in most coals [48,49].The organic S may occur in aromatic or heterocyclic macromolecular S structures, thiol moieties, or thioether units, among others [46].When coal and coal mining waste are exposed to oxygen-rich water during weathering, both organic and inorganic S-bearing compounds are oxidized and/or dissolved causing release of SO 4 2− , the end oxidation product of S-bearing compounds, in order to subsequently migrate into ground and surface water.Murphysboro and Mt.Rorah coal seams in the Illinois Basin have variable total S contents ranging from ~2% in low-S coals to more than 7% in high-S coals [48].Coal mining solid waste (coal refuse) can be even more enriched in S with values up to 20% [49].The ratio of organic to inorganic S can be quite variable, though in many low-S coals, the two S fractions contribute about equally to the total S content.Besides pyrite, Zn-and Pb-sulfides have also been identified [48].Additional inorganic constituents include sulfate minerals (i.e., gypsum, jarosite), elemental sulfur, as well as clay minerals and quartz [46,48].The ubiquitous presence of S in coal and coal mining waste assures a constant influx of SO 4 2− in drainages associated with coal mining.
At Tab-Simco, the AMD overall chemistry reflects the composition of Murphysboro and Mt.Rorah coals; however, the component ratios are dissimilar (Figure 4).In coal, which contains high amounts of clay minerals and quartz, Al and Si predominate, whereas, in AMD, SO 4 2− is the dominant dissolved species followed by Fe, Al and Si [49].Likewise, even though S represents only ~17% of the inorganic molar fraction in Murphysboro coal samples, it makes up ~70% of the total dissolved load in Tab-Simco AMD [18] (Figure 4).This discrepancy is due to weathering patterns of coal and coal mining waste as well as to specific methods employed to measure contaminant transport in AMD.In coal, weathering of pyrite and other S-bearing compounds generate coal mine drainages enriched in SO 4 Murphysboro and Mt.Rorah coal seams in the Illinois Basin have variable total S contents ranging from ~2% in low-S coals to more than 7% in high-S coals [48].Coal mining solid waste (coal refuse) can be even more enriched in S with values up to 20% [49].The ratio of organic to inorganic S can be quite variable, though in many low-S coals, the two S fractions contribute about equally to the total S content.Besides pyrite, Zn-and Pb-sulfides have also been identified [48].Additional inorganic constituents include sulfate minerals (i.e., gypsum, jarosite), elemental sulfur, as well as clay minerals and quartz [46,48].The ubiquitous presence of S in coal and coal mining waste assures a constant influx of SO4 2− in drainages associated with coal mining.
At Tab-Simco, the AMD overall chemistry reflects the composition of Murphysboro and Mt.Rorah coals; however, the component ratios are dissimilar (Figure 4).In coal, which contains high amounts of clay minerals and quartz, Al and Si predominate, whereas, in AMD, SO4 2− is the dominant dissolved species followed by Fe, Al and Si [49].Likewise, even though S represents only ~17% of the inorganic molar fraction in Murphysboro coal samples, it makes up ~70% of the total dissolved load in Tab-Simco AMD [18] (Figure 4).This discrepancy is due to weathering patterns of coal and coal mining waste as well as to specific methods employed to measure contaminant transport in AMD.In coal, weathering of pyrite and other S-bearing compounds generate coal mine drainages enriched in SO4 2− and metals such as Al, Fe, Mn, Zn, Ni and Cr.However, SO4 2− sequestration is typically low in the coal mine waste, since most of the SO4 2− bearing minerals (e.g., gypsum, jarosite and schwertmannite) are relatively unstable and prone to dissolution, which promotes SO4 2− partition into the AMD [49].

Examination of pH Variations
The pH of the AMD influent measured during a six-year monitoring period at the Main Seep displayed values ranging between 2.3 and 3.2 and an average of 2.7 (Figure 5).The pH range was similar in well B-1 (2.7-3.5) and well B-2 (2.7-3.2) [5].This result is surprising since during the same time period large variations of the hydroclimatic parameters were recorded including periods of extreme temperatures, droughts, and intense precipitation events.Because, at Tab-Simco, the AMD originates mostly in the underground mine workings, we can assume that steady-state conditions were maintained in the relatively thick cover of the reclaimed surface coal mine spoil throughout the duration of this study.

Examination of pH Variations
The pH of the AMD influent measured during a six-year monitoring period at the Main Seep displayed values ranging between 2.3 and 3.2 and an average of 2.7 (Figure 5).The pH range was similar in well B-1 (2.7-3.5) and well B-2 (2.7-3.2) [5].This result is surprising since during the same time period large variations of the hydroclimatic parameters were recorded including periods of extreme temperatures, droughts, and intense precipitation events.Because, at Tab-Simco, the AMD originates mostly in the underground mine workings, we can assume that steady-state conditions were maintained in the relatively thick cover of the reclaimed surface coal mine spoil throughout the duration of this study.Similar ranges of pH were recorded at both Bioreactor Inlet and Main Seep suggesting that the limestone channel had a limited capacity to buffer the influent AMD (Figure 5).An analogous finding was documented in our field experiments conducted to evaluate the effects of simple versus complex carbon sources on stimulating microbial sulfate reduction under low-pH conditions, where the limestone-only reactor, which lacked an organic substrate matrix, developed acidic conditions immediately after exposure to AMD [18].In both cases, the passivation of the limestone by both dNP, dominated by clay minerals, and nNP, dominated by goethite, restricted limestone dissolution and alkalinity production [6,18].
The effluent pH measured at Bioreactor Outlet and System Outlet was initially higher than that of Main Seep, with values as high as 6.9 in the Bioreactor Outlet and 7.6 in the System Outlet (Figure 5).A pH > 6 was measured year-round during the first two years of operation (2008-2010) in the Bioreactor Outlet and System Outlet, and the pH subsequently declined to <6 during winter and early spring.However, similar to trends observed in our field experiments [18], the advent of higher temperatures in the following summers led to higher pH values in the Bioreactor Outlet but not in the System Outlet (Figure 5).
Two main processes are responsible for creating alkalinity in the bioreactor and wetland, namely limestone dissolution and the presence of bacterial sulfate reduction.Low temperatures during winter months had a minimal influence on limestone dissolution; nevertheless, they strongly affected biological processes [18].Thus, the decrease in pH in the Bioreactor Outlet and System Outlet during winter was most probably related to minimal bacterial activity and decreased BSR-generated bicarbonate alkalinity during the winter months.The pH temporal trends in Bioreactor Outlet also suggest that the microbial processes responsible for producing alkalinity within the bioreactor were less important after 2011.The overall gradual decrease of pH in Bioreactor Outlet and the fading and actually reversing winter-to-summer pH trend in Bioreactor Outlet through 2013 indicates that the bioreactor was able to buffer the AMD effectively only for the first four out of six years of treatment (Figure 5).In the wetland, lower organic matter availability to sustain BSR processes had a significant impact on its buffering capacity, which was maintained only for the first three years of Tab-Simco PBS operation.Similar ranges of pH were recorded at both Bioreactor Inlet and Main Seep suggesting that the limestone channel had a limited capacity to buffer the influent AMD (Figure 5).An analogous finding was documented in our field experiments conducted to evaluate the effects of simple versus complex carbon sources on stimulating microbial sulfate reduction under low-pH conditions, where the limestone-only reactor, which lacked an organic substrate matrix, developed acidic conditions immediately after exposure to AMD [18].In both cases, the passivation of the limestone by both dNP, dominated by clay minerals, and nNP, dominated by goethite, restricted limestone dissolution and alkalinity production [6,18].
The effluent pH measured at Bioreactor Outlet and System Outlet was initially higher than that of Main Seep, with values as high as 6.9 in the Bioreactor Outlet and 7.6 in the System Outlet (Figure 5).A pH > 6 was measured year-round during the first two years of operation (2008-2010) in the Bioreactor Outlet and System Outlet, and the pH subsequently declined to <6 during winter and early spring.However, similar to trends observed in our field experiments [18], the advent of higher temperatures in the following summers led to higher pH values in the Bioreactor Outlet but not in the System Outlet (Figure 5).
Two main processes are responsible for creating alkalinity in the bioreactor and wetland, namely limestone dissolution and the presence of bacterial sulfate reduction.Low temperatures during winter months had a minimal influence on limestone dissolution; nevertheless, they strongly affected biological processes [18].Thus, the decrease in pH in the Bioreactor Outlet and System Outlet during winter was most probably related to minimal bacterial activity and decreased BSR-generated bicarbonate alkalinity during the winter months.The pH temporal trends in Bioreactor Outlet also suggest that the microbial processes responsible for producing alkalinity within the bioreactor were less important after 2011.The overall gradual decrease of pH in Bioreactor Outlet and the fading and Minerals 2017, 7, 41 10 of 20 actually reversing winter-to-summer pH trend in Bioreactor Outlet through 2013 indicates that the bioreactor was able to buffer the AMD effectively only for the first four out of six years of treatment (Figure 5).In the wetland, lower organic matter availability to sustain BSR processes had a significant impact on its buffering capacity, which was maintained only for the first three years of Tab-Simco PBS operation.

Examination of SO 4 2− Concentration Variations
The SO 4 2− concentrations measured during the six-year period at the Main Seep varied between 2500 and 6200 mg/L with an average value of 4500 mg/L (Figure 6).Lower SO 4 2− concentrations were measured in samples from the two monitoring wells B-1 and B-2, which ranged from 1000 to 3300 mg/L with an average value of 2400 mg/L [5].

Examination of SO4 2− Concentration Variations
The SO4 2− concentrations measured during the six-year period at the Main Seep varied between 2500 and 6200 mg/L with an average value of 4500 mg/L (Figure 6).Lower SO4 2− concentrations were measured in samples from the two monitoring wells B-1 and B-2, which ranged from 1000 to 3300 mg/L with an average value of 2400 mg/L [5].The large temporal variations in SO4 2− concentrations correlated with seasonal variations in hydroclimatic conditions at the site.Specifically, higher SO4 2− concentrations were recorded at the end of extended drought periods (e.g., 24 August 2012) and lower SO4 2− concentrations were measured after high-precipitation events (e.g., 24 March 2013).Previous investigations have also highlighted large seasonal variations of SO4 2− concentrations in AMD due to intense evaporation during drought periods and dilution by rainwater [37,50,51].
The SO4 2− concentration recorded in Bioreactor Inlet samples ranged between 3200 and 4500 mg/L with an average of 3800 mg/L (Figure 6), suggesting that some dissolved SO4 2− may have precipitated out of the AMD along the limestone channel.Although precipitation of S-bearing nNP was documented between Main Seep and Bioreactor Inlet, the lower SO4 2− levels at the Bioreactor Inlet may have been also due to: (1) SO4 2− contaminant transport as particulate (NP) and colloidal forms [6]; or (2) dilution with additional mine pool seepage with lower SO4 2− content [5].
Lower dissolved SO4 2− concentrations were systematically recorded at the Bioreactor Outlet and to an even greater extent at the System Outlet locations as compared to those measured at the Bioreactor Inlet (Figure 6).Overall, SO4 2− concentrations in the Bioreactor Outlet and System Outlet effluent steadily increased over time and, in the last year of PBS operation, approached the Bioreactor Inlet values.The overall contrasting trends of pH and SO4 2− in the Bioreactor Outlet and System Outlet suggest that the presence of bacterial sulfate reduction was critical in producing alkalinity and increasing SO4 2− sequestration in the PBS.

Examination of Eh-pH Diagram
Precipitate formation is the main S sequestration mechanism in Tab-Simco PBS.To gain insight into such processes, we plotted the effluent datasets from Main Seep, Bioreactor Inlet, Bioreactor The large temporal variations in SO 4 2− concentrations correlated with seasonal variations in hydroclimatic conditions at the site.Specifically, higher SO 4 2− concentrations were recorded at the end of extended drought periods (e.g., 24 August 2012) and lower SO 4 2− concentrations were measured after high-precipitation events (e.g., 24 March 2013).Previous investigations have also highlighted large seasonal variations of SO 4 2− concentrations in AMD due to intense evaporation during drought periods and dilution by rainwater [37,50,51].The SO 4 2− concentration recorded in Bioreactor Inlet samples ranged between 3200 and 4500 mg/L with an average of 3800 mg/L (Figure 6), suggesting that some dissolved SO 4 2− may have precipitated out of the AMD along the limestone channel.Although precipitation of S-bearing nNP was documented between Main Seep and Bioreactor Inlet, the lower SO 4 2− levels at the Bioreactor Inlet may have been also due to: (1) SO 4 2− contaminant transport as particulate (NP) and colloidal forms [6]; or (2) dilution with additional mine pool seepage with lower SO 4 2− content [5].
Lower dissolved SO 4 2− concentrations were systematically recorded at the Bioreactor Outlet and to an even greater extent at the System Outlet locations as compared to those measured at the Bioreactor Inlet (Figure 6).Overall, SO Outlet suggest that the presence of bacterial sulfate reduction was critical in producing alkalinity and increasing SO 4 2− sequestration in the PBS.

Examination of Eh-pH Diagram
Precipitate formation is the main S sequestration mechanism in Tab-Simco PBS.To gain insight into such processes, we plotted the effluent datasets from Main Seep, Bioreactor Inlet, Bioreactor Outlet, and System Outlet into the Eh-pH diagram for Fe-S-K-O-H system at 25 • C (Figure 7).Previous reports have listed highly variable thermodynamic data for schwertmannite and ferrihydrite [52][53][54] so the Eh-pH stability fields are largely symptomatic.Outlet, and System Outlet into the Eh-pH diagram for Fe-S-K-O-H system at 25 °C (Figure 7).Previous reports have listed highly variable thermodynamic data for schwertmannite and ferrihydrite [52][53][54] so the Eh-pH stability fields are largely symptomatic.The examination of the diagram (Figure 7) reveals that the effluent data the Main Seep and Bioreactor Inlet are distributed among the stability fields corresponding to Fe 2+ (aq), schwertmannite, and goethite.The prevailing presence of Gallionella species, a Fe-oxidizing chemolithotrophic bacteria, in the well B-1 (Figure 3) suggest that active oxidation processes, sustained AMD production, and likely precipitation of Fe(III)-rich nNP were occurring in the coal mine waste [17].
When AMD from the coal mine waste, enriched in SO4 2− and Fe 2+ , reaches the surface at the Main Seep, Fe 2+ (aq) is oxidized to Fe 3+ both abiotically by molecular oxygen and microbially by Fe-oxidizing bacteria such as Ferrovum species, which were detected in the Tab-Simco open limestone channel (Figure 3).Hydrolysis of Fe 3+ (aq) with subsequent precipitation of nNP such as goethite, schwertmannite, jarosite, or ferrihydrite has been documented by our field studies [6,18].The particular nNP type depends on the Eh-pH conditions and the availability of key elements such as potassium and sulfur (Figure 7).These initially-precipitated nNP are metastable and in time tend to transform to well-crystalized goethite [6].The hydrolysis and precipitation of Fe-rich nNP produce protons and thus are acid generating reactions [6].
Samples from the Bioreactor Outlet plot in the stability fields of pyrite, goethite, schwertmannite, and nano-crystalline Fe(OH)3, suggesting that there is heterogeneity in terms of bioreactor treatment performance.The Eh-pH values in the Bioreactor Outlet are not directly representative for the redox conditions inside the bioreactor because the partially-treated AMD discharge was unavoidably exposed to the atmosphere due to the design of the discharge pipe [5].
With this caveat, the data suggest that the bioreactor promoted the sequestration of SO4 2− both as sulfide and SO4 2− -rich nNP.Microorganisms related to species capable of sulfur (e.g., anaerobic Desulfuromonas) and sulfate (e.g., Desulfotomaculum, and Desulfobacteraceae) reduction were detected in the Bioreactor Outlet (Figure 3), suggesting that active BSR processes occurred in the bioreactor cell [17].Furthermore, the presence of S-bearing precipitates was confirmed when the bioreactor media was replaced in August 2013.The organic substrate contained micron-sized iron sulfides as The examination of the diagram (Figure 7) reveals that the effluent data for the Main Seep and Bioreactor Inlet are distributed among the stability fields corresponding to Fe 2+ (aq) , schwertmannite, and goethite.The prevailing presence of Gallionella species, a Fe-oxidizing chemolithotrophic bacteria, in the well B-1 (Figure 3) suggest that active oxidation processes, sustained AMD production, and likely precipitation of Fe(III)-rich nNP were occurring in the coal mine waste [17].
When AMD from the coal mine waste, enriched in SO 4 2− and Fe 2+ , reaches the surface at the Main Seep, Fe 2+ (aq) is oxidized to Fe 3+ both abiotically by molecular oxygen and microbially by Fe-oxidizing bacteria such as Ferrovum species, which were detected in the Tab-Simco open limestone channel (Figure 3).Hydrolysis of Fe 3+ (aq) with subsequent precipitation of nNP such as goethite, schwertmannite, jarosite, or ferrihydrite has been documented by our field studies [6,18].The particular nNP type depends on the Eh-pH conditions and the availability of key elements such as potassium and sulfur (Figure 7).These initially-precipitated nNP are metastable and in time tend to transform to well-crystalized goethite [6].The hydrolysis and precipitation of Fe-rich nNP produce protons and thus are acid generating reactions [6].
Samples from the Bioreactor Outlet plot in the stability fields of pyrite, goethite, schwertmannite, and nano-crystalline Fe(OH) 3 , suggesting that there is heterogeneity in terms of bioreactor treatment performance.The Eh-pH values in the Bioreactor Outlet are not directly representative for the redox conditions inside the bioreactor because the partially-treated AMD discharge was unavoidably exposed to the atmosphere due to the design of the discharge pipe [5].
With this caveat, the data suggest that the bioreactor promoted the sequestration of SO 4 2− both as sulfide and SO 4 2− -rich nNP.Microorganisms related to species capable of sulfur (e.g., anaerobic Desulfuromonas) and sulfate (e.g., Desulfotomaculum, and Desulfobacteraceae) reduction were detected in the Bioreactor Outlet (Figure 3), suggesting that active BSR processes occurred in the bioreactor cell [17].Furthermore, the presence of S-bearing precipitates was confirmed when the bioreactor media was replaced in August 2013.The organic substrate contained micron-sized iron sulfides as well as SO 4 2− -rich nNP found primarily as coating the organic matter and limestone surfaces.Additionally, a sediment layer with an average thickness of 0.5 m was found on top of the organic substrate.This layer developed due to the accumulation of both transported and within-cell precipitated NP [6].Samples from the System Outlet plot predominantly in the stability field of nNP, namely goethite, schwertmannite, and nano-crystalline Fe(OH) 3 with some samples hosted in the Fe 2+  (aq) stability field.This is in accordance with active Fe 2+  (aq) oxidation processes occurring in the wetland mediated by either molecular oxygen and/or the Fe-oxidizing Betaproteobacteria Sideroxydans (Figure 3) with subsequent precipitation of SO 4 2− -rich nNP.Additional reduction of Fe 3+ (aq) and/or Fe 3+ -bearing nNP in the wetland by members of the Acidiphilium genus could have entered some Fe 2+  (aq) in the System Outlet effluent.The S-bearing phases in Murphysboro and Mt.Rorah coals exhibit δ 34 S ranging from −2.7 to +19.8 [48], which are within the range (δ 34 S = −10.2 to +26.8 ) reported for coals from the Illinois Basin [46][47][48][49]55,56].The organic sulfur (S-or) and elemental sulfur (S-el) fractions in both for Murphysboro and Mt.Rorah coals were somewhat enriched in 34 S (δ 34 S = +3.4 to +19.8 ) whereas pyrite (S-py), water-soluble SO 4 2− (S-ws) and acid soluble SO 4 2− (S-as) fractions were 34 S-depleted (δ 34 S = −2.7 to +14.1 ) (Figure 8).The variability in δ 34 S for pyritic sulfur (δ 34 S = −1.1 to +12.7 ) was due to the presence of several generations of pyrite formed during the different stages of coalification as well as the impact of hydrothermal fluids that precipitated reduced sulfur in Illinois coals [48,49].Detailed information about the sequential extraction method used to extract different S fractions from coal samples and an in-depth discussion of the results were presented by Singh [48].
Minerals 2017, 7, 41 12 of 20 well as SO4 2− -rich nNP found primarily as coating the organic matter and limestone surfaces.Additionally, a sediment layer with an average thickness of 0.5 m was found on top of the organic substrate.This layer developed due to the accumulation of both transported and within-cell precipitated NP [6].Samples from the System Outlet plot predominantly in the stability field of nNP, namely goethite, schwertmannite, and nano-crystalline Fe(OH)3 with some samples hosted in the Fe 2+ (aq)  stability field.This is in accordance with active Fe 2+ (aq) oxidation processes occurring in the wetland mediated by either molecular oxygen and/or the Fe-oxidizing Betaproteobacteria Sideroxydans (Figure 3) with subsequent precipitation of SO4 2− -rich nNP.Additional reduction of Fe 3+ (aq) and/or Fe 3+ -bearing nNP in the wetland by members of the Acidiphilium genus could have entered some Fe 2+ (aq) in the System Outlet effluent.

Sulfur Isotope Composition of Dissolve Sulfate in AMD
During the six-year monitoring period, AMD samples from Main Seep, well B-1, and well B-2 showed narrow δ 34 SSO4 ranges (δ 34 S = +5.3‰ to +7.6‰) (Figure 9a), despite the wide δ 34 S ranges of S-bearing phases in Murphysboro and Mt.Rorah coals (Figure 8).A similar pattern was reported at an abandoned coal mine site in Indiana, where the weathering of coal containing S-bearing phases

Sulfur Isotope Composition of Dissolve Sulfate in AMD
During the six-year monitoring period, AMD samples from Main Seep, well B-1, and well B-2 showed narrow δ 34 S SO 4 ranges (δ 34 S = +5.3 to +7.6 ) (Figure 9a), despite the wide δ 34 S ranges of S-bearing phases in Murphysboro and Mt.Rorah coals (Figure 8).A similar pattern was reported at an abandoned coal mine site in Indiana, where the weathering of coal containing S-bearing phases characterized by large δ 34 S variations (δ 34 S = −24.4 to +9.9 ) produced an AMD with a δ 34 S SO 4 narrow range (δ 34 S = +8 to +10 ) [50].Most probably, the gradual weathering of the two main S-bearing phases in the coal and coal mine waste, namely the 34 S-enriched organic-S and 34 S-depleted pyritic-S, resulted in a homogeneous δ 34 S signature of dissolved SO 4 2− in AMD.
the S cycling was mostly dominated by processes that resulted in minor isotope fractionation such as abiogenic or biogenic sulfide oxidation in the coal mine waste [20,24], adsorption/desorption of SO4 2− onto available surfaces, and precipitation/dissolution of SO4 2− -rich nNP [6], which produced relatively small S isotope fractionation between reactants and products [24,25,56].In contrast, the wider δ 34 SSO4 ranges of Bioreactor Outlet and System Outlet effluents, which also overlap and contained the most positive (Figure 9a) and variable isotopic values (Figure 10a), suggested that BSR occurred in these systems.The δ 34 SSO4 range was +5.9‰ to +10.8‰ for Bioreactor Outlet and +5.9‰ to +10.4‰ for System Outlet effluent samples.The S cycle was further investigated in 2013 when the organic substrate of the bioreactor was replaced.At that time, we collected water samples from the bioreactor acid pond and Bioreactor Outlet and samples from the AMD sediments and the organic substrate layer [6].Distinct δ 34 S ranges were determined for the dissolved SO4 2− in the acidic water layer that overlies the organic substrate (δ 34 S = +6.1‰ to +6.4‰), porewater SO4 2− in the upper Fe-rich sediments (δ 34 S = +4.4‰ to +5.9‰), porewater SO4 2− in the lower Al-rich sediments (δ 34 S = +3.7‰ to +4.7‰), sulfide minerals embedded in the organic substrate (δ 34 S = −2.3‰ to +4.6‰), and the porewater SO4 2− in the organic substrate (δ 34 S = +6.9‰ to +7.8‰) (Figure 9a).Within the Tab-Simco PBS, the widest range of δ 34 S (δ 34 S = −2.3‰ to +10.8‰) was associated with the bioreactor (Figure 9).The observed δ 34 S patterns can be best explained by an active, biologically-mediated S cycle that involved both oxidizing and reducing pathways.The restricted presence of 34 S-depleted sulfide minerals in the organic substrate suggests that BSR processes were limited to this layer within the bioreactor.Within the organic substrate, BSR-produced H2S reacted with Fe 2+ to precipitate sulfides and left behind a 34 SSO4-enriched pool (Figure 9b).Partial re-oxidation of 34 S-depleted H2S back to SO4 2− also occurred in the bioreactor as representative chemolithoautotrophic, S The narrow and overlapping δ 34 S SO 4 ranges of wells B-1 and B-2, Main Seep, and Bioreactor Inlet (Figure 9) suggest that, in the underground mine pool, coal mine waste, and open limestone channel, the S cycling was mostly dominated by processes that resulted in minor isotope fractionation such as abiogenic or biogenic sulfide oxidation in the coal mine waste [20,24], adsorption/desorption of SO 4 2− onto available surfaces, and precipitation/dissolution of SO 4 2− -rich nNP [6], which produced relatively small S isotope fractionation between reactants and products [24,25,56].In contrast, the wider δ 34 S SO 4 ranges of Bioreactor Outlet and System Outlet effluents, which also overlap and contained the most positive (Figure 9a) and variable isotopic values (Figure 10a), suggested that BSR occurred in these systems.The δ 34 S SO 4 range was +5.9 to +10.8 for Bioreactor Outlet and +5.9 to +10.4 for System Outlet effluent samples.The S cycle was further investigated in 2013 when the organic substrate of the bioreactor was replaced.At that time, we collected water samples from the bioreactor acid pond and Bioreactor Outlet and samples from the AMD sediments and the organic substrate layer [6].Distinct δ 34 S ranges were determined for the dissolved SO 4 2− in the acidic water layer that overlies the organic substrate (δ 34 S = +6.1 to +6.4 ), porewater SO 4 2− in the upper Fe-rich sediments (δ 34 S = +4.4 to +5.9 ), porewater SO 4 2− in the lower Al-rich sediments (δ 34 S = +3.7 to +4.7 ), sulfide minerals embedded in the organic substrate (δ 34 S = −2.3 to +4.6 ), and the porewater SO 4 2− in the organic substrate (δ 34 S = +6.9 to +7.8 ) (Figure 9a).
Temporal trends of sulfur isotopes were monitored both as δ 34 SSO4 at the four monitoring points (Figure 10a) and as sulfur isotopic differences between the Main Seep and the Bioreactor Inlet, Bioreactor Outlet, and System Outlet, which are reported in capital delta notation Δ 34 S (Figure 10b).The initial Δ 34 S values were relatively large (Δ 34 SBio Out-MS ≥ +4‰ and Δ 34 SSystem Out-MS ≥ +3.8‰), which gradually decreased to negative Δ 34 S in the winter of 2013, in contrast to the small variations (Δ 34 SBio In-MS ≤ +1.3‰) recorded at the Bioreactor Inlet over the same time period (Figure 10b).Within the Tab-Simco PBS, the widest range of δ 34 S (δ 34 S = −2.3 to +10.8 ) was associated with the bioreactor (Figure 9).The observed δ 34 S patterns can be best explained by an active, biologically-mediated S cycle that involved both oxidizing and reducing pathways.The restricted presence of 34 S-depleted sulfide minerals in the organic substrate suggests that BSR processes were limited to this layer within the bioreactor.Within the organic substrate, BSR-produced H 2 S reacted with Fe 2+ to precipitate sulfides and left behind a 34 S SO 4 -enriched pool (Figure 9b).Partial re-oxidation of 34 S-depleted H 2 S back to SO 4 2− also occurred in the bioreactor as representative chemolithoautotrophic, S oxidizing bacteria of genus Sulfuricurvum were found in the Bio Outlet (Figure 3).Similarly, when unreacted H 2 S upwelled from the organic substrate, as it passed through the overlaying AMD sediment layer deposited on the top of the organic substrate [6], it was oxidized by S oxidizing microorganisms back to SO 4 2− .The 34 S-depleted porewater SO 4 2− in the Fe-and Al-rich layers (Figure 9b) were most probably the result of mixing between the 34 S-depleted porewater SO 4 2− resulting from H 2 S oxidation and the SO 4 2− from the downward infiltration of acidic water that pool on top of the bioreactor.
Overall, the magnitude of 34 S enrichment of the effluent SO 4 2− is an indication of the presence of bacterial sulfate reduction as the AMD was treated in the Tab-Simco bioreactor [15,16].

Sulfur Isotope Temporal Trends
Temporal trends of sulfur isotopes were monitored both as δ 34 S SO 4 at the four monitoring points (Figure 10a) and as sulfur isotopic differences between the Main Seep and the Bioreactor Inlet, Bioreactor Outlet, and System Outlet, which are reported in capital delta notation ∆ 34 S (Figure 10b).The initial ∆ 34 S values were relatively large (∆ 34 S Bio Out-MS ≥ +4 and ∆ 34 S System Out-MS ≥ +3.8 ), which gradually decreased to negative ∆ 34 S in the winter of 2013, in contrast to the small variations (∆ 34 S Bio In-MS ≤ +1.3 ) recorded at the Bioreactor Inlet over the same time period (Figure 10b).
These trends are consistent with the S cycling being dominated by the inorganic precipitation processes in the limestone channel and the presence of bacterial sulfate reduction in the bioreactor and possibly in the wetland.Significantly, the ∆ 34 S Bio Out-MS trends (Figure 10b), which showed sharp decreases during low-temperature periods and the overall gradual decreases in ∆ 34 S Bio Out-MS values, suggest that BSR processes were characterized by large temporal and spatial variabilities.The seasonal trends in S isotope are related to those of pH (decreasing) and SO 4 2− (increasing) that can be explained by the temperature dependence of BSR processes [18,37,57,58].The negative ∆ 34 S Bio Out-MS recorded in winter of 2012 and 2013 are indicative of the addition of a 34 S SO 4 -depleted pool resulted probably from re-oxidation of 34 S-depleted sulfide minerals [59] within the bioreactor.The presence of diverse S oxidizing bacteria communities in the Bioreactor Outlet (Figure 3) suggests that sulfide re-oxidation was an important component of the S cycle within the bioreactor (Figure 2).The steady build-up of Fe(III)-nNP in the bioreactor could also favor the re-oxidation of H 2 S and thus the decrease in bioreactor capacity to sequester the SO 4 2− contaminant.The gradual accumulation of Fe(III)-nNP precipitates in the bioreactor was probably substantial during the winter time when the low-temperature and low-pH conditions negatively affected the BSR activities but not the inorganic precipitation of SO 4 2− bearing nNP [18,60].Significantly, the positive ∆ 34 S Bio Out-MS values recorded during the warm periods suggest that BSR rebounded every spring with the advent of higher temperatures.One of the most notable results of our investigation is that even in the summer of 2013, before the bioreactor organic substrate was replaced, the ∆ 34 S Bio Out-MS rebounded from −0.5 during winter to ≥+1.0 in the summer.However, the steep decrease in ∆ 34 S Bio Out-MS from +4.2 in April 2008 to +1.2 in April 2013 suggest that the contribution of BSR processes in sequestrating S within the bioreactor diminished over time.This trend was superimposed on the overall diminished SO 4 2− sequestration in the Tab-Simco PBS.
In the case of the oxidation pond and aerobic wetland, the persistent lower ∆ 34 S System Out-MS compared to ∆ 34 S Bio Out-MS (Figure 10b) suggest that BSR processes, if present, had minimal influence on δ 34 S SO 4 values.On the contrary, these isotopic differences indicate that oxidation of dissolved and particulate sulfides originating in the bioreactor was an important part of the S cycling due to the continuous additions of a 34 S−depleted pool to dissolved SO 4 2− already present in the wetland.
Additional inorganic processes involving dissolved SO 4 2− , such as precipitation and dissolution of nNP were concurrently taking place; however, such processes cannot be quantified using S isotopes because they have had a minimal impact on S isotope variations.

Using Sulfur Isotopes to Fingerprint BSR Processes
The ability to identify the presence of bacterial sulfate reduction in a complex and dynamic S cycle in a bioremediation system and quantify the formation of various reaction products is critical in monitoring the SO 4 2− remediation efficiency in any PBS.This study revealed that during AMD treatment in the Tab-Simco bioreactor a significant fraction of SO 4 2− was reduced through BSR, thus leading to a significant increase in the δ 34 S of the residual SO 4 2− .This is a very promising result and suggests that S isotope monitoring can provide an additional tool for investigating S cycling in a bioremediation system by enabling the detection and monitoring BSR processes.Significantly, this approach integrates, in a single assessment, both the SO 4 2− concentrations and the δ 34 S SO 4 , which can be further used to estimate the contribution of BSR to S sequestration in a PBS.The competition for S sequestration among SO 4 -bearing nNP will have a minor effect on such an approach since sulfate mineral precipitation produces minor inorganic-induced shifts in δ 34 S.
Quantitative estimates of BSR involved in S sequestration can maximize the usefulness of S isotopes approach and require knowledge of the isotope fractionation factors (ε 34 S SO 4 -H 2 S ).Previous research has shown that the extent of S isotope fractionation between sulfate and sulfide during BSR varies and depends on (i) the rate of metabolism by bacteria; (ii) the type and abundance of microbial sulfate reducers; (iii) the concentration and nature of the available electron donors; (iv) the thermal regime; and (v) the rate of sulfide removal.
Laboratory experiments that determined ε 34 S SO 4 -H 2 S produced by BSR have found that most of the values cluster around ε 34 S SO 4 -H 2 S ≈ +25 ± 10 [57].Additional research on pure cultures found an inverse relationship between the cell specific rate of sulfate reduction (csSRR, fmol/cell•day) and the fractionation factor [57].Large ranges of ε 34 S SO 4 -H 2 S were found to be a function of the available electron donor type and its rate of delivery.For example when a single organism isolated from marine coastal sediments (Desulfovibrio sp.strain DMSS-1) utilized different electron donors at BSR cell specific rates spanning two orders of magnitude, ε 34 S SO 4 -H 2 S varied from −6 to −66 [61,62].Furthermore, by varying the delivery rate of a single electron donor (lactate) for a single organism (Desulfovibrio vulgaris Hildenborough), Leavitt et al. [63]  These estimates are broad, as many additional abiotic and microbial processes could have contributed to a complex S cycling in the bioreactor.Further work is needed to specifically investigate the relationships between ε 34 S SO 4 -H 2 S and environmental parameters in highly metalliferous AMD environments, including the types of BSR communities and electron donors, the cell specific rate of SO 4 2− reduction, and temperature, before they can be incorporated into robust quantitative models.Nonetheless, the result of the isotope mass-balance calculations underscores the importance of BSR processes as well as their decreasing contribution over time to S sequestration in the Tab-Simco bioreactor.

Conclusions
This study highlights the key advantages of incorporating S isotopes in monitoring S cycling and sequestration in a PBS.By corroborating the sulfur isotope composition of dissolved sulfate and sulfide minerals with water chemistry, microbiology, and mineralogy data, we were able to differentiate between the two main mechanisms of S sequestration, namely (1) the precipitation of SO 4 2− -bearing nNP and (2) the microbially-mediated sulfate reduction to sulfide and subsequent precipitation of sulfide minerals.Simultaneous monitoring of both SO 4 2− concentrations and δ 34 S SO 4 helped us (1) identify the areas within the PBS where BSR processes were actively important and had contributed to S sequestration; (2) highlight the seasonality of BSR processes and therefore their dependence on temperature; and (3) estimate that the contribution of the BSR processes to S sequestration in the Tab-Simco bioreactor decreased from up to 30% initially to less than 10% at the end of the six-year study period.
To increase the accuracy of the S isotope approach and its use as indicator of BSR processes in PBS, additional investigations are needed into the complex S redox pathways in the AMD-impacted environments, including the role and fate of intermediate S species in redox processes and S sequestration pathways.Further work is also needed on constraining the magnitude of isotope fractionations associated with specific microbial communities across different temperatures and with varying relevant electron donors.Additionally, incorporating oxygen isotopes of dissolved SO 4 2− could help to better refine redox S pathways including fingerprinting disproportionation of intermediate S species in the bioreactor since S isotopes alone are insufficient to discriminate among these processes.Our results suggest that the further applications and development of this technique are warranted and can be employed in monitoring PBS treating drainages associated with coal and metal mining operations worldwide.

Figure 1 .
Figure 1.Plan view of the Tab-Simco site with sample locations indicated: the Main Seep (MS), Bioreactor Inlet (Bio In), Bioreactor Outlet (Bio Out), and the System Outlet (System Out).

Figure 1 .
Figure 1.Plan view of the Tab-Simco site with sample locations indicated: the Main Seep (MS), Bioreactor Inlet (Bio In), Bioreactor Outlet (Bio Out), and the System Outlet (System Out).

Figure 2
Figure 2 presents a simplified conceptual model of S cycling in a typical PBS, such as the Tab-Simco PBS.In the open limestone channel, S sequestration occurs through precipitation of amorphous and crystalline SO 4 2− -bearing nNP such as schwertmannite [Fe 8 O 8 (OH) 6 SO 4 ] and jarosite[KFe 3 (SO 4 ) 2 (OH) 6 ] and by adsorption and incorporation into the Fe(III)-rich nNP such as goethite (α-FeOOH) and lepidocrocite (γ-FeOOH)[6] (Figure2a).The presence in AMD of the detrital nano-and micro-scale particles (dNP), originated in the coal mine waste, promotes heterogeneous nucleation and growth of SO 4 2− -bearing nNP[6] and thus contributes to increased SO 4 2− sequestration.

Figure 2 .
Figure 2. Simplified sulfur cycling and sequestration in a typical passive bioremediation system components: (a) the open limestone channel; and (b) the bioreactor and/or the wetland.

Figure 2 .
Figure 2. Simplified sulfur cycling and sequestration in a typical passive bioremediation system components: (a) the open limestone channel; and (b) the bioreactor and/or the wetland.

Minerals 2017, 7 , 41 6 of 20 Figure 3 .
Figure 3. Abundance of bacterial 16S rRNA gene sequences in water samples collected at four sampled sites at the Tab-Simco site (data from Burns et al. [17]).

Figure 3 .
Figure 3. Abundance of bacterial 16S rRNA gene sequences in water samples collected at four sampled sites at the Tab-Simco site (data from Burns et al. [17]).
2− and metals such as Al, Fe, Mn, Zn, Ni and Cr.However, SO 4 2− sequestration is typically low in the coal mine waste, since most of the SO 4 2− bearing minerals (e.g., gypsum, jarosite and schwertmannite) are relatively unstable and prone to dissolution, which promotes SO 4 2− partition into the AMD [49].Minerals 2017, 7, 41 8 of 20

Figure 4 .
Figure 4. Percentage distribution and average concentrations of the main inorganic elements in Murphysboro coal and Tab-Simco acid mine drainage.

Figure 4 .
Figure 4. Percentage distribution and average concentrations of the main inorganic elements in Murphysboro coal and Tab-Simco acid mine drainage.

Minerals 2017, 7 , 41 9 of 20 Figure 5 .
Figure 5. Temporal trends for pH measured in water samples collected at the Main Seep, Bioreactor Inlet (Bio In), Bioreactor Outlet (Bio Out), and the System Outlet (System Out).

Figure 5 .
Figure 5. Temporal trends for pH measured in water samples collected at the Main Seep, Bioreactor Inlet (Bio In), Bioreactor Outlet (Bio Out), and the System Outlet (System Out).

Figure 6 .
Figure 6.Temporal trends for the concentrations of dissolved sulfate (SO4 2− ) in water samples collected at the Main Seep, Bioreactor Inlet (Bio In), Bioreactor Outlet (Bio Out), and the System Outlet (System Out).

Figure 6 .
Figure 6.Temporal trends for the concentrations of dissolved sulfate (SO 4 2− ) in water samples collected at the Main Seep, Bioreactor Inlet (Bio In), Bioreactor Outlet (Bio Out), and the System Outlet (System Out).

Figure 7 .
Figure 7. Eh-pH diagram for Fe-S-K-O-H system at 25 °C.Stability fields for the mineral phases were plotted using available thermodynamic data from Bigham and Nordstrom [29].Plotted data represent water samples collected at the Main Seep, Bioreactor Inlet (Bio In), Bioreactor Outlet (Bio Out), and System Outlet (System Out).

Figure 7 .
Figure 7. Eh-pH diagram for Fe-S-K-O-H system at 25 • C. Stability fields for the mineral phases were plotted using available thermodynamic data from Bigham and Nordstrom [29].Plotted data represent water samples collected at the Main Seep, Bioreactor Inlet (Bio In), Bioreactor Outlet (Bio Out), and System Outlet (System Out).

Figure 8 .
Figure 8. Range of sulfur isotopes values of sulfur fractions from Murphysboro and Mt.Rorah coal samples: organic sulfur (S-or), elemental sulfur (S-el), acid-soluble sulfur (S-as), water-soluble sulfur (S-ws) and pyritic sulfur (S-py).Data S-py are tabulated in Appendix A1 and previously published data for S-or, S-el, S-as, and S-ws are from Singh [48].

Figure 8 .
Figure 8. Range of sulfur isotopes values of sulfur fractions from Murphysboro and Mt.Rorah coal samples: organic sulfur (S-or), elemental sulfur (S-el), acid-soluble sulfur (S-as), water-soluble sulfur (S-ws) and pyritic sulfur (S-py).Data S-py are tabulated in Appendix A1 and previously published data for S-or, S-el, S-as, and S-ws are from Singh [48].

Figure 9 .
Figure 9. Ranges of sulfur isotopes values of (a) dissolved sulfate (SO4 2− ) in water samples collected at different sampling points across the Tab-Simco PBS; and (b) dissolved and solid sulfur fractions from the Tab-Simco bioreactor collected in August 2013 when the bioreactor media was replaced.

Figure
Figure Ranges of sulfur isotopes values of (a) dissolved sulfate (SO 4 2− ) in water samples collected at different sampling points across the Tab-Simco PBS; and (b) dissolved and solid sulfur fractions from the Tab-Simco bioreactor collected in August 2013 when the bioreactor media was replaced.

Figure 10 .
Figure 10.Temporal trends for (a) the sulfur isotope values of dissolved sulfate (δ 34 SSO4) in water samples collected from the Main Seep (MS), Bioreactor Inlet (Bio In), Bioreactor Outlet (Bio Out), and the System Outlet (System Out); and the (b) difference between the sulfur isotopic values of the Bio In, Bio Out, and System Out effluent dissolved sulfate and MS influent dissolved sulfate (Δ 34 SSO4).

Figure 10 .
Figure 10.Temporal trends for (a) the sulfur isotope values of dissolved sulfate (δ 34 S SO 4 ) in water samples collected from the Main Seep (MS), Bioreactor Inlet (Bio In), Bioreactor Outlet (Bio Out), and the System Outlet (System Out); and the (b) difference between the sulfur isotopic values of the Bio In, Bio Out, and System Out effluent dissolved sulfate and MS influent dissolved sulfate (∆ 34 S SO 4 ).
found a ~50-fold change in BSR rate with ε 34 S SO 4 -H 2 S varying from −11 to −55 .Temperature is also an important player in BSR since it controls both the SO 4 2− reduction rates and the magnitude of ε 34 S SO 4 -H 2 S .Using two strains of Archaeoglobus fulgidus, a hyperthermophilic sulfate reducer that was isolated from hot oil field production waters in the North Sea, Mitchell et al. [58] determined ε 34 S SO 4 -H 2 S, which ranged between −27 and −0.5 , with the largest fractionations found at intermediate temperatures and the smallest fractionations at the lowest and highest temperatures.In natural environments such as naturally occurring sediments and water bodies, most S fractionation factor values are clustered at +45 ± 10 [20].Generally, the laboratory and field experiments have stressed the intricate relationships between the measured ε 34 S SO 4 -H 2 S values and specific environmental conditions.To estimate the contribution of BSR processes to S sequestration in the Tab-Simco bioreactor, we performed a relatively simple isotope mass-balance calculation using the SO 4 2− concentrations for the Bioreactor Inlet and Bioreactor Outlet sampling points and assuming that the ∆ 34 S Bio Out-Bio In values are the sole result of BSR processes.Then, we ran our simplified model under two scenarios for which we selected two different ε 34 S SO 4 -H 2 S values, one representative for the laboratory experiments and the other for natural settings, namely +25 and +45 , respectively.These values were chosen because they span the temperature range of our field system.We are not proposing a new fractionation regime for BSR, but rather illustrating how the models respond to different assumptions about ε 34 S SO 4 -H 2 S .Under the first scenario (ε 34 S SO 4 -H 2 S = +25 ), our calculations indicate that the contribution of BSR to S sequestration had decreased from 30% on 27 August 2008, to 27% on 18 July 2010, to 25% on 11 November 2011 and finally to 15% on 24 May 2013.Under the second scenario (ε 34 S SO 4 -H 2 S = +45 ), the BSR contribution to S sequestration decreased from 21% on 27 August 2008, to 18% on 18 July 2010, to 17% on 11 July 2011 and finally to <10% on 24 May 2013.