Rapid Iron Reduction Rates Are Stimulated by High-AmplitudeRedox Fluctuations in a Tropical Forest SoilBrian Ginn,† Christof Meile,‡ Jared Wilmoth,† Yuanzhi Tang,§ and Aaron Thompson*,†

†University of Georgia, Crop and Soil Sciences, Athens, Georgia 30602, United States‡University of Georgia, Marine Sciences, Athens, Georgia 30602, United States§Georgia Institute of Technology, Earth and Atmospheric Sciences, Atlanta, Georgia 30332, United States

*S Supporting Information

ABSTRACT: Iron oxides are important structural and biogeochemicalcomponents of soils that can be strongly altered by redox-drivenprocesses. This study examined the influence of temporal oxygenvariations on Fe speciation in soils from the Luquillo Critical ZoneObservatory (Puerto Rico). We incubated soils under cycles of oxic-anoxic conditions (τoxic:τanoxic = 1:6) at three frequencies with andwithout phosphate addition. Fe(II) production, P availability, and Femineral composition were monitored using batch analytical andspectroscopic techniques. The rate of soil Fe(II) production increasedfrom ∼3 to >45 mmol Fe(II) kg−1 d−1 over the experiment with aconcomitant increase of an Fe(II) concentration plateau within eachanoxic period. The apparent maximum in Fe(II) produced is similar inall treatments, but was hastened by P-amendment. Numerical modelingsuggests the Fe(II) dynamics can be explained by the formation of a rapidly reducible Fe(III) phases derived from the progressivedissolution and re-oxidation of native Fe(III) oxides accompanied by minor increases in Fe reducer populations. The shift inFe(III) reactivity is evident from Fe-reducibility assays using Shewanella sp., however was undetectable by chemical extractions,Mossbauer or X-ray Absorption spectroscopies. More broadly, our findings suggest Fe reduction rates are strongly coupled toredox dynamics of the recent past, and that frequent shifts in redox conditions can prime a soil for rapid Fe-reduction.

■ INTRODUCTION

Although iron comprises only 1−3% of total soil mass,1 it has aprofound influence on soil properties. The high reactivity andlarge surface area of Fe-(oxyhydr)oxide minerals influencemany biogeochemical processes through the sorption of keynutrients or contaminants.2−10 When exposed to low oxygenconditions, Fe minerals serve as important electron acceptorsfor microbial respiration,11 leading to the dissolution of solidphases and release of sorbed and incorporated constituents.12,13

This reductive dissolution of Fe has significant implications forthe global C cycle in soils where a substantial fraction of thetotal soil organic C oxidation is coupled to the microbialreduction of Fe-(oxyhydr)oxides (e.g., 44% in humid tropicalforest soils14), which can occur in upland soils with Fe(II)production rates ranging from 0.5 to 50 mmol kg−1 d−1.15,16

The availability of Fe minerals toward reduction is stronglycorrelated with their surface area and particle size.17,18 Thecrystallinity of Fe-(oxyhydr)oxides also strongly influences thekinetics of microbial Fe(III) reduction.19 Microbes can rapidlyreduce ferrihydrite and other short-range ordered (SRO) Feminerals (typically within hours), but reduce well-orderedhematite (α-Fe2O3), goethite (α-FeOOH), and lepidocrocite(γ-FeOOH) at slower rates (e.g., several months).18,20

In upland soils, Fe reduction takes place within micrositeswhere high-levels of organic matter combine with ephemeral

water saturation and low O2 diffusion rates to yield anoxicconditions.21,22 Since O2 often reinfuses such microsites as thesoils dry, Fe(II) produced during the anoxic period viamicrobial Fe reduction is often reoxidized to Fe(III) solidphases. This Fe redox cycling can influence the rapid cycling ofother elements (e.g., C, N, and metals23). Redox fluctuationscan drive repeated dissolution and precipitation of Fe(oxyhydr)oxide minerals with potential influence on thecrystallinity of those phases.21,24−27 During anoxic periods,the sorption of the Fe2+(aq) ions to Fe mineral surfaces canaccelerate the process of Ostwald ripening from SRO tosecondary crystalline mineral phases (e.g., goethite, magnetite)that are less chemically reactive and bioavailable thanferrihydrite.26,28,29 However, there are also studies illustratingSRO phases form or persist in fluctuating redox environ-ments,30−34 indicating that the dynamics of the fluctuations arelikely important in determining the trajectory of mineraltransformations.Changes in the crystallinity of Fe-(oxyhydr)oxides may also

impact the fate of phosphorus (P) in natural systems because

Received: November 11, 2016Revised: February 9, 2017Accepted: February 13, 2017Published: February 13, 2017

Article

pubs.acs.org/est

© 2017 American Chemical Society 3250 DOI: 10.1021/acs.est.6b05709Environ. Sci. Technol. 2017, 51, 3250−3259

Fe-(oxyhydr)oxide surfaces have a great capacity for adsorbingP.4,5 Ferrihydrite has a significantly higher sorption capacitythan the crystalline (oxyhydr)oxides because of its large surfacearea, but can form weaker surface complexes with aqueousspecies than the more crystalline iron phases.35−37 As aconsequence, microbial Fe(III) reduction may simultaneouslyreduce the overall capacity of soils to adsorb metal/metalloidswhile strengthening the remaining sorbate−sorbent interac-tions.Given the potential importance of Fe redox cycling for soil

systems, the influence of redox oscillations on Fe reductionrates needs to be quantified. A redox fluctuation can becharacterized by (a) the frequency of the redox oscillation; (b)the oscillation amplitude; and (c) the time spent under oxic(τoxic) or anoxic (τanoxic) conditions. Here we hypothesize thatshorter frequency oscillations will lead to faster Fe reductionrates with associated changes in Fe-(oxyhydr)oxide mineralcomposition and the distribution of phosphate. We test this byexposing a tropical soil to different redox oscillation frequencieswhile holding the oscillation amplitude and ratio of τoxic: τanoxicconstant. In order to minimize spatial heterogeneity, soils werecontinuously mixed in slurriesseparating the effects of redoxfluctuations from transport to the extent possible. We thenmeasure changes in HCl-extractable Fe(II), Fe mineralcomposition and P distribution, and simulate the experimentaldata in a numerical model to calculate Fe reduction rates.

■ MATERIALS AND METHODSField Site and Soil Preparation. Intact soil blocks were

collected from the upper 10 cm of soil at a site similar to thatdescribed by Peretyazhko and Sposito38 in the Bisley watershedof the Luquillo Experimental Forest (LEF) Puerto Rico(approximate GPS location 18.31553N,-65.74574W), which ispart of the Luquillo Long Term Ecological Research site andthe Critical Zone Observatory. Samples were placed in 4-milpolyethylene ziplock bags and transported without coolinginside an insulated container to the University of Georgia.Within 24 h of sampling, all moist soils were processed by air-drying at 30°C for 24 h followed by 2 mm sieving/homogenization (to remove large roots and gravel) underfield soil moisture conditions within a 95%:5% N2:H2 glovebox(Coy). All soils were stored at 22°C under oxic conditions inthe dark. Ginn et al.39 showed these soils maintained thehighest levels of microbial Fe-reduction when air-dried prior tolong-term storage. Soils in the LEF are classified predominatelyas Ultisols formed from volcanic parent material with quartzdiorite intrusions.40,41 The major primary minerals consist ofquartz and plagioclase feldspars with lesser amounts of biotite,hornblende, K-feldspar, and accessory magnetite, sphene,apatite, and zircon.42 Prior work by Silver et al.21 has identifiedperiodic fluctuations in soil O2 concentrations sufficient togenerate anoxic conditions within the soil. Further informationon this soil collection and characterization are describedelsewhere.39,43

Redox Oscillation Experiments. Three oscillation treat-ments were performed simultaneously. In each experiment, 4.5g (dry-weight) of live soil (not sterilized) was suspended in abuffer solution containing 2 mM KCl and 10 mM solution of 2-(N-morpholino)ethanesulfonic acid (MES) buffer with a 45 gfinal suspension mass. The suspension was supplemented with200 mg kg−1 of P (i.e., 6.46 mmol kg−1 soil and 0.58 mmol L−1

suspension). The slurries were incubated for 56 days in aglovebox (95% N2, 5% H2) on an end-overend shaker at 250

rpm. Although these experiments likely had sufficient labileorganic carbon to support Fe reduction rates (see below), thepresence of H2 in the glovebox may have served as anadditional electron donor for Fe reduction.44 In each of theexperiments, the soil slurries were exposed to alternating oxic-anoxic conditions at frequencies of 3.5, 7, and 14 days with theratio of time under oxic/anoxic conditions (τoxic:τanoxic)maintained at 1:6 for all treatments (i.e., the 3.5 day oscillationhad 3 days of anoxic time and 0.5 days of oxic time; the 7 dayperiod had 6 days anoxic and 1 day of oxic; and the 14 daysperiod had 12 days anoxic and 2 days oxic time). Samples weretransferred between shaker tables inside and outside of theglovebox to create anoxic and oxic conditions, respectively.Control soil slurries were maintained at constant anoxic or oxicconditions for 56 days.Samples were extracted at the beginning and end of each

oscillation cycle using wide-orifice pipette tips that allowedcomplete collection of soil particles in the slurry. AqueousFe(II) was extracted from the soil slurries by centrifuging thesamples at 11 000 rcf for 30 min, which will remove particles>40 nm based on Stokes’ law with an assumed mean particledensity of 2.72 g cm−3 for basalt-influenced soils.45 Acid-extractable Fe(II) was solubilized by suspending the remainingpellet in 1.0 mL of 0.5 M HCl and shaking it for 2 h on ahorizontal shaker at 150 rpm. The extracts were thencentrifuged at 11 000 rcf for 30 min and the supernatantsanalyzed for Fe(II) using a modified ferrozine protocol24 on aShimadzu-1700 UV−vis spectrophotometer at λ = 562 nm. Thedry weight of each pellet from the centrifugation step wasmeasured to correct for minor differences in suspension densitybetween treatments and allow accurate dry-weight normal-ization. All values are reported as averages of triplicatemeasurements with error bars indicating standard deviation.

FeIII-Reduction Rate Experiment. To quantify Fe(III)reduction rates midway through the experiment, additionalsamples were oscillated at the three frequencies given above for28 days and then exposed to 7 days of continuously anoxicconditions during which they were sampled for aqueous andacid-extractable Fe(II) every 0.5 days as described above. Thisdata captured the iron reduction dynamics during a reducingperiod and was used to parametrize the reduction kinetics inthe numerical model described below.

Citrate-Ascorbate Extractable Fe. To quantify the short-range-ordered solid Fe phases in the samples, we conductedcitrate-ascorbate extractions46 on freeze-dried soils from theunreacted samples or at the end of the oscillation experiment(freeze-dried immediately upon completion of the experiment).Briefly, the samples were suspended in a solution containing 0.2M Na-citrate and 0.05 M ascorbic acid (pH 6) at a 60:1reagent:sample weight ratio. The suspensions were shaken for16 h on a horizontal shaker and centrifuged at 11 000 rcf for 30min. The supernatants were analyzed for total Fe on an atomicabsorption (AA) spectrometer (PerkinElmer, AAnalyst 200).

Mossbauer and X-Ray Absorption Spectroscopies. Fespeciation of the soil samples was characterized via 57FeMossbauer spectroscopy and X-ray absorption spectroscopy(XAS). Details on data collection, analysis, and interpretationare given in the Supporting Information (SI) Sections 2 and 3.

Fe Reduction Assay: Shewanella Incubations. Wequantified the availability of Fe(III) for microbial reductionby incubating the soil with a culture of Shewanella oneidensisMR-1. The purpose of this inoculation experiment was todetermine whether redox fluctuations altered the iron oxide

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b05709Environ. Sci. Technol. 2017, 51, 3250−3259

3251

availability for microbial Fe-reduction. S. oneidensis MR-1 wasused because it is a good model Fe-reducer, capable of fast Fe-reduction and can live in both oxic and anoxic conditions. Thisexperiment is performed solely to assess Fe availability acrosstreatments, and is not designed to quantify the in situ ironreduction rate. The assay was performed on each replicate ofsoils freeze-dried from the end of the three different oscillationfrequencies, the anaerobic and aerobic controls, and unreactedsoil. Postoxidation, freeze-dried soil was used to maintainconsistency across treatments. Shewanella oneidensis MR-1 wasgrown to stationary phase in media containing (per liter): 0.5 gKH2PO4, 1.0 g Na2SO4, 2.0 g NH4Cl, 1.0 g yeast extract, 0.5mM CaCl2, 0.1 mM MgSO4, 10 mM Na-lactate, and 50 mMFe-citrate. The cells were washed by centrifugation at 3000 rcffor 30 min and resuspended in fresh media without yeastextract. The washing step was repeated twice to thoroughlyremove spent media, after which 0.9 g of washed cell culturewas added to 2 mL centrifuge tubes containing 0.1 g of freeze-dried soil so each sample contained the same ratio of bacterialculture to soil mass. The cultures were incubated for a weekunder anoxic conditions. Freeze-dried soils with no S. oneidensisinoculation were used as controls. The cultures were sampledfor acid-extractable Fe(II) by pipetting 0.1 mL of sample fromeach culture and centrifuging at 11 000 rcf for 30 min. Thesupernatants were poured off and the pellets were resuspended

in 0.1 mL of 0.5 M HCl. The suspensions were shaken for 2 hon a horizontal shaker. The extracts were then centrifuged at11 000 rcf for 30 min. Samples were analyzed for acid-extractible Fe(II) using the ferrozine protocol described above.

Phosphate Extractions. Sequential extractions wereperformed according to Hedley et al.47 A soil mass of 0.01 gwas sequentially extracted for 16 h (each) on a horizontalshaker with 0.6 mL of ultrapure water, 0.5 M NaHCO3, 0.1 MNaOH, and 1 M HCl. After each extraction, the tubes werecentrifuged at 11 000 rcf for 30 min. In addition, we quantifiedmicrobial biomass P following Hedley and Stewert.48 Briefly, 1mL of CHCl3 was added to 0.01 g of soil in a centrifuge tubeand capped. The solution was shaken periodically over 30 minand then uncapped so that the CHCl3 could evaporateovernight. In all cases, supernatants were analyzed for P usinga molybdate blue protocol49 on a Shimadzu-1700 spectropho-tometer at λ = 880 nm.

Phosphate Adsorption Isotherms. To assess theadsorption capacity of the soil, we constructed phosphateadsorption isotherms on unamended soils. Briefly, 1 mL of 0,80, 160, 320, and 640 ppm phosphate was added to 0.1 g ofeach freeze-dried soil in a centrifuge tube. The suspensionswere shaken on a horizontal shaker for 2 h and then centrifugedfor 30 min at 11 000 rcf. The supernatants were collected andmeasured as described above.

Figure 1. Fluctuations in Fe(II) concentrations due to redox oscillations with periods of 3.5 days (bottom), 7 days (middle) or 14 days (top). Alltreatments have the same oxic:anoxic exposure time ratio of 6:1 and the oxic cycles involve exposure to 21% O2. Graphs on the left side are forunamended slurries and those on the right are for treatments with added phosphorus. Symbols indicate measured values with 1 SD error bars basedon three replicates. The solid lines represent the numerical model.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b05709Environ. Sci. Technol. 2017, 51, 3250−3259

3252

Numerical Model. Experiments were simulated withMATLAB (version R2015b) using a simple, parsimonious,process-based kinetic model to describe and test mechanisms ofmicrobial Fe(III) reduction, growth of the Fe-reducingmicrobial community, abiotic Fe(II) oxidation by O2, andsolid phase partitioning of Fe minerals into distinct pools.Redox oscillations are imposed by setting O2 concentrations to200 μM during oxic periods (slightly below saturation) and 0during anoxic periods, with Fe(II), Fe-reducing microbialcommunity, and a fast and a slowly reacting Fe(III) solid phase(FeOxfast and FeOxslow, respectively) as state variables. Duringanoxic periods, the rate of microbial Fe(III) reduction is givenas

∑=+

R kK

[BM][FeOx ]

[FeOx ]ii

i

i ired

m, (1)

where ki is a cell specific rate constant, [BM] is theconcentration of Fe reducers, Km,i is a half saturation constantand i represents the fast and slowly reacting Fe(III) solidphases, respectively. The net growth rate of Fe-reducingmicrobes (RBM) is set proportional to the Fe reduction rate,RBM = γRred, with a growth efficiency γ set to 0.06 (based ongrowth yields from Kostka et al.50). As the oxic periods arerelatively short in our experiments, no net growth or death ofthe Fe reducers is considered during oxic periods. Under oxicconditions, Fe reduction is set to 0, and the oxidation of Fe(II)is implemented following Stumm and Morgan:51

=R k [Fe ][O ]ox oxII

2 (2)

where the Fe(II) oxidation rate constant kox is set to 3 × 102

M−1 min−1, 1000 fold higher than the value reported forhomogeneous oxidation kinetics at the pH of 5.5 in our theexperiments to accommodate heterogeneous reaction rates.52

Fast oxidation can promote the production of less ordered andmore reactive Fe oxides,53,54 thus we partition all the oxidizedFe(II) to the fast pool given the rapid oxidation rate observedin our experiment. During our experiments, we routinelyobserved that a portion of the HCl-extractable Fe(II) formedduring the anoxic half-cycle is resistant to oxidation, resulting inan apparent accumulation of Fe(II) over the course of theexperiment. We represented this in the numerical model byremoving a small portion of the Fe(II) from the reactive pool,which we hypothesize results from physical protection(encasement) against oxidation. The rate of Fe(II) encasedwas set to 2% (q) of the rate of iron oxidation to fit theexperimental data. Thus, the dynamics of fast and slowlyreacting Fe(III) solid phases are given by

= − −t

q R RdFeOx

d(1 )fast

ox redfast

(3)

= −t

RdFeOx

dslow

redslow

(4)

Initial reducible Fe(III) solid concentrations were based on theobserved values of 10 and 140 mmol L−1 for fast and slowlyreacting Fe oxides, respectively, with zero initial Fe(II). The fastpool represents Fe that can be reduced in less than 3 days;39

while the slow pool Fe is set to reproduce the Fe(II) plateaulevels observed in the oscillation experiments. To match the Fereduction rates across all treatments, the two model parametersrepresenting kfast *[BM]t=0 and kslow*[BM]t=0 were set to 6 and1.5 mmol kg−1 d−1 in experiments without P additions, and 10

and 2.3 mmol kg−1 d−1 in experiments with P addition,reflecting alleviation of P limitation. The fast and slow rateparameters were determined by fitting the data from FeIII-reduction rate experiment conducted after 28 days ofincubation (see above). All other model parameters wereidentical across all simulations.

■ RESULTS AND DISCUSSIONFe Reduction Dynamics. Fe(II) concentrations responded

to changing redox conditions (Figure 1) with peak values at theend of the anoxic periods and the lowest values after an oxicperiod. Peak Fe(II) concentrations increased from 10 to 148−165 mmol kg−1 soil in the P-amended treatments as theexperiment progressed through successive redox cycles (Figure1). The peak Fe(II) concentrations plateaued near 150 mmolFeII kg−1 soil after about 1 month and increased only slowly forthe remainder of the anoxic half-cycles. The phosphate-amended treatments reached this ∼150 mmol Fe(II) kg−1

soil plateau at earlier oscillation cycles than the unamendedtreatments, but by the end of the experiment all treatmentsexhibited similar amplitudes of Fe(II) oscillation, except the 3.5days unamended control soil (Figure 1).The increase in peak Fe(II) concentrations was similar across

most oscillation frequency treatments, despite different anoxicincubation lengths (Supporting Information (SI) Figure SI-1).Thus, the time-averaged iron reduction rate is lower intreatments with shorter oscillation frequencies, but the similarpeak Fe(II) concentrations between treatments suggests totalFe reduction extent is independent of oscillation frequency. Toassess how the frequency of redox oscillations affected thereduction kinetics, we repeated our oscillation experiments (forthe phosphorus addition treatments only) initiating acontinuous 2-week anoxic incubation after one month ofredox oscillations (Figure 2). By tracking the accumulation of

Fe(II) over time (Figure 2), we surmise there are two phases inthe production of Fe(II) during anoxic incubations: (1) aninitial, rapid Fe(II) production interval followed by (2) aninterval with much slower Fe(II) production that brings theFe(II) concentrations into proximity of the peak Fe(II)concentrations at the end of each anoxic cycle (Figure 1).Irrespective of oscillation frequency, a slow Fe-reduction phase

Figure 2. Fe(II)-concentrations under anoxic conditions following 28days of redox oscillations with periods of 3.5, 7, or 14 days intreatments amended with phosphorus. Error bars are 1 SD This datawas used to fit the dynamics during the reduction period in thenumerical model across all treatments.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b05709Environ. Sci. Technol. 2017, 51, 3250−3259

3253

(or plateau) occurs after 4 days of anoxic incubation. Thisplateau could explain why all of the phosphate-amended soilslurries had similar peak Fe(II) concentrations (even the 3.5day treatment), as 90% of the plateau Fe(II) concentration isachieved within 3 days of incubation (Figure 1). Although thesecond experiment was only conducted with the P-amendedsoils, we suspect this plateau may not manifest as prominentlyin the 3.5 days unamended treatment because the overall slowerFe reduction rates in the unamended controls.Our experiments were conducted on rewetted soils that were

oxidized without subsequent drying during the redoxoscillations. This experimental design minimized heterogeneityduring the experiment and tightened reproducibility betweenreplicates, however it may have impacted the iron dynamics intwo important ways. First, the rewetting of soils is well-knownto produce a flush of organic carbon,55 and thus our treatmentslikely contain more soluble, labile organic carbon than if we hadinitiated the experiments from a field moist state. This likelyprovided an excess of electron donor in the form of solubleorganic carbon and it may have increased iron availabilitythrough complexation or by facilitating electron shuttlingrelative to the field moist state.56−60 Second, oxidizing thesamples under saturated conditions represents the commonsituation in moist soils or near-surface sediments61 in whichpulses of O2 accompany the infiltration of rainwater; however,our design does not represent as closely soil-drying events, inwhich oxidation fronts follow the major fluid flow paths andcreate a heterogeneous distribution of FeIII phases.62

Fe Mineral Composition. To quantify the changes in theavailability of Fe for microbial reduction,63 we inoculated thesame freeze-dried soils (before and after treatment) with S.oneidensis MR-1 in a defined media to quantify the rapidlyreducible Fe(III) content (Figure 3). The purpose of this

experiment was not to simulate Fe-reduction that occurs in thefield, but to determine the bioavailability of Fe after exposure tofluctuating redox conditions. The S. oneidensis culturesproduced FeII more rapidly following the 3.5 day, 7 day, 14day, and anoxic control slurry incubations than in the oxiccontrol and slurries with soils prior to the oscillationexperiments. We found Fe(III) reduction in the noninoculated

soils was negligible compared to that occurring in theinoculated soils. Thus, outside of a possible stimulatory effectof Shewanella on the growth of the native soil bacteria,32 weassume the Fe(III)-reduction assays were not biased bydifferences in soil microbial population or activity.The 3.5 day, 7 day, 14 day, and anoxic control soils exhibited

similarly higher rates of Fe(III) reduction in the S. oneidensisassays than the oxic and initial soils (Figure 3, SI section 4).Although each treatment had different redox oscillation periods,all are likely to have similar abundances of recently oxidizedFe(III), as each oscillation treatment reached similar Fe(II)concentration plateaus in their last anoxic cycle, whereas theoxic control and the starting soil had much lower HCl-extractable Fe(II) concentrations. Since we terminated eachtreatment following an oxidation half cycle (including theanoxic control), the rapid introduction of O2 oxidizes nearly allof the recently produced Fe2+(aq), likely forming a veryreactive, short-range-ordered (SRO) phase.54

However, we did not find significant changes in Fe mineralcomposition between treatments using citrate-ascorbate chem-ical extractions (SI Figure S2), 57Fe Mossbauer spectroscopy(MBS, SI section 2), or X-ray adsorption spectroscopy (XAS, SIsection 3). This suggests one or more of the following: (1) theproportional change in the SRO pool is too small relative to thetotal background soil Fe in the soil for these chemical andspectroscopic measures; (2) the reoxidized Fe is more proximalto the microbial iron reducersperhaps precipitating as thinsurface coatings on the original SRO-Fe phases, while retainingsimilar local structure or by forming complexes withextracellular polymeric substances on the bacterial surface;64−66

(3) the reoxidized Fe coprecipitated with organic electronshuttling agents that can increase the availability for microbialreduction without yielding detectable changes in mineralcomposition.Prior work from redox fluctuation experiments illustrates that

Fe mineral crystallinity can increase or decrease,24,32,33,67 orremain undetected,68 although the specific conditions drivingthe trajectory of mineral change are unclear. Documentedincreases in Fe mineral crystallinity might be associated withslow or incomplete Fe oxidation. Liptzin and Silver67 foundincreases in Fe crystallinity when exposing intact, unhomogen-ized soil samples to redox fluctuations and Thompson et al.24

similarly found Fe crystallinity increases when redox fluctuatingsoil slurries were exposed to progressive oxidation intended toapproximate conditions within an aggregate. Likewise, if theinitial phase has very low crystallinity, then exposure toFe2+(aq) during the anoxic periods may promote increases incrystallinity as Tomaszewski et al.26 showed when exposingferrihydrite to a redox fluctuating environment. However,decreases in Fe crystallinity have also observed during redoxfluctuations, especially when accompanied by soil drying,69,70 orwhen the starting material is of higher crystallinity.32

Furthermore, there are other reports, akin to our study,where changes in Fe crystallinity cannot be detected followingredox fluctuations.68

Process Identification. We identified process dynamicsconsistent with the observational data by implementingdifferent processes in a numerical model. Our implementationis able to reproduce the observed increase in Fe(II)concentrations (aqueous plus 0.5 M HCl extractable) duringthe anoxic cycles, as well as the plateau phase revealed in oursecond experiment (Figure 1). The basic structure of the modelincludes two pools of iron with different cell-specific Fe

Figure 3. Fe(II) concentrations in a liquid culture of S. oneidensis MR-1 incubated with the sole source of ferric iron deriving from freeze-dried soil samples from the indicated redox oscillation treatments andcontrols (all of which ended with oxic exposure, including the anoxiccontrol). Each data point is the difference in acid-extractable FeII

concentrations between inoculated soils and the noninoculatedcontrols. Error bars are 1 SD. See SI Section 4 for more details.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b05709Environ. Sci. Technol. 2017, 51, 3250−3259

3254

reduction rates (kfast and kslow). This was critical for reproducingthe high Fe(II) production rates at the beginning of each anoxicperiod, followed by slower Fe(II) production rates associatedwith reduction of the less reactive iron pool.A key feature of the whole data set (P-amended and

nonamended treatments) is that net Fe(II) production ratesincrease over the course of the experiment (Figure 1). Theinitial rates of Fe reduction were ∼3 mmol Fe(II) kg−1 soil d−1,which are similar to rates measured previously on this soil.39

However, during the later half of the experiment, the Fereduction rates increased to 17−66 mmol Fe(II) kg−1 soil d−1

across all treatments (Figure 4). This increase in peak Fe(II)concentrations over the course of the experiment can beexplained by the combined effect of an increased pool ofreactive iron oxideformed during rapid oxidation of Fe(II)accumulated during the anoxic periodsand a minor increasein the activity of the iron reducing community. We measuredFe reduction population densities at the beginning and end ofthe experiment (SI Table S1) and found no difference withinthe log unit precision of our methods, whichconsistent withthe model simulationsimplies the Fe reducer populationincreased less than 10-fold over the course of the experiment.Fe(II) concentrations also increase slightly during the oxic half-cycles over the course of the experiment, amounting to whatappears as a upward drifting Fe(II) baseline (Figure 1). Thismay reflect the formation of a mixed-valence iron mineral, astable Fe(II)-organic matter complex or an Fe sulfide phase,represented in the model by the partitioning a portion of Fe(II)into a nonoxidizable form that is still extractable via 0.5 M HCl

(see Materials and Methods). Thus, in our final model, weassumed that the “active Fe”defined as Fe that could bemobilized via Fe reduction or other processes within the timescale of our experimentmoved between four reservoirs: (1) arapidly reducible Fe(III) solid; (2) a slowly reducible Fe(III)solid; (3) dissolved or adsorbed Fe(II); and (4) a nonoxidizableFe(II) phase. Both Fe(III) solid phases contributed to the trendof increasing Fe(II) peaks over the course of the oscillationexperiment. During each oxidation cycle, we partitioned all ofthe Fe(II) oxidized into the rapidly reducing Fe(III) solidphase, hence providing each subsequent simulated reductioncycle with a larger pool of rapidly reducible Fe(III).An alternative explanation for the plateau in Fe(II)

concentrations is that sorption of Fe(II) to Fe mineral surfacesmay pacify the Fe(III) toward further reductive dissolution.Roden and Zachara2 attributed a microbial Fe(III)-reductionplateau they observed in pure culture experiments to this effect.However, in our experiment, peak Fe(II) concentrations clearlyincrease with subsequent oscillation cycles, suggesting Fereduction rates across all treatments increased as the experi-ment progressed (Figure 4).Imposing the above constraints and processes does not yield

a model fit to all frequency treatments unless the modelparameters are tuned to each treatment. However, by imposinga short suppression of simulated microbial growth at thebeginning of each anoxic cyclerepresenting a predominanceof intracellular metabolic adjustments over the formation ofnew biomasswe can achieve a consistent match with theexperimental Fe(II) data. A single global parameter set matches

Figure 4. Net Fe(II) production rates calculated at each time point from the numerical model for the unamended (left graphs) and P-amended(right graphs) treatments. Values shown only for anoxic periods.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b05709Environ. Sci. Technol. 2017, 51, 3250−3259

3255

the observations across all oscillation frequencies (Figure 1),except that the P-amended treatments required a higher cell-specific Fe reduction rate than the unamended treatments.Considering first the nonamended treatments, the simulated

peak rates of Fe reduction are similar for the 3.5 days and 7days treatments (peak rates of Fe(II) production ∼30 mmolkg−1 soil d−1), but substantially slower in the 14 days treatment(peak rates of Fe(II) production ∼17 mmol kg−1 soil d−1).Because the iron reduction rates are only high during the first∼4 days of each anoxic period (Figure 4), the 14 daystreatment has longer period in which the microbes are largelyinactive. With microbial growth linked to metabolic activity,this leads to less growth. Over the course of the experiment, thesimulated biomass increased 4.3× in both the 3.5 and 7 daystreatments, compared with an increase of only 2.7× in the 14day treatment (SI Figure S3), which in turn leads to lowerFe(II) peak concentrations (Figure 1).Phosphorus Considerations. Augmenting the soil slurries

with phosphate enhanced the production of Fe(II) during theanoxic cycles (Figure 1), suggesting that microbial activity wasstimulated. Consistent with this hypothesis, the amount ofphosphorus released following exposure to CHCl3 increasedover the course of the experiment (SI Figure S5), suggesting Paccumulated in the microbial biomass. No other changes in thephosphorus distributions (based on the Hedley protocol47)were observed in the treatments, with nearly all of the Presiding in the NaOH extractable pool (negligible P was foundin the aqueous, NaHCO3, and HCl extracts and thus were notreported). NaOH extractable P is considered to be associatedwith Fe and Al oxide surfaces in the standard interpretation ofthe Hedley protocol.47 Previous measurements of Luquillo soilsalso found P to be predominately associated with mineralsurfaces.38,67 We found there was substantial potential for Puptake in these soils, but the amount and shape of thephosphate adsorption isotherms (which followed typicalLangmuir curves) did not differ between treatments andcontrols (SI Figure S4, Table S2).The phosphorus addition treatments had higher simulated

rates of Fe reduction than the nonamended controls, with peakrates of Fe(II) production ∼66, ∼52, and ∼31 mmol kg−1 soild−1 for the 3.5 day, 7 day, and 14 day treatments, respectively.Higher cell-specific Fe reduction rate constants are required tomatch the temporal evolution of Fe(II) oscillations in the P-amended treatments than are required in the nonamendedcontrols (an increase of 66.7% and 50.3% for fast and slowlyreacting iron oxides, respectively). Similar to the nonamendedtreatments, our simulations increase microbial biomass more inthe shorter frequency treatments as a direct consequence oftheir proportionally longer time with access to rapidly reducibleiron [Fe(III)RR] than the longer frequency treatments. Over thecourse of the experiment, the simulated biomass in the P-amended treatments increased 7.0× , 5.1×, and 3.2× in the 3.5day, 7 day, and 14 day treatments, respectively.This may be the first study illustrating that additions of P

promote an increase in Fe(II) production rates in native soils orsediments, however, there is a large body of work based onsynthetic minerals. In these pure-system studies, addition ofphosphate often alters the rate of Fe(II) production, althoughthe direction and magnitude of change is inconsistent,depending on P concentrations, Fe mineral composition andother factors.71−78 Our P-amended soils had 6.46 mmol kg−1

soil (or 0.58 mmol L−1 suspension) of P added, which is lowerthan most of the P amendments used in prior synthetic iron

mineral studiestypical P amendments are 2−20 mmol L−1 inthe suspension,74,75 with 0.4 mmol L−1 often considered a lowP treatment.73 Thus, given the low amount of P added in ourtreatments, we suspect the increase in Fe(II) production ratesresults from an alleviation a P limitation on microbial growth,which is consistent with P uptake to the biomass (SI FigureS5).

■ ECOSYSTEM IMPLICATIONSUpland ecosystems are subjected to short periods of anoxia thatoften lead to some ephemeral iron reduction, which caninfluence C cycling and the fate of many nutrients andchemicals. We have shown that successive periods of redoxfluctuations over a month time scale with ample organic C canincrease soil Fe reduction rates, regardless of fluctuationfrequency. Thus, Fe reduction rates during the onset of anoxiaare strongly coupled to physiochemical dynamics of the recentpast. Terrestrial systems exposed to rain events that yield ironreducing conditions and a buildup of Fe(II) thus might beprimed for more rapid iron reduction rates than systemsemerging from a long drought or with continuously wet soils.The amount of Fe(II) produced following a prolonged

anoxic incubation is often predicted through selectiveextractions (oxalate, citrate-ascorbate, hydroxylamine, etc.)targeting the short-range-ordered (SRO) phases. Our work isconsistent with this in that the total amount of Fe(II) producedduring an anoxic cycle later in the experiment (or in the anoxiccontrol) was similar to the citrate-ascorbate extractable iron(∼30% of total Fe). However, SRO abundance is a poorpredictor of iron reduction rate, because the SRO abundancedoes not change throughout the experiment, while ironreduction rate changes dramatically. Instead, we propose theavailability of SRO phases for Fe reduction is enhanced throughrapid oxidation events by atmospheric air (∼21% O2), creatinga subset of rapidly reducible SRO-Fe [Fe(III)RR]. We foundShewanella cultures could reduce Fe(III) much faster from soilsexposed to redox fluctuations than from the oxic controls,however we could not detect any corresponding shifts inmineral composition via selective extractions, Mossbauerspectroscopy, or X-ray adsorption spectroscopy. Regardless,by the later half of the experiment nearly the entire SRO Fepool undergoes reduction within 4 days.Soils that are frequently exposed to redox fluctuations have a

large pool of rapidly reducible iron to transfer electrons toduring periods of oxygen depletiona biogeochemicalbattery79 that is recharged by rapid oxidation of previouslyanoxic pore waters. Our work highlights the importance ofidentifying the reaction time scales and limiting reactants/catalysts (microbes) in predicting iron reduction rates in soils.Prior work suggests this soil does not have significant C-limitations,39 but we did find the soil responded to P-amendments with more rapid increases in iron reduction ratethan unamended soils. The P-amended treatments required afaster cell-specific Fe reduction rate when P was added andsuggests iron reducer activity (population growth, CO2production, etc.) is higher when redox fluctuations occur ontime scales similar to the process time scales (length of time tooxidize most of the Fe(II) or the length of time to a Fe(II)concentration pseudoplateau). Soils that are exposed to near-weekly redox pulsing (due to variable rain events or water tableshifts) may therefore be more dominated by iron biogeochem-ical dynamics than soils that experience redox shifts onlyseasonally.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b05709Environ. Sci. Technol. 2017, 51, 3250−3259

3256

■ ASSOCIATED CONTENT

*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.est.6b05709.

Section 1 contains two additional tables and fiveadditional figures cited in the main text; Section 2contains a detailed description of the Mossbauer (MBS)characterization of soil Fe; Section 3 contains a detaileddescription of the XAS characterization of soil Fe and acomparison of the XAS and MBS findings; Section 4contains a description of the statistical test of Shewanellasp. assay (PDF)

■ AUTHOR INFORMATION

Corresponding Author*Phone: (01) 706-410-1293; e-mail: AaronT@uga.edu.

ORCIDYuanzhi Tang: 0000-0002-7741-8646Aaron Thompson: 0000-0001-6301-7377NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

We thank Whendee Silver for providing access to the Bisleysite; Michelle Scherer, Drew Latta, and Tim Pasakarnis forMossbauer measurements; Chris Gorski and Prachi Joshi forpreliminary electrochemical analyses of these soils; Qing Ma(APS 5-BM-D) and Ryan Davis (SSRL 4-1) for assistance withXAS analysis; and Nehru Mantripragada for laboratoryassistance. This work was funded by USDA-NIFA SoilProcesses Program AFRI- NIFA Grant no. 2009-65107-05830to AT and CM; NSF grants EAR-1331841, EAR-1053470,EAR- 1451508, and DEB-1457761 to AT; and NSF OCE-1559087 to YT. Portions of this research were conducted at theStanford Synchrotron Radiation Lightsource (SSRL) and theAdvanced Photon Source (APS).

■ REFERENCES(1) Loeppert, R. H.; Inskeep, W. P., Iron. In Methods of Soil Analysis:Part 3-Chemical Methods, Sparks, D. L., Ed. Soil Science Society ofAmerica: 1996; pp 639−664.(2) Roden, E. E.; Zachara, J. M. Microbial reduction of crystallineiron(III) oxides: Influenceof oxide surface area and potential for cellgrowth. Environ. Sci. Technol. 1996, 30, 1618−1628.(3) Frenzel, P.; Bosse, U.; Janssen, P. H. Rice roots andmethanogenesis in a paddy soil: Ferric iron as an alternative electronacceptor in the rooted soil. Soil Biol. Biochem. 1999, 31, 421−430.(4) Ann, Y.; Reddy, K. R.; Delfino, J. J. Influence of redox potentialon phosphorus solubility in chemically amended wetland organic soils.Ecol. Eng. 2000, 14, 169−180.(5) Miller, A. J.; Schuur, E. A. G.; Chadwick, O. A. Redox control ofphosphorus pools in hawaiian montane forest soils. Geoderma 2001,102, 219−237.(6) Fimmen, R.; Richter, D.; Vasudevan, D.; Williams, M.; West, L.;Rhizogenic, Fe−C redox cycling: A hypothetical biogeochemicalmechanism that drives crustal weathering in upland soils. Biogeochem-istry 2008, 87, 127.(7) Sulzberger, B.; Suter, D.; Siffert, C.; Banwart, S.; Stumm, W.Dissolution of Fe(III) (hydr)oxides in natural waters; laboratoryassessment on the kinetics controlled by surface coordination. Mar.Chem. 1989, 28, 127−144.

(8) Bruemmer, G. W.; Gerth, J.; Tiller, K. G. Reaction kinetics of theadsorption and desorption of nickel, zinc and cadmium by goethite. I.Adsorption and diffusion of metals. J. Soil Sci. 1988, 39, 37−52.(9) Sparks, D. L. Kinetics of Soil Chemical Processes; Academic Press,2013.(10) Spadini, L.; Manceau, A.; Schindler, P. W.; Charlet, L. Structureand stability of Cd2+ surface complexes on ferric oxides: 1. Resultsfrom EXAFS spectroscopy. J. Colloid Interface Sci. 1994, 168, 73−86.(11) Canfield, D. E.; Thamdrup, B.; Hansen, J. W. The anaerobicdegradation of organic-matter in danish coastal sediments: Ironreduction, manganese reduction, and sulfate reduction. Geochim.Cosmochim. Acta 1993, 57, 3867−3883.(12) Lovley, D. R. Organic-matter mineralization with the reductionof ferric iron - a review. Geomicrobiol. J. 1987, 5, 375−399.(13) McBride, M. B. Environmental Chemistry of soils; OxfordUniversity Press, NY, 1994.(14) Dubinsky, E. A.; Silver, W. L.; Firestone, M. K. Tropical forestsoil microbial communities couple iron and carbon biogeochemistry.Ecology 2010, 91, 2604−2612.(15) Yang, W. H.; Liptzin, D. High potential for iron reduction inupland soils. Ecology 2015, 96, 2015−2020.(16) Hall, S. J.; Liptzin, D.; Buss, H. L.; DeAngelis, K.; Silver, W. L.Drivers and patterns of iron redox cycling from surface to bedrock in adeep tropical forest soil: A new conceptual model. Biogeochemistry2016, 130, 177−190.(17) Bonneville, S.; Behrends, T.; Van Cappellen, P. Solubility anddissimilatory reduction kinetics of iron(III) oxyhydroxides: A linearfree energy relationship. Geochim. Cosmochim. Acta 2009, 73, 5273−5282.(18) Bonneville, S.; Van Cappellen, P.; Behrends, T. Microbialreduction of iron(III) oxyhydroxides: Effects of mineral solubility andavailability. Chem. Geol. 2004, 212, 255−268.(19) Roden, E. E.; Wetzel, R. G. Organic carbon oxidation andsuppression of methane production by microbial Fe(III) oxidereduction in vegetated and unvegetated freshwater wetland sediments.Limnol. Oceanogr. 1996, 41, 1733−1748.(20) Roden, E. E. Geochemical and microbiological controls ondissimilatory iron reduction. C. R. Geosci. 2006, 338, 456−467.(21) Silver, W. L.; Lugo, A. E.; Keller, M. Soil oxygen availability andbiogeochemistry along rainfall and topographic gradients in uplandwet tropical forest soils. Biogeochemistry 1999, 44, 301−328.(22) Sexstone, A. J.; Revsbech, N. P.; Parkin, T. B.; Tiedje, J. M.Direct measurement of oxygen profiles and denitrification rates in soilaggregates. Soil Sci. Soc. Am. J. 1985, 49, 645−651.(23) Roden, E. E. Analysis of long-term bacterial vs. Chemical Fe(III)oxide reduction kinetics. Geochim. Cosmochim. Acta 2004, 68, 3205−3216.(24) Thompson, A.; Chadwick, O. A.; Rancourt, D. G.; Chorover, J.Iron-oxide crystallinity increases during soil redox oscillations.Geochim. Cosmochim. Acta 2006, 70, 1710−1727.(25) Mejia, J.; Roden, E. E.; Ginder-Vogel, M. Influence of oxygenand nitrate on Fe (hydr) oxide mineral transformation and soilmicrobial communities during redox cycling. Environ. Sci. Technol.2016, 50, 3580−3588.(26) Tomaszewski, E. J.; Cronk, S. S.; Gorski, C. A.; Ginder-Vogel,M. The role of dissolved Fe (II) concentration in the mineralogicalevolution of Fe (hydr) oxides during redox cycling. Chem. Geol. 2016,438, 163−170.(27) Vogelsang, V.; Kaiser, K.; Wagner, F. E.; Jahn, R.; Fiedler, S.Transformation of clay-sized minerals in soils exposed to prolongedregular alternation of redox conditions. Geoderma 2016, 278, 40−48.(28) Zachara, J. M.; Kukkadapu, R. K.; Fredrickson, J. K.; Gorby, Y.A.; Smith, S. C. Biomineralization of poorly crystalline Fe(III) oxidesby dissimilatory metal reducing bacteria (DMRB). Geomicrobiol. J.2002, 19, 179−207.(29) Hansel, C. M.; Benner, S. G.; Neiss, J.; Dohnalkova, A.;Kukkadapu, R. K.; Fendorf, S. Secondary mineralization pathwaysinduced by dissimilatory iron reduction of ferrihydrite under advectiveflow. Geochim. Cosmochim. Acta 2003, 67, 2977−2992.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b05709Environ. Sci. Technol. 2017, 51, 3250−3259

3257

(30) Chadwick, O. A.; Gavenda, R. T.; Kelly, E. F.; Ziegler, K.; Olson,C. G.; Elliott, W. C.; Hendricks, D. M. The impact of climate on thebiogeochemical functioning of volcanic soils. Chem. Geol. 2003, 202,195−223.(31) Chorover, J.; Amistadi, M. K.; Chadwick, O. A. Surface chargeevolution of mineral-organic complexes during pedogenesis inhawaiian basalt. Geochim. Cosmochim. Acta 2004, 68, 4859−4876.(32) Coby, A. J.; Picardal, F.; Shelobolina, E.; Xu, H.; Roden, E. E.Repeated anaerobic microbial redox cycling of iron. Appl. Environ.Microbiol. 2011, 77, 6036−6042.(33) Li, Y.; Yu, S.; Strong, J.; Wang, H. Are the biogeochemical cyclesof carbon, nitrogen, sulfur, and phosphorus driven by the “FeIII−FeIIredox wheel” in dynamic redox environments? J. Soils Sediments 2012,12, 683−693.(34) Thompson, A.; Rancourt, D. G.; Chadwick, O. A.; Chorover, J.Iron solid-phase differentiation along a redox gradient in basaltic soils.Geochim. Cosmochim. Acta 2011, 75, 119−133.(35) Dzombak, D. A.; Morel, F. M. M. Surface ComplexationModeling: Hydrous Ferric Oxide; John Wiley & Sons, 1990.(36) Axe, L.; Anderson, P. R. Experimental and theoreticaldiffusivities of Cd and sr in hydrous ferric oxide. J. Colloid InterfaceSci. 1997, 185, 436−448.(37) Trivedi, P.; Axe, L. Modeling Cd and Zn sorption to hydrousmetal oxides. Environ. Sci. Technol. 2000, 34, 2215−2223.(38) Peretyazhko, T.; Sposito, G. Iron(III) reduction andphosphorous solubilization in humid tropical forest soils. Geochim.Cosmochim. Acta 2005, 69, 3643.(39) Ginn, B. R.; Habteselassie, M. Y.; Meile, C.; Thompson, A.Effects of sample storage on microbial Fe-reduction in tropicalrainforest soils. Soil Biol. Biochem. 2014, 68, 44−51.(40) Beinroth, F. H. Some highly weathered soils of Puerto Rico, 1.Morphology, formation and classification. Geoderma 1982, 27, 1−73.(41) Frizano, J.; Johnson, A. H.; Vann, D. R.; Scatena, F. N. Soilphosphorus fractionation during forest development on landslide scarsin the luquillo mountains, Puerto Rico. Biotropica 2002, 34, 17−26.(42) White, A. F.; Blum, A. E.; Schulz, M. S.; Vivit, D. V.;Stonestrom, D. A.; Larsen, M.; Murphy, S. F.; Eberl, D. Chemicalweathering in a tropical watershed, luquillo mountains, Puerto Rico: I.Long-term versus short-term weathering fluxes. Geochim. Cosmochim.Acta 1998, 62, 209−226.(43) Tishchenko, V.; Meile, C.; Scherer, M. M.; Pasakarnis, T. S.;Thompson, A. Fe2+ catalyzed iron atom exchange and re-crystallization in a tropical soil. Geochim. Cosmochim. Acta 2015,148, 191−202.(44) Achtnich, C.; Bak, F.; Conrad, R. Competition for electrondonors among nitrate reducers, ferric iron reducers, sulfate reducers,and methanogens in anoxic paddy soil. Biol. Fertil. Soils 1995, 19, 65−72.(45) Thompson, A.; Chadwick, O. A.; Boman, S.; Chorover, J.Colloid mobilization during soil iron redox oscillations. Environ. Sci.Technol. 2006, 40, 5743−5749.(46) Reyes, I.; Torrent, J. Citrate-ascorbate as a highly selectiveextractant for poorly crystalline iron oxides. Soil Sci. Soc. Am. J. 1997,61, 1647−1654.(47) Hedley, M. J.; Stewart, J. W. B.; Chauhan, B. Changes ininorganic and organic soil phosphorus fractions induced by cultivationpractices and by laboratory incubations. Soil Sci. Soc. Am. J. 1982, 46,970−976.(48) Hedley, M. J.; Stewart, J. W. B. Method to measure microbialphosphate in soils. Soil Biol. Biochem. 1982, 14, 377−385.(49) Murphy, J.; Riley, J. P. A modified single solution method forthe determination of phosphate in natural waters. Anal. Chim. Acta1962, 27, 31−36.(50) Kostka, J. E.; Dalton, D. D.; Skelton, H.; Dollhopf, S.; Stucki, J.W. Growth of iron(III)-reducing bacteria on clay minerals as the soleelectron acceptor and comparison of growth yields on a variety ofoxidized iron forms. Appl. Environ. Microbiol. 2002, 68, 6256−6262.

(51) Stumm, W.; Morgan, J. J. Aquatic Chemistry: ChemicalEquilibrium and Rates in Natural Waters, 3rd ed.; John Wiley &Sons, Inc.: New York, NY, 1996; p 1022.(52) Van Beek, C.; Hiemstra, T.; Hofs, B.; Nederlof, M.; VanPaassen, J.; Reijnen, G. Homogeneous, heterogeneous and biologicaloxidation of iron (II) in rapid sand filtration. J. Water Supply Res. T.2012, 61, 1−13.(53) Senko, J. M.; Dewers, T. A.; Krumholz, L. R. Effect of oxidationrate and Fe (II) state on microbial nitrate-dependent Fe (III) mineralformation. Appl. Environ. Microbiol. 2005, 71, 7172−7177.(54) Steefel, C. I.; Van Cappellen, P. A new kinetic approach tomodeling water-rock interaction: The role of nucleation, precursors,and ostwald ripening. Geochim. Cosmochim. Acta 1990, 54, 2657−2677.(55) Lundquist, E. J.; Jackson, L. E.; Scow, K. M. Wet−dry cyclesaffect dissolved organic carbon in two california agricultural soils. SoilBiol. Biochem. 1999, 31, 1031−1038.(56) Kleber, M.; Eusterhues, K.; Keiluweit, M.; Mikutta, C.; Mikutta,R.; Nico, P. S., Mineral−organic associations: Formation, properties,and relevance in soil environments. In Advances in Agronomy; Donald,L. S., Ed.; Academic Press: 2015; Vol. 130, pp 1−140.(57) Kappler, A.; Newman, D. K. Formation of Fe(III)-minerals byFe(II)-oxidizing photoautotrophic bacteria. Geochim. Cosmochim. Acta2004, 68, 1217−1226.(58) Taillefert, M.; Beckler, J. S.; Carey, E.; Burns, J. L.; Fennessey,C. M.; DiChristina, T. J. Shewanella putrefaciens produces an Fe(III)-solubilizing organic ligand during anaerobic respiration on insolubleFe(III) oxides. J. Inorg. Biochem. 2007, 101, 1760−1767.(59) Royer, R. A.; Burgos, W. D.; Fisher, A. S.; Unz, R. F.; Dempsey,B. A. Enhancement of biological reduction of hematite by electronshuttling and Fe(II) complexation. Environ. Sci. Technol. 2002, 36,1939−1946.(60) Royer, R. A.; Burgos, W. D.; Fisher, A. S.; Jeon, B.-H.; Unz, R.F.; Dempsey, B. A. Enhancement of hematite bioreduction by naturalorganic matter. Environ. Sci. Technol. 2002, 36, 2897−2904.(61) Couture, R.-M.; Charlet, L.; Markelova, E.; Made, B. T.;Parsons, C. T. On−off mobilization of contaminants in soils duringredox oscillations. Environ. Sci. Technol. 2015, 49, 3015−3023.(62) Schulz, M.; Stonestrom, D.; Lawrence, C.; Bullen, T.;Fitzpatrick, J.; Kyker-Snowman, E.; Manning, J.; Mnich, M., Structuredheterogeneity in a marine terrace chronosequence: Upland mottling.Vadose Zone J. 2016, 15.010.2136/vzj2015.07.0102(63) Eusterhues, K.; Neidhardt, J.; Hadrich, A.; Kusel, K.; Totsche, K.Biodegradation of ferrihydrite-associated organic matter. Biogeochem-istry 2014, 119, 45−50.(64) Warren, L. A.; Ferris, F. G. Continuum between sorption andprecipitation of Fe (III) on microbial surfaces. Environ. Sci. Technol.1998, 32, 2331−2337.(65) Frankel, R. B.; Bazylinski, D. A. Biologically inducedmineralization by bacteria. Rev. Mineral. Geochem. 2003, 54, 95−114.(66) Mikutta, C.; Mikutta, R.; Bonneville, S.; Wagner, F.; Voegelin,A.; Christl, I.; Kretzschmar, R. Synthetic coprecipitates ofexopolysaccharides and ferrihydrite. Part I: Characterization. Geochim.Cosmochim. Acta 2008, 72, 1111−1127.(67) Liptzin, D.; Silver, W. L. Effects of carbon additions on ironreduction and phosphorus availability in a humid tropical forest soil.Soil Biol. Biochem. 2009, 41, 1696−1702.(68) Parsons, C. T.; Couture, R. M.; Omoregie, E. O.; Bardelli, F.;Greneche, J. M.; Roman-Ross, G.; Charlet, L. The impact of oscillatingredox conditions: Arsenic immobilisation in contaminated calcareousfloodplain soils. Environ. Pollut. 2013, 178, 254−263.(69) McGeehan, S. L.; Fendorf, S. E.; Naylor, D. V. Alteration ofarsenic sorption in flooded-dried soils. Soil Sci. Soc. Am. J. 1998, 62,828−833.(70) Mansfeldt, T.; Schuth, S.; Hausler, W.; Wagner, F.; Kaufhold, S.;Overesch, M. Iron oxide mineralogy and stable iron isotopecomposition in a gleysol with petrogleyic properties. J. Soils Sediments2012, 12, 97−114.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b05709Environ. Sci. Technol. 2017, 51, 3250−3259

3258

(71) Fredrickson, J. K.; Zachara, J. M.; Kennedy, D. W.; Dong, H. L.;Onstott, T. C.; Hinman, N. W.; Li, S. M. Biogenic iron mineralizationaccompanying the dissimilatoryreduction of hydrous ferric oxide by agroundwater bacterium. Geochim. Cosmochim. Acta 1998, 62, 3239−3257.(72) Zachara, J. M.; Fredrickson, J. K.; Li, S. M.; Kennedy, D. W.;Smith, S. C.; Gassman, P. L. Bacterial reduction of crystalline Fe3+oxides in single phasesuspensions and subsurface materials. Am.Mineral. 1998, 83, 1426−1443.(73) Glasauer, S.; Weidler, P. G.; Langley, S.; Beveridge, T. J.Controls on Fe reduction and mineral formation by a subsurfacebacterium. Geochim. Cosmochim. Acta 2003, 67, 1277−1288.(74) Kukkadapu, R. K.; Zachara, J. M.; Fredrickson, J. K.; Kennedy,D. W. Biotransformation of two-line silica-ferrihydrite by adissimilatory Fe(III)-reducing bacterium: Formation of carbonategreen rust in the presence of phosphate. Geochim. Cosmochim. Acta2004, 68, 2799−2814.(75) Borch, T.; Masue, Y.; Kukkadapu, R. K.; Fendorf, S. Phosphateimposed limitations on biological reduction and alteration offerrihydrite. Environ. Sci. Technol. 2007, 41, 166−172.(76) O’Loughlin, E. J.; Boyanov, M. I.; Flynn, T. M.; Gorski, C. A.;Hofmann, S. M.; McCormick, M. L.; Scherer, M. M.; Kemner, K. M.Effects of bound phosphate on the bioreduction of lepidocrocite (γ-FeOOH) and maghemite (γ-Fe2O3) and formation of secondaryminerals. Environ. Sci. Technol. 2013, 47, 9157−9166.(77) O’Loughlin, E. J.; Gorski, C. A.; Scherer, M. M.; Boyanov, M. I.;Kemner, K. M. Effects of oxyanions, natural organic matter, andbacterial cell numbers on the bioreduction of lepidocrocite (γ-FeOOH) and the formation of secondary mineralization products.Environ. Sci. Technol. 2010, 44, 4570−4576.(78) Zachara, J. M.; Kukkadapu, R. K.; Peretyazhko, T.; Bowden, M.;Wang, C. M.; Kennedy, D. W.; Moore, D.; Arey, B. The mineralogictransformation of ferrihydrite induced by heterogeneous reaction withbioreduced anthraquinone disulfonate (AQDS) and the role ofphosphate. Geochim. Cosmochim. Acta 2011, 75, 6330−6349.(79) Byrne, J. M.; Klueglein, N.; Pearce, C.; Rosso, K. M.; Appel, E.;Kappler, A. Redox cycling of Fe (II) and Fe (III) in magnetite by Fe-metabolizing bacteria. Science 2015, 347, 1473−1476.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.6b05709Environ. Sci. Technol. 2017, 51, 3250−3259

3259