The Fe(III) oxides employed in this study included a variety of pure synthetic phases [40] as well as three previously characterized Fe(III) oxide-bearing subsoil or subsurface materials (HC, CP, and Oyster). The synthetic oxides were prepared according to standard methods [50]. The HC and CP natural materials are Fe(III) oxide/layered silicate mixtures obtained from Ultisols in Tennessee and North Carolina, respectively. The Oyster material is Fe(III) oxide-coated sand from Pleistocene Age Atlantic Coastal Plain sediments. More detailed descriptions of the properties of these materials are available elsewhere [42,54,55]. The synthetic Fe(III) oxides were freeze-dried and passed through a 100-μm sieve, and their specific surface area determined by multipoint BET N₂ adsorption. The natural materials were air dried and passed through a 2-mm sieve prior to use in experiments.

Two well-characterized dissimilatory Fe(III)-reducing bacteria (FeRB) were employed in Fe(III) reduction experiments, Shewanella putrefaciens strain CN32 [14], and Geobacter sulfurreducens strain PCA [3,31]. The procedures used for growth and handling of these organisms for Fe(III) reduction experiments are described in detail in Roden [40] and Jeon et al. [19], respectively.

The synthetic and natural Fe(III) oxide-bearing solids were suspended in sterile, anaerobic Pipes buffer (10 mM, pH 6.8) contained in sealed serum vials to obtain Fe(III) concentrations of 5 to 250 mmol l⁻¹. Other experiments were conducted with soluble Fe(III) citrate as an electron acceptor. The electron donor for metal reduction was either 100% H₂ in the vial headspace, or 10 mM of either sodium lactate or sodium acetate. The medium was inoculated with ca. 10⁸ cells ml⁻¹ of either S. putrefaciens or Geobacter sulfurreducens cells. Samples for determination of dissolved (0.2-μm filtration and Ferrozine analysis) and total Fe(II) (0.5M HCl extraction and Ferrozine analysis) and pH were collected at 1–10-d intervals.

The synthetic and natural Fe(III) oxide-bearing solids were suspended in anaerobic 10 mM ascorbic acid or 10 mM AH₂DS (the reduced form of AQDS, anthroquinone-2,6-disulfonate, prepared by reacting AQDS with 100% H₂ gas in the presence of a palladium catalyst) in 10 mM Pipes buffer. The oxide suspensions were incubated at room temperature on a rotary shaker (250 rpm) and samples were removed with a N₂-flushed plastic syringe at regular intervals. A portion of the sample was passed through a 0.2-μm filter into Ferrozine for Fe(II) analysis, and the remainder used for determination of pH. The final (after reduction ceased) dissolved Fe(II) concentration and pH values achieved in the AH₂DS reduction experiments were used in conjunction with the $E_{\mathrm{h}}^0$ of the AQDS/AH₂DS couple (+0.23 V) [5] to estimate the reduction potential ($E_{\mathrm{h}}^0$) of the synthetic Fe(III) oxide phases.

Initial (0–3 d) surface area-specific rates of synthetic Fe(III) oxide reduction by S. putrefaciens and G. sulfurreducens did not vary systematically across a wide range of oxide surface area and $E_{\mathrm{h}}^0$ (Fig. 1A). In contrast, surface area-specific rates of abiotic Fe(III) oxide reduction by ascorbate and AH₂DS were significantly correlated with oxide surface area and $E_{\mathrm{h}}^0$ (Fig. 1B). These results indicate that initial rates of bacterial Fe(III) oxide reduction are not strongly controlled by oxide crystal thermodynamic properties. An explicit illustration of the relative influence of oxide crystal structural properties on initial rates of biological vs. chemical reduction can be drawn from the results obtained with synthetic lepidocrocite (diamonds circled in Fig. 1). Lepidocrocite possesses a lower degree of crystal order (less negative $\mathrm{\Delta }{G}_{\mathrm{f}}$, higher $E_{\mathrm{h}}^0$) than goethites of comparable particle size and surface area [51]. Consequently, lepidocrocite yielded a ca. two-fold higher initial rate of bacterial reduction compared to goethites with similar surface areas (Fig. 1A). However, the effect on the rate of reductive dissolution by ascorbate or AH₂DS was much more dramatic: lepidocrocite with a surface area of 64 m² g⁻¹ was reduced at a rate ca. two orders of magnitude greater than goethites with surface areas of 62 and 73 m² g⁻¹ (Fig. 1B).

Since detachment of a metal ion from an oxide surface site is generally viewed as the rate-limiting step in oxide mineral dissolution [52], it could be argued that because of the tendency for Fe(II) to reassociate (or never become detached in the first place) with oxide surfaces during enzymatic reduction at circumneutral pH (discussed in detail in Roden and Urrutia [47]), the kinetics of the enzymatic reduction system is not controlled by the presence of an obvious leaving group for which the detachment energy (related to the energy required for metal-ligand bond formation and breaking processes [1,47]) is affected by the thermodynamic properties of the oxide phase. However, during the bacterial Fe(III) oxide reduction experiments conducted under simplified aqueous geochemical conditions (i.e. in the absence of ions such as $\mathrm{HCO}_3{}^{-}$ and $\mathrm{PO}_4{}^{3-}$, which can induce formation of Fe(II) surface precipitates, more than 65% of total (0.5M HCl-extractable) Fe(II) production was accounted for by aqueous Fe(II) accumulation (data not shown). Hence, enzymatic Fe(III) oxide reduction was mainly a reductive dissolution process in these experiments. If Fe(II) detachment from the oxide surface during enzymatic reduction was affected by the thermodynamic properties of the oxide and thus controlled the bulk reduction rate, we would have expected to see a significant correlation between initial surface area normalized reductive dissolution rate and oxide $E_{\mathrm{h}}^0$ – as was clearly the case for reductive dissolution by ascorbate and AH₂DS. Since this was not the observed, we conclude that the mechanism and/or the rate-limiting step during enzymatic Fe(III) oxide reduction are fundamentally different than that for abiotic reductive dissolution. The simplest explanation is that the rate of electron transfer, rather than Fe(II) detachment, is the rate-limiting step during enzymatic reduction, and that rates of enzymatic electron transfer are not strongly affected by oxide thermodynamic properties.

Ongoing studies with Shewanella [10,12,32–35] and recent studies with Geobacter [24,29,30] have provided direct evidence that low redox potential, outer membrane-associated c-type cytochromes are involved in electron transfer from FeRB to Fe(III) oxides. In addition, a recent atomic force microscopy study by Lower et al. [28] demonstrated apparent molecular “recognition” of Fe(III) oxide surface sites by a putative ca. 150 kDa outer membrane protein of the dissimilatory FeRB Shewanella oneidensis (formerly S. putrefaciens strain MR-1), a close relative of the organism used in this study. Together these findings suggest the possibility that the similarity of surface area normalized electron transfer rates across a broad range of oxide minerals results from the fact that dissimilatory FeRB ‘recognize’ different Fe(III) oxide surfaces more or less equally independent of the underlying crystal structure, such that initial rates of electron transfer subsequent to recognition are not strongly dependent on crystal structure. This suggestion is consistent with an argument presented by Fischer [13] to account for the relatively minor influence of oxide solubility on rates of synthetic Fe(III) oxide reduction by Corynebacteria at pH 7. Fischer [13] reasoned that if the redox potential of the bacterial cells (i.e., their outer membrane c-type cytochromes) is sufficiently negative for reduction of well-crystallized oxide phases like hematite or goethite, each collision (or ‘recognition’) of a FeRB cell with an oxide particle will trigger reduction of a Fe(III) surface site. Therefore, the amount of Fe(II) produced during each collision event will not increase markedly with increasing oxide solubility.

Similar patterns of Fe(II) accumulation were observed during long-term (3-week incubation) reduction of the various synthetic Fe(III) oxides by G. sulfurreducens (Fig. 2A): after an initial period of rapid Fe(II) accumulation, rates of reduction decreased and Fe(II) levels approached an asymptote. pH values correlated directly with aqueous Fe(II) concentrations during Fe(III) oxide reduction (Fig. 2B), as expected from the stoichiometry of Fe(III) oxide reduction coupled to H₂ oxidation:

It could be argued that the similar surface area-specific rates of reduction observed for the different crystalline Fe(III) oxide phases (Fig. 1A), as well as the correlation between oxide surface area and extent of reduction (Fig. 4), was due to the presence of easily reducible amorphous Fe(III) oxide impurities in the synthetic crystalline Fe(III) oxides, whose abundance could have scaled with oxide surface area. If this were the case, the arguments put forward here and in other recent papers [39,40] regarding surface area vs. thermodynamic control of crystalline Fe(III) oxide reduction would be invalid. Data on initial rates of ascorbate-catalyzed Fe(III) oxide reduction were used to evaluate this question.

Fig. 5 shows time course data for reductive dissolution of hematite, two goethite phases (‘Goethite 30C’ and ‘High Surface Area’ (HSA) goethite), and HFO in the presence of excess (10 mM) of ascorbic acid at pH 3. HFO was fully dissolved within 4 h, whereas much smaller amounts of the crystalline Fe(III) oxide phases were dissolved over a 26-h period. As a result of the much faster kinetics of HFO dissolution compared to the other oxides, it was possible to use the time course data to quantitatively estimate the amount of amorphous HFO impurities present in the crystalline solids. The time course data for the crystalline phases indicated a relatively rapid initial accumulation of Fe(II) during the first few hours of reaction, followed by constant rate of oxide dissolution between 4 and 26 h. This initial rapid Fe(II) production can be attributed to dissolution of HFO impurities. The overall accumulation of Fe(II) during the initial period thus represents a combination of crystalline oxide and HFO impurity dissolution. A simple mathematical model that simulated parallel dissolution of a mixture of crystalline oxide and HFO was developed to estimate the abundance of the HFO impurity. The rate of crystalline oxide dissolution (in % total Fe(III) per hr) was set equal to the linear rate observed between 4 and 26 h, as defined by the least-squares regression analyses shown in Fig. 5A–C:

The rate of HFO dissolution was modeled as a first-order reaction, using the rate constant (1.37 h⁻¹) derived from the nonlinear least-square regression analysis shown in Fig. 5D:

Eqs. (1) and (2) were integrated numerically using a standard Runge–Kutta algorithm [38], and total % Fe(III) dissolution was computed as the cumulative sum of crystalline oxide and HFO reduction. The initial fractional abundance of HFO was adjusted to achieve agreement between the observed and simulated % Fe(III) dissolution vs. time data. The results of the analysis are illustrated by the dashed lines in Fig. 5A–C, which reflect estimated HFO impurities of 0.04, 0.10, and 0.32% of the total Fe(III) content of hematite, ‘goethite 30C’, and ‘HSA goethite’, respectively.

The estimated relative abundance of HFO impurities in the crystalline Fe(III) oxides can be compared to the amount of Fe(III) oxide reduction that took place in the G. sulfurreducens reduction experiments in order to constrain the potential contribution of such impurities to the observed Fe(III) reduction activity. The percent of Fe(III) subject to enzymatic reduction ranged from 2.3 to 9.7% for the crystalline hematite and goethite phases; these values are 5- to 95-fold higher than the estimated abundance of HFO impurity in the respective oxide phases. These results verify that enzymatic reduction of crystalline Fe(III) oxide surfaces – rather than reduction of amorphous Fe(III) oxide impurities – was the dominant mode of Fe(II) production in the cultures. The conclusions reached above regarding thermodynamic vs. surface-bound Fe(II) control of long-term crystalline Fe(III) oxide reduction are therefore valid, and they suggest that thermodynamic calculations should be used with caution when interpreting controls on enzymatic Fe(III) oxide reduction in soils and sediments.

Data from long-term experiments on bacterial and abiotic reduction of natural Fe(III) oxides (Fig. 6) were interpreted in relation to a standard a generalized rate law for mineral dissolution [22,36]:

Kinetic and thermodynamic considerations indicated that neither the abundance of electron donor (lactate) nor the accumulation of aqueous end-products of oxide reduction (Fe(II), acetate, dissolved inorganic carbon) are likely to have posed significant limitations on the long-term kinetics of oxide reduction [42]. Rather, accumulation of biogenic Fe(II) on residual oxide surfaces appeared to play a dominant role in governing the long-term kinetics of natural crystalline Fe(III) oxide reduction. This assertion is fully supported by the results of the synthetic Fe(III) oxide reduction experiments (Figs. 3 and 4).

The experimental findings summarized here point to a conceptual model of bacterial Fe(III) oxide reduction kinetics that differs fundamentally from established models of abiotic Fe(III) oxide reductive dissolution in that oxide surface area, rather than crystal structure and thermodynamic stability, exerts primary control on both the initial rate and the long-term extent of reduction. Numerical simulations of surface area-controlled biotic vs. abiotic Fe(III) oxide reduction indicate that this conceptual model can account for the pseudo-first-order kinetics of reduction of the operationally defined ‘microbially reducible’ fraction of the sediment Fe(III) oxide pool [42]. The explicit surface area control of the initial rate and extent of oxide reduction leads to a general rate law for oxide reduction as a function of electron acceptor and FeRB abundance that differs from those for reduction of chelated Fe(III) and other soluble metal species. As illustrated for synthetic goethite in Fig. 7, initial rates of Fe(III) oxide reduction are a linear function of oxide loading up to relatively high (200 mmol l⁻¹) bulk Fe(III) concentrations, a hyperbolic (Monod-style) function of total FeRB cell density, and a linear function of attached FeRB cell density. The latter relationships are analogous to the well-recognized dependence of abiotic reductive dissolution rate on total and surface-associated ligand concentration [16]. These relationships between oxide reduction rate and oxide/FeRB cell abundance are the opposite of those for reduction of soluble metals, as illustrated for Fe(III)-citrate in Fig. 8: rates of soluble metal reduction are a hyperbolic function of electron acceptor concentration, and a linear function of FeRB cell abundance. Similar patterns hold for reduction of other soluble metals such as U(VI), Co(III), Cr(VI), and Tc(VI) [25,27,43,45,53]. Table 2 provides a summary of rate laws that are appropriate for use in modeling solid-phase and dissolved Fe(III) (and other oxidized metals) reduction.

Virtually all experimental work to date on bacterial synthetic Fe(III)-oxide reduction indicates that oxide mineral heterogeneity in natural soils and sediments is likely to affect initial rates of bacterial reduction (e.g., during the early stages of anaerobic metabolism following the onset of anoxic conditions) mainly via an influence on reactive surface site density. Although variations in oxide thermodynamic properties may alter rates of enzymatic reduction to some extent (as illustrated by the studies of lepidocrocite vs. goethite reduction), this is likely to represent a secondary effect in relation to the primary controlling influence of oxide surface area. Hence, the surface area of different oxides phases present in a soil or sediment will exert primary control on initial rates of enzymatic reduction.

During the later stages of bacterial Fe(III) oxide reduction in permanently reduced sediments, accumulation of aqueous and surface-bound Fe(II) is expected to exert a dominant control on apparent Fe(III) oxide reactivity toward enzymatic reduction, particularly in situations where removal of Fe(II) end-products is slow compared to the kinetics of reduction [46]. This conceptual model is consistent with a recent analysis of the kinetics of hematite reduction by S. putrefaciens strain CN32 [2,48], which showed that initial rates of reduction were under kinetic control (presumably limited by the rate of electron transfer from FeRB cells to the oxide), whereas the long-term extent of reduction was limited by mass transfer of Fe(II) away from oxide/FeRB surfaces. As discussed above (see Fig. 4), there is a general relationship between oxide surface area and long-term extent of oxide reduction in closed reaction systems, which results from the function of oxide surfaces as a repository for sorbed and/or surface-precipitated biogenic Fe(II). Although the existence of this relationship implies a connection between extent of bacterial reduction and oxide thermodynamic properties, evidence suggests that this connection is not directly related to thermodynamic properties such as $E_{\mathrm{h}}^0$ or $\mathrm{\Delta }{G}_{\mathrm{f}}$, but rather results mainly from the correlation between these properties and oxide surface area.

An important implication of the above findings is that inferences regarding the ability of bacterial Fe(III) oxide reduction to compete with other terminal electron accepting processes (TEAPs) in soils and sediments should be based on estimates of bulk reactive (i.e. microbially accessible) surface site density – rather than the thermodynamic properties of the oxide(s) identified as the dominant phase(s) in a particular soil or sediment. This line of reasoning leads to the provisional conclusion that recent thermodynamic explanations for the coexistence of bacterial Fe(III) oxide reduction and other TEAPs (e.g., sulfate reduction and methanogenesis) in sediments and subsurface environments [17,18,37] have produced reasonable results for mechanistically incorrect reasons. The ability of thermodynamic considerations to explain the coexistence of bacterial Fe(III) oxide reduction and other TEAPs in sediments is likely the fortuitous result of the correlation between oxide thermodynamic properties and surface properties, which, based on current experimental information, actually control the initial reduction rate and long-term availability of Fe(III) oxides as competing electron acceptors for anaerobic respiration. Experiments with a mixed culture of wetland sediment bacteria and a range of synthetic Fe(III) oxides indicate that Fe(III)-reducing bacteria can outcompete methanogens for acetate with equal effectiveness when the different oxides are present at comparable surface area loadings – despite major differences in computed ΔG values for acetate oxidation coupled to Fe(III) oxide reduction [39]. These results emphasize the need for more accurate and robust wet-chemical (e.g., Hacherl et al. [15]) and/or spectroscopic techniques for assessing the surface properties (e.g., specific surface area and reactive site density) of natural Fe(III) oxide assemblages, including ones in which the presence of sorbed or surface precipitated Fe(II) limits the potential for enzymatic electron transfer and thereby controls apparent oxide reactivity toward microbial reduction.