Effects of aqueous complexation on reductive precipitation of uranium by Shewanella putrefaciens

We have examined the effects of aqueous complexation on rates of dissimilatory reductive precipitation of uranium by Shewanella putrefaciens. Uranium(VI) was supplied as sole terminal electron acceptor to Shewanella putrefaciens (strain 200R) in defined laboratory media under strictly anaerobic conditions. Media were amended with different multidentate organic acids, and experiments were performed at different U(VI) and ligand concentrations. Organic acids used as complexing agents were oxalic, malonic, succinic, glutaric, adipic, pimelic, maleic, citric, and nitrilotriacetic acids, tiron, EDTA, and Aldrich humic acid. Reductive precipitation of U(VI), resulting in removal of insoluble amorphous UO2 from solution, was measured as a function of time by determination of total dissolved U. Reductive precipitation was measured, rather than net U(VI) reduction to U(IV), to assess overall U removal rates from solution, which may be used to gauge the influence of chelation on microbial U mineralization. Initial linear rates of U reductive precipitation were found to correlate with stability constants of 1:1 aqueous U(VI):ligand and U(IV):ligand complexes. In the presence of strongly complexing ligands (e.g., NTA, Tiron, EDTA), UO2 precipitation did not occur. Our results are consistent with ligand-retarded precipitation of UO2, which is analogous to ligand-assisted solid phase dissolution but in reverse: ligand exchange with the U4+ aquo cation acts as a rate-limiting reaction moderating coordination of water molecules with U4+, which is a necessary step in UO2 precipitation. Ligand exchange kinetics governing dissociation rates of ligands from U(VI)-organic complexes may also influence overall UO2 production rates, although the magnitude of this effect is unclear relative to the effects of U(IV)-organic complexation. Our results indicate that natural microbial-aqueous systems containing abundant organic matter can inhibit the formation of biogenic amorphous UO2.


I. INTRODUCTION
Uranium ͑U͒ is a prominent environmental radiocontaminant at a local to regional scale throughout much of the developed world. Most historic releases of U into natural aquatic systems have occurred through the agencies of U mining, ore processing, isotopic separation, and waste disposal related to the production of fission weapons, and the operation of commercial nuclear reactors. Other major sources of environmental U dispersal include coal combustion, releasing U through fly ash or stack emissions, and the military use of 235 U-depleted U͑IV͒ oxide in kinetic penetrators.
Under oxic conditions the stable form of uranium is U͑VI͒, which is highly soluble and in aqueous solution tends to adsorb onto available mineral and organic surfaces. [1][2][3][4][5][6][7][8][9][10][11] Uranium͑VI͒ adsorption onto sedimentary minerals may retard its propagation through aquifers, however dispersal could be facilitated by adsorption onto transportable organic matter, such as bacteria. 12,13 Transport in the aqueous phase is also facilitated by the tendency of U͑VI͒ to stably complex with a wide range of inorganic and organic ligands. 14,15 Reduction of soluble U͑VI͒ to relatively insoluble U͑IV͒ may serve as one potential method, under anoxic conditions, of removing uranium from the aqueous phase, or of sequestering uranium adsorbed onto organic or mineral surfaces. Uranium͑IV͒ is sparingly soluble, and will tend to precipitate forming amorphous UO 2 (s). Abiotic methods exploiting U͑VI͒ reduction as a means of U removal from groundwater have been approached, including permeable groundwater barriers incorporating zero-valent iron or sodium dithionite as reducing agents. 16 -18 These methods seek to isolate U from the biotic environment through chemical reductive precipitation, and could presumably offer a more permanent remedial solution than sedimentary U͑VI͒ adsorption, which would be vulnerable to reversibility via shifts in groundwater pH ͑e.g., industrial acidification͒ or ionic strength ͑e.g., salt water intrusion͒. Bacterial dissimilatory U͑VI͒ reduction, leading to U͑IV͒ precipitation as UO 2 (am), has been proposed as an alternative means of promoting U mineralization under industrial or in situ environmental conditions. 19 Diverse bacteria are capable of dissimilatory U͑VI͒ reduction, in many cases obtaining energy for growth by coupling U͑VI͒ reduction to the oxidation of organic carbon or H 2 . 20,21 Bacteria such as Shewanella putrefaciens, 22,23 Geobacter metallireducens and G. sulfurreducens 24 -26 are capable of employing U͑VI͒ as sole terminal electron acceptor ͑TEA͒, coupled a͒ Author to whom correspondence should be addressed; electronic mail: johnson.haas@wmich.edu with use of organic carbon or H 2 as electron donor. Other bacterial groups capable of dissimilatory U͑VI͒ reduction include Desulfovibrio [27][28][29][30][31][32] Pseudomonas, 23,33 Thermus, 34 and some halophilic Archaea, 35 although many bacteria are not apparently capable of growth using U͑VI͒ as sole terminal electron acceptor.
Previous studies have examined the influence of aqueous organic complexation on U͑VI͒ bioreduction by S. putrefaciens and D. desulfuricans 36,37 and report that high concentrations of strongly chelating organic ligands can have an appreciable effect on the rate of U͑VI͒ reduction and precipitation by bacteria. Ligands that weakly complex U͑VI͒, such as acetate, result in overall U͑VI͒ reductive precipitation rates similar to control media containing unchelated U͑VI͒, whereas strongly chelating ligands, such as oxalate, citrate, or tiron ͑4,5-dihydroxy-1,2-benzene disulfonic acid͒ tend to retard microbial U͑VI͒ reduction. However, most previous studies 36,38,37 do not offer predictive estimates of U removal rates, therefore it is difficult to extend the available data to diverse geochemical conditions.
Haas and DiChristina 39 report a systematic relationship between the equilibrium stabilities of aqueous organic Fe͑III͒ complexes and rates of microbial dissimilatory Fe͑III͒ reduction by S. putrefaciens under laboratory conditions. These results demonstrate that competitive coordination reactions, involving aqueous ligands and enzymatic coordination sites at the bacteria-water interface, can govern the bioavailability of Fe͑III͒ in a systematic and predictive fashion. It is unknown if similar relationships hold with respect to U͑VI͒. Uranium in waste from industrial isotopic separations, or in spent nuclear fuel, is often accompanied by high concentrations of organic ligands ͑e.g. EDTA͒ used in cleansing of machine components. 33,40 Concentrations of EDTA up to 5876 g/g have been reported in the aqueous phase of Hanford mixed high-level nuclear waste, along with 5965 g/g citric acid, 591 g/g oxalic acid, and 260 g/g NTA. 41 Thus, understanding the effects of organic chelation on microbial U͑VI͒ reductive precipitation is crucial in forecasting the efficacy of proposed bioremedial designs, and in predicting the long-term fate and transport of existing and future U groundwater contamination.
In the present study we examine the effects of aqueous organic complexation on rates of U͑VI͒ reductive precipitation by S. putrefaciens under controlled laboratory conditions. A suite of low molecular weight polycarboxylic acids and one polysulfonic acid were selected, including maleic acid, citric acid, nitrilotriacetic acid ͑NTA͒, 4,5-dihydroxy-1,3-benzene-disulfonic acid ͑tiron͒, ethylenediaminetetraacetic acid ͑EDTA͒ and saturated aliphatic dicarboxylic ligands of carbon number two to seven ͑oxalic, malonic, succinic, glutaric, adipic, and pimelic acids͒. These particular acids were chosen to systematically examine speciation effects on U precipitation rates, as a function of chelate ring size and aqueous U͑VI͒ complex stability. Aldrich humic acid was used as a natural organic matter ͑NOM͒ analog, and to compare with data obtained from low molecular-weight ligands. Rates of U reductive precipitation, as UO 2 (am), were measured in media containing different initial U͑VI͒-organic aqueous complexes. Experimental rates were considered alongside properties of U͑VI͒ and U͑IV͒ organic complexes, to develop a generalized and predictive understanding of aqueous chelation effects on U biomineralization.

A. Bacteria and cultivation methods
S. putrefaciens ͑strain 200R͒ was used in this study. Bacteria of the genus Shewanella are widely recognized and used in biogeochemical research as model bacteria for examining dissimilatory metal reduction., see e.g., Refs. 42-46, 21, 39, 2, 22, 47. S. putrefaciens is a versatile, facultatively aerobic, Gram-negative metal-reducing bacterium, capable of obtaining energy for growth using Fe͑III͒, Mn͑IV͒, U͑VI͒, NO 3 Ϫ , SO 3 Ϫ , Cr͑VI͒, Tc͑VII͒ and fumarate, each as sole TEA. 22,48 -51 In this study cultures were prepared from frozen glycerol-treated stock maintained at Ϫ80°C, transferred to agar plates of Luria-Bertani ͑LB͒ media, and cultured at 30°C aerobically for at least 24 h prior to use. Single colonies from viable, growing LB-agar plates were transferred to sterile liquid LB media and cultivated in 50 mL volumes aerobically at 30°C in a heated, shaking environmental chamber for 24 h. These cultures, harvested at midexponential phase, were used as inocula for U͑VI͒ reduction experiments. Initial cell density in all experiments was approximately 10 9 cells/mL. Surveys of subsurface aquifers report microbial populations on the order of 10 5 -10 8 cells/cm 3 , and even higher populations in eutrophic surface waters. [52][53][54] The relatively high cell density used in the experiments, relative to subsurface natural conditions, was adopted to facilitate rapid U͑VI͒ reduction kinetics that would allow experiments of shorter duration. This was done to minimize potential cell lysis problems associated with lengthy batch cultivations.
Cell density was measured as mass per unit volume in initial inocula, using procedures outlined in Haas. 55 Cell number per unit volume of solution was calculated using an average wet mass value of 7.3ϫ10 12 bacteria/g, derived from replicate measurements of S. putrefaciens bulk mass under a range of cultivation conditions, combined with cell population values obtained via direct-counts of acridine orange stained bacteria, measured using epifluorescence microscopy. 55,2

B. Sample preparation and media formulation
Active aerobically grown cultures of S. putrefaciens were transferred to a Coy® glove box anaerobic chamber and degassed with recirculated interior atmosphere (95%N 2 /5%H 2 ) for 15 min prior to inoculation to remove dissolved oxygen from growth media. Experiments were carried out in 50 mL sterile polypropylene centrifuge tubes, into each of which was introduced 50 mL of sterile experimental growth medium and 100 L of inoculant. Sterile experimental media were allowed to equilibrate in the anaerobic chamber at 30°C for 24 h prior to inoculation. Bacteria were washed twice in sterile anoxic 0.1 M NaCl prior to inoculation into experimental media.
Experimental media consisted of a minimal basal salts solution modified from Lovely et al. 26 to exclude phosphate and bicarbonate, but to include dissolved uranyl acetate ͑1 mM to 10 M͒ and selected chelating ligands ͑200 to 20 mM͒. The common media composition was ͑per L͒: NH 4 Cl 0.25 g, KCl 0.1 g, Na-lactate ͑60% syrup͒ 5 mL, Sigma® RPMI-1640 vitamins solution 0.1 mL, trace elements as per Lovely et al., 26 and adjusted to pH 7 using sterile trace metal grade NaOH and HCl. Additional pH buffering agents were excluded to avoid potential interferences due to complexation of buffer species with U͑VI͒.
All experimental media were prepared from reagentgrade materials, autoclaved or filter-sterilized ͑0.2 m͒, and degassed inside the anaerobic chamber to remove dissolved oxygen. The anaerobic chamber was thermostated at 30°C throughout all experiments. Trials were carried out in parallel sets of 50 mL polypropylene reaction vessels, each containing a specific U͑VI͒-ligand combination, each set inoculated from the same source culture, and incubated under gentle agitation on a rocking stage inside the anaerobic chamber.

C. Ligand selection and speciation
Ligands selected for this study include a set of compounds broadly representing low molecular weight organic acids in natural systems, as well as some organic contaminants found with anthropogenic U contamination. These include six saturated dicarboxylic acids of carbon number two to seven: oxalic, malonic, succinic, glutaric, adipic and pimelic acids. Experiments also examined maleic acid ͑an unsaturated aliphatic dicarboxylic acid͒, citric acid ͑a saturated aliphatic dicarboxylic, monohydroxylic acid͒, NTA ͑saturated aliphatic tricarboxylic acid͒, tiron ͑an aromatic disulfonic acid͒, and EDTA ͑an aliphatic tetracarboxylic acid͒. Humic acid ͑Aldrich͒ was also used to compare the effects of refractory NOM with those of low molecular weight organic ligands.
Concentrations of U͑VI͒ and organic ligands were tailored to ensure that U͑VI͒-ligand complexes ͑of varying sto-ichiometry͒ dominated the speciation of U in all trials. The aqueous speciation of U in experimental media was calculated using the algorithm JCHESS 56 and its associated database, together with values for U͑VI͒-ligand and U͑IV͒-ligand complexation stability constants derived from the literature, 57,58 and corrected where necessary to infinite dilution. 59 Speciation calculations accounted for complexation of U͑VI͒ with inorganic ͑e.g., hydroxide, chloride͒ and organic components ͑lactate, added ligand͒ present in the experimental media at their appropriate concentrations, organic acid dissociation reactions, and complexation equilibria involving pertinent aqueous cations and inorganic anions. Ion activity coefficients were accounted for using the extended Debye-Huckel model.
Aqueous complexation of U͑IV͒ with strongly chelating organic ligands can result in stabilization of dissolved U͑IV͒, however a lack of reliable stability constants for U͑IV͒ complexation prevents a quantitative estimation of the precise magnitude of this phenomenon in the current study. Fujiwara et al. 60 and Allard et al. 58 provide equilibrium constants for U͑IV͒ hydrolysis, and the compilation of Smith et al. 57 provides data for U͑IV͒ complexation with some organic ligands at elevated ionic strength. 61,62 However, there is a general paucity of literature values for U͑IV͒ aqueous complexation. As a result, it is difficult to assess quantitatively the combined impacts of U͑VI͒ aqueous complexation and U͑IV͒ complexation on bulk U mineralization in a microbialaqueous context. This is one motivation for the current project, which seeks to evaluate the combined influence of both processes on U͑VI͒ reductive removal from solution. Further work is required to quantify the stabilities of U͑IV͒organic complexes generally, but that work is beyond the intended scope of the present study.
Uranium͑VI͒-ligand aqueous complexes accounted for at least 99% of dissolved U in each solution at each set of concentration conditions employed in this study. In all cases the predominant form of dissolved U in the experimental media was an aqueous complex of U͑VI͒ with the appropriate added ligand. Lactic acid was present in all experiments as a bacterial carbon source, but does not strongly complex with U͑VI͒ in comparison with other ligands used in this study, and under none of the experimental conditions other than the inoculated control were U͑VI͒-lactate complexes expected to account for a significant fraction of dissolved U͑VI͒.

D. Analytical methods
At time intervals throughout the experiments, samples were extracted from reaction vessels and total dissolved U analyzed using ICP-MS. Samples were taken via syringe at time intervals throughout each experiment, immediately filtered inside the anaerobic chamber to 0.2 m using a fresh 5 mL disposable syringe and nylon syringe filter, and stabilized by addition of trace metal grade HNO 3 . Microbial U͑VI͒ reduction can result in formation of nanoparticulate UO 2 (am), 63 which could potentially pass a 0.2 m filter. For this reason both filtration ͑0.2 m͒ and ultracentrifugation ͑14 000 rpm for 30 min in a microcentrifuge͒ were tested in this study. Both methods yielded similar results ͑data not shown͒.
Acid-stabilized filtered samples were analyzed by ICP-MS, using Dy as an internal standard, and calibrated against U standards spanning the experimental concentration range. Ten replicate analyses of U were performed during each analysis, which for most samples provided approximately Ϯ0.5% analytical relative standard deviation. Overall uncertainty for most experiments is estimated to be ϳ5% including analytical and sampling errors. For experiments involving humic acid, sampling and filtration uncertainty is larger (ϳ15%), based on replicate sample determinations. Each diagram presents data for dissolved U concentrations in experimental media as a function of time in hours, and includes data representing an inoculated control trial containing only lactate as a bacterial carbon source, but no additional chelating ligands. Sterile control trials were conducted separately, and showed no evidence of UO 2 (am) precipitation after 30 days, the maximum period tested in this study ͑data not shown͒. Figure 1 shows data for U͑VI͒ reduction experiments in which total available U was 10 M and total available ligand was 20 mM. For humic acid, 0.2 g/50 mL were used. The inoculated control trial shows rapid loss of U from the aqueous phase, as U͑VI͒ is reduced by the bacteria to U͑IV͒ and precipitated as UO 2 (am). Production of solid UO 2 (am) was visually evident in this and succeeding experiments by the appearance of a black precipitate. At an initial U͑VI͒ concentration of 10 M U, little black precipitate was evident in bulk solution, but the solid became visually apparent when it accumulated on filter membranes during sample filtration. In Fig. 1 the inoculated control trial demonstrates that, in the absence of additional U-chelating ligands, U reductive removal is rapid, removing ϳ95% of available U from solution after approximately 48 h. From 48 h to the maximum sampling period of 96 h, little additional U precipitation was evident. The rate of UO 2 (am) precipitation in the inoculated control trial was essentially linear with time for the first 3 h of the experiment, after which the rate diminished asymptotically. These results are similar to those reported previously in studies of microbial U reductive precipitation-see e.g., Ref.

26.
Trials with several other ligands display behavior similar to the inoculated control. Curves for pimelic, adipic, glutaric, succinic, and maleic acids are broadly similar in shape to that of the inoculated control. In contrast, the data for malonate show slower overall U͑VI͒ reductive removal, and even slower rates for humic acid. Values for malonic acid and humic acid further demonstrate that U͑VI͒ reductive removal, up to 96 h, does not proceed to the extent of completion observed for the control ͑and many other ligands͒. Approximately 15% of available U͑VI͒ remains in solution at the end of 96 h in the presence of 20 mM malonic acid, while half of the available U͑VI͒ remains in solution in the presence of 4 g/L humic acid. Citric acid, oxalic acid, and NTA do not show clear evidence of any U removal, although data at 96 h for NTA show lower concentrations that may indicate incipient partial U removal.
Similar results are obtained from experiments in which the concentration of initial U͑VI͒ is increased to 100 M, maintaining ligand concentrations at 20 mM ͑or 4 g/L humic acid͒. Under these conditions, experimental data ͑shown in Fig. 2͒ show rapid U precipitation in the inoculated control, with similarly rapid U precipitation in trials containing pimelic and adipic acids. In trials containing glutaric, succinic, and malonic acids, U precipitation rates are marginally slower, while maleic acid appears to allow initially rapid reduction but ultimately stalls at ϳ50% of initial U͑VI͒ remaining in solution. For oxalic acid, citric acid, NTA, and humic acid, no measurable U removal is observed up to 96 h. At 1 mM available U͑VI͒ and 200 mM ligand, results broadly follow the above-observed patterns, but with some significant differences ͑Fig. 3͒. The inoculated control and trials containing pimelic, adipic, and glutaric acids exhibit rapid reduction of U͑VI͒, leaving less than 1% of available U in solution at the end of 96 h. Trials containing succinic, malonic, and maleic acids show UO 2 (am) precipitation at initial rates that are slower than the inoculated control. Both malonic and maleic acids limit the extent of overall U removal to about 35%-50% by the end of the experiment. Oxalic acid trials possibly show some initial U removal, although after 9 h little additional U removal occurs, leaving ϳ85% of initial U in solution at the end of 96 h. No U precipitation is observed in trials containing citric acid or NTA.
In general, the experimental results demonstrate that the presence of organic chelating ligands can significantly influence the rate of microbial U͑VI͒ reductive removal, depending on ligand type and concentration. Some ligands, such as pimelate, adipate, glutarate and succinate, did not significantly retard the rate of UO 2 (am) precipitation by S. putrefaciens, while the ligands maleate, malonate, and oxalate curtailed UO 2 (am) formation, and in some cases appeared to arrest this process prior to completion. Strongly chelating ligands, such as citrate or NTA, and including Aldrich humic acid, appeared to effectively prohibit UO 2 (am) formation by S. putrefaciens.

IV. DISCUSSION
The observed effects of aqueous U chelation on UO 2 (am) production rates likely derive from a combination of ͑1͒ aqueous complexation of initial uranyl with organic ligands, and ͑2͒ aqueous organic complexation of U͑IV͒ produced by microbial catalysis of uranyl reduction. To quantify the influence of aqueous complexation reactions on U reductive removal, initial linear U removal rates were obtained from experimental data shown in Figs. 1-3, using only those initial portions of ligand-specific trends that could be best approximated as linear. For most trials UO 2 (am) precipitation rates could be approximated as linear within the first 8 -9 h of each experiment, when electron donor and acceptor concentrations were not limiting. Uranium removal rates further into experiments were slower, consistent with expectations based on Michaelis-Menten kinetics. Initial U removal rates correlated with initial U͑VI͒ concentration, as shown in Fig. 4, which shows results for inoculated control experiments containing no added ligands beyond lactate. Initial U removal rates are linear in initial U͑VI͒ concentration, indicating that the maximum Michaelis-Menten U͑VI͒ conversion rate was not approached in the experiments, where the maximum initial U͑VI͒ concentration was 1 mM. Thus, U reductive removal rates measured in ligand-amended media-which were slower than rates in control mediawere not constrained relative to each other by the ceiling imposed by the maximum enzymatic turnover rate. It is therefore expected that variations among rates in different media result principally from U speciation effects, and not the bacterial processing rate for U͑VI͒.
Absorption of U͑VI͒ onto bacterial surfaces was not a significant sink for U during experiments, and is not likely to have contributed measurably to removal of U from solution during trials. Using experimental equilibrium constants for U͑VI͒ adsorption onto S. putrefaciens, 2 along with data for media solute and cell concentrations, calculations conducted using a constant capacitance model and the speciation algorithm JCHESS indicate that even at the lowest U concentration employed in experiments, 10 M U, less than 0.1% of available U adsorbs onto the bacterial surfaces. Thus, removal of U during experiments can be attributed primarily to reductive precipitation of UO 2 (am), and not to adsorption onto bacterial cell membranes.
Previous work examining dissimilatory Fe͑III͒ reduction by S. putrefaciens 39 demonstrates that, under laboratory conditions in pure culture, initial Fe͑III͒ bioreduction rates depend strongly on the stability constant of the relevant 1:1 aqueous Fe͑III͒-organic complex. This relationship was observed even under conditions where the predominant Fe͑III͒ aqueous species was a higher-order ͑i.e., 1:2, 1:3 cation:li-gand͒ complex. It was interpreted from this observation that, under conditions where aqueous ligands and bacterial coordination sites compete for available dissolved Fe͑III͒, the rate-limiting step constraining Fe͑III͒ coordination with the bacterial surface is the leaving rate of the last conjoined ligand of an aqueous Fe͑III͒-organic complex. This leaving rate is equivalent to the rate of ligand exchange from the cation:ligand complex, which under isothermal conditions where one cation is considered, is dependent primarily on the equilibrium stability constant. A similar relationship is observed in the present study with respect to U͑VI͒; in Fig. 5, initial linear U precipitation rates are compared with stability constants for aqueous U͑VI͒-organic 1:1 complexes. Rates shown in Fig. 5 are normalized to the maximum rate for each U͑VI͒:ligand concentration condition.
The trend observed in Fig. 5 demonstrates that U reductive removal by S. putrefaciens is strongly dependent on U͑VI͒ aqueous speciation, in a manner similar to that observed for Fe͑III͒. Under conditions where U͑VI͒ is weakly complexed, for example by lactate or high carbon-number ͑i.e., large chelate ring͒ dicarboxylic aliphatic acids, UO 2 (am) formation is rapid. Where U͑VI͒ is strongly complexed, as by citrate, NTA, tiron, or EDTA, little to no UO 2 (am) formation is observed. At intermediate values of U͑VI͒-ligand complex stability, the UO 2 (am) formation rate is a function of the 1:1 complex-formation log K value. Values at 10 M, 100 M, and 1 mM initial U͑VI͒ fall on the same trend, indicating that the relationship is dependent pri- FIG. 4. ͑Color͒ Initial U reductive removal rates in lactic acid trials ͑inoculated controls͒ as a function of initial U͑VI͒ concentration. Line represents a best-fit linear regression of the data. marily on the extent of U chelation, and the stability of the chelate formed. For the organic ligands tested, the equilibrium stability constant of the first-order U͑VI͒-ligand complex is a good indicator of the efficacy of U reductive mineralization, regardless of absolute U activity. This facilitates application of the observed relationship to natural conditions, which are typically characterized by low M concentrations of aqueous U in all but the most contaminated settings. However, even in highly contaminated settings, with aqueous U concentrations in the high g/g range, the experimental data would appear to remain applicable. The maximum U concentration tested in this study ͑1 mM͒ is 238 g/g, which is higher than U concentrations recorded in most severely perturbed aquatic systems, including U mining or high-level waste release sites. 64 An alternative explanation for the trends observed in Figs. 1-3 could involve chelation of the added ligands not with U͑VI͒, but with U͑IV͒. Stable aqueous complexation of U͑IV͒ would tend to raise the solubility of UO 2 (am) by diminishing the equilibrium activity of the aquo U 4ϩ cation. A result would be retention of U in solution, despite successful reduction of U͑VI͒ to U͑IV͒ by the bacteria. This mechanism would, for example, undermine bioremedial efforts that hinge on UO 2 (am) formation. It might also inhibit the formation of uraninite ore bodies in subsurface settings where microbial activity is the principal mediator of ore formation. High activities of dissolved organic matter, in such a situation, would retard U mineralization, potentially acting as a feedback limitation on microbially promoted uraninite ore accumulation. This hypothesis is supported by Ganesh et al. 36 and Robinson et al., 37 who report aqueous U͑VI͒ reduction by S. putrefaciens and D. desulfuricans in laboratory cultures in the presence of citric acid and tiron, but little to no UO 2 (am) precipitation. A quantitative assessment of the dependence of U reductive precipitation rates on the stability of U͑IV͒-organic complexes is problematic, however, due to the general paucity of reliable U͑IV͒ aqueous-complexation equilibrium constants in the literature.
Experimental data for U͑IV͒ aqueous complexation with lactic acid are provided in the critical review and compilation of Smith et al. 57 but only at an ionic strength of 1.0 M. Carey and Martell 61 provide data for U͑IV͒ complexation with EDTA and CDTA, while Carey and Martell 62 provide data for mixed organic-ligand complexes with U͑IV͒. More data are available for Th͑IV͒ complexation with organic ligands ͑referenced in Smith et al. 57 ͒. Thermodynamic data for Th͑IV͒ complexation were used in this study to estimate complexation stability constants for U͑IV͒ complexation, using a linear free energy correlation. Figure 6 illustrates a covariation between aqueous complex-formation equilibrium constants for Th͑IV͒ and U͑IV͒, where each value corresponds to the logarithm of the first-order formation constant of a respective cation:ligand aqueous complex. Values in Fig.  6 are taken from the literature 57,61,62 and are uncorrected for ionic strength, on the basis that activity coefficient terms correcting Th͑IV͒-ligand and U͑IV͒-ligand equilibrium constants to infinite dilution, where cation, ligand, and complex charges are constant with respect to each Th/U pair of data, are identical, resulting only in displacement of values along the observed trend. The trend in Fig. 6 was used to predict values for U͑IV͒ complexation constants for oxalic, malonic, succinic, maleic, glutaric and citric acids. These predicted values, along with literature values for U͑IV͒ complexation with EDTA 61 and lactic acid, 57 corrected to infinite dilution 59 are compared with rate data in Fig. 7.
In Fig. 7 formation constants for 1:1 U͑IV͒-ligand aqueous complexes are plotted against UO 2 (am) formation rates normalized to initial U͑VI͒ concentration. Figure 7 is similar in structure to Fig. 5, and shows a similar trend with respect to the dependence of UO 2 (am) formation rates on U complexation. In Fig. 7, rates are seen to vary in a systematic fashion with the stability of U͑IV͒ complexation, with weak U͑IV͒ complexation promoting more rapid U reductive precipitation, strongly stable U͑IV͒ complexation preventing to an equivalent set of ligands ͑e.g., Refs. 65 and 66͒.
The trend observed in Fig. 7 supports a model of ligandretarded precipitation of UO 2 (am). This interpretation is theoretically consistent with expectations based on previous work describing ligand-assisted dissolution of metal-oxide phases. For example, Ludwig et al. 67,68 report experimental results for ligand-assisted dissolution of bunsenite ͑NiO͒, showing a positive linear correlation between the stability of aqueous Ni:L organic complexes and measured NiO dissolution rates. This relationship arises due to mediation of water exchange with solid-phase or aqueous cations and a coordinating ligand. An inverse of this process would be expected to occur with respect to solid phase precipitation under supersaturated conditions: strong ligand coordination with aqueous cations should retard the rate of precipitation, resulting in a negative slope on a log K ML ͑equilibrium formation constant of a metal:ligand complex͒ versus r precip ͑initial linear rate of precipitation͒ diagram. This relationship is observed in Fig. 7.
Rates of UO 2 (am) precipitation drop to zero in the presence of strongly chelating ligands, and this result would be anticipated under conditions where rates are either immeasurably slow, or where the solution is no longer saturated with U͑IV͒. The latter possibility is difficult to ascertain, because of the paucity of reliable values for U͑IV͒ aqueous complexation constants. For weakly complexing ligands, UO 2 (am) precipitation rates would be expected to increase as a linear function of decreasing values for log K ML , however the precipitation rate will not exceed the maximum rate imposed by the rate of bacterial U͑VI͒ to U͑IV͒ turnover, which is itself a function of intrinsic Michaelis-Menten kinetics, and is also likely to depend on U͑VI͒ speciation. These envisaged constraints on UO 2 (am) precipitation rates are shown qualitatively in Fig. 7, as ͑1͒ a sloping dashed line, denoting the effect of ligand-retardation on precipitation rates, ͑2͒ a horizontal dashed line, denoting the maximum rate of bacterial U͑VI͒ conversion, ͑3͒ the x axis at an intercept of 0 along the y axis, denoting a rate of zero under conditions of likely UO 2 (am) undersaturation, and ͑4͒ the dotted curve which interpolates a smooth relationship bounded by constraints ͑1͒ through ͑3͒.
It is significant that ͑Aldrich͒ humic acid was found in this study to suppress dissimilatory UO 2 (am) precipitation rates. Gu and Chen 46 found that polycondensed aromatic humic materials can enhance the rate of microbial dissimilatory U͑VI͒ reduction, presumably by functioning as effective electron shuttles. Although some humic substances may prove effective at promoting electron delivery to U͑VI͒ from bacteria, our results suggest that humics can also suppress UO 2 (am) production via ligand-retarded precipitation. In Fig. 5, U removal rates in the presence of ͑Aldrich͒ humic acid are plotted using literature values for 1:1 U͑VI͒:humic complexation. 69 Corresponding values for U͑IV͒ complexation with humic acid are unavailable, but because the trends in Figs. 5 and 7 are similar, U͑VI͒ complexation constants can suffice as an approximation.
Values from Lenhart et al. 69 were adjusted to neutrality using the pH dependence of the original data, and approximated to infinite dilution using Debye-Huckel parameters assuming the ligand charge behavior of citric acid, yielding a conditional log K ͑1:1͒ for U͑VI͒:humic acid of ϳ9. Using this approach, humic acid falls on the same overall trend as other low molecular weight organic acids. This result suggests that natural waters containing high concentrations of dissolved NOM could effectively suppress microbial UO 2 (am) formation. This observation is consistent with field data showing no apparent U͑IV͒ precipitation in naturally U-enriched freshwater peat bog sediments. 70 In contrast, marine sediments commonly show evidence for U reductive mineralization as a function of increasing sediment depth. 30,[71][72][73][74][75][76][77] This may occur due to weaker U͑VI͒ or U͑IV͒ complexation with fulvic acids 69 in marine sediments where organic matter degradation is more complete, or may be due to an intrinsic difference in the U͑VI͒ reductive capabilities of the respective microbial flora in each setting. Further studies investigating the effects of NOM complexation on microbial dissimilatory UO 2 (am) production would be required to better assess these possibilities.
At present a quantitative extrapolation of the experimental results to field conditions is problematic, because of ͑1͒ limited data describing U͑IV͒ aqueous complexation and solubility, ͑2͒ an incomplete understanding of the effects of U͑VI͒ aqueous speciation on dissimilatory U͑VI͒ reduction rates, and ͑3͒ an almost complete lack of information on ligand-impacted dissolution and precipitation rates for biogenic UO 2 (am). The experimental data suggest that under natural conditions where chelating organic ligands are abundant, production of UO 2 (am) is constrained by ligand complexation with dissolved U͑IV͒. Aqueous U͑VI͒ complexation will also tend to impact rates of dissimilatory U͑VI͒ reduction, however, and the combination of these two largely unquantified processes is likely to complexly govern U biogeochemical speciation in natural settings. Further work is required to assess these factors.

ACKNOWLEDGMENTS
This research was supported by funds provided by Western Michigan University. Our work benefited greatly from many helpful discussions with Carla Koretsky, Charlie Moore, David Fowle, and Tom DiChristina. Technical assis- FIG. 7. ͑Color͒ Equilibrium formation constants for aqueous U͑IV͒:ligand complexes, plotted vs initial linear U reductive precipitation rates from experiments. Rates are normalized to the maximum rate from each set of results at constant U:organic acid concentrations. The dashed and dotted lines denote potential constraints on UO 2 precipitation rates, and are discussed in the narrative. tance was provided by Nancy Morgan. This is contribution number 4 of the Microbial Aqueous Geochemistry, Biogeochemistry and Thermodynamics ͑MAGBAT͒ laboratory at WMU.