Original Article

Subject Category: Integrated genomics and post-genomics approaches in microbial ecology

The ISME Journal (2011) 5, 305–316; doi:10.1038/ismej.2010.117; published online 29 July 2010

Genome-scale dynamic modeling of the competition between Rhodoferax and Geobacter in anoxic subsurface environments

Kai Zhuang1, Mounir Izallalen2, Paula Mouser2, Hanno Richter2, Carla Risso2, Radhakrishnan Mahadevan1 and Derek R Lovley2

  1. 1Department of Chemical Engineering and Applied Chemistry, Institute of Biomaterials and Biomedical Engineering, University of Toronto, Toronto, Ontario, Canada
  2. 2Department of Microbiology, University of Massachusetts, Amherst, MA, USA

Correspondence: R Mahadevan, Department of Chemical Engineering and Applied Chemistry, Institute of Biomaterials and Biomedical Engineering, University of Toronto, 200 College Street Room 326, Toronto, Canada M5S3E5. E-mail: krishna.mahadevan@utoronto.ca

Received 14 January 2010; Revised 14 June 2010; Accepted 17 June 2010; Published online 29 July 2010.



The advent of rapid complete genome sequencing, and the potential to capture this information in genome-scale metabolic models, provide the possibility of comprehensively modeling microbial community interactions. For example, Rhodoferax and Geobacter species are acetate-oxidizing Fe(III)-reducers that compete in anoxic subsurface environments and this competition may have an influence on the in situ bioremediation of uranium-contaminated groundwater. Therefore, genome-scale models of Geobacter sulfurreducens and Rhodoferax ferrireducens were used to evaluate how Geobacter and Rhodoferax species might compete under diverse conditions found in a uranium-contaminated aquifer in Rifle, CO. The model predicted that at the low rates of acetate flux expected under natural conditions at the site, Rhodoferax will outcompete Geobacter as long as sufficient ammonium is available. The model also predicted that when high concentrations of acetate are added during in situ bioremediation, Geobacter species would predominate, consistent with field-scale observations. This can be attributed to the higher expected growth yields of Rhodoferax and the ability of Geobacter to fix nitrogen. The modeling predicted relative proportions of Geobacter and Rhodoferax in geochemically distinct zones of the Rifle site that were comparable to those that were previously documented with molecular techniques. The model also predicted that under nitrogen fixation, higher carbon and electron fluxes would be diverted toward respiration rather than biomass formation in Geobacter, providing a potential explanation for enhanced in situ U(VI) reduction in low-ammonium zones. These results show that genome-scale modeling can be a useful tool for predicting microbial interactions in subsurface environments and shows promise for designing bioremediation strategies.


Geobacter; Rhodoferax; community modeling; bioremediation; systems microbiology



A wide phylogenetic diversity of microorganisms that are capable of dissimilatory metal reduction has been recovered from subsurface environments (Lovley et al., 2004; Lovley, 2006). The factors controlling which species predominate in a given subsurface environment are poorly understood, but may have important environmental consequences. For example, some dissimilatory metal-reducing microorganisms, such as those from the Geobacteraceae family, are capable of reducing U(VI) to U(IV), which can impact the mobility of uranium in the subsurface (Lovley, 1991, 2001, 2006; Lovley et al., 1993, 2004; Wall and Krumholz, 2006), whereas others do not reduce U(VI) (Lovley et al., 2004; Lovley, 2006). Stimulating dissimilatory metal reduction to promote the reductive precipitation of uranium shows promise as a bioremediation strategy for uranium-contaminated groundwater (Anderson et al., 2003; Vrionis et al., 2005), but relies on stimulating the appropriate dissimilatory metal-reducing microorganisms.

Field studies on in situ uranium bioremediation at a uranium-contaminated site in Rifle, CO, have shown that Rhodoferax and Geobacter species, two phylogenetically distinct groups of dissimilatory metal-reducing microorganisms, are important components of the naturally anoxic subsurface at this site (Mouser et al., 2009). Multiple field experiments at the Rifle site have shown that when dissimilatory metal reduction is artificially stimulated with the addition of acetate to the subsurface, Geobacter consistently become the predominant dissimilatory Fe(III) reducing microorganisms (Holmes et al., 2002, 2007; Anderson et al., 2003; Chang et al., 2005; Wilkins et al., 2009). This is despite the fact that similar to Geobacter species (Lovley et al., 1993, 2004; Caccavo et al., 1994), the one described Fe(III)-reducing Rhodoferax species, Rhodoferax ferrireducens, is also capable of oxidizing acetate with the reduction of Fe(III) (Finneran et al., 2003). In contrast to the Geobacter species, R. ferrireducens has not been shown to be capable of U(VI) reduction, and cannot fix nitrogen in the absence of ammonium (Figure 1b; Finneran et al., 2003). Therefore, the ability of Geobacter species to outcompete Rhodoferax species may be critical to the success of stimulated in situ bioremediation at the Rifle site, and presumably other similar uranium-contaminated environments. The relative abundance of Rhodoferax and Geobacter species in different subsurface locations within the site has been measured and suggests that the competition among these Fe(III) reducers is related, at least in part, to the availability of ammonium in those locations (Figure 1c) (Mouser et al., 2009).

Figure 1.
Figure 1 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Conceptual model of uranium bioremediation. (a) Before acetate injection, acetate is generated in the subsurface primarily through fermentation. During bioremediation, acetate is injected to the subsurface at high concentrations through the injection galleries. The artificial flow of acetate is combined with natural acetate flow, and follows the direction of the ground water. This acetate stimulates the growth of multiple microbial species downstream of the injection galleries, including Geobacter and Rhodoferax species. Groundwater samples are collected periodically at the test wells downstream of the injection galleries, including wells D02, D05 and D08. Various tests were performed on these groundwater samples, including ammonium concentration measurements and 16S rRNA-based analysis of the relative microbial abundance (Mouser et al. 2009). (b) Both Geobacter and Rhodoferax oxidize the dissolved acetate, and carry out the reduction of Fe(III) by attaching to a Fe(III) surface. Geobacter is also capable of reducing U(VI) in its planktonic phase. Given that Fe(III) is the primary electron acceptor for Geobacter, this implies that the organisms compete for both electron donor and acceptor. The reduction of Fe(III) creates dissolved Fe(II), whereas the reduction of dissolved U(VI) creates U(IV), which precipitates. The reductive precipitation of uranium effectively removes uranium from the groundwater. Abbreviations: Rf, Rhodoferax; GsA, Geobacter (attached to sediment); GsP, Geobacter (planktonic). (c) Ammonium concentrations at wells D02, D05 and D08 have been measured periodically (Mouser et al., 2009) Here, the initial and average ammonium concentrations are shown. The Geobacter and Rhodoferax columns show the number of Geobacter and Rhodoferax 16S rRNA genes as a percentage of the total genes in the sample. The Geobacter fraction is calculated using equation (1). The relative abundance of these two organisms seems to be related to the concentration of ammonium.

Full figure and legend (265K)

Constraint-based metabolic models (Fell and Small, 1986; Watson, 1986; Varma and Palsson, 1994) offer the possibility of predicting the physiological responses of microorganisms to a diversity of environmental conditions, as well as the interactions of microorganisms with each other and their environments (Lovley, 2003; Scheibe et al., 2009). Flux balance analysis is a constraint-based modeling technique that is able to predict metabolic fluxes in microorganisms under different growth conditions without information on the kinetic parameters for each individual metabolic reaction (Feist and Palsson, 2008; Oberhardt et al., 2009). Genome-scale flux balance analysis models of pure cultures, such as Escherichia coli (Ibarra et al., 2002; Reed and Palsson, 2003; Reed et al., 2003; Feist et al., 2007; Feist and Palsson, 2008) and Geobacter species (Mahadevan et al., 2006; Segura et al., 2008; Sun et al., 2009), are capable of predicting the detailed physiology including growth yields and respiration rates. This modeling approach has been useful for both understanding the behavior of biological systems in complex environments and for engineering purposes (Burgard et al., 2003; Pharkya et al., 2003, 2004; Pharkya and Maranas, 2006; Hjersted et al., 2007; Anesiadis et al., 2008; Izallalen et al., 2008). The availability of the genome sequences of Geobacter (Methe et al., 2003; Sun et al., 2009) and Rhodoferax (Risso et al., 2009) species provides the opportunity to apply genome-scale modeling techniques to examine the interaction and competition between Geobacter and Rhodoferax species under naturally occurring and artificially stimulated Fe(III)-reducing conditions.

It has been previously shown that it is possible to model the syntrophic growth of two microorganisms with a small-scale metabolic model (Stolyar et al., 2007). However, in that model, the stoichiometric matrices of the two partners, Desulfovibrio vulgaris and Methanococcus maripaludis, were coupled directly and dynamic changes in the biomass concentrations of the individual species were not considered. Although this approach may be appropriate when the microorganisms are interdependent, it is inappropriate in ecological settings where the community composition is dynamic. To predict metabolic behavior of such microbial systems, in which other interactions—such as competition—are important, we have developed the dynamic multi-species metabolic modeling (DMMM) framework based on the dynamic flux balance analysis (Mahadevan et al., 2002). The advantages of the DMMM framework over a Monod kinetics-based model and the model of Stolyar et al. are discussed in more detail in the Supplementary Information.

In this paper, we present a dynamic metabolic model of Geobacter sulfurreducens and R. ferrireducens, implemented using the DMMM framework (Figure 2), and show how this approach can provide important insights into the factors controlling the relative abundance of Geobacter and Rhodoferax species under different subsurface conditions. This is the first model to dynamically integrate multiple genome-scale metabolic models.

Figure 2.
Figure 2 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

DMMM of Geobacter and Rhodoferax community. The in silico representation of a minimal community whose growth depends on the oxidation of acetate coupled to Fe(III) reduction. Here, μ is the biomass growth rate, ViGs is the flux of metabolite, i predicted by the G. sulfurreducens model (Mahadevan et al. 2006), ViRf is the flux of metabolite, i predicted by the R. ferrireducens model (Risso et al. 2009), X is the biomass concentration, [Si] is the concentration of ith metabolite, Ksi is the saturation constant for ith metabolite. The simulations in this paper assume the field to be a spatial homogeneous chemostat to simplify the transport and geochemical process.

Full figure and legend (146K)


Materials and methods

The dynamic multi-species metabolic modeling framework

We have developed the DMMM framework, a computational framework capable of integrating multiple pre-existing genome-scale metabolic models into a community metabolic model. The DMMM framework was used to integrate the previously published models of G. sulfurreducens and R. ferrireducens into a community model of subsurface iron-reducers (Figure 2). The formulation and implementation of the DMMM framework is described in detail in the Supplementary Information.

Modeling resource availability

In anoxic sedimentary environments, complex organic matter is fermented with acetate as the primary fermentation product (Lovley and Phillips, 1989; Lovley and Chapelle, 1995; Lovley et al., 2004). Thus, the slow steady release of acetate from fermentation is expected to be the primary natural source of acetate in the subsurface sediments at the Rifle site.

The rates of microbial processes have been studied in various organics-poor sediments. In a Danish shallow sandy aquifer similar to the Rifle aquifer, the rate of acetate oxidation associated with iron-reduction, sulfate-reduction and methanogenesis was reported to be between 0 and 0.5μMh−1 (Hansen et al., 2001). Similar values have been reported for oligotrophic lake, marine and salt marsh sediments (Crill and Martens, 1986) reported that the acetate turnover rate in a methane-producing marine sediment to be 0.40μMh−1 at 30cm depth. Kuivila et al. (1989) reported that the acetate production rate from fermentation was between 0.035 and 0.17μMh−1. Balba and Nedwell (1982) reported that the acetate utilization rate in salt marsh sediment is in the range of 0–4.8μMh−1. Lovley and Klug (1983) reported that the total rate of microbial metabolism in a lake sediment to be 6.4μMh−1, which provides an upper limit for the acetate turnover in that environment. In a previous sediment model, Lovley and Klug (1986) have assumed 3μMh−1 to be the rate of acetate turnover in oligotrophic sediments. The acetate turnover rate in a deep aquifer is measured to be 0.0135μMh−1 at 35m depth (Chapelle and Lovley, 1990), which is assumed to be the lower bound for our simulations. On the basis of these previous findings, we assumed the acetate flux into the subsurface at Rifle is between 0 to 0.54μMh−1 before the stimulation of metabolism through the addition of acetate.

The ammonium concentrations were set to range from 5 to 400μM, which correspond to the ammonium concentrations previously observed at the Rifle site (Mouser et al., 2009).

Previous sediment sampling at bioremediation sites reported that the Fe(III) concentration to be in the range of 5–40μMolg−1 of sediment (Anderson et al., 2003; Vrionis et al., 2005; Yabusaki et al., 2007; Komlos et al., 2008), which includes both microbially reducible Fe(III) and Fe(III) that is resistant to reduction. If it is assumed that the sediment density is 2gml−1 (Petrie et al., 2003) and that ~50% of Fe(III) is bioavailable, then 2.5–20mmol of Fe(III) per liter of groundwater is available for microbial reduction. Therefore, the initial Fe(III) concentration was assumed to be 10mmol per liter for both pre-injection and during-injection simulations.

The initial cell density for both G. sulfurreducens and R. ferrireducens were assumed to be 105cellsl−1 based on previous results for the Rifle site (Holmes et al., 2007). The average cell mass of G. sulfurreducens is experimentally measured to be 10−12 gcell−1 (Izallalen et al., 2008). On the basis of the volume ratio between G. sulfurreducens and R. ferrireducens, and assuming their intracellular contents have a similar physical density, the average cell mass of R. ferrireducens is calculated to be 6.25 × 10−12gcell−1 (see Supplementary Information). Recent studies have shown that a high proportion of Geobacter species do not firmly attach to subsurface sediments at the Rifle site (Strycharz et al., 2009) and therefore, it was assumed that cells were planktonic and that the biomass was fully affected by dilution due to groundwater flow.

The simulations model the field as a spatially homogeneous chemostat, which simplifies the transport and geochemical process. The groundwater flow in the subsurface is modeled using a dilution rate of 0.034day−1, or 0.0014h−1 (see Supplementary Information for calculation). As this study was not designed to model field-scale events, we focused solely on community competition dynamics. However, efforts are underway to integrate this dynamic community model with a reactive transport model of the Rifle site similar to a previously described method (Scheibe et al., 2009).

The injection of acetate to promote U(VI) reduction at the Rifle site results in acetate concentrations of 3–5mM (Anderson et al., 2003; Vrionis et al., 2005). The acetate injection rate was calculated by multiplying the acetate concentration of 3mM by the dilution rate, resulting in an acetate flux of 0.0042mMh−1. This value was used to simulate the acetate flux during acetate injection. As the measured ammonium concentrations in situ did not change significantly during the course of bioremediation, the ammonium concentration during acetate amendment was fixed at 400μM for the ammonium-excess simulation and fixed at 5μM for the ammonium-limiting simulation. The parameters used in the simulation are summarized in the Supplementary Information.

Treatment of experimental data and model assessment

Data from the 2007 field experiment at Rifle, CO (Mouser et al., 2009) were used to assess the accuracy of the community dynamics predicted by the model. These data include the initial and average ammonium concentrations, as well as the in situ abundance of 16S rRNA of G. sulfurreducens and R. ferrireducens before the start of the acetate injection (day 0) and 9, 14, 18 days after the start of the injection (Figure 1c). To compare the relative abundance of Geobacter and Rhodoferax species mathematically, we defined a metric called the ‘Geobacter fraction’ (equation (1)). A higher Geobacter fraction value indicates a greater abundance of Geobacter species relative to Rhodoferax. Both 16S rRNA abundance (experimentally measured using reverse transcription-quantitative PCR; Mouser et al., 2009) and computationally predicted cell densities were converted to Geobacter fraction value using equation (1).

Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

The prediction accuracy of the model was accessed for both natural conditions and during injection conditions by comparing simulated Geobacter fractions with experimental Geobacter fractions at test wells D02, D05 and D08. For both cases, the ammonium concentrations were set to the measured concentrations before acetate injection at these wells.


Results and Discussion

Community competition under natural conditions

Before addition of acetate to the groundwater, fermentation of complex organic matter is expected to be the primary source of acetate to Rhodoferax and Geobacter species. The rate of this natural supply of acetate at the Rifle site is not known, but is expected to fall within the range of acetate turnover rates that have been observed in various sedimentary environments (Crill and Martens, 1986; Lovley and Klug, 1986; Kuivila et al., 1989; Chapelle and Lovley, 1990; Hansen et al., 2001). The steady-state Geobacter fractions generated using low acetate turnover rates of 0.2, 0.25 and 0.38μMh−1 were compared with the Geobacter fraction values calculated from experimental measurement at day 0 (Figure 3). This comparison suggests that the acetate turnover rates are close to 0.2μMh−1 at well D02, 0.38μMh−1 at well D05 and 0.25μMh−1 at well D08 (Figure 3). All three inferred acetate turnover rates fall within the range measured in similar environments (Figure 3; Chapelle and Lovley, 1990; Hansen et al., 2001). These results also suggest that this model could be used to predict the community dynamics under natural conditions more accurately if measurements of the acetate turnover rates at the three wells before injection were available.

Figure 3.
Figure 3 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Comparison between predicted and measured Geobacter fractions under natural conditions. Predicted Geobacter fractions at D02, D05 and D08 before the acetate addition are generated using acetate turnover rates of 0.2, 0.25 and 0.38μMh−1. The predictions are compared with the Geobacter fraction values calculated from experimental 16S rRNA measurements at the respective wells, at day 0 (Mouser et al. 2009). This comparison suggests that the acetate turnover rates are close to 0.2μMh−1 at well D02, close to 0.38μMh−1 at well D05, and close to 0.25μMh−1 at well D08. All three inferred acetate turnover rates fall within the range measured in similar environments, which supports the predicted competition dynamics at natural conditions.

Full figure and legend (55K)

The community competition was simulated over three different (low, medium and high) acetate turnover ranges (Figure 4). The simulations suggest that at low acetate turnover rates, Rhodoferax species are likely to be more abundant than Geobacter species, especially when ammonium is in excess. This prediction is consistent with the results of the analysis of 16S rRNA gene sequences at the Rifle site before the addition of acetate (Mouser et al., 2009).

Figure 4.
Figure 4 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Relative composition of Geobacter and Rhodoferax in Fe(III) reducing community before acetate amendment. The steady-state community compositions at three different acetate turnover rate ranges are simulated. The Geobacter fraction is used to measure the relative success of Geobacter to Rhodoferax species. The competition with respect to the ammonium concentration and acetate turnover rate is viewed at three different scales with respect to the acetate turnover rate. (a) The acetate turnover rates range from 0 to 0.2μMh−1. At this scale, the nonlinearity of the competition is highlighted. (b) The acetate turnover rates range from 0 to 0.54μMh−1, corresponding to the range of subsurface acetate availability reported by Hansen et al. (2001). (c) The acetate turnover rates range from 0 to 0.1μMh−1, corresponding to the rates measured in an aquifer extremely poor in acetate (Chapelle and Lovley, 1990).

Full figure and legend (223K)

Risso et al., 2009 analyzed the energetics of R. ferrireducens by comparing predicted yields with experimental yields under multiple conditions, and found that the proton translocation stoichiometry at cytochrome reductase of 2H+/2e was consistent with the data for R. ferrireducens, whereas the stoichiometry at cytochrome reductase that could explain experimental data was found to be 1H+/2e for G. sulfurreducens (Mahadevan et al., 2006). The doubling of cytochrome reductase energetic efficiency led to a significant increase in the efficiency of the electron transport chain and the predicted Rhodoferax biomass yield during Fe(III) respiration. Model simulations show that under acetate-limited growth on Fe(III), the yield of R. ferrireducens is 0.0798gDWmmol−1 of acetate, nearly two times the yield of G. sulfurreducens (0.0437gDWmmol−1 of acetate). Thermodynamic analysis of cellular growth shows that the free energy of substrate oxidation can either be used to drive metabolic reactions at higher rates or to produce biomass; in other words, an organism can either optimize for substrate uptake rate or optimize for energetic efficiency (Pfeiffer et al., 2001; Pfeiffer and Bonhoeffer, 2003, 2004; Pfeiffer and Schuster, 2005; Schuster et al., 2008). Therefore, by choosing to optimize for efficiency, Rhodoferax has a higher biomass yield at the expense of growth rate, resulting in a significantly lower substrate uptake rate compared with Geobacter. This is exemplified by the observation that the maximum acetate uptake rate of G. sulfurreducens (18mmolgDW−1h−1) is more than 10-fold higher than that of R. ferrireducens (1.71mmolgDW−1h−1).

It was previously shown that yield strategists are favored over rate strategists under low substrate flux conditions (Pfeiffer et al., 2001). The same rationale can be applied to the anoxic Fe(III)-reducing microbial community. The growth rate is the product of the substrate uptake rate and yield (μ=Vs × Y). At low acetate turnover conditions, the low environmental acetate flux limits the uptake rate of Geobacter to a fraction of its maximum. If both organisms have similar uptake rates, then the growth rate of Rhodoferax will be significantly higher, as its yield is twice that of Geobacter. This explains the abundance of Rhodoferax at low acetate turnover conditions when ammonium is readily available (Figure 4c). As acetate turnover rate increases, the acetate no longer limits growth. When the acetate turnover rate is sufficiently high to allow Geobacter uptake rates to become more than two times that of Rhodoferax, the growth rate of Geobacter becomes greater than that of Rhodoferax. Therefore, Geobacter becomes more competitive at higher acetate fluxes (Figures 4a and b).

The modeling suggests that the availability of ammonium in the sediments also has a major role in the relative abundance of Geobacter and Rhodoferax under natural conditions in the sediments, with Geobacter favored at low ammonium conditions (Figures 3 and 4). The community composition varied nonlinearly with respect to ammonium concentration and acetate turnover rate (Figure 4a). The region of coexistence (Geobacter fraction between 0.4 and 0.6) of Rhodoferax and Geobacter under steady-state conditions was narrow; minute changes in acetate and ammonium fluxes led to the complete dominance of one organism (Figures 4b and c). The simulations predicted that as the availability of ammonium decreased, Geobacter would become more competitive with Rhodoferax at lower acetate turnover rates (Figure 4). This can be attributed to the ability of Geobacter to fulfill its nitrogen requirements from fixation of nitrogen, whereas Rhodoferax is incapable of nitrogen fixation. Availability of acetate in excess, results in the limitation in ammonium for Rhodoferax, favoring the growth of Geobacter. A high expression of genes for nitrogen fixation in Geobacter species living in a diversity of subsurface sediments (Holmes et al., 2004, 2007; Mouser et al., 2009), has suggested that they are actively involved in nitrogen fixation in the subsurface. Geobacter nitrogen fixation genes may even be expressed when low (<50μM) concentrations are available in sediments (Holmes et al., 2009), suggesting that such subsurface sediments may well contain microzones that are ammonium depleted. The presence of such microzones would further facilitate the coexistence of Geobacter and Rhodoferax, with growth of Geobacter favored in the ammonium-depleted microzones. However, the current model does not incorporate spatial heterogeneity and therefore cannot predict such spatial variations in community composition.

The insights derived from our simulations can be applied to understand the competition between Geobacter and Rhodoferax in other environments. For example, Geobacter is found to dominate in a landfill leachate-contaminated Dutch aquifer with high ammonium content (Lin et al., 2007). In this case, the leachate contamination contains 50–100mgl−1 of dissolved organic carbon (Lin et al., 2007), which is much higher than the natural dissolved organic carbon content at the Rifle site (<5mgl−1; Barlett, Personal Communication). Therefore, the leachate contamination itself provides a rich organic nutrient source, presumably leading to an increased rate of acetate production through fermentation in the Dutch aquifer relative to that of Rifle. The increase in the acetate production rate favors the rate optimizing Geobacter species, leading to their dominance.

Community competition when acetate is added to the subsurface

Our simulation predictions of the community composition during the acetate addition at wells D02, D05 and D08 show the same trend as the values calculated from the 16S rRNA measurements (Figure 5). These simulations (Figure 5) agree with previous observations that the addition of millimolar concentrations of acetate to the subsurface to stimulate in situ uranium bioremediation at the Rifle site significantly influences the relative proportions of Geobacter and Rhodoferax, with Geobacter species consistently becoming the predominant microorganisms regardless of ammonium availability (Anderson et al., 2003; Vrionis et al., 2005; Holmes et al., 2007; Mouser et al., 2009). Furthermore, the predicted number of Geobacter cells is within an order of magnitude relative to the measured number of Geobacter cells at day 19 (Figure 6; Holmes et al., 2007).

Figure 5.
Figure 5 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Comparison between predicted and measured Geobacter fractions during acetate addition. The predicted Geobacter fractions at wells D02, D05 and D08 are compared with the experimentally measured fractions at the respective wells. Simulations are initiated with the Geobacter fractions and ammonium concentrations measured at day 0 in wells D02 (58%, 40μM), D05 (64%, 21μM) and D08 (20%, 400μM) (Mouser et al. 2009). The predicted Geobacter fractions (solid lines) are compared with the Geobacter fractions calculated from the experimental 16S rRNA measurements ().

Full figure and legend (89K)

Figure 6.
Figure 6 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Simulation of the competition dynamics in a Fe(III)-reducing community during acetate addition. Biomass concentrations of G. sulfurreducens and R. ferrireducens (a), Geobacter fractions (b), acetate concentration (c) and Fe(III) concentration (d) under both ammonium-limiting and ammonium-excess conditions are shown. Under both conditions, Geobacter outcompetes Rhodoferax soon after acetate addition begins. The blue lines are the ammonium-excess simulation results; the red lines are the ammonium-limiting simulation results. The black bar in (a) represents the experimentally measured range of the Geobacter cell density at day 19 (Holmes et al., 2007). Simulations are initialized with equal concentrations of Geobacter and Rhodoferax. The acetate injection rate of 3mMday−1 (4.2μMh−1) is used for both ammonium-limiting ((ammonium)=0.005mM) and ammonium-excess conditions ((ammonium)=0.4mM). Note: the y axis of (a) is in log scale, whereas the y axis of (b, c and d) are in linear scale.

Full figure and legend (129K)

The reasons for this were readily apparent from the dynamic genome-scale modeling (Figure 6). Here, the acetate injection rate was calculated to be 4.2μMh−1 based on the assumed geometry (Anderson et al., 2003; Vrionis et al., 2005). At this acetate flux, the ammonium flux in the ammonium excess case (estimated by multiplying the dilution rate and the steady ammonium concentration to be 0.56μMh−1) is more than sufficient for biomass synthesis and is not limiting. In the ammonium limiting case, the ammonium flux (0.035μMh−1) is a primary limitation of growth at the same level of acetate availability. Geobacter rapidly responded to increased acetate availability with a substantial increase in biomass, whether or not sufficient ammonium was available to completely support all of the acetate-dependent growth (Figures 6a and b). In contrast, the growth rate of R. ferrireducens was just above the dilution rate in the ammonium excess case, whereas there is a loss in Rhodoferax biomass in the ammonium limiting case. This reflects the absolute requirement (Figures 6a and b) of ammonium to support the growth of Rhodoferax. Low ammonium delayed, but did not prevent extensive growth of Geobacter. However, the maximum amount of Geobacter biomass was lower under ammonium-limiting conditions because of the increased energetic demands for nitrogen fixation (Figure 6a). With extended time, Fe(III) was depleted and Geobacter declined (Figures 6a and c). The acetate concentration is briefly reduced to zero before rising again after the occurrence of Fe(III) depletion under both conditions (Figure 6d).

The high flux of acetate from the artificial addition allows both species to achieve their maximum acetate uptake rate. As the maximum acetate uptake rate of Geobacter is more than 10 times higher than that of Rhodoferax, whereas the yield of Rhodoferax is only twice higher than that of Geobacter, the growth rate of Geobacter is significantly higher than that of Rhodoferax when both organisms use acetate at their respective maximum uptake rates. Therefore, during bioremediation, Geobacter's advantage in uptake rate outweighs Rhodoferax's advantage in yield, leading to the overwhelming success of Geobacter.

The cell densities and the Fe(III) concentration before acetate injection can vary greatly in the field before acetate amendment. To evaluate the sensitivity of the competition dynamics to variations in the initial cell densities and Fe(III) concentrations, simulations with different initial Fe(III) concentrations (2.5, 5, 10 and 20mmoll−1) and cell densities (104, 105 and 106cellsl−1) were performed. These simulations showed no significant changes in the dynamics of the competition, suggesting that the variations in these two parameters would not affect the increase in the Geobacter fraction following acetate amendment (data not shown). Despite the lack of spatial and geochemical descriptions in our model, we were able to predict the trend changes in the community composition; this suggests that biological interactions are perhaps the dominant factor for influencing the microbial ecology in the Rifle subsurface.

Predicted changes in metabolic states and its implications for bioremediation

An additional feature of the genome-scale metabolic models, relative to models based on Monod kinetics with constant yield parameters, is their capability to predict changes in the metabolic states of microorganisms under different environmental conditions. For organisms with few physiological states, traditional models, which depend on experimentally measured yields for each state, are sufficient. As the number of physiological states increase, the complexity of implementing a Monod kinetics-based model increases as additional data for the different physiological states are required and each state has to be addressed individually (see Supplementary Information). When acetate is the donor, the metabolism of Rhodoferax is relatively simple and can be modeled with traditional methods, but Geobacter has many physiological states, such as acetate-limiting ammonium utilization mode, iron-limiting ammonium utilization mode, acetate-limiting nitrogen fixation mode and iron-limiting nitrogen fixation mode, which require constraint-based models to predict. The genome-scale model predicted that the biomass yields of Geobacter can vary between 1.95 and 4.05gDW per mol acetate (Scheibe et al., 2009). Under ammonium-depleted conditions, R. ferrireducens is predicted to experience a maintenance phase with no associated growth because of the lack of ammonium, but with the ability to generate ATP by oxidizing acetate coupled to Fe(III) reduction. In contrast, G. sulfurreducens is able to grow in the absence of ammonium by nitrogen fixation.

The necessity for nitrogen fixation is predicted to have a number of environmentally relevant physiological consequences for Geobacter (Figure 7). The genome-scale model predicts that to meet increased energetic demands associated with nitrogen fixation, more acetate enters the tricarboxylic acid cycle, resulting in higher fluxes through NADH dehydrogenase and extracellular electron transfer, and ATP synthase (Figure 7). Under nitrogen-fixing conditions, less biomass is produced; this is reflected in the increased acetate activation through acetyl-CoA transferase instead of acetate kinase, as well as less pyruvate production through pyruvate oxidoreductase for biomass synthesis. These changes in metabolic fluxes are predicted to result in a 30% reduction in growth yield (Figure 7 and Supplementary Figure S8 in Supplementary Information). A reduction of Geobacter biomass yield under nitrogen-fixation conditions has been observed experimentally (Methe et al., 2005).

Figure 7.
Figure 7 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Comparison of the predicted flux distribution during growth with ammonium vs nitrogen-fixation dependent growth of G. sulfurreducens. The predicted flux distributions of G. sulfurreducens during growth with unlimited ammonium uptake and nitrogen-fixation dependent growth are compared. The biomass flux is measured in h−1; all other fluxes are measured in mmolgDW−1h−1. The red fluxes are increased during nitrogen fixation; the blue fluxes are decreased during nitrogen fixation. The first number represents the flux through the reaction when ammonium is acquired through environmental uptake; the second number represents the flux through the reaction when ammonium is provided through nitrogen fixation.

Full figure and legend (98K)

The predicted shift in metabolism associated with nitrogen fixation could have a significant impact on the effectiveness of in situ uranium bioremediation. The rate of U(VI) reduction depends on both the specific uranium reduction rate and the biomass concentration. During nitrogen fixation, the rate of U(VI) reduction per cell may increase because of the up-regulation of respiration, at the expense of biomass production. This prediction of the model is consistent with field results at the Rifle site. In the 2007 field experiment, U(VI) was much more effectively removed from the groundwater at sites low in ammonium than at sites with relatively high-ammonium concentrations (Mouser et al., 2009). This trade-off between the specific uranium reduction rate and the biomass yield can be evaluated using the DMMM framework, and it may be possible with dynamic optimization techniques to predict from genome-scale metabolic models the optimal ammonium concentrations for maximal rates of U(VI) reduction. This would be analogous to the recent design of a G. sulfurreducens strain with increased respiration rates and lower biomass yields using genome-scale metabolic modeling (Izallalen et al., 2008).


The simulations suggest that Geobacter and Rhodoferax species have adopted very different strategies for growth in subsurface environments. Geobacter sacrificed maximizing yield from substrate utilization in favor of rapid growth and the ability to grow in the absence of ammonium, whereas Rhodoferax are more optimized for higher growth yields. Both strategies seem to be adaptive for growth under natural subsurface conditions at the Rifle site. Geobacter and Rhodoferax are found within the same sampling zones. This probably reflects the growth of Geobacter and Rhodoferax species in different microenvironments within the heterogeneous subsurface environment. The ability of Geobacter to grow faster than Rhodoferax at high acetate concentrations and to multiply in the absence of ammonium permits it to rapidly respond to the artificial conditions imposed when acetate is added to promote in situ uranium bioremediation. It is fortuitous that Geobacter outcompete Rhodoferax under these conditions, because pure culture studies have suggested that Geobacter (Lovley, 1991; Caccavo et al., 1994), but not Rhodoferax (Finneran et al., 2003) species are capable of U(VI) reduction. If this distinction were true for all Rhodoferax species, then in situ uranium reduction would be much less effective if Rhodoferax could readily compete with Geobacter under high acetate/low ammonium conditions.

There are other microbial interactions that are likely to impact the effectiveness of in situ uranium bioremediation. Most notably, the consumption of added acetate by acetate-oxidizing sulfate reducers that are ineffective in U(VI) reduction (Anderson et al., 2003) may limit U(VI) reduction by Geobacter species, decreasing the efficacy of the bioremediation (Anderson et al., 2003; Vrionis et al., 2005). As genome-scale models for these sulfate reducers become available, it should be possible to evaluate this interaction in a manner similar to that reported here for Geobacter/Rhodoferax interactions. Furthermore, the increasing availability of the genome sequences of environmentally relevant microorganisms should make it feasible to apply genome-scale metabolic modeling to further investigate the ecology of a wide diversity of microbial ecosystems.



  1. Anderson RT, Vrionis HA, Ortiz-Bernad I, Resch CT, Long PE, Dayvault R et al. (2003). Stimulating the In Situ activity of Geobacter species to remove uranium from the groundwater of a uranium-contaminated aquifer. Appl Environ Microbiol 69: 5884–5891. | Article | PubMed | ISI | ChemPort |
  2. Anesiadis N, Cluett WR, Mahadevan R. (2008). Dynamic metabolic engineering for increasing bioprocess productivity. Metab Eng 10: 255–266. | Article | PubMed | ChemPort |
  3. Balba MT, Nedwell DB. (1982). Microbial metabolism of acetate, propionate and butyrate in anoxic sediment from the Colne Point Saltmarsh, Essex, UK. J Gen Microbiol 128: 1415–1422. | ChemPort |
  4. Burgard AP, Pharkya P, Maranas CD. (2003). Optknock: a bilevel programming framework for identifying gene knockout strategies for microbial strain optimization. Biotechnol Bioeng 84: 647–657. | Article | PubMed | ISI | ChemPort |
  5. Caccavo F, Lonergan DJ, Lovley DR, Davis M, Stolz JF, McInerney MJ. (1994). Geobacter sulfurreducens sp. nov., a hydrogen- and acetate-oxidizing dissimilatory metal-reducing microorganism. Appl Environ Microbiol 60: 3752–3759. | PubMed | ISI | ChemPort |
  6. Chang YJ, Long PE, Geyer R, Peacock AD, Resch CT, Sublette K et al. (2005). Microbial incorporation of 13C-labeled acetate at the field scale: detection of microbes responsible for reduction of U(VI). Environ Sci Technol 39: 9039–9048. | Article | PubMed | ISI | ChemPort |
  7. Chapelle FH, Lovley DR. (1990). Rates of microbial metabolism in deep coastal plain aquifers. Appl Environ Microbiol 56: 1865–1874. | PubMed | ISI | ChemPort |
  8. Crill PM, Martens CS. (1986). Methane production from bicarbonate and acetate in an anoxic marine sediment. Geochim Cosmochim Acta 50: 2089–2097. | Article | ChemPort |
  9. Feist AM, Henry CS, Reed JL, Krummenacker M. (2007). A genome-scale metabolic reconstruction for Escherichia coli K-12 MG1655 that accounts for 1260 ORFs and thermodynamic information. Mol Syst Biol 3: 121. | Article | PubMed | ChemPort |
  10. Feist AM, Palsson B. (2008). The growing scope of applications of genome-scale metabolic reconstructions using Escherichia coli. Nat Biotechnol 26: 659–667. | Article | PubMed | ChemPort |
  11. Fell DA, Small JR. (1986). Fat synthesis in adipose tissue. An examination of stoichiometric constraints. Biochem J 238: 781–786. | PubMed | ISI | ChemPort |
  12. Finneran KT, Johnsen CV, Lovley DR. (2003). Rhodoferax ferrireducens sp. nov., a psychrotolerant, facultatively anaerobic bacterium that oxidizes acetate with the reduction of Fe(III). Int J Syst Evol Microbiol 53: 669–673. | Article | PubMed | ISI | ChemPort |
  13. Hansen LK, Jakobsen R, Postma D. (2001). Methanogenesis in a shallow sandy aquifer, Rømø, Denmark. Geochimica et Cosmochimica Acta 65: 2925–2935. | Article | ChemPort |
  14. Hjersted JL, Henson MA, Mahadevan R. (2007). Genome-scale analysis of Saccharomyces cerevisiae metabolism and ethanol production in fed-batch culture. Biotechnol Bioeng 97: 1190–1204. | Article | PubMed | ChemPort |
  15. Holmes DE, Bond DR, O’Neil RA, Reimers CE, Tender LR, Lovley DR. (2004). Microbial communities associated with electrodes harvesting electricity from a variety of aquatic sediments. Microb Ecol 48: 178–190. | Article | PubMed | ChemPort |
  16. Holmes DE, Finneran KT, O’Neil RA, Lovley DR. (2002). Enrichment of members of the family Geobacteraceae associated with stimulation of dissimilatory metal reduction in uranium-contaminated aquifer sediments. Appl Environ Microbiol 68: 2300–2306. | Article | PubMed | ISI | ChemPort |
  17. Holmes DE, O’Neil RA, Chavan MA, N’Guessan LA, Vrionis HA, Perpetua LA et al. (2009). Transcriptome of Geobacter uraniireducens growing in uranium-contaminated subsurface sediments. ISME J 3: 216–230. | Article | PubMed | ChemPort |
  18. Holmes DE, O’Neil R, Vrionis HA, N’guessan L, Ortiz-Bernad I, Larrahondo MJ et al. (2007). Subsurface clade of Geobacteraceae that predominates in a diversity of Fe(III)-reducing subsurface environments. ISME J 1: 663–677. | Article | PubMed | ChemPort |
  19. Ibarra RU, Edwards JS, Palsson BO. (2002). Escherichia coli K-12 undergoes adaptive evolution to achieve in silico predicted optimal growth. Nature 420: 186–189. | Article | PubMed | ISI | ChemPort |
  20. Izallalen M, Mahadevan R, Burgard A, Postier B, Didonato R, Sun J et al. (2008). Geobacter sulfurreducens strain engineered for increased rates of respiration. Metab Eng 10: 267–275. | Article | PubMed | ChemPort |
  21. Komlos J, Peacock A, Kukkadapu RK, Jaffé PR. (2008). Long-term dynamics of uranium reduction/reoxidation under low sulfate conditions. Geochim Cosmochim Acta 72: 3603–3615. | Article | ChemPort |
  22. Kuivila KM, Murray JW, Devol AH, Novelli PC. (1989). Methane production, sulfate reduction and competition for substrates in the sediments of Lake Washington. Geochim Cosmochim Acta 53: 409–416. | Article | ChemPort |
  23. Lin B, Braster M, Röling WFM, van Breukelen BM. (2007). Iron-Reducing Microorganisms in a Landfill Leachate-Polluted Aquifer: Complementing Culture-Independent Information with Enrichments and Isolations. Geomicrobiol J 24: 283–294. | Article | ChemPort |
  24. Lovley DR. (1991). Dissimilatory Fe(III) and Mn(IV) reduction. Microbiol Rev 55: 259–287. | PubMed | ISI | ChemPort |
  25. Lovley DR. (2001). Bioremediation. Anaerobes to the rescue Science 293: 1444–1446. | ChemPort |
  26. Lovley DR. (2003). Cleaning up with genomics: applying molecular biology to bioremediation. Nat Rev Microbiol 1: 35–44. | Article | PubMed | ISI | ChemPort |
  27. Lovley DR. (2006). Dissimilatory Fe(III)- and Mn(IV)-Reducing Prokaryotes. In: (eds). The Prokaryotes. Springer: New York, pp 1143.
  28. Lovley DR, Chapelle FH. (1995). Deep subsurface microbial processes. Rev Geophys 33: 365–381. | Article | ISI
  29. Lovley DR, Giovannoni SJ, White DC, Champine JE, Phillips EJ, Gorby YA et al. (1993). Geobacter metallireducens gen. nov. sp. nov., a microorganism capable of coupling the complete oxidation of organic compounds to the reduction of iron and other metals. Arch Microbiol 159: 336–344. | Article | PubMed | ISI | ChemPort |
  30. Lovley DR, Holmes DE, Nevin KP. (2004). Dissimilatory Fe(III) and Mn(IV) reduction. Adv Microb Physiol 49: 219–286. | Article | PubMed | ISI | ChemPort |
  31. Lovley DR, Klug MJ. (1983). Sulfate reducers can outcompete methanogens at freshwater sulfate concentrations. Appl Environ Microbiol 45: 187–192. | PubMed | ISI | ChemPort |
  32. Lovley DR, Klug MJ. (1986). Model for the distribution of sulfate reduction and methanogenesis in freshwater sediments. Geochim Cosmochim Acta 50: 11–18. | Article | ChemPort |
  33. Lovley DR, Phillips EJ. (1989). Requirement for a microbial consortium to completely oxidize glucose in Fe(III)-reducing sediments. Appl Environ Microbiol 55: 3234–3236. | PubMed | ChemPort |
  34. Mahadevan R, Bond DR, Butler JE, Esteve-Nuñez A, Coppi MV, Palsson BO et al. (2006). Characterization of metabolism in the Fe(III)-reducing organism Geobactersulfurreducens by constraint-based modeling. Appl Environ Microbiol 72: 1558–1568. | Article | PubMed | ISI | ChemPort |
  35. Mahadevan R, Edwards JS, Doyle FJ. (2002). Dynamic flux balance analysis of diauxic growth in Escherichia coli. Biophys J 83: 1331–1340. | Article | PubMed | ISI | ChemPort |
  36. Methe BA, Nelson KE, Eisen JA, Paulsen IT. (2003). Genome of Geobactersulfurreducens: metal reduction in subsurface environments. Science 302: 1967–1969. | Article | PubMed | ISI | ChemPort |
  37. Methe BA, Webster J, Nevin K, Butler J, Lovley DR. (2005). DNA microarray analysis of nitrogen fixation and Fe(III) reduction in Geobacter sulfurreducens. Appl Environ Microbiol 71: 2530–2538. | Article | PubMed | ChemPort |
  38. Mouser PJ, N’Guessan AL, Elifantz H, Holmes DE, Williams KH, Wilkins MJ et al. (2009). Influence of heterogeneous ammonium availability on bacterial community structure and the expression of nitrogen fixation and ammonium transporter genes during in situ bioremediation of uranium-contaminated groundwater. Environ Sci Technol 43: 4386–4392. | Article | PubMed | ChemPort |
  39. Oberhardt MA, Palsson BO, Papin JA. (2009). Applications of genome-scale metabolic reconstructions. Mol Syst Biol 5: 320. | Article | PubMed
  40. Petrie L, North NN, Dollhopf SL, Balkwill DL, Kostka JE. (2003). Enumeration and characterization of iron(III)-reducing microbial communities from acidic subsurface sediments contaminated with uranium(VI). Appl Environ Microbiol 69: 7467–7479. | Article | PubMed | ISI | ChemPort |
  41. Pfeiffer T, Bonhoeffer S. (2003). An evolutionary scenario for the transition to undifferentiated multicellularity. Proceedings of the National Academy of Sciences 100: 1095–1098. | Article | ChemPort |
  42. Pfeiffer T, Bonhoeffer S. (2004). Evolution of cross—feeding in microbial populations. Am Nat 163: E126–E135. | Article | PubMed
  43. Pfeiffer T, Schuster S. (2005). Game-theoretical approaches to studying the evolution of biochemical systems. Trends Biochem Sci 30: 20–25. | Article | PubMed | ISI | ChemPort |
  44. Pfeiffer T, Schuster S, Bonhoeffer S. (2001). Cooperation and competition in the evolution of ATP-producing pathways. Science 292: 504–507. | Article | PubMed | ISI | ChemPort |
  45. Pharkya P, Burgard AP, Maranas CD. (2003). Exploring the overproduction of amino acids using the bilevel optimization framework OptKnock. Biotechnol Bioeng 84: 887–899. | Article | PubMed | ISI | ChemPort |
  46. Pharkya P, Burgard AP, Maranas CD. (2004). OptStrain: a computational framework for redesign of microbial production systems. Genome Res 14: 2367–2376. | Article | PubMed | ChemPort |
  47. Pharkya P, Maranas CD. (2006). An optimization framework for identifying reaction activation/inhibition or elimination candidates for overproduction in microbial systems. Metab Eng 8: 1–13. | Article | PubMed | ChemPort |
  48. Reed JL, Palsson B. (2003). Thirteen years of building constraint-based in silico models of Escherichiacoli. J Bacteriol 185: 2692–2699. | Article | PubMed | ISI | ChemPort |
  49. Reed JL, Vo TD, Schilling CH, Palsson B. (2003). An expanded genome-scale model of Escherichiacoli K-12 (iJR904 GSM/GPR). Genome Biol 4: R54. | Article | PubMed
  50. Risso C, Sun J, Zhuang K, Mahadevan R, DeBoy R, Ismail W et al. (2009). Genome-scale comparison and constraint-based metabolic reconstruction of the facultative anaerobic Fe(III)-reducer Rhodoferaxferrireducens. BMC Genomics 10: 447. | Article | PubMed | ChemPort |
  51. Scheibe TD, Mahadevan R, Fang Y, Garg S. (2009). Coupling a genome-scale metabolic model with a reactive transport model to describe in situ uranium bioremediation. Microb Biotechnol 2: 274–286. | Article | ChemPort |
  52. Schuster S, Pfeiffer T, Fell DA. (2008). Is maximization of molar yield in metabolic networks favoured by evolution? J Theor Biol 252: 497–504. | Article | PubMed | ChemPort |
  53. Segura D, Mahadevan R, Juárez K, Lovley DR. (2008). Computational and experimental analysis of redundancy in the central metabolism of Geobacter sulfurreducens. PLoS Comput Biol 4: e36. | Article | PubMed | ChemPort |
  54. Stolyar S, Van Dien S, Hillesland KL, Pinel N, Lie TJ, Leigh JA et al. (2007). Metabolic modeling of a mutualistic microbial community. Mol Syst Biol 3: 92. | Article | PubMed | ChemPort |
  55. Strycharz A, Peacock A, Williams KH, N’Guessan L, Lovley DR. (2009). Geobacter species are predominately planktonic during growth in uranium-contaminated subsurface sediments. American Society for Microbiolgy 109th General Meeting 14: 447.
  56. Sun J, Sayyar B, Butler JE, Pharkya P, Fahland TR, Famili I et al. (2009). Genome-scale constraint-based modeling of Geobacter metallireducens. BMC Syst Biol 3: 15. | Article | PubMed | ChemPort |
  57. Varma A, Palsson BO. (1994). Stoichiometric flux balance models quantitatively predict growth and metabolic by-product secretion in wild-type Escherichia coli W3110. Appl Environ Microbiol 60: 3724–3731. | PubMed | ISI | ChemPort |
  58. Vrionis HA, Anderson RT, Ortiz-Bernad I, Ortiz-Bernad I, O’Neil KR, Resch CT et al. (2005). Microbiological and geochemical heterogeneity in an in situ uranium bioremediation field site. Appl Environ Microbiol 71: 6308–6318. | Article | PubMed | ISI | ChemPort |
  59. Wall JD, Krumholz LR. (2006). Uranium reduction. Annu Rev Microbiol 60: 149–166. | Article | PubMed | ChemPort |
  60. Watson MR. (1986). A discrete model of bacterial metabolism. Comput Appl Biosci 2: 23–27. | PubMed | ChemPort |
  61. Wilkins MJ, Verberkmoes NC, Williams KH, Callister SJ, Mouser PJ, Elifantz H et al. (2009). Proteogenomic monitoring of Geobacter physiology during stimulated uranium bioremediation. Appl Environ Microbiol 75: 6591–6599. | Article | PubMed | ChemPort |
  62. Yabusaki SB, Fang Y, Long PE, Resch CT, Peacock AD, Komlos J et al. (2007). Uranium removal from groundwater via in situ biostimulation: field-scale modeling of transport and biological processes. J Contam Hydrol 93: 216–235. | Article | PubMed | ChemPort |



Mathematical symbols

Table A1



This research was supported by the Office of Science (BER), the US Department of Energy, Cooperative Agreement No. DE-FC02-02ER63446 and Grant No. DE-FG02-07ER64367; this research was also supported by the Canada Foundation for Innovation and the University of Toronto Open Fellowship, as well as the Government of Canada through Genome Canada and the Ontario Genomics Institute (2009-OGI-ABC-1405). We thank Caitlin O’Connel for her artistic illustration, Laurence Yang for his insightful discussions and Emma Janssen for her grammatical corrections.

Supplementary Information accompanies the paper on The ISME Journal website