Identification and closure of network gaps.
The initial database was used to construct a preliminary metabolic model for G. sulfurreducens
, and simulations were performed to determine whether this collection of independent reactions could synthesize a full complement of DNA, RNA, amino acids, lipids, carbohydrates, and cofactors from acetate and a defined mineral medium containing vitamins (29
). These simulations revealed gaps in the network that required additional evaluation. For example, G. sulfurreducens
is typically cultured with ammonia as the sole nitrogen source, and yet the network was unable to synthesize lysine, serine, alanine, and threonine. Synthesis of lysine by the network necessitated the incorporation of two non-gene-associated reactions (tetrahydropicolinate succinylase and succinyl-diaminopimelate desuccinylase) to allow production of the key metabolite meso-2,6-diaminopimelate from 2,3,4,5-tetrahydrodipicolinate.
In order to close gaps detected in the network and enable synthesis of all growth precursors, a total of 55 non-gene-associated reactions were added. These included reactions to complete menaquinone and fatty acid biosynthesis and allow exchange of diffusible metabolites and gases (e.g., diffusion of H2
, and CO2
), as well as those required for amino acid biosynthesis. A list of these non-gene-associated reactions is presented in Table S3 in the supplemental material. However, even after the addition of these reactions, there were network gaps in poorly characterized biochemical pathways. Reactions in such pathways (denoted as blocked reactions) were predicted to be inactive in all model simulations even though they were incorporated into the model (10
Previous analysis of the G. sulfurreducens
genome suggested the potential for ammonia oxidation and autotrophic growth on one-carbon compounds (62
). However, the pathways for these two types of metabolism appeared to be incomplete. Furthermore, there is no physiological evidence that G. sulfurreducens
can grow with NH4
as an electron donor or grow in the absence of an additional carbon source if either formate or H2
is provided as the electron donor (12
). Therefore, reactions that would lead to completion of the pathways for autotrophic growth and ammonia oxidation were not incorporated into the model.
The completed reconstructed metabolic network of G. sulfurreducens contained 588 genes (or 17% of a total of 3,467 ORFs), 522 biochemical reactions, and 541 unique metabolites. A full list of all gene-associated reactions is provided in Table S1 in the supplemental material.
Evaluation of the proton translocation stoichiometry during fumarate reduction.
Once a network capable of synthesizing biomass in minimal medium under standard conditions was constructed, unique aspects of Geobacter
energy metabolism and physiology were integrated. The reactions in the model included explicit details of proton production and the movement of protons between the cytoplasmic and extracellular compartments. Hence, the model provided a unique platform for evaluating mechanisms of energy generation. This characteristic was especially advantageous for studying Geobacter
species, which are completely dependent upon electrogenic electron transport for ATP production and cannot generate ATP from acetate via substrate-level phosphorylation (58
Fumarate reduction was selected as the first energy generation mode to be examined with the metabolic model. Extensive literature is available regarding microbial fumarate respiration, and fumarate, in contrast to extracellular electron acceptors typically exploited by Geobacter
species, is an intracellular electron acceptor involving a relatively simple electron transport chain (18
). In addition, experimental observations of growth in the presence of fumarate indicate that G. sulfurreducens
can neither oxidize fumarate nor utilize fumarate as a carbon or energy source, even when grown in the presence of noncarbon electron donors such as hydrogen (12
Genetic studies indicate that G. sulfurreducens
utilizes a single, three-subunit membrane-bound complex (FrdABC) with a cytoplasmic active site for both the reduction of fumarate and the oxidation of succinate (11
). The bifunctional fumarate reductase/succinate dehydrogenase of G. sulfurreducens
is homologous to the characterized, di-heme, menaquinone-reducing succinate dehydrogenase of Bacillus subtilis
and the menaquinone-oxidizing fumarate reductases of Wolinella succinogenes
and belongs to heterotrimeric B-type family of fumarate reductases and succinate dehydrogenases (11
). During fumarate reduction by G. sulfurreducens
, the succinate dehydrogenase is not required for fumarate production, since fumarate is present in the media and the TCA cycle operates as an “open loop” (11
). Therefore, only six electrons, in the form of NADH, NADPH, and ferredoxin, are derived from oxidation of acetate during fumarate reduction with the fumarate reductase serving as the terminal electron accepting step.
In the metabolic model, all reducing equivalents derived from the TCA cycle were assumed to be interconverted to NADH and enter the menaquinone pool via the two 14-subunit type I NADH dehydrogenase complexes that are encoded in the G. sulfurreducens
). Type I NADH dehydrogenases typically translocate at least two charge equivalents per NADH oxidized and catalyze the consumption of two cytoplasmic protons during menaquinone reduction (33
). Whether menaquinol oxidation results in electrogenic translocation at the fumarate reductase was less clear. The succinate dehydrogenase of B. subtilis
has been suggested to be electrogenic (84
). However, more recent biochemical evidence indicates that the fumarate reductases of W. succinogenes
, Rhodothermus marinus
, and B. subtilis
are not electrogenic (6
). Thus, three possible mechanisms for generating energy during growth with fumarate were examined: translocation at both the NADH dehydrogenase and the fumarate reductase, translocation at the NADH dehydrogenase alone, and translocation only by fumarate reductase. Simulations were performed assuming an H+
/ATP ratio of 4.
Only one of these scenarios, in which proton translocation occurred at the NADH dehydrogenase alone, was consistent with both thermodynamic considerations and available experimental data. The lower boundary for the ΔG
′ of acetate oxidation coupled to fumarate reduction was calculated to be approximately −186 kJ/mol acetate, based on the steady-state concentrations of substrates and products present in the medium during acetate-limited growth in chemostats (μ ∼ 0.05 h−1
). Given a ΔG
′ for ATP synthesis of −60 kJ/mol (47
), a maximum of 3 mol of ATP synthesized/mol of acetate oxidized was estimated. If the NADH dehydrogenase was considered to be the sole electrogenic site, then the maximum electron transport stoichiometry was 2H+
, where n
is the number of electrons, and a total of 1.5 mol of ATP could be generated per mol of acetate, a value well within the thermodynamic boundary of 3 mol of ATP per mol of acetate established above. If both NADH dehydrogenase and fumarate reductase were electrogenic, the electron transport stoichiometry increased to 4H+
, resulting in 3 mol of ATP/mol of acetate, which was at the thermodynamic boundary, and hence unlikely to occur in vivo. Translocation at the fumarate reductase alone was excluded based on the observation that growth yields per mole of succinate produced were similar during growth with acetate and that with hydrogen (19
). If the fumarate reductase operated as an electrogenic redox loop, hydrogen oxidation would result in translocation of 4H+
for acetate oxidation, which would manifest as a significantly higher growth yield per electron.
Evaluation of the proton translocation stoichiometry during Fe(III) reduction.
Analysis of the in silico growth of G. sulfurreducens
under Fe(III)-reducing conditions uncovered an important energetic dilemma facing Fe(III)-reducing bacteria. To adapt the model of fumarate reduction to Fe(III) reduction, Fe(III) reduction was first modeled as a reaction that occurred outside the cell, consistent with numerous experimental observations (18
) and the fact that insoluble Fe(III) oxides are the predominant form of Fe(III) in most soils and sediments (53
). Under Fe(III)-reducing conditions, the TCA cycle operated as a closed loop (34
) and produced eight electrons per mole of acetate oxidized. Menaquinone reduction was considered to occur as a consequence of a nonelectrogenic succinate dehydrogenase and a proton-pumping NADH dehydrogenase as described for fumarate reduction, resulting in the translocation of six H+
per acetate oxidized. However, model simulations using this electron transport scheme indicated that cells would not be capable of growth (in silico) under Fe(III)-reducing conditions.
The inability of a single 2H+/2e− NADH dehydrogenase coupling site to support simulated Fe(III)-dependent growth was traced to the fact that the site of Fe(III) reduction was extracellular. The eight cytoplasmic protons that were produced from each mole of acetate oxidized in the cytoplasm were consumed in the cytoplasm when fumarate was the electron acceptor. In contrast, during Fe(III) reduction, electrons were transported outside the cell, while leaving protons in the cytoplasm, effectively dissipating the membrane potential and acidifying the cytoplasm. In order to generate sufficient energy to compensate for the production of protons in the cytoplasm, an additional coupling step was required.
The most likely mechanism for additional membrane potential generation during Fe(III) reduction was during transfer of electrons into the periplasmic cytochrome pool. Based on the fact that cytochromes implicated in Fe(III) reduction have midpoint potentials in the range of −190 mV (omcB
) and −136 to −155 mV (ppcA
), the energy available for coupling at this site could support translocation of 1H+
. This reaction was modeled as the release of menaquinol protons back to the cytoplasm by a protein capable of translocating 1 H+
per pair of electrons transferred to the cytochrome pool. Inclusion of this reaction and accounting for all of the protons produced and consumed during metabolism resulted in a theoretical maximum yield with Fe(III) as the electron acceptor of 0.5 mol of ATP/mol of acetate compared to the 1.5 mol of ATP/mol of acetate during fumarate reduction. This output of the model provides an explanation for the experimental finding that growth yields of G. sulfurreducens
are ∼3-fold higher when fumarate serves as the terminal electron acceptor than during growth with Fe(III)-citrate (29
). This experimental result was initially surprising because it is not consistent with expectations based on available energy, since the mid-point potential of the fumarate-succinate redox couple (at pH 7) is 0.03 V, whereas that of the Fe(III)-citrate/Fe(II) couple is 0.37 V.
These results suggest that reducing extracellular electron acceptors such as Fe(III) oxides, Fe(III)-citrate, elemental sulfur (S°), or electrodes will result in the generation of less biomass per electron transferred than growth with intracellularly reduced electron acceptors. This may be an important consideration for applications such as bioremediation and electricity harvesting from waste organic matter, in which electron transfer to metals or electrodes, rather than production of biomass, is the primary goal.
Incorporation of physiological parameters.
Once the proton translocation stoichiometry was determined, the next step in the development of the model was incorporation of physiological parameters, namely, the biomass component demands and maintenance energy requirements. First, a reaction representing growth-associated biosynthetic demands was constructed based on measurements of G. sulfurreducens biomass composition (see Materials and Methods). This reaction takes into account the amounts of 58 metabolites, cofactors, precursors, and ions required to synthesize each gram (dry weight) (gdw) of biomass, as well as proton consumption for reductive reactions, and the calculated ATP costs for the polymerization (peptide bond formation, DNA replication, and RNA polymerization) and biosynthesis of precursors and metabolites (see Table S2 in the supplemental material).
The ATP demand due to non-growth-associated energy functions, such as maintenance of ion gradients, turnover of RNA, and regulatory metabolism, was estimated by plotting acetate consumption during acetate-limited growth in chemostats with fumarate serving as the electron acceptor versus growth rate, extrapolating acetate consumption at a growth rate of zero (66
), and calculating the amount of ATP that would be produced from oxidation of this amount of acetate. This flux (ATPM: 0.45 mmol of ATP/gdw/h) was then used in conjunction with the biomass demand equation to predict the in silico growth yield of G. sulfurreducens
over a range of growth rates. Comparison of these predictions with chemostat-derived growth yields (29
) indicated that an additional flux of 46.7 mmol of ATP/gdw was consumed for growth-associated energy demands. This value was included as part of the biomass synthesis equation for all further growth simulations.
Evaluating model robustness.
The metabolic reaction network, combined with demand reactions for biomass synthesis, correctly predicted growth yields and acetate consumption rates for growth in standard acetate-limited chemostats (μ = 0.06 h−1
) with Fe(III)-citrate or fumarate as the electron acceptor (Fig. 1
). Perturbations in variables used to construct the model, such as the biomass composition, which was derived from batch cultures of fumarate grown cells, had minimal effect on predicted growth yields for G. sulfurreducens.
For instance, when a range of biomass composition equations (e.g., reflecting a range from 0.4 to 0.55 g of protein/gdw) were incorporated into the model, predicted yields were not significantly affected (1.5 to 2.5% differences). By comparison, a change in electron acceptor from fumarate to Fe(III) citrate caused threefold differences in yield or respiration rates (29
). This revealed that the model predictions were robust to changes in biomass composition and nutrient availability, which was consistent with other work showing that variations in biomass composition produce only subtle effects on predicted growth yields or fluxes through central metabolic pathways (23
). Hence, it is possible to assume that even significant changes (10 to 20%) in biomass composition would not affect the nature of metabolic predictions.
Model-based characterization of G. sulfurreducens metabolism.
Examination of the reconstructed metabolic network revealed that G. sulfurreducens
has multiple pathways for acetate utilization (acetyl-coenzyme A [acetyl-CoA] transferase, acetate kinase, and phosphotransacetylase), interconversion of pyruvate to acetyl-CoA (pyruvate formate lyase, pyruvate ferredoxin oxidoreductase, and pyruvate dehydrogenase), and anapleurotic reactions (phosphoenolpyruvate [PEP] carboxykinase and pyruvate carboxylase). Simulations predicted that during acetate-limited growth with Fe(III)-citrate (acetate uptake rate of 13.63 mmol/gdw/h for a growth rate of 0.06 h−1
), 93.6% of all acetate transported into the cell was utilized for oxidation and ATP generation via the TCA cycle (as flux through citrate synthase, or 12.76 mmol/gdw/h [Fig. 2
Flux from acetyl-CoA to pyruvate via pyruvate-ferredoxin oxidoreductase was predicted to be the sole source of carbon fixation in G. sulfurreducens
, and in silico, 4% of consumed acetate (0.55 mmol/gdw/h) was utilized in this fixation reaction when Fe(III)-citrate was the electron acceptor. These values compared favorably with radiolabeling experiments in a close marine relative of G. sulfurreducens
, Desulfuromonas acetoxidans
, in which acetate oxidation linked to reduction of extracellular sulfur resulted in 4% of acetate being fixed into cell carbon via a labeling pattern consistent with the use of pyruvate ferredoxin oxidoreductase (35
Activation of acetate via acetyl-CoA transferase only provided acetyl-CoA at a rate equal to the flux through the TCA cycle due to the dual role of the transferase in completing the TCA cycle (succinate to succinyl-CoA) and activating acetate. Thus, flux through an additional acetate activation pathway (acetate kinase [0.94 mmol/gdw/h]) was required to provide sufficient acetyl-CoA for pyruvate synthesis to meet gluconeogenic and anapleurotic demands. The major demand for pyruvate (54%) in G. sulfurreducens was predicted to be PEP synthesis for gluconeogenic reactions. Activation of pyruvate to PEP consumed 5.1% of the ATP flux in the cell, a value that doubled if the cost of activating acetate to acetyl-phosphate was considered. These distributions indicated that the availability of exogenous compounds with three or more carbons would eliminate this significant acetate and ATP demand and demonstrated the metabolic specializations which enable G. sulfurreducens to use acetate as both a carbon source and electron donor.
Although a complete pathway for glycolysis could be reconstructed, there is no physiological evidence that G. sulfurreducens
can grow with glucose as an electron donor (12
). This is most likely due to the fact that no sugar transporters appear to be present in the genome of G. sulfurreducens
. Incorporation of the appropriate transporters into the model suggested that G. sulfurreducens
could be genetically engineered to grow not only on glucose but also on substrates such as threonine, malate, and glycerol. This possibility is currently under experimental investigation.
As described above, G. sulfurreducens
can neither oxidize fumarate for the production of energy nor utilize fumarate as a carbon source (12
). The only potential fumarate transporter that could be conclusively identified in the G. sulfurreducens
genome was a homolog of the dicarboxylate exchanger of Wolinella succinogenes
), which can catalyze the exchange of fumarate and succinate (DcuB encoded by GSU2751). This gene is absent from the genome of a closely related species, G. metallireducens
, which cannot grow with fumarate as an electron acceptor. Expression of G. sulfurreducens
DcuB in G. metallireducens
renders it capable of exploiting fumarate as an electron acceptor, suggesting that DcuB is indeed involved in fumarate uptake (11
). Model simulations with DcuB as the only route of fumarate entry indicated that a possible explanation for the inability of G. sulfurreducens
to assimilate fumarate as a carbon source is the requirement for equimolar succinate secretion by the exchanger in order to allow fumarate uptake. Since these results appear to be consistent with experimental data, the only mode of fumarate uptake that was incorporated into the model was this dicarboxylate exchanger.
Comparisons of amino acid synthesis by G. sulfurreducens and E. coli metabolic networks.
The abilities of the E. coli
) and G. sulfurreducens
metabolic networks to synthesize amino acids with acetate serving as the sole carbon source and electron donor were compared. The reactions associated with Fe(III) reduction in the in silico model of G. sulfurreducens
were incorporated into the E. coli
model. The ATP maintenance parameters and the electron transport chain in E. coli
were replaced with the corresponding parameters and reactions from G. sulfurreducens
in order to isolate the role of central metabolism and eliminate the effect of differences in electron transport on the ability to synthesize amino acids. The amino acid production rate was maximized for each individual amino acid at a fixed acetate uptake rate of 10 mmol/gdw/h. By analyzing the differences in the predicted yields under identical growth conditions, it was possible to detect differences in the metabolic capabilities of the two networks and explore the sources of any disparities. One key difference found during the present study was that the metabolic network of G. sulfurreducens
was more efficient at synthesizing most amino acids when acetate was the electron donor (Fig. 3
, efficiency determined as moles of amino acid synthesized per mole of acetate consumed). This discrepancy was not due to differences in the amino acid biosynthetic pathways themselves and was most dramatic in amino acids belonging to the pyruvate and aspartate families.
The key reaction that accounted for these differences was the production of 3-carbon units from activated acetate. In the case of G. sulfurreducens
, pyruvate ferredoxin oxidoreductase catalyzed the synthesis of pyruvate from one molecule each of acetyl-CoA and CO2
, whereas in E. coli
pyruvate synthesis required the use of the glyoxalate bypass in which 2 moles of acetyl-CoA are consumed and 1 mole of CO2
is produced (22
). Interestingly, incorporating this key reaction along with reactions enabling production of ferredoxin into the E. coli
network led to amino acid yields equal to those of G. sulfurreducens
. These simulations illustrate how metabolism in G. sulfurreducens
is adapted for acetate utilization and how growth with acetate can be selective for the presence of pathways that fix carbon dioxide compared to selection for growth with larger carbon compounds that reduce carbon fixation demands.
Analysis of the metabolic cost of extracellular quinones.
It has been hypothesized that one reason why Geobacter
species predominate over other Fe(III)-reducing microorganisms in many subsurface environments is that Geobacter
species expend less energy to reduce Fe(III) oxides than other Fe(III)-reducing organisms (58
). Current evidence suggests that Geobacter
species directly contact Fe(III) oxides (68
), whereas other species, such as Geothrix
) and possibly Shewanella
) species, secrete a quinone-like water-soluble compound that serves as an electron shuttle between the surface of the cell and Fe(III), although this possibility is not without controversy (63
In order to gain insight into the cost of producing a shuttle, menaquinone was selected as a model shuttling compound. The minimal concentration of electron shuttle that must be maintained to be effective was assumed to be 2 μM (49
). In subsurface environments shuttles released extracellularly will be lost via diffusion and advection. A very conservative estimate of the rate of this loss is 1% per hour. At cell densities on the order of 106
per ml, a secretion rate of 0.02 mmol/gdw/h would be required to maintain minimal shuttle concentrations. Even with these conservative estimates and assuming a typical acetate consumption rate of 0.01 mmol/liter/h, synthesis of the quinone group of menaquinone reduced the growth rate by 9% (Fig. 4
). As the size of the menaquinone molecule was increased with the addition of progressively larger side chains, the growth rate decreased by as much as 40% for the largest shuttling molecule considered.
These results suggest that the energy cost associated with synthesizing a secreted compound for electron shuttling can lead to significant reductions in biomass yields for an acetate-oxidizing organism. However, it is important to note that this analysis was done with the menaquinone as a representative compound and that additional analyses will be required to assess the biosynthetic demands associated with the production of actual electron shuttles once their structures and secretion rates are determined.
Functional analysis of G. sulfurreducens mutant phenotypes.
The availability of a genome scale model also enabled the characterization of systems level properties of the metabolic network. One such property is the set of genes and reactions that are essential to support growth in a defined medium. This information is important for genetic investigations since it can provide insight into which mutations may or may not have an observable phenotype.
In silico deletion analysis (26
) for growth with acetate as the electron donor and Fe(III)-citrate or fumarate as the electron acceptor indicated that most mutations were predicted to have either lethal [139 for fumarate and 143 for Fe(III)] or silent [440 for fumarate and 437 for Fe(III)] phenotypes (Fig. 5
and Table S4 in the supplemental material). Lethal mutations (e.g., deletion of acetyl-CoA transferase and pyruvate carboxylase) reflected the inability of the perturbed network to synthesize essential components from acetate, a relatively simple two-carbon compound, or the fact that a nonfermentable substrate such as acetate presents few alternative energy-yielding oxidative mechanisms.
Some silent phenotypes predicted by this analysis corresponded to reactions associated with seemingly redundant enzymes. The presence of functionally similar (but nonorthologous) enzymes could be due to selection for genetic robustness, in order to protect against mutations in essential reactions. Alternatively, this redundancy could reflect a need for metabolic robustness, in which different enzymes are needed to favor flux in opposite directions or are optimized for oxidation of different substrates. For instance, model simulations indicated that a mutation in any component of pyruvate-ferredoxin oxidoreductase would be compensated for by activity of pyruvate dehydrogenase or pyruvate-formate-lyase. However, since pyruvate-formate-lyase strongly favors function in the oxidative direction, it is unlikely that this enzyme can substitute for pyruvate-ferredoxin oxidoreductase in vivo, and the redundancy at this node likely reflects the presence of enzymes specialized for different tasks. Mutational and biochemical investigations are under way to test these hypotheses.
Large-scale in silico deletion analysis identified only 17 reactions (of 522 [3.25%]) that when deleted would be nonlethal but would have an effect on growth rate during acetate-limited growth (acetate uptake of 5 mmol/gdw/h) with fumarate as the acceptor. In contrast, in silico deletion of 59 reactions (6.4% of total) in the E. coli
network resulted in an intermediate growth rate during glucose-limited aerobic growth (79
). The lack of many intermediate modes of growth again reflects the fact that G. sulfurreducens
utilizes electron donors (acetate) and acceptors that cannot be metabolized by alternative pathways such as partial oxidation, fermentation, or routing of metabolites through pathways that would consume ATP, which is already in short supply.
Because of its importance to the energetics and TCA cycle function of Geobacter
, the mutant phenotype resulting from deletion of the bifunctional succinate dehydrogenase-fumarate reductase (FrdA) was investigated further. Since a mutant in the catalytic subunit for this enzyme already has been characterized, it was possible to compare in silico and in vivo phenotypes (11
). The model correctly predicted that the FrdA-deficient mutant would be unable to reduce fumarate and would be unable to grow with Fe(III)-citrate as the electron acceptor and acetate as the electron donor, since cells lacking the succinate dehydrogenase would be unable to complete the TCA cycle. Likewise, the model predicted that the mutant could grow with Fe(III)-citrate as the electron acceptor and H2
as the electron donor, if acetate was provided as a carbon source. In fact, this was the condition under which the FrdA-deficient strain was recovered.
The FrdA-deficient strain was previously reported to grow by oxidizing acetate with Fe(III)-citrate as the electron acceptor if the medium was supplemented with fumarate to allow completion of the TCA cycle (11
). When “rescuing” the TCA cycle in this manner, the FrdA-deficient strain grew at a faster rate and obtained higher yields than those of the wild-type strain growing in unsupplemented medium. In model simulations, an identical phenotype was observed (Fig. 6
). In this case, fumarate was not used as a carbon or energy source since G. sulfurreducens
can utilize fumarate only in the fumarate reductase due to the requirement for equimolar exchange of fumarate and succinate. Thus, the increase in growth yield for the fumarate-supplemented mutant was found to be due to the elimination of the energy cost associated with the cytosolic protons produced during succinate oxidation under Fe(III)-reducing conditions.
To further evaluate this phenotype, steady-state growth experiments in chemostats were conducted that confirmed the magnitude of the growth yield increase predicted in silico (Fig. 6
). The individual flux predictions for each TCA cycle reaction illustrate how the wild-type cells, which achieved a lower ATP yield per acetate oxidized, respired more rapidly to maintain the same growth rate as the fumarate-supplemented mutant. These results demonstrated the ability of the in silico model to make quantitative predictions and further highlight the energetic impact of cytosolic proton production during Fe(III) reduction by G. sulfurreducens.