Reversible interconversion of carbon dioxide and formate by an electroactive enzyme

Abstract

Carbon dioxide (CO₂) is a kinetically and thermodynamically stable molecule. It is easily formed by the oxidation of organic molecules, during combustion or respiration, but is difficult to reduce. The production of reduced carbon compounds from CO₂ is an attractive proposition, because carbon-neutral energy sources could be used to generate fuel resources and sequester CO₂ from the atmosphere. However, available methods for the electrochemical reduction of CO₂ require excessive overpotentials (are energetically wasteful) and produce mixtures of products. Here, we show that a tungsten-containing formate dehydrogenase enzyme (FDH1) adsorbed to an electrode surface catalyzes the efficient electrochemical reduction of CO₂ to formate. Electrocatalysis by FDH1 is thermodynamically reversible—only small overpotentials are required, and the point of zero net catalytic current defines the reduction potential. It occurs under thoroughly mild conditions, and formate is the only product. Both as a homogeneous catalyst and on the electrode, FDH1 catalyzes CO₂ reduction with a rate more than two orders of magnitude faster than that of any known catalyst for the same reaction. Formate oxidation is more than five times faster than CO₂ reduction. Thermodynamically, formate and hydrogen are oxidized at similar potentials, so formate is a viable energy source in its own right as well as an industrially important feedstock and a stable intermediate in the conversion of CO₂ to methanol and methane. FDH1 demonstrates the feasibility of interconverting CO₂ and formate electrochemically, and it is a template for the development of robust synthetic catalysts suitable for practical applications.

Carbon dioxide (CO₂) is a kinetically and thermodynamically stable molecule that is difficult to chemically activate. As the unreactive product of the combustion of carbon-containing molecules, such as fossil fuels, and biological respiration, it is accumulating in the atmosphere and is a major cause of concern in climate-change scenarios (1, 2). Consequently, catalysts that are able to sequester CO₂ from the atmosphere rapidly and efficiently, to generate reduced carbon compounds for use as fuels or chemical feedstocks, have long been sought after. Organometallic complexes that insert CO₂ into an M–H bond and the use of supercritical CO₂ to facilitate its hydrogenation are two promising approaches to the development of a homogeneous catalytic process (3–5). Extensive efforts to develop electrode materials for the direct, electrochemical reduction of CO₂ have been made also, but so far, they require the application of extreme potentials (are inefficient energetically) or they are nonspecific and produce mixtures of products (4, 6, 7). An efficient and specific method for reducing CO₂ electrochemically has obvious applications in a future energy economy based on the cyclic reduction and oxidation of simple carbon compounds, as an alternative to the well-known hydrogen economy model.

Here, we focus on the two-electron reduction of CO₂ to formate (see Fig. 1). Formate is the first stable intermediate in the reduction of CO₂ to methanol or methane, but it is increasingly recognized as a viable energy source in its own right. Thermodynamically, formate and hydrogen are oxidized at similar potentials (the reduction potential for the reduction of CO₂ to formate is close to that for the reduction of protons to hydrogen, see below), and fuel cells that use formate are being developed (8). It is not possible, at present, to reduce CO₂ to formate efficiently and specifically at an electrode surface.

Fig. 1.

Schematic representation of the electrocatalytic interconversion of CO₂ and formate by a formate dehydrogenase adsorbed on an electrode surface. Two electrons are transferred from the electrode to the active site (buried inside the insulating protein interior) by the iron–sulfur clusters, to reduce CO₂ to formate, forming a C–H bond. Conversely, when formate is oxidized, the two electrons are transferred from the active site to the electrode. The structure of FDH1 (which contains at least nine iron–sulfur clusters) is not known, so the structure shown is that of the tungsten-containing formate dehydrogenase from Desulfovibrio gigas [PDB ID code 1H0H (12)].

Formate dehydrogenases are enzymes that catalyze the oxidation of formate to CO₂. The most common class catalyze the direct transfer of a hydride moiety from formate to NAD(P)⁺, but it is difficult to drive them in reverse because the reduction potential of NAD(P)⁺ is more positive than that of CO₂ (9). In some prokaryotes, however, the formate dehydrogenases are complex enzymes that contain molybdenum or tungsten cofactors to transfer the electrons from formate oxidation to an independent active site, to reduce quinone, protons, or NAD(P)⁺ (10–13). These enzymes are suitable for adsorption onto an electrode, so that the electrode accepts the electrons from formate oxidation, and it may also donate electrons and drive CO₂ reduction. Therefore, they are potential electrocatalysts for the reduction of CO₂ (see Fig. 1). Tungstoenzymes catalyze low-potential reactions (14), so tungsten-containing formate dehydrogenases [in which the tungsten is coordinated by two pyranopterin guanosine dinucleotide cofactors and a selenocysteine (12)] are those most associated with CO₂ reduction. Here, we focus on FDH1, one of the tungsten-containing formate dehydrogenases from Syntrophobacter fumaroxidans, an anaerobic bacterium that oxidizes propionate to acetate, CO₂, and six reducing equivalents (15). The reducing equivalents are used to reduce protons to hydrogen or to reduce CO₂ to formate. During syntrophic growth S. fumaroxidans optimizes its metabolism by transferring the formate (and/or the hydrogen) to its syntrophic partner (16). Therefore, the tungsten-containing formate dehydrogenases of S. fumaroxidans must be both thermodynamically and kinetically efficient, and so they are good candidate electrocatalysts for the electrochemical reduction of CO₂.

Results

Molecular Properties of FDH1 from S. fumaroxidans.

FDH1, tungsten-containing formate dehydrogenase 1 isolated from S. fumaroxidans, comprises three different subunits [see supporting information (SI) Figs. S1 and S2]. The C-terminal domain of the largest subunit contains the conserved selenocysteine that ligates the tungstopterin active site (17), and it belongs to a large family of structurally related molybdo- and tungstoenzymes, including nitrate reductase (18), DMSO reductase (19), and the structurally characterized formate dehydrogenases (10–12). Between them, the N-terminal domain of the largest subunit and the other two subunits contain binding motifs for at least nine iron–sulfur clusters (see Fig. S2), deduced from their homology to proteins of known structure, in particular three subunits of the hydrophilic domain of respiratory NADH ubiquinone oxidoreductase (complex I) (20). Despite the conservation of residues that bind (or are predicted to bind) the flavin and NADH in complex I, purified FDH1 does not catalyze the reduction of NAD(P)⁺ by formate (15). However, it does catalyze the rapid interconversion of formate and CO₂ in catalytic activity assays using methyl viologen (MV) as the artificial redox partner; our preparations of FDH1 exhibited turnover rates of up to 3,380 s⁻¹ for formate oxidation at pH 8, and 282 s⁻¹ for CO₂ reduction at pH 7.5. Thus, considering FDH1 as a homogeneous catalyst alone, it exhibits a rate of CO₂ reduction that is more than two orders of magnitude higher than any known nonbiological catalyst for the same reaction (see below). Note that, for simplicity, we use the term “CO₂” to encompass the pH-dependent equilibrium of dissolved CO₂, H₂CO₃, HCO₃⁻, and CO₃²⁻ formed in aqueous solution under gaseous CO₂, or upon the addition of sodium carbonate (21).

FDH1 Adsorbed on an Electrode Surface Catalyzes the Reversible Interconversion of CO₂ and Formate.

Fig. 1 illustrates our experimental strategy (22) for the electrochemical interconversion of CO₂ and formate. Molecules of FDH1 are adsorbed to a freshly polished pyrolytic graphite edge electrode to activate the surface. When the electrode is placed into an electrochemical cell containing the enzyme's substrates, catalysis can be driven and controlled by applying a potential and quantified by measuring the current flowing. Thus, the catalytic properties and capability of the adsorbed enzyme are defined under precisely controlled conditions. Ideally, substrate supply to the activated surface and electron transfer between the enzyme's active site and the electrode are fast and efficient and do not hinder turnover. Fig. 1 highlights how the many iron–sulfur clusters in FDH1 may enable efficient electron transfer through the insulating protein interior to the buried active site.

Voltammograms of FDH1 catalyzing the interconversion of formate and CO₂ are presented in Fig. 2. Both CO₂ reduction and formate oxidation are initiated at approximately −0.4 V, close to the reduction potential (see below), and then their rates increase sharply as the driving forces increase. Qualitatively, the voltammograms resemble those exhibited by Allochromatium vinosum NiFe hydrogenase, a focus of research in the development of biohydrogen fuel cell technology (23). The voltammetric waveshapes display a sigmoidal onset that changes into a linear dependency as the driving force (overpotential) is increased. This reflects the influence of the interfacial electron transfer process that, even at the highest driving force applied, lags behind the rapid active-site turnover (24). CO₂ reduction reaches 0.08 mA cm⁻² at −0.8 V (pH 5.9), whereas formate oxidation increases to 0.5 mA cm⁻² at +0.2 V (pH 8.0) (Fig. 2 A and C).

Fig. 2.

Electrocatalytic voltammograms showing CO₂ reduction and formate oxidation by FDH1. Shown are the reduction of 10 mM CO₂ (pH 5.9) (A); electrocatalysis in 10 mM CO₂ and 10 mM formate (pH 6.4) (B), showing the points of intersection (marked with crosses) that define the reduction potential for the interconversion of CO₂ and formate; and the oxidation of 10 mM formate (pH 7.8) (C). The first voltammetric cycles are shown in black, subsequent cycles are in gray; background cycles recorded in the absence of substrate are also shown (gray). Note that the background cycle in B is offset slightly from the catalytic scans because of variation in the electrode capacitance. Substrates were added as sodium formate or sodium carbonate. For A and C, 25 mV s⁻¹, for B, 100 mV s⁻¹; 37°C, electrode rotation 1,000 rpm.

Fig. 2B displays formate oxidation and CO₂ reduction in a single experiment. The electrocatalytic activity gradually decreases, because of enzyme desorption or denaturation, so that a potential at which all of the catalytic voltammograms intersect is a potential of zero net catalytic current. Net formate oxidation occurs at all potentials above the intersection, and net CO₂ reduction occurs at all potentials below it. The potentials of zero net catalytic current are the same in both scan directions, and they define the reduction potential for the interconversion of CO₂ and formate in the solution. Importantly, the current varies continuously across the reduction potential, showing that catalysis is thermodynamically reversible (even the smallest displacement in potential drives turnover) and that the system interconverts electrical and chemical bond energy efficiently.

The pH-Dependent Reduction Potential of CO₂.

Fig. 3 shows how the reduction potential of CO₂, measured as the potential of zero net catalytic current in 10 mM CO₂ and 10 mM formate, varies between pH 4 and pH 8 (the region in which both oxidation and reduction are observed). The data have been fit by using the Nernst equation (Eq. 1). Because the total concentration of solution carbonate species (CO_2(T) = CO_2(aq) + H₂CO₃ + HCO₃⁻ + CO₃²⁻) present under 1 atm of CO₂ increases dramatically at higher pH, Eq. 1 has been calculated by using a closed system containing 1 M CO_2(T) instead. It includes the pK values for both the oxidized and reduced states, defined in Fig. 3.

The Pourbaix diagram (see Fig. 3) shows that at low pH the pH-dependence reflects the two-electron, one-proton interconversion of CO₂ (or H₂CO₃) and formate and, at higher pH, the two-electron, two-proton interconversion of HCO₃⁻ and formate. Thus, the data define an effective pK_ox1 of 6.39 for the equilibrium between CO₂ (or H₂CO₃) and HCO₃⁻, matching the known value (21). The reduction potential at pH 7 is −0.389 V, a value that is consistent with that commonly adopted at neutral pH (−0.42 or −0.43 V, for example, refs. 6 and 25). By using the known value of pK_red1 (3.75) and Eq. 1, the measured reduction potentials extrapolate to −0.042 V at pH 0. Previously, values for the standard reduction potential for CO₂ to formic acid (pH 0, 1 atm CO₂, 289 K) have been calculated from calorimetric data (because irreversible electrochemical responses preclude direct measurements). Perhaps surprisingly, given the current level of interest in CO₂ reduction, no consensus has been reached, and two different values are quoted, −0.199 V (26) and −0.11 V (4). By using 34 mM atm⁻¹ for Henry's constant (21), the two standard potentials are redefined as −0.157 V (26) and −0.067 V (4) for 1 M CO_2(T) instead of 1 atm of CO₂. By using the published pK values (21), they give −0.496 and −0.406 V at pH 7, respectively. The values reported here were recorded under nonstandard conditions, in 10 mM formate and 10 mM CO_2(T), at 310 K. They are most consistent with the standard potential of −0.11 V under 1 atm of CO₂ (equivalent to −0.067 V in 1 M CO_2(T) or −0.406 V at pH 7). Note that experimental measurement of the standard reduction potential is complicated by the highly imperfect behavior of CO₂ (particularly at high pH) and by the impracticality of measuring at pH 0 by using an enzyme-based system.

Fig. 3.

The pH-dependent reduction potential of CO₂, measured by using the electrocatalytic voltammetry of FDH1. The reduction potentials (open circle, measured as illustrated in Fig. 2B), were recorded in 10 mM formate and 10 mM CO₂ at 310 K. They have been fitted to the Nernst equation for the reduction potential of CO₂ (Eq. 1) by using the parameters pK_red1 = 3.75, pK_ox1 = 6.39, pK_ox2 = 10.32 (21) and E^0′ = −0.042 V The pH-dependent reduction potential is superimposed on the Pourbaix diagram, showing which species are most stable thermodynamically at different pH values and potentials.

Rate of Catalysis by FDH1.

The rate of formate oxidation by FDH1 is greatest at pH 8.4, both in conventional solution assays and on the electrode surface (see Fig. 4). In conventional assays, the rate of CO₂ reduction increases at lower pH but cannot be measured below pH 6.5 because the absorption coefficient of the MV radical decreases significantly. Electrocatalytic CO₂ reduction is greatest at pH 5.5. These observations are consistent with CO₂ being the species that is reduced (rather than HCO₃⁻ or CO₃²⁻) (27). Direct comparisons of the rates of catalysis in solution and on the electrode are difficult. They may be confounded by kinetic limitations from the artificial redox partner and from interfacial electron transfer, respectively. Furthermore, the amount of electrocatalytically active adsorbed enzyme is too low to be detected in the absence of substrate and may vary with pH. During a 1-hour incubation in a graphite electrocatalysis pot, the total amount of FDH1 that adsorbed to the electrode surface (whether in an electrocatalytically active conformation or not) was estimated from the FDH1 activity lost from solution, and it corresponded to a densely packed monolayer (3.7 ± 0.5 × 10⁻¹² mol cm⁻², calculated value 3.5 × 10⁻¹² mol cm⁻², assuming a protein density of 1.4 g cm⁻³). Using the turnover numbers from conventional assays suggests that a monolayer of electrocatalytically active FDH1 should produce limiting current densities of 0.2 mA for CO₂ reduction at pH 7.5 and 2.4 mA for formate oxidation at pH 8 [comparable with the current density from H₂ oxidation at either platinum or hydrogenase-modified electrodes (23)]. The maximum current densities observed so far in the voltammetric experiments (as shown in Fig. 2) are significantly lower (0.015 and 0.5 mA cm⁻², respectively), showing that the number of electrocatalytically active molecules adsorbed on the surface is much lower than that required to constitute a complete monolayer. The low currents observed also reflect the need to optimize the interfacial electron transfer characteristics, to contract the potential-dependent domain and realize the true limiting current.

Fig. 4.

The kinetics of catalysis by FDH1 as a function of pH. Electrocatalytic current densities from the first voltammetric scan in 10 mM formate and 10 mM CO₂ at 0.25 V overpotential (left-hand axis) are compared to data from conventional assays in solution in 10 mM formate or 10 mM CO₂ (monitored by formation or consumption of the methyl viologen radical, right-hand axis). Lines are guidelines only (dashed lines and open symbols, and solid lines and filled symbols). Open circle, rate of formate oxidation in solution; filled circle, electrocatalytic formate oxidation; open square, rate of CO₂ reduction in solution; filled square, electrocatalytic CO₂ reduction.

Importantly, electrocatalytic CO₂ reduction is observed to increase 6-fold from pH 7.5 to 5.5 (see Fig. 4). The highest current density attained so far in a single experiment (0.080 mA cm⁻² at pH 5.9 at −0.8 V) corresponds to an average turnover number of 112 s⁻¹ from a monolayer of FDH1. This value is a known underestimate for the true rate of electrocatalysis by FDH1 (much less than a monolayer of electrocatalytically active enzyme is actually present (see above), and the current at −0.8 V has not achieved its potential independent limit). Nevertheless, it is still more than two orders of magnitude faster than CO₂ hydrogenation by ruthenium complexes in supercritical CO₂ (5) and 10 times faster than photosynthetic CO₂ activation during the carboxylation of d-ribulose 1,5-bisphosphate by rubisco (28).

Formate Is the Only Product of CO₂ Reduction by FDH1.

To establish the products of electrochemical CO₂ reduction by FDH1, a graphite “pot” working electrode was used for bulk electrocatalysis. FDH1 was adsorbed from the solution [containing 20 mM sodium carbonate (pH 6.5)] during electrocatalysis. This method sustained a small population of adsorbed electroactive molecules but provided a lower maximum current (typically 15 μA cm⁻² after 15 min.) than direct application of a concentrated solution. The overpotential and time of electrocatalysis were varied; typically, in a 1-hour experiment the current at −0.8 V averaged 8.5 μA cm⁻² (30.4 mC cm⁻² h⁻¹), and the FDH1 molecules present catalyzed, on average, 42,500 transformations. The total charge passed was defined by integrating the current–time curve, and the formate produced was quantified by ion-exclusion chromatography. The data (see Table 1) reveal that formate is the sole product of electrocatalytic CO₂ reduction by FDH1, with a measured Faradaic efficiency of ≈100% at modest overpotentials. At higher overpotentials, the efficiency decreases slightly (to 97% at 0.45 V overpotential) because nonspecific electrode reactions begin to contribute.

Table 1.

Quantitative electrochemical reduction of CO₂ to formate catalyzed by FDH1

Applied potential, V	Faradaic efficiency, %	Thermodynamic efficiency, %
−0.41	102.1 ± 2.2	96.6
−0.51	98.6 ± 3.6	89.5
−0.61	98.0 ± 1.7	83.4
−0.81	97.3 ± 1.1	73.3

The Faradaic efficiency is the ratio of the formate detected to the charge consumed (accounting for two electrons per formate). The thermodynamic efficiency is the ratio of the chemical bond energy formed (E_H2O–E_CO2) to the potential difference applied (E_H2O–E_app), assuming that the oxidation of H₂O to O₂ is 100% efficient [E_H2O = 0.844 V, E_CO2 = −0.368 V (pH 6.5)]. Experimental conditions: 20 mM sodium carbonate (pH 6.5).

Discussion

There have been many attempts to drive the electrochemical reduction of CO₂ to formate (4, 6, 7). However, excessive overpotentials were required so that a significant proportion of the potential applied was dissipated as heat. Furthermore, mixtures of products were formed, resulting from the formation of the CO₂⁻ radical and from competing reactions such as hydrogen formation. Notably, the direct reduction of aqueous CO₂ at lead, mercury, indium, and thallium electrodes produces formate with >90% Faradaic efficiency but only at potentials below −1.5 V. Table 1 reports the “thermodynamic efficiency” of CO₂ reduction (coupled to H₂O oxidation) at the overpotentials used here. For comparison, at −1.5 V, ≈50% of the potential is dissipated as heat. Preliminary attempts to use a ruthenium catalyst to reduce CO₂ to formate electrochemically required similarly high overpotentials and also led to a mixture of products (29). Formate dehydrogenase from Pseudomonas oxalaticus was applied also to the electrochemical reduction of CO₂ (25). However, it was present in solution at high concentration, and electron transfer between the enzyme and the electrode was mediated by millimolar concentrations of methyl viologen (comparable with the concentrations of formate generated), with variable Faradaic efficiency. The only previous demonstration of efficient electrocatalytic CO₂ reduction is from another adsorbed enzyme, CO dehydrogenase I from Carboxydothermus hydrogenoformans, which catalyzes the rapid and reversible interconversion of CO₂ and CO (not CO₂ and formate) (30).

Virtually nothing is known about the mechanism of CO₂ activation and reduction in enzymes such as FDH1. A number of rudimentary bis(dithiolene)tungsten^IV,VI thiolate and selenolate compounds have been synthesized to mimic the active sites of formate dehydrogenases (31), but, perhaps unsurprisingly, they exhibit no catalytic activity. However, although molybdenum and tungsten are not generally associated with biological hydride transfer reactions (14, 32), CO₂ is known to insert into the W–H bond in [HW(CO)₅]⁻ to form a stable formate adduct (33), and considerable progress has been made in developing homogeneous organometallic catalysts based on M–H insertion reactions (3, 4). For example, CO₂ insertion into Ru(terpy)(bpy)H⁺ proceeds at 8.5 s⁻¹ in 10 mM CO₂ (although displacement of the formate by H₂O to complete a catalytic cycle is significantly slower) (34).

FDH1 is the first example of a molecular structure that can activate an electrode to reduce CO₂ (and form a C–H bond) specifically, rapidly, reversibly, and under thoroughly mild conditions. Although, as a large and oxygen-sensitive protein complex it is not itself a realistic target for technical exploitation, FDH1 establishes an important paradigm. It is accessible to electrochemical, spectroscopic and structural studies, to provide detailed mechanistic information and elucidate structure-function relationships. It is hoped that these will aid in the rational evaluation and development of robust synthetic compounds that host similar chemical and electrochemical capabilities to FDH1 and that would be suitable for future applications in carbon capture, energy storage, and fuel cell devices.

Materials and Methods

All procedures were carried out under strictly anaerobic conditions. FDH1 was isolated from S. fumaroxidans cells grown on 20 mM propionate and 60 mM fumarate by using the protocol described (15) with minor modifications, except that FDH1 eluted from the hydroxyapatite column (40 μm type II CHT ceramic; Bio-Rad) in a linear gradient of sodium phosphate (pH 7). In conventional activity assays at 37°C, oxidation was initiated by 10 mM sodium formate, after a 2-min incubation of FDH1 in 1 mM MV, 10 μM dithionite, 50 mM Tris-Cl (pH 8), and reduction was initiated by adding 10 mM sodium carbonate to FDH1 in 100 mM sodium phosphate (pH 7.5), 0.1 mM MV, 0.045 mM dithionite. The reactions were followed spectrometrically by the reduction of MV or by the oxidation of the MV radical, respectively (ε₅₇₈ = 9.7 mM⁻¹ cm⁻¹). pH values were checked after experimentation.

Cyclic voltammetry was performed at 37°C as described (24), by using a rotating disk pyrolytic graphite edge (PGE) electrode. pH was controlled by using a mixed-buffer system [10 mM potassium acetate, 2-morpholinoethane sulfonic acid (Mes), N-cyclohexyl-2-aminoethane sulfonic acid (Hepes), and N-tris(hydroxymethyl)methyl-3-aminopropanesulfonic acid (TAPS)] and checked following experimentation. All potentials are reported relative to the standard hydrogen electrode. Electrocatalysis was carried out at 25°C in 100 mM Mes (final pH 6.5), 1 M NaCl, 20 mM sodium carbonate, and ≈5 μg FDH1, in a PGE pot working electrode (0.75 ml, surface area 3.94 cm²) and monitored amperometrically. The Ag/AgCl reference electrode and platinum mesh counter electrode were behind a Vycor glass frit. Formate concentrations were determined by ion-exclusion chromatography (using a Dionex AS15 column (2 × 250 mm), IonPac AG15 guard column (2 × 50 mm), and suppressed-conductivity ICS-2000 detector). Formate was eluted isocratically in 20 mM KOH, and concentrations >0.25 ppm (5.4 μM) were determined with a standard deviation of 0.5%.