Efficient Electrocatalytic CO2 Fixation by Nanoconfined Enzymes via a C3-to-C4 Reaction That Is Favored over H2 Production

Reduction of CO2 and its direct entry into organic chemistry is achieved efficiently and in a highly visible way using a metal oxide electrode in which two enzyme catalysts, one for electrochemically regenerating reduced nicotinamide adenine dinucleotide phosphate and the other for assimilating CO2 and converting pyruvate (C3) to malate (C4), are entrapped within its nanopores. The resulting reversible electrocatalysis is exploited to construct a solar CO2 reduction/water-splitting device producing O2 and C4 with high faradaic efficiency.

N anoconfinement and compartmentalization are essential characteristics of living cells. A key feature ensuring that multistep catalytic processes are fast and efficient is that enzymes along a cascade are highly concentrated at a local level, and diffusion distances of intermediates are very short. For instance, in biological photosynthesisin plants and algaethe reactions of the Calvin cycle along with NADP + reduction and adenosine 5′-triphosphate generation occur in the chloroplast stroma over distances < 100 nm. 1,2 The inspiration offered by biological nanoconfinement thus suggests unique advantages to help drive the development of artificial photosynthesis. Recently, we reported that the ubiquitous photosynthetic enzyme ferredoxin NADP + reductase (FNR) trapped within the nanopores of a porous indium tin oxide (ITO) layer deposited on a conducting support is highly active and stable for NADP(H) recycling and coupling to a second enzyme (E2) that is also nanoconfined ( Figure  1). 3−5 Crucially, regeneration of NAD(P)H is a subject of intense research, as hundreds of oxidoreductases catalyzing important reactions require stoichiometric amounts of NAD-(P)H as a mobile redox cofactor. 6 Recycling continuously is essential to make any biocatalytic process economically feasible. Cofactor recycling has been extensively reviewed. 7−14 Several protocols most relevant to this paper are also mentioned later.
The electrode system we are developing is denoted (FNR + E2)@ITO/support. When trapped alone at ITO, FNR displays diagnostic reversible electrochemistry of its flavin adenine dinucleotide active site and of NADP + when the latter is introduced. 3 The highly efficient nanoconfined electrochemical recycling of NADP + /reduced nicotinamide adenine dinucleotide phosphate (NADPH) both drives catalysis and allows continuous monitoring of the overall catalytic rate as the electrical current. 5,15 Electrochemical CO 2 reduction exploiting the efficiency of enzymes as reversible electrocatalysts has long focused on C1 transformations using CO dehydrogenase or formate dehydrogenase. 16−22 By introducing, as E2, the enzyme L-malate:-NADP + oxidoreductase (oxaloacetate-decarboxylating), EC 1.1.1.40 (MaeB from Escherichia coli: hereafter abbreviated as ME), the electrode offers an instructive way to drive CO 2 reduction with a very low overpotential, by incorporating it into pyruvate (C3) to form malate (C4) (Figure 1). The small Δ r G (−7 kJ mol −1 at 25°C, pH 7.5) for decarboxylation of malate by NADP + allows the reaction to be run in reverse, so it presents an ideal system by which to achieve specific CO 2 reduction, provided a continuous supply of NADPH cofactor is available. 23 Use of ME to incorporate CO 2 into pyruvate was first reported over 30 years ago and featured photoelectrochemical cofactor regeneration by FNR using methyl viologen as a low-potential electron mediator. 14,24−26 Here, we show how nanoconfinement is exploited to perform C3-to-C4 CO 2 incorporation with ease, electrochemical clarity, and high efficiency with regard to overpotential and competition with H 2 evolution.
The stationary cyclic voltammograms (CVs) shown in Figure 2 demonstrate the reversibility of the electrocatalytic incorporation of CO 2 into pyruvate and the inhibitory effect of malate. A (FNR + ME)@ITO/Ti electrode (1 cm 2 ) was prepared by placing a ITO/Ti foil electrode, preloaded with FNR (see Supporting Information for details) into a CO 2saturated solution of pyruvate (50 mM) then adding ME to give a total concentration of 2.5 μM ME. The solution was buffered at pH 7.5 (25°C) using 0.20 M N-(2-hydroxyethyl)piperazine-N′-ethanesulfonic acid (HEPES) and 0.10 M KHCO 3 (see Supporting Information for details), and contained 4 mM MgCl 2 (required for activity of ME) and 20 μM NADP + . The voltammogram labeled as "pyruvate + CO 2 " (black trace) was recorded after 6 cycles at 1 mV s −1 (approx. 100 min after introducing ME): it reveals a large reductive current (reaching a limit at approximately −0.4 V) due to the incorporation of CO 2 into pyruvate. At a much more negative potential (below −0.55 V), the current increases again, because of H + reduction. After adding malate (final concentration also 50 mM) and stirring briefly, a new voltammogram (blue trace) is obtained, in which the reduction current is suppressed and a sizeable oxidation current has appeared. Under the conditions of equal concentrations of C3 and C4, the trace cuts through the zero-current axis at the formal potential for the C3−C4 interconversion (taking an average for scans in both directions): the markers that signify the potentials of relevance for this work include this value. Notably, CO 2 reduction commences at a significantly more positive electrode potential than the thermodynamic value for H 2 evolution: the C3-to-C4 potential is approximately +0.1 V versus the reversible 2H + /H 2 potential (RHE). High Faradaic efficiency for aqueous CO 2 reduction is therefore predicted. No catalysis is observed when NADP + or ME are not included (see Supporting Information, Figure S1) and the electrode remains stable for several days without any significant loss. Figure 3A shows a timecourse (chronoamperogram) for electrochemical CO 2 fixation [−0.45 V vs standard hydrogen electrode (SHE) at 25°C, pH 7.5, stirring] with a constant flow of CO 2 through the cell headspace. A larger ITO/Ti electrode (16 cm 2 ) preloaded with FNR was used for this experiment to increase the amount of the product formed in the reactionthe solution now containing 20 mM pyruvate, 50 mM phosphate as buffer (to clarify product analysis by 1 H NMR), 0.10 M KHCO 3 , and 4 mM MgCl 2 . After injecting NADP + (to 20 μM) which results immediately in a small catalytic current, introduction of ME (final concentration 2.5 μM) initiates the growth of a large catalytic current, which reaches a maximum after 2 h before slowly decreasing over the course of 24 h. The current increase, commencing from zero, corresponds to the rate at which ME enters the electrode nanopores to couple with the NADP(H) recycling as it occurs at FNR molecules. (At this stage, we do not know the optimum FNR/ME ratio that will be active at a local level, and further experiments must be devised to answer this question.) After 14 h, the quantity of malate formed (17.6 mM = 88% conversion) was confirmed by 1 H NMR ( Figure S2), the resulting total turnover number (TTN, moles malate formed/ moles NADP + present) being 880. The current decreases slowly over the course of time, which could be related to various factors, such as instability of the FNR/ME catalytic zones and/or product inhibition. To identify the origins of current decrease, we performed a series of interventions. In panel A, the reaction was recharged with pyruvate after 15 h, the injection of which (to give a further 20 mM) increased the current. Notably, the concentration of the unconverted substrate after 14 h of reaction had dropped well below K M (6.21 mM for pyruvate). 27 To investigate the effect of product inhibition a solution exchange was performed after 18.3 h. The result is shown in an enlarged form in panel B. Briefly, the cell compartment was washed 10 times with purified H 2 O, then fresh buffer solution (without NADP + ) was introduced: consequently, the new solution did not contain any enzyme and any product that had accumulated was removed. Reinjection of NADP + (to 20 μM) caused an increase in current, 100 μA higher than the current (rate) before buffer exchange. This result demonstrated that product inhibition is an important contributor to the current decrease with time.
An electrode was also prepared by placing it in a stirred solution of FNR (15.7 μM) and ME (7 μM) in borate buffer (pH 8) for 2 h, then thoroughly rinsing with ultrapure H 2 O (Milli-Q, Millipore) to remove unbound proteins. The (FNR + ME)@ITO/Ti electrode (1 cm 2 ) was then placed in buffer solution (4 mL) containing 50 mM phosphate, 0.10 M KHCO 3 , 4 mM MgCl 2 , and 80 mM pyruvate at pH 7.5. Panel C shows the result of the chronoamperometry (−0.45 V vs SHE, stirring, 25°C) with the resulting electrode. NADP + was added to a final concentration of 20 μM and a sizeable reduction current (current density: 20 μA/cm 2 ) was obtained. The amount of FNR adsorbed in this experiment, obtained by peak integration, was 9.9 pmol/cm 2 : this quantity allowed us to determine an absolute rate on a per-FNR basis of  , under 100% CO 2 ), with 50 mM pyruvate (black), and after addition of 50 mM malate (blue). Scan rate = 1 mV s −1 .

ACS Catalysis
Research Article approximately 7 s −1 , which is instructive, although we lack information on what local FNR/ME ratio is optimal. The result obtained using a much lower cofactor concentration is shown in panel D. Chronoamperometry (−0.45 V vs SHE, stirring, 25°C) of a (FNR + ME)@ITO/Ti electrode (16 cm 2 ) prepared as for panel A was carried out in 50 mM phosphate, 0.10 M KHCO 3 , 4 mM MgCl 2 , and 20 mM pyruvate (pH 7.5, volume 4 mL). In this experiment, NADP + was injected to a final concentration of 1 μM. The current trace differs from panel A in that it only starts to decrease after 7 h, and more than 60% of the maximum current is retained after 24 h when, based on coulometry, a conversion of 31% (6210 TTN) has been achieved. The lower conversion/current decrease is again consistent with lower product inhibition. The fact that a reasonable rate is still achieved with such a low NADP + concentration highlights the practical possibilities for a (FNR + E2)@ITO/Ti electrode technology in cofactor recycling for organic synthesis. 8 Investigations were made to gauge the affinity of the twoenzyme system for CO 2 , noting that natural photosynthesis occurs in air containing approximately 0.04% CO 2 . Results are shown in Figure 4. Panel A shows a chronoamperometry  The 2%Mo:BiVO 4 /FeOOH photoanode oxidizes water and the electrons combine with holes produced in the single junction Si solar cell upon light absorption. The electrons are used by the (FNR + ME)@ITO/Ti electrode which continuously regenerates NADPH for use by malic enzyme. The "carrier" is pyruvate; the product is malate.

ACS Catalysis
Research Article experiment (stirred), conducted at −0.45 V versus SHE, 25°C , pH 7.5, in which a FNR@ITO/Ti electrode (62.3 pmol/ cm 2 ) was placed in a cell containing 0.20 M HEPES, 4 mM MgCl 2 , and 80 mM pyruvate. Initially, 100% argon was bubbled into the solution. After introducing NADP + (final concentration 20 μM), ME was added to a final concentration of 2.5 μM. After about 80 min, small aliquots of a saturated CO 2 stock solution were injected to increase the CO 2 concentration in stages, from 0.4 to 100%. Concurrently, the cell headspace gas composition was adjusted accordingly using two mass flow controllers connected to Ar and CO 2 cylinders. Each addition resulted in a sharp increase in reduction current which decreased to a new steady level. The area under the initial current burst was similar in all cases and attributable to the reduction of trace O 2 in the stock solution. An experiment was also carried out using cyclic voltammetry, allowing time for removal of O 2 , and similar results were obtained. A titration curve was prepared by plotting the normalized current increase against % CO 2 ( Figure 4B). The results showed that 50% of the maximum catalytic rate was attained with 4.0 ± 0.5% CO 2 (see Supporting Information and Figure S3).
The low overpotential needed to drive CO 2 reduction suggested the feasibility of constructing a device mimicking the essential features of photosynthesis by using H 2 O as the

ACS Catalysis
Research Article electron donor. We chose Mo-doped BiVO 4 as a photoanode because of its high light absorption, high photocurrent, and straightforward synthesis. 28 A fluorine-tin oxide electrode (1 cm 2 ) was coated with Mo-doped BiVO 4 (Mo 2%) and a layer of O 2 evolution catalyst, FeOOH (goethite), was added by photo-electrodeposition (ESI). Although the band gap of BiVO 4 (2.4 eV) is suited for visible light excitation, the conduction band potential is too positive to reduce NADP + . 29−31 This situation is clarified in Figure 5 which compares two voltammogramsthe reductive carboxylation of pyruvate to malate by the (FNR + ME)@ITO/Ti electrode (blue trace, using Pt as the counter) and water oxidation by the 2%Mo:BiVO 4 /FeOOH photoanode (using Pt as the counter) in the dark (black trace) and during illumination (red trace). It is clear that an additional bias of at least 0.4 V is required in order to drive the C3-C4 reaction. The situation resembles that of oxygenic photosynthesis in which two photosystems are required. Here, the problem was solved by adding a 1 V Sisolar cell (Figure S4A), the output voltage of which could be attenuated. Panel B of Figure 5 shows the result obtained when water photo-oxidation is driven by the photoanode coupled to the 1 V silicon solar cell (purple trace). Given the low overpotential requirement expected, the voltage was cut to 0.5 V by including an additional resistance ( Figures S4B and 5B blue trace). In both cases, the (FNR + ME)@ITO/Ti electrode could be driven using visible light and water as

ACS Catalysis
Research Article electron donors. Scheme 1 shows the setup used for the lightdriven reductive carboxylation of pyruvate.
Photoconversion experiments were carried out in a two compartment cell separated by a Nafion 115 membrane. The anode compartment (0.5 M borate, 4 mM MgCl 2 , pH 9) was initially flushed with Ar (for 10 min) and then sealed to allow quantification of photogenerated O 2 with an O 2 sensing electrode (see Supporting Information and Figure S5). 32 Illumination was provided by a 300 W Xe lamp equipped with a 420 nm UV cut-off filter (see Supporting Information for more details) and held at a distance of 6.5 cm from the anode compartment. The cathode compartment (volume 4 mL, 50 mM phosphate, 0.10 M KHCO 3 , 4 mM MgCl 2 , pH 7.5) was continuously flushed with high-purity CO 2 at a rate of 10 sccm in order to minimize additional current due to O 2 leaking in from the atmosphere. The cathode compartment was protected by a piece of Al foil, as the (FNR + ME)@ITO/ Ti cathode was unstable at high light intensity. Figure 6A shows the result obtained with a FNR@ITO/Ti cathode (14 cm 2 ), a 2%Mo:BiVO 4 /FeOOH photoanode (1 cm 2 ) and an auxiliary solar cell with the output capped at 0.5 V. As expected, no current was detected in the dark, whereas upon illumination a background current was observed, which was verified by gas chromatography (sample taken from the cathode compartment) to be due to H 2 evolution. After the current had reached a steady value, an aliquot of NADP + solution was added to the cathode compartment (final concentration 50 μM), and then after a further 30 minutes, ME was added (final concentration 2.5 μM). A higher concentration of the cofactor was used in the light-driven experiments to compensate for the lower current provided by the photo-anode. The current rapidly reached a maximum value, and then it was monitored for 23 h after which it had decreased to 47% of the original value. The malate concentrations measured by 1 H NMR (Supporting Information, Figures S6 and 6B) were 0 mM (0 h), 3.4 mM (8 h), and 14.8 mM (24 h) thus resulting in 18.5% conversion and a TTN of approximately 300. The latter value is 1−2 orders of magnitude higher than other light-driven cofactor regeneration systems (Table 1). 14,31,33 The Faradaic efficiency (charge used to produce malate/total charge passed) was approximately 70% after 24 h. In another experiment (see Supporting Information, Figure S7), the full 1 V of the solar cell was used: this resulted in 34% conversion and a TTN of 544. Tests carried out in different experiments showed that the decrease in photocurrent was due at least in part to the accumulation of malate which acts as a product inhibitor, consistent with the result shown in Figure 2 (Figure S8).
A more detailed experiment to quantify Faradaic efficiency is shown in Figure 6C. In this case, a 1 cm 2 photoanode and a 2 cm 2 cathode were used and all components apart from NADP + and ME were initially present. A bias of 0.5 V was applied and a small current appeared upon illumination, corresponding to H 2 evolution. Addition of NADP + (to 50 μM) produced a small increase in current (accompanied by an O 2 spike); ME was then added (to give 2.5 μM) and a large increase in current was observed for several minutes. Three interventions were made over a 6 h period, each signified with a blue square. In the first of these interventions (3 h after commencing the experiment), the light was switched off, the cathode compartment was washed with H 2 O, fresh buffer was added (without pyruvate, NADP + , ME), and a bias of 0.8 V was applied. Restoring illumination revealed a much larger background current, as expected because H 2 can now be produced at a high rate. Re-addition of NADP + and pyruvate (N + and P) to achieve the same concentrations as before resulted in an increase in photocurrent. The procedure was repeated with applied biases of 1 and 0.4 V: for each condition, the faradaic efficiency was calculated from the total photocurrent and that observed without NADP + and pyruvate. Values were: 75 ± 8% at 0.5 V, 50 ± 6% at 0.8 V, 33 ± 10% at 1 V. No significant current increase was observed when 0.4 V was applied.
In conclusion, direct, reductive CO 2 incorporation into organic molecules becomes highly efficient when two cooperating catalysts, one recycling transferable "hydride" using electrons and the other using the hydride to perform carboxylation, are confined within the nanopores of an electrode. While the system uses enzymes and has little or no practical use, enzymes provide the ultimate benchmarks for rates, efficiencies, and selectivities, and their inspirational values cannot be ignored. Cyclic voltammetry demonstrates that the two enzymes function in unison as a reversible electrocatalyst for the C3-to-C4 interconversion and the reaction is easily detectable at low (<1%) CO 2 levels. The thermodynamics are favorable with respect to H 2 evolution and a biomimetic artificial photosynthesis device based on the (FNR + ME)@ITO/Ti electrode evolves O 2 and reduces CO 2 with 70% Faradaic efficiency at minimal (0.5 V) bias.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acscatal.9b03532.
Material and methods, control CV of a FNR@ITO/Ti electrode with no ME, spectra ( 1 H-NMR) for CO 2 reduction, CV of (FNR+ME)@ITO/Ti at different CO 2 percentages (0.4% to 100%), current/voltage (i/V) curve of the 1 V silicon solar cell and chronopotentiometry of the solar cell, continuous detection of photogenic O 2 , 1 H-NMR spectra obtained for lightdriven CO 2 reduction, light-driven pyruvate reductive carboxylation to malate, and addition of malic enzyme and ferredoxin NADP + -reductase after 20 hours of lightdriven reductive carboxylation of pyruvate to malate (PDF) The TTN is defined as moles of product formed/moles of cofactor added at the beginning of the experiment.