Hydrogenation of CO₂ at ambient pressure catalyzed by a highly active thermostable biocatalyst

Conversion of CO₂ into other chemicals has gained increased attention in recent times primarily caused by the problem of increasing atmospheric CO₂ concentrations caused by anthropogenic emissions and the need for sustainable technologies for energy conversion and storage, carbon sequestration, as well as chemical production [1]. Among the possible reactions, direct hydrogenation of CO₂ to formic acid or the conjugate base formate is of special interest. Formate can serve as substrate for production of other carbon containing products by chemical or biotechnological processes or utilized directly for leather processing, food preservative, or as deicing agent [2, 3]. As substrate for biological fermentations, it overcomes mass transfer limitations of gaseous substrates such as synthesis gas [4]. In addition, formate can be utilized as hydrogen storage system coupled to downstream hydrogen release or serve directly as electron source for a fuel cell device [5, 6]. Replacing fossil fuels as energy carrier requires alternatives that combine sustainable production, high volumetric energy density, easy and fast refueling for mobile applications, and preferably low risk of hazard. Molecular hydrogen (H₂) has been considered as promising alternative technologies either for mobile applications or to store excess electricity generated at off-peak times by wind or solar power. However, the practical application is struggling because of the low volumetric energy density and the explosion hazard when stored in large amounts [7, 8]. One way to overcome these limitations is the transient conversion of H₂ into other chemicals, for example so-called liquid organic hydrogen carriers (LOHC), with elevated volumetric energy density. Among the multiple technologies envisioned for overcoming this problem, CO₂ hydrogenation to formic acid represents a promising solution [7, 9]. Formic acid combines its liquid nature with low toxicity and easy handling, prerequisites for the storage and further utilization.

The thermodynamic stability of CO₂ makes hydrogenation a challenging reaction. Increased attention to this conversion in the last years has led to important advances. Homogenous or heterogenous chemical catalysts have been developed for CO₂ hydrogenation and studied under multiple conditions [3, 9, 10]. However, many solutions suffer from low turnover frequencies (TOF), dependency on high pressure or temperature, or very expensive additives emphasizing the need for the discovery of better catalysts and a fundamental understanding of the process.

In contrast to the large number of chemical catalysts, recently the first biocatalyst for direct and reversible hydrogenation of CO₂ has been described. It operates under ambient pressures and temperatures with a TOF of 101,600 h⁻¹ which is significantly better than chemical catalysts [10–12]. The enzyme, named hydrogen-dependent CO₂ reductase (HDCR), consists of four subunits with a [FeFe] hydrogenase, a molybdenum/tungsten-bis pyranopterin guanosine dinucleotide (bis PGD) cofactor containing formate dehydrogenase and iron–sulfur cluster containing electron transferring subunits (Fig. 1). In contrast to the membrane-bound formate hydrogenlyase (FHL) of E. coli, a long well-known enzyme that also connects the electron carriers formate/CO₂ and H₂, the HDCR is a soluble cytoplasmic enzyme complex [13–15]. FHL has the physiological function of formate oxidation, explaining its strong bias in catalyzing this direction of the reaction, whereas the HDCR of A. woodii showed very similar catalytic rates in both directions [11, 16]. By using catalytic protein film electrochemistry, the hydrogenase subunit was characterized as first completely CO-tolerant [FeFe] hydrogenase with the highest apparent affinity for H₂ ever reported for a [FeFe] hydrogenase [17]. Otherwise, the enzyme showed the typical phenotype of (standard) [FeFe] hydrogenases. The formate dehydrogenase subunit shows, based on the primary sequence, high similarity to other metal-containing formate dehydrogenases, raising the question why this enzyme is very active towards CO₂ reduction, whereas many closely related enzymes such as the formate dehydrogenase subunit of the FHL are strongly biased for formate oxidation [18]. In addition to the catalytic properties, the HDCR showed the surprising ability to polymerize into large filamentous structures (around 0.2 µm in length) [19].

The exceptional TOF of the first described HDCR together with the questions raised by this discovery led to the motivation to search for other enzymes of this class with potentially superior catalytic properties and/or increased stability. We focused our search on enzymes from thermophilic sources. The increased temperature should be connected with increased conversions rates, and the adaptation to high temperatures could result in increased stability of the enzyme, both desirable properties for industrial application. So far, the number of known HDCR gene clusters is very limited, especially in thermophilic organisms. However, we could identify a putative gene cluster coding for a thermophilic HDCR in the bacterium Thermoanaerobacter kivui. This organism belongs to the group of the acetogenic bacteria as A. woodii but has a growth optimum at 66 °C [20–23]. T. kivui can utilize among others H₂ + CO₂, CO, formate, pyruvate or glucose as growth substrate. This organism utilizes the reductive acetyl-CoA pathway (Wood–Ljungdahl pathway) for reduction of two molecules of CO₂ to acetyl-CoA that is converted to acetate as major end product [20, 21]. In this study, we were successful in isolating the HDCR from T. kivui, demonstrating that this first thermostable enzyme for direct hydrogenation of CO₂ shows remarkable catalytic properties with TOFs of 9,556,000 h⁻¹ at 1 bar H₂ + CO₂ at 60 °C. Even at 30 °C, the enzyme from T. kivui shows an 18-fold higher conversion rate compared to the only other characterized HDCR from A. woodii. The enzymes seem to be responsible for the first step of the reductive acetyl-CoA pathway of T. kivui catalyzing the essential step of CO₂ reduction to formate.

To screen for novel HDCRs, we used the gene cluster of the HDCR from A. woodii as query to screen available genome sequence data for novel HDCRs. This gene cluster consists of 7 genes containing two isogenes coding for the formate dehydrogenase subunit (fdhF1 and fdhF2), the putative formate dehydrogenase maturation protein fdhD, the [FeFe] hydrogenase hydA2, and three putative electron transferring proteins (hycB1, hycB2, hycB3) (Fig. 2a). Homologs of each gene can be found individually in a multitude of organism making identification of genuine HDCR gene clusters more difficult. Therefore, we used the amino acid sequence of fdhF1, hydA2, hycB1, and hycB3 from A. woodii to search for organisms encoding these proteins in their genome using MultiGeneBlast and manually checked these candidates to have the corresponding genes organized in an arrangement resembling the one for A. woodii. We identified putative HDCR gene clusters in 25 organisms (Additional file 1: Table S1). A complete HDCR gene cluster was identified in two thermophilic organisms (Symbiobacterium thermophilum and T. kivui). S. thermophilum can only grow in coculture with Geobacillus sp. [24], we therefore focused on T. kivui, a Gram-positive acetogenic bacterium that can grow with H₂ + CO₂ as substrates and has a growth optimum of 66 °C [20]. The putative HDCR gene cluster of this organism consists of only 5 genes (Fig. 2a). The two missing genes compared to A. woodii are isogenes of the formate dehydrogenase subunit and the corresponding electron transferring Hyc subunit that potentially arose from gene duplication resulting in two variants of the formate dehydrogenase subunit, one with cysteine and one with selenocysteine at the same position, otherwise showing an identity of 80% on the level of the amino acid sequence. In T. kivui, only one gene coding for the formate dehydrogenase is present that, based on the primary sequence, does not contain selenocysteine as FhdF2 of A. woodii. Two genes hycB1 and hycB2 coding for small putative electron transferring subunits separate the formate dehydrogenase gene from the putative hydrogenase gene hydA2. At the end of the gene cluster, fdhD codes for a putative maturation protein followed by a putative formate transporter fdhC.

To verify the presence of a functional HDCR, T. kivui was grown with the substrates H₂ + CO₂, glucose, formate or pyruvate. Using cell free extracts of these cells, it could be demonstrated that hydrogen was indeed produced from formate as substrate with specific activities up to 5 µmol H₂ min⁻¹ mg⁻¹ (specific hydrogen productivity of 300 mmol H₂ h⁻¹ g⁻¹) with cell free extracts from cells grown on formate or pyruvate showing the highest specific activities (Fig. 2b). To isolate the HDCR from T. kivui, the organism was grown with pyruvate as substrate because this substrate led to the highest specific hydrogen productivity combined with higher cell yields compared to formate grown cells, both requirements to achieve high enzyme yields for subsequent enzyme characterization and applications. Cells were grown to the late exponential growth phase, harvested and disrupted with a French pressure cell press. Membranes were removed by ultracentrifugation, and the HDCR complex was isolated using anoxic conditions from the cytoplasm by hydrophobic interaction chromatography on Phenyl-Sepharose followed by ion exchange chromatography on Q-Sepharose and gel filtration on Superose 6. Formate oxidation coupled to reduction of methylviologen was used as assay to identify formate dehydrogenase activity in all purification steps. This procedure yielded an apparently homogeneous preparation and the enzyme was purified 55-fold compared to the cell free extract. The isolated enzyme was separated by SDS polyacrylamide gel electrophoresis yielding four different subunits (Fig. 2c). LC–MS/MS identified these subunits as being indeed encoded by the genes of the putative HDCR gene cluster as described before (TKV_c19950–TKV_c19990). The isolated enzyme contained the subunits FdhF, HydA2, HycB3, and HycB4. FdhD and FdhC could not be identified to be part of the enzyme. The calculated cumulative molecular mass of the enzyme complex is 177.3 kDa based on a stoichiometry of α₁β₁γ₁δ₁. We used ICP-MS to analyze the metal content of the purified enzyme. Based on the primary sequence, the predicted cofactor content of the entire complex is 12 [4Fe4S] clusters, the di-iron center of the [FeFe] hydrogenase subunit and the bis PGD cofactor of the formate dehydrogenase subunit. We detected 42 mol Fe, < 0.05 mol Mo, 0.34 mol W, 4.5 Zn per mol of enzyme. The iron content is lower than the expected 50 mol Fe per mol of enzyme indicating that during protein purification, part of the cofactors could have been lost. Interestingly, molybdenum could not be detected in the enzyme but tungsten. In contrast to the HDCR from A. woodii, the enzyme from T. kivui seems to contain a formate dehydrogenase of the tungstoenzyme family and not molybdenum bound to the bis PGD cofactor.

Next, we wanted to analyze the coupling between hydrogen oxidation and CO₂ reduction therefore incubated the isolated enzyme under an atmosphere of H₂ + CO₂ (80:20 [v:v], 1 bar) at 60 °C. Samples were taken under anoxic conditions and the formate concentration was determined by an enzymatic assay. Formate was produced with an initial turnover frequency of 9,565,000 h⁻¹ (specific activity of 900 µmol formate min⁻¹ mg⁻¹) (Fig. 3a). This rate outcompetes the enzyme from A. woodii by a factor of 96 (Additional file 1: Table S2) and is roughly 1200 times faster than chemical catalysts under comparable conditions [25] and almost ten times faster than the best-known chemical catalysts (p(H₂/CO₂) 30 bar/10 bar; 120 °C) [9, 11, 26]. The reverse reaction, oxidation of formate to CO₂ coupled to H₂ evolution was catalyzed in this assay with a TOF of 4,814,000 h⁻¹ (450 µmol H₂ min⁻¹ mg⁻¹) (Fig. 3c). Noticeably, we observed using this assay that the HDCR did not show an expected substrate proportionality. By decreasing the amount of enzyme below 10 µg, the specific activity further increased but the catalyst sharply lost the activity below a concentration of 3 µg ml⁻¹. Therefore, for stability reasons, we used increased enzyme concentrations for further measurements.

To analyze the catalytic properties of part activities of the HDCR, we utilized methylviologen as artificial electron acceptor and formate or H₂ as electron donor with the isolated HDCR as catalyst. Reduction of methylviologen was followed with UV/Vis spectrophotometry at 604 nm. Correlation of enzyme activity and hydrogen concentration followed a typical Michaelis–Menten kinetic with an apparent K_M for H₂ of 0.13 mM (Fig. 4a). This value is lower than K_M values reported for other typical [FeFe] hydrogenases [17, 27]. The specific activity for H₂ oxidation was 14,400 µmol H₂ min⁻¹ mg⁻¹ (TOF 42,266 s⁻¹). The formate oxidation showed an apparent Michaelis–Menten kinetic at low formate with an apparent K_M of 0.55 mM for formate and a specific activity for formate oxidation of 455 µmol formate min⁻¹ mg⁻¹ (TOF 1335 s⁻¹) (Fig. 4b). This result does not apply to high formate concentrations as will be shown later. The most striking feature of the novel HDCR isolated from T. kivui is its thermostability. Higher temperatures are typically connected to higher catalytic rates and provide process advantages especially considering whole cell catalysis. Higher temperatures reduce the energy costs necessary to cool down the fermentation and decrease the risk of contamination. The thermostability of the isolated HDCR was tested first using the artificial electron acceptor methylviologen to study the part reactions separately. The hydrogenase activity showed a maximum at 70 °C and decreased drastically above this temperature (Fig. 5a). Decreasing the temperature had a less severe effect with almost 50% of the maximum activity still present at 30 °C. The formate dehydrogenase activity also showed a maximum at 70 °C but a steeper decrease in activity at lower temperatures with only 10% of the maximum activity present at 30 °C (Fig. 5b). The coupled activity, hydrogen-dependent reduction of CO₂ showed the highest TOF at 70 °C (11,623,000 h⁻¹) (Fig. 3b). Surprisingly, at 30 °C, the TOF was still 1,856,000 h⁻¹. This rate is 18 times faster than the one observed for the mesophilic A. woodii enzyme with an optimum at 30 °C. The same was true for the reverse reaction, H₂ production from formate (Fig. 3d). Next, the optimum pH range for catalysis was determined. The activity profiles differed for the part activities with a pH optimum for hydrogen oxidation at a pH of 8 and an optimum of pH 7 for the formate dehydrogenase (Fig. 5c, d). However, at pH 7, the hydrogenase activity is only at 50% of its maximum but in absolute terms still one order of magnitude faster than the formate dehydrogenase activity thus not limiting the coupled reaction.

The HDCR of T. kivui is not only the fastest catalyst known for CO₂ hydrogenation, but as shown before also for the reverse reaction, hydrogen evolution from formate. We wanted to analyze the maximum formate concentrations that can be used before inhibitory effects reduce the catalytic rates. We tested concentrations up to 1M of sodium formate. To our surprise, increasing the substrate concentrations did not only show no inhibition of the enzyme but increased the activity significantly (Additional file 1: Figure S1). The tested substrate concentrations are over 2 orders of magnitude above the K_M value; therefore, the increase in activity must result from other effects, potentially a second low affinity formate binding site. At 1M sodium formate, hydrogen was produced with an activity of 650 µmol H₂ min⁻¹ mg⁻¹ (TOF 6,940,000 h⁻¹). As a control if the effect stems from the increasing salt concentration, we tested 50 mM sodium formate with 950 mM sodium chloride but this did not result in increased activity. We also observed a sharply increasing specific activity when using enzyme concentrations below 6 µg ml⁻¹ as described before at constant formate concentrations of 150 mM. The maximum activity measured at 3 µg ml⁻¹ was 930 µmol H₂ min⁻¹ mg⁻¹ (TOF 9,892,000 h⁻¹). However, decreasing the concentration further resulted in an immediate total loss of activity (data not shown). We can only speculate that this unusual behavior of the enzyme is connected to the quaternary structure. For the A. woodii enzyme, it has been described that the enzyme polymerizes in the presence of divalent cations in vitro into filamentous structures of up to 0.2 µm in length and that this behavior is connected to the enzyme activity. Polymerization of the T. kivui enzyme into high molecular weight structures was indicated by the observed elution volume of the isolated HDCR on the gelfiltration resembling the behavior of the A. woodii enzyme.

For application of the isolated enzyme, the stability is a critical factor. The enzyme stability was tested by incubating the isolated enzyme at two different temperatures and determining the specific activity over time. At 60 °C, the activity dropped to 50% of the initial activity after 9 h (Fig. 6b). Afterwards, the decrease was slower with 25% of the activity still retained after 34 h. At room temperature, 54% of the initial activity was still present after over 52 days. This is worth paying attention since as reported before, at 30 °C the HDCR of T. kivui is still much faster than the best-known catalyst at this temperature so far. For storage, we tested storing the enzyme at 4 °C and using repeated freezing (− 21 °C) and thawing cycles. The enzyme was not strongly affected by freezing and retained almost 80% of the initial activity after 52 days and 8 freezing/thawing cycles (Additional file 1: Figure S2). Storage at 4 °C led to a decrease to 50% of the activity after 27 days.

In summary, the HDCR from T. kivui is the first description of a thermostable member of this new enzyme class of CO₂ reductases that are directly coupled to hydrogen oxidation. The enzyme was isolated from a thermophilic organism that has a growth optimum at 66 °C and showed a similar temperature optimum in vitro of 70 °C. The enzyme is a complex of four subunits, a [FeFe] hydrogenase, a formate dehydrogenase, and two putatively electron transferring subunits connecting the two redox reactions. The enzyme displayed remarkable catalytic properties with a TOF for formate formation from hydrogen and CO₂ of 9,556,000 h⁻¹ and 9,892,000 h⁻¹ for the reverse reaction. This is, in both directions, significantly faster than state-of-the art chemical catalysts while needing only ambient pressure and moderate temperatures. The novel HDCR is also superior to biological catalysts described for this reaction and the thermostability as well as the broad temperature range for activity offers more flexibility for process development (Additional file 1: Table S2). Given the important problem of hydrogen storage, the reversible catalyst described in this study offers a promising step towards the efficient conversion of hydrogen into the liquid organic hydrogen carrier formic acid/formate and the subsequent release of the gas for the following utilization as energy source. The characterization of more highly active CO₂ reductases will also help to attain a better understanding of the factors required for efficient catalysis. Noticeably, the highly active HDCR enzymes contain a formate dehydrogenase subunit that, based on the primary sequence, does not show an obvious difference from other metal-dependent formate dehydrogenases of the DMSO reductase family of which most show only low activity towards CO₂ reduction and have a strong bias for formate oxidation. Comparing the two HDCRs described so far, it strikes the attention that the more active enzyme from T. kivui does not contain selenocysteine as does the less active A. woodii enzyme. The presence of selenocysteine is typically connected to increased catalytic activities due to the lower pK_a value and reduction potential of the selenol group compared to the thiol group of cysteine [28, 29]. However, the molybdenum bound to the bis PGD cofactor has been replaced by tungsten in T. kivui, a metal often found in highly active CO₂ reductases and especially in enzymes operating at increased temperatures [30, 31]. To shed more light on the catalysis of the HDCR enzymes a structure of a complete HDCR or the formate dehydrogenase subunit is urgently needed to reveal the architecture of the active site and to identify the factors determining its high activity. This could also help to design more active bio-inspired chemical catalysts, since utilization of CO₂ as substrate is at the moment obstructed by the availability of efficient catalysts.