1 The Relentless Rise of Carbon Dioxide

In 2018 alone, more than 36Gt of CO2 [1] were dumped into the atmosphere as waste material from fossil resources-based energy and chemical industries! In that year, the global atmospheric CO2 concentration reached an annual average value of 407 ppm, an increase of 150% since pre-industrial times (277 ppm in 1750) (Fig. 1) [1]. Yet, new records are being set, and a monthly average of 416 ppm was already observed this March 2020 [2]. This ever-increasing atmospheric CO2 concentration is causing large and unpredictable impacts on the Earth climate, due to the CO2 significant greenhouse effect, besides being responsible for the ocean acidification, with consequent huge impacts in our daily lives and in all forms of life.

Fig. 1
figure 1

Source NOAA Climate.gov, based on EPICA Dome C data provided by NOAA NCEI Paleoclimatology Program (https://www.climate.gov/news-features/understanding-climate/climate-change-atmospheric-carbon-dioxide)

Relentless rise of carbon dioxide. Global atmospheric CO2 concentrations in parts per million (ppm) for the past 800,000 years. The peaks and valleys track ice ages (low CO2) and warmer interglacials (higher CO2). During these cycles, CO2 was never higher than 300 ppm. In 2018, it reached 407.4 ppm. On the geologic time scale, the increase (blue dashed line) looks virtually instantaneous

Atmospheric CO2 concentration results from the balance between CO2 emission and uptake [3]. CO2 is emitted from human activities, such as fossil fuel combustion and oxidation from other energy and industrial processes (10.0Gt of carbon in 2018 [1]) and deliberate activities on land, mainly deforestation (1.5Gt of carbon in 2018 [1]), as well as, from natural processes, such as volcanic eruptions and biological emissions. On the other plate of the scale, the small “CO2 sinks” are mainly provided by physical and biological processes in oceans (2.6Gt of carbon in 2018 [1]) and land (3.5Gt of carbon in 2018 [1]). To break this largely unfavourable imbalance (more than 5Gt in 2018), we must actively reduce the CO2 emissions and develop new and more efficient “CO2 sinks”—new individual actions and political decisions are needed, as reviewed by Seixas and Ferreira in this book [3].

Until recently, the debate often focused only on “passive CO2 mitigation”, searching for strategies for CO2 capture and sequestration (CCS). Instead, we should be looking at the opportunities for the energy and chemical industries provided by exploiting this novel and huge carbon feedstock (Fig. 2), such as (a) storage of “intermittent” renewable energy sources (RES) (wind, solar and hydropower energy, which are now rapidly growing and becoming economically viable), (b) conversion of RES-derived electricity into fuels (mainly for mobility and transport sector, in particular aviation and heavy freight over long distances, the major polluters), (c) production of added-value compounds (VC) and feedstock chemicals for making all the modern-world chemical commodities (from bulk chemicals to plastics, fertilisers and even pharmaceuticals). Regarding atmospheric CO2 reduction, points (a) and (b) (energy industry) are of major relevance, as the different scales of energy and chemical industries impede the VC production to function as a quantitative “sink” for the massive fossil fuels-dependent CO2 emissions. Together, these three axes, storage/conversion/production, will certainly provide a straightforward way to actively reduce the CO2 emissions, while actively consuming the CO2 already released—”two-in-one solution”.

Fig. 2
figure 2

Directing CO2 into the axes storage/conversion/production through the formate route. (icons from www.icons8.com)

But, how to direct CO2 into the storage/conversion/production axes? Formic acid/formateFootnote 1 offers key advantages (Fig. 2)!

2 Formic Acid—The Stepping Stone Towards Carbon Dioxide Utilisation

Formic acid was identified in fifteenth century as an acidic vapour in ant hills, from where its name derivates—”formica”, the Latin word for ant. It was first synthesised only in the nineteenth century from hydrocyanic acid by the famous French chemist and physicist Joseph Gay-Lussac and also from carbon monoxide by another French chemist, Marcellin Berthelot. However, formic acid received little industrial attention until the last quarter of twentieth century, when it started to be used as a preservative and antibacterial in livestock feed due to its low toxicity (LD50 of 1.8 g/kg). More recently, formic acid regained a new interest, due to some features that are key to the longed-for “post-fossil era” (Fig. 2).

  1. (a)

    Formic acid is a stable product that can be formed by the “simple” two-electron reduction of CO2 (Eq. 1), what, noteworthy, resulted in a considerable attention to the electrochemical CO2 reduction in the last decade (see Sect. 3.).

  2. (b)

    Besides formic acid, also carbon monoxide is a stable product of the two-electron CO2 reduction (Eq. 2). However, its high toxicity, low solubility and low mass transfer rate make the carbon monoxide subsequent utilisation challenging. In contrast, formic acid is a highly soluble and stable liquid, easy to store and transport.

  3. (c)

    Formic acid is also not explosive, what represents an important advantage relatively to dihydrogen, an ideal “clean” fuel (see below).

  4. (d)

    Formic acid is already used as a “building block” in chemical industry.

  5. (e)

    Formic acid can be a substrate for further reduction to a carbon-based fuel, methanol and methane, what might not be the most obvious option regarding CO2 consumption.

  6. (f)

    Formic acid, formed from CO2 and dihydrogen (Eq. 3), can be used as a “storage form” of dihydrogen, an ideal “clean” fuel (potentially zero contribution to the global carbon cycle, with a high gravimetric energy density) [4,5,6,7,8,9,10,11,12,13]. Although formic acid is not a perfect dihydrogen “storage medium”, due to its relatively small hydrogen content (4.4%(m/m) or 5.3%(m/v)), it is currently still one of the best options to circumvent the technical difficulties associated with dihydrogen handling, storage and transport. For this purpose, formic acid produced by CO2 hydrogenation or any other approach is converted back to dihydrogen when needed.

  7. (g)

    Moreover, formic acid fuel cells are being the centre of a renewed interest [14,15,16,17,18].

  8. (h)

    From a biotechnological point of view, formate can be both produced and assimilated by many natural and biotechnologically engineered organisms and, unlike dihydrogen (that is “just” oxidised to form reducing power), act as a carbon source for “formatotrophic” organisms, thus, enabling considerably higher biomass formation and VC and fuels production yields [19].

$${\text{CO}}_{ 2} + {\text{ 2e}}^{ - } + {\text{ 2H}}^{ + } \rightleftharpoons \,{\text{HCOOH}}$$
$${\text{CO}}_{ 2} + {\text{ 2e}}^{ - } + {\text{ 2H}}^{ + } \rightleftharpoons \,{\text{CO}} + {\text{H}}_{ 2} {\text{O}}$$
$${\text{CO}}_{ 2} + {\text{ H}}_{ 2} \rightleftharpoons \,{\text{HCOOH}}$$

3 How to Convert Carbon Dioxide to Formic Acid/Formate?—The Chemical Way

CO2 is a kinetically and thermodynamically stable molecule, with a high negative value of the reduction potential of the CO2/HCOOH pair (highly pH dependent), what makes its activation and reduction a difficult task [8]. Hence, perhaps the first answer that comes to mind to reduce CO2 to formate is: electrochemically [20,21,22,23,24,25,26,27,28,29,30,31,32,33]. However, the feasibility—meaning essentially the economic viability—of this process, that is currently the centre of intense research, depends on the Faradaic efficiency and energetic efficiency of CO2 reduction (avoiding high electrochemical overpotentials) and on the rate and selectivity (purity) of formate production. The other “cost” to be considered is obviously the environmental one, and this depends on the use of a RES-derived electricity and on the sustainability of the electrodes (composition and durability).

The second answer is probably going to be photoreduction, which is the most straightforward way to use a RES to convert CO2. Solar energy (photogenerated electrons) can be used to drive chemical reactions, and this solar-to-chemical energy conversion followed by storage in the form of chemical bonds is generally called “artificial photosynthesis” (as it is a mimic of photosynthetic process used by living organism to fix CO2). The progress in this field has been quite remarkable, and several highly efficient and promising systems have been developed for CO2 reduction (as well as water oxidation and hydrogen evolution), and formic acid can be produced with high rates and selectivity [34,35,36,37,38,39,40,41,42,43,44]. However, some problems have yet to be solved. In a very simplified way, artificial photosynthesis needs two fundamental components: an ideal light absorber/photosensitiser (for light harvest, charge separation and charge transfer) and an ideal catalyst (with high intrinsic activity and stability and low overpotential). Therefore, the heterogenisation of the molecular catalysts and engineering of applicable devices are the main challenges towards the development of effective artificial photosynthesis devices (practical problems, such as density and exposure of the catalyst active sites, conductivity, mass transport and stability of the catalyst-derived material or electrode, all make the catalyst intrinsic activity and efficiency quite different from the device performance numbers). In addition, scale-up feasibility and whole device long-term stability (also associated with the costs of using expensive high-purity semiconductors to achieve high efficiency) have to be attained. Nevertheless, artificial photosynthesis devices might become economically viable sooner than many anticipate: considering the energy consumption forecasted for 2050, the future solar energy devices only need a ≈ 10% solar-to-fuel efficiency (already achievable in proof of concept devices!) if 1% of the Earth’s surface is covered [44].

Formic acid can also be produced chemically from CO2 and dihydrogen. This CO2 hydrogenation is just the thermal overall CO2 reduction by dihydrogen using molecular catalysts (Eq. 3), as an alternative to direct electrochemical or photoelectrochemical reduction of CO2 [510, 1213, 45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71]. Hence, here, it is the rational design of the catalytic systems (efficiency and selectivity) that must be attained and the systems that have been developed to date exhibit selectivity and yield lower than desirable, besides requiring a high temperature and/or high pressure. Dihydrogen is a “clean” fuel (potentially zero contribution to the global carbon cycle), but its real environmental impact depends on how it is produce. The industrial production of dihydrogen (primarily from methane) requires harsh temperatures and emits as much CO2 into the atmosphere as natural gas burning [72]. To be environmentally friendly, dihydrogen must be produced by electrolysis of water using a RES and selected heterogeneous or homogeneous catalysts or biological systems [73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93]. As noted above, besides producing formic acid itself, CO2 hydrogenation is thought as a relevant way to storage dihydrogen. Therefore, also the reversible interconversion of formic acid to CO2 and dihydrogen must be carefully considered.

4 How to Convert Carbon Dioxide to Formic Acid/Formate?—Exploiting the Power of Formate Dehydrogenases (Enzymes for Solving Humankind’s Problems)

4.1 The Biochemical Way

In contrast to purely physicochemical, biological processes are substrate and product-specific (life requires a well-defined metabolism) and occur under truly “green”, sustainable conditions, at ambient temperature and pressure and close to neutral pH. Biological catalysts—enzymes—offer selectivity and specificity, coupled with high specific activity (in terms of active sites) and maximal rate (under the respective cellular context). Enzymes have evolved to become perfect catalystsFootnote 2, comprising (a) specific surface patches to establish contact with specific biomolecules (for cell localisation, integration into metabolic complexes, crosstalk or simple electron transfer), (b) channels where only (or mainly) the correct substrates come in and well-determined products come out, as well as, (c) highly defined active sites, assembled to promote the formation of key transition states and intermediates and, thus, lower the reaction energy barriers and energy loss. For that, active sites are built with precise steric features, electrostatic and hydrogen bonding interactions, fine-tuned reduction potentials and pKa values and optimised (and often synchronised) electron and proton transfer paths. The power and efficiency of biological catalysis is such that enzymes cascades are the cornerstone of all metabolic pathways that sustain life on Earth. Hence, there is a growing interest in making use of all the advantages the “biochemical way” can provide. Numerous hybrid systems have been (are being) designed to merge the best of the two worlds—chemical and biochemical—and the CO2 reduction to formate is no exception.

4.2 Formate Dehydrogenases—Enzymatic Machineries

To interconvert CO2 and formate, living organisms use formate dehydrogenase (FDH) enzymes. FDHs are a heterogeneous and broadly distributed group of enzymes that catalyse the reversible two-electron interconversion of formate and CO2 (Eq. 1) [94,95,96,97,98,99,100,101]. These enzymes evolved to take part in diverse metabolic pathways, being used by some prokaryotic organism to fix (reduce) CO2 into formate, while other prokaryotes use FDHs to derive energy, by coupling the formate oxidation (which has a very low reduction potential value, Eº′(CO2/HCOO) = −0.43 V) to the reduction of several terminal electron acceptors; FDHs are also broadly used by both prokaryotes and eukaryotes in C1 metabolism.

FDHs can be divided into two major classes, based on their cofactor content and the consequent chemical strategy used to carry out the formate/CO2 interconversion. One class comprises FDHs that have no metal ions or other redox-active centres—the metal-independent FDHs class [102,103,104,105,106,107,108,109]. These enzymes are widespread, being found in bacteria, yeasts, fungi and plants, are all (as far as is known) NAD-dependent and belong to the D-specific dehydrogenases of 2-oxyacids family. The other class—the metal-dependent FDHs class Footnote 3—comprises only prokaryotic enzymes that hold different redox-active centres (Table 1) and whose active site harbours one molybdenum or one tungsten centre (molybdenum-containing FDH (Mo-FDH) or tungsten-containing FDH (W-FDH), respectively) [94,95,96,97,98,99,100,101, 110,111,112].

Table 1 Summary of the features of some representative formate dehydrogenases and N-formyl-methanofuran dehydrogenases

4.2.1 The Metal-Independent Formate Dehydrogenases

Comparatively, the metal-independent FDHs are quite simple enzymes, generally forming homodimers, containing a NAD(H) and a formate-binding pockets in a close vicinity of each other (Fig. 3) [102,103,104,105,106,107,108,109]. The formate-binding site harbours a conserved arginine and asparagine residues, while an aspartate and serine residues make contacts to the nicotinamide ring, with another arginine residue binding the phosphate moiety linker of NAD(H).

Fig. 3
figure 3

C. boidinii formate dehydrogenase. A Three-dimensional structure view of the homodimer. B Arrangement of NAD and azide shown in the same orientation (but not same scale) as in (A). C Enzyme active site, with azide and NAD bound. The structures shown are based on the PDB file 5DN9 [109] (\(\alpha\) helices and \(\beta\) sheets are shown in red and cyan, respectively)

4.2.2 The Metal-Dependent Formate Dehydrogenases

Because the metal-dependent FDHs are involved in diverse metabolic pathways (energy and C1 metabolism), for which different “interfaces” are needed, this class is extraordinarily heterogeneous, comprising enzymes with diverse redox-active centres, such as iron–sulfur centres (Fe/S), haems and flavins, besides the characteristic molybdenum or tungsten active sites, organised in different subunit compositions and quaternary structures (Table 1) [94,95,96,97,98,99,100,101, 110,111,112]. This structural diversity is well exemplified by Escherichia coli, that expresses one “simple” monomeric cytoplasmatic enzyme, containing only the molybdenum centre and one [4Fe–4S] centre (the FDH H; Fig. 4) [113,114,115,116,117], and two “complex” heteromeric ((αβγ)3) membrane-bound respiratory enzymes that harbour seven additional redox-active centres ([4Fe–4S] centres and b-type haems) in addition to the molybdenum centre (the FDH N [118,119,120] (Fig. 5) and FDH O [121,122,123]). Also, the sulfate-reducing bacteria of the Desulfovibrio genus contain diverse Mo-FDHs and W-FDHs [124,125,126,127,128,129], such as the heterodimeric (αβ) periplasmatic W-FDH of D. gigas [130] or D. vulgaris [129, 131, 132], with “only” four [4Fe–4S] centres and one tungsten centre (Fig. 6) [133, 134], or the more “complex” heteromeric (αβγ) Mo-FDH of D. desulfuricans [135,136,137] or D. vulgaris [129, 131] that contains eight redox-active centres ([4Fe–4S] centres and c-type haems) in addition to the molybdenum centre. Remarkably, the overall protein fold of the molybdenum—and tungsten-containing subunits, including the arrangement of Fe/S centre, is highly conservedFootnote 4 [116, 117, 119, 130, 132, 138,139,140].

Fig. 4
figure 4

E. coli formate dehydrogenase H. A Three-dimensional structure view. B Arrangement of the redox-active centres shown in the same orientation (but not same scale) as in (A). C Molybdenum catalytic centre of oxidised enzyme. D Molybdenum catalytic centre of reduced enzyme as suggested by Boyington et al. in 1997 [116]. E Molybdenum catalytic centre of reduced enzyme as suggested by Raaijmakers and Romão in 2006 [117]. The structures shown are based on the PDB files 1FDO (A, B, C), 1AA6 (D) [116] and 2IV2 (E) [117] (\(\alpha\) helices and \(\beta\) sheets are shown in red and cyan, respectively)

Fig. 5
figure 5

E. coli formate dehydrogenase N. A Three-dimensional structure view. B Arrangement of the redox-active centres shown in the same orientation (but not same scale) as in (A). C Molybdenum catalytic centre of oxidised enzyme. The structures shown are based on the PDB file 1KQF [119] (\(\alpha\) helices and \(\beta\) sheets are shown in red and cyan, respectively)

Fig. 6
figure 6

D. gigas (A, B, E) and D. vulgaris (C, D, F, G) formate dehydrogenases. A Three-dimensional structure view of D. gigas W-FDH. B Arrangement of the redox-active centres of D. gigas W-FDH, shown in the same orientation (but not same scale) as in (A). C Three-dimensional structure view of D. vulgaris W-FDH. D Arrangement of the redox-active centres of D. vulgaris W-FDH, shown in the same orientation (but not same scale) as in (C). E Tungsten catalytic centre of oxidised D. gigas W-FDH. F Tungsten catalytic centre of oxidised D. vulgaris W-FDH. G Tungsten catalytic centre of formate-reduced D. vulgaris W-FDH. The structures shown are based on the PDB files 1H0H (A, B, E) [130], 6SDR (C, D, F) and 6SDV (G)  [132] (\(\alpha\) helices and \(\beta\) sheets are shown in red and cyan, respectively)

The diversity of metal-dependent FDHs is also observed through their “molecular plasticity”. Some FDHs take part in formate-hydrogen lyase systems, as is the case of FDH H from E. coli (Mo-FDH) [141], Pectobacterium atrosepticum (Mo-FDH) [142] or C. carboxidovorans (W-FDH) [143,144,145,146,147]. Physiologically, the E. coli formate-hydrogen lyase is a membrane-bound system involved in formate oxidation and dihydrogen formation under fermentative growth conditions [141, 148,149,150,151]. The system comprises two enzymes, the cytoplasmatic Mo-FDH (described above) and a membrane-bound, cytoplasm-faced nickel/iron-containing hydrogenase (Ni/Fe-Hase); FDH oxidises formate to CO2 and the resulting reducing equivalents are transferred, through three Fe/S proteins, to the Ni/Fe-Hase that reduces protons to dihydrogen (Fig. 7).

Fig. 7
figure 7

Adapted with permission from Ref. [149]

Predicted architecture of the E. coli formate-hydrogen lyase. B, F and G represent three Fe/S proteins. See text for details

A different rearrangement of the same basic features (FDH plus Hase) is found in cytoplasmatic dihydrogen-dependent FDHs (better denominated as CO2 reductases), that physiologically catalyse the reduction of CO2 to formate with the simultaneous and direct oxidation of dihydrogen, that is, without the intervention of an external electron-transfer protein or molecule, as reviewed by Litty and Müller in this Book [152] and also [153,154,155,156,157,158,159]. The dihydrogen-dependent CO2 reductase of the acetogen A. woodii is a tetramer (αβγδ), holding one FDH-like subunit comprising one molybdenum and one [4Fe–4S] centres, where CO2 is reduced; the necessary electrons are transferred intramolecularly from an iron–iron hydrogenase-like (Fe/Fe-Hase) subunit (second active site), via two small electron-transfer subunits (each with four [4Fe–4S] centres) (Fig. 8) [153]. A tungsten-containing homologue enzyme is found in Thermoanaerobacter kivui [156].

Fig. 8
figure 8

Adapted with permission from Ref. [156]. http://creativecommons.org/licenses/by/4.0/

Structural organisation of A. woodii (molybdenum-dependent) and T. kivui (tungsten-dependent) dihydrogen-dependent CO2 reductases. ET represents two small electron-transfer subunits. See text for details

A further example of the “plasticity” of FDH-like proteins is provided by N-formyl-methanofuran dehydrogenases (FMFDH) that also have two physically separated active sites: one catalyses the reduction of CO2 to formate, which is then intramolecularly transferred to the second active site, where it is condensed with methanofuran to form N-formyl-methanofuran (Eq. 4) [140, 160, 161]. The FMFDHs are even more complex than FDHs and the enzyme from the methanogen M. wolfeii is a tetramer of (αβγδεω) units, whose CO2-reducing subunit shares the tungsten and [4Fe–4S] centres, as well as, the protein fold of the W-FDHs and Mo-FDHs (Fig. 9).

Fig. 9
figure 9

M. wolfeii N-formyl-methanofuran dehydrogenase. A Three-dimensional structure view. B Arrangement of the redox-active centres shown in the same orientation (but not same scale) as in (A). C Tungsten catalytic centre of oxidised enzyme. The structures shown are based on the PDB file 5T5I [140] (\(\alpha\) helices and \(\beta\) sheets are shown in red and cyan, respectively)

In contrast to the structural and organisational diversity, the active site of all presently known metal-dependent FDHs and FMFDH is very well conserved [94,95,96,97,98,99,100,101, 110,111,112, 140]. In the oxidised form, the active site harbours one molybdenum ion (in the case of Mo-FDHs and Mo-FMFDHs) or one tungsten ion (in W-FDHs and W-FMFDHs) coordinated by the cis-dithiolene (–S–C = C–S–) group of two pyranopterin cofactor molecules (Fig. 10), as is characteristic of this family of mononuclear molybdenum and tungsten enzymes [97, 110,111,112, 162,163,164,165]. The metal first coordination sphere is completed by one terminal sulfido group (Mo/W = S)Footnote 5 plus one sulfur or selenium atom from a cysteine or selenocysteine residue (Mo/W–S(Cys) or Mo/W–Se(SeCys)) (abbreviated as Cys–Mo–FDH, Cys–W–FDH, SeCys–Mo–FDH and SeCys–W–FDH), in a distorted trigonal prismatic. Noteworthy, there is no apparent relation (as far as is presently known) between the metal and the bound amino acid residue (examples of the four combinations Cys–Mo–FDH, Cys–W–FDH, SeCys–Mo–FDH and SeCys–W–FDH are known for long; Table 1) or the enzyme activity. The active site also comprises two other residues that are strictly conserved to all known FDHs and FMFDHs and are thought to be crucial to the catalytic cycle (as discussed below), one arginine and one histidine (this linked (C-terminal side) to the selenocysteine or cysteine that coordinates the molybdenum or tungsten ion) [116, 117, 119, 130].

Fig. 10
figure 10

Active site of formate dehydrogenase and N-formyl-methanofuran dehydrogenase. Structure of the pyranopterin cofactor (top). The pyranopterin cofactor molecule is formed by pyrano(green)-pterin(blue)-dithiolene(red)-methylphosphate(black) moieties; in all so far characterised FDHs, the cofactor is found esterified with a guanosine monophosphate (dark grey). The dithiolene (–S–C = C–S–) group forms a five-membered ene-1,2-dithiolene chelate ring with the molybdenum or tungsten ion, here indicated as M (from metal). Structure of the molybdenum/tungsten centre in the oxidised state (middle). For simplicity, only the dithiolene moiety of the pyranopterin cofactor is represented. Structure of the molybdenum/tungsten centre in the reduced state (bottom). For simplicity, only the dithiolene moiety of the pyranopterin cofactor is represented. Contrary to the oxidised state (that is consensually accepted), the structure of the reduced state is still under debate, as discussed below, under Sect. 4.3.2b. The two hypotheses under debate are represented, with the cysteine or selenocysteine residue bound to the metal and with the residue dissociated from the metal (Sect. 4.3.2b)

4.3 Formate Dehydrogenases—Mechanism of Action

4.3.1 The Metal-Independent Formate Dehydrogenases

The metal-independent FDHs are NAD-dependent enzymes, whose chemical strategy to interconvert formate and CO2 is surprisingly simple and well established (Fig. 11) [102,103,104,105,106,107,108,109]: the enzyme binds formate and NAD+ in close proximity of each other (1.4 Å distance between H-(formate) and C4-(pyridine ring)) and makes NAD+ acquire a bipolar conformation, which increases its electrophilicity and, thus, facilitates the hydride transfer. The reaction, then, proceeds by straightforward hydride transfer from formate to NAD+. In accordance, the rate-limiting step of the catalytic cycle is the formate C–H bond cleavage, as shown by kinetic studies of the 2H-labelled formate isotopic effect), and the enzyme operate via a ternary complex (FDH-formate-NAD+) kinetic mechanism [107, 170,171,172,173,174,175,176,177,178].

Fig. 11
figure 11

Hydride transfer mechanism proposed for metal-independent NAD-dependent formate dehydrogenases

4.3.2 The Metal-Dependent Formate Dehydrogenases

Several experimental and some computational approaches have been exploited to elucidate how the metal-dependent FDHs carry out the formate/CO2 interconversion and over the years a few mechanistic proposals have been put forward [116, 117, 137, 166, 169, 179,180,181,182,183,184,185]. Presently, several key points are well established, but two remain a matter of debate, and all are discussed below, before the currently accepted mechanistic hypotheses are introduced.

  1. (a)

    Presently, several key points are well established

  1. (i)

    The formate/CO2 interconversion occurs at the molybdenum or tungsten centre, in a reaction that is intermediated by the metal, which cycles between the +6 and +4 oxidation states (Eq. 5a5d), as demonstrated by numerous spectroscopic and kinetic studies. The electrons necessary to carry out CO2 reduction or released from formate oxidation are intramolecularly transferred from the physiological partner (electron donor or electron acceptor), through the different redox centres of each enzyme (Fe/S centres, haems, FAD (see above)) that act like a “wire” to facilitate the fast and effective electron transfer. Therefore, the intramolecular electron transfer is, thus, an integral aspect of the global reaction. Depending on the enzyme (on the biochemical pathway where the enzyme is involved in), the physiological redox partner can be membrane quinols, cytoplasmatic and periplasmatic cytochromes, ferredoxins, NAD(P) or coenzyme F420 [94,95,96,97,98,99,100,101]. For those enzymes like the dihydrogen-dependent CO2 reductase (see above), the electrons are directly provided by the co-substrate oxidation (dihydrogen in this case) that occurs in the enzyme second active site. As a consequence of the physical separation of the oxidation and reduction half-reactions (that occur at different enzyme centres), all these enzymes operate via a ping-pong kinetic mechanism, as observed experimentally.

$${\text{HCOO}}^{ - } + {\text{Mo}}/{\text{W}}^{ 6+ } \rightleftharpoons \,{\text{CO}}_{ 2} + {\text{ H}}^{ + } + {\text{ Mo}}/{\text{W}}^{ 4+ }$$
$${\text{Mo}}/{\text{W}}^{ 4+ } + {\text{Physiol}}.\,{\text{Partner}}^{\text{oxidised}} \rightleftharpoons \,{\text{Mo}}/{\text{W}}^{ 6+ } + {\text{Physiol}}.{\text{ Partner}}^{\text{reduced}}$$
$${\text{CO}}_{ 2} + {\text{ H}}^{ + } + {\text{ Mo}}/{\text{W}}^{ 4+ } \rightleftharpoons \,{\text{HCOO}}^{ - } + {\text{ Mo}}/{\text{W}}^{ 6+ }$$
$${\text{Mo}}/{\text{W}}^{ 6+ } + {\text{ Physiol}}.{\text{ Partner}}^{\text{reduced}} \rightleftharpoons \,{\text{Mo}}/{\text{W}}^{ 4+ } + {\text{ Physiol}}.{\text{ Partner}}^{\text{oxidised}}$$
  1. (ii)

    Although the formate/CO2 interconversion occurs at the molybdenum or tungsten centre, the reaction is not one of oxygen atom transfer, as is characteristic of many molybdoenzymes and tungstoenzymes (Fig. 12) [94, 97, 110,111,112, 162,163,164,165]: the substrate for “CO2 reduction” is in fact CO2 and not hydrogencarbonate (Eq. 6) [186], and the product of formate oxidation is CO2 and not hydrogencarbonate (Eq. 7), as was clearly demonstrated by the formation of 13C16O2 gas during the oxidation of 13C-labelled formate in 18O-enriched water [166]. Therefore, to catalyse the formate oxidation, FDH has to abstract one proton plus two electrons (Eq. 8) or one hydride (Eq. 9) from the formate molecule (or the reverse for CO2 reduction).

    Fig. 12
    figure 12

    Oxygen atom transfer in molybdo—and tungstoenzymes. Typically, these enzymes catalyse the transfer of an oxygen atom from water to product—oxygen atom insertion (blue arrows)—or from substrate to water—oxygen atom abstraction (green arrows)—in reactions that entail a net exchange of two electrons, in which the molybdenum/tungsten atom cycle between Mo/W6+ and Mo/W4+, and, most importantly, where the metal is the direct oxygen atom acceptor or donor. This feature was coined by Holm and others in the 1980s as the “oxo transfer hypothesis”

  1. (iii)

    A simple chemical reasoning, based on the pKa values of formic acid (HCOOH/HCOO = 3.77; HCOO/CO22− ≫ 14), demonstrates that it is much more difficult to abstract the Cα proton from formate (Eq. 8) than abstract a hydride (Eq. 9), which, in addition, lead to the formation of a stable product (CO2), instead of a carbonanion (CO22−) (Scheme 1). The simple mechanistic strategy followed by the metal-independent FDHs (Sect.  4.3.1.), that is, direct reaction with NAD+, with no enzyme cofactors involved, further confirms that it must be exceptionally facile (thermodynamics) to abstract a hydride from the formate molecule. On its turn, CO2, with an electronic structure O  −δC+2δ− Oδ and a carbon-localised LUMO, is susceptible to attack by nucleophiles and to reduction, being a good hydride acceptor, as supported by the chemistry of several synthetic transition metal-hydride complexes that mimic the FDH catalysis [65, 67, 187,188,189,190,191,192].

    Scheme 1
    scheme 1

    Products formed by proton abstraction (ruled by a pKa2 >> 14) or hydride abstraction from formate

  2. (iv)

    The terminal sulfido group of the active site (Fig. 10) is well documented as a hydride acceptor/donor. Since the 1970s, the sulfido group is established as the hydride acceptor in the oxidised molybdenum centre (Mo6+=S) of xanthine oxidase and aldehyde oxidaseFootnote 6 [110,111,112, 162,163,164,165, 193,194,195,196,197,198,199,200,201,202], as well as, the hydride donor in the reduced centre (Mo4+–SH) of hydroxybenzoyl-CoA reductaseFootnote 7 [203, 204]. This “twin” behaviour (oxidised/hydride acceptor versus reduced/hydride donor) is supported by a remarkable characteristic of the Mo/W-ligands: the pKa values of the coordinated ligands change dramatically with the oxidation state of the metal and determine that the higher oxidation states should hold deprotonated ligands, that is Mo/W6+=S, while the lower oxidation states should hold protonated ligands, that is Mo/W4+–SH [205,206,207]. This behaviour (Eq. 10) enables the metal-sulfido to act as a hydride acceptor/donor and is supported by the high covalency of the terminal sulfur atom in the metal sulfido group, with an available S π-bond well suited to accept a hydride.

$${\text{Mo}}/{\text{W}}^{ 6+ } = {\text{S}} + \left( { 2 {\text{e}}^{ - } + {\text{ H}}^{ + } } \right) \rightleftharpoons \,{\text{Mo}}/{\text{W}}^{ 4+ } - {\text{SH}}$$

The involvement of the sulfido group as the hydride acceptor during FDH catalysis was demonstrated by electron paramagnetic resonance (EPR) spectroscopic studies that showed that, in formate-reduced FDH, the formate Cα hydrogen atom is transferred to an acceptor group located within magnetic contact to the molybdenum atom of FDHs from different sources (E. coli [166], D. desulfuricans [136] or Cupriavidus necator (previously known as Ralstonia eutropha) [184]). The observation of a strongly coupled, solvent-exchangeable and substrate-derived proton, with a hyperfine constant of 20–30 MHz, is consistent with the hydrogen atom being transferred to a ligand in the first coordination sphere of the molybdenum atom upon its reduction [196, 197, 200, 201, 208]. Similar hyperfine constant values were determined in model complexes [209] and also in real enzymes, as in xanthine oxidase, where the strong coupled hydrogen is originated from the xanthine C8 hydrogen atom (the position that is hydroxylated by that enzyme (see Footnote 6) [196,197,198, 200,201,202, 208]. It should be noted that a hyperfine interaction of this magnitude could not arise from the transfer of the formate Cα hydrogen atom to an acceptor in the second coordination sphere of the metal, for example, transfer to the conserved histidine residue, as initially proposed [116, 117, 166, 169, 180,181,182,183], or transfer to the selenocysteine/cysteine residue if it had been dissociated from the molybdenum/tungsten ion [117, 169, 180,181,182,183]—an hypothesis discussed below in point (vi). Photolysis assays with 77Se-enriched FDH, described below in point (vi) and Footnote 8, further confirmed that the selenocysteine residue cannot be the hydrogen atom acceptor [166].

Further evidence that the sulfido group becomes protonated upon reduction was also provided by a recent X-ray absorption spectroscopy (XAS) study with the R. capsulatus Mo-FDH [210].

  1. (v)

    The terminal sulfido group is essential to both formate oxidation and CO2 reduction. It is well established, by numerous spectroscopic and kinetic studies, that cyanide reacts with the active site sulfido group of different molybdoenzymes, such as xanthine oxidase, and abstracts it in the form of thiocyanate, yielding a desulfo enzyme form that harbours an oxo group in the place of the native sulfido group [110,111,112, 193,194,195, 198, 199, 202]. The sulfido by oxo replacement renders xanthine oxidase and other enzymes inactive, because its active site is no longer able to accept a hydride (see Footnote 6). The same and complete cyanide inhibition is observed in several FDH, such as the ones from Methanobacterium formicicum [211], Alcaligenes eutrophus [212], E. coli [167], R. capsulatus (where the sulfido was observed to be replaced by an oxo group) [210], or D. desulfuricans (where thiocyanate formation accounted to 0.87 per molybdenum atom) [137]. Together with the experimental evidences that support the involvement of the sulfido group as a hydrogen atom acceptor during FDH catalysis (described above), these inhibitory results demonstrate that the sulfido group acts as a hydride acceptor/donor in FDH catalysis.

  1. (b)

    Two interrelated points are not yet consensual

  1. (vi)

    Does the active site cysteine or selenocysteine residue dissociate from the metal during catalysis?

If the configuration of the oxidised active site is consensually accepted, the reduced form still finds a few contradictory experimental evidences (Fig. 10).

X-ray crystallography: In a reinterpretation of the crystallographic data of the reduced E. coli SeCys–Mo–FDH H originally obtained by Boyington et al. in 1997 [116], Raaijmakers and Romão in 2006 [117] suggested that the polypeptide loop containing the selenocysteine was not properly traced in the original work and that the selenocysteine residue is not bound to the metal, but, instead, is found dissociated from the molybdenum ion and shifted away (12 Å) (Fig. 4). Therefore, those authors suggested that, while in the oxidised state the selenocysteine residue is coordinated to the metal, the enzyme reduction triggers the residue dissociation, resulting in a square pyramidal penta-coordinated centre, where the molybdenum ion is coordinated by the cis-dithiolene (–S–C = C–S–) group of two pyranopterin cofactor molecules (in the equatorial positions) plus the terminal sulfido group (in the axial position) (Fig. 10). Regardless of this reinterpretation, all other crystallographic structures so far available, for FDH and FMFDH (Figs. 4, 5, 6 and 9, and references herein), show a stable hexa-coordination, with the cysteine or selenocysteine always bound to the molybdenum/tungsten ion. This is also the case of the recently solved structure of the formate-reduced D. vulgaris SeCys–W–FDH [132] and also of the NADH-reduced R. capsulatus Cys–Mo–FDH, whose structure was determined by cryo-electron microscopy [139].

XAS: Two recent XAS studies at the Mo K-edge suggested that, in R. capsulatus Cys–Mo–FDH, the Mo5+ state holds the cysteine residue bound to the metal, as the oxidised Mo6+ one, with a Mo-S(Cys) bond of 2.63 Å, while the Mo4+ state of formate-reduced enzyme has its cysteine displaced form the metal [169, 210]. However, contrary results were obtained with SeCys–Mo–FDHs from E. coli [213] and D. desulfuricans [214], which showed a metal bound residue in all oxidised and reduced states: the E. coli enzyme EXAFS data at both the Mo and Se K-edges were interpreted as indicating the presence of one Mo–Se bond of 2.62 Å, plus one Se–S bond of 2.19 Å (between the sulfido group and the selenocysteine selenium) [213].

EPR spectroscopy: Further experimental evidence for the stable molybdenum/tungsten hexa-coordination came from EPR spectroscopy that clearly showed that the selenocysteine/cysteine must remain bound to the Mo5+ centre of formate-reduced enzyme [208]. When the EPR spectrum is obtained from 77Se-enriched enzyme, a very strong and anisotropic interaction with selenium is observed (A1,2,3(77Se) = 13.2, 75, 240 MHz) [166]. This interaction and the observation of the expected 95,97Mo hyperfine coupling confirms that the selenium atom of the selenocysteine is directly coordinated to the Mo5+ and further suggests that the unpaired electron is delocalised over the selenium (17–27%) and molybdenum atoms (73–83%) [166]. Also, the hydrogen atoms of the β-methylene carbon of the selenocysteine residue are thought to be in the close proximity of the molybdenum atom, being responsible for an interaction with a not solvent-exchangeable protons (A1 = 35.1 MHz) [136]. Photolysis assays additionally confirmed that the selenium/sulfur ligation is retained in the FDH Mo5+ centre (the light beam did not affect the strong selenium–molybdenum EPR interaction observed in 77Se-enriched FDH)Footnote 8 [166]. The Mo5+ hexa-coordination (resulting from having the selenocysteine/cysteine residue bound to the molybdenum ion) was also supported by theoretical calculations on the signals-giving species of FDHs [215].

Inhibition assays: A different type of experimental evidence came from inhibition studies with iodoacetamide, an alkylating agent that reacts with “free” ionised selenocysteine or cysteine residues (carboxamidomethylation). Native E. coli SeCys–Mo–FDH H and its cysteine mutant [216] and native R. capsulatus Cys–Mo–FDH [183] are not inhibited by iodoacetamide treatment. However, when the preliminary iodoacetamide treatment (incubation) is carried out in the presence of formate (not under turnover), both native and cysteine-containing mutant E. coli FDHs are inhibited [216]. Inhibition is also observed in the R. capsulatus FDH, but only when the iodoacetamide treatment (incubation) is carried out in the presence of nitrate; in this case, the cysteine carboxamidomethylation was confirmed by mass spectroscopy [183]. On the other hand, native D. vulgaris SeCys–W–FDH is inhibited by iodoacetamide, but mass spectroscopy clearly showed that the inhibition is not due to the carboxamidomethylation of the active site selenocysteine residue (but of 9 other cysteine residues not present in the active site) [132]. In addition, other FDHs are not at all affected by iodoacetamide [135, 136]. Hence, the inhibition results available were not obtained under formate/CO2 turnover conditions and the inconsistency of the results, once more, do not contribute to provide a definitive answer.

Overall, the majority of experimental evidences points towards a molybdenum/tungsten stable hexa-coordination, with the cysteine/selenocysteine residue always bound to the metal. Regarding the results showing a metal penta-coordination, with unbound cysteine/selenocysteine residue, it is possible that the crystallisation/irradiation had induced some artefacts that are not relevant to the enzyme activity; but it is also possible that the species crystallographically characterised, being catalytically relevant, bear no relation to the species observed by XAS and EPR (with these being not catalytically relevant). Certainly, high-resolution structures are needed to confirm the existence of the two alternating conformations of the selenocysteine/cysteine-containing polypeptide loop and to discuss the catalytic relevance of each conformation.

  1. (vii)

    Do formate/CO2 bind directly to the molybdenum/tungsten ion during catalysis?

Inspired by the oxotransfer chemistry displayed by several molybdenum—and tungsten-dependent enzymes (Fig. 12) [110,111,112, 162,163,164,165], and in particular by periplasmatic nitrate reductasesFootnote 9, it was suggested that FDH catalysis necessarily involves the formate/CO2 direct binding to the molybdenum/tungsten ion [180, 218].

To begin the discussion of this point, it should be noted that the direct formate/CO2 binding would involve an unprecedented hepta-coordinated molybdenum/tungsten centre or it would depend on the dissociation of the cysteine/selenocysteine residue from the metal, in order to create a vacant position (penta-coordinated centre) for substrate binding. While the hypothesis of the hepta-coordinated metal centre was (so far) never perused, the dissociation of the cysteine/selenocysteine is, as discussed above, controversial.

The two recent XAS studies mentioned above in point (vi) suggested that, in the presence of formate, the cysteine ligation of (active) R. capsulatus Cys–Mo–FDH is replaced by a long Mo–O bond of 2.15 Å, which was interpreted as arising from the Mo–OCO(H) complex [169, 210]. The strong and competitive inhibition of E. coli SeCys–Mo–FDH H-catalysed formate oxidation by azide, cyanate, thiocyanate, nitrite and nitrate (1–00 μM range) was also evoked to support that formate, as well as those inhibitors, bind directly to the molybdenum ion [219]. Yet, competitive inhibition can arise if the inhibitor binds in the active site, but not directly to the metalFootnote 10 [220]; this seems to be the case at least of azide, a well documented inhibitor of both metal-independent [104, 171, 177] and metal-dependent [136, 166] FDHs (as suggested by EPR [136, 215]) and nitrite (as suggested by crystallographyFootnote 11). Regarding nitrate, (once more) several other FDH enzymes are not inhibited or the inhibition constants are 2–3 orders of magnitude higher than Km(formate) [132, 135, 136, 166], or it is though as a substrate (even though a very poor one; see Footnote 9 [182]). Moreover, the same study [219] showed that the inhibition of the E. coli SeCys-Mo-FDH H-catalysed CO2 reduction by those anions is very weak (in the range of 1–50 mM) and not competitive in nature, results that are contradictory to the hypothesis that the reduced active site (the one that reacts with CO2) becomes penta-coordinated, with an unbound selenocysteine residue, and with an available position to bind inhibitors and CO2.

Therefore, except from the abovementioned XAS data, there are no other direct experimental evidences of the direct formate or CO2 binding to the FDH molybdenum/tungsten ion; namely, there are no crystallographic structures showing the formate molecule in the active site and there are no EPR signals showing the presence of formate in the first coordination sphere of molybdenum/tungsten [136, 166, 184].

Theoretically, it can be argued that, since the FDH-catalysed reaction does not involve the transfer of an oxygen atom (as explained above in point (ii)), there is no need to form the otherwise expected Mo/W6+–OCO(H) or Mo/W4+–OCO complexes (follow Mo4+–OR and Mo4+–OQ in Fig. 12). It can also be argued in the opposite way: if formate/CO2 binds directly to the molybdenum/tungsten ion, why there is no oxygen atom transfer to form hydrogencarbonate (Eq. 7)? Overall, in the absence of more definitive experimental evidences, we must continue to ask: Does the direct formate/CO2 binding to the metal occur? Is it necessary or desirable to interconvert formate and CO2? Is the penta-coordinated metal centre with unbound cysteine/selenocysteine catalytically relevant?

  1. (c)

    Currently accepted mechanistic hypotheses

In accordance with the well-established points (i) to (v) highlighted above, the FDH-catalysed formate oxidation and CO2 reduction are presently recognised to occur through hydride transfer (Eq. 9), with the oxidised and reduced active site sulfido group, Mo/W6+=S and Mo/W4+–SH, acting as the direct hydride acceptor and donor, respectively. Yet, points (vi) and (vii) still raise questions to some authors regarding the coordination of the active site and substrates binding during FDH catalysis.

As originally proposed by Niks et al. [184] for formate oxidation and shortly after also for CO2 reduction [137], we suggest that FDH catalysis proceeds as follows (the reaction mechanism is suggested to be identical in Mo–FDH and W–FDH, as well as in FMFDH):

Formate oxidation (Fig. 13, blue arrows) is initiated with the formate binding to the oxidised active site, but not directly to the molybdenum/tungsten atom. Following the example provided by the metal-independent FDH, where the formate-binding site harbours arginine and asparagine residues [102,103,104,105,106,107,108,109], it is suggested that the conserved arginine residue is essential to drive the formate Cα hydrogen towards the sulfido ligand, by establishing hydrogen bond(s) with its oxygen atom(s). Also, azide (N3, isoelectronic with CO2) is suggested to bind (tightly) to the same site and not directly to the molybdenum/tungsten ion (as had been previously suggested for the D. desulfuricans FDH inhibition by azide [136, 215]). The binding of azide and formate to the same site, and not to the molybdenum/tungsten atom itself, explains why azide is a powerful inhibitor of both metal-independent (Ki = 40 nM for Candida boidinii) [104, 171, 177] and metal-dependent FDHs [136, 166]. A similar reasoning applies to the inhibitor nitrite (isoelectronic with formate). Formate oxidation, then, proceeds by a straightforward hydride transfer from formate to the sulfido group of the oxidised molybdenum/tungsten centre, Mo/W6+=S, leading to the formation of Mo/W4+–SH and CO2. The re-oxidation of Mo/W4+ to Mo/W6+ (via intramolecular electron transfer to the enzyme other(s) redox centre(s) and, eventually, to the physiological partner) and the release of CO2 close the catalytic cycle. The now oxidised Mo/W6+ favours the sulfido group deprotonation (dictated by the ligand pKa [205,206,207]), and the initial oxidised molybdenum/tungsten centre, Mo/W6+=S, is regenerated. Under non-steady-state catalytic conditions (as the ones created in EPR experiments) the molybdenum/tungsten one-electron oxidation should be favoured (Mo/W4+→Mo/W5+), leading to the formation of the EPR detectable Mo5+–SH species.

Fig. 13
figure 13

Reversible hydride transfer mechanism proposed for metal-dependent formate dehydrogenase and N-formyl-methanofuran dehydrogenase [137, 184]. Reaction mechanism proposed for formate oxidation (blue arrows) and CO2 reduction (green arrows). For simplicity, the mechanism is represented only for a molybdenum, selenocysteine-containing enzyme, but it should be similar for tungsten and cysteine-containing enzymes. See text for details. A similar hydride transfer mechanism can also take place with a penta-coordinated reduced metal centre, with a dissociated selenocysteine/cysteine residue (see text for details)

CO2 reduction is suggested to follow the reverse reaction mechanism (Fig. 13, green arrows). First, CO2 binds to the reduced active site, not directly to the molybdenum/tungsten ion, but at the same site as formate (and azide), with the conserved arginine anchoring its oxygen atom(s) through hydrogen bond(s) and orienting its carbon atom towards the protonated sulfido ligand. In an approximated way, based on the inhibition and Michaelis–Menten constants for the D. desulfuricans FDH, the “binding strength” is suggested to follow the order CO2 (Km ≈ 15 μM [137]) > azide (Ki ≈ 30 μM [136]) > formate (Km ≈ 60 μM [137]). Then, the reaction proceeds through straightforward hydride transfer from the protonated sulfido group of the reduced molybdenum/tungsten centre, Mo/W4+–SH, to the CO2 carbon, whose LUMO have predominant C–π orbital character, prone to nucleophile attack and reduction. This yields a formate moiety and Mo/W6+=S. The subsequent re-reduction of Mo/W6+ to Mo/W4+ (via intramolecular electron transfer from the enzyme physiological partner, through its redox centre(s)) and formate release closes the catalytic cycle. The now reduced Mo/W4+ favours the sulfido group protonation and the initial reduced molybdenum/tungsten centre, Mo/W4+–SH, is regenerated.

The FDH-catalysed reaction is reversible and the equilibrium between formate oxidation versus CO2 reduction is determined by the availability of formate versus CO2 and the ability to maintain the active site oxidised (Mo/W6+) versus reduced (Mo/W4+), which, in its turn, determines the protonation state of the metal sulfido group in a concerted and straightforward way.

Overall, the chemical strategy herein suggested is exactly the same as the one proposed for the metal-independent FDHs: both bind formate in a close proximity to an oxidised, electrophilic, hydride acceptor, which in metal-independent enzymes is a NAD+ molecule and in metal-dependent enzymes is the M6+=S group; both bind CO2 in a close proximity of a reduced, nucleophilic, hydride donor, a NADH molecule or the M4+–SH group.

As expected, this mechanistic proposal faces some criticism and the most relevant one concerns the role of the active site selenocysteine/cysteine residue. In fact, although the mechanism is suggested to operate in a hexa-coordinated metal centre (Fig. 13), it can also take place in a penta-coordinated centre (Fig. 10), with an unbound selenocysteine/cysteine—the sixth ligand does not seem to interfere with the hydride transferFootnote 12. Even though there are experimental evidences (as discussed above) and mechanistic arguments can be envisaged to support the necessity of having a bound selenocysteine/cysteine (as discussed in [96, 98]), in the absence of clear and definitive experimental evidences, both scenarios—dissociated and bound selenocysteine/cysteine—seem to be possible and this is an aspect that will remain in open for now. Certainly, future research will shed light in these aspects of the FDH reaction, allowing a critic evaluation of this mechanistic proposal.

4.4 Formate Dehydrogenases in the Context of Carbon Dioxide Utilisation

The majority of currently known FDHs function in vivo to oxidise formate, with only a few participating in metabolic pathways to reduce (fix) CO2—the reaction direction that is interesting to “solve humankind’s problem” with atmospheric CO2. However, the CO2/formate interconversion is thermodynamically reversible (E º′(CO2/HCOO) = −0.43 V) and in vitro there is no a priori reason for a FDH to be unable to catalyse the CO2 reduction, as long as there is sufficient reducing power availableFootnote 13. Regarding the metal-independent FDHs, this simply means the adequate NADH/NAD+ ratio (Eq. 11); in what concerns the metal-dependent FDHs, this means that the molybdenum/tungsten active site centre must be kept reduced at the proper reduction potential (Eq. 12). Although at first sight obvious, the necessity to keep the enzyme active site reduced is most often overlooked, what may explain why there are so many reports in the literature of FDHs unable to reduce CO2. This is particularly true for metal-dependent FDHs: if the reduction potential of one (or more) of the FDH redox centres is (are) relatively high, it could be difficult to “push” the electrons into the active site (the centre with the higher reduction potential could stay reduced, “blocking” the electron transfer to the other(s) centre(s) with lower reduction potentials and the active site in particular). In addition to thermodynamics, also the kinetics has to be taken into account to evaluate if the CO2 reduction is going to be efficient, or too slow relatively to the formate oxidation to be relevant (rate of formate oxidation versus CO2 reduction, Eqs. 13, 14). The key point here is that FDH kinetics is determined by four parameters, the Km and kcat for the two substrates (CO2 and formate), and the reaction can be run under different regimes (mainly forward, mainly reverse and equilibrium, as determined by kcat/Km and imposed conditions). However, except for protein engineering (a difficult task on its own), there is not much that can be done to modify the kinetic parameters to further favour the CO2 reduction.

$${\text{CO}}_{ 2} + {\text{NADH}} \to {\text{HCOO}}^{ - } + {\text{NAD}}^{ + }$$
$${\text{CO}}_{ 2} + {\text{FDH }}\left( {{\text{Mo}}/{\text{W}}^{ 4+ } } \right) \to {\text{HCOO}}^{ - } + {\text{FDH }}\left( {{\text{Mo}}/{\text{W}}^{ 6+ } } \right)$$
$${\text{CO}}_{ 2} + {\text{A}}_{\text{red}} \,\underrightarrow {k^{{{\text{CO}}_{2} }} }\,{\text{HCOO}}^{ - } + {\text{A}}_{\text{ox}}$$
$${\text{HCOO}}^{ - } + {\text{A}}_{\text{ox}} \,\underrightarrow {k^{{{\text{HCCO}} - }} }\,{\text{CO}}_{ 2} + {\text{A}}_{\text{red}}$$

Also, the enzymes stability and potential interfering compounds must be well thoughtout. The lifetime of a CO2 converter device is a critical issue, and it would greatly depend on the time the enzyme maintains its full activity. In this respect, it should be emphasised that the purifying processes often decrease the enzymes stability and even make them unstable (while taken out of their biological environment), thus hampering their usage in a sustained (“real life”) way or just making the scale up process unviable. The inhibition and inactivation by compounds that might be present in the “substrate reaction mixture”; for example, dioxygen or carbon monoxide, are pitfalls that must be considered to avoid the need (additional cost) of using purified CO2. The inhibition or inactivation (or, on the contrary, improved stability) by the materials used to build the device cannot be overlook also.

The enzyme–material “communication” is another major challenge in hybrid systems. It is necessary to properly orient and “link” the enzyme to the material (for example, electrode or light absorber), via electrostatic or covalent interactions to maximise the charge transfer. In this respect, features as the enzyme size (in the nanoscale) and local-specific surface charge/hydrophilicity/hydrophobicity must be taken into consideration when choosing the materials and its functionalisations.

Having these general points tackled, any FDH could be used to build a device to promote the CO2 reduction. The same would be true for whole-cell devices, but considering in that case the organism whole metabolism (carbon and energy needs).

4.5 Formate Dehydrogenases in Action

The use of enzymes and whole-cells systems to convert CO2 into VC is growing exponentially due to the “green” advantages the “biochemical way” can offer, namely substrate and product specificity (ability to discriminate the substrate in a complex mixture and to produce only the product of interest) in reactions at ambient temperature and pressure and neutral pH. Numerous hybrid systems are currently being exploited to convert CO2 into formate, following the same master lines as described in Sect. 3 (Fig. 2). Like electrochemistry, bioelectrochemistry is currently under intense research, as is reviewed in [221,222,223,224,225,226,227] and references herein (below). Most interest is also being focused on the biophotoreduction of CO2, as solar light represents the most straightforward way to use a RES to convert CO2. Semi-artificial photosynthesis systems have been devised, where enzymes and also entire metabolic pathways within cells are interfaced with synthetic materials to develop new solar-to-VC and solar-to-fuel devices, which would not be feasible with natural or artificial systems alone [228 and references herein (below)]. The direct CO2 hydrogenation is also getting enormous attention, mimicking metabolic pathways, where the formate-hydrogen lyase systems are the most explored examples, but using also whole-cells systems. Most important are the breakthroughs achieved by exploiting the recently identified metabolic pathways of acetogens and its dihydrogen-dependent CO2 reductase enzymes (Sect. 4.2.2.), as is reviewed by Litty and Müller in this Book [152] and also [153,154,155,156,157,158,159] and references herein (below).

Herein (below), a few promising studies and successful proof of concepts of FDH-dependent CO2 reduction to formate and beyond are discussed, to highlight the power of FDHs and the challenges this CO2 bioconversion still faces.

One of the most efficient CO2 reducers so far described (along with the T. kivui enzyme described below) is a SeCys–W–FDH from the Synthrobacter fumaroxidans that displays an impressive rate of CO2 reduction of ≈ 2.5 × 103s−1 (reported as 900Umg−1; K CO2m not determined, assays with 10 mM hydrogencarbonate), with a slightly lower formate oxidation rate (≈ 1.9 × 103s−1 (reported as 700Umg−1); K HCOO−m of 40 μM) [229,230,231]. This enzyme is also a good electrocatalyst to carry out the electrochemical reduction of CO2 to formate, using mild conditions and applying small overpotentials, with a maximum current density of ≈ 80 μAcm−2 that corresponds to ≈ 110 s−1 (from a monolayer of enzyme) [232]. Intriguingly, while in homogeneous catalysis in solution the CO2 reduction is slightly faster than the formate oxidation, in the electrochemical-assisted reduction/oxidation is the formate oxidation that is more than 2 times faster (with a current density of ≈ 200 μAcm−2 [232]). S. fumaroxidans expresses another very fast CO2 reducer SeCys-W-FDH, with a rate of ≈ 200 s−1 (reported as 90Umg−1) [229,230,231], but its CO2 reduction activity cannot kinetically compete with its highly efficient formate oxidation, rate of ≈ 5.6 × 103s−1 (value reported as 2700Umg−1) and K HCOO−m of 10 μM. Unfortunately, these enzymes are extremely oxygen-sensitive, and no further studies towards a biotechnological application were pursuit, as far as we know.

Several other FDHs have been described to be able to reduce CO2, but at considerably lower rates. Numerous studies have been conducted with metal-independent FDHs, many of which relying on sacrificial electron donors [233,234,235,236,237,238,239,240,241,242,243,244,245,246,247,248]. This is the case of the C. boidinii NAD-dependent metal-independent FDH, that, in spite of its considerably low k HCO3−cat value of only 0.009 s−1 (K HCO3−m  ≈  27.3 mM [240]; kHCO3− ≈ 0.3 M−1s−1; k HCOO−cat  ≈ 5.0 s−1; K HCOO−m  ≈ 5.0 mM; kHCOO− ≈ 1.0 × 103M-1s-1 [109]), has been largely exploited for its ability to reduce CO2. To push the reaction in the desired, but thermodynamically unfavourable, directionFootnote 14 is important to remove NAD+/regenerate NADH (also essential for the process to become cost-effective, since NADH is a very expensive reducing agent). Four selected examples of different strategies to force the reaction towards the CO2 reduction are: (a) an electroenzymatic cell where NADH is electrochemically regenerated through a rhodium complex, with which a formate formation rate of  ≈  3.2 × 10−4 μmolmin−1mg−1 was achieved [240]; (b) electrochemical NADH regeneration, but with an electropolymerised mediator-regenerator (neutral red) in a novel cathode with immobilised FDH, which is able to produce formate at a rate of ≈ 60 μMmin−1) [249]; (c) photocatalytical NADH regeneration using a rhodium complex and a visible light-active photocatalyst) that enabled a formate formation rate of ≈ 1 μmolmin−1 [250]; (d) and enzymatic regeneration, using glutamate dehydrogenase with NAD(H) being covalently attached to micro-particles, to be easily recovered and reused, in an approach that allowed to improve the reaction yield from 0.12 to 1.27 methanol formed/NADH consumed (in this study, formate was further reduced to methanol) [251]. The Thiobacillus sp KNK65MA NAD-dependent metal-independent FDH exhibit an as well low kcat value (k HCO3−cat  ≈ 0.32 s−1; K HCO3−m  ≈ 9.2 mM; kHCO3− ≈ 35 M−1s−1), but its specificity for formate is only 3 times superior (k HCOO−cat  ≈ 1.8 s−1; K HCOO−m  ≈ 16 mM; kHCOO− ≈ 110 M.1s.1) [252]. This Thiobacillus enzyme was successfully used to reduce CO2 by coupling it with a NADH photoelectrochemical regeneration system (Fig. 14), with a formate production rate of 2 μMmin−1 (current density  ≈  3.5mAcm−2) [245].

Fig. 14
figure 14

Reproduced (from Ref. [245]) by permission of the Royal Society of Chemistry. All rights reserved. https://doi.org/10.1039/c6gc02110g

Schematic diagram of enzymatic photosynthesis of formic acid using Thiobacillus FDH coupled with photoelectrochemical regeneration of nicotinamide cofactors. Co-Pi, cobalt phosphate. See text and Ref. [245] for details.

The metal-dependent FDHs display a wide range of CO2 reduction rates. The Clostridium carboxidivorans NAD-dependent SeCys-W-FDH exhibits a considerably low k COcat 2 value (only 0.08 s−1; K HCO3−m  ≈ 50 μM) [144, 145, 147]. Nevertheless, it enabled the photoelectrochemical CO2 reduction to formate within an enzyme cascade that led to the methanol production at  ≈  4 μMmin−1) [253]. The Methylobacterium extorquens AM1 NAD-dependent Cys-W-FDH has also been exploited with different approaches to drive the electrochemical CO2 reduction [242, 254,255,256,257,258]. Using mediated enzymatic bioelectrocatalysis with gas diffusion electrodesFootnote 15, current densities of 15–20mAcm−2 were attained [254, 255]; in a whole-cell catalyst, M. extorquens was able to electrochemically produced formate concentrations of up to 60 mM [242].

The R. capsulatus [182] and Cupriavidus oxalaticus [259] NAD-dependent Cys–Mo–FDH enzymes, on the other hand, have k COcat 2 values, of 1.5 s−1 and ≈ 3 s−1, respectively, but ≈ 25 and ≈ 30 times (respectively) lower than the one for formate oxidation (R. capsulatus K HCOO−m  ≈ 280 μM and K COm 2 not determined, assays with 100 mM hydrogencarbonate; C. oxalaticus K HCOO−m ≈ 100 μM and K HCO3−m ≈ 40 mM). The C. necator NAD-dependent Cys–Mo–FDH, on the contrary, catalyses the reduction of CO2 with a k COcat 2 ≈ 11 s−1 (K COm 2 ≈ 2.7 mM; K NADHm ≈ 45 μM) [260, 261]. To fulfil the potential industry application of this oxygen-tolerant and robust enzyme, it is necessary to implement a NADH regenerating system that pushes the reaction towards CO2 reduction (as discussed above; see Footnote 14) [262]. With the C. necator FDH, this was successfully achieved with the inclusion of glucose dehydrogenase in the system (Fig. 15), which, while catalysing the re-reduction of NAD+ to NADH, enabled the continuous electron delivery to drive the CO2 reduction and, therefore, improved the reaction yield from 0.2 to 1.8 formate formed/NADH consumed [262].

Fig. 15
figure 15

Schematic diagram of the enzymatic cascade reaction C. necator FDH and glucose dehydrogenase. See text and Ref. [262] for details

The E. coli SeCys-Mo-FDH H was also shown to be able to reduce CO2 [263], but at rates considerably lower than the ones of formate oxidation, < 1 versus 160 s−1 [263]. Interestingly, when the reaction is driven electrochemically (protein film voltammetry), the formate oxidation was only two times higher than the CO2 reduction, with current densities of 180 versus 80 μAcm−2, respectively [263]. This E. coli enzyme feature has been exploited in fuel cell devices (FDH immobilised in redox mediators-functionalised redox polymers) [15, 16], where CO2 could be reduced with a very high Faradaic efficiency (99%) and a current density of ≈ 60 μAcm−2 (K HCO3−m  ≈ 2.5 mM) [16]. Thanks to the FDH H-containing formate-hydrogen lyase system (Sect. 4.2.2.)Footnote 16, engineered E. coli whole cells were also used as a “cell factory” to very efficiently produce formate from a gaseous mixture of CO2 and dihydrogen (56:44; up to 10 bar) (Fig. 16); an 100% of CO2 conversion was achieved, with formate (more than 500 mM) being accumulated outside the bacterial cells [264]. Intact E. coli cells were also used in a microbial electrolysis system using an iron-modified carbon cathode, with which a formate production rate of ≈ 10 μMmin−1, with a Faradaic efficiency of ≈ 60%, was attained [265].

Fig. 16
figure 16

Adapted with permission from Ref. [264]. http://creativecommons.org/licenses/by/4.0/

Schematic diagram of a “cell factory” to produce formate using E. coli whole-cells. FHL, formate-hydrogen lyase. See text and Ref. [264] for details

FDHs from sulfate-reducing bacteria constitute other very interesting systems to exploit, exhibiting high rates of CO2 reduction. The D. desulfuricans SeCys–Mo–FDH is a strikingly efficient CO2 reducer. With a \(k_{\text{cat}}^{{{\text{CO}}_{2} }}\)  ≈ 50 s−1 and particularly low \(K_{m}^{{{\text{CO}}_{2} }}\) ≈ 15  μM, this enzyme has a superior specificity for CO2 (\(k^{{{\text{CO}}_{2} }}\)  ≈ 3.3 × 106M−1s−1) [137]. The high Km value for formate (K HCOO−m  ≈ 55 μM; k HCOO−cat  ≈ 550 s−1; kHCOO− ≈ 10 × 106M−1s−1) enables D. desulfuricans SeCys–Mo–FDH to be a powerful CO2 reducer, as long as the formate concentration is kept low (is removed from the system). In addition, once the catalysis is initiated (occurring at steady-state rates), this enzyme robustness allows the reaction to fully proceed even in the presence of dioxygen [139]. Moreover, the D. desulfuricans SeCys–Mo–FDH is also a good electrocatalyst (unmediated electrochemistry) to carry out the electrochemical reduction of CO2 with good catalytic currents being attained [266]. The ability of D. desulfuricans to produce formate was also demonstrated in whole-cells catalysis, where the continuous formate production exhibited a maximum specific formate production rate of 14 mM formate/gdcwh, and more than 45 mM of formate were obtained with a production rate of 0.40mMh−1 [267].

The D. vulgaris contains also several interesting FDHs. One SeCys–Mo–FDH is able to catalyse the CO2 reduction at a rate of ≈ 3.4 s−1 (reported as 1Umg−1) [129, 131, 268]. However, its extremely low Km value for formate (K HCOO−m of 8 μM) and higher rate of formate oxidation (k HCOO−cat  ≈ 260 s−1) makes this enzyme a very interesting biocatalyst to oxidise formate instead, namely to be coupled to dihydrogen production. The proof of concept that D. vulgaris is able to produce dihydrogen at high volumetric and specific rates (0.125 dm3 H2/dm3h1 and 2.5 dm3 H2/gdcwh) was obtained recently, with the demonstration that whole cells are able to grow by catalysing the oxidation of formate to hydrogencarbonate and dihydrogen, in the absence of sulfate or a syntrophic partner [268, 269].

The D. vulgaris SeCys–W–FDH, on the other hand, is better suited for CO2 reduction, with a k COcat 2 ≈ 315 s−1 (\(K_{m}^{{{\text{CO}}_{2} }}\) ≈ 420 μM; \(k^{{{\text{CO}}_{2} }}\) ≈ 0.75 × 106M−1s−1) [132]. Even though the CO2 specificity of this enzyme is considerably lower (100 times) than that for formate (k HCOO−cat  ≈ 1310 s−1; K HCOO−m ≈ 17 μM; kHCOO− ≈ 77.5 × 106M−1s−1 [132]), this D. vulgaris W-FDH is at the base of several very well succeeded proof-of-principle devices for semi-artificial photosynthesis and production/storage of dihydrogen. A D. vulgaris W-FDH-containing cathode wired to a T. elongatus photosystem II-containing photoanode with a synthetic dye with complementary light absorption was successfully employed to drive light-dependent CO2 conversion to formate, using water as an electron donor (Fig. 17) [270]. In this photoelectrochemical tandem device, electrons are photogenerated in the photosystem II, which oxidises water to dioxygen, and transferred to the FDH cathode (this biocathode catalyses the formate formation with a current density of 240 μAcm−2 (at—0.6 V versus SHE) and a Faradaic efficiency of ≈ 80%). The whole system is able to efficiently produce formate at 0.185 μmolcm−2, with Faradaic efficiency of ≈ 70%, but progressive photosystem II photodegradation (due to prolonged irradiation) resulted in an irreversible decrease in the CO2 photoreduction. A different D. vulgaris W-FDH-material configuration was recently devised, based on a ruthenium dye [271]. The employment of this dye-sensitised TiO2-adsorped FDH enable the visible light-driven CO2 reduction to formate with a turnover frequency of 11 s−1, in the absence of a soluble redox mediator (Fig. 18) [271] (comparatively, this bioelectrode reached a current density of 100 μAcm−2 (at −0.6 V versus SHE), with a Faradaic efficiency of 92.5%). Furthermore, the D. vulgaris FDH-mediated electroenzymatic CO2 reduction to formate was attained using a redox viologen-based polymer/enzyme-modified gas diffusion electrode [272]. The D. vulgaris W-FDH has also been exploited to drive the dihydrogen formation/storage in a system that mimics the natural formate-hydrogen lyase systems (see Sect. 4.2.2.) [273]. The semi-artificial formate-hydrogen lyase system consists of the D. vulgaris W-FDH and D. vulgaris Ni/Fe-Hase immobilised on a conductive scaffold of indium tin oxide that acts as an electron relay. This configuration enables the overall reaction to proceed reversibly towards formate conversion into CO2 plus dihydrogen or towards formate formation, with minimal bias in either direction (Fig. 19), thus allowing the longed-for dihydrogen storage and release on demand. The system is able to produce dihydrogen (upon formate addition) at a rate of 4nmolmin−1 (turnover number of 23 × 103 and turnover frequency of 6.4 s−1 for the Hase) or to produce formate (in the presence of dihydrogen) at a rate of 22nmolmin−1 (turnover number of 16 × 103 and turnover frequency of e.4 s−1 for the FDH) for 8 h (this bioelectrode system reached current densities of 185 and 450 μAcm−2, for CO2 and H+ reduction, respectively (at −0.6 V versus SHE) and of 300 and 440 μAcm−2 for formate and H2 oxidation, respectively (at −0.2 V versus SHE), with Faradaic efficiencies for H2 and formate production of 77 and 76%, respectively). Moreover, this semi-artificial formate-hydrogen lyase concept can be deployed in either an electrochemical cell or a self-assembled colloidal suspension, thus providing versatility for applications in different contexts.

Fig. 17
figure 17

Adapted with permission from Ref. [270]

Schematic diagram of a semi-artificial photosynthetic tandem PEC cell coupling water oxidation to CO2 reduction by D. vulgaris FDH. dpp, phosphonated diketopyrrolopyrrole dye, POs, [poly(1-vinylimidazole-coallylamine)-[Os(bipy)2Cl]Cl redox polymer], PS II, photosystem II. See text and Ref. [270] for details

Fig. 18
figure 18

Adapted with permission from Ref. [271]

Schematic diagram of a photocatalyst system for CO2 conversion using a dye-semiconductor-D. vulgaris FDH arrangement. ATR-IR, attenuated total reflection infrared, PFV, protein film voltammetry, QCM, quartz crystal microbalance, TEOA, triethanolamine. See text and Ref. [271] for details

Fig. 19
figure 19

Adapted with permission from Ref. [273]

Schematic diagram of a semi-artificial formate-hydrogen lyase system for the reversible and selective interconversion of dihydrogen and CO2 into formate using D. vulgaris FDH. The concept can be deployed in either an electrochemical cell (top) or a self-assembled colloidal suspension (bottom). H2ase, hydrogenase, ITO. indium tin oxide, NP, nanoparticle. See text and Ref. [273] for details

Acetogens and methanogens are organisms that reduce (fix) CO2 in vivo [274, 275], and, as such, they have been the focus of intense research to develop new CO2 converter devices, enzymatic and whole-cell systems, as is reviewed by Litty and Müller in this Book [152] and also [153,154,155,156,157,158,159]. Herein, we only highlight the dihydrogen-dependent CO2 reductases (Sect. 4.2.2.) from Acetobacterium woodii and T. kivui: the former is a SeCys–Mo–FDH that catalyses the CO2 hydrogenation with a kcat of 28 s−1 (reported as 10Umg−1; K HCO3−m  ≈ 37 mM) and displays slightly higher rates of formate oxidation (CO2 plus dihydrogen formation with a kcat  ≈  39 s−1, reported as 14Umg−1; K HCOO−m  ≈ 1 mM) [153, 154]; the second is a outstanding Cys–W–FDH that catalyses the CO2 hydrogenation with a kcat of 2.5 × 103s−1 (900 μmol formate min−1mg−1; K H2m  ≈ 130 μM—one of the fastest CO2 reducers so far described), with the reverse reaction being catalysed with a kcat of 2.7 × 103s−1 (930 μmol dihydrogen min−1mg−1; K HCOO−m  ≈ 550 μM) [156]. The CO2 hydrogenation equilibrium constant close to one (∆G º′ = 3.5KJmol−1) makes these systems ideal biocatalysts for dihydrogen storage and production. A. woodii was successfully used as a whole-cell biocatalyst to produce dihydrogen from formate, reaching a specific dihydrogen formation rate of ≈ 70 mmol g −1protein 1h−1 (≈ 30 mmol g −1cdw h−1) and a volumetric dihydrogen evolution rate of ≈ 80 mMh−1, with yields up to 1 mol dihydrogen per mol formate [155]. T. kivui was successfully exploited in a whole-cell system to convert dihydrogen plus CO2 (hydrogencarbonate) into formate, achieving a specific formate formation rate of ≈ 235 mmol g −1protein 1h−1 (≈ 150 mmol g −1cdw h−1) and a volumetric formate production rate of ≈ 270mMh−1; high titres up to 130 mM of formate were reached, with the key advantage of having the unwanted acetate formation abolished [159].

5 Outlook

The global energy demand and the present high dependence on fossil fuels have caused the increase in the atmospheric CO2 concentration for the highest values since records began. Due to its significant greenhouse effect, CO2 rise is responsible for large and unpredictable impacts on the Earth climate, besides being responsible for ocean acidification (its major sink). While some authors defend that these alterations are no longer reversible, the CO2 emissions must be greatly decelerate and new and more efficient “CO2 sinks” must be developed to avoid worsen this (already huge) “carbon crisis”. The three axes, storage/conversion/production, are envisaged by many authors as the best strategy to actively reduce the CO2 emissions, while actively consuming the CO2 already released—”two-in-one solution”. Along with chemical strategies, the “biochemical way” is proving is high value in different hybrid and biological systems to convert CO2 into fuels and VC. FDHs are efficient catalysts to reduce CO2 to formate and are in the right way to became key partners in the longed-for safe energy/stable climate solution.