Structural Insights Into CO₂ Transport Pathways in a W‐Formate Dehydrogenase: Structural Basis for CO₂ Reduction

ABSTRACT

Mo/W‐dependent formate dehydrogenases (Fdhs) catalyze the reversible reduction of CO₂ to formate and are key biocatalysts with high potential for CO₂ capture/conversion technologies. Although previous studies have suggested the presence of two substrate‐access tunnels in Fdhs, experimental evidence for CO₂‐specific pathways has been lacking. Here, we present an integrated study of Nitratidesulfovibrio vulgaris FdhAB combining crystallography, molecular dynamics simulations, mutagenesis, and kinetic assays. NvFdhAB crystals pressurized with Kr, O₂, and CO₂ were used to map gas diffusion routes and uncovered a substrate‐retention site consistently occupied by small molecules in multiple crystal structures. Our results indicate that both substrates mostly use the main tunnel to reach this retention site, but H₂O and CO₂ can also enter through a novel side branch before following a shared route to the buried W active site. The retention site, located at the junction of both tunnels, plays a synergistic role in enhancing CO₂ reduction by increasing substrate concentration near the catalytic center, thereby improving catalytic efficiency. Notably, variants affecting this site showed a selective effect for CO₂ reduction, with no impact on formate oxidation. These findings provide experimental evidence of a CO₂‐specific pathway and identify structural determinants underpinning efficient CO₂ reduction in this enzyme family.

Integrated structural analysis of N. vulgaris formate dehydrogenase AB, an enzyme with applications in climate change mitigation, is reported. The substrate/product diffusion pathways were fully mapped, and a retention site was identified that transiently holds substrates and has a key role in CO₂‐reducing activity.

1. Introduction

Climate change mitigation strategies are in high demand. Conversion of atmospheric CO₂ to added value products is one of the most challenging chemical reactions, due to CO₂ inherent stability, and requires extensive improvement toward the development of energy‐efficient industrial applications [1, 2, 3]. Metal‐dependent formate dehydrogenases (Mo/W‐Fdhs) are highly active, efficient, and selective enzymes capable of catalyzing CO₂ reduction to formate with remarkable effectiveness [4, 5, 6, 7]. There is an increasing interest in the development and deployment of natural and optimized versions of these enzymes, leading also to the design of inorganic bio‐inspired model compounds [8, 9, 10, 11, 12, 13].

Fdhs harboring a Mo/W catalytic site are significantly more active than their metal‐independent counterparts [3]. In metal‐dependent Fdhs, a Mo/W ion is coordinated by four sulfur atoms from the dithiolene moieties from two molybdopterin guanine dinucleotides (MGD), a (seleno)cysteine (SeCys/Cys) ligand from the polypeptide chain, and a terminal sulfido ligand (‐SH/═S). Although their active site is structurally characterized with several high‐resolution structures available, the catalytic mechanism remains controversial, with conflicting theories regarding the (Se)Cys dissociation during catalysis [14, 15] and the hydride (H⁺+2e⁻) transfer mechanism [16] still under debate. Several of these mechanistic hypotheses have also been explored with the help of computational approaches [17, 18, 19, 20].

FdhAB from Nitratidesulfovibrio vulgaris Hildenborough (previously Desulfovibrio vulgaris Hildenborough) harbors a W ion coordinated by a SeCys and is among the most efficient CO₂‐reducing enzymes characterized so far [5]. The main access tunnel (sometimes also referred to as the formate tunnel) is positively charged, suitable to accommodate formate molecules and reaches the solvent‐exposed catalytic pocket [5] (Figure S1). A secondary independent CO₂‐specific tunnel was initially proposed for NvFdhAB based on the existing structures and by comparison with the homologous formylmethanofuran dehydrogenase [5, 21]. However, to the best of our knowledge, no experimental evidence has been reported so far that could confirm the presence of a CO₂‐specific tunnel in metal‐dependent Fdhs.

Here, we report a crystallographic study on gas‐pressurized (Kr, O₂, and CO₂) NvFdhAB crystals, supported by molecular dynamics (MD) simulations, mutagenesis, and kinetic data. The results reveal a novel, unexplored tunnel that converges with the main tunnel at a single access point, leading to the W active site. In addition, the structures obtained from gas‐pressurized crystals unveiled a hydrophobic site within the positively charged tunnel, at the bifurcation point, which is found to be consistently occupied by small molecules (formamide, glycerol, and formate) in several other crystal structures of NvFdhAB [5, 22, 23]. We propose that this site functions as a CO₂ retention site, temporarily holding the substrate and playing a pivotal role in binding and catalysis. It synergistically enhances CO₂ reduction by increasing the local concentration of substrate near the active site, thereby boosting catalytic efficiency. Interestingly, this effect appears exclusive to CO₂ reduction, as formate oxidation remains unchanged in variants affecting this site.

2. Results and Discussion

2.1. NvFdhAB Crystals Pressurized With Kr, CO₂, and O₂ Give Insights Into Gas Access Pathways

In this work, we used the as‐isolated NvFdhAB for the crystallization assays, as reported by Oliveira et al. [5]. This form of NvFdhAB exhibits lower catalytic activity and substrate affinity [22, 23, 24], but its higher stability, combined with the use of high substrate concentrations [5, 22, 24], facilitated experiments aimed at experimentally mapping the enzyme access routes and elucidating how small molecules reach the buried active site (Tables 1 and S1). We employed a “soak‐and‐freeze” method that allows crystals to be processed under carbon dioxide, oxygen, and krypton atmospheres, using the HPMX (High Pressure Macromolecular Crystallography) laboratory at the ESRF [25, 26, 27].

TABLE 1.

Summary of the different in stillo (in the drop) and high‐pressure gas “soak‐and‐freeze” experiments performed with NvFdhAB.

Structure (resolution)	Anaerobic crystallization and crystal handling	In stillo soaking (conc; time)	High‐pressure soaking	Flash cooled	RMSD with NvFdhAB WT (#Cα)
Fdh_CO2_red PDB_ID: 9RJT (1.83 Å)	Yes	Sodium dithionite (1 mM; 3 min)	CO2 (58 bar)	He (58 bar)	0.33 Å (1177)
Fdh_CO2_ox PDB_ID: 9RJU (2.07 Å)	Yes	—	CO2 (50 bar)	He (50 bar)	0.27 Å (1176)
Fdh_O2 PDB_ID: 9RJV (1.83 Å)	No	—	O2 (55 bar)	O2 (55 bar)	0.21 Å (1177)
Fdh_Kr PDB_ID: 9RJW (1.64 Å)	No	—	Kr (100 bar)	Kr (100 bar)	0.23 Å (1178)
Fdh_FMD PDB_ID: 9RJX (1.73 Å)	No	Formamide (7 mM; 17 min)	—	Normal atmosphere	0.30 Å (1172)
Fdh_V197S PDB_ID: 9RJY (2.22 Å)	No	—	—	Normal atmosphere	0.28 Å (1176)
Fdh_Q447A PDB_ID: 9RJZ (2.12 Å)	No	—	—	Normal atmosphere	0.26 Å (1174)
Fdh_S194A PDB_ID: 9RK0 (2.24 Å)	No	—	—	Normal atmosphere	0.29 Å (1177)
Fdh_W491E PDB_ID: 9RK1 (1.99 Å)	No	—	—	Normal atmosphere	0.26 Å (1176)

The crystallization and CO₂ derivatization of as‐isolated (Fdh_CO₂_ox) and dithionite‐incubated (Fdh_CO₂_red) crystals was performed under anaerobic conditions and the crystals were transferred in a closed system using a miniature airlock device [28]. These conditions are required to prevent enzyme inactivation due to dissociation of the SeCys ligand from the metal, which was shown to occur when the enzyme is incubated with CO₂ or formate in the presence of oxygen [24]. Importantly, none of the gas derivatizations had any impact on the overall folding of the protein backbone (Table 1‐last column).

A total of 18 and 22 CO₂ molecules were modeled in Fdh_CO₂_ox and Fdh_CO₂_red structures, respectively (Figure S2), based on their distinct electron density shape. Additionally, 13 oxygen molecules and 13 Kr atoms were modeled in the Fdh_O₂ and Fdh_Kr structures, respectively (Figure S2) [25, 26, 27], providing structural details on the binding of gases to NvFdhAB. Kr was employed because it is easily identified (anomalous signal) and commonly used as a probe to identify CO₂ tunnels and binding sites in proteins [29, 30, 31]. The locations of the Kr atoms bound to the protein could be unambiguously identified through peaks in the anomalous difference Fourier map. Unlike the placement and modelling of the CO₂ and Kr ligands, unequivocal confirmation of the identity of all oxygen ligands was not possible at 1.8 Å resolution, due to the lack of anomalous signal and the similarity of their electron density shape to that of water molecules. Despite the added difficulty in unambiguously identifying O₂ binding sites at the available resolution, comparison of the modeled O₂ binding sites with previously reported ones in hydrogenases [31, 32], shows that, although diverse in nature, there is a preference for interactions with short and medium chain aliphatic residues (Ala, Val, Leu, and Ile) and hydroxyl‐containing residues (Ser and Thr); the same was found in Fdh_O₂, particularly at the retention site. Most of the modeled gas molecules are located close to the protein surface, probably constituting low‐affinity binding sites (Figure S2), some of which are found at the entrance of the main tunnel. In contrast, no gas molecules were detected in the previously proposed “CO₂ tunnel” [5] in any of the four gas‐soaking experiments (Figure 1).

FIGURE 1.

Superposition of the modeled ligand molecules in the main access tunnel (orange) and the previously proposed “CO₂ tunnel” (shaded green) in NvFdhAB [5]. In all panels, the W ion, sulfido group and U192 selenium atom are shown as spheres (light blue, yellow, and orange, respectively) and bonds as sticks. (a) Fdh_CO₂_ox with three CO₂ molecules (carbon in black and oxygen in red), one of which is in the retention site, highlighted with a violet dashed circle. (b) Fdh_CO₂_red structure featuring one single CO₂ molecule present in site 2, closer to the W site and highlighted with a red dashed ellipse. (c) Fdh_Kr structure showing the two Kr atoms (in violet) located at the bottom of the main tunnel, one of which also in the retention site. (d) Fdh_O₂ with two O₂ molecules shown in red, one of which is also in the retention site.

Structural analysis of the gas‐soaked crystal structures of NvFdhAB, together with tunnel calculations using CAVER [33] and AQUA–DUCT [34], allowed for the identification of a new offshoot of the main tunnel (Figure 2), designated as “new branch”. The conformations of the residues lining this branch do not change between the oxidized and the reduced forms of the enzyme (Figure S3). Interestingly, in previous structures (PDB_IDs: 8BQH, 8BQI, 8BQK, 8BQL, 8RC8, 8RCA, 9QM0, 9QM1) [23, 24, 35] glycerol, PEG, and ethylene glycol molecules, along with neatly lined water molecules, were consistently modeled within this tunnel. Despite not being present in all gas‐soaked structures, a CO₂ molecule was modeled inside this tunnel in the Fdh_CO₂_ox structure (Figure S3). Unlike the positively charged main tunnel, this branch is mostly lined with hydrophobic residues (e.g., W491, W492, F714, W728, V767, W765, I817, M818), suggesting that nonpolar CO₂ molecules can diffuse through this tunnel.

FIGURE 2.

Calculated tunnels for NvFdhAB. The main access tunnel (orange) (that includes the retention site) (calculated for the Fdh_CO₂_ox structure) and the novel CO₂‐specific branch (violet), highlighting the three CO₂ molecules, two in the main tunnel, one of which bound at the retention site (highlighted with a violet dashed circle) and another located in the novel CO₂ branch (in violet). The W active site and the residues lining the tunnels are shown as sticks; the W ion, sulfido group, and U192 selenium atom are shown as spheres (in light blue, yellow, and orange, respectively).

In the Fdh_CO₂_ox structure, another CO₂ molecule is found at the bifurcation point between the new branch and the main access tunnel in a hydrophobic patch (hereafter called retention site, Site 1) surrounded by highly conserved residues H193, R441, V197, Q447, and T450 (Figure 3a). Moreover, two O₂ molecules and two Kr atoms were also modeled near this region, with one O₂ and one Kr bound at this retention site (Figure 3b–d). In the case of Fdh_CO₂_red, no CO₂ molecule was observed in the retention site, now occupied by the side chain of the catalytic H193, adopting a different rotamer in the NvFdhAB reduced form (Figure 3e) [5]. Surprisingly, in the Fdh_CO₂_red structure (Figures 3e and S4), while H193 and SeCys side chains were found in the reduced form conformation, as expected (PDB ID: 6SDV) [5], residues E443, Q890, and the MGD2 were already found in the oxidized form conformation (Figure S4b), probably due to partial oxidation of the enzyme by CO₂. The resulting structure likely corresponds to an intermediate state of the CO₂ reduction reaction. Notably, this form resembles one of the intermediate states captured (5_min (PDB ID: 8BQJ)) in our previously reported cryo‐trapping experiments of NvFdhAB reduction with formate [23]. Furthermore, in this structure (Fdh_CO₂_red), one CO₂ molecule is found closer to the active site (5 Å away from the W atom) (Site 2) (Figure S5). This location is new, as no other structure or theoretical calculation has placed a substrate molecule in this region of the outer coordination shell of the metal (W/Mo). It likely corresponds to an artifact resulting from the intense CO₂ pressure, as this site is occupied by the hydrophobic side chain of M405 in the activated state of NvFdhAB, studied using the variant C872A as a proxy [22]. In fact, M405 plays a crucial role in conveying the stimulus from the surface exposed S–S bond to the buried active site, by altering the conformation of the active site, thus being integral for the correct operation of the previously described redox switch [22]. The classification of this site as an artifact is further supported by the existence of only nonspecific interactions established between the CO₂ molecule and the protein (L440 carbonyl, side chain of A404, one water molecule and several long‐range contacts (>3.5 Å)). These findings contribute to rule out any specific physiological role for Site 2.

FIGURE 3.

Environment and interactions between the modeled ligand molecules and the surrounding residues close to the active site in nine crystal structures of ligand‐bound NvFdhAB. In all panels, the W active site and the surrounding residues are shown as sticks; the W ion, sulfido group, and U192 selenium atom are shown as spheres (light blue, yellow, and orange, respectively). 2Fo–Fc electron density maps, at 1σ, are shown as blue mesh. Ligand‐W distance, hydrogen bonds and electrostatic interactions are shown as black, teal, and orange dashes, respectively. Distances are in Å. (a) Fdh_CO₂_ox (red) with one CO₂ molecule bound at the retention site. (b) Fdh_Kr (cyan) with two Kr atoms (violet spheres) at the main tunnel site shown, the 2Fo–Fc map (blue mesh, contoured at 1σ) and the anomalous map peaks for the two Kr atoms, the W ion, and the Se atom shown (green mesh, contoured at 3σ). (c) Superposition of seven crystal structures at the retention site to highlight the conservation of the side chains and ligands (native NvFdhAB (PDB ID: 6SDR) (green) [5], Fdh_Kr (cyan), Fdh_O₂ (pink), Fdh_CO₂_ox (red), Fdh_CO₂_red (black), Fdh_FMD (brown), Fdh_FMT (blue)). (d) Fdh_O₂ (pink) with two dioxygen molecules shown (red sticks) bound at the retention site. (e) Fdh_CO₂_red (black) with H193 sidechain occupying the retention site. (f) Fdh_FMD (brown) with a formamide (FMD) molecule at the retention site. (g) Fdh_FMT (PDB ID: 8BQG) (blue) with one formate (FMT) molecule at the retention site [23]. (h) Native NvFdhAB (PDB ID: 6SDR) (green) with one glycerol (GOL) molecule at the retention site [5]. (i) Fdh_C872A_FMD variant, in complex with formamide (PDB ID: 8CM6) (dark green), also at the retention site [22], superimposed with Fdh from Methylorubrum extorquens AM1 (PDB ID: 8J83) (light brown) [36]. (j) [5]Same as in (c), showing the protein surface (gray) and highlighting the main substrate access tunnel and the retention site.

In addition to the static structural analysis of gas‐soaked crystal structures, we employed water, formate and CO₂ molecules as molecular probes to characterize tunnel architecture and flow in the dynamic structure of NvFdhAB using AQUA–DUCT software [34] based on MD simulations of its active form (PDB ID: 8CM6), conducted in water with either CO₂ or sodium formate present in separate systems (Supporting Information Note 1, Figures 4 and S6–S13 and Tables S2 and S3). In these analyses, all CO₂, HCOO⁻, and water molecules were traced passing through the protein core—defined as a 6 Å spherical zone from the center of mass of residue V197 (Figure 4d).

FIGURE 4.

Structural and tunnel analysis of NvFdhA using CAVER (structure) and AQUA–DUCT (1^st replica of MD simulations). The W ion is shown as a dark blue sphere; NvFdhA protein is shown as a gray cartoon or surface. (a) Main access tunnel (orange) and a new branch (violet) calculated using CAVER from the Fdh_CO₂_ox crystal structure (PDB_ID: 9RJU). (b) AQUA–DUCT analysis using CO₂ molecules as probes. Clusters of inlets marking tunnel entries: main/formate (orange), and new branch (violet). Traced molecule paths indicating tunnel shapes are shown as gray lines. (c) AQUA–DUCT analysis using formate molecules as probes. Clusters of inlets marking tunnel entries: main (orange). Traced molecule paths are shown as gray lines. (d) Passage of a single smoothed CO₂ path (path no. 14) entering the new branch, passing through the retention site (marked as a spherical zone of 6 Å around V197 shown as an orange shape), and exiting via the main tunnel. Retention site residues are shown as black sticks. CO₂ molecules modeled in the Fdh_CO₂_ox structure are shown as spheres (red and gray). Clusters of tunnel entries/exits are color‐coded as above. The single path is color‐coded: red—incoming, green—within the retention site, blue—outgoing. Clusters of inlets of CO₂ are shown together with path no. 18 and two CO₂ molecules from the Fdh_CO₂_ox structure. (e) Clusters of CO₂ inlets on the surface of NvFdhA, together with pie chart showing the sizes of CO₂ inlet clusters, indicated by the corresponding percentage of CO₂ molecules passing through a given tunnel entry. (f) Clusters of HCOO⁻ inlets on the surface of NvFdhA, together with a pie chart showing the sizes of formate inlet clusters, indicated by the corresponding percentage of formate molecules passing through a given tunnel entry. (g) Clusters of water inlets on the surface of NvFdhA—from MD simulations of NvFdhA with CO₂, together with a pie chart showing the sizes of water inlet clusters, indicated by the corresponding percentage of water molecules passing through a given tunnel entry. (h) Clusters of water inlets on the surface of NvFdhA—from MD simulations of NvFdhA with HCOO⁻, together with a pie chart showing the sizes of water inlet clusters, indicated by the corresponding percentage of water molecules passing through a given tunnel entry.

The analyses confirmed the presence of the main tunnel (orange) with a hydrophobic branch (violet), both accessible to water and CO₂ (Figures 4b,e,g,h and S12), whereas formate was transported exclusively via the main tunnel (Figures 4c,f and S13), as marked by clusters of inlets on the surface of the NvFdhAB. Interestingly, CO₂ molecules were observed to pass through the “new branch” only in the first of three MD simulation replicas (Figures S12 and S17), whereas water was able to permeate the “new branch” in the second and third replicas, albeit with roughly 50% fewer inlets (Table S2). The total number of inlets for both CO₂ and water passing through the defined object was similar across all replicas (Table S2), which indicated good reproducibility of simulation results.

The main tunnel (in orange) is very wide and is used by 96.3%–97.1% of water and by at least 90% of CO₂ molecules passing through the defined protein core, respectively. The “new branch” tunnel (in violet) is the second most transversed, used by up to 3.6% of water molecules (Tables S2 and S3). Quantitatively, more water molecules passed through both tunnels than CO₂ or HCOO⁻ molecules, as the analyzed MD simulations contained only 100 CO₂ or 58 HCOO⁻ molecules, respectively. However, for all ligands (water, CO₂, and HCOO⁻), the orange main tunnel is the foremost route used for molecular transport (Figure 4 g,h). Nevertheless, in the case of the first replica of MD with CO₂, the violet “new branch” tunnel is used more frequently (7.4% of molecules) than in the case of water (2.8%–3.6%) (Figure 4e,g). These observations suggest that the additional tunnel may play only a minor role in CO₂ delivery but is responsible for the transport of H₂O to and from the active site.

The MD simulation also confirmed occupancy of the retention site by CO₂ molecules, as shown by a selected single path of CO₂ molecule showing clear retention in the place where CO₂ was modeled in the retention site of the Fdh_CO₂_ox structure (Figures 4d and S17). The presence of the second CO₂ in the crystal structure new branch suggests that the passage through that channel can be hindered by gating residues or strong interactions with tunnel‐lining residues. In the AQUA–DUCT analysis, in some of the CO₂ passage events, it is visible that the molecule was trapped in this channel in the vicinity of the modeled CO₂ residue (Figure 4d).

Lastly, the results of the structural analysis of gas‐soaked NvFdhAB crystals experiments, combined with MD simulations, agree that the previously proposed independent “CO₂ tunnel” [5] is not functional in NvFdhAB. Furthermore, this tunnel appears to be obstructed in the active form of the enzyme as previously reported (PDB ID: 8CM6) [22]. Nonetheless, the current results cannot fully exclude the possibility that an allosteric mechanism, induced by the transit of substrate/product molecules through the previously proposed “CO₂ tunnel”, may transiently open or stabilize it.

2.2. The Retention Site in the Main Tunnel of NvFdhAB Is Supported by Additional Experiments

NvFdhAB crystals were also soaked with formamide (FMD), a substrate analogue. The resulting structure revealed a formamide molecule present also in the retention site, forming a hydrogen bond with R441 (R441 NH₂—Formamide NH₂, 3.5 Å) and contacting H193 (3.1 Å) (Figure 3f). Structures obtained in previous works [5, 22, 23] (native NvFdhAB (PDB ID: 6SDR), NvFdhAB in complex with formate (PDB ID: 8BQG), and NvFdhAB in the activated form (C872A variant) in complex with formamide (PDB ID: 8CM6)) also show the presence of small molecules binding to this unique site and interacting similarly with the same conserved residues (R441, H193, Q447, V197, T450) from the main tunnel (Figure 3g,h,i).

Considering the binding of six different ligands in the very same region of the main tunnel (Figure 3c,j) and the almost complete conservation of the residues lining this site (Table 2), we propose that the retention site has an important functional role. This site may serve to transiently hold the substrate near the active site, functioning as a pre‐chamber that locally accumulates the substrate and delivers it when required.

TABLE 2.

Degree of conservation of the residues lining the proposed retention site cavity and interacting with ligand molecules in the main tunnel of NvFdhAB. The right‐handed column lists the degree of conservation of these residues on the subset of Fdh enzymes that have a k _cat higher than 11 s⁻¹, for the reduction of CO₂.

Residue	Degree of conservation in all Mo/W‐Fdhs [5]	Degree of conservation on highly active Fdhs for CO2 reduction (9 seqs) (this work)
H193	96%	100%
R441	96%	100%
V197	96%	89% V/11% G
Q447	100%	100%
T450	88%	33% T/45% C/22% S

Interestingly, this structurally conserved region is also present in most Fdhs, such as Methylorubrum extorquens FDH1 [36] and the Fdh subunit of Thermoanaerobacter kivui hydrogen‐dependent CO₂ reductase [37], which do not contain the double cysteine motif required for activation [22], (Figure 3i), thus hinting that this retention site is likely a general feature among metal‐dependent Fdhs.

In the NvFdhAB reduced forms (Fdh_CO₂_red and formate‐reduced NvFdhAB (PDB 6SDV)) the H193 side chain displaces these molecules by occupying the retention site and opening room in the vicinity of the metal (Figure 3e). The hypothesis that the electron densities found in this region of the tunnel are not ligands, but a minority of populations of the reduced conformation of H193, can be discarded as it is supported by the distinct electron density of several ligands (formate and formamide), as well as by the presence of a clear anomalous signal in Fdh_Kr, that marks the presence of Kr atoms (Figure 3b).

2.3. The Retention Site Plays a Key Role in CO₂ Reduction Activity

The identification of a nonpolar tunnel, leading to the retention site, led us to propose its direct involvement in mediating CO₂ diffusion from the solvent to the buried active site. Seeking confirmation, we mutated W491 to a glutamate residue, to induce a salt bridge with R769, located across the entry of the tunnel, thus aiming to block the putative CO₂ pathway. The crystal structure for the W491E variant reveals an almost perfect superposition with the native NvFdhAB (Table 1 and Figure 5a), retaining the backbone conformation. As expected, E491 rotates inward to form an electrostatic interaction with R769 at 4.3 Å and mediated by two ordered water molecules (Figure 5b). These water molecules likely create further hindrance to the flow of CO₂ through this bottleneck.

FIGURE 5.

The engineered partial blockade at the solvent‐exposed entry of the newly identified branch (violet) caused by the mutation of W491 to a glutamate residue. (a) Superposition of the Fdh_CO₂_ox (red) and Fdh_W491E (dark blue) structures. The W active site and the residues lining the tunnels are shown as sticks; the W ion, sulfido group, and U192 selenium atom are shown as spheres (in light blue, yellow, and orange, respectively). (b) Same as in (a), showing a zoomed‐in view of the end region of the new CO₂ tunnel (green dashed circle), highlighting the E491 and R769 interaction mediated by two water molecules. Electrostatic and hydrogen bonding interactions are shown as black and teal dashes, respectively. Distances are in Å.

Consistent with this, activity assays show that this substitution affects the CO₂ reduction reaction (53‐fold lower catalytic efficiency than the WT protein), while the reverse reaction is affected to a lesser extent (Table 3 and Figure 6).

TABLE 3.

Kinetic parameters of NvFdhAB and variants for both formate oxidation and CO₂ reduction. Values represent mean ± s.d. (n ≥ 3 technical assay replicates).

	Formate oxidation			CO2 reduction
Variant	k cat (s−1)	K M (µM)	k cat/K M (s−1 mM−1)	k cat (s−1)	K M (µM)	k cat/K M (s−1 mM−1)	Reference
WT	1 310 ± 50 (100% WT)	16.9 ± 2.8	77 515	340 ± 40 (100% WT)	320 ± 50	1 063	[5, 22]
W491E	1180 ± 85 (90% WT)	120 ± 30	9833	230 ± 25 (68% WT)	11 600 ± 1 800	20	This work
Q447A	370 ± 90 (28% WT)	67 ± 10	5 522	23 ± 2 (7% WT)	5 800 ± 875	4	This work
S194A	220 ± 20 (16% WT)	790 ± 10	278	60 ± 20 (18% WT)	680 ± 70	88	This work
H457R	330 ± 60 (25% WT)	70 ± 20	4 459	80 ± 20 (24% WT)	490 ± 95	163	This work
V197S	1 080 ± 15 (82% WT)	314 ± 3	3 439	11 ± 2 (3% WT)	5 800 ± 1 570	2	This work

FIGURE 6.

Relative activities of the NvFdhAB variants compared to the WT. Relative activities of the variants V197S, Q447A, S194A, H457R, and W491E compared to WT (data from [22]) for both formate oxidation (in black) and CO₂ reduction (in gray). Data are presented as mean values ± s.d. (n ≥ 3 assay technical replicates). Standard deviations are shown.

To elucidate the structural mechanism underlying this profound impact (which is remarkable given the ∼30 Å distance between the mutated residue W491 and the active site), we performed MD simulations in a water box containing CO₂ (Figures S14–S16).

The analysis of E491 geometry confirmed its interaction with R769, which can be even tighter than that observed in the crystal structure (Figure S15). This results in restricted CO₂ penetration and decreased water flow through the new branch (Figure S14 and Table S4). However, E491 is also able to make contact with R206, which does not hamper water flow through the side channel. Additionally, the mutation W491E increases the flexibility of two surface loops 300–350 and 475–491 (Figure S16), which seems to result in an increased H₂O and CO₂ flow through the main channel, especially in 2^nd replica of the MD simulation, where the E491–R769 salt bridge persisted for most of the trajectory. As a result, the mutant conformation with a closed side tunnel exhibits 20%–30% increased water and CO₂ transfer through the main channel compared to the structures with an open side tunnel (Figure S14 and Table S4).

As a result, the reported impact on catalysis, specifically on CO₂ reduction, is most likely associated with changes in k _on and k _off rate for CO₂ through the enzyme channel. Consequently, our hypothesis is that this new hydrophobic branch can assist the main substrate tunnel in transporting CO₂ to the active site, offering a parallel route and thus providing a moderate enhancement in catalytic performance, as supported by the MD simulations, but is also a facile route for H₂O diffusion to and from the active site.

The consistently similar binding of different molecules at the retention site prompted us to investigate its putative role. The sequence alignment [5] shows a high degree of conservation among several residues lining this region, particularly those interacting with the ligands, H193, R441, V197, Q447, and T450 (Table 2). While H193 and R441 are well known to be catalytically essential, we selected V197 and Q447 (96% and 100% conservation, respectively) to perform mutagenesis studies and produced the variants V197S and Q447A. The crystal structures for both variants show a virtually complete superposition with the native NvFdhAB (Table 1), preserving the backbone conformation of the mutated residues, while only altering the chemical nature of their side chains (Figure 7a,b,c). In the case of the Q447A variant, the absence of the Gln side chain results in a less constricted tunnel (Figure 7d,e).

FIGURE 7.

Impact of the NvFdhAB variants V197S and Q447A in the main tunnel. The displayed surfaces are colored according to their hydrophobicity, ranging from highly hydrophilic (orange), to highly hydrophobic (teal). The green arrow indicates the location and orientation of the main access tunnel (from the surface toward the W active site). (a) The solvent‐exposed opening of the main access tunnel of NvFdhAB as‐isolated (PDB ID: 6SDR) is shown as a surface. (b) Surface representation of the main access tunnel of NvFdhAB as‐isolated (PDB ID: 6SDR), where V197 is highlighted in magenta. (c) Surface representation of the main access tunnel of Fdh_V197S, where S197 is highlighted in magenta. (d) NvFdhAB as‐isolated (PDB ID: 6SDR) shown with color‐coded atoms and hydrophobicity‐colored surface. Q447 is highlighted in purple. (e) Fdh_Q447A is shown both as sticks and as a hydrophobicity‐colored surface. A447 is highlighted in purple. (f) Superposition of NvFdhAB WT oxidized (gray) (PDB ID: 6SDR) and variant V197S (violet). The 2Fo–Fc electron density map for the S197 residue is shown contoured at 1σ (blue mesh). (g) Superposition of NvFdhAB WT oxidized (gray) (PDB ID: 6SDR) and variant Q447A (orange). The 2Fo–Fc electron density map for the A447 residue is shown contoured at 1σ (blue mesh).

Remarkably, activity assays revealed an almost complete loss of CO₂ reduction activity for both V197S (3% of WT) and Q447A (7% of WT), while a significant formate oxidation activity was retained (V197S and Q447A 82% and 28% of WT, respectively) (Table 3 and Figure 6). Additionally, both variants presented a markedly reduced affinity for CO₂ as reflected by a drastic increase in K _M values that are identical: 5800 µM for V197S and Q447A (Table 3 and Figure S18). The K _M for formate also increases for both variants: 67 µM for Q447A, and 314 µM for V197S (Table 3 and Figure S18). MD simulations on the V197S variant show that this mutation affects the binding of CO₂ to the retention site, making it less favorable and with a shorter interaction time (Figures S8c,d and S10).

Two additional variants, the S194A and H457R, both located very close but outside of the retention site, were also constructed and characterized structurally (for S194A only) and kinetically (Supporting Information Note 2 and Figure S19); these variants show a loss of activity for both reactions but do not exhibit any selectivity.

Moreover, a sequence alignment (Table S5) was performed using highly active Fdhs with k _cat values above 11 s⁻¹ for CO₂ reduction, which is the k _cat of V197S, and represents a conservative estimate of the maximum activity achievable without a fully functional retention site. This alignment shows that, in addition to H193 and R441, residue Q447 is also fully conserved within this highly active subset of Fdhs, while V197 is replaced by a Gly only in Rhodobacter capsulatus Fdh [38]. T450 is 33% conserved, while the remaining sequences contain 45% of cysteine and 22% of serine in position 450, residues that are expected to retain the role of a threonine, by interacting with the CO₂ molecule through a thiol/hydroxyl group.

These results strongly suggest that conserved residues V197 and Q447 play a key role in the catalytic conversion of CO₂ to formate, supporting our hypothesis on the importance of the retention site in facilitating CO₂ loading and concentration close to the active site [39, 40]. Given the distance between the retention site and the active site (approximately 9 Å from the W atom), it is unlikely that V197 and Q447 are directly involved in the catalytic mechanism. The identification of a retention site on NvFdhAB does not support nor exclude any of the two current hypotheses, of whether the reaction occurs on the first or the second coordination spheres of the W, nor does it provide any particular data to indicate if the hydride transfer occurs to the sulfido ligand or to the metal itself. The retention site acts as a substrate‐holding reservoir positioned close enough to the active site to enable rapid transfer into the catalytic orientation. The presence of this pre‐catalytic staging ground reduces the down time of the active site, the interval during which the catalytic center would otherwise be unoccupied and diffusion‐limited, being significantly faster than recruitment from bulk solvent and thus effectively decoupling catalytic turnover from substrate diffusion. This benefit applies regardless of whether catalysis occurs in the first or the second coordination sphere as the retention site ensures that substrate molecules are locally concentrated, thereby accelerating transition into the reactive orientation without requiring a full diffusion‐rebinding event.

The results support that the ingress of both substrates (CO₂ and formate) to the W site is conducted through the main tunnel. The oxidation of formate is much less affected by these mutations due to the positively charged nature of the main tunnel, which may result in a higher affinity for the negatively charged formate (at physiological pH), without the need for highly specific transient binding sites. Regarding the neutral CO₂, this molecule is less efficiently guided by the positively charged tunnel and requires more specific diffusion routes (CO₂ tunnel and retention site). The presence of the V197 and Q447 residues appears necessary to effectively steer CO₂ toward the active site. While the new hydrophobic branch may not be strictly essential to allow access of CO₂ to the active site, as the main tunnel handles most of the CO₂ transport, it does offer a moderate enhancement, as suggested by the mutagenesis studies.

Our hypothesis is that the main substrate tunnel is mostly responsible for the transport of both substrates to the proposed retention site, with the new branch assisting in CO₂ and H₂O transport in parallel. The retention site acts as a pre‐catalytic staging ground, with V197 and Q447 residues necessary to effectively steer CO₂ toward the active site, while both substrates share the final segment of the main tunnel to access the buried tungsten active site. This mechanism bears resemblance to the tunnel that mediates efficient CO channeling in the bifunctional CO‐dehydrogenase/acetyl‐CoA synthase complex [39].

3. Conclusion

In conclusion, we have experimentally mapped gas diffusion routes in N. vulgaris FdhAB, and we disclosed a hydrophobic offshoot branching of the main tunnel. The main tunnel was shown to handle formate transport and most of the CO₂ transport, while the new branch is also used by CO₂ and H₂O in parallel to the main tunnel, offering a moderate enhancement in CO₂ reduction, as supported by mutagenesis studies and MD simulations. Furthermore, we found a highly conserved hydrophobic site region at the branching region, which exhibits affinity toward different small ligands (including CO₂ and substrate analogues). We propose that this region functions as a retention site that transiently holds the substrate prior to catalysis, thereby synergistically increasing the efficiency of the CO₂ reduction reaction. This hypothesis is strengthened by the discovery that two highly conserved residues involved in such interactions, V197 and Q447, are crucial for CO₂ reduction activity, but not for formate oxidation. These mechanistic insights into substrate diffusion in NvFdhAB contribute to the clarification of how the enzyme handles formate and CO₂, and how this relates to catalysis and specificity. This knowledge can now be leveraged to engineer NvFdhAB variants with enhanced CO₂ reduction activity.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

The necessary files required for re‐run of the MD simulation can be found as Supporting Information to this article. The raw trajectories reported are available from the authors upon reasonable request. The authors have cited additional references within the Supporting Information [5, 9, 25, 26, 27, 34, 41, 42, 43, 44, 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, 72]. Supplementary file 1: anie71673‐sup‐0001‐SuppMat.pdf.

Supplementary file 2: anie71673‐sup‐0002‐Data.zip.

Data Availability Statement

The atomic coordinates and structure factors for the x‐ray crystal structures reported in this study have been deposited in the Protein Data Bank (https://www.rcsb.org) under accession codes 9RJT, 9RJU, 9RJV, 9RJW, 9RJX, 9RJY, 9RJZ, 9RK0 and 9RK1. The respective raw data are deposited at the Xtal Raw Data Archive (https://xrda.pdbj.org) under the same accession codes. Additional data are available from the corresponding author upon reasonable request.