Spin Polarization Reveals the Coordination Geometry of the [FeFe] Hydrogenase Active Site in Its CO-Inhibited State.

The active site of [FeFe] hydrogenase features a binuclear iron cofactor Fe2ADT(CO)3(CN)2, where ADT represents the bridging ligand aza-propane-dithiolate. The terminal diatomic ligands all coordinate in a basal configuration, and one CO bridges the two irons leaving an open coordination site at which the hydrogen species and the competitive inhibitor CO bind. Externally supplied CO is expected to coordinate in an apical configuration. However, an alternative configuration has been proposed in which, due to ligand rotation, the CN- bound to the distal Fe becomes apical. Using selective 13C isotope labeling of the CN- and COext ligands in combination with pulsed 13C electron-nuclear-nuclear triple resonance spectroscopy, spin polarization effects are revealed that, according to density functional theory calculations, are consistent with only the "unrotated" apical COext configuration.

[FeFe] hydrogenases are the most active enzymes catalyzing the reversible conversion of protons and electrons into molecular hydrogen.[1] The active site of these enzymes is highly conserved and is termed the “H-cluster” (see Figure ). It consists of a unique binuclear iron center [2Fe]H coordinated by two terminal CN– ligands and three CO ligands, one of which bridges the two iron atoms. In addition, the two iron atoms are bridged by an aza-propane-dithiolate ligand containing a secondary amine moiety in its bridgehead. This amine acts as a proton shuttle connecting the [2Fe]H cofactor to the proton channel of the enzyme. The [2Fe]H cofactor is covalently linked to a generic [4Fe-4S] cluster through the bridging thiol group of one of the coordinating cysteines. This [4Fe-4S]H subcluster is part of the H-cluster and connects the active site to the electron transport chain of the enzyme. It is generally assumed that the oxidized state of the H-cluster, “Hox”, is the entry point to the catalytic cycle.[2] Externally supplied CO acts as a competitive inhibitor[3] of the enzyme and binds to the open coordination site at the distal iron Fed in the oxidized state forming the “Hox-CO” state (Figure ).

Figure 1

Structure of the H-cluster with coordinating protein residues according to Protein Data Bank entry 4XDC (CpI from Clostridium pasteurianum).[4] COext is overlaid from the structure of Protein Data Bank entry 1C4C.[5] “Fep” and “Fed” are the irons proximal and distal to the [4Fe-4S]H subcluster, respectively.

Both Hox and Hox-CO are paramagnetic and characterized by a mixed valence Fe(I)Fe(II) configuration of binuclear subsite [2Fe]H. The characteristic axial electron paramagnetic resonance (EPR) spectrum of Hox-CO is displayed in Figure a. It is generally assumed that the extraneous CO of the Hox-CO state binds in the open coordination site in an apical coordination.[1,6] Some spectroscopic data [Fourier transform infrared (FTIR) and X-ray emission spectroscopy] in combination with DFT/QM-MM analysis, however, seem to suggest an alternative coordination in which the extraneous CO rotates to the basal (equatorial) position and the distal CN– ligand adopts an apical (axial) coordination.[7−9] It is known from our previous studies that the H-cluster in its Hox and Hox-CO states exhibits complex magnetic interactions featuring extensive spin delocalization and polarization over the iron core[10] and limited amounts of spin delocalization and polarization over its ligands.[11,12] Therefore, we investigated whether these magnetic interactions can be used to distinguish apical versus basal coordination modes of the CO/CN– ligands by focusing on the signs (positive or negative) of their 13C electron–nuclear hyperfine interactions (HFIs).

Figure 4

(A) EPR spectrum of Hox-CO. (B) The black trace is the Mims-TRIPLE spectrum of the 13CN–-labeled H-cluster in the Hox-CO state (g∥, 1185 mT) with pumping pulse RF1 positioned off resonance at the 13C Larmor frequency. The red trace is from the same experiment with RF1 positioned at 14.5 MHz (high-frequency line of the inner doublet). The TRIPLE effect on the high-frequency line of the outer doublet is denoted with an asterisk. The blue trace is the difference spectrum (black minus red). (C) The black trace is the Davies TRIPLE of the 13CN–- and 13COext-labeled H-cluster in the Hox-CO state (g⊥, 1209 mT) with pumping pulse RF1 positioned off resonance at the 13C Larmor frequency. The dotted red trace is from the same experiment with RF1 positioned at 11 MHz (low-frequency line of the inner 13CN– doublet). The TRIPLE effect on the low-frequency line of the outer 13CN– doublet as well as the high-frequency line of the 13COext doublet is denoted with an asterisk. The blue trace is the difference spectrum (black minus red). All spectra were recorded at 17 K. Mims sequence: MW pulse = 20 ns, τ = 200 ns, and TRF = 60 μs. Davies sequence: preparation pulse = 100 ns, π/2 pulse = 40 ns, τ = 200 ns, and TRF = 60 μs.

To this end, we constructed several in silico molecular models for each of the two CO/CN– configurations. The simplest models in which the [4Fe-4S]2+ subcluster is truncated to an SH-ethyl moiety are displayed in Figure . Extended models including the [4Fe-4S]H subcluster as well as selected amino acid side groups from the protein pocket are discussed in the Supporting Information. The quantum chemical calculations on these extended models provide additional insights into the electronic structure of the H-cluster but essentially predict the same spin distribution over the [2Fe]H subcluster as the small (truncated) models. Model S-COa represents the coordination configuration in which the extraneous CO is coordinated in an apical mode, while model S-COb represents the alternative basal coordination of the extraneous CO ligand, i.e., with the CNd ligand in apical coordination mode. These models are used in quantum chemical calculations aimed at distinguishing these ligand configurations through their effect on the magnitudes and signs of the magnetic parameters of the H-cluster. The calculations were performed using ORCA 4.11 on the density functional theory (DFT) level and used the BP86 functional and zora-def2-TZVP basis set for the geometry optimizations. EPR property calculations were performed using the same basis set and the B3LYP as well as the BP86 functional, including relativistic and dispersion corrections (details in the Supporting Information).

Figure 2

DFT models. S-COa represents the apical coordination of COext, while S-COb depicts the basal coordination of COext.

According to earlier DFT studies of the H-cluster, both irons in [2Fe]H have a distorted octahedral coordination geometry and are in a low-spin configuration, while the spin density is mainly concentrated on the iron centers with smaller contributions on the ligands.[13,14] For Hox-CO, both Fep and Fed contribute to the SOMO that mainly consists of an antisymmetric linear combination of the d orbitals from both Fe centers.[13] Our quantum chemical calculations on both ligand configurations (S-COa and S-COb) are fully consistent with the analysis of Fiedler and Brunold[13] and provided, in addition, the HFI values for the basal 13CO/13CN– ligands. Figure a shows the SOMO for model S-COa confirming that the carbon centers of the basal CO/CN– ligands do not participate in the SOMO. The SOMO for model S-COb is virtually identical to that of S-COa (not shown). Figure b depicts the relevant spin populations on the selected Fe and carbon centers. Indeed, the centers participating in the SOMO (Fep, Fed, COb, and COext) have positive spin population. The spin population on the basal CO/CN– carbons is negative due to “spin polarization”. One can imagine that the Fe-CN– σ-bonds are polarized by the magnetic iron d-orbitals (participating in the SOMO) in a manner similar to that of the proton σ-bonds in aromatic radicals.[15]

Figure 3

(a) SOMO of model S-COa that shows only density on COext. (b) Mulliken spin population on the Fe centers and 13CO/13CN ligands. The carbons of the basal ligands have a negative spin population, while the carbon of the apical COext has a positive spin population.

The calculated g tensor and HFIs of the 57Fe and 13C nuclei are listed in Table . All models predict negative isotropic HFIs for the basal 13CO/13CN– ligands (colored blue in Figure and Table ), while the apical 13CO/13CN– ligand has positive HFIs (colored red in Figure and Table ). This sign difference in basal versus apical HFI allows us to experimentally distinguish the two proposed ligand configurations for Hox-CO displayed in Figure .

Table 1

Experimental and DFT-Predicted Magnetic Parameters (isotropic 57Fe, 13C HFI in megahertz) for Hox-CO with Apical and Basal COext (models S-COa and S-COb)a

HFI Aiso values for apical and basal ligands are colored red and blue, respectively. The experimental 57Fe HFIs are obtained from ref (19).

The magnitudes of the 13C HFI values for the CN– and CO ligands in the Hox-CO state have previously been experimentally determined (Table ), but their relative signs have not.[16−18] The two 13CN– couplings are on the same order of magnitude (7.3 MHz vs 4.0 MHz) and seem to reflect the relative spin populations on the two iron centers (0.62 vs 0.24). The 13C HFI of the extraneous CO is significantly larger (17.1 MHz), which is consistent with the DFT predictions for conformation “COa” (basal CN–). Overall, however, the DFT calculations overestimate the magnitude of the 13CN– and 13CO HFI tensors. Interestingly, the relative magnitudes of the 13CO/13CN– HFIs are better reproduced in the BP86 calculations than for the B3LYP results. Overall, the trend reflected in the DFT calculations clearly favors the conserved conformation in which the extraneous CO binds and remains in an apical position on Fed.

To further corroborate the assignment of Hox-CO to conformation “S-COa” with COext in the apical position, we set out to determine the relative signs of the 13C HFI tensors and performed electron–nuclear–nuclear triple resonance (TRIPLE)[20] experiments (see the Supporting Information and Scheme S1) using Chlamydomonas reinhardtii (CrHydA1) samples obtained through artificial maturation with synthetic precursors labeled with 13CN–[21,22] and inhibited with either natural abundance CO or 13CO (see Figure S1 for FTIR spectra of the samples). In a TRIPLE experiment, an additional stationary RF pulse is applied during the regular Mims[23] or Davies sequence[24] (see the Supporting Information). If this pulse is resonant with one of the nuclear spin transitions, the ENDOR efficiency of all nuclear spin transitions from the same mS manifold is reduced. In the difference (off resonance minus on resonance) spectrum, these transitions show up as (weak) positive contributions. For 13C nuclei (I = 1/2), the observed ENDOR transitions are given as νENDOR = νI + mIA, resulting in a ν± = νI ± A/2 doublet. Transitions with a TRIPLE effect on the same side of the doublet as the pumped transition (i.e., both larger or both smaller than the center frequency) belong to nuclear spins having an HFI with the same sign as the pumped nucleus. If a TRIPLE effect occurs on the opposite branch of the doublet, the HFI signs are opposite. Figure B presents a Mims TRIPLE experiment on (13CN–)2-labeled CrHydA1. The spectra show the two 13C doublets of similar HFI splitting originating from the two CN– ligands. The standard Mims ENDOR spectrum[23] is colored black but is partly hidden by the overlaid TRIPLE experiment indicated as a red dotted line. The regular ENDOR spectrum is obtained by setting the additional RF1 pulse off resonance [in this case at ν(13C) = 13 MHz]. In the TRIPLE experiment, the additional RF1 pumping pulse was applied at the high-frequency line of the inner doublet (indicated by the red arrow). From the difference spectrum (colored blue, i.e., the black minus the red spectrum), it can be seen that only the high-frequency line of the outer doublet is affected (slightly reduced in intensity). This signifies that the HFIs of the two 13C doublets have the same sign, which is consistent with configuration “S-COa”.

To verify that the 13C HFI of the extraneous CO ligand has a different sign from the two CN– ligands, we performed a Davies TRIPLE experiment on a CrHydA1 preparation in which 13CN–-labeled Hox was exposed to 13CO gas in the dark (to prevent CO scrambling). The Davies ENDOR experiment is better suited for relatively large hyperfine interactions like the one observed for the extraneous 13CO. By choosing a long (selective) preparation pulse, we could also observe the weaker coupling from the 13CN– ligands. The resulting spectra (Figure C) show, in addition to the two 13CN– doublets, the high-frequency line of the 13COext ligand at 23.5 MHz. The low-frequency line at 7 MHz is not displayed because its intensity is significantly lower, which makes it unfavorable to observe the “triple effect” (see below). In the Davies TRIPLE experiment, the RF1 pumping pulse was set to the low-frequency line of the inner 13CN– doublet. From the difference spectrum colored blue, one can see that now the “triple effect” occurs on the low-frequency line of the 13CN– outer doublet as well as on the high-frequency line of the 13COext ligand. This confirms the Mims TRIPLE result that the two 13CN– ligand HFIs have the same sign. Additionally, the Davies TRIPLE shows that the sign of the 13COext HFI is opposite to that of the two 13CN– ligands.

It is, therefore, concluded that the observed magnitude and signs of the 13C ligand HFIs are consistent only with DFT model “S-COa” (apical COext) and not with DFT model “S-COb” (basal COext). The “unrotated” structure was also proposed in the original crystallographic study of Hox-CO.[6] Here, illumination at low temperatures exposed the photolability of the COext ligand leading to formation of the Hox state.

One possibility is that the distal CN– ligand coordinates in the apical position only at room temperature where most published FTIR spectra were measured while at low temperatures (where our EPR and ENDOR spectra are recorded) it, somehow, rearranges to a basal position. As observed by Roseboom et al.[25] for the [FeFe] hydrogenase from Desulfovibrio desulfuricans, the FTIR spectrum of Hox-CO is temperature-independent between 15 K and room temperature, and thus, a temperature-dependent ligand rearrangement is highly unlikely. In the available crystal structures (see Figure b), the CN– bound to the distal iron atom is hydrogen bonded to a lysine residue as well as a backbone amide. These interactions can be expected to exert a significant kinetic barrier preventing ligand rotation already at room temperature, thus stabilizing the open coordination site in the Hox state. The fact that externally supplied 13CO (in the dark) is not scrambled to give 13CO in the basal and bridging positions further supports this notion. We can, therefore, conclude that at all temperatures the Hox-CO state is characterized by an extraneous CO ligand that is positioned in an apical coordination with respect to Fed.

The rigidity of the CO/CN– coordination configuration in the H-cluster was recently demonstrated in studies of its reduced states (Hred and Hsred), showing that the bridging CO ligand is retained under all circumstances.[26,27] While some groups[7−9] proposed that the H-cluster structure allows for extensive ligand dynamics and rotation, potentially leading to the formation of inactive configurations containing bridging hydrides and rotated (apical) CO ligands, the former studies as well as our study clearly show that the CO and CN– ligand configuration is rigid and stabilized by H-bonds. This ensures minimal reorganization energy during the catalytic cycle.

Overall, we have demonstrated how advanced EPR spectroscopic techniques such as TRIPLE can be extremely useful in investigating the ligand structure around the metal center at the active site of the [FeFe] hydrogenase. This technique can potentially be applied to numerous other enzymes as well as molecular catalysts to gain more detailed insight into their mechanism of action.