Caught in the Hinact: Crystal Structure and Spectroscopy Reveal a Sulfur Bound to the Active Site of an O2‐stable State of [FeFe] Hydrogenase

Abstract [FeFe] hydrogenases are the most active H2 converting catalysts in nature, but their extreme oxygen sensitivity limits their use in technological applications. The [FeFe] hydrogenases from sulfate reducing bacteria can be purified in an O2‐stable state called Hinact. To date, the structure and mechanism of formation of Hinact remain unknown. Our 1.65 Å crystal structure of this state reveals a sulfur ligand bound to the open coordination site. Furthermore, in‐depth spectroscopic characterization by X‐ray absorption spectroscopy (XAS), nuclear resonance vibrational spectroscopy (NRVS), resonance Raman (RR) spectroscopy and infrared (IR) spectroscopy, together with hybrid quantum mechanical and molecular mechanical (QM/MM) calculations, provide detailed chemical insight into the Hinact state and its mechanism of formation. This may facilitate the design of O2‐stable hydrogenases and molecular catalysts.

photons/s and the beam size was about 1 (height) × 4 (width) mm 2 . Incoming X-rays were selected using a Si(220) double-crystal monochromator. A Rh-coated mirror was used for harmonic rejection. Samples were kept at ~10 K in a liquid helium flow cryostat. Data were calibrated using an iron foil, with the first inflection point set to 7111.2 eV. Spectra were collected on various spots in the sample with a scan time of 27 min/spot. No damage was observed within 30 min collection on a single spot. Final data analysis was performed with scans showing no radiation damage. All PFY-EXAFS and PFY-XANES spectra shown in the main text and here are averages of 6 to 8 scans and the final Hox and Hinact spectra presented in all the cases are the apo subtracted spectra. Data were processed using Athena. [12] Spectra were normalized in energy (eV) space, by fitting a third-order polynomial to the pre-edge region and subtracted throughout the entire EXAFS spectrum. A three-region cubic spline (with the AUTOBK function within Athena) was employed to model the background function to a minimum of k = 16 Å −1 for all spectra. Fourier transformation (FT) was performed over a windowed k range indicated in the figure captions, and all FT spectra are shown without phase shift correction. Theoretical EXAFS fittings were calculated using Artemis [12] by using the multiple scattering FEFF6 code. [13] ADT, Hox and Hinact models were generated from the ADT crystal structure, [2c] the PDB crystal structure of Hox (PDB ID 1HFE) [14] and our resolved Hinact crystal structure (PDB: 6SG2). EXAFS scattering paths were calculated with FEFF6 [12] and employed the FEFF cards: SIG2 0.001; RMAX 5; NLEGS 4. The Fourier-transform spectrum of each were fit over a R range of 1 -3 Å (non-phase shift corrected). The FT arises from a transform of k 3 -weighted EXAFS spectrum with a Hann window of k = 2 to 12 Å -1 for Hox, k = 2.5 to 12.5 Å -1 for Hinact and k = 2 to 12.717 Å -1 for ADT. A single ΔE0 variable was used for all paths in a given fit. S0 2 was set to 0.9 for all the fits. The E0 for ADT, Hox and Hinact was set to 7118.210 eV. The quality (goodness) of the final fits was examined by their R-value.

QM/MM calculations of XAS, IR and NRVS spectra
QM/MM model generation is detailed in a separate chapter below. A QM/MM numerical partial Hessian (QM-atom only displacements) for each state was calculated using DL-FIND in Chemshell. [15] NRVS partial vibrational densities of states (PVDOS) were calculated using the orca_vib program (part of ORCA) [16] and the orca_mapspc program was used to create broadened vibrational spectra (Gaussian lineshape with 12 cm -1 (FWHM)). XAS spectra were calculated on QM/MM-optimized geometries using TDDFT and ZORA-TPSSh. RIJCOSX was used in TDDFT calculations with a decontracted auxiliary basis set (SARC/J). Up to 400 roots were calculated using the Fe 1s orbitals as donor orbitals while the acceptor space consisted of all virtual orbitals. Electric dipole, magnetic dipole and quadrupole contributions were included. The TDDFT calculations were polarized by MM point charges. Orca_mapspc was used for spectral broadening (Gaussian lineshape with 1 eV (FWHM)).
Nuclear resonance vibrational spectroscopy. NRVS spectra were recorded at SPring-8 BL19LXU using a Si(111) double crystal high heat load monochromator (HHLM) to produce 14.414 keV radiation with ≈1.0 eV resolution, followed by a high energy resolution monochromator (HRM) [Ge(422)x2Si(975)] to increase the resolution to ≈0.8 meV. The beam flux was ≈5.4 x 10 9 photons/s and the beam size was about 0.6 (height) × 1 (width) mm 2 . A 2 x 2 avalanche photodiode (APD) detector array collected the delayed nuclear fluorescence and K fluorescence. The temperature at the base of the sample was maintained at 10 K with a liquid He cryostat. The Stoke/anti-Stoke imbalance derived real sample temperatures were 40-70 K. NRVS spectral analysis was performed using the PHOENIX software package executed through spectratools. [17] Energy scale calibration was achieved with [NEt4][ 57 FeCl4].

Resonance Raman spectroscopy.
Resonance Raman spectra were collected on a LabRam HR-800 Jobin Yvon confocal Raman spectrometer connected to a liquid-N2-cooled charge-coupled device (CCD) as previously described. [10] The 514 nm emission line of an Ar + -ion laser with 2 mW power was used for excitation. Temperature was set to 80 K by a liquid-N2-cooled cryo-stage (Linkam Scientific instruments). Data were processed using Bruker OPUS software.

Supplementary Text
Crystallography Analysis and Supplementary Discussion. To verify that the state being measured in the X-ray diffraction experiments is indeed Hinact, we performed IR spectroscopy on crystals from the same crystal drop used for X-ray diffraction data collection. Crystals were transferred to MgF2 plates, as much liquid was removed as possible, and the plates were frozen in liquid nitrogen. The IR spectra of the crystals show the active site is indeed in the Hinact state. Differences in the relative intensities of the peaks may be related to the orientation of the anisotropic crystal in the IR beam, as the extinction coefficients are orientation dependent. Small differences of the band positions may additionally correlate with temperature induced shifts. A Native dataset collected at 12 keV on an Hinact crystal. B Anomalous dataset collected at 6 keV on the same Hinact crystal as for A after the native data collection. C Additional anomalous dataset collected at 6 keV on a fresh Hinact crystal obtained under identical conditions.
As explained in the main text, the structure of Hinact was solved using molecular replacement with the structure published by Nicolet et al. (PDB ID 1HFE) as a starting model, and was refined to a resolution of 1.65 Å with the crystal parameters and refinement statistics shown in the Table above. The overall structure of DdHydAB in the Hinact state is essentially identical to the starting model with a root mean square deviation (RMSD) of 0.631 Å (calculated for all Cα atoms of residues 2-397 without outlier rejection, Figure 1 of the main text). Detailed analysis of the atomic coordinates at the [2Fe] sub-site shows a few small differences in the atomic positions, in particular at the bridging ligands, ADT and CO. Most likely, these deviations arise from the restraints introduced by the ligand description. In Nicolet's structure one of the bridging ligands is modelled as a propane 1,3-dithiolate, whereas our model contains a 2-azapropane 1,3-dithiolate ligand. The differences in C-C and C-N bond angles may result in the different position of the bridgehead. This has been observed experimentally in the CpHydA1 ADT vs PDT crystal structures reported by Esselborn and coworkers. [18] The more notable difference in the location and orientation of the bridging CO also arises from the different ligand restraints. While Nicolet et al. modelled the CO ligand as non-bonded; we used the ligand description that was also employed by Duan et al. for the [FeFe] hydrogenase from Clostridium pasteurianum (CpHydA1). [19] When taking a careful look at the omit map calculated for the model in absence of the [2Fe] sub-site and the additional S ligand, we observe that the electron density of the ligand is slightly shifted towards Fed, as was also observed for some of the CpHydA1 crystal structures. [18]

Supplementary Discussion on Ligand Occupancy
As stated in the main text, the diiron site and exogenous ligand were both modelled with an occupancy of 0.6. This is likely due to a combination of incomplete artificial maturation of the H-cluster despite the long maturation times used, and some loss of the diiron subcluster during the 3 days of crystallization. Despite the lower [2Fe] content the electron density is well-defined and the B-factors for the [2Fe] subcluster are reasonably low (15 -29 Å 2 ), except for the additional sulfide that has a higher B-factor (45 Å 2 ) indicating some intrinsic disorder. Most importantly, the FO-FC maps generated with different occupancies of the additional ligand and assuming a sulfur ligand indicate that the best fit between model and experimental data is achieved with identical occupancy for the subcluster and the additional ligand, and thus every [2Fe] subcluster has an exogenously bound ligand.

Force Field Parameters:
The CHARMM36 protein force-field [20] was used in the MM preparation and the QM/MM calculations with modifications to account for the metal clusters. Non-bonded parameters for the metal clusters were derived and the metal clusters were kept fixed during all MM optimizations and MD simulations. Atomic charges for the metal clusters were derived from ZORA-TPSSh [21] /ZORA-def2-TZVP [22] calculations (including the CPCM continuum) using Hirshfeld population analysis [23] . Appropriate Lennard-Jones parameters from the CHARMM force-field were used for the clusters. The cysteines bound to the iron-sulfur clusters were modelled as deprotonated cysteines using parameters available in CHARMM36.

MM model preparation and solvation:
The whole protein was modelled classically and the initial structure was based on the crystal structure from this work. GROMACS, version 5.0.4 [24] was used to set up the original MM model and add missing hydrogens. Multiple occupancies were removed (Glu53, Met54, Met170, Cys178, Pro195, Pro319 and Leu395). Protonation states of titratable residues were determined using manual inspection of hydrogen bonding patterns. All aspartate and glutamate residues were modelled as deprotonated. All lysine residues were modelled as protonated.
Water molecules (as oxygen atoms) present in the crystal structure were kept and hydrogens added using GROMACS. The system was solvated by placing the protein inside a 90 x 90 x 90 Å box and filling the box with TIP3P waters. The system has a total positive charge of +7 after the hydrogenation step. To balance the charge, 7 Clcounter-ions were added to the solvent. The final total system size was 71491 atoms ( Figure S6). The structure was optimized (metal clusters and Cys178 kept frozen) and then subjected to a 5 ns MD simulation (metal clusters and Cys178 kept frozen). Bond constraints (LINCS algorithm [25] ) were applied to all bonds in order to maintain a 1 fs time-step during the simulation. A 4-chain Nosé-Hoover thermostat [26] with coupling to the whole system, was used for heating and maintaining a simulation temperature of 300 K. The system was gradually heated from 50 K to 300 K in 0-500 ps. The RMSD (with respect to crystal structure) of all heavy protein atoms converged at ~0.3 Å during the simulation.
QM/MM calculation details: The MM model was used directly in the QM/MM calculations (without periodic boundary conditions). Chemshell version 3.7 [15] was used for all QM/MM calculations and the system was imported using scripts previously described [27] . The ORCA quantum chemistry code (version 4.1.1) [16a] was interfaced to Chemshell via a modified Chemshell-ORCA interface. All calculations used electrostatic embedding and link atoms were used to terminate the QM-MM border together with the charge-shift procedure as implemented in Chemshell [28] . For the QM part in the QM/MM optimizations, the TPSSh hybrid density functional [29] with D3BJ dispersion correction [30] and the ZORA relativistic approximation [31] was used with the relativistically recontracted def2-TZVP basis sets [20,32] . The RIJCOSX approximation [33] was used to speed up Coulomb and Exchange integrals. The QM region used for H-cluster calculations is shown in Figure S7. The MM part was calculated using DL_POLY [34] as implemented in Chemshell using the modified CHARMM36 force-field. The QM/MM geometry optimizations were done using the DL-FIND [35] program inside Chemshell. An active region of 1081 atoms was used (approximately spherical region around the H-cluster). The HDLC coordinate system was used in all optimizations. Numerical QM/MM frequency calculations (IR, NRVS) were performed at the same level of theory.

Computational models for Hinact
The primary models for Hinact that were considered (with SH, Cl or OH ligands) are shown in Figure S8 along with the model for Hox. Only the [FeFe] cluster geometry is shown in Figure S8, but all calculations used for the QM-region are shown in Figure S7. A deprotonated sulfide model Hinact-S (charge of -3) was calculated but this gave a highly reactive sulfide that subsequently reacted with a carbonyl group, resulting in the implausible structure shown in Figure S9. The Hinact-SH model in Figure S8 (and Figure S10 left) features the NH-group of the ADT ligand making a hydrogen bond to the SH group. This was found to be the energetically most favorable conformer (by >10 kcal/mol). Alternative conformations of the Hinact-SH model that were explored (including different conformations of Cys178), are shown in Figure S10. Based on the energetics (polarized QM energies) the alternative conformers are less plausible models for Hinact and they result in overall very similar [FeFe] structures.  ] clusters needed to be removed from the total spectra. Therefore, the spectrum of the apo-protein (before maturation with the [2Fe] subcluster) was measured and subtracted from the Hinact and Hox spectra after normalization in energy (eV) space as follows:

Hinact-S
Hox apo-subtracted = Hox -12/14 x Apo Hinact apo-subtracted = Hinact -12/14 x Apo For comparison, a sample of the diiron precursor complex (Et4N)2[Fe2(2-aza-propane 1,3dithiolate)(CO)4(CN)2] (referred to simply as ADT) dissolved in 100 mM Tris-HCl, 150 mM NaCl, pH 8 buffer was also measured. Figure S11A shows the raw Fe K-edge XANES spectra for the apo-enzyme, holo-Hox, and holo-Hinact. In Figure S11B the Hinact and Hox apo-subtracted spectra can be seen with error bars included, indicating the standard deviation in the intensity over 8 scans. It is clear that the raw spectra of Hox and Hinact are highly dominated by the features of the apo-enzyme, where the Fe centers have a non-centrosymmetric tetrahedral geometry allowing 3d−4p mixing, giving an intense pre-edge feature around 7112 eV. The edge energy is influenced by the sulfur ligands, which are heavy scatterers that shift the edge toward lower energies [36] . Figure S11 B shows apo-subtracted spectra, also reported in Figure 4A of the main text, but including error bars generated as reported previously [37] . This clearly indicates the significance of the differences in the pre-edge and edge regions of the XANES spectra (see Tables S2 and S3 for fitting of the pre-edge features). Table S2. Parameters derived from the pre-edge fitting of the Hox apo-subtracted spectrum. The fitting was done in blueprintXAS. Table S3. Parameters derived from the pre-edge fitting of the Hinact apo-subtracted spectrum. The fitting was done in blueprintXAS.

EXAFS Supplementary Discussion:
EXAFS fittings reported in the main text have been performed only on the Hox and Hinact apo-subtracted spectra. The EXAFS data plotted in "k-space" ( Figure S12 and insets of Figures S13, S14 and S15) are fourier transformed (FT) to give the final plots shown in Figure 3 of the main text and Figures S13, S14 and S16 of the SI. FT is the product of a transform of the k 3 -weighted EXAFS spectrum with a Hann window over the range of k = 2 to 12 Å -1 for Hox and k = 2.5 to 12.5 Å -1 for Hinact. [14] , that the Fep-Scys bond is quite long (~ 2.5 Å) compared with the Fe-SADT bonds (~2.2 Å). As the EXAFS scattering intensity decreases with the square of the distance between the absorber and the scatterer, it was not clear whether inclusion of the contribution of the Fep-Scys was necessary in the fit and in case it is, whether it would be possible to group it with the same average distance as the Fe-SADT contributions. Thus, for the Hox data a set of fits were carried out in which the coordination number (degeneracy of the path, N) of the Fe-sulfur-path (Fe-S path) were altered. A degeneracy of the Fe-S path of N = 2.5 (Figure 3 A, Figure S12 A, Table S4) gave a reasonable Debye-Waller (DW or σ 2 ) factor (a measurement of the static and thermal disorder) of the Fe-S path. When the degeneracy of the Fe-S path was set to N = 2 (excluding the contribution from Fep-Scys), the DW factor of the Fe-S path became negative and that of the Fe-Fe path became unrealistically small (<0.002 Å 2 , see Figure S13 A and Table S6). This indicates that the contribution of Fep-Scys is essential. A larger degeneracy of N = 3 (Figure S13 B and Table S7) gave unreasonably large DW factors.

Intensity peak STD
Position peak STD HWHM peak STD Gaussian Fract peak STD Total Area Integra 0. 2 Figure  3 in the main text. Fitting parameters are shown in Table S4 and S5. Table S4. EXAFS fit parameters for Figure 3A of DdHydAB in the Hox state from the main text and Figure S12 A.   (Table S6). B: increasing the degeneracy (N) of the Fe-S path to N = 3 (Table S7). Figure S13 A. For Hinact EXAFS fits, when the sulfur degeneracy was set to N = 3 (Figure 3 B, Figure S12 B, Table S5) the DW factor of the Fe-S path is very small (σ 2 ≈ 0.001 Å 2 ), suggesting that the fit is strongly dominated by the S-scattering path. Such a small DW value normally suggests that the degeneracy of the path should be even higher. When a set of fits changing the degeneracy of the Fe-S path (but fixing the DW of the Fe-Fe-path to σ 2 = 0.002 Å 2 ) were examined, it was clearly observed that: 1) when the degeneracy of the Fe-S path is decreased to N = 2.5, its DW factor becomes negative and the quality of the fit (evaluated by the R-value) also decreases ( Figure S14 A and Table S8), and 2) when N = 3.5, the DW factor increases to a more reasonable value (see Figure S14 B and Table S9) but the overall quality of the fit decreases. Since the DW factor of the Fe-S path is strongly correlated with both the Fe-Fe distance and the DW factor of the Fe-Fe path, the EXAFS analysis can only give a range on the Fe-S coordination number of 3 to 3.5. Thus, while the N = 3.5 fit cannot be excluded, it is certain that N (Fe-S path) > 2.5 and, therefore, there are more sulfur scatterers in Hinact than Hox. Similarly, and as mentioned in the main text, fits attempting to separate the Fe-S scattering paths into shorter and longer Fe-S distances (as observed in the crystal structure) resulted in the fit paths converging to the same distance. Thus, the separation of the two sulfur contributions is beyond the resolution of our data (~0.16 Å). Similar to the protein crystallography, we also note the EXAFS cannot distinguish between Cl or S as the additional ligand. Due to the strong correlation between the various parameters of the various paths, more accuracy is not possible. However, taken in the context of the crystal structure and the other spectroscopic data, the EXAFS analysis are consistent with the notion of a single additional sulfur ligand bound to the [2Fe]H subcluster in the Hinact state relative to Hox, and the distances obtained from the EXAFS fits are in reasonable agreement, within the associated errors (~0.1 Å), with the crystal structure. [38] Fig. S14. k 3 -EXAFS and FT spectra for the set of Hinact fits changing the degeneracy of the Fe-S path. A: decreasing the degeneracy (N) of the Fe-S path to N = 2.5 (Table S8). B: increasing the degeneracy (N) of the Fe-S path to 3.5 (Table S9). As a validation of the apo-hydrogenase subtraction approach, we also decided to measure XANES and EXAFS on the diiron precursor complex used for artificial maturation of the apo-hydrogenase, referred to here as ADT, for which a high resolution crystal structure is available [2c] .  Attempts to fit the EXAFS data from ADT ( Figure S16) demonstrated rather large uncertainties of the fitted parameters. The scattering paths are so strongly correlated with each other (Table S10) that the error associated with the fitted parameters is large (0.015 -0.035 Å for the distances and 1.2 x 10 -3 -4.7 x 10 -3 Å 2 for the DW factors). Fitting EXAFS data for such a complex system as the [FeFe] hydrogenase, where all the parameters are strongly correlated to each other, cannot give accurate values for all the parameters. Even though EXAFS analysis has its limitations, it does provide valuable information about the type of scatterers and their contribution to each state. The Hinact vs Hox fits clearly show that more sulfur scatterers are present in the former.  Table S10 below).

TD-DFT Analysis of the pre-edge Fe XAS:
TD-DFT calculations were performed using the ⍵B97X functional (on top of TPSSh-optimized geometries) for Hox and Hinact-SH QM/MM models and gave good agreement with the experimental Fe K-edge XAS spectra in the pre-edge region (Figure 4 in main article). Importantly, clear differences could be seen between the calculated spectra for Hox and Hinact, which were also observable in the experimental spectra. According to our TD-DFT calculations, the first transition observed in the pre-edge of the experimental Hox spectrum (which is not observed in the Hinact spectrum), corresponds to a transition into an empty low-lying dz 2 orbital, which is low in energy because of the unoccupied coordination site in Hox (see Figure S17). At higher energy, the next pre-edge transitions, present in both Hox and Hinact-SH models, correspond to excitations to the other "d-holes". The higher energy features of the pre-edge region in both models arise from metal ligand charge transfer (MLCT) transitions into bridging CO π * orbitals. These MLCT transitions occur at the same energy for both metals in the [2Fe]H sub-cluster in the Hinact-SH model since it harbors a Fe(II)Fe(II) core. However, this feature is split in energy for Hox since it has an Fe(I)Fe(II) core. The MLCT transitions occurring at Fe(I) take place at lower energy, explaining why the pre-edge of Hox is broader than for Hinact. Other Hinact models with OH and Cl ligands were also calculated. The TD-DFT calculated XAS spectra in Figure S18 shows that both Hinact-Cl and Hinact-OH give almost identical spectra as Hinact-SH. The pre-edge XAS spectrum is thus not sensitive to the nature of the ligand. The positions of the MLCT transitions were found to be sensitive to the HF-exchange in the functional. The range-separated hybrid functional ⍵B97X (that has more correct long-range behavior and less self-interaction error) was found to give MLCT positions at slightly higher energy than the d-transitions that gave a spectrum in better agreement with experiment (see Figure S19).  ; 50 mM Tris-H2SO4, 500 mM K2SO4, pH 8 (middle two spectra); or 50 mM Tris-H2SO4, 500 mM LiF, pH 8 (bottom two spectra) was oxidized under N2 with 1 mM thionine acetate (Hox spectra). This was followed by addition of 10 mM Na2S and 10 mM thionine acetate, and exposure to air (Hinact spectra). In each case, the Hinact state could be formed indicating that the presence of chloride is not required for Hinact formation. Some of the Hox-CO state (from H-cluster decomposition) contributes to the Hox spectra.

Fig. S21. Correlation of the experimental and calculated IR frequencies for Hox and Hinact.
Experimental FTIR frequencies for the Hox and Hinact states (y-axis) are plotted against the calculated frequencies (x-axis) using the Hox (black circles) and Hinact-SH (green circles) models described in the QM/MM calculations section. Regression lines are plotted with an x-axis intercept at y = 0, and the formulae of the regression lines are shown on the chart. A value of 0.964 was taken as the scaling factor for the calculated frequencies used to derive the spectra presented in the main text and Figure S22.  The close-up view of the calculated spectra shows a shift to lower energy of the peak at ≈ 350 cm -1 (marked with an asterisk) when going from 32 S (green) to 34 S (royal blue) as it should for a heavier atom. This results in the broad green peak ( 32 S, asterisk) splitting into two blue peaks ( 34 S), or an overall intensity reduction. The calculated isotope shifts are also shown in Table S12. The experimental spectrum is consistent with the intensity reduction and the shoulder at ≈ 350 cm -1 (asterisk) is also shifted toward lower energy. Since the shifts are very subtle, it is difficult to observe them in the experimental spectra because they are within the noise. Resonance Raman spectra were measured on 3 mM DdHydAB samples at 80 K using 514 nm excitation. Experimental spectra of Hinact prepared with natural abundance Na2S (green trace) and 34 S-labelled Na2S. The spectra are normalized to modes at 622 cm -1 and 644 cm -1 corresponding to the amino acid side chains phenylalanine and tyrosine, respectively. [39] Calculated resonance Raman band positions using the Hinact-SH QM/MM model with 32 S and 34 S bound at open coordination site of Fed are presented in Table S12 and shown as Movies S1 and S2.

Possible Hinact Formation Mechanism
We note that it is not possible to provide accurate ΔG values in the last two steps of the mechanism. The reliable accurate calculation of redox potentials and protonation/deprotonation reaction energies is very difficult in computational chemistry due to the big environmental effect difference between the two charged states.

Fig. S26
. Proposed scheme for Hinact formation from Hox. First H2S binds to the apical coordination site on Fed in the Hox state forming Hox(SH2). Binding has been calculated to be thermoneutral with an estimated ΔG of +0.2 kcal mol -1 . This is followed by proton transfer to the nitrogen base of the ADT bridge giving HoxH + (SH -). This step is calculated to be thermodynamically favorable with an estimated ΔG of -4.9 kcal mol -1 . Deprotonation of the ADT bridge via the proton transfer channel leads to electron transfer from [2Fe]H to [4Fe-4S]H, yielding the spectroscopically characterized Htrans state. Finally, Htrans is converted to Hinact by oxidation of [4Fe-4S]H. An alternative proposal of H2O binding gives a binding energy of ΔG= -6 kcal/mol but an H2O ligand cannot be deprotonated by the ADT ligand (attempted optimizations resulted in spontaneous proton transfer back), unlike an H2S ligand.