|Home | About | Journals | Submit | Contact Us | Français|
[Fe-Fe]-hydrogenases are enzymes that reversibly catalyze the reaction of protons and electrons to molecular hydrogen, which occurs in anaerobic media. In living systems, [Fe-Fe]-hydrogenases are mostly used for H2 production. The [Fe-Fe]-hydrogenase H-cluster is the active site, which contains two iron atoms. The latest theoretical investigations1,2 advocate that the structure of di-iron air inhibited species are either FepII-FedII-O-H-, or FepII-FedII-O-O-H, thus O2 has to be prevented from binding to Fed in all di-iron subcluster oxidation states in order to retain a catalytically active enzyme. By performing residue mutations* on [Fe-Fe]-hydrogenases, we were able to weaken O2 binding to distal iron (Fed) of Desulfovibrio desulfuricans hydrogenase (DdH). Individual residue deletions were carried out in the 8 Å apoenzyme layer radial outward from Fed to determine what residue substitutions should be made to weaken O2 binding. Residue deletions and substitutions were performed for three di-iron subcluster oxidation states, FepII-FedII, FepII-FedI, and FepI-FedI of [Fe-Fe]-hydrogenase. Two deletions (ΔThr152 and ΔSer202) were found most effective in weakening O2 binding to Fed in FepII-FedI hydrogenase (ΔGQM/MM = +5.4 kcal/mol). An increase in Gibbs' energy (+2.2 kcal/mol and +4.4 kcal/mol) has also been found for FepII-FedII, and FepI-FedI hydrogenase respectively. π-backdonation considerations for frontier molecular orbital and geometrical analysis corroborate the Gibbs's energy results.
[Fe-Fe]-hydrogenases are enzymes that catalyze the reversible reduction of protons to molecular hydrogen (2H+ + 2e- H2) in anaerobic media2,3, and are considered one of the oldest enzymes in nature4. The eventual elucidation of the catalytic mechanism of hydrogen synthesis may avail researches produce clean hydrogen fuel, using certain prokaryotes and eukaryotes5-14.
The active site of hydrogenase, the H-cluster (Scheme 1) is comprised of two iron atoms (Fep-Fed, i.e., proximal and distal iron). The di-iron atoms are coordinated by endogenous ligands, i.e., two cyanides, two terminal carbonyls, and a bridging carbonyl (COb). Also, 1,3-di(thiomethyl)amine (DTMA) and propanedithiolate (PDT) are bidentate ligands of the di-iron subsite15-17. A cubane cluster, [4Fe-4S] (which also belongs to the H-cluster), is bonded to Sγ of Cys382, while the former (Sγ) is bound to Fep of the H-cluster.
Previous Density Functional Theory (DFT) as well as hybrid quantum mechanics/molecular mechanics (QM/MM) calculations2,18-24 have been successful in clarifying some aspects of the catalytic properties of the H-cluster. As in previous computational studies2,19, CH3-S is substituted for cysteine, and a H+ replaces the proximal cubane. Furthermore, computational and experimental research1,2,15,25-52 on [Fe-Fe]-hydrogenase H-cluster and synthetic H-cluster-like compounds sheds light on the potential redox states of the [Fe-Fe]-hydrogenase H-cluster subunit, Fep-Fed; FepI-FedI is the reduced hydrogenase H-cluster subunit, FepII-FedI is the partially oxidized enzyme subunit, and FepII-FedII is the fully oxidized, inactive enzyme H-cluster subunit.
The oxidized H-cluster, FepII-FedII, has OOH-, H2O molecule, or OH- bound to the FedII site 1,19. In our previous investigation31, we have inferred that a vacant FepII-FedII could also be a viable intermediate in H2 synthesis. Regardless of the redox states in [Fe-Fe]-hydrogenase H-cluster, the proximal cubane always retains a 2+ oxidation state, [Fe4S4]2+. The partially oxidized H-cluster (containing FepII-FedI), Hox, is the active species of the hydrogenase enzyme having the tendency for protonation (Liu and Hu19).
The crystal structures of hydrogenase obtained from Clostridium pasteurianum (CP)16 and Desulfovibrio desulfuricans (Dd)17 led to a more detailed understanding of the biochemical role of these enzymes. The crystal structure of CPI hydrogenase shows an oxygen species (i.e. either OH- or H2O) bound to Fed. However, according to X-ray crystal structure, of CPI and based on computational results of Tye et al.1, the enzyme has OOH- bound to Fed in its inactive form. Hence, we endeavor to ascertain if oxygen binding to distal iron (Fed-O2) can be hindered by residue mutations within the surrounding apoprotein of the catalytic site.
The current investigation is comprised of two subdivisions. (1) The thermodynamic analysis of the reactions of O2 binding to wild-type and mutated [Fe-Fe]-hydrogenase for the three different oxidation states of the di-iron subcluster: FepI-FedI, FepII-FedI, and FepII-FedII. (2) Geometrical analyses of inter-atomic distances in Fed-O2, the extrinsic ligand, O-O, and of COb carbon distances to the di-iron atoms. In the Supplementary Information the NBO charges and an electronic analysis discussing the frontier molecular orbitals are presented.
The QM/MM [DFT/UFF53] ONIOM54 method (DFT for the QM region, and the universal force field, UFF, for the MM region, implemented in Gaussian0355) has been used to calculate the reaction thermodynamics, i.e., ΔG, for the reactions of oxygen binding to [Fe-Fe]-hydrogenase. The electronic structure of H-cluster (except the proximal cubane) was described by QM using the DFT Hamiltonian (B3LYP functional56,57), with 6-31+G(d,p) basis set. In accordance with experimental and in silico data, low spin states (singlet, and doublet) and low oxidation states (I, and II) have been selected for the Fe atoms2,24.
The Gromacs program58,59 was employed to add hydrogen atoms, water, and counter ions to the crystal structure of DdH [Brookhaven Protein Data Bank (id.1HFE)]. A 1 nm layer of water has been added to the DdH structure. Sodium cations have been randomly inserted into the solvent to neutralize the negative charges encountered in DdH, e.g., the -2 a.u. found on the cubane/cysteine moieties, or in the H-cluster (when needed)60. For both basic and acidic amino acids, charges were assigned by the Gromacs program pdb2gmx to be at pH 7. The low layer consists of all metalloenzyme residues as well as its constituent cubanes, i.e., proximal, medial, and distal. The high layer is comprised of 2Fe subunit, (which is a moiety of the H-cluster), and Cβ and Sγ (of the bridging Cys382). Two linking hydrogen atoms were added between Cα and Cβ of Cys382, and between Sγ and a Fe atom of the proximal cubane. The charge equilibration method of the UFF was used to describe the electrostatic interactions within the low layer of the system61. The DdH partial charges were obtained using the charge equilibration method (QEq), whereas the solvent charges were acquired from literature61 (qO = -0.706 a.u. and qH = 0.353 a.u.).
ONIOM geometry optimizations have been performed on DdH, with the low layer (MM region) being frozen*, with the exception of the proximal cubane; for the high layer (QM), only the iron atoms, Fep-Fed, and the N (of the DTMA bridge) have been kept frozen**.
Residue mutations were carried out within the adjacent apoenzyme environment to the H-cluster in order to hinder O2 from binding to the open coordination site (Fed) of DdH H-cluster. Residue mutations are comprised of deletions and substitutions, which are performed 8 Å radially outward from Fed. In order to screen the thirty polar residues located in the 8 Å apoenzyme layer, individual residue deletions were carried out followed by calculations to ascertain what residue substitutions should be made to impede O2 binding to Fed. The deletion of a residue was performed by removing the point charges of its atoms from the Gaussian input file. Residue deletions and substitutions were performed for all three di-iron subcluster oxidation states of [Fe-Fe]-hydrogenase H-cluster: FepII-FedII, FepII-FedI, and FepI-FedI.
The results of the QM/MM calculations for the reactions of O2 binding to [Fe-Fe]-hydrogenase are shown in Table 1. Two values of the Gibbs' energy are given for wild-type DdH. In the first row are Gibbs' energies (ΔGcQM/MM) obtained by taking into account neighboring charges of the 2Fe subunit: i.e. point charges from the proximal cubane*, the MM region of Cys382, C of the peptide bond in Gly381, and N of Val383. The values (ΔGQM/MM) in the second row of Table 1 were obtained without including the neighboring charges of the 2Fe subunit in the calculation. Sometimes the deletion of MM charges in the vicinity of the QM system is necessary to avoid distortions (artificial polarization) in the wave function induced by these charges. The difference in the wave function polarization (i.e. with or without neighboring charges) is quantified by difference in the natural bond orbital charges (NBO, Figure 1a, and 1b). The strongest effect of the neighboring MM charges is on the NBO charges of Sγ of Cys382 and the linking atom (HL) attached to it. A detailed analysis of these charges is given in the Supplementary Information.
Figure 2 shows O2 inhibition pathways for all three oxidation states of the H-clusters: FepII-FedII (1), FepII-FedI (3), and FepI-FedI (5), where (1), (3), and (5), are the cluster identifiers. In gas phase, the 1st reaction, 1 → 2, (O2 binds to the fully oxidized H-cluster (1)) is endergonic (ΔGgas = +9.8 kcal/mol; gas = gas phase**). Hybrid QM/MM calculations, on the other hand, show that O2 binding in protein environment occurs exergonically (ΔGcQM/MM = -16.6 kcal/mol; ΔGQM/MM = -10.6 kcal/mol), confirming the affinity of hydrogenases for O262.
In gas phase, the 2nd reaction, 3 → 4, starts with the partially oxidized H-cluster (3), FepII-FedI, and the bonding of O2 to FedI occurs rather exergonic (ΔGgas = -36.1 kcal/mol). ONIOM results show that in protein environment O2 binds spontaneously to the H-cluster when neighboring charges are included (FepII-FedI, ΔGcQM/MM = -7.9 kcal/mol), but the binding reaction is non-spontaneous (ΔGQM/MM = +2.6 kcal/mol) when the neighboring charges are not included.
The 3rd reaction in gas phase, 5 → 6, starts with the reduced H-cluster (5), FepI-FedI; the reaction occurs spontaneously (ΔGgas = -36.0 kcal/mol) and has almost identical Gibbs' energy as reaction 3 → 4, probably because both loci of oxygen binding (FedI-O2) are on similar oxidized species, FedI. However, the QM/MM calculations for 5 → 6 show a difference between the enzyme (ΔGcQM/MM = -20.7 kcal/mol; ΔGQM/MM = -20.5 kcal/mol) and gas phase results (ΔGgas = -36.0 kcal/mol) Gibbs' energies.
In Figure 2 the gas phase protonation reaction 5 → 7 is very exergonic (ΔGgas = -220.6 kcal/mol), essentially because the charge on H-cluster 5 is -2 a.u. ONIOM calculations also show a very high H+ affinity (ΔGcQM/MM = -219.2 kcal/mol) for the hydrogenase H-cluster, which is close to the gas phase result (ΔGgas = -220.6 kcal/mol). The thermodynamic results in Figure 2 demonstrate that most reactions considered for the hydrogenase H-cluster proceed exergonically with the exception of 1 → 2 in gas phase. Thus, the calculations on the wilde-type [Fe-Fe]-hydrogenase are in agreement with experimental observations and confirm the inhibition of [Fe-Fe]-hydrogenase by molecular oxygen.
The thermodynamic calculations of O2 binding to the H-cluster in gas phase and protein environment suggest that the enzyme electric field modulates the reactivity of the H-cluster toward O2, and thus is responsible for the inhibition of [Fe-Fe]-hydrogenase by molecular oxygen. Thus thoughtful modification of the enzyme electric field by residue mutation may advert or weaken the O2 binding to the H-cluster. This idea is the basis of the study presented here.
Residue screening in a protein layer (8 Å) surrounding the H-cluster has been carried out to gauge the effect of point mutations on the strength of O2 binding to H-cluster. First we performed residue deletions to gauge the effect of the electric field of the residues targeted for substitution on the reaction of O2 binding, then we performed residue substitutions for residues that showed significant effect upon deletion.
For O2 binding to FedII (of the oxidized biferrous hydrogenase H-cluster subsite, FepII-FedII, (1)), the results obtained are a function of stereoelectronic effects from the juxtaposed residues on the catalytic site. Both neutral polar and charged residue deletions provided good results, e.g., ΔSer62s*, ΔAsp144, ΔGlu146, ΔAsp150, ΔThr152, and ΔSer202 gave ΔGQM/MM = -9.0 kcal/mol, ΔGQM/MM = -8.4 kcal/mol, ΔGQM/MM = -8.7 kcal/mol, ΔGQM/MM = -9.4 kcal/mol, ΔGQM/MM = -9.2 kcal/mol, and ΔGQM/MM = -7.9 kcal/mol, respectively.
O2 is hindered from binding to FedI of the partially oxidized di-iron subsite (FepII-FedI). Endergonic binding reactions have been identified all tried residue deletions (Table 1), except for the following: ΔTyr112, ΔLys237, ΔThr257, ΔThr259, ΔSer289, and ΔThr299, which gave ΔGQM/MM = -3.2 kcal/mol, ΔGQM/MM = -2.7 kcal/mol, ΔGQM/MM = -2.9 kcal/mol, ΔGQM/MM = -3.0 kcal/mol, ΔGQM/MM = -2.9 kcal/mol, and ΔGQM/MM = -2.8 kcal/mol, respectively.
An improved trend towards impeding O2 binding to FedI, of the fully reduced di-iron subsite (FepI-FedI), has been observed for residue deletions (Table 1): ΔThr152 and ΔSer202, which gave ΔGQM/MM = -18.6 kcal/mol, and ΔGQM/MM = -17.3 kcal/mol, respectively. Table 1 shows that the deletion of charged amino acid residues like Asp144 and Glu374 has a similar effect (ΔGQM/MM = +4.6 and +4.7 kcal/mol, respectively) on O2 binding to FepII-FedI, as the deletion of polar but neutral amino acid residues like Thr152 and Ser202 (ΔGQM/MM = +4.1 and +3.9 kcal/mol, respectively), which seems rather intriguing. A closer look at the distribution of the amino acid residues targeted for deletion in the 8 Å layer around the di-iron catalytic unit reveals that the charged residues Asp144 and Glu374 are about 10 Å away from distal Fe, while Thr152 and Ser202 are about 3 Å closer. This difference in location with respect to the H-cluster explains why the deletion of two charge residues, further away, has a similar effect on O2 binding as the deletion of two polar residues, which are much closer to H-cluster. The residues interlaced between Asp144/Glu374 and the catalytic site screen the bare charge of these charged residues, thus diminishing its effect on the electronic structure of the di-iron unit. Because charge residues have a stronger effect on their local environment than polar residues, it is expected that polar/nonpolar residues are better targets for mutations because their replacement should be less detrimental to the overall folding of the enzyme. Thus, we selected neutral polar residues for mutations.
With the positive results obtained from residue deletions, we were now able to carry out residue substitutions (Table 2). Two residue deletions (ΔThr152 and ΔSer202), that showed significant effect on the Gibbs' energy of O2 binding for all oxidation states of the di-iron H-cluster subunit, were followed by mutations to alanine, i.e., Thr152Ala, and Ser202Ala. The dual residue deletions, ΔThr152 and ΔSer202 impede further O2 binding to the H-cluster subunit FepII-FedI (ΔGQM/MM = +5.4 kcal/mol). However, for the oxidation states FepII-FedII and FepI-FedI only a slight enhancement in O2 inhibition has been observed (+2.2 kcal/mol and +4.4 kcal/mol, respectively). The simultaneous mutations to alanine (Thr152Ala, and Ser202Ala) give better O2 inhibition results (ΔGQM/MM = -9.2 kcal/mol for FepII-FedII, ΔGQM/MM = +4.2 kcal/mol for FepII-FedI, and ΔGQM/MM = -18.1 kcal/mol for FepI-FedI).
Additionally, it is known that certain organisms containing [Fe-Fe]-hydrogenases thrive around suboceanic thermal vents4. Hence, at a temperature of 100 °C intercalated with hydrogenase mutations (Table 2), QM/MM results indicate that the extrinsic O2 binding to metalloenzyme is further reduced (ΔGQM/MM = -5.6 kcal/mol for FepII-FedII, ΔGQM/MM = +7.9 kcal/mol for FepII-FedI and ΔGQM/MM = -12.9 kcal/mol for FepI-FedI).
In this section the Gibbs' energies for the reaction of O2 binding to wild-type DdH are correlated with geometrical parameters, such as interatomic distances and bond angles (Table 3). The iron-carbon distances in the Fep-COb subunit are investigated for all three oxidation states of the di-iron subunits.
The iron-carbon distance in Fep-COb (FepII-FedII) becomes smaller, 1.925 Å (1) → 1.807 Å (2), upon O2 binding concomitant with Fed-COb bond elongation, 1.942 Å (1) → 2.287 Å (2), which generally indicates an increased bonding strength for an exogenous ligand31 (i.e. O2). For FepII-FedI, the bond distance Fep-COb becomes smaller, 1.939 Å (3) → 1.924 Å (4), upon O2 binding while a bond elongation is observed for Fed-COb, 1.908 Å (3) → 1.924 Å (4). For the reduced di-iron subcluster (FepI-FedI) the bond distance Fep-COb becomes smaller, 1.942 Å (5) → 1.935 Å (6), upon O2 binding, while Fed-COb increases, 1.826 Å (5) → 1.945 Å (6). The above geometrical analysis concludes that for all oxidation states of Fep-Fed the bond between the carbon of the bridging carbonyl (COb) and Fed becomes longer and the bond between COb and Fep becomes shorter, upon O2 binding to the catalytic site.
Next, an analysis is presented for the interatomic distances between distal iron and oxygen, and between oxygen atoms, relative to Gibbs' energy of O2 binding to all three di-iron oxidation states.
For the FepII-FedII subcluster, the iron-oxygen distance, Fed-OI, is rather small (1.729 Å; Table 3), which suggests a strong bonding (ΔGQM/MM = -10.6 kcal/mol; 1 → 2) between the distal iron and the oxygen atom bound to it. The inter-oxygen (OI-OII) bond distance is 1.276 Å, which corresponds to a bond order between a single and double bond.
In the case of the active di-iron subcluster, FepII-FedI, the Fed-OI bond distance is ca. 6% longer (1.840 Å) than Fed-OI interatomic distance in FepII-FedII-OI, giving rise to a weaker bond (ΔGQM/MM = +2.6 kcal/mol; 3 → 4) between the distal iron and oxygen. The OI-OII bond distance is 1.281 Å, which is relatively close to the OI-OII bond for FepII-FedII subcluster.
In FepI-FedI the OI-OII bond distance is relatively larger, 1.373 Å, which suggests that π-backdonation occurs between a filled d-orbital of Fed and the empty π* orbital of O2. Out of the three di-iron oxidation states, only the reduced di-iron subcluster (FepI-FedI) has attributes of π-backdonation1, i.e., the OI-OII bond order is intermediate between a double and a single bond order, and the OI-OII bond is elongated (1.373 Å).
Finally, the Fed-OI-OII angle varies as the oxidation states decrease: FepII-FedII-OI-OII (137.0°), FepII-FedI-OI-OII (160.6°), and FepI-FedI-OI-OII (126.0°) in conjunction with effects of the nearby electric field of the apoprotein.
The QM calculations on the gas phase H-cluster in different oxidation states of Fep-Fed subunit show that O2 binding to the fully oxidized H-cluster (FepII-FedII) is exergonic, and endergonic for the partially oxidized (FepII-FedI) and reduced (FepI-FedI) H-clusters.
On the other hand, the QM/MM calculations on the wild-type [Fe-Fe]-hydrogenase confirm that the resting state of the enzyme (FepII-FedII) is inhibited by O2. In addition, the O2 binding to the partially oxidized (FepII-FedI) hydrogenase is slightly endergonic, but is exergonic for the reduced oxidation state (FepI-FedI). The contrast between the calculation results in gas phase and protein environment suggest a dramatic effect of the enzyme electric field on the O2 binding reaction. Thus, we explored the modulation of this electric field by performing point mutations in an 8 Å protein layer surrounding the H-cluster. First, residue deletions have been performed, one by one. Then, from clues obtained from these residue deletions, residue substitutions have been carried out. For FepII-FedII both neutral polar residue and charged residue deletions (ΔSer62s, ΔAsp144, ΔGlu146, ΔAsp150, ΔThr152, and ΔSer202) led to a decreased binding of O2. For FepII-FedI residue deletions ΔGlu374, ΔAsp144, ΔSer177, and ΔThr152 and for FepI-FedI residue deletions ΔThr152 and ΔSer202 hinder O2 binding to FedI.
As expected, residue substitutions Thr152Ala and Ser202Ala affect O2 binding to the same extent as the corresponding deletions (ΔThr152 and ΔSer202).
Finally, the fact that one substitutes nonpolar amino acid residues (Thr152Ala, and Ser202Ala) for some polar ones juxtaposed to the catalytic site (the H-cluster) seems to be sufficient cause to hinder the binding of O2 to hydrogenase H-cluster. This shows that well thought modulation of the enzyme electric field accomplished by point mutations can be used to advert hydrogenase inactivation by molecular oxygen. Hence, DdH mutations open up new research opportunities along these lines.
This work was supported by funds from the Department of Energy, grant: DE-FG02-03ER15462 and National Institutes of Health, grant: 1R15GM070469-01. Computational resources have been provided by the National Center for Supercomputer Applications (University of Illinois) and the Ohio Supercomputer Center.
*In this investigation, QM/MM [DFT/UFF] hybrid method (Gaussian03) has been used.
*Where “frozen” means that x, y, z atom coordinates are kept fixed; “freezing atoms” is practiced to reduce computational time.
**For the fully and partially oxidized vacant di-iron subunits, additional optimizations have been carried out by freezing these atoms: Fep-COt (where COt stands for terminal carbonyl; COt is bound to Fep). The extra optimizations have been done because the above mentioned di-iron subunits are more likely to undergo COb migration.
*Except for the cubane sulfur (Sc,d) situated diagonally from the cubane Fe (bound to cysteinyl sulfur of Cys382).
**The gas phase results are reproduced from reference 18.
*s = DdH small chain
1The π-backdonation agrees with the Gibbs' energy results form Table 1. For example, the FepII-FedII subcluster has an exergonic Gibbs' energy (ΔGQM/MM = -16.6 kcal/mol), which can be improved however by DdH mutations such as residue deletions and substitutions, since there is only slight π-backdonation present. The bond Fed-OI is still relatively weak (ΔG = -7.9 kcal/mol) for FepII-FedI subsite. However, for the reduced FepI-FedI subsite, the π-backdonation makes the oxygen bond very strong, thus making its elimination rather difficult (even by means of DdH mutations, Table 2).