Regioselectivity of H cluster oxidation.

The H(2)-evolving potential of [FeFe] hydrogenases is severely limited by the oxygen sensitivity of this class of enzymes. Recent experimental studies on hydrogenase from C. reinhardtii point to O(2)-induced structural changes in the [Fe(4)S(4)] subsite of the H cluster. Here, we investigate the mechanistic basis of this observation by means of density functional theory. Unexpectedly, we find that the isolated H cluster shows a pathological catalytic activity for the formation of reactive oxygen species such as O(2)(-) and HO(2)(-). After protonation of O(2)(-), an OOH radical may coordinate to the Fe atoms of the cubane, whereas H(2)O(2) specifically reacts with the S atoms of the cubane-coordinating cysteine residues. Both pathways are accompanied by significant structural distortions that compromise cluster integrity and thus catalytic activity. These results explain the experimental observation that O(2)-induced inhibition is accompanied by distortions of the [Fe(4)S(4)] moiety and account for the irreversibility of this process.

Introduction

Hydrogenases are enzymes capable of catalyzing the reversible oxidation of n class="Chemical">hydrogen[1−7] and have therefore attained attention in the field of clean energy research.[8−10] However, a generally high sensitivity to molecular oxygen complicates their technological application.

Hydrogenases are classified into three different groups according to phylogenetic relations and composition of their active sites. There exist [FeFe], [NiFe], and [Fe] hydrogenases, which vary in their catalytic activity as well as in their sensitivity against dioxygen.

[Fe] n class="Chemical">hydrogenase is the only air-stable form, but oxygen-tolerant [NiFe] hydrogenases have also been discovered (see below). In its catalytic mechanism [Fe] hydrogenase differs substantially from [NiFe] and [FeFe] hydrogenases, as the cofactor methylenetetrahydromethanopoterin acts as a hydride acceptor following heterolytic cleavage ozf H2.[11,12] Oxygen-induced inhibition is best understood for [NiFe] hydrogenases and involves two states. The less oxidized Ni–B state is readily reactivated upon reduction by one electron and, according to X-ray data, harbors a hydroxy ligand that bridges the Ni and the Fe center of the active site. In the unready Ni–A state, which is hard to reactivate, the same position is probably occupied by a hydroperoxo species.(13)

Interestingly, apart from enzyme variants with amino acid substitutions in the gas diffusion channel,[14,15] remarkably oxygen tolerant [Nin class="Chemical">Fe] hydrogenase variants were identified in R. eutropha and A. aeolicus.[16−19] Analysis of the membrane-bound hydrogenase of R. eutropha revealed the presence of six instead of the usual four cysteine residues in the vicinity of the H clusters cubane moiety which are crucial for the increased O2 tolerance of this enzyme.(18) These variants provide the first direct link between changes in the primary sequence of the protein and altered reactivity of the active site with respect to O2. However, a mechanistic explanation for this observation and evidence that this structural feature can be transferred to other hydrogenases are still lacking.

In terms of hydrogen formation, [n class="Chemical">FeFe] hydrogenases show the highest catalytic activity and are therefore the most interesting candidates for clean energy production. At the same time, however, they are most sensitive to molecular oxygen. O2-tolerant [FeFe] hydrogenases have not yet been identified or designed, and oxygen exposition of the active enzyme even at low oxygen levels rapidly leads to irreversible inactivation accompanied by the loss of H cluster signals in electron paramagnetic resonance (EPR) spectroscopy.[20,21] Interestingly, species-dependent differences in O2 sensitivity exist(22) which point to the possibility of obtaining enzyme variants with increased O2 tolerance by means of protein engineering. A prerequisite for this undertaking is to understand the mechanism of O2-induced inhibition which up to now has not been unambiguously elucidated.

In a theoretical study we addressed this question and showed that O2 exothermically coordinates to the distal n class="Chemical">Fed atom of the H cluster,(23) which is in line with experimental findings obtained by protein film voltammetry.(22) Protonation and water abstraction can lead to a highly reactive oxo compound that could further react by disintegrating the ligand environment of the 2[Fe]H subsite.(23) Dioxygen coordination to the H cluster was also studied in a quantum mechanics/molecular mechanics (QM/MM) approach by Dogaru et al.,(24) in which the QM region comprised only the 2[Fe]H subcluster but not the [Fe4S4] cubane which we find to be of crucial importance because of magnetic coupling subtleties within the H cluster. Of course, reactions that occur at the cubane subcluster also require a QM description.

Remarkably, O2 binding is clearly determined by the precise ligand arrangement around the active n class="Chemical">Fed atom. We showed that the active site of [Fe] and [FeFe] hydrogenases share a common structural principle that allows one to relate both.(25) In sharp contrast to [FeFe] hydrogenase, coordination of O2 to the active iron atom of [Fe] hydrogenase is endothermic[26,27] (also found by Dey(28)), which accounts for the oxygen tolerance of this enzyme. When we swap the first-shell ligand environment of the Fed atom into that of the central Fe atom of [Fe] hydrogenase, we obtain almost an inversion of the sign of the O2 coordination energy: Fed binds O2 less exothermically, whereas the binding to the Fe atom of [Fe] hydrogenase in a swapped ligand sphere becomes exothermic.(27)

Experimental studies on the [Fen class="Chemical">Fe] hydrogenase from C. reinhardtii led to the suggestion that oxygen-induced inactivation is connected with dramatic structural changes in the cubane subsite of the H cluster caused by reactive oxygen species such as O2–.(29) Such a mechanism was shown to be crucial for the enzyme aconitase, where oxidation is supposed to induce release of a Fe2+ ion from the cubane in the active site to produce a [Fe3S4] cluster.(30) Because of the numerous sulfur-containing ligands surrounding the H cluster an alternative mechanism of oxygen-induced degradation consists of sulfoxygenation, as discussed in a recent review by Darensbourg and Weigand.(31)

Here, we employ density functional theory to investigate possible reaction pathways for the formation of reactive oxygen species and their subsequent attack at the cubane subsite of the H cluster (for details on the computational methodology see the Appendix, where we also discuss possible shortcomings of our approach). We focus on the isolated n class="Chemical">metal cluster (anchored by thiomethanolate ligands in positions as in the full protein) because a comprehensive analysis of its reactivity with respect to O2 is the prerequisite to analyze and understand the influence of the protein environment in which it is embedded. Because of the huge multitude of possible reaction pathways determined by the specific attack sites at the H cluster, the type of the reactive oxygen species, and the overall charge and spin states as well as by the modulation of amino acid residues in the vicinity of the active site, a direct attempt to tackle the problem from the start with a full structural model of the protein is hardly feasible and may obstruct the discovery of the principles of the process, as the results would have been obtained only for a specific enzyme. In contrast, the present study of the generic active site will open up the path for protein engineering attempts to design a more oxygen-tolerant enzyme variant. Still, the elucidation of all possible oxidation pathways at the generic (anchored) H cluster is already a formidable task, as will become evident in the discussion to follow.

Reactivity of O2 along the H Cluster

While we have already studied oxygen-induced decomposition reactions at the distal n class="Chemical">iron atom in detail in refs (23, 27, and 32), we here consider a detachment of the O2 ligand to facilitate its diffusion along the cluster toward the [Fe4S4] cubane. In our quantum chemical calculations we choose an H cluster model as described in ref (27): i.e., cysteine ligands have been replaced by thiomethanolate molecules where one C–H bond was spatially fixed according to the Cα–Cβ bond as observed in the crystal structure of the [FeFe] hydrogenase of Cl. pasteurianum.(33) We focus on the active oxidized form Hoxcat of the H cluster with a total electric charge set to 3– elementary charges. In nature, the high negative charge of the active site is compensated by the protein scaffold. In the case of our model, the continuum solvation model COSMO(34) was used to mimic electrostatic effects, which would otherwise be induced by the protein environment, and to minimize charge artifacts, which could arise when species with different charges are compared. EPR studies show that the Hoxcat form of the enzyme should correspond to a doublet state.(3) In order to obtain more insight into the spin-dependent reactivity of the H cluster, we also compare with results obtained for the quartet state.

For our study of the reactivity of the isolated H cluster with oxygen species, we should, however, keep in mind which atoms of the H cluster are spatially accessible for the n class="Chemical">dioxygen species. For this purpose, we first assess the space available for oxygen moving along the H cluster when embedded in the protein. The published crystal structures of [FeFe] hydrogenase generally feature a high content of crystal water molecules, especially in the vicinity of the H cluster. Figure 1 shows a water-filled path present in the crystal structure of the [FeFe] hydrogenase from C. pasteurianum that connects the 2[Fe]H subcluster and the associated cubane. This could facilitate the diffusion of reactive oxygen species, provided the dynamics of the protein allows for subtle rearrangements of side chains such as Cys299 and Met497 (Figure 1). Unfortunately, no crystal structure exists for the hydrogenase from C. reinhardtii, for which the oxidative damage of the cubane was proposed,(29) but the high degree of sequence similarity between the clostridial and the algal protein suggests the presence of a similar hydrophilic channel also for the enzyme from C. reinhardtii. This is also supported by the crystal structure of a 2[Fe]H subcluster-free form of this hydrogenase.(35) Therefore, water-filled tracks around the H cluster where single water molecules can be displaced by reactive oxygen species (ROS) appear to be an effective means by which these compounds can reach the [Fe4S4] moiety.

Figure 1

(top) Structure of the H cluster model studied in this work. Element color code: red, O; gray, C; blue, N; yellow, S; brown, Fe; white, H. The introduced numbering of the iron atoms is used consistently through the paper. (bottom) The H cluster in the crystal structure of [FeFe] hydrogenase from C. pasteurianum (PDB entry 3C8Y(33)). Water molecules from the crystal structure that connect the two subclusters are depicted as blue spheres.

(top) Structure of the H cluster model studied in this work. Element color code: red, O; gray, C; blue, N; yellow, S; brown, Fe; white, H. The introduced numbering of the n class="Chemical">iron atoms is used consistently through the paper. (bottom) The H cluster in the crystal structure of [FeFe] hydrogenase from C. pasteurianum (PDB entry 3C8Y(33)). Water molecules from the crystal structure that connect the two subclusters are depicted as blue spheres.

3O2 Binding Sites

If we now consider oxygen species moving along the H cluster, we must first study possible binding sites. Figure 2 compares the reaction energies for the formation of difn class="Chemical">ferent O2 adducts of the H cluster in the Hoxcat form and confirms[23,32] that the 2[Fe]H subsite is the primary site of O2 attack also for thermodynamic reasons (apart from the kinetic reason that it is the first site reached from the gas diffusion channel). To compare all isomers, we choose the added energies of 3O2 and the isolated Hoxcat form as energy reference. In the BP86 calculations, the low-spin (doublet) state is always energetically favored over the quartet state (see Figure 2), irrespective of the O2 adduct formed. However, care must be taken in all cases where the quartet state is close in energy, because pure density functionals such as BP86 favor low-spin states over high-spin states and the ordering might reverse when the energy gap for a given species is small.[36−38] However, this would then hardly change the overall energetic picture of the reactions considered. Consequently, it is well justified to assume a 2[Fe]H subsite oxygen adduct as a starting structure and to investigate possible subsequent reaction events that lead to an irreversible destruction of the H cluster.

Figure 2

Comparison of different possible dioxygen adducts of the active oxidized Hoxcat form of the H cluster (total charge is 3– elementary charges). The structures have been optimized with BP86/RI/TZVP/COSMO in order to account for the high negative charge. All clusters are described in doublet (blue) and quartet (red) spin states. The point of zero-energy reference has been chosen as the free 3O2 and the isolated Hoxcat form. Shaded ovals highlight the coordinated O2 molecule. Element color code: red, O; gray, C; blue, N; yellow, S. brown, Fe; white, H.

Comparison of different possible n class="Chemical">dioxygen adducts of the active oxidized Hoxcat form of the H cluster (total charge is 3– elementary charges). The structures have been optimized with BP86/RI/TZVP/COSMO in order to account for the high negative charge. All clusters are described in doublet (blue) and quartet (red) spin states. The point of zero-energy reference has been chosen as the free 3O2 and the isolated Hoxcat form. Shaded ovals highlight the coordinated O2 molecule. Element color code: red, O; gray, C; blue, N; yellow, S. brown, Fe; white, H.

As it is a priori not certain that the initial attack of O2 on the H cluster takes place when the enzyme is in its oxidized form, we also considered formation of difn class="Chemical">ferent O2 adducts of the H cluster in the reduced Hredcat form (Figure 3). Again, the 2[Fe]H subsite is the primary site of O2 coordination. Interestingly, the dioxygen adduct to the distal iron atom Fed of the 2[Fe]H subsite is energetically the most favored one for the reduced form of the enzyme. To compare all isomers, we choose as energy reference 3O2 and the isolated Hredcat form. Here, for the oxygen adduct 6 in the reduced state, the triplet configuration is more stable than the singlet one because of a ligand rearrangement that occurs during structure optimization. Hydrogen bonding from the NH group of the dithiomethylamine (dtma) bridge of the dinuclear subsite to the CN– ligand of the iron atom Fed is responsible for stabilizing this adduct, although this is not likely to happen in the protein environment and is thus an artifact of our minimal model system.

Figure 3

Comparison of different possible dioxygen adducts of the active reduced Hredcat form of the H cluster (total charge is 4– elementary charges). The structures have been optimized with BP86/RI/TZVP/COSMO, and for the sake of simplicity, we use the same numbering of structures as in the case of the oxidized form in Figure 2. All clusters are described in singlet (blue) and triplet (red) spin states. The point of zero-energy reference has been chosen as the free 3O2 and the isolated Hredcat form. Shaded ovals highlight the coordinated O2 molecule. Element color code: red, O; gray, C; blue, N; yellow, S; brown, Fe; white, H.

Comparison of different possible n class="Chemical">dioxygen adducts of the active reduced Hredcat form of the H cluster (total charge is 4– elementary charges). The structures have been optimized with BP86/RI/TZVP/COSMO, and for the sake of simplicity, we use the same numbering of structures as in the case of the oxidized form in Figure 2. All clusters are described in singlet (blue) and triplet (red) spin states. The point of zero-energy reference has been chosen as the free 3O2 and the isolated Hredcat form. Shaded ovals highlight the coordinated O2 molecule. Element color code: red, O; gray, C; blue, N; yellow, S; brown, Fe; white, H.

Possibility for Catalytic Formation of Reactive Oxygen Species

In the previous work published by our group, we studied a series of protonation events and water abstraction that could possibly lead to a breakdown of the 2[n class="Chemical">Fe]H subsite.(23) However, for the [FeFe] hydrogenase of Cl. pasteurianum the aforementioned experimental data of Armstrong, Happe, and co-workers(29) suggest decomposition of the [Fe4S4] cubane following the conversion of the coordinating O2 to a reactive oxygen species (ROS).

As reported previously, 3O2 addition to the distal n class="Chemical">Fed atom of the 2[Fe]H subsite in the present model cluster is exothermic by −17.5 kcal/mol(27) (here, we find a slightly different value of −16.2 kcal/mol, as we have applied tighter convergence thresholds throughout; see Appendix). If the Hoxcat form of the enzyme is considered (see Figure 2), the coordinated O2 ligand in 2 features an O–O bond length of 1.31 Å, which points to the superoxide species O2– in support of the hypothesis by Armstrong, Happe, and co-workers.(29) Here, we find that dissociation of the bound superoxide from the Fed atom is endothermic by +12.7 kcal/mol and is therefore unlikely to occur.

This situation changes if a one-electron reduction of the cluster is considered which is energetically favored by −21.5 kcal/mol and leaves an overall charge of 4– elementary charges on the H cluster, corresponding to the Hredcat form. Notably, the release of O2– in this redox state is exothermic by −19.6 kcal/mol. As an alternative to n class="Chemical">O2– release, a protonation of the terminal oxygen atom of the attached O2 species could precede a dissociation of OOH•. With Cys299 (nomenclature of the enzyme from C. pasteurianum) as proton donor, we obtain a slightly exothermic reaction energy of −2.2 kcal/mol. However, we find the dissociation of OOH• to be endothermic by +31.5 kcal/mol for the Hoxcat form and by +36.0 kcal/mol for the Hredcat form, respectively. Therefore, in terms of reaction energies reduction and superoxide release from the H cluster in Hredcat form is the only feasible way to produce O2– at the catalytic center. Remarkably, these results suggest that a catalytic cycle for the formation of superoxide can be proposed, provided that rereduction of the H cluster can still be sustained by suitable electron donors (see Figure 4).

Figure 4

Possible catalytic reaction cycle for the formation of reactive oxygen species (ROS) at the H cluster. The given charges correspond to the active oxidized Hoxcat and the active reduced Hredcat forms of the H cluster. All clusters are considered to be in low spin state. Reaction energies in kcal/mol are calculated for fully optimized structures (BP86/RI/TZVP/COSMO) with respect to the isolated Hoxcat form and 3O2, 2O2−, 2OOH•, 1OOH−, 1H2O2, respectively. For reaction energies obtained with the B3LYP functional, see Table 1. Protonation energies were calculated by assuming an energy of −262.4 kcal/mol for a solvated proton.[39,40]

Possible catalytic reaction cycle for the formation of reactive oxygen species (n class="Chemical">ROS) at the H cluster. The given charges correspond to the active oxidized Hoxcat and the active reduced Hredcat forms of the H cluster. All clusters are considered to be in low spin state. Reaction energies in kcal/mol are calculated for fully optimized structures (BP86/RI/TZVP/COSMO) with respect to the isolated Hoxcat form and 3O2, 2O2−, 2OOH•, 1OOH−, 1H2O2, respectively. For reaction energies obtained with the B3LYP functional, see Table 1. Protonation energies were calculated by assuming an energy of −262.4 kcal/mol for a solvated proton.[39,40]

Table 1

Reaction Energies for the Reactions Shown in Figure 44 with the Active Oxidized Hoxcat and the Oxidized Inactive Hoxinact Forms of the H Cluster as the Starting Pointsa

	Hoxcat—Hredcat		Hoxinact—Hoxcat
reacn	BP86	B3LYP	BP86	B3LYP
1a	–16.2	–8.6	–13.6	–6.0
2a	–23.4	–23.5	–58.4	–81.4
1b	–21.5	–10.3	–55.8	–78.8
2b	–18.1	–21.9	–16.6	–8.6
3a	–19.6	–25.5	+12.9	+29.7
3bext	–100.6	–99.4	–66.6	–50.1
3bint	–2.2	+16.8	+8.5	+19.8
1c	–31.5	–23.9	–30.0	–23.9
4	–26.0	–27.0	–57.4	–72.9
5a	–11.2	–16.4	+18.6	+35.4
5bext	–86.9	–89.3	–60.9	–60.5
5bint	+6.9

B3LYP energies (kcal/mol) were obtained for single-point calculations with the COSMO solvation model (ε = 4) on BP86/RI/TZVP/COSMO optimized structures. Protonation energies marked with the superscript “ext” were calculated by assuming, for a solvated proton, an energy of −262.4 kcal/mol.[39,40] For comparison, the energies for internal protonation with proton transfer from Cys299 are marked with the superscript “int”. In the case of 5bint and Hoxinact—Hoxcat the proton moves back to the cysteine in our structure optimization with tight convergence thresholds.

B3LYP energies (kcal/mol) were obtained for single-point calculations with the COSMO solvation model (ε = 4) on BP86/RI/TZVP/COSMO optimized structures. Protonation energies marked with the superscript “ext” were calculated by assuming, for a solvated proton, an energy of −262.4 kcal/mol.[39,40] For comparison, the energies for internal protonation with proton transfer from n class="Chemical">Cys299 are marked with the superscript “int”. In the case of 5bint and Hoxinact—Hoxcat the proton moves back to the cysteine in our structure optimization with tight convergence thresholds.

Interestingly, when one considers further protonation and reduction events, the formation of H2O2 and n class="Chemical">OOH– also appears to be feasible at the 2[Fe]H subsite, as the scheme shown in Figure 4 demonstrates. In this case protonation of the reduced adduct 2 (total charge is 4– elementary charges) can take place as a competing pathway to O2– release. This induces reduction of the superoxide ligand to a peroxide-like species, as can be deduced from the increased oxygen–oxygen bond length of 1.47 Å. We study the protonation reaction both by considering the transfer of an external solvated proton and by choosing Cys299 as the proton donor (see Table 1). Protonation is exothermic by −100.6 kcal/mol if a solvated proton is considered and by −2.2 kcal/mol for the internal proton transfer from Cys299. Subsequent reprotonation of Cys299 is energetically favored by −98.4 kcal/mol (external protonation by a solvated proton).

A subsequent reduction of the active site by one electron is then exothermic by −26.0 kcal/mol and can be followed by one of two reaction events that close the catalytic cycle for peroxide formation. The first possibility is release of a n class="Chemical">OOH– ion (−11.2 kcal/mol), which can be later protonated to finally form H2O2. The second option is protonation of the OOH adduct to form H2O2, which is found to spontaneously dissociate from the active site. The reaction energy for this step is −86.9 kcal/mol for the addition of H+ from solution and +6.9 kcal/mol if Cys299 is considered as a proton donor. Hence, the former possibility appears to be more likely.

A final reprotonation of the cysteine residue 299 is again exothermic by −93.7 kcal/mol. Both reaction paths close the catalytic cycle, leaving the active site in its oxidized, catalytically active form Hoxcat. Interestingly, H+ transn class="Chemical">fer from Cys299 to the OOH adduct cannot take place before the system is reduced by one electron, because otherwise the proton returns back to the sulfur atom of the cysteine residue in the course of structure optimization. Hydrogen peroxide can also be formed by rebinding of protonated O2– to the distal iron atom (Hoxcat redox state of the cluster), which is exothermic by −31.5 kcal/mol, and can be followed by the same reaction events as described previously starting with the one-electron reduction of the system. On the basis of the reaction scheme proposed in Figure 4, production of O2– and OOH– appears to be a feasible toxic pathway catalyzed by the H cluster.

The surprising observation(2) of O2 tolerance of the otherwise extremely n class="Chemical">oxygen sensitive [FeFe] hydrogenase in its anaerobically generated inactive form Hoxinact can now be explained. Let us consider Figure 4 (for reaction energies see Table 1) with a starting charge state of the H cluster of 2– instead of 3– elementary charge. The first step—coordination of an oxygen molecule to the Fed atom—is less exothermic than for the Hoxcat form (−13.6 kcal/mol) but still feasible. However, all subsequent steps involving proton transfer from Cys299 are not likely to occur. The proton returns to Cys299 when internal protonation is considered or, when considering a solvated proton (see reactions 3b and 5b in Table 1), the protonation is almost 30 kcal/mol less exothermic than for the Hoxcat form of the enzyme. The alternative pathway with O2– release (reaction 3a) is endothermic by +12.9 kcal/mol. Finally, the dissociation of OOH– (reaction 5a) is endothermic by +18.6 kcal/mol. Therefore, for the anaerobically prepared Hoxinact state, the formation of ROS at the 2[Fe]H subsite is not likely and cannot lead to irreversible cluster destruction.

Finally, we should note that the reaction energies obtained with the B3LYP functional (given in Table 1) are in line with the BP86 data discussed.

Possible Oxidants

Summarizing what we have found so far, triplet O2 binds to the distal n class="Chemical">iron atom of the H cluster and can exothermically be transformed into superoxide or hydroperoxide anion. Thereafter, O2– and OOH– can be protonated to become an OOH• radical and H2O2, respectively. Such protonation events can be facilitated in principle by protic residues such as cysteine, tyrosine, histidine, and glutamic acid. For these amino acids we calculate +8.5, +6.6, −40.4, and −4.4 kcal/mol, respectively, for proton transfer from an isolated amino acid to O2– to form OOH•. Analogously, for proton transfer from cysteine, tyrosine, histidine, and glutamic acid to OOH– to form H2O2, we obtained −14.7, −16.5, −63.5, and −27.6 kcal/mol, respectively. Hence, protonation of O2– from a cysteine is unlikely to occur, while a protonated histidine or glutamic acid in the vicinity of the H cluster could accomplish this. However, there is no histidine residue in the cluster environment but glutamic acid (Glu495) is located in close proximity to the active site and could in principle act as a proton donor. In contrast, protonation of OOH– from any of the residues considered (leaving an anionic residue behind) is very exothermic and thus likely to take place. The cysteine residue Cys299 is located just above the postulated site of OOH– dissociation and could therefore easily act as a proton donor for the formation of H2O2. To determine which ROS species are most likely to attack the active site, it is important to note that O2– can undergo spontaneous disproportionation:(41)This is the fourth possible way to produce H2O2 around the cluster. Therefore, hydrogen peroxide appears to be the most convincing reactant that could be responsible for the irreversible inhibition of the H cluster. Nevertheless, we continue to investigate the action of all four reactive oxygen species which can be catalytically produced at the active site: O2–, OOH•, OOH–, and H2O2.

At first, we study the coordination of the negatively charged species O2– and n class="Chemical">OOH–. Superoxide adducts to the H cluster are presented in Figure 3 for O2 addition, but the formation is always impeded by endothermic reaction energies (not shown). For instance, the energetics of O2– coordination to the Fe1 and Fe2 atoms of the cubane in the Hoxcat state of the cluster (see adducts 5Fe and 7Fe in Figure 3) are +33.0 and +33.2 kcal/mol. All attempts to coordinate O2– to a sulfur atom of the cubane resulted in dissociation of the superoxide, leaving the cluster intact. In general, formation of OOH– adducts at iron or sulfur atoms of the cubane is endothermic by more than +20 kcal/mol.

As a consequence, we consider O2– and n class="Chemical">OOH– to be important in the migration step from the distal iron atom toward the cubane cluster, as these species can dissociate from the distal iron atom. Thus, reaction at the cubane most likely occurs only when superoxide is protonated to yield OOH• or again reduced to form H2O2. Therefore, we now exclude O2– and OOH– for the subsequent mechanism of H cluster inhibition.

The OOH• Radical As Damaging Agent

As already outlined above, the direct attack of O2 or n class="Chemical">O2– at the [Fe4S4] cubane is almost thermoneutral and therefore cannot explain the decomposition of the H cluster. This situation changes dramatically when we consider the protonated superoxide ion that yields the radical OOH•.

While we are primarily interested in reactions that take place at the cubane subcluster, we should briefly discuss what OOH• can induce at the dinuclear n class="Chemical">iron subsite. Simple rebinding of superoxide in the protonated OOH• form to the distal iron atom Fed of the 2[Fe]H subsite (Figure 5a) is energetically favored by −31.5 kcal/mol and can be—next to the OOH• binding to the proximal iron atom Fep (Figure 5b) with a reaction energy of −26.7 kcal/mol—the preferred site of OOH• coordination to the active site. The OOH ligand can then be further reduced and may dissociate as hydrogen peroxide, as shown in the scheme of Figure 4. It is also possible that OOH• oxidizes sulfur atoms from the dtma ligand linking the two irons of the 2[Fe]H subsite. This reaction is exothermic by −20.4 kcal/mol; the product is depicted in Figure 5. Nevertheless, OOH• reactions with the cubane will be of importance when the protonation of superoxide is facilitated in close proximity to the cubane rather than to the 2[Fe]H subsite, which is the case when Glu495 is considered as the proton donor. Note that Figure 5b shows a structure where the S–Fe bond, which links the subsite to the cubane through Cys503, is cleaved. A similar process has been found in the context of the inhibitor CO reacting with the H cluster.(42)

Figure 5

Structures of OOH adducts of the 2[Fe]H subsite optimized with BP86/RI/TZVP/COSMO for a total charge of 3– elementary charges in all cases: (a) distal adduct; (b) proximal adduct; (c) oxidized dtma ligand. The clusters are considered in their low spin states (singlet). Shaded ovals highlight coordinated OOH• or reaction products thereof.

Note that the variety of additions and reaction events which can follow the initial OOH• coordination to the cubane is too large to be investigated systematically. For example, three options exist to approach a single n class="Chemical">iron center by the OOH• species. Some adducts feature intact cubane clusters, while in others the cubane is opened. These structures are very similar in energy (within a range of up to 3 kcal/mol). Here, we report an intact cubane structure for Fe-OOH and an open one for Fe-OOH (see Figure 6). All other structures obtained are presented and discussed in detail in the Supporting Information.

Figure 6

Structures of OOH• coordination to the Fe1 (a) and Fe2 (b) iron atoms of the [Fe4S4] cubane, Fe1-OOH and Fe2-OOH, respectively, optimized with BP86/RI/TZVP/COSMO (see Figure 1 for the labeling of atoms). Structure Fe2-OOH(OH)(μ-O) depicted in (c) was obtained after coordination of a second OOH•. Structure Fe2-(μ-O)[(S)OH] in (d) emerges after activation of the OOH• bound in Fe2-OOH by bending the FeOO angle. The charge is 3– elementary charges in all cases. The clusters are described in the low-spin states. Shaded ovals highlight the coordinated OOH• molecule.

Structures of OOH• coordination to the n class="Chemical">Fe1 (a) and Fe2 (b) iron atoms of the [Fe4S4] cubane, Fe1-OOH and Fe2-OOH, respectively, optimized with BP86/RI/TZVP/COSMO (see Figure 1 for the labeling of atoms). Structure Fe2-OOH(OH)(μ-O) depicted in (c) was obtained after coordination of a second OOH•. Structure Fe2-(μ-O)[(S)OH] in (d) emerges after activation of the OOH• bound in Fe2-OOH by bending the FeOO angle. The charge is 3– elementary charges in all cases. The clusters are described in the low-spin states. Shaded ovals highlight the coordinated OOH• molecule.

Coordination of OOH• to iron atom Fe1 of the cubane, for instance, is exothermic by −18.6 kcal/mol, with similar energies to be found for the addition to the remaining three Fe atoms.

We now proceed to investigate possible pathways for irreversible cubane disintegration following OOH• attack. First, we study the addition of a second n class="Chemical">OOH• species to the newly formed free coordination site at iron atom Fe2 in Fe-OOH, which resulted after cleavage of an iron–sulfur bond upon attack of the first OOH• radical. Structure optimization of this species yields Fe-OOH(OH)(μ-O) (Figure 6c) featuring a broken O–OH bond with an energy of −51.5 kcal/mol forIn Fe2-OOH(OH)(μ-O), the oxo bridge is formed between Fe2 and Fe4 (see Figure 1 for the labeling of atoms).

Another possible decomposition pathway of the [Fe4S4] cubane may open up in an intramolecular reaction, when we distort (activate) the optimized structure of the n class="Chemical">OOH adduct Fe2-OOH by decreasing the O···S distance with respect to the coordinating cysteine. If it is considered as an elementary reaction, this distortion may require an activation energy that is too high to be thermally activated. However, structure optimization of a distorted species with an O–S distance of 1.6 Å results in splitting of the O–OH bond and significant distortion of the cubane in the structure Fe2-(μ-O)[(S)OH], in which the OH fragment binds to a sulfur atom (see Figure 6d). These structural changes are accompanied by an overall exothermic reaction of −38.0 kcal/mol. Attempts to locate transition states for such a reaction turned out to be very difficult. Nevertheless, these attempts have led us to the conclusion that a direct intramolecular attack in an elementary (one-step) reaction requires an energy that is too high to be important for cluster decomposition reactions. However, such an intramolecular decomposition reaction does not need to be considered as an elementary reaction step. Instead, it may be feasible if it proceeds stepwise.

In order to explore this latter option, we calculated the intrinsic bond energies of O–O and Fe–S in n class="Chemical">Fe2-OOH: i.e., the electronic energy differences for bond breaking if the products are taken in the structure they adopt in Fe2-OOH. These (intrinsic) reaction energies for structurally frozen fragments are a good estimate for an activation barrier of such a bond-breaking reaction which produces two independent molecular fragments. We obtained an intrinsic bond energy of 37.5 kcal/mol for the O–O bond and of 26.9 kcal/mol for the Fe–S bond (remember that we break the Fe–S bond to the Fe atom which already coordinates the OOH species). Hence, to first activate the Fe–S bond appears to be feasible and a subsequent structure optimization shows that the actual dissociation energy is only +0.8 kcal/mol for breaking the Fe–S bond with subsequent structural relaxation to yield the intermediate Fe2-OOH···[S] (for the sake of comparison, taking structural relaxation in the O–O bond breaking of Fe2-OOH into account reduces the endothermicity of this process only to +26.7 kcal/mol). The Fe–S bond-breaking step is thus basically thermoneutral and can be considered a first step toward the production of Fe2-(μ-O)[(S)OH]. In order to understand whether subsequent steps are also feasible, we calculated the dissociation of OH from the intermediate Fe2-OOH···[S] to form the second intermediate Fe2-O···[S], which requires an energy of only +24.0 kcal/mol. The reaction of this dissociated OH radical with the noncoordinating sulfur atom of the second intermediate Fe2-O···[S] to yield the final product Fe2-(μ-O)[(S)OH] is exothermic by −31.8 kcal/mol.

As a final remark, we note that the energy for coordination of a second OOH• to Fe3 in the distorted cubane Fe2-(μ-O)[(S)OH] is −17.7 kcal/mol. Note that the reaction energy for an addition of a second OOH• to Fe1 in the intact cubane of Fe1-OOH is exothermic by −21.7 kcal/mol, while the coordination of a second OOH• to Fe4 of Fe1-OOH yields −25.5 kcal/mol.

Addition of further OOH• radicals are exothermic and lead to even more pronounced structural changes of the n class="Chemical">iron–sulfur cluster. It is clearly visible that already the [Fe4S4] unit of the structure depicted in Figure 6d is no longer of cuboidal shape. This structural deformation is in line with the fact of changed iron–sulfur distances observed in the X-ray absorption spectroscopy (XAS) spectra of [FeFe] hydrogenase after oxygen exposure(29) with a still intact 2[Fe]H subsite.

Release of an iron atom from an oxidized [Fe4S4] cubane was proposed as a possible decomposition pathway for the iron–sulfur cluster of the enzyme aconitase.(30) At this stage, it is not possible to decide whether the iron atom Fe2 can be dissolved, but we may refer to the analogous case in the next section on sulfoxygenation. It is, however, reasonable that the final structure, which is a result of a highly exothermic reaction, may release a solvated iron ion.

Clearly, the uncharged OOH• is much more reactive than n class="Chemical">superoxide and OOH• binding could account for the irreversible destruction of the cubane. The OOH• radical is known to be very reactive and usually is created shortly before the actual reaction takes place. The presented results confirm that OOH• is capable of [Fe4S4] cubane decomposition, provided that protonation of superoxide to produce this highly reactive species can be accomplished near the cubane.

Sulfoxygenation

The catalytic formation of hydrogen peroxide at the distal n class="Chemical">iron atom opens the possibility for several competing decomposition reactions at the active center. It is well-known that H2O2 can oxidize thiolate and thioether sulfur atoms. Hence, it is no surprise that the formation of oxidized cysteines, i.e., of CysSO groups, was also suggested for the oxidized form of [NiFe] hydrogenase on the basis of a combined XAS, EPR, Fourier transform infrared (FTIR) spectroscopy, and broken-symmetry (BS) DFT study.(17) Such reactions can also take place in the vicinity of the H cluster. For this section we optimized possible reaction products of H2O2 attack by placing this molecule in the close vicinity of various atoms of the H cluster. Of course, this may create high-energy structures that can then relax by O–O bond cleavage.

We should stress that H2O2 does not bind to the distal n class="Chemical">Fed iron atom of the 2[Fe]H subsite in the Hoxcat state. If we place H2O2 in close vicinity of Fed, the proton moves from the oxygen atom proximal to Fed to the second O atom such that water is formed with a reaction energy of –35.6 kcal/mol. The remaining oxygen atom is then a ligand to Fed (see top left structure in Figure 7 and compare also ref (23)). The water formation reaction, however, is 8 kcal/mol less exothermic than the competing oxidation of the thiolate sulfur atoms, which can take place either at the dtma bridge directly connected to the distal iron atom Fed or at the closest cysteine residue Cys503. Since both the distal iron atom and the sulfur atoms are easily accessible for H2O2, the more exothermic reactions with thiolates are more likely to take place than the oxidation of the iron atom Fed. The latter reaction would lead to the competing degradation pathway discussed in the previous study reported by our group.(23)

Figure 7

Reaction products of hydrogen peroxide addition to the 2[Fe]H subcluster. Structures were optimized with BP86/RI/TZVP/COSMO considering a total charge of 3– elementary charges in all cases. The clusters are described in their low-spin states (doublet). Shaded ovals highlight the oxygen-based reaction products.

Generally, we observe that H2O2 pren class="Chemical">ferentially binds and reacts with thiolate groups, while we were not able to converge adducts of H2O2 with Fe atoms of the cubane. Sulfur atoms of the [Fe4S4] subsite are also not likely to be the primary site of H2O2 attack, because in all calculations performed H2O2 dissociates from the cluster during structure optimization, indicating that the cluster would at least require activation, if feasible at all. This result is in line with the observation of the sensitivity of charged Cys residues against H2O2. However, a sulfur atom at the cubane might be oxidized to form a sulfoxide compound after H2O2 is first placed close to a thiolate group of a coordinating cysteine ligand (see Figure 8d).

Figure 8

Reaction products for the addition of hydrogen peroxide to cysteines coordinating the [Fe4S4] cubane: (a) product IFe1 of H2O2 addition to Cys503; (b) IFe2 for H2O2 addition to Cys355; (c) IIFe2 to Cys355; (d) product IFe3 of H2O2 addition to Cys300. Structures were optimized with BP86/RI/TZVP/COSMO for a total charge of 3– elementary charges in all cases. The clusters are considered to be in their low-spin states (doublet). Shaded ovals highlight the oxygen-based reaction products.

From the proposed site of H2O2 formation at the distal n class="Chemical">Fed iron atom, the thiolate groups that are most easily accessible belong to the bidentate dtma ligand. Here, H2O2 brought in close proximity to the dtma bridge leads to three different products, depending on the initial possible structure from which a calculation was started. The parameter that can be chosen completely freely upon attack of H2O2 to the dtma bridge is the dihedral angle among the oxygen atoms of hydrogen peroxide, the sulfur atom of the dtma bridge, and the distal iron atom Fed. The most exothermic among the products obtained (−43.2 kcal/mol) converges by cleavage of the O–O bond and water formation, while the sulfur atom is oxidized to a sulfoxide group (see Figure 7). Importantly, these results are obtained upon structure optimization without any activation of the H cluster.

For a complete picture of the attack of bridging sulfur atoms by n class="Chemical">H2O2, we need to investigate transition-state barriers. Figure 9 shows a transition state for the reaction of H2O2 with the subcluster-bridging sulfur atom of cysteine 503. The transition state was obtained by starting from an optimized structure of the free H cluster with overall charge of 3– elementary charges and varying the distance between one H2O2 O atom and the bridging sulfur atom between 1.67 and 2.57 Å. For each case a full structural optimization was carried out while the Cartesian coordinates of the oxygen and the sulfur atom were kept fixed. For the maximum energy structure a more refined search was carried out by varying the O–S distance between 2.25 and 2.45 Å. The structure that featured the highest energy was obtained for a O–S distance of 2.35 Å. A vibrational analysis of this structure revealed six imaginary frequencies (i179.44, i121.52, i119.62, i104.78, i68.04, and i46.71 cm–1) with the second frequency corresponding to the O–O stretch vibration. The second normal mode was then considered in an eigenvector-following approach as implemented in the Turbomole package to search for the transition state (see Appendix). Here, convergence could only be achieved by fixing the O–S distance to 2.35 Å during the search. Care was taken to converge a broken-symmetry solution that corresponds to the experimentally observed spin coupling of the oxidized H cluster. Subsequent frequency analysis revealed two imaginary frequencies (i80.66 and i4.90 cm–1), with the first one corresponding to the O–O stretching vibration and the second one being an artifact of the seminumerical analysis. For determination of the activation barrier using this approximate transition state the electronic energy was re-evaluated by employing the COSMO solvent screening model and assuming a dielectric constant of 4.0. The activation barrier is only 7.6 kcal/mol and provides a reasonable upper bound for the activation energy of H2O2-induced oxidation of the subcluster-bridging sulfur atom of cysteine residue 503 (according to the C. pasteurianum sequence). Hence, the reaction of H2O2 with bridging sulfur atoms is clearly kinetically feasible.

Figure 9

Transition state for the reaction of H2O2 with the subcluster-bridging S atom of Cys 503 (C. pasteurianum).

Oxidation of sulfur atoms from bridging n class="Chemical">thiolates in compounds biomimetic to the active site of [FeFe] hydrogenase was discussed in a recent review by Darensbourg and Weigand.(31) The molecular structures with oxidized sulfur atoms presented in their review are comparable to the structures predicted by our DFT calculations (Figure 7). However, our study shows that an alternative product can be formed that features a S–OH moiety at the dtma ligand. The second OH group of the attacking H2O2 forms a water molecule after protonation by Cys299 (see Figure 7d). The energy for this reaction is −36.5 kcal/mol. The least exothermic (−29.0 kcal/mol) reaction product that could be formed by the reaction of H2O2 with the dtma bridge consists of a S–OH moiety and an OH ligand at the distal Fed atom (Figure 7b).

After the dtma ligand, the next accessible n class="Chemical">thiolate group is at the cysteine residue that connects the 2[Fe]H subsite with the cubane (Cys503). In this case, the most exothermic product (−40.8 kcal/mol) features a split O–O bond of H2O2 which is accompanied by breaking of the bond between the cysteine S atom and the Fe1 atom of the cubane, thereby leaving oxidized S and Fe1 atoms both carrying an OH group (mode IFe1; see Figure 8a). Notably, for [Fe4S4]-carrying proteins as aconitase or HiPIPs (high potential iron–sulfur proteins) the formation of Fe–OH species is considered to be the initial step in [Fe4S4] degradation, with the possibility that the oxidized iron atom can be dissolved by water, leading to the formation of a [Fe3S4] cluster.(30) Interestingly, the type of reaction product obtained for the remaining three cysteines anchoring the [Fe4S4] cluster depends on the corner of the cubane considered. At the site of the Fe2 atom (compare Figure 1), which is coupled ferromagnetically to the Fe1 atom in the Hoxcat state of the active site, H2O2 can coordinate in mode IFe2, similar to that described for the Fe1 atom (see Figure 8), with a reaction energy of −42.4 kcal/mol. This type of product was not observed for cysteines connected to iron atoms Fe3 and Fe4. In contrast, when H2O2 attacks the sulfur atoms of cysteines coordinating iron atoms Fe3 and Fe4, it breaks into OH moieties and reattaches to the sulfur atoms from the cubane, oxidizing them as shown in Figure 8d. The byproduct of this reaction is water.

There is a possibility for an alternative adduct formation, common to all the cysteines coordinating the [n class="Chemical">Fe4S4] cluster. As depicted in Figure 8c for Cys355, H2O2 oxidizes the sulfur atoms to sulfoxide groups while the second oxygen atom is released as water. This type of reaction resembles the oxidation of the dtma ligand and delivers a similar overall reaction energy of −41.4 kcal/mol for the oxidation of Cys355, with similar energies for the oxidation of the remaining cysteines. For the iron atom Fe2 of the cubane, both binding modes IFe2 and IIFe2 are energetically nearly degenerate, while for Fe1 the preferred product is IFe1. For Fe4, oxidation of the sulfur atom of the coordinating cysteine residue is energetically preferable over oxidation of a sulfur atom of the cubane by almost 10 kcal/mol.

In contrast to the oxidized dtma ligand or n class="Chemical">sulfoxides formed at the cubane, the oxidized cysteine residues can be further oxidized by bringing a second H2O2 molecule into close proximity. This second reaction step results in formation of a sulfone and water and is somewhat more exothermic than the first step, with −59.8 kcal/mol for product IFe2, −63.7 kcal/mol for IIFe2, and −61.7 kcal/mol for IFe1 (see Figure 10). Of course, while doubly oxidized sulfur atoms, i.e., sulfones, are well-known oxidation products of thioethers, optimization of a transition-state structure would be needed to confirm that sulfone formation is possible under oxygen inhibition conditions.

Figure 10

Reaction products for the addition of a second and a third H2O2 molecule to the H cluster: (a) addition of a second hydrogen peroxide to the oxidized sulfur atom of the structure IFe1; (b, c) H2O2 addition to the sulfoxide of the structures IIFe2 (b) and IFe2 (c); (d) final product of sulfoxygenation of the S atom. The total charge of the clusters is 3– elementary charges in all cases. The clusters are considered to be in their low-spin states (doublet). Shaded ovals highlight the oxygen-based reaction products. Note that the structure shown in (a) would differ substantially if the optimization was performed without the COSMO solvation model.

Reaction products for the addition of a second and a third H2O2 molecule to the H cluster: (a) addition of a second n class="Chemical">hydrogen peroxide to the oxidized sulfur atom of the structure IFe1; (b, c) H2O2 addition to the sulfoxide of the structures IIFe2 (b) and IFe2 (c); (d) final product of sulfoxygenation of the S atom. The total charge of the clusters is 3– elementary charges in all cases. The clusters are considered to be in their low-spin states (doublet). Shaded ovals highlight the oxygen-based reaction products. Note that the structure shown in (a) would differ substantially if the optimization was performed without the COSMO solvation model.

If it would be kinetically possible to oxidize the sulfone for a third time, an R–SO3 group bound to the protein would be produced which no longer can coordinate a n class="Chemical">metal atom of the H cluster. Then, another water molecule is coordinated to the iron atom Fe2 of the cubane (see Figure 10d, where the R–SO3 group appears as Me–SO3 in our structural model), which might subsequently be dissolved as mentioned before in the previous section on distingration induced by OOH•.

When we compare products of H2O2 addition to the H cluster with oxidation products that result from the attack of other n class="Chemical">ROS, we notice that the primary coordination preference that precedes subsequent reaction events is shifted from Fe to S atoms with increasing exothermicity in the series O2, OOH•, H2O2. The oxidation energies for all reactions of hydrogen peroxide with thiolate groups are considerably more exothermic than for other reactive oxygen species and are close to −40 kcal/mol (for the first H2O2 addition). These results are in line with the work of Roth and Jordanov,(43) who showed that H2O2 readily decomposes model compounds of iron–sulfur cubanes in a much more aggressive way than molecular oxygen. Judging on the basis of reaction energies, there is no preferred site of H2O2 attack on the active site. However, the dtma bridge and Cys503 are located in the closest distance to the assumed H2O2 formation site and thus are the easiest to reach.

Note, however, that selectivity of H2O2-induced oxidation of nucleophilic n class="Chemical">cysteines was reported by Weerapana et al.(44) Kim et al. proposed an experimental technique to identify those Cys residues that are prone to oxidation by H2O2,(45) which could be used to validate our findings experimentally. Interestingly, when we tried to oxidize protonated Cys299, H2O2 did not bind to this residue. Hence, not all cysteines have the same sensitivity to H2O2. The cysteine residues ligating the iron–sulfur cubane belong to the group which is more readily oxidized. Thus, sulfoxidation is an important, if not the key, factor of O2 tolerance of hydrogenase enzymes and attack of hydrogen peroxide at the bridging cysteine Cys503 might be the first step for the oxidative decomposition of the [FeFe] hydrogenase H cluster.

Do Partially Destroyed H Clusters Support the Catalytic Production of ROS?

After having shown possible pathways for the decomposition of the H cluster, an important question remains to be answered: can a partially decomposed H cluster support the production of ROS so that the decomposition process is maintained?

The feasibility of catalytic formation of n class="Chemical">ROS at partially destroyed H clusters was investigated for the reaction products of H2O2 addition to the 2[Fe]H subsite (dtmaSO and dtmaSOH structures presented in Figure 7c,d) and to Cys503 coordinating [Fe4S4] cubane (structure IFe1 depicted in Figure 8a). From OOH-derived degradation products we consider these structures: Fe2-OOH and Fe2-OOH(OH)(μ-O) of Figure 6b,c. The results obtained are summarized in Figure 11 and in Table 2. It turned out that the catalytic formation of ROS remains feasible for all of the considered clusters.

Figure 11

Reactions for the formation of ROS at the partially destroyed H cluster, where “[cluster]” represents either the structural unit of the reaction products dtmaSO and dtmaSOH presented in Figure 7c,d (top) or the structural unit of reaction products IFe1, Fe2-OOH, and Fe2-OOH(OH)(μ-O) of Figure 8a and Figure 6b,c (bottom): Reaction energies (kcal/mol) for the top part corresponding to the starting structure dtmaSO are given in Roman type, while for the reaction cycle starting from the dtmaSOH structure, the reaction energies are given in boldface type. Reaction energies (kcal/mol) for the bottom part corresponding to the starting structure IFe1 are given in Roman type, while for the reaction cycle starting from the Fe2-OOH structure, the reaction energies are given in boldface type. Reaction energies of Fe2-OOH(OH)(μ-O)-based clusters are given in italics. Protonation energies were calculated by assuming, for a solvated proton, an energy of −262.4 kcal/mol.[39,40] Given charges correspond to the active oxidized Hoxcat and the active reduced Hredcat forms of the modified H cluster. All clusters are considered in their low-spin state. Reaction energies are calculated for fully optimized structures (BP86/RI/TZVP/COSMO) with respect to 3O2, 2O2−, 2OOH•, 1OOH−, or 1H2O2 and dtmaSO and dtmaSOH, respectively (top), and with respect to 3O2, 2O2−, 2OOH•, 1OOH−, or 1H2O2 and IFe1, Fe2-OOH, and Fe2-OOH(OH)(μ-O), respectively (bottom).

Table 2

Reaction Energies for the Reactions of Inhibited Forms of the H Cluster Shown in Figure 11: i.e., dtmaSO, dtmaSOH, IFe1, Fe2-OOH, and Fe2-OOH(OH)(μ-O) as the Starting Pointsa

reacn	dtmaSO	dtmaSOH	IFe1	Fe2-OOH	Fe2-OOH(OH)(μ-O)
1a	–18.0	–9.0	–18.3	–13.4	–19.1
2a	–29.1	–26.0	–22.5	–30.3	–37.3
1b	–29.4	–37.3	–37.5	–22.5	–27.3
2b	–17.7	2.3	dissoc	–21.3	–29.1
3a	–12.1	–24.2	–18.4	–15.5	–2.7
3bext	–95.7	–100.7	–104.1	–95.4	–86.5
1c	–34.1	–26.9	–36.2	–30.3	–34.2
4	–16.5	–15.4	–27.7	–26.1	–29.1
5a	–18.2	–26.4	–4.8	–12.3	–5.4
5bext	–121.0	–108.2	–87.5	–85.7	–85.6

Structures were optimized with BP86/RI/TZVP/COSMO (ε = 4) for the total charge of 3–/4– elementary charges, as indicated in Figure 11. Protonation energies were calculated by assuming, for a solvated proton, an energy of −262.4 kcal/mol.[39,40]

Reactions for the formation of ROS at the partially destroyed H cluster, where “[cluster]” represents either the structural unit of the reaction products n class="Chemical">dtmaSO and dtmaSOH presented in Figure 7c,d (top) or the structural unit of reaction products IFe1, Fe2-OOH, and Fe2-OOH(OH)(μ-O) of Figure 8a and Figure 6b,c (bottom): Reaction energies (kcal/mol) for the top part corresponding to the starting structure dtmaSO are given in Roman type, while for the reaction cycle starting from the dtmaSOH structure, the reaction energies are given in boldface type. Reaction energies (kcal/mol) for the bottom part corresponding to the starting structure IFe1 are given in Roman type, while for the reaction cycle starting from the Fe2-OOH structure, the reaction energies are given in boldface type. Reaction energies of Fe2-OOH(OH)(μ-O)-based clusters are given in italics. Protonation energies were calculated by assuming, for a solvated proton, an energy of −262.4 kcal/mol.[39,40] Given charges correspond to the active oxidized Hoxcat and the active reduced Hredcat forms of the modified H cluster. All clusters are considered in their low-spin state. Reaction energies are calculated for fully optimized structures (BP86/RI/TZVP/COSMO) with respect to 3O2, 2O2−, 2OOH•, 1OOH−, or 1H2O2 and dtmaSO and dtmaSOH, respectively (top), and with respect to 3O2, 2O2−, 2OOH•, 1OOH−, or 1H2O2 and IFe1, Fe2-OOH, and Fe2-OOH(OH)(μ-O), respectively (bottom).

ROS can still be formed at the H cluster even when there are changes in structure of the 2[n class="Chemical">Fe]H subsite. In the worst case one path of the small cycle (O2– release) is energetically not likely to be feasible. For instance, for IFe1-derived clusters coordination of molecular oxygen must take place before one-electron reduction, because O2 does not coordinate to the reduced form of IFe1 due to changes in the position of ligands at the distal iron atom (in the reduced 4– form of IFe1, the CO ligand moves from a bridging to a terminal position and blocks the free coordination site on Fed; O2 does not bind in a bridging position between Fed and Fep). Note, however, that due to interactions with, for example, a conserved lysine residue in the protein environment the ligands are not likely to move so that the catalytic activity of the H cluster should be preserved when the whole protein is considered.

Reactions of O2 at the 2[n class="Chemical">Fe]H subsite of a partially destroyed H cluster can support further destruction of the cubane structure. For instance, for Fe2-OOH(OH)(μ-O) a coordination of O2 to the Fed atom (−19.1 kcal/mol) and subsequent one-electron reduction (−37.3 kcal/mol) triggers the breaking of one of the bonds between iron and sulfur atoms of the cubane, leaving an iron atom considerably more outside of the cubane structure (see Figure 12).

Figure 12

Structures of reaction products of Fe2-OOH(OH)(μ-O)-based ROS formation reaction cycle, optimized with BP86/RI/TZVP/COSMO: (a) Fe2-OOH(OH)(μ-O) (b) O2-Fe2-OOH(OH)(μ-O) with O2 coordinated to the distal iron atom Fed of Fe2-OOH(OH)(μ-O); (c, d) two alternative structures of reduced O2-Fe2-OOH(OH)(μ-O) (4– elementary charges) optimized for two different spin coupling schemes of the iron atoms.

We should, however, note that for products of OOH-driven degradation the energetics of n class="Chemical">ROS generation is highly affected by the coupling scheme of the spins at the iron atoms of the cubane (see details given in the Supporting Information).

General Implications for Oxidation of Iron–Sulfur Clusters in Proteins

The reaction cycle for the formation of ROS described here agrees well with experimental results.[29,22] On the basis of spectn class="Chemical">roscopic measurements performed on [FeFe] hydrogenase isolated from C. reinhardtii, they proposed that enzyme deactivation takes place by the initial reversible formation of an O2 adduct followed by production of ROS and irreversible modification of the [Fe4S4] cubane. Our work provides mechanistic insight into the formation of ROS at the H cluster and discusses possible reaction events that lead to the irreversible changes at the iron–sulfur cubane. We found that OOH• and H2O2 are the most probable damaging agents and identified their preferable coordination sites. This part of the results is summarized in Figure 13. The picture of dioxygen-induced inhibition of the H cluster emerging from considerations presented here is complementary to the aforementioned earlier work by us,(23) where the decomposition of the first ligand sphere of the Fed atom was proposed. In both studies we investigated double protonation of oxygen adduct 2 (see Figure 2). However, in the first paper protonation of the distal oxygen atom and water release is considered with Fed=O being a stable intermediate. In the present work, we focus on the possibility that a second proton is transferred to the oxygen atom proximal to the Fed atom and subsequent release of H2O2 takes place. Interestingly, the same alternative Fe=O versus H2O2 formation was studied for oxidation of the active site of cytochrome P450 by Thiel and co-workers.(46) Their QM/MM study showed that preferred choice of the accepting oxygen atom for the transfer of a second proton depends on the amino acids surrounding the active site. Therefore, it is the logical next step to investigate the influence of the protein environment of the H cluster on the two alternative reaction pathways proposed by our group. These investigations can now be carried out after having established a clear picture of the principles of oxygen inhibition at the generic H cluster.

Figure 13

Overview of possible centers of ROS attack. Arrows indicate ROS binding sites. Their thickness qualitatively correlates with the energy released.

It is also instructive to compare the degradation pathway proposed here with studies on the decomposition of [Fe4S4] cubanes upon n class="Chemical">oxygen exposure in other enzymes such as aconitase or ferredoxins.[47−49] The main difference between the aconitase iron–sulfur cluster and the [Fe4S4] cubane in [FeFe] hydrogenase is in the number of coordinating cysteine residues. For aconitase, one of the iron atoms of the cubane plays a role at the active center and is not coordinated by cysteine. Our calculation shows that coordination of dioxygen to a free iron atom of the [Fe4S4] cubane would be energetically favored by more than −15 kcal/mol. Therefore, the decomposition mechanism of the [Fe4S4] cubane in aconitase follows most probably a different path, where oxygen can be reduced at the cubane. This is in line with the fact that the final product of oxidation of the [Fe4S4] cubane in aconitase is a [Fe3S4] cuboidal structure where one Fe atom is removed, most probably in a Fenton reaction (see ref (50) and references therein), while for [FeFe] hydrogenase the final product has not yet been experimentally identified. Another important difference is that aconitase interaction with dioxygen is subsite specific and reversible through [Fe3S4]/[Fe4S4] cluster interconversion,(50) whereas for [FeFe] hydrogenases dioxygen-induced inhibition is irreversible.

Because of greater structural similarities between hydrogenase and n class="Chemical">ferredoxins (the iron–sulfur cubane is coordinated by four cysteine ligands in both proteins), the oxygen sensitivity of ferredoxins is easier to compare. No details are known about a possible decomposition pathway, but it is assumed by analogy to aconitase that the degradation of the [Fe4S4] cluster should proceed through a [Fe3S4] intermediate. However, no spectroscopic evidence for formation of [Fe3S4] was found.(51)

A very recently studied example of an O2-sensitive protein containing a [n class="Chemical">Fe4S4] cluster is the transcriptional regulator FNR.[48,49,52] A crystal structure of this protein has not yet been solved, and the discussion in the literature is based on a homology model. The distance between Cys122 and the cubane is proposed to be longer than for three other ligating residues. However, in the work of Jervis et al.(49) the site of Cys23 at the cubane is considered to be the primary site of dioxygen attack. Jervis et al. show that shielding an iron–sulfur bond between the Cys23 and an iron atom from the cubane can slow down destructive O2 interaction. ROS attack at an iron–sulfur bond in one of the cysteine residues ligating the cubane is in line with the findings of this study.

Conclusions

In this work we have studied the O2-induced deactivation of the H cluster by uncovering the formation of n class="Chemical">reactive oxygen species at this active site and subsequent attack of the cubane structure. The most important findings of this study are summarized as follows.

Triplet dioxygen does not bind to the cubane subsite of the H cluster. We find, however, that catalytic formation of n class="Chemical">ROS (O2– and OOH–) at Fed is feasible—even after parts of the H cluster have already been degraded—and consider it an intrinsic pathological property of the H cluster. At the same time these results explain the protection of the enzyme by the reversible inhibitor CO.(29)

For the inactive oxidized form Hoxinact the formation of ROS is energetically not n class="Chemical">feasible. This explains the experimentally observed O2 tolerance of the cluster in this state.(2)

O2– and OOH– are not likely to directly act as damaging agents, but we find that OOH• and H2O2 can be formed in close proximity to the [Fe4S4] cubane by protonation of O2– and OOH–, respectively. It is shown that OOH• and H2O2 are able to oxidize the iron–sulfur cubane, most likely leading to its experimentally observed(29) decomposition.

We identify the primary site of H2O2 attack to be the n class="Chemical">cysteine ligands coordinating the [Fe4S4] unit, which is in contrast to OOH•, which shows preferable binding directly to the iron atoms of the H cluster (see Figure 13). Therefore, it should be possible to detect sulfoxides in the oxygen-inhibited form of [FeFe] hydrogenase when H2O2 is the oxidizing agent. Oxidation of sulfur atoms of the dtma ligand is also energetically feasible; however, it does not lead to any major changes in the structure of the H cluster.

For the sake of convenience, we collect the reaction energies obtained for the attack of up to three OOH• and n class="Chemical">H2O2 molecules, respectively, to a selected structure of the H cluster in Table 3. Despite many years of intense research on the decomposition of iron–sulfur cubanes upon dioxygen exposure, no detailed mechanism for the oxidation has yet been established. We find pronounced structural changes in the [Fe4S4] cubane induced by ROS. The results obtained here for the H cluster might not only apply to hydrogenases but could be generalized to other enzymes harboring iron–sulfur cubanes such as aconitase and ferredoxins. In order to investigate such a generalization, work on these enzymes has been initiated in our laboratory.

Table 3

Collection of BP86/RI/TZVP Reaction Energies (in kcal/mol) for OOH• Coordinated to an Iron Atom in Fe2-OOH (Middle Column) and for H2O2 Reacted with a Sulfur Atom in IFe2 (Right Column), Respectivelya

reacn	OOH•	H2O2
A	–20.2	–42.4
B	–51.5	–59.8
C	–21.5	–66.8

Reaction A implies addition of the first ROS molecule, reaction B considers addition of a second molecule, and reaction C finally considers addition of a third molecule. Note that the coordination of the first two OOH• radicals is to Fe2, while the third one is to Fe4—in accordance with the discussion in the text.

This work on the isolated H cluster forms the basis for understanding the influence of the protein structure on cluster reactivity toward oxygen inhibition which we will study in our laboratory for a large QM model of the active site that allows us to consider amino acid exchanges in the vicinity of the H cluster.

Table 4

Comparison of Reaction Energies (kcal/mol) Obtained for Reactions 1–3a

	reacn
	1	2	3
BP86/def-TZVP	40.6	–126.3	–98.3
B3LYP/def-TZVP	42.2	–98.6	–75.6
B2PLYP/def-TZVP	42.3	–130.6	–76.6
TPSS/def-TZVP	37.4	–122.8	–94.0
MP2/aug-cc-pVDZ	44.7	–145.9	–70.8
SCS-MP2/aug-cc-pVDZ	44.0	–124.6
CCSD/aug-cc-pVDZ	46.0	–83.8	–61.2
CCSD(T)/aug-cc-pVDZ	44.1	–109.0	–74.4
MP2/aug-cc-pVTZ	42.2	–147.5	–78.6
SCS-MP2/aug-cc-pVTZ	41.5	–125.6
CCSD/aug-cc-pVTZ	44.1	–82.9	–64.5
CCSD(T)/aug-cc-pVTZ	42.1	–109.3	–76.6

The energies were obtained for single-point calculations on BP86/RI/TZVP-optimized structures. As correlation methods such as MP2 and CCSD(T) may strongly depend on the basis set, we report results for aug-cc-pVDZ and aug-cc-pVTZ basis sets.