Solvent Tuning of Electrochemical Potentials in the Active Sites of HiPIP Versus Ferredoxin

See allHide authors and affiliations

Science  30 Nov 2007:
Vol. 318, Issue 5855, pp. 1464-1468
DOI: 10.1126/science.1147753


A persistent puzzle in the field of biological electron transfer is the conserved iron-sulfur cluster motif in both high potential iron-sulfur protein (HiPIP) and ferredoxin (Fd) active sites. Despite this structural similarity, HiPIPs react oxidatively at physiological potentials, whereas Fds are reduced. Sulfur K-edge x-ray absorption spectroscopy uncovers the substantial influence of hydration on this variation in reactivity. Fe-S covalency is much lower in natively hydrated Fd active sites than in HiPIPs but increases upon water removal; similarly, HiPIP covalency decreases when unfolding exposes an otherwise hydrophobically shielded active site to water. Studies on model compounds and accompanying density functional theory calculations support a correlation of Fe-S covalency with ease of oxidation and therefore suggest that hydration accounts for most of the difference between Fd and HiPIP reduction potentials.

Proteins containing Fe4S4 clusters catalyze one-electron transfer processes and are ubiquitous in nature (14). These proteins have evolved into two classes that have large differences in their electrochemical potentials: high potential iron-sulfur proteins (HiPIPs) and bacterial ferredoxins (Fds) (58). Physiological conditions support a reduction potential window ranging from about –600 to +500 mV (9). Spectroscopic and electrochemical measurements have shown that the resting state in both classes of protein is [Fe4S4]2+ and that HiPIPs react by oxidation to [Fe4S4]3+ (with an associated oxidation potential ranging from –150 mV to –350 mV) whereas Fds are reduced to [Fe4S4]+ (with an associated reduction potential ranging from –250 mV to –400 mV) (1014). The corresponding Fd oxidation and HiPIP reduction potentials lie outside the physiological window, and so the associated processes are not biologically active.

The geometric and spectroscopic features of the [Fe4S4]2+ resting states are identical in both protein classes (1518), and no significant differences in their electronic structures emerged from quantum mechanics (QM) (19) and molecular mechanics (MM) simulations (20, 21). Thus, the difference in electrochemical behavior appears to arise from environmental factors at the respective active sites. Crystallographic data indicate that the Fe4S4 cluster is buried in a hydrophobic core in HiPIP with ∼five conserved H-bonding interactions to the cluster ligands from the backbone, whereas Fds are exposed to solvent, with eight protein H-bonding interactions, and have a highly conserved CXXCXXC loop (where C is Cys and X is any amino acid) (18). Structurally congruent synthetic inorganic complexes have been reported for the three oxidation states observed in HiPIPs and Fds (3, 22, 23). Whereas the oxidation of synthetic [Fe4S4]2+ complexes has a similar potential to those observed in HiPIPs, the reduction of the [Fe4S4]2+ model complex has ∼1 V more negative potential than that of the Fds (24). Holm and co-workers reported significant variations in the [Fe4S4]2+ reduction potential for these complexes with varying solvent dielectric, nature of the thiolate, and presence of charged residue near the cluster; however, the reduction potentials of these models were still very negative and mostly below the physiological range (25). Electrostatic and MM calculations suggested that large dipolar interactions from peptide residues and intercalated water molecules in the Fd environment shift the [Fe4S42+ reduction potential by 1.4 V above that in HiPIP while at the same time disfavoring oxidation by shifting that potential 1.5 V below the corresponding HiPIP couple (21, 26). However, so far there are no experimental data that determine the relative contributions of these effects on the electrochemical potentials of these proteins.

Direct determination of electrochemical potentials in perturbed (lyophilized, unfolded, etc.) protein active sites using voltametric techniques is challenging at best. Our previous results have shown that, in a series of model complexes, the electrochemical potential varies linearly with the extent of Fe-S covalency in the cluster, which can be directly measured experimentally with use of sulfur K-edge x-ray absorption spectroscopy (XAS). In this study, we systematically varied the solvent exposure to the active sites of these proteins in an effort to quantify the impact of hydration of these active sites on their electronic structure and electrochemical potential.

With use of S K-edge XAS, we previously observed a large decrease in Fe-S bond covalency in Fd relative to HiPIP, although the origin of this difference was not determined at that time (27). Sulfur K-edge XAS is a direct experimental probe of ligand-metal bond covalency (28). The primary feature at the S K edge is the S 1s → 4p transition. The lower energy transitions from ligand 1s orbitals to the unoccupied metal-based 3d antibonding orbitals are disallowed, but they obtain absorption intensity if these 3d orbitals mix with S3p orbitals because the sulfur-based 1s → 3p (ΔL = 1, where L represents the orbital angular momentum) transition is electric-dipole allowed (29, 30). This pre-edge intensity is directly proportional to the percent of S3p mixing in the Fe3d orbitals and hence provides an experimental measure of ligand-metal bond covalency (31). This method was previously used to analyze the electronic structure and bonding of models and active sites of mononuclear Fe-S (rubredoxins), binuclear Fe2S2 (spinach ferredoxin and Rieske protein), trinuclear (bacterial ferredoxin), and tetranuclear (HiPIPs and ferredoxins) proteins (27, 28, 30, 32). In most cases, the Fe-S bonds in the protein active sites were found to be less covalent than those in the structurally congruent corresponding model complexes (30, 33, 34).

The S K-edge spectrum of (NEt4)2[Fe4S4(SEt)4] (Fig. 1A), where Et indicates an ethyl group, serves as an appropriate reference for the Fe4S4 clusters in proteins because an aliphatic ethyl thiolate ligand has a bonding interaction with the cluster similar to that of the cysteine ligands in the proteins. Its absorption spectrum (Fig. 1A black) shows a pre-edge peak at 2470.2 eV. Previous S K-edge data on clusters with S2– substituted by Se2– and RS (where R is an alkyl or an aryl substituent) substituted by Cl have shown that the four Ssulfide and the four Sthiolate pre-edge transitions are centered at 2470.0 eV and 2471.0 eV, respectively (30). From fits to the black spectrum in Fig. 1A and its second derivative, the peak intensity reflects 168% thiolate and 480% sulfide 3p characters mixed into the unoccupied valence orbitals of the Fe4S4 cluster (28). In the following text, the total intensity of this peak, i.e., the total covalency of this model complex, is used as a reference for the decreases observed in the proteins. The additional rising-edge feature at 2472 to 2473 eV represents the background S1s → C-Sσ* transitions for all cysteine and methionine residues in the model or protein (thus, these vary between different models and proteins).

Fig. 1.

(A) S K-edge x-ray absorption spectra of (NEt4)2[Fe4S4(SEt)4] (black), HiPIP from C. vinosum (blue), HiPIP from E. halophila I (light blue), Fd from Bt (red), and Fd from Pf (orange). (B) Pre-edge region of the absorption spectra for the model (black), WT C. vinosum HiPIP (blue), unfolded C. vinosum HiPIP (dashed brown), WT Bt Fd (red), and lyophilized Bt Fd (dashed green). (C) X-ray absorption spectra of (NBu4)2[Fe4S4(SPh)4] as a solid (black) and in acetonitrile (green), N,N-dimethylformamide (blue), and N-methylformamide (red) solutions.

The XAS data for representative HiPIPs from Chromatium vinosum (Fig. 1A, blue) and Etothiorhodospira halophila I (Fig. 1A, light blue) show pre-edge intensities very similar to each other and only slightly lower than that of the model complex. The fits to the HiPIP data indicate that the total Fe-S covalencies are 95% to 96% of that of the model complex (Table 1). In contrast, the S K-edge data for representative Fds from Bacillus thermoproteolyticus (Bt) (Fig. 1A, red) and Prococcus furiosus (Pf) (35) (Fig. 1A, orange) have significantly less pre-edge intensity than the model complex (Fig. 1A, black) and the HiPIPs (Fig. 1A, blue and light blue). These data indicate that the Fe-S bonds in Fds are less covalent than those of the model complex and of the HiPIPs. The fitted intensities for the Fds give total Fe-S covalencies of 86% to 85% of that of the model complex. Both the μ3Ssulfide and the Sthiolate covalencies are significantly reduced in the Fd active sites relative to those of the HiPIP active sites (Table 1).

Table 1.

Fe-S covalency in Fe4S4 and proteins and a model from S K-edge XAS. Covalencies were summed over 1800% total Fe3d-based orbitals, where each of the 18 unoccupied orbitals is worth 100%.

View this table:

The weakening of Fe-S bonds in Fds relative to HiPIPs could stem from differences in the H bonding from the peptide backbone (five in HiPIP versus eight in Fd) as well as differences in solvent exposure (HiPIP is buried in a hydrophobic core, and Fd is solvent-exposed) (10). To gauge the effect of hydration in Fe-S covalency, we subjected a lyophilized sample of Bt Fd to XAS. The data (Fig. 1B, dashed green) show a marked increase in pre-edge intensity relative to the pre-edge of Bt Fd in aqueous solution (Fig. 1B, red), although the rising edge remains relatively unaffected (for details, see fig. S1A). The reduction of the pre-edge intensity in buffered aqueous solution is reversible (fig. S1A) and indicates that the Fd cluster forms H bonds directly with water molecules, thereby reducing its covalency. Quantification indicates that the total covalency increases from 86% in Bt Fd (Table 1, row 5) to 95% of that of the model complex upon lyophilization (Table 1, row 6).

In a complementary experiment, the C. vinosum HiPIP was treated anaerobically with 4 M guanidinium hydrochloride, which has been shown to partially and reversibly unfold this HiPIP, to expose the cluster to water without further chemical modification (36). The XAS data for unfolded HiPIP (Fig. 1B, dashed brown) show a decrease in the pre-edge intensity relative to that of folded C. vinosum HiPIP (Fig. 1B, blue). This process is reversible (fig. S1B), and the similar rising edges between these spectra (Fig. 1B, blue line and brown dashed line) indicate that the cluster in HiPIP remains intact during this conversion. Fits to the data show that the total Fe-S covalency decreases from 96% of the model in wild-type (WT) C. vinosum HiPIP to 89% in the unfolded protein and that the Fe-μ3Ssulfide and Fe-Sthiolate bond covalencies are equally affected. The result of unfolding the HiPIP protein is complementary to the lyophilization of the Bt Fd protein, with both indicating that hydration of the cluster is the major source of reduction of Fe-S bond covalency.

To gauge the relative impacts of solvent dielectric and H bonding independently on Fe-S covalency, we dissolved a well-characterized Fe4S4 model complex (NBu4)2[Fe4S4(SPh)4], where Bu is a butyl group and Ph is a phenyl group, in a variety of solvents with different dielectric constants (acetonitrile and N,N-dimethylformamide, DMF) and capacities for H bonding (N-methyl formamide, NMF). The model complex is stable in all three solvents. The S K-edge data for the solid (NBu4)2[Fe4S4(SPh)4] complex (Fig. 1C, black) are very similar to the data for the complex in acetonitrile (Fig. 1C, green) and DMF (Fig. 1C, blue). These results suggest that there is little impact on covalency from changes in dielectric constant around the cluster. However, the data for the same complex in NMF (Fig. 1C, red), which can act as an H-bond donor, show a large decrease in the pre-edge intensity. Fits to the data indicate that the total Fe-S covalency decreases by 15% in NMF (fig. S2). This decrease in covalency due to H bonding from solvent parallels the results observed on perturbing water access to the protein active sites (lyophilized Bt Fd and unfolded C. vinosum HiPIP) and confirms that H bonding of the water to the S ligands is the direct contribution to this effect.

The above results were compared to density functional theory (DFT) calculations on the model complex and protein active sites. The (NBu4)2[Fe4S4(SPh)4] complex was modeled with an [Fe4S4(SMe)4]2– (where Me indicates a methyl group) cluster (Fig. 2A) to which eight water molecules were added (Fig. 2B). The HiPIP active site was approximated with a 210-atom molecule whose coordinates were taken from a high-resolution (1.2 Å) crystal structure of the C vinosum protein from the Protein Data Bank (PDB identification code 1CKU) (Fig. 2C). A 140-atom model of the Bt Fd active site with and without seven H2O molecules was similarly simulated (PDB identification code 1IQZ) (Fig. 2D) (31).

Fig. 2.

Computational structures simulating (A) isolated model complex, (B) H-bonded model complex, (C) HiPIP active site (the NH–S distances are indicated), and (D) Fd active site (the CXXCXXC loop of the backbone is highlighted).

Upon forming H bonds to eight water molecules, the simulated [Fe4S4(SMe)4]2– compound evidenced a total Fe-S covalency reduction of 10% S3p (table S1), which is in reasonable agreement with the experimentally observed decrease of 15% S3p in an H-bonding solvent (vide supra). The calculations also indicate that both the Fe-μ3Ssulfide and the Fe-Sthiolate bond covalencies are reduced as observed in the experiment (fig. S2B). The calculation on the Bt Fd (without H2O molecules) and HiPIP active sites show 3% and 2% reduction in the Fe-S bond covalency (table S1), respectively, relative to the reference model. This result is consistent with the limited 5% and 4% decrease of Fe-S bond covalency observed in lyophilized Bt Fd and HiPIP (Table 1, rows 6 and 2). These decreases are largely localized in Fe-Sthiolate moieties (table S1), which is consistent with the larger number of H bonds donated to the Sthiolate atoms relative to the Ssulfide atoms from the backbone. When the amide groups in the simulated backbone are replaced with ester groups (i.e., the dipolar interaction of the C=O is still present, but there is no H bonding from the N-H groups), no decrease in Fe-S bond covalency is observed (table S1, row 4). Including seven H2O molecules in the solvent-exposed side of the Fd model shows a 10% decrease in Fe-S covalency [in both Fe-μ3Ssulfide and Fe-Sthiolate (table S1, row 5)] relative to the model complex, which is close to (although less than) the experimentally observed 15% decrease in Fe-S covalency in Fds relative to the model complex (Table 1, row 5).

Thus, both the experimental and theoretical results show that Fe-S covalency is a sensitive probe of electrostatic interactions present in protein active sites. This decrease of covalency due to H bonding would correspond to the solute polarization term in a total energy calculation (37, 38). In both HiPIPs and Fds, H bonding from the peptide backbone causes only a limited reduction in Fe-S bond covalency. Rather the hydration causes two-thirds of the large decrease in the Fe-S bond covalency that is observed in the solvent-exposed Fd active site relative to an alkyl thiolate model.

Relative to the model complex [in DMF scaled to the normal hydrogen electrode (NHE)], the reduction potential of Fd is raised from –1100 mV to –300 mV (black and red points in Fig. 3A). From our past studies, the electrostatic effect of water H-bonded to S ligands will decrease their covalent interaction with the Fe and weaken the Fe-S bonds. This will raise the reduction potential by destabilizing the oxidized state more than the reduced state (40). An estimate of this effect on the reduction potential of the Fe4S4 clusters is given in Fig. 3A, in which the covalencies derived from S K-edge XAS intensities for a series of [Fe4S4(SR)4]2– complexes with varying thiolates (solid black squares) are plotted relative to their corresponding [Fe4S4]2+ one-electron reduction potentials (measured in DMF scaled to NHE). This plot shows a linear correlation, and the negative slope indicates that it is easier to reduce the [Fe4S4]2+ cluster as the Fe-S covalency decreases (34). For the Fe4S4 cluster in HiPIP, the experimental covalency decreases by 5% relative to the model, which corresponds to a +150 mV shift in its [Fe4S4]2+ reduction potential (open blue diamond). Consistent with this extrapolation, the one-electron reduction of the [Fe4S4]2+ resting form of HiPIP from C. vinosum has been experimentally estimated to be –910 mV, which is outside the physiological window (39).

Fig. 3.

Plot of total Fe-S covalency as a function of [Fe4S4]2+ (A) reduction potential and (B) oxidation potential versus NHE for a series of [Fe4S4(SR)4] complexes (solid black) and Bt Fd (solid red). The open squares represent the predicted (using covalency) [Fe4S4]2+ electrochemical potentials. The [Fe4S4]2+ reduction-oxidation potential for the reference complex [Fe4S4(SEt)4]2– is indicated by the horizontal dashed gray line.

For Fd in aqueous solution, the electrostatic effect of the local H bonds on the thiolate and sulfide ligands of the Fe4S4 cluster decreases the covalency by 15% relative to the model complex. The plot in Fig. 3A gives an estimated change of the reduction potential of +450 mV associated with this Fe-S covalency change. The remaining +350 mV contribution to the observed reduction potential of Fd would result from the additional stabilization of the reduced state by the asymmetric electrostatic field around the H-bonded cluster. One face of the cluster is exposed to the high dielectric of water and the other to the dipoles in the CXXCXXC peptide loop, which is highly conserved in all the bacterial Fds (Fig. 2D, highlighted in bold, and fig. S3) (40). The CXXCXXC dipoles were found to make the dominant contribution to the protein dipolar stabilization term in the calculations of Warshel and co-workers (21, 26).

For the HiPIP [Fe4S4]2+ oxidation couple, an analogous plot of the measured total S3p character for a series of [Fe4S4(SR)4]2– complexes with different thiolate ligands versus the [Fe4S4]2+ oxidation potential (41, 42) also shows a linear correlation (Fig. 3B). The protein data points on this plot show essentially the same features as observed for the [Fe4S4]2+ reduction couple in Fig. 3A). The predicted [Fe4S4]2+ oxidation potential for HiPIP (Fig. 3B, blue open diamond at –265 mV) shows a limited shift because of the 5% decrease in covalency relative to the alkyl thiolate model couple (–200 mV). Alternatively, the [Fe4S4]2+ oxidation potential in the Fd protein active site is lowered by –700 mV relative to the model complex, into the nonphysiological –800 to –900 mV range (Fig. 3B, solid red) (41); –300 mV because of the electrostatic effect of local H bonds, which decrease the covalency; and –400 mV because of nonlocal protein electrostatics.

Thus, the same factors that shift the [Fe4S4]2+ reduction potential in the Fd active site by +∼600 mV relative to HiPIP (from –900 mV to –300 mV), bringing it into physiological range, also shift the [Fe4S4]2+ oxidation potential of Fds by ∼–500 mV relative to HiPIP (from –350 mV to –850 mV) and out of physiological range.

Our results indicate that H bonding from water in a solvent-exposed Fd site can tune the [Fe4S4]2+ reduction potential up by at least +300 mV. This phenomenon may provide a regulation mechanism for electron transfer in Fe-S cluster proteins. As examples, Fds are proposed to catalyze the reduction of formyl dehydrogenase (E° = –530 mV) by hydrogenase (E° = –414 mV) in methanogenic bacteria (43), and DNA binding is proposed to lower the [Fe4S4]2+ reduction potential by >300 mV in endonuclease III (44). As illustrated in Fig. 4, interaction with an electron transfer partner may lead to a partial desolvation of the active site that can shift the reduction or oxidation potential of the Fd [Fe4S4]2+ cluster. Indeed, the large surface area of docking (1000 Å2) and the close contact between the active sites (10 Å to 12 Å) in a nuclear magnetic resonance structure of a protein or protein complex of a Fd with its redox partner cytochrome c553 (from Desulfovibrio norvegicum) (45) are consistent with this possibility. Thus, solvent tuning may play an important role in protein-protein and protein-DNA interactions.

Fig. 4.

Schematic representation of [Fe4S4]2+ reduction potential tuning by desolvation.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article