Protein Control of Iron-Sulfur Cluster Redox Potentials”

The relationship between the three-dimensional structures of iron-sulfur proteins and the redox potentials of their iron-sulfur clusters is of fundamental importance. We report calculations of the redox potentials of the [Fe4S4(S-~ys)4]-2”S couple in four crystal- lographically characterized proteins: Azotobacter uinelandii ferredoxin I, Peptococcus aerogenes ferre- doxin, Bacillus thermoproteolyticus ferredoxin, and Chromatium uinosum high potential iron protein (HiPIP). Our calculations use the “protein dipoles Langevin dipoles” microscopic electrostatic model, which includes both protein and solvent water. The variations in calculated redox potentials are in excellent agreement with experimental data. In particular, our results confirm the important role of amide groups close to the cluster in separating the potential of C. uinosum HiPIP from those of the other three proteins. However, the potentials of these latter exhibit a substantial range despite extremely similar amide group environments of their clusters. Our results show that the potentials in these proteins are tuned in part by varying the access of solvent water to the neighborhood of the cluster. Our calculations provide the first successful quantita-tive modeling of the protein control of iron-sulfur cluster redox potentials.

The relationship between the three-dimensional structures of iron-sulfur proteins and the redox potentials of their iron-sulfur clusters is of fundamental importance. We report calculations of the redox potentials of the [Fe4S4(S-~ys)4]-2"S couple in four crystallographically characterized proteins: Azotobacter uinelandii ferredoxin I, Peptococcus aerogenes ferredoxin, Bacillus thermoproteolyticus ferredoxin, and Chromatium uinosum high potential iron protein (HiPIP). Our calculations use the "protein dipoles Langevin dipoles" microscopic electrostatic model, which includes both protein and solvent water. The variations in calculated redox potentials are in excellent agreement with experimental data. In particular, our results confirm the important role of amide groups close to the cluster in separating the potential of C. uinosum HiPIP from those of the other three proteins. However, the potentials of these latter exhibit a substantial range despite extremely similar amide group environments of their clusters. Our results show that the potentials in these proteins are tuned in part by varying the access of solvent water to the neighborhood of the cluster. Our calculations provide the first successful quantitative modeling of the protein control of iron-sulfur cluster redox potentials.
Proteins containing iron-sulfur ([Fe-SI) clusters occur ubiquitously in nature and play a major role in biological electron transport (1). Thus, the understanding of the control of the redox potential of an [Fe-SI cluster by its protein environment is of fundamental importance. Here, we address this challenge by investigating the origin of the enormous variation in the midpoint potential of the -2/-3 couple of the [Fe4S4(S-~y~)4] clusters of four small, structurally well-characterized proteins: Azotobacter uinelandii ferredoxin I (AuFdI),' Peptococcus aerogenes ferredoxin (PaFd), Bacillus thermoproteolyticus ferredoxin (BtFd), and Chromatium uinosum high potential iron protein (CuHiPIP). The electrostatic interaction of these [Fe-SI clusters with their protein and water environment is cal-culated using the "protein dipoles Langevin dipoles" (PDLD) approach (2-4). The predicted variation in the electrostatic contribution of protein and water together to the cluster redox potential is in excellent agreement with experiment. These calculations constitute the first successful modeling of the protein control of [Fe-SI cluster redox potentials.
In this work we evaluate the effect of the protein on the redox potential of the [Fe4S4(S-~y~)4] clusters in these four proteins using the PDLD approach (2-4) as implemented in the program POLARIS (17). The electrostatic interaction of a charged moiety: in this case, the [Fe4S4(S,)4]-2/-3 (S, = cysteine S) portion of the [Fe-SI cluster, with its surroundings is expressed as the sum of four terms: VQ,, VQ,, VL and VB. VQ and VQ, are the interactions of the cluster with the partial charges and induced dipoles of protein atoms, respectively. VL and VB are the interactions with solvent water, treated microscopically within a sphere of radius rL surrounding the cluster, and macroscopically beyond, respectively. Microscopic water molecules are represented by point dipoles placed on a grid and oriented self-consistently in the combined electrostatic field arising from cluster charges, protein charges and induced dipoles and all other water dipoles.
Calculations have been carried out for the -2/-3 couple of the [Fe4S4(S-~y~)4] clusters of AuFdI, BtFd and CuHiPIP and of one of the two clusters of PaFd (that ligated by cysteines 8, 11, 14, and 46); an error in the protein sequence used in the PaFd structure determination in the neighborhood of the second cluster has recently come to light (18), and we have excluded this cluster from consideration. The differential solvation energies of the -2 and -3 clusters obtained are reported in Table I (which also includes additional details of the calculations) and Fig. 1. Redox potentials obtained thence are compared to experiment in Fig. 2. In principle, the absolute redox potential can be evaluated using the potential of the cluster in aqueous solution as a reference (4). Since this is not available, we here address the trend within the different proteins and we arbitrarily set the calculated and experimental redox potentials of AuFdI to be identical. We predict an ordering of potentials: CuHiPIP << AuFdI < PaFd< BtFd, in agreement with experiment. The spread for AuFdI, PaFd, and BtFd is 447 mV compared to an experimental range of 370 mV.
The contributions of VQ,, VQ,, VL, and VB to the total differential solvation energies of BtFd, PaFd, AuFdI, and CuHiPIP are given in Table I and are diagrammed in Fig. 1  This approach has been justified and validated elsewhere (2,3,29). At the same time the effects of an ionic strength of 0.1 M also is modeled using the approach of Ref. 30. The ionization state of the ionizable protein groups were evaluated by assigning to each group its intrinsic aqueous pK, and calculating self-consistently their interactions using the above t(R). A similar trend was obtained by repeating the calculations while estimating the intrinsic pK, of groups which are not exposed to water using the PDLD method. AGQQ is relatively small and its addition does not change the trend obtained in its absence. Brookhaven coordinate file: 1FXB; B. stearothermophilus ferredoxin, from which the redox potential was obtained, contains just two Asp for Glu substitutions relative to Bt ferredoxin (7).
AuFdI than in CuHiPIP is the primary cause of the much higher potential of the -2/-3 couple of the former three proteins. The crystallographic studies of BtFd, PuFd, AuFdI, and CuHiPIP have identified hydrogen (H)-bonding by amide NH groups to the [Fe4S4(S-~y~)4] clusters. The first three proteins exhibit highly conserved cluster environments with H-bonding from 8 amide groups (5,8,18,19). In the case of CuHiPIP 5 amide groups are involved in H-bonding (10,18). Elimination of the charges of these amide groups leads to the results given in Table I. The total differential solvation energies are substantially reduced in BtFd, PuFd, and AuFdI; in contrast, in CuHiPIP very little change occurs. These results demonstrate the importance of both the presence and the orientation of the amide groups in the immediate environment of the [Fe4S4(S-~y~)4] clusters of BtFd, PuFd, and AuFdI in elevating their redox potentials above that of CuHiPIP. The difference in cluster-amide H-bonding between CuHiPIP and PuFd was recognized early as a likely major contributor to their very different redox properties (9,15,19,20). Our results confirm, refine, and quantitate this expectation.
Although the contribution of the local amide groups accounts for the difference between CuHiPIP and PuFd, it does not explain the variation in redox potential within the BtFd, PuFd, and AuFdI group. This variation originates in changes in the contributions of VQ,, VQ,, and VL which are comparable are PuFd < AuFdI -BtFd, while for VL AuFdI <BtFd < PuFd. Thus, the lower potential of AuFdI compared to PuFd is attributable to a substantially smaller VL contribution, which outweighs the greater VQ, and VQ, contribution. In contrast, the lower potential of PuFd relative to BtFd is attributable to the substantially lower VQ, and VQ, contribution, which outweighs the greater VL contribution. The interrelationship of the variations in the protein structure and in VQ,, VQ,, and VL contributions to the redox potential if complex. Calculations in which protein groups more than 9 A from the cluster center are discarded exhibit very similar variations, showing that variations in redox potential originate predominantly in changes in protein structure in the neighborhood of the cluster. Of particularly significant interest is the fact that the water contribution VL is the dominant factor in causing the difference in redox potential between AuFdI and PuFd. The environments of the clusters in these two proteins exhibit strong homology (5,18). Apparently, however, the extent of water penetration to the neighborhood of the clusters is significantly different, as illustrated in Fig.  3. This aspect of the calculations is being studied in more detail.
Our results provide additional support for the PDLD approach in describing electrostatic energetics in aqueous solutions of proteins. We note that in our calculations: (i) there is no change in structure with oxidation state, (ii) all residues (excepting cysteine cluster ligands) are uncharged (see, however, below), and (iii) the second cluster in PaFd and AuFdI are uncharged. The inclusion of protein reorganization is expected to improve the agreement of calculation and experiment (see, e.g., Ref. 21). The effects of including cluster interactions with ionized residues (and other clusters) are expected to be relatively small and can be conveniently estimated using a macroscopic model (18,(22)(23)(24)(25) with the results given in Table I. Although the factors governing the redox potentials of [Fe-S] clusters have been widely discussed (18), microscopic calculations of protein tuning of [Fe-S] cluster redox potentials have not been reported previously. Furthermore, the crucial role of the solvent in establishing the trend in redox potentials has not been demonstrated. In view of the agreement between our calculations and experiment, we are extending our studies to encompass other [Fe-S] clusters/redox couples (1) and to examine the results of site-specific mutations in AuFdI (12,26,27). More detailed accounts of these studies will be forthcoming.