The Quaternary Structure of the Saccharomyces cerevisiae Succinate Dehydrogenase

Succinate dehydrogenases and fumarate reductases are complex mitochondrial or bacterial respiratory chain proteins with remarkably similar structures and functions. Succinate dehydrogenase oxidizes succinate and reduces ubiquinone using a flavin adenine dinucleotide cofactor and iron-sulfur clusters to transport electrons. A model of the quaternary structure of the tetrameric Saccharomyces cerevisiae succinate dehydrogenase was constructed based on the crystal structures of the Escherichia coli succinate dehydrogenase, the E. coli fumarate reductase, and the Wolinella succinogenes fumarate reductase. One FAD and three iron-sulfur clusters were docked into the Sdh1p and Sdh2p catalytic dimer. One b-type heme and two ubiquinone or inhibitor analog molecules were docked into the Sdh3p and Sdh4p membrane dimer. The model is consistent with numerous experimental observations. The calculated free energies of inhibitor binding are in excellent agreement with the experimentally determined inhibitory constants. Functionally important residues identified by mutagenesis of the SDH3 and SDH4 genes are located near the two proposed quinone-binding sites, which are separated by the heme. The proximal quinone-binding site, located nearest the catalytic dimer, has a considerably more polar environment than the distal site. Alternative low energy conformations of the membrane subunits were explored in a molecular dynamics simulation of the dimer embedded in a phospholipid bilayer. The simulation offers insight into why Sdh4p Cys-78 may be serving as the second axial ligand for the heme instead of a histidine residue. We discuss the possible roles of heme and of the two quinone-binding sites in electron transport.

the reduction of fumarate to succinate, the reverse of the SDH reaction.
The structures of the Escherichia coli and the Wollinella succinogenes FRDs have been solved to resolutions of 2.7 and 2.2 Å, respectively (9,10). More recently, the structure of the E. coli SDH was solved to 2.6 Å resolution (11). SDH and FRD enzymes consist of a hydrophilic catalytic dimer that protrudes into the mitochondrial matrix or the bacterial cytoplasm and either one or two integral membrane subunits. Table I lists the subunit and cofactor compositions of the relevant enzymes. The catalytic dimers are comprised of two subunits that exhibit a high degree of sequence conservation across species and as expected, the structures of the catalytic subunits are very similar. The larger subunit carries a covalently bound flavin adenine dinucleotide (FAD), while the smaller subunit contains three iron-sulfur clusters. In contrast, the membrane portions of SDH and FRD enzymes show considerable variability in the primary structures of their subunits and in cofactor composition (3,7). In the E. coli SDH and FRD enzymes, two hydrophobic subunits are present. The E. coli FRD was crystallized without heme and with two bound menaquinone molecules, while the E. coli SDH was crystallized with one heme and one ubiquinone. The W. succinogenes FRD has a single membrane subunit, which contains two hemes, but the crystals did not contain any bound quinone. Although, this latter enzyme is known to oxidize quinone, the positions of any quinone-binding sites have yet to be determined. The structures of the membrane subunits of each enzyme are considerably different, although in all three x-ray structures, four antiparallel helices form a central helix bundle. It is not possible to superimpose the membrane subunits without changing the position, orientation, and tilt of the transmembrane helices (12). The variability among membrane subunits and the cofactors they harbor are major determinants of the distinct properties of SDH and FRD enzymes in different biological systems.
Mammalian SDH functions not only in mitochondrial energy generation, but also has a role in oxygen sensing and tumor suppression. Mutations in the human SDHA gene result in mitochondrial encephalopathy or optic atrophy, while mutations in the SDHB, SDHC, or SDHD genes are associated with paraganglioma, benign tumors in the head and neck (13)(14)(15). The mev-1 mutation in the Caenorhabditis elegans SDHC hom-olog results in decreased lifespan and the increased production of superoxide ions (16). The striking variability of clinical presentations associated with SDH mutations and a possible role for the enzyme in aging have greatly stimulated interest in understanding the protein structure and the molecular mechanisms that elicit disease.
The Saccharomyces cerevisiae SDH is a tetrameric (Sdh1p-Sdh4p), single-heme containing enzyme that has been extensively studied as a model for the mitochondrial enzymes (5,17). Based on the kinetics of inhibition by the quinone analog, sec-butyl-4,6-dinitrophenol (s-BDNP) and on mutagenesis studies, the yeast enzyme is proposed to harbor two non-equivalent quinone-binding sites (18,19). In addition, an unusual extension of the Sdh4p carboxyl terminus is essential for the optimal catalysis of quinone reduction and for the assembly of a stable holoenzyme (20,21).
In this report, we present a model for the quaternary structure of the S. cerevisiae SDH. FAD, iron-sulfur clusters, heme b, quinones, and the quinone analog inhibitor, s-BDNP, were docked into the yeast structure. In addition, we used molecular dynamics on the membrane dimer embedded in a hydrated phospholipid bilayer to investigate Sdh4p Cys-78, which may serve as an unusual heme axial ligand. We discuss the possible role of heme b and the involvement of two quinone-binding sites in electron transport.

EXPERIMENTAL PROCEDURES
Model Building-The structure of the SDH apoenzyme was constructed by homology modeling using the program Modeler6v2 (22)(23)(24). Templates used for modeling were the E. coli SDH, the E. coli FRD, and the W. succinogenes FRD (Protein Data Bank (PDB) entry codes 1NEK, 1LOV, and 1QLA, respectively). Templates and target sequences were aligned using ClustalW (25) and manually optimized in the Swiss-Pdb Viewer v3.7 (26 -28) by monitoring the mean force potential energy (29,30) of template-sequence alignments. Since the flavoprotein and the iron-sulfur subunits have high sequence identities (Table II), all three templates (1NEK, 1QLA, and 1LOV) were used for modeling the catalytic dimer. However, it is not possible to model the membrane subunits in this way due to the low primary sequence conservation (3,7). Of the three structures, 1NEK (11) provided the best template for modeling the yeast membrane dimer. Alignments of the membrane subunits were guided by the positions of the three transmembrane segments for each subunit (18,19), the heme ligands, and conserved residues that may interact with ubiquinone ( Fig. 1). For the Sdh3p alignment with the E. coli SdhC (Fig. 1A), an alternative alignment that introduced three additional identities was rejected because it resulted in the disruption of the second transmembrane helix. None of the three new identities involved conserved amino acids that could be identified in a multiple sequence alignment of over 60 Sdh3p related sequences (not shown). An ensemble of 10 structures, nine of which were clustered (rmsd Ͻ 0.5 Å), was generated, and the one with the lowest objective function (energy) was chosen for further analysis.
None of the three templates used for modeling provides any structural information for building the Sdh4p C terminus (residues 127-150). Secondary structure analysis suggests that residues 127-144 are slightly amphipathic with a hydrophobic moment of 5.23. We searched for compatible templates to use in modeling using the "scan loop data base" command of the Swiss-Pdb Viewer. The template with the lowest energy (GROMOS96 force field) and maximum sequence similarity was selected, modeled, and connected to the Sdh4p C terminus with the program O v8.0 (31) using Gly-126 as a flexible hinge. The model was energy-minimized (100 cycles of steepest descent) using GROMACS v3.1.1 (32,33). We guided the placement of the C-terminal segment with the results of our mutagenesis studies, which indicated that Lys-132 may interact with Asp-67 (21). The coordinates for the model (PDB entry code 1PB4) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (www.rcsb.org/). The Sdh1p, Sdh2p, Sdh3p, and Sdh4p subunits correspond to chains A-D in the model, respectively.
Docking-The locations of FAD, the three iron-sulfur clusters, and heme b were determined by superimposition with 1NEK using the program Modeler6v2. Docking of UQ and its inhibitor analog, s-BDNP, into the SDH model (the flavoprotein subunit was removed since it does not contribute to the formation of the quinone or heme binding sites) was performed with the program Autodock v3.0.5 (36 -39), using the graphical front-end AutoDockTools. The Lamarckian genetic algorithm (LGA) was used to perform an automated molecular docking (39). Parameters for the number of generations, energy evaluations, and docking runs were set to 27,000, 250,000, and 50, respectively. The docked protein-ligand complexes were evaluated on the basis of interaction energy combined with geometrical matching of the ligands with functional residues at the docking site.
Polar hydrogens and partial charges for SDH residues and cofactors were assigned in AutoDockTools using the Kollman United Atom and Gasteiger charges, respectively. Flexible torsions in the cofactors were assigned with AutoTors and all dihedral angles were allowed to rotate freely. In general, these were all acyclic, non-terminal single bonds (excluding amide bonds) in a given ligand molecule. Affinity grid fields were generated, using the auxiliary program AutoGrid. For docking into the proximal quinone-binding site (Q P ), the grid field was centered at the middle of SDH, while for docking into the distal quinone-binding site (Q D ), the grid field was centered at the lower half of the macromolecule. The positioning of the grid fields was guided by the results of mutagenesis studies of the quinone-binding sites (18,19). For UQ2 docking, the grid was a 60 Å cube with grid points separated by 0.375 Å. For s-BDNP docking, the grid was a 37.5 Å cube with grid points separated by 0.375 Å. The cofactors were initially placed at a random position away from the docking site.
Binding Free Energy Prediction-Free energies of binding were estimated using the empirical scoring function included in AutoDock (39). The scoring function included the van der Waals interaction repre-  sented as a Lennard-Jones 6 -12 dispersion/repulsion term, the hydrogen bonding represented as a directional 12-10 term, and the Coulombic electrostatic potential. The scoring function was further refined by including desolvation and entropic terms and empirically-calibrated coefficients derived from regression analysis of protein-inhibitor complexes whose structures are known at high resolutions.
Model Evaluation and Validation-The Modeler energy function was used to assess the degree of violation of the template-derived restraints. Stereochemical evaluations were performed with the program PRO-CHECK v3.4.4 (40,41). Quality evaluation of the model was done by the program ProsaII (29), which relies on statistical potentials involving single residues and pairs of residues (mean force potentials). When evaluation tools indicated problems with the model, it was recalculated after the manual repositioning of side chains or the editing of a region of the alignment.
Molecular Dynamics-Molecular dynamics simulation was performed in GROMACS v3.1.1 (32, 33) with the GROMOS87 force field. The flavoprotein subunit was not used in the simulation since it does not directly contact the Sdh3p and Sdh4p subunits. The polypeptides were solvated in a rectangular box filled with dipalmitoylphosphatidylcholine (DPPC) and simple point charge (SPC) waters. The starting model for the lipid bilayer is a pre-equilibrated box of 128 solvated DPPC molecules (dppc128; Ref. 42). Briefly, an empty rectangular box (130 Å ϫ 125 Å ϫ 68 Å) with the protein positioned in the center was generated using the "editconf " command of GROMACS. The box was then filled with dppc128, using the "genbox" command. The resulting simulation box contained 484 DPPC molecules (about 242 molecules in each leaflet) solvated on both sides with SPC water. The total system contained ϳ80,000 atoms. Periodic boundary conditions were employed to avoid edge effects.
After solvation, the simulation system was energy-minimized with 300 iterations of the steepest descent algorithm to remove bad van der Waals contacts. These configurations served as the starting structures for 500 ps position-restraint simulations in which the protein was harmonically restrained with an isotropic force constant of 1000 kJ⅐mol Ϫ1 ⅐nm Ϫ2 . This allows further equilibration of the bilayer and solvent, while keeping the conformation of the protein unchanged. The bilayer, solvent, and peptides were separately coupled to the temperature and pressure baths. Constant temperature and pressure were maintained at 300 K and 100 kPa using the weak coupling technique (43) with coupling and relaxation parameters of 0.1 and 1.0 ps, respectively. The van der Waals interactions were modeled using a 6 -12 Lennard-Jones potential, cutoff at 1.0 nm, with the first and second neighbor exclusions and scaled third neighbor Lennard-Jones coefficients. Electrostatic interactions were included via the particle mesh Ewald summation (44,45) with a cut-off of 0.9 nm for the real space calculation and a fast Fourier transform algorithm for the reciprocal space calculation. All covalent bond lengths were constrained using the LINCS algorithm to allow a time step of 2 fs.
Molecular dynamics simulations were carried out on a cluster of 6 processors running Red Hat Linux 7.2 patched with MOSIX v1.9.0 (46). Simulations lasted for 5 ns, with the first nanosecond taken for equilibration. Trajectories were taken every 50 ps and visualized using VMD v1.7.2 (47). Fig. 2A depicts the structural model of the S. cerevisiae SDH constructed by Modeler, which generates structures by the satisfaction of spatial restraints (22)(23)(24). It extracts distance and dihedral angle restraints on the target sequence from its alignment with the template structures. These template-derived restraints, in combination with the CHARMM force field, were used to obtain a full objective function. Optimization of the target function with conjugate gradient and simulated annealing via molecular dynamics leads to a model that best satisfies all the spatial restraints.

A Model of the S. cerevisiae SDH-
In the model, the structures of the Sdh1p and Sdh2p subunits closely resemble the structures of the templates (Fig. 2B). The model of the catalytic dimer is an average structure based on restraints derived from the three coordinate sets, E. coli SDH (11), E. coli FRD (10), and W. succinogenes FRD (9). Several attempts were made to model the anchor subunits using the three available coordinate sets as templates. However, this approach failed due to the differences in the position, orientation, tilt, and number (in the case of W. succinogenes FRD) of transmembrane helices. The crystal structure of the E. coli SDH shows that SdhC Arg-31 is within 4 Å of UQ, while SdhD Tyr-83 is hydrogen bonded to the UQ. Interestingly, these residues are conserved in the S. cerevisiae SDH subunits (Fig. 1, A and B). Using these residues, the predicted heme ligands, and the observed transmembrane segments as guides, we generated sequence alignments of the membrane subunits ( Fig. 1 and Refs. 18 and 19). The result is a model quality that is comparable to the E. coli SDH (root mean square deviation of 0.62 Å (Fig. 2B) for 4040 backbone atoms as determined with the program Swiss-Pdb viewer). The stereochemistry of the model is as good as those of the starting templates (Table III).
Docking Calculations-The FAD cofactor and the three ironsulfur clusters could be docked by superimposing the catalytic dimer of the E. coli SDH onto that of the S. cerevisiae SDH since cofactor locations are highly conserved among the catalytic dimers from the three crystal structures (Fig. 2, C and D). Similarly, heme b could be docked into the S. cerevisiae SDH by superimposition (Fig. 3A). The locations of quinone and its   inhibitor analog, s-BDNP, were determined with the program AutoDock v3.0.5 (Fig. 3, A and D). The docking program AutoDock produces docking simulations in excellent agreement with crystallographically determined structures of ligand-protein complexes (36,39). To validate its use in the current study, heme b was docked into the crystal structure of the E. coli SDH (code 1NEK), after removing all cofactors from the coordinate file. Fig. 3B shows the docked heme superimposed upon the heme in the crystal structure; the positions of the two hemes are in close agreement.
In our docking calculations for the S. cerevisiae SDH, each experiment consisted of 50 docking runs, with each run producing a single docked structure. The ensemble of 50 structures was first sorted geometrically in terms of their structural similarities. A set of structures having a root mean square deviation in atomic positions of less than 0.5 Å was designated a cluster (Table IV). A small number of clusters indicate strong and specific binding, while a large number of clusters indicates a weaker or lower specificity of binding as the solutions sort into many different binding conformations or orientations. The preferred cofactor structure was identified as the conformation with the lowest energy and was chosen for further analysis.
FAD and the Iron-Sulfur Clusters-As predicted from sequence analysis, the environments of FAD and the three ironsulfur clusters in the S. cerevisiae SDH model (Fig. 2, B and C) are similar to those observed for the E. coli SDH (11), the E. coli FRD (10), and the W. succinogenes FRD (9). In the three crystal structures, FAD is associated with the flavoprotein subunit through a covalent bond between the flavin C8␣-methyl group and the N⑀2 of a histidine residue. In the S. cerevisiae SDH model, Sdh1p His-62 is in proximity to the flavin C8␣-methyl group. FRDs from E. coli and W. succinogenes and SDH from E. coli contain three iron-sulfur clusters, which are coordinated by cysteine residues in the FrdB or the SdhB subunits. The only exception is the E. coli SDH [2Fe-2S] center where Asp-63 residue is a ligand. In the S. cerevisiae SDH, the following cysteine residues are conserved: Heme Binding- Fig. 3, A and C show the docked heme b in the S. cerevisiae SDH. In this model, the N⑀2 atom of Sdh3p His-106 and the S␥ atom of Sdh4p Cys-78 are correctly oriented to form coordinating bonds with the central iron atom of the heme. The distance between the iron atom and the N⑀2 atom is 2.21 Å, which is within the range of coordination distances observed in the E. coli SDH (11) and W. succinogenes FRD crystal structures (9). However, the S␥ atom of Cys-78, the putative second heme axial ligand in the yeast SDH, is 2.96 Å away from the heme Fe; longer than S␥-Fe bond lengths observed in other crystal structures. This is not surprising since the model is based on restraints derived from the E. coli SDH crystal structure where SdhD His-71 is the ligand. Molecular dynamics simulations were employed to sample alternative low energy conformations for the Cys-78 containing model (see below).
Quinone Binding Sites-UQ can be docked into two spatially separated sites with an edge-to-edge distance of 25.4 Å (Fig. 3,  A and D), similar to the observed sites in the E. coli FRD. The   FIG. 4. Quinone-binding environments. A, polar amino acid residues at the Q P site. B, non-polar amino acid residues at the Q P site. C, amino acid residues at the Q D site.  free energies of binding (⌬G BIND ) at the Q P and Q D sites are Ϫ99.7 and Ϫ11.3 kcal mol Ϫ1 , respectively (Table V). Amino acid residues near each of the two quinone-binding sites are shown in Fig. 4, A-C.
The 2-substituted derivatives of 4,6-dinitrophenol are known quinone analog inhibitors of the respiratory chain (48,49). s-BDNP inhibits the S. cerevisiae SDH with non-linear, noncompetitive kinetics (18,19). The s-BDNP inhibition kinetics, coupled with the results of our mutagenesis studies, have led us to propose a two quinone-binding site model for the yeast enzyme. Two molecules of s-BDNP were docked into the Q P and Q D sites (Fig. 3C). The ⌬G BIND values for the two inhibitor binding sites differ by approximately one order of magnitude (Table V), correlating well with the previously determined s-BDNP inhibition constants (18,19).
Molecular Dynamics-Although the membrane subunits of the three template structures share a similar fold (a bundle of four anti-parallel helices), the positions, orientations, and tilts of individual helices can vary considerably. We used molecular dynamics simulations to discover whether lower energy conformations of the membrane dimer exist for the yeast enzyme. Simulations lasting for 5 ns were performed with the membrane dimer embedded in a periodic box filled with a solvated pure lipid bilayer (Fig. 5). The trajectories from the last 500 ps of simulation showed the least structural fluctuation when superimposed upon the starting structure (Fig. 6). This was taken to represent the equilibrium ensemble, and the conformations within this ensemble were averaged. Fig. 7A shows the structure of this averaged ensemble generated by the MD sim-ulation superimposed on the starting structure. The rmsd is 0.21 Å involving 2176 atoms of the Sdh2p, Sdh3p, and Sdh4p subunits. The most mobile sequences are transmembrane helix II and the C terminus of Sdh4p. The mobility of the helix brings the S␥ atom of Sdh4p Cys-78 to within 2.42 Å of the iron atom of heme b (Fig. 7B). The Cys S␥-heme iron bond lengths in the diheme cytochrome c from the Rhodovulum sulfidophilum SOXAX complex (PDB code 1H31) and the Bacillus megaterium cytochrome P450 (PDB code 1BU7) are 2.46 Å and 2.38 Å, respectively (50,51). DISCUSSION Our understanding of the structure and function of SDH and FRD enzymes has increased tremendously since the publication of the E. coli SDH, the E. coli FRD, and the W. succinogenes FRD crystal structures (9 -11). However, the variability in heme and quinone content leaves questions regarding the pathway of electron transfer through the membrane dimer (3). There is also considerable interest in how SDH function is linked with disease. The structure and function of the membrane dimer is particularly interesting in light of the newly discovered roles for SDH as a generator of superoxide ions and as a tumor suppressor (13,16,52,53).
Mutagenesis and biochemical studies in the E. coli FRD and the S. cerevisiae SDH have established the importance of the Q D site, though its exact role in the reaction mechanism has yet to be established (5,54,55). The computational approaches followed in this study have yielded considerable insight into the structure of the quinone and heme binding sites of a mitochondrial SDH.
In our previous reports, we proposed that the S. cerevisiae SDH has two quinone-binding sites (5) . Fig. 4, A and B show the residues in the vicinity of the Q P site. Sdh4p Tyr-89 and Sdh2p Trp-174 are in direct contact with O1 atom of UQ. These residues are equivalent to SdhD Tyr-83 and SdhB Trp-164 of the E. coli SDH. In our model, we observed that the O␥ atom of Sdh3p Ser-44 is within hydrogen bonding distance of the O4 atom of the UQ ring and may provide additional interactions with the quinone bound at the Q P site. Interestingly, there is no corresponding polar side chain in the proximity of the UQ O4 carbonyl oxygen in the E. coli SDH; in that enzyme, SdhC Ile-28 corresponds to the Sdh3p Ser-44. The additional interaction provided by the Ser-44 side chain may indicate that the quinone is more tightly bound in the yeast SDH than it is in the E. coli enzyme. As most clearly revealed by the C. elegans mev-1 mutation (16), SDH can be a generator of superoxide radicals. Ser-44 may be a eukaryotic adaptation to ensure that the S. cerevisiae SDH produces fewer free radicals. We are currently exploring this idea through mutagenesis studies. Recently, we showed that the S. cerevisiae SDH generates superoxide when residues near the Q P site are mutated suggesting that the free radicals may originate from impaired quinone binding or reduction (52).
Sdh3p His-113 and Trp-116 are residues proposed to be in the vicinity of the Q P site, since mutation of these residues significantly impairs quinone reductase activities (18). However, the locations of these two residues in our model suggest that they do not directly interact with quinone at the Q P site but rather that they are necessary to properly position residues that do.
An additional quinone molecule can be docked into the membrane dimer model; this corresponds to the Q D site (Fig. 3, A  and D). In the E. coli SDH crystal structure, this Q D site is filled with a cardiolipin molecule (11). Sdh4p Ser-71 is within hydrogen bonding distance of the UQ O4 atom (Fig. 4C). A S71A mutation greatly impairs quinone reductase enzymatic activity and results in a large decrease in quinone affinity for the Q D site (19). Another polar residue placed in the vicinity of the Q D site in our model is Sdh4p Ser-68, which may form a hydrogen bond with the quinone methoxy group. A series of aromatic residues, including Sdh3p Phe-57, Phe-60, Phe-99, and Phe-103, is involved in forming the Q D site and determining its rather hydrophobic character. We have previously identified Sdh3p Phe-103 as an important residue for quinone reduction and binding (18).
Clearly, the environment of the Q D site (Fig. 4C) is more hydrophobic than that of the Q P site (Fig. 4, A and B). In addition, there are fewer residues forming hydrogen bonds with UQ at the Q D site. These observations provide an explanation for the lower affinity of the Q D site for quinone and inhibitor. The s-BDNP apparent inhibition constants at the Q P and the Q D sites differ by about one order of magnitude (18,19). The free energies of binding calculated using the AutoDock empirical scoring function are in good agreement with the experimental observations (Table V). The hydrophobicity of the Q D site and the small number of polar interactions with UQ bound at this site should facilitate the exchange of bound quinone with the quinone pool.
Sdh3p His-106 and Sdh4p Cys-78 are in close proximity to the iron atom in the heme and as such are the best candidates for the axial ligands (Figs. 3C and 7B). To our knowledge, this is the first non-histidine axial heme ligand to be proposed for the SDH and FRD family of enzymes. The E. coli, Proteus vulgaris, and Haemophilus influenzae FrdD subunits also have a cysteine residue at this position, but these enzymes do not contain heme (3). It is not possible to mutate Cys-78 to a histidine residue; the larger side chain leads to severe steric clashes with heme b. Thus, in contrast to the E. coli SDH and the W. succinogenes FRD, the heme-binding site of the yeast SDH is relatively restricted in size. This may provide an explanation for why Cys with its smaller side chain is the axial ligand in the S. cerevisiae enzyme instead of histidine. We have used site-directed mutagenesis and spectroscopic analysis to explore this prediction. 2 The results strongly indicate that Sdh3p His-106 and Sdh4p Cys-78 are the heme axial ligands. Furthermore, substitution of Sdh4p Cys-78 for a His residue impairs enzyme assembly and catalysis.
The arrangement of the redox centers in the membrane dimer suggests that the heme and the two quinone-binding sites are involved in electron transfer from the catalytic dimer to the quinone pool. Fig. 7C shows the proposed arrangement of all redox centers in the S. cerevisiae SDH model. All edge-to-2 K. S. Oyedotun and B. D. Lemire, submitted publication. FIG. 7. A, averaged structure of the Sdh2p-Sdh3p-Sdh4p trimer calculated after MD simulation and the pathway of electron transfer. A, superimposition of the SDH trimer structure after molecular dynamics (blue) with the starting conformation (yellow). Sdh4p residues that are most mobile during molecular dynamics simulation are colored red. The average structure was calculated from the equilibrium ensemble of the last 500 ps using the g_rmsf utility of GROMACS. B, structure of the heme-binding site after a 5-ns molecular dynamics simulation. C, cofactor location and pathway of electron transfer. Edge-to-edge distances (Å) between the cofactors are indicated. edge distances between the redox centers are within the physiological range of electron transfer (56). Electron transfer between the Q P and the Q D sites, which are about 25 Å apart, will only occur through the heme b intermediate. All SDH identified to date contain at least one heme, but its role in electron transfer remains an open question. The higher affinity of the Q P site for ubiquinone may serve to optimize the site for quinone reduction, which proceeds through a semiquinone intermediate, while minimizing the release of radical species. The lower affinity of the Q D site suggests that it is an efficient site from which to release ubiquinol to the quinone pool. In this model, SDHs contain two electronically coupled quinone binding sites to balance the need for efficient catalysis with the need to minimize the production of damaging free radicals. Further investigations addressing the roles of the Q P and Q D sites and of heme b will be required to test this hypothesis.
The quaternary structure model of the S. cerevisiae SDH described in this study has identified new amino acid residues that may determine the structural or catalytic properties of each of the two quinone binding sites. The model has also provided insight into the unusual use of a cysteine as the second heme ligand instead of the histidine residues seen in all other members of the SDH and FRD family. Predictions arising from this model will stimulate further biochemical investigations and lead to a better understanding of the catalytic properties of a mitochondrial SDH and their connection to disease states.