Inhibitor-complexed Structures of the Cytochrome bc1 from the Photosynthetic Bacterium Rhodobacter sphaeroides*

The cytochrome bc1 complex (bc1) is a major contributor to the proton motive force across the membrane by coupling electron transfer to proton translocation. The crystal structures of wild type and mutant bc1 complexes from the photosynthetic purple bacterium Rhodobacter sphaeroides (Rsbc1), stabilized with the quinol oxidation (QP) site inhibitor stigmatellin alone or in combination with the quinone reduction (QN) site inhibitor antimycin, were determined. The high quality electron density permitted assignments of a new metal-binding site to the cytochrome c1 subunit and a number of lipid and detergent molecules. Structural differences between Rsbc1 and its mitochondrial counterparts are mostly extra membranous and provide a basis for understanding the function of the predominantly longer sequences in the bacterial subunits. Functional implications for the bc1 complex are derived from analyses of 10 independent molecules in various crystal forms and from comparisons with mitochondrial complexes.

The cytochrome bc 1 complex (bc 1 ) is a major contributor to the proton motive force across the membrane by coupling electron transfer to proton translocation. The crystal structures of wild type and mutant bc 1 complexes from the photosynthetic purple bacterium Rhodobacter sphaeroides (Rsbc 1 ), stabilized with the quinol oxidation (Q P ) site inhibitor stigmatellin alone or in combination with the quinone reduction (Q N ) site inhibitor antimycin, were determined. The high quality electron density permitted assignments of a new metal-binding site to the cytochrome c 1 subunit and a number of lipid and detergent molecules. Structural differences between Rsbc 1 and its mitochondrial counterparts are mostly extra membranous and provide a basis for understanding the function of the predominantly longer sequences in the bacterial subunits. Functional implications for the bc 1

complex are derived from analyses of 10 independent molecules in various crystal forms and from comparisons with mitochondrial complexes.
A central component of the cellular respiratory chain is the cytochrome bc 1 complex (cyt bc 1 or bc 1 ) 2 that catalyzes the electron transfer (ET) from quinol to cytochrome c (cyt c) and simultaneously pumps protons across the membrane, contributing to the electrochemical potential that drives ATP synthesis and many other cellular activities (1). In chloroplasts and cyanobacteria a related membrane protein complex, the cytochrome b 6 f (cyt b 6 f), bridges photosystem I and II, enabling oxygenic photosynthesis and conversion of light energy into a proton gradient for ATP generation (2). For non-oxygenic photosynthetic bacteria, such as R. sphaeroides (Rs), which can grow both aerobically and photosynthetically under anaerobic condition, the bc 1 complex is involved in both growth modes; however it is essential only under anaerobic conditions (3).
The critical importance of bc 1 has made it a target for numerous antibiotics, fungicides, and anti-parasitic agents. As a result, resistance to these agents has been documented in a wide variety of organisms (4 -8). Disorders that are related to defects in bc 1 complex are manifest clinically as mitochondrial myopathy (9), exercise intolerance (10), and Leber's optical neuropathy (11). Mounting evidence suggests a correlation between aging and the production of reactive oxygen species from defective bc 1 complexes (12,13). The elucidation of the molecular mechanisms underlying these phenomena requires a combination of experimental approaches and in particular, structural investigations that can provide a molecular framework for further experiments.
Significant advances in elucidating architectural features of this complex have been made by crystal structure determinations of mitochondrial bc 1 (14 -17) and b 6 f from a bacterium (18) and an alga (19). In particular, crystal structures of mitochondrial bc 1 in complex with various bc 1 inhibitors provide important mechanistic insights (20 -27), leading to a significant increase in the number of experimental studies and analyses of this enzyme. However, most recent functional investigations have been conducted with bacterial bc 1 complexes, especially those of non-oxygenic photosynthetic purple bacteria such as R. sphaeroides and R. capsulatus (Rc). These bacterial systems contain simpler bc 1 complexes consisting of either three (Rc) or four (Rs) subunits whose sequences have remained close to their mitochondrial counterparts. Chromatophore vesicles are easy to isolate in large quantities. Site-specific mutants can be readily prepared and tested in an optical pulse mode due to its coupling to the photosynthetic reaction center (28). Because of the importance of bacterial bc 1 in functional studies, a highresolution structure has been actively pursued for many years. The crystal structure of the bc 1 complex from R. capsulatus (Rcbc 1 ) reported at 3.8 Å resolution represented the first step toward this goal (29), though it lacks sufficient resolution of structural details that distinguish the bacterial form from the mitochondrial one. Here, we report the crystal structures of the wild type and mutant bc 1 complex from the R. sphaeroides (Rsbc 1 ) with bound inhibitors ranging in resolution from 3.1 to 2.4 Å.

EXPERIMENTAL PROCEDURES
Protein Purification and Crystallization-Cyt bc 1 complexes of both wild-type and double mutant S287R Cytb /V135S ISP from R. sphaeroides were prepared as described (30). The concentrated protein solution (90 mg/ml) was diluted by a factor of six with a buffer containing 50 mM Tris, pH 7.5, 200 mM NaCl, 10% glycerol, 5 mM NaN 3 , 0.5% ␤-OG (Anatrace), 200 mM histidine, and 2 mM diheptanoyl phosphatidyl choline (Avanti). The solution was left on ice for 12 h after adding a 5-fold molar excess of stigmatellin (Fluka). A second detergent, sucrose monocaprate (Fluka), was added to a final concentration of 0.12% followed by 10 mM strontium nitrate and 10% PEG400. The resulting solution was incubated overnight at 4°C. A small amount of precipitation was centrifuged off and the supernatant was used in sitting drop crystallization experiments that yielded small, red, translucent crystals after 2 months of incubation at 15°C. The reservoir solution was prepared separately and contained 100 mM Tris, pH 8.0, 600 mM NaCl, 20% glycerol, 5 mM NaN 3 , and 26% PEG 400.
Structure Determination-Crystals of Rsbc 1 were frozen without additional cryoprotectants but showed decay during the low temperature (100 K) data collection at beamline ID22D (SER-CAT) of the Advanced Photon Source (APS). Diffraction intensities were integrated with the program Denzo and merged and scaled with Scalepack (HKL2000 package) (31). The structures of Rsbc 1 inhibitor complexes were solved by molecular replacement (MR) using a dimeric Rsbc 1 model based largely on the structure of bovine bc 1 (32) with minor modification as input for the program MolRep (33) of the CCP4 (34) program package. The successful solution consisted of a set of three dimers related by non-crystallographic symmetry (NCS). This model was subjected to thorough rigid body refinement and simulated annealing (35), followed by cycles of standard crystallographic refinement and model building in O (36). Clear electron density for residues that were missing from the initial model in particular for insertions into cyt b and a rapid drop of R free confirmed the correctness of the MR solution. The final model contains continuous polypeptide chains of cyt b, cyt c 1 , and the ISP but none of the supernumerary subunit IV. Difference Fourier maps showed positive density for six lipid molecules, six ␤-OG molecules, six molecules of the substrate ubiquinone, and nine strontium ions. Throughout the refinement, NCS restraints were maintained except for a few regions at crystal contacts or in places of apparent local disorder. The structure of the wild type complex (P2 1 ) was solved by MR using a refined dimeric polypeptide-only model of the double mutant Rsbc 1 . A difference Fourier synthesis revealed the position of all prosthetic groups as well as the presence of both inhibitors stigmatellin and antimycin.

RESULTS AND DISCUSSION
Structure Determination and Overall Structure of the Rsbc 1 -The presence of the Q P site inhibitor stigmatellin, the use of the amino acid histidine and a mixture of ␤-octyl glucopyranoside (␤-OG) and sucrose monocaprate are important for obtaining high quality crystals. A batch of Rsbc 1 with the double muta- tions S287R cytb /V135S ISP appeared to be particularly suitable for growing well-behaved crystals (37). These crystals (space group C2) diffracted x-rays to 2.35 Å resolution (Table 1). Three bc 1 dimers occupy the crystallographic asymmetric unit (ASU). The wild type enzyme crystallizes with two dimers per ASU (space group P2 1 ) and diffracted x-rays to 2.6 Å resolution (Table 1). Surprisingly, only the three core subunits are present in both wild type and mutant Rsbc 1 crystals; apparently, subunit IV was lost upon crystal formation. The assembly of the three-subunit Rsbc 1 (Fig. 1A) resembles closely that of the corresponding subunits in bovine mitochondrial bc 1 (Bos taurus bc 1 or Btbc 1 ) (14), and the remarkable conservation in architectural features not only pertains to a single monomer but also to an assembled homodimer. The root-mean-square (r.m.s.) deviation between the cyt b dimers of R. sphaeroides and bovine bc 1 is less than 1.1 Å for 726 C␣ atoms (Table 2). Consequently, the distances between prosthetic groups are virtually identical, implying functional conservation (Supplemental Table  S1). As in mitochondrial bc 1 , the extrinsic domain of the iron-sulfur protein subunit in Rsbc 1 crosses over, connecting one molecule of cyt b to the adjacent one. In contrast to the seven or eight supernumerary subunits in mitochondrial enzymes, Rsbc 1 has only one. Thus, it has been speculated that supernumerary subunits represent functional or structural equivalents of the insertions, extensions, and deletions found in the sequences of the catalytic subunits of bacterial bc 1 (38).
Structure of the Cytochrome b Subunit-The cyt b subunit of Rsbc 1 has eight membrane spanning helices named A to H, forming two helical bundles (A-E and F-H) (Fig.  1B). The two heme groups, b L and b H , reside within the first bundle. Extra membranous loops connect pairs of transmembrane (TM) helices and those that are longer than 20 residues are the AB, CD, DE, and EF loop. The quinol oxidation site (Q P , positive side, P-side) near the periplasmic side of the membrane and quinone reduction site (Q N , cytoplasmic or negative side, Nside) on the opposite side can be identified with bound stigmatellin and antimycin, respectively. When compared with structures of Btbc 1 , Rs cyt b features two terminal extensions and two major insertions. The N-and C-terminal extensions are 22 and 29 residues long, respectively. Both contain helices named a0 and i, respectively (Figs. 1B and 3A). One insertion (de helix) is in the cytoplasmic DE loop and another (ef1 helix) inserts after the ef helix on the periplasmic side.
Structure of the Cytochrome c 1 Subunit-The cyt c 1 subunit folds in a manner similar to that of its mitochondrial counter- part, having a C-terminal TM helix (Fig. 1C) and featuring the Cys 36 -X-X-Cys 39 -His 40 motif characteristic for c-type cytochromes ( Fig. 3B) with the heme iron atom being coordinated by the side chains of His 40 and Met 185 as 5th and 6th ligand, respectively. The heme group is located and positioned identically to that of mitochondrial enzymes. Crystals of Rsbc 1 grown in the presence of strontium ions revealed a metal ion-binding site on cyt c 1 (Fig. 1, C and D), which is not present in mitochondrial bc 1 but appears to be conserved in photosynthetic bacteria (Fig. 3B). The strontium ion, confirmed by the appearance of a strong anomalous signal from the data set collected above the strontium absorption edge, is accessible from the periplasm and coordinated by side chains of Asp 8 , Glu 14 , and Glu 129 as well as by the backbone carbonyl oxygen atom of residue Val 9 in a distorted octahedron. To our knowledge, this metal ion-binding site has not been described previously and its possible physiological role is currently under investigation.
Structure of the Iron-Sulfur Protein Subunit-The ISP subunit has a C-terminal periplasmic head domain (extrinsic domain, ISP-ED), which connects through a flexible hinge region to its N-terminal TM helix (Fig. 1E). The ISP-ED is predominantly a ␤-structure consisting of three ␤-sheets arranged in three parallel layers with the 2Fe2S cluster located at the apex of the ISP-ED between the 2nd and 3rd ␤-sheet. The conserved ADV motif in the hinge region ( Fig. 3C) adopts an ␣-helical (HA) conformation, unlike the random coil secondary structure of bovine bc 1 in the space group I4 1 22. One insertion with respect to the bovine sequence is located between Thr 96 and Ala 109 , filling a surface depression that would otherwise exist between ␤-sheets 2 and 3.
Inhibitor Binding Sites-Stigmatellin is a potent Q P site inhibitor (39) and has frequently been used to arrest the mobile head domain of the ISP (26). This may be of particular importance in crystallizing the bare-bone bc 1 complexes that cannot rely on lattice contacts formed by large hydrophilic core subunits (14,15), or by antibody fragments (17). In all Rsbc 1 structures described in this work, stigmatellin is clearly visible in the difference Fourier maps and oriented in a way that bridges the side chains of the residues Glu 295 (cyt b) and His 152 (ISP) ( Fig. 2A). The hydrophobic tail, known to contribute to the small K d (0.4 -1 nM) of stigmatellin binding (40), is fully visible (Fig. 3A) and makes the same contacts as in complexes from mitochondria (17,26) or in b 6 f complexes (19,41).
Antimycin molecules were refined in all four (P2 1 ) or six copies (C2) of cyt b in the asymmetric unit. The antimycin complex of bovine cyt b superimposes very well with the corresponding Rsbc 1 subunit (Fig. 2B), and the inhibitors fit nicely into the difference density observed in the latter. Antimycin binds strongly to mitochondrial bc 1 with a K d of 32 pM (42), but the K d for Rsbc 1 FIGURE 2. Inhibitor binding sites in cyt b. A, stereoscopic diagram showing electron density for stigmatellin (white) and its binding environment (blue). Stigmatellin and its surrounding residues are portrayed in the stick model with yellow carbon, red oxygen and blue nitrogen, green sulfur and brown iron atoms. His 161 of ISP is 2.64 Å away from the atom O4 of stigmatellin and its phenolic oxygen atom O8 is 3.01 Å away from OE1 of Glu 295 . The electron density is contoured at 1.5. B, stereoscopic view of the Q N site. The bound antimycin in yellow forms two hydrogen bonds (red dotted lines) with the 3-FASA moiety of the inhibitor. The b H heme is shown as a stick model in gray. The inhibitor displays anti conformation as in 1PPJ (27). is not known. As prominent contributors to the binding energy, the formyl amide and the phenolic OH group of antimycin are in excellent hydrogen bonding distance to Asp 252 (Asp 228 , Btbc 1 ). However, two amino acid changes could strongly influence the binding of antimycin. The substitution of Ser 35 in Btbc 1 to Val 49 in Rsbc 1 eliminates the H-bond between the hydroxyl group of Ser and the central amide group (N2) of antimycin. It is likely that this loss of an H-bond reduces the inhibitor's binding affinity to Rs cyt b because a similar mutation (S35I) in Leishmania tarentolae cyt b increases the IC 50 for antimycin by 10 -20-fold (43). The impact of the change from Ser 205 (Bt) to Asn 221 (Rs) is difficult to judge and may be restricted to a modulation of the steric interference with the aromatic moiety of antimycin. Bound antimycin was modeled in anti conformation with respect to the positions of atoms O2 (OH) and O3 (OϭC), abolishing the intramolecular hydrogen bond that was proposed to be essential for its recognition (44) and indeed found in the free structure (45). It appears that the conclusion that this intramolecular hydrogen bond is essential, drawn from a synthetic analogue of antimycin, cannot be maintained, as recent high resolution structures (27) 3 agree better with the anti conformation. Insertions and Deletions in Rsbc 1 Relative to Mitochondrial bc 1 -When sequences of mitochondrial bc 1 and Rsbc 1 are compared, the latter often possesses more insertions than deletions. Remarkably, the insertions occur only on or near the periplasmic or cytoplasmic side and not within the transmembrane region (Fig. 1). An understanding of the functions of these additions or deletions may provide insight into the evolutionary process that transformed the bacterial enzyme into its mitochondrial equivalent. Characteristic of this process is the addition of supernumerary subunits that possibly provide structural stability and functional integrity to the enzyme (3).
Roles of the Additions in Cyt b-Cyt b of Rsbc 1 is 66 residues longer than its bovine mitochondrial equivalent. As seen in the 3 D. Xia  structure-based sequence alignment and from its structure (Figs. 1B and 3A), bacterial bc 1 has extensions at both termini (a0 and i helices), a helical insertion between the D and E helices (de helix) as well as an insertion between the ef-loop and the F helix (ef1 helix). In the structure, the C terminus of cyt b is visible up to residue 430, consistent with the observation that deletion of the last 15 residues does not affect the function of Rsbc 1 (38). Except for the ef1 helix, all extensions and insertions are located on the N-side of the membrane (Figs. 1B and 4A), which likely function to maintain the structural integrity of the quinone reduction site by preventing potential electron leakages and by safeguarding channels for proton influx (24). Indeed, without the supernumerary subunits, especially core1 and core2, the heme b H (and with it the Q N site) of the mito-    chondrial cyt b is only weakly shielded from the aqueous matrix by a thin layer of protein side chains. In contrast, the Q N site of Rsbc 1 is well protected by an additional layer consisting of the de-helix insertion and the two terminal extensions. The location of the de-helix permits interaction with its own C-terminal extension and with the end of the N terminus from the neighboring cyt b through a network of hydrogen bonds (Fig. 1, A and  B, Supplemental Table S2). The a 0 -helix reaches to the cyt b of its symmetry mate and forms a pair of salt bridges between Arg 22 and Glu 126 of the symmetry-related cyt b and a number of hydrogen bonds as well as van der Waals interactions (Fig. 1, A  and B). Mutational studies have shown that C-terminal truncations as far as residue 421 lead to increasing detergent sensitivity, loss of ISP and subunit IV during purification and lowering the potentials of both heme groups, leading to eventual inactivation of Rsbc 1 (38). The structure qualitatively explains these observations by demonstrating the interaction of the C terminus of cyt b (via the i helix) with the C terminus of cyt c 1 (indirectly to the N terminus of the ISP) and with the de-helix, which is in close proximity to the b H heme (38).
On the periplasmic side, there is one large insertion of 18 residues (310 -327) between Pro 285 and Asn 286 (Bt cyt b) containing the ef1-helix (Fig. 4B), which protrudes from cyt b laterally and runs parallel to the membrane surface (Fig. 1B). This insertion occurs only in species that belong to the phylum proteobacteria (Fig. 3A). However, it is functionally important, as the point mutation S322A or deletion of residues 309 -326 significantly lowers the enzyme activity (46). The ef1-helix may play an important role in lipid binding, as features of several potential lipid molecules are visible in the electron density (Figs. 4B and 5C). It also enhances crystal contacts through aromatic stacking interaction between Trp 313 of adjacent cyt b subunits.
Insertions and Deletions in Cyt c 1 -The structure-based sequence alignment (Fig. 3B) shows that cyt c 1 of Rsbc 1 has undergone both insertions and deletions relative to mitochondrial complexes. Apart from the two small insertions in the Rs cyt c 1 after Glu 52 (4 residues) and Ala 146 (3 residues), there is one large insertion between Gly 109 and Gly 127 . It features a short helix (H1d) that protrudes from cyt c 1 into the lipid bilayer sealing off a compartment between cyt c 1 and cyt b (Fig.  1, A and C). In Btbc 1 , the absence of this insertion creates a niche at the interface between the end of the Helix E of cyt b and cyt c 1 . A possible function of this insertion may relate to lipid binding (next section). The only insertion in cyt c 1 that may replace the function of a supernumerary subunit is the 18-residue insertion starting at position 162, which is spatially close to the head domain of ISP (Fig. 4C). Containing a short helix H2a, this region is characterized by an increased disorder (high B-factor) but features a stabilizing disulfide bridge (Cys 145 -Cys 169 ), whose existence is in agreement with recently published data (37). Approaching the ISP-ED within ϳ8 Å (C␣ distance from cyt c 1 Asn 173 to ISP Asp 143 ), this insertion presumably functions as an extended arm to limit the motion of the ISP-ED (Fig. 4C). However, the intrinsic flexibility and extent of solvent exposure renders it susceptible to proteolytic attack and, conceivably, places it in an evolutionarily disadvantageous position, possibly leading to the replacement of its function by the supernumerary subunit VIII in mitochondrial enzymes (Supplemental Fig. S1).
Compared with mitochondrial cyt c 1 , two large deletions, near residues Thr 77 and Ser 92 , respectively (Fig. 3B), result in the loss of bridging interactions between the two cyt c 1 subunits within the dimer (Fig. 4C). The absence of these contacts in Rsbc 1 creates a large continuous groove (13 Å wide) on the P-side surface. Beyond possible functional implications, the closure of the gap improves stability around the heme group in mitochondrial cyt c 1 .
Insertions in the ISP-Structure-based sequence alignment shows one insertion in the sequences of Rs ISP (Fig. 3C). This insertion (residues 97-108) is located on the surface of ISP-ED distal to cyt c 1 and stays 20 -25 Å away from the 2Fe2S cluster as predicted (47); it forms a globular structure containing three ␤-turns and one inverse ␥-turn (Fig. 4D). There is an intricate network of interactions employing both main chain and side chain atoms, suggesting a stabilizing role for this insertion (Supplemental Table S3). Disruption of this network of interactions by more than one point mutation led to the loss of the ISP subunit in the complex (47). From a morphological point of view, the insertion 97-108 in Rsbc 1 and 97-107 in Rcbc 1 help maintain the globular shape of the ISP-ED as compared with its mitochondrial homologues (Fig. 4D).
Structures of Bound Lipid and Detergent Molecules-Membrane proteins depend on the presence of lipids to maintain their functional and/or structural integrity. Bovine and yeast mitochondrial bc 1 are inactivated through delipidation (21,48), a process that can be reversed by restoring specific lipids (49). In crystals of Rsbc 1 , characteristic features of several lipid and detergent molecules appeared in difference Fourier maps. Ordered lipid molecules are often found between symmetryrelated dimers, at the dimer or subunit interfaces, and in surface depressions. One lipid molecule was positively identified on the N-side of the membrane and included in the model. However, additional lipids that are only partially recognizable at both sides of the membrane were excluded. The lipid molecule bound at the cytoplasmic surface of cyt b is modeled as a lauryl oleoyl phosphatidyl ethanolamine (PE); its head group aligns with the surface plane of the cytoplasmic leaflet of the membrane and its fatty acid chains flank the TM helices B and G of cyt b (Fig. 5A). The exact identities of the fatty acids are unknown but the assignment as PE is supported by comparing it to the lipids present in bovine 3 and yeast (21) mitochondrial bc 1 . The phosphate group is hydrogen bonded to two highly conserved consecutive tyrosine residues (Tyr 117 and Tyr 118 ), and the lipid head group is further stabilized by the side chain of Arg 358 by forming an ion pair with the lipid phosphate. Common to all structures is the lining of the groove between TM helices B and G (cyt b) with one of the fatty acid chains. In contrast to the lipid in mitochondrial bc 1 whose terminal n-alkyl (n ϳ6 -9) moieties of both chains reach into the groove, the bulky side chain of Phe 113 prevents this interaction in Rs cyt b.
On the periplasmic side, a few clusters of residues are involved in lipid binding. The 17-residue insertion, including the H1d helix, in the cyt c 1 subunit (110 -126, Figs. 1C and 3B) is positioned parallel to the plane of lipid head groups of the membrane and protrudes laterally into the membrane bilayer, creating a cavity bounded additionally by the TM helices E and G and the ef-loop (including PEWY) of cyt b. The C-terminal end of this helical insertion is ϳ5 Å away from the cyt c 1 metal binding site and ϳ5 Å away from the sugar ring of a detergent molecule (␤-OG). There is weak electron density in the void, which resembles the head group of another lipid molecule with fatty acid hydrocarbon chains extending toward the cytoplasmic side (Fig.  5B). Weak density is present in all six copies of the bc 1 monomers but is insufficient to build and refine a lipid molecule with confidence.
At the N-terminal end of the ef1 insertion in cyt b, the side chain of Trp 313 forms a stacked pair with its symmetry-mate from a neighboring dimer at a distance of 3.8 Å. This pair is symmetrically flanked by at least six pieces of extra electron density, most likely stemming from bound lipid or detergent molecules (Fig. 5C). A strontium ion, clearly confirmed by its anomalous signal, sits right above the indole rings of the tryptophan pair. Its exact coordination environment cannot be resolved, but might involve the head groups of two pairs of putative lipid molecules. We observed the tryptophan pair formation in all crystal forms, and the presence of strontium ions seems to strengthen the interaction but is not required (Fig. 5C).
Fate of the Subunit IV-Purified Rsbc 1 , both wild type and mutant, contains one additional 14.4-kDa subunit (subunit IV), which has been shown to enhance the activity of the core subunits by 68% (50) but is not essential for the function of the complex or the survival of the organism. The same observations have been made about the non-essential 6-kDa subunit in Rhodovulum sulfidophilum bc 1 (51). In fact, many of the known bacterial forms of bc 1 , including Rhodobacter capsulatus and Paracoccus denitrificans, contain only the required three core subunits, cyt b, cyt c 1 and the ISP. In the crystal structure, however, subunit IV is missing from the complex, indicating that the crystallization medium (including PEG400, detergents, etc.) must have caused the detachment of subunit IV. A SDS-PAGE gel revealed the presence of subunit IV in solution, but showed no detectable amount in crystals (data not shown). It is not uncommon to lose a supernumerary subunit during crystallization of mitochondrial bc 1 complexes (17,27). To test whether subunit IV is indirectly required for the crystallization of Rsbc 1 , we purified the ⌬-subIV mutant (32) and subjected it to the same crystallization conditions. Triclinic crystals grew readily, displaying the same plate-like morphology as the monoclinic ones, and diffracted x-rays to 3.1-Å resolution. The structure could be readily solved and refined (Data not shown), demonstrating that the subunit IV is not required for crystallization.
Functional Implications of the Rsbc 1 Structure-Despite lacking supernumerary subunits, bacterial bc 1 operates in exactly the same way as mitochondrial enzymes and has therefore been widely used as a model system for mechanistic studies. The modified Q cycle mechanism for bc 1 function has received most experimental support (1); it defines separate Q P and Q N sites and requires an obligatory bifurcated ET pathway at the Q P site. Previous crystallographic studies on mitochondrial bc 1 complexes revealed the physical locations for the Q P and Q N site, respectively (14), demonstrated the importance of the ISP-ED mobility in the electron bifurcation at the Q P site (15), and outlined a possible mechanism for the ISP-ED conformation switch (46). The structures of Rsbc 1 in this work contain additional dynamic structural information encoded in multiple copies of the complex that promises further insight into its mechanism. For example, the crystal with C2 symmetry consists of six copies of Rsbc 1 monomers in a crystallographic asymmetric unit, and crystals of space group P2 1 have four copies. By properly superimposing various parts of the structure, dynamic information with respect to substrate binding and subunit motion is revealed.
Multiple Binding Positions of Quinone at the Q N Site-Toward the end of the refinement of the Rsbc 1 structure in the absence of the Q N site inhibitor antimycin, non-crystallographic symmetry (NCS) restraints for side chains of the quinone interacting residues (His 217 , Asp 252 , and Asn 221 ) as well as for the bound substrate ubiquinone (UQ) were released. This permitted a more realistic estimate of ϳ70% occupancy of the bound natural substrate UQ based on the comparison of average B factor of UQ to side chain atoms of those of surrounding, interacting residues. The long isoprenoid tail of UQ falls rapidly into discontinuous electron density and was therefore modeled with only two isoprenoid repeats. The quinone molecules are roughly perpendicular to the parallel planes of Phe 216 and heme b H on one side and parallel to the plane of Phe 244 on the other side (Fig. 6A). The side chains of three polar residues, His 217 , Asp 252 , and Asn 221 , are within contact distances to the bound UQ.
When the six independent cyt b monomers were superimposed in pairs, the average r.m.s. deviation of the superposition is 0.14 Å for 428 residues. Except for the ef1 helix, the rest of the main chain atoms align almost perfectly. Additionally, the chromone rings of stigmatellin molecules at the Q P pockets superimpose well. At the Q N pocket, hydrophobic residues lining the wall and the b H heme groups are also well aligned, whereas the positions and orientations of bound quinone substrates and side chains of the three interacting residues, His 217 , Asp 252 , and Asn 221 , were spread over a small range (Fig. 6A) (52). Asn 221 and Asp 252 (Ser 205 and Asp 228 in bovine) were observed to H-bond to UQ in Btbc 1 (24) but the bonding distance in the independent Rsbc 1 monomers varied from 3.6 to 4.5 Å for Asn 221 and from 3.8 to 5.5 Å for Asp 252 . The motions exhibited by bound UQ demonstrate a weak binding of the Q N pocket for the substrate UQ; such a variation was not observed for the binding of the Q N site-specific inhibitor antimycin. Furthermore, this observation provides a direct structural support for the notion of low-binding affinity of the substrate during catalysis (24) or functional conformation (53). In one of the six monomers, Asp 252 H-bonds to a well-ordered water molecule. The conserved Lys 251 , proposed previously to be important for proton uptake, displays considerable conformation variations.
Positional Anisotropy in the Subunits of Rsbc 1 -The presence of six independent copies of each subunit (C2 form) permits an analysis of the degree to which cyt b, cyt c 1 and the ISP show flexibility when they assemble into the complex. Of particular interest is the question of the position of the ISP-ED, which is known to undergo large scale conformational changes as part of the mechanism of bc 1 function (15,46). However, in this inhibited complex the ISP-ED is firmly locked down in the Q P site by stigmatellin. By superimposing only cyt b subunits and calculating the r.m.s. deviations between the position of pairs of cyt B, stereo diagram of the structure and environment of the P-side insertion (ef1-helix) in cyt b. The C␣ trace of Rsbc 1 is yellow and that of Btbc 1 is black. At the N-terminal side of the ef1-helix is a large piece of electron density shown in green. Labels in italics refer to the bovine cyt b sequence. C, stereo view of a superposition of C␣ traces of Rs (green) and bovine (brown) cyt c 1 looking down the molecular 2-fold axis from the periplasmic side into membrane bilayer. Also shown is the Rs ISP (gray) in the hypothetical c 1 position obtained from the position of ISP found in the 1BE3 structure. Mitochondrial cyt c 1 extends the hair-pin structure around residues 73-79 (bovine numbering) to reach out to the helix around residues 92-108. The contact area between the two elements that reach across the dimer is shown in transparent cyan and beige surfaces. In contrast, equivalent residues (77, 95) in Rsbc 1 are more than 13 Å apart. D, stereo view of the ISP insertion. The insertion of residues 97-108 (in stick model) is shown under a transparent surface. The locations of the 2Fe2S cluster with its ligands (His 131 , His 152 , Cys 129 , Cys 149 ), the C terminus Gly 187 and N-terminal Ala 48 that connects to the TM helix are marked on the surface in green, cyan, and yellow, respectively. While the r.m.s. deviations hint at the increased propensities of cyt c 1 and the ISP to adopt different positions, these values alone do not reveal the nature of the underlying distributions. To visualize the character of the distributions, each sextet of C␣ positions was subjected to a trivariate Gaussian analysis (See Supplemental information) (Fig.  6B). The surface of the ellipsoids is drawn at a constant probability density encompassing a volume representing 90% of the total probability and shows clearly anisotropic spread of the ISP-ED C␣ positions that has a pivotal region around the 2Fe2S cluster, which grows larger with distance. We interpret the anisotropic shape and size of the probability ellipsoids of the ISP-ED not as an actual movement of the ISP, as each molecule is locked into the Q P position and restrained by crystal packing forces, but as a qualitative measure for the ability of the ISP-ED to undergo directional movements when unrestrained. Although the mobility of ISP-ED has been established by a number of experimental approaches (15, 20, 26, 54 -56), the ability of the ISP-ED to undergo positional adjustments qualitatively consistent with its function to move toward cyt c 1 has never been shown in a single structure before.
The extrinsic domain of cyt c 1 (cyt c 1 -ED) also showed considerable "motion". Unlike ISP-ED, the displacements of cyt c 1 -ED are largely isotropic (Fig. 6B).). The increased mobility that arises from the absence of inter cyt c 1 contacts FIGURE 5. Lipid and detergent molecules in Rsbc 1 . A, modeled lipid at the N-side and its binding environment. Transparent van der Waals spheres indicate the volume of the fatty acid chains. B, modeled detergent and lipid molecules at the P-side near the Sr 2ϩ binding site in cyt c 1 . The cyt b is drawn as green ribbons and the cyt c 1 is blue. The c 1 insertion (110 -125) is shown as yellow stick model. C, stereo pair: lipid-protein interactions at the interface between symmetry-related dimers. The contacts at the interface between two Rsbc 1 dimers in the crystal are provided entirely by residues on the periplasmic side and the contribution from the ef1 helix of cyt b is illustrated here. This view from the periplasmic side shows portions of symmetry related cyt b around the ef1 helix. Residues contributing to the interface are drawn in stick models and are labeled. Anomalous difference electron density for the Sr 2ϩ ion is contoured at 4 (magenta) and the difference density for putative lipids is displayed as green wire cages contoured at 3. due to the two deletions in the Rsbc 1 dimer (Fig. 4c) might result in unsuccessful docking attempts of ISP-ED to cyt c 1 , leading to reduced efficiency in ET.
Atomic coordinates of the refined inhibitor-bound Rsbc 1 structures have been deposited in the Protein Data Bank with accession codes: 2QJP (wild type, stigmatellin and antimycin), 2QJY (double mutant, stigmatellin), 2QJK (double mutant, stigmatellin and antimycin). FIGURE 6. Observed variability in substrate binding and in subunit conformation. A, stereoscopic pair showing that after superimposing the six copies of cyt b, the UQ molecules bound at the Q N site are in slightly different orientations and positions. The secondary structure elements delimiting the Q N pocket are shown in ribbon form and labeled. Residues potentially interacting with bound quinone are labeled and drawn as stick models. The bound UQ molecules are shown as stick models in various shades of gray representing the origins of different monomers. B, after superimposing cyt b from ten different Rsbc 1 models, the spread of the C␣ positions of all three subunits is represented by ellipsoids drawn at 90% probability level. Note that there is a significant angular displacement in the ISP with a pivotal point near the 2Fe2S cluster.