Structure of the Yeast Cytochrome bc 1 Complex with a Hydroxyquinone Anion Q o Site Inhibitor bound

5- -tridecyl-6-hydroxy-4,7-dioxobenzothiazole;

The mechanism of the enzyme known as the protonmotive Q cycle (8) involves via the low and the high potential b type hemes to ubiquinone. The resulting stable semiquinone is fully reduced after a second ubiquinol molecule is oxidized at the Q o site.
While the main features of catalysis are understood, the molecular mechanism of ubiquinol oxidation is not clear. Also, pathways for proton uptake and release are hypothetical (7,9,10). Several hypotheses have been proposed to explain the divergent transfer of electrons into thermodynamically different pathways (see Ref. 2). The double occupancy model suggests synergistic interaction between two quinone molecules which occupy the Q o site simultaneously (11)(12)(13). The proton-gated charge transfer mechanism proposes that the activation barrier is a function of the deprotonation of ubiquinol (14), but this mechanism is not supported by other kinetic studies (15). Single occupancy models include simultaneous as well as sequential electron transfer to the acceptors. For the latter a proton-gated affinity change mechanism claims the presence of a relatively stable intermediate in the transition state with the rate-limiting step at the second electron transfer (16). Since a semiquinone radical has not been detected at the Q o site, this has been explained by an EPR silent anti-ferromagnetically coupled semiquinone-[2Fe-2S] reduced pair (16,17). Another explanation for the undetectable semiquinone is provided by Crofts and colleagues, who suggest rapid dissociation of the product after the first electron transfer and movement of the semiquinone within the bilobal Q o binding pocket to allow rapid reduction of heme b L (10, 18). Recent kinetic data show that the midpoint potentials of b type hemes control the rate of cytochrome c 1 reduction. This is consistent with the view that ubiquinol oxidation is a concerted reaction (19).
Inhibitors are important tools to analyze the molecular mechanism of Q o site catalysis. Three types of inhibitors can be distinguished: ligands binding at the proximal domain and therefore perturbing the spectroscopic properties of heme b L (Q o -I e.g. myxothiazol, MOA-stilbene), those binding to the distal domain and affecting the Rieske [2Fe-2S] EPR lineshape (Q o -II e.g. UHDBT), or compounds exhibiting both effects (Q o -III e.g. stigmatellin) (20). Kinetic data indicates that occupation of these inhibitors at the Q o site is mutually exclusive, suggesting overlapping binding sites. This was observed in crystal structures where these inhibitors are found to bind in different but overlapping domains of the bilobal Q o site, termed distal and proximal to heme b L (20).
Analysis of anomalous scattering data indicated a high occupancy of the catalytic Rieske domain in the b-position in the presence of inhibitors that bind to the distal domain, such as stigmatellin and UHDBT (20). However, in previous crystallographic studies on UHDBT binding (18, 20) the inhibitor could not be refined and high resolution structural information about UHDBT binding at the active site has not been available up to now. The substrate ubiquinol has not been detected in the Q o site by X-ray structural analysis. Therefore, the analysis of structural analogs of the substrate which function as competitive inhibitors is important.
Here, we present the three-dimensional structure of a UHDBT analog, HHDBT, inhibited bc 1 complex from the yeast Saccharomyces cerevisiae at 2.5 Å resolution. This hydroxyquinone binds in its ionized form and its binding is discussed as resembling an intermediate step of ubiquinol oxidation. Conformational changes at the binding site confirm the previously postulated proton transfer pathway and reveal plasticity at the active site.

EXPERIMENTAL PROCEDURES
Protein Purification and Crystallization -The bc 1 complex from the yeast S. cerevisiae was purified and a co-complex with the antibody fragment Fv 18E11 was formed and crystallized as previously described with the following minor modifications (6,21).
The buffer volume for detergent exchange in the second DEAE anion exchange chromatographic step was reduced by 5-fold. HHDBT was added at a final concentration of 100 µM to the purified bc 1 complex-Fv co-complex after size exclusion chromatography (TSK4000, TosoHaas) prior to crystallization. The final purification step was performed at pH 7.5. The crystals were obtained using micro seeding and vapor diffusion technique against PEG4000 at 4°C. The protein solution (50 mg/ml) was mixed with precipitation agent [5 % PEG4000, Tris-HCl pH 7.5 (adjusted at room temperature), 0.05% n-undecyl-β-D-maltopyranoside, 10 µM HHDBT] resulting in a pH of 8.0 at 4 °C, that is 0.5 pH unit lower than the structure of the stigmatellin inhibited enzyme (6).
Crystals grew within a few days to a size suitable for X-ray analysis (~ 0.5 x 0.5 x 1.0 mm).
Total protein determination was performed with a modified Lowry procedure, using the BC Assay protein quantitation kit (Uptima) (22). The bc 1 complex content was estimated as half the amount from spectroscopic quantification of the two b-type hemes using an extinction coefficient of 28.5 mM -1 cm -1 for the dithionite-reduced minus ferricyanide-oxidized difference spectra (262-275 nm). Enzyme activity was determined by monitoring cytochrome c reduction in a spectrophotometric assay at 550 nm using an extinction coefficient of 18.5 mM -1 cm -1 for cytochrome c. Turnover numbers refer to mol cytochrome c reduced per mol bc 1 complex per second under conditions of continuous turnover where the catalytic reaction is zero order with respect to decylubiquinol and cytochrome c. A detailed description is reported elsewhere (21). The concentration of decyl-ubiquinol was varied and the activity measured in the absence or presence of UHDBT.
addition to the previously assigned phospholipids (7). Finally, water molecules were included according to peaks observed in the Fo-Fc electron density map contoured at 3σ. Their positions were refined yielding 326 molecules of which 203 are the same as in the original model (1KB9) and their numbering was kept. New water molecules were numbered as starting from Wat 500 . Refinement resulted in final R factor and free R factor of 22.8% and 25.2%, respectively ( Table 1). Coordinates of the HHDBT inhibited enzyme have been deposited in the database (PDB entry 1P84).
For comparison stigmatellin-inhibited bc 1 complex was crystallized at the same pH as the HHDBT containing enzyme. The control data-set was collected with 2. Analysis of neighbouring atoms and hydrogen bond interactions was performed using the programs HBPlus (27) and contact analysis from CNS. Accessibility was PROCHECK (V.3.2) analysis verified the stereochemical quality of the coordinates (Table 1) (29). Hydrogen bonds were assigned according to appropriate distance and geometry. For analysis of weak hydrogen bonds an estimation of hydrogen atom position was made by generating a structural model with hydrogens added using CNS (V.1.0). Criteria for identifying weak hydrogen bonds were extracted from (30). Hydrogen bond angle is denoted as θ (X-H … A) and the bending angle at acceptor atom φ (H … A-C).
Illustrations were prepared using program O (25), LIGPLOT V.4.0 (31), MolScript V.1.4 (32), BobScript (33) and Raster3D (34). 1 Complex -The optimized purification of yeast bc 1 complex resulted in a pure and more active membrane protein complex with a higher turnover number of 82 s -1 compared with the previously reported activity of 64 s -1 (21). The increase is most likely due to higher phospholipid content of the modified protein preparation ( Fig.1) (35,36). Effective inhibition of the complex has been shown for 5-n-alkyl-6-hydroxy-4,7-dioxobenzothiazoles containing 7-15 carbon alkyl side-chains (37). Here, the shorter heptyl side-chain analog of UHDBT was used for crystallization, in order to avoid non-specific binding that might occur with the longer alkyl side-chains at the high concentrations used.

Crystallization of HHDBT Inhibited bc
The inhibitory efficacy of UHDBT was shown to depend on the oxidationreduction poise of the catalytic subunits, demonstrated by enhanced binding when the Rieske protein is reduced (37). The purified bc 1 complex used in this study has a partially reduced Rieske and is fully inhibited by the applied amount of HHDBT (results not shown). A pK a of 6.5 has been determined for the weakly acidic hydroxy group of UHDBT, which was measured in phosphate buffer containing 1% ethanol, and deprotonation of the hydroxy group is manifested by a color change from yellow to roseviolet (38). Here, the complex was crystallized at pH 8.0 as a protein-detergent complex, therefore the acidity of the hydroxy group was measured in detergent micelles by monitoring the blue shift in the optical spectrum upon ionization. The pK a determined by spectrophotometric titration in detergent micelles was 6.1 (results not shown). This suggests that 98% of the inhibitor is ionized in the crystallization mixture, demonstrated by the violet color of the solution, and finally a purple tint of the crystals. Hydroxyquinone binding to cytochrome bc 1 complex A difference spectrum of bc 1 complex with bound inhibitor versus the complex alone shows that the inhibitor is ionized when bound to the enzyme while the pH of the buffer is close to the pK a of the unbound inhibitor (Fig.2). The difference spectrum of the complex at pH 6.0 with inhibitor bound at substoichiometric amount is similar to that of the inhibitor alone at pH 8.7. As inhibitor is added in molar excess, the mixture consists of bound and unbound inhibitor and the absorbance maximum shifts to longer wavelengths. Crystallographic analysis of stigmatellin binding clearly showed that Nε2 is protonated at pH 8.5, as the fixation of the Rieske protein in the b-position is stabilized by a hydrogen bond between this atom and the carbonyl group of stigmatellin (6). For crystallization of the bc 1 complex in the presence of HHDBT the pH was lowered by half a unit, therefore, Nε2 is expected to be protonated under these conditions (see  UHDBT is a Competitive Inhibitor -To analyze if hydroxy-dioxobenzothiazoles compete with substrate at the Q o site, the apparent K m for decyl-ubiquinol binding to yeast bc 1 complex was determined at varying concentrations of UHDBT. The K m varied with inhibitor concentration, whereas V max remained constant, clearly demonstrating that UHDBT is a competitive inhibitor (Fig.4). Therefore, hydroxy-dioxobenzothiazoles may be regarded as substrate analogs. High structural similarity to the substrate and competitive inhibition by HHDBT suggest that it is a substrate analogue in which the ring methyl group is replaced by a deprotonated hydroxy group, the two methoxy groups are replaced by a fused thiazole ring and the isoprenoid tail is replaced by a short saturated side chain (see insert, Fig.3).

DISCUSSION
Hydroxy-dioxobenzothiazoles efficiently inhibit the bc 1 complex in a pH dependent manner (38). It was argued that low efficacy at alkaline pH could be attributed to ionization of the inhibitor due to restricted access to the binding site or to protonation of a functional group within the enzyme (38,44). Notably, the apparent pK a of 7.5 for inhibitor efficacy is more alkaline than the pK a of UHDBT but closely matches the pK a suggested for the imidazole nitrogen of His 181 of the oxidized Rieske protein His 181 is expected to be protonated, because it was shown to donate a hydrogen bond to the carbonyl group of stigmatellin at half a pH unit above the one used for HHDBT crystallization. In addition, crystallization is performed close to the pK a of His 181 . Furthermore, the enzyme preparation is partially reduced and the pK a of the reduced Rieske imidazole nitrogen is far above (11.5) the crystallization conditions (45). The presence of a hydrogen bond from His 181 to HHDBT is shown by the fact that ligand binding fixes the mobile Rieske domain in the b-position. Spectroscopic analysis shows that hydroxy-dioxobenzothiazole is deprotonated upon binding when added in substoichiometric amounts to the complex at pH > pK a (Fig. 2). The hydrogen bond to the protonated His 181 in combination with the bifurcated weak hydrogen bond on-edge with Tyr 279 stabilizes the charge on oxygen atom O6.
In the nondissociated form HHDBT can exist as ortho-and para-quinone tautomers (38). It can be excluded that HHDBT is bound as ortho-quinone with its functional groups in the same orientation as stigmatellin, because the hydroxy group would be facing Glu 272 , but in the HHDBT structure this residue is rotated out of the binding pocket. There is also no indication that the binding site is occupied with a mixed population of tautomers, because all interacting residues show a defined orientation as judged from the clear cut electron density and B-factor distribution.
Stigmatellin (46) and HHDBT (Fig.4) are both competitive inhibitors. Kinetic studies show that stigmatellin is more tightly bound than UHDBT (44,47). Furthermore, stigmatellin binding raises the midpoint potential (E m ) of the Rieske protein by 250 mV The difference in polarization of the hydrogen bond to His 181 will add to the effect on the midpoint potential. The hydrogen bonding pattern to the negatively charged hydroxy group creates a lower electron withdrawing effect. Obviously, the tight binding interactions of stigmatellin involve the whole ligand, thus explaining why its binding is not pH dependent. In contrast, stabilization of the charged HHDBT is strongly dependent on the hydrogen bond from the protonated His 181 . This is in agreement with the higher affinity of hydroxy-dioxobenzothiazoles to the reduced Rieske protein independent of chain-length (7-15 carbon alkyl side chains), binding 15 times more strongly to the reduced Rieske protein (37). Experimental evidence and theoretical calculations suggest that reduction of the [2Fe-2S] cluster shifts the pK a of the cluster ligands, His 161 and His 181 , to values above 10, obviously favoring protonation of the histidines (49,50).
Consequently, pH dependence of HHDBT binding is related to the protonation state of a functional group within the protein, namely His 181 .
From structural analysis of the hydroxyquinone anion inhibitor HHDBT and stigmatellin binding at the Q o site, the following events for electron and proton transfer are deduced (Fig.7). Upon entry into the binding pocket the substrate ubiquinol is Link has proposed a sequential mechanism in which a stable semiquinone is formed and is anti-ferromagnetically coupled to the Rieske center until it is oxidized by heme b L . The concerted mechanism assumes that neither electron is transferred independently, but rather the semiquinone is so unstable that ubiquinol can not reduce the Rieske center unless the semiquinone reduces heme b L (17,19). In such a mechanism the concentration of ubisemiquinone is so low as to be almost nonexistent.
We suggest that stigmatellin and alkyl-hydroxy-dioxobenzothiazoles mimic either Glu 272 will accept a proton from ubiquinol or ubisemiquinone and consequently rotate towards heme b L as seen for HHDBT where it is not liganded to the carbonyl group. If a stable ubisemiquinone anion is formed, it is stabilized by localization of its negative charge on the oxygen atom interacting with protonated His 181 . After transfer of the second electron the product is no longer stabilized and will leave the binding pocket and give more mobility to Tyr 279 , thus breaking its hydrogen bond to the Rieske protein ( Fig. 7, Panel 6). Accordingly, the latter is found to be rotated into the binding pocket in the chicken and bovine structures with an empty Q o site (4,5).
The importance of Tyr 279 is supported by its full conservation in mitochondrial cytochrome b, and it is only replaced with a phenylalanine in a few chloroplast homologs (39). We propose that the on-edge weak hydrogen bonds from the aromatic side-chain of Tyr 279 are crucial for positioning ubiquinol in the active site. This is supported by recent mutagenesis studies in Rhodobacter sphaeroides where the conservative mutation of the homologous residue Tyr 303 to phenylalanine has no effect on enzyme activity, whereas exchange to Leu, Gly and Gln resulted in 3, 40, and 50 fold decrease, respectively (41).
There Our studies strongly support that the rotation of Glu 272 is an initial step for the release of the second proton as previously proposed (6,10,18). Glu 272 is completely conserved in mitochondrial cytochrome b (39) and the importance of the residue for proton transfer is indicated by mutagenesis studies as alteration to glutamine abolishes ubiquinol oxidation in Rhodobacter sphaeroides (40). Also, kinetic studies recently showed that protonation of a group with pK a of 5.7 blocked catalysis and this effect was assigned to displacement of Glu 272 upon HHDBT binding provide experimental evidence for the previously postulated proton transfer pathway (6). Glu 272 can deliver the proton directly to the heme propionate A via a single water molecule (Fig.6). The subsequent proton release is mediated by a hydrogen bonded water chain stabilised by cytochrome b residues: Arg 79 , Asn 256 , Glu 66 and Arg 70 .
Finally, the observed binding mode of the hydroxyquinone anion and the deduced proton transfer pathway as well as the demonstration of competitive inhibition by HDBT suggests a feasible catalytic mechanism, which is in line with a single occupancy model for ubiquinol oxidation.   Difference spectra were recorded as the inhibitor was added to one cuvette. Lowering THDBT concentration to sub-stoichiometric amounts shifts the absorbance maximum to 272 nm, indicating ionization of the bound inhibitor.