Hydrogen Bonding to the Substrate Is Not Required for Rieske Iron-Sulfur Protein Docking to the Quinol Oxidation Site of Complex III*

Complex III or the cytochrome (cyt) bc1 complex constitutes an integral part of the respiratory chain of most aerobic organisms and of the photosynthetic apparatus of anoxygenic purple bacteria. The function of cyt bc1 is to couple the reaction of electron transfer from ubiquinol to cytochrome c to proton pumping across the membrane. Mechanistically, the electron transfer reaction requires docking of its Rieske iron-sulfur protein (ISP) subunit to the quinol oxidation site (QP) of the complex. Formation of an H-bond between the ISP and the bound substrate was proposed to mediate the docking. Here we show that the binding of oxazolidinedione-type inhibitors famoxadone, jg144, and fenamidone induces docking of the ISP to the QP site in the absence of the H-bond formation both in mitochondrial and bacterial cyt bc1 complexes, demonstrating that ISP docking is independent of the proposed direct ISP-inhibitor interaction. The binding of oxazolidinedione-type inhibitors to cyt bc1 of different species reveals a toxophore that appears to interact optimally with residues in the QP site. The effect of modifications or additions to the toxophore on the binding to cyt bc1 from different species could not be predicted from structure-based sequence alignments, as demonstrated by the altered binding mode of famoxadone to bacterial cyt bc1.

The ubiquinol-cytochrome c oxidoreductase, also known as complex III of mitochondrial respiratory chain or cytochrome (cyt) 3 bc 1 complex, catalyzes the reaction of electron transfer (ET) from ubiquinol (QH 2 ) to cyt c and concomitantly translocates protons across the inner membrane of mitochondria or the plasma membrane of photosynthetic purple bacteria, contributing to a cross-membrane potential important for cellular function (1,2). Although catalyzing the same enzymatic reaction, cyt bc 1 complexes isolated from different organisms have very different subunit compositions. Prokaryotic bc 1 complexes often consist of 3-4 subunits, whereas mitochondrial enzymes have 10 -11 different subunits (1). Nevertheless, only three subunits are essential for the ET function ( Fig. 1A): cyt b, cyt c 1 , and the Rieske iron-sulfur protein (ISP) (3). The cyt b subunit contains two b-type hemes, b L and b H , for the low and high potential hemes, respectively, and is entirely embedded in the membrane with eight transmembrane (TM) helices (4). The cyt c 1 subunit, anchored to the membrane by a single TM helix, has a c-type heme covalently attached to its active domain that is located in the intermembrane space of mitochondria or periplasm in bacteria. The ISP subunit features an integrated 2Fe-2S cluster in its extrinsic domain (ISP-ED) that is on the same side as the cyt c 1 subunit and also anchored to the membrane by a helix.
The ET-coupled proton translocation function, as modeled by the Q-cycle mechanism (Fig. 1B), requires two active sites: a QH 2 oxidation (Q P ) site and a ubiquinone (Q) reduction (Q N ) site (5)(6)(7), which were shown to exist by crystal structures of mitochondrial and bacterial bc 1 complexes in complexes with various bc 1 -specific inhibitors (4, 8 -12). A characteristic feature of the Q-cycle mechanism is the bifurcated ET pathway at the Q P site, in which the two electrons of the substrate QH 2 take two separate routes (Fig. 1B): one takes a high potential route going from ISP to cyt c 1 and to substrate cyt c, and the other follows the low potential route traveling to the b L and b H hemes sequentially, ending in ubiquinone/ubisemiquinone bound at the Q N site. Strikingly, structural studies of bovine mitochondrial cyt bc 1 (Btbc 1 ) in complex with various respiratory inhibitors revealed the inhibitor-type dependent conformation switch of ISP-ED. This observation not only suggested a mechanism for the QH 2 oxidation at the Q P site but also offered a means for inhibitor classification (13)(14)(15). Thus, inhibitors that immobilize the ISP-ED to the Q P site are called P f -type inhibitors, and those that mobilize it are termed P m -type inhibitors (13).
High resolution structural studies of cyt bc 1 have so far failed to show the binding of substrate QH 2 at the Q P site. Studies of cyt bc 1 in the presence of inhibitors that are substrate QH 2 homologs such as stigmatellin (4, 11-13, 16, 17), 3-Undecyl-2hydroxydioxobenzothiazol (13), 5-n-heptyl-6-hydroxy-4,7-dioxobenzothiazole (18), and 2-nonyl-4-hydroxyquinoline N-oxide (13) revealed conformational immobilization or docking of the ISP-ED to the Q P site and a hydrogen bond (H-bond, 2.8 -FIGURE 1. Structure and mechanism of cyt bc 1 complex. A, ribbon representation of the structure of dimeric cyt bc 1 in complex with famoxadone from photosynthetic bacterium R. sphaeroides. The cyt b subunit is shown in green, the cyt c 1 is in blue, and ISP is in yellow. Heme groups are shown as the ball and stick models with carbon in black, nitrogen in blue, and oxygen in red; they are labeled in black as b H , b L , and c 1 , respectively. The iron-sulfur cluster (FeS) of ISP is shown as spheres with Fe in orange-brown and S in green. The two parallel horizontal lines represent the cytoplasm membrane bilayer. The bound famoxadone (Fam) at the Q P site is shown as the CPK model with carbon in orange, oxygen in red, and nitrogen in blue. One detergent molecule (␤-OG) bound at cyt c 1 is drawn as a brown/red ball and stick model. A bound strontium ion (Sr 2ϩ ) bound to cyt c 1 is also shown together with its amino acid ligands Asp 8 , Glu 14 , and Glu 129 in the stick models. B, Q-cycle mechanism. The cyt bc 1 consists of two reaction sites: Q P and Q N sites. The Q P site is near the P side, where the two electrons of QH 2 diverge. The first electron goes to the high potential chain via the ISP and cyt c 1 , ending in cyt c 2 (cyt c in mitochondrial cyt bc 1 ), which can be interrupted by stigmatellin. The second electron enters the low potential chain via hemes b L and b H , end in substrate Q or radical Q ⅐ bound at the Q N site. The low potential path can be inhibited by myxothiazol at the Q P site and antimycin at the Q N site.
3.4 Å depending on the structure) between the histidine ligand to the 2Fe-2S cluster of ISP (His 141 , bovine sequence) and oxygen moieties of the inhibitor, leading to the proposal that docking of the ISP-ED to the Q P site is likely the consequence of this H-bond (11,19,20). Despite evidence that suggests the docking of the ISP-ED to the Q P site may be independent of the H-bond (21), systematic study is lacking.
The dire consequence of obstructing cellular respiration has made cyt bc 1 one of the most frequent targets of antibiotics. However, allelopathic inhibition has been observed as a strategy employed by many organisms to gain survival advantage (22). Consequently, toxin-producing organisms are intrinsically resistant to their own toxin. Although many mechanisms may be at work to render this intrinsic resistance, changes in target structures by amino acid substitutions are believed to contribute significantly to the phenomenon, leading to amino acid sequence diversification at the site of inhibition across organisms. Inhibitors of bc 1 have been used not only for the purpose of disease control but also in studies of the mechanisms of bc 1 function, inhibition, and resistance (13,14,(23)(24)(25)(26)(27)(28). Famoxadone (5-methyl-5-(4-phenoxy-phenyl)-3-phenylamino-2,4-oxazolidinedione) ( Fig. 2A) is an oxazolidinedione-type Q P site inhibitor (29). It has reportedly displayed various IC 50 values for submitochondrial particles from different species (30). However, evidence is mounting that some fungal species are intrinsically more resistant (30,31), and field applications of famoxadone for fungal disease control have led to rapid development of resistance that renders the inhibitor ineffective (32)(33)(34). How inhibitors such as famoxadone react to fluidity in its binding environment has not been fully explored experimentally, let alone understood.
It is known that the ET between the ISP and cyt c 1 requires movement of the ISP-ED, which could be controlled, as we proposed earlier (14), by the bimodal conformation switch of the ISP-ED. Thus, the question regarding the involvement of the H-bond in this control process needs to be addressed. In the present study, we used three oxazolidinedione-type inhibitors to form complexes with either mitochondrial or bacterial cyt bc 1 or both and showed that all three inhibitors can induce docking of the ISP-ED at the Q P site without forming a direct H-bond with the ISP, which is consistent with the notion that the ISP conformation switch does not require a direct H-bond. We established the structural basis for the toxophore of oxazolidinedione-type inhibitors and showed the effect of alterations to the toxophore on inhibitor binding to the cyt bc 1 , both structurally and biochemically. These observations have strong implications concerning the development of drugs that are designed to target a broad spectrum of pathogens.

Results
Differential Inhibition of Mitochondrial and Bacterial cyt bc 1 Complexes by Famoxadone-Previously, we reported that binding of famoxadone to Btbc 1 induces docking of the ISP to the Q P site (21). Unlike other Q P site inhibitors such as stigmatellin and 3-Undecyl-2-hydroxydioxobenzothiazol, famoxadone does not make a direct H-bond with the ISP-ED because the shortest distance between His 141 of the ISP and the phenylamino moiety of famoxadone is 6.1 Å (Fig. 3A). However, whether the fixation of the ISP-ED by famoxadone binding can be interpreted as part of the Q-cycle mechanism remains unclear. Thus, to understand this phenomenon, one would have to investigate the effect of famoxadone on cyt bc 1 of other organisms, which are evolutionarily remote from mammals or yeast, such as bacteria. Doubts that the capture of the ISP-ED induced by famoxadone binding is indeed part of a conserved mechanism were initially cast by the observation that the experimentally determined IC 50 values of 1.4 and 418 nM of isolated cyt bc 1 from photosynthetic bacterium Rhodobacter sphaeroides (Rsbc 1 ) and mitochondrial cyt bc 1 (Btbc 1 ), respectively, differ by a factor of nearly 300 (Fig. 4A). Further differences were noticed in the shapes of the inhibition curves for Rsbc 1 and Btbc 1 , suggesting that Rsbc 1 might exhibit a different set of binding interactions with famoxadone.
Paradoxically, the difference in IC 50 values is not supported by the apparent conservation in the Q P site, because all residues in contact with famoxadone in Btbc 1 are conserved, except for two residues: Phe 276 and Ala 277 (Pro 300 and Phe 301 in Rsbc 1, Fig.  5). The closest distances between these two Btbc 1 residues and bound famoxadone are 4.53 and 3.72 Å, respectively (Fig. 3A). Judging by the alignment between Btbc 1 (PDB code 1L0L) and Rsbc 1 (PDB code 2FYN) structures, it seems unlikely that these two amino acid substitutions would result in a significant change in the binding of famoxadone, because modeling exercises indicate residue Phe 301 in Rsbc 1 appears to be in a good position to replace the function of Phe 276 in Btbc 1 .
Distinct Conformation of Famoxadone as the Q P Site Occupant of Rsbc 1 -To answer questions about the binding mode of the inhibitor famoxadone and its local and global effect, we determined the crystal structure of Rsbc 1 in complex with famoxadone ( Fig. 1A, Table 1). There are four copies of cyt bc 1 in the Rsbc 1 crystal per asymmetric unit, each binding one mol- ecule of famoxadone. Even though no non-crystallographic symmetry (NCS) restraints were applied to the inhibitor during refinement, the four famoxadone molecules superpose within 0.198 Å root mean square (RMS) deviation, allowing our analysis to focus on just one molecule.
The structure of the Rsbc 1 ⅐famoxadone complex reveals that famoxadone binds in the same location as in Btbc 1 (Fig. 3B). Like Btbc 1 , the chiral environment of the Q P site of Rsbc 1 selects the S-(Ϫ)-isomer of famoxadone for binding. Most importantly for mechanistic implications, the ISP-ED is found docking in the Q P position, as evidenced by the strong peak of the anomalous difference Fourier density at the location of the 2Fe-2S cluster of ISP. At the same time, residue His 152 (ligand to the 2Fe-2S cluster in Rsbc 1 and equivalent of His 141 of Btbc 1 ) has a distance of 5.7 Å to the phenylamino group of famoxadone. This result confirms that recruitment of the ISP to the Q P site does not depend on the H-bond formation between the ISP-ED and the inhibitor.
As in the Btbc 1 ⅐famoxadone structure, the inhibitor bound to Rsbc 1 is capable of forming an H-bond between atom O6 and the backbone amide nitrogen atom of Glu 295 over a distance of 2.69 Å and between atom N1 and the conserved water molecule (W1) (Fig. 3B). Unexpectedly, the terminal phenoxy group adopts a position essentially opposite to that found in Btbc 1 ⅐famoxadone, which is quantifiable by a change in torsion angle (C12-C11-O14-C15) from Ϫ116°to ϩ145° (Fig. 6A). However, this difference in positions of the phenoxy group is consistent with the observation that the residues lining the entrance to the Q P pocket are quite variable between cyt b subunits of different organisms. The adaptation of the terminal phenoxy group to a different environment appears to be based on both exclusion and attraction. The exclusion is caused by the substitution of Ala 277 of Btbc 1 with Phe 301 in Rsbc 1 (Fig. 5), which would not allow the phenoxy group to be in the same position as in Btbc 1 . Being flexible to swing around, the aromatic phenoxy group of famoxadone finds stabilization by aromatic attractions from Phe 166 , Phe 337 , and Phe 144 (Figs. 3B and 5A), which correspond to Ile 150 , Ala 295 , and Phe 128 in the bovine sequence.
Famoxadone has been noted to prefer aromatic environments (21), which is demonstrated most convincingly by the 13-fold increase in IC 50 value for the F129L mutation in yeast (30). Thus, the combined effect of the naturally occurring aromatic residues Phe 166 and Phe 337 in the cyt b subunit of Rsbc 1 (as opposed to the respective residues Leu 150 and Ala 295 in Btbc 1 ) might cause the stabilization of famoxadone in the present conformation and explain the difference in IC 50 values compared with Btbc 1 (Fig. 4A).
Mapping the interactions between Q P site residues and bound famoxadone, it became immediately apparent that Rsbc 1 uses an overlapping but non-identical set of residues from Btbc 1 for famoxadone binding (Fig. 5). Most notably, Rsbc 1 uses Phe 166 , Met 336 , and Phe 337 to engage famoxadone, whereas the equivalent residues Leu 150 , Leu 294 , and Ala 295 in Btbc 1 remain uninvolved with the inhibitor. Conversely, not all interactions observed in Btbc 1 are also found in the Rsbc 1 ⅐famoxadone complex, as seen in residues Phe 91 , Tyr 95 , Phe 276 , Ala 277 , and Ile 298 in Btbc 1 (Phe 105 , Tyr 109 , Pro 300 , Phe 301 , and Ile 340 in Rsbc 1 ) (Fig.  5). Thus, these structural observations qualitatively explain the differences in IC 50 values for famoxadone binding to Btbc 1 and Rsbc 1 , which could not be predicted based on sequence alignment.
Oxazolidinedione-type Inhibitors Induce Docking of the ISP-ED to the Q P Site without Direct H-bonding to the ISP and Inhibitor-bound Btbc 1 Structures Reveal a Conserved Corebinding Motif-Superposition of the cyt b subunits of Rsbc 1 and Btbc 1 complexed with famoxadone shows a good structural alignment of the oxazolidinedione core of famoxadone and its two directly connected aromatic rings (Fig. 6A), which raises the question as to whether the difference in binding affinity of famoxadone between Rsbc 1 and Btbc 1 is entirely due to the difference in interaction with the terminal phenoxy group. To evaluate the amount of binding energy contributed by various portions of famoxadone other than the phenoxyl group, we measured the binding of two additional oxazolidindione-type compounds, 5-methyl-5-(4,6-difluorophenyl)-3-phenylamino-2,4-oxazolidinedione (jg144) and 5-methyl-2-(methylsulfanyl)-5-phenyl-3-(phenylamino)-3,5-dihydro-4H-imidazol-4-one (fenamidone), to both Btbc 1 and Rsbc 1 (Fig. 4, B and C). The chemical structures of jg144 and fenamidone closely resemble that of famoxadone except that they do not have the conformationally flexible terminal phenoxy group (Fig. 2, B and C). Although jg144 replaces the phenoxy group and the hydrogen atom H9 of famoxadone each with fluorine, fenamidone features a substituted imidazole ring in place of the oxazolidinedione ring.
The IC 50 values for jg144 inhibition were determined to be 1.1 and 1.2 nM, respectively, for Rsbc 1 and Btbc 1 (Fig. 4B). Similarly, the IC 50 values for fenamidone were 3.5 and 8.3 nM, respectively, for Rsbc 1 and Btbc 1 (Fig. 4C). More importantly, the profiles of the jg144 and fenamidone inhibition curves are very similar regardless of whether they bind to Rsbc 1 or Btbc 1 , indicating that similar interactions are employed for their binding.
We determined and refined the structure for the complex Btbc 1 ⅐fenamidone at 2.65 Å resolutions (Table 1) and reanalyzed the structure of Btbc 1 ⅐jg144 (PDB code 2FYU) at 2.25 Å resolution. In both cases, their respective ISP-EDs are docked at the Q P site, as indicated by the high anomalous peak heights of the 2Fe-2S clusters ( Table 2). The arrest of the ISP-ED is a Binding of cyt bc 1 by Oxazolidinedione-type Inhibitors NOVEMBER 25, 2016 • VOLUME 291 • NUMBER 48 characteristic trait of P f -type inhibitors (13), and here the observed shortest distances between His 141 of the ISP and the bound inhibitors are 6.34 and 6.74 Å, respectively, for jg144 and fenamidone (Fig. 3, C and D, Table 2). Thus, no direct H-bond is formed between the ISP and the inhibitors.
The structures of both Btbc 1 ⅐jg144 and Btbc 1 ⅐fenamidone showed no significant differences in side chain rotamers from Btbc 1 ⅐famoxadone (Fig. 6B). This in turn suggests that the terminal phenoxy group of famoxadone is capable of sampling the environment around the entrance of the Q P site without perturbing any side chains in the Q p site. The consistency of interactions and the measurements showing little difference in IC 50 values between Btbc 1 and Rsbc 1 for jg144 and fenamidone firmly establish the structural basis for the three-ring unit as the toxophore of the oxazolidinone-type inhibitors (29).

Discussion
Mechanistic Implications Concerning the Bifurcated Electron Transfer at the Q P Site-Quinol oxidation at the Q P site requires strict, consecutive bifurcated ET steps that are subject to a high degree of control, an idea that has received support over the years by a substantial amount of experimental evidence and features a bimodal ISP-ED conformation switch at its core (14,15). Although most structural data on this switch were first obtained from Btbc 1 , structural information of several avian mitochondrial cyt bc 1 (Ggbc 1 ) complexes recently made public in the Protein Data Bank (PDB codes 3L74 and 3L73) is in perfect agreement with the former. However, the conformational switch of the ISP-ED in bacterial bc 1 has been difficult to observe crystallographically. This is because all structures of bacterial cyt bc 1 complexes obtained so far were crystallized with bound stigmatellin, and it has been difficult to crystallize them in a different P f -type inhibitor such as famoxadone. Conceivably, crystallizing bacterial cyt bc 1 bound with a P m -type inhibitor would be even more difficult. Prior to this work, complexes of bacterial bc 1 were crystallized only in the presence of the P f -type inhibitor stigmatellin with the ISP-ED immobilized at the Q P position (4,12,17). In this work, we show for the first time that another P f -type inhibitor, namely famoxadone, is able to fix the conformation of ISP-ED without providing a direct H-bond to the ISP subunit, which allows successful crystallization of Rsbc 1 .
Quinol oxidation at the Q P site is initiated with the recruitment to and/or stabilization and proper alignment of the ISP-ED at the Q P site, which was proposed to take place via a direct H-bond between one of the histidine ligands to the 2Fe-2S cluster of ISP (His 141 in Btbc 1 and His 152 in Rsbc 1 ) and the hydroxyl group of its substrate ubiquinol. This proposal was based on the observation that in high resolution crystal structures, polar atoms from the enzyme's ISP-ED and the inhibitor (specifically stigmatellin and 5-n-heptyl-6-hydroxy-4,7-dioxobenzothiazole (11,19)) appeared within H-bond distance. Contrary to this view, structural studies with Btbc 1 demonstrated that famoxadone is capable of immobilizing the ISP-ED at the Q P site in the absence of a direct H-bond to the ISP (13,21). The question becomes whether the presence or absence of the H-bond can support the required bimodal conformation switch of the ISP-ED. The bifurcated ET at the Q P site begins with the binding of substrate QH 2 , which also triggers the docking of ISP-ED. An H-bond forms between a hydroxyl group of QH 2 and a histidine of the ISP-ED. After deprotonation, the QH 2 transfers its first electron to the 2Fe-2S cluster of the ISP, which is an energetically favorable process. The second electron of QH 2 , however, has to travel to the Q N site via the hemes, b L and b H , eventually reducing a ubiquinone (Q) or ubisemiquinone (Q ⅐ ) molecule (Fig. 1B). It is important to emphasize (i) that to avoid short circuit reactions, the 2Fe-2S cluster cannot be allowed to leave the Q P site and return oxidized until the product has been completely removed and/or replaced by QH 2 and (ii) that although the ISP-ED remains at the Q P site, the H-bond should always exist between the reduced ISP and the product ubiquinone (Q) after the completion of the two-electron transfer reaction (Fig. 7). In other words, the H-bond between the substrate and the ISP cannot control the conformation switch of the ISP-ED. Only the residues lining the Q P pocket are shown. Blue rectangles placed above sequences are part of helices and are labeled accordingly. Residues whose mutations are known to confer fungicide resistance are in red, and the fungicides they are resistant to are indicated below the alignment. FaR is short for famoxadone resistance, and AzR is for azoxystrobin resistance. Residues that are in direct contact with famoxadone in Btbc 1 are highlighted yellow. Famoxadone-contacting residues in the Rsbc 1 are shaded in blue. Differences important for famoxadone binding in avian bc 1 are highlighted in magenta.

TABLE 1
Statistics on the quality of diffraction data sets of cyt bc 1 crystals and structural models fen, fenamidone; fam, famoxadone; stg, stigmatellin; ant, antimycin. In this work, we present evidence from the famoxadonebound Rsbc 1 structure and two Btbc 1 structures with bound oxazolidinedione-type inhibitors, jg144 and fenamidone, showing that docking of the ISP-ED to the Q P site, both in mitochondrial and bacterial cyt bc 1 , does not require a direct H-bond between the ISP and the inhibitor. This lends strong support for the hypothesis that electron bifurcation at the Q P site is achieved through a controlled bimodal conformational change of the ISP-ED (14,15).

Crystal Contacts Do Not Significantly Influence the Conformation of ISP-ED-
The head domain of ISP in Rsbc 1 is involved in crystal-packing contacts, whereas that of Btbc 1 is not. This observation provides a good opportunity to qualitatively assess how much influence crystal contacts have on the conformation of the intrinsically moving domain. For the purpose of an unbiased comparison, we also determined the structure of the doubly inhibited (stigmatellin and antimycin) ternary complex of Rsbc 1 (Table 1). Antimycin is a Q N site inhibitor and is not known to influence ISP-ED conformation; thus, its presence is of no consequence for this analysis. Stigmatellin inhibited Rsbc 1 has been structurally characterized before (4). However, the newly determined structure that we are using as a reference was obtained from the same batch of protein under the same conditions, and importantly, the crystals features the same space group P1, the same number of molecules per asymmetric unit (two dimers) and has the same cell dimensions and crystal packing. To quantify how much the ISP-ED moves in response to a change in the Q P site occupant, we superposed only the C␣ atoms of the cyt b subunits of the different complexes (stigmatellin versus famoxadone) and calculated the respective RMS deviations of the head domain of the ISP, cyt b, and cyt c 1 subunits (Table 3). Although the C␣ traces of cyt b superimpose very well (Ͻ0.4 Å), the traces of the ISP-ED no longer coincide: they display RMS deviations as large as 2.55 Å (139 C␣) and 2.86 Å (124 C␣), respectively, for Rsbc 1 and Btbc 1 . This is significant, because the corresponding deviations of the head domain of cyt c 1 (a domain that is spatially close to ISP-ED and located in the same aqueous phase) amounted to an RMS deviation of only 0.60 Å (Rsbc 1 ) and 0.93 Å (Btbc 1 ), respectively. It is worth noting that the RMS deviations of ISP-ED in Rsbc 1 and Btbc 1 are nearly the same despite the fact that the ISP-ED of the Rsbc 1 is clearly involved in crystal contacts, unlike the ISP-ED in Btbc 1 , which has none. The similarities in RMS deviations between Rsbc 1 and Btbc 1 suggest that the crystal contacts in Rsbc 1 play only a secondary role in determining the final conformation of the ISP-ED once ISP-ED is docked at the Q P site.
The angular displacement of the ISP-EDs between Btbc 1 and Rsbc 1 when bound with different inhibitors can also be visualized using vectors (Fig. 8A), which is calculated based on superposition of cyt b subunits only between the structures with bound antimycin and famoxadone. The red vectors in the ISP trace (yellow trace) of the stigmatellin-inhibited Rsbc 1 represent the connection to the C␣ atoms of the famoxadone-inhibited complex. The blue vectors represent the equivalent repositioning of the C␣ atoms of the ISP-ED in Btbc 1 . This vector diagram illustrates qualitatively that the amount of change (vector length) is approximately the same in both sets. However, the direction in which the ISP-ED is displaced in Rsbc 1 , FIGURE 7. H-bond between substrate and histidine ligand of 2Fe-2S cluster before and after ET reaction at the Q P site.  1 5.73 1.0 P f a The peaks are normalized against anomalous signals of their respective b H heme irons. b Anomalous signals were calculated using diffraction data in the range of 20 -5 Å. c These numbers were taken from Esser et al. (13) for comparison. d These two inhibitors fix ISP-ED to the b site allowing Rsbc 1 to be crystallized.
However, because of crystal contact, the anomalous peak heights are not considered reflecting a true state of equilibration in solution. versus Btbc 1 , is characterized by an angle of 58°(represented by half-transparent triangles). It is conceivable that, as a consequence of packing forces in Rsbc 1 crystals, the ISP-ED undergoes an angular adjustment but importantly maintains the proper distance from the Q P site of cyt b. Implications Concerning the Design and Use of Broad Spectrum Antifungal Agents-Despite considerable pressure that must exist to maintain the intricate function of cyt b and its active sites, several residues, individually or in groups, may have changed in the course of evolution. These changes are seen as the root cause of intrinsic drug resistance by target site mutations. By determining the structure of Rsbc 1 in complex with famoxadone and comparing it with that of Btbc 1 , we uncovered the structural basis for the effects of sequence polymorphisms on the binding of inhibitors to cyt bc 1 of different organisms, which were manifested by varied efficacies of inhibitors against cyt bc 1 complexes. This structural observation is consistent with measurements of IC 50 values, indicating that famoxadone inhibits Rsbc 1 more strongly than Btbc 1 (Fig. 4A); it agrees also with the binding energy of Ϫ12.0 kcal/mol and Ϫ12.8 kcal/mol, respectively, estimated by molecular docking experiments for mitochondrial and bacterial cyt bc 1 complexes. The conformation of famoxadone in the binding pocket shows sensitivity to the environment and is characterized by a tendency to optimize Ar-Ar interactions (21).
Because of the absence of an apo Rsbc 1 structure, the only comparisons that can be made are between complexes of Rsbc 1 with famoxadone and stigmatellin (Fig. 8B) and the equivalent pair of Btbc 1 complexes (Fig. 8C). The comparison made it immediately clear that the accommodation of famoxadone in the Q P pocket requires adjustments of torsion angles in both the inhibitors and cyt b residues. Direct steric interactions force Phe 298 (Phe 274 in Btbc 1 ) to adopt two distinct conformations depending on the binding of either stigmatellin or famoxadone in both Rsbc 1 and Btbc 1 . Similarly pronounced are changes in the positions of Glu 295 (Glu 271 in Btbc 1 ) in response to the inhibitors. In the Rsbc 1 ⅐famoxadone structure, Glu 295 returns to the solvent channel or to the "native" position as observed in Btbc 1 structures because of the loss of a hydrogen bond to the inhibitor (8).
From the Rsbc 1 structure, Phe 166 , Met 336 , and Phe 337 (Leu 150 , Leu 294 , and Ala 295 in Btbc 1 ) appear to play a major role in repositioning the terminal phenoxy moiety of famoxadone. Although this is convincing based on qualitative energy considerations, the situation seems subtler in the recently published Ggbc 1 structure with famoxadone (35), showing the same moiety in an orientation nearly identical to the one found in Rsbc 1 but in contrast to that in Btbc 1 . Despite the higher sequence identity between Ggbc 1 and Btbc 1 (Fig. 5), residue Phe 151 in the avian Gg cyt b is also a phenylalanine, as in Rsbc 1 (Rsbc 1 Phe 166 ). However, residue Phe 337 of Rs cyt b coincides with the non-aromatic Ala 296 in Gg cyt b, implying that a single change in the amino acid composition (Leu 150 Phe) at the entrance to the Q P pocket is sufficient to stabilize an alternate conformation of famoxadone to the one that was first found in Btbc 1 . This observation underscores the importance of Ar-Ar interaction in inhibitor binding.
The fact that famoxadone is able to adapt to the different microenvironments of the target sites of a wide range of organisms suggests two important principles for designs of broad spectrum inhibitors: one is the flexibility to adopt different conformations as needed, and the other is the ability to fully engage in Ar-Ar interactions. A single pair of Ar-Ar interactions could provide as much as 1.3 kcal/mol of free energy (36), which can be exploited either for the enhancement of specificity or for the evasion of resistance and is also consistent with previous observations that the phenoxyphenyl group of famoxadone is most tolerant to variations (29).

Experimental Procedures
Expression of ⌬-sub IV Rsbc 1 -The purification procedure for Rsbc 1 , published earlier (37), was used with minor changes to obtain protein for crystallization. Briefly, chromatophore membranes were prepared from cells harboring the plasmid (pRKDfbcFBC H ) coding for ⌬-sub IV wild type Rsbc 1 protein, by disrupting cells with a French press followed by differential centrifugations. To purify His 6 -tagged Rsbc 1 , the chromatophore suspensions were adjusted to a cyt b concentration of 25 M with 50 mM Tris-Cl (pH 8.0 at 4°C), containing 20% glycerol, 1 mM MgSO 4 , and 1 mM PMSF. Dodecyl-D-maltopyranoside (DDM) solution (10% w/v) was added dropwise to a final detergent to protein ratio of 0.57 (mg/mg). After centrifugation, the supernatant was loaded on a nickel-nitrilotriacetic acid column, which was washed with 6 column volumes of buffer A (50 mM Tris-HCl, pH 8.0, at 4°C, 200 mM NaCl, 0.01% DDM), 6 column volumes of buffer A in the presence of 5 mM histidine, and 4 column volumes of buffer B (50 mM Tris-HCl, pH 8.0, at 4°C, 200 mM NaCl, 0.5% ␤-OG) in the presence of 5 mM histidine. The desired protein fractions were eluted with buffer B containing 200 mM histidine and concentrated with Centriprep-50 to a final concentration of 300 M.
Crystallization of Btbc 1 in Complex with Inhibitors-The final concentration of purified Btbc 1 complex was adjusted to 20 mg/ml in a solution containing 50 mM MOPS buffer at pH 7.2, 20 mM ammonium acetate, 20% (w/v) glycerol, and 0.16% sucrose monocaprate. This solution was further incubated with a 5ϫ molar excess amount of the desired inhibitor and set up for crystallization as described in previous publications (8,21). Crystals were treated with a cryoprotectant in 5% increments until the final glycerol concentration reached 42% and then cryo-cooled in liquid propane. Crystallization of Famoxadone-inhibited Rsbc 1 -A solution of ⌬-sub IV Rsbc 1 (57 mg/ml) in a buffer containing 0.5% ␤-OG was treated overnight with a 3-fold molar excess of famoxadone (Chem. Service Inc.). The solution was diluted 4-fold with buffer consisting of 50 mM Tris-HCl, pH 8.0, 0.3% ␤-OG, 200 mM histidine, 150 mM NaCl, 10 mM sodium ascorbate, and 10% glycerol. The resulting solution of 14 mg/ml Rsbc 1 in 0.35% ␤-OG was augmented with sucrose monocaprate (0.06%, ϳ0.5 critical micelle concentration) and 10 mM strontium nitrate. PEG 400, added to a final concentration of 7%, served as the precipitant. The solution was allowed to stand overnight at 4°C. Any precipitate was centrifuged off, and 5 l of the supernatant was used in sitting drop vapor diffusion crystallization trays over 1 ml of 100 mM Tris-HCl, pH 8.0, 20% glycerol, 600 mM NaCl, 26% PEG 400, and 5 mM NaN 3 . After several weeks, small red crystals appeared which froze cleanly in liquid propane without requiring additional cryoprotectants.
Diffraction Data Collection, Structure Determination, and Refinement-Crystals of the tetragonal Btbc 1 ⅐fenamidone and the triclinic famoxadone-inhibited Rsbc 1 were stable when cryo-cooled to 100 K, permitting several hours of data collection at the SER-CAT Beamline (ID22) of the Advanced Photon Source, Argonne National Lab. Raw diffraction frames were processed with the program package HKL2000 (38). A diffraction data set for the Btbc 1 crystal was phased with coordinates of the 11-subunit apo structure of Btbc 1 (PDB code 1NTM) (23). The program REFMAC (39) was employed for initial rigid body refinement followed by iterative maximum likelihood and TLS (Translation, Libration, and Screw tensor) refinement. In between REFMAC runs, sigma A (40) weighted 2F o Ϫ F c and F o Ϫ F c Fourier maps were calculated and used to identify and build bound ligand and solvent molecules manually with the program Coot (41). The data set for the Rsbc 1 was phased with molecular replacement using an edited version of the Rsbc 1 dimer (PDB code 2QJY) in MolRep (CCP4) (42), producing a solution with two dimers/unit cell. Care was taken in the subsequent refinement to allow the head domain of the ISP to adjust to a position possibly different from that in stigmatellin-inhibited complexes. Rigid body refinement, simulated annealing, positional, and atomic displacement (ADP, TLS) refinement were carried out with phenix.refine 1.8 -1512 (43). 4-fold NCS restraints were applied throughout the refinement. However, no NCS restraints were imposed on the four famoxadone molecules during refinement, yet they are superposed within 0.2 Å RMS deviation (average from mean structure), allowing our analysis to focus on just one molecule. Manual adjustments of the coordinates were carried out in O (44) and Coot 0.6. As the last step, famoxadone was fitted into the difference density. Final refinement statistics are given in Table 2.
Measurement of IC 50 Values for bc 1 Inhibitors-The activities of Mtbc 1 and Rsbc 1 were assayed following the reduction of substrate cyt c. The bc 1 preparation was diluted to a final concentration of 0.1 and 1 M, respectively, for Btbc 1 and Rsbc 1 in a buffer containing 50 mM Tris-HCl, pH 8.0, 0.01% ␤-DDM, and 200 mM NaCl. To a 2-ml assay mixture containing 100 mM phosphate buffer, pH 7.4, 0.3 mM EDTA, and 80 M cyt c, Q 0 C 10 BrH 2 was added to a final concentration of 25 M, and the solution was split evenly into two cuvettes. To one cuvette, 5 l of diluted bc 1 solution was added, and we immediately began recording the cyt c reduction at 550-nm wavelengths for 100 s in a two-beam Shimadzu UV-2250 PC spectrophotometer at room temperature. The amount of cyt c reduced at a given period of time was calculated using a millimolar extinction coefficient of 18.5 mM Ϫ1 cm Ϫ1 . To measure the effect of bc 1 inhibitors, bc 1 was preincubated with various concentrations of inhibitors for 5 min on ice prior to the measurement of bc 1 activity. The IC 50 value for each inhibitor was calculated by a least squares procedure fitting the equation (Y ϭ A min ϩ (A max Ϫ A min )/(1 ϩ 10 (X-logIC50) )) implemented in a commercial package Prism, where A max and A min are maximal and minimal activities, respectively.
Author Contributions-D. X. conceived and coordinated this study and wrote most of this paper. L. E. wrote significant parts of this paper, prepared Figs. 1 and 8, and crystallized and solved the structures of RS⅐fam and RS⅐stg/an. F. Z. expressed and purified R. sphaeroides bc 1 needed in this study and performed the assays in Fig. 4. Y. Z., Y. X., and Z. Q. contributed to the crystallization and structure determination of Btbc 1 ⅐fenamidone complex. W.-K. T. advised on the interpretation of the kinetics of the binding studies. C.-A. Y. contributed to the discussion and interpretation of the effect of inhibitors in bc 1 .