Structure of the cytochrome aa₃-600 heme-copper menaquinol oxidase bound to inhibitor HQNO shows TM0 is part of the quinol binding site

Significance

The proton-pumping quinol oxidases, close orthologs of the cytochrome c oxidases, directly oxidize membrane-bound quinols to reduce oxygen to water. As one of the major contributors of proton-motive force, they exist widely in bacteria, including many pathogens. Here, we describe the structure of a quinol oxidase that clearly defines the protein features that have evolved to allow oxidation of a membrane-bound quinol as a substrate. TM0, a transmembrane helix in subunit I that is not present in the related cytochrome c oxidases, forms part of a cleft that would accommodate the menaquinol-7 substrate. Cocrystal structure with the quinol-analog inhibitor HQNO (N-oxo-2-heptyl-4-hydroxyquinoline) or 3-iodo-HQNO reveals a single binding site where the catalytic intermediate binds.

Abstract

Virtually all proton-pumping terminal respiratory oxygen reductases are members of the heme-copper oxidoreductase superfamily. Most of these enzymes use reduced cytochrome c as a source of electrons, but a group of enzymes have evolved to directly oxidize membrane-bound quinols, usually menaquinol or ubiquinol. All of the quinol oxidases have an additional transmembrane helix (TM0) in subunit I that is not present in the related cytochrome c oxidases. The current work reports the 3.6-Å-resolution X-ray structure of the cytochrome aa₃-600 menaquinol oxidase from Bacillus subtilis containing 1 equivalent of menaquinone. The structure shows that TM0 forms part of a cleft to accommodate the menaquinol-7 substrate. Crystals which have been soaked with the quinol-analog inhibitor HQNO (N-oxo-2-heptyl-4-hydroxyquinoline) or 3-iodo-HQNO reveal a single binding site where the inhibitor forms hydrogen bonds to amino acid residues shown previously by spectroscopic methods to interact with the semiquinone state of menaquinone, a catalytic intermediate.

The mitochondrial aerobic respiratory chains of virtually all aerobic eukaryotes terminate with a proton-pumping heme-copper oxygen reductase, cytochrome aa₃, which is an aa₃-type cytochrome c oxidase (1). In mammals, the enzyme contains 13 subunits, 3 of which (subunits I, II, and III) are encoded in the mitochondrion and 10 in the nucleus (2). Most aerobic prokaryotic organisms also contain a proton-pumping heme-copper oxygen reductase as part of the respiratory chain, but these enzymes typically contain fewer subunits (2 to 4) (1, 3). All heme-copper oxygen reductases contain a homolog of subunit I, which usually contains 12 transmembrane (TM) helices, along with the binding sites for a low-spin heme (e.g., heme a) and the high-spin heme-copper binuclear center (e.g., heme a₃-Cu_B). The subscript “3” designates the heme where O₂ binds and is reduced to water. Subunit I contains the proton-conducting channels which facilitate the translocation of protons from the electrically negative side of the membrane either to the active site for the chemical conversion of O₂ to 2 H₂O or to the positive side of the membrane, i.e., pumped protons (4, 5). Prokaryotic heme-copper oxygen reductases, besides having fewer subunits than the eukaryotic homologs, also can utilize different hemes (3, 6). Whereas the mitochondrial enzyme contains 2 equivalents of heme A (a chemical modification of heme B), prokaryotic heme-copper oxygen reductases can utilize different combinations of heme B, heme O, and heme A. For example, cytochrome bo₃ from Escherichia coli contains 1 equivalent of heme B and 1 equivalent of heme O, which binds to O₂.

Most prokaryotic heme-copper oxygen reductases (A, B, and C families) (1) use reduced cytochrome c as a substrate, but a subfamily of the A family utilizes a membrane-bound quinol, usually ubiquinol or menaquinol, as a substrate in place of cytochrome c. The quinol oxidases can be distinguished from the homologous cytochrome c oxidases by the absence of both the cytochrome c binding domain and the Cu_A binding motif in subunit II. In addition, subunit I of all of the quinol oxidases contains 15 TM helices rather than the canonical 12 TM spans (1). These extra membrane-spanning helices are designated as TM0 at the N terminus and TM13 and TM14 at the C terminus of subunit I. The presence of TM0 is unique to the quinol oxidases, while TM13 and TM14 are found in some cytochrome c oxidases, e.g., cytochrome caa₃ from Bacillus subtilis (7).

While there are numerous structures of cytochrome c oxidases from a variety of species (2, 8–12), there is only 1 structure reported for a heme-copper quinol oxidase, E. coli cytochrome bo₃ ubiquinol oxidase (13). This enzyme contains 4 subunits, 3 of which are homologous to the 3 mitochondrial encoded subunits of the eukaryotic cytochrome c oxidases (14). The structure of cytochrome bo₃ is missing ∼25% of the expected amino acid residues, including TM0, TM13, and TM14 in subunit I, and does not contain ubiquinone. Despite these deficiencies, the structure did reveal a cluster of amino acid residues in subunit I that are conserved among ubiquinol oxidases (R71^I, D75^I, H98^I, and Q101^I; E. coli numbering, and the superscript I designates subunit I), suggesting that these might constitute a portion of an ubiquinol binding site (13). Mutagenesis showed their functional importance, and pulsed electron paramagnetic resonance (EPR) studies have clearly demonstrated that R71^I, D75^I, H98^I, and, to a lesser extent, Q101^I directly interact with and stabilize the ubisemiquinone (SQ-•) catalytic intermediate (15–19).

In the current work, the X-ray structure of the cytochrome aa₃-600 proton-pumping menaquinol oxidase from B. subtilis (20–22) is reported at 3.6-Å resolution. This enzyme contains 4 subunits and is a close ortholog of E. coli cytochrome bo₃. The sequences of subunit I of these 2 enzymes are 50% identical, even though the enzymes use different quinol substrates and contain different hemes. Cytochrome aa₃-600 contains 2 equivalents of heme A in place of 1 equivalent each of heme B and heme O found in cytochrome bo₃. The B. subtilis and E. coli enzymes utilize menaquinol-7 and ubiquinol-8, respectively. The “600” in the name of the menaquinol oxidase designates the peak of the alpha absorption band due to heme a. In contrast to the structure of cytochrome bo₃, cytochrome aa₃-600 is crystallized with 1 equivalent of the native substrate (menaquinol-7). Importantly, TM0 was resolved and appeared to form part of a cleft facing the lipid bilayer into which the hydrophobic poly-isoprene side chain of the substrate, menaquinol-7, can fit while being oxidized. In addition, structures have also been determined with 1 equivalent of the inhibitors HQNO (N-oxo-2-heptyl-4-hydroxyquinoline) or 3-iodo-HQNO bound to the protein. The inhibitor is hydrogen bonded to residues R70^I, D74^I, and H94^I (B. subtilis numbering), which have been shown spectroscopically to stabilize the semiquinone state of the menaquinol substrate (21, 22).

Results

Additional TM Helices in Subunit I Compared to the Canonical Cytochrome c Oxidases.

Whereas subunit I of most cytochrome c oxidases contains 12 TM helices, subunit I of both cytochrome bo₃ and aa₃-600 (and all quinol oxidases) contains 15 TM helices; 1 additional TM span, TM0, at the N terminus; and 2 additional TM spans, TM13 and TM14, at the C terminus. Although not resolved in the crystal structure of cytochrome bo₃ (Protein Data Bank [PDB] ID code 1FFT), these 3 helices are confidently modeled in the structure of cytochrome aa₃-600 (Fig. 1). The superposition of the cytochrome aa₃-600 menaquinol oxidase structure and the structure of the aa₃-type cytochrome c oxidase from Rhodobacter sphaeroides (PDB ID code 1M56) shows that TM13 and TM14 overlap with the first 2 helices of subunit III of the R. sphaeroides enzyme. TM0 has no counterpart in any of the cytochrome c oxidases and is unique to the heme-copper quinol oxidases.

Fig. 1.

Surface representations (A and B) and cartoon representations (C and D) of the structure of cytochrome aa₃-600 from B. subtilis. The four subunits are color-coded: subunit I (blue), subunit II (red), subunit III (gray), and subunit IV (brown). TM0 has no counterpart in any of the cytochrome c oxidases and is unique to the heme-copper quinol oxidases.

Significantly, the side chains of TM0 are not immediately adjacent to the nearby TM helices, but form a groove along with TM1 and TM2 which appears to be well placed to facilitate the binding of menaquinol-7 from the membrane bilayer (Fig. 2). The residues that are hydrogen bonded to the menasemiquinone species, R70^I, D74^I, and H94^I, are located at the “top” of TM1 and TM2 (external face of the membrane). Two consecutive lysine residues on the cytoplasmic side of TM0, K38^I, and K39^I are in good positions to interact with the phospholipids to anchor TM0 to the membrane.

Fig. 2.

Side view (A) and top view (B) of TM0, TM1, and TM2 of subunit I of cytochrome aa₃-600, showing the conserved cluster of residues that are conserved in all heme-copper menaquinol oxidases, R70^I, D74^I, H94^I, and E97^{I I}. The groove lined by TM0, TM1, and TM2 forms the binding site for menaquinol-7 within the membrane bilayer. Two consecutive lysine residues, K38^I and K39^I, are proposed to interact with phospholipids in the inner leaf of the membrane bilayer.

Similar Features of Cytochrome aa₃-600 and Cytochrome bo₃.

Subunit I.

The structures of the binuclear center and the porphyrin rings of the high- and low-spin hemes are essentially identical for both cytochrome bo₃ and cytochrome aa₃-600, apart from the chemical modifications differentiating heme B, heme O, and heme A. The farnesyl tail of heme a, which is not present in cytochrome bo₃, has a similar orientation as in the caa₃-type cytochrome c oxidase from Thermus thermophilus (PDB ID code 2YEV), but different from the aa₃-type cytochrome c oxidase from R. sphaeroides (PDB ID code 1M56). The farnesyl tails of heme a₃ and heme o₃ from cytochrome aa₃-600 and cytochrome bo₃, respectively, have similar orientations.

The 2 proton-conducting channels, named the D channel and the K channel, that are present in A-family heme-copper oxygen reductases (23) are also present in cytochrome aa₃-600, which is also in the A family. The D channel starts from a highly conserved aspartate residue, D131^I, at the entrance of the channel; includes polar residues T207^I, N138^I, N120^I, N203^I, S141^I, T200^I, and T197^I; and ends at E282^I, which is the branching point of chemical protons and pumped protons, as determined from previous studies. The K channel features a conserved lysine residue K358^I and contains a series of polar residues S295^I, T355 ^I, and Y284 ^I. These key residues in both channels exist at the same places as in the cytochrome bo₃ quinol oxidase (SI Appendix, Fig. S3).

Subunit II.

Unlike cytochrome c oxidases, there is no Cu_A-binding site or cytochrome c docking site in subunit II of quinol oxidases. Whereas subunit I of B. subtilis cytochrome aa₃-600 subunit I has ∼50% amino acid sequence identity with that of the E. coli cytochrome bo₃, the sequences of subunit II have only 31% identity.

Subunits III and IV.

Subunits III of cytochrome aa₃-600 and cytochrome bo₃ are 44% identical, and in both enzymes, subunit III has 5 TM helices. The positions of these 5 helices overlap with helices TM3 to TM7 in canonical cytochrome c oxidases (e.g., from R. sphaeroides PDB ID code 1M56). As previously noted, TM13 and TM14 from subunit I of cytochrome aa₃-600 overlap with TM1 and TM2 of subunit III of canonical cytochrome c oxidases, suggesting an evolutionary “shift” of these TM helices from the N terminus of subunit III in the canonical cytochrome c oxidases to the C terminus of subunit I. The gene encoding subunit III is immediately downstream from that encoding subunit I in most organisms.

The function of subunit IV is not known, and this subunit is the least conserved, with 29% identity between cytochrome aa₃-600 and cytochrome bo₃. The side chains of subunit IV are not resolved in the structure of cytochrome bo₃. Subunit IV of cytochrome aa₃-600 contains 3 TM helices, as also modeled for cytochrome bo₃ (13).

Bound menaquinone-7.

Cytochrome aa₃-600 oxidase was solubilized, isolated, and crystallized by using the detergent dodecylmaltoside (DDM). With this detergent, the enzyme was copurified with 1 equivalent of menaquinone-7 (22), the oxidized form of the endogenous substrate. In the omit map calculated by using a ligand-free model, electron density was observed in the groove defined by TM0, TM1, TM2, and TM3 of subunit I. The electron density was insufficient to confidently fit a molecular model. Assuming this density represents the bound menaquinol-7, the electron density could be modeled with the menaquinol head group plus 3 of the 7 isoprene units of the hydrophobic tail. The menaquinone head group in this model was near V73^I (TM1, subunit I) and F156^I (TM4, subunit I), and the isoprenyl tail established hydrophobic interactions with the hydrophobic residues in TM3 and TM0 (Fig. 3A). Higher-resolution data will be required to confirm this interpretation of the observed electron density. It cannot be ruled out that this density is due to a detergent or lipid molecule.

Fig. 3.

(A) Menaquinone molecule in the groove. The menaquinone head group is near V73^I and F156^I. (B) Superposition of the native and inhibitor-complexed structures. Green, native; magenta, cytochrome aa₃-600 complexed with 3-iodo-HQNO. The maps are omit maps, calculated by using models without the substrate and contoured at 1.0 σ.

Site-directed mutagenesis and inhibition by HQNO.

HQNO is a potent inhibitor of cytochrome bo₃ (24) and is generally considered to mimic the semiquinone state of both menaquinone and ubiquinone. Wild-type cytochrome aa₃-600 was completely inhibited in the presence of 1 μM HQNO. In addition, similar to what was found with cytochrome bo₃ (24), the formation of the menasemiquinone by cytochrome aa₃-600 was abolished in the presence of 1 μM HQNO. This is shown in SI Appendix, Fig. S2.

The residues that interact with and stabilize the menasemiquinone in cytochrome aa₃-600 have been identified both by analogy to cytochrome bo₃ (24), but also by pulsed EPR methods (21, 22): R70^I, D74^I, and H94^I (Fig. 2). It was reported (21) that the R70^IH mutant has a perturbed interaction with the menasemiquinone species, but does not eliminate the EPR signal of the radical, as does the equivalent mutation in cytochrome bo₃ (25). Furthermore, the R70^IH mutant is more active than wild-type cytochrome aa₃-600, indicating a greater plasticity or tolerance to structural perturbation for stabilization of the menasemiquinone species compared to cytochrome bo₃. To further confirm the identity of the residues that interact with the menasemiquinone species in cytochrome aa₃-600, additional mutations were generated, and the effects on the steady-state activities with both DMNH₂ and with menadiol are shown in Table 1. The mutations of residues D74^I and H94^I were the most potent in reducing the activity of the enzyme, though the H94^IF mutant retained 62% of the activity with DMNH₂ as the substrate. The results strongly suggest that D74^I and H94^I interact directly with one of the menasemiquinone species and contribute to stabilizing the menasemiquinone, but that stabilization of the menasemiquinone is less sensitive to changes in the surrounding amino acid residues than the stabilization of the ubisemiquinone species by cytochrome bo₃ (25). In addition, similar to the results obtained with cytochrome bo₃ (25), each of the mutants shown in Table 1 is bound to 1 equivalent of menaquinone-7 after solubilization and purification in the detergent DDM. Hence, menaquinone-7 does not require any interaction with R70^I, D74^I, or H94^I to remain bound to the enzyme in the presence of DDM.

Table 1.

Activities of the wild-type and mutant variants of cytochrome aa₃-600 using soluble menaquinol analogs

Enzyme	Enzyme turnover, s−1 (%)
	With DMNH2 (50 μM)	With menadione (130 μM)
WT aa3-600	115 (100)	33 (100)
R70H*	177 (154)	51 (155)
E97Q	192 (167)	12 (36)
H94F	71 (62)	24 (73)
H94D	18 (16)	2 (6)
D74H	6 (5)	0 (0)
D74N	6 (5)	6 (18)

^*Results reported in ref. 21.

The mutations of these active-site residues also make cytochrome aa₃-600 less sensitive to inhibition by HQNO. For example, in the presence of 30 μM HQNO, R70^IH, H94^ID, H94^IF, and E97^IQ have substantial residual activity: 8%, 45%, 64%, and 20%, respectively. These results suggest that these residues are at or near the binding site for HQNO, similar to the observations made with cytochrome bo₃ (26). The structure of cytochrome aa₃-600 with the 3-iodo-HQNO derivative of HQNO confirms that the inhibitor binds directly to R70^I, D74^I, and H94^I, as discussed below.

Binding site for inhibitors HQNO and 3-iodo-HQNO.

After soaking the crystals with HQNO, the initial F_o − F_c map showed extra density near residues R70^I, D74^I, and H94^I (Fig. 4A). However, due to the limited resolution, it was not possible to locate the exact position of individual atoms of the HQNO molecule. To better understand how the inhibitor interacts with cytochrome aa₃-600, 3-iodo-HQNO was synthesized to take advantage of the strong electron density of the iodine atom. After soaking the crystals with 3-iodo-HQNO, the initial F_o − F_c map clearly showed a positive peak stronger than 7 σ representing the position of the iodine atom. At 3 σ, the density of the rest of the inhibitor analog was also visible (Fig. 4B). The 3-iodo-HQNO was hydrogen bonded to R70^I, D74^I, and H94^I. H94^I has well-defined electron density and underwent a significant conformational change compared with the native enzyme (Fig. 3B). All heme-copper menaquinol oxidases have a conserved glutamate (E97^I) in TM2 (subunit I), near the bound 3-iodo-HQNO, but the electron density of E97^I is not well defined, suggesting conformational flexibility of this side chain.

Fig. 4.

(A) F_o − F_c map of the cytochrome aa₃-600–HQNO complex contoured at 2.5 σ. (B) F_o − F_c map of the cytochrome aa₃-600–3-iodo-HQNO complex. The magenta map representing the position of iodine atom is contoured at 7.0 σ, and the white map is contoured at 3.0 σ.

The residual electron density that was fit to a portion of menaquinol-7 in the crystal of the native enzyme in the absence of the inhibitor was still observed in the presence of either HQNO or 3-iodo-HQNO.

Discussion

Proton-pumping heme-copper oxygen reductases that directly oxidize membrane-bound quinols are present in the aerobic respiratory chains of many bacterial species. By far, the best characterized is the cytochrome bo₃ ubiquinol oxidase from E. coli (26, 27), and the current work presents the X-ray structure of a very close ortholog, the cytochrome aa₃-600 menaquinol oxidase from B. subtilis (22). Although the hemes and substrates are different for these 2 enzymes, the sequences of subunit I, which contains the binding sites for the hemes and the catalytic sites, are 50% identical.

The 2 most important conclusions from the current work are that

• 1)TM0, the N-terminal TM helix in subunit I that is found only in heme-copper oxygen reductases that utilize quinol substrates, is at a location where it can be reasonably assumed to form part of the binding site for the membrane-bound quinol substrates. This substrate is menaquinol-7 in the case of cytochrome aa₃-600. The groove facing the lipid bilayer that is formed by portions of TM0, TM1, TM2, and TM3 is located just below the cluster of residues known to stabilize the semiquinone state of the menaquinol-7 substrate, R70^I, D74^I, and H94^I (21, 22). These residues are located within TM1 and TM2.
• 2)The semiquinone-mimic HQNO that inhibits cytochrome aa₃-600, binds stoichiometrically to the enzyme and is hydrogen bonded to the 3 residues known to stabilize the semiquinone form of menaquinone-7 bound to the protein, R70^I, D74^I, and H94^I (21, 22). Notably, H94^I undergoes a significant conformational change to accommodate the hydrogen bonding to 3-iodo-HQNO (and HQNO). A similar change in conformation of H94^I would also be necessary to hydrogen bond to the semiquinone state of the substrate.

Although it is known that the DDM-solubilized cytochrome aa₃-600 that is crystallized contains 1 equivalent of menaquinone-7 (22), the weak electron density in the groove defined by TM0, TM1, TM2, and TM3 cannot be definitely assigned as being due to the bound substrate. Assuming that this electron density is due to menaquinone-7, the location is consistent with the oxidized form of the substrate being trapped bound to the enzyme in the presence of DDM in a way that neither alters the binding of HQNO nor is perturbed by the binding of HQNO. The simultaneous binding of HQNO and a long-chain quinone species is also observed with long-chain ubiquinone in DDM-solubilized cytochrome bo₃ and is an artifact due to the poor solubility of the long-chain quinones in the detergent micelle (26). Under native conditions in a phospholipid bilayer, one would not expect the oxidized quinone to remain tightly bound to the enzyme, since it is the product of the reaction catalyzed by the enzyme. In DDM-solubilized cytochrome aa₃-600 without the HQNO inhibitor, a 1-electron reduction of the bound (trapped) menaquinone-7 results in the formation of a bound semiquinone stabilized by direct interaction with R70^I, D74^I, and H94^I (21, 22). This could occur easily by a small change of the position of the bound menaquinone within the binding groove, along with the shift in the side chain of H94^I (Fig. 3)

In summary, the current work clarifies the major structural modifications that have evolved to allow the respiratory heme-copper oxygen reductases that utilize reduced cytochrome c as a substrate to directly oxidize quinol from the membrane-bound pool. The critical structural alterations are 1) the addition of TM0, which forms part of a cleft that facilitates binding of the quinol from the membrane bilayer; and 2) the addition of the cluster of residues at the “top” of TM1 and TM2 that stabilizes the semiquinone state of the quinone, a required catalytic intermediate.

Experimental Methods

Purification of Cytochrome aa₃-600 Menaquinol Oxidase.

The construct containing cytochrome aa₃-600 genes, qoxABCD, was generated by PCR from the B. subtilis genome, and a hexahistidine tag was added at the C terminus of qoxA. The gene was subcloned into the pDR111 plasmid (28) and eventually integrated onto the genome of B. subtilis PY79 strain via homologous recombination (both the pDR111 plasmid and PY79 strain were gifts from Yunrong Chai, College of Science, Northeastern University, Boston). B. subtilis PY79 cells containing the incorporated construct were grown in Luria–Bertani medium, collected, and resuspended in 50 mM potassium phosphate buffer (pH 8.0), with the addition of phenylmethylsulfonyl fluoride (Sigma), 10 mM (ethylenedinitrilo)tetraacetic acid, and DNase I (Sigma). Homogenized B. subtilis cells were disrupted by using a high-pressure homogenizer (ATS Engineering), followed by a 10-min centrifugation at 20,000 × g. The membrane fraction was separated after 3-h ultracentrifuge at 180,000 × g. The isolated membranes were resuspended in 50 mM potassium phosphate buffer (pH 8.0) and solubilized with 1% DDM on ice for 1 h. Insoluble material was removed by centrifugation at 180,000 × g for 30 min. The solubilized fraction was loaded onto an Ni-nitriloacetic acid (NTA) (Qiagen) column preequilibrated with 50 mM potassium phosphate buffer and 0.1% DDM (pH 8.0). A gradient wash was applied with increasing concentration of imidazole (5 to 25 mM) in the same buffer in order to remove nonspecific bound protein. His-tagged cytochrome aa3-600 was eluted with 50 mM imidazole in 50 mM potassium phosphate buffer and 0.1% DDM (pH 8.0). The peak fractions from the Ni-NTA column were pooled and concentrated 5-fold to ∼200 μL by using a 100-kDa molecular-mass-cutoff Vivaspin centrifugal concentrator. For subsequent crystallization, the protein was twice diluted with 5 mL of buffer containing 20 mM Tris⋅HCl (pH 7.8) and 50 mM NaCl and concentrated. In the presence of aa₃-600, DDM micelles could not pass through the 100-kDa cutoff concentrator, and no detergent was needed in the exchanged buffer. The ultraviolet-vis spectra of the isolated enzyme, along with the sodium dodecyl sulfate/polyacrylamide gel electrophoresis gel, are shown in SI Appendix, Fig. S1.

Steady-State Activity.

The steady-state activity of purified cytochrome aa₃-600 was measured by using an oxygen meter (Clark electrode) in the presence of menaquinone-reducing type II NADH dehydrogenase from Oceanobacillus iheyensis. About 200 μM NADH was used to keep 10 μM 2,3-dimethyl-1,4-naphtoquinone reduced in the presence of the type II NADH dehydrogenase. By adding 0.05 μM cytochrome aa₃-600, the reaction was performed under conditions where the reaction of cytochrome aa₃-600 was rate-limiting (22).

Protein Crystallization.

Crystallization was performed in MRC-Maxi 48-well plates by using the sitting-drop vapor-diffusion method. Two microliters of protein solution containing 7 mg/mL cytochrome aa₃-600 in 20 mM Tris (pH 7.8), 50 mM NaCl, and ∼0.5% DDM was mixed with 2 μL of well solution containing 0.1 M calcium chloride, 0.1 M Tris (pH 6.3), and 10 to 13% polyethylene glycol monomethyl ether 2000 (PEG 2000 MME) (vol/vol). The crystallization drop was incubated at 22 °C against 100 μL of well solution. The best plate-like crystals of a maximum size of 200 × 80 × 30 μm appeared after 20 d and reached final size after 30 d.

Data Collection and Processing.

Crystals were transferred to a cryoprotectant solution comprising 0.1 M calcium chloride, 0.1 M Tris (pH 6.3), and 27% PEG 2000 MME (vol/vol), then flash frozen in liquid nitrogen. Each of the 3 X-ray datasets was collected from a single crystal at 100 K at BL17U1 of the Shanghai Synchrotron Radiation Facility. Images were processed with Mosflm (29). The crystals belonged to space group P2₁, with 2 molecules (complexes) per asymmetric unit (SI Appendix, Table S1).

Phasing, Model Building, and Refinement.

The structure of cytochrome aa₃-600 was solved by molecular replacement implemented in Phaser (30). Coordinates of subunits I and III of cytochrome bo₃ from E. coli (PDB ID code 1FFT) were used as the original search model. Well-defined electron density was revealed around the 2 subunits in the search model and fitted by using a poly-alanine model from the corresponding model (PDB ID code 1FFT). The combined model was used as the starting model for Phenix Autobuild (31). The phase was gradually improved, indicated by the better electron density of the Heme farnesyl tail and the side chains of subunits II and IV. Subsequently, models were manually built into the well-defined electron density by using COOT (32). The final model of 2 complexes contained 2,228 residues, accounting for 85% of the total residues. Structures of the cytochrome aa₃-600/HQNO complex and the cytochrome aa₃-600/iodoHQNO complex were solved by molecular replacement using the native cytochrome aa₃-600 structure as search model, and the ligands were manually fitted by using COOT (32). Structure refinement was performed by using Phenix (31). Membrane bilayer was generated with CHARMM-GUI Membrane Builder (33), and figures of the models were created with PYMOL (34).

Data Availability.

The coordinates for the model of cytochrome aa₃-600 from B. subtilis have been deposited in the Protein Data Bank, https://www.rcsb.org (PDB ID codes 6KOB, 6KOC, and 6KOE).