The cryoEM structure of cytochrome bd from C. glutamicum provides novel insights into structural properties of actinobacterial terminal oxidases

Cytochromes bd are essential for microaerobic respiration of many prokaryotes including a number of human pathogens. These enzymes catalyze the reduction of molecular oxygen to water using quinols as electron donors. Their importance for prokaryotic survival and the absence of eukaryotic homologs make these enzyme ideal targets for antimicrobial drugs. Here, we determined the cryoEM structure of the menaquinol-oxidizing cytochrome bd-type oxygen reductase of the facultative anaerobic Actinobacterium Corynebacterium glutamicum at a resolution of 2.7 Å. The obtained structure adopts the signature pseudosymmetrical heterodimeric architecture of canonical cytochrome bd oxidases formed by the core subunits CydA and CydB. No accessory subunits were identified for this cytochrome bd homolog. The two b-type hemes and the oxygen binding heme d are organized in a triangular geometry with a protein environment around these redox cofactors similar to that of the closely related cytochrome bd from M. tuberculosis. We identified oxygen and a proton conducting channels emerging from the membrane space and the cytoplasm, respectively. Compared to the prototypical enzyme homolog from the E. coli, the most apparent difference is found in the location and size of the proton channel entry site. In canonical cytochrome bd oxidases quinol oxidation occurs at the highly flexible periplasmic Q-loop located in the loop region between TMHs six and seven. An alternative quinol-binding site near heme b 595 was previously identified for cytochrome bd from M. tuberculosis. We discuss the relevance of the two quinol oxidation sites in actinobacterial bd-type oxidases and highlight important differences that may explain functional and electrochemical differences between C. glutamicum and M. tuberculosis. This study expands our current understanding of the structural diversity of actinobacterial and proteobacterial cytochrome bd oxygen reductases and provides deeper insights into the unique structural and functional properties of various cytochrome bd variants from different phylae.

Cytochromes bd are essential for microaerobic respiration of many prokaryotes including a number of human pathogens. These enzymes catalyze the reduction of molecular oxygen to water using quinols as electron donors. Their importance for prokaryotic survival and the absence of eukaryotic homologs make these enzyme ideal targets for antimicrobial drugs. Here, we determined the cryoEM structure of the menaquinol-oxidizing cytochrome bd-type oxygen reductase of the facultative anaerobic Actinobacterium Corynebacterium glutamicum at a resolution of 2.7 Å. The obtained structure adopts the signature pseudosymmetrical heterodimeric architecture of canonical cytochrome bd oxidases formed by the core subunits CydA and CydB. No accessory subunits were identified for this cytochrome bd homolog. The two b-type hemes and the oxygen binding heme d are organized in a triangular geometry with a protein environment around these redox cofactors similar to that of the closely related cytochrome bd from M. tuberculosis. We identified oxygen and a proton conducting channels emerging from the membrane space and the cytoplasm, respectively. Compared to the prototypical enzyme homolog from the E. coli, the most apparent difference is found in the location and size of the proton channel entry site. In canonical cytochrome bd oxidases quinol oxidation occurs at the highly flexible periplasmic Q-loop located in the loop region between TMHs six and seven. An alternative quinol-binding site near heme b 595 was previously identified for cytochrome bd from M. tuberculosis. We discuss the relevance of the two quinol oxidation sites in actinobacterial bd-type oxidases and highlight important differences that may explain functional and electrochemical differences between C. glutamicum and M. tuberculosis. This study expands our current understanding of the structural Introduction An estimated 2.4 billion years ago, the "Great Oxidation Event", presumably originating from the activity of the photosynthetic ancestors of cyanobacteria, raised atmospheric O 2 levels over four orders of magnitude (to about 10% of current levels) and drastically changed the evolutionary course of life while also leading to one of the greatest mass extinction events on earth (Schirrmeister et al., 2013;Lyons et al., 2014). Certain organisms managed to escape oxidative damage by inhabiting oxygen-free niches. However, the more important adaption to these extreme environmental conditions was the evolution of metalloproteins that catalyze the reduction of highly toxic dioxygen to harmless water and at the same time harness the oxidizing power of oxygen for organismal energy metabolism (Hoganson et al., 1998).
Among these enzymes, terminal oxygen reductases catalyze the reduction of molecular oxygen to water employing electrons from the oxidation of either cytochrome c (cyt. c) or membrane dissolved quinols (QH 2 ). They are classified into i) heme-copper oxidases (HCO), which contain, as a unifying characteristic, a heme-copper binuclear center, ii) alternative oxidases (AOX), first identified in plants, and iii) cytochrome bd oxidases, which are only present in prokaryotes (May et al., 2017;Borisov and Siletsky, 2019). All so-far experimentally characterized bd-type oxidases use exclusively quinols as substrates, whereas HCOs comprise both cyt. c-and QH 2 -oxidizing members.
The cytochrome bd enzyme was first described nearly a century ago and since then has been identified as a membrane-integrated respiratory terminal oxidase unique to the prokaryotic domain, including a number of human pathogens (Cook and Poole, 2016;Borisov et al., 2020). While bd-type enzymes catalyze the reduction of molecular oxygen to water, they do not actively pump protons across the cytoplasmic membrane as do members of the HCO superfamily. Instead, cytochrome bd generates an electrochemical proton gradient via charge separation by consuming substrate protons from the cytoplasmic space and the release of protons upon quinol oxidation to the perisplasm. Thus, the contribution to the proton motive force (pmf) is smaller than that of protonpumping HCO-type oxidases (Puustinen et al., 1991;Borisov et al., 2011;Borisov and Verkhovsky, 2015).
Canonical cytochrome bd oxidases share a common core architecture of two subunits, denoted CydA and CydB, which may be accompanied by up to two additional single-transmembrane helix subunits (Miller and Gennis, 1983;VanOrsdel et al., 2013;Hoeser et al., 2014;Safarian et al., 2016;Safarian et al., 2019;Theßeling et al., 2019;Grund et al., 2021;Safarian et al., 2021;Wang et al., 2021;Friedrich et al., 2022). As implied by designation, cytochromes bd contain two b-type hemes (low-spin b 558 , high-spin b 595 ) and one d-type heme involved in quinol oxidation, electron transfer and oxygen reduction. Substrate-binding occurs within CydA at the quinol binding domain (Q-loop) located on the periplasmic side of the membrane. Members of the canonical cytochrome bd oxidases are further subdivided into those containing either a short or a long hydrophilic Q-loop domain (S-and L-subfamily, respectively) (Osborne and Gennis, 1999;Sakamoto et al., 1999;Kusumoto et al., 2000). The role of the N-terminal insertion of the Q-loop domain has been studied previously, yet it still remains elusive whether this extension fulfils a purely structural or additionally also a functional role within cytochrome bd oxidases (Goojani et al., 2020;Theßeling et al., 2020).
Cytochromes bd are characterized by high oxygen affinity, as well as their resistance to cyanide inhibition (IC 50 concentration for KCN of cytochrome bo 3 from E. coli: 10 μM; IC 50 concentration for KCN of cytochrome bd-I from E. coli: 2 mM) (Kita et al., 1984;Borisov and Verkhovsky, 2015). Extreme cyanide insensitivity as well as an apparent lack of a heme d signal, led to the sub-classification of cytochromes bd into the non-canonical cyanide-insensitive quinol oxidases (CIO) and canonical bd-type enzymes (Borisov et al., 2011). In CIOs, heme d is replaced by a second high-spin b-type heme (Azarkina et al., 1999;Jackson et al., 2007;Mogi et al., 2009). CIO-type terminal oxidases are encoded by the cio operon. The deduced amino acid sequences of CioA and CioB are homologous to CydA and CydB (Cunningham et al., 1997;Quesada et al., 2007). CIOs are found inter alia in Pseudomonas aeruginosa and Pseudomonas pseudoalcaligenes, and are assumed to enable aerobic respiration under cyanogenic and microaerobic growth conditions. This is of relevance for the opportunistic pathogen P. aeruginosa to attain full pathogenicity in the cyanide-mediated paralytic killing of nematodes (Cunningham et al., 1997;Quesada et al., 2007).
Here, we determined the cryoEM structure of the actinobacterial cytochrome bd from Corynebacterium glutamicum to 2.7 Å which is closely related to the mycobacterial cytochrome bd oxidases and at the same time displays spectroscopic and electrochemical properties closely matching those of cytochrome bd-I from E. coli (Nikolaev et al., 2021).

Results
The menaquinol oxidizing cytochrome bd-type oxygen reductase of the facultative anaerobic Actinobacterium C. glutamicum (Cgbd) was produced under the activity of its native promoter without an affinity tag as described previously (Kabashima et al., 2009) Figure  S1). The structure of the natively purified oxidase was determined via single-particle analysis cryoEM. A total of 13,390 movies were collected on a Titan Krios G3i microscope equipped with a Gatan K3 camera. The final map refined to 2.7 Å resolution (Supplementary Figures S2-S4, Supplementary  Table S1).
Cytochrome bd from C. glutamicum is characterized by the canonical cytochrome bd architecture comprising the larger CydA and the slightly smaller CydB subunits. Each subunit is composed of two four-helix bundle motifs and an additional peripheral helix. The heterodimeric Cgbd forms a pseudosymmetrical complex with C2 symmetry (Figure 1). The dimerization interface is composed of the symmetry related TMHs 2, 3, and 9 of CydA and CydB, respectively. Subunit interactions are mostly characterized by van-der-Waals interactions. Unlike the E. coli and G. thermodenitrificans cytochrome bd oxidases, Cgbd does not contain small accessory subunits which resembles the composition of the closely related mycobacterial oxidases (Safarian et al., 2021;Wang et al., 2021). C. glutamicum and M. tuberculosis both belong to the phylum of Actinobacteria. The amino acid sequences of their core subunits share a sequence identity of 62% and 48% for CydA and CydB, respectively.

FIGURE 1
CryoEM structure of C. glutamicum cytochrome bd at 2.7 Å resolution. (A and C) Side view representations of the cryoEM map density and its corresponding ribbon model of the heterodimeric C. glutamicum cytochrome bd structure. The substrate quinol binding site 1 at the Q-loop in proximity to heme b 558 is highlighted in orange. The second putative quinol binding site next to heme b 595 is shown in purple. (B and D) Top view representation of the cryoEM map density and the corresponding tube model of the Cgbd structure highlighting the pseudosymmetrical arrangement of the two four helix bundles and the additional peripheral helix per subunit. Color code: CydA, green; CydB, pink; b-type hemes, cyan; heme d, dark blue.
The catalytically active subunit CydA of Cgbd harbors three prosthetic heme groups (b 558 , b 595 , and d) and contains the quinol oxidizing domain (Q-loop). The Q-loop is located between TMHs six and seven and protrudes into the periplasm. While the Q C region is fully resolved, residues 268 to 317 which constitute the Q N region could not be traced with high confidence. The Q-loop starts with the short helix Qh1 comprising the conserved residues Lys 263.A Cgbd and Glu 268.A Cgbd . These residues have been implicated in quinol binding and oxidation as well as coordination of propionate A from heme b 558 in cytochromes bd from Proteobacteria (Matsumoto et al., 2006;Mogi et al., 2006). The Q N region of Cgbd appears to be characterized by increased movement dynamics similar to what has been observed for cytochrome bd from Mycobacterium smegmatis (Msbd) (Wang et al., 2021). The Q C region of Cgbd is fully resolved and consists of the periplasm exposed helix Qh3 that is followed by a 180°turn and leads back to TMH7. In the cytochrome bd structure of Mycobacterium tuberculosis (Mtbd), a hydrophobic cluster stabilizing the interaction of Qh3 with the periplasmic loop connecting TMHs eight and nine (PL8) was identified (Safarian et al., 2021 (Sviriaeva et al., 2020).
The two b-type hemes b 558 and b 595 as well as the chlorintype heme d are organized in the well-characterized triangular cytochrome bd-specific geometry and are located in proximity of the periplasmic surface ( Figure 2). The coordinating and vicinal residues, as well as the arrangement of the b-type hemes and heme d are conserved between cytochromes bd from Proteobacteria and Actinobacteria (Safarian et al., 2019;Theßeling et al., 2019;Grauel et al., 2021;Grund et al., 2021;Safarian et al., 2021;Wang et al., 2021). A noteable variation in the triangular heme arrangement is however found in the structure of cytochrome bd from Geobacillus thermodenitrificans (Gtbd), in which the high-spin heme b and heme d have "switched" positions.
The structural subunit CydB does not harbor cofactors or structural quinone molecules as observed in the proteobacterial cytochromes bd-I and bd-II (Safarian et al., 2019;Theßeling et al., 2019;Grauel et al., 2021;Grund et al., 2021). This is in agreement with the structures of the closely related Mtbd and Msbd that also lack structural quinones in the smaller subunit CydB (Safarian et al., 2021;Wang et al., 2021). Analogous to Mtbd, TMHs six and seven in CydB are connected by a short linker of three amino acids (CydB Cgbd : amino acids 220-222), whereas the proteobacterial enzymes contain an elongated β-sheet extending along the periplasmic surface of CydB and a short α-helix oriented in a roughly 90°angle to TMH6 (Supplementary Figure S5).
The structural framework of a cytochrome bd-type enzyme requires, in addition to domains and cofactors for substrate oxidation and electron transfer, specific channels and cavities for allowing the access of protons and dioxygen to the reduction site at heme d, which is buried deeply within the core of the enzyme at the interface of CydA and CydB. In the Cgbd structure, both an oxygen and a proton pathway were identified ( Figures 3A, B) (Sehnal et al., 2013;Pravda et al., 2018). The proton channel (H-channel) runs perpendicular to the membrane plane and connects the cytoplasm to the active site at heme d. The H-channel is predicted to start as a shallow cavity at the cytoplasmic site in proximity to TMHs two and three of CydB and subsequently constricts and elongates at the interface of CydA and CydB towards the reduction site. The entrance of the H-channel is shifted towards CydB compared to the proteobacterial channels ( Figure 3C). The typical opening in the proteobacterial structures is obstructed by the extended C-terminal stretch of CydA in Cgbd, while the entrance of the H-channel of Cgbd is blocked by the short cytoplasmic C-terminal helix of CydB/AppB in cytochromes bd of E. coli. On the other hand, the location of the hydrophobic oxygen channel (O-channel) is matching well with the pathways observed in the proteobacterial and mycobacterial enzymes. The O-channel is predominantly lined by hydrophobic and uncharged residues and starts at the lipid interface between TMHs one and nine of CydB and extends parallel to the membrane plane, connecting the lipid bilayer to the reduction site at the chlorin-type heme d.

Discussion
Two substrate quinone binding sites have been characterized for bd-type oxidases (for a comparison of all quinone sites please see Supplementary Figure S6). One at the Q-loop next to heme b 558 which is best characterised in Ecbds (site 1). A second quinol binding-site composed of TMH1, TMH9 and heme b 595 has been identified exclusively in Mtbd (site 2) (Safarian et al., 2019;Grauel et al., 2021;Safarian et al., 2021). At first glance, the overall structure of Cgbd is showing strong resemblance to that of Mtbd, however the structure of Qh3-PL8 region near site 1 is slightly different.
In Mtbd, the Qh3-PL8 region is mostly stabilized by a hydrophobic cluster composed of aromatic side chains as described previously (Safarian et al., 2021). In addition, a hydrogen bond between the side chain carbonyl group of Gln 412.A Mtbd and Arg 324.A Mtbd seems to strengthen the interaction between PL8 and Qh3 ( Figure 4). Furthermore, electrostatic interactions between Arg 324.A , Glu 320.A , and Gln 312.A seem to rigidify the conformation of the Mtbd Qc , it is conceivable that the Qh3-PL8 interaction in Cgbd might be weaker than in Mtbd. In consequence, the local charge and steric environment in close proximity of the wellconserved Gln 260.A Cgbd and Lys 263.A Cgbd residues of Qh1, which have been suggested to be important for quinol oxidation in Ecbd-I, is significantly different (Mogi et al., 2006). These structural dissimilarities may be first indications that site 1 could represent the quinol binding and oxidation region in Cgbd. The increased movement of the Cgbd Q N region reflected by unclearly-resolved densities of residues 268 to 317 (CydA) may support a model in which site 1 represents the

FIGURE 4
Comparison of the Qh3-PL8 region in (A) Mtbd and (B) Cgbd. The stabilizing residues E320, Y321 and R324 of Qh3 in Mtbd (7NKZ) (Safarian et al., 2021) are replaced by alanines in Cgbd, presumably leading to a weaker interaction between Qh3 and PL8. Frontiers in Chemistry frontiersin.org electrochemically active substrate-binding domain. In fact, quinol oxidase activity of Cgbd is about 200-400 s −1 (Kusumoto et al., 2000), considerably higher than that of Msbd (22 s −1 ) (Wang et al., 2021). In contrast to the structures of Cgbd and Ecbd-I, the Q-loop domain is fully resolved in Mtbd (Safarian et al., 2021). This may be because the Q N region of Mtbd contains an internal disulfide bond which contributes to an overall increased stability and rigidity. Based on this unique finding, we previously suggested that the existence or absence of a disulfide bond within the Q-loop domain may correlate with utilization of either site 1 or site 2 as the electron entry route from substrate quinols (Safarian et al., 2021). Interestingly, these two cysteine residues (Cys 271.A Cgbd and Cys 290.A Cgbd ) are conserved in Cgbd, yet our structure suggests that they are not forming a covalent bond, furthermore raising the question about the mechanism of disulfide bond generation and breakage within Q-loops of cytochromes bd in a physiological and metabolic context.
In Mtbd, site 2 is occupied by a MK-9 group, while it is appearing empty in Cgbd ( Figure 5) (Safarian et al., 2021). At higher contour levels a weak density might be present, yet as judged by its size and shape at the current resolution, it more likely represents a bound lipid molecule rather than a quinone group (Supplementary Figure S3B). Nevertheless, the local environment formed by TMH1, TMH9 and heme b 595 in Cgbd is almost identical to that of Mtbd. The residues forming the binding pocket for the naphthoquinone head group of MK-9 in Mtbd are conserved in Cgbd (Arg 8.A Cgbd , Trp 9.A Cgbd and Met 405.A Cgbd ), and thus site 2 may still represent a presumable secondary MK-9 biding site. In Ecbd-I, site 2 is occupied by the accessory subunit CydH, which forms a different surface at this peripheral site of the cytochrome bd complex and consequently does not form any cavities which would allow for quinol binding close to the heme b 595 group. The absence of the CydH subunit and the resulting access to site 2 might furthermore be the reason why we could previously show that the addition of menaquinones modulates the activity of Ecbd-II in a concentration dependent manner (Grund et al., 2021).
We have previously reported that preincubation of Cgbd and Gtbd with quinones prior to induction of catalytic oxygen reduction activity enhances the turnover of the enzyme (Sakamoto et al., 1999;Kusumoto et al., 2000). The activity enhancement observed in these cases may be explained by the binding of quinones to site 2, the putative regulatory or acceleratory quinone binding site. It appears that the access of quinones to site 2 is possible in Gtbd and Cgbd since no accessory subunits are found at this position in the X-ray and cryoEM structures, respectively. However, Trp 9.A Cgbd/Mtbd , which seems important for forming stacking interactions with the naphthoquinone head group of bound MK-9, is not conserved in Gtbd. The absence of the Trp residue could explain the different preference of preincubated quinone species observed for Cgbd and Gtbd (Sakamoto et al., 1999;Kusumoto et al., 2000).
The MK-9 that binds to site 2 of Mtbd is derived from endogenous sources, and its amount may vary depending on cultivation conditions and sample preparation procedures such as purification methods and the choice of detergents. Future studies are required to characterize site 2 interactions between the protein environment, the heme b 595 group, and exogenously added quinones in greater detail.
In conclusion, the atomic structure of the cytochrome bd oxidase from C. glutamicum confirms common features among actinobacterial cytochrome bd oxidases: (i) the arrangement of three prosthetic heme cofactors in a triangular manner, (ii) the absence of accessory subunits, (iii) the location of the H-and the O-channels, (iv) structural features of a second quinone binding site (site 2). On the other hand, it is also reaffirmed that there are considerable differences between cytochrome bd oxidases of Actinobacteria and those of Firmicutes and Proteobacteria. To gain a deeper and more precise understanding of the catalytic mechanisms of the so-far characterized cytochrome bd homologs of Actinobacteria, follow-up structural studies of cytochromes bd in the presence of substrate quinols or substrate-derived inhibitors are indispensable.

Materials and methods
Materials and production of cytochrome bd from C. glutamicum Chemicals were purchased from Sigma unless otherwise stated. Cytochrome bd from C. glutamicum was produced in C. glutamicum subsp. lactofermentum (ΔctaD/pPC4-cydABCD). This production strain was established as described previously (Kabashima et al., 2009). In brief, the entire cydABDC genes including its authentic promoter region were amplified from genomic DNA of C. glutamicum by polymerase chain reaction (PCR) and inserted into the E. coli-C. glutamicum shuttle vector, pPC4. The cytochrome aa 3 -deficient mutant of C. glutamicum (ΔctaD) was transformed with the resulting plasmid, pPC4-cydABDC. Transformants were selected on solid medium containing 50 μg/ml kanamycin and 10 μg/ml chloramphenicol. Individual colonies were cultured in liquid medium for 30 h and the clone with the highest heme d content in membrane fractions was selected. Expression of cytochrome bd coding genes is induced at late stationary phase, in response to the low oxygen level due to its native promoter. After harvesting, cells were resuspended in 10 mM NaPi buffer at pH 7.0 containing 0.5% (w/v) NaCl. Cell disruption was performed via vigorous mixing with glass beads (diameter: 0.18 mm) in a cell disrupting mixer (Bead-Beater, Biospec). Subsequently, low-speed centrifugation (5,000 g, 10 min, 4°C) was carried out to remove undisrupted cells and cell debris. The supernatant was then subjected to ultracentrifugation (100,000 g, 30 min, 4°C). Pelleted membranes Frontiers in Chemistry frontiersin.org were resuspended in 10 mM NaPi buffer at pH 7.0 and used for further protein purification.
Purification of cytochrome bd from C. glutamicum The membrane fraction was resuspended at 10 mg/ml in a first washing buffer [10 mM NaPi (pH 7.4), 0.5 M NaCl, 0.1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM benzamidine] and stirred for 30 min at 4°C. The suspension was subjected to ultracentrifugation (100,000 g, 20 min, 4°C) and the pelleted membranes were resuspended at 10 mg/ml in a second washing buffer [10 mM NaPi (pH 7.4), 1.5% Na-cholate, 0.5% Nadeoxycholate and 0.1 M NaCl] and stirred for 30 min at 4°C. The membranes were sedimented at 100,000 g for 20 min and the pellet was subsequently resuspended at a concentration of 5 mg/ml in a solubilization buffer [10 mM NaPi (pH 7.4), 10% glycerol, 1% n-dodecyl-β-D-maltoside (DDM), 0.1 mM PMSF and 1 mM benzamidine]. Solubilization was carried out for 90 min at 4°C. The solubilized membrane fraction was separated from remaining crude membranes by centrifugation at 100,000 g for 20 min at 4°C.
For sample polishing, the sample was thawed on ice, subjected to centrifugation (21,130 g, 10 min, 4°C) and applied to a Superdex ™ 200 Increase 10/300 GL column (GE Healthcare), which had been pre-equilibrated with 3 CV of a buffer containing 20 mM Tris/HCl (pH 7), 100 mM NaCl and 0.002% (w/v) LMNG. The chromatography was performed on an Äkta ™ pure system (GE Healthcare) with a flow rate of 0.5 ml/min and a fraction volume of 0.2 ml. Peak fractions were pooled and concentrated to 2.5 mg/ml for downstream cryoEM sample preparation.
Single-particle cryoEM sample vitrification and data acquisition Sample vitrification Quantifoil R1.2/1.3 300 mesh Au-carbon grids were washed in chloroform and subsequently glow discharged with a PELCO easiGlow device at 15 mA for 90 s. A volume of 4 μL sample was applied on a grid immediately before plunge freezing. Samples were vitrified at 4°C, 100% humidity, and a blot force of 20 using a Vitrobot IV device (Thermo Scientific). Blotting was carried out for 4 s before plunge-freezing in liquid ethane.

Data collection
Electron micrographs were recorded using a Titan Krios G3i microscope operated at 300 kV (Thermo Scientific). Data were acquired automatically using EPU software (Thermo Scientifc) with aberration free image shift (AFIS) in faster acquisition mode using a K3 direct electron detector in combination with a postcolumn energy-imaging filter (Gatan) at a nominal magnification of ×105,000, corresponding to a calibrated pixel size of 0.837 Å. Dose-fractionated movies were recorded for 4.99 s at an electron flux of 15 e − x pixel −1 x s −1 , corresponding to a total dose of about 107 e − x A −2 . A focus range of −1.1 to −2.1 μm was applied throughout data collection.

Data processing
Movies were motion corrected and dose weighted with MotionCor2 using a 5 × 5 patch (Zheng et al., 2017). Initial CTF (contrast transfer function) parameters for each movie were determined with Gctf (Zhang, 2016). At this step images with estimated poor resolution (>3.2 Å) and high astigmatism (>250) were removed. Autopicking was performed using crYOLO (Wagner et al., 2019) and all further data processing including 2D classification, 3D classification, 3D refinement and Bayesian polishing was performed using Relion 3.1 (Zivanov et al., 2018). A total of 10,586,211 particles were picked and cleaned up with four rounds of two-dimensional (2D) classification. One round of 3D classification provided the final particle stack of 410,010 particles which was used for all subsequent rounds of auto-refinement, CTF refinement and Bayesian polishing. Final maps were generated after subtraction of the detergent micelle followed by CTF refinement. An overview of the processing workflow is given in Supplementary Figure S2.

Model building and map validation
All model-building steps were performed using COOT (version 0.9.6) (Emsley et al., 2010). The cryoEM structure of cytochrome bd from Mycobacterium tuberculosis (pdb: 7NKZ) was used as starting model (Safarian et al., 2021). After backbone fitting and side chain docking, we performed real-space refinement in Phenix (version 1.18.2). Refinement results were manually inspected and corrected if necessary. Map-to-model cross validation was performed in Phenix Frontiers in Chemistry frontiersin.org (version 1.18.2). FSC 0.5 was used as cut-off to define resolution. Model parameters and corresponding cryoEM map statistics are summarized in Supplementary Table S1. The finalized model was visualized using ChimeraX (Goddard et al., 2018). Tunnels and interior cavities were mapped with MOLE2.5 (bottleneck radius: 1.1 Å, bottleneck tolerance: 3 Å, origin radius 5 Å, surface radius 8 Å, probe radius 5 Å, interior threshold 1.1 Å) (Sehnal et al., 2013;Pravda et al., 2018).

Data availability statement
The cryoEM map file of cytochrome bd from C. glutamicum was deposited in the EMD database and can be found under accession number EMD-15851. The model file of cytochrome bd from C. glutamicum was deposited to the PDB under accession number 8B4O.

Author contributions
TG optimized sample conditions, prepared grids, collected cryoEM data, processed cryoEM data, refined the structure, built the model, co-drafted the manuscript, and prepared figures. YK and TK produced enzyme, established purification, and codrafted the manuscript. DW collected cryoEM data. SW calibrated and aligned the microscope. JS, HM, and SS initiated, designed and supervised the project. JS, HM, and SS evaluated data, and drafted the manuscript.

Funding
This work was supported by the Max Planck Society, the Nobel Laureate Fellowship of the Max Planck Society, and by a Grant-in-Aid for Scientific Research (C) (16K07299 to JS) from the Japan Society for the Promotion of Science.