Recent Advances in Structural Studies of Cytochrome bd and Its Potential Application as a Drug Target

Cytochrome bd is a triheme copper-free terminal oxidase in membrane respiratory chains of prokaryotes. This unique molecular machine couples electron transfer from quinol to O2 with the generation of a proton motive force without proton pumping. Apart from energy conservation, the bd enzyme plays an additional key role in the microbial cell, being involved in the response to different environmental stressors. Cytochrome bd promotes virulence in a number of pathogenic species that makes it a suitable molecular drug target candidate. This review focuses on recent advances in understanding the structure of cytochrome bd and the development of its selective inhibitors.


Introduction
The terminal oxidase cytochrome bd (EC 7.1.1.7) is so far found only in the electron transport chains (also known as the respiratory chains) of bacteria and archaea [1]. Notably, even some microorganisms considered to be strict anaerobes possess the active bd oxidase to respire, which enhances their growth [2,3]. Cytochrome bd reduces molecular oxygen to water at the expense of the concomitant oxidation of species-specific quinols such as ubiquinol, menaquinol, and possibly plastoquinol [4,5]. The chemical energy released in this redox reaction is conserved in the form of a proton motive force across a protonimpermeable prokaryotic membrane [6][7][8]. The proton electrochemical gradient is used by the cell to drive ATP synthesis and to do other useful work. The bd oxidase is made up of two to four polypeptide chains, depending on the species. The catalytically active subunit bears the three heme prosthetic groups, b 558 , b 595 , d, and the Q-loop. [9][10][11][12][13][14][15]. The latter is a quinol-binding domain located in the hydrophilic loop region between the transmembrane helices 6 and 7. Until recently, the classification of the cytochrome bd family was based on the size of the Q-loop. Accordingly, it was divided into two subfamilies: L (long Q-loop) and S (short Q-loop) [16,17]. However, very recently the phylogeny of the bd family has been reassessed by Murali et al. [1]. As a result, three families and several subfamilies within the cytochrome bd superfamily have been identified (see [1] for details). Heme d is the site for O 2 binding, usually with a very high affinity, and activation for reduction to 2H 2 O [18,19]. Heme b 595 facilitates heme d to carry out the catalytic reaction, and there is significant interaction between these hemes [20][21][22][23][24]. Cytochrome bd is phylogenetically unrelated to the heme-copper oxidases and alternative oxidases, the two other groups of terminal oxidases which use O 2 as the final electron acceptor [16,. Unlike the heme-copper oxidases, the bd enzyme does not pump protons and generates the proton motive force solely by transmembrane charge separation [48][49][50][51]. The role of most heme-copper enzymes is likely limited to energy production, while cytochrome bd has other, alternative functions [52]. The ability of cytochrome bd to consume O 2 at high Structural analyses of bacterial cytochrome bd oxidases by means of X-ray crystallography (Geobacillus thermodenitrificans, pdb IDs 5DOQ & 5IR6, [9]) and cryoelectron microscopy (cryo-EM) (Escherichia coli bd-I, pdb IDs 6RKO [11] and 6RX4 [10]; Mycobacterium smegmatis, 7D5I [12]; Mycobacterium tuberculosis, 7NKZ [13] and E. coli bd-II, 7OSE [14] and 7OY2 [15]) have shed light on the architecture of this class of oxidases.With the exception of the Geobacillus structures 5DOQ and 5IR6 that were obtained by X-ray crystallography, all structural data are based on cryo-EM. Resolutions range from 3.8 Å (5IR6, X-ray crystallography) to 2.5 Å (7NKZ, cryo-EM), providing nearatomic details (for a detailed discussion of optical resolutions in X-ray crystallography and cryo-EM, we suggest the work of Dubach and Guskov [85]). This work has unveiled common features shared by all members of the bd oxidase class that are distinctly different from the members of the HCO and AOX classes. At the same time, however, the bd oxidase structures also disclosed intriguing peculiarities that seem to reflect the differing demands on these enzymes in dependence of the metabolic context, in which they are active. It appears that the bd oxidases found in different bacteria represent a variation of a common theme: beyond the shared, basic architecture they differ in the number of the additional subunits, in the arrangement of the hemes and in the position of the quinol binding site.

Common Architecture of Cytochrome bd Oxidases
Overall, in phyla with available structural data on bd oxidases, i.e., Actinobacteria, Firmicutes, and Proteobacteria, the structural core consists of two large subunits with an Structural analyses of bacterial cytochrome bd oxidases by means of X-ray crystallography (Geobacillus thermodenitrificans, pdb IDs 5DOQ & 5IR6, [9]) and cryo-electron microscopy (cryo-EM) (Escherichia coli bd-I, pdb IDs 6RKO [11] and 6RX4 [10]; Mycobacterium smegmatis, 7D5I [12]; Mycobacterium tuberculosis, 7NKZ [13] and E. coli bd-II, 7OSE [14] and 7OY2 [15]) have shed light on the architecture of this class of oxidases.With the exception of the Geobacillus structures 5DOQ and 5IR6 that were obtained by X-ray crystallography, all structural data are based on cryo-EM. Resolutions range from 3.8 Å (5IR6, X-ray crystallography) to 2.5 Å (7NKZ, cryo-EM), providing near-atomic details (for a detailed discussion of optical resolutions in X-ray crystallography and cryo-EM, we suggest the work of Dubach and Guskov [85]). This work has unveiled common features shared by all members of the bd oxidase class that are distinctly different from the members of the HCO and AOX classes. At the same time, however, the bd oxidase structures also disclosed intriguing peculiarities that seem to reflect the differing demands on these enzymes in dependence of the metabolic context, in which they are active. It appears that the bd oxidases found in different bacteria represent a variation of a common theme: beyond the shared, basic architecture they differ in the number of the additional subunits, in the arrangement of the hemes and in the position of the quinol binding site.

Common Architecture of Cytochrome bd Oxidases
Overall, in phyla with available structural data on bd oxidases, i.e., Actinobacteria, Firmicutes, and Proteobacteria, the structural core consists of two large subunits with an invariant fold, denoted as CydA (approximately 60 kDa) and CydB (approximately 40 kDa) in most bd oxidases and as AppC and AppB in the second bd oxidase (bd-II) in E. coli. These subunits are arranged in a two-fold rotational pseudo-symmetry, with both subunits comprising nine transmembrane (TM) helices each. These nine TM helices are arranged in two membrane-spanning four-helix bundles and an additional peripheral transmembrane (TM) helix ( Figure 1). Beyond this, all three heme groups, denoted as hemes b 558 , b 595 , and heme d, are located in CydA (or AppC, respectively). They are arranged in a triangle with two heme groups oriented linearly in a plane and the third one sitting orthogonally above this plane (Figure 2A,B). Despite its structural homology to CydA/AppC, the second large subunit CydB/AppB does not carry any metal-containing cofactors. Instead, for CydB of E. coli bd-I and AppB of E. coli bd-II, it has been found that a ubiquinone-8 or demethylmenaqinone-8 occupies a position equivalent to the heme binding sites in CydA/AppC ( Figure 2C) [10,11,14,15], perfectly tracing the positions of the hemes in the opposing subunit. Hence, it has been postulated that this quinone has mainly stabilizing properties. This assumption is supported by the fact that mutations of amino acid residues in the ubiquinol-binding groove of E. coli CydB lead to a loss of activity [10]. The quinol must be very tightly bound to CydB because ubiquinone is not displaced by a high excess of aurachin D, a specific quinol-site inhibitor [10], and demethylmenaquinone is not replaced from AppB by a high excess of menaquinone-8 [15]. However, the presence of demethylmenaquinone-8 in one of the published E. coli bd-II structures (pdb ID 7OY2) was additionally suggested to indicate a close association of bd-II biosynthesis with menaquinone synthesis pathways [15]. In that study, menaquinones were discussed as co-substrates of bd-II in aerobic respiration, when molecular oxygen is limited, thus extending the role of menaquinones in bd-II biosynthesis and enzyme catalysis. The observed divergence of the type of quinol found in AppB of E. coli bd-II, with binding of ubiquinol in one structure [14] and demethylmenaquinone in the other [15], is most likely due to different conditions of cell growth and protein production. Importantly, the bd oxidases of Mycobacteria lack these quinones bound to CydB [12,13]. Instead, they contain multiple tryptophan and phenylalanine residues that take this place, creating a stabilizing network of van-der-Waals contacts and making quinone binding redundant ( Figure 2D).
All bd oxidases feature an additional structural element that is, in most cases, crucial for substrate binding, that being the Q loop [16,86]. It is inserted between TM helix 6 and 7 of CydA/AppC in close proximity to heme b 558 ( Figure 3). Accordingly, it has been proposed that the Q-loop is involved in the binding of the quinol substrate (from which the name is derived). The N-terminal region of the Q-loop shows a higher amount of conserved amino acid residues than the C-terminal portion [87,88]. Hydrogen/deuterium exchange mass spectrometry measurements demonstrated that the common N-terminal domain of the Q-loop is intrinsically flexible [11], while its C-terminal extension present in the members of the L-subfamily is rather rigid and extends all over the periplasmic surface of CydA to the CydAB interface [10,11]. This implies a functional role for the N-terminal domain of the Q-loop and a structural role for its C-terminal extension. These differences might explain why exchanging the Q-loop between various forms of bd oxidases is not possible, with the only exception being the Q-loop of E. coli bd-II that may replace that of E. coli bd-I [89,90]. Importantly, bd oxidase from M. tuberculosis and M. smegmatis are usually classified as belonging to the S-subfamily [12]. However, they contain an extra insertion of eight amino acid residues in a periplasmic loop of CydA, which represents a rigid unit [13]. These extra residues interact with a small helix called Qh3 of the C-terminal domain of the Q-loop via a cluster of hydrophobic amino acid residues [13]. It is known that some of these residues are essential for the activity of the mycobacterial enzyme [91]. This arrangement is further stabilized by several H-bonds, implying that these interactions are most likely not involved in quinol oxidation. Furthermore, the N-terminal part of the mycobacterial Q-loop is fixed by a unique disulfide bond. Altogether, the distinctive flexibility of the Q-loop that is expected to be essential for its function in quinol oxidation in Proteobacteria is drastically diminished in the mycobacterial enzyme. Thus, the restricted conformational dynamics of the Q-loop most likely prevents the binding of quinol ( Figure 3) [13]. [15]. However, the presence of demethylmenaquinone-8 in one of the published E. coli bd-II structures (pdb ID 7OY2) was additionally suggested to indicate a close association of bd-II biosynthesis with menaquinone synthesis pathways [15]. In that study, menaquinones were discussed as co-substrates of bd-II in aerobic respiration, when molecular oxygen is limited, thus extending the role of menaquinones in bd-II biosynthesis and enzyme catalysis. The observed divergence of the type of quinol found in AppB of E. coli bd-II, with binding of ubiquinol in one structure [14] and demethylmenaquinone in the other [15], is most likely due to different conditions of cell growth and protein production. Importantly, the bd oxidases of Mycobacteria lack these quinones bound to CydB [12,13]. Instead, they contain multiple tryptophan and phenylalanine residues that take this place, creating a stabilizing network of van-der-Waals contacts and making quinone binding redundant ( Figure 2D).  Although at least parts of the Q-loop were successfully modelled in all available structures, only for a single member of the L-subfamily, E. coli bd-II, has the Q-loop been structurally resolved, and this is only because aurachin D was added, which fixated the flexible loop in a defined conformation [14]. However, the structural work available to date shows that the lid-like position on subunit CydA/AppC on the periplasmic or extracellular side seems to be a general feature of the Q-loop over all phyla ( Figure 3). mycobacterial enzyme [91]. This arrangement is further stabilized by several H-bonds, implying that these interactions are most likely not involved in quinol oxidation. Furthermore, the N-terminal part of the mycobacterial Q-loop is fixed by a unique disulfide bond. Altogether, the distinctive flexibility of the Q-loop that is expected to be essential for its function in quinol oxidation in Proteobacteria is drastically diminished in the mycobacterial enzyme. Thus, the restricted conformational dynamics of the Q-loop most likely prevents the binding of quinol ( Figure 3) [13].  . Instead, a quinol binding site was identified close to heme b 595 , where menaquinone-9 was found to interact with CydA (green surface, residual protein surface in light salmon).

Additional Subunits
Proteobacterial cytochrome bd oxidases contain at least one additional single-helix TM subunit [92,93] that was shown to be crucial for the assembly, stability, and activity of the enzyme [93]. In G. thermodenitrificans bd oxidase, this subunit (approximately 4 kDa) is called CydS, while in E. coli, it has been denoted as CydX (bd-I) or AppX (bd-II), respectively. It was found to be oriented alongside the TM helices of CydA/AppC and to bind to that subunit through numerous interactions with TMs 1, 4, 5, 6, and 7 ( Figure 4A-C). The intricate interaction pattern appears to be the cause for the requirement of CydS/CydX/AppX for the stable assembly and activity of proteobacterial bd oxidases. In one of the cryo-EM structures, bd-II reconstituted into amphipols is mainly present as a dimeric species [14]. Dimerization is mediated by several weak hydrophobic interactions between the small AppX subunits. It is noteworthy that none of the other structures reported a dimeric version of a bd oxidase. However, the ability of E. coli bd-II to dimerize does not change the overall position of AppX with respect to AppC as compared to the homologous pair of CydX/CydA in E. coli bd-I, nor does it affect any of the protein-protein interactions. Interestingly, the homologous subunit CydS in G. thermodenitrificans is slightly tilted towards TM6 of CydA ( Figure 4A), resulting in a maximum shift of 4-5 Å at the periplasmic side. Although this tilt excludes TM4 from binding, the general function of CydS was suggested to be equivalent to the E. coli bd oxidases [9].
In addition to CydX/CydS/AppX, E. coli bd-I oxidase contains another single-helix subunit called either CydY or CydH [10,11]. This small subunit (approximately 3 kDa) binds in a cleft between TM helices 1 and 9 of CydA ( Figure 4D). This subunit is encoded by the former orphan gene ynhF that is neither part of the cyd nor the app operon encoding both E. coli bd oxidases. Homologues of this subunit have previously been identified in the proteobacterial clade, where they coincide with the presence of the L-subfamily bd oxidases [11]. In contrast to CydX/CydS/AppX that mainly seems to play a structural role, CydY/CydH appears to be of functional importance, as it blocks the putative oxygen access site of G. thermodenitrificans bd oxidase (see below). This already implies a different oxygen entry to the E. coli and the G. thermodenitrificans bd oxidases.
reported a dimeric version of a bd oxidase. However, the ability of E. coli bd-II to dimerize does not change the overall position of AppX with respect to AppC as compared to the homologous pair of CydX/CydA in E. coli bd-I, nor does it affect any of the protein-protein interactions. Interestingly, the homologous subunit CydS in G. thermodenitrificans is slightly tilted towards TM6 of CydA ( Figure 4A), resulting in a maximum shift of 4-5 Å at the periplasmic side. Although this tilt excludes TM4 from binding, the general function of CydS was suggested to be equivalent to the E. coli bd oxidases [9].  (B,C) CydX/AppX of E. coli bd-I and bd-II bind in a nearly identical manner to CydA/AppC, underlining the close homology of both bd oxidases. As compared to G. thermodenitrificans bd, the N-terminus (top) of CydX/AppX is tilted further away from heme b 558 . (D) Subunit CydH (limon), exclusive to E. coli bd-I, is found at the opposite side of CydA (light green) and appears to be further stabilised by a glycerophospholipid (shown as sticks). Again, mainly hydrophobic interactions contribute to binding.

Quinol Binding Site(s)
The substrate quinol is usually bound and oxidized at the interface of the Q-loop and TM helices 6 and 7 of CydA. It has been established that the Q-loop fulfils a critical function in the structural arrangement of the propionate group of heme b 558 relative to the quinol [87]. In addition, the Q-loop contains a conserved glutamate and lysine residue that are essentially involved in the binding and oxidation of the quinol [87]. The high flexibility of the N-terminal part of the Q-loop is most likely needed for the fast and transient binding of the quinol. So far, there is no structure of a bd oxidase with a quinol tightly bound to the Q-loop, which is most likely due to its extremely dynamic binding. Several attempts were made to obtain a structure with a bound Q-site inhibitor or a bound quinone by incubating the enzyme preparation with a huge molar excess of aurachins, quinolone-type inhibitors, and ubiquinone and menaquinone [10,11,15]. However, none of the substances gave rise to an additional electron density close to the Q-loop after cryo-EM. The only exception is the binding of aurachin D to E. coli bd-II oxidase [14]. Here, an extra electron density was detected that unambiguously derived from aurachin D ( Figure 3A). This might be due to the extremely high affinity of aurachin D to this oxidase of only about 7 nM [14] and, thus, the only available inhibitor with an IC 50 below 10 nM. The addressed protein surface features a pronounced complementarity to the inhibitor. Binding is mostly mediated by hydrophobic interactions, but a single H-bond to an aspartate, which is conserved within Proteobacteria but not in Mycobacteria, adds to the strong binding. Mutation of this single residue to an asparagine residue led to a more than fivefold decrease of the IC 50 and diminished the enzyme activity to one fourth [14]. Unexpectedly, a menaquinone-9 molecule in its oxidized form was detected in the unusual Q-loop of the M. tuberculosis oxidase close to the unique disulfide bond ( Figure 3C) [13]. This interaction is further supported by the porphyrin scaffold of heme b 595 and several residues from TM helices 1 and 9 [13]. The naphtoquinone headgroup interacts with a methionine residue close to the extra eight amino acid residues inserted in a periplasmic loop of CydA and a tryptophan residue at the N-terminus of CydA. These residues are only conserved among Myobacteria. It is noteworthy that this menaquinone binding site is fully occupied by CydY/CydH in E. coli bd oxidases. Molecular dynamic (MD) simulations imply that either an oxidized or a reduced menaquinone may bind at this position [13]. Thus, it seems that in Mycobacteria, the Q-loop is structurally fixed and incapable of quinol binding and reduction, while the open access to heme b 595 might be used as an alternative electron pathway leading directly to heme d, thus bypassing heme b 558 [13].

Heme Arrangement and Electron Transfer
Despite their common architecture and although the heme binding sites themselves are conserved in all available structures, bd oxidases further differ in the heme arrangement within the triangle (Figure 2A,B). While the low-spin heme b 558 , the first of the two hemes arranged on a plane, is always found closest to the Q-binding site near the Q-loop and its axial ligands, histidine and methionine are conserved, and the second and the third heme groups, high-spin hemes b 595 and d, are located at different positions within the triangle depending on the species.
The bd oxidase from G. thermodenitrificans harbors heme b 595 in the plane of heme b 558 [9]. Here, both axial coordination sites of the central Fe-atom are occupied by histidine and glutamate residues (Figure 2A). The glutamate residue homes in onto the Fe atom from the distal side of heme b 558 and the Q-binding site to fully occupy the axial ligand site at an O-Fe distance of 2.1 Å. Oxygen, the terminal electron acceptor, is prohibited any access to that heme. The orthogonally placed heme d, however, does feature an oxygen cavity lined by a threonine and a leucine residue. Heme d is situated closely enough to the protein-membrane interface to allow for rapid diffusion of oxygen from the membrane. Hence, this heme most likely is the site where molecular oxygen is reduced to water.
Unexpectedly, the heme orthogonally placed to heme b 558 , which is heme d in G. thermodenitrificans [9], is heme b 595 in all other structures [10][11][12][13][14][15]. Again, the strictly conserved glutamate residue serves as an axial ligand proximal to heme b 558 ( Figure 2B) and, equivalent to G. thermodenitrificans heme d, the distal coordination site of the Fe atom is always found unoccupied. However, in all species harbouring heme b 595 at this position, a phenylalanine approaches the coordination site to a distance of 3.4 Å, enough for hydrophobic contacts to the porphyrin moiety to occur, but too little for binding of molecular oxygen. Hence, heme b 595 found in that position is not the dioxygen reduction site. Instead, this is heme d, coordinated by the well conserved histidine residue as first axial ligand proximal to heme b 558 ( Figure 2B) [10][11][12][13][14][15]. The distal side of heme d, however, does not have a proteinaceous ligand, but was found to be either unoccupied or to contain an extra electron density that was interpreted as molecular oxygen [10][11][12][13][14][15]. Here, the coordinating glutamate residue is located more remotely at a distance of at least 4.8 Å to the central Fe atom. The shift of the coordinating glutamate originates from an inserted leucine residue (Leu101 in E. coli bd-I and bd-II and M. smegmatis bd; Leu100 in bd of M. tuberculosis; this residue is lacking in G. thermodenitrificans bd oxidase) into TM helix 3 of CydA, resulting in its stronger curvature. This generates a voluminous cavity for the binding of dioxygen at the axial position of heme d, which is enclosed by hydrophobic isoleucine and phenylalanine residues ( Figure 2B). Thus, an ideal environment for binding a dioxygen molecule is provided. Since this site is buried deeply inside the enzyme core, it requires an oxygen channel, and indeed, such a channel leads straight from the membrane through to heme d (see below).
For the reduction of dioxygen to water, four protons and four electrons are needed. From the triangular arrangement of and the distances between the hemes in bd oxidases, conclusions were drawn on the electron transfer and the mechanism of dioxygen reduction. As stated above, the heme arrangement in G. thermodenitrificans differs from that in all other bd oxidases. Here, the short distance between heme b 558 close to the quinol binding site and the active site heme d implies a direct electron transfer from b 558 to d. The electron would then equilibrate between the d and the b 595 hemes [9]. In the reduced enzyme, two electrons may derive from heme d via the oxoferryl-state, one from b 595 and the fourth electron from the macrocycle of heme d [9,23,24]. This very rapid or simultaneous four-electron transfer mechanism is reminiscent of that of HCOs preventing the formation of reactive oxygen species [9,23]. The other bd oxidases feature a short distance between the two b-type hemes, implying a sequential electron transfer from b 558 via b 595 to heme d. Reduction of dioxygen is then catalyzed by a short-lived peroxide intermediate in agreement with spectroscopic data [10,11,19,23].

Oxygen Access
The bd oxidases from M. tuberculosis and M. smegmatis consist only of subunits CydA and CydB [12,13]. Although it cannot be completely excluded that additional small subunits being part of these oxidases are encoded in the respective genomes, no orphan genes have been identified in the mycobacterial databases that may code for such subunits [13]. Furthermore, these oxidases are fully assembled, stable and catalytically active, although they do not contain another subunit [12,13]. But while in E. coli bd-I oxidase, oxygen access to heme b 595 is blocked by CydY/CydH ( Figure 4C) [10,11], and, instead, molecular oxygen is directed through a hydrophobic channel directly to heme d ( Figure 5A), E. coli bd-II oxidase and the mycobacterial bd oxidases, which do not have such a subunit, rely on an intrinsic barrier that is present in all bd oxidases, an isoleucine residue (Ile143 in M. smegmatis and M. tuberculosis, Ile144 in E. coli bd-I and bd-II, Ile146 in G. thermodenitrificans, Figure 5C-E). Hence, although heme b 595 is freely accessible from the membrane due to the lack of a small subunit, all enzymes with the exception of that from G. thermodenitrificans utilize an oxygen channel, leading from CydB to heme d bound to CydA ( Figure 5B), just as described for E. coli bd-I [10,11,14,15]. Accordingly, this small hydrophobic channel, always starting above a conserved tryptophan residue on CydB/AppB and extending further to CydA/AppC, is a common feature of all oxidases except for the one from G. thermodenitrificans. Common to these oxygen channels is a constriction that may act as a selectivity filter disabling the passage of angled molecules. The small channel leading directly to heme b 595 was proposed as a second oxygen access channel in the structure of the M. smegmatis bd oxidase [12]. However, experimental evidence for this hypothesis has not yet been provided. It is worthy of note that this is the position for which it is assumed that it binds menaquinol in M. tuberculosis bd oxidase (see above) [13]. As opposed to this, oxygen access to bd oxidase from G. thermodenitrificans is principally different due to the re-arrangement of the b 595 and d heme groups [9][10][11]. Here, heme d is freely accessible from the membrane via a very short channel, making an additional channel as found in all other oxidases superfluous ( Figure 5B). although they do not contain another subunit [12,13]. But while in E. coli bd-I oxidase oxygen access to heme b595 is blocked by CydY/CydH ( Figure 4C) [10,11], and, instead molecular oxygen is directed through a hydrophobic channel directly to heme d ( Figur  5A), E. coli bd-II oxidase and the mycobacterial bd oxidases, which do not have such subunit, rely on an intrinsic barrier that is present in all bd oxidases, an isoleucine residu (Ile143 in M. smegmatis and M. tuberculosis, Ile144 in E. coli bd-I and bd-II, Ile146 in G thermodenitrificans, Figure 5C-E). Hence, although heme b595 is freely accessible from th membrane due to the lack of a small subunit, all enzymes with the exception of that from G. thermodenitrificans utilize an oxygen channel, leading from CydB to heme d bound to CydA ( Figure 5B), just as described for E. coli bd-I [10,11,14,15]. Accordingly, this smal hydrophobic channel, always starting above a conserved tryptophan residue on CydB/AppB and extending further to CydA/AppC, is a common feature of all oxidase except for the one from G. thermodenitrificans. Common to these oxygen channels is constriction that may act as a selectivity filter disabling the passage of angled molecules The small channel leading directly to heme b595 was proposed as a second oxygen acces channel in the structure of the M. smegmatis bd oxidase [12]. However, experimenta evidence for this hypothesis has not yet been provided. It is worthy of note that this is th position for which it is assumed that it binds menaquinol in M. tuberculosis bd oxidase (se above) [13]. As opposed to this, oxygen access to bd oxidase from G. thermodenitrificans i principally different due to the re-arrangement of the b595 and d heme groups [9][10][11]. Here heme d is freely accessible from the membrane via a very short channel, making an additional channel as found in all other oxidases superfluous ( Figure 5B).

Proton Pathways
In addition to electrons delivered from quinol, the reduction of dioxygen to water requires protons. As the reaction of bd oxidases is electrogenic [8], protons for dioxygen reduction are taken up at the cytosolic side of the membrane, while the protons generated during quinol oxidation are released to the periplasmic side of the membrane. To enable a fast oxygen reduction, the oxygen channel and the proton pathway(s) should meet at the open coordination side of heme d. It was proposed that the oxygen channel may also conduct protons by a connection with a proton pathway [94]. In G. thermodenitrifcans bd oxidase, the proton pathway leads to heme b 595 , and from there the protons are most likely further transferred to the d heme [9]. That implies that all proton pathways lead to the heme group that is located the closest to the cytosol. And indeed, the proton pathways in the various bd oxidases are very similar to each other and start at a broad and shallow hydrophilic cavity at the cytosolic side of the CydAB/AppCB interface [10][11][12][13][14][15]. This cavity narrows to a channel of about 4 Å diameter and runs perpendicular to the membrane, flanked by the TM helices 2 and 3 of CydA/AppC and CydB/AppB, respectively ( Figure 6). A series of serine, glutamate and aspartate residues and several water molecules lead directly to the propionate group of heme d [10][11][12][13][14][15]. Remarkably, this proton pathway splits up to create a second branch that leads to the oxygen channel in CydB/AppB. This branch ends at a conserved aspartate residue that is essential for the activity of E. coli bd-I, a member of the L-subfamily of bd oxidases [95]. Due to the large distance of 20 Å of this residue to heme d, its direct participation in dioxygen reduction is unlikely. It rather acts as proton storage or is required for charge compensation [15]. An additional solvent-accessible area was detected on the periplasmic side of the M. tuberculosis oxidase by MD simulations [13]. Here, water molecules located between the TM helices 5 and 6 of CydA connect the propionate group of b 558 with a conserved glutamate residue close to b 595 over a distance of about 12 Å. It was proposed that this solvent filled area functions as a "dielectric well" and may facilitate charge compensation [13] as expected for the second branch of the proton pathway from the cytoplasmic side described above. For the G. thermodenitrificans oxidase, two different proton pathways were proposed [9]. One is located in the four-helix bundle formed by the TM helices 1-4 of CydA and the second one in the symmetry-related helices in CydB [9]. Accordingly, they have been named CydA and CydB pathways. Both pathways lead to heme b 595 with the CydA pathway ending directly at the conserved glutamate, the ligand of b 595 . However, the route of the additional proton transfer to heme d needed for the reaction of dioxygen to water is still unclear [9]. Future studies will have to address this point by means of the site-directed mutagenesis of putatively involved sidechains. In addition, recent advances in the field of cryo-EM may make high-resolution structures available that will support a refined identification and characterization of water, proton, and oxygen pathways and in particular the involved protein sidechains. It is anticipated that the upcoming years will bring about a wealth of novel data, which will deepen our understanding of the catalytic principles of bd oxidases. thermodenitrificans, is lined by several hydrophilic amino acid sidechains (shown as thick lines) and directs water molecules (i.e., protons) to the propionate of the active site heme d. Numerous water molecules (blue spheres) have been found in that channel, highlighting its full accessibility. (C) E. coli bd-II features a comparable channel. But due to the much wider cavity protruding deeply into the protein core, the actual channel requires only two glutamate sidechains (shown as sticks) to coordinate water molecules and guide them to the active site. Here again, the presence of water molecules illustrates the accessibility of solvent molecules.

Cytochrome bd as a Prospective Target for the Development of New Antibiotics
When bacterial pathogens colonize host cells and tissues, they encounter adverse environmental conditions, such as hypoxia or the presence of reactive oxygen and nitrogen species generated by the host immune system as weapons against the invaders. An increase in the cytochrome bd expression is a mechanism for survival used by the pathogenic microorganisms under these conditions. This is likely due to the unique structural features of the bd oxidase, which allows it to function actively in these and other unfavorable environments. Accordingly, the promotion of virulence by cytochrome bd was observed in Listeria monocytogenes, M. tuberculosis, uropathogenic E. coli (UPEC), Shigella sp., Group B Streptococcus, Salmonella enterica serovar Typhimurium, Clostridia species, Staphylococcus aureus, and Burkholderia pseudomallei [52,96,97]. The importance of cytochrome bd for pathogenic bacteria justifies its choice as a potential drug target.
Particularly noteworthy is the fact that the bd oxidase protects mycobacteria from antibiotic stress by allowing them to respire. M. tuberculosis is the causative agent of tuberculosis, the 13th leading cause of death, and the second leading infectious killer after COVID-19 [98]. Most troubling is the continued rise in drug-resistant forms of the disease. Multidrug-resistant (MDR), extensively drug-resistant (XDR), and totally drug-resistant (TDR) strains of M. tuberculosis pose a serious threat to public health. Thus, new antimicrobials with novel mechanisms of action are urgently needed.
The M. tuberculosis electron transport chain has recently gained interest as a target space for next-generation antibacterials [91,[99][100][101][102][103][104][105][106][107][108][109]. M. tuberculosis possesses a branched aerobic respiratory chain. The menaquinone pool receives electrons from various dehydrogenases including one proton-pumping NADH dehydrogenase, Nuo (type I NDH-1), two non-proton-pumping NADH dehydrogenases, Ndh and NdhA (type II NDH-2), and two succinate dehydrogenases, SDH-1 and SDH-2 [110]. The electrons from the reduced menaquinone (menaquinol) can then be transferred to O2 via two different terminal respiratory enzymes. Usually, the choice depends on the oxygen tension. Under aerobic conditions, this is a bcc-aa3 supercomplex composed of cytochrome bcc (complex III or qcrBCD) and the aa3-type cytochrome c oxidase (complex IV or ctaBCDE). Under Figure 6. Proton pathways in bacterial bd oxidases to the heme triangle. (A) G. thermodenitrificans features a pathway along the CydA/CydB interface, termed the CydB pathway (red). A second pathway running through CydA has been proposed as well, termed CydA pathway (light beige). (B) A pathway in E. coli bd-I (olive green), largely equivalent to the CydB pathway in G. thermodenitrificans, is lined by several hydrophilic amino acid sidechains (shown as thick lines) and directs water molecules (i.e., protons) to the propionate of the active site heme d. Numerous water molecules (blue spheres) have been found in that channel, highlighting its full accessibility. (C) E. coli bd-II features a comparable channel. But due to the much wider cavity protruding deeply into the protein core, the actual channel requires a only a glutamate and an aspartate sidechain (shown as sticks) to coordinate water molecules and guide them to the active site. Here again, the presence of water molecules illustrates the accessibility of solvent molecules.

Cytochrome bd as a Prospective Target for the Development of New Antibiotics
When bacterial pathogens colonize host cells and tissues, they encounter adverse environmental conditions, such as hypoxia or the presence of reactive oxygen and nitrogen species generated by the host immune system as weapons against the invaders. An increase in the cytochrome bd expression is a mechanism for survival used by the pathogenic microorganisms under these conditions. This is likely due to the unique structural features of the bd oxidase, which allows it to function actively in these and other unfavorable environments. Accordingly, the promotion of virulence by cytochrome bd was observed in Listeria monocytogenes, M. tuberculosis, uropathogenic E. coli (UPEC), Shigella sp., Group B Streptococcus, Salmonella enterica serovar Typhimurium, Clostridia species, Staphylococcus aureus, and Burkholderia pseudomallei [52,96,97]. The importance of cytochrome bd for pathogenic bacteria justifies its choice as a potential drug target.
Particularly noteworthy is the fact that the bd oxidase protects mycobacteria from antibiotic stress by allowing them to respire. M. tuberculosis is the causative agent of tuberculosis, the 13th leading cause of death, and the second leading infectious killer after COVID-19 [98]. Most troubling is the continued rise in drug-resistant forms of the disease. Multidrug-resistant (MDR), extensively drug-resistant (XDR), and totally drugresistant (TDR) strains of M. tuberculosis pose a serious threat to public health. Thus, new antimicrobials with novel mechanisms of action are urgently needed.
The M. tuberculosis electron transport chain has recently gained interest as a target space for next-generation antibacterials [91,[99][100][101][102][103][104][105][106][107][108][109]. M. tuberculosis possesses a branched aerobic respiratory chain. The menaquinone pool receives electrons from various dehydrogenases including one proton-pumping NADH dehydrogenase, Nuo (type I NDH-1), two non-proton-pumping NADH dehydrogenases, Ndh and NdhA (type II NDH-2), and two succinate dehydrogenases, SDH-1 and SDH-2 [110]. The electrons from the reduced menaquinone (menaquinol) can then be transferred to O 2 via two different terminal respiratory enzymes. Usually, the choice depends on the oxygen tension. Under aerobic conditions, this is a bcc-aa 3 supercomplex composed of cytochrome bcc (complex III or qcrBCD) and the aa 3 -type cytochrome c oxidase (complex IV or ctaBCDE). Under low oxygen concentrations, cytochrome bd (cydAB) serves as a terminal oxidase [111]. Each of the terminal segments of the respiratory chain generates a proton motive force, albeit with different efficiency, to drive ATP synthesis via the F 1 F o -ATP synthase (atpBEFHAGDC).
Pharmacological targeting of the mycobacterial respiratory enzymes holds significant clinical promise. The first antimycobacterial drug of this kind approved by the FDA and EMA is bedaquiline (Sirturo™). The drug selectively inhibits the F 1 F o -ATP synthase [112]. The other compound that targets M. tuberculosis cellular energy production is Q203 (telacebec). The compound targets the cytochrome b subunit (QcrB) of cytochrome bcc [113]. Q203 has gone through three clinical studies, the most recent being a phase 2a efficacy trial [114]. Cytochrome bd appears to contribute to the defense of mycobacteria against the stress induced by each of the two antibiotics. The defense mechanisms seem to be different. In the case of bedaquiline, the bd oxidase possibly detoxifies reactive oxygen species generated by the drug. As to Q203, cytochrome bd provides an efficient alternate respiratory route for electrons transferring from menaquinol to O 2 (see [53] and references therein). Accordingly, genetic or chemical inhibition of the bd oxidase has synthetic lethal interactions in M. tuberculosis with the bacteriostatic Q203, leading to rapid cell death against both replicating and non-replicating cells in vitro and in a mouse model of tuberculosis [115,116]. These findings raise hopes that combinations of respiratory chain inhibitors, which include a cytochrome bd inhibitor, may have a rapid and high killing capacity towards M. tuberculosis and other pathogens containing the bd oxidase.
The search for cytochrome bd inhibitors suitable for clinical purposes is at the very beginning. The quinone-analogs aurachins and their derivatives have attracted particular attention. Aurachins are isoprenoid quinoline alkaloids originally extracted from myxobacteria. Meunier et al. first reported that aurachin C and aurachin D are powerful inhibitors of the terminal quinol oxidases of E. coli [117]. The addition of 214 nM aurachin C to the E. coli membranes containing either cytochrome bd-I (strains GL101 or GL105) or the heme-copper cytochrome bo 3 (strain RG145) inhibits the duroquinol:O 2 oxidoreductase activity of each oxidase by 90%. 400 nM aurachin D, in turn, inhibits the membrane-bound bd-I by 93% and the membrane-bound bo 3 by as little as 5%. Thus, aurachin C is effective on both quinol oxidases, whereas aurachin D displays selectivity for inhibition of cytochrome bd-I [117] (see also Figure 7). Theßeling et al. determined the apparent IC 50 of aurachin C and aurachin D towards the duroquinol:O 2 oxidoreductase activity of the isolated cytochrome bd-I from E. coli (strain BL21 ∆cyo/pET28a cydA h BX) to 12 and 35 nM, respectively [10]. Aurachin C and aurachin D also strongly inhibit the duroquinol:O 2 oxidoreductase activity of the isolated E. coli cytochrome bd-II (strain BL21 ∆cyo/pET28b(+) appC his BX) with the apparent IC 50 of 7.1 and 11.1 nM, respectively [14].
Radloff et al. studied the inhibitory effects of aurachin C and new aurachin D derivatives on the ubiquinol-1:O 2 oxidoreductase activity of the isolated cytochromes bd-I, bd-II, and bo 3 from E. coli (strains C43 ∆bo 3 /pET17b-cydABX-StrepII, C43 ∆bo 3 /pET17b-appCBX-StrepII, and GO195 pIRHisA, respectively) [118]. Long-(C10, decyl or longer) and shortchain (C4, butyl to C8, octyl) aurachin D derivatives were synthesized. Their inhibitory potency and selectivity were assessed. The authors confirmed the strong inhibition of all three quinol oxidases by aurachin C derivatives in a nanomolar range and the fact that none of these compounds selectively inhibits a certain oxidase. Earlier data showed that the replacement of N-OH with an N-H group in aurachin C and its analogs decreases the inhibitory potential only for the bo 3 enzyme, keeping a strong inhibitory effect on the bd-I enzyme [119]. Correspondingly, all aurachin D derivatives tested clearly show the inhibition of both cytochrome bd-I and cytochrome bd-II ( Figure 7). As the inhibitory effect on the bd-I enzyme was higher than that on the bd-II enzyme, the former was investigated in most experiments. Two short-chain aurachin D derivatives, 2-(2-heptyl)-3-methyl-4(1H)quinolone (AD7-1) and unsaturated AD7-1 with a double-bond in the heptyl-side chain (AD7-1*), were found to be highly selective towards the bd-I oxidase. Their apparent K i values appeared to be similar to that of the natural aurachin D. Furthermore, the inhibitory activity was shown to increase with increasing chain length at position R1 of the 2-methyl-4-quinolones backbone. Among the inhibitors tested, AD7-1 combines properties of high inhibitory potency and selectivity for cytochrome bd-I while causing no inhibition of cytochrome bo 3 in the low nanomolar to micromolar range. It was concluded that AD7-1 could be the promising candidate for trials on a physiological level [118]. Radloff et al. studied the inhibitory effects of aurachin C and new aurachin D derivatives on the ubiquinol-1:O2 oxidoreductase activity of the isolated cytochromes bd-I, bd-II, and bo3 from E. coli (strains C43 Δbo3/pET17b-cydABX-StrepII, C43 Δbo3/pET17b-appCBX-StrepII, and GO195 pIRHisA, respectively) [118]. Long-(C10, decyl or longer) and short-chain (C4, butyl to C8, octyl) aurachin D derivatives were synthesized. Their inhibitory potency and selectivity were assessed. The authors confirmed the strong inhibition of all three quinol oxidases by aurachin C derivatives in a nanomolar range and the fact that none of these compounds selectively inhibits a certain oxidase. Earlier data showed that the replacement of N-OH with an N-H group in aurachin C and its analogs decreases the inhibitory potential only for the bo3 enzyme, keeping a strong inhibitory effect on the bd-I enzyme [119]. Correspondingly, all aurachin D derivatives tested clearly show the inhibition of both cytochrome bd-I and cytochrome bd-II ( Figure 7). As the inhibitory effect on the bd-I enzyme was higher than that on the bd-II enzyme, the former was investigated in most experiments. Two short-chain aurachin D derivatives, 2-(2heptyl)-3-methyl-4(1H)-quinolone (AD7-1) and unsaturated AD7-1 with a double-bond in the heptyl-side chain (AD7-1*), were found to be highly selective towards the bd-I oxidase. Their apparent Ki values appeared to be similar to that of the natural aurachin D. Furthermore, the inhibitory activity was shown to increase with increasing chain length at position R1 of the 2-methyl-4-quinolones backbone. Among the inhibitors tested, AD7-1 combines properties of high inhibitory potency and selectivity for cytochrome bd-I while causing no inhibition of cytochrome bo3 in the low nanomolar to micromolar range. It was concluded that AD7-1 could be the promising candidate for trials on a physiological level [118].
Makarchuk et al. screened a target-focused library of small molecules that contains a set of quinones, naphthoquinones, phenols, quinolones, coumarins, and flavonoids to identify potential inhibitors of cytochrome bd-I from E. coli (strain BL21 Δcyo/pET28a cydAhBX) using protein film voltammetry [120]. As such, quinolones with alkyl or iodine Makarchuk et al. screened a target-focused library of small molecules that contains a set of quinones, naphthoquinones, phenols, quinolones, coumarins, and flavonoids to identify potential inhibitors of cytochrome bd-I from E. coli (strain BL21 ∆cyo/pET28a cydA h BX) using protein film voltammetry [120]. As such, quinolones with alkyl or iodine substituents in positions C-2 and C-3 were identified. The most active inhibitors were produced by chemical modification of the quinolone core and the introduction of an isoprenyl chain in position C-3. The authors showed that the inhibitory efficiency of these compounds increases from one to two isoprene repetitive units and decreases for longer chains [120].
The effect of aurachin D on the activity of the bd oxidase in mycobacteria was also investigated. It was shown that the compound inhibits O 2 consumption of cytochrome bd in inverted membrane vesicles of M. smegmatis (strain mc 2 155 ∆qcrCAB::hyg) with an IC 50 of 400 nM [121] (Figure 7). In these experiments, the reaction was started by the addition of NADH as the electron donor. Accordingly, under similar conditions, aurachin D decreases the O 2 consumption rates of the bd oxidase (IC 50 of 0.158 µM) in inverted membrane vesicles of M. tuberculosis recombinantly produced in a M. smegmatis mc 2 155 ∆cydAB strain [13]. Although a significant inhibitory effect of aurachin D was shown for the membrane vesicles, the compound when applied alone does not effectively inhibit mycobacterial growth. The minimal inhibitory concentrations (MICs) for inhibition of growth of M. smegmatis and M. tuberculosis were reported to be >85 µM and >100 µM, respectively [121,122]. This finding indicates that aurachin D is not able to efficiently permeate the mycobacterial cell wall [121]. However, the presence of aurachin D significantly enhances the growth inhibition of M. tuberculosis (strain H37Rv) by Q203, a selective inhibitor of cytochrome bcc [122] (Figure 7). The Q203 MIC decreases from 10 nM when used alone to 1.25 nM when in combination with 25 µg/mL aurachin D. Furthermore, the addition of aurachin D to M. tuberculosis H37Rv converts the bacteriostatic activity of Q203 (30 × MIC) into bactericidal activity in a dose-dependent fashion [122]. Consistently, the almost identical bactericidal activity of Q203 was observed within cytochrome bd knockout strains [122]. Thus, the bd oxidase inhibitor aurachin D can considerably stimulate the impact of a companion drug targeting cytochrome bcc.
Harikishore et al. employed an in silico screening to identify a novel inhibitor 3-[[2-(4chlorophenyl)ethylamino]methyl]-1-ethyl-indole-2-carboxylic acid (MQL-H 2 ) that likely targets mycobacterial cytochrome bd at the menaquinol binding site [123]. The compound was shown to inhibit the ATP synthesis of inverted membrane vesicles of the wild-type M. smegmatis (strain mc 2 155) driven by both NADH and succinate with IC 50 of 60 and 75 µM, respectively (Figure 7). Similarly, the addition of MQL-H 2 causes the inhibition of NADH-driven ATP synthesis in inverted membrane vesicles of the cytochrome bcc deficient M. smegmatis mutant (strain mc 2 155 ∆bcc) with an IC 50 of 34 µM. These new data may pave the way for medicinal chemistry-based hit optimization efforts of MQL-H 2 .
Lee et al. identified a small molecule, N-(4-(4-(trifluoromethyl)phenoxy)phenyl) quinazolin-4-amine (ND-011992), that seems to target cytochrome bd [116]. ND-011992 is ineffective on its own, however, its combination with Q203 inhibits O 2 consumption and ATP synthesis in M. tuberculosis H37Rv and M. bovis BCG (Figure 7). The addition of ND-011992 to Q203 does not accelerate the frequency of spontaneous resistant mutations in the pathogen. The ND-011992/Q203 combination appeared to be bactericidal against clinical isolates of various phylogenetic lineages, against MDR and XDR isolates, and kills replicating and antibiotic-tolerant non-replicating mycobacteria in vitro. Furthermore, Q203 being supplemented with ND-011992 achieves better killing than Q203 alone in a mouse model of tuberculosis infection. Although the in vivo study shows that the presence of ND-011992 enhances the potency of Q203, the enhanced potency is limited by ND-011992 s less-than-optimal pharmacokinetic properties. Lead optimization is clearly needed to optimize the potency of the compound and improve its pharmacokinetic properties [116]. In general, these data suggest that inhibitors of the terminal oxidases could be part of a short sterilizing drug combination for tuberculosis.
Hopfner et al. found that another class of compounds, thieno[3,2-d]pyrimidin-4amines, can inhibit cytochrome bd in mycobacteria [124]. The authors reported an initial structure-activity-relationship of 13 compounds in M. bovis BCG, M. tuberculosis H37Rv, and M. tuberculosis clinical isolate N0145 in the measurements of ATP depletion in the presence and absence of Q203. All compounds reveal activity against M. bovis BCG and M. tuberculosis N0145 with ATP IC 50 from 6 to 54 µM, as determined by ATP depletion in the presence of Q203 (Figure 7). None of the compounds alone inhibit ATP. This clearly suggests that they do target the bd oxidase, since potency is observed only if cytochrome bcc is selectively blocked by Q203. The tested compounds turned out to be much less potent against M. tuberculosis H37Rv as compared to M. tuberculosis N0145: the ATP IC 50 went from 24 to >100 µM vs. 9-52 µM, respectively. A lower potency of the inhibitors in the case of M. tuberculosis H37Rv may be due to the increased level of cytochrome bd expression in this laboratory-adapted strain. N-(4-(tert-butyl)phenethyl)thieno[3,2-d]pyrimidin-4-amine (named compound 19) appeared to be the most potent compound, having ATP IC 50 of 6 to 18 µM against all three strains in the presence of Q203. Based on their results, the authors will aim to develop new members of this class of compounds with improved potency and acceptable pharmacokinetics to warrant in vivo evaluation [124].

Concluding Remarks
New structural information on cytochrome bd should be very helpful for the design and discovery of new antibacterial agents which do not affect the energy metabolism of human organs and tissues. The studies reviewed above show that at present the main efforts are concentrated on the search for inhibitors specific to the cytochrome bd quinol oxidation site. We assume that other structural segments of the enzyme, such as the oxygenbinding site, specific intraprotein oxygen channels, and proton transfer pathways, should also be considered as promising targets when creating effective inhibitors of the bd oxidase. In general, the development of next-generation antibiotics targeting the respiratory chain enzymes including cytochrome bd will contribute to overcoming a great global public health challenge-antibiotic resistance.