Research Article

An electron transfer path connects subunits of a mycobacterial respiratory supercomplex

See allHide authors and affiliations

Science  30 Nov 2018:
Vol. 362, Issue 6418, eaat8923
DOI: 10.1126/science.aat8923

An electron bridge in place of a ferry

Respiratory complexes are massive, membrane-embedded scaffolds that position redox cofactors so as to permit electron transfer coupled to the movement of protons across a membrane. Gong et al. used cryo–electron microscopy to determine a structure of a stable assembly of mycobacterial complex III–IV, in which a complex III dimer is sandwiched between two complex IV monomers. A potential direct electron transfer path stretches from the quinone oxidizing centers in complex III to the oxygen reduction centers in complex IV. A loosely associated superoxide dismutase may play a role in detoxifying superoxide produced from uncoupled oxygen reduction.

Science, this issue p. eaat8923

Structured Abstract


Cellular respiration is a core feature in the metabolism of many organisms that allows for the generation of a proton gradient across a membrane. During respiration, electrons are transferred from electron donors to oxygen through an electron transport chain. The energy created allows protons to be pumped across a membrane (cellular or mitochondrial). In electron transport chains, quinones and cytochrome c are two of the electron carriers that shuttle electrons to and from large macromolecular structures that are embedded in the membrane. The components that allow respiratory chains to function in the mitochondria are well characterized, but the situation is less clear and more varied in prokaryotic systems. A soluble cytochrome c pathway for electron transfer similar to that in mitochondria is commonly found in Gram-negative bacteria. Gram-positive bacteria such as Mycobacteria are devoid of a soluble cytochrome c but instead possess cytochrome c proteins that are anchored onto the membrane or have a fused cytochrome c domain to mediate electron transfer between two of the major complexes, which are referred to as CIII and CIV.

Structures of eukaryotic respiratory supercomplexes have been reported, but cytochrome c is not visible in any of these structures. Thus, a complete pathway for electron flow has not yet been visualized. CIII–CIV supercomplexes have been isolated from Mycobacterium smegmatis, Corynebacterium glutamicum, and Mycobacterium tuberculosis and shown to couple quinol oxidation to oxygen reduction without an external electron shuttle, suggesting that the flow of electrons is internalized in this type of complex. The determination of the structure of this complex reveals a path for electron transfer between the subunits of these supercomplexes.


The structural information provided here is required to understand the molecular details of electron transport in Mycobacteria. We have selected the supercomplex CIII–CIV from M. smegmatis because it is highly similar to the CIII–CIV complex from the human pathogen M. tuberculosis. This complex was amenable to expression and purification and analysis by means of cryo–electron microscopy (cryo-EM).


We have determined a cryo-EM structure of a respiratory supercomplex isolated from M. smegmatis. The structure allows the complete visualization of 20 subunits that associate to form the complex. Central to the supercomplex is a CIII dimer that is flanked on either side by individual CIV subunits. Fused c-type cytochrome domains bridge and mediate electron transfer from CIII to CIV. The structure also reveals three previously unidentified associated subunits that contribute to the stability of the supercomplex and the presence of superoxide dismutase (SOD), which may be responsible for the detoxification of superoxide formed by CIII.


This study of a respiratory supercomplex in Mycobacteria reveals cofactors positioned at distances that permit electron tunneling, enabling direct intrasupercomplex electron transfer from menaquinol to oxygen without the need for a separate cytochrome c electron shuttle. The presence of a bound SOD to the respiratory supercomplex suggests a mechanism of mycobacterial resistance against exogenous and endogenous oxidative stress in macrophages and host immune responses. The structure of the quinone binding sites provides a framework for rational structure-based M. tuberculosis drug discovery. A binding site can be proposed for the candidate antimycobacterial drug Q203, which acts by inhibiting the activity of this supercomplex.

Structure of mycobacterial respiratory supercomplex CIII2CIV2SOD2.

Overall architecture of the bcc-aa3–type respiratory CIII–CIV supercomplex from M. smegmatis. The cryo-EM map of the supercomplex shows a linear twofold dimerized form of CIV1–CIII2–CIV1.


We report a 3.5-angstrom-resolution cryo–electron microscopy structure of a respiratory supercomplex isolated from Mycobacterium smegmatis. It comprises a complex III dimer flanked on either side by individual complex IV subunits. Complex III and IV associate so that electrons can be transferred from quinol in complex III to the oxygen reduction center in complex IV by way of a bridging cytochrome subunit. We observed a superoxide dismutase-like subunit at the periplasmic face, which may be responsible for detoxification of superoxide formed by complex III. The structure reveals features of an established drug target and provides a foundation for the development of treatments for human tuberculosis.

In cellular respiration, chemical energy is extracted by coupling the oxidation of an energy source (such as sugars, fatty acids, or amino acids) and the reduction of an electron acceptor (such as oxygen, sulfur, nitrate, or sulfate) to synthesize adenosine triphosphate (ATP), which powers cellular reactions. In aerobic organisms, electrons are transferred from electron donors to oxygen, the terminal acceptor, through the electron transport chain (ETC) to pump protons across a membrane (cellular or mitochondrial). This creates a transmembrane proton gradient [proton motive force (PMF)] that drives ATP synthesis (1). In ETCs, quinones and cytochromes are two types of electron carriers that shuttle electrons to and from large macromolecular structures embedded in the membrane. In the mitochondrial respiratory chain, four membrane oxidoreductases are involved in electron transfer: complex I [reduced form of oxidized nicotinamide adenine dinucleotide (NADH):ubiquinone oxidoreductase] (CI), complex II (succinate:ubiquinone oxidoreductase) (CII), complex III (bc1-type ubiquinol:cytochrome c oxidoreductase) (CIII), and complex IV (aa3-type cytochrome c oxidase) (CIV). CIII oxidizes ubiquinol to ubiquinone and passes the electrons to soluble cytochrome c, which then shuttles them to CIV, where oxygen is reduced to water (Fig. 1A). The transmembrane PMF is generated by proton pumping in CI, CIII, and CIV.

Fig. 1 Respiration in Actinomycetes and overall architecture of the mycobacterial respiratory machine CIII2CIV2SOD2.

(A) The respiratory electron transfer chain in Actinomycetes (left) and the five major prokaryotic cytochrome c pathway variants with the organization schemes in representative organisms (right). The M. smegmatis cytochrome c pathway has its cytochrome c fused with complex III, forming a bcc-type complex III that interacts with the aa3-type complex IV to assemble into a CIII–CIV supercomplex. MK, menaquinone/menaquinol. (B) Overall architecture of the bcc-aa3–type respiratory CIII–CIV supercomplex from M. smegmatis. The cryo-EM map of the supercomplex shows a linear twofold dimerized form of CIV1–CIII2–CIV1 with dimensions 200 Å by 70 Å by 120 Å. CIII is colored in orange, CIV is in magenta and the association factors, PRSAF1 is in green, LpqE is in blue, and SOD is in gray. (C) Cartoon representation of the side view of the supercomplex (top) and a cross-sectional view (bottom). The MK is presented as bright green–colored solid spheres, and the phospholipids are shown as yellow sticks. In the cross-sectional view (bottom), the boundaries of CIII, CIV, and the association factor PRSAF1 are depicted with dashed lines in color (orange for CIII, magenta for CIV, and green for PRSAF1).

The situation is more complicated in prokaryotic respiratory chains (Fig. 1A) (2). A soluble cytochrome c pathway similar to that in mitochondria is common in Gram-negative bacteria. Variations include a membrane-anchored cytochrome c mediating electron transfer from CIII to CIV (3) and a caa3-type CIV with a fused cytochrome c domain (4). Gram-positive bacteria usually possess cytochrome c proteins that are anchored onto the membrane (5), or a fused cytochrome c domain (bcc-type CIII or caa3-type CIV) to mediate electron transfer between CIII and CIV (6, 7). Mycobacteria and other Actinobacteria such as Corynebacterium glutamicum are inherently devoid of a soluble cytochrome c in their genomes (8) but contain a bcc-type CIII with a di-heme cytochrome c domain fused to CIII (9, 10). Variations are also observed within CIII and CIV. Alternative complex IIIs (ACIIIs), structurally unrelated to bc1-type CIII, mediate quinol oxidation (11). Quinol oxidases couple quinol oxidation to oxygen reduction without the need for cytochrome c (12). Alternative oxidases (Aoxs) catalyze quinol oxidation/oxygen reduction without proton pumping (13).

Despite variation with ETCs, homologs or analogs of CI through to CIV are the most common components of respiratory chains in aerobic organisms. Structures of both prokaryotic and eukaryotic CI (1416), CII (17, 18), CIII (1921), and CIV (2224) have been determined, elucidating the flow of electrons within these individual complexes. Structural information for the mitochondrial respirasome CI1CIII2CIV1 and preliminary electron microscopic study of a CIII2CIV2 supercomplex from yeast have been reported (25, 26). However, cytochrome c is not visible in any of these structures. Thus, a complete pathway for electron flow is yet to be fully elucidated. Understanding the precise details of the structural assembly for a CIII–CIV supercomplex will greatly assist in this endeavor (27, 28). It has been reported recently that respiratory supercomplexes in situ have a conserved core of CI and a dimer of CIII, but otherwise, their stoichiometry and structure vary (29). Up to two copies of monomeric CIV were found associated with the CI1CIII2 assembly in bovine heart and the yeast Yarrowia lipolytica, but their positions varied (30). The conserved features of supercomplex assemblies such as CI1CIII2 and CIII2CIV2 suggest that these types of associations have important roles in respiratory electron transfer.

The bcc-type CIII from Actinobacteria has a di-heme c subunit (7). It has been suggested that one cytochrome c domain is the donor for the aa3-type CIV, and the other is the acceptor for the CIII Rieske Fe-S protein (31). In support of this concept of intrasupercomplex electron transfer, CIII–CIV supercomplexes have been isolated from Mycobacterium smegmatis, C. glutamicum, and Mycobacterium tuberculosis and shown to couple quinol oxidation to oxygen reduction without an external electron shuttle (9, 10, 32). Therefore, structural data for the bcc-aa3–type CIII–CIV supercomplex (SC III–IV) can provide answers as to how CIII and CIV are coupled and how electrons are transferred from CIII to CIV.

Purification and characterization of SC III–IV

To isolate the SC III–IV in a functional form, we engineered the genome of M. smegmatis to incorporate a 10× His tag at the C terminus of the QcrB subunit of CIII and extracted and purified the CIII–CIV complex by means of nickel–nitrilotriacetic acid (Ni-NTA) affinity chromatography and gel filtration. Gel filtration and blue native polyacrylamide gel electrophoresis (BN-PAGE) showed a single peak and a single band, suggesting a highly ordered supramolecular assembly (fig. S1, A and B). SDS-PAGE and mass spectrometry confirmed the presence of all the known components of CIII and CIV as well as several previously unknown components (fig. S1D and table S1). Native Orbitrap mass spectrometry gave a molecular weight of 873.4 kDa ± 10.4 Da for the complex (fig. S1E). Both the electronic absorption spectrum and the electron paramagnetic resonance (EPR) spectrum showed peaks expected from the various hemes, copper centers, and [2Fe-2S] prosthetic groups in CIII and CIV (fig. S1K). Because the bcc:aa3 preparations from M. smegmatis are active with the more soluble menadiol (2-methyl-1,4-naphthoquinol) as substrate (33), the quinol:oxygen oxidoreductase activity of SC III–IV was assayed by measuring the rate of O2 consumption in the presence of menadiol. SC III–IV oxidized menadiol and reduced O2 with an apparent catalytic rate constant (kcat) of 2.80 ± 0.05 s–1 for O2 consumption or 11.20 ± 0.20 e s–1 and a Michaelis constant (Km) of 120.70 ± 3.72 μM for menadiol (fig. S2, A to C). The kcat value is comparable to the 65 e s–1 reported for this complex, with 2,3-dimethyl-1,4-naphthoquinol (DMNQH2) as the electron donor (9); the difference is likely due to DMNQH2 being more reducing and a closer structural analog of the natural substrate menaquinol (MKH2) than menadiol. The data confirm that the purified sample is a functioning supercomplex containing CIII and CIV and capable of directly coupling quinol oxidation to oxygen reduction.

Overall architecture of SC III–IV

The structure of SC III–IV was determined by means of cryo–electron microscopy (cryo-EM) to an overall resolution of 3.5 Å (Fig. 1B; table S2; fig. S3, A to G; and movie S1). The dimensions of the supercomplex are 200 Å by 70 Å by 120 Å, with a linear dimeric CIV1–CIII2–CIV1 arrangement in which individual CIVs flank the central CIII dimer on either side (Fig. 1, B and C, and Movie 1). This C2 symmetrized linear architecture is completely different from those previously reported for respiratory supercomplexes (fig. S4). CIII is composed of canonical three subunits as a homodimer (Fig. 2A and fig. S5A). In addition to the four known subunits of M. smegmatis CIV, two subunits were observed that match two of the newly identified proteins, CtaI and CtaJ (Fig. 2B and fig. S3H), showing a similar topology and binding schema to those of subunit Va and IV in mitochondrial CIV (fig. S5J).

Movie 1. The overall architecture of the supercomplex.

Cartoon representation of the supercomplex. The menaquinone/menaquinol (MK) is presented as bright green colored solid spheres and the phospholipids as yellow sticks.

Fig. 2 Structure of CIII2 and CIV from M. smegmatis.

(A) Overall structure of the CIII dimer (left) and the spatial location (right) of prosthetic groups. QcrA, QcrB, and QcrC are colored pink, blue, and gold, respectively. The twofold symmetry of the dimer is depicted by the black axis. The zoom-in view shows the heme c binding domains (D1 and D2) of QcrC. The heme groups (bH, bL, cD1, and cD2) and [2Fe-2S] clusters are shown as spheres, and menaquinone/menaquinol (MK) are shown as sticks. The regions of ridge roof, ceiling junction, and base plate of CIII dimer are marked with dashed ellipses. (B) Overall structure of CIV (left) and the spatial location (right) of prosthetic groups. CtaC, CtaD, CtaE, CtaF, CtaI, and CtaJ are colored in magenta, dark green, yellowish brown, cyan, brown, and violet, respectively. Prosthetic groups are shown as spheres.

There is extra density within the interface between CIII and CIV as well as at the top of the CIII dimer (Fig. 1B). The density within the interface could be modeled by two proteins, LpqE and PRSAF1 (prokaryotic respiratory supercomplex association factor 1) (Fig. 1C and Movie 1). LpqE was found to be a N-terminal triacylated lipoprotein, with a N-acylated-S-diacylated modification of Cys24 within the lipobox (–21Lxx24C–) sequence (34). On the periplasmic side, the density on top of the CIII dimer consisted of a peptide fragment linked with a region of bulk density that could be visualized in a low-pass filtered map. The peptide fragment was modeled with residues Cys21–Pro45 of the N-terminal sequence of superoxide dismutase (SOD) SodC of M. smegmatis. The EM map indicated side chain modifications, including triacylation at Cys21 that was part of a lipobox (–18Lxx21C–) sequence and possible glycosylation sites (fig. S3H). Native mass spectrometry further identified that SodC is a component of SC III–IV (fig. S1F). The stoichiometry of SodC was confirmed as a dimer through the collisional dissociation of the SodC complex. Extensive glycosylation and copper-binding of SodC was observed in the mass spectrum (fig. S1F). The bulk density was therefore fitted with a dimer of SodC (Fig. 1, B and C, and fig. S3F). SC III–IV possesses SOD activity, with a specific activity of 132.56 ± 12.57 IU/mg-SOD, assuming 100% occupancy (fig. S2E). SodC association with SC III–IV was also confirmed through isolation of the supercomplex by means of Ni-affinity chromatography when a His-tag was introduced only to the C terminus of SodC and not any of the CIII and CIV subunits. The SodC-tagged form of the complex showed a higher specific activity of 957.36 ± 23.34 IU/mg-SOD but still lower than the 1000 to 6000 IU/mg-SOD that is typically observed with soluble SOD enzymes. It also indicates an ~14% (SodC)2 occupancy in the purified QcrB-tagged SC III–IV particles for cryo-EM study. Thus, it appears that there might be some dissociation of SodC from SC III–IV during detergent solubilization. It is also possible that the SOD occupancy is growth-regulated because an up-regulation of sodC (the gene encoding the SOD here) in response to phagocytosis by human macrophages has been reported (35). Further work on the role of SOD in association with SC III–IV is in progress.

All the prosthetic groups predicted from the canonical CIII and CIV subunits were clearly resolved and found to be coordinated with conserved canonical residues (figs. S3J and S8). Menaquinone (MK) molecules were observed at the quinone binding sites in CIII (fig. S7, A and B). The calculated molar ratio between iron atoms and copper atoms in the final model is 1.6, which is in excellent agreement with the value of 1.5 determined with atomic absorption spectroscopy (AAS) (fig. S1C). SodC from M. tuberculosis does not contain zinc (36), and AAS analysis showed that zinc was absent from SC III–IV (fig. S1C). In total, we were able to build 34 phospholipids (fig. S3I) and 10 MK molecules (fig. S7) in SC III–IV. The total molecular mass of the model, including the new identified subunits, is ~760 kDa, which is lower than the 873.4 kDa ± 10.4 Da determined with native Orbitrap mass spectrometry (fig. S1E). This difference may be accountable on the basis of contributions from the detergents and lipids and the possibility of the presence of additional unidentified subunits.

Structure of CIII and CIV in the supercomplex

The cryo-EM map clearly shows QcrA, QcrB, and QcrC of CIII in a dimeric form with all their prosthetic groups visualized (Fig. 2A; fig. S3, H and J; and Movie 2). QcrA has a “U”-shaped structure within its N-terminal domain, whereas the equivalent subunit in bc1-type CIII has only one transmembrane helix (TMH) (equivalent to QcrATMH3) (Fig. 2A and fig. S5A). The region linking the two arms is located near the cytoplasmic side. The C-terminal domain of QcrATMH3 is on the periplasmic side and holds the [2Fe-2S] cluster. QcrA here also has a roof-like structure on the periplasmic side that is involved in the dimerization of CIII (Fig. 2A), whereas the bc1-type CIIIs do not have this feature. The heme bH and heme bL cofactors are bound within four TMHs in QcrB (Fig. 2A). The N-terminal periplasmic portion of QcrC can be divided into two heme-containing cytochrome c domains, D1 and D2 (equivalent to the c1 domain in bc1 complex) (Fig. 2A and fig. S5N). These two domains are in close contact with and face each other in an antiparallel orientation. The D2 domain interacts extensively with QcrA and QcrB, whereas the additional D1 domain protrudes out of the core of CIII and is involved in direct intrasupercomplex electron transfer. Overall, although the bcc-type CIII in the supercomplex shares a similar dimeric association as that of the bacterial bc1-type CIII and mitochondrial CIII, the structural details are markedly different (fig. S5B).

Movie 2. The composition and structure of CIII dimer.

Cartoon representation of the complex III. The menaquinone/menaquinol (MK) is presented as bright green colored solid spheres and the phospholipids as yellow sticks.

CIV in SC III–IV belongs to the type A heme-copper oxidase (HCO) family (Fig. 2B and Movie 3) (37). The central cavity of the barrel-like arrangement of the 12 TMHs in the CtaD subunit holds heme a, heme a3, and CuB. The C-terminal hydrophilic β barrel domain of CtaC holds the two CuA ions. The four protons required for oxygen reduction by heme a3:CuB in CIVs are transferred to the catalytic center through two pathways, denoted D and K (38). Because of the limited resolution, water molecules are not observed in our model. However, structure comparison revealed that the D and K pathways are conserved in SC III–IV (fig. S5, K and L).

Movie 3. The composition and structure of CIV.

Cartoon representation of the complex IV. The phospholipids are shown as yellow sticks.

Interaction between CIII and CIV and the contribution of association subunits and lipids

The linear form of SC III–IV arises from the dimerization of CIII, which is mediated by contacts between subunits QcrA and QcrB (Fig. 2A and fig. S6, A to D). There are extensive contacts between CIII and CIV on both the cytoplasmic and periplasmic sides of the membrane (Fig. 3). Of the three association subunits LpqE, PRSAF1, and SOD in SC III–IV, both LpqE and PRSAF1 form numerous contacts with CIII and CIV, suggesting an important role for both in supercomplex stability (Fig. 3B and fig. S6, E to G). SOD uses its lipid-modified N-terminal fragment to associate with CIII (Fig. 4E) and forms a dimer on the periplasmic side of the CIII dimer. There are no direct interactions between the main body of the SOD dimer and CIII or CIV. This might allow a flexible orientation for this subunit, which could be important for efficient clearance of reactive oxygen species (ROS) generated by side reactions when electrons are transferred from CIII to CIV.

Fig. 3 Interaction between CIII and CIV and roles of association subunits.

(A) A section profile of the interaction interface between CIII and CIV. The left and right images show the surface of CIII and CIV and the bound structural segments from CIV and CIII subunits and the association subunits, LpqE and PRSAF1. (B) Interactions between the association subunits (PRSAF1, LpqE, and SOD) and the subunits of CIII and CIV. The corresponding subunits of CIII and CIV are shown in surface representation and colored as indicated. The association subunits are shown in ribbon and colored as indicated. The catalytic domain of SOD was fitted in map and shown as a dimer (SOD2). The lipid modification of SOD is shown in yellow sticks (detail provided in Fig. 4E).

Fig. 4 Important roles of phospholipids in the stability and assembly of the supercomplex.

(A) The distribution of phospholipids in the membrane region (middle, view from periplasmic side) at the junction between the CIII monomers (left) and the interfaces between CIII and CIV (right). (B) A polyethylene (PE) molecule mediates the interaction between PRSAF1 (TMH2) and CIV (CtaC and CtaD). (C) A polyimide (PI) molecule binds to the interface between QcrCTMH1 and CtaFTMH4. (D) Four CL molecules are bound in the groove between the TM regions of CIII and CIV. (E) The N-terminal lipid modification of SOD at Cys21 binds to QcrB via hydrophobic interactions. (F) The N-terminal lipid modification of LpqE at Cys24 mediates the interactions between CtaD and PRSAF1.

Phospholipids, especially cardiolipin (CL), are known to contribute to both the assembly and stability of respiratory complexes and supercomplexes (39). In the structure of SC III–IV, phospholipids are identified in the transmembrane space of CIII and CIV and the interface between CIII and CIV (Fig. 4). Of particular note are the four CL molecules in the large groove between CIII and CIV (Fig. 4D); the 16 fatty acid chains fill most of the space in the groove and may play an important role in stabilizing the supercomplex. The N-terminal–modified lipid tails of the lipoproteins SOD and LpqE mediate intersubunit hydrophobic interactions and contribute to the stability of the supercomplex (Fig. 4, E and F). These lipid modifications described here are similar to those observed in a recently described structure of alternative complex III (11).

Quinone and quinone binding pockets

Quinone binding sites of respiratory complexes are of great interest because they are part of the Q-cycle hypothesis. They have varied sequences and specificities between species and are often the sites for inhibitor binding and thus are important for drug discovery. We have identified the two quinone binding sites (QP and QN) in SC III–IV (Fig. 5). The quinol oxidation site (QP site) responsible for MKH2 oxidation is near heme bL, whereas the quinone reduction site (QN site) responsible for MK reduction is close to heme bH.

Fig. 5 Structures of the MK/MKH2 binding sites in M. smegmatis CIII.

(A) The QP binding site. (B) The QN binding site. The residues potentially involved in the binding of MK/MKH2 are shown with side chains in a stick representation. The MK molecules are colored in green. The heme and [2Fe-2S] groups are shown in spheres and labeled accordingly. (C) Sequence alignment for the QP site and QN site in QcrB with other Actinobacteria and Homo sapiens ETC systems. The red and green dots indicate whether H. sapiens shares a common conservative site with Actinobacteria (red) or not (green).

The QP site near heme bL is at the center of an inverted triangle structure and surrounded by helices (Fig. 5A). Residues at this site are not conserved compared with the bc1 complex (Fig. 5C). The typical “PEWY” motif in the bc1 complex is replaced by “PDFY” (Fig. 5C). One MK molecule is identified at this site with its naphthoquinone ring surrounded mainly by hydrophobic residues. The edge-to-edge distance from MK to heme bL is 16 Å. Thus, we speculate that the endogenous electron donor MKH2 would bind closer to heme bL to facilitate electron transfer, and what we observe here might be a representation of the oxidized product as it leaves the QP site. Furthermore, there are no observed hydrogen bonds to the carbonyl groups of MK (Fig. 5A). It is known that proton abstraction from MKH2 is coupled to electron donation to CIII. Thus, hydrogen bonds are needed between the binding residues and the hydroxyl group of MKH2 to help deprotonate the substrate. Hence, MKH2 should bind deeper inside the pocket, close to polar residues such as QcrBTyr159, QcrBThr308, and QcrBAsp309 (Fig. 5A). Structural superposition with the inhibitor-bound bc1 complex (40) shows that the MK at the QP site binds deeper into the pocket in the bc1 complex than in this SC III–IV complex.

By contrast, the head group of MK at the QN site is bound in a similar fashion to ubiquinone in the structures of bc1 complexes (Fig. 5B). As found for the QP site, these residues are not conserved compared with the bc1 complexes from bacteria to eukaryotes (Fig. 5C). In particular, the carbonyl groups of ubiquinone are coordinated by the conserved His and Asp amongst bc1 complexes; these residues are substituted by QcrBTrp231 and QcrBSer261 in the bcc-type CIII of M. smegmatis. The carbonyl groups of MK in SC III–IV interact with the side chains of QcrBTyr48 and QcrBSer261 that may supply protons for MK reduction. The MK head group is ideally placed for electron transfer, being within 5 Å of the A-edge of heme bH.

Besides finding quinone molecules at the Q sites, we also observed map signals for another three possible MK/MKH2 molecules in the supercomplex (fig. S7, C to E). However, further experiments are needed to clarify the identity and function of these molecules.

The new family of candidate antimycobacterials, the imidazo[1,2-a]pyridines (IP) represented by Q203, operate by competing with MK for binding at the QP site of CIII of M. tuberculosis (41). Sequence alignments indicate a high similarity between the QP sites of CIIIs from M. tuberculosis and M. smegmatis (Fig. 5C), thus suggesting that Q203 would also have a similar binding mechanism and a similar effect on the activity of M. smegmatis CIII. Indeed, recent studies of the antimycobacterial activity of Q203 on M. tuberculosis and M. smegmatis demonstrated that Q203 targets the bcc complex in both with similar affinity (42). We investigated the in vitro inhibition of M. smegmatis SC III–IV by Q203 by means of the menadiol/oxygen oxidoreductase activity assay and compared the effect with a hybrid supercomplex of M. tuberculosis bcc-CIII and M. smegmatis aa3-CIV. Q203 showed inhibition of menadiol-induced oxygen consumption, with median inhibitory concentration (IC50) values of 0.84 ± 0.22 μM and 0.61 ± 0.16 μM for SC III–IV and the hybrid supercomplex, respectively (fig. S2D).

Prosthetic groups and implication for direct electron transfer

The prosthetic groups of CIII (heme bH/bL, [2Fe-2S] clusters, and heme cD1/cD2) and CIV (CuA, CuB, and heme a/a3) are clearly identified from the cryo-EM map (Fig. 2, A and B, and fig. S3J). The redox centers in CIII and CIV are within distances that allow long-range electron transfer (Fig. 6A). Both heme bL and heme bH are found in QcrBCIII (figs. S3J and S8A). The edge-to-edge distance between the two heme groups is 12 Å, allowing rapid inter-heme electron transfer. The shortest distance from bL to MK at the QP site and from bH to MK at the QN site are 16 and 5 Å, respectively (Fig. 6A). Previous studies have proposed that CIIIs from different species adopt a dimeric architecture and form an H-shaped electron transfer system that distributes electrons between four quinone oxidation-reduction sites within the CIII dimer (43). Consistent with this hypothesis, upon dimerization of M. smegmatis CIII, the two bL heme groups from the CIII monomers are 14 Å apart (Fig. 6A), allowing electron tunneling between the two hemes.

Fig. 6 The complete electron transfer pathway in the bcc-aa3–type respiratory CIII/CIV supercomplex.

(A) The prosthetic groups in the supercomplex are shown in sticks or spheres and labeled accordingly, with the corresponding midpoint potentials shown in parentheses. These values are based on the measurement of that of the bcc-aa3–type supercomplex from C. glutamicum (10). The edge-to-edge distances between adjacent prosthetic groups are shown in black dashed lines, with the numbers in the parentheses representing the center-to-center distances. (B) A schematic diagram showing the entire electron transfer pathway from CIII to CIV and the relevant proton translocations in CIII and CIV. The potential role of associated SOD for the clearance of ROS is also proposed.

The heme cD1-containing D1 domain of QcrCCIII protrudes into the periplasm and interacts with the CuA-containing periplasmic domain of CtaCCIV (Fig. 6A). The edge-to-edge distance from heme cD1 to the CuA center is 12 Å. At the interface between these two domains is a gating residue CtaCTrp138 on the CuA binding loop. The distance between CtaCTrp138 and the CuA center and that between CtaCTrp138 and heme cD1 are approximately equal. It has been proposed, on the basis of mutagenesis studies, that the corresponding Trp121 (44) of Paracoccus denitrificans CIV is the electron entry site from cytochrome c. CIVs from Bos taurus also possesses a tryptophan residue at the equivalent location (45). Thus, the heme cD1-containing D1 domain of QcrCCIII here interacts with CIV on a similar electron entry site as in other respiratory complexes.

Within CIV, the heme a group is just below the CuA center, and the heme a3 group is beside heme a (Fig. 6A). CuB is coordinated to three conserved His residues. The space between the iron of heme a3 and CuB is the catalytic center for reduction of oxygen. In proximity to propionate groups of heme a3, a CuC ion was modeled according to the cryo-EM map (fig. S3J). In other reported CIV structures, the equivalent density at this site is occupied by a water molecule (46) or a Mg2+ ion (47). However, we did not detect magnesium in our AAS analysis (fig. S1C). The edge-to-edge distance between CuC and the CuA center and that between CuC and heme a3 are 12 and 9 Å, respectively.

From the positions of the prosthetic groups in M. smegmatis SC CIII2CIV2SOD2 and the redox center separations, it is possible to trace an uninterrupted pathway for the flow of electrons within the supercomplex starting from the electron donor at the QP site in CIII to the final site of oxygen reduction in CIV (Fig. 6B). MKH2 from the Q-pool binds at the QP site near heme bL and transfers one electron to the [2Fe-2S] cluster and the other to heme bL, which passes the electron via heme bH to a MK bound at the QN site, generating the highly reactive intermediate menasemiquinone (MK•). The reduced [2Fe-2S] cluster is within electron tunneling distance to transfer an electron to heme cD2, which rapidly passes the electron to heme cD1. A second MKH2 then binds at the QP site and repeats the process. The MK• at the QN site is fully reduced to MKH2 and released to the Q-pool. This completes the Q cycle. An electron path between CIIIs in the dimer is also possible through tunneling between the adjacent bL heme groups, albeit with low efficiency (43). The reduced [2Fe-2S] cluster again transfers the other electron to heme cD2/cD1. Once heme cD1 is reduced, two CuA ions of CtaCCIV accept the electron through contacts forged with the D1 domain of QcrCCIII. At this point, an electron is transferred from CIII to CIV. Within CIV, the electron is transferred through heme a (or possibly the CuC center) and finally reaches the terminal heme a3:CuB reaction center for oxygen reduction. As a consequence of the electron transfer, protons are translocated to the periplasm, forming a transmembrane PMF. Throughout the entire pathway from the QP site to the terminal oxygen reduction center, electrons tunnel between prosthetic groups that are all buried inside this integral complex.

Quinone reduction at the QN site to complete the Q cycle can be bypassed or “short-circuited” if both electrons from MKH2 oxidation at the QP site are transferred to the [2Fe-2S] cluster and then to CIV for oxygen reduction. Hence, competent energy transduction requires that electron transfer from the QP site has to be bifurcated between reduction of the [2Fe-2S] cluster and heme bL. In essence, the pathway for transferring electrons from the [2Fe-2S] cluster eventually to CIV has to be sufficiently slow for electron transfer to the QN site for quinone reduction to occur. In the bc1 complex, the [2Fe-2S] cluster domain cycles between “b” and “c1” states. In the b-state, the [2Fe-2S] cluster is close to the QP site to accept an electron from quinol oxidation but too far away (26 Å) (fig. S5E) to transfer an electron to the c1 heme at an appreciable rate. The cluster domain undergoes a “head displacement” conformation change to the c1-state, in which the [2Fe-2S] cluster moves to within 11 Å of the c1 heme to facilitate electron transfer. This head displacement or “gating” step occurs with a rate constant of 6 × 104 s–1 (48).

We cannot rule out the possibility that SC III–IV could adopt a different conformation and cycle between states similar to the b and c1 states of bc1 complexes. However, the dimer of QcrA is held firmly in place by the periplasmic roof-like structure and further surrounded by QcrB and QcrC, which might limit the space for the potential conformational change (fig. S5D). Structural superposition shows that the position of the [2Fe-2S] in QcrA is similar to that of bc1-type CIII in the b state—rapid cluster reduction by MKH2 can occur. The [2Fe-2S] cluster is at a distance of 16 Å from heme cD2. We used this distance, the reported midpoint potentials of the prosthetic groups (10), and a reorganization energy (λ) of 0.7 eV to calculate a rate constant of electron transfer of 5.6 × 102 s–1 from [2Fe-2S] to heme cD2 (49). This is much slower than the head displacement gating step in bc1 complexes (6 × 104 s–1). We conclude that short-circuiting in SC III–IV is likely to be rate-limited by slow electron transfer between the [2Fe-2S] cluster and heme c made possible by positioning of the chain of redox centers, rather than by conformational changes as found in bc1 complexes.

Role of SOD association

In aerobic organisms, the respiratory ETC not only generates the energy needed to fuel biological functions but is also a major source of intracellular ROS that can cause damage to cellular structures and components (50). Complex III is one of the major sites of ROS production (51). Although ROS are emerging as important elements in the bacterial response to lethal stress (52), it is well documented that they can disturb respiratory activity through oxidative damage of ETC complexes, which are in turn protected by the ROS scavengers SOD and catalase (5355). Up-regulation of M. tuberculosis sodC (the gene encoding the SOD here) in response to phagocytosis by human macrophages has been observed (35). This mycobacterial Cu, Zn SOD SodC was identified as a membrane-bound enzyme and proposed to protect specific membrane-associated targets from oxy-radical damage, thus facilitating mycobacterial intracellular growth (35). A null sodC mutant of M. tuberculosis was shown to be readily killed by externally generated superoxide and by activated macrophages producing oxidative bursts (56). In this work, we found that catalytically active SodC is an integral part of a respiratory supercomplex CIII2CIV2SOD2. SodC could serve to scavenge ROS generated locally or by other ETC complexes as well as ROS released by the immune response of the host. Its recruitment has the potential to make this critical respiratory machinery a robust system even under the high oxidative stress inside macrophages. Immunoblotting of Caenorhabditis elegans respiratory supercomplexes separated by means of BN-PAGE showed that mitochondrial SOD-2 (mtSOD-2) is associated with the respirasome CI–CIII2–CIV, suggesting that the mtSOD might also provide similar local protection against ROS damage (57).


The cryo-EM structure of a CIII–CIV respiratory supercomplex from M. smegmatis has revealed a complete intracomplex electron transfer pathway from quinol oxidation in CIII to oxygen reduction in CIV, a new mechanism for bifurcating electron transfer to ensure completion of the Q cycle for energy transduction, and the association of a SOD that can provide protection against oxidative damage by ROS. The structure of the quinone binding sites also provides a framework for structure-based antimycobacterial drug discovery.

Materials and methods

Bacterial strain

A M. tuberculosis–like highly hydrogen peroxide-resistant M. smegmatis mutant strain, mc2 51 (58), was used in this study. The draft genome sequence data can be obtained through the GenBank accession no. JAJD00000000.1. The code for the 10× His tag was introduced at the C terminus of the QcrB or SodC genome loci through homologous recombination. This modification allowed the metal affinity purification step. The presence of the histidine tag was confirmed by PCR and Western blot.

M. smegmatis culture and membrane isolation

Culturing of the cells and membrane isolation were performed as described in previously published protocols, with some modifications (9, 32). The cells were grown in LB media supplemented with hygromycin (50 μg/mL), carbenicillin (20 μg/mL) and Tween 80 (1 mL/L), at 37°C with shaking to maintain oxygenation. A total of 10× 1 L bacteria solutions were cultured until the OD600 reached ~1.5 and then harvested by centrifugation for 30 min at 4,000 rpm. The obtained bacteria were resuspended in buffer A (20 mM MOPS, pH 7.4, 100 mM NaCl, 1 mM EDTA, 1 mM PMSF). Cell lysis was achieved by three passes through a high pressure cell disrupter at 4°C and 1200 bar. The lysate was centrifuged at 14,000 rpm for 10 min to remove cell debris and non-lysed cells. The resulting supernatant was centrifuged at 36,900 rpm for 1 hour in a Ti45 rotor (Beckman). The membrane pellets were harvested and stored at –80°C.

Supercomplex purification

Membranes were thawed and homogenized in buffer A. Respiratory supercomplexes were extracted from the membrane by 1% (w/v) digitonin for three hours with slow stirring at 4°C. Insoluble materials were removed by centrifugation at 18,000 rpm for 30 min at 4°C. The supernatant was loaded onto a Ni-NTA column followed by gravity feeding three column volumes of buffer A through the resin. The resin was further washed by buffer B (20 mM MOPS, pH 7.4, 100 mM NaCl, 1 mM EDTA, 0.1% (w/v) digitonin, 50 mM imidazole). The protein was eluted with buffer C (20 mM MOPS, pH 7.4, 100 mM NaCl, 1 mM EDTA, 0.1% (v/w) digitonin, 500 mM imidazole). Eluted protein was concentrated using a 100-kDa cut-off centrifugal concentrator (Millipore) and this sample was then loaded onto a Superose 6 (10/300 GL, GE Healthcare) column equilibrated in a buffer containing 20 mM MOPS, pH 7.4, 100 mM NaCl, 1 mM EDTA and 0.1% (v/w) digitonin. The peak fractions (elution volume between 12.25 mL and 13 mL) were pooled and concentrated to 5.5 mg mL−1 by using a 100-kDa cutoff centrifugal concentrator.

Characterization of the respiratory supercomplex CIII2CIV2SOD2

The supercomplex was characterized by optical spectroscopy, mass spectrometry (MS) and 3,3′-diaminobenzidine (DAB) staining. To identify the heme groups, selected fractions were analyzed by recording spectra from 250 to 700 nm before and after reduction with dithionite according to previously described methods (9, 32). To detect the protein components, the concentrated supercomplex sample was subjected to MS analysis at the National Center for Protein Science (Shanghai, China). The protein sample was analyzed by BN-PAGE and then a separate strip of the gel was stained with DAB (5961).

Pyridine hemochrome assay

For denaturing redox difference spectra samples were taken up in 20% (v/v) pyridine, 0.1 M NaOH. Redox difference spectra were recorded in the range 500–620 nm using potassium ferricyanide for oxidation and sodium dithionite for reduction. Potassium ferricyanide (250 μM) was added to the sample and the spectrum of this oxidized form was recorded. Sodium dithionite (5 mM) was then added to this sample, the solution was mixed well and the spectrum was scanned repeatedly until it remained unchanged, giving the spectrum of the reduced form. The heme a, b, and c contents were determined from the reduced-minus-oxidized difference spectrum using the following extinction coefficients Δε587–620 = 21.7 mM–1·cm–1, Δε557–540 = 23.98 mM–1·cm–1, Δε550–535 = 23.97 mM–1·cm–1.

Native mass spectrometry of supercomplex

Purified supercomplex was buffer exchanged into 200 mM ammonium acetate buffer pH 7.5 containing DDM at 2 times critical micelle concentration. The sample was immediately introduced into a modified Q-Exactive mass spectrometer (Thermo) and ions were transferred into the Higher-energy collisional dissociation (HCD) cell following a gentle voltage gradient (injection flatapole, inter-flatapole lens, bent flatapole, transfer multipole: 7.9, 6.94, 5.9, 4 V respectively). The capillary voltage was maintained at 1.2 to 1.4 eV at temperature of 200°C. The optimized acceleration voltage for intact supercomplex is 150 V and 100-120 V in the source and HCD cell, respectively. SodC subunits were dissociated from the complex with high acceleration voltage (250 to 300 V) in the source and the collisional dissociation of dimeric SodC was performed with the voltage ramp from 0-150 V in HCD cells. Backing pressure was maintained at ~1.20 × 10−9 mbar and data was analyzed using Xcalibur 2.2 SP1.48.

Independent lipidomics analysis 1

Co-purified lipids from supercomplex were extracted by chloroform/methanol (2:1, v/v) and lyophilized and re-dissolved in 60% acetonitrile (ACN). For LC-MS/MS analysis, the extracted lipids were loaded into a 5-μl sample loop using an autosampler by a full-loop method with an overfill factor of 1.4, and then transferred to a 1-μl injection loop using a loading pump at a flow rate of 5 μl/min with 70% solution A and 30% solution B. The lipids in the injection loop were injected onto a C18 column (Acclaim PepMap 100, C18, 75 μm × 15 cm, Thermo Fisher Scientific) using a nanopump at a flow rate of 300 nl/min, 30% solution B. The lipids were separated on the C18 column at 40°C by a gradient starting from 30% solution B. After 10 min, solution B was ramped to 65% over 1 min, then 80% over 6 min, before being held at 80% for 10 min, then ramped to 99% over 6 min and held for 7 min. [Solution A: (ACN: H2O (60:40), 10 mM ammonium formate, 0.1% formic acid] and solution B [IPA: ACN (90:10), 10 mM ammonium formate, 0,1% formic acid]. The column eluent was delivered via a dynamic nanospray source to a hybrid LTQ Orbitrap mass spectrometer (Thermo Scientific). Typical MS conditions were: spray voltage (1.8 kV) and capillary temperature (175°C). The LTQ-Orbitrap XL was operated in negative ion mode and in data-dependent acquisition with one MS scan followed by five MS/MS scans. Survey full-scan MS spectra were acquired in the orbitrap (m/z 350 to 2000) with a resolution of 60,000. Collision-induced dissociation (CID) fragmentation in the linear ion trap was performed for the five most intense ions at an automatic gain control target of 30,000 and a normalized collision energy of 38% at an activation of q = 0.25 and an activation time of 30 ms.

Independent lipidomics analysis 2

Homogenized membranes and protein crystals were extracted with 800 μL of chloroform:methanol:water (1:1:0.1) in glass vials. After shaking the samples vigorously at 1500 rpm for 30 min at 4°C, 350 μL of water was added to break the phases. Samples were centrifuged and the lower organic phase was transferred to new glass vials. The remaining membrane and/or protein pellets were re-extracted with 400 μL of chloroform. The combined extracts were dried using a SpeedVac (Genevac, UK). Samples were stored at –80°C until mass spectrometric analysis.

Liquid chromatography-mass spectrometric (LCMS) analyses were carried out on an Exion UPLC coupled with a SciexQTRAP 6500 Plus mass spectrometer. Analysis of phospholipids and free mycolic acids were carried out using normal-phase LCMS as described previously (62, 63). Levels of MK species were quantified using a reverse-phase LCMS method (64).

Quinone reduction

Menadiol was prepared as previously described, with minor modifications (65). To prepare reduced menadiol, 3.4 mg 2-methyl-1,4-naphthoquinone (Sigma M5625) was dissolved in 1 mL N2-saturated anhydrous cyclohexane to yield a 20 mM solution. The solution was mixed with 5 mL N2-saturated 1 M sodium dithionite solution (in H2O) and shaken vigorously. After phase separation, the organic phase containing the reduced menadiol was removed and transferred to a 15 mL centrifuge tube (all steps performed under a stream of N2). The cyclohexane was evaporated under an N2 stream, while the sample was kept at 40°C in a water bath. Subsequently, the reduced quinol was dissolved in N2-saturated, acidified ethanol (ethanol with 10 mM HCl), aliquoted, flash frozen in liquid nitrogen, and stored at –20°C.

Oxygen consumption assay

Since the bcc:aa3 preparations from M. smegmatis are active with menadiol (2-methyl-1,4-naphthoquinol) as substrate and electron donor, which has better solubility in water than MKH2 (33), the supercomplex was routinely characterized by its menadiol:O2 oxidoreductase activity. Rates of oxygen consumption were determined using a Clark-type oxygen electrode (Hansatech) in a magnetically stirred chamber at 25°C in 20 mM MOPS pH 7.4, 100 mM NaCl, 0.005% LMNG, 5 mM DTT, following a method previously described, with minor modifications (33). The rate of menadiol autoxidation was measured in parallel as a reference and subtracted from the supercomplex-dependent rate.

Enzyme kinetics of supercomplex CIII2CIV2SOD2

The Michaelis-Menten curves were obtained by measuring the initial supercomplex-dependent oxygen consumption rates as a function of menadiol concentration. The initial rate data were fitted to the Michealis-Menten equation using a non-linear fitting program (GraphPad Prism 6.0).

In vitro inhibition of the supercomplex activity by Q203

The supercomplexes were preincubated for 10 min with inhibitors, menadiol was added as electron donor to a final concentration of 200 μM and the oxygen respiration was measured for 3 min. Data were normalized relative to solvent (Ethanol) control for full activity and to a sample with 160 μM Antimycin A for complete inhibition.The supercomplex-dependent initial rates were used to generate an inhibition curve and determine the IC50 value.

Determination of SOD activity in the supercomplex

The SOD activity was measured using a SOD Assay Kit-WST (19160; Sigma-Aldrich) following the kit instruction. Since the absorption maximum of WST-1 formazan is 450 nm and it is proportional to the amount of superoxide anion, the SOD activity (inhibition rate %) as an inhibition activity can be quantified by measuring the decrease in the color development at 450 nm. The SOD activity of the complex purifications and a reference SOD sample with a known specifice activity in IU/mg (international unit per milligram) from bovine erythrocytes (Catalog number S5395; Sigma) were both determined in this assay. Based on the linear relationship between 1/(inhibition rate) and 1/(enzyme activity), the IC50 (50% inhibition activity) of the complex and bovine SOD can be determined. By comparing the IC50 of each, the specific SOD activity of the complex purification was determined and normalized into international unit per milligram. To be noted, the IU here refers to the MF unit (the McCord-Fridovich unit) which is determined by measuring the reduction of cytochrome c by the xanthine oxidase/hypoxanthine system at 550 nm and is the international unit for SOD activity (EC measurement (66). To avoid the potential interference from the cytochrome c domains of QcrC subunite, the McCord–Fridovich test was not used here.

EPR spectroscopy

Low temperature EPR spectra were recorded on a Bruker X-band (9.4 GHz) EMX plus 10/12 spectrometer equipped with an Oxford Instrument EPR 910 liquid Helium continuous-flow cryostat. A cylindrical resonator (ER4119hs TE011) was used for data collection. 100 μL of supercomplex solution at 10 μM was transferred to an EPR tube (Wilmad 707-SQ-250M Quartz EPR sample tube) and rapidly frozen in liquid nitrogen before inserting into the resonator. For the reduced sample, sodium dithionite (100mM) was added and mixed with the supercomplex to give them the spectrum of the reduced form. Specific parameters for EPR experiments were summarized here: temperature 10 K, microwave power 2.0 mW, microwave frequency 9.39 GHz, modulation amplitude 0.9 mT, modulation frequency 100 kHz, time constant 163.84 ms, and scan rate 3.125 mT/s. Multiple scans were accumulated to obtain a good S/N ratio.

Cryo–electron microscopy analysis

Uranyl acetate (1% w/v) was used for negative staining. 5 μL of the supercomplex sample at concentration of 0.05 mg mL−1 was applied to glow discharged copper grids supported by a thin layer of carbon film (Zhongjingkeyi Technology Co. Ltd) for one minute and stored at room temperature. Images were taken on an FEI Tecnai Spirit microscope operating at 120 kV. This data allowed initial model building.

Aliquots (4 μL) of freshly purified supercomplex at a concentration of 0.5 mg mL−1 were applied to glow-discharged holey carbon grids (Quantifoil Au R1.2/1.3). The glow discharge followed the standard recipes of H2 and O2 mixture in Gatan Solarus 950 for 1 min. Grids were blotted for 2.5 s and flash-frozen in liquid ethane cooled by liquid nitrogen using an FEI Mark IV Vitrobot operated at 8°C and 100% humidity. Images were taken using an FEI Titan Krios electron microscope operating at 300 kV with a Gatan K2 Summit detector at a nominal magnification of 18,000×. Images were recorded in super-resolution mode and binned to a pixel size of 1.35 Å. Automated single-particle data acquisition was performed with SerialEM (67). Defocus values varied from 1.3 to 2.7 μm. 32 frames per stack were collected with a total exposure time of 11.4 s. The dose rate was set to ~8 e/pixel/s and the total dose was ~50 e2.

Image processing

We first generated a low-resolution reconstruction of the supercomplex from 53 micrographs of the negative-stained sample. Particle picking was performed with the EMAN2 (68) subroutine in an interactive boxing mode, yielding 25,287 particles. Reference-free classification was performed with, generating 36 good classes from a total of 80 classes. The reconstruction model was generated by the good classes using, which served as an initial model for the subsequent cryo-EM image processing.

For cryo-EM image processing, a total of 7,600 good micrographs were manually selected from 8200 original micrographs. All processing steps were performed using Relion 1.4 or Relion 2.0 (6972). A diagram of the procedures for data processing is presented in fig. S3C. At the whole-image level motion-corrected stacks were produced and binned 2 fold by MotionCorr (73) to produce motion-corrected stacks. Further motion correction and dose weighting was performed by MotionCorr2 (74) to average the output stacks. The contrast transfer function parameters of each image were estimated using Gctf (75). A total of ~1,294,000 particles were automatically picked using Gautomatch ( The particles were extracted using a 3362 pixel box and sorted by two rounds of reference-free 2D classification, resulting in ~844,000 particles selected from good 2D classes. These were subjected to 3D classification with the initial model that was low-pass filtered to 60 Å. The 3D classification was performed several times with different K values resulting in ~400,000 particles of good quality. The relative orientation of the QcrC subunit was variable, producing 3D classes with two different states. After 3D classification, the sets of particles within class-I and class-II were re-extracted and re-centered using a 2922 pixel box and subjected to the final refinement, respectively. The resolution of class-I particles was then corrected using the “high-resolution noise substitution” method (76) and estimated as 3.9 Å according to the gold-standard FSC (Fourier shell correlation) 0.143 criterion. The pixel size was also calibrated to 1.30 Å/pixel provided by the Center for Biological Imaging (CBI), Institute of Biophysics, Chinese Academy of Science. Then the resolution of the class-II particles was estimated to be 4.9 Å. Another 3D classification with class-I particles was performed by applying C2 symmetry, resulting in ~202,000 particles with good symmetry. The particles with C2 symmetry were further refined to yield a final reconstruction with the resolution of 3.5 Å according to the gold-standard FSC 0.143 criterion. Prior to visualization, all density maps were corrected for the modulation transfer function (MTF) of the detector, and then sharpened by applying a negative B-factor (77) that was estimated using automated procedures. Local resolution variations were estimated using ResMap (78). The orientation distribution of the particles used in the final reconstruction was calculated using RELION 2.0.

Model building and refinement

All the subunits were initially built as poly-alanine chains, except for the residues that coordinate the prosthetic groups. Then residue assignment was performed, which was guided largely by secondary structure analysis carried out using PSIPRED (79) and Phyre2 (80).

It is noteworthy that the sequence of PRSAF1 in the model has an extra 16 amino acids before the N terminus of WP_003893930.1 but it is consistent with the revised complete ORF in the genome of M. smegmatis str. MC2 51 (genome locus tag AD56_RS0106885 added with the immediately preceding 64 bp). Model building was performed using Coot 0.8 (81) and further refined using the real space method in Phenix (82), except for the SOD catalytic domain (47Ala-236Gly) which was homology modeled (83) based on the structure of Sodc from M. tuberculosis (PDB Code: 1PZS) (36) and fitted into the map with the aid of the “Fit in Map” feature of Chimera (84). The completeness of the final resulting model was summarized in table S3.

For the phospholipids, the results of three independent mass spectrometry analyses confirmed that the majority variants of lipids in the purified sample are PE, PI and CL (fig. S1, L and N). No PC or other branching phospholipids was detected. Since most of the phospholipids are localized through the neighboring pipetides and are located in the transmembrane space which falls in the 3.2to 3.5 Å high local resolution range, the CL moleculars were readily identified based on the four alkyl groups and the PI and PE molecules were distinguishable through the significent difference in their “head” group size. A CL molecular was also distinguishable from two PI or PE moleculars through the differnet distances between the two phosphatidic acid moieties. Finially, all the identified and assigned lipid backbones were modeled according to the EM potential map with necessary truncations and the results are summarized and illustrated against the context of the covering map in fig. S3I.

All the figures were created using PyMOL (85) or UCSF Chimera (84).

Supplementary Materials

Figs. S1 to S8

Tables S1 to S4

References (8694)

Movie S1

References and Notes

Acknowledgments: We thank J. Zhang, Y. Zhang, and L. Wang from C.-C. Wang’s Research Group (National Laboratory of Biomacromolecules, Institute of Biophysics, CAS) for their technical support on Clark-type oxygen electrode and oxygen consumption assay and K. Mi (CAS Key Laboratory of Pathogenic Microbiology and Immunology, Institute of Microbiology, CAS) for sharing the strain M. smegmatis mc2 51. We are also grateful to B. Zhu, X. Huang, and G. Ji from Center for Biological Imaging (CBI), Institute of Biophysics, CAS, and staff members from National Center for Protein Science Shanghai (NCPSS) for their technical support on cryo-EM and C. Peng from the Mass Spectrometry System of NCPSS for his technical support. We also thank T. Niu from the HPC-Service Station in CBI and X. Jian and X. Meng from TianHe-1(A) at the National Supercomputer Center in Tianjin for computational support. Funding: This work was supported by grants from the National Key Research and Development Program of China (grant 2017YFC0840300), the Strategic Priority Research Program of the Chinese Academy of Sciences (grant XDB08020200), the State Key Development Program for Basic Research of the Ministry of Science and Technology of China (973 Project Grants 2014CB542800 and 2014CBA02003 to Z.R. and 2014CB910700 to F.S.), and the National Natural Science Foundation of China (Grant Nos. 813300237 and 81520108019). Author contributions: Z.R. conceived, initiated, and coordinated the project. H.G., Ju.L., and S.W. purified the M. smegmatis SC III-IV supercomplex; H.G., Y.T., and Ju.L. characterized the spectroscopic features of the samples; Ji.L., H.-Y.Y., C.V.R., S.M.L., and G.S. preformed mass spectrometry analysis and identified the contents of the complex; Y.T. and Q.W. set up the biochemical assays and measured the menadiol:O2 oxidoreductase activity of the complex; H.G. measured the SOD activity of the complex; L.Y. and C.T. performed the EPR experiments and data analysis; A.X. and Q.W. collected and processed cryo-EM data; Q.W. reconstructed the 3.5 Å resolution map and supervised cryo-EM structure determination; Q.W., A.X., R.G., and W.J. built and refined the structure model; Ju.L., Q.W., F.S., H.G., A.X., X.Y., Y.S., X.L., M.J., C.T., C.Y., B.J., Z.L., L.W.G., L.-L.W., and Z.R. analyzed the structure and discussed the results; and the manuscript was written by F.S., Q.W., Ju.L., W.J., A.X., H.G., R.G., Z.L., L.W.G., L.-L.W., and Z.R. Competing interests: The authors declare no competing interests. L.W. is a founder and consultant of Oxford Biotrans, UK. C.V.R. is a founding director and consultant of OMass Technologies, UK. Data and materials availability: All data are available in the manuscript or the supplementary materials. The accesion no. for the 3D cryo-EM density map reported in this paper is EMD-9610. The PDB accession no. for the coordinates of the CIII-CIV complex is 6ADQ.
View Abstract

Stay Connected to Science

Navigate This Article