Structures of Respiratory Supercomplex I+III<sub>2</sub> Reveal Functional and Conformational Crosstalk

The overall arrangement of the SC I+III2 shown here is similar to that in the full respirasome (Letts et al., 2016b) as well as in situ (Davies et al., 2018). CIII2 contacts CI at two main sites: (1) in the IMM, where CI subunit NDUFA11 contacts UQCRB and UQCRQ of the adjacent CIII protomer (Figure 6A), and (2) in the mitochondrial matrix, where CI subunits NDUFB4 and NDUFB9 (B22) contact UQCRC1 of that same CIII protomer (Figure 6B). At the interface of the mitochondrial inner membrane and matrix, only van der Waals contacts can be seen between the top of NDUFA11 (Leu49) and the main chain of the adjacent UQCRB helix (between Lys11 and Trp12). However, a cluster of positively charged residues extending from both NDUFA11 (Arg103 and Arg105) and UQCRQ (Arg42) suggests the possibility of lipid-mediated bridging between CI and CIII2 (Figure 6A). On the inter-membrane space side of the membrane, several charged and polar residues are in close contact (NDUFA11: Glu16, His18, Arg19, and Gln78; UQCRQ: Tyr56, Gln64, and Lys68), which allows the formation of stabilizing hydrogen bonds or salt bridges (Figure 6A). Within the mitochondrial matrix, salt-bridging interactions can be seen between NDUFB9 and NDUFB4 of CI and UQCRC1 of the adjacent CIII protomer (Figure 6B). Possible salt bridges include NDUFB9 Lys54 and UQCRC1 Asp20, NDUFB9 Asp55 and UQCRC1 Arg24, NDUFB4 Glu27 and UQCRC1 Lys51, and NDUFB4 Arg29 and UQCRC1 Glu225 (Figure 6B). Although these interacting residues are conserved in mammals, they diverge quickly in other lineages, likely reflecting different ways of forming SC I+III2 across eukaryotes (Davies et al., 2018).Figure 6Interactions, CoQ Density, and Overall Arrangement of the Amphipol-Stabilized SC I+III2 Particles(A) Interaction between CI and CIII2 in the membrane. CI subunit NDUFA11 in cyan, CIII2 subunits UQCRB in green, and UQCRQ in light green.

The putative “positive cluster” lipid-binding site is circled and labeled. Side chains of important residues are shown as sticks and colored by atom, with nitrogen blue, oxygen red, and carbon the same color as the subunit. Models shown as cartoons throughout.(B) Interaction between CI and CIII2 in the mitochondrial matrix. CI subunits NDUFB4 are in cyan and NDUFB9 in light blue, and CIII2 subunit UQCRC1 are in light green. Important side chains are shown and colored as in (A). Putative salt-bridging interactions indicated by dashed ovals.(C) Density for CoQ-10 binding in the proximal QN-site from the CIII2 focused maps. CIII2 subunits are in green, and CoQ, heme bH, and the side chain of MT-CYB His201 are shown as sticks and colored by atom, with carbon gray, nitrogen blue, oxygen red, and iron orange.(D) Density for CoQ-10 binding in the distal QN site from the CIII2 focused maps. CIII2 subunits, CoQ, heme bH, and the MT-CYB side chain His201 are shown as in (C).(E) Density for CoQ binding in the distal QP site from the CIII2 focused maps. CIII2 subunits, CoQ, and heme bL are shown as in (C).(F) Slice through the membrane domains of CI and CIII2 within the closed structure of the SC looking from the mitochondrial matrix. Amphipol-lipid belt is shown at low contour in red, and CI (blue) and CIII2 (green) are shown at a higher contour. Each CIII protomer is colored a different shade of green. CoQ active sites are marked: solid wedges for location on the matrix leaflet of the membrane and dashed wedges for inter-membrane space leaflet. Black and yellow ovals indicate lipid binding pockets of CI; black and yellow circle indicates lipid binding pocket of NDUFA11; black and yellow X indicates barrier to CoQ diffusion caused by contact site between CI and CIII2 in the membrane.

Yellow arrow illustrates the shortest path for CoQ diffusion from the CI Q-tunnel to the proximal QP site of CIII2.

Given the hydrophobic nature of CoQ-10, it would remain in the fragment of lipid bilayer that is co-purified with the SC I+III2 particles. Hence, even if only limiting amounts of CoQ-10 were co-purified within the SC, its local concentration would be high. Accordingly, although no exogenous CoQ was added to our cryoEM sample, clear density for the CoQ-10 head group can be seen in the CIII2 focused maps at three of the four CoQ binding sites (Figures 6C–6E). Density could be seen at both QN sites, which are expected to have a higher affinity for the oxidized CoQ-10 (Figures 6C and 6D) and the QP site distal from the CI Q-tunnel (Figure 6E). However, the QP site proximal to the CI Q-tunnel did not display any density for CoQ-10, suggesting different affinities for CoQ-10 in the QP sites of the different MT-CYB protomers.

At low contour, the cryoEM maps showed density for a disordered layer of amphipols and lipids forming a belt around the hydrophobic surfaces of the SC (Figure 6F). On the CIII2-proximal side of CI, the four-TMH-containing subunit NDUFA11, which directly interacts with CIII2, is shaped like a lipid-filled arch (Figure 6F), contacting CI with the ends (springers) and CIII2 with the top (crown). The dominance of lipid-mediated contacts between CI and NDUFA11 may explain why this subunit is so easily lost or disordered by detergent when CI is extracted. The rest of the CI-CIII2 interface within the membrane involves two large lipid-filled cavities, made discontinuous by the close interactions among NDUFA11, UQCRB, and UQCRQ. These interactions likely act as a barrier to the free diffusion of CoQH2 as it exits the CI Q-tunnel (Figure 6F).

The two Q cavities of CIII2 penetrate deep into the core of the dimer and are separated from each other by an interaction between the two MT-CYB protomers (Figure 6F). Each Q cavity contains a CoQH2-oxidizing QP site and a CoQ-reducing QN site; however, the QP and QN sites within each cavity derive from the opposite CIII protomer (Figure 6F) and, hence, the QP and QN sites of one MT-CYB subunit face opposite cavities. One of the Q cavities is proximal to the Q-tunnel of CI, while the other is distal on the opposite side of the dimer and much less accessible (Figure 6F). This arrangement led to the hypothesis that symmetry-breaking within CIII2 may result in the specialization of the Q cavities, with the proximal Q cavity specialized for CoQH2 oxidation and the distal cavity specialized for CoQ reduction (Letts and Sazanov, 2017, Letts et al., 2016b). In the amphipol-stabilized SC, trapping of CoQ would limit the diffusion path from the CI Q-tunnel to CIII2 and vice versa to the lipid-amphipol belt surrounding the SC (Figure 6F). Given the overall structure of the SC (Figure 6F) and evidence of poor CoQ exchange in the amphipol-stabilized particle (Figure 3L), functional symmetry-breaking of CIII2 and specialization of the Q-cavities is highly likely.

Clear differences can be seen in the local resolution of specific regions of CI and CIII2 between the different reconstructions (Figure 7; Figures S7B and S7C). In the SC I+III2 class containing the closed state of CI, the density corresponding to ND4 is of lower resolution compared with the surrounding ND2 and ND5. This lower resolution patch is not seen in the local resolution maps of the open states (Figures 7B and 7D). A major difference between the closed and open states is that contact between the CI peripheral arm and membrane arms via NDUFA5 and NDUFA10 is seen only in the closed state (Figures 4F and 5C). Thus, in the closed state, conformational flexibility in the peripheral arm may be transmitted into the membrane arm. This interaction may also help facilitate coupling of CoQ reduction and H+ pumping during CI turnover.Figure 7Local Resolution Maps of the SC I+III2 ReconstructionsLocal resolution maps of the three 4.2 Å SC I+III2 structures shown with the closed state (left), open state 1 (middle), and open state 2 (right) throughout.(A) CI side view.(B) Slice through CI from the same view as in (A).(C) View from the mitochondrial matrix.(D) Slice through the membrane domains of CI and CIII2 from the same view as in (C).(E) Slice through the membrane domains of CI and CIII2 viewed from the “heel” of CI looking at the position indicated by the dashed lines in (C).See also Figure S7.

Surprisingly, the CIII2 MT-CYB that is in close contact with CI showed lower overall resolution in the SC class containing closed CI than in those with open CI (Figures 7D and 7E). This indicates that this MT-CYB subunit has increased motion or is more disordered in presence of closed CI. The MT-CYB subunit of the CIII protomer in close contact to CI contains the CoQH2 oxidation site most proximal to the Q-tunnel of CI. Hence, this MT-CYB subunit is likely responsible for oxidizing the majority of CoQH2 produced by CI within the SC. The decrease in the local resolution of this subunit only in the closed state of CI strongly suggests that there is crosstalk between the two complexes and that CIII2 may be primed for CoQH2 oxidation upon the closing of CI. Nonetheless, at this resolution it remains unclear how this crosstalk between CI and CIII2 occurs.

We present here the first chromatographic isolation of functional SC I+III2 from mammalian mitochondria, followed by its detailed functional and structural characterization. Previous isolations of this SC from mitochondrial membranes displaying NADH:cyt c oxidoreductase activity have been reported (Hatefi and Stiggall, 1978, Hatefi et al., 1961). However, the purity of these preparations is unclear, and their stability is poor. The preparation presented here displays a homogeneous single peak by SEC (Figure 1C) and appears as a single major band on BN-PAGE gels (Figure 1D). The characterization of amphipol-stabilized SC I+III2 NADH:cyt c oxidoreductase activity demonstrates that it is possible to study the transfer of electrons from NADH to cyt c, both in the absence or presence of added CoQ analogs (Figure 2). The rates of NADH oxidation and cyt c reduction observed when the SC was supplemented with 10 μM DQ were lower than what has been observed for CI and CIII2 turnover in membranes (Letts et al., 2016a, Ripple et al., 2013) (Figures 2E and 2F). However, the activity of both complexes increased to native levels upon the addition of lipid and detergent (Figures 2H and 2I). With the purity and inhibitor sensitive activity achieved here (Figures 1, 2, and S3), this preparation can be used for the study of SC I+III2 by diverse biophysical methods.

This preparation is not without its limitations. Because of the lack of a sealed membrane compartment, there is no Δψ or ΔpH present in the system. This is significant, as CI normally operates near equilibrium with the proton motive force (Ripple et al., 2013), and as the redox poise of the b-hemes in CIII2 is affected by both Δψ and ΔpH (Kim et al., 2012). Also, because of the nature of the amphipol-based stabilization of the SC in solution, CoQ-10 and to a lesser extent DQ are trapped in the SC particle, and their diffusion is constrained to the lipid-amphipol belt (Figure 6F). CoQ-10 is highly hydrophobic; hence, when extracted into an aqueous solution, CoQ-10 will partition into the hydrophobic environment of the deep lipid pockets present in the SC. Together with the affinities of CI and CIII2 for CoQ-10, a concentrating effect of CoQ-10 within the extracted SC is expected. Hence, it does not follow from CoQ-10 trapping in this context that CoQ-10 is not free to diffuse and exchange with the bulk membrane pool in the hydrophobic environment of the IMM.

Nevertheless, the fact that CoQ is “trapped” by the amphipol-stabilized SC I+III2 was advantageous, as it allowed us to test the hypothesis that CoQ-10 trapping by SC I+III2 in situ would increase the rate of the ETC. This system allowed us to determine how CoQ trapping affects CI and CIII2 activity and to place what we learned into a structural context. Given the arrangement of CI and CIII2 in the SC (Figure 6F), it is likely that only one side of CIII2 would be readily supplied with CoQH2 from CI and that the symmetry of CIII2 would be functionally broken. Nonetheless, given the Hill coefficients of 2 for both the QP- and QN-site CIII2 inhibitors (Figure S3), it is clear that the CoQ sites in the distal CIII2 Q-cavity also play an important role in turnover. Rate limitation of CI by CIII2 can be understood in the context of the SC I+III2 structure (Figure 6F). Trapping the CoQ between CI and CIII2 would limit the ability of CIII2 to oxidize CoQH2 to a single QP site. In the context of the structure, the ∼2-fold increase in CIII2 activity may be due mainly to a re-establishment of two accessible QP sites caused by the free exchange of DQH2 in the LD mixture (Table S3). This indicates that CIII2 within the amphipol-stabilized SC I+III2 particles is operating at near maximal activity in the experimental context but that only one Q-cavity has access to reduced CoQH2. The rate coupling in these contexts is nearly 2.0 (Table S3), demonstrating that CI activity is limited by CIII2.

Thus, this argues against the view that trapping of CoQ-10 by SCs may increase the flux through the ETC but instead indicates that CI can function more effectively if CoQ is able to exchange with the membrane pool and is not dependent solely on the proximal CIII2 QP site for regeneration of oxidized CoQ. The native arrangement within the IMM may allow both CI and CIII2 to operate at the highest rates via exchange with the bulk pool and not via CoQ trapping. Given the structure of the SC, the only way for CoQ trapping to result in higher rates would be if the turnover of CI and the single CIII protomer were well matched. Because of the nature of the Q-cycle, CIII2 must oxidize two CoQH2 for each CoQ reduced by CI. Given the complexity of the Q-cycle mechanism—involving electron bifurcation, conformational transitions of UQCRFS1, and three simultaneously bound substrates—this strict requirement for rate matching may be difficult to accommodate. The mismatch in rates between CI and CIII2 turnover may in part explain why there is an ∼3-fold excess of CIII2 compared with CI in mitochondrial membranes (Schägger and Pfeiffer, 2001).

We demonstrate here that the CI closed-to-open state transition involves not only a rotation of the CI peripheral and membrane arms and the ordering of the NDUFS2 β1-β2 and ND3 TMH1-TMH2 loops (Figure 4G) but also a notable rotation of ND6’s TMH3 by ∼100° that is driven by the conversion of a π-bulge into an α-helix (Figures 4H and 4I). This ND6 TMH3 π-to-α transition was also seen in mouse mitochondrial CI between active and deactive forms (Agip et al., 2018). Although the ovine closed state is likely to be identical to the mouse active form, this is not necessarily the case for the ovine open states and the mouse deactive form. The deactive form of CI is obtained upon incubating the enzyme at 37°C without substrates and then converts back to active upon turnover (Vinogradov, 1998). No significant amount of deactive form was detected biochemically in the ovine enzyme as prepared (data not shown). This is consistent with mouse CI, which is ∼90% in the active form as purified (Agip et al., 2018). Given that our preparation was largely in the active state, but many SC particles were in open states, it is likely that CI undergoes conversion between open and closed states while active as part of its catalytic cycle. In this case, the deactive form would represent one of the open-like states with likely more extensive unfolding of loops around the Q-site. A direct link between closed-to-open transitions and the catalytic cycle was also proposed in a recent cryoEM study of Y. lipolytica CI (Parey et al., 2018). Given that the ND6 TMH3 π-bulge is conserved in the structures of yeast and bacterial CI (Baradaran et al., 2013, Zickermann et al., 2015), this rotation is likely a conserved feature of CI turnover and may be a novel way of achieving efficient conformational coupling by the enzyme. Although the rotation was not observed in the Y.

lipolytica or bacterial structures to date, this may be due to limited resolution or crystal contacts.

By careful classification of the SC particles, we show that there are state-dependent differences in local resolution (Figure 7). In the closed state of CI, conformational flexibility in the peripheral arm may be transmitted to ND4 in the membrane arm via interaction between NDUFA5 and NDUFA10 (Figure 7), possibly facilitating coupling of CoQ reduction and H+ pumping during CI turnover. How this state-dependent conformational flexibility is further communicated to CIII2 remains unclear. However, our structures of multiple states of CI within the SC at similar resolutions suggest that the state of CI affects CIII2’s MT-CYB conformational flexibility (Figures 7D and 7E). The fact that CoQ was found at only three of the four possible binding sites within CIII2 (Figures 6C–6E) also suggests a functionally relevant effect of CI on CIII2. Overall, our data further support the hypothesis of functional symmetry breaking in the CIII2 dimer by CI.

Our pure, biochemically defined system allowed us to demonstrate that in contrast to the long-proposed substrate-channeling role for the SC, CoQ trapping in the SC limits CI turnover and would reduce the overall rate of the ETC in vivo. If CoQ trapping is inefficient, what could be the functional role of the SCs? Possible roles in the stabilization of the individual complexes, the reduction of ROS production, and the prevention of non-specific protein aggregation such that the IMM retains suitable CI/CIII2/CIV ratios remain feasible. The SCs may provide kinetic advantages by bringing the active sites into close proximity, but this advantage would exist even in the absence of CoQ trapping by the SCs, as it is only dependent on the distance between the active sites. Moreover, under high turnover, because of the structure of the mitochondrial membranes, individual CIs may outpace the capacity for CoQ pool equilibration, in which case the local CoQH2 concentration may become higher than in the bulk membrane (Budin et al., 2018). By ensuring close association of CI and CIII2 in SCs, but still allowing free exchange of CoQ with the bulk, the SC would prevent any local buildup of CoQH2, while at the same time not limiting CI’s activity by making it entirely dependent on the adjacent CIII2 for the re-oxidation of CoQH2. We show here, for the first time, functional crosstalk between CI and CIII2 within the SC. Although more work is needed to confirm and characterize the nature of this coordination, our results and our new experimental framework have significant implications for the physiological roles of respiratory SCs, suggesting more subtle functionally relevant interactions between CI and CIII2.

REAGENT or RESOURCESOURCEIDENTIFIERBiological SamplesOvine heartsC Humphrey’s & Sons (Chelmsford, UK)N/AChemicals, Peptides, and Recombinant ProteinsNADHSigma-AldrichCat# 1246440005; CAS 606-68-8DecylubiquinoneSanta Cruz BiotechnologyCat# 358659; CAS 55486-00-5Cytochrome c from bovine heartSigma-AldrichCat# 30398; CAS 9007-43-6BSACell signaling technologiesCat# 9998SLMNGAnatraceCat# NG310Amphipol A8-35AnatraceCat# A835DOPCSigma-AldrichCat# P6354; CAS 4235-95-4CardiolipinSigma-AldrichCat# 21979; CAS 200-578-6Critical Commercial AssaysNativePAGE 3-12% Bis-Tris GelInvitrogenCat# BN1003NativePAGE Cathode Buffer Additive (20X)InvitrogenCat# BN2002NOVEX 4-20% Tris-Glycine GelInvitrogenCat# XP04205Pierce BCA Protein Assay KitThermo Fisher ScientificCat# 23227MonoQ 5/50 GLGE HealthcareCat# 17-5166-01Q-Sepharose Fast FlowGE HealthcareCat# 17-0510-01Superose 6 prep grade, XK 6/100GE HealthcareCat# 90-1003-36Deposited DataOvine CI modelFiedorczuk et al., 2016PDB: 4HEABovine CIII2 modelHuang et al., 2005PDB: 1PPJCoordinates of ovine CI Peripheral armThis studyPDB: 6Q9DCryoEM map of ovine CI Peripheral ArmThis studyEMDB: 4480Coordinates of ovine CI Membrane ArmThis studyPDB: 6Q9BCryoEM map of ovine CI Membrane ArmThis studyEMDB: 4479Coordinates of ovine CIII2This studyPDB: 6QE9CryoEM map of ovine CIII2This studyEMDB: 4481Coordinates of ovine SC I+III2 closed classThis studyPDB: 6QBXCryoEM map of ovine SC I+III2 closed classThis studyEMDB: 4493Coordinates of ovine SC I+III2 open class 1This studyPDB: 6QC3CryoEM map of ovine SC I+III2 open class 1This studyEMDB: 4495Coordinates of ovine SC I+III2 open class 2This studyPDB: 6QC2CryoEM map of ovine SC I+III2 open class 2This studyEMDB: 4494Coordinates of ovine SC I+III2 open class 3This studyPDB: 6QC4CryoEM map of ovine SC I+III2 open class 3This studyEMDB: 4496Coordinates of ovine isolated CI classThis studyPDB: 6QA9CryoEM map of ovi

ne isolated CI classThis studyEMDB: 4482Coordinates of ovine CI FRC closed classThis studyPDB: 6QC5CryoEM map of ovine CI FRC closed classThis studyEMDB: 4497Coordinates of ovine CI FRC open class 1This studyPDB: 6QC6CryoEM map of ovine CI FRC open class 1This studyEMDB: 4498Coordinates of ovine CI FRC open class 2This studyPDB: 6QC8CryoEM map of ovine CI FRC open class 2This studyEMDB: 4500Coordinates of ovine CI FRC open class 3This studyPDB: 6QC7CryoEM map of ovine CI FRC open class 3This studyEMDB: 4499Coordinates of ovine CI FRC open class 4This studyPDB: 6QC9CryoEM map of ovine CI FRC open class 4This studyEMDB: 4501Coordinates of ovine CI FRC open class 5This studyPDB: 6QCACryoEM map of ovine CI FRC open class 5This studyEMDB: 4502Coordinates of ovine CI FRC open class 6This studyPDB: 6QCFCryoEM map of ovine CI FRC open class 6This studyEMDB: 4505CryoEM map of ovine CI FRC poor closed classThis studyEMDB: 4506CryoEM map of ovine CI FRC poor open classThis studyEMDB: 4507Software and AlgorithmsFEI EPUFEIhttps://www.fei.com/software/epu/MotionCor2Zheng et al., 2017http://msg.ucsf.edu/em/software/motioncor2.htmlGctfZhang, 2016https://www.mrc-lmb.cam.ac.uk/kzhang/Gctf/RELION-2.0Kimanius et al., 2016https://www2.mrc-lmb.cam.ac.uk/relionUCSF ChimeraPettersen et al., 2004https://www.cgl.ucsf.edu/chimera/UCSF ChimeraXGoddard et al., 2018http://www.rbvi.ucsf.edu/chimerax/CootEmsley et al., 2010https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/PHENIXAdams et al., 2010https://www.phenix-online.org/MolProbityChen et al., 2010http://molprobity.biochem.duke.edu/PyMolSchrödinger, LLC.https://pymol.org/2/PRISM 5GraphPad Softwarehttps://www.graphpad.com/scientific-software/prism/REPD’Imprima et al., 2017https://github.com/rkms86/REPOtherEM grid R 0.6/1 on 300 mesh CuQuantifoilItem# N1-C11nCu30-01

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Leonid A. Sazanov ([email protected])

Ovis aries (sheep) hearts were purchased from C Humphreys & Sons (Chelmsford, UK).

Mitochondria were isolated and membranes prepared as previously described (Letts et al., 2016a). Complex (C)I purification in LMNG was initially performed as previously reported (Letts et al., 2016a), consisting of a single anion-exchange step followed by SEC. In short, 5% (w/v) LMNG was added drop-wise to the washed mitochondrial membranes to a final concentration of 1% LMNG, followed by stirring for ∼30 min at 4°C and centrifugation at 48,000 g for 45 min. The supernatant was filtered (0.45 μm pore size polyethersulfone membrane) and loaded onto a pre-equilibrated 45 mL Q-Sepharose FF anion-exchange column (GE Healthcare). The Q-Sepharose buffers (A and B) contained 20 mM Tris-HCl, pH 7.4, 10% (v/v) glycerol, 1 mM EDTA, 1 mM DTT and 0.04% LMNG; additionally, buffer B contained 1 M NaCl; the Q-Sepharose column was pre-equilibrated at 5% buffer B in buffer A. After application of the mitochondrial extract, the Q-Sepharose column was washed with 50 mL of 5% buffer B, then with a 30-mL linear gradient of 5%–22%, and finally with 150 mL of 22% buffer B. CI was then eluted with a 300-mL linear gradient of 22%–30.5% buffer B (CI elution gradient; Figure S1A). Any remaining protein was then eluted with 100% buffer B. The Q-Sepharose gradient was run overnight at 1.0 mL/min at 4°C. CI-containing fractions were pooled based on NADH:FeCy activity and concentrated to 1.5-2.0 mL. Once it was determined that this Q-Sepharose protocol resulted in a significant proportion of CI in the high-salt wash (Figure S1A), the CI elution gradient was extended to 22%–60% buffer B over 450 mL, resulting in the isolation of two distinct peaks containing NADH:FeCy (CI) activity (Figure S1B). As a further optimization, the elution gradient was changed to 22%–50% over 250 mL. This resulted in the abundant isolation of SC I+III2 (Figure 1A).

It was also essential to increase the flow rate over the column to 3.0 mL/min throughout to minimize the amount of time the protein was on the Q-Sepharose column.

Samples eluted from the Q-Sepharose column were then applied either to the same column again and run with a similar gradient to separate CI and CIII2 (Figure S1C), or to a Superose 6 prep grade XK 6/100 size exclusion chromatography column equilibrated in buffer S (20 mM HEPES, pH 7.4, 2 mM EDTA, 10% glycerol, 50 mM NaCl and either 0.05% LMNG (Figure S1D) or no LMNG for the amphipol samples (Figure 1B)), and eluted overnight at 0.35 mL/min at 4°C for either step. For the amphipol sample, amphipol A8-35 was added to a final concentration of 0.3% (w/v) immediately following elution from the Q-Sepharose gradient. This mixture was then tumbled at 4°C for ∼30 min before concentrating to ∼1.5-2.0 mL and loading onto the Superose 6 column. Subsequent SEC runs were done using the same column either in 0.5% LMNG (Figure S1E) or in no LMNG (Figure 1C) buffer S. After elution, the fractions containing NADH:FeCy activity were pooled and concentrated. These samples were either used directly for cryoEM grid preparation or made to 30% glycerol and stored in liquid nitrogen. Samples stored in this way maintained high activity over several months.

Samples diluted with 4X sample buffer (50 mM BisTris-HCl, pH 7.2, 50 mM NaCl, 10% w/v glycerol, and 0.001% Ponceau S) were loaded onto a NativePAGE 3%–12% Bis-Tris Gel (Invitrogen) and run in the cold room at 4°C. The cathode buffer was 50 mM Tricine, 50 mM BisTris-HCl, pH 6.8 plus 1X NativePAGE Cathode Buffer Additive (Invitrogen) and the anode buffer was 50mM Tricine, 50 mM BisTris-HCl, pH 6.8. Gels were run at 150 V for ∼60 min and then the voltage was increased to 250 V for the remainder of the run ∼90 min.

Mass spectrometry (MS) analysis was performed essentially as previously described (Letts et al., 2016a). Briefly, polyacrylamide gel slices (1–2 mm) containing bands of the purified proteins (Figure S1H) were prepared for mass spectrometric analysis by manual in situ enzymatic digestion. The excised protein gel pieces were placed in wells of a 96-well microtiter plate and destained with 50% (v/v) acetonitrile and 50 mM ammonium bicarbonate, reduced with 10 mM DTT, and alkylated with 55 mM iodoacetamide. After alkylation, proteins were digested with 6 ng/μL trypsin (Promega, UK) overnight at 37°C. The resulting peptides were extracted in 2% (v/v) formic acid, 2% (v/v) acetonitrile. The digest was analyzed by nano-scale capillary LC-MS/MS using an Ultimate U3000 HPLC (ThermoScientific Dionex, San Jose, CA) to deliver a flow of ∼300 nL/min. A C18 Acclaim PepMap100 5 μm, 100 μm x 20-mm nanoViper (ThermoScientific Dionex) trapped the peptides prior to separation on a C18 Acclaim PepMap100 3 μm, 75 μm x 250 mm nanoViper (ThermoScientific Dionex). Peptides were eluted with a gradient of acetonitrile. The analytical column outlet was directly interfaced via a nanoflow electrospray ionization source, with a hybrid dual pressure linear ion trap mass spectrometer (Orbitrap Velos, ThermoScientific). Data-dependent analysis was carried out, using a resolution of 30,000 for the full MS spectrum, followed by 10 MS/MS spectra in the linear ion trap. MS spectra were collected over an m/z range of 300-2000. MS/MS scans were collected using the threshold energy of 35 for collision-induced dissociation. LC-MS/MS data were then searched against a protein database (mammalian subset of UniProt KB) using the Mascot search engine program (Matrix Science, UK) (Perkins et al., 1999).

Database search parameters were set with a precursor tolerance of 10 ppm and a fragment ion mass tolerance of 0.8 Da. One missed enzyme cleavage was allowed, and variable modifications for oxidized methionine and carbamidomethyl cysteine were included. MS/MS data were validated using the Scaffold program (Proteome Software) (Keller et al., 2002). All data were additionally interrogated manually. See Table S1.

Measurement of organic phosphate content was performed according to the protocol of Anderson and Davis without modification (Anderson and Davis, 1982). Lipid was extracted from protein samples (∼20 μL) by addition of 200 μL of 2:1 chloroform/MeOH followed by vortexing. Then, 50 μL of 125 mM NaCl was added, and the samples were vortexed again. The phases were separated by centrifugation for 3 min at ∼13,000 x g, and the lower chloroform phase was desiccated and used for measurement. See Table S2.

CI NADH:FeCy activity was measured by spectroscopic observation of NADH oxidation at 340 nm wavelength using a Shimadzu UV-2600 UV-VIS spectrophotometer with CPS-100 thermoelectrically temperature-controlled cell positioner and modified by Rank Brothers for continuous sample stirring. SC I+III2 NADH:cyt c oxidoreductase activity was measured at 340 nm for NADH oxidation and 550 nm for cyt c reduction using either the same Shimadzu UV-2600 UV-VIS spectrophotometer as above or a Spectramax M2e Plate-Reader (Molecular Devices). The buffer used for NADH:FeCy activity was 20 mM HEPES, pH 7.4, 50 mM NaCl, 2 mM EDTA, 0.1% DDM, 100 μM NADH and 1 mM potassium ferricyanide (KFeCy). NADH:Cyt c oxidoreductase activity of the isolated SC I+III2 were done in the standard buffer (SB): 100 mM HEPES, pH 7.4, 50 mM NaCl, 10% glycerol and 4 μM KCN, 50 U/mL SOD; or lipid-detergent (LD) buffer: SB plus 0.1% CHAPS, 0.1% LMNG, 0.25 mg/mL 4:1 DOPC:CL and 10 μM BSA. Both NADH oxidation activity (340 nm, top) and cyt c reduction activity (550 nm, bottom) were monitored in separate experiments for each condition. Activity is reported as Units (1 U = 1 μmol substrate per min per mg of SC I+III2). Where Michaelis-Menten curve fitting is appropriate, the Vmax and apparent Km are shown on the plots ± standard error. Where the concentration-activity curves are not hyperbolic, the maximum activity (Actmax) is shown, mean ± SEM. SC I+III2 was added to a final concentration of 1.3-10 nM depending on the experimental conditions and amount of activity. Measurements done using the Shimadzu UV-2600 UV-VIS spectrophotometer were carried out using disposable 2 mL polystyrene cuvettes (LLG Labware) with constant stirring. Measurements done using the Spectramax M2e Plate-Reader were done using 200 μL reaction buffer in 96-well plates with 10-20 s of stirring before beginning to record.

All activity measurements were performed at 30°C and initiated by addition of NADH. Concentrations of NADH stocks were determined using the Shimadzu UV-2600 UV-VIS spectrophotometer and the known extinction co-efficient of 6.22 mM-1 cm-1. These standards were then used to generate a standard curve on the Spectramax M2e Plate-Reader under experimental conditions in the presence of 100 μM cyt c. Standard curves for oxidized and reduced (by dithionite) cyt c were also generated in a similar fashion in situ under experimental conditions in the presence of 100 μM NADH. All protein concentrations were determined with the Pierce bicinchoninic acid (BCA) assay kit using BSA standards (Thermo Fisher, Waltham, MA). All samples for protein concentration determination were diluted at least 10-fold into 20 mM HEPES, pH 7.4, 0.1% DDM, and 50 mM NaCl buffer to reduce interference from glycerol in the sample buffers. See Figures 2, 3, and S2 and Table S3.

Initial grid preparation using the amphipol samples resulted in only broken particles. We hypothesized that this was due to instability of the SC at the large air-water interface present during blotting (Glaeser and Han, 2017, Glaeser et al., 2016) and found it necessary to add back detergent to reduce the surface tension and preserve the particles. Using the polyoxyether detergent Brij-35 generated high-quality grids with a high-density of particles in the holes, without solubilizing SCs. The final optimal blotting conditions were from application of 2.7 μL aliquots of ∼2 mg/mL amphipol stabilized SC I+III2 (in 250 mM NaCl, 20 mM HEPES, pH 7.7, 0.02% Brij-35) to glow-discharged holey carbon grids (Quantifoil R0.6/1 CU) followed by blotting for 30 s at 4°C, 95% humidity and flash freezing in liquid ethane using an FEI Vitrobot IV. CryoEM data acquisition was performed on a 300 kV FEI Titan Krios electron microscope with a Falcon II camera. Automated data collection was performed with the FEI EPU package. Micrographs were recorded at a nominal magnification of 100,000 X, resulting in a pixel size of 1.40 Å. Defocus values varied from −1.5 to −3.0 μm. The dose rate was ∼50 electrons per pixel per second. Exposures of 2 s were dose-fractionated into 34 frames, leading to a dose of 1.5 electrons per Å2 per frame and a total accumulated dose of 51 electrons per Å2. A total of 1,854 micrographs were collected. See Table S4.