Functional Dissection of the Proton Pumping Modules of Mitochondrial Complex I

A catalytically active subcomplex of respiratory chain complex I lacks 14 of its 42 subunits yet retains half of its proton-pumping capacity, indicating that its membrane arm has two pump modules.


Introduction
Respiratory complex I (proton-pumping NADH:ubiquinone oxidoreductase) is a very large membrane integral multiprotein complex found in most energy converting membranes of bacteria and eukaryotes [1]. It plays a central role in cellular energy metabolism, and complex I defects have been implicated in a number of bioenergetic diseases including major neurodegenerative disorders [2] and biological aging [3]. The enzyme of the inner mitochondrial membrane contains one FMN and eight ironsulfur clusters as redox prosthetic groups and is composed of .40 different subunits with a total molecular mass of almost 1,000 kDa. Fourteen ''central'' subunits are conserved among pro-and eukaryotes and form the functional core of complex I. Little is known about the function of the remaining .26 ''accessory'' subunits [1]. Mitochondrial complex I links the electron transfer from NADH to ubiquinone to the pumping of four protons from the matrix into the intermembrane space [4]. Recent progress in the X-ray structural analysis of prokaryotic [5] and eukaryotic [6] complex I confirmed that the redox reactions are confined entirely to the hydrophilic peripheral arm [7] of the L-shaped molecule [8] and take place at a remarkable distance from the membrane domain [9,10]. While this clearly implies that the proton pumping within the membrane arm of complex I is driven indirectly via long-range conformational coupling, the molecular mechanism and the number, identity, and localization of the pump-sites remain unclear.
The 180 Å long membrane arm or P module of complex I is subdivided into two subdomains of approximately equal size, called the proximal or P P and the distal or P D module [6].
Structural analysis revealed a very long helix running in parallel to the membrane that has been implicated as a mechanical transmission element exerting conformational coupling between these two subdomains [5,6]. Based on their homology to bacterial Mrp-type Na + /H + antiporters, the three largest subunits of the membrane arm ND2, ND4, and ND5 have been discussed as prime candidates for harbouring the proton pumps [11]. Indeed subunits ND4 and ND5 form the core of the P D module, suggesting that this part of complex I located well over 100 Å apart from the closest redox-center contributes to proton pumping. However, it cannot be concluded from homology alone that the Mrp-type subunits are functional in complex I. A recent mutagenesis study of subunit ND5 in Escherichia coli showed the functional importance of this protein for redox-linked proton pumping [12], but otherwise the contribution of the different parts of the membrane arm of complex I to proton pumping and the number of functional pumping modules has not been addressed experimentally. Here, we report that a subcomplex lacking specifically the P D module of the membrane arm is still capable of pumping protons at half the stoichiometry of holo-complex I.

Results and Discussion
The Deletion of Subunit NB8M Results in a Stable Subcomplex Lacking Specifically the P D Module When we deleted the gene for the accessory subunit NB8M from the genome of the strictly aerobic yeast Yarrowia lipolytica by homologous recombination, a defined subcomplex of complex I was found in mitochondria from knock-out strain nb8mD. Subunit NB8M has a molecular mass of 11.1 kDa [13] and is the homologue of subunit B18 of bovine heart complex I [14]. It has no predicted transmembrane helices and carries a twin CX 9 C motif characteristic for mitochondrial proteins imported via the so called MIA-pathway [15]. It has been proposed that it resides at the intermembrane space side of the P D module of complex I [16]. NADH:hexamineruthenium (HAR) oxidoreductase activity that monitors the amount of functional N-module of complex I of mitochondrial membranes prepared from deletion strain nb8mD was normal. Also no significant changes in the signatures of the iron-sulfur clusters were observed by electron paramagnetic resonance spectroscopy (unpublished data). In contrast, inhibitor sensitive dNADH:decylubiquinone (DBQ) oxidoreductase activity reflecting the physiological function of complex I was reduced to 30% of that of the parental strain (Table S1).
Further analysis by two-dimensional blue-native polyacrylamide electrophoresis (2D BN/SDS-PAGE) revealed that instead of fully assembled complex I strain nb8mD contained a defined subcomplex ( Figure 1) migrating slightly above complex V (F 1 F O -ATP synthase) at an apparent molecular mass of about 600 kDa [17]. The identity of this band as a subcomplex of complex I was confirmed by in-gel activity staining and Western blotting and a very small amount of fully assembled complex I was found to be still present in strain nb8mD ( Figure S1).
Since we had constructed the deletion of the gene for subunit NB8M in a strain expressing the 30-kDa subunit with a his-tag, we could purify the subcomplex using dodecyl-maltoside as a detergent by the same affinity-purification protocol established previously for holo-complex I [18]. The subcomplex eluted as a symmetrical peak from the final size exclusion column and as judged by its specific NADH:HAR oxidoreductase activity was of similar purity (unpublished data). To determine the subunit composition of subcomplex nb8mD we applied doubled sodium dodecyl sulfate polyacrylamide electrophoresis (dSDS-PAGE; [19]) and laser induced liquid bead ion desorption (LILBID) and electrospray ionisation (ESI) mass spectrometry (MS) as three complementary proteomic approaches that together allowed reliable identification of all subunits present. dSDS-PAGE is especially suited to separate highly hydrophobic proteins that migrate above the diagonal in these gels ( Figure 2). LILBID-MS at high laser intensity dissociates non-covalently bound subunits of enzyme complexes and generates a complete mass fingerprint in a single experiment ( Figure S2). Most subunits present in the subcomplex were clearly identified by all three methods. In cases where there were weak or overlapping spots in dSDS-PAGE, nonseparated peaks in LILBID-MS, or difficulties in identifying specific subunits by ESI-MS, the other methods still allowed to decide whether a subunit was present or not. We found that 14 subunits were missing in subcomplex nb8mD (Table 1) adding to a total molecular mass of 281,189 Da. Most notably the two mitochondrially coded Na + /H + antiporter homologous subunits ND4 and ND5 were absent. As these are the only central subunits previously assigned to the distal portion of the membrane arm [5], this suggested that subcomplex nb8mD specifically lacked the P D module.
To determine the total mass of purified subcomplex nb8mD, we next applied non-destructive LILBID-MS at ultrasoft conditions. As shown previously, under these conditions the peak maxima indicate the mass of multiprotein membrane protein complexes including detergents and phospholipids that are still bound to the ionized species [13]. Therefore, the lower mass onset of the peaks reflects the contribution of the protein to the total mass. In the case of subcomplex nb8mD this offset corresponded to ,30 kDa. Thus the predominant series of LILBID-peaks from the subcomplex indicated a protein mass of ,645 kDa ( Figure 3, red ticks). However, a series of minor peaks showed that, as previously observed for holo-complex I [13], the rather loosely attached subunit ST1 was dissociated during ionisation from a major fraction of the complexes even under ultrasoft conditions. Indeed a prominent peak at m/z 35 kDa matched the single-charged form of this protein ( Figure 3). The positions of the maxima of the minor peaks were consistent with the then expected mass of ,680 kDa for subcomplex nb8mD (Figure 3, green ticks). This fitted nicely

Author Summary
Mitochondria-the power plants of eukaryotic cellsproduce energy in the form of ATP. More than one-third of this energy production is driven by a gradient of protons across the mitochondrial membrane created by the pumping action of a very large enzyme called complex I. Defects in complex I are implicated in numerous pathological processes like neurodegeneration and biological aging. Recent X-ray structural analyses revealed that complex I is an L-shaped molecule with one arm integrated into the membrane and the other sticking into the aqueous interior of the mitochondrion; the chemical reactions of the enzyme take place in this hydrophilic arm, clearly separated from proton pumping that must occur somewhere in the membrane arm. To assign the pump function to structural domains, we created a stable subcomplex of complex I by deleting the gene encoding one of its small subunits in a yeast called Yarrowia lipolytica. This subcomplex lacked half of the membrane arm; it was still catalytically active but it pumped only half the number of protons as the full complex. This indicates that complex I has two functionally distinct pump modules operating in its membrane arm.   Table 1 with the proteomic data from which a residual mass of 682.5 kDa was calculated (Table 1).

3D Electron Microscopy Structure of Subcomplex nb8mD
To further investigate the structure of subcomplex nb8mD, we performed electron microscopic single particle analysis. Figure 4 shows the 3D reconstruction of the subcomplex at a resolution of 23 Å calculated from 10,586 tilt images from a random conical data set with a tilt angle of 55u. We compared the 3D model of subcomplex nb8mD to that of holo-complex I. The fit of the 3D model of the subcomplex into the outline of the 3D reconstruction of holo-complex I from [9] nicely showed that indeed the P Dmodule of the membrane arm was missing and that the remaining parts exhibited no major structural rearrangements ( Figure 4). With the distal end of the membrane arm, also the central and distal membrane arm protuberances were missing, which we had putatively assigned to be in close proximity to the proton translocating subunits ND4 and ND5.

Subcomplex nb8mD Pumps Protons at Half the Stoichiometry of the Complete Enzyme
After lipid-reactivation with asolectin [20] the subcomplex purified from strain nb8mD exhibited an inhibitor sensitive NADH:DBQ oxidoreductase activity of 2.7 mmol?min 21 ?mg 21 . At an activity of 5.8 mmol?min 21 ?mg 21 for purified holo-complex I and taking into account the lower molecular mass of the subcomplex, this indicated a reduction of the turnover number of the subcomplex to 34%. This fitted well to the values determined for mitochondrial membranes (Table S1) and indicated that subcomplex nb8mD had remained stable during the purification procedure. To test whether the inhibitor sensitive ubiquinone reductase activity of subcomplex nb8mD was still coupled to proton pumping, we reconstituted it into proteoliposomes and monitored the DpH dependent fluorescence quench of 9-amino-6-chloro-2-methoxyacridine (ACMA) as previously described [21]. As shown in Figure 5a, addition of NADH and the ubiquinone analogue DBQ to subcomplex nb8mD proteoliposomes resulted in the formation of a pH gradient. However, reflecting the lower ubiquinone reductase activity of only 38%, the rate by which it was reached was significantly lower than for holo-complex I. The gradient was maintained at a steady-state plateau over several minutes and was abolished by the addition of the specific inhibitor 2-n-decyl-4quinazolinyl-amine (DQA) (Figure 5a) or the uncoupler carbonylcyanide-p-trifluoro-methoxy-phenylhydrazone (FCCP) ( Figure S3). This indicated that even the purified subcomplex nb8mD had retained proton pumping activity, although at a reduced rate. To determine whether the reduced rate of proton pumping by subcomplex nb8mD was just due to its reduced electron transfer activity or also reflected a lower pumping stoichiometry, we went on to determine the pumping efficiencies of reconstituted parental complex I and subcomplex nb8mD. When the ACMA fluorescence quench reaches a steady-state plateau, the rates of proton pumping and backflow are equivalent [21]. This fluorescence level should thus reflect the pump rate, provided that the proton leak is not changed. Indeed we found that the rates of proton backflow that can be fitted by a first order exponential function after the addition of DQA ( Figure S5) were very similar for proteoliposomes with parental enzyme and with subcomplex nb8mD (Figure 5b). Also the maximal fluorescence %FL max was essentially identical for the enzymes from both strains (Figure 5b). Importantly, the decrease in fluorescence turned out to be proportional to electron transfer activity up to the point where a complete quench of the ACMA fluorescence was observed (Figure 5c; Figure S4). We concluded that the slope of this plot can be used as a measure for the pumping stoichiometry of complex I. The ACMA fluorescence quench signal is difficult to calibrate in absolute terms, however the pumping stoichiometry for Y. lipolytica complex I was previously determined at 4 H + /2 e 2 [22]. Thus the observation that for subcomplex nb8mD proteoliposomes the slope of the %FL max over activity plot (Figure 5c) was about half of that for parental complex I indicated that the absence of the P D module resulted in a reduced pumping stoichiometry of 2 H + /2e 2 .

Conclusion
We concluded that we have functionally dissected two pumping modules of complex I and that these modules correspond to the P P and P D domains of the membrane arm previously identified by X- Figure 3. LILBID mass spectra of purified subcomplex nb8mD. The predominant series of LILBID-peaks from the subcomplex indicated a protein mass of ,645 kDa (red ticks). A series of minor peaks (green ticks) indicate that the rather loosely attached subunit ST1 was dissociated during ionisation from the major fraction even under ultrasoft conditions. doi:10.1371/journal.pbio.1001128.g003 ray structural analysis ( Figure 6; [6]). The fact that the pump in the P P module was still functional in subcomplex nb8mD provides experimental support for the proposal that the P D domain of complex I indeed harbours a proton pump connected in series to the primary pump in the P P domain. This is consistent with the idea that the long helical transmission element bridging the two subdomains is involved in operating the proton pumps of complex I [5,6].

Construction of Deletion Strain nb8mD of Yarrowia lipolytica
The Y. lipolytica deletion strain nb8mD (NUGM-Htg2, ndh2i, lys11 2 , leu2 2 , ura3 2 , nb8m::URA3, MatB), in which a 976 bp fragment corresponding to the complete NB8M gene including the intron and up-and downstream sequences was replaced by the Y. lipolytica URA3 gene (1.6 kb) oriented in opposite direction to the original NB8M open reading frame, was constructed by the double homologous recombination strategy as previously published [23].
Purification of Complex I from Mitochondrial Membranes of Y. lipolytica Y. lipolytica strains PIPO and strain nb8mD were grown in modified YPD medium (2.5% glucose, 2% bactopeptone, 1% yeast extract). Both strains contain a chromosomal copy of the modified NUGM gene, encoding a C-terminally his-tagged version of the 30-kDa subunit of complex I [23]. Mitochondrial  The halftime t for the proton backflow after addition of DQA (left panel) was determined (see Figure S5) and, like the maximal fluorescence FL max (right panel), was found to be essentially the same for complete complex I and subcomplex nb8mD. (c) The fluorescence at the steady state quench plateau of 5 (filled symbols, holo-complex I) and 3 (open symbols, subcomplex nb8mD) independent reconstitution experiments were plotted against the corresponding DQA-sensitive NADH:DBQ oxidoreductase activities ( Figure S4). The uninhibited electron transport rates in the presence of membranes were prepared following the protocol of [24] with the modification detailed in [18]. Purification of n-dodecyl-ß-Dmaltoside solubilized complex I was achieved by Ni-affinity chromatography, followed by gel filtration as detailed in [18]. The yield of each preparation was determined by measuring the NADH:HAR oxidoreductase activity while the specific NADH: DBQ oxidoreductase activity was measured after lipid-activation as described [20].

Electrophoretic and Associated Techniques
Mitochondrial membranes from Y. lipolytica were solubilized by 1.5% laurylmaltoside and separated by two-dimensional bluenative electrophoresis (2D BN-PAGE) as described [17]. For the analysis of the subunit composition of the purified nb8mD subcomplex, dSDS-PAGE [19]) was applied. In-gel catalytic activity stain was performed with solubilized membranes after BN-PAGE as detailed previously [17]. For Western blot analysis of mitochondrial membranes, proteins were transferred to polyvinylidendifluoride membranes and incubated with a monoclonal antibody against the 49-kDa subunit [10]. The secondary antibody was a peroxidase conjugated rabbit antimouse IgG and the signal was detected by enhanced chemiluminescence.

Mass Spectrometry
Laser Induced Liquid Bead Ion Desorption Mass Spectrometry (LILBID-MS) was done as described previously [25]. Proteins from gel-spots were identified by nano-ESI-LC-MS/MS on a Thermo Scientific LTQ Orbitrap XL mass spectrometer essentially as described in [26] except that database searches were done against an in-house compiled version of the Y. lipolytica protein database of the Genolevures consortium [27] containing annotations of all known complex I subunits.

Electron Microscopy
For electron microscopy the subcomplex was prepared on holey carbon coated grids and embedded in stain (NanoW, Nanoprobes, Yaphank, NY), following the deep staining methods described in [28,29]. For single particle 3D reconstruction using the method of Random Conical Tilting [30], 100 tilt pairs were recorded at a nominal magnification of 67 k6, and a tilt angle of 55u. 68 tilt pairs were used for the reconstruction. The images were scanned with an Intergraph SCAI flatbed scanner (Z/I Imaging Corporation, Huntsville, AL) and reduced to a calibrated pixel size of 3.136 Å . Particle pairs were picked from the 0u-and tiltmicrograph. The microscope contrast transfer function was determined according to Radermacher et al. [31], and all images uncoupler for these preparations were 6.560.7 mmol min 21 mg 21 for the parental strain and 2.560.3 mmol min 21 mg 21 for strain nb8mD. For parental complex I data for activities greater than ,1.3 mmol min 21 mg 21 were omitted since at these high rates ACMA fluorescence was quenched completely. Measurements performed in parallel are marked with symbols of the same shape (e.g., m, n). The ratio between the slopes for holocomplex I (27567%?l?min mmol 21 ) and subcomplex nb8mD (23663%?l?min mmol 21 ) indicated a proton pumping stoichiometry of 2.1 H + /2 e 2 for the subcomplex. When analyses with two additional preparations of subcomplex nb8mD were included (unpublished data), the overall average stoichiometry for a total of five independent pairs of experiments was calculated as 1.860.3 H + /2 e 2 . doi:10.1371/journal.pbio.1001128.g005 Figure 6. Schematic model of complex I and its pumping modules. Our data are consistent with the dissection of complex I into four functional modules [6] and allow the assignment of subunits and proton pumps to the proximal (P P , green) and distal (P D , cyan) domains of the membrane arm. The FMN containing N-module (light yellow), the Q-module harbouring the ubiquinone binding site (dark yellow), the chain of seven iron-sulfur clusters (orange spheres), and the extended transmission element (magenta) are indicated. doi:10.1371/journal.pbio.1001128.g006 were corrected by smooth phase flipping. A total of 10,897 image pairs were processed, using a sequence of reference free alignment followed by several rounds of correspondence analysis, followed by classification and multi-reference alignment. The last step in the processing of the 0u images was a classification after correspondence analysis using Diday's method of moving centers and hierarchical ascendant classification [32]. The processing yielded 10 classes of subcomplex volumes that mostly differed in their orientation to the supporting carbon and could be combined into a single reconstruction. The final volume was calculated from 10,586 images. The resolution was determined using the Fourier Shell Correlation with a cutoff of 0.3 [33]. The subcomplex structure was superimposed on the holocomplex from earlier studies [8]. The handedness of all volume representations of complex I has been adjusted to the handedness presented in [5].

Reconstitution of Complex I into Proteoliposomes
Parental complex I and the subcomplex purified from strain nb8mD were reconstituted into proteoliposomes, generally following the protocol described in [21] with the modifications detailed in Dröse et al. [34]. 10 mg/ml asolectin solubilized in 16 mg/ml octylglucoside (resulting in ''total solubilization'' of lipids) were mixed at a protein-to-lipid ratio of 1:50 (w/w) with purified complex I (0.8-1.0 mg per experimental set) in 20 mM K + /Mops pH 7.2, 80 mM KCl and the detergents were removed by BioBeads. Note that due to the reduced molecular mass of the subcomplex, the corresponding molar protein-to-lipid ratio was somewhat lower for subcomplex nb8mD (i.e., less phospholipids per complex I). Usually, subcomplex nb8mD and parental complex I were reconstituted and analyzed in parallel on the same day.

Proton Transport and Activity Determination of the Reconstituted Enzyme
Proteoliposomes containing ,30 mg of complex I were added to 2 ml buffer (20 mM K + /Mops pH 7.2, 80 mM KCl, 0.5 mM valinomycin) in a stirred cuvette. H + -translocation was monitored as fluorescence change of ACMA that was added to a final concentration of 0.5 mM. Measurements were performed in a Shimadzu RF-5001 fluorescence spectrophotometer at 30uC (l ex = 430 nm, l em = 475 nm, band pass 5 nm, integration time 0.2 s). To gradually reduce the catalytic activity of the applied reconstituted complex I, increasing concentrations of the ubiquinone binding site inhibitor DQA were added prior to the start of the experiment. Each measurement of one dataset followed a strict time protocol. After starting the recording, ACMA was added after 10 s to determine the somewhat variable background fluorescence. At 30 s, 100 mM DBQ were added which caused significant fluorescence quenching due to spectral overlap [21]. This interference stabilized after ,45 s, and at 75 s 100 mM NADH were added to start complex I driven proton-pumping. At 285 s, this activity was stopped completely by adding DQA to a final concentration of 10 mM. The measurement was terminated after 6 min to allow for backflow analysis (see below). To determine the catalytic activity of the reconstituted complex I, NADH-oxidation rates were recorded using a Shimadzu Multi Spec-1501 diode array spectrophotometer (e 340-400 nm = 6.1 mM 21 cm 21 ) under the same experimental conditions and time regime except that the total volume was reduced to 1 ml.

Data Analysis and Determination of H + /e 2 Stoichiometry
Relative proton pumping rates were determined indirectly from the fluorescence level of the ACMA quench assay assuming that during this steady-state the pump and the leak rates should be equivalent. This approach was found to be valid since the average haftimes of proton backflow were very similar (Figure 5b). We had also considered determining the rate of proton pumping in a more direct way from the initial ACMA quench rate. However, this turned out to be not feasible, since complex I requires a few turnovers for its transition to full activity from its deactive state [21].
To minimize variations from the background fluorescence, the mean value of 48-52 data points between start of the measurement and the addition of ACMA was subtracted. As the starting (maximal) fluorescence value, the average of 11 data points before the addition of NADH (and after settling of the DBQ quench effect) was then calculated and set to 100%. Obvious ''outliers'' were excluded from these calculations. The fluorescence [%FL max ] reached at the steady state maximal quench plateau representing the equilibration between the complex I catalyzed H + -pumping and the DpH-enforced H + -leak [21] was determined as the mean value of at least 60 data points and plotted against the NADH:DBQ oxidoreductase activity of the reconstituted enzyme measured in parallel under the same conditions (see above). For complex I from the parental strain data above an activity of ,1.3 mmol min 21 mg 21 could not be plotted, since ACMA fluorescence was quenched completely. The activities were corrected for the small residual activity after the final addition of a saturating amount of DQA. Finally the data were fitted by linear regression analysis. Figure S1 Blue native electrophoresis, in-gel activity staining, and Western blot analysis of mitochondrial membranes from strain nb8mD. Mitochondrial membranes of the parental strain and the deletion strain nb8mD were solubilized with dodecylmaltoside (1.5 g/g) and separated by BN-PAGE. Two lanes of the BN-gel were subsequently used for the NADH dehydrogenase activity staining assay and two lanes were subjected to Western blot analysis with a monoclonal antibody directed against the 49-kDa subunit. The positions of respiratory chain complexes (I, V, III, and IV) and the subcomplex nb8mD (I*) are indicated. #, unidentified band with NADH dehydrogenase activity that has been previously observed in Y. lipolytica membranes 17 . (TIF) Figure S2 LILBID mass fingerprint spectra of complex I and subcomplex nb8mD. The LILBID anion mass spectra reveal individual subunits of purified holo-complex I and subcomplex nb8mD. In the range of 6-8 m/z the cation spectra are also shown in red. The 42 known subunits are numbered as in Table 1 and the masses are indicated with vertical lines. Subunits that were absent in subcomplex nb8mD are marked by ''X''. Peaks corresponding to doubly charged subunits are also assigned. X, unidentified contaminating protein; *, unidentified possible 43 rd subunit. (TIF) Figure S3 Proton pumping of reconstituted complex I and subcomplex nb8mD. The proton pumping activity was monitored by ACMA quenching. Additions of 0.5 mM ACMA, substrates (60 mM DBQ, 100 mM NADH), inhibitor (10 mM DQA), and uncoupler (1 mM FCCP) are indicated. Note that these control measurements were performed at a lower DBQ concentration resulting in somewhat lower electron transfer activities and a reduced extent of fluorescence quenching for subcomplex nb8mD. Left panels, holo-complex I; right panels, subcomplex nb8mD. (TIF) Figure S4 Quantification of proton pumping efficiencies by inhibitor titration. Representative dataset that was included in the analysis shown in Figure 5c. The figure shows the quality of the original data. Measurements were started in the presence of the indicated DQA concentrations (0-10 mM) to gradually reduce the activities of the reconstituted holo-complex I and subcomplex nb8mD. Note that somewhat different DQA concentrations had to be added to the holo-complex I (upper panel) and subcomplex nb8mD (lower panel) to achieve distribution of quench levels over the entire range. After starting the experiments 0.5 mM ACMA, 100 mM DBQ, 100 mM NADH, and finally DQA to a total concentration of 10 mM (e.g., if 1 mM was present at the start, 9 mM had to be applied) were subsequently added. The NADH:DBQ oxidoreductase activity was monitored in parallel under identical experimental conditions (not shown) and only data within the linear range of the activity/quench dependence were included in the analysis. (TIF) Figure S5 Determination of the H + -leak of proteoliposomes with reconstituted holocomplex I and subcomplex nb8mD. After addition of the inhibitor DQA to a final concentration of 10 mM, the ACMA fluorescence increased exponentially due to the passive backflow of protons ( Figure S4). To obtain the halftime t of the proton leak, the backflow curves were fitted using the Origin 6.0 software package by the function:

Supporting Information
Only data above ,70% FL max were included in the fit to remove the mixing artefact that also prevented fitting signals returning from a smaller maximal quench amplitude. Representative analysis of two measurements from Figure S4 is shown. The half times (t) obtained for the shown examples were t = 15.0 s for holocomplex I (R 2 = 0.992) and t = 13.6 s for subcomplex nb8mD (R 2 = 0.994). (TIF)