Research article

A scaffold of accessory subunits links the peripheral arm and the distal proton-pumping module of mitochondrial complex I

Heike Angerer, Klaus Zwicker, Zibiernisha Wumaier, Lucie Sokolova, Heinrich Heide, Mirco Steger, Silke Kaiser, Esther Nübel, Bernhard Brutschy, Michael Radermacher, Ulrich Brandt, Volker Zickermann


Mitochondrial NADH:ubiquinone oxidoreductase (complex I) is a very large membrane protein complex with a central function in energy metabolism. Complex I from the aerobic yeast Yarrowia lipolytica comprises 14 central subunits that harbour the bioenergetic core functions and at least 28 accessory subunits. Despite progress in structure determination, the position of individual accessory subunits in the enzyme complex remains largely unknown. Proteomic analysis of subcomplex Iδ revealed that it lacked eleven subunits, including the central subunits ND1 and ND3 forming the interface between the peripheral and the membrane arm in bacterial complex I. This unexpected observation provided insight into the structural organization of the connection between the two major parts of mitochondrial complex I. Combining recent structural information, biochemical evidence on the assignment of individual subunits to the subdomains of complex I and sequence-based predictions for the targeting of subunits to different mitochondrial compartments, we derived a model for the arrangement of the subunits in the membrane arm of mitochondrial complex I.

  • membrane protein
  • mitochondria
  • NADH dehydrogenase
  • protein import
  • respiratory chain
  • subcomplex


Proton-pumping NADH:ubiquinone oxidoreductase (complex I, EC is the largest membrane protein complex in the respiratory chain of many bacteria and mitochondria [13]. Coupling the two-electron transfer reaction from NADH to ubiquinone to the vectorial translocation of four protons, mitochondrial complex I provides a major portion of the protonmotive force that drives ATP synthesis. The enzyme is a source of ROS (reactive oxygen species) and has been implicated in the pathogenesis of a number of human neurodegenerative diseases [4].

Mitochondrial complex I is composed of more than 40 different subunits subdivided into central and accessory subunits. The fourteen central subunits carry out the bioenergetic function of complex I and are conserved throughout species, from bacteria to mammals. They can be assigned to three functional modules [3] that harbour the bioenergetic core functions: NADH oxidation (N-module), ubiquinone reduction (Q-module) and proton translocation (P-module). The substrate-binding sites are connected by a chain of iron–sulfur clusters. The function of the accessory subunits of mitochondrial complex I is largely unknown. Individual subunits were suggested to play a role in assembly, regulation and stabilization of the complex [3,57]. In complex I from the strictly aerobic yeast Yarrowia lipolytica 28 accessory subunits with a total mass of 421.0 kDa have so far been identified corresponding to 43.7% of the total mass of 963.7 kDa [8] (E. Nübel, S. Kaiser, B. Brutschy, S. Kerscher, B. Meyer, M. Karas, L. Sokolova, K. Zwicker, H. Angerer and U. Brandt, unpublished work).

Electron microscopy has revealed that complex I from mammalian and fungal mitochondria, like that from bacteria, is L-shaped with a hydrophilic peripheral arm and a hydrophobic membrane arm [9]. Recently, X-ray crystallographic analysis provided insight into the overall architecture of prokaryotic [10] and eukaryotic [11] complex I at an intermediate resolution. The X-ray structure of a bacterial peripheral arm fragment had been solved earlier at up to 3.1 Å (1 Å=0.1 nm) resolution [12]. It comprises the N- and the Q-module that together perform all of the redox chemistry. As had been suggested previously [13,14], the terminal iron–sulfur cluster N2, which is the immediate electron donor for ubiquinone, was found at a remarkable distance from the membrane domain [10,11]. The membrane arm or P-module is subdivided into a proximal (PP) and a distal (PD) domain [11]. Several lines of evidence suggest that both parts of the P-module contribute to proton translocation. An elongated, presumably α-helical, element that is laterally associated with the membrane arm was discovered that connects the proximal and distal domains of the P-module and has been suggested to play a key role in energy transmission within the membrane arm [10,11]. Overall, the relative arrangement of functional modules in complex I indicates a spatial separation of redox chemistry and proton translocation, essentially excluding direct mechanisms of energy transfer. This rather suggests that complex I operates by a conformational coupling mechanism.

Complex I from many bacteria is too unstable to be prepared in an intact form. This has only been achieved with complexes from Escherichia coli [15], Paracoccus denitrificans [16] and some thermophilic bacteria [10,17]. Complex I purified from these species still tends to disintegrate very easily into the peripheral arm comprising the Q- and N-modules and the membrane integral P-module. Also the compositions of several subcomplexes that have been prepared biochemically from bovine complex I by applying mild chaotropic conditions [1821] reflect the modular architecture of complex I, although there are some notable discrepancies. Subcomplex Iα comprises the Q- and N-module of the peripheral arm and a sector of the PP-module of the membrane arm, whereas subcomplex Iβ essentially corresponds to the distal membrane arm module PD. Further disassembly of subcomplex Iα leads to formation of subcomplex Iλ that is more hydrophilic, but still contains two accessory subunits with predicted transmembrane helices.

A subcomplex similar to bovine subcomplex Iλ has been generated previously by disintegration of complex I from the strictly aerobic yeast Y. lipolytica [22]. In the present study we extended these previous studies and prepared a new type of subcomplex, still comprising the Q-, N- and PD-module of complex I, but lacking a significant portion of the PP-module. The results of the present study suggest that a rather stable connection must exist in mitochondrial complex I between the peripheral arm and the distal membrane arm domain that involves specific accessory subunits. Combining available structural information, biochemical data and sequence analysis, we have derived a topological model for the arrangement of the central and accessory subunits in the membrane arm of complex I.


Preparation of subcomplex Iδ from Y. lipolytica, catalytic activity and EPR spectroscopy

Complex I from Y. lipolytica strain PIPO (histidine-tagged 30 kDa subunit) was prepared as described previously [23]. A strain variant of PIPO expressing subunit NUJM-Strep-tagII (histidine-tagged 30 kDa subunit and StrepII-tagged NUJM) was generated as described elsewhere (E. Nübel, S. Kaiser, B. Brutschy, S. Kerscher, B. Meyer, M. Karas, L. Sokolova, K. Zwicker, H. Angerer and U. Brandt, unpublished work). Subcomplex Iδ was prepared by incubation of 2 mg of purified complex I in 5 ml of incubation buffer [20 mM sodium phosphate, 400 mM NaCl, 20 mM imidazole, 1 mM MgCl2 and 1% LDAO (lauryl dimethylamine oxide), pH 7.2] on ice for 2 h. The sample was loaded on to a Ni-NTA (Ni2+-nitrilotriacetate) column and washed with 12 column volumes of LDAO washing buffer [20 mM sodium phosphate, 400 mM NaCl, 20 mM imidazole and 1% LDAO (pH 7.2)]. The detergent was exchanged to DDM (β-dodecylmaltoside) with 9 column volumes of DDM washing buffer [20 mM sodium phosphate, 400 mM NaCl, 20 mM imidazole and 0.025% DDM (pH 7.2)]. Subcomplex Iδ was eluted with 7 column volumes of DDM elution buffer [20 mM sodium phosphate, 400 mM NaCl, 140 mM imidazole and 0.025% DDM (pH 7.2)]. The concentrated sample (Amicon concentrator 50 kDa cut-off, Millipore) was applied to a gel filtration column (TSK 4000 column 7.8 mm×30 cm) in gel filtration buffer [20 mM sodium Mops, 100 mM NaCl, 1 mM EDTA and 0.025% DDM (pH 7.2)] and concentrated to a final protein concentration of 15 mg/ml. The protein concentration was determined according to a modified Lowry protocol using BSA as the protein standard.

NADH:HAR (hexammineruthenium), NADH:DBQ (decylubiquinone) and NADH:Q1 (ubiquinone-1) oxidoreductase activity was measured essentially as described previously [24]. X-band EPR spectra were obtained essentially as described previously [25].


BN (blue-native)-PAGE of Y. lipolytica mitochondria was performed using DDM (1 g/g of protein) as the detergent as described in [26]. Molecular masses of complex I, complex V and complex III dimers from Y. lipolytica mitochondrial membranes were assigned as described previously [27].

Subunits of protein complexes were separated by dSDS/PAGE (doubled SDS/PAGE) which was carried out as described previously [28] with a minor change that protein samples were reduced using 130 mM DTT (dithiothreitol). Briefly, lanes from one-dimensional SDS/PAGE gels (10% polyacrylamide, 6 M urea) were incubated in acidic solution containing 100 mM Tris/HCl and 150 mM HCl (pH<2) for 30 min and analysed by second-dimension SDS/PAGE using 16% polyacrylamide.

ELISA and identification of subunit NUPM by Western blot analysis and LC-MS/MS (liquid-chromatography tandem MS)

The standard ELISA using alkaline conditions for binding of complex I or subcomplex Iδ to polystyrene microtitre plates was carried out as described previously [13]. Primary monoclonal antibodies against several subunits were used for subunit identification [anti-(49 kDa), anti-PSST, anti-NUPM, two different anti-NESM and anti-NIAM antibodies] [13] (V. Zickermann, unpublished work).

We used a three-dimensional electrophoresis technique [BNE (blue-native electrophoresis) followed by dSDS/PAGE] to separate the protein subunits of complex I from Y. lipolytica and to characterize the monoclonal antibody 31A8. To prepare complex I for BN-PAGE, 0.3 g/g of DDM was added to 1 mg of chromatographically purified complex I [26]. The repurified complex I band was excised from the BN gel and the subunits were separated using dSDS/PAGE. Following electroblotting of the dSDS/PAGE gels on PVDF membranes [26], monoclonal antibody 31A8 was used for immunological detection. A gel piece from a duplicate Coomassie-Blue-stained gel was cut out for LC-MS/MS analysis revealing binding of monoclonal antibody 31A8 to subunit NUPM.


LILBID-MS (laser-induced liquid bead ion desorption MS) was performed as described previously [29]. Protein bands were analyzed by ESI-MS (electrospray ionization MS) after dSDS/PAGE or BN-PAGE. Gel spots were excised and in-gel digested with trypsin, essentially as described by Collins et al. [30]. In addition complete purified complex I was digested in solution (see Supplementary Figure S1 at Briefly, approximately 70 μg of complex I was reduced with DTT (5 mM) and alkylated with iodoacetamide (15 mM) in 50 mM ammonium bicarbonate buffer, and subsequently digested overnight with 1 μg of trypsin or chymotrypsin in 0.01% ProteaseMAX™ surfactant (Promega). The peptides were extracted from the digest using cation exchange. The dried peptides were redissolved in 5% acetonitrile/0.5% formic acid for subsequent analysis by ESI-LC-MS/MS.

The extracted peptides were separated on a nano-HPLC column (PicoTip™, New Objective), packed with reversed-phase silica (Hypersil Gold C18, 3 μm, Thermo Scientific) and analysed by a LTQ Orbitrap XL mass spectrometer as described previously [31], except that gradients were adapted to the complexity of the sample.

Electron microscopy

Complex I decorated with the anti-NUPM monoclonal antibody 31A8, was analysed by electron microscopy and image processing. The sample was first diluted to 0.48 mg/ml [in 25 mM sodium Mops (pH 7.2), 100 mM NaCl and 0.025% DDM] and mixed with a 2-fold molar excess of antibody. After dilution to 0.03 mg/ml, the mixture was applied to carbon-coated 400 mesh copper grids and deep-stain-embedded with NanoW (Nanoprobes) [32,33]. The micrographs were scanned on a SCAI flatbed scanner (Z/I Imaging) with a calibrated pixel size of 4.02 Å on the specimen scale. In total, 755 single particle images were boxed out of the micrographs, using as the only criterion that the particles showed the approximate size of complex I and were separated from their neighbour particles, without differentiation between labelled and unlabelled protein. The boxed particles were first centred then processed by iterations of rotational translational alignments [34], correspondence analysis [35], classification [36] and multireference alignments, classification being the final step.

Sequence analysis

Prediction of N-terminal MTSs (mitochondrial-targeting sequences) was done using the MITOPROT algorithm [37]. Alignments were done with default parameters using the program ClustalW2 [38]. For prediction of transmembrane segments we used the programs TMHMM and hmmtop [39,40].


Subcomplex Iδ lacks part of the proximal P-module

Exposure of purified complex I to 1% of the zwitterionic detergent LDAO reproducibly resulted in formation of a defined subcomplex as determined by BN-PAGE (Figure 1A) and size-exclusion chromatography (Figure 1B). Expanding the established nomenclature of biochemically prepared subcomplexes [18,20] we termed the new subcomplex Iδ. We noted that in BN-PAGE the subcomplex migrated as a rather diffuse band and exhibited a pronounced tendency to form oligomers. The subunit composition of monomer and oligomers was found to be essentially identical (see below and Supplementary Figure S2 at A total mass of ~850 kDa at peak maximum can be deduced from a series of prominent peaks in non-destructive LILBID-MS under ultrasoft conditions (Figure 1C). As shown previously, detergents and phospholipids are still bound to the ionized complexes under these conditions and the mass of the protein portion is 40–50 kDa lower [8], corresponding to the lower mass onset of the peak at approximately 808 kDa (blue, Figure 1C).

Figure 1 Exposure of native complex I to LDAO generates subcomplex Iδ

(A) BN-PAGE of solubilized mitochondrial membranes from Y. lipolytica (left-hand lane) and subcomplex Iδ (right-hand lane). The migration behaviour of subcomplex Iδ (Iδ) indicated a lower mass compared with native complex I (CI). Note that complex I subunit ST1 is detached under BN-PAGE conditions resulting in a total mass of ~930 kDa for the control. Complex V monomer (CV ~580 kDa), complex III dimer (CIII2, ~420 kDa). The bands in the high molecular mass range in the right-hand lane were identified as oligomers of subcomplex Iδ (n Iδ). (B) Subcomplex Iδ (light grey trace) eluted as a symmetrical peak with slightly increased retention volume compared with native complex I (black trace) in size-exclusion chromatography. (C) Non-destructive LILBID-MS analysis of subcomplex Iδ at low laser intensity. A prominent peak series indicated m/z values (Iδ+ND2, blue) of ~ 850 kDa including detergent and lipids. As for native complex I, the protein mass was found to be 40–50 kDa offset from the peak maximum and using the m/z values of the left peak edges the mass of subcomplex Iδ+ND2 was best approximated to be ~808 kDa (blue). A second peak series corresponding to ~755 kDa was assigned to subcomplex Iδ without ND2 (Iδ, red). A third peak series of ~413 kDa can be assigned to the peripheral arm (Q/N-module, green) which is generated by disintegration of laser-irradiation-sensitive subcomplex Iδ (red) (see the text for details).

Consistent with a reduction in molecular mass the specific NADH:HAR oxidoreductase activity of the subcomplex Iδ was 90±4 μmol·min−1·mg−1 as compared with 65±2 μmol·min−1·mg−1 for purified holo-complex I. The non-physiological assay is indicative for a functional N-module, as it solely depends on the presence of an intact NADH oxidation site and FMN [41]. In contrast, inhibitor-sensitive NADH:ubiquinone oxidoreductase activity using ubiquinone analogues such as DBQ or Q1 was completely abolished in the subcomplex and could not be recovered by addition of lipids (see Supplementary Table S1 at This suggested that either the Q-module itself or the access path to its Q-binding pocket was affected. EPR spectroscopy indicated that a largely intact peripheral arm was present in the subcomplex (Figure 2). The spectral signatures of iron–sulfur clusters N1, N3 and N4 appeared essentially unchanged. However, iron–sulfur cluster N2 showed a slight distortion in peak shape of the gz signal and a shift of the gxy signal, suggesting a somewhat modified environment for this redox centre as observed previously for bovine subcomplex Iλ [19].

Figure 2 EPR spectra of complex I and subcomplex Iδ

Spectra were recorded using the following parameters: microwave frequency 9.47 GHz, modulation amplitude 0.64 mT, modulation frequency 100 kHz. Complex I or subcomplex Iδ (15 mg/ml) were reduced with NADH (2 mM) and the spectra were recorded at 12 K. *Organic radical signal, probably representing an impurity.

Next we asked which of the central subunits had been retained in the subcomplex. Determining the subunit composition of a membrane integral multiprotein complex is a difficult task and we therefore applied three complementary techniques, dSDS/PAGE, LILBID-MS and ESI-MS of individual subunits and, if necessary, additional immunological and molecular genetics approaches. Please note that we use the bovine nomenclature for central subunits throughout (Table 1). Figure 3 shows silver-stained dSDS/PAGE and Figure 4 shows LILBID subunit spectra of complex I and subcomplex Iδ. LILBID-MS at high laser intensity dissociates non-covalently bound subunits of enzyme complexes and generates a complete mass fingerprint in a single experiment [8]. The analysis of these data and the results from ESI-MS analysis of subcomplex Iδ monomer and oligomers excised from BN gels are summarized in Supplementary Figure S2. Using all three techniques, the seven central subunits of the peripheral arm and subunits ND4 and ND5 of the membrane were detected in the subcomplex, whereas the hydrophobic subunits ND1 and ND3 were clearly absent (Table 1). The presence of the 49 kDa and PSST subunit was further confirmed by ELISA (see Supplementary Figure S3 at In intact complex I, subunits ND6 and ND4L could only be detected reliably by LILBID-MS. However, for ND6 (molecular mass of 20.76 kDa) this required shifting the almost identical mass of accessory subunit NUJM (molecular mass of 20.83 kDa) using a StrepII tag. This separated the two peaks in the spectrum of the tagged complex I, allowed unambiguous identification of both subunits and indicated that ND6 was not missing in the subcomplex (Figure 4, inset). In contrast, the LILBID spectra showed that ND4L had been removed. The spot for subunit ND2 partially overlapped with that of subunit ND4 in dSDS/PAGE, but it appeared that this subunit was at least markedly reduced in the subcomplex (Figure 3). This was confirmed by LILBID-MS showing that subunit ND2 had been removed partially by LDAO treatment (Figure 4, peak 9).

Figure 3 dSDS/PAGE of purified complex I (A) and subcomplex Iδ (B)

Subunits missing in subcomplex Iδ are highlighted by a circle. In the subcomplex an extra spot was visible near the 49 kDa subunit (49-kDaΔN). It was identified by ESI-MS as a degradation product of the 49 kDa subunit lacking a 3 kDa fragment at the N-terminus. Insets 1 and 2 show the section of a dSDS gel around subunit NUJM of complex I or subcomplex Iδ prepared from a strain carrying a StrepII-tagged version of this subunit. Inset 3 shows a section of a Coomassie-Blue-stained BN-dSDS/PAGE gel of complex I (left-hand panel) and the corresponding Western blot (right-hand panel). Subunit NUPM was recognized by monoclonal antibody (mab) 31A8 (anti-NUPM). Note that the assignment for some of the accessory subunits was updated and is therefore different from [22].

Figure 4 LILBID mass fingerprint spectra of complex I and subcomplex Iδ

The peaks are numbered according to Table 1. Inset, detail from mass spectra of a complex I variant carrying a StrepII tag on subunit NUJM (black, broken line) and the corresponding subcomplex Iδ (grey, broken line). The mass shift permitted identification of subunits NUJM (18) and ND6 (13) in the subcomplex (*). Taking into account new evidence on C-terminal peptides and prediction of targeting sequences (Table 1) we reassigned the peaks for subunits NUMM and NI9M (compare with [8]). The peak at 7.7 kDa previously assigned to NI9M remains unassigned. To account for these changes and the newly discovered subunits NUUM and NEBM, the assignment of peaks 27–29 and peaks 37–42 was updated to match decreasing molecular masses of the corresponding subunits. The spectrum of subcomplex Iδ exhibited a few additional peaks that correspond either to higher charged species of the 75 kDa (1), 51 kDa (2) and 49 kDa (3) subunits, or to the truncated version of the 49 kDa subunit (3ΔN, see Figure 3).

View this table:
Table 1 Subunits of complex I from Y. lipolytica highlighting proteins missing in subcomplex Iδ (in bold)

mtDNA, mitochondrial DNA.

With subunits ND1, ND3 and ND4L, a significant portion of the proximal part of the PP-module was completely missing in the subcomplex (Table 1). As subunit ND1 is the central subunit that, according to the recently published X-ray structural analysis of bacterial complex I [10] forms the major connection between the membrane arm and the peripheral arm, we concluded that in mitochondrial complex I another connection between the two parts must exist. Electron microscopic analysis of subcomplex Iδ showed particles of the expected size with a three-domain substructure. Unfortunately, the very high flexibility of the particles limited the three-dimensional structural analysis of the complex to very low resolutions of only ~60 Å (results not shown), precluding a more detailed interpretation.

Subcomplex Iδ lacks seven accessory subunits

To determine which of the accessory subunits were missing in the subcomplex we again applied dSDS/PAGE, LILBID-MS, and ESI-MS (Figures 3 and 4, and see Supplementary Figure S2). For most proteins the results obtained with the three methods complemented each other; e.g. subunits with similar masses that resulted in overlapping peaks in LILBID-MS, were well-enough separated in dSDS/PAGE. In the case of subunit NUJM we obtained an unambiguous result using a complex I variant where this subunit was modified with a StrepII tag. As discussed above, subunit NUJM has a molecular mass almost identical with that of the central subunit ND6, resulting in overlapping peaks in LILBID-MS. In dSDS/PAGE the spot for subunit NUJM overlapped with that of subunit PSST making it again impossible to decide whether it was present in subcomplex Iδ. The spot for StrepII-tagged subunit NUJM was shifted away from that of subunit PSST, but now partly overlapped with the spot corresponding to subunit NUPM (Figure 3, inset 1). However, as subunit NUPM was detached by LDAO treatment, as clearly evident from LILBID-MS, the remaining spot at this position could be assigned unambiguously to the StrepII-tagged NUJM subunit in subcomplex Iδ [Figure 3, inset 2, and Western Blot analysis (results not shown)]. Peptides indicating the presence of subunit NUXM were found in the oligomeric form of subcomplex Iδ by ESI-MS. However, LILBD-MS and dSDS/PAGE clearly excluded this subunit as a stoichiometric component. Using ESI-MS we confirmed the presence of the recently discovered subunit NUUM in the holo-enzyme and in subcomplex Iδ [42]. However, we derived a different splice pattern for this subunit with a longer N-terminal sequence (see Supplementary Figure S1).

To further confirm the presence or absence of individual subunits, ELISA experiments using specific monoclonal antibodies against the NUPM, NIAM and NESM subunits were performed. In contrast with the anti-NUPM antibody, all other antibodies detected the respective subunit in the subcomplex (see Supplementary Figure S2 and S3). Overall, the evidence generated by dSDS/PAGE, LILBID-MS, BN-PAGE, ESI-MS and ELISA on the subunit composition of complete complex I and subcomplex Iδ indicated that the seven accessory subunits NUPM, NUXM, NB6M, NIPM, NIMM, NI9M and ST1 were detached by LDAO treatment of holo-complex I (Table 1).

Subunit ND2 is present in a subpopulation of subcomplex Iδ

The total mass of the subunits not found in subcomplex Iδ was 177.4 kDa. This corresponds to a residual mass of the subcomplex of 786.3 kDa. As we have indications for the presence of at least one small, as yet uncharacterized, subunit (E. Nübel, S. Kaiser, B. Brutschy, S. Kerscher, B. Meyer, M. Karas, L. Sokolova, K. Zwicker, H. Angerer and U. Brandt, unpublished work) this fits well with the total mass of ~808 kDa determined experimentally by LILBID-MS (blue, Figure 1C). However, this includes 53.3 kDa of subunit ND2 that was found to be present only in a fraction of the subcomplex Iδ sample. Closer inspection revealed that indeed a second peak series corresponding to a mass of ~755 kDa was present in the LILBID-MS spectra obtained in ultrasoft mode (red, Figure 1C). The difference of the experimentally determined masses is consistent with a subcomplex that, in addition to the other ten subunits, had lost subunit ND2. However, the peak intensities for the 808 kDa species suggested a much higher fraction of complexes containing this subunit than was expected based on LILBID-MS single subunit spectra and dSDS/PAGE (Figures 3 and 4). A straightforward explanation for this discrepancy is that the subcomplexes lacking subunit ND2 were much less stable under the conditions of the LILBID experiment, and for the most part were disintegrated even in the ultrasoft LILBID mode, making the 808 kDa subcomplex the dominant species in the spectra. The observed series of peaks indicative of a mass of 413 kDa can be consistently explained by decomposition of laser-irradiation-sensitive subcomplex Iδ into the peripheral arm (green, Figure 1C, Q/N module) and smaller fragments that are not resolved in the spectrum. We conclude that subunit ND2 was present only in a smaller fraction of the sample that, however, was more resistant towards laser irradiation and thus appeared as the major spectral component.

The arrangement of the accessory subunits can be deduced by combining biochemical structural and sequence data

Subcomplex Iδ lacked those central subunits of the membrane arm that, according to the X-ray data [10,11], form the major interface with the peripheral arm. Yet it was stable enough to be prepared biochemically, suggesting that most probably accessory subunits are involved in connecting the two major parts of mitochondrial complex I. We therefore compiled the available information on the 28 accessory subunits of Y. lipolytica complex I to assign them to the different domains (Table 1) and find out how they may stabilize the core of central subunits of the P-module and how the P-module could be interconnected with the peripheral arm.

From their presence in the hydrophilic subcomplex Iλ [1820,22], seven hydrophilic accessory subunits could be assigned to the Q- and N-module of the peripheral arm (Table 1). The sulfurtransferase subunit ST1 is not predicted to be imported into mitochondria (Table 1) and therefore most probably resides in the intermembrane space. Because it is only loosely attached to complex I and has not been found in other species [43] we did not consider it here any further.

Of the remaining 20 accessory subunits, 12 contain at least one predicted transmembrane segment identifying them as integral membrane proteins of the P-module [6] (Table 1). We then reasoned that a high probability for mitochondrial import and the presence of a MTS should indicate that a subunit resides on the matrix side of the membrane arm. This was clearly the case for the acyl-carrier-protein-like subunits ACPM1 and ACPM2. For subunit NI2M the MITOPROT algorithm predicts a targeting sequence and a 100% import probability. However, we could unambigously identify the N-terminal peptide of the precursor protein just lacking the initial methionine residue by ESI-MS (Supplementary Figure S1). We concluded that NI2M was also imported into the mitochondrial matrix, but that for some unknown reason the targeting sequence was not cleaved off. For the remaining five accessory subunits the MITOPROT algorithm predicted a very low import probability score and no MTS. This suggested that they are probably located on the side of the membrane arm facing the intermembrane space. Proteins imported to the intermembrane space are characterized by internal targeting signals and specific patterns of cysteine residues [44]. Indeed, three of the subunits, NUPM, NIPM and NB8M, exhibited canonical twin Cx9C motifs (see Supplementary Figure 4 at This indicated import to the intermembrane space via the Mia40 pathway [45]. In subunit NUPM the twin Cx9C motif is repeated. Also, subunit NIDM may be imported via this pathway, although it contains only one pair of cysteine residues with a spacing of ten residues. NB4M contains no cysteine motif and it remained unclear how it reaches the intermembrane space side of complex I.

Next we assigned the 20 accessory subunits of the membrane arm to the different subdomains defined by X-ray structural analysis [11]. With the exception of subunits NEBM, NUXM, NUNM and NUUM, all subunits have orthologues in bovine complex I. Thus information from biochemically prepared subcomplexes Iα, Iβ and Iλ of the bovine enzyme [1820,22] sub-complexes of Y. lipolytica [22] (S. Dröse, S. Krack, L. Sokolova, K. Zwicker, H.-D. Barth, N. Morgner, H. Heide, M. Steger, E. Nübel, V. Zickermann, S. Kerscher, B. Brutschy, M. Radermacher and U. Brandt, unpublished work and E. Nübel, S. Kaiser, B. Brutschy, S. Kerscher, B. Meyer, M. Karas, L. Sokolova, K. Zwicker, H. Angerer and U. Brandt, unpublished work) and the proteomic analysis of subcomplex Iδ reported here was included in this analysis.

Bovine subcomplex Iβ corresponds to the PD-module of the membrane arm comprising central subunits ND4 and ND5 [18,21]. As it contains homologues of the STMD (single transmembrane domain) subunits [6] NIAM, NESM and NB2M, and the hydrophilic accessory subunits NIDM, NI2M and NB8M, we can assign these subunits to the distal membrane arm module of Y. lipolytica complex I (Figure 5). Bovine subunit B15, the homologue of Y. lipolytica NB5M, is detected in subcomplexes Iβ and Iα, but it is, like STMD subunit NUNM, absent in a Y. lipolytica subcomplex comprising the peripheral arm and the PP-module (E. Nübel, S. Kaiser, B. Brutschy, S. Kerscher, B. Meyer, M. Karas, L. Sokolova, K. Zwicker, H. Angerer and U. Brandt, unpublished work). This suggested that these two subunits are located in the PD-module, with subunit NB5M residing near its interface with the PP-module. The homologue of the transmembrane subunit NUJM has been found in bovine subcomplex Iλ, suggesting a direct connection with the peripheral arm. On the other hand, deletion of subunit NB8M in Y. lipolytica complex I was found to generate a subcomplex concomitantly lacking the PD module and subunit NUJM (S. Dröse, S. Krack, L. Sokolova, K. Zwicker, H.-D. Barth, N. Morgner, H. Heide, M. Steger, E. Nübel, V. Zickermann, S. Kerscher, B. Brutschy, M. Radermacher and U. Brandt, unpublished work). We thus assigned a position in the PD module at the interface to the peripheral arm and the PP module. Considering the absence of intermembrane space subunit NIPM in the two Y. lipolytica subcomplexes described above (Nübel et al., unpublished work; Dröse et al., unpublished work) while the bovine homologue has been assigned to subcomplex Iα [20], we suggest a position at the interface of the PP- and PD-module. In contrast with all other species analysed so far, complex I from Y. lipolytica contains two subunits qualifying as mitochondrial acyl carrier proteins. ACPM1 is more similar to the bovine orthologue SDAP that was also detected in subcomplexes Iβ and Iα. In the Y. lipolytica subcomplex lacking the PD-module (E. Nübel, S. Kaiser, B. Brutschy, S. Kerscher, B. Meyer, M. Karas, L. Sokolova, K. Zwicker, H. Angerer and U. Brandt, unpublished work), ACPM1 is retained and ACPM2 is missing. We thus assigned ACPM1 to the PP-module and ACPM2 to the PD-module and propose that they reside at the interface between of the two domains. Taken together, we concluded that the PD-module consists of 14 proteins, the two central subunits ND4 and ND5, seven accessory subunits with single or multiple transmembrane helices, and two hydrophilic accessory subunits at the matrix side and three at the intermembrane space side of complex I (Figure 5).

Figure 5 Subunit arrangement in complex I (A) and subcomplex Iδ (B) from Y. lipolytica

Accessory subunits form an extended scaffold and connect the peripheral arm (Q/N-module, white) and the membrane arm (PP-module, light grey; PD-module, dark grey). See text for details. The long helical transmission element [11] implicated in conformational energy transfer is shown in light grey. In the subcomplex, subunit ND2 is transparent because it was present in substoichiometric amounts. Subunit ST1 is not shown. The positions of subunits NB4M, NIDM and NI2M have not been confirmed experimentally. IMS, intermembrane space.

Including subunit ACPM1, this leaves five central and eight accessory subunits for the PP-module. The results obtained here by proteomic analysis of subcomplex Iδ provided further information on the arrangement of these subunits within the PP-module. The absence of the four membrane integral subunits NUXM, NB6M, NIMM, NI9M and the cysteine-containing subunit NUPM from subcomplex Iδ identified these proteins, together with the central subunits ND1, ND3 and ND4L, as constituents of the proximal part of the PP-module located directly under the peripheral arm (Figure 5). The predicted position of subunit NUPM at the intermembrane space side of complex I could be confirmed by electron microscopy. The monoclonal antibody 31A8 of the IgG subtype has been shown to bind with high affinity to complex I from Y. lipolytica in native and denaturing ELISA variants [46]. However, it failed to recognize any complex I polypeptide in a standard denaturing dSDS/PAGE/Western blot analysis. In contrast, under milder BN-dSDS/PAGE (three-dimensional PAGE) conditions the integrity of the presumably discontinuous epitope was sufficiently conserved to allow binding of the antibody to the NUPM subunit (Figure 3, inset 3). The identity of the corresponding spot in dSDS/PAGE was confirmed by ESI-MS (results not shown). When we analysed the co-complex of complex I and antibody 31A8 by electron microscopic single-particle analysis (Figure 6), the mass corresponding to the antibody was clearly connected to the proximal end of the membrane arm on the intermembrane space side. This confirmed the predicted position of subunit NUPM. Together with the central subunits ND6 and ND2, the remaining two accessory subunits, the membrane integral subunit NEBM and the subunit NB4M on the intermembrane space side, could be assigned to the part of the PP-module orientated towards the PD-module (Figure 5).

Figure 6 Localization of the NUPM subunit on the intermembrane space side of the membrane arm

Two-dimensional average of single-particle images of complex I decorated with monoclonal antibody 31A8 against the NUPM subunit. Alignment and classification of 749 images yielded six major classes. Class 3, shown here and containing 16% (118) of the particles, most clearly shows the bound antibody. The resolution of the average is 30 Å as measured by the Fourier Ring Correlation with a cut-off criterion of 0.3 [53]. Scale bar=100 Å.


Despite recent progress in X-ray crystallographic analysis of mitochondrial complex I [11], much of the architecture of this membrane integral multiprotein complex of >40 different subunits remains unknown. Notably, the arrangement of the numerous accessory subunits, and how this may relate to the function and stability of complex I remained obscure. In the present paper we report the preparation and proteomic analysis of subcomplex Iδ of complex I, providing clues on how the peripheral and the membrane arm are connected.

The most remarkable feature of subcomplex Iδ was that it lacked subunits ND1 and ND3. From X-ray structural analysis of bacterial complex I [10] and cross-linking experiments [47] it seemed that these two central subunits are the only proteins of the membrane arm making a direct connection with the peripheral arm via the 49 kDa and PSST subunits where the extended ubiquinone-binding pocket has been located by site-directed mutagenesis [48]. The observation that electron transfer to ubiquinone was completely abolished and that the EPR signature of the electron donor for ubiquinone, iron–sulfur cluster N2, was slightly altered in subcomplex Iδ fits with inhibitor-labelling studies suggesting that ND1 might be important for guiding ubiquinone to the active site [49,50]. The fact that subcomplex Iδ still contained the complete Q- and N-module and much of the P-module (Figure 5B) indicated that additional connections between the two arms of mitochondrial complex I must exist.

To obtain further insight into the arrangement of the accessory subunits we performed a sequence-based analysis to predict the import pathways and the sub-compartment localization of each subunit (Table 1). Remarkably, three complex I subunits, NUPM, NIPM and NB8M, contain CX9C motifs that are a hallmark of proteins imported to the intermembrane space by the Mia40 pathway [45,51] (see Supplementary Figure S4). The sulfhydryl oxidase Erv1 and the redox-activated import acceptor Mia40 form a disulfide relay system which is located in the intermembrane space of mitochondria from fungi, plants and animals. Y. lipolytica protein YALI0E21373g (Genolevures databank) is similar to Saccharomyces cerevisiae Mia40 (Uniprot accession number P36046), and protein YALI0D25894g (Genolevures databank) is similar to S. cerevisiae Erv1 flavin-linked sulfhydryl oxidase (Uniprot accession number P27882), indicating the presence of Erv1 and Mia40 as part of a disulfide relay system in Y. lipolytica that promotes retention of imported proteins in the intermembrane space. We consider it remarkable that co-ordination of complex I assembly not only involves two genomes, as complex I subunits are encoded by nuclear and mitochondrial DNA, but also multiple import pathways that direct nuclear-encoded subunits to either the side of the matrix or intermembrane space.

Based on the predictions derived from this sequence analysis and the results obtained for subcomplex Iδ, together with earlier results obtained with other subcomplexes of complex I from bovine heart and Y. lipolytica, we developed a model for the arrangement of the central and accessory subunit of the membrane arm or P-module of complex I (Figure 5). We compared our model with the electron density map of complex I from Y. lipolytica (Figure 7) [11]. In line with our model, electron density on the matrix side of the membrane arm is rather limited and concentrated in the DMP (distal membrane arm protrusion) that had been already identified by electron microsocopy [52]. The DMP may, in part, correspond to subunit NI2M and the N-terminal hydrophilic domains of STMD subunits NIAM and NESM [6]. In contrast, a prominent layer of electron density at the intermembrane space side harbours a number of elongated density features probably corresponding to α-helices oriented in parallel to the membrane plane (Figure 7). This fits well with our prediction that subunits NUPM, NB4M, NIPM, NB8M and NIDM are arranged along this side of complex I. We propose that this ensemble of accessory subunits functions as a platform stabilizing the membrane arm that is connected via the peripherally arranged transmembrane segments of the twelve membrane integral accessory subunits to the matrix side of complex I. Therefore these accessory subunits seem to form a scaffold around the central subunits of the P-module. Our observation that the PD-module stayed connected to the peripheral arm in subcomplex Iδ even in the absence of subunit ND2, suggested that subunit NUJM may be of particular importance to keep the peripheral arm connected to the residual membrane arm. Remarkably, this scaffold of accessory subunits seemed to be partially retained even under conditions that detached several central subunits from the membrane arm. It is tempting to speculate that the pronounced sensitivity of the central subunits in the proximal membrane arm domain towards disintegration reflects a functionally relevant flexibility inherent to a conformational coupling mechanism.

Figure 7 Membrane arm of mitochondrial complex I

Section from electron density map (blue mesh) of Y. lipolytica complex I and structural model for transmembrane segments (yellow) [11]. The contour line of a complex I monomer is tentatively indicated by a broken blue line; the red line separates the PP and PD-module. A prominent continuous layer of extrinsic protein mass is attached to the intermembrane space side (IMS) of the membrane arm (compare with Figure 5).


Heike Angerer designed and performed research, produced complex I and subcomplex samples, performed activity assays, BN-PAGE, dSDS/PAGE, ELISA and alignments, analysed data and wrote the paper. Klaus Zwicker performed EPR spectroscopy. Zibiernisha Wumaier performed Western blot analysis from 3D-PAGE. Esther Nübel identified subunit NEBM. Heinrich Heide and Mirco Steger performed ESI-MS and analysed data. Silke Kaiser provided the StreptII–NUJM. Bernhard Brutschy and Lucie Sokolova performed LILBID-MS and analysed data. Michael Radermacher performed electron microscopy single-particle analysis. Ulrich Brandt and Volker Zickermann analysed data and wrote the paper.


This work was supported by the Deutsche Forschungsgemeinschaft [grant number ZI 552/3-1 (to V.Z.)], the Cluster of Excellence Frankfurt “Macromolecular Complexes” at the Goethe University Frankfurt [grant number EXC115], and the National Institutes of Health [grant number 2RO1 GM068650 (to M.R.)].


We thank Ilka Wittig and Hermann Schägger for fruitful discussions and Andrea Duchene, Karin Siegmund, Maximilian Mattil and Gudrun Beyer for excellent technical assistance. We thank Stefan Kerscher for analysis of the splice pattern of subunit NUUM.

Abbreviations: BN, blue-native; DBQ, decylubiquinone; DDM, β-dodecylmaltoside; DMP, distal membrane arm protrusion; dSDS/PAGE, doubled SDS/PAGE; DTT, dithiothreitol; ESI-MS, electrospray ionization MS; HAR, hexammineruthenium; LC-MS/MS, liquid-chromatography tandem MS; LDAO, lauryl dimethylamine oxide; LILBID-MS, laser-induced liquid bead ion desorption MS; MTS, mitochondrial-targeting sequence; Q1, ubiquinone-1; STMD, single transmembrane domain


View Abstract