A three-way proteomics strategy allows differential analysis of yeast mitochondrial membrane protein complexes under anaerobic and aerobic conditions

Abstract

To investigate the effect of anaerobiosis on the Saccharomyces cerevisiae mitochondrial proteome and the formation of respiratory chain and other protein complexes, we analyzed mitochondrial protein extracts that were enriched from lysates of aerobic and anaerobic steady-state chemostat cultures. We chose an innovative approach in which native mitochondrial membrane protein complexes were separated by 1-D blue native PAGE, which was combined with quantitative analysis of each complex subunit using stable isotope labeling. LC-FT(ICR)-MS/MS analysis was applied to identify and quantify the mitochondrial proteins. In addition, to establish if changes in mitochondrial complex composition occurred under anaerobiosis, we investigated the 1-D blue native PAGE protein migration patterns by Pearson correlation analysis. Surprisingly, we discovered that under anaerobic conditions, where the yeast respiratory chain is not active, the respiratory chain supercomplexes, such as complex V dimer, complex (III)₂(IV)₂ and complex (III)₂(IV) were still present, although at reduced levels. Pearson correlation analysis showed that the composition of the mitochondrial complexes was unchanged under aerobic or anaerobic conditions, with the exception of complex II. In addition, this latter approach allowed screening for possible novel complex interaction partners, since for example protein Aim38p, with a yet unknown function, was identified as a possible component of respiratory chain complex IV.

1 Introduction

The intriguing ability of Saccharomyces cerevisiae to adapt easily to anaerobic conditions 1 is a key contributor to its success in industrial applications such as the leavening of bread, brewing of beer and ethanol production. The effect of anaerobiosis on S. cerevisiae has been studied at the level of the transcriptome 2–5 and recently at the combined levels of the transcriptome and proteome 6. In this latter study, a data set of 474 reliably quantified proteins was obtained of which 249 showed oxygen-dependent changes in expression level. Furthermore, changes in mRNA levels were compared with those of the corresponding protein levels, which surprisingly revealed that many responses were regulated post-transcriptionally, such as for pathways as glycolysis, amino-acyl-tRNA synthesis, purine nucleotide synthesis and amino acid biosynthesis 6. Since previous research was based on the total yeast lysates, key structural aspects of mitochondria operating under anaerobic conditions, such as organization of their respiratory chain complexes, have not yet been investigated.

Mitochondria are eukaryotic organelles of endosymbiotic origin that are presumably derived from respiring α-proteobacteria which, during evolution, were assimilated by anaerobic hosts 7, 8. They still possess their own protein biosynthesis machinery but a lot of the original symbiont genes have been transferred to the nucleus of the host so that extensive interaction between these two organelles is necessary to assemble important mitochondrial structures such as the mitochondrial ribosomes and the respiratory chain complexes 8. The mitochondrial respiratory chain, which couples electron transport to the generation of proton motive force 9, is an example of membrane protein architecture in which specific protein complexes assemble into functional entities. In S. cerevisiae this respiratory chain consists of complexes II, -III, -IV and -V, whereas complex III can form supercomplexes with complex IV with different stoichiometries, and complex V exists both as a monomer and as a dimer 10. Although S. cerevisiae lacks a “classical” complex I NADH-dehydrogenase complex, oxidation of mitochondrial and cytosolic NADH can be coupled to ubiquinone via several alternative dehydrogenases, and shuttle mechanisms 11. As shown in Fig. 1, complex II reduces ubiquinone by oxidizing succinate to fumarate. Subsequently, ubiquinol is oxidized by complex III. The electrons are then transferred to cytochrome C, which is then oxidized by complex IV by reducing O₂ to H₂O 12. Complexes III and IV can translocate protons across the inner mitochondrial membrane throughout electron transfer, resulting in a proton gradient that drives ATP synthesis by complex V. In respiring cultures of S. cerevisiae, this mechanism is the major source of energy to sustain the various anabolic processes in the cell.

Figure 1. Protein complexes of the respiratory chain. Under aerobic conditions, the yeast respiratory chain is organized in four main protein complexes and supercomplexes that generate ATP by chemi-osmotic energy conservation. Respiratory chain complex II is the succinate dehydrogenase complex that catalyzes the oxidation of succinate to fumarate in the TCA cycle, after which it transfers the electrons to the ubiquinone pool. Respiratory chain complex III is the cytochrome bc₁ complex that transfers electrons from ubiquinol to cytochrome C. Respiratory chain complex IV or cytochrome oxidase is the terminal electron acceptor of the respiratory chain that removes two electrons from two molecules of cytochrome C and produces H₂O by transferring electrons to O₂. It translocates two protons across the membrane and produces a proton gradient for ATP generation. Respiratory chain complex V is F₁F₀-ATP synthase, of which the F₁ part is involved in ATP synthesis and hydrolysis reactions, whereas the membrane-bound F₀ mediates proton transport. Respiratory supercomplexes are formed by complexes III and IV in different stoichiometries, and by dimerization of complex V.

The mitochondrial proteome of aerobically grown yeast has been studied extensively by various approaches 13, 14, in contrast, little is known about the function of mitochondria in anaerobic cultures, in which ATP is exclusively dependent on alcohol fermentation. It has been shown by electron microscopy that these organelles undergo major structural changes 15, and it was suggested that the metabolism of anaerobic mitochondria is widely reprogrammed to cope with the absence of respiration 16.

Proteins are generally organized in complexes 17, and several strategies have been developed for large-scale analysis of protein interactions, e.g. by using affinity purification combined with MS 18, by profiling the co-sedimentation of proteins through rate zonal centrifugation gradients 19 or by in silico prediction 20. Other studies have demonstrated that protein correlation profiling is a valid method to introduce a functional dimension into proteomic approaches 21. In particular 2-D blue native/SDS-PAGE has shown to be suited to study membrane protein complexes 22, and this is a well-established method to identify respiratory chain protein complexes of yeast mitochondria 10.

We applied a three-way proteomic strategy to investigate the role of the mitochondria in both aerobic and anaerobic S. cerevisiae. First, we focused on the changes of the mitochondrial proteome during anaerobiosis by employing ¹⁴N/¹⁵N metabolic labeling 6, 23, 24 to aerobic and anaerobic steady-state chemostat cultures. Chemostat cultivation was used as it enables the accurate manipulation of individual culture parameters (e.g. the presence of oxygen) while keeping other cultivation conditions constant 25. Second, we used a novel approach that allowed us to accurately quantify the effect of anaerobiosis on the protein complex levels in the mitochondria, i.e. we used relative quantification through ¹⁴N/¹⁵N metabolic labeling in combination with separation of the complexes using 1-D blue native PAGE (BN-PAGE), making the regular second SDS-PAGE dimension superfluous. Third, to investigate whether the composition of the complexes changes under anaerobic conditions, the proteins separated by 1-D BN-PAGE were analyzed upon correlated protein migration patterns. This last approach is also suited to detect possible new associations of proteins with known mitochondrial protein complex members.

2 Materials and methods

2.1 Experimental design

Three lines of investigation of the response of the mitochondrial proteome to anaerobiosis were chosen, as shown in Fig. 2. In short, S. cerevisiae was grown in glucose-limited chemostat cultures under aerobic or anaerobic conditions, and metabolic labeling was applied using ¹⁴N or ¹⁵N ammonium sulfate as sole nitrogen source. The cell lysates were enriched for the mitochondria. First, all these lysate proteins were relatively quantified, i.e. the samples were mixed at a ratio of 1:1 based on the protein content. After digestion, peptides were separated using strong cation exchange (SCX) chromatography and RPLC followed by on-line detection by FT(ICR)-MS and relative quantification by MSQuant (Fig. 2, I). Second, complexed proteins in this mitochondrial sample were relatively quantified. Aerobic and anaerobic mitochondrial samples were mixed at a ratio of 1:1 based on the protein content. Subsequently, protein complexes were separated by BN-PAGE. Prominent bands were excised, digested and subjected to RPLC-FT(ICR)-MS followed by quantitative analysis (Fig. 2, II). Third, protein migration correlation analysis was performed to assess protein complex formation and/or stability. BN-PAGE was applied for the anaerobic and aerobic mitochondrial samples separately. Lanes with samples of the BN-PAGE gel were cut into 35 slices, and the proteins therein were identified by RPLC-FT(ICR)-MS. The number of identified spectra versus the migration distance was determined for each of the identified proteins. Pearson correlation analysis was performed to locate patterns of protein clusters (Fig. 2, III).

Figure 2. Flowchart of the three experimental designs. Three lines of investigation into the response of the mitochondrial proteome to anaerobiosis were chosen. (I) Protein quantification of anaerobic versus aerobic samples enriched for mitochondria. Aerobic and anaerobic samples were mixed 1:1 based on protein content. After digestion peptides were separated using SCX and RPLC followed by on-line detection by FT(ICR)-MS and relative quantification by MSQuant software. (II) Relative quantification of mitochondrial protein complexes. Aerobic and anaerobic samples were again mixed at a ratio of 1:1 based on the protein content. Subsequently, protein complexes were separated by 1-D BN-PAGE. Prominent bands were excised, digested and subjected to RPLC-FT(ICR)-MS, followed by quantitative analysis using MSQuant. (III) Protein complex correlation analysis. BN-PAGE was performed for both the anaerobic and the aerobic samples separately. Lanes with samples of the BN-PAGE gel were cut into 35 slices, and the proteins therein identified by RPLC-FT(ICR)-MS. The number of identified spectra versus the migration distance was determined for each of identified proteins. Correlation analysis was performed, resulting in patterns of protein clusters.

2.2 Culturing

The S. cerevisiae strain CEN.PK113-7D (MATa, MAL2-8^cSUC2) was cultured and metabolically labeled using light (¹⁴N) and heavy (¹⁵N) nitrogen source (NH₄)₂SO₄ as described earlier 6. Duplicate experiments were performed, using reversed labeling, thus anaerobic and aerobic cultures were each labeled once with ¹⁴N and once with ¹⁵N.

2.3 Enrichment of the mitochondrial fraction under aerobic and anaerobic conditions

Approximately 1.5 g of cells (wet weight) was harvested for each sample. To avoid contamination of the anaerobic samples by oxygen, the sampling from the fermenters was performed under argon environment. Used protocols for isolation of fractions enriched for mitochondria were based on the previously published procedures 26, using 1 μg/mL cycloheximide to prevent mitochondrial protein synthesis during fractionation 27. To maintain anaerobic conditions, all isolation steps of mitochondria from anaerobic chemostat cultures were performed in an oxygen-free cabinet. The mitochondrial fraction was kept in buffer containing 25 mM potassium-phosphate pH 7.5, 1 mM MgCl₂, 1 mM EDTA, 0.65 M sorbitol, supplemented with protease inhibitors (using protease inhibitor cocktail tablets according to the instructions of the manufacturer; Roche Diagnostics) and snap-frozen in liquid nitrogen until further use.

2.4 BN-PAGE

BN-PAGE was carried out as previously described 22. The applied detergent/protein ratio was 5 mg digitonin (Calbiochem, San Diego, CA, USA) per milligram of protein. For the relative quantification experiments about 100 μg of protein of the (¹⁵N-)aerobic and of the (¹⁴N-)anaerobic mitochondria were mixed prior to solubilization with 1 mg digitonin. The samples were incubated for 50 min on ice and then for 20 min at room temperature. Subsequently, membrane protein complexes were separated on a 3–16% acrylamide gradient gel (dimensions: approx. 16×16 cm). The proteins in the gels were fixed with 40% methanol and 10% acetic acid, and then restained with Gelcode Blue Stain Reagent (Pierce, IL, USA). Protein bands were excised, reduced with DTT, alkylated with iodoacetamide and in-gel digested with trypsin 24. For comprehensive analysis of protein complex formation and/or stability in the anaerobic compared with the aerobic mitochondria, BN-PAGE gel lanes of each of both samples were cut into 35 equal slices.

2.5 Digestion and peptide separation

For direct digestion of the enriched mitochondria, organelles were pelleted at 20 000×g for 20 min and resuspended in 8 M urea. Samples were reduced with 200 mM dithiothreitol for 1 h, and subsequently alkylated with 500 mM iodoacetamide in the dark. Digestion with LysC (1:50; LysC:mitochondrial proteins) was performed overnight. Afterward, the urea concentration was reduced to 2 M and a subsequent digestion step with trypsin (1:50; trypsin:mitochondrial proteins) was performed. Urea and salts were removed by binding the peptides on a Resprosil C18-AQ, 3 μm (Ammerbuch-Entringen, Germany) column prepared in a gel loader tip 28. The sample was then dried in a vacuum centrifuge and resuspended in 20% ACN with 0.05% formic acid. SCX was performed on Agilent Zorbax Bio-SCX Series II columns using a linear NaCl gradient from 0 to 0.5 M to elute the peptides from the SCX column. A flow rate of 0.3 mL/min was applied during peptide fractionation using buffers containing 20% ACN and 0.05% formic acid. We collected 1-min fractions and analyzed the peptides from fractions 1 to 20 with LC-FT(ICR)-MS.

2.6 LC-MS/MS and data analysis

Nanoflow LC-MS/MS was performed on an Agilent 1100 nanoflow system (Agilent Technologies) connected to a Finnigan LTQ-FT(ICR) mass spectrometer (Thermo Electron, San Jose, CA, USA) equipped with a nanoelectrospray ion source. The sample was loaded on a C18 precolumn (100 μm id; 375 μm od; Resprosil C18-AQ, 3 μm (Ammerbuch-Entringen) using a flow rate of 5 μL/min. Sequential elution of peptides was accomplished using a linear 1 h gradient from buffer A (0.1 M acetic acid) to 50% of buffer B (80% ACN; 0.1 M acetic acid) over the precolumn inline with a 20–25 cm resolving column (50 μm id; 375 μm od; Resprosil C18-AQ, 3 μm (Ammerbuch-Entringen)). Survey full scans were acquired in a mass range of m/z 300–1500 with a resolution of R=100 000 at m/z 400 and the two most intense ions were subjected to collision-induced dissociation in the linear ion trap.

Bioworks software (Thermo Electron) was used to centroid and merge all MS/MS spectra. Searches were performed against the S. cerevisiae database 29, containing 5779 entries, (http://www.yeastgenome.org/2007) using MASCOT version 2.2 with a precursor mass tolerance of 10 ppm and a MS/MS mass tolerance of 0.9 Da. All data are stored in the public depository PRIDE under accession number XXX and the project description “Blue native electrophoresis combined with stable isotope labeling reveals that S. cerevisiae mitochondrial respiratory chain supercomplexes are retained under anaerobic conditions” (currently only accessible for reviewers; username: review58631, password: XgC-qjHz). Quantification of ¹⁵N/¹⁴N ratios was performed using MSQuant (http://msquant.sourceforge.net). A MASCOT protein cut-off score of 30 was used, corresponding to a false-positive rate of 1%, which was assessed by using a decoy reverse database search strategy. The MASCOT searches of the relevant SCX-fractions were combined prior to MSQuant analysis. Quantification data were normalized on the bulk of unchanged proteins and thus the determined normalization factors were also used to normalize the data obtained by the BN-PAGE experiments. Only proteins identified and quantified in both experiments (¹⁴N aeaerobic/¹⁵N aerobic and ¹⁵N anaerobic/¹⁴N aerobic) were accepted in the final data set, Subsequently, the two ratios were averaged, and propagation of the SD was performed based on the standard mathematical rules, i.e. SD_tot=sqrt(SD+SD). Protein ratios from 0.5 to 2 were considered as unchanged, whereas with ratios outside these ranges were considered to be significantly changed. PCA was carried out on the replicates using the multivariate analysis function in MiniTAB version 14 (http://www.minitab.com/), after which protein hits outside a 95% confidence interval were removed, as described by de Groot et al.6. Some other proteins were not within the 95% confidence interval, but these were included since their ratios were consistent, as these pointed at either up- or down-regulation. Further, a number of proteins were detected solely for one of the growth conditions, these were also considered as interesting and added to the data set. Quantified proteins from the robust data set were further analyzed using the software program BiNGO 30 to assess the overrepresentation of Gene Ontology “cellular component” category “mitochondrion” against the S. cerevisiae proteins database. A hypergeometric test was performed, and the p-value was adjusted for the false discovery rate through Benjamini–Hochberg correction. The chosen significance level was 0.05.

In the last set of BN-PAGE experiments (experiment III, Fig. 2), proteins were identified by LC-FT(ICR) MS/MS and protein migration distances in the gel were analyzed (Fig. 2, III). MASCOT results were loaded into Scaffold (Proteome Software), with the following settings; a peptide confidence level of 95% and a protein confidence level of 99%. This revealed the number of identified spectra for each protein of the BN-PAGE gel lanes. The number of identified spectra and the migration distance for each protein were used as input for Pearson correlation analysis. This was performed using Spotfire version 19.1.977 (http://spotfire.tibco.com) for hierarchical clustering 31.

3 Results

3.1 Experimental approach

We are particularly interested in the S. cerevisiae mitochondrial proteome response to anaerobiosis, for which the cells were grown in glucose-limited chemostats under either aerobic or anaerobic conditions. Metabolic labeling by ¹⁴N or ¹⁵N sources were applied to obtain adequate relative quantification of protein levels using FT(ICR) MS/MS and MSQuant. The three basic approaches that were used for investigation of the effect of oxygen to the mitochondrial proteome are shown in Fig. 2. First, we relatively quantified the proteome from 1:1 mixed anaerobic and aerobic mitochondrial fractions, for which the corresponding mitochondrial peptides were separated by 2-D chromatography, using SCX and reversed phase separation, respectively (Fig. 2, I). Second, to assess the impact of anaerobiosis to mitochondrial protein complex formation and stability, BN-PAGE was applied to separate these complexes under native conditions, followed by quantification (Fig. 2, II). Finally, the correlation of BN-PAGE mitochondrial protein migration patterns was investigated to detect if protein complex composition is maintained in anaerobic mitochondria (Fig. 2, III).

3.2 Anaerobic versus aerobic mitochondrial protein quantification

In total, 1112 proteins were identified in all replicates of the mitochondrial samples. Next, proteins were relatively quantified using MSQuant, resulting in 623 quantified proteins common to both replicates. To determine which of these proteins were reliably quantified, PCA was performed using a 95% confidence interval in a similar manner to what has been described earlier 6, which resulted in 549 reliably quantified proteins (Supporting Information Fig. 1). This set was combined with the 84 proteins that were uniquely detected in samples from one of the two growth conditions, i.e. 18 proteins for the anaerobic and 66 proteins for the aerobic growth condition, respectively. Thus, the final robust data set consists of 633 quantified proteins, which is summarized in Supporting Information Table 1.

The subcellular localization of the quantified proteins was determined by retrieving annotations from the comprehensive yeast genome database (http://mips.gsf.de/genre/proj/yeast/), which revealed that 362 of the 633 quantified proteins in this robust data set could be annotated as mitochondrial (Supporting Information Table 1). To establish if mitochondrial enrichment had occurred, hypergeometrical distribution analysis was performed on the quantified proteins of the robust data set, using “BiNGO” 30. This showed that the sample was indeed significantly enriched for mitochondrial proteins (corrected p-value of 3.2 exp −24), indicating that our enrichment procedure is well suited to study the mitochondrial proteome. “STRING” 32 was used to analyze associations between the mitochondrial proteins in our robust data set, showed a complex network of proteins in which interesting protein complexes and metabolic pathways could be recognized (Supporting Information Fig. 2). For the pathways, proteins involved in ubiquinone biosynthesis, branched amino acid biosynthesis, heme biosynthesis, organic acid catabolism and TCA cycle were detected and reliably quantified. Further, other associations were found for the proteins that form known mitochondrial complexes are the respiratory chain complexes III, IV and V, and the outer mitochondrial membrane protein translocase subfamily.

As can be concluded from our data, mitochondrial proteins with a higher level under anaerobiosis are Coq1p, Coq9p and Cat5p in the ubiquinone biosynthesis pathway and Hem1p, Hem14p and Hem15p in the heme biosynthesis pathway (Supporting Information Table 2). For the branched amino acid biosynthesis pathway, elevated mitochondrial protein levels were found for Bat1p, Ilv3p, Ilv5p and Mmf1p, whereas the other protein levels were unchanged under anaerobiosis, except for Bat2p and Leu3p, which were not detected, which is largely in accordance with earlier data 6. For the organic acid catabolism, only the level of Ald5p was slightly elevated, whereas the Acs1p and Ald4p levels were reduced. In the TCA cycle, the effect of anaerobiosis was only visible for the levels of Aco2p as being raised 2.4 times, and of Cit3p, Fum1p, Kgd2p, Lsc1p, Lsc2p and the YJL045w protein product as being reduced more than two times under anaerobiosis, which is in good agreement with a reduced activity of the TCA cycle in the absence of oxygen 33. Mitochondrial protein levels for the detected members of the respiratory chain complexes III, IV and V were all reduced under anaerobic conditions. Most ribosomal proteins that are located in the mitochondria showed unchanged levels, as only for 7 of the 43 detected proteins more than twofold changes were detected. Finally, of the outer mitochondrial membrane protein translocase subfamily, only the Tom70p level was reduced.

3.3 Quantitative analysis of mitochondrial membrane protein complexes

BN-PAGE was applied to investigate if mitochondrial protein complexes are still present under anaerobiosis (Fig. 2, II). For relative quantification, ¹⁵N anaerobic and ¹⁴N aerobic samples were mixed at a ratio of 1:1 based on the protein content and separated by BN-PAGE, the nine visible bands of the BN-PAGE gel were excised and proteins were identified using LC-FT(ICR)-MS/MS. As shown in Supporting Information Table 3, we could quantify most known components of the respiratory chain complexes and supercomplexes 22. In order of BN-PAGE migration distance, these are supercomplexes V dimer, (III)₂(IV)₂, and (III)₂(IV) and also complex V monomer. These supercomplexes are still present under anaerobiosis; however, their levels are reduced (Fig. 3; Supporting Information Table 3). We could relatively quantify Sdh1p (band 9; its level was unchanged), which is a member of respiratory complex II.

Figure 3. Relative quantification of the major mitochondrial complexes formed under anaerobic or aerobic conditions. 1-D BN-PAGE was applied to the mitochondrial protein samples to investigate the effect of anaerobiosis to mitochondrial protein complex formation (Fig. 2, II). To accurately quantify changes in complex member levels, ¹⁵N anaerobic and ¹⁴N aerobic mitochondrial samples were dissolved in digitonin, mixed at a ratio of 1:1 based on the protein content, and separated by BN-PAGE. At the left side the BN-PAGE gel image is shown, of which the nine prominent bands were excised and the complex subunits were identified and relatively quantified using LC-FT(ICR)-MS. Histograms of protein level ratios (i.e. ratio anaerobic/aerobic) are shown for the following mitochondrial complexes: Prohibitin (1), F₁F₀-ATP synthase dimer (2), supercomplex of III and IV with stoichiometry (III)₂(IV)₂ (3), and with stoichiometry (III)₂(IV) (4), F₁F₀-ATP synthase monomer (5), vacuolar ATPase (no histogram shown) (6), isocitrate dehydrogenase (7), HSP60 (8), dehydrogenases of which some form supramolecular complexes with complex II (9). The numbers 1–9 on top of the histograms correspond to the numbers of the BN-PAGE excised bands, and the size of the bars corresponds to the detected ratios. For each band the protein ratios are given in exactly the same order as in Supporting Information Table 3, with increasing ratios, thus for band 1, bars 1–4 represent Afg3p (ratio 1.21±0.12), Yta12p (ratio 1.28±0.07), Phb2p (ratio 1.65±0.09) and Phb1p (ratio 1.68±0.09), respectively. The histograms show that the prohibitin complex ratios (gel band 1) did not change under anaerobiosis, in contrast to, e.g. the supercomplexes (III)₂(IV)₂ (gel band 3) and (III)₂(IV) (gel band 4) that show overall strongly reduced levels.

The largest complex in the BN-PAGE gel is that of prohibitin, of which the four co-migrating protein levels show no significant change in anaerobic mitochondria (Fig. 3, band 1; Supporting Information Table 3). Detected levels of vacuolar ATPases were mostly increased (Supporting Information Table 3), but since this complex is not localized in the mitochondrial membrane, this result may be an artifact from the fractionation, which was not aimed at selection for the vacuoles, thus it is not further discussed. The levels of the two isocitrate dehydrogenase complex members (Fig. 3, band 7; Supporting Information Table 3) and the tetradecameric mitochondrial chaperonin HSP60 (Fig. 3, band 8; Supporting Information Table 3) did not change. Many dehydrogenases that are part of known monomeric or multimeric complexes, such as for the complex II proteins and the pyruvate dehydrogenase complex proteins were detected (Fig. 3, band 9; Supporting Information Table 3). Drastically reduced levels for anaerobic mitochondria were detected for dehydrogenases Fum1p, Gut2p, Ald4p, unchanged levels for Ndi1p, Kgd1p, Ald6p, Sdh1p, Idp1p and Pdb1p, and elevated levels for Ald5p, Pda1p, Tdh1p and Tdh3p; however, the latter two proteins are not localized in the mitochondria.

Finally, prediction analyses of protein transmembrane helices were performed to the results of two methods, i.e. SCX-RPLC and BN-PAGE. As shown in Supporting Information Fig. 3, 49% of the 810 proteins separated by BN-PAGE have at least one transmembrane domain, whereas only 36% of the 1112 proteins separated by SCX-RPLC have predicted transmembrane domains. BN-PAGE has been developed to separate in particular membrane–protein complexes 22, although few proteins were detected due to the loss of uncomplexed proteins that migrate off the gel. This result indicates that the use of digitonin and BN-PAGE is indeed better suited for the analysis of membrane proteins than 2-D SCX-RPLC separation of the mitochondrial sample peptides.

3.4 BN-PAGE mitochondrial protein migration correlation analysis

To establish if the composition of mitochondrial complexes and supercomplexes is changed under anaerobiosis, samples of digitonin-dissolved aerobic and anaerobic mitochondrial membrane protein complexes were also independently separated by BN-PAGE (Fig. 2, III). The gel lanes were cut into 35 equal pieces, after which the proteins were digested and analyzed by LC-FT(ICR)-MS/MS. After evaluating the number of identified spectra per protein versus the migration distance in the BN-PAGE gel, similar profiles were obtained for interacting proteins, as shown in Fig. 4. Although, for most complex subunits fewer identified spectra per protein were detected in the sample of the anaerobic mitochondria, which is in line with the BN-PAGE quantification results (Fig. 2, II) that also showed reduced levels of respiratory complex proteins. Both complex V dimer and -monomer can be distinguished, similar to the results shown in Fig. 3. Supercomplexes (III)₂(IV)₂ and (III)₂(IV) are detected, and the data of band 20–28 suggest that a (III)(IV) supercomplex is also present. Moreover, separate analysis of both mitochondrial samples reveals that all components of complex II are detected. This is in contrast to the results of the BN-PAGE quantification experiments, where the 1:1 mixed anaerobic and aerobic samples were analyzed, for which only Sdh1p could be relatively quantified. Strikingly, when comparing the complex II patterns for anaerobic and aerobic mitochondria, the maxima for the number of identified spectra shift from gel band 32 to 28, suggesting that a change in complex composition occurred. A possible explanation is that complex II associations with other dehydrogenases, such as Gut2p, Ald4p and Fum1p may be lost in the anaerobic mitochondrial membranes, since these dehydrogenase levels drastically reduced (Supporting Information Table 3). Pearson correlation analysis on these profiles shows that known interacting partners in respiratory chain complexes, such as in complex III, -IV and -V indeed have related profiles, as shown in Fig. 5 (details in Supporting Information Table 4). Thus, these BN-PAGE correlation data suggest that protein complex and supercomplex compositions are unchanged for all complexes and supercomplexes in anaerobic mitochondria, with an exception for complex II, that besides the five known members, Sdh1-4p and Sdh1Bp, may form complexes with other dehydrogenases.

Figure 4. BN-PAGE protein migration profiles reveal co-migration of respiratory chain complexes. Membrane protein complexes of aerobic and anaerobic mitochondria were separated in individual lanes by BN-PAGE. Gel lanes were cut into 35 pieces; the numbers are indicated below the gel lanes and on the x-axes. Protein migration profiles, as assessed by the number of identified spectra versus migration distance in BN-PAGE, show good correlation for the mitochondrial respiratory complexes, i.e. complexes III, IV and V. However, for complex II this is not true, as the migration of complex II members shift from gel band 32 to 28 for the anaerobic mitochondria. Also, supercomplexes (III)₂(IV)₂ and (III)₂(IV) co-migrate well.

Figure 5. Correlation of BN-PAGE migration and mitochondrial protein complex formation. Heatmap showing correlation of the number of identified spectra versus migration distance in BN-PAGE. Proteins that are known to form stable complexes in the mitochondrial membranes were detected in clusters, both for anaerobic and for aerobic mitochondria. Clusters of corresponding complexes and supercomplexes in aerobic and anaerobic mitochondria are closest, indicating that complex composition was not changed upon anaerobiosis. Names of mitochondrial protein complexes are given at the right-hand side, indicated in regular or italic letters are the names of the complexes detected in the aerobic or the anaerobic mitochondria, respectively. Correlation profiles of possibly new interaction partners of complex III/IV are indicated with symbols at that right axis; proteins Aim38p, Aim31p, YDR119W-A and YBR255C-A. The color code of the heatmap ranges from dark gray to white; dark gray indicates a large number of identified spectra for that particular protein, white indicates no spectra.

Surprisingly, Pearson correlation analysis also revealed that the migration profile of some proteins with yet unknown function is highly correlated with the subunit profiles of respiratory chain complexes. For example, the protein Aim38p, localized in mitochondria 14, showed a highly correlated migration profile to complex IV proteins (indicated with ^* in the heatmap of Fig. 5). Figure 6 shows how well BN-PAGE migration profiles of protein Aim38p match with those of complex IV members, in both the aerobic and the anaerobic mitochondria. Other examples of proteins that correlate with the (III)(IV) supercomplexes are Aimp31p, and the proteins encoded by YDR119W-A and YBR255C-A, these are also indicated in the heatmap of Fig. 5.

Figure 6. Protein Aim38p is a potential new interaction partner of complex IV. Protein migration profiles of all detected members of respiratory chain complex IV are indicated with gray lines, and of protein Aim38p with a black line. Pearson correlation analysis revealed that the migration profile of hypothetical protein Aim38p correlates well to the complex IV protein migration profiles, indicating that this protein might be an interaction partner of this complex IV.

4 Discussion

For functioning without oxygen as terminal electron acceptor, eukaryotes such as S. cerevisiae have to maintain redox balance without aerobic respiration. Aerobic respiration occurs in the mitochondria, and some eukaryotes such as protists and metazoa have functional mitochondria without respiration, in case of which terminal electron acceptors other than O₂ are utilized 34. S. cerevisiae is rather unique among eukaryotes and even among other yeast species in that it can grow fast in the presence as well as in the absence of oxygen. The NADH, produced throughout glycolysis, is consumed in the reaction of pyruvate to lactate or ethanol, end products that are subsequently secreted. Despite a plethora of scientific publications addressing mitochondria and their role in yeast physiology, comparatively little is known about how the absence of oxygen and therefore of respiratory chain activity, affects mitochondrial structure and function. This study focusses on the S. cerevisiae mitochondria proteome with an emphasis on membrane protein complex formation and stability.

4.1 Overall quantitative analysis of the mitochondrial proteome

S. cerevisiae was cultured in chemostats to optimally control growth conditions, using either aerobic or anaerobic environment. The lysates were enriched for the mitochondria by differential ultracentrifugation, whereas anaerobic samples were strictly kept oxygen-free during fractionation 26. However, the commonly applied last step of density centrifugation using sucrose gradients to further purify the mitochondria was omitted for several reasons. This was avoided since the density centrifugation step might have introduced oxygen into the anaerobic samples, which may have resulted in superfluous effects on mitochondrial proteome levels or protein complex composition. Most importantly, earlier studies revealed that the morphology of mitochondria changes upon anaerobic growth conditions, resulting in dissimilarly shaped mitochondria 15, which is not compatible with proper purification of mitochondria using density gradient purification.

We compared the current results using mitochondrial fractionation with our earlier results on the relative quantification of total cellular yeast protein lysates 6. This earlier robust data set contained only 30% mitochondrial proteins, whereas for the current set this improved to 58%, and in addition, the number of reliably quantified mitochondrial proteins increased from 143 to 362. It is thus apparent that our relatively simple protocol used for mitochondria enrichment was proven successful.

Further, comparing the quantified mitochondrial proteins of the current data set with this former data set 6 by Spearman rank (SR) correlation analysis resulted in an SR coefficient of 0.692, which is rather good, considering that the character of both samples, the total lysate and the mitochondrial-enriched fraction, is fairly different. Accordingly, we detected an SR coefficient of 0.486 for the non-mitochondrial proteins.

4.2 Mitochondrial membrane protein complex formation and/or stability under anaerobiosis

We successfully separated and relatively quantified mitochondrial protein complexes using a novel combination of 1-D BN-PAGE, a non-denaturating electrophoresis method that keeps protein complexes intact, and ¹⁴N/¹⁵N-labeling for anaerobic and aerobic mitochondrial samples for accurate quantification of the relative levels of all protein complex members individually (Fig. 3, Supporting Information Table 3). Usually BN-PAGE is followed by a second dimension, i.e. SDS-PAGE, to separate the complex members based on the overall size, though by using our innovative approach, we obtain a more accurate evaluation of relative protein levels and specific protein interactions, thus a second dimension is no longer needed. These quantification data are very reliable and robust, and it has been shown earlier that S. cerevisiae metabolic labeling with ¹⁴N/¹⁵N in combination with chemostat culturing provides highly reproducible quantitative data 6, 24. Kolkman et al., have also shown that these protein quantitative data can be validated with Western blot and quantitative PCR 24.

The complex with the highest M_r is the prohibitin complex that is assembled into a membrane-associated ring-shaped supercomplex of approximately 1 MDa, consisting of Phb1p and Phb2p, and acts as a chaperone for newly synthesized membrane proteins 35–37. This complex is also known to be associated with certain m-AAA proteases 37, which were detected with a similar ratio (Supporting Information Fig. 3).

The mitochondrial protein respiratory chain complexes were still present when yeast cells were grown under anaerobic conditions, but their levels were reduced to an average of 55–60% for complex V dimer and monomer, and even to an average of 20–30% for the two complex III/IV supercomplexes, (III)₂(IV)₂ and (III)₂(IV). Moreover, the composition of these complexes is unchanged, since migration patterns of all supercomplex members are highly correlated. S. cerevisiae can grow rapidly under anaerobiosis, whereas it can also adapt quickly when growth conditions are changed back into aerobiosis, which may be facilitated when the respiratory chain complexes are still present in the anaerobic mitochondria 38, 39. Furthermore, although oxygen is not available as final electron acceptor under anaerobiosis, it has been reported that some remaining activity exists in the mitochondrial respiratory chain, suggesting that other electron acceptors are being utilized 38.

A remarkable change for anaerobic mitochondria, however, was found for complex II that showed a significant change in migration for all five complex II proteins Sdh1-4p and Sdh1Bp. The ratio anaerobic/aerobic in the BN-PAGE experiment could only be determined for Sdh1p, which was 1.3±0.1 (Supporting Information Table 3), thus no change in mitochondrial level was detected for this complex II member. Still, the difference in migration points at a difference in complex II composition between aerobic and anaerobic mitochondria, which may be due to the association with other proteins or protein complexes. It has been found earlier through other native electrophoresis experiments that yeast mitochondrial dehydrogenases can form supramolecular complexes 36. A model was proposed that linked the mitochondrial membrane-bound dehydrogenases Gut2p, Nde1p, Nde2p and Ndi1p to a.o. Sdh1p and Fum1p as TCA cycle enzymes that are located close to the membrane 36. Except for Nde1p and Nde2p, we detected all these proteins in our BN-PAGE quantification experiment, all with reduced levels under anaerobiosis, i.e. ranging from 20 to 50% reduction (Supporting Information Table 3), which might explain this change in migration distance within the 1-D BN-PAGE gel. Gut2p is localized in the inner mitochondrial membrane, where it oxidizes glycerol-3-phosphate to dihydroxyacetone phosphate. Together with Gut1p, it acts as a glycerol-3-phosphate shuttle that is responsible for oxidation of NADH under aerobic conditions 40. It has been proposed that under anaerobiosis the aldehyde dehydrogenase Adh3p, collaborates together with the fumarate reductase and the glycerol-3-phosphate dehydrogenase to maintain redox balance 41. We detected down-regulation of Gut2p, which points to reduced participation of the glycerol shuttle in the control of the NAD/NADH redox balance during anaerobiosis. Adh3p, however, was found at higher levels (Supporting Information Table 1), indicating its importance for maintaining the redox balance during anaerobiosis.

Correlation of protein profiles has been successfully used earlier to identify unknown members of organelles 21, 42, 43 and here a variant of this principle was demonstrated by analyzing the protein migration patterns in BN-PAGE. Pearson correlation analysis to the BN-PAGE protein migration profiles of all proteins in the mitochondrial samples revealed clusters of proteins with known interactions (Figs. 4 and 5). Interestingly, by using this approach we could assign protein Aim38p, with unknown function, as possible member of complex IV (Fig. 6). Although further evidence should be obtained that Aimp38p is indeed a complex IV member, this protein was identified as mitochondrial protein 14, whereas deletion of the AIM38 gene leads to impaired growth on non-fermentable carbon sources 44, which points at its involvement in respiration.

5 Concluding remarks

Here, we demonstrate that metabolic labeling can be successfully combined with either 2-D peptide separation by SCX-RPLC or 1-D protein complex separation by BN-PAGE, which forms a powerful multi-faceted technology that yields accurate quantitative proteomic data on intact membrane protein complexes. Our novel approach to combine 1-D BN-PAGE with quantification by protein stable isotope labeling allows accurate analysis of relative changes in yeast mitochondrial complex levels, in particular for each subunit separately, without the need for a second dimension separation by SDS-PAGE. Further, the 1-D BN-PAGE migration correlation analysis permits not only the analysis of changes in complex composition, but also reveals possible new complex members. Our strategy provided us with new insights into the changes that occur in anaerobic mitochondria and also demonstrates a new approach to study membrane complex dynamics and protein networks in general.

Acknowledgements

The authors thank Dr. Bas van Breukelen for his support with PCA analysis and with other bioinformatic tools. Both the Kluyver Centre for Genomics of Industrial Fermentation (http://www.kluyvercentre.nl/) and the Netherlands Proteomics Centre (http://www.netherlandsproteomicscentre.nl/) provided financial support; both programs are embedded in the Netherlands Genomics Initiative.

The authors have declared no conflict of interest.