Defining the protein complex proteome of plant mitochondria.

A classical approach, protein separation by two-dimensional blue native/sodium dodecyl sulfate-polyacrylamide gel electrophoresis, was combined with tandem mass spectrometry and up-to-date computer technology to characterize the mitochondrial "protein complex proteome" of Arabidopsis (Arabidopsis thaliana) in so far unrivaled depth. We further developed the novel GelMap software package to annotate and evaluate two-dimensional blue native/sodium dodecyl sulfate gels. The software allows (1) annotation of proteins according to functional and structural correlations (e.g. subunits of a distinct protein complex), (2) assignment of comprehensive protein identification lists to individual gel spots, and thereby (3) selective display of protein complexes of low abundance. In total, 471 distinct proteins were identified by mass spectrometry, several of which form part of at least 35 different mitochondrial protein complexes. To our knowledge, numerous protein complexes were described for the first time (e.g. complexes including pentatricopeptide repeat proteins involved in nucleic acid metabolism). Discovery of further protein complexes within our data set is open to everybody via the public GelMap portal at www.gelmap.de/arabidopsis_mito.


Introduction
The proteome of Arabidopsis thaliana mitochondria has been extensively characterized by gel-based and gel-free strategies (Kruft et al., 2001;Millar et al., 2001;Heazlewood et al., 2004;Heazlewood et al., 2007;Lee et al., 2011;Taylor et al., 2011). Based on these projects, more than 500 proteins were assigned to mitochondria in Arabidopsis. However, targeting prediction software tools assign more than 1500 proteins encoded by the Arabidopsis genome to this subcellular compartment. It therefore is concluded that most mitochondrial proteins, especially those of low abundance and/or high hydrophobicity, remain to be discovered. Many mitochondrial proteins form part of protein complexes, e.g. the complexes of the respiratory chain.
However, again, only the complexes of high abundance could be characterized while the ones of low abundance so far are not known.
Meanwhile, mass spectrometry and appendant software tools have much improved.
We therefore decided to re-use BN / SDS PAGE for a first broad-scale proteomic analysis of plant mitochondria based on this experimental system. Overall, 471 unique proteins were identified in 200 gel spots excised from a single blue native gel.
The new 2D gel presentation software tool GelMap was further developed to allow a systematic exploration of the "protein complex proteome" of plant mitochondria. Evidence for the occurrence of at least 35 different protein complexes in Arabidopsis mitochondria is presented, several of which were described for the first time, e.g. protein complexes including the plant-specific "PPR" proteins. The "protein complex proteome" of Arabidopsis mitochondria as defined by our study so far consists of more than 200 distinct proteins. Results of our investigation are presented at www.gelmap.de/arabidopsis_mito.

Results and Discussion
Separation of mitochondrial proteins from Arabidopsis by 2D BN / SDS PAGE Mitochondrial fractions of ten independent mitochondrial isolations from Arabidopsis thaliana cell suspension cultures were separated by 2D Blue native / SDS PAGE as described in the Material and Methods section. Resulting gels were highly reproducible as revealed by gel comparisons using the Delta 2D TM 4.2 software package (data not shown). A typical gel (Figure 1) was selected for further analyses. The most prominent 200 spots were excised and analyzed by electrospray tandem mass spectrometry. On average, six different proteins were identified per spot. Identifications were accepted if supported by MS reliability scores (Mascot) and by at least two matching peptides per proteins (for details see Supp. Table I Figure 2.

Evaluation of organelle purity
To assess the purity of mitochondrial fractions, the subcellular localizations of all identified proteins were evaluated using the "Arabidopsis SubCellular Proteomic Database" (SUBA II, http://suba.plantenergy.uwa.edu.au/). This database summarizes all experimental and computer prediction data for subcellular localizations of proteins in Arabidopsis (Heazlewood et al. 2005, Heazlewood et al. 2007. Of the overall 200 analyzed gel spots, 194 of the "first hit" identifications (97%) represent mitochondrial proteins according to experimental evidences given in the SUBA II database (Supp. of the identified unique proteins represent known mitochondrial proteins and another 20% are assigned to other cellular compartments like plastids (9%), peroxisomes (2%) or the plasma membrane (8%). However, most of the unassigned proteins are localized in mitochondria according to targeting prediction computer programs and therefore represent candidates for new constituents of this subcellular compartment.
Furthermore, identifications of proteins assigned to other subcellular compartments in most cases are based on only few peptides (resulting in a comparatively low MS reliability score) which indicates that these proteins are of rather low abundance in our fraction. Therefore, evaluation of our MS dataset using SUBA II was repeated based on the identified peptides of all proteins (Supp. Table 2.3). Of the overall 6992 peptides, 6204 peptides belong to known mitochondrial proteins (89%). A comparatively low number of 346 peptides (5%) is assigned to known non-mitochondrial proteins.
Another 442 peptides (6%) are of unknown subcellular localization because experimental data are lacking so far. However, most of the corresponding proteins are predicted to be localized in mitochondria (Supp. Table III). Therefore, the unassigned peptides / proteins represent candidates for newly identified mitochondrial proteins.
Based on the different evaluation results we finally conclude that our organellar fraction has a purity in the range of 93-95%.
Annotation of the "protein complex proteome" of Arabidopsis mitochondria The novel GelMap annotation software package was used for online data presentation. GelMap was recently developed to functionally annotate 2D IEF / SDS gels (Rode et al., 2011). The original software proved not to be suitable for the annotation of 2D BN / SDS gels because spots in general include several different proteins. We therefore upgraded the application range of GelMap for the assignment of lists of proteins to individual spots. Since all proteins are annotated according to distinct functions and protein complexes, the software now allows to selectively display unknown protein complexes based on the vertical positioning of their subunits on a 2D Blue native gel.
For GelMap annotation, the image file of the 2D BN / SDS gel (Figure 1)  The How-to area of the GelMap website provides all information necessary to prepare and upload a dataset (http://gelmap.de/howto). Based on the GelMap software package, the content of the columns of the "coord" table is automatically displayed on the map in pop-ups linked to the spots or in a menu to the right of the gel image, which includes information concerning assignment of the proteins to protein com-  Figure 4) and by clicking on a protein name further information is displayed, including a "more protein details" link at the bottom which gives access to the "coord" table including all detail information available for the protein of interest. A "more peptide details" link leads to a second table indicating peptide specific information (e.g. ion score of a peptide, its amino acid sequence, and its modifications). In case a spot only includes a single protein, the additional information pop up is displayed directly.
Features best can be followed if directly tested in the internet (www.gelmap.de/arabidopsis_mito). Proteins detected within the same spot are sorted according to their Mascot scores (column 2 in Supp. Table I). In general, we interpret that high scores (resulting from high scores for the identified peptides and/or many identified unique peptides) reflect high abundance of a protein. However, this assumption not always is correct because the Mascot scores also depend on the biochemical properties of proteins.
The "protein complex proteome" of Arabidopsis mitochondria The annotation of Arabidopsis mitochondria as generated by GelMap offers several advantages. First of all, it allows easy access to protein identification data of large proteome data sets based on functional categorization. But more importantly, it specifically allows searching for protein complexes of low abundance, which are positioned at the same location on the gels as protein complexes of high abundance, like the complexes of the Oxidative Phosphorylation system. The low abundant protein complexes in most cases have comparatively low MS reliability scores, which nevertheless are clearly above the threshold. Using the categorization tools offered by GelMap, their positioning along a vertical line becomes visible, allowing to deduce new protein-protein interactions. The following 35 protein complexes were found in the course of the project (summarized in Table II): The I+III 2 supercomplex (1500 kDa) and complex I (1000 kDa). Subunits present within plant complex I recently were systematically characterized (Klodmann et al., 2010;Klodmann and Braun, 2011). Currently, 48 different subunits are known, seven of which occur in pairs of isoforms. In the frame of the present study, 44 subunits were detected in monomeric complex I. Four subunits were not found (ND4L [AtMg00650], 13 kDa subunit [At3g03070], and the plant specific subunits At5g14105 and At1g68680. Subunit ND4L was never described in any proteomic study, probably due to its extreme hydrophobicity. The plant specific subunit At1g68680 (8.3 kDa protein), which only was described by Meyer et al., 2008, is present on our 2D gel map but at a position far away from complex I (spot 168; 60 kDa range with respect to the native gel dimension). It could have been detached from complex I during solubilization or it may represent an erroneously identified complex I subunit. Most of the identified subunits of complex I also were found in the 1500 kDa I+III 2 supercomplex.
Four additional proteins were discovered by our study, which represent candidates for newly identified complex I subunits: At3g10110 (spots 17), At1g18320 (spot 186), At1g72170 (spots 149 and 151), and At2g28430 (spot 154). The first two subunits represent isoforms and resemble proteins of the TIM preprotein translocase family.
Interestingly, another complex I subunit, the so-called "B14.7" protein, exhibits similarity to the proteins of the TIM family (Carrol et al., 2002). Three of the four new subunits were found both in the 1500 kDa supercomplex and in monomeric complex I, supporting their assignment as complex I subunits (one new subunit, At2g28430, only was detected in the supercomplex).
Complex I subunits were additionally detected in 27 spots not forming part of the vertical rows of spots representing the I+III 2 supercomplex or monomeric complex I. All these subunits migrate at smaller molecular mass with respect to the first gel dimen- This illustrates the suitability of the GelMap software tool to identify protein complexes of low abundance, e.g. assembly intermediates of OXPHOS complexes.
The ATP synthase complex (complex V). Based on our molecular mass calibration ( Figure 2), complex V runs at 660 kDa on the native gel dimension. Fourteen subunits have been reported for Arabidopsis (Heazlewood et al., 2003c;Meyer et al., 2008), all of which are included in our map except for subunit c, which is small and very hydrophobic. Additionally, a homologue of subunit g from other groups of eukaryotes was detected in the low native mass range (14 kDa protein; present in three isoforms: At4g29480, At2g19680, At4g26210; spots 164 and 165). In yeast, this subunit only is associated to ATP synthase within the dimeric ATP synthase supercomplex (Arnold et al., 1998). Under the conditions applied during our investigation, ATP synthase dimers are not stabilized. Since subunit g is completely detached from the ATP synthase monomer on our 2D gel, it probably also represents a dimer-specific protein in Arabidopsis.
The more hydrophobic subunits of complex V, especially subunits a and b, were identified in several neighboring spots and therefore are difficult to precisely localize on our map. Our interpretation of assignment of these subunits to the spots on the 2D gel is given in Supp. Figure 1.
On our 2D gel (Figure 1), the known F 1 and F 0 subcomplexes run at 300 and 260 kDa. The relatively high abundances of these subcomplexes indicate that they rather represent breakdown products than assembly intermediates. F 1 includes subunits α, β, γ, δ, and ε, F 0 the subunits a, d, 8, OSCP, ATP17 (plant specific) and 6 (plant specific). The FAD subunit does not form part of the two subcomplexes but only was detectable within the holo-complex or as a singular protein (spots 132, 133).
Subunits of complex V additionally were detected within several other spots not forming part of the vertical rows of spots representing the holo-complex or the F 1 and F 0 subcomplexes, especially in the low native mass range. These proteins may form part of smaller complex V subcomplexes or assembly intermediates.
Cytochrome c reductase (complex III; 500 kDa). This central complex of the OXPHOS system has 10 subunits in plants (Braun and Schmitz 1995;Meyer et al., 2008), all of which are included in our map. For the first time, the so-called "hinge" subunit (QCR6) was detected for Arabidopsis. Six of the subunits are present in pairs of isoforms (α-MPP, Cyt c 1 , FeS protein, QCR6, QCR7, and QCR8). All 10 subunits also form part of the I+III 2 supercomplex. No subcomplexes of complex III were detected, further supporting its extraordinary stability (Heinemeyer et al., 2009).
Cytochrome c oxidase (complex IV). On BN / SDS gels, complex IV usually is found in a larger form (230 kDa) and a smaller form (200 kDa Vb subunits are visible on the gel but peptides could not be detected, most likely due to the extreme hydrophobicity of these proteins. Currently, further efforts are made to identify these subunits and include them into the online-version of GelMap. The Cox X6 subunit was found in gel regions representing larger native masses and therefore might only be included in the larger form of complex IV. Some of the complex IV subunits occur in two or more isoforms (Millar et al., 2004b), several of which were detected in the frame of the present project.
Succinate dehydrogenase (complex II). The holo-complex has a size of 160 kDa and consists of 8 subunits termed SDH1-SDH8, four of which are plant specific (SDH5-SDH8) (Eubel et al., 2003;Millar et al., 2004b). All proteins except SDH8 are included in our map (SDH8 migrates at the dye-front on the 2D BN / SDS gel). Some proteins occur in isoforms. All subunits except SDH5 also are present in a smaller form of complex II, which runs at about 100 kDa. Analyses by BN / BN PAGE revealed that complex II gets dissected into two subcomplexes which co-migrate at this position on the native gel dimension (Sunderhaus et al., unpublished).
Cytochrome c. The intermembrane space protein cytochrome c, which translocates electrons from complex III to complex IV, occurs as monomer of 12 kDa but also was detected in the native range of about 100 kDa. It is positioned on a vertical line together with the cytochrome c-binding enzyme L-galactono-1,4-lactone dehydrogenase. We speculate that the two proteins form a protein complex of about 100 kDa.
Alternative NAD(P)H dehydrogenases. Alternative NAD(P)H dehydrogenases are encoded by two gene families in Arabidopsis termed NDA (2 genes) and NDB (4 genes) (Michalecka et al., 2003). On our map, we found the NDA2, NDB2 and NDB4 proteins (spots 96 and 94). Molecular masses of the monomeric precursor proteins are in the range of 56 to 65 kDa. All proteins were detected in the 160 kDa range of the native gel dimension. Similar results were previously reported for NDA2 (Ras-musson and Agius 2001). We postulate that the NDA and NDB proteins might form a protein complex which probably has a hetero-trimeric structure.
Protein complexes of TCA cycle enzymes. All enzymes of the TCA cycle are included in our map. Due to horizontal and vertical smearing effects of the BN / SDS gel, most proteins were identified within several spots. However, based on their Mascot scores, conclusions can be drawn concerning their main location on the map. In accordance with reports in the literature, three enzymes, fumarase, isocitrate dehydrogenase and malate dehydrogenase, have a homo-tetrameric structure (Table II) Pyruvate dehydrogenase complex. Pyruvate dehydrogenase probably is the largest protein complex of the mitochondrial matrix. It is composed of numerous copies of the E1-alpha, E1-beta, E2 and E3 subunits and has an overall molecular mass of about 9.5 MDa (Zhou et al., 2001). The holo-enzyme was outside of the molecular mass limit of our BN / SDS gel. However, subcomplexes were found. A tetramer composed of two E1-alpha (spot 110) and two E1 beta (spot 104) subunits runs at about 140 kDa on the native gel dimension (Table II). Furthermore, much larger protein associations were detected which include the E3 subunit (spots 3, 31, 96, 172). The preprotein translocase of the outer mitochondrial membrane (the TOM complex).
The core of the TOM complex consists of 6 distinct subunits: TOM40, TOM20, TOM22 (TOM9), TOM7, TOM6, and TOM5 (Werhahn et al., 2001;Werhahn et al., 2003). Its native molecular mass is 260 kDa on our 2D gel, slightly smaller than reported before. All subunits are included in our map, except for the very small TOM5 and TOM6 proteins which migrate within the dye-front of the second dimension gel.
The preprotein receptor TOM20 occurs in several isoforms. One protein of unknown function co-migrates with the TOM complex on the 2D gel (At3g49240, spot 62). It has a size of 70.2 kDa and contains a tetratricopeptide motif like the TOM70 prepro- within the oligomers. The ADP/ATP translocase is known to have a large aminoterminal extension which is removed upon import but dispensable for protein targeting (Emmermann et al., 1991;Winning et al., 1992;Mozo et al., 1995). Possibly, this precursor protein should be considered to represent a polyprotein which is cleaved into two separate polypeptides, the smaller of which is of unknown function. Similar to the results obtained for the ADP/ATP translocase, other members of the mitochondrial metabolite carrier family were predominantly found in the 60-120 kDa region, e.g. members of the dicarboxylate-and tricarboxylate carrier subfamilies, the phosphate translocase and the plant uncoupling protein (PUMP).
Porin complexes. Also "porin" of the outer mitochondrial membrane, which is desig- Changes in abundance of protein complexes will be analysed in the future and results will be implemented into the GelMap project.
The identification of protein complexes by combination of 2D BN / SDS PAGE, tandem mass spectrometry and GelMap provides comprehensive insights into the plant mitochondrial protein complex proteome. Many protein complexes could be identified for the first time. Biochemical studies will follow to validate the presence and composition of these protein complexes in the future.

Conclusion
Our study allowed us to identify 471 distinct proteins on a 2D BN / SDS gel and to assign more than 150 of these proteins to 35 different mitochondrial protein complexes. Due to the application of Blue Native / SDS PAGE, hydrophobic proteins are not discriminated. Therefore, broad range insights into the "protein complex proteome" of plant mitochondria have been possible. Several protein complexes were described for the first time, e.g. complexes including the plant specific PPR proteins. Nevertheless, our experimental strategy to identify protein complexes has to be considered to represent a biased approach because the 200 most abundant spots were selected for MS analysis and thereby protein complexes of low abundance only were detected if positioned at a gel region of a protein complex of high abundance. If the "protein complex proteome" of plant mitochondria shall be completely analyzed it would be necessary to systematically scan a 2D BN / SDS gel by MS, e.g. within 100 horizontal gel stripes by 100 spots each (= 10.000 spots). In principle, such an analysis would be feasible and annotation by GelMap would allow visualizing all protein complexes above the detection limit for MS-based protein identifications. Already for the study presented in this publication, the GelMap software package proved to be very helpful for the exploration of a 2D BN / SDS gel. Its option to differentially display protein complexes should be of even increased value for evaluating BN / SDS gels of non-mitochondrial fractions because resolution of the resulting gels often is of lower quality. We believe that GelMap in general will prove to be a valuable tool for gelbased proteomics.

Material and Methods
Preparation of Mitochondria and 2D Blue native / SDS PAGE Mitochondria were isolated from Arabidopsis cell suspension cultures as described in Werhahn et al., 2001. Organelle fractions were divided into aliquots of 100 µl (10 µg mitochondrial protein / µl). For sample preparation prior to Blue native PAGE, freshly prepared organelles of one aliquot were sedimented by centrifugation for 10 minutes at full speed (Eppendorf centrifuge) and resuspended in 100 µl "Digitonin solubilization buffer" (30 mM HEPES, 150 mM potassium acetate, 10 % [v/v] glycerol, 5 % [w/v] digitonin). After incubation for 10 minutes on ice, insoluble material was discarded by another centrifugation step (10 minutes, full speed, Eppendorf centrifuge) and the supernatant was supplemented with 5 µl Coomassie-blue solution (750 mM aminocaproic acid, 5% [w/v] Coomassie 250 G). Fractions were directly loaded onto a Blue native gel. Blue native gel electrophoresis was carried out as described before (Wittig et al., 2006). Basic parameters were the following: (i) Protean II electrophoresis chamber (Biorad, München, Germany), (ii) gel dimensions: 16 x 16 cm, (iii) polyacrylamide concentration of the first dimension gradient gel: 4.5 to 16% (top to bottom), (iv) second gel dimension: Tricine SDS-PAGE system (Schägger and von Jagow, 1987), constant polyacrylamide concentration in the separation gel (16.5%) which was overlaid with 2.5 cm spacing gel as described in Jänsch et al., 1996. Gels were stained according to the Coomassie-colloidal protocol (Neuhoff et al., 1988).

Protein identification by mass spectrometry
Spots were cut out from a 2D gel using a manual spot picker (Genetix, New Milton, Hampshire, United Kingdom; spot diameter: 1.4 mm) and were numbered consecutively from 1 to 200. In-gel digestion was carried out as described in Klodmann et al., 2010. Tryptic peptides were resolved in 22 µl LC sample buffer (2% Acetonitrile, 0.1% FA in H 2 O) and 15 µl were injected into the LC-MS system. Nano high-performance liquid chromatography (nLC) electrospray ionization quadrupole time of flight mass spectrometry (ESI Q-ToF MS) analyses were carried out with an Easy-nLC system (Proxeon, Thermo Scientific, Dreieich, Germany) coupled to a micrOTOF Q II MS (Bruker Daltonics, Bremen, Germany). For LC separation, a two column setup with a C18 reversed phase for hydrophobic interaction of peptides was used.
Solution B then was again continuously increased for 20 minutes to finally 95% B and 5% A. MS/MS fragmentation was carried out automatically. For this, the software selected up to 3 peptides of highest intensity (with a minimal intensity of 3000 counts) in the MS precursor scan.
For data processing DataAnalysis software from Bruker Daltonics was used.
Database search was carried out with ProteinScape 2.0 (Bruker Daltonics) and the Mascot search engine in the TAIR10 database (www.arabidopsis.org). The search was carried out with the following parameters: enzyme: trypsin/P, with up to 1 missed cleavage allowed; global modification: carbamidomethylation (C), variable modifications: acetyl (N), oxidation (M); precursor ion mass tolerance 20 ppm, fragment ion mass tolerance 0.05 Da, Peptide charge 1+, 2+ and 3+, Instrument type ESI QUAD TOF. Minimum peptide length was set to 4, protein and peptide assessments were carried out if the Mascot Score was >25 (for proteins as well as peptides).
Proteins were further validated by comparison with the SUBA II database (Heazlewood et al., 2005, Heazlewood et al., 2007 to determine the proportion of nonmitochondrial proteins in our sample (see Supplementary Tables II and III). Therefore, accession numbers of all identified proteins (according to the TAIR 10 database) were used for a search in the SUBA II database. Available proteomic data (MS and GFP data) given in SUBA II were used to assign each protein to its most likely subcellular localization. For four different data subsets (first hit proteins, unique proteins, all proteins and all peptides) the subcellular distribution, as well as the proportion of mitochondrial and non-mitochondrial proteins was determined (see Supp. Table II Gels were scanned using the Image Scanner III (GE Healthcare, Munich, Germany) and stored as JPEG files with minimal compression. Image files were loaded into the Delta 2D TM 4.2 software package (Decodon, Greifswald, Germany). Spot detection was carried out by automatic mode of the software package and corrected manually.
A file including the coordinates of all spots was generated ("coord file") and exported into Excel (Microsoft, Unterschleißheim, Germany). Finally, gel image (.jgp) and coord (.txt) files were exported into the GelMap software package available at www.gelmap. de (Rode et al. 2011). Information how to use GelMap are given in the

Supplementary Material
Supp. Figure 1: Assignment of subunits of complex V (ATP synthase) Supp. Figure 2: Protein complex visualization by GelMap Supp.     names of all its subunits become available (see Supp. Figure 2). Detail information on spots is given upon "clicking" on encircled spots on the 2D image (see Figure 4).
The GelMap software package also has different search options, e.g.   Table I). The "more peptide details" link leads to a table which includes details on the peptides identified by MS. Besides the characteristics already given in the pop-up window, it contains further information about spot characteristics (like coordinates and apparent masses), the MS analysis, and the physiological context of the proteins (see Table I).

Legends of Supplementary Material
Supp. Spots were numbered (column A), cut out from the gel and analyzed by LC-ESI-MS/MS. Database searches were performed using Mascot (see Material and Methods). The table shows the Mascot score for each protein (column B), the sequence coverage by the identified peptides (column C) and the number of unique peptides found for a protein (column D). The calculated molecular mass of a protein is given in column E, its apparent mass in the second (vertical) gel dimension (column F) and in the first (horizontal) and native gel dimension (column G) were calculated by a formula deduced from the positions of known proteins / protein complexes on the gel (see Figure 2). General information about the protein is given to the right of the analytical parameters: protein name (column I), accession number in the TAIR database tion" (W) and "remarks" (X). In the "subcellular localization" column, assignment of all proteins was carried out with respect to the "MS data" and "GFP data" column (Y and Z, respectively). A protein was assigned to the cellular compartment with the most experimental evidence. In some cases, the "subcellular localization" differs from the data found in SUBA II, if a protein is known to be found in another compartment than the results in the SUBA II database indicate. Therefore, a comment is added in the "remarks" column. In case no proteomic data are given in the SUBA II database, the protein is dedicated to mitochondria (unless there is evidence for the protein to be located in another compartment, which is then indicated in the "remarks" column). Those newly identified mitochondrial proteins are called "NEW mitochondria" in the "subcellular localization" column. Several dual-targed proteins are assigned to "mitochondria&plastids". For the evaluation (Supp . Table II) these are only counted for mitochondria. The "remarks" column shows short informations that helped in the evaluation of the data. T10 = localization/information given the TAIR10 database.
Supp. Figure 1: Assignment of subunits of complex V (ATP synthase).
The figure shows a detail view of the monomeric ATP synthase complex in the BN / SDS gel from Arabidopsis thaliana mitochondria (Figure 1). Proteins were assigned to the spots based on MS analyses. Subunits indicated in red form part of the F 1 subcomplex, subunits in blue are either part of the F 0 subcomplex or the stator stalk. The position of subunit g is indicated in brackets because it is specific for the dimeric ATP synthase supercomplex and has not been identified in the monomeric complex. Subunit c (gray) was not found on the gel but can be assigned due to results of earlier studies (Heazlewood et al., 2003). Subunits a and b could not be assigned properly due to their high hydrophobicity but can be detected in the range between OSCP and delta subunit (orange line). A molecular mass standard (kDa) is given on the left (gray).
A) The menu at the right of the gel image shows a list of all protein complexes present on the gel, sorted by physiological function. The number of proteins identified for a complex is given in brackets (e. g. 14 proteins for the TOM complex). In the gel all analyzed spots are circled. B) By clicking on a complex (here: TOM complex), a list of all identified subunits of this complex appears (light blue), the number in www.plantphysiol.org on August 18, 2017 -Published by Downloaded from Copyright © 2011 American Society of Plant Biologists. All rights reserved. brackets indicates how often a subunit was found on the gel. When a complex in the right menu is chosen, also the spots on the gel map are automatically filtered: only spots are visible which contain subunits of the chosen complex. The pop-up windows on the map (also see Figure 4) also exclusively show proteins belonging to the complex (compare the pop-up windows on the gel of A and B). Thus, it is easy to visualize the position of any protein complex on the gel, even if it is of low abundance.
www.plantphysiol.org on August 18, 2017 -Published by Downloaded from Copyright © 2011 American Society of Plant Biologists. All rights reserved. Table I: Column content of the results table presented in the supplementary materials  (Supp. Table I) column (from left to right) content comment 1 spot ID spot ID on the 2D BN / SDS gel 2

Tables
Mascot score probability score for the protein identification www.matrixscience.com 3 No. of peptides number of unique peptides matching to a protein hit 4 coverage sequence coverage of a protein by the identified peptides 5 calc. MM calculated molecular mass of the protein. Note: in many cases, presequences are cleaved off from nuclear encoded mitochondrial proteins after their transfer into mitochondria is completed. As a consequence, the apparent MM is 1-8 kDa smaller 6 app. MM 2nd apparent molecular mass as determined on the second gel dimension (see Figure 2 for mass determination) 7 app. MM 1st apparent molecular mass as determined on the first gel dimension (see Figure 2 for mass determination) 8 accession