Figure 1.
Enrichment for Mitochondrial Proteins by MS
(A) shows the 546 proteins (in rows) identified from 28 datasets (columns). The proteins are sorted in decreasing order down rows by the number of experiments in which peptide tags were identified by MS and binned into three classes of detection frequency. The number at the bottom of each class indicates the total number of proteins in the class. Proteins that are part of the reference set, and thus are previously known mitochondrial proteins (M), are marked to the left. The experiments are divided according to fermentable (F) and nonfermentable (NF) mitochondrial preparations.
(B) Proportions of proteins identified in membrane and matrix fractions. Whether a protein was detected predominantly in either the membrane or matrix fraction, or equal in both fractions, was determined based on where it was detected with an average higher tag number. Shown are the proportions for all 546 proteins, for known matrix proteins (i.e., matrix and intermembrane space, n = 109), for known membrane proteins (i.e., inner and outer membrane, n = 101), and for detected proteins not previously known to be mitochondrial (n = 290).
(C) Distribution of proteins identified under fermentable and nonfermentable conditions by proteomics, and overlap with previously known mitochondrial proteins. Total numbers are given in parentheses.
(D) Breakdown by localization of the 546 proteins identified. For mitochondrial localization the reference set was chosen; for localization outside mitochondria the GFP fusion protein data were used (Huh et al. 2003). The inner circle represents the distribution for all proteins in yeast.
(E) Distribution of median mRNA expression under fermentable and nonfermentable conditions, protein abundance under fermentable conditions (Ghaemmaghami et al. 2003), and protein length across bins of confidence of identification (maximum number of tags identified in any of the 28 datasets). The bars indicate fold differences from the median for the known mitochondrial proteins that were not detected by MS (“M not det.”).
Figure 2.
Evaluation of Proteomic Data for Protein Abundance and Mitochondrial Localization
(A) Coverage of known mitochondrial proteins (Mref) by two MS proteome studies (this study and Sickmann et al. [2003]). We evaluated the 340 proteins of the mitochondrial reference set for which protein abundance data existed (Ghaemmaghami et al. 2003). The x-axis represents the median protein abundance of ten consecutive, equally sized bins of proteins.
(B) Distribution and overlap of proteins identified by the two MS studies and known mitochondrial proteins. The total number of entries for each dataset is indicated in parentheses outside each circle. The number inside each circle indicates the number of proteins in each of the categories. In addition, the percentage of proteins that were localized to mitochondria by GFP tagging (Huh et al. 2003) is given in parentheses for each category.
Figure 3.
Functional Categories and Cellular Localization of Our Proteomic, Deletion, and Expression Datasets
Each field shows the proportion of proteins found by the experiment out of the total number of proteins known with a given combination. Fields are color coded by the level of coverage gained by the experiment (color scale upper right). Localization outside mitochondria was based on the GFP fusion protein data (Huh et al. 2003). Fields with less than three identified proteins were not evaluated and left blank. In the upper left corner is the distribution and overlap of proteins identified by each experiment. In parentheses are the known mitochondrial proteins based on the reference set.
Figure 4.
Specificity and Sensitivity of Systematic Approaches with Regard to Mitochondria
Various datasets were benchmarked against the mitochondrial reference set. Each dot in the graph represents an entire dataset: PSORT (Nakai and Harton 1999), hap4 expression (Lascaris et al. 2003), deletion phenotype screen (Steinmetz et al. 2002), tag localization (Kumar et al. 2002), GFP localization (Huh et al. 2003), MitoProt greater than 90 (Scharfe et al. 2000), Bayesian prediction (Drawid and Gerstein 2000), pet phenotypes (Dimmer et al. 2002), three MS proteome studies (Pflieger et al. 2002; Sickmann et al. 2003; Ohlmeier et al. 2003), mitochondria localized ribosomes (Marc et al. 2002), Predotar (Small et al. 2004), and yeast proteins with known human mitochondrial orthologs (MitoP2 database). High-throughput protein–protein interaction datasets (PPI) were combined and divided into confidence classes (von Mering et al. 2002). Medium and high confidence PPI datasets were defined by interactions with known mitochondrial proteins (MitoP2 database). The predictive score for a mitochondrial protein (MitoP2) was based on the integration of 22 datasets, most of which are shown, and was calculated for different thresholds. Specificity and sensitivity are current best estimates owing to the incompleteness of the reference set.
Figure 5.
Verification of Proteomic Candidates by Mitochondrial Import
Samples were incubated in the presence or absence of a membrane potential (MP) and of proteinase K (PK). Cases where import was accompanied by removal of the signal peptide (SP) are marked as “SP-processing” (+). Su9(1–69)DHFR and AAC serve as positive controls for a processed matrix protein and a nonprocessed inner membrane protein, respectively. The bar graphs indicate if a protein was more likely to be found in either the membrane or the matrix fractions of our proteomic data. The height of the bar corresponds to the number of samples in which a protein was identified with higher tag number—in the mitochondrial membrane or mitochondrial matrix fractions, respectively.
Figure 6.
Verification of Prediction in Selected Mitochondrial Protein Complexes
The assignment of complexes to mitochondrial compartments is based on known localizations of the protein subunits. Complexes are shown as clusters of circles, where each circle represents one protein. Red denotes a protein that was detected under fermentable and green under nonfermentable growth conditions by our proteomic dataset; white indicates proteins that were not detected. The numbers indicate the MitoP2 predictive score. For proteins without a number, no predictive score was assigned by the integrative analysis. Ac, acetyl; CoA, coenzyme A; α-KG, α-ketoglutarate; GDC, glycine decarboxylase; NDH, NADH-oxidoreductase; OAA, oxaloacetate; PDH, pyruvate dehydrogenase; RCC, respiratory chain complex; TIM, transport across inner membrane; TOM, transport across outer membrane; MOM and MIM, mitochondrial outer and inner membrane, respectively. A list of the genes for the plotted complexes is available in Table S4.