For an organelle that once seemed to be loved only by hard core biophysicists, the mitochondrion has come a long way. Severe disorders of mitochondrial oxidative phosphorylation (OXPHOS) are now recognized as the most common group of inborn errors of metabolism, affecting at least 1 in 5,000 individuals1. Mitochondria are best known as the cell's energy source as producers of ATP. But, they also have pivotal roles in generating reactive oxygen species, calcium metabolism and cell death. Thus, it is not surprising that mitochondrial dysfunction contributes to diverse pathologies including neurodegeneration, diabetic complications and tumorigenesis. A recent study by Vamsi Mootha and colleagues2 in Cell provides a basis for connecting mitochondrial pathologies with molecular etiology by identifying new mitochondrial proteins and profiling the extent of their tissue-specific diversity.

Compiling the list

Over the past few years, different strategies have been used in attempting to determine the total number of proteins in mitochondria. These include epitope-tagging, systematic functional screening of whole-genome pools of mutants and proteomic analyses of highly purified mitochondria. Bioinformatic analyses have also been used to predict proteins with a classical mitochondrial targeting sequence or genes that are coregulated with genes encoding known mitochondrial proteins. These complementary approaches are necessary to overcome the various limitations in sensitivity and specificity of each method. Recent studies using such approaches in yeast are consistent with an estimate of 800–1,000 different mitochondrial proteins3,4,5,6.

Yeast mitochondria are relatively simple and uniform, but mammalian mitochondria vary widely between different tissues and often don't resemble the text-book version (Fig. 1) in morphology or composition7 (Table 1). The mitochondrial proteome of mammals is probably much larger and more diverse than that of yeast, but its size and variability have not been ascertained. Computational analysis predicts up to 4,000 mitochondrial proteins in humans3, which may be a true reflection of the complexity and diversity of our mitochondria or an overestimate caused by a systematic artifact. The first large-scale proteomic analysis of mammalian mitochondria, using human heart tissue, was published recently and concluded that at least 615 different proteins were present8.

Figure 1
figure 1

© Photo Researchers, Inc.

Colored high resolution scanning electron micrograph of a single mitochondrion in the cytoplasm of an intestinal epithelial cell.

Full size image
Table 1 Examples of the extent of mitochondrial variation between tissues7
Full size table

Mootha and colleagues now report a proteomic survey of mitochondria purified from mouse brain, heart, kidney and liver2. They identified 399 mitochondrial proteins using liquid chromatography tandem mass spectrometry. Of these proteins, 75 were new and 88 had recently been identified as mitochondrial proteins8. Although some could be contaminants, most seem authentic based on multiple lines of evidence. Mootha et al. compiled these proteins with known mitochondrial proteins and obtained a list of 591 distinct mitochondrial proteins, called, in Hollywood parlance, the mito-A list.

Variation and cooperation

Only 40% of the previously known mitochondrial proteins were present in all four mouse tissues, and after correcting for imperfect sensitivity and reproducibility, it was predicted that ∼85% of the mitochondrial proteins detected in one tissue would be detected in a different tissue. Analysis of published RNA expression levels also suggested that about half the mitochondrial genes were expressed in all four tissues. The protein and RNA data were consistent with a simple model in which half the proteins in the mitochondrial proteome are ubiquitous and half are tissue-specific, with a 50% probability of being expressed in a given tissue. The true situation is undoubtedly more complex, with the tissue-specific component likely to include some proteins expressed in perhaps a single tissue and others expressed in most but not all tissues. But this model gives a framework on which to base further investigations.

For two-thirds of the mito-A genes, previously published expression data were available for 45 mouse tissues, and a number of clusters or subnetworks of genes were identified with correlated expression. Not surprisingly, the largest cluster was enriched for genes related to OXPHOS. This group would be expected to have quantitative, rather than absolute, differences in expression and was most prominent in tissues with the highest oxidative capacity. Other clusters corresponded to tissue-specific metabolic pathways, such as steroidogenesis in adrenal cortex and heme synthesis in bone marrow. Clustering allows preliminary functional annotation of proteins with unknown or unanticipated roles; for example, the OXPHOS subnetwork included 11 proteins not previously associated with OXPHOS.

Analysis of published expression data for 10,000 nuclear and mitochondrial genes identified 470 non-mito-A genes whose expression profiles were most similar to genes in the mito-A list; comprising the 'mitochondrial neighborhood'. Many of these probably encode authentic mitochondrial proteins not detected by proteomics, and others may encode nuclear transcription factors, cytosolic chaperones and other proteins and RNAs whose expression is linked to mitochondrial genes. Notably, and in keeping with the authors' previous results from a human expression data set9, the mitochondrial neighborhood contains only about half of the mito-A genes. This might be expected if there are large numbers of mitochondrial proteins expressed in a relatively unique manner (e.g., with high tissue-specificity) shared by only small numbers of other mitochondrial proteins.

Deducing origins

A popular theory for the origin of mitochondria holds that they are derived from a eubacterial symbiont. Some mitochondrial proteins are thought to have derived from genes introduced by this ancient bacterial invader to the host nuclear genome, whereas others are thought to have been recruited to the evolving mitochondrion from the host genome to allow for protein import, ATP export and other (tissue-)specific functions. The ancestral proteins in the mito-A list—those with homologs in eubacteria—tended to share their pattern of gene expression with more mitochondrial proteins than others, that is, they had larger numbers of close neighbors. Previous studies suggest that, at least in yeast, these ancestral proteins are more likely to be translated on polysomes in the immediate vicinity of mitochondria than on free cytosolic polysomes10. So, it seems that a core group of ubiquitous mitochondrial proteins may have retained a distinct mechanism of coregulated gene expression.

Combining the two recent proteomics studies results in a list of more than 700 proteins in the mammalian mitochondrial proteome2,8. Further studies are clearly needed to define the true number, which is probably at least twice this value, given the extent of differences between different tissues, the likelihood that many proteins have low abundance and the number of potential mitochondrial proteins predicted by gene expression analyses. Thus, it is reasonable to conclude that at least 5% of mammalian genes, and perhaps up to 10%, encode proteins located in the mitochondria of one or more tissues.

The expanding spectrum of mitochondrial proteins offers a new range of candidate genes for involvement in OXPHOS disorders and risk factors for many common diseases. Analysis of coregulated mitochondrial genes, and the transcription factors regulating the networks, also offers an approach to understanding the pathogenesis of mitochondrial dysfunction in conditions such as diabetes11.