This may be music to your ears. Researchers in the U.S. and Italy have developed a new tool for identifying genes that encode mitochondrial proteins linked to disease. Vamsi Mootha, at the Broad Institute of MIT and Harvard University, and colleagues have orchestrated eight different data sets used to predict mitochondrial sequences into a single suite, dubbed Maestro, that performs much better than any of the individual methods alone. Maestro delivers few false positives and has already helped to identify the gene responsible for a rare, mitochondrial DNA depletion syndrome that leads to liver damage, neurologic problems, and multiple brain lesions. Maestro and its first big hit are detailed in two papers in the April 2 Nature Genetics online.

Through their roles in basic metabolism, apoptosis, and the generation of reactive oxygen species, mitochondria have been linked to various neurodegenerative diseases including Alzheimer and Parkinson diseases. In fact, Parkinson disease (PD) can be induced by chemicals that inhibit mitochondrial respiration (see ARF related news story), and mutations that cause inherited forms of PD have been found in Pink1, DJ-1, and parkin, proteins that have been linked to the mitochondria (see ARF related news story). In the case of Alzheimer disease, increased oxidation of mitochondrial DNA in patient tissue samples also hints of a mitochondrial connection to the disease (see ARF related news story and ARF live discussion).

But there are obstacles to studying mitochondrial disease etiologies, the principal one being that we do not know enough about the proteins that localize to mitochondria. Hidden among the 30,000-plus genes of the human genome, there are a predicted 1,500 or so that code for proteins that are delivered to these organelles. But as Mootha and colleagues point out, only about half of the 1,500 have been identified. This is because there are no clear ways to finger mitochondrial proteins. While the presence of N-terminal sequences that guide delivery to the organelles can be a dead giveaway, those signal sequences are not always easy to detect and, in fact, are absent from many mitochondrial proteins that are imported by other mechanisms. Fortunately, there are other clues that pinpoint mitochondrial proteins, and the authors put these together to develop Maestro.

First author Sara Calvo and colleagues combined scores from eight data sets to arrive at a Maestro score that predicts if a gene codes for a mitochondrial protein. The eight data scores include mitochondrial signal sequence score; protein domain score—based on whether domains in the protein are exclusively mitochondrial; cis-motif score—based on the presence of promoter regions common to nuclear-encoded mitochondrial genes; yeast homology score—derived from experimental evidence of mitochondrial localization in yeast; ancestry score—based on similarity to proteins in the bacteria Rickettsia prowazekii, the closest living relative of mitochondrial ancestors; coexpression score—reflecting simultaneous expression with known mitochondrial genes; a tandem mass spectrometry score—based on the number of tissues in which the protein was detected in the mitochondrial proteome from the mouse; and induction score—derived from whether or not the gene is upregulated in a model of mitochondrial biogenesis.

When Calvo and colleagues ran Maestro against more than 33,000 proteins in the human genome, it predicted 1,451 proteins were mitochondrial, with an estimated false discovery rate of only 10 percent. Of the predicted genes, 490 had no apparent prior link to mitochondria, and so are novel predictions. These include a large number of uncharacterized proteins.

To test how good Maestro is, Calvo and colleagues put 39 predicted proteins to the test—30 in a mass spectrometry analysis of purified mitochondria from mouse liver, and nine in a green fluorescent protein-tagged expression test. The mass spectrometry data revealed that all 30 predicted proteins were indeed found in liver mitochondria, while in the expression test, eight of the proteins were visible in the organelles. The one protein that was not detected in the latter test also had the lowest Maestro score.

To determine if Maestro may be useful in pinpointing mitochondrial proteins involved in disease, Calvo and colleagues compared Maestro scores with known mitochondrial disorders that have been linked to large genomic intervals, but for which no reliable gene candidate has yet emerged. Maestro delivered candidates for eight different mitochondrial diseases, including three known and two novel candidates for Friederich Ataxia 2, a disease most commonly caused by expansion of a GAA trinucleotide in the gene frataxin, but which has other, unknown causes.

Collaborators Massimo Zeviani and colleagues at the National Neurological Institute "C. Besta," Milan, Italy, took this analysis one step further. In the second paper, first author Antonella Spinazzola used Maestro to prioritize genes on human chromosome 2p21-23 that have been linked to a hepatocerebral mitochondrial DNA depletion syndrome (MDDS).

Maestro analysis of this region identified the human homolog of the mouse Mpv17 gene as being a mitochondrial gene and therefore a potential disease candidate. When Spinazzola and colleagues analyzed DNA from three MDDS families, they found that three different mutations in Mpv17 segregate perfectly with the disease. In two families, MDDS is caused by single homozygous point mutations (149G-A and 498C-A), while in the third family the disease is caused by a single point mutation in one allele (148C-T) and a 25bp deletion (116-141) in the other.

The role of Maestro in identifying these mutations is even more impressive, given that Mpv17 has been described as a peroxisomal protein and therefore may have been passed over as being involved in a mitochondrial disease. Spinazzola and colleagues found that the protein is, in fact, delivered to the inner mitochondrial membrane where it is essential for oxidative phosphorylation. Without it, mitochondrial DNA is not maintained in individuals with MDDS or in Mpv17-negative mice.—Tom Fagan

Comments

No Available Comments

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. A New Link Between Pesticides and Parkinson's Disease
  2. Pink Mutations Link Parkinson’s Disease to Mitochondria
  3. Promoter Bashing—Mitochondrial Ones Damaged in AD Brain

Other Citations

  1. ARF live discussion

Further Reading

No Available Further Reading

Primary Papers

  1. . MPV17 encodes an inner mitochondrial membrane protein and is mutated in infantile hepatic mitochondrial DNA depletion. Nat Genet. 2006 May;38(5):570-5. PubMed.
  2. . Systematic identification of human mitochondrial disease genes through integrative genomics. Nat Genet. 2006 May;38(5):576-82. PubMed.