A deep look into the mitochondrial "proteome"

Genes, like people, often flock to the center of life's hustle and bustle. But this kind of "suburban flight" — in particular, leaving the mitochondria to be city slickers in the cell's nucleus — has left genes unevenly distributed and created questions about the logistics of the arrangement.

Now, an international team of scientists reports progress in deciphering these lopsided demographics, with the use of an integrated, genome-wide approach to enlarge the current catalogue of human proteins that are contributed by the nucleus to its mitochondrial neighbors. As described in the April 2 online edition of Nature Genetics, this expanded list — the most comprehensive one to date — enabled the identification of causative mutations in a nuclear gene, MPV17, which underlie a human mitochondrial disease. Because the lion's share of mitochondrial diseases result from aberrations in nuclear genes, rather than mitochondrial-based genes, this work will help to unearth additional gene culprits and to advance our understanding of mitochondria in normal biology and disease.

Mitochondria, the tiny energy-producing factories within eukaryotic cells, come equipped with their very own genome. But this gene collection, a vestige of the organelles' evolutionary past, is dwarfed by the one housed in the nucleus and contributes only a fraction of the proteins needed for proper mitochondrial function. Accordingly, mitochondria selfishly gather most of their proteins from the genetic code of their cellular companion. Yet these imports are not easily identified simply by their molecular characteristics, which makes it difficult at best for scientists to compile an accurate list of all the mitochondria's proteins. This endeavor, an important prelude to determining the molecular basis of diseases associated with mitochondria, remains unfinished: Of the ~1,500 proteins predicted to function in mitochondria, only half have so far been identified.

To expand our knowledge of the components of the mitochondrial "proteome," Broad scientists developed an integrated approach that combined eight genome-wide computational methods under an inclusive program coined "Maestro." Each method individually queries sequence and expression data for the full complement of human genes and proteins. Some look for structural hallmarks of mitochondrial proteins or for short protein sequences that, based on experimental studies, are localized within mitochondria. Others search for DNA sequence similarities, either to known mitochondrial genes in yeast, or to genes in Rickettsia prowakzekii, a relative of the bacterial predecessor of human mitochondria. The remaining methods screen for specific trends in gene expression and in gene regulatory sequences that suggest a link to mitochondria.

"Our work illustrates the strength that combining diverse computational methods can deliver to genomic studies, particularly for discovering the full suite of proteins in mitochondria," said Vamsi Mootha, the senior author of the study, an assistant professor at Massachusetts General Hospital and Harvard Medical School, and an associate member working with the Broad's Metabolic Disease Initiative. "With hundreds of new additions to the mitochondrial proteome, this approach has already had important applications to human disease, which will certainly be expanded in the future."

On its own, each component of Maestro proved weak in pinpointing mitochondrial proteins, but when employed in concert, the approach proved dramatically more effective. This indication first stemmed from the examination of two "training" sets: a group of proteins known to function expressly within mitochondria and a group of proteins active elsewhere in the cell. Maestro accurately predicted 71% of the known mitochondrial proteins. When subjected to the full array of human proteins, numbering more than 30,000, Maestro earmarked nearly 800 additional proteins for mitochondria. Roughly half of these proved novel, with no prior link to mitochondria. The researchers verified a handful of these novel predictions using both computational and experimental methods, and found evidence of mitochondrial localization for 18 proteins, out of the 19 analyzed. This improved inventory of the mitochondria's proteins, which now stands at 1,451 proteins (represented by 1,080 genes), signifies the most comprehensive one so far.

The scientists, including first author Sarah Calvo, a first-year computational biology PhD student in the Harvard-MIT Division of Health Sciences and Technology, sought to determine the utility of this approach, specifically in prioritizing large genomic regions associated with a disease. Such regions can often include hundreds of genes and, for mitochondrial diseases in particular, could be more quickly winnowed if the genes were ranked by their likelihood of being localized to mitochondria. Therefore, the researchers applied Maestro to eight mitochondrial diseases, for which a candidate region, but yet no specific gene, had been identified. In each case, the approach narrowed the candidate list to a testable number of likely mitochondrial genes.

The researchers pursued one of these diseases, called hepatic mitochondrial DNA (mtDNA) depletion syndrome, in a collaboration with Massimo Zeviani at the National Neurological Institute in Milan, Italy. Upon sequencing a series of candidate genes, which lay in the chromosomal region previously linked to the disease, they discovered mutations in MPV17 in patients from three unrelated families. As described in an accompanying article in Nature Genetics, further experimental work supports the role of MPV17 in hepatic mtDNA depletion syndrome and in particular, predicts that its gene product resides within the inner membrane of the mitochondria. Earlier studies had mistakenly pegged the protein a resident of the peroxisome, so researchers may not have been inclined to otherwise investigate it, were it not for Maestro's skill in directing scientists in the right direction.