Canvassing Protein Complexes

Two yeast studies begin to identify protein interactions on a genome-wide scale.

Sep 1, 2008
David Secko
<figcaption>Protein complex in yeast cytoplasm that contributes to mRNA decay. Credit: © embl</figcaption>
Protein complex in yeast cytoplasm that contributes to mRNA decay. Credit: © embl

The drama of biology is played out through thousands of protein-protein interactions. Historically, researchers could only examine these interactions one by one, but genomic sequences and high throughput methods have opened up the "interactome" - the complete list of all protein interactions in an organism.

With a sequenced eukaryotic genome, lists of every protein, and genetic amenability, Saccharomyces cerevisiae was an obvious model with which to examine cell-wide protein interactions. In 2000, Peter Uetz and colleagues, then at the University of Washington, made the first attempt at systematically mapping protein-protein interactions in yeast, reporting 957 putative interactions from 1,004 yeast proteins.

A smattering of similar studies soon followed, but they shared a flaw: the use of expression vectors that overproduced tagged proteins. It became clear that future work needed to maintain endogenous protein levels, says Jack Greenblatt, from the University of Toronto.

In 2006, back-to-back Hot Papers did this using tandem affinity purification (TAP). The method involves inserting a 'tag' by homologous recombination at one end of every yeast gene, with resulting proteins purified and identified by mass spectrometry. Tagged proteins pull down complexes of interacting proteins that together give a broad view of the interactome. Giulio Superti-Furga's group at Cellzome and colleagues tagged all 6,466 open reading frames in S. cerevisiae, purified 1,993 unique proteins, and found 257 novel protein complexes.1 The other study, from Greenblatt's group, purified 2,357 proteins and catalogued 547 protein complexes, with again about half being novel.2 Although each team used different algorithms to define complexes based on mathematical models, each complex had at least two proteins. And the studies found the average number of proteins in a complex to be 3.1 and 4.9 proteins, respectively.

"The two studies are definitely milestones," says Uetz, who is now at The J. Craig Venter Institute and was not involved in either Hot Paper. Researchers have gone on to scrutinize some novel complexes, expand their efforts to map still unknown protein complexes, and continue to debate the biological relevance of the uncovered interactions.

Defining complexes

The TAP method has complications. Protein complexes can vary in composition depending on which tagged proteins researchers use for purification, because experimental conditions can vary, and the effect of the tag on different proteins can vary. And proteins can also be part of several different complexes. Plus, "most of the yeast proteins were forming protein complexes," says Anne-Claude Gavin, first author of the Superti-Furga study. "This seemed to be the rule rather than the exception." Add it up and researchers need to interpret raw mass spectrometry data to determine which proteins are truly parts of a protein complex.

Illustrating this, despite the similar methodology used by the Hot Papers, "the outcomes were surprisingly different," Uetz says. In a comparison of the two, he found only six identical protein complexes.3

However, "most of the divergence turned out not to be in the mass spectrometry data," says Greenblatt, but in the mathematical models. Greenblatt and his colleagues later compared both sets of raw purification data with a novel approach based on the hierarchical clustering of protein-protein interactions and strength of protein interactions, finding greater agreement between the studies.4

Despite the discrepancies, the genome-wide interaction data has proved insightful. For example, Greenblatt's study uncovered the yeast histone acetyltransferase Rtt109 in a novel complex with histone chaperone Vps75. Previous genetic studies showed that Rtt109 works with the chaperone Asf1, "raising the question of why Vps75 is associated with [Rtt109]," says Greenblatt. Recently, the group found that Rtt109 needs Vps75 to carry out a novel histone acetylation.5 Similarly, the 26 European lab ?3D Repertoire' project is using data from the first Hot Paper with the aim of solving the structures of all the protein complexes in S. cerevisiae by electron microscopy, X-ray crystallography, and in silico methods.

Tip of the interactome

A good deal of uncharted territory on yeast protein complexes still exists, says Stephen Michnick, from the University of Montreal. Michnick chalks this up to technical issues, notably that extraction conditions are not always optimal. But, "the point of these large-scale studies isn't to be comprehensive," he says, "but to get a general idea about the overall organization of protein interaction networks and what it tells us about the way the cell works."

Michnick's group recently developed a protein-fragment complementation assay (PCA) to examine protein-protein interactions in vivo. Where other techniques give physical interactions, PCAs can be used to study changes in interactions in intact cells. "We wanted to be able to see the dynamics of these interactions under particular conditions," says Michnick. In a genome-wide screen in yeast, they found 2,770 interactions involving 1,124 proteins, 80% of which had never been reported.6

Recent studies estimate the number of yeast protein-protein interactions to be 25,000-30,000.7 In humans the estimate rises to 650,000. If the numbers are accurate, it means that less than 0.3% of protein interactions have been identified. For Michnick, this suggests that: "We are going to need a wide range of techniques to drill down on cellular processes of this complexity."


A-C. Gavin et al., "Proteome survey reveals modularity of the yeast cell machinery," Nature, 440:631-6, 2006. (Cited in 300 papers) N. Krogan et al., "Global landscape of protein complexes in the yeast Saccharomyces cerevisiae," Nature, 440:637-43, 2006. (Cited in 288 papers) 1. A-C. Gavin et al., "Proteome survey reveals modularity of the yeast cell machinery," Nature, 440:631-6, 2006. (Cited in 300 papers) 2. N. Krogan et al., "Global landscape of protein complexes in the yeast Saccharomyces cerevisiae," Nature, 440:637-43, 2006 (Cited in 288 papers) 3. J. Goll and P. Uetz, "The elusive yeast interactome," Genome Biol, 7:223, 2006. 4. S. R. Collins et al., "Toward a comprehensive atlas of the physical interactome of Saccharomyces cerevisiae," Mol Cell Proteomics, 6:439-50, 2007. 5. J. Fillingham et al., "Chaperone control of the activity and specificity of the histone H3 acetyltransferase Rtt109," Mol Cell Biol, 28:4342-53, 2008. 6. K. Tarassov et al., "An in vivo map of the yeast protein interactome," Science, 320:1465-70, 2008. 7. M. P. Stumpf et al., "Estimating the size of the human interactome," Proc Natl Acad Sci, 105:6959-64, 2008.