Protein-Protein Interactions

These pages have been contributed by Cosmin Saveanu (Institut Pasteur, France)

Cells need interactions
Imagine a cell in which, suddenly, the specific interactions between proteins would disappear. This unfortunate cell would become deaf and blind, paralytic and finally would disintegrate, because specific interactions are involved in almost any physiological process. Sensing extracellular signals is a matter of receptor to adaptor interactions and the shape of the cell is maintained by an intricate network of structural protein interactions. Finding interactions between proteins involved in common cellular functions is a way to get a broader view of how they work cooperatively in a cell. In the following paragraphs, for reasons of brevity, I will call protein-protein interactions any type of meaningful interaction that proteins are capable of (protein-RNA, receptor-ligand, etc.)

Functional genomics "puzzle"

Even if I feel tempted to place the techniques that identify protein-protein interactions at the heart of the functional genomics puzzle, a balanced view is more accurate. Bioinformatics helps scientists to handle the vast amount of data generated by different genome-wide approaches. Interesting conclusions about genome evolution and the relationships between genes and function may be drawn by comparative genomics. Predictions of function, based on genome sequence analysis, will become more and more accurate as the number and diversity of sequenced genomes steadily increases. A recent review of computational methods for function hunting is offered by Pellegrini (2001). Systematic deletions (physical or by RNA interference analysis, see for example Kamath et al., (2000) and systematic localization studies are first steps in functional characterization of genes and may be of great help in improving the quality of protein interaction maps. For example, a two-hybrid interaction between a nuclear and a mitochondrial protein is not likely to be biologically significant (even if it could be, but that's another story). Clustering of genes by mRNA expression profiles (a comprehensive study of expression profiles in yeast under a variety of conditions was described by Hughes et al., (2000)) seem to be an effective way to predict protein function. Protein quantitation and differences in protein expression under different experimental conditions may be analysed by 2D-electrophoresis but the ICAT technique that makes extensive use of mass spectrometry might be more effective in proteome-wide projects (Gygi et al., 1999). I should also mention here, the equivalent of DNA chips in the protein world and the projects of constructing libraries of specific antibodies raised against every single protein of a cell (for a review see Holt et al., 2000). Biochemical genomics, where the entire ORFeome (see Reboul et al., 2001 for the term ORFeome) of an organism is cloned, expressed and individually purified, allow the discovery of proteins when a biochemical assay for their function has already been developed (for a review see Grayhack & Phizicky, 2001). By a different approach, structural genomics is trying to decipher biochemical protein function by structural similarity to known proteins, however this method is limited to predicting the cellular function of a given protein. Not mentioned in the previous picture but important as always are the tools of genetics, and genetic screening is a continuous source of marvelous discoveries.

What are we going to study today?
In the pre-genomic era it was quite difficult to ask questions like: "I have this gene, looks nice, has a zinc-finger motif, let's see what the function is...". However this type of "fishing expedition" (quote from Marc Vidal's review in Cell, 2001) may be quite succesful even if, as one of my colleagues told me: "this is not like are not asking a functional question...". In fact, the hunting for function may reveal interesting, and sometimes unexpected results. Additionally, by the development of automated, large scale projects an important amount of functional data may accumulate to be used by the scientific community.

Alternative ways of studying protein function

One, two, three...infinity
I have the sequence of a predicted protein and I would like to know if it collaborates with other proteins in a functional pathway (and I hope that additionally my protein is THE essential one in this pathway). One possibility is to test whether my protein:
1. Has a reasonable level of expression.
2. Is associated to a complex in which at least one of the components has a known function.

For these purposes, I will isolate the protein fused with an affinity tag, either by fine-tuning the conditions of isolation such that the protein is highly enriched in a single affinity purification step or by using a "generic" method such as that described by Rigaut et al., 1999 under the name of TAP (Tandem Affinity Purification). The figure below shows different kinds of results when yeast proteins of unknown function were tested by TAP isolation. Two successive affinity purifications are done using two tags fused to the protein of interest; the first step involves binding of the protein A tag to an IgG column. Under mild conditions the putative complex is released from the column by a specific viral protease and rebound to a second column, where calcium dependent interaction of calmodulin with a calmodulin binding peptide (CBP) will allow the enrichment of the complex. Recently, a similar technique was used in a multiple affinity purification (MAFT) using three different tags in the succesion: calmodulin binding peptide, histidines, hemagglutinin (Honey et al., 2001).

Number of proteins isolated by using the TAP strategy in yeast with different unknown proteins as "baits"

It is a good idea to test whether the depletion of the studied protein gives rise to a particular phenotype. If the tagged protein used for complex purification is functional, this phenotype should not be observed. That's why it is an advantage to study one particular protein by the affinity purification of the associated complex in a yeast strain where the functional protein is essential. This way, we can be sure that the tag did not interfere with the protein function and that the purified complex might be the 'physiological one'.

Two hybrids are better than one
A long time ago, in a galaxy far, far away... the two-hybrid system was described in yeast (Fields and Song, 1989). The original two-hybrid underwent many technical modifications but the original idea of reconstructing a functional factor (transcriptional, enzymatic, signal transduction) from two pieces that come close together, being in "hybrids" with proteins that interact specifically, is always present (see figure below).

If an interaction is established between the two hybrid proteins the transcription of a reporter gene is activated and one can select the cells by testing the activity of the reporter protein

Reconstituting a functional transcription factor requires that the hybrids are transported into the nucleus and it was only ten years after the original description of the two-hybrid system that a nuclear localization signal was uncovered in the sequence of the bacterial LexA, the bait DNA binding domain (Rhee et al., 2000). Moreover, a one hybrid system with a mutated LexA sequence was developed. In this system a protein of interest is fused with both LexA and Gal4 activator domain reconstituting a stable transcriptional activator. However this activator is not functional if it is not transported through the nuclear pore complex at the site of transcription. If the protein of interest has a nuclear localization signal (NLS) it will be transported and the activation of transcription will take place.

(page last updated: August 15th, 2001)