by Jeffrey M. Perkel
"By their friends, you shall know them," the saying goes. The same is true of proteins. The majority of cellular proteins are cogs in large macromolecular machines comprising anywhere from two to dozens of members. For those tasked with figuring out the function of novel polypeptides, identifying a protein's associates can often reveal something of its function.
There are a variety of approaches researchers can use to map protein-protein interactions, each with its own unique strengths and weaknesses. Some work in vivo (that is, they report interactions that occur inside the cell), others in vitro (identifying interactions that form when purified proteins or extracts are mixed). Some require prior knowledge of potential partners, while others are unbiased. The best approach, says Ryan Bomgarden, research scientist in R&D for Thermo Fisher Scientific, is to cover your bases.
"There isn't any perfect method for protein-protein interactions," says Bomgarden. "Researchers need several different techniques in their arsenal."
Chemical crosslinking
One approach to detecting protein-protein interactions involves physically coupling the two proteins with a chemical crosslinker. Thermo Fisher Scientific offers a variety of such tools through its Pierce Protein Research Products division.
Last fall, the company launched a line of photo-activated crosslinking reagents that are incorporated into proteins in vivo. L-Photo-Leucine and L-Photo-Methionine, analogs of leucine and methionine, respectively, contain a diazirine ring in the amino acid side chain that covalently attaches to associated proteins when irradiated with ultraviolet light.
The reactive group is a "zero-length crosslinker," Bomgarden explains: it does not increase the amino acids' side-chain length. As a result, it should have minimal impact on protein folding and interactions.
"It is an intimate crosslinker," he says, "it potentially reduces false positives by only crosslinking within protein-protein interaction domains."
The analogs are incorporated into proteins during translation in vivo, using a specially formulated growth medium lacking the two amino acids. UV irradiation of the culture crosslinks proteins prior to cell lysis; the crosslinked proteins can then be analyzed by SDS-PAGE and Western blotting.
"This is just an exciting technology, because you are tricking the translation machinery of the cell to build proteins that contain these photoreactive groups," Bomgarden says.
Pull-down assays
Another way to detect protein-protein interactions is the pull-down assay. In coimmunoprecipitation, for instance, an antibody to one protein, physically coupled to a bead, is used to isolate that protein -- and any other protein(s) with which it might be associated. These proteins can then be identified by SDS-PAGE, Western blotting, or mass spectrometry.
Promega offers a range of pull-down products, including the MagneGST Pull-Down System (based on the interaction between glutathione S-transferase (GST) and glutathione-coupled beads) and its HaloTag product line.
HaloTag is a 33 kD modified enzyme. When a HaloTag fusion protein (a fusion of the tag with some protein of interest) is incubated with HaloLink Resin, which contains the enzyme's ligand, the enzyme covalently couples to the bead, producing a far stronger interaction than between, say, GST and glutathione or a polyhistidine tag and nickel. Coming along for the ride are any proteins that are interacting with the fusion protein (interactions that, whether generated in vivo or in vitro, remain non-covalent). These interacting partners – which may be associated with the fusion protein directly or through neighbors – can be identified via Western blotting or mass spectrometry.
The obvious implication, says strategic marketing manager Pat Bresnahan, is that washing and purification steps can be much more stringent. The tag also can be used to build immobilized protein arrays. "But for visualization, I think the implications are probably the strongest," she says, "because with covalent attachment, you can label a protein in vivo and follow it through the cell." Indeed, Promega offers a variety of fluorescent HaloTag ligands, which can be used, for instance, in pulse-chase experiments.
Surface plasmon resonance
Another approach to in vitro interaction detection involves surface plasmon resonance (SPR). In SPR, a laser is directed through a prism at a thin gold film at an oblique angle such that all the energy in the beam is transferred to the film, creating a standing evanescent wave. The resulting loss of reflectivity is called the "SPR dip," says Nguyen (Win) Ly, R&D director at Biosensing Instrument (BI), one of the companies that sells instruments based on this principle (others include GE Healthcare's Biacore division and Horiba Jovin Yvon).
If the film is coated with a protein, the angle at which the dip occurs will change. And if a second test protein interacts with the first, the film's reflective properties, and thus the SPR angle, will change yet again. Thus, protein-protein interactions on the gold surface can be detected as changes in the film's SPR characteristics.
"If you held everything constant but changed the refractive index of the media on the film, then the dip will move, the angular position will move," Ly explains. "So if a protein binds to the sensor surface, its additional mass will result in a change of the effective refractive index."
The advantage is that the process is both fast and label-free; there's no need to build fusion proteins or apply covalent tags. But SPR is also decidedly unbiased – you must generally know which proteins you are working with (though it is possible to probe the chip following SPR via mass spectrometry, to determine just what molecules bound).
Fluorescence resonance energy transfer
Moving in vivo, one approach to detecting protein-protein interactions is FRET, or fluorescence resonance energy transfer. In FRET, two proteins, tagged with different fluorophores, are expressed in a cell. If those two proteins interact in the cell, bringing their fluorescent domains into close proximity, and if the fluorophores are chosen properly, then irradiation of one dye can lead to excitation of the other, resulting in a detectable loss of excitation energy.
"A common misconception [about FRET] is you transfer energy by light," explains Jim Mattheis, fluorescence applications manager at Horiba Jobin Yvon, which offers a number of FRET-capable instruments. "It is actually a through-space energy transfer."
In other words, instead of exciting the second dye with light emitted from the first, energy is directly transmitted from one dye to the other. "It’s like a Wi-Fi connection with a very tight spatial requirement," he says. "There is no radiation of light. FRET is almost a magical event."
Yeast two-hybrid
But when it comes to detecting interactions in vivo, the gold standard is the yeast two-hybrid (Y2H) assay, or one of its many variants. Y2H is based on the modular design of eukaryotic transcription factors, which often have separate DNA-binding and transcriptional activation domains.
Suppose a researcher suspects two proteins interact, and wants to know if that occurs in vivo. In the classic Y2H experiment, the researcher creates two fusion constructs: protein A with a DNA binding domain, and protein B with a strong activation domain. Both constructs are expressed inside a single yeast cell, where they will translocate to the nucleus. In the absence of an interaction between A and B, a reporter gene, controlled by transcription factor binding sites to match the DNA-binding domain, will remain silent, as there will be nothing to recruit the transcriptional activation domain to the promoter. But if A and B interact, the transactivation domain will be recruited to the reporter gene's promoter, turning the gene on.
The advantages of this approach are that it occurs in vivo and is amenable to high-throughput screening. The disadvantages are that Y2H occurs in yeast (which may not be the origin of the proteins in question, and which may produce different post-translational modifications); the proteins are overexpressed; it forces proteins into the nucleus (membrane-bound proteins cannot be used); and it precludes analysis of transcription factors themselves, or any other protein that can autoactivate the reporter in the absence of a binding partner.
Several companies have devised workarounds for these shortcomings. Promega's CheckMate system, for instance, works in mammalian cells, not yeast, and so is more likely to have the proper folding and post-translational modifications. But it still cannot work with membrane proteins, and still requires expression in the nucleus. More to the point it, like most mammalian systems, is not really amenable to library screening, and so is typically used to confirm Y2H results, says Bresnahan.
Dualsystems Biotech offers several Y2H-based systems designed to circumvent these problems. Dualsystems' DUALmembrane and DUALhunter systems are both based on the so-called split-ubiquitin system.
In the split-ubiquitin system, ubiquitin (a small cellular protein that targets other proteins for destruction) is split into two pieces, disabling it. If these pieces are then coupled to interacting proteins, that association reconstitutes ubiquitin, enabling a specific protease to cleave it. That cleavage reaction releases an attached transcription factor, which then migrates to the nucleus to activate transcription, just as in the classical Y2H.
"The nice thing is you can detect this interaction at the membrane and get readout in the nucleus," says chief executive officer Daniel Auerbach. "So you combine the advantages of yeast two-hybrid with the fact that you can screen in a compartment outside the nucleus."
DUALmembrane probes for interactions between two membrane-bound or associated proteins; DUALhunter forces cytosolic proteins to interact at the membrane by coupling one partner to the membrane (as a result, even autoactivator proteins can be screened by this method).
The final word?
Yet that's hardly the final word in interaction research. For one thing, there are many other available approaches, including protein arrays (in vitro) and FRET-based assays (in vivo). According to Marc Vidal, associate professor of genetics at the Dana-Farber Cancer Institute, who uses Y2H extensively in his work mapping protein-protein interaction networks in eukaryotic cells, no one has ever systematically put all the various interaction methods through their paces to see which works best.
Vidal has taken a stab at it, however. He and his team used Invitrogen's recombinase-based Gateway technology to clone 200 well-characterized (positive control) interaction pairs from human, worm, yeast, and Arabidopsis, as well as 200 negative-control pairs. Then they tested these 400 interactions using five different approaches, including Y2H (but none of the other techniques described above).
"The bottom line was that none of them can detect all positive interactions, which is not really unexpected," he says. Instead, each detected between one-third and one-fourth of the expected interactions, overlapping sets that in aggregate recapitulated the entire set of positive controls.
"They have overlapping profiles," he says, "so use them all and use Gateway so that your favorite genes are cloned once and for all, and then run them all to remove bias from your work."