BiFC (bimolecular fluorescence complementation) is a tool for investigating interactions between proteins. Non-fluorescent fragments of, for example, GFP (green fluorescent protein) are fused to the interacting partners. The interaction brings the fragments together, which then fold, reassemble and fluoresce. This process can be carried out in living cells and provides information both on the interaction and its subcellular location. We have developed a split-GFP-based BiFC assay for use in the budding yeast Saccharomyces cerevisiae in which the modifications are carried out at the genomic level, thus resulting in the tagged yeast proteins being expressed at wild-type levels. The system is capable of detecting interactions in all subcellular compartments tested (the cytoplasm, mitochondria and nucleus) and makes a valuable addition to techniques for the investigation of protein–protein interactions in this model organism.
Protein–protein interactions play a key role in many biological processes. The identification and characterization of these interactions is crucial to our understanding of these cellular processes. While many systems have been described to detect interactions, quite often their use is limited . As a strong favourite with molecular biologists due to its simple rapid screening approach, the classic Y2H (yeast two-hybrid) system often requires the overexpression of proteins that must interact in the nucleus. The TAP (tandem affinity purification) method allows detection of interactions at native levels in a variety of cellular compartments, but is limited to long-lived complexes. In vivo FRET (fluorescence resonance energy transfer) is an extremely sensitive detection system that offers instantaneous readout throughout the cell; however, it requires expensive equipment and complicated data analysis.
PCAs (protein-fragment complementation assays)
PCAs have been developed as a means to overcome such problems . First described by Johnsson and Varshavsky  in 1994 using a split-ubiquitin screen, PCAs have since been developed using other proteins including β-lactamase, β-galactosidase and dihydrofolate reductase. While the technology behind PCAs is simple and straightforward, their development has allowed the analysis of interactions that had proved difficult with other methods. The split-ubiquitin system allowed for the detection of membrane-associated interactions , whereas the β-galactosidase assay provides a sensitive signal due to its amplified enzymatic readout, allowing the detection of interactions of low affinity .
BiFC (bimolecular fluorescence complementation) assays
BiFC describes a specific group of PCAs that involve the reassembly of fluorescent proteins, such as GFP (green fluorescent protein) and its variants, to detect protein–protein interactions in living cells. GFP is a member of a growing family of β-can fluorescent proteins found in marine invertebrates; it is a small protein (238 amino acids) that has the unusual ability to fluoresce spontaneously without the presence of enzymatic substrates or cofactors . Mature GFP is very robust and exhibits great thermal and pH stability. This has allowed it to be fused successfully to a wide variety of proteins. Its stability is in part due to the compact β-can fold surrounding residues 65–67, which spontaneously form the fluorophore in the presence of oxygen. These properties of GFP make it a good candidate for BiFC assays. The principle of BiFC is that a pair of fragments alone exhibits no fluorescence. However, when tagged through suitable linkers to interacting protein partners, this interaction drives the functional reassembly of the split GFP fragments, allowing a fluorescent readout (Figure 1). The split GFP fragments are relatively small and therefore unlikely to interfere with protein complex formation. On protein interaction, the autocatalytic formation of the chromophore generates a directly observable signal with the added advantages of (i) providing information about the subcellular localization of the protein–protein interaction and (ii) not needing the addition of fluorogenic reagents.
Reassembly of split-GFP fragments in a BiFC
Basic BiFC principles suggest that split fragments should not recombine spontaneously without the protein–protein interaction and, furthermore, should not drive forward an otherwise unfavourable interaction. Early work on split-GFP BiFC assays  described the application of a split GFP system in bacterial cells and supported the requirement for protein–protein (or peptide–peptide) interactions to generate a specific fluorescent signal by the reassembly of split fragments. Advantages of the split-GFP-based BiFC system over previously described methods include its ability to detect interactions between weakly associating proteins (KD of approx. 1 mM) and detecting transient interactions . Furthermore, unlike the classic Y2H, BiFC is not restricted to the nucleus, allowing detection of interactions in their native compartments.
Modifications and improvements of the original split-GFP screen have broadened its use. The introduction of linker sequences has been applied in various BiFC studies to allow flexibility between fragments reducing the potential of steric hindrance (for example ). In addition, genetic manipulation of GFP has generated stronger readouts. EGFP (enhanced GFP) is a double mutant that exhibits a 35-fold increase in fluorescence . The development of spectral variants of GFP, such as YFP (yellow fluorescent protein), blue fluorescent protein and cyan fluorescent protein, has allowed for multicolour BiFC analysis of simultaneous or competing interactions in living cells . Since its introduction, BiFC has been implemented in a range of organisms including mammalian cells , yeast  and plants .
Applications of split-GFP-based BiFC assays in yeast
In 2007, two South Korean groups developed modified BiFC assays to detect protein–protein interactions in the budding yeast, Saccharomyces cerevisiae. Park et al.  described a plasmid-based BiFC system inspired by the Y2H strategy. They merged the principles of the Y2H and PCA technologies to design an EGFP reporter system with potential for high-throughput screening of interactions in yeast. Plasmid vectors were designed with N- and C-terminal EGFP fragments using the yeast alcohol dehydrogenase promoter, which enabled constitutive expression. Using these vectors, the group successfully detected the interaction between the Gal4p dimerization domain and Gal11p in yeast. The benefits of their design include the use of low-copy-number plasmids to reduce expression levels, a drawback of the classic Y2H system, while also overcoming the autoactivation problem commonly incurred using Y2H.
The second group employed homologous recombination to enable N-terminal and C-terminal labelling of yeast proteins with split fragments of YFP . Unlike the plasmid-based system , upon labelling, the proteins were expressed at native levels under the control of their native promoters. The chaperone proteins Sis1p and Ssa1p are known to form heterodimers . Sis1p in MATa yeast cells was labelled with a C-terminal YFP fragment and Ssa1p in MATα yeast cells was labelled with an N-terminal YFP fragment. On mating the two strains, the resulting diploids were found to fluoresce. This protocol can be applied to the investigation of homodimer interactions in yeast by mating MATa and MATα strains in which the same protein has been labelled with N- and C-terminal fragments of YFP.
Recently, we described a similar BiFC system for the detection of interactions in haploid yeast . The method combines the features of the TAP tagging methodology with basic PCA features. As with the method described by Sung and Huh , our system also benefits from detection of interaction at native protein expression levels. However, it is not amenable to homodimer interaction detection. Plasmid vectors were constructed containing N- and C-terminal EGFP fragments and selectable markers suitable for insertion into yeast cells. PCR-generated linear DNA fragments encoding split EGFP and the marker were homologously recombined into the yeast DNA genome such that they were inserted at the 3′-end of the genes of interest (Figure 2). To test the system, the known interaction between the subunits of the glycolytic enzyme Pfk (phosphofructokinase) were used. Pfk1p and Pfk2p subunits were labelled with N- and C-terminal EGFP fragments respectively. Fluorescence microscopy of recombinant yeast cells showed cytoplasmic fluorescence as expected for a glycolytic enzyme (Figure 3) . Further interaction studies were carried out to test the system in other cellular compartments. Interactions were confirmed by fluorescence in the nucleus, mitochondrial membrane and mitochondrial matrix . In contrast with the system described by Sung and Huh , 100% of the interacting partners are labelled (compared with an estimated 25%). This contributes additional sensitivity and we were able to detect interactions between extremely poorly expressed proteins such as those involving the mitochondrial succinate dehydrogenase subunit Sdh3p (approx. 200 copies per cell ).
Strategy for the modification of yeast genes by homologous recombination
The split-GFP-based BiFC assay works in a variety of subcellular compartments
The systems described comprise a suite of GFP-based methodologies that are applicable under different circumstances. The plasmid-based assay  is most readily adaptable to global screens, whereas the method of Sung and Huh  is ideal for homodimer detection. Our own method  offers the possibility of highly sensitive interaction detection. To get the best from any BiFC assay, a variety of approaches may have to be tried. While no system is perfect, each method described complements existing methods and provides an alternative approach to investigate and dissect the yeast proteome.
Ubiquitin and Ubiquitin-Like Modification in Health and Disease: Biochemical Society Irish Area Section Annual Meeting held at Royal College of Surgeons in Ireland, Dublin, Ireland, 22–23 November 2007. Organized and Edited by Caroline Jefferies (Royal College of Surgeons in Ireland).
E.B. is in receipt of a studentship from the European Social Fund and her attendance at the conference was funded in part by the Irish Area Section of the Biochemical Society.