Current Proteomics, 2007, 4, 17-27 17 1570-1646/07 $50.00+.00 ��2007 Bentham Science Publishers Ltd. Detection of Protein-Protein Interactions Using Protein-Fragment Complementation Assays (PCA) Emma Barnard, Neil V. McFerran, John Nelson and David J. Timson* School of Biological Sciences, Queen���s University Belfast, Medical Biology Centre, 97 Lisburn Road, Belfast BT9 7BL, UK Abstract: Protein-protein interactions play a central role in many cellular processes. Their characterisation is necessary in order to analyse these processes and for the functional identification of unknown proteins. Existing detection methods such as the yeast two-hybrid (Y2H) and tandem affinity purification (TAP) method provide a means to answer rapidly questions regarding protein-protein interactions, but have limitations which restrict their use to certain interaction net- works furthermore they provide little information regarding interaction localisation at the subcellular level. The develop- ment of protein-fragment complementation assays (PCA) employing a fluorescent reporter such as a member of the green fluorescent protein (GFP) family has led to a new method of interaction detection termed Bimolecular Fluorescent Com- plementation (BiFC). These assays have become important tools for understanding protein interactions and the develop- ment of whole genome interaction maps. BiFC assays have the advantages of very low background signal coupled with rapid detection of protein-protein interactions in vivo while also providing information regarding interaction compartmen- talisation. Modified forms of the assay such as the use of combinations of spectral variants of GFP have allowed simulta- neous visualisation of multiple competing interactions in vivo. Advantages and disadvantages of the method are discussed in the context of other fluorescence-based interaction monitoring techniques. Key Words: Protein-protein interactions, green fluorescent protein, protein complementation assay, bimolecular fluorescent complementation, EGFP, two hybrid screen. INTRODUCTION Many processes in living cells are controlled by the for- mation of stable or transient protein-protein interactions these include transcription, translation, signal transduction and cell division. The ability to identify these interactions is crucial for the characterisation of such cellular processes. Furthermore it has been shown that data gained from protein interaction studies can also be used to determine functions of unknown proteins (Schwikowski et al., 2000). There are many methods currently in use for the detection of protein interactions with new methods continually being developed (Piehler, 2005). Major existing methods, for in vivo detection of interactions, include the yeast two hybrid (Y2H), tandem affinity purification (TAP) tagging, fluorescence energy transfer (FRET) and more recently protein-fragment com- plementation assays (PCA) (Reviewed by Figeys, 2003 Meng et al., 2005). YEAST TWO HYBRID (Y2H) Designed and developed in 1989 by Fields and Song, the original yeast two-hybrid system detects the interaction be- tween two proteins tagged to functionally essential domains of the yeast transcriptional factor, Gal4p. Gal4p contains an N-terminal DNA binding domain (DBD) and a C-terminal transcription activation domain (TAD) (Fields and Song, 1989). The DBD of Gal4p is fused to ���bait��� protein X and *Address correspondence to this author at the School of Biological Sci- ences, Queen���s University Belfast, Medical Biology Centre, 97 Lisburn Road, Belfast BT9 7BL, UK Tel: +44-(0)28 9097 5875 Fax: +44-(0)28 9097 5877 E-mail: d.timson@qub.ac.uk TAD fused to ���prey��� protein Y. Alone TAD is inactive how- ever association with DBD upon X and Y interaction re- stores Gal4p activity (Fields and Song, 1989). Since its invention the Y2H system has been used exten- sively to detect protein-protein interactions. The yeast two hybrid technology is a powerful and broadly applied system that has since been implemented in studies of interactions within yeast (Uetz et al., 2000 Ito et al., 2001), bacteria (McCraith et al., 2000 Joung et al., 2000) and the fruit fly (Formstecher et al., 2006). It has the advantage of detecting interactions in vivo (McAllister-Henn et al., 1999) in a sim- ple and sensitive manner. A series of adaptations of the original system (Chien et al., 1991 Cagney et al., 2000 Uetz, 2001) has allowed for the wide-scale analysis of whole libraries of interactions in yeast permitting the investigation of a higher level of proteome organisation and the assembly of interaction network maps (Mayer and Heiter, 2000). The Y2H system has progressed rapidly over the years to become perhaps the most widely used interaction detection system. Despite its popularity with molecular biologists, the Y2H system has several limitations. Split domains of a transcriptional factor, such as Gal4p may be the standard choice for studying protein interactions (Immink and Angenent, 2002). However Ma and Ptashne (1987), found that random ���bait��� sequences when fused to the Gal4p DNA-binding domain were enough to activate the transcription factor without the need for an interaction with the activation domain or associated ���prey��� protein. Autoacti- vation of the transcription factor gives rise to the generation of ���false positives���, perhaps the most common problem with

18 Current Proteomics, 2007, Vol. 4, No. 1 Barnard et al. the Y2H system. False positives are brought about by inter- actions of a non-biological origin, which occur independ- ently of any physiologically relevant protein-protein interac- tion that is those which only occur in the context of the screen, and would not normally occur under normal, non- Y2H, physiological conditions (Semple et al., 2002 Droit et al., 2005 Fields, 2005). The use of a transcription factor also restricts protein interactions to the nucleus (Von-Mering et al., 2002). There- fore many proteins are not in their native compartments or recognized physiological environment and no information regarding interaction localization can be deduced. This shortcoming of the Y2H also makes membrane and mem- brane-associated proteins which account for 30% of the pro- teome (Piehler, 2005) inaccessible to the standard version of the screen (McAllister-Henn et al., 1999). There are other problems associated with the Y2H screen. The original system cannot detect interactions be- tween three or more proteins at any one time some interac- tions therefore may be missed if a third ���bridge��� protein is involved. An adapted system developed by Zhang and Lau- ter (1996), overcame this problem and allowed for detection of interactions within a ternary complex. The two proteins are fused to DBD and TAD domains while the third protein is fused to a nuclear localisation signal. This modification also carries its own problems such as the detection of inter- actions that don���t require the third protein (Causier and Da- vies, 2002). Furthermore the third protein may promote or inhibit formation of the two-hybrid interaction complex (Drees, 1999). A chief concern when choosing any approach to explore an interactome is its reliability. Two large Y2H screens car- ried out in 2000 to analyse the Saccharomyces cerevisiae proteome resulted in remarkably different sets of data, which in turn triggered a considerable amount of criticism. A screen carried out by Ito et al. (2000) generated 842 interac- tions of high relevance, while a similar screen by Uetz et al. (2000), resulted in 641 interactions of which only 141 over- lapped with the former study. However before being too du- bious about the method one must bear in mind its value in answering the fundamental question, ���does A interact with B?��� Its ability to do just that has allowed the yeast two hy- brid screen to remain a strong favourite with molecular bi- ologists. TANDEM AFFINITY PURIFICATION (TAP) First described by Rigaut et al., 1999, the TAP tagging method is based on affinity purification of a protein of inter- est and its associated interacting partners. The original TAP method utilized two tags ��� the immunoglobulin (IgG) bind- ing domains of Staphylococcus aureus protein A and a calmodulin binding peptide (CBP). In the sequence joining the tags is a tobacco etch virus (TEV) protease cleavage site. The gene encoding the protein of interest is fused to a se- quence encoding the TAP tag. Tagged proteins are expressed at native levels under normal physiological conditions. Inter- acting proteins form a complex which is then subject to a two-step affinity purification under non-denaturing condi- tions. The first purification step involves the binding of the S. aureus protein A to an IgG matrix. Incubation with TEV protease cleaves the tag and elutes the bound protein com- plex. The second purification step is carried out with calmodulin agarose beads which remove contaminants and remaining protease. Finally the protein complex is eluted using a calcium ion chelating agent such as EGTA. Recov- ered complexes are then separated by SDS-PAGE and ana- lysed by mass spectrometry (Puig et al., 2001 Bauer and Kuster, 2003). Variations of TAP have since been developed using successive rounds of purifications with several other tags. This ���multiple affinity purification��� or MAFT produces even purer preparations of protein complexes (Honey et al., 2001). As well as having the advantages of allowing the detec- tion of interactions at their native levels (Suter et al., 2006) the TAP method also permits purification of very large com- plexes - as shown in comprehensive yeast proteome studies by Gavin et al. (2002) and Ho et al., (2002) and Krogan et al., (2006). In the study by Gavin et al. (2002), 78% of 589 tagged proteins were purified along with associated partners, whilst the remaining 22% were unable to be purified or iden- tified. Gavin et al. (2002) deduced that the tagging can im- pair protein function or localization and interfere with com- plex formation due to the 20 kDa tag size. In a similar study by Ho et al. (2002), 493 complexes were purified, of which 93 overlapped with the Gavin study, 48 (52%) of these com- plexes contained interacting partners detected by both groups and the remaining 45 (48%) complexes showed no similari- ties. The study by Krogan et al. (2006) detected over 7,000 interactions involving 2,708 proteins. Each protein complex was prepared using two independent methods and analysed by mass spectrometry. The need for mass spectrometry to analyse the complexes retrieved by affinity purification bi- ases the results found in the studies towards long lived and stable complexes (Ito et al., 2002). Krogan et al. (2006) de- scribe the results from the study as ���a snapshot of interac- tions and complexes in that particular yeast strain subjected to particular growth conditions���. Some complexes may not have been present under those given conditions and some less stable or loosely-associated complexes or those with short life times may not have endured such time consuming purification steps. Like the yeast two hybrid system, the original TAP method designed for yeast has seen many refinements allow- ing its use to study interactions within bacteria (Escuela et al., 2006), plants (Rohila et al., 2004, 2006 Rubio et al., 2005) and mammalian cells (Knuesel et al., 2003 Li et al., 2004). The TAP tag system may well be the most accurate and efficient system for the detection of multiprotein complexes (Von-Mering et al., 2002). However it requires much ex- perimental effort and extensive data analysis for each de- tected interaction complex and cannot detect transient inter- actions. It also provides no information regarding interaction localisation. FLUORESCENCE RESONANCE ENERGY TRANS- FER (FRET) FRET is a biophysical phenomenon in which energy is transferred between two fluorescent molecules. For this transfer to take place, the molecules must be close in space