Analysing protein-protein interactions with the yeast ... - Springer Link

Plant Molecular Biology 50: 855–870, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.

855

Analysing protein-protein interactions with the yeast two-hybrid system Barry Causier∗ and Brendan Davies School of Biology, University of Leeds, Leeds LS2 9JT, UK (∗ author for correspondence; e-mail [email protected]) Received 3 July 2001; accepted in revised form 21 December 2001

Key words: interaction trap, protein-protein interactions, proteomics, yeast two-hybrid system

Abstract Plant research is moving into the post-genomic era. Proteomic-based strategies are now being developed to study functional aspects of the genes predicted from the various genome-sequencing initiatives. All biological processes depend on interactions formed between proteins and the mapping of such interactions on a global scale is providing interesting functional insights. One of the techniques that has proved itself invaluable in the mapping of protein-protein interactions is the yeast two-hybrid system. This system is a sensitive molecular genetic approach for studying protein-protein interactions in vivo. In this review we will introduce the yeast two-hybrid system, discuss modifications of the system that may be of interest to the plant science community and suggest potential applications of the technology. Abbreviations: 3-AT, 3-amino-1,2,4-triazole; 5FOA, 5-fluoroorotic acid; AD, activation domain; BD, DNA-binding domain; FRET, fluorescence resonance energy transfer; GEF, guanyl nucleotide exchange factor; GFP, green fluorescent protein; GST, glutathione S-transferase Introduction Over recent years, plant research has moved through the genomic era into the post-genomic era. Plant genome sequencing projects have identified a large number of genes for which a function has yet to be assigned. Consequently, plant research is moving into the fields of functional genomics and proteomics (the global analysis of the proteins expressed by a cell at a particular time). Key to many biological processes, such as signal transduction and transcription, are the physical interactions between proteins. Proteomic-based strategies are now being used to determine the network of interactions between all the proteins expressed in a cell. Numerous techniques have been developed to study protein-protein interactions, from biochemical approaches such as coimmunoprecipitation and affinity chromatography, to molecular genetic approaches such as the yeast twohybrid system. The yeast two-hybrid system is proving itself to be a powerful tool for proteomic-based inves-

tigations. The technology has already been employed to investigate the protein-protein interactions between many of the full-length open reading frames predicted from the yeast (Saccharomyces cerevisiae) genome sequencing initiative (Uetz et al., 2000). A similar approach has also been taken for the large-scale mapping of protein-protein interactions in Caenorhabditis elegans (Walhout et al., 2000) and Helicobacter pylori (Rain et al., 2001). It is only a matter of time before similar strategies are used to study the network of protein-protein interactions that occur at different developmental stages of Arabidopsis and other plant species, once the various genome sequencing projects are completed. Fields and Song (1989) described the first yeast two-hybrid system. It was based on the fact that many eukaryotic transcription factors have discrete and separable DNA-binding and transcriptional activation domains. In their system, protein-protein interactions were tested by fusing one test protein to the DNA-binding domain of the yeast GAL4 transcription

856

857 Figure 1. The yeast two-hybrid system and its modifications. A. The classical yeast two-hybrid system. Two plasmids are constructed, one encoding protein X fused in-frame with the DNA-binding domain (BD) of a transcription factor such as GAL4, the second encoding protein Y fused in-frame with a transcription activation domain (AD). The plasmids are co-transformed into a suitable yeast strain where the fusion proteins are expressed. Interaction between proteins X and Y reconstitutes an active transcription factor which specifically binds to elements upstream of the reporter genes and activates their expression. B. The hSos recruitment system. Two plasmids are constructed, one encoding protein X fused in-frame to the human Ras guanyl nucleotide exchange factor (Ras GEF) hSos, the second encoding protein Y fused in-frame with a v-Src myristylation signal. The plasmids are co-transformed into a temperature-sensitive yeast mutant containing a lesion in the Cdc25 gene which encodes a yeast Ras GEF. The protein Y fusion is targeted to the yeast plasma-membrane. Interaction between proteins X and Y recruits hSos to the yeast plasma membrane where it complements the cdc25 mutation by activating the Ras signalling cascade. Interactions are detected through growth of yeast cells at the restrictive temperature (37 ◦ C). The Ras recruitment system is based on similar principles to the hSos recruitment system except that instead of hSos, mRas is recruited to the yeast plasma membrane where it triggers the Ras signal transduction pathway (see text). C. Three protein system. This system is based on the classical yeast two-hybrid system. Proteins X and Y are expressed in-frame with a transcription factor DNA-binding domain and transcription activation domain, respectively. A third protein, Z, is expressed, without any added domains, in the yeast nucleus. Protein Y may only interact with X in the presence of Z. A domain formed through the interaction between X and Z may provide an interaction interface for protein Y (i). Alternatively, protein Z may act as a bridge between proteins X and Y (ii). D. The dual-bait system. Two test proteins (X1 and X2) are fused to two different DNA-binding domains (LexA and λcI, respectively). The two fusion constructs are co-expressed in the same yeast cell and tested for interaction with proteins fused to the B42 transcription activation domain. Interaction with X1 induces expression of LexA-dependent reporter genes (lacZ and LEU2). Interaction with X2 induces expression of λcI-dependent reporter genes (LYS2 and gusA). E. The reverse two-hybrid system. This system is based on the classical two-hybrid system except that expression of the reporter gene is toxic to the yeast cells under certain conditions. In this example, the reporter gene URA3 allows for selection of protein-protein interactions between X and Y on media minus uracil and counter-selection of disrupted protein-protein interaction between X and Y on media containing 5FOA. This system can be used to identify residues required for protein-protein interaction by making use of a mutagenised copy of the cDNA encoding one of the proteins. cDNAs encoding proteins no longer able to interact can be sequenced to reveal amino acids essential for interaction.

factor, and the second protein to the GAL4 activation domain. The fusion proteins were expressed in a suitable yeast strain and the interaction detected by assaying for expression of a GAL4 responsive reporter gene (see Figure 1A). The yeast two-hybrid system offers a number of advantages over many of the biochemical procedures often used for the analysis of protein-protein interactions. Yeast two-hybrid technology is relatively inexpensive since it avoids costly procedures such as antibody production and protein purification. cDNA expression libraries can easily be screened to isolate proteins interacting with a protein of interest. In this way, not only are the interacting proteins identified, but also the cloned cDNAs encoding them become available, so simplifying downstream studies. The system is often more sensitive than many in vitro techniques, and may be more suited to the detection of weak or transient interactions. Proteins expressed using in vitro techniques, or in bacterial cells, often lack key post-translational modifications that may be important for certain protein-protein interactions. In addition, the proteins may not fold correctly or may not be stable in the buffer conditions used. The yeasttwo-hybrid system simplifies mutational analyses allowing investigations such as the mapping of motifs or residues required for the protein-protein interaction. The yeast two-hybrid system has become a routine tool for the study of protein-protein interactions. Over the twelve or so years since the first report of

the system, it has been modified in numerous ways and has been adapted for the study of not only proteinprotein interactions but also DNA-protein interactions (the yeast one-hybrid system) and RNA-protein interactions (the yeast three-hybrid system; reviewed by Brent and Finley, 1997). Yeast cells offer a convenient system for these types of interaction studies but the system has also been adapted to use bacterial and mammalian cells. In this review we will give an overview of the yeast two-hybrid system, its modifications and examples of its use. We will review some of the current methodologies and suggest some applications of the two-hybrid system that may be particularly suited to plant research.

Yeast two-hybrid technologies The most widely used yeast two-hybrid systems utilise the reconstitution of an active transcription factor to assay for protein-protein interactions. However, such systems are not problem free and complementary systems, which do not rely on a transcriptional readout, have also been developed. Further developments of the system allow for the detection of ternary complex formation (three-protein systems) and for the dissection of residues involved in particular protein-protein interactions (such as the reverse two-hybrid system). The main two-hybrid technologies are described below.

858 Classical yeast two-hybrid systems The early yeast two-hybrid systems were based on the finding that many eukaryotic transcription factors have separable DNA-binding and transcription activation domains. Fusion of test proteins to each separate domain reconstitutes an active transcription factor providing that the test proteins interact. The expression of reporter genes, that contain upstream elements to which the DNA-binding domain (BD) binds, can be monitored to detect the interaction (see Figure 1A). The most commonly used systems are the GAL4 system (in which the DNA-binding and activation domains of the yeast GAL4 protein are used; Fields and Song, 1989) and the LexA system (DNA-binding domain of the bacterial repressor protein LexA used in combination with the Escherichia coli B42 activation domain; Gyuris et al., 1993). The LexA system is known generally as the interaction trap system. The yeast strains used for two-hybrid experiments carry mutations in a number of genes required for amino acid biosynthesis, such as TRP1, LEU2, HIS3 and URA3. If these amino acids are omitted from the growth medium the yeast strain will fail to grow. Many of the two-hybrid plasmids carry genes that complement these mutations and allow for selection of the transformant yeast. In the original system only one reporter gene, lacZ, was used (Fields and Song, 1989). However, as the technology developed, yeast strains containing a number of reporter genes were generated. Genes such as HIS3 (in the case of the GAL4 system) and LEU2 (in the LexA system) are now regularly used as reporter genes, in conjunction with lacZ, to provide a more stringent assay for protein-protein interactions. In the case of the HIS3 reporter gene, the assay for proteinprotein interaction can be optimised to reduce falsepositives by the inclusion of 3-amino-1,2,4-triazole (3-AT), at millimolar concentrations, in the media. 3AT is a competitive inhibitor of the HIS3 gene product and reduces the background due to basal HIS3 expression. In the LexA system, the sensitivity of the reporter genes is governed by the number of LexA operator sequences within the promoters of the reporter genes (reviewed by Brent and Finley, 1997). The most sensitive yeast strains (e.g. EGY48) have six LexA-binding sites compared with the least sensitive strains (e.g. EGY188) which have only two LexA-binding sites. The classical system has two major applications. It can be used to determine whether two known proteins interact with one another, or it can be used to

identify unknown proteins, encoded by a cDNA library, that interact with a protein of interest. In the latter case, the yeast two-hybrid system becomes a powerful tool for investigating the network of interactions that form between proteins involved in particular biological processes. Although the system offers many advantages over biochemical methods, such as cost, convenience and sensitivity, it still has several associated problems. False-positives are often generated in two-hybrid library screens. The use of two or more reporter genes to assay for an interaction eliminates a large number of false-positives but true interactions are generally confirmed with alternative procedures (see below). The system relies on proteins localising to the yeast cell nucleus. Many of the current two-hybrid vectors encode nuclear localisation signals to encourage non-nuclear proteins into the nucleus. However, proteins carrying strong signals for localisation to other parts of the cell, or that contain strongly hydrophobic domains (such as integral membrane proteins), may prove problematic in this system. In such cases truncations of the problematic proteins may make such proteins more accessible to two-hybrid analyses. Proteins that are not stably expressed or are toxic to yeast cells may prove difficult. In the latter case, plasmids with modified promoters driving low-level expression of the fusion proteins, or plasmids which are maintained at a low copy number (e.g. ARS4/CEN6 plasmids) and give reduced expression levels, may provide a solution. Alternatively, plasmids that drive expression of the fusion proteins from an inducible promoter (such as the yeast GAL1 promoter, which is galactoseinducible) result in the transient expression of toxic proteins. One other advantage of driving the expression of fusion proteins at low levels is that the number of false-positives may be reduced. The classical yeast two-hybrid system relies on the transcriptional activation of reporter genes and so may not be suited to the identification of proteins such as transcriptional repressors. In most two-hybrid vector systems the DNAbinding domain and the activation domain are fused to the amino-terminal end of the test protein. In some cases, where interactions occur at the N-terminus of the test protein(s), the presence of the DNA-binding or activation domain may cause problems. Vectors that provide a carboxy-terminal fusion may be more suitable in such cases (e.g. pNLexA, available commercially from OriGene). Finally, some proteins fused to the DNA-binding domain may possess a transcription activation domain that may activate reporter gene

859 expression. Indeed, up to 10% of randomly generated cDNAs inserted into a DNA-binding domain vector have been shown to auto-activate the reporter genes (Fashena et al., 2000). Therefore, it is important that the DNA-binding domain fusion proteins are tested for auto-activation of the reporter genes prior to testing for protein-protein interactions. If auto-activation is a problem, simple truncations of the test protein may be required to remove this activity. Alternatively, if HIS3 is being used as one of the reporter genes, increased concentrations of 3-AT in the media may reduce the level of auto-activation. If URA3 is being used as a reporter gene, yeast can be plated on media containing 5FOA to test for auto-activation (see below). Commercial yeast two-hybrid kits are now available from companies such as Clontech (Matchmaker systems; www.clontech.com), Stratagene (HybriZAP systems; www.stratagene.com), Invitrogen (Hybrid Hunter and ProQuest systems; www.invitrogen.com) and OriGene (DupLex A system; www.origene.com). These systems generally contain the appropriate yeast strains, vectors and controls. Some of these companies also provide cDNA libraries, from a number of species, constructed in yeast two-hybrid vectors (see Table 1). Other two-hybrid technologies hSos/Ras recruitment systems Three disadvantages of the classical yeast two-hybrid system, discussed above, are the potential difficulties with transcriptional repressors, auto-activation of reporter genes by bait constructs and the problem of certain proteins not localising to the yeast cell nucleus. Recently, two cytoplasm-based yeast two-hybrid systems, that do not rely on a transcriptional readout, have been developed to overcome these problems. These systems rely on the activation of the Ras signalling pathway to rescue a yeast cdc25 mutant that shows a temperature sensitive growth phenotype. Cdc25 encodes a Ras guanyl nucleotide exchange factor (GEF). The cdc25 phenotype can be rescued by human mRas, or by a human Ras GEF, hSos, if it is recruited to the yeast plasma membrane where it activates yRas. Two separate yeast two-hybrid systems have been developed around the hSos and mRas proteins. In the hSos system (available commercially from Stratagene as the CytoTrap system; see www.stratagene.com), hSos is fused to one test protein. The second test protein is fused to the v-Src myristylation signal which targets it to the yeast plasma membrane. Interaction

between the two test proteins results in recruitment of hSos to the plasma membrane and activation of the Ras signalling pathway through the exchange of GDP for GTP (Aronheim et al., 1997; see Figure 1B). In the mRas system, the farnesyl attachment sequence (CAAX box) is deleted thus preventing mRas locating to the plasma membrane. The altered mRas is fused to one test protein and is recruited to the plasma membrane through interaction with the myristylated partner of the test protein (Broder et al., 1998). In both cases, protein-protein interactions are detected by growth of the cdc25 yeast strain at the restrictive temperature. Since neither of these systems depends on a transcriptional readout for the detection of protein-protein interactions, the system is suitable for the analysis of proteins such as transcriptional repressors and activators. Similarly, since the system is based in the yeast cytoplasm, proteins that may not target well to the nucleus may be used successfully in this system. There are a number of problems with this particular yeast two-hybrid system. One is the yeast strain. In our experience the cdc25 yeast strain reverts (i.e. is able to grow at the restrictive temperature regardless of the plasmids it carries). This means that the yeast strain must be tested for temperature sensitivity both before transformation and after transformation, prior to any assays for protein-protein interaction. Secondly, the detection of protein-protein interactions relies solely on the growth of the yeast at the restrictive temperature, there is no other marker to confirm an interaction. The problem of reversion of the yeast strain and the lack of additional markers for protein-protein interactions suggests that the hSos/Ras recruitment system may not be suited to library screening. However, the system may find a role in confirming protein-protein interactions detected by other means. As with all two-hybrid approaches, any interactions detected with this system must be confirmed using alternative procedures (see below). Split-ubiquitin systems In the split-ubiquitin system proteins of interest are fused to separate domains of the ubiquitin polypeptide. Interaction between the proteins of interest reconstitutes an active ubiquitin molecule resulting in the proteolytic cleavage of a reporter protein from one of the ubiquitin domains. In the original system (Johnsson and Varshavsky, 1994) the released reporter protein was detected using immunological techniques. Consequently, its use in library screening was lim-

860

Table 1. Sources of yeast two-hybrid vector systems, yeast strains, reporter genes and plant cDNA libraries. Two-hybrid systems

Classical GAL4 Clontech MatchMaker2

Vectors (and selection)

Examples of yeast strains and reporter genes

Plant libraries Arabidopsisa

BD: pGBT9 (Trp) AD: pGAD424 (Leu)

HF7c: LacZ, HIS3

MatchMaker3

BD: pGBKT7 (Trp) AD: pGADT7

AH109: LacZ, HIS3, ADE2, MEL1

LifeTechnologies/ Invitrogen ProQuest

BD: pDBTrp (Trp)d ; pDBLeu (Leu); AD: pPC86 (Trp) Gateway ready: pDEST32 (BD; LEU); pDEST22 (AD; Trp)

MaV203: LacZ, HIS3, URA3

Stratagene HybriZap

BD: pBD-GAL4 CAM (Trp) AD: pAD-GAL4-2.1 (Leu)

YRG-2: LacZ, HIS3

Classical LexA Invitrogen Hybrid Hunter

BD: pHybLex/Zeo (Zeocin) AD: pYESTrp2 (Trp)

L40, EGY48: LacZ, HIS3

Clontech MatchMaker

BD: pGilda (His) AD: pB42AD (Trp)

EGY48c : LacZ, LEU2

OriGene DupLexA

BD: pGilda, pEG202 (His) AD: pJG4-5 (Trp)

EGY48c : LacZ, LEU2

MoBiTec Grow ’n’ Glow

BD: Bait Plasmid (His) AD: Prey Plasmid (Trp)

EGY48c : GFP, LEU2

hSos/Ras recruitment Stratagene CytoTrap

pMyr (URA) pSos (LEU)

Cdc25H: temperaturesensitive

Arabidopsis; tomato/P. syringaeb

None

a GAL4 Arabidopsis libraries are available from a number of sources. Clontech provide an Arabidopsis library in the pGAD10. The Arabidopsis Biological Resource Center (ABRC) have stocks of two separate Arabidopsis cDNA libraries in pACT2 (stock numbers CD4-10 and CD4-22; see www.biosci.ohio-state.edu/∼plantbio/Facilities/abrc/abrchome.htm). All of these libraries can be used with each of the GAL4-based two-hybrid systems listed in the table, but are not compatible with the LexA or hSos/Ras systems. b LexA Arabidopsis and Tomato/P. syringae libraries constructed in the pJG4-5 vector (see www.fccc.edu/research/labs/golemis/InteractionTrapInwork.html). c Many of the LexA systems are supplied with a number of yeast strains that carry reporter genes of varying sensitivities (e.g. EGY48, YM4271, YM187, EGY194, EGY188). MoBiTec do not name their yeast strains in their literature but they are likely to be similar to strains in other systems. d The Proquest system supplies two BD vectors. One has a LEU2 marker that can be used to screen libraries in the pPC86 vector (with TRP1 marker). However, libraries in a LEU2 plasmid (e.g. pGAD424) can be screened with BD-fusion proteins encoded by the pDBTrp vector (with the TRP1 marker). BD, DNA-binding domain plasmid; AD, transcription activation domain plasmid.

861 ited. In a variation of the system, primarily designed for the study of interactions between membrane proteins, the transcription factor A-LexA-VP16 was used as the reporter protein (Stagljar et al., 1998). With classical two-hybrid yeast strains (in this case L40), protein-protein interactions can be detected through the activation of the lacZ and HIS3 reporter genes. While the split-ubiquitin systems are ingenious in design, their use in the analysis of protein-protein interactions has thus far been limited. Three-protein systems The yeast two-hybrid system is designed to test for the interaction between two proteins. However, many cellular processes are dependent on the formation of complexes between numerous proteins. Interactions may be missed in a two-hybrid assay if the interaction between two proteins requires the presence of a third protein or if a protein interacts with a domain formed through the interaction of two other proteins. Yeast protein-protein interaction systems have been developed to study the formation of these ternary complexes (Zhang and Lautar, 1996; Egea-Cortines et al., 1999). They are based on the same principles as the classical yeast two-hybrid system. Two of the proteins are fused to the DNA-binding and activation domains of the GAL4 protein while the third is expressed with a nuclear localisation signal. The system is designed such that reporter gene activity is only detected when all three proteins come together to form a complex at the GAL4 upstream activation sequence (see Figure 1C). A problem with three-protein systems is that interactions, independent of the third protein, may be detected. Thorough testing of the interactions, both in vivo and in vitro, must be performed to confirm that the interaction requires all three proteins. These would include testing the interactor in a yeast strain lacking the third partner plasmid (i.e. DNA-binding domain fusion against activation domain fusion), and against the third partner fused to the DNA-binding domain. Not only will these tests determine the specificity of the three-way interaction, they may also shed light on whether the interactor acts as a bridge or binds to a domain formed through the interaction of the other two proteins. A commercial system for the study of ternary complex formation is now available from Clontech (pBridge vector system; www.clontech.com). Small ligand-dependent systems One adaptation of the yeast two-hybrid system that may be of interest to the plant science community, is

the detection of protein-protein interactions that rely upon non-protein molecules (reviewed by Brent and Finley, 1997). One could imagine a scenario in which a particular protein-protein interaction, say between a receptor and an effector, only occurs in the presence of a certain ligand, such as a phytohormone. Two-hybrid screens in the presence and absence of the ligand may allow such interactions to be investigated. Dual-bait systems The dual bait system is based on two sets of reporter genes bound by different DNA-binding domains. One set of reporters (lacZ and LEU2) contains promoter elements to which LexA binds, while the promoters of the second set (LYS2 and gusA) are bound by the λcI DNA-binding domain (Serebriiskii et al., 1999; see Figure 1D). Two DNA-binding domain constructs are prepared. The first encodes one test protein fused to LexA, the second expressing a different test protein fused to λcI. The constructs are co-transformed into an appropriate yeast strain and interaction studies are performed with proteins fused to the B42 transcription activation domain. The dual-bait system has a number of applications. Two test proteins can be analysed for protein-protein interaction partners in a single library screening. The system can be used to test the specificity of a proteinprotein interaction amongst evolutionarily conserved proteins and can be used to identify domains or residues required for interaction with one partner but not another (reviewed by Fashena et al., 2000). The dual-bait system is now commercially available from Invitrogen (Dual Bait Hybrid Hunter; www.invitrogen.com). Reverse two-hybrid systems Once a protein-protein interaction has been detected, the investigator may wish to dissect the domains or residues involved in forming the interaction. Generally speaking this involves creating deletions or point mutations within the cDNA encoding one of the proteins and determining whether the proteins still interact in two-hybrid assays. With the conventional reporter genes such as lacZ and HIS3, protein-protein interactions are detected by testing for β-galactosidase activity and growth on media lacking histidine (plus 3AT). If the interaction is disrupted the yeast no longer demonstrate reporter gene activity. To isolate yeast in which the protein-protein interaction is perturbed requires replica plating of the yeast on general maintenance media and on media to test for reporter gene

862 activity. Those yeast showing no reporter gene activity can then be recovered from the general maintenance plate. Not only is this procedure time-consuming, it also makes the screening of libraries of point mutations very difficult. The whole procedure can be simplified using a reverse two-hybrid approach (Figure 1E) which is used to select against certain proteinprotein interactions. This system utilises a reporter gene whose expression is toxic to the yeast cells under specific conditions. For example, one reverse twohybrid system uses URA3 as a reporter gene (Vidal et al., 1996a, b). The URA3 gene product converts the compound 5-fluoroorotic acid (5FOA) into the toxic compound 5-fluorouracil. Yeast cells that have URA3 as one of its reporter genes can be plated directly onto media containing 5FOA. If a protein-protein interaction occurs, the URA3 gene is expressed and the yeast cells die. However, if the protein-protein interaction is disrupted, URA3 is not up-regulated and the yeast cells survive. Colonies can be selected from the 5FOA plate and analysed to determine how the protein-protein interaction has been disrupted. Other reverse two-hybrid systems, based on different reporter genes, have also been developed for use in yeast (Leanna and Hannink, 1996; Brent and Finley, 1997). Mammalian and bacterial cell two-hybrid systems All the two-hybrid systems discussed above use yeast cells to perform the interaction assays and for most plant research yeast is probably the best system. However, a number of mammalian cell- and bacterial cellbased systems have recently become commercially available. Mammalian cell-based systems rely on the same principles as the yeast two-hybrid system. Test proteins are expressed as fusions to transcription factor DNA-binding and activation domains. Protein interactions result in expression of the CAT reporter gene that can be detected in standard assays (Luo et al., 1997). It is difficult to see how this system would be of benefit to the plant molecular biologist since it would be expensive to set up and is not necessarily suited to library screening. Bacterial cell-based systems are also based on transcription activation (Joung et al., 2000; Serebriiskii et al., 2000a). In the commercially available system from Stratagene (the BacterioMatch two-hybrid system), one test protein is fused to the bacteriophage λcl protein while the other is fused to the amino-terminal domain of the RNA polymerase α-subunit. Interaction between the two test proteins recruits the RNA polymerase to the promoter which results in expression of the AmpR and lacZ reporter

genes. The main advantages of the bacterial system over the yeast system are speed and the potential to screen larger libraries. However, the system is prokaryote based and may present problems if certain post-translational modifications are important for a particular protein-protein interaction.

Applications of the yeast two-hybrid system The yeast two-hybrid system has been used extensively in plant research to analyse known interactions and to isolate new interacting partners. It has been used to study protein-protein interactions involved in a number of diverse processes such as floral development (e.g. Davies et al., 1996), self-incompatibility mechanisms (e.g. Mazzurco et al., 2001), the circadian clock (e.g. Jarillo et al., 2001), plant disease resistance (e.g. Ellis et al., this issue) and phytohormone signalling (e.g. Ouellet et al., 2001). A yeast three protein system has been used to demonstrate the formation of protein complexes between plant transcription factor dimers (Egea-Cortines et al., 1999; Honma and Goto, 2001) and the hSos recruitment assay has been used to confirm interactions between plant MADS-box transcription factors (Causier and Davies, unpublished). The various genome sequencing projects are generating large amounts of new data. However, for a large proportion of the genes predicted from these projects, no function has yet been assigned. Proteinprotein interaction studies may begin to give clues about which processes and pathways these functionally unassigned genes may play a role. Perhaps one of the most interesting applications of two-hybrid technology, is the ability to map protein-protein interactions on a global scale (i.e. analysis of all the possible interactions amongst a population of proteins). Already these approaches are bearing fruit in other species (Uetz et al., 2000; Walhout et al., 2000; Rain et al., 2001) and it is envisaged that such analyses will soon be conducted for plant species. However, before such screens are initiated, it may be possible to make some informed predictions about the possible interaction partners of some plant proteins. The data generated through the analysis and visualisation of the large number of protein-protein interactions detected in the yeast experiment is publicly available at http://portal.curagen.com (Uetz et al., 2000). Proteins conserved between species, and their interaction partners, can be analysed using the databases and software

863 at the Curagen website. Not only is it now possible to perform protein-protein interaction studies in vivo, but it is now possible to think about predicting particular protein-protein interactions in silico.

Methodologies A flow diagram, describing a typical two-hybrid library screen, is presented in Figure 2. Some of the main techniques are described in the following sections. In these sections we concentrate mainly on the classical yeast two-hybrid system. However, many of the techniques described are applicable to other systems. In Table 1 are listed the common two-hybrid vectors, yeast strains and available plant two-hybrid libraries. Preparation of yeast two-hybrid constructs Generally speaking, simple in-frame cloning of PCR fragments into the appropriate vectors is sufficient to generate the two-hybrid vectors. Once generated, the constructs should be sequenced to confirm that no PCR errors have been generated and that the test protein is in-frame with its fusion partner. After yeast transformations, protein expression may be tested by immunological techniques such as western blotting. Some of the vector systems commonly used for two-hybrid analyses encode an epitope tag (such as haemagglutinin) which can be used for immunological detection of the expressed fusion proteins and for later analyses of protein-protein interactions (see below). Alternatively, commercially available antibodies, raised to the DNA-binding or transcription activation domains, may be used. In some systems, the expression of the fusion proteins is often too low to detect on western blots. However, the low level of expression does not preclude a fusion protein from working well in these systems. To accomplish the task of the global analysis of protein-protein interactions in yeast, Uetz et al. (2000) cloned the yeast ORFs into the appropriate two-hybrid vectors by homologous recombination in yeast cells. The vectors were modified to contain a region of sequence that precisely matched 70 bp tails at the ends of the PCR amplified ORFs. Linearised vector was transformed into yeast cells in combination with the PCR products. This allowed highly efficient homologous recombination between the vector and the PCR prod-

uct and generation of the appropriate fusion constructs in vivo. Recently, Life Technologies/Invitrogen launched a new system for cloning that relies on in vitro homologous recombination rather than restriction digest and ligation reactions (the Gateway system). The system uses the same mechanism that bacteriophage λ employs to integrate its DNA site-specifically into the E. coli genome, and to later excise it. The system is based on the formation of an ‘entry’ vector. Gatewaytagged PCR products are cloned into the ‘entry’ vector by using an in vitro homologous recombination step. From this ‘entry’ vector a whole series of other vectors can be generated such as expression vectors, binary vectors, yeast two-hybrid vectors and the like. These ‘destination’ vectors are generated in an in vitro homologous recombination step which includes the ‘entry’ vector, the empty ‘destination’ vector (essentially any vector that has been modified to act as a ‘destination’ vector in this system) and components involved in the excision of λ from the E. coli genome. The reading frame of the cloned DNA insert is maintained from ‘entry’ vector to ‘destination’ vector. Only the ‘entry’ vector needs to be sequenced since the reaction to form the ‘destination’ vector from the ‘entry’ vector is base perfect. These steps are quick to perform and are carried out wholly in vitro. In theory, ‘destination’ vectors can be generated within a matter of days. Gatewayready yeast two-hybrid vectors are now available from Life Technologies/Invitrogen (see Table 1). Not only will systems such as Gateway simplify the construction of two-hybrid vectors, they should also simplify downstream applications. cDNAs, present in a Gateway two-hybrid vector, which encode proteins that interact with a test protein, can be shuttled into another ‘destination’ vector (e.g. a GSTfusion vector for testing protein-protein interactions in vitro), using the Gateway homologous recombination steps, within a matter of days. This negates the need to identify restriction sites suitable for cloning, the possible need to design and synthesise oligonucleotides, and the subsequent PCR reactions, restriction digests and ligation reactions which may take several weeks. Two-hybrid libraries In the classical yeast two-hybrid system libraries are usually generated in the activation domain plasmid. Full-length cDNAs (produced using oligo-dT) or random-primed libraries (consisting of truncated

864

Figure 2. A typical yeast two-hybrid library screen. The diagram describes a typical yeast two-hybrid library screen using the GAL4 protein system with lacZ and HIS3 reporter genes. The first step is to transform the appropriate yeast strain with the construct encoding the test protein (X) fused in-frame with the GAL4 DNA-binding domain (BD). Transformant yeast colonies should be checked for activation of reporter genes by the BD-X fusion protein. A weak auto-activation may be eliminated by increased 3-AT concentrations in the medium. Alternatively, deletions may be required to remove the domain of X responsible for activating reporter gene expression. If the BD-X construct passes the auto-activation test it can be transformed with the GAL4 activation domain (AD)-cDNA fusion library, using the protocol detailed in the text. Transformants are selected on media lacking histidine (and containing 3-AT if required). Colonies that grow at a faster rate compared to the background of yeast growth are putative positives for a protein-protein interaction and are transferred to a fresh plate lacking histidine – the master plate. From this master plate yeast can be plated for β-galactosidase assays and onto media minus histidine +3-AT to confirm reporter gene activity. The activation domain plasmid can be rescued from yeast cells demonstrating both reporter gene activities. The AD plasmid can be transformed back into yeast cells (of the same or different strain) bearing the BD-X construct, and re-assayed for reporter gene activity. If the interaction is detected, the cDNA cloned into the AD vector can be sequenced and additional analyses performed (see text).

865 Table 2. Protocol for small-scale yeast transformations.a 1.

From a freshly grown plate inoculate 1–2 large colonies into 10 ml of liquid media (either YPD (2% tryptone, 1% yeast extract, 2% glucose, pH 5.8) if the yeast strain contains no plasmids or the appropriate selective media if the strain carries a plasmid) and incubate overnight at 30 ◦ C/200 rpm 2. Inoculate into 100 ml YPD to give an OD600 of ca. 0.2, and incubate for 4–4.5 h at 30 ◦ C/200 rpm 3. Pellet the cells at 3000 rpm for 5 min and wash in 100 ml sterile water 4. Wash cells in 1.5 ml 1× Li/TE (prepared fresh from 10× stocks, see below) and resuspend in 0.5 ml 1× Li/TE 5. Each transformation consists of the following components: 1 µg plasmid DNA 160 µg ssDNA (disperse salmon sperm DNA at 10 mg/ml in TE by repeated pipetting and dissolve by stirring overnight at 4 ◦ C; aliquot and store at −20 ◦ C. Before use boil for 5 min and cool rapidly on ice) 100 µl yeast cell suspension 10 µl DMSO 600 µl 1× PEG/Li/TE (prepared fresh from stocks, see below) Gently mix components together until homogenous mixture results 6. Incubate, in a water bath, at 30 ◦ C for 30 min with occasional mixing and then at 42 ◦ C for 15–30 min in a water bath with occasional mixing 7. Pellet cells and resuspend in 1 ml sterile water 8. Plate 100 µl onto appropriate selective media and incubate at 30 ◦ C until colonies visible (about 3 days)b Solutions: 1× Li/TE – prepare from autoclaved 10× TE (100 mM Tris.Cl pH 7.5, 10 mM EDTA pH 7.5) and 10× LiAc (4 M LiAc pH 7.5) solutions PEG/Li/TE – prepare 1× Li/TE using an autoclaved 50% polyethylene glycol (Mr 4000) solution. a For example, transformation of a single construct, or co-transformation of both two-hybrid vectors, into

yeast. b Small-scale transformation protocol should yield about 105 transformants per µg of plasmid DNA. Co-

transformations typically yield less transformants than sequential transformations.

cDNAs) have both been used successfully in yeast two-hybrid screens. cDNA libraries are generally directionally cloned into the relevant vector. Directional cloning ensures that one-in-three of the cDNAs will be in-frame with the appropriate fusion partner compared with non-directionally cloned libraries where only one in six of the cDNAs will be in-frame. Since yeast transformations are generally inefficient (compared to E. coli transformations) directionally cloned libraries require that less yeast transformants need to be generated to give a representative screen. The tissue from which the library is generated may be of importance. Libraries generated from cells/tissues in which the test protein is known to be expressed will reduce the number of false-positives generated and so produce more biologically relevant interactions. Yeast transformations A large number of protocols have been published for the efficient transformation of yeast cells, ranging

from heat-shock methods to electroporation. The procedures used in our laboratory for small-scale transformations are detailed in Table 2, and for library-scale transformations in Table 3. Testing for auto-activation and nuclear localisation Auto-activation of reporter gene expression may be a problem with some proteins (e.g. transcription factors) fused to the DNA-binding domain. It is critical that all DNA-binding domain fusions are tested for autoactivation prior to library screening (see Figure 2). As discussed above, this is simply a matter of testing for reporter gene activity. In addition to auto-activation, it is also important to ensure that the fusion proteins are localised to the yeast cell nucleus and bind to the appropriate operators. In the GAL4 system it may be possible to test this with a positive control. In the LexA system it is generally tested using a repression assay (Brent and Ptashne, 1984). In the repression assay the yeast strain harbour-

866 Table 3. Protocol for library-scale yeast transformations.a 1. 2. 3. 4. 5.

6. 7. 8. 9. 10.

From a freshly grown plate inoculate several large colonies into 100 ml of appropriate selective medium and incubate at 30 ◦ C/200 rpm overnight. Determine the cell number and pellet the volume of culture corresponding to 1.5 × 109 cells (3000 rpm for 5 min). Resuspend the cells in 300 ml of pre-warmed (30 ◦ C) YPD (see Table 2) and incubate at 30 ◦ C/200 rpm until the cell number reaches 2 × 107 cells/ml (ca. 3–4 h). Pellet the cells and wash with 150 ml sterile water To the cell pellet add the following (in order): 14.4 ml 50% PEG 4000 2.16 ml 1 M LiAc 1.5 ml ssDNA (2 mg/ml, prepared as described in Table 2) 60 µg plasmid DNA sterile water to 21 ml. Vortex until a homogenous mixture results and incubate at 30 ◦ C for 30 min and then at 42 ◦ C for 30 min with occasional mixing. Pellet cells and resuspend in 20 ml sterile water. Plate a small aliquot (0.5–1 µl) onto media lacking the appropriate amino acids to select for the plasmids (to assess the transformation efficiency). Plate the remainder (300 µl per 15 cm petri dish) onto media lacking the appropriate amino acids to select for the two plasmids and for reporter gene expression. Incubate plates at 30 ◦ C until colonies appear (3 days or more).b

a After Gietz and Schiestl, 1995; see www.umanitoba.ca/faculties/medicine/biochem/gietz/ b In our hands this procedure yields >5 × 106 total transformants.

ing the DNA-binding domain construct is transformed with the pJK101 plasmid. pJK101 contains a GAL1 upstream activating sequence, and between this and the GAL1 transcription start is a LexA operator which can be bound by LexA fusion proteins. In the presence of galactose β-galactosidase activity is strong in yeast cells containing pJK101. However, if a non autoactivating LexA fusion protein localises to the nucleus and binds to the pJK101 LexA operator, significant repression of β-galactosidase activity results. Repression is never complete and appropriate positive and negative controls should be included to gauge the level of repression. Yeast mating A number of the commercial yeast two-hybrid kits offer yeast mating strains. Haploid yeast strains of opposite mating type (a and α), one containing, for example, a DNA-binding domain fusion construct, the other containing an activation domain fusion construct, can be simply mated to obtain the diploid yeast strain containing both constructs (see Figure 3). Yeast mating can be used to test the interaction between a great number of proteins from a relatively few yeast transformations. Indeed, it was the basis of the global

analysis of the interactions between many of the fulllength open reading frames predicted from the yeast genome (Uetz et al., 2000). Interestingly, it has been suggested that interactions might be detected in a mating screen that are missed in standard screens (Uetz et al., 2000). A simple yeast mating procedure is detailed in Table 4. Selecting putative positives from a library screen Several days after a library-scale yeast transformation, colonies larger than the overall background of colonies, begin to become apparent. These fast growing colonies are putative positives for a protein-protein interaction. Over the period of about a week these colonies can be selected and replica plated onto media that selects for the protein-protein interaction (e.g. −His + 3-AT) and onto general maintenance media (to enable assays for the second reporter gene, such as lacZ). Yeast colonies that demonstrate activity of both reporter genes can be analysed further (see below and Figure 2). Quantitative liquid assays have been described to measure β-galactosidase activity. While these may be useful for comparing the relative strength of various interactions with a single protein they are not a

867

Figure 3. Classical yeast-two hybrid assays using yeast mating. Constructs encoding putative interacting proteins (X and Y) fused to the DNA-binding (BD) and transcription activation (AD) domains, respectively, are transformed separately into yeast strains of opposite mating type (a and α). After mating, both constructs are present in the same yeast cell and, if proteins X and Y interact, reporter gene activity is detected.

Table 4. Yeast mating procedure. 1. 2. 3. 4. 5. 6. 7.

Resuspend a single colony (from a fresh plate) of the a-type strain in 30 µl sterile water. Resuspend a single colony of the α-type strain in 30 µl sterile water Onto a fresh YPD plate (YPD + 2% bacteriological agar) pipette 2.5 µl of the a-type yeast suspension and allow it to become absorbed into the medium. Pipette 2.5 µl of the α-type yeast suspension on top of the dried a-type spot. Incubate at room temperature until yeast growth is present (1–2 days). Patch a small amount of the yeast growth onto the appropriate selective media (to select for diploid yeast cells containing both plasmids). Grow at 30 ◦ C as normal.

Table 5. Isolation of nucleic acids from yeast cultures. 1. 2. 3. 4. 5. 6.

Pellet approximately 3 ml of a yeast overnight culture (in YPD). Resuspend the yeast cell pellet in 0.2 ml lysis buffer (2% Triton X-100; 1% SDS; 100 mM NaCl; 10 mM Tris.Cl pH 8, 1 mM EDTA). Add 0.2 ml chloroform and 0.3 g acid washed glass beads (425–600 µm). Vortex for 2 min and spin at top speed for 5 min in a microcentrifuge. Ethanol precipitate the supernatant (200 µl) and dissolve the pellet in 20 µl sterile water. Transform ca. 2 µl, by electroporation, into the appropriate E. coli strain.

868 definitive measure of the affinity of one protein for another. Isolation of library plasmids from yeast A simple protocol for the isolation of crude nucleic acids from yeast cultures is described in Table 5. In the GAL4 system, the activation domain plasmids (the library plasmid) generally carry the LEU2 gene for selection (see Table 1). The library plasmids can be selected by transformation in a leuB E. coli strain, such as HB101, that can be rescued by the LEU2 gene on minimal media lacking leucine. In other systems the E. coli antibiotic selection marker on the DNAbinding domain and activation domain plasmids are different. The library plasmids can easily be isolated by transformation into E. coli and selection with the appropriate antibiotic. Another option is to use DNAbinding domain plasmids that carry the CYH2 gene that confers sensitivity to cycloheximide. Only yeast cells that have lost the DNA-binding domain plasmid are able to grow in the presence of the antibiotic. Eliminating false-positives Library screens invariably result in the isolation of false positive interactions. False-positives are those proteins which are not expected to interact with the test protein or could not possibly interact with it in the cell (Colas and Brent, 1998; Serebriiskii et al., 2000b). Such false-positives may include activation domain fusion proteins that directly bind to and activate the reporter gene. The use of different reporter genes, with different promoter structures, may eliminate many of these false-positives. Other false positives may include proteins that contain domains commonly involved in forming protein-protein interactions, but for which there is no physiological context for the interaction. Such false-positives may be discarded on the basis of information such as the expression patterns of the proteins being tested. Many false-positives can be simply eliminated by isolating the activation domain plasmid from the appropriate yeast colony (see below) and re-transforming it back into a yeast strain containing the DNA-binding domain fusion. In addition, it is often considered good practice to test the interaction with reciprocal hybrids. If an interaction between proteins X and Y is detected when X is fused to the DNA-binding domain (BD) and Y to the transcription activation domain (AD), is the interaction still detectable if Y is fused to BD and X is

fused to AD? However, if one fails to detect an interaction in the opposite direction, it does not necessarily prove that the proteins do not interact. A database of commonly isolated false positives has been compiled by Serebriiskii and Golemis (http://www.fccc.edu/research/labs/golemis/InteractionTrapInWork.html) and includes ribosomal proteins and heat shock proteins (Colas and Brent, 1998; Serebriiskii et al., 2000b). Scanning such databases may provide information on whether a particular interaction is likely to be significant or not. In addition, it is good practice to test whether proteins isolated from library screens interact with proteins unrelated to the test protein to eliminate inherently ‘sticky’ proteins. The number of false-positives can be kept to a minimum by simple microbiological procedures. Firstly, library transformations should not be plated too densely. In the protocol given in Table 3 about 70 plates (15 cm diameter) are required to plate the entire transformation. We recommend that this is the minimum number of plates one should use. Problems often arise when interpreting reporter gene activity if large amounts of yeast have been patched onto the relevant media. If such problems arise we recommend that a small amount of yeast growth be dispersed in a small volume of sterile water (50 µl) and serial dilutions spotted onto the appropriate media. Generally speaking, alternative methods are used to confirm protein-protein interactions detected using a two-hybrid system. Such procedures are described in the following section. Confirming interactions in vitro Interactions detected in yeast two-hybrid assays should always be confirmed using in vitro approaches to eliminate the possibility that yeast proteins mediated the interaction. In situ hybridisations (for protocols refer to Long et al., 1996), using probes corresponding to each interacting protein, will demonstrate whether the corresponding genes have similar spatial (at the cell level) and temporal expression patterns, and indicate whether the interaction is likely to be biologically relevant. Similarly, promoter-GUS/-GFP fusion studies can be used to determine the expression patterns of the genes of interest. Particular protein-protein interactions may suggest a biological role, based on the literature, for one or both of the interacting partners. For certain proteins it may be possible to confirm the biological relevance of the interaction by over-expressing the proteins in

869 transgenic plants and analysing the phenotype. For example, the MADS-box family of transcription factors are involved in many aspects of floral development, including floral organ identity (reviewed by Causier et al., 1999). These factors form homo- and heterodimers and it was assumed that expression of the particular dimers were sufficient for the development of particular floral organs. However, this was not borne out in transgenic experiments since the ectopic expression of these floral organ identity MADS-box dimers had no effect on vegetative organs. It was clear other factors were required to specify floral organs. Subsequent studies identified these additional factors (the SEPALLATA (SEP) MADS-box proteins) (Pelaz et al., 2000). Using a yeast three-protein system, Honma and Goto (2001) demonstrated that the SEP3 protein formed a higher-order complex with floral organ identity MADS-box dimers. As predicted, ectopic expression of these protein complexes in transgenic Arabidopsis plants resulted in the conversion of leaves to floral organs, so confirming the formation of a complex that was suggested by the yeast experiments. Other procedures, such as in vitro co-immunoprecipitation or GST pull-downs, are relatively easy to set up and perform. Techniques such as these are covered in more detail elsewhere in this issue (Vitale, this issue). Interactions may also be confirmed in vitro by assaying for biological activity. For example, the formation of enzyme complexes may be monitored by enzymatic assay; interactions involving transcription factors can be analysed using DNA electrophoretic mobility shift assays (for protocols, see Dent et al., 1999), and so on. More elaborate techniques may also be used to determine whether the protein-protein interaction occurs in plant cells. One such technique, which is becoming increasingly popular, is FRET (fluorescence resonance energy transfer). FRET is discussed elsewhere in this issue (Visser, Hink and Bisseling).

Concluding remarks The yeast two-hybrid system, and its various incarnations, has proved itself an invaluable tool in the study of protein-protein interactions. Twelve years after its first report, the yeast two-hybrid system really has come of age. Its use in plant research has been continual but now that plant genome sequencing projects

have been completed, or are nearing completion, the real power of this system will be realised. Proteinprotein interaction studies, on a global scale, will place uncharacterised genes within a functional context. Networks of protein-protein interactions will be realised and be compared with similar networks from other species.

Acknowledgements We would like to thank Dr Martin Kieffer and Dr Irene Weir for critical comments during the preparation of the manuscript. B.C. is supported by a grant from the BBSRC.

References Aronheim, A., Zandi, E., Hennemann, H., Elledge, S.J. and Karin, M. 1997. Isolation of an AP-1 repressor by a novel method for detecting protein-protein interactions. Mol. Cell Biol. 17: 3094– 3102. Brent, R. and Finley, R.L. 1997. Understanding gene and allele function with two-hybrid methods. Annu. Rev. Genet. 31: 663–704. Brent, R. and Ptashne, M. 1984. A bacterial repressor protein or a yeast transcriptional terminator can block upstream activation of a yeast gene. Nature 312: 612–615 Broder, Y.C., Katz, S. and Aronheim, A. 1998. The Ras recruitment system, a novel approach to the study of protein-protein interactions. Curr. Biol. 8: 1121–1124. Causier, B., Weir, I. and Davies, B. 1999. MADS-box factors in hermaphrodite flower development. In: C.C. Ainsworth (Ed.) Sex Determination in Plants, Bios Scientific Publishers, Oxford, pp. 1–23. Colas, P. and Brent, R. 1998. The impact of two-hybrid and related methods on biotechnology. Trends Biotechnol. 16: 355–363. Davies, B., Egea-Cortines, M., de Andrade Silva, E., Saedler, H. and Sommer, H. 1996. Multiple interactions amongst floral homeotic proteins. EMBO J. 15: 4330–4343. Dent, C.L., Smith, M.D. and Latchman, D.S. 1999. The DNA mobility shift assay. In: D.S. Latchman (Ed.) Transcription Factors: A Practical Approach, Oxford University Press, Oxford, pp. 1–25. Egea-Cortines, M., Saedler, H. and Sommer, H. 1999. Ternary complex formation between the MADS-box proteins SQUAMOSA, DEFICIENS and GLOBOSA is involved in the control of floral architecture in Antirrhinum majus. EMBO J. 18: 5370–5379. Ellis et al. - this issue Fashena, S.J., Serebriiskii, I. and Golemis, E.A. 2000. The continued evolution of two-hybrid screening approaches in yeast: how to outwit different preys with different baits. Gene 250: 1–14. Fields, S. and Song, O.-K. 1989. A novel genetic system to detect protein-protein interactions. Nature 340: 245–246. Gietz, R.D. and Schiestl, R.H. 1995. Transforming yeast with DNA. Meths Mol Cell Biol 5: 255–269. Gyuris, J., Golemis, E., Chertov, H. and Brent, R. 1993. Cdi1, a human G1 and S phase protein phosphatase that associates with Cdk2. Cell 75: 791–803.

870 Honma,T. and Goto, K. 2001. Complexes of MADS-box proteins are sufficient to convert leaves into floral organs. Nature 409: 525–529. Jarillo, J.A., Capel, J., Tang, R.H., Yang, H.Q., Alonso, J.M., Ecker, J.R. and Cashmore, A.R. 2001. An Arabidopsis circadian clock component interacts with both CRY1 and phyB. Nature 410: 487–490. Johnsson, N. and Varshavsky, A. 1994. Split ubiquitin as a sensor of protein interactions in vivo. Proc. Natl. Acad. Sci. USA 91: 10340–10344. Joung, J.K., Ramm, E.I. and Pabo, C.O. 2000. A bacterial twohybrid selection system for studying protein-DNA and proteinprotein interactions. Proc. Natl. Acad. Sci. USA 97: 7382–7387. Leanna, C.A. and Hannink, M. 1996. The reverse two-hybrid system: a genetic scheme for selection against specific protein/protein interactions. Nucl. Acids Res. 24: 3341–3347. Long, J.A., Moan, E.I., Medford, J. and Barton, M.K. 1996. A member of the KNOTTED class of homeodomain proteins encoded by the SHOOTMERISTEMLESS gene of Arabidopsis. Nature 379: 66–69. Luo, Y., Batalao, A., Zhou, H. and Zhu, L. 1997. Mammalian two-hybrid system. A complementary approach to the yeast two-hybrid system. Biotechniques 22: 350–352. Mazzurco, M., Sulaman, W., Elina, H., Cock, J.M. and Goring, D.R. 2001. Further analysis of the interactions between the Brassica S receptor kinase and three interacting proteins (ARC1, THL1 and THL2) in the yeast two-hybrid system. Plant Mol. Biol. 45: 365–376. Ouellet, F., Overvoorde, P.J. and Theologis, A. 2001. IAA17/AXR3: Biochemical insight into an auxin mutant phenotype. Plant Cell 13: 829–841. Pelaz, S., Ditta, G.S., Baumann, E., Wisman, E. and Yanofsky, M.F. 2000. B and C floral organ identity functions require SEPALLATA MADS-box genes. Nature 405: 200–203. Rain, J.-C., Selig, L., De Reuse, H., Battaglia, V., Reverdy, C., Simon, S., Lenzen, G., Petel, F., Wojcik, J., Schächter, V., Chemama, V.Y., Labigne, A. and Legrain, P. 2001. The protein-protein interaction map of Helicobacter pylori. Nature 409: 211–215.

Serebriiskii, I., Khazak, V. and Gomelis, E.A. 1999. A twohybrid dual bait system to discriminate specificity of protein interactions. J. Biol. Chem. 274: 17080–17087. Serebriiskii, I.G., Toby, G.G. and Golemis, E.A. 2000a. Two-hybrid system for the characterization of protein-protein interactions in E. coli. Biotechniques 29: 288–296. Serebriiskii, I.G., Estojak, J., Berman, M. and Golemis, E.A. 2000b. Approaches to detecting false positives in yeast two-hybrid systems. Biotechniques 28: 328–336. Stagljar, I., Korostensky, C., Johnsson, N. and Te Heesen, S. 1998. A genetic system based on split-ubiquitin for the analysis of interactions between membrane proteins in vivo. Proc. Natl. Acad. Sci. USA 95: 5187–5192. Uetz, P., Giot, L., Cagney, G., Mansfield, T.A., Judson, R.S., Knight, J.R., Lockshon, D., Narayan, V., Srinivasan, M., Pochart, P., Qureshi-Emili, A., Li, Y., Godwin, B., Conover, D., Kalbfleisch, T., Vijayadamodar, G., Yang, M., Johnston, M., Fields, S. and Rothberg, J.M. 2000. A comprehensive analysis of protein-protein interactions in Saccharomyces cerevisiae. Nature 403: 623–627. Vidal, M., Brachmann, R.K., Fattaey, A., Harlow, E. and Boeke, J.D. 1996a. Reverse two-hybrid and one-hybrid systems to detect dissociation of protein-protein and DNA-protein interactions. Proc. Natl. Acad. Sci. USA 93: 10315–10320. Vidal, M., Braun, P., Chen, E., Boeke, J.D. and Harlow, E. 1996b. Genetic characterization of a mammalian protein-protein interaction domain by using a yeast reverse two-hybrid system. Proc. Natl. Acad. Sci. USA 93: 10321–10326. Visser - this issue Vitale - this issue Walhout, A.J.M., Sordella, R., Lu, X., Hartley, J.L., Temple, G.F., Brasch, M.A., Thierry-Mieg, N. and Vidal, M. 2000. Protein interaction mapping in C. elegans using proteins involved in vulval development. Science 287: 116–122. Zhang, J. and Lautar, S. 1996. A yeast three-hybrid method to clone ternary protein complex components. Anal. Biochem. 242: 68– 72.