Utilizing techniques common to modern molecular biology laboratories, two-hybrid systems offer a number of advantages over other biochemical methods for detecting interactions between proteins, such as immunoprecipitation, biochemical copurification, and affinity chromatography. One significant advantage is that two-hybrid systems require neither purified proteins/antibodies nor prior knowledge of a protein's function. They offer the ultimate in flexibility: Pairs of known proteins can be tested against each other for possible interactions, groups of proteins can be tested in pairwise combinations, or entire libraries of genes (cDNAs or genomic) can be tested against either an individual protein or another library, enabling the detection of completely novel sets of interacting proteins. Two-hybrid systems are marvelously efficient in that they provide the opportunity to screen entire libraries for interactions in a single experiment and in fact, have been used to map interactions within the entire genome/proteome of bacteriophage T7 of E. coli.3 A further advantage is that the gene sequences for the interacting proteins are contained in the clones that score positively in two-hybrid screens, facilitating the retrieval of genes of interest. And finally, as experiments are conducted in living eukaryotic cells, the proteins are more likely to be in their native state than proteins expressed in bacteria.
All two- (and one- and three- for that matter) hybrid systems exploit the fact that many transcriptional activators in species as diverse as bacteria and humans are bipartite, containing domains for DNA binding (referred to as DB domains) and transcriptional activation or transactivation domains (AD or TA domains). While in nature these two domains coexist on the same protein, they also work when they exist separately as long as the two domains are brought into proximity. And herein lies the magic of two-hybrid systems. Using a set of vectors that carry what are termed the bait and target (or prey), genes to be tested are fused to the two domains of transcriptional activators separately. These systems utilize host cells (most often yeast, but several mammalian systems are available as well) that bear reporter gene constructs. The reporter genes are under control of the relevant transcriptional activator. In short, the reporter gene is hooked up to a sequence responsive to the AD domain used by a given two-hybrid system. The host cells are transformed with a set of bait (DB domain) and target (AD domain) vectors, and the fusion protein products are made in vivo, where they can interact in the host cell's cytoplasm. When they do, the two domains of the transcriptional activator are brought together and transcription of the reporter gene ensues.
There are two basic yeast systems, GAL4 and LexA, each of which was developed independently by two different groups--Stanley Fields (University of Washington), and S.J. Elledge (Baylor College of Medicine) derived the GAL4 system and Roger Brent (Massachusett General Hospital) and Stanley Hollenberg (Oregon Health Science University) the LexA. Kits and reagents that incorporate elements of all four systems are available commercially. Roughly half the kits in today's market are based on the system first used in the seminal experiments of Fields and Song. These kits employ GAL4, a yeast transcriptional activator, which regulates the transcription of a number of genes when yeast are grown on a galactose-containing medium. The original experiments employed a yeast strain deleted for GAL4 containing the lac Z gene of ß-galactosidase fused to a gal-responsive gene (GAL1). Activation by the reconstituted GAL4 was detected using the X-Gal blue/white screening technique. Since then, it has become standard to use dual reporters, lac Z and HIS3, particularly for screening cDNA libraries. More recently, other reporter genes have been incorporated into two-hybrid systems, among them the genes encoding luciferase and green fluorescent protein.
The GAL4 two-hybrid system pioneered by Fields and Song was nothing short of revolutionary. Second-generation GAL4 two-hybrid systems are frequently and successfully used today. However, nothing is perfect. Preliminary screenings of cDNA libraries generally pull up quite a few potential interacting partners. One must perform a series of tests to determine which clones represent the bona fide interacting partners and which are simply false positives. False positives caused by autoactivation and other factors are reduced in the interaction trap (IT) two-hybrid system developed independently and concurrently by Gyuris et al.4 The lower levels of false positives enjoyed by this system are due to the fact that it uses a heterologous transcriptional activator (bacterial LexA) instead of a yeast transcriptional activator (such as GAL4). The Golemis group has collected and compiled statistics on the success rate of the IT system that indicate for well-controlled bait (that is to say, expressed at reasonable levels), the success rate of the IT system is greater than 70%. These results, as well as tables of successful "hunts" (biologically relevant interactions) and "trash" (proteins that give rise to false positives) can be viewed on the group's Web site (www.fccc.edu/research/labs/golemis/InteractionTrapInWork.html)..
A common reason for failure of two-hybrid screens is auto- or transactivation by bait alone, which subverts the necessity of interaction. A significant fraction (roughly 5-10%) of randomly generated cDNA sequences act as transcriptional activators on their own, which should give one pause when contemplating using libraries as baits in two-library screens. (It's not as bad as it sounds--a simple prescreening of bait-containing plasmids in the absence of prey can eliminate such clones from a library.) Another source of false positives is the spurious interaction of one protein with another; among trash are found proteins that have a penchant for interacting with other proteins, such as heat shock proteins. While the presumption is that because interactions are tested in vivo, the proteins are in their native state, often only portions of a protein are expressed, and domains not normally exposed might be made available to other proteins. Finally, it is also worth noting that in two-hybrid screens, proteins might be exposed to species not normally present in the same time and in the same cellular space in vivo.
Two-hybrid systems have proven to be quite sensitive, a decided virtue when working with weakly interacting proteins, often missed by some of the more commonly used alternatives (coimmunoprecipitation, for example). However, with high sensitivity come high backgrounds, as weak and nonspecific interactions can lead to positives as well. For this reason, a number of two-hybrid kits have incorporated ways to modulate the sensitivity, either biochemically or genetically, by varying the number of responsive sites upstream of the reporter gene, whatever it may be. Host yeast strains with different numbers of LexA binding sites are contained in the kits from Bio 101 and Origene. Other companies (Life Technologies and CLONTECH) provide low copy number plasmids to reduce the likelihood of overexpression of the proteins, which might bring with it nonspecific binding; yet others have incorporated multiple reporting systems to override any spurious interaction between the protein and a single site and eliminate single-site mutations as a cause of false positives.
Several one-hybrid kits are available for detecting DNA-protein interaction (CLONTECH and Display Systems Biotech). These kits work basically the same way as two-hybrid systems except in the one-hybrid kits, a DNA binding motif is positioned upstream of a reporter gene, which must interact with a protein expressed on an AD vector in order to score positively. Hence, the first step is to clone a target DNA element upstream of a reporter gene in the yeast genome. A protein or library is then cloned into a vector containing an activation domain and transformed into the yeast strain containing the target element insertion. In both kits currently on the market, multiple copies of the target DNA sequence are introduced upstream of the reporter to enable the detection of weak interactions.
DNA, protein--only RNA is missing in this equation. But not so. A three-hybrid system has been described,2 a kit for which is now available from Invitrogen, which tests for RNA-protein interactions. Once again, based on the separation of the transcriptional activator domain from DNA binding domain, this system positions a third component, a bifunctional RNA, in the mix. The bait vector contains a fusion of a DNA binding domain with an RNA binding domain (MS2 coat protein in the original system as well as the Invitrogen kit); the target vector contains a second RNA binding domain (target), and interposed between the two is a third vector containing a "hybrid" RNA, part known interacting RNA (MS2 RNA in this example) and part unknown or bait. (See diagram above.) In the original report, two different known RNA-protein interactions were shown to be detected in this system: The iron response element and its iron regulatory protein, as well as HIV Tat protein interaction with its target RNA, TAR.
Which two-hybrid system is best for you depends on the kinds of interactions you are interested in, if in fact you know--whether you are looking for weak or strong interactions, for example, or potentially toxic proteins, or whether you are screening libraries or well-known proteins. For weakly interacting proteins, systems that express bait and targets at high levels would be preferred, and/or systems with sensitive reporters. For strongly interacting species, low copy number plasmids and tightly regulated reporters would be a better choice. Vectors specifically designed for use with potentially toxic bait have also been developed. Prescreens, especially important when screening libraries, are built into some systems, and while this is an extra step, it can save time in the later stages of characterizing clones. And don't try this at home, as without the specialized yeast strains found in two-hybrid kits, developed by researchers in academia and in companies, there would be no two-hybrid systems.
A number of kits, detailed below, have been developed for two-hybrid screens. Many of them use virtually identical reagents, but most have incorporated some different twist--modified vector, different numbers of and kinds of reporters, a choice of strains--for addressing the various tricky points in the system.
Bio 101 offers two kits, GrowNGlow, a LexA based system, and Split Hybrid, a novel kit for detecting disruptions in protein interactions. GrowNGlow uses green fluorescent protein (GFP) as the sole reporter gene, which can be assayed directly on the plate with a UV lamp. The kit has some useful features, such as yeast strains of differing sensitivities to reduce background, and a bait vector with a fusion tag to facilitate the isolation of the bait for verification and further study.
The Split Hybrid kit, called by some a reverse two-hybrid system, turns the system upside down; instead of detecting protein interactions, it detects disruptions in protein interaction. Based on recent work from the labs of Robert Goodman and Merl Hoekstra,5 this kit is designed to screen for mutations in proteins known to be part of a scheme of interacting proteins. It employs a novel set of yeast strains to do this, as well as novel vectors. In the Split Hybrid system, the LexA binding protein carries the bait, and the transcription activator, in this case VP16 from herpesvirus, binds to the promoter for the tetracycline repressor, which in turn controls the expression of the HIS3 gene. When the bait and target interaction is intact, the tetracycline (tet) repressor is made, rendering the host dependent on histidine in the medium. On the other hand, when one of the proteins carries a mutation that interferes with the normal binding interaction, tet repression is released, allowing for a positive selection (histidine minus medium) for noninteracting proteins.
The Split Hybrid system is compatible with other LexA/VP16 libraries. The system can be modulated with tet levels and includes strains with different histidine inhibition sensitivities (weak, medium, and strong) to match the strength of interaction of the proteins under scrutiny.
CLONTECH offers a complete line of kits, libraries, and supporting reagents for conducting searches for protein interactions in its MATCHMAKER series. This includes one-, two- and three-hybrid systems for screening in yeast; in the case of the two-hybrid system, both GAL4 and LexA systems are offered. A large selection of premade MATCHMAKER libraries can be purchased, including ones that are already pretransformed into yeast ready for screening by yeast mating. CLONTECH's two-hybrid systems utilize different reporter genes (see table); hence, a GAL4 library can only be used with GAL4 systems employing the LEU2 reporter, and the LexA libraries can be used with LexA systems using the TRP reporter.
In the latest version of the MATCHMAKER GAL4 system, CLONTECH includes a new yeast host strain, AH109, and new bait and activator vectors. AH109 includes distinct, integrated reporter constructs (Ade, His, LacZ, Mel1), each controlled by different regulatory sequences, which allows researchers to vary the stringency of selection and directly confirm colonies harboring interacting proteins on the selection plate. In addition, the new vectors include T7 promoters and universal HA or c-Myc tags to facilitate downstream analysis.
The LexA system comes with dual reporting capability, lacZ, which is carried on a high copy number plasmid, and LEU2, which is integrated into the host genome. This system is inducible and contains multiple LexA binding sites, enabling the detection of weakly interacting proteins.
CLONTECH's Mammalian two-hybrid system requires cotransfection of host cells with three vectors: bait (GAL4 based), target (VP16 protein of herpes simplex), and a reporter vector with CAT activity under the control of the GAL4 responsive element and adenovirus E1b promoter.
CLONTECH offers a one-hybrid system with a double reporter system, HIS3 and/or lac Z to reduce background (false positives), though the company claims that with this system, the backgrounds are low.
Display Systems Biotech offers a line of displayGREEN kits using two reporter genes, LEU2 and a variant of the green fluorescent protein (GFP) that is 18 times brighter than the wild-type GFP protein from Aquorea victoria. Based on the LexA system, the basic two-hybrid system uses three vectors: displayBAIT, with the bait expressed from the strong alcohol dehydrogenase (ADH) promoter; displayTARGET, from which controlled expression of inserts can be induced with galactose; and displayREPORTER, which has eight copies of LexA operator upstream of GFP. The company claims that the tight control of Leu2 coupled with the easy detection of GFP in a LexA background combine to give few false positives. Also, the use of GFP as a reporter facilitates the procedure by eliminating the need to do biochemical lacZ assays to confirm interactions.
Display Systems has several other hybrid kits in the displayGREEN series: one for cloning of cDNAs produced by displayPROFILE Restriction Fragment Differential Display System (displayGREEN-PROFILE Kit), one for replacing lacZ with GFP reporter (displayGREEN-Replace Kit), and a one-hybrid system that uses the GFP reporter. The company also offers premade libraries and recommends its cDNA library construction kit for making a library for screening. To complete the two-hybrid product line, the company offers the media required for the displayGREEN system and complete kits for yeast transformation (the displayGREEN Yeast Transformation Kit) and yeast plasmid isolation (the displayGREEN Yeast DNA Isolation Kit).
Invitrogen offers a LexA-based system sporting smaller vectors and expanded multiple cloning sites for ease of use and improved efficiency of tranformation. The bait vector carries the unique resistance marker Zeocin® and hence is compatible with any target vector currently available. Zeocin resistance also allows selection of the vector in bacteria facilitating the isolation of the prey vector after screening. The Hybrid Hunter prey vector carries a nuclear localization signal and a fusion tag. The kit comes with two host strains: one that has constitutive expression for simple, one-step two-hybrid screening and a second that allows inducible expression useful for validating bait expression if desired, and for detecting interactions with toxic proteins. Invitrogen also offers premade libraries that can be used with any LexA system and a selection of reporter vectors and strains that allow the researcher to alter the stringency of the protein-protein interactions detected.
In addition to the Hybrid Hunter Yeast Two-Hybrid System, Invitrogen also offers a unique system for detecting RNA-protein interactions called the RNA-Protein Hybrid Hunter System. In this system, a bait RNA sequence is transcribed as a 3' or 5' fusion to the MS2 RNA. The host strain, which bears reporter genes controlled by a series of LexA binding sites, carries the gene for the MS2 coat protein fused to the LexA DNA binding domain. Transcription of reporters is activated when an RNA-protein interaction ensues: The bait RNA-MS2 transcript binds the MS2 coat protein-LexA fusion, which can do its usual magic on the reporter gene.
Life Technologies ProQuest System is a yeast GAL4-based system that incorporates improvements as described by Marc Vidal and Ed Harlow (at Massachusetts General Hospital), and Jef Boeke, Pierre Chevray, and Daniel Nathans (at Johns Hopkins University). The system incorporates modifications to the strains, vectors, and protocols to reduce false positives and false negatives.
The system utilizes three reporter genes regulated by three independent promoters, which present four different phenotypes. Together these three unrelated promoters and four phenotypes eliminate false positives resulting from interaction of library-encoded fusions with the promoter or proteins bound there. ProQuest uses ARS/Cen low-copy number vectors to bring levels of protein expression closer to physiological and to address the critical issues of false positives due to nonspecific binding and false negatives due to toxicity. The use of ARS/Cen vectors also provides a consistent copy number (rather than the variable copy number exhibited by 2 µm-based vectors), providing reproducible phenotypes, which is important when subtle changes in interactions need to be detected. A cycloheximide-sensitive bait vector facilitates the recovery of the target library.
Life Technologies offers a number of supporting reagents to ProQuest including a continually expanding list of ProQuest catalog and custom cDNA libraries. In addition, Life Technologies also offers a two-hybrid screening service and has made its reverse two-hybrid technology available through its technology access program.
OriGene Technologies offers a LexA-based system with a number of improvements: reporter genes with different sensitivities (based on the number of LexA binding sites upstream of the reporter gene), the capability of putting the bait at either the 5' or 3' end of the LexA DNA binding domain, and inducible expression. The company also offers the pGILDA, a low-copy number plasmid, for toxic bait. A large number of premade libraries are available that are compatible with several other LexA two-hybrid systems (Invitrogen and CLONTECH). A system for checking that the bait is expressed and capable of binding to proteins is also included. For a brief review of OriGene's system, see www.the- scientist.library.upenn.edu/yr1999/jan/tools1_990104.html.
Promega offers a mammalian two-hybrid system for testing selected protein interactions in mammalian cells. This system is designed to test putative interacting proteins that may have originally been identified in yeast. The vectors contain a DNA binding domain derived from yeast (GAL4) and an activation domain from herpes simplex virus (VP16). The reporter gene, firefly luciferase, provides a sensitive read-out and is located on a third plasmid (pG5luc) that bears five GAL4 binding sites. The bait vector also carries the Renilla luciferase gene, which provides an internal control for transfection efficiency.
The pBIND and pACT vectors contain the strong cytomegalovirus promoter to improve expression of the GAL4 DNA binding domain and VP16 activation domain, respectively, along with the regulatory regions and binding sites for T3 and T7 RNA polymerase to facilitate the production of sense and antisense RNAs. The T7 RNA polymerase promoter is useful in combination with in vitro translation to verify sizes of the expected fusion proteins. The system also contains a pair of positive control fusion protein vectors that provide high levels of luciferase activity.
Unique among the two-hybrid kits, Quantum Biotechnologies Q-PID Two-Hybrid System is based on an RNA polymerase III (pol III) protein-interaction detection system. This system offers advantages over the transcriptional activators-based two-hybrid systems as it involves fewer components and activates fewer genes in yeast, providing a simpler system with fewer false positives. At the center of this system is a small nuclear RNA required for growth in yeast, SNR5, which has intragenic control regions required for activation by pol III. When the intragenic regions of this RNA are interrupted, transcription can not take place; however, hybrid molecules containing a pol III element (the T138 subunit of TRIIIC) placed close to the DNA binding domains of GAL4 have been shown to restore transcription. The Quantum kit utilizes a special strain of yeast with a temperature-sensitive mutation in the SNR5 locus, rendering it dependent on activation by a hybrid complex. This system also has a built-in check for false positives: The SNR5 locus is carried in the host strain on a plasmid along with URA3 gene. Growing positives in the presence of an analog that is incorporated by URA 3 (5-fluoroorotic acid) selects against those colonies that are positive in the hybrid screen by virtue of reactivation of the chromosomal copy of SNR5.
Stratagene offers two unique kits: (1) a lambda vector-based two-hybrid system that uses the GAL4 yeast transcriptional activator and (2) a system that looks at interactions in the cytoplasm (as opposed to the nucleus in the case of all other two-hybrid kits). HybridZAP 2.1 vector is used to construct the cDNA or genomic library and, by in vivo excision, fuses with the target vector containing the activation domain of the GAL 4 transcriptional activator. Library construction kits can be purchased separately, as can the vectors themselves.
The CytoTrap system for detecting protein interactions uses membrane-bound proteins as the trap and bait. The human Sos gene, which activates membrane-bound ras protein and which is a homolog to a temperature-sensitive cdc25 gene in the host, is used for bait, and the target is attached to a myristlation signal that binds it to the membrane. When the bait contains a protein that interacts with the membrane-bound target, hSOS activates ras, leading to temperature-insensitive growth. This system is unique in that the interactions take place in the cytoplasm, making it possible to look at the effect of post-translational modification on the process.
The success of two-hybrid systems has spawned a number of modifications and enhancements to broaden the reach of this type of system. For example, a drawback to yeast two-hybrid systems, which do not employ the full range of post-translational modifications found in higher eukaryotes, is that they can not be used to assess the effects of post-translational modifications on protein interactions. Necessity is the mother of invention, and some solutions to problems of this type have already been developed: Cotransformation of host yeast cells with a bait protein and a protein tyrosine kinase has been used to analyze the ability of the tyrosine-phosphorylated bait to interact with other proteins. Another extension of two-hybrid systems is the use of peptide libraries in two-hybrid systems to identify common motifs, much as peptide display libraries do. Protocols for defining three- component systems have been described, where the third component can be a protein or even a small molecule or drug. And dual-bait systems that utilize distinct reporter genes are under development; these can be used to look for patterns of interacting proteins or to perform multiple screens with a single transformation. All in all, it seems that many exciting developments using one/two/three (and more?) hybrid technologies are on the horizon.
- S, Fields and O. Song, "A novel genetic system to detect protein-protein interactions," Nature, 340:256-6, 1989.
- D.J. SenGupta et al., "A three-hybrid system to detect RNA-protein interactions in vivo," Proceedings of the National Academy of Sciences (PNAS), 93:8496-501, 1996.
- P.L. Bartel et al., "A protein linkage map of Escherichia coli bacteriophage T7," Nature Genetics, 12:72-7, 1996.
- J. Gyuris et al., "Cdi1, a human G1 and S phase protein phosphatase that associates with Cdk2," Cell, 75:791-803, 1993.
- H.M. Shiu et al., "A positive genetic selection for disrupting protein-protein interactions. Identification of CREB mutations that prevent association with the coactivator CBP," PNAS, 93:13896-901, 1996.