Applications of interaction traps/ two hybrid systems to

biotechnology research

Andrew R. Mendelsohn and Roger Brent

Summary

Two hybrid methods provide a simple and sensitive means to detect the interaction between two proteins in living cells. Their use has resulted in the isolation of new proteins, and has facilitated characterization of particular protein-protein interactions. These techniques have already resulted in the identification of important targets for pharmaceutical intervention, and it is likely that their extension in coming years will allow the development of new drugs.

Introduction

Two hybrid systems are based on the finding that most eukaryotic transcription activators are modular. Brent and Ptashne showed that the activation domain of yeast GAL4, a yeast transcription factor, could be fused to the DNA binding domain of E. coli LexA to create a functional transcription activator in yeast [1] . These and similar experiments [2] formally define activation domains as portions of proteins, often acidic [3] , that activate transcription when brought to DNA by DNA binding domains. The DNA binding domain does not have to be physically on the same polypeptide as the activation domain. Ma and Ptashne demonstrated that the yeast protein Gal80 binds to Gal4, obscuring Gal4's major activation domain [4] . They then deleted the activation domain from Gal4, fused an activation domain to Gal80, and showed that, under conditions where the two proteins interacted, derivatized Gal80 caused the inactive Gal4 to stimulate transcription [4] . Similarly, Triezenberg, McKnight and coworkers found that although HSV VP16 does not bind DNA well on its own, when recruited to DNA by virtue of its interaction with a DNA bound protein (now known to be Oct-1), it strongly stimulates transcription of HSV immediate early genes [5, 6] . These findings raised the possibility that transcription of reporter genes could be used as a phenotype to detect protein interactions. In this view, when proteins encoded by libraries interact with fusion proteins bound to DNA upstream of reporters, they might obscure an activation domain on the DNA bound fusion, like GAL80 obscures an activation domain on GAL4. Alternatively, interacting proteins might carry their own activation domains, like VP16 and stimulate transcription of the downstream reporter.

Meanwhile, working independently, Fields and his coworkers made the seminal suggestion that protein interactions could be detected if two potentially-interacting proteins were expressed as chimeras. In their suggestion, the first protein contains a DNA binding domain and is bound to DNA upstream of a reporter gene, the second contains an activation domain [7] . They demonstrated this concept with SNF1 and SNF4, yeast proteins that interact in vitro. They showed that SNF1 fused to a DNA binding domain and SNF4 fused to an activation domain together could activate transcription of a reporter gene. This idea and corresponding experiment were important for two reasons. First, the use of the activation tag on the interacting protein increases the likelihood that its binding to the DNA-bound protein will detectably increase transcription of the reporter. Second, the experiment demonstrated that transcription could be used as a tool to study interactions between proteins not involved in the transcription process. All current systems [8-11] for detecting intracellular protein protein interactions make use of these results [7] .Here we describe the application of the two hybrid system to biotechnology research.

Two-hybrid systems

All systems (see [8-14] share common elements. All use 1) a plasmid that directs the synthesis of a "bait": a known protein which is brought to DNA by being fused to a DNA binding domain, 2) one or more reporter genes ("reporters") with upstream binding sites for the bait, and 3) a plasmid that directs the synthesis of proteins fused to activation domains and other useful moieties ("activation tagged proteins", or "prey" ). All current systems direct the synthesis of proteins that carry the activation domain at the amino terminus of the fusion, facilitating the expression of open reading frames encoded by cDNAs. Figure 1 illustrates many of these elements.

The systems differ in their specifics. These details (see [8-12] ) are typically relevant to their successful use. Baits differ in their DNA binding domains. Some systems use baits that contain native E. coli LexA repressor protein [10, 11] . LexA binds tightly to appropriate operators [15,16] , and carries a dimerization domain at its C terminus [17-20] . In yeast, LexA and most LexA derivatives enter the nucleus [1, 21] , but are not necessarily nuclear localized [22] . Others use baits that contain a portion of the yeast GAL4 protein [8, 9, 13] . This portion, encoded by residues 1-147, is sufficient to bind tightly to appropriate DNA binding sites [23] , localize fused proteins to the nucleus [24] , and direct dimerization [25] ; it also contains a domain that weakly activates transcription from mammalian cell extracts in vitro [26] , and it is thus conceivable that this domain may increase transcription resulting from weakly interacting proteins.

Reporter genes differ in the phenotypes they confer. The products of some reporter genes (e.g., HIS3 [8, 9, 11, 13] , LEU2 [10] ) allow cells expressing them to be selected by growth on appropriate medium, while the products of others (e.g. lacZ [8-12] ) allow cells expressing them to be visually screened. Reporters also differ in the number and affinity of upstream binding sites (e.g., lexA operators) for the bait, and in the position of these sites relative to the transcription startpoint [10] . Finally, they differ in the number of molecules of the reporter gene product necessary to score the phenotype. These differences affect the strength of the protein interactions the reporters can detect (J. Estojak, E. Golemis, and R. Brent).

Preys differ in the activation domains they carry, and in whether they contain other useful moieties such as nuclear localization sequences and epitope tags [8-12] . Some activation domains are stronger than others. Although strong activation domains should allow detection of weaker interactions, their expression can also harm the cell due to poorly understood transcriptional effects, either by titration of cofactors necessary for transcription of other genes ("squelching") [27] or by toxic effects that result when strong activation domains are brought to DNA [28] . Thus, it is possible that strong activation domains may prevent detection of some interactions. Activation tagged proteins also differ in whether they are expressed constitutively [8, 9, 11, 13] , or conditionally [10] . Conditional expression allows the transcription phenotypes obtained in selections (or "hunts") for interactors [10] (see below) to be ascribed to the synthesis of the tagged protein, thus reducing the number of false positive cells that grow because their reporters are aberrantly transcribed.

Although most two hybrid systems use yeast, two mammalian variants deserve mention. In one, interaction of activation tagged VP16 derivatives with a Gal4-derived bait drives expression of reporters that direct the synthesis of Hygromycin B phosphotransferase, Chloramphenicol acetyltransferase, or CD4 cell surface antigen [29] . In the other, interaction of VP16-tagged derivatives with Gal4-derived baits drives the synthesis of SV40 T antigen, which in turn promotes the replication of the prey plasmid, which carries a SV40 origin [30] .

Applications

Several industrially significant uses of two hybrid systems have emerged. The first is to identify new protein targets for pharmaceutical intervention. For example, Seol et al. [31] used the interaction trap system of Gyuris et al. ( [10] and see Figure 1 and legend) to identify new nuclear hormone receptors. They expressed in yeast a LexA-RXR (9-cis-retinoic acid receptor) bait, verified that it did not stimulate transcription, and determined that it bound operator by performing a "repression" assay (see [21] ). They then introduced the bait plasmid into a strain that contained a sensitive LexA operator-LEU2 (LexAop-LEU2) reporter and a medium sensitive LexAop-lacZ reporter to create the selection strain. Next, they introduced an activation tagged liver cDNA library, harvested 3x106 independent transformants, took 2 X 10ee7 cells from this collection, induced prey synthesis by growth of the cells in galactose, and selected 100 LEU+ colonies. They screened these for those that were also blue on Xgal and in which the LEU+ and lacZ+ phenotypes depended on the growth of the cells on galactose (expression of the prey). They identified 40 different clones of cells that carried potential interactors. To verify that the encoded proteins interacted specifically with this retinoic acid receptor, prey plasmids were rescued from these cells and introduced into strains that carried different LexA fusion protein baits. 4 new interacting proteins were identified, two of which, RIP14 and RIP15, encoded nuclear hormone receptors of novel sequence type. Ligands for these receptors are likely to be biologically active and may well have pharmaceutical significance.

A second industrial application is to determine the specific residues involved in a given protein-protein interaction. In two hybrid systems, strength of activation generally correlates with strength of interaction, and mutations in either interacting protein that diminish the strength of the reporter phenotype binding can indicate residues involved in protein-protein contact (J. Gyuris and R. Brent, unpublished; and [32] ). Second site suppresser mutations can sometimes be selected; such compensatory mutations can identify amino acids that interact with the affected residues in the first protein [32] .

A third industrial use is to find compounds that modulate protein interactions. A search for such compounds would likely use reporters, such as Aequorea victoria Green Fluorescent Protein [33] , or lacZ together with a fluorescent substrate such as C12FDG [34] , whose expression can be easily quantitated by automated equipment. Compounds that weaken a given interaction would diminish expression of reporters. However, S. cerevisiae are relatively impermeable to many chemicals [35] so that some inhibitory compounds might go undetected. One solution to this problem is to increase the permeability of yeast, for example, by selection of appropriate mutants (e.g., [36] ), by treatment with lytic enzymes to dissolve the cell wall (e.g., [37] ), or by treatment with specific reagents such as polymixin B [35] . Another would be to circumvent the permeability problem by developing robust two hybrid systems in mammalian cells or that rely on transcription in vitro. Alternatively,it may be possible to circumvent the problem by searching for possible inhibitors among proteins or RNAs encoded by combinatorial libraries. For example, we have used the interaction trap to select peptides that interact with human cyclin dependent kinase 2 (Cdk2) from a library of conformationally constrained random peptides (T. Jessen, B. Cohen, P. Colas, A. Mendelsohn, J. McCoy, and R. Brent, in preparation), some of which appear to inhibit the interaction of Cdk2 with its partners. Reporters, such as URA3,LYS2 or CYH2, that allow selection against transcription activation will facilitate selection of such inhibitors.

Interaction technology may also be of utility for assigning function in genome applications; for exampleto assign function to unknown proteins, to assign proteins to ordered genetic pathways, and even to find genes altered in disease states (R. Finley and R. Brent, PNAS, in press). These possibilities arise from the idea that that a given protein interacts only with specific types of bait moieties which may provide hints as to its function. Currently, there are two variant schemes to test large numbers of individual binary protein interactions. In one, (figure 2a) an investigator mates a lawn of a haploid strain that contains an activation tagged unknown protein with a grid of haploid strains of the opposite mating type, each member of which contains a different bait. In another, (figure 2b) the investigator mates bait strains in horizontal stripes with prey strains plated in vertical stripes. In each case, the investigator selects diploid exconjugants by replica plating, and by a subsequent replica plating step, determines which members of the grid of diploid strains ("interaction matrix") have active reporters. Inspection of the interaction matrix reveals which proteins contact one another. The use of reporters of different sensitivities should allow such patterns to be determined for interactions of different affinities. A great deal of information resides in these patterns, for example, observation of appropriate individual binary interactions among a group of mutually interacting proteins can suggest that they form a multi-protein complex (Finley and Brent).

In the near future, it is likely that many important proteins (for example, new cell cycle regulators) may be isolated by combining two hybrid methods with simultaneous genetic selections. Other possible future developments include development of systems in which transcription depends upon protein interactions that occur only at specific phases of the cell cycle, or times during development, or in particular subcellular compartments, or that persist for a restricted length of time, or that depend on particular protein modifications. These new systems may permit theselection of compounds that modulate such interactions. Furthermore, it is likely that systems will be developed that do not depend on transcription, since, in principle, any biological effect that depends on protein interactions can be used to detect them [38].

Conclusion

Interaction technology has already had a large impact on basic and applied biological research. In industry, it is being used to isolate and characterize new targets for drug development. In the near future, it will allow researchers to isolate small organic molecules, peptides, and nucleic acids that may lead to new drugs. Future applications for genome characterization and for modulation of specific protein-protein interactions are on the horizon. The ramifications of this technology promise to be exciting.

Acknowledgements.

We are grateful to Barak Cohen, Russ Finley, Tod Gulick, Ed Lavallie, John McCoy, and Jason Morris for comments on the manuscript.

References and Recommended Reading

* papers of great interest

** papers of outstanding interest

* 1. Brent R,Ptashne M: A eukaryotic transcriptional activator bearing the DNA specificity of a prokaryotic repressor. Cell 1985, 43:729-736.

Eukaryotic transcription factors may be modular with respect to their DNA binding domain and transcriptional activating domain. Activation domain of GAL4 fused to bacterial LexA repressor, a DNA binding protein, can activate transcription downstream of LexA binding sites in yeast.

* 2. Keegan L, Gill G, Ptashne M: Separation of the DNA binding from the transcription-activating function of a eukaryotic regulatory protein. Science 1986, 231:699-704.

* 3. Ma J,Ptashne M: A new class of transcriptional activators. Cell 1987, 51:113-119.

*4. Ma J,Ptashne M: Converting a eukaryotic transcriptional inhibitor into an activator. Cell 1988, 55:443-446. A eukaryotic DNA binding domain need not be covalently linked to an activation domain to activate transcription. Gal4 DNA binding domain fused to a Gal80 interacting domain activates transcription in yeast when co-expressed with Gal80 fused to the Gal4 activation domain.

5. Triezenberg SJ, Kingsbury RC, McKnight SL: Functional dissection of VP16, the trans-activator of herpes simplex virus immediate early gene expression. Genes & Development 1988, 2:718-29.

6. Triezenberg SJ, LaMarco KL, McKnight SL: Evidence of DNA: protein interactions that mediate HSV-1 immediate early gene activation by VP16. Genes & Development 1988, 2:730-42.

**7. Fields S,Song O: A novel genetic system to detect protein-protein interactions. Nature 1989, 340:245-246.

In this paper Fields makes the seminal suggestion that results in the development of the two hybrid system (see above).

*8. Chien CT, Bartel PL, Sternglanz R, Fields S: The two-hybrid system: a method to identify and clone genes for proteins that interact with a protein of interest. Proceedings of the National Academy of Sciences of the United States of America 1991, 88:9578-82.The first published implementation of the two hybrid system. This version has been used widely to isolate important protein interactors.

9. Durfee T, Becherer K, Chen PL, Yeh SH, Yang Y, Kilburn AE, Lee WH, Elledge SJ: The retinoblastoma protein associates with the protein phosphatase type 1 catalytic subunit. Genes & Development 1993, 7:555-69.

*10. Gyuris J, Golemis E, Chertkov H, Brent R: Cdi1, a human G1 and S phase protein phosphatase that associates with Cdk2. Cell 1993, 75:791-803.

The "Interaction Trap", our version of the two hybrid system is described and used to clone a human tyrosine phosphatase that interacts with human cyclin dependent kinases.

*11. Vojtek AB, Hollenberg SM, Cooper JA: Mammalian Ras interacts directly with the serine/threonine kinase Raf. Cell 1993, 74:205-14.

Two hybrid system developed by Hollenberg, R. Sternglanz, and H. Weintraub uses LexA DNA binding protein and sites.

12. Brent R, Jessen T, Golemis E, Finley R: MGH Molecular Biology Gopher Server. 1993-1994,

*13. Harper JW, Adami GR, Wei N, Keyomarsi K, Elledge SJ: The p21 Cdk-interacting protein Cip1 is a potent inhibitor of G1 cyclin-dependent kinases. Cell 1993, 75:805-16.

The most up to date description of the Elledge lab version of the two hybrid system which is used to select a protein kinase inhibitor that interacts with cyclin dependent kinases.

14. Zervos AS, Gyuris J, Brent R: Mxi1, a protein that specifically interacts with Max to bind Myc-Max recognition sites. Cell 1993, 72:223-232.

15. Golemis E,Brent R: Fused protein domains inhibit DNA binding by LexA. Mol. Cell. Biol. 1992, 12:3006-3014.

16. Ebina BA, Takahara Y, Kishi F, Nakazawa A, Brent R: LexA is a repressor of the colicin E1 gene. J. Biol. Chem. 1983, 258:13258-13261.

17. Brent R: Regulation and autoregulation by lexA protein. Biochimie 1982, 64:565-569.

18. Little JW,Mount DW: The SOS regulatory system of Escherica coli. Cell 1982, 29:11-22.

19. Schnarr M, Pouyet M, Granger-Schnarr M, Daune M: Large-scale purification, oligomerization equilibria, and specific interaction of the LexA repressor of Escherica coli. Biochemistry 1985, 24:2812-2818.

20. Thliveris A, Little J, DW M: Repression of the E. coli recA gene requires at least two lexA protein monomers. Biochime 1991, 73:449-455.

21. Brent R,Ptashne M: A bacterial repressor protein or a yeast transcriptional terminator can block upstream activation of a yeast gene. Nature 1984, 312:612-615.

22. Silver PA, Brent R, Ptashne M: DNA binding is not sufficient for nuclear localization of regulatory proteins in Saccharomyces cerevisiae. Mol. Cell. Biol. 1986, 6:4763-4766.

23. Keegan L, Gill G, Ptashne M: Separation of the DNA binding from the transcription-activating function of a eukaryotic regulatory protein. Science 1986, 231:699-704.

24. Silver PA, Keegan L, Ptashne M: Amino terminus of the yeast GAL4 gene product is sufficient for nuclear localization. Proc. Natl. Acad. Sci. U.S.A. 1984, 81:5951-

5955.

25. Carey M, Kakidani H, Leatherwood J, Mostashari F, Ptashne M: An Amino-terminal Fragment of GAL4 binds DNA as a dimer. J. Mol. Biol. 1989, 209:423-432.

26. Lin Y-S, Carey M, Ptashne M, Green MR: GAL4 Derivatives Function Alone and Synergistically with Mammalian Activators In Vitro. Cell 1988, 54:659-664.

27. Gill G,Ptashne M: Negative effect of the transcriptional activator GAL4. Nature 1988, 334:721-724.

28. Berger SL, Cress WD, Cress A, Triezenberg SJ, Guarente L: Selective inhibition of activated but not basal transcription by the acidic activation domain of VP16: evidence for transcriptional adaptors. Cell 1990, 61:1199-208.

29. Fearon ER, Finkel T, Gillison ML, Kennedy SP, Casella JF, Tomaselli GF, Morrow JS, Van DC: Karyoplasmic interaction selection strategy: a general strategy to detect protein-protein interactions in mammalian cells. Proceedings of the National Academy of Sciences of the United States of America 1992, 89:7958-62.

30. Vasavada HA, Ganguly S, Germino FJ, Wang ZX, Weissman SM: A contingent replication assay for the detection of protein-protein interactions in animal cells. Proceedings of the National Academy of Sciences of the United States of America 1991, 88:10686-90.

31. Seol W, Choi H-S, Moore D: Isolation of Proteins That Interact Specifically with the Retinoid X Receptor: Two Novel Orphan Receptors. submitted 1994,

32. Hardy CF, Sussel L, Shore D: A RAP1-interacting protein involved in transcriptional silencing and telomere length regulation. Genes & Development 1992, 6:801-14.

33. Chalfie M, Tu Y, Euskirchen G, Ward WW, Prasher DC: Green fluorescent protein as a marker for gene expression. Science 1994, 263:802-5.

34. Zhang YZ, Naleway JJ, Larison KD, Huang Z, Haugland RP: Detecting LacZ Gene Expression in Living Cells with New Lipophilic, Fluorogenic b-Glactosidase Substrates. FASEB J. 1991, 5:3108.

35. Boguslawski G: Effects of polymixin B sulfate and polymixin B nonapeptide on growth and permeability ot the yeast Saccharomyces cerevisiae. Mol. Gene. Genet. 1985, 199:401.

36. Sclafani R,Fangman W: Thymidine utilization by tut mutants and facile cloning of mutant alleles by plasmid conversion in S. cerevisiae. Genetics 1986, 114:753-767.

37. Hinnen A, Hicks J, Fink G: Transformation of Yeast. Proceedings of the National Academy of Sciences of the United States of America. 1978, 75:1929-33.

38. Fritz CC, Green, MR: Fishing for Partners Current Biology 1992, 2:403-405.

Figure Legends

Figure 1. A full-featured two hybrid system, such as the 'interaction trap'. The bait (dark grey) contains a DNA binding portion, which in turn consists of a DNA binding domain, such as the LexA amino terminal region, separated by a hinge region from a dimerization domain, such as the LexA carboxy terminus, whose interaction (grey stubble) with another such region directs the homodimerzation of the protein. The DNA binding portion is separated by a hinge region from the bait moiety. The bait is expressed from a plasmid that carries a constitutive promoter, such as the ADH1 promoter and a selectable marker, such as HIS3. The bait is bound to DNA binding site, such as a LexA operator, upstream of a reporter gene (reporter 1), such as a LEU2 reporter integrated into the chromosome. Many two hybrid systems also employ a second reporter, such as lacZ, carried either on plasmid or inserted into the chromosome, to provide a confirming phenotype and to allow visual screening in addition to selection.

The prey (light grey) contains a cDNA encoded protein fused to an activation tag (rays); it may also contain an epitope tag, and a nuclear localization sequence. It may be expressed from a conditional promoter, such as the GAl1 promoter. In the interaction trap, the cDNA encoded protein is fused to a B42 activation sequence, the HA epitope tag, and the SV40 nuclear localization signal, and is expressed from a plasmid. Detection of an interaction occurs when the prey interacts with the bait moiety at interacting surfaces (grey stubble), to activate transcription of the reporter(s) (large arrow). Figure shows the interaction occurring "somewhere inside a yeast nucleus".

Figure 2. Interaction mating schemes to test individual protein protein interactions.

2a) Top. A lawn of a haploid strain that carries the prey (top right) is mated with a grid of different colonies of the opposite mating type, each of which carries a different bait (top left). Middle. Diploids (grey circles) are selected. The plate is then replica plated to medium on which transcription of the reporter genes can be selected. Bottom. Patches of diploid cells in which interaction is occurring (black circles) grow on selective medium. Inspection of this grid ("interaction matrix") reveals proteins contacted by the prey.

2b) Top. In this similar scheme, stripes of haploid yeast carrying different baits (vertical stripes), are mated with stripes of haploid yeast of opposite mating type (horizontal stripes). Middle. Diploids (grey squares) are selected, and transferred to a plate on which transcription of the reporters can be selected. Patches in which interaction is occuring grow on this medium. As described in the text, inspection of the resulting matrix of binary protein interactions matrix may reveal information about the tested proteins, for example whether they may associate into multiprotein complexes.

Andrew Mendelsohn and Roger Brent

5 October 1994

Department of Molecular Biology

Massachusetts General Hospital

50 Blossom Street

Boston, Massachusetts 02114 and

and Department of Genetics

Harvard Medical School

mendelsohn@opal.mgh.harvard.edu

brent@opal.mgh.harvard.edu