Two-hybrid systematic screening of the yeast proteome

Transcription

1 Two-hybrid systematic screening of the yeast proteome Nicolas Lecrenier, Françoise Foury, and Andre Goffeau* Summary The yeast two-hybrid system is a genetic method that detects protein-protein interactions. One application is the detection by library screening of new interactors of a protein of known function. In the August issue of Nature Genetics, Fromont-Racine et al. 1 showed for the first time that the construction of the protein interaction map of a complex pathway, such as that of the mrna splicing machinery, is now possible, because of the combination of recent technical improvements elaborated in several laboratories. With a yeast cell mating procedure that increases screen efficiency, they used their complex yeast genomic library of clones to test interactions against 15 proteins. They identified and classified 170 potential interactors, including approximately 70 proteins of previously unknown function. More than 25% of the interactors are probably biologically relevant. The achievements of Fromont-Racine et al. have opened the way to the systematic analysis of the protein interaction networks of the 6,000 open reading frames-yeast proteome. This task requires, however, automation of the library screens and creation of a two-hybrid library database. BioEssays 20:1 5, John Wiley & Sons, Inc. INTRODUCTION By the fall of 1997, the complete genomic sequences of a dozen of microorganisms, including Saccharomyces cerevisiae, Escherichia coli and Bacillus subtilis had already been published, and many more are expected in the coming years. 2 One surprising finding is that more than one-third of the genes encode unknown functions. It is becoming more Laboratoire de Biochimie Physiologique. Place Croix du sud 2/ Louvain-La-Neuve. Belgium. Contract grant sponsor: Pôle d Attraction Inter-Universitaire. *Correspondence to: Andre Goffeau, Laboratoire de Biochimie Physiologique. Place Croix du sud 2/ Louvain-La-Neuve. Belgium. Goffeau@fysa.ucl.ac.be Abbreviations: AD, Transcription Activation Domain; BD, DNA Binding Domain; ORF, Open Reading Frame; UAS, Upstream Activating Sequence and more evident that the vast majority of the metabolic functions are achieved by large protein complexes. Therefore, the systematic identification of interactions occurring with a protein, or a set of proteins of already known function, would be of great help to identify functions of the newly unravelled proteins. Methods to detect protein-protein interactions have been reviewed by Phizicky and Fields. 3 Among these, the yeast two-hybrid system 4 exploits the finding that many transcription activators are composed of two domains that are physically separable and remain active provided they are in close proximity. The binding domain (BD) tethers specific DNA sequences localized in the promoter region of genes regulated by the transcription activator, whereas the transcription activation domain (AD) in concert with the RNA polymerase machinery induces transcription (Fig. 1A). In the two-hybrid system, a given protein is expressed as a fusion with a DNA-binding domain, such as that of Gal4p (Fig. 1B) or LexAp (bait fusion), whereas a putative interacting protein is fused to a transcription activation domain, such as that of Gal4p (Fig. 1C) or Herpes simplex VP16 protein (prey fusion). The bait and the putative prey, borne by distinct plasmids, are co-expressed into yeast. Their interaction, BioEssays 20:1 6, 1998 John Wiley & Sons, Inc. BioEssays

2 Figure 1. Scheme of the two-hybrid system. A: The entire Gal4 protein binds to upstream activating sequence (UAS) of GAL promoter and activates the transcription. B: Bait protein binds to Gal UAS but cannot activate the transcription without transcription activation domain. C: The prey cannot bind to Gal UAS and does not activate transcription. D: The interaction between a bait and a prey reconstitutes Gal4p properties and activates transcription of the reporter genes. BD, DNA Binding Domain of Gal4p; AD, transcription Activation Domain of Gal4p. which brings the transcription activator moiety in proximity of the transcription complex, is detected by using reporter genes such as lacz or prototrophic markers (e.g. HIS3, ADE2) that are under the control of upstream activating sequences (UAS) recognized by the DNA binding domain moiety of the bait (Fig. 1D). Only those transformants in which bait-prey interactions occur will grow on media lacking the selective histidine or adenine markers or develop a blue color in the presence of X-gal. 5-9 The large scale systematic search for protein-protein interactions implies the construction of highly redundant cdna or genomic multi-copy plasmid libraries of preys so that interactions with several overlapping prey sequences can be tested. So far, despite the abundant literature devoted to the two-hybrid system, only small genomes, such as that of the bacteriophage T7 10, have been exhaustively screened. The technical improvements described by Fromont-Racine et al. 1 have built up new foundations for the systematic screening of complex libraries and more especially that of the S. cerevisiae genome. TECHNICAL STRATEGY FOR A LARGE-SCALE TWO-HYBRID SCREENING The yeast library constructed by Fromont-Racine et al. 1 contains 5x10 6 clones with inserts of 700-bp average size fused to the 3 end of Gal4AD. Statistically, a new fusion event occurs every fourth nucleotide. Thus, even though 2 BioEssays 20.1

3 theoretically, only one of six fusions contains an open reading frame (ORF) expressed in frame with Gal4AD, this library is impressively redundant and fulfills the requirements for recovering several overlapping fusions. This is an important feature because, due to instability or incorrect folding, not all of the expected fusions are expressed and recovered. In this respect, the small size of the library inserts constructed by Fromont-Racine et al. 1 may be a limitation to the recovery of preys requiring a large domain for their interaction. The library constructed by James et al. 11, which is also highly complex, may have the advantage of containing inserts of an expected average size of 1,000 to 1,500 bp (Philip James, unpublished results). Libraries are usually kept in E. coli or as a DNA pool; however, Fromont-Racine et al. 1 have preferred to keep their two-hybrid library in yeast (13 x 10 6 transformants) in the form of a large number of cell aliquots stored at 70 C. Consequently, a large collection of homogenous ready-to-use samples is available. However, the yeast transformation step must be carried out at a three- or four-fold multiplicity to avoid creating a bias in the library so that the final number of yeast diploids that must be tested is considerably increased. In former two-hybrid strategies, bait- and prey-expressing plasmids were introduced into the same cell by sequential or co-transformations. However, the number of interactions that could reasonably be tested was low. More recently, the yield was considerably increased by mating the haploid yeast strain, which contains the bait, with a representative sampling of the prey-containing transformants that were stored at 70 C. 10,12,15 Fromont-Racine et al. 1 mated their strains on filters and directly spread the mating mixture on selective media, avoiding the traditional tedious cross-replica-platings and huge number of Petri-plates required to select all of the diploids that express interactions (Fig. 2). The prey candidates were selected first for their ability to induce the expression of the HIS3 reporter gene and second for the blue color in the presence of X-gal produced by the expression of lacz. False-positive preys, which mostly are promoter-dependent transcription activators, are common problems of the two-hybrid system. Many methods have been developed for their identification, but they are all extremely tiresome for a systematic screening. In fact, false positives can be progressively identified by comparing several screenings of the same library with unrelated baits. Moreover, the use of a yeast strain, such as PJ69-4A 11, which contains three different reporters (ADE2, HIS3 and lacz), under the control of different GAL4p activated promoters (Gal2, Gal1 and Gal7 promoters, respectively), considerably decreases the number of false positives. In general, a prey that can activate these three reporter genes is a true positive. ANALYSIS OF THE PRE-mRNA SPLICING NETWORK To build up an interaction map of the yeast pre-mrna splicing proteins, ten baits of well known function involved in this pathway were chosen by Fromont-Racine et al. 1 Four preys identified in this first screen were, in turn, used as baits for second-round screens followed by a third round using as a bait a prey identified in the second screen. Altogether, 15 screens were carried out, testing nearly interactions. Even though the library has not been completely covered for each bait, its high redundancy has certainly allowed representative data to be obtained. Depending on the bait, from 58 up to 17,000 His colonies were recovered in each screen. However, the number of double-positive clones, His LacZ varied from 10 to 86, a more reasonable range. Altogether, 15 screens identified approximately 700 His LacZ candidates, most of which were redundant. For most of them, a partial sequence of the fusion and identification of the ORF were performed. It must be noted that ORFs which are not fused in frame with Gal4AD need not necessarily be depicted as false positives. Indeed, frameshift events occurring during translation can restore the production of an in-frame functional prey capable of interaction with the bait The preys were divided into different categories by Fromont-Racine et al. 1 Group B includes all those fused polypeptides that have no biological significance, because they are located in intergenic regions or are in the reverse orientation of an ORF. Group A, which includes 170 preys located in coding regions and corresponding to 145 distinct ORFs, has been subdivided into four categories. Group A1 includes 32 preys that were recovered as several independent overlapping clones, groups A2 and A3 contain 55 preys recovered as single fusions predicted as rare events, whereas 83 preys that did not match these criteria were listed in group A4. Nine preys corresponded to known pre-mrna-splicing factors, and five revealed new splicing factors that also could have been identified by searching databases. Eight were other RNA processing factors and 45 were involved in other functions, highlighting possible new connections between splicing and other metabolic pathways. Ten were transcriptional activators, mostly false positives. The function of 50% of the preys is unknown. Based on the preys that have a biological function, we have estimated that approximately 60% and 40% of the preys in categories A1 and A2-A3, and less than 25% in category A4, respectively, are probably biologically relevant. These estimations would confirm the observation that 93% of the splicing factors were recovered in categories A1 and A2-A3, whereas only 7% were listed in category A4. Therefore, the classification used by Fromont- Racine et al. 1 is prone to select the most relevant preys. BioEssays

4 Figure 2. Two-hybrid screening strategy using cellular mating on filter. The bait and the prey are expressed by the MAT and MATa strains, respectively. Of course, several exceptions can be mentioned. Mth1p, a repressor of hexose transport, which was classified in A2-A3, also has been found in our laboratory by using an unrelated bait. Swip1p, a transcriptional activator which is most likely a false positive, was listed in A1; however, it is a potential prey for so many different baits that it is immediately highly suspect. More troublesome is the case of Gin10p, classified in the dubious category A4, which has recently been involved in pre-mrna splicing. 19 However, erroneous assignments have been smoothed down considerably by the use of several baits in related pathways and reciprocal bait-prey interaction analysis. For instance, the Prp11p, Prp21p and Prp9p mrna splicing factors, used as initial baits, also were recovered as preys in the A1 or A2-A3 categories. Similarly, the Ynr053c ORF of unknown function, listed in A2-A3, has been identified as an interactor of the Prp11p, Prp9p and Prp21 splicing factors and several other snrnr components. The central position 4 BioEssays 20.1

5 of Ynr053c in the RNA splicing machinery is of special interest because it bears 55% sequence identity with a human protein responsible for auto-immune responses. PERSPECTIVES With the use of a complex high quality library, a strategy based on a small set of carefully selected baits and several rounds of library screens, Fromont-Racine et al. 1 showed for the first time, that it is possible to identify among the 6000 ORFs of the S. cerevisiae genome 20 quite a large number of interactors in a given metabolic pathway. The estimate that approximately 25% to 30% of the double positive His LacZ clones contain interactors of biological interest is quite encouraging. Future studies will tell us whether this protein interaction map of components of the mrna splicing pathway published in Nature Genetics is exhaustive; nevertheless, the present data will lead to a large set of new experimental strategies to better understand the physiological significance of the predicted interactions and the potential links to other metabolic pathways. The improved two-hybrid system allows the performance of a complete screening with one bait in less than three weeks. Moreover, several screenings, using different baits, can be carried out simultaneously. It is clear, however, that even though a small group of scientists can achieve the interaction map of a given metabolic pathway in a reasonable amount of time, the exhaustive analysis of the yeast proteome by the two-hybrid system would still be a titanic task. Special efforts are obviously required for automation of the screening procedure, in particular for identification of the preys, possibly by hybridation on high density oligonucleotide arrays, and also for the creation of a public database that collects the two-hybrid data from the scientific community. ACKNOWLEDGMENTS We thank Dr. J.-C. Jauniaux and R. Poirey (German Cancer Research Center, Heidelberg, Germany) for their expertise on the two-hybrid system. We are grateful to Dr. M. Ghislain for his critical reading of the manuscript and S. Pirlet for her assistance. This work was supported by the Pôle d Attraction Inter- Universitaire. F.F. is Directeur de Recherche at the Fonds National pour la Recherche Scientifique. N.L. is supported by a fellowship from the Fonds pour la Formation à la Recherche dans l Industrie et dans l Agriculture. REFERENCES 1 Fromont-Racine, M., Rain, J.-C., and Legrain, P. (1997). Toward a functional analysis of the yeast genome through exhaustive two-hybrid screens. Nat Genet 16, Pennisi, E. (1997). Laboratory workhorse decoded. Science 277, Phizichy, E.M., and Fields, S. (1995). Protein-protein interactions: Methods for detection and analysis. Microb Rev 59, Fields, S., and Song, O. (1989). Novel genetic system to detect protein-protein interactions. Nature 340, Chevray, P.M., and Nathans, D. (1992). Protein interaction cloning in yeast: Identification of mammalian proteins that react with the leucine zipper of Jun. Proc Natl Acad Sci USA 89, Durfee, T., Becherer, K., Chen, P., Yeh, S., Yang, Y., Kiburn, A.E., Lee, W., and Elledge, S.J. (1993). The retinoblastoma protein associates with the protein phosphatase type 1 catalytic subunit. Genes Dev 7, Hannon, G.J., Demetrick, D., and Beach, D. (1993). Isolation of the Rb-related p130 through its interaction with CDK2 and cyclins. Genes Dev 7, Zervos, A.S., Gyuris, J., and Brent, R. (1993). Mxi1, a protein that specifically interacts with Max to bind Myc-Max recognition sites. Cell 72, Chiu, M.I., Katz, H., and Berlin, V. (1994). RAPT1, a mammalian homologue of yeast Tor, interacts with the FKBP12/rapamycin complex. Proc Natl Acad Sci USA 91, Bartel, P.L., Roecklein, J.A., SenGupta, D., and Fields, S. (1996). A protein linkage map of Escherichia coli bacteriophage T7. Nat Genet 12, James, P., Halladay, J., and Craig, E.A. (1996). Genomic libraries and a host strain designed for highly efficient two-hybrid selection in yeast. Genetics 144, Bendixen, C., Gangloff, S., and Rothstein, R. (1997). A yeast mating-selection scheme for detection of protein-protein interactions. Nucleic Acids Res 22, Gangloff, S., McDonald, J.P., Bendixen, C., Arthur, L., and Rothstein, R. (1994). The yeast type I topoisomerase Top3 interacts with Sgs1, a DNA helicase homolog: a potential eukaryotic reverse gyrase.mol Cell Biol 12, Bendixen, C., Gangloff, S., and Rothstein, R. (1994). A yeast mating-selection scheme for detection of protein-protein interactions. Nucleic Acids Res 9, Finley, R.L. Jr., and Brent, R. (1994). Interaction mating reveals binary and ternary connections between Drosophila cell cycle regulators. Proc Natl Acad Sci USA 91, Farabaugh, P.J. (1993). Alternate reading of the genetic code. Cell 74, Sourdive, D.J.D., Transy, C., Garbay, S., and Yaniv, M. (1997). The bifunctional DCOH protein binds to HNF1 independently of its 4- carbinolamine dehydratase activity. Nucleic Acids Res. 25, Stahl, G., Bidou, L., Rousset, J.-P., and Cassan M. (1995). Versatile vectors to study recoding: conservation of rules between yeast and mammalian cells. Nucleic Acids Res 23, Fabrizio, P., Laggerbauer, B., Lauber, J., Lane, WS., and Lührmann, R. (1997). An evolutionarily conserved U5 snrnp-specific protein is a GTP-binding factor closely related to the ribosomal translocase EF-2. EMBO J 16, Goffeau, A., Barrell, B.G., Bussey, H., Davis, R.W., Dujon, B., Feldmann, H., Galibert, F., Hoheisel, J.D., Jacq, C., Johnston, M., Louis, E.J., Mewes, H.W., Murakami, Y., Philippsen, P., Tettelin, H., and Oliver, S.G. (1996). Life with 6000 genes. Science 274, BioEssays