A Two-hybrid Dual Bait System to Discriminate Specificity of Protein Interactions*

Biological regulatory systems require the specific organization of proteins into multicomponent complexes. Two hybrid systems have been used to identify novel components of signaling networks based on interactions with defined partner proteins. An important issue in the use of two-hybrid systems has been the degree to which interacting proteins distinguish their biological partner from evolutionarily conserved related proteins and the degree to which observed interactions are specific. We adapted the basic two-hybrid strategy to create a novel dual bait system designed to allow single-step screening of libraries for proteins that interact with protein 1 of interest, fused to DNA binding domain A (LexA), but do not interact with protein 2, fused to DNA binding domain B (λ cI). Using the selective interactions of Ras and Krev-1(Rap1A) with Raf, RalGDS, and Krit1 as a model, we systematically compared LexA- and cI-fused baits and reporters. The LexA and cI baitr reporter systems are well matched for level of bait expression and sensitivity range for interaction detection and allow effective isolation of specifically interacting protein pairs against a nonspecific background. These reagents should prove useful to refine the selectivity of library screens, to reduce the isolation of false positives in such screens, and to perform directed analyses of sequence elements governing the interaction of a single protein with multiple partners.

Biological regulatory systems require the specific organization of proteins into multicomponent complexes. Two hybrid systems have been used to identify novel components of signaling networks based on interactions with defined partner proteins. An important issue in the use of two-hybrid systems has been the degree to which interacting proteins distinguish their biological partner from evolutionarily conserved related proteins and the degree to which observed interactions are specific. We adapted the basic two-hybrid strategy to create a novel dual bait system designed to allow single-step screening of libraries for proteins that interact with protein 1 of interest, fused to DNA binding domain A (LexA), but do not interact with protein 2, fused to DNA binding domain B ( cI). Using the selective interactions of Ras and Krev-1(Rap1A) with Raf, RalGDS, and Krit1 as a model, we systematically compared LexA-and cI-fused baits and reporters. The LexA and cI baitr reporter systems are well matched for level of bait expression and sensitivity range for interaction detection and allow effective isolation of specifically interacting protein pairs against a nonspecific background. These reagents should prove useful to refine the selectivity of library screens, to reduce the isolation of false positives in such screens, and to perform directed analyses of sequence elements governing the interaction of a single protein with multiple partners.
To understand and manipulate the function of a particular protein of biological interest, it is generally useful to identify other proteins with which it associates. Although identification of protein interactions initially proceeded almost solely by technically difficult biochemical methods, in recent years yeast two-hybrid systems (1) have developed as a powerful genetic tool to rapidly select previously uncharacterized proteins that specifically interact with a target protein of interest from a suitable library (2)(3)(4)(5). In this schema, a protein of interest is synthesized in yeast as a fusion to a DNA binding domain (DBD), 1 which is typically the bacterial repressor protein LexA or the amino-terminal end of the yeast transcription factor GAL4. Interaction of this DBD protein fusion (a "bait") with a transcriptional activation domain-fused partner protein (either a defined partner or a novel protein screened from a library) allows the activation of reporter genes (lacZ, HIS3, LEU2) responsive to the cognate DBD. More recently, interest has focused on expanding the utility of two-hybrid systems to enable the detection of interactions between proteins and RNA (6,7), proteins and nonprotein ligands (8), proteins and peptides (9,10), and proteins and multiple partners (11,12). A second thrust has been to enable whole-genome applications (13)(14)(15), leading to the generation of maps of protein interaction networks with the potential to complement the vast resource of sequence information now being developed as part of the Genome Project. Finally, there has been interest in developing two-hybrid systems as tools in high throughput drug discovery screening strategies to identify agents regulating the activity of biologically important target proteins.
As two-hybrid technologies have evolved to more complex applications, a question of mounting importance has been the degree to which library screens performed in these systems yield partners specific for the utilized bait, as opposed to proteins of broad interaction capability ("false positives"). Although the large number of published two-hybrid papers indicates that many specific partners are obtained, a recent survey has suggested that the majority of library screens isolate at least some cDNAs that are nonspecific. 2 As a related issue, it is clear that many biologically important proteins are organized into families of evolutionarily related members that conserve substantial sequence similarity (e.g. Refs. [17][18][19]. Thus, the degree to which two-hybrid systems isolate proteins partners absolutely specific for individual baits, rather than those that interact generally with a class of protein ("familial positives"), is also an issue. Although existing two-hybrid systems allow discrimination of uniquely specific interactors from false positives or familial positives through use of various methods of specificity testing performed subsequent to a screen (20), these methods are frequently laborious, particularly when many possible interactors must be tested. For this reason, it has been of considerable interest to devise a method to eliminate such clones before selection.
In this report, we describe a novel adaptation of the twohybrid system designated the dual bait system. This system incorporates controls for false positive or nonspecific interactions in a single step and allows the simultaneous assay of a protein interaction with two related or unrelated partners in a single cell, which should also be useful for a variety of high throughput and genome-oriented studies. We demonstrated that these reagents are effective at selectively identifying two * This work was supported by an award from the Merck Genome Research Institute, by National Institutes of Health Core Grant CA06927, and by a Small Business Innovative Research program grant subcontract from Invitrogen. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

EXPERIMENTAL PROCEDURES
Molecular Biology and Genetic Techniques-DH5␣ Escherichia coli was grown on LB medium (21)where appropriate, and antibiotics were added at the concentrations recommended by the suppliers. Standard DNA manipulation techniques were as described in (21). Yeast were grown on yeast-peptone-glucose or minimal medium and manipulated using standard techniques (22). Two-hybrid experiments and ␤-galactosidase assays were performed as described (23), with six independent colonies assayed for each value presented; for sensitive plate-based X-Gal and X-Gluc assays, the procedure of Duttweiler (24) was used.
Dual Bait System Reagents-Relevant properties of all strains and plasmids are described in the text. The bacteriophage repressor protein cI (25) was used as the basis of reagent development, as its size, structure, and DNA binding properties suggested it might behave comparably as a DNA binding domain to the pre-existing two-hybrid system DBD LexA (26 -28).
cI-responsive LacZ Reporters-A 68-base pair fragment of the bacteriophage genome (LAMCG, nucleotides 37950 -38018) containing 3 naturally occurring cI operators was amplified, and XhoI ends were added by PCR. The resulting product was digested with XhoI and inserted into the XhoI site of the plasmid LR1⌬1 (parent of all currently utilized interaction trap lexAop-lacZ reporters (20)) in either orientation upstream of a basal GAL1 promoter directing expression of the lacZ gene. The resulting plasmids pcIop-LacZA and pcIop-LacZB have a 2-m origin of replication and use a URA3 marker for selection in yeast; they differ only in the orientation of the cI operator cassette.
As a separate approach, as SK10 proved refractory to deletion of the URA3 gene, the PstI (475045)-XbaI (474175) fragment of Saccharomyces cerevisiae chromosome II, located 295 base pairs upstream of the wild-type LYS2 gene, and a PstI-HindIII fragment of pcIL-1, encompassing the cI-responsive cassette, were assembled on a pUC-based plasmid to yield pcIL-2. pcIL-2 reconstructed about 6 kilobases of yeast genomic sequence with the native LYS2 promoter replaced by GAL1 promoter/cI operator sequence. This construct was used for integrative transformation of the EGY48 (4) and EGY191 (23) yeast strains, followed by Lys selection (ability to grow on 0.2% DL(␣) aminoadipic acid (a toxic metabolite of the LYS2 protein (29)) as the sole source of nitrogen. 24 stable integrants in each of the parent strains were selected and confirmed to be Lys-Leu-His-Ura-Trp-; representative integrants into each parent have been designated SKY48 and SKY191 and have the genotype (MAT␣ trp1 ura3 his3 lexAop-LEU2 cIop-LYS2), with 6 or 2 lexA operators upstream of LEU2, respectively.
cI-responsive gusA Reporters-The novel yeast vector pRG00 was used as the basis of cIoperator-gusA reporter vector construction. The detailed map and sequence files for pRG00 are available upon request; this vector contains the colE1 origin of replication and kanamycin resistance gene for selection in bacteria, a 2-m origin and URA3 gene for selection in yeast, and the complete coding sequence for E. coli ␤-glucuronidase (gusA; a gift of J. Vossen, University of Amsterdam). An EcoRI fragment containing a minimal GAL1 promoter, cI operator cassette (containing three naturally occurring cI operators), and the translational start of the GAL1 gene was excised from cIop-lacZA and recloned immediately upstream of the gusA sequences in pRG00 to create the plasmid pRG2 (cIop-gusA).
cI Fusion Bait Vectors-A DNA fragment containing the complete coding sequence (with no stop codon) for bacteriophage cI repressor protein (LAMCG, nucleotides 37230 -37940) was amplified by PCR and cloned into the plasmid pUC19 to yield pUC-cI. Separately, a HIS3containing fragment of pEG202 was removed by AatII-ClaI digestion and replaced by a synthetic AatII-ClaI linker, to create pGK202. Subsequent HindIII digestion, fill-in reaction, and EcoRI digestion were used to remove the lexA gene from pGK202, followed by replacement with the cI gene on a BglII (filled-in)-EcoRI fragment excised from pUC-cI. The resulting plasmid pGK302, was digested with BamHI and AatII and ligated to a BamHI-AatII fragment of pEG202 to create pGKS3, a pEG202 "sibling" with the cI gene exactly replacing the lexA gene. pGKS3 has a 2-m origin of replication, carries a HIS3 marker for yeast selection, and was used in control experiments.
Subsequently, a BsaBI-EcoRI fragment of pGKS3 (encompassing cI) was used to replace the lexA gene in the plasmid pLexZeo (Invitrogen), which had been digested with HpaI-EcoRI. The resulting plasmid, pGKS6, used the ADH1 promoter to express a cI fusion. It had a 2-m origin of replication and used Zeocin (Invitrogen) resistance for selection in yeast and bacteria. Expression of proteins was assayed by standard lysis of cells expressing appropriate constructs (20) followed by SDS-polyacrylamide gel electrophoresis and Western analysis with antibodies to Krev-1 (Transduction Labs, Inc.), LexA, or cI repressor (a gift of G. Kalmar).
cI Fusion Bait/cIop-gusA Reporter Dual Purpose Vectors-To allow introduction of all system components into a single yeast strain, a combined cI bait expression/cIop-gusA reporter with a single selectible marker (Zeo R ) was constructed. To create this plasmid, pGKS8, the 3cIop-gusA reporter cassette was excised from pRG2 as a KpnI-KpnI fragment and inserted into the BsrGI site of pGKS6, destroying both restriction sites. The map of pGKS8 is available on request; this plasmid retained EcoRI, SacI, BglII, PvuII, KpnI, SacII, and NotI sites for the insertion of coding sequences for expression as cI fusion proteins.
Assaying Reporter Activation-Activation of LacZ reporters was assessed qualitatively by on-plate overlay assays (24) using the substrates X-Gal or Magenta-Gal (Diagnostic Chemicals Ltd) and quantitatively using ␤-galactosidase assays as described in (21). Activation of gusA reporters was assessed qualitatively using the same overlay procedure as for LacZ but with X-Gluc (Diagnostic Chemicals Ltd) as a substrate. A quantitative assay was performed as for a ␤-galactosidase assay but using 4-nitrophenyl-␤-D-glucuronic acid instead of 2-nitrophenyl-␤-Dgalactopyranoside as the substrate. For both assays, determination was made in a plate reader; in a standard procedure, 100 l of exponential phase yeast cultures were harvested by centrifugation in a microtiter plate. Yeast pellets were resuspended in 50 l of Z-buffer, and the A 590 was determined in a plate reader, after which the plate was frozen at Ϫ70°C. After thawing, 50 l of 4-nitrophenyl-␤-D-glucuronic acid (2 mg/ml) or 2-nitrophenyl-␤-D-galactopyranoside (2 mg/ml) in Z-buffer, as appropriate, was added to each well. Immediately, the A 405 was measured, and the plate was then placed in a 30°C incubator. Additional A 405 readings were taken between 5 and 45 min, depending on the intensity of the reaction. To calculate ␤-glucuronidase or ␤-galactosidase activity, the following formula was used: where v is volume in ml, t is time, and (A 405 t Ϫ A 405 0 ) is a net increase of optical density at 405 nm between zero time and time t, and A 590 is optical density at 590 nm.
Activation of LYS2 or LEU2 reporters was accomplished by streaking or replica-plating yeast to plates lacking leucine or lysine and monitoring growth over 4 days. In assessing activation of the LYS2 reporter by bait-prey combinations (as in Fig. 5), this was done either on medium lacking uracil, histidine, tryptophan, and lysine in the presence of zeocin, or alternatively, on medium lacking only uracil, histidine, and lysine in the absence of zeocin, with the selection for growth on lysine medium providing positive selective force for the retention of cI bait and activation domain-fused prey. Qualitatively identical patterns of protein interaction are obtained with both media; however, because of minor nonspecific growth inhibitory effects of zeocin, growth is faster (ϳ24 h to point of large colonies) with zeocin omitted.
For mixing experiments, yeast were pregrown in medium selective for plasmid retention and inducing for activation domain fusion expression (ura-his-trp-galactose/raffinose ϩ zeocin), diluted as described in results, and plated to selective medium; colony outgrowth was monitored over 4 days.
In the dual bait system described here, three further components are added (Fig. 1B). The first of these is a cI-fused alternate bait, expressed from the novel Zeo R , 2-m plasmid, pGKS8. The second is a (cIop) n -gusA (␤-glucuronidase) (n ϭ 3) reporter gene cassette, additionally borne on the plasmid GKS8. The third is an additional integrated reporter system in which cI operators direct the expression of the LYS2 gene; (cIop) n -LYS2 (n ϭ 3) in the yeast strains SKY48 or SKY191 (derivatives of EGY48 and EGY191, respectively). These reagents can be utilized in multiple ways to enhance measurement of protein interactions over currently existing approaches.
As a first example, in a library screen, if an activation domain-fused interacting protein associates uniquely with a LexA-fused primary bait but not with a cI-fused alternate bait, SKY48 or SKY191 yeast containing the appropriate bait and reporter constructs would turn blue on medium containing X-Gal but not on medium containing X-Gluc, and grow on medium lacking leucine, but fail to grow on medium lacking lysine; in contrast, promiscuously interacting clones would be revealed by their growth on medium lacking both leucine and lysine and blue color with both X-Gal and X-Gluc. False positives would be eliminated simultaneously with isolation of true positive clones. As a second example, in targeted examination of the interaction of a single activation domain-fused protein with two defined partners (for example, interaction of activation domain-fused cyclin D with LexA-fused CDK4 and cI-fused CDK6), a randomly mutagenized pool of activation domainfused partners could be screened to identify mutations that disrupt interaction with either one or both of the partner proteins. As a third example, one source of interest in two-hybrid systems is their use in drug screening approaches to identify compounds that disrupt interactions between discrete pairs of interacting proteins (8,32,33); dual bait reagents would apply a simultaneous control to the specificity of such interactions.
Parallel Performance of LexA and cI Expression and Reporter Systems-Given that assessment of protein interactions in twohybrid systems is a factor of bait expression levels (34) and stringency of reporter systems (23), for these hypothetical uses to be practicable, the two-bait-reporter combinations utilized in the dual bait system must express respective baits to similar levels and possess similar sensitivities to transcriptional activation. Therefore, an initial step was to carefully measure these parameters. To this end, we constructed equivalent pEG202 (LexA) and pGKS3 (cI) fusions to the protein Krev-1/ rap1A (31), a human Ras-family GTPase. These and parent vectors were transformed in parallel into EGY48 yeast, and expression of the synthesized proteins was assayed by Western analysis using antibodies to Krev-1, LexA, or cI (Fig. 2). Expression of the two Krev-1 fusion constructs was found to be comparable in 4 randomly chosen colonies, with slightly higher levels (ϳ2-3-fold) in the cI constructs. Furthermore, expression of the fusion protein was in each case similar to the matching unfused DNA binding domain, indicating that cI tolerated attachment of a fusion domain without loss of stability. Finally, essentially identical expression levels were observed using pGKS6-Krev-1, a Zeo R instead of HIS3 version of pGKS3 (not shown), indicating the selectable marker could be exchanged without gross alteration of plasmid copy number.
We next determined the degree to which activation through cI operators was comparable with activation through lexA operators. As a conservative first step, we constructed analogous fusions of pGKS3 and pEG202 to Krit1 (a Krev-1-interacting protein (30) that fortuitously functions as a transcriptional FIG. 1. Interaction trap and dual bait systems. A, an activation (Act.) domain-fused prey interacts with a LexA-fused bait to drive transcription of lexAop-responsive LEU2 and lacZ reporters. B, an activation domain-fused prey interacts with a LexA-fused bait to drive transcription of lexAop-responsive LEU2 and lacZ reporters but does not interact with a cI-fused bait and, thus, does not turn on transcription of cIop-responsive LYS2 and gusA reporters. Note, as shown here, that cI bait is drawn as representing a nonspecific partner; the system can also be configured so prey interacts with both baits; see text.
activator of moderate strength) and assayed activation of the closely related cIop-and lexAop-lacZ reporters. Parallel transformations were performed with pGKS3-Krit1 (cI-Krit1) plus cIop-lacZA and cIop-lacZB, which contained the three naturally occurring cI operators O R1 -O R3 of phage in either orientation (25). In addition, for comparison, yeast were transformed with pEG202-Krit1 (LexA-Krit1) plus pRB1840, pJK103, or pSH18 -34 (1, 2, or 8 lexAop-lacZ) (23), and as a negative control, with pEG202-Krit1 plus cIop-lacZA. ␤-Galactosidase assays were used to measure activation of the lacZ reporters (Fig. 3). In these tests, the cI-Krit1 fusion protein activated the two cIop-lacZ constructs to equivalent levels, which were closely comparable with that obtained using the combination of LexA-Krit1 and pJK103. As a negative control, the LexA-Krit1 construct was also shown not to activate the cIop-lacZ reporters, as expected. Based on this result, we next used the cIop/ minimal promoter cassette to develop a cIop-gusA reporter, pRG2, to be used in conjunction with the standard lexAop-lacZ reporters. Yeast were transformed with the reporter and an activating cI-Krit1 fusion protein or LexA-fused Ras (which does not activate transcription) as a negative control, and gusA transcription was assessed with a quantitative ␤-glucuronidase assay, analogous to a ␤-galactosidase assay (see "Experimental Procedures"); as with the cIop-lacZ reporters, a high degree of specific activation was observed against a much lower background (6300 ␤-glucuronidase units for cI-Krit1 versus 85 for the negative control, a 75-fold difference), although overall the gusA reporter was more sensitive than lacZ, as reflected in the higher units (700 ␤-galactosidase units for cI-Krit1). In plate overlay assays, positive and negative were similarly clearly distinct, with cI-Krit1 medium to dark blue, whereas the negative LexA-fused control was pale blue or white (not shown; also, see below).
Finally, we compared direct activation of the LEU2 versus LYS2 auxotrophy reporters, again using analogous LexA-and cI-fused Krit1 (Fig. 4). Using SKY48 and SKY191 as hosts, we determined that LexA-Krit1 activates the LEU2 reporter of these strains, whereas cI-Krit1 does not (Fig. 4, second panel  from top). Conversely, cI-Krit1 is capable of activating the LYS2 reporter of SKY strains, whereas LexA-Krit1 is not (Fig.  4, third panel from top). Finally, each fusion activated the appropriate lacZ reporter to comparable degrees, independent of growth properties on Leu or Lys medium (Fig. 4, bottom   panel, middle two rows). Notably, positive growth dependent on activation of the LEU2 and LYS2 reporters could be assessed in similar time frame, with results detectable at 24 -48 h after plating yeast on selective media. Based on visual estimation of growth rate, the sensitivity of the cIop-LYS2 reporter in these strains appears to be intermediate between that of the LEU2 reporters in EGY48 and EGY191.
Cumulatively, these results indicated that the cI-and LexAbased expression and reporter constructs yielded results in a similar sensitivity range, making them suitable for comparative purposes. One point remaining was the development of reagents suitable for expressing all baits and reporters in the same strain, to allow simultaneous assay. SKY strains (MAT␣ trp1 ura3 his3 lexAop-LEU2 cIop-LYS2) utilize the LEU2 and LYS2 markers for reporter genes. The activation domain fusion expression plasmid (pJG4 -5) uses a TRP1 marker; the lexAop-LacZ reporter (pJK103) uses a URA3 marker; and the LexA fusion bait plasmid (pEG202) uses a HIS3 marker; the cI fusion bait plasmid (pGKS6) utilizes a Zeo R marker. To introduce the cIop-gusA reporter, we took advantage of the fact that the cIop-gusA cassette is only 2.6 kilobases, whereas the plasmid backbone for the cI fusion plasmid pGKS3 is unusually small, as the Zeo R marker is used for both bacterial and yeast selection. The cIop-gusA cassette was introduced into pGKS6, resulting in a new plasmid, pGKS8, which encompassed both cI-bait and cIop-gusA reporter. Control experiments similar to those outlined above demonstrated that this bait-reporter-combined plasmid yielded results similar to those obtained with the combination of pGKS3 and pRG2 (not shown). This construct was used for the experiments described in the following sections.
Specificity of the Dual Bait System in Controlled Two-hybrid Assay-The major criterion for effective use of a dual bait system is that it should effectively discriminate interactions of a partner with related but distinct proteins. Ras and Krev-1 possess 56% amino acid identity and are known to interact with an overlapping set of protein partners (35)(36)(37). In experiments described elsewhere, we determined that Raf preferentially interacts with Ras by a two-hybrid system assay, whereas Krit1 preferentially interacts with Krev-1 (30). The Ral guanine nucleotide dissociation stimulator RalGDS interacts with both Krev-1 and Ras (36). Neither Ras nor Krev-1 activates transcription when expressed as a DNA binding domain fusion.
All yeast grew on nonselective plates (uraϪ hisϪ trpϪ, glucose, or galactose, Fig. 5, panel A). No strains grew on either Leu or Lys plates when glucose was present as the carbohydrate source. However, under galactose induction, strains containing pJG4 -5-Raf were able to grow preferentially on Leu medium (Fig. 5, panel F) but only minimally on Lys medium, based on the association between Raf and LexA-Ras; conversely, strains containing pJG4 -5-Krit1 grew well on Lys medium but only weakly on Leu medium, based on the interaction between Krit1 and cI-Krev-1 (Fig. 5, panel E). Strains containing pYesTrp2-RalGDS grew well on both Lys and Leu medium (Fig. 5, panels E and F), whereas a negative control (strains containing empty plasmid pJG4 -5) did not grow on any selective plates (Fig. 5, panels E and F). Interaction of RalGDS with both baits could be also detected on the double auxotrophic Lys-Leu plate, where this was the sole plasmid combination resulting in growth (Fig. 5, panel G). Interactors that associated with only the cI-fused bait or nonselectively, with both the cI-and LexA-fused baits, could be counterselected by inclusion of ␣-aminoadipate in the growth medium as the sole source of nitrogen (Fig. 5, panel H).
The results of X-Gal and X-Gluc assay on the plates are in good correspondence with the auxotrophic selection assay, with Raf-Ras positive with X-Gal (Fig. 5, panel C), Krev1-Krit1 positive with X-Gluc (Fig. 5, panel B), and Ral-GDS positive with both (Fig. 5, panel B and C). Note: using a complementary set of color-producing substrates (Magenta-Gal ϩ X-Gluc), both LacZ and GusA activities can also be assayed on a single plate (Fig. 5, panel D). These results paralleled those previously obtained using a conventional two-hybrid selection (30) and confirmed that the dual bait system can be used to distinguish interactions between two closely related potential partner proteins. We note that although the ␣-aminoadipate counterselection works well in the controlled situation shown above, over time background colonies arose on the counterselected plates, suggesting this particular application may be more useful in targeted disruption of known interactions than in conjunction with library screens.
Selection of Specifically Interacting Protein Pairs from a Nonspecific Pool-The previous results demonstrated that direct streaking of uniform populations of yeast containing predetermined combinations of baits, activation domain fusions, and reporters yields expected results. A more rigorous test of the power of the ability of these reagents to discriminate specific interactions was performed using a mixing experiment, as outlined in Fig. 6. Four populations of SKY191 yeast were generated. Each expressed LexA-Ras and cI-Krev-1 and contained lexAop-lacZ and cIop-gusA reporters and an activation domain fused to either 1) Raf, 2) RalGDS, 3) Krit1, or 4) nonspecific (a fragment of hsp90). 10 -100 colony forming units (ϳ30 -300 cells), each of populations 1-3, were mixed together with 2 ϫ 10 6 cells containing the nonspecific control and parallel samples of the pooled cells plated to media selective for the lexAop-LEU2 (ura-his-trp-leu-Zeo) or cIop-LYS2 (ura-his-trplys-) reporters.
Approximately 50 colonies arose on each of these plates, in good accord with the number anticipated based on the seed. Of these, 24 were chosen from each of the Leu and Lys plates and transferred to a master plate, then retested for growth on both leuϪ and lysϪ medium, as well as activation of lacZ or gusA reporters. 43 of the 48 analyzed colonies were specific for growth on Leu or Lys medium, whereas 5 of the 48 total colonies analyzed grew on both Leu and Lys medium. In tests with the colorigenic reporters, 45 of the 48 displayed expected patterns for lacZ and gusA; LEUϩ colonies were blue with X-Gal, LYSϩ colonies were blue with X-Gluc, LEUϩLYSϩ colonies were blue with both; the remaining 3 colonies were white with both substrates. Finally, 5 colonies for each group (LEUϩ LacZϩ LYS-gusAϪ, LEUϩ LacZϩ LYSϩ gusAϩ, LEUϪLacZ-LYSϩ gusAϩ) were selected at random and used for PCR using primers containing sequences complementary to library vector sequence-flanking inserts to identify inserts based on characteristic product size (Fig. 6, bottom). Separately, each individual library plasmid was amplified to provide a size standard for Krit1, Raf, and RalGDS). As shown, each of the selected LEUϩLYSϪ colonies contained Raf, each of the LEU-LYSϩ colonies contained Krit1, and each of the LEUϩLYSϩ colonies contained RalGDS as predicted, based on their reporter activation profile. In contrast, PCR from the three colonies with an inappropriate pattern of activation (e.g. LYSϩ but gusAϪ) revealed the presence of the hsp90 fragment (data not shown). In numerous tests of simple two-hybrid system reagents, such as the interaction trap, we previously found that a common source of false positives is selection of strains mutated so as to transcribe the auxotrophic reporter (LEU2) in the absence of contribution from either bait or prey; these false positives are standardly detected because of their inability to additionally activate the second chromogenic reporter (20). The detection of the nonspecific hsp90 fragment in the LYSϩ gusAϪ cells suggests that these colonies represent similar such selected mutant strains and supports the idea that the presence of two distinct reporters continues to provide a useful specificity for the system. DISCUSSION In this report, we describe the development and characterization of novel dual bait reagents that can be used to study the interaction of a protein with two distinct partners in a single yeast cell. The cI repressor/cI operator system utilized in the SKY yeast strains and cIop-LacZ plasmids is demonstrated to function with a sensitivity range closely comparable with the pre-existing LexA repressor/lexA operator system in the interaction trap, facilitating their combined use. In a model system assaying the interaction of the related GTPases Ras and Krev-1 with their preferred partners Raf and Krit1 and their shared partner RalGDS, the dual bait system clearly differentiates higher affinity versus lower affinity interactions. In addition to effectiveness in discriminating interactions in grids of yeast streaked to selective plates, the discrimination observed is sufficiently robust to allow the isolation of yeast containing specifically interacting protein pairs against a vast excess of noninteracting pairs. These properties support the idea that these reagents will be useful in library screening and genomescale applications. The reagents described here offer the option of performing two independent, simultaneous screens in a single yeast, with one set of positives registering through lexAop activation of LEU2 and lacZ and a second set through cIop activation of LYS2 and gusA, both negatively controlled against each other. Together, these developments have the potential to greatly expand two-hybrid system contributions to studies of biological interactions. Several groups have recently described the use of two simultaneously expressed baits to identify mutations that selectively affect interactions of an activation domain-fused protein with one of two partners (38 -40). In each case, introduction of a second bait-reporter system was obtained by eliminating one of the two reporters used for the primary bait, greatly reducing the screening power available to the system. However, in these novel reagents, both baits retain two distinct reporters, greatly facilitating the screening process. The value of having two separate reporters is evident even in the mixing experiment performed here (Fig. 6), as their use allowed the immediate discrimination of the noninteracting background clones from specific partners, which activated both reporters. The dual bait reagents described here can be similarly used for mutational analysis and have been recently used to successfully identify mutations in Pak1 kinase, which selectively reduce interaction affinity for either of two partners, the Cdc42 or Rac GTPases. 3 Finally, there is preliminary evidence that these reagents possess the power necessary to perform library screens in organisms with complex genomes; as in several recent library screens using the above reagents, cI-and the LexA-fused baits have yielded specific partners. 4 FIG. 5. Discrimination of interaction specificity by dual bait. All yeast colonies shown express LexA-Ras (from pEG202) and cI-Krev-1 (from pGKS8) and contain lexAop-lacZ (pJK103) and cIop-gusA (pGKS8) reporters and integrated lexAop-LEU2 and cIop-LYS2 (SKY191) reporters. Yeast additionally contain the AD fusion vector (top row), AD-Krit1 (second row), AD-RalGDS (third row), AD-Raf (fourth row). Panel A, growth on nonselective medium; panel B, growth on plate assay for ␤-glucuronidase activity with X-Gluc; panel C, growth on plate assay for ␤-galactosidase activity with X-Gal; panel D, dual assay for for ␤-glucuronidase and ␤-galactosidase activity with X-Gluc and Magenta-Gal; panel E, growth on plates lacking lysine; panel F, growth on plates lacking leucine; panel G, growth on plates lacking both leucine and lysine; panel H, growth on plates lacking leucine but containing ␣-aminoadipic acid (␣AA) (counterselective for LYS2). Note: less yeast are generally plated when assaying for ␤-glucuronidase than for ␤-galactosidase because of the increased sensitivity of the gusA reporter; similarly, when performing dual assays, an initial overlay is done with agar containing Magenta-Gal, followed by a subsequent addition of X-Gluc.
On a more basic level, the system allows considerable savings in the effort required by individual investigators wishing to perform multiple two-hybrid screens without invoking the dual bait selectivity function. Instead of having to perform two separate library transformations or matings (generally the most laborious step in a screening process) and subsequent selection of positive clones, only one such step is required for any two baits. Furthermore, in the case where only one-half of a dual bait screen of a previously untested cDNA library is positive (e.g. if a LexA bait yields interactors but a cI bait does not), the fact that positive interactors are obtained for at least one bait will provide useful information toward the determination of whether the library utilized was of acceptable quality. Finally, it has previously been noted that some proteins of interest for cDNA library screening perform better with particular fusion domains (i.e. are productively utilized as LexA fusions but not as GAL4 fusions; or vice versa (23)). In cases where specific DNA binding domain/fusion domain problems are suspected, an investigator could express a bait of interest both as a LexA and as a cI fusion and screen with the bait in both configurations to maximize chances of obtaining valid positive interacting partners.
The dual bait reagents are built upon the interaction trap form of two-hybrid system (4). cI and LexA are similar in size (237 versus 202 amino acids) and structure (41) and use related amino-terminal helix-turn-helix domains to bind palindromic operator sites with similar K m values, ranging from 200 pM to 20 nM for LexA (discussed in Estojak et al. (23)) similar to cI (42). Because of these many similarities, it is clear that the LexA and cI systems are well matched. However, in addition to use in the current interaction trap shell, the cI "add-on" parts of this system have been constructed to potentially supplement any of the currently existing two-hybrid variants. Thus, the reporter system developed in this study purposely uses a DNA binding domain (cI), reporter genes (gusA and LYS2), and plasmid marker (zeocin resistance) not in use in any other two-hybrid-based system (2,3,5), including the recently described membrane-based Sos recruitment system (16). Thus, these reagents could readily be integrated with any of the other screening systems operating on two-hybrid principles; in the case of the Sos recruitment system, this raises the possibility that with minor modifications of the library vector, a single bait could be simultaneously used to identify interactors using either a membrane-based or a transcriptional activation-based selection strategy, enlarging the potential pool of interacting proteins obtained. An additionally useful feature of the gusA reporter is that it is assayed using similar protocols even on the same plates as the lacZ reporter standardly used in two-hybrid systems, again contributing to ease of use. Finally, although the dual bait reagents here described have been optimized for use in conjunction with LexA fusions, parameters have been previously established to test and vary sensitivity levels (23), making merging of two-hybrid systems a relatively simple and certainly useful effort that should contribute to efforts to understand complex protein-protein interactions on the genome scale.