Gγ recruitment systems specifically select PPI and affinity-enhanced candidate proteins that interact with membrane protein targets

Protein-protein interactions (PPIs) are crucial for the vast majority of biological processes. We previously constructed a Gγ recruitment system to screen PPI candidate proteins and desirable affinity-altered (affinity-enhanced and affinity-attenuated) protein variants. The methods utilized a target protein fused to a mutated G-protein γ subunit (Gγcyto) lacking the ability to localize to the inner leaflet of the plasma membrane. However, the previous systems were adapted to use only soluble cytosolic proteins as targets. Recently, membrane proteins have been found to form the principal nodes of signaling involved in diseases and have attracted a great deal of interest as primary drug targets. Here, we describe new protocols for the Gγ recruitment systems that are specifically designed to use membrane proteins as targets to overcome previous limitations. These systems represent an attractive approach to exploring novel interacting candidates and affinity-altered protein variants and their interactions with proteins on the inner side of the plasma membrane, with high specificity and selectivity.

Scientific RepoRts | 5:16723 | DOI: 10.1038/srep16723 membrane proteins as targets. The updated method allows the Gγ recruitment system to be used in the analysis of both cytoplasmic and membrane target proteins.

Results
Selection of candidate proteins interacting with membrane protein targets using a previously established PPI-detecting Gγ recruitment system. First, we tested whether the previous Gγ recruitment system could target membrane proteins. In the previous system, the Fc protein of human immunoglobulin G (IgG) and the Z domain of Staphylococcus aureus protein A (Z WT ) 46 were used for the PPI models. Several Z variants (Z WT , Z K35A , Z I31A and Z 955 ) with varied affinities for the Fc protein were also used for the PPI models (Z WT , 5.9 × 10 7 M −1 ; Z K35A , 4.6 × 10 6 M −1 ; Z I31A , 8.0 × 10 3 M −1 ; and Z 955 , none) 47,48 . In contrast to the previous system, target protein 'X' was set to localize to the inner leaflet of the plasma membrane (previously, target 'X' was fused to Gγ cyto in the cytosol), and candidate protein 'Y 1 ' was fused to Gγ cyto (previously, candidate protein 'Y 1 ' was artificially localized to the inner leaflet of the membrane) (Fig. 1A,B). As the fictive model of target protein 'X, ' the Fc fragment was fused to the lipidation motifs in this study (Fig. 1B). It was also notable that the lipidation motifs were fused to the Fc fragment at both the N-terminus (Gpa1p motif; Gpa1N) and the C-terminus (Ste18p motif; Ste18C) to test the accessibility between the Fc fragment and the Z variants (the C-terminal Ste18p motif was used to express the Z variants as the candidate 'Y 1 ' proteins described in the previous study) (Fig. 1B). As the models of 'Y 1 ' proteins for the candidate library, the Z variants were fused to the C-terminus of Gγ cyto to express Gγ cyto -Y 1 fusion proteins in the cytosol (Fig. 1B).
To express the target membrane proteins, the genes encoding the Fc fragment attached to artificial lipidation motifs were stably integrated into the ste18 locus of an a-type haploid yeast chromosome, Figure 1. Schematic diagram of Gγ recruitment systems to detect PPIs of cytosolic or membrane target proteins. (A) Schematic outline of the previously established Gγ recruitment system for cytosolic target proteins. When target protein 'X' fused to Gγ cyto interacts with candidate protein 'Y 1 ' , the Gβ and Gγ cyto complex (Gβ γ cyto ) migrates to the inner leaflet of the plasma membrane and restores the signaling function. If protein 'X' cannot interact with protein 'Y 1 ' , Gβ γ cyto is released into the cytosol, and signaling is blocked. (B) Schematic outline of the Gγ recruitment system for membrane protein targets. When membrane target protein 'X' interacts with candidate protein 'Y 1 ' fused to Gγ cyto , the Gβ and Gγ cyto complex (Gβ γ cyto ) migrates to the inner leaflet of the plasma membrane and restores the signaling function. If membrane protein 'X' cannot interact with protein 'Y 1 ' , Gβ γ cyto is released into the cytosol, and signaling is blocked.
Scientific RepoRts | 5:16723 | DOI: 10.1038/srep16723 resulting in MC-FC and MC-FN yeast strains (Table 1). For the candidate proteins, autonomous replication plasmids for the expression of the four different Z variants (Z WT , Z K35A , Z I31A and Z 955 ) fused to Gγ cyto (Gγ cyto -Y 1 ) (pGK413-Gγ -EZWT, pGK413-Gγ -EZK35A, pGK413-Gγ -EZI31A and pGK413-Gγ -EZ955) ( Table 2) were introduced into the MC-FC and MC-FN yeast cells ( Fig. 3A and Fig. S2A). Flow cytometric analysis of the transformants was conducted after incubation in medium containing the α -alpha-cell mating pheromone (α -factor) ( Fig. S1; left). The engineered yeast strains expressing the Gγ cyto -Z WT and Gγ cyto -Z K35A fusion proteins as candidates slightly induced the transcription of GFP reporter genes via interaction with the membrane-associated Fc fragment, although the fluorescence levels were extremely low ( Fig. 3B and Fig. S2B). In mating selection with intact α -type yeast cells ( Fig. S1; right), the strains expressing Gγ cyto -Z WT and Gγ cyto -Z K35A exhibited specific but negligible cell growth on selective medium ( Fig. 3C and Fig. S2C). In both GFP transcription assays and mating growth selection, interactions of Gγ cyto -Z I31A (very low affinity for Fc) and Gγ cyto -Z 955 (negative control) with the membrane-associated should be expressed as a fusion with Gγ cyto in the cytosol. Protein 'Y 1 ' should be anchored to the plasma membrane, whereas 'Y 2 ' should be expressed in the cytosol. By establishing 'Y 1 ' and 'Y 2 ' as the parental (known) proteins originally bound to target 'X' and the candidate variant proteins, respectively, 'Y 1 ' and 'Y 2 ' compete to bind to target 'X. ' When 'X' has higher affinity for 'Y 2 , ' G-protein signaling is prevented due to the inability of Gγ cyto to migrate to the plasma membrane. When 'X' has higher affinity for 'Y 1 , ' G-protein signaling is transmitted into the yeast cells and invokes the mating process. Thus, affinity-enhanced proteins or affinity-attenuated proteins can be screened in a specific manner. (B) Schematic outline of competitorintroduced Gγ recruitment system for membrane protein targets. Target protein 'X' is a transmembrane or membrane-associated protein. Protein 'Y 1 ' should be fused to Gγ cyto , whereas 'Y 2 ' should be expressed in the cytosol. By establishing 'Y 1 ' and 'Y 2 ' as the parental (known) proteins originally bound to the membrane target 'X' and the candidate variant proteins, respectively, 'Y 1 ' and 'Y 2 ' compete to bind to target 'X. ' When 'X' has a higher affinity for 'Y 2 , ' G-protein signaling is prevented due to the inability of Gγ cyto to migrate to the plasma membrane. When 'X' has higher affinity for 'Y 1 ' fused to Gγ cyto , G-protein signaling is transmitted into the yeast cells and initiates the mating process.   Fc fragment were not detected. These results showed that the previous protocol was not sufficient to screen the interactions between membrane-associated target 'X' and candidate 'Y 1 '-fused Gγ cyto proteins.
PPI-detecting Gγ recruitment system for the selection of candidate proteins interacting with membrane protein targets. Next, we tested the new protocol, in which we changed the method used to introduce the Gγ cyto -Y 1 candidate genes. The DNA cassettes for cytosolic expression of the Gγ cyto -fused candidate Z variants (Z WT , Z K35A , Z I31A and Z 955 ) as a library were stably integrated into the MC- Competitive selection of affinity-enhanced protein variants interacting with membrane protein targets using a previous protocol. Previously, we established the competitor-introduced Gγ recruitment system for selective screening of protein variants that exceed a specified affinity threshold 41 ( Fig. 2A). In the conventional Gγ recruitment system, additional expression of a cytosolic parental (known) protein (Y 2 ) that binds to Gγ cyto -fused target protein 'X' competes with artificially membrane-associated protein variants as a candidate library (Y 1 ), thereby permitting the selective screening of affinity-enhanced protein variants ( Fig. 2A).
To test whether the previous competitor-introduced Gγ recruitment system allows for the use of membrane proteins as target 'X' (Fig. 2B and S3A), we consistently used the membrane-associated Fc fragment and the Gγ cyto -fused Z variants as target 'X' and candidate 'Y 1 ' proteins, respectively (Figs. 4A and S4A). Z I31A (low affinity for Fc; 8.0 × 10 3 M −1 ) was utilized as the model of the competitive parental 'Y 2 ' protein (Figs. 4B and S4B). Therefore, the Z WT and Z K35A candidate proteins (Y 1 ), with higher affinities, should have outcompeted the interaction between membrane-associated Fc (X) and cytosolic Z I31A (Y 2 ), recovering the signaling in the system (Fig. S3A). In the previous system, the DNA cassette for Z I31A expression as a competitor 'Y 2 ' protein in the cytosol was stably integrated into the yeast chromosome of MC-FC, in which the C-terminally membrane-associated Fc fragment (X) (with the Ste18p lipidation motif) was expressed, generating an FC-I strain (Table 1). Autonomous replication plasmids for expression of the Gγ cyto -fused Z variants as candidate 'Y 1 ' (pGK413-Gγ -EZWT, pGK413-Gγ -EZK35A, Scientific RepoRts | 5:16723 | DOI: 10.1038/srep16723 pGK413-Gγ -EZI31A and pGK413-Gγ -EZ955) ( Table 2) were then introduced into the FC-I strain. However, both flow cytometric analysis and mating selection were barely able to detect the interactions between the membrane-associated Fc fragment (target 'X') and the Gγ cyto -fused Z variants (candidate 'Y 1 ') relative to the interactions between the membrane-associated Fc fragment and cytosolic Z I31A in all transformants (Fig. 4B). Additionally, when using an FN-I strain chromosomally expressing an N-terminally membrane-associated Fc fragment (X) (with a Gpa1p lipidation motif) and competitive Z I31A protein (Y 2 ) (Table 1), the transformants in which the candidate autonomous plasmids were introduced to express the Gγ cyto -fused Z variants (Y 1 ) provided similar results to the C-terminally membrane-associated Fc fragment (Fig. S4B). These results showed that the previous system was unable to screen the interactions between membrane-associated target 'X' and candidate 'Y 1 '-fused Gγ cyto proteins relative to the interactions between membrane target 'X' and the cytosolic 'Y 2 ' competitor.
Competitor-introduced Gγ recruitment system that specifically selects affinity-enhanced protein variants interacting with membrane protein targets. Similar to what was described in the previous section, we attempted to change the protocol by introducing the expression cassettes for Gγ cyto -Y 1 candidate genes into the competitor-introduced Gγ recruitment system (Figs. 4C and S4C). As competitive parental 'Y 2 ' proteins, the genes for expressing the four different Z variants (Z WT , Z K35A , Z I31A and Z 955 ) in the cytosol were integrated into the MC-FC yeast chromosome (also expressing the C-terminally membrane-associated Fc fragment with the Ste18p lipidation motif as target 'X'), generating FC-W, FC-K, FC-I and FC-9. The DNA cassettes for expressing the Gγ cyto -fused candidate Z variants as model library Y 1 proteins were then stably integrated into the chromosome of the four yeast strains, generating 16 engineered yeast strains (FC-GWW through FC-G99; Table 1) (Fig. 4C).
Both flow cytometric analysis and mating selection revealed the interactor combinations between membrane-associated Fc and the Gγ cyto -fused Z variants serving as candidate 'Y 1 ' proteins, with higher affinities than when the cytosolic Z variants served as competitor 'Y 2 ' proteins (e.g., Y 1 and Y 2 : Z WT and Z K35A ; Z WT and Z I31A ; and Z K35A and Z I31A ), although the very weak interactions between Fc and Gγ cyto -fused Z I31A (Y 1 and Y 2 : Z I31A and Z 955 ) could not be detected (Fig. 4D,E). These results clearly showed that the strains recovered signal transmission only when interactions between the membrane-associated Fc fragment (target 'X') and the Gγ cyto -fused Z variants (candidate 'Y 1 ') overcame the competitive interactions between Fc (target 'X') and the cytosolic Z variants (competitor 'Y 2 '). Additionally, when using a strain chromosomally expressing the N-terminally membrane-associated Fc fragment (X) (with the Gpa1p lipidation motif) (FN-GWW through FN-G99; Table 1), similar results were obtained (Fig. S4C-E).
Thus, Gγ cyto -fused 'Y 1 ' candidate proteins should be stably integrated into the yeast chromosome to specifically select the affinity-enhanced protein variants against membrane-associated protein 'X' in the competitor-introduced Gγ recruitment system. This modification of the method made the competitor-introduced Gγ recruitment system able to screen affinity-enhanced protein variants by using membrane proteins as the target proteins.
Competitive selection of affinity-attenuated protein variants interacting with membrane protein targets using a previous protocol. Previously, we also established a system that permits the selective screening of affinity-attenuated protein variants. In the conventional Gγ recruitment system, by setting the cytosolic protein (Y 2 ) as the candidate library and the artificially membrane-associated protein (Y 1 ) as the parental (known) competitor that binds to Gγ cyto -fused target protein 'X, ' the system permits the selective screening of affinity-attenuated protein variants ( Fig. 2A).
To test whether the previous competitor-introduced Gγ recruitment system allows for the use of membrane proteins as target 'X' (Fig. 2B and S3B), we consistently used the membrane-associated Fc fragment and the cytosolic Z variants as target 'X' and candidate 'Y 2 ' proteins, respectively ( Fig. 5A and S5A). Z WT was utilized as the model of the competitive parental 'Y 1 ' protein. Therefore, Gγ cyto -fused Z WT (Y 1 ) should have outcompeted the interactions between membrane-associated Fc (X) and the Z K35A , Z I31A and Z 955 candidate proteins (Y 2 ), which have lower affinities, recovering the signaling in the system. In the previous system, autonomous replication plasmids for expression of the Z variants in the cytosol as candidate 'Y 2 ' proteins (pGK-LsZWTc, pGK-LsZK35Ac, pGK-LsZI31Ac and pGK-LsZ955c) ( Table 2) were introduced into the FC-GW strain, which chromosomally expresses Fc-Ste18C as 'X' and Gγ cyto -Z WT as competitor 'Y 1 ' (Table 1). Both flow cytometric analysis and mating selection revealed the interactor combinations between membrane-associated Fc and the cytosolic Z variants serving as candidate 'Y 2 ' proteins, whose affinities were lower than that of Gγ cyto -fused Z WT as the competitor 'Y 1 ' protein ( Fig. 5B,C). Additionally, when using the FN-GW strain chromosomally expressing Gpa1N-Fc as 'X' and Gγ cyto -fused Z WT as competitor 'Y 1 ' (Table 1), the transformants in which the candidate autonomous plasmids were introduced to express the Z variants in the cytosol (Y 2 ) provided similar results (Fig. S5B,C). In contrast to the affinity-enhanced system, these results showed that the previous competitor-introduced Gγ recruitment system was able to screen affinity-attenuated protein variants using membrane proteins as the target proteins.
Demonstration of applicability of our system using intracellular domain of EGFR and Grb2. To demonstrate the applicability of our system, we selected the intracellular domain of EGFR (EGFR cyto ), which contains a tyrosine kinase domain and tyrosine phosphorylation sites, and the adaptor protein Grb2 protein for the PPI pair 49 . In normal cells, binding of the epidermal growth factor (EGF) to the extracellular domain of EGFR leads to dimerization of the receptor and autophosphorylation of the receptor intracellular domain 50,51 . Grb2 binds to the phosphotyrosines of EGFR and links to the activation of subsequent intracellular signaling cascades 52,53 . In yeast, the intracellular domain of EGFR and its mutant derivatives have been often used to test the interaction with Grb2 protein [54][55][56] . To assay the interaction between EGFR and Grb2 in yeast, we used the intracellular domain of EGFR with L834R mutation (EGFR L834R,cyto ; that is constitutively dimerized and activated even in the absence of EGF 49,57 ) as the membrane protein by fusing several types of lipidation motifs at both the N-terminus (Gpa1p motif; Gpa1N) and the C-terminus (Ras1p motif; Ras1C and Ste18p motif; Ste18C). The Grb2 adaptor was fused to Gγ cyto at the N-terminus and the C-terminus to test the accessibility between the membrane-associated EGFR L834R,cyto and the cytosolic Gγ cyto -fused Grb2.
To express the membrane-associated EGFR L834R,cyto protein, the genes encoding the EGFR L834R,cyto attached to the artificial lipidation motifs (Ras1C, Ste18C and Gpa1N) were stably integrated into the ste18 locus of an a-type haploid yeast chromosome, resulting in MC-ErC, MC-EsC and MC-EgN yeast strains (Table 1). For the candidate proteins, the DNA cassettes for cytosolic expression of the Gγ cyto -fused Grb2 at the N-terminus and the C-terminus (Gγ cyto -Grb2 and Grb2-Gγ cyto ) were stably integrated into the MC-ErC, MC-EsC and MC-EgN yeast chromosomes, generating ErC-Ggrb(grbG), EsC-Ggrb(grbG) and EgN-Ggrb(grbG) ( Table 1) (Fig. S6A,D). As a consequence of GFP transcription assays and mating selection, the engineered strains co-expressing the EGFR L834R,cyto with C-terminal lipidation motifs (Ras1C and Ste18C) and the C-terminally Gγ cyto -fused Grb2 (Grb2-Gγ cyto ) specifically showed apparent fluorescence and cell growth on the selective medium (Fig. S6A-F). The accessibility between the phosphotyrosines of membrane-associated EGFR L834R,cyto and the SH2 domains of Grb2 or the distance of Gβ γ cyto complex from the membrane might have influenced the interactions of these proteins or to the subsequent membrane-anchored effector molecule 49,52 . Compared with the MC-ErC strain introducing the Grb2-Gγ cyto -expressing autonomous replicating plasmid (pGK413-Grb2-Gγ ) ( Table 2), the ErC-grbG strain that chromosomally expressed Grb2-Gγ cyto was determinably more suitable for recovering the signaling (Fig. 6A-E).
To further test whether the competitor-introduced Gγ recruitment system that has designed to select the affinity-enhanced protein variants interacting with membrane target proteins is applicable to the intracellular domain of EGFR, we consistently used the membrane-associated EGFR L834R,cyto and the Gγ cyto -fused Grb2 as membrane target 'X' and candidate 'Y 1 ' proteins, respectively (Fig. 6F). Several Grb2 variants (Grb2, Grb2 E89K and Grb2 R86G ) with different affinities for the phosphotyrosines of EGFR were utilized for the competitive parental 'Y 2 ' proteins (K a ; Grb2 > Grb2 E89K > Grb2 R86G ) 58 . Similar to what was described in the previous section, we tested the new protocol by chromosomally integrating the expression cassettes for Y 1 -Gγ cyto candidate genes (Fig. 6F). As competitive parental 'Y 2 ' proteins, the genes for expressing the three different Grb2 variants (Grb2, Grb2 E89K and Grb2 R86G ) in the cytosol were integrated into the ErC-grbG yeast chromosome (also co-expressing the membrane-associated EGFR L834R,cyto with the Ras1p lipidation motif as target 'X' and the Grb2-Gγ cyto fusion protein as candidate 'Y 1 -Gγ cyto '), generating ErC-grbG-grb, ErC-grbG-E89K and ErC-grbG-R86G (Table 1). ErC-grbG-LEU yeast strain never expressing any competitor proteins was also generated as positive control (Table 1).
Both flow cytometric analysis and mating selection displayed the consistent results with the Z variants as expected (Fig. 6G,H). When using the strains respectively expressing Grb2 E89K and Grb2 R86G as the competitive parental 'Y 2 ' proteins (ErC-grbG-E89K and ErC-grbG-R86G), the Gγ cyto -fused Grb2 expressed as candidate 'Y 1 ' (Grb2-Gγ cyto ) predictably recovered the signaling in accordance with the order of difference in the affinity strengths between the competitive proteins and the candidate proteins. Similarly, the strain co-expressing the same Grb2 protein as the candidate 'Y 1 ' and the parental 'Y 2 ' proteins (ErC-grbG-grb) barely showed GFP fluorescence and cell growth on the selective medium. Thus, we demonstrated that our systems were applicable to the membrane protein, which linked to the cellular states involved in various diseases.

Discussion.
In this study, we found that the previously established Gγ recruitment systems 41,42 were basically unable to utilize membrane proteins as target protein 'X. ' The new systems described here successfully enable the use of membrane proteins as target 'X, ' both in the conventional (for screening  of PPI candidate 'Y 1 ' proteins) and competitor-introduced (for screening of affinity-enhanced candidate 'Y 1 ' protein variants) Gγ recruitment systems. In the new systems, only the protocol for expression of Gγ cyto -fused candidate 'Y 1 ' proteins was changed: instead of autonomous replicating plasmids,

Figure 3. Selection of Z variants binding to a membrane-associated target Fc protein using previous and new Gγ recruitment systems. (A) Previous Gγ recruitment system for membrane proteins as targets. (B,C)
Flow cytometric analyses and mating growth assay. The fluorescence and growth intensities of the engineered strains expressing C-terminally membrane-associated Fc via stable integration into the yeast chromosome as well as cytosolic Z variants fused to Gγ cyto 'Y 1 ' via autonomous replication plasmids. The control yeast shows the strain without the expression of 'Y 1 ' fused to Gγ cyto (transformed with pGK413 mock vector). (D) New Gγ recruitment system for membrane proteins as targets. (E,F) Flow cytometric analyses and mating growth assay. The fluorescence and growth intensities of the engineered strains expressing C-terminally membraneassociated Fc and cytosolic Z variants fused to Gγ cyto via stable integration into the yeast chromosome. The control yeast shows the strain without the expression of 'Y 1 ' fused to Gγ cyto (MC-FC in Table 1).

Figure 4. Competitive selection of Z variants with higher affinities for membrane-associated target Fc using previous and new methods for affinity-enhanced systems. (A)
Previous affinity-enhanced system for membrane proteins as targets. (B) Flow cytometric analyses and mating growth assay. The fluorescence and growth intensities of the engineered strains expressing C-terminally membrane-associated Fc and competitor Z I31A as cytosolic 'Y 2 ' via stable integration into the yeast chromosome as well as cytosolic Z variants 'Y 1 ' fused to Gγ cyto via autonomous replication plasmids. Control yeast strains lacked the expression of 'Y 1 ' fused to Gγ cyto (transformed with pGK413 mock vector). (C) New affinity-enhanced system for membrane proteins as target. (D,E) Flow cytometric analyses and mating growth assay. The fluorescence and growth intensities of the engineered strains expressing C-terminally membrane-associated Fc, competitor cytosolic Z variants 'Y 2 ' and cytosolic Z variants 'Y 1 ' fused to Gγ cyto via stable integration into the yeast chromosome. The control yeast shows the strain without the expression of 'Y 1 ' fused to Gγ cyto . chromosomal integration was employed. These new systems are therefore very simple but highly useful. The results of the intracellular domain of EGFR and Grb2 interaction showed that our Gγ recruitment systems could be exploited as a convenient heterologous system to discern the strong binders to the phosphotyrosines in the intracellular domain of EGFR, and therefore would provide the basis for studying other receptor tyrosine kinases as well. In this manner, the screening of binding partners and affinity-enhanced variants targeted to the inner domains of these membrane proteins has great potential for applications in the treatment of human diseases. Previously, we demonstrated that Gγ recruitment systems enabled extremely reliable screening that could completely exclude false-positive candidates 41,42 . Generally, membrane yeast two-hybrid systems 30,31,33,34 and protein fragment complementation assays 23,38 sometimes exhibit background readouts 23,59 due to the use of directly fused artificial transcription factors and automatic self-associations of the split proteins. These background readouts are a critical problem, even when they are negligible, especially in the case of growth screening using a large-scale library 23 . The exclusive selection in Gγ recruitment systems is made possible by using the signal transduction machinery, which requires the localization of Gβ /Gγ in GFP transcription assays and mating selection (Figs. 3-5). This extremely disciplined selection machinery makes Gγ recruitment systems worth using.
In the Gγ recruitment system that has designed for membrane proteins as the target, Z I31A with extremely low affinity could not be detected in both cases of the flow cytometric analysis and the mating selection (Fig. 3). Due to the very low affinity between Z I31A and the Fc region (8.0 × 10 3 M −1 ), the migration of Gγ cyto to the membrane was likely insufficient for the recovering of the signal transduction. This affinity (8.0 × 10 3 M −1 ) seems to be less than a lower limit of our present system, although it is unlikely that a protein mutant exhibiting such extremely low affinity would be required.
From the perspective of screening for a target membrane protein 'X' , the new methods that chromosomally integrate the DNA cassettes expressing Gγ cyto -fused candidate 'Y 1 ' proteins might have a handicap in constructing a library. Specifically, the transformation efficiencies of homologous integrations into the yeast chromosome are commonly 10 1 -10 2 fold lower than those of autonomous replicating plasmids (approximately 10 5 -10 6 cfu/μ g) [60][61][62] . Therefore, constructing a large-scale library might require a little ingenuity to increase the transformation efficiencies, such as via the use of large amounts of DNA, the electroporation method 61,63 , the spheroplasting method 64 , and use of I-SceI meganuclease 65 . Even allowing for this additional effort, however, the conventional Gγ recruitment system is a powerful tool because of its extremely reliable selection of binding partners. In addition, the competitor-introduced Gγ recruitment system, which allows for the specific screening of affinity-enhanced protein variants (specifically excluding protein variants showing equal or lower affinities 41 ), is valuable as a unique and irreplaceable growth selection technique.
A similar approach for screening for affinity-attenuated protein variants among membrane proteins serving as target 'X' made it possible to apply the previous method using autonomous replicating plasmids to express the candidate 'Y 2 ' in the cytosol (Fig. 5). We believe that the unstable expression of 'Y 1 '-fused Gγ cyto using autonomous replicating plasmids rendered the Gγ recruitment system useless. Because it has been reported that plasmid retentions become unstable during signal-promoted states 66 , 'Y 1 '-fused Gγ cyto might be more affected by this unstable plasmid retention than cytosolic 'Y 2 ' is. In any event, the chromosomal expression of 'Y 1 '-fused Gγ cyto is favorable in our Gγ recruitment systems.
In summary, new Gγ recruitment systems make it possible for membrane proteins to be target protein 'X. ' These systems permit reliable and specific screens for binding partners and affinity-enhanced protein variants. We envision that our selection method will provide a powerful, broadly applicable tool for studying biological processes, creating new opportunities to develop new drugs targeting a wide range of membrane-associated proteins and inner domains of transmembrane proteins.

Methods
Strains and media. The genotypes of Saccharomyces cerevisiae BY4741 67 , MC-F1 43 , and BY4742 67 and the other recombinant strains used in this study are provided in Table 1. The yeast strains were grown in YPD medium containing 1% (w/v) yeast extract, 2% peptone and 2% glucose or in SD medium containing 0.67% yeast nitrogen base without amino acids (BD Diagnostic Systems, Sparks, MD, USA) and 2% glucose. The SD medium was supplemented with amino acids and nucleotides (20 mg/L histidine, 60 mg/L leucine, 20 mg/L methionine, or 20 mg/L uracil), as required by the auxotrophic strains. Agar (2%; w/v) was added to the medium to produce YPD and SD solid media.

Construction of plasmids.
All plasmids and primers used in this study are listed in Table 2 and   Table S1. Plasmids inserting lipidation motifs were constructed as follows. The fragments of the PGK1 promoter (P PGK1 ) fused to the lipidation motif of Gpa1p (9 a.a. of N-terminus) and the multi-cloning site were amplified from pGK425 68 using primer 1, primer 2 and primer 3 and inserted into the XhoI-BglII sites of the autonomous replication plasmid pGK425 68 , yielding plasmid pGK425-Gpa1N. The fragments of the PGK1 promoter fused to the lipidation motif of Ste18p (9 a.a. of C-terminus) and the multi-cloning site were amplified from pGK425 68 using primer 1, primer 4 and primer 5 and inserted into the XhoI-BglII sites of the autonomous replication plasmid pGK425 68 , yielding plasmid pGK425-Ste18C. The fragments of the PGK1 promoter fused to the lipidation motif of Ras1p (10 a.a. of C-terminus) and the multi-cloning site were amplified from pGK425 68 using primer 1, primer 6 and primer 7 and inserted into the XhoI-BglII sites of the autonomous replication plasmid pGK425 68 , yielding plasmid pGK425-Ras1C.
The plasmids used for the expression of the Fc fragment on the membrane were constructed as follows. The fragments encoding the Fc protein were amplified from pUMGP-Gγ MFcH 42 using primers 8 and 9 or primers 10 and 11 and inserted into the SalI-BamHI sites of the autonomous replication plasmid pGK425-Gpa1N or pGK425-Ste18C, yielding pGK425-Gpa1N-Fc and pGK425-Fc-Ste18C, respectively. The cassettes for expression of the membrane-associated Fc protein for integration at the ste18 locus on the yeast chromosome were then amplified from pGK425-Gpa1N-Fc or pGK425-Fc-Ste18C using primer 12 and primer 13 and inserted into the XhoI sites of pGK426-GPTK 42 using an In-Fusion HD Cloning Kit (Clontech Laboratories -Takara Bio, Shiga, Japan), yielding pUMGPTK-Gpa1N-Fc and pUMGPTK-Fc-Ste18C, respectively.
The cassettes for expression of the cytosolic Z variants as competitors for integration upstream of the HOP2 gene locus (P HOP2 : HOP2 promoter region) on the yeast chromosome were constructed as follows. The fragments encoding P HOP2 were amplified using primer 24 and primer 25 and inserted into the NotI-SacI sites of pGK-LsZWTc, pGK-LsZK35Ac, pGK-LsZI31Ac and pGK-LsZ955c 41 , yielding plasmids pGK-LsZWTc-HOP, pGK-LsZK35Ac-HOP, pGK-LsZI31Ac-HOP and pGK-LsZ955c-HOP, respectively.
The plasmids used for the expression of the Grb2-Gγ cyto in the cytosol were constructed as follows. The fragment encoding the Grb2-Gγ cyto was amplified from B1U-GL 49 using primer 30 and primer 31 and inserted into the SalI-EcoRI sites of the autonomous replication plasmid pGK413 68 using an In-Fusion HD Cloning Kit, yielding plasmid pGK413-Grb2-Gγ . Subsequently, the cassettes for expression of the Grb2-Gγ cyto for integration at the his3 locus on the yeast chromosome were constructed as follows. The fragment containing the STE18 promoter (P STE18 ) was amplified from pUMGP-Gγ MFcH 42 using primer 32 and primer 33 and inserted into the XhoI-NheI sites of pGK416 68 , yielding plasmid Ste18p-416. The fragment containing the gene encoding Gγ cyto were amplified from pUMGP-Gγ MFcH 42 using primer 34 and primer 35 and inserted into the XbaI-EcoRI sites of Ste18p-416, yielding plasmid pUSTE18p-c-Gγ cyto. The fragment encoding HIS3 terminator (T HIS3 ) was amplified from the BY4741 genome using primer 20 and primer 21 and inserted into the NotI-SacI sites of pUSTE18p-c-Gγ cyto, yielding plasmid pUSTE18p-c-Gγ cyto-HIS3t. Finally, the fragment encoding Grb2 was amplified from pGK413-Grb2-Gγ using primer 36 and primer 37 and inserted into the NheI-XmaI sites of pUS-TE18p-c-Gγ cyto-HIS3t, yielding plasmid pUSTE18p-Grb2-Gγ cyto-HIS3t.
The plasmids used for the expression cassettes of the Gγ cyto -Grb2 for integration at the his3 locus on the yeast chromosome were constructed as follows. The fragment encoding Grb2 was amplified from pGK413-Grb2-Gγ using primer 38 and primer 39 and inserted into the NheI-XmaI sites of pUS-TE18p-Gγ cyto-HIS3t, yielding plasmid pUSTE18p-Gγ cyto-Grb2-HIS3t.
The cassettes for expression of the cytosolic Grb2 variants as competitors for integration at the upstream of the HOP2 gene locus (P HOP2 : HOP2 promoter region) on the yeast chromosome were constructed as follows. The fragments encoding P HOP2 were amplified using primer 24 and primer 25 and inserted into the NotI-SacI sites of pGK415 68 , yielding plasmid pGK415-HOP2p. The fragment encoding Grb2 was amplified from pGK413-Grb2-Gγ using primers 38 and 39 and inserted into the SalI-XmaI sites of pGK415-HOP2p using an In-Fusion HD Cloning Kit, yielding plasmid pGK415-Grb2-HOP2p. The fragment encoding Grb2 R86G mutant was amplified from pGK413-Grb2-Gγ using primers 40 and 42, primers 41 and 43 and the fragments encoding the Grb2 R86G mutant was amplified from these two fragments by overlap PCR using primer 40 and primer 41, and inserted into the SalI-XmaI sites of pGK415-HOP2p using an In-Fusion HD Cloning Kit, yielding plasmid pGK415-Grb2 R86G -HOP2p. The fragment encoding Grb2 E89K mutant was amplified from pGK413-Grb2-Gγ using primers 40 and 44, primers 41 and 45 and the fragments encoding the Grb2 E89K mutant was amplified from these two fragments by overlap PCR using primer 40 and primer 41, and inserted into the SalI-XmaI sites of pGK415-HOP2p using an In-Fusion HD Cloning Kit, yielding plasmid pGK415-Grb2 E89K -HOP2p.
Construction of yeast strains. All strains used in this study are listed in Table 1. Integration of the DNA cassettes for expressing the membrane-associated Fc protein was achieved as follows. The DNA fragments containing P STE18 -P PGK1 -Fc-Ste18C-T PGK1 -kanMX4-T STE18 and P STE18 -P PGK1 -Gpa1N-Fc-T PGK1 -kanMX4-T STE18 were amplified from pUMGPTK-Fc-Ste18C and pUMGPTK-Gpa1N-Fc using primer 46 and primer 47. The amplified DNA fragments were then used to transform MC-F1 43 using the lithium acetate method 69 . The transformants were selected on a YPD + G418 plate to yield MC-FC and MC-FN ( Table 1).
Integration of the DNA cassettes for the Gγ cyto -Z domain variants (Z WT , Z K35A , Z I31A and Z 955 ) in the cytosol was achieved as follows. The DNA fragments containing URA3-P PGK1 -Gγ cyto -Z WT (-Z K35A , -Z I31A and -Z 955 )-T PGK1 -T HIS3 were amplified from pUSTE18p-Gγ cyto-ZWT(-ZK35A, -ZI31A and -Z955)-HIS3t using primer 48 (containing the homologous regions of the HIS3 promoter) and primer 49. The amplified DNA fragments were used to transform MC-FC and MC-FN using the lithium acetate method 69 . The transformants were then selected on an SD-Ura plate (containing leucine, histidine and methionine) to yield FC-GW, FC-GK, FC-GI, and FC-G9 and FN-GW, FN-GK, FN-GI and FN-G9 (Table 1).
Integration of the DNA cassettes for the Grb2-Gγ cyto in the cytosol was achieved as follows. The DNA fragments containing URA3-P PGK1 -Grb2-Gγ cyto -T PGK1 -T HIS3 was amplified from pUS-TE18p-Grb2-Gγ cyto-HIS3t using primer 48 (containing the homologous regions of the HIS3 promoter) and primer 49. The amplified DNA fragments were used to transform MC-ErC, MC-EsC and MC-EgN using the lithium acetate method 69 . The transformants were the selected on an SD-Ura plate to yield ErC-grbG, EsC-grbG and EgC-grbG ( Table 1). Integration of the DNA cassettes for the Gγ cyto -Grb2 in the cytosol was achieved as follows. The DNA fragments containing URA3-P PGK1 -Gγ cyto -Grb2-T PGK1 -T HIS3 was amplified from pUSTE18p-Gγ cyto-Grb2-HIS3t using primer 48 (containing the homologous regions of the HIS3 promoter) and primer 49. The amplified DNA fragments were used to transform MC-ErC, MC-EsC and MC-EgN using the lithium acetate method 69 . The transformants were then selected on an SD-Ura plate to yield ErC-Ggrb, EsC-Ggrb and EgC-Ggrb ( Table 1).
All transformants were obtained by introducing the autonomous replicating plasmids ( Table 2) into these yeast strains using the lithium acetate method 69 .
GFP reporter expression analysis. GFP reporter expression analysis basically followed previous methods 41 , with certain modifications. In the case of the previous method, the engineered yeast a-cells were grown in 5 mL of SD-His medium (for the PPI detection system), SD-His/-Leu medium (for the affinity-enhanced system) or SD-Leu/-Ura medium (for the affinity-attenuated system) at 30 °C overnight. The cultured cells were then inoculated in 2 mL of fresh SD-His, SD-His/-Leu or SD-Leu/-Ura medium containing 5 μ M α -factor (Zymo Research, Orange, CA, USA) to obtain an initial OD 600 of 0.1 (OD 600 = 0.1). In the case of the new method, the engineered yeast a-cells were grown in 5 mL of YPD medium (for the PPI detection system and affinity-enhanced system) at 30 °C overnight. The cultured cells were then inoculated in 2 mL of fresh YPD medium containing 5 μ M α -factor (Zymo Research, Orange, CA, USA) to obtain an initial OD 600 of 0.1 (OD 600 = 0.1). The expression of the FIG1-EGFP fusion reporter gene was then stimulated by growth at 30 °C for 6 hours.
The fluorescence intensities of the cultured cells were measured using a BD FACSCanto II flow cytometer equipped with a 488-nm blue laser (BD Biosciences, San Jose, CA, USA) 70 . The GFP fluorescence signal was specifically collected through a 530/30-nm band-pass filter. The mean fluorescence intensity was defined as the GFP-A mean of 10,000 cells. The data were analyzed using BD FACSDiva software (version 5.0, BD Biosciences).
Mating growth spotting assay. The mating growth spotting assay basically followed a previous method 41 , with certain modifications. For the previous method, each engineered yeast a-cell was grown in 5 mL of SD-His media (for PPI detection system), SD-His/-Leu medium (for the affinity-enhanced system) or SD-Leu/-Ura medium (for the affinity-attenuated system) at 30 °C overnight and then cultivated in 5 mL of YPD medium with the mating partner, or the BY4742 α -cell 67 , at 30 °C for 3 hours. The initial OD 600 of each haploid cell was set at 0.1 (OD 600 = 0.1). For the new method, each engineered yeast a-cell was grown in 5 mL of YPD medium (for the PPI detection system and the affinity-enhanced system) at 30 °C overnight and then cultivated in 5 mL of YPD medium with the mating partner, or the BY4742 α -cell 67 , at 30 °C for 3 hours. The initial OD 600 of each haploid cell was again set at 0.1 (OD 600 = 0.1). After cultivation, the yeast cells were harvested, washed, and resuspended in distilled water. To quantify the mating ability of each strain, a dilution series of each yeast cell suspension was prepared (OD 600 = 1.0, 0.1, 0.01, 0.001 and 0.0001), and 40 μ L of each dilution was then spotted on a selective SD-Ura/Leu plate (lacking methionine, lysine and histidine; for the PPI detection system generated by the previous method), SD-Ura plate (lacking methionine, lysine, histidine and leucine; for the Scientific RepoRts | 5:16723 | DOI: 10.1038/srep16723 affinity-enhanced system generated by the previous method), SD-His plate (lacking methionine, lysine, leucine and uracil; for the affinity-attenuated system generated by the previous method), SD-His/Leu plate (lacking methionine, lysine and uracil; for the PPI detection system generated by the new method) or SD-His plate (lacking methionine, lysine, uracil and leucine; for the affinity-enhanced system generated by the new method).