Yeast One-Hybrid Gγ Recruitment System for Identification of Protein Lipidation Motifs

Fatty acids and isoprenoids can be covalently attached to a variety of proteins. These lipid modifications regulate protein structure, localization and function. Here, we describe a yeast one-hybrid approach based on the Gγ recruitment system that is useful for identifying sequence motifs those influence lipid modification to recruit proteins to the plasma membrane. Our approach facilitates the isolation of yeast cells expressing lipid-modified proteins via a simple and easy growth selection assay utilizing G-protein signaling that induces diploid formation. In the current study, we selected the N-terminal sequence of Gα subunits as a model case to investigate dual lipid modification, i.e., myristoylation and palmitoylation, a modification that is widely conserved from yeast to higher eukaryotes. Our results suggest that both lipid modifications are required for restoration of G-protein signaling. Although we could not differentiate between myristoylation and palmitoylation, N-terminal position 7 and 8 play some critical role. Moreover, we tested the preference for specific amino-acid residues at position 7 and 8 using library-based screening. This new approach will be useful to explore protein-lipid associations and to determine the corresponding sequence motifs.


Introduction
A wide range of proteins are modified by covalent attachment of fatty acids and/or isoprenoid groups, modifications that play major roles in regulating protein structure and function [1]. Moreover, aberrant expression of lipidated proteins or their biosynthetic enzymes is associated with many diseases, ranging from cancer to neurological disorders [2]. Because recruitment of lipidated proteins to the plasma membrane influences complex signaling pathways that regulate specific physiological functions, there is a great interest in gaining a better understanding of protein trafficking via lipid modification.
Dual lipidation of proteins, which consist of myristoylation and palmitoylation at N-terminal residues, a posttranslational modification that has been conserved from yeast to humans, leads to transport of the modified proteins toward plasma membrane. For some lipid modification, the consensus motifs have been identified. For example, after removal of the N-terminal methionine residue by methionine aminopeptidase, myristate is attached to the Nterminal glycine of protein substrates with the consensus motif Met 1 -Gly 2 -Xaa 3 -Xaa 4 -Xaa 5 -Ser/Thr 6 -Xaa 7 -Xaa 8 (where Xaa indicates any amino acid residue) [3] in a reaction catalyzed by myristoyl-CoA:protein N-myristoyltransferase (NMT) [4]. Further refinement of this consensus might be possible, however, as it has been shown that not all amino acid residues are allowable at the positions indicated by Xaa [4][5]. Cysteine residues within proteins can be acylated with the 16-carbon fatty acid palmitate. Although protein acyltransferases (PATs) have a common DHHC Cys-rich domain [6], the consensus motif for palmitoylation of proteins is still unclear, due at least in part to the diversity of substrates that can be recognized by multiple PATs, including 7 DHHC proteins in yeast and 23 in humans [7].
We previously developed the Gc recruitment system (GRS), which uses yeast G-protein signaling (pheromone signaling) to detect protein-protein interactions [8][9][10]. This system is based on the observation that transduction of the signal requires localization of the c subunit of G-proteins (Gc) to the inner leaflet of the plasma membrane [11]. Deletion of lipidation sites in yeast Gc (Gc cyto ) completely disrupts G-protein signaling [8]; however, protein-protein interactions between a Gc cyto -fused target and a membrane-bound binding partner can restore of G-protein signaling.
Here we suggest a new approach to the investigation of proteinlipid associations, yeast one-hybrid GRS (Fig. 1). If a Gc cyto -fused hybrid protein is localized to the plasma membrane, G-protein signaling is recovered, inducing a mating response. Hybrid proteins positive in the assay can then be detected by diploid growth selection [9][10]. Thus, using this system, it is possible to ask if a domain fused to Gc cyto associates with lipid molecules that localize the protein to the plasma membrane.
In the current study, we attached short signal sequences, from 6 to 10 amino acid residues (AA), derived from G-protein a (Ga) subunits to N-terminus of Gc cyto to append dual lipid modifications, i.e., myristoylation and palmitoylation, resulting in trafficking to the plasma membrane. Ga subunits play a key role in signal transduction that is mediated by lipid modification. It has been reported that lipidation-defective Gpa1 mutants (G2A and C3A) have lost their intrinsic ability to localize to membranes [12][13]. Among human Ga proteins, most subunits in the G i a subfamily receive both myristoylation and palmitoylation at N-terminal residues, similar to yeast Gpa1. Indeed, human G i2 a contains the same sequence at N-terminal 6 AA (Met 1 -Gly 2 -Cys 3 -Thr 4 -Val 5 -Ser 6 ) as yeast Gpa1 [14][15]. We have evaluated the membranetargeting ability of several N-terminal short sequences thought to receive myristoylation and palmitoylation using yeast one-hybrid GRS.

Strains and Media
Detailed information about Saccharomyces cerevisiae strains BY4741 [16] and MCF4741 [17], as well as about other strains used in this study, including genotypes, are shown in Table 1. Yeast cells were grown in YPD medium containing 1% yeast extract, 2% peptone and 2% glucose, or in SD medium without histidine (SD -His) containing 0.67% yeast nitrogen base without amino acids (Becton Dickinson and Company, Franklin Lakes, NJ, USA), 2% glucose, 20 mg/l uracil, 30 mg/l leucine and 20 mg/l methionine. In either case, 2% agar was added to make solid media (plates).

Construction of Plasmids
The sequences of oligonucleotides used in this study are shown in Table 2. The plasmids shown in Table 1 were made as follows. Using BY4741 genomic DNA as a template, the promoter of the PGK1 gene was amplified with oligonucleotide pair o1 and o2, and inserted in place of the promoter of the STE2 gene at the SacII-NotI sites of pHY-2GA [17], yielding the plasmid pHY-PGA.

Construction of Yeast Strains and Yeast Transformation
The plasmid pUYG-Cre [17], which expresses Cre recombinase under the control of the promoter of the GAL1 gene, was introduced into MCF4741 using the lithium acetate method [18]. To exclude the kanMX4 gene between the two loxP sites, a transformant harboring pUYG-Cre was cultivated in SGal media (containing 6.7% yeast nitrogen base without amino acids; Becton Dickinson and Company, Franklin Lakes, NJ, USA, and 2% galactose) without uracil but containing 20 mg/l histidine, 30 mg/ l leucine and 20 mg/l methionine. Then the cells were maintained on YPD media containing 5-fluoroorotic acid (5-FOA) to select ura3 2 strains that have lost the plasmid pUYG-Cre, yielding MCF-B1.
Deletion of STE18 gene by substitution with LEU2 was carried out as follows. The upstream region of STE18 was termed STE5', and the downstream region STE3'. The LEU2 gene was amplified with oligonucleotide pair o19 and o20, using pLY-3GC [17] as a template. Using genomic DNA from strain BY4741 as the template, STE5' was amplified with oligonucleotide pair o21 and o22, and STE3' was amplified with oligonucleotide pair o23 and o24. DNA fragments containing STE5'-LEU2-STE3' were combined using oligonucleotide pair o21 and o24 from the amplified fragments STE5', LEU2 and STE3', and then used to transform MC-F1 using the lithium acetate method. Transformants were selected on SD plates without leucine (SD -Leu) but containing 20 mg/l histidine, 20 mg/l uracil and 20 mg/l methionine. The resulting strain was named MCF-B1L. Each plasmid shown in Table 1 was introduced into MCF-B1L using the lithium acetate method and the transformants were used for the following analyses.

Diploid Growth Assay
Evaluation of mating ability was performed as follows. Each engineered yeast strain was cultivated in 1 ml of YPD medium with the mating partner BY4742 at 30uC for 1.5 h setting the Wild-type Gc is lipid-modified at its C-terminus, and localized at plasma membrane to transmit the intracellular signal. An engineered Gc lacking membrane association (Gc cyto ) is fused to the target protein domain or peptide motif, yielding a Gc cyto hybrid protein. When the target protein domain or peptide motif does not confer membrane association, Gprotein signaling is not restored. In contrast, when a Gc cyto hybrid protein confers plasma membrane localization, G-protein signaling is restored, leading to induction of the mating response and generation of diploid cells. doi:10.1371/journal.pone.0070100.g001 initial OD 600 of each haploid cell at 0.1. After cultivation, yeast cells were harvested, washed, and resuspended in distilled water. Setting the OD 600 at 1, 0.1 and 0.01, 10 ml of each cell suspension were spotted on an SD plate without methionine and lysine but containing 20 mg/l, histidine, 30 mg/l leucine, and 20 mg/l uracil (SD -Met, Lys plate) for growth selection of diploid cells. After incubation at 30uC for 2 days, we recorded image data of diploid colonies generated on solid medium. For quantitative evaluation, cell suspensions were spread on SD -Met, Lys plates using the appropriate dilution factor for each strain. After incubation at 30uC for 2 days, the colony number was determined for each strain and multiplied by its dilution factor to estimate the number of diploid cells generated in an equivalent volume of 1 ml of cell suspension, with the OD 600 set at 1.0.

Microscopic Observation
The plasmid pHY-Gcc-G, pHY-6Gcc-G, pHY-6LGcc-G, pHY-10YGcc-G, pHY-10HiiGcc-G, pHY-10HiiiGcc-G, pHY-8YLGcc-G, pHY-6L2YGcc-G, pHY-G2A-Gcc-G, pHY-C3A-Gcc-G or pHY-GCA-Gcc-G was introduced into diploid strain BY4743 (Table 1) to indicate localization of each Gc cyto hybrid protein in yeast cells. The cells were incubated in SD -His medium at 30uC for 18 h, harvested and washed with distilled water. The localization patterns of GFP-fused Gc cyto hybrids were observed using a confocal microscopy (Zeiss LSM Pascal 5), using a 636 oil PlanApo objective lens (Zeiss). Images were collected with the same laser power and detector settings using Pascal software version 3.2. To observe green fluorescence, 488 nm laser was used.

Preparation of Library and Diploid Growth Screening
DNA fragments encoding Met 1 -Gly 2 -Cys 3 -Thr 4 -Val 5 -Ser 6 -Xaa 7 -Xaa 8 -Thr 9 -Ile 10 -Gc cyto were amplified from plasmid pHY-M10YGc cyto with the oligonucleotide pair o25 containing tandem NNK codons and o5, and inserted in place of EGFP at the NotI-BamHI sites of pHY-PGA, yielding a plasmid library. The library was introduced into the MCF-B1L strain and transformants were grown in SD -His medium or plated at 30uC for 2 days to yield an initial yeast library expressing Met 1 -Gly 2 -Cys 3 -Thr 4 -Val 5 -Ser 6 -Xaa 7 -Xaa 8 -Thr 9 -Ile 10 -Gc cyto .
Then the initial yeast library grown in SD -His medium was cultivated in 1 ml of YPD medium with the mating partner BY4742 at 30uC for 1.5 h setting the initial OD 600 of each haploid cell at 0.1. After cultivation, yeast cells were harvested, washed, and resuspended in distilled water. Cell suspensions were spread on SD -Met, Lys, His plates using the appropriate dilution factor for each strain. After incubation at 30uC for 2 days, plasmids were extracted from 100 randomly picked diploid colonies and sequences corresponding to Xaa 7 -Xaa 8 were analyzed using the 3500 Genetic Analyzer (Applied Biosystems Inc, CA, USA), together with plasmids extracted from 100 randomly picked colonies from the initial library.

Membrane-targeting Ability of the N-terminal Motif Derived from Yeast Ga
Our strategy to investigate protein-lipid association is outlined in Figure 1. The ste18D strain MCF-B1L (lacking endogenous Gc) was used as a host strain for yeast one-hybrid GRS. To confirm STE18 gene deletion and restoration of signal transduction by recruitment of Gc to the plasma membrane, a diploid growth assay was carried out using MCF-B1L with the plasmid pHY-Gcc, which encodes Gc cyto (a mutant lacking C-terminal lipidation site), or with the plasmid pHY-Gc, which encodes wild-type Gc (Fig. 2). Introduction of Gc cyto did not result in diploid colonies, consistent with a failure in signaling in the absence of membrane localization. In contrast, introduction of wild-type Gc, which is expected to localize to the plasma membrane due to the C-terminal Table 2. Sequence of oligonucleotides used to construct plasmids and yeast strains. lipidations, restored signaling, as evidenced by the production of diploid colonies. We next attached the N-terminal 10 AA from the N-terminus of Gpa1 (yeast Ga subunit) to Gc cyto (MP10Y-Gc cyto ) in order to create a potential lipidation site receiving myristoylation and palmitoylation. As shown in Figure 2, MP10Y-Gc cyto successfully restored signaling, resulting in diploid cells, although the number of diploid cells generated from MP10Y-Gc cyto was lower than from wild-type Gc. These results suggest that it is possible to evaluate the membrane-targeting ability of additional motifs fused to Gc cyto as described in Figure 1. We additionally investigated the contribution of myristoylation and palmitoylation to the membrane-targeting ability of MP10Y-Gc cyto . Neither lipidationdefective mutants G2A-(lacking myristoylation-acceptor site), C3A-(lacking palmitoylation-acceptor site) and G2A-C3A-Gc cyto (lacking both sites) restored the signal transduction at all, suggesting that both lipid modifications are required.

Contribution of the N-terminal Sequence from Position 7 to 10 to Membrane Targeting
Yeast Gpa1 and human G i2 a, receiving both myristoylation and palmitoylation, has the common N-terminal 6 AA (MP6 sequence; Met 1 -Gly 2 -Cys 3 -Thr 4 -Val 5 -Ser 6 ). To confirm its function for Nterminal dual lipidations, MP6 sequence was attached to the Nterminus of Gc cyto (MP6-Gc cyto ). As shown in Figure 3A, yeast cells expressing MP6-Gc cyto generated diploid cells, although the number of diploid cells recovered was lower than what was observed for MP10-Gc cyto (Fig. 2). Next, an irrelevant linker sequence, Gly-Gly-Gly-Ser, was inserted before Gc cyto (MP6L-Gc cyto ; Fig. 3A). In the diploid growth assay, MP6L-Gc cyto did not recover signal transduction. These results suggest that the Nterminal 6 AA alone are not sufficient to complete dual lipidation.
To find a sequence that influences N-terminal dual lipid modification, 10 AA from the N-terminus of human G i2 a or G i3 a (G i1 a contains the same 10 AA at the N-terminus as G i3 a) were attached to Gc cyto , yielding MP10Hii-Gc cyto or MP10Hiii-Gc cyto . In diploid growth assay, neither Gc cyto hybrid resulted in diploid cell formation (Fig. 3A). These results are in agreement with past reports that the N-terminal sequence of human G i2 a and G i3 a are not substrates of yeast NMT [4][5].
To further investigate the contribution of the sequence from position 7 to position 10 to N-terminal dual lipid modifications, two kinds of Gc cyto hybrids were prepared (Fig. 3B). MP6L2Y-Gc cyto contains an irrelevant linker sequence (Gly-Gly) at position 7 and 8 of the N-terminal 10 AA sequence derived from Gpa1 and MP8YL-Gc cyto contains an irrelevant linker sequence (Gly-Ser) at position 9 and 10. In the diploid growth assay, MP8YL-Gc cyto restored diploid cell formation but MP6L2Y-Gc cyto resulted in only a few diploid colonies (Fig. 3B). This suggests that in addition to the MP6 sequence, the N-terminal positions 7 and 8 also contribute largely to dual lipid modifications, and that positions 9 and 10 have subtle but definite effects on these modifications.

Quantitative Assessment of the Membrane-targeting Ability of Gc cyto Hybrids
To compare expression levels among Gc cyto hybrids shown in Figure 4A, a GFP reporter was fused to the C-terminus of Gc cyto or wild-type Gc. Figure 4B shows the fluorescence intensity of each yeast cell suspension. All Gc cyto hybrids were successfully synthesized in yeast. Moreover, based on GFP fluorescence, we found that the levels of expression of the hybrid proteins were higher than that observed for wild-type Gc. We next used the number of colonies generated in the diploid growth assay as a quantitative measure of the membrane-targeting ability of each Gc cyto hybrid (Fig. 4C). With the exception of wild-type Gc, MP10Y-Gc cyto produced the most diploid colonies, suggesting that the N-terminal 10 AA derived from Gpa1 lead to efficient attachment of both myristate and palmitate.
To confirm a direct relationship between membrane-localization of Gc cyto hybrids and recovery of signal transduction, we observed yeast cells expressing Gc cyto hybrid-GFP using a confocal microscopy (Fig. 4D). While G2A-Gc cyto , G2A-C3A-Gc cyto , MP6L-Gc cyto , MP10Hii-Gc cyto and MP10Hiii-Gc cyto were localized in cytoplasm as well as Gc cyto that completely lacks membrane-targeting ability, other Gc cyto hybrids were detected at plasma membrane with different degrees of intensity. The intensity of green fluorescence from the juxtamembrane region, except for C3A-Gc cyto , relatively reflected the efficiency of MP6-Gc cyto , MP10Y-Gc cyto , MP8YL-Gc cyto and MP6L2Y-Gc cyto at which signal transduction was restored. Exploration of Amino Acid Residue Preference via Screening of a Library Randomized at N-terminal Positions 7 and 8 A randomized DNA library encoding Met 1 -Gly 2 -Cys 3 -Thr 4 -Val 5 -Ser 6 -Xaa 7 -Xaa 8 -Thr 9 -Ile 10 -Gc cyto was prepared by PCR amplification using primers with NNK codons (N represents an equal mixture of A, G, C, and T; K represents an equal mixture of G and T) covering all 20 amino acids, followed by introduction into the MCF-B1L strain (initial yeast library). After dilution and isolation of initial yeast library colonies on solid media, the population of amino acid residues present at positions 7 and 8 was determined by DNA sequencing of plasmids from 100 randomly picked colonies ('initial' in Fig. 5).
We next randomly picked 100 colonies grown in the diploid growth assay and determined the frequency of each amino acid residue at positions 7 and 8 ('final' in Fig. 5). The ratio of enrichment for each amino acid residue was defined as the value for final population dived by that for initial population (Table 3). Uncharged hydrophilic residues, i.e., Ser, Thr, Asn and Gln, were preferred at positions 7 and 8. No preference was observed for charged residues, i.e., Asp, Glu, Arg, Lys and His, with the exception of a preference for His at position 8. We never observed Asp at position 7, or Arg or Lys at position 8 (ratio of enrichment of 0 in all three cases), suggesting that those residues are incompatible with lipid modification. Cys is known to be a potent acceptor for palmitoylation. In this assay, Cys was enriched at position 7 but not at position 8. Aromatic residues Phe and Tyr, but not Trp (the largest amino acid), were preferred at positions 7 and 8. Among hydrophobic residues, Ala, Val, Ile and Met were preferred at position 7, and Val, Leu and Met were enriched at

Verification of Amino Acid Residue Preferences Conferring Gc cyto Membrane-targeting Ability
To verify the results of the preference test, we next looked at the membrane-targeting ability of the set of Gc cyto -hybrids listed in Table 4 using the diploid growth assay (Fig. 6). DW-, GE-and WA-Gc cyto hybrids were used as representatives expected to confer low efficiency of lipid modification (Asp, Gly and Trp were enriched the least at position 7). FN-, QT-and TM-Gc cyto hybrids were used as representatives expected to confer efficient lipid modification (Phe, Gln and Thr were the most enriched residues at position 7). As shown in Figure 6, DW-, GE-and WA-Gc cyto rarely generated diploid cells. In contrast, FN-, QT-and TM-Gc cyto restored signal transduction with a higher efficiency than MP10Y-Gc cyto , which contains the wild-type 10 AA from Gpa1. These results suggest that the amino acid residue biases identified using the yeast one-hybrid GRS (see Figure 5 and Table 3) are reliable predictions of protein activity in cells.

Discussion
The aim of this study was to establish a genetic system, yeast one-hybrid GRS, for analyzing membrane-targeting ability of proteins in a living cell (Fig. 1). Our system is based on the observation that G-protein signaling requires localization of Gc to the inner leaflet of the plasma membrane [11]. In the current study, we focused the sequence involved in N-terminal dual lipid modifications, i.e., myristoylation and palmitoylation. Protein lipidation is a critical modification that in most cases results in protein trafficking to the membrane. Because the attachment of a single lipid is often not sufficient for membrane targeting, an additional lipidation is typically required [19]. Modification via both myristoylation and palmitoylation at the N-terminus is an example of dual lipidation that can enhance the membranetargeting ability of a protein.
We succeeded in construction of yeast one-hybrid GRS to investigate lipid-protein associations. Use of diploid growth caused by restoration of G-protein signaling has a potency to explore the sequences responsible for protein lipidations with high-throughput. In our system, both myristoylation and palmitoylation of Gc cyto hybrids are required to restore the signal transduction (See Fig. 2). As shown in Figure Table 3. Frequency of each amino acid residue at positions 7 and 8 before and after diploid growth screening. myristoylation partially localized at the plasma membrane, it did not restore the signal transduction at all (Fig. 2). It has been reported that G-proteins, G-protein-coupled receptors and other regulators are accumulated in lipid rafts in order to trigger intracellular signal transduction [20][21][22]. Because the combination of myristoylation and palmitoylation functions as a defined lipid raft-targeting signal in protein trafficking [23], C3A-Gc cyto lacking palmitoylation might have failed to localize in lipid rafts. Therefore, the number of generated diploid cells seems to reflect the amount of dually-lipidated Gc cyto hybrids in a cell. N-terminal residues position 7 and 8 play some critical role, but it is unclear if these residues are required for either the myristoylation, palmitoylation, or both as shown in Figure 3. Because it has been reported that Ala 7 -Glu 8 within MP10Hii-Gc cyto and MP10Hiii-Gc cyto were not suitable as substrates of yeast NMT [4][5], these positions are probably involved in myristoylation directly. Also these positions are at least indirectly involved in palmitoylation because myristoylation is a prerequisite for palmitoylation [11].
Using yeast one-hybrid GRS, we investigated sequences at position 7 and 8 required for dual lipid modifications which can give membrane-targeting ability to proteins ( Fig. 5 and 6). We classified the unsuitable or unacceptable residues for dual lipid modifications into three broad categories. Charged residues are common in N-terminal positions 8 to 10 of human Ga subunits in the G i a subfamily [14]. As described above, differences in substrate specificity of NMT between yeast and humans suggest that Glu at position 8 is unacceptable, at least for myristoylation. The second category comprised residues with the highest (Gly) or lowest (Pro) degree of rotational freedom [24], which can have extreme effects on main-chain structure. Finally, Trp, the largest amino acid residue, can be classified as the sole member of a last category of amino acids with a high steric barrier.
In conclusion, we have established a new approach, yeast onehybrid GRS, to facilitate investigation of protein-lipid associations. Using N-terminal sequences derived from Ga subunits as a model case, we showed that it is possible to test membrane-targeting ability of N-terminal motifs for their ability to receive dual lipid modification, i.e., myristoylation and palmitoylation. Our method approved to be reliable and versatile, and as such, use of the method should help promote further advancement in function analyses of proteins within a living cell. Table 4. N-terminal sequences of Gc cyto hybrids used to verify amino acid residue preferences.