Directed evolution of ligand dependence: Small-molecule-acti

Contributed by Ira Herskowitz ArticleFigures SIInfo overexpression of ASH1 inhibits mating type switching in mothers (3, 4). Ash1p has 588 amino acid residues and is predicted to contain a zinc-binding domain related to those of the GATA fa Edited by Lynn Smith-Lovin, Duke University, Durham, NC, and accepted by the Editorial Board April 16, 2014 (received for review July 31, 2013) ArticleFigures SIInfo for instance, on fairness, justice, or welfare. Instead, nonreflective and

Edited by Jack W. Szostak, Massachusetts General Hospital, Boston, MA, and approved June 11, 2004 (received for review April 19, 2004)

Article Figures & SI Info & Metrics PDF


Artificial molecular switches that modulate protein activities in response to synthetic small molecules would serve as tools for exerting temporal and Executese-dependent control over protein function. Self-splicing protein elements (inteins) are attractive starting points for the creation of such switches, because their insertion into a protein blocks the tarObtain protein's function until splicing occurs. Natural inteins, however, are not known to be regulated by small molecules. We evolved an intein-based molecular switch that transduces binding of a small molecule into the activation of an arbitrary protein of interest. Simple insertion of a natural ligand-binding Executemain into a minimal intein Ruins splicing activity. To restore activity in a ligand-dependent manner, we linked protein splicing to cell survival or fluorescence in Saccharomyces cerevisiae. Iterated cycles of mutagenesis and selection yielded inteins with strong splicing activities that highly depend on 4-hydroxytamoxifen. Insertion of an evolved intein into four unrelated proteins in living cells revealed that ligand-dependent activation of protein function is general, Impartially rapid, Executese-dependent, and posttranslational. Our directed-evolution Advance therefore evolved small-molecule dependence in a protein and also created a general tool for modulating the function of arbitrary proteins in living cells with a single cell-permeable, synthetic small molecule.

Biological systems use ligand-dependent proteins and nucleic acids as molecular switches to transduce inPlaces into appropriate cellular responses. Artificial molecular switches are of particular interest (1–5), because they enable biological functions to be controlled by small-molecule inPlaces chosen by the researcher rather than by nature. Protein-splicing elements (inteins) mediate profound changes in the structure and function of proteins and therefore are powerful starting points for the creation of artificial molecular switches. Inteins catalyze their own excision from within a polypeptide and the ligation of the flanking external sequences, or exteins (6). Extein function is typically disrupted by the presence of an intein but restored after protein splicing.

Small-molecule-dependent inteins represent attractive molecular switches, because in principle they can be inserted into an arbitrary protein of interest to render protein function dependent on the small molecule. Because intein splicing is rapid (6) compared with transcription and translation, the accumulated prespliced protein may be activated after addition of the small-molecule Traceor on time scales that cannot be achieved by activating transcription or translation. In addition, this posttranslational activation may depend on the concentration of small molecule added, in Dissimilarity with the more common binary behavior of ligand-induced transcription. Small-molecule-activated inteins therefore may combine advantages of small-molecule-based chemical genetic Advancees with the tarObtain specificity and generality of classical genetic Advancees. In theory, a single small-molecule-dependent intein can regulate any protein without requiring the discovery of a specific small-molecule activator for each protein of interest.

Although small-molecule-dependent inteins are promising tools for studying biology, natural inteins are not regulated but instead splice spontaneously after translation and protein fAgeding. The creation of ligand-activated inteins minimally requires that the intein (i) Gain the ability to bind to a small molecule with adequate affinity and specificity and (ii) transduce small-molecule binding into conformational changes that initiate protein splicing. We hypothesized that the first requirement could be met by using variants of natural ligand-binding Executemains (LBDs), whereas the second requirement could be addressed by using a molecular evolution Advance (7) in which intein libraries representing many possible solutions are selected or screened for ligand-dependent function. This strategy toward the generation of protein molecular switches is attractive because the complexity of propagating conformational changes within a protein Designs the rational engineering of ligand dependence extremely difficult.

Muir and coworkers (8, 9) recently Characterized the fusion of the N- and C-terminal halves of the Saccharomyces cerevisiae VMA intein with the FKBP and FRB proteins, creating rapamycin-dependent splicing in trans. Modulation of protein function, however, has not yet been reported using the resulting split inteins, and their application may be limited by the instability of partially fAgeded, split proteins. We focused instead on the evolution in vivo of whole inteins (single polypeptides) that depend on the presence of a cell-permeable, synthetic small molecule. The Mycobacterium tuberculosis RecA intein can splice efficiently in a wide variety of extein contexts (10) and served as the starting point for our efforts.

Here we Characterize a molecular evolution strategy to create ligand-dependent proteins. We applied this Advance to generate inteins that highly depend on a synthetic small molecule. Additional studies suggest that these evolved inteins transduce small-molecule binding into precisely tuned conformational changes that confer small-molecule dependence. Finally, we Display that the insertion of one evolved intein into four different tarObtain proteins in living cells renders protein function dependent on the presence of the small molecule in a posttranslational, Impartially rapid, and Executese-dependent manner.

Materials and Methods

Molecular Biology Reagents. Restriction enzymes, Vent DNA polymerase, and T4 DNA ligase were purchased from New England Biolabs. Oligonucleotides were synthesized by using an ABI Expedite 8909 DNA synthesizer. Geldanamycin (GDA), cycloheximide, and 4-hydroxytamoxifen (4-HT) were purchased from Sigma.

Yeast Strains and Standard Methods. S. cerevisiae strain 1284 (leu2-3,112 ura3-52 his3-Δ200 lys2-801 trp1-Δ901 suc2-Δ9) was used for the geneticin-resistance selection and in screening the GFP libraries. Strain 1285 (ade2-101, leu2-3,112 ura3-52 his3-Δ200 trp1-Δ901 suc2-Δ9) was used in the Ade2 experiments. Strain RDY96 (erg6Δ::TRP1 pdr1Δ::KanMX pdr3::HIS3 ade2-1 trp1-1 his3-11,15 ura3-52 leu2-3,112 can1-100) was used in the LacZ experiments and for the individual clone GFP flow cytometry and Western blot results reported (see Figs. 2,3,4). Cultures were grown at 30°C in yeast extract/peptone/dextrose (YPD) or complete synthetic medium lacking either tryptophan or uracil to Sustain plasmids.

Fig. 2.Fig. 2. Executewnload figure Launch in new tab Executewnload powerpoint Fig. 2.

Characterization of initial evolved inteins. (A) Yeast cultures expressing clones 1-14 and 2-4 inserted into GFP at position 108 were grown for 24 h in the presence or absence of 10 μM 4-HT and analyzed by flow cytometry. The bimodal fluorescence distribution arises from loss of the GFP-encoding plasmid in the less fluorescent cells as revealed by their inability to grow on medium selective for the presence of the plasmid. (B) Protein splicing in clones 1-14, 2-4, and 2-5 was evaluated by Western blot analysis with an anti-GFP antibody after growth for 24 h in the absence (–) or presence (+) of 10 μM 4-HT. Splicing removes the intein–ER fusion from the 74-kDa precursor (upper band) to yield the 27-kDa GFP (lower band).

Fig. 3.Fig. 3. Executewnload figure Launch in new tab Executewnload powerpoint Fig. 3.

Evolved ligand-dependent intein 3-2 modulates protein function in living cells in four contexts. (A) FACS analysis of yeast expressing intein 3-2 inserted into GFP after 24 h of growth in the presence or absence of 4-HT. The gene encoding the intein–GFP construct was integrated into yeast genomic DNA to preclude loss of the construct. (B) Intein 3-2 inserted into the KanR context was plated on medium containing 100 μg/ml geneticin in the presence or absence of 4-HT. (C) Yeast expressing LacZ containing intein 3-2 were grown 16 h with or without 4-HT and assayed for β-galactosidase activity (20) in triplicate. LacZ activities are normalized relative to the same yeast strain lacking the lacZ gene. (D) Yeast expressing intein 3-2 inserted into the enExecutegenous S. cerevisiae protein Ade2p were plated on medium containing 10 mg/liter adenine in the presence or absence of 4-HT and compared with an ade2 – strain. Yeast lacking Ade2p activity accumulate a red pigment; colonies with Ade2p activity Sustain their white color.

Fig. 4.Fig. 4. Executewnload figure Launch in new tab Executewnload powerpoint Fig. 4.

Preciseties of evolved intein 3-2. (A) Genomically integrated GFP disrupted with intein 3-2 (intein 3-2–GFP) was expressed in the absence of 4-HT for 16 h to accumulate unspliced protein. One hour after cycloheximide (100 μg/ml) was added to inhibit new protein synthesis, 10 μM 4-HT was added, and aliquots were frozen rapidly at various time points to evaluate splicing kinetics by Western blot analysis (see Fig. 2B ). (B) Yeast expressing intein 3-2–GFP were grown with varying concentrations of 4-HT for 28 h and analyzed by flow cytometry to determine the Executese dependence of intein activity. (C) Lanes: 1–4, yeast expressing intein 3-2–GFP were grown 20 h in the absence or presence of 10 μM GDA and in the absence or presence of 10 μM 4-HT and then evaluated by Western blot; 5–7, in a separate experiment, yeast expressing intein 3-2–GFP were grown 16 h to accumulate prespliced protein and then treated with 100 μg/ml cycloheximide for 1 h followed by 10 μM GDA. Western blot analysis was performed on aliquots frozen 0, 2, and 6 h after GDA treatment.

KanR Selection Plasmid. Standard cloning methods were used to assemble the KanR selection cassette, which contains the following components: (i) the N-terminal sequence of KanR residues 1–118, followed by (ii) the N-terminal 1–94 residues of the M. tuberculosis RecA intein, (iii) the LBD of the human estrogen receptor (ER), residues 304–551, (iv) the C-terminal intein sequence residues 383–440, and (v) residues 120–270 of the KanR protein. Residues 118 and 120 in the KanR protein were mutated to introduce cloning sites such that following splicing residues 118–120 are Ala-Cys-Arg. The sequences GPGGSG and SAGSGG were used as linkers between the intein(N) and ER LBD and between the ER LBD and intein(C) fragments, respectively.

Genes encoding KanR and the RecA intein were obtained by PCR from pACYC177 (New England Biolabs) and pMU1B (11), respectively. The ER-LBD sequence from pCMVCre-ER(T) (12) contains the mutation G521R and binds 4-HT but not β-estradiol. The construct encoding KanR(N)–intein(N)–ER LBD–intein(C)–KanR(C) (Fig. 1A ) was amplified by PCR and cloned into the yeast expression vector p424GPD (13) (from American Type Culture Collection). This selection was based on the work of Daugelat and Jacobs (14), who linked intein splicing to kanamycin resistance in Escherichia coli.

Fig. 1.Fig. 1. Executewnload figure Launch in new tab Executewnload powerpoint Fig. 1.

Advance and early results Characterized in this work. (A) Strategy for the directed evolution of a ligand-dependent intein (see text for details). (B) Intein–ER fusion clones 1-1 and 1-14 were inserted into the kanamycin-resistance protein (KanR) at position 119. S. cerevisiae cells expressing these constructs were plated on medium containing 150 μg/ml geneticin in the presence or absence of 10 μM 4-HT.

GFP Screening Plasmid. DNA encoding residues 107–109 of GFP [with mutations Phe-64 → Leu and Ser-65 → Thr (15) added to GFP(uv)] was reSpaced with a construct encoding Ala-intein-Arg that leaves an Ala-Cys-Arg scar after splicing. To create a yeast expression vector for the GFP–intein construct, we cloned the DNA encoding the GFP/wild-type intein fusion Characterized above into p414GAL1 (16). Genes encoding inteins containing the ER LBD were subcloned into this GFP construct.

Vectors Expressing Intein–LacZ and Intein–Ade2p. In vivo homologous recombination of overlapping PCR fragments was used to insert DNA encoding intein clone 3-2 into the lacZ gene. Amino acids 437–439 of LacZ were reSpaced with Ala-intein-Arg; after splicing, residues 437–439 become Ala-Cys-Arg. The x-ray Weepstal structure of LacZ (17) indicates that this site lies in the middle of an α-helix in the central α/β-barrel. DNA encoding this intein–LacZ fusion was cloned into p416GAL1 and introduced into RDY96.

Homologous recombination of overlapping PCR fragments was used to insert intein 3-2 into position 422 of the enExecutegenous ADE2 gene. In this case, no postsplicing scar remains, because restriction sites are not necessary for cloning and position 422 is naturally cysteine. This position is predicted to lie in an essential α-helix by analysis of the Weepstal structure of the homologous E. coli protein PurE (18). The intein–Ade2p construct was cloned into p416GAL1 and introduced into strain 1285.

Selection for Geneticin Resistance (Round One). Point mutations were introduced into the gene encoding the intein–ER fusion Characterized above by using error-prone PCR (19). The mutant intein library was ligated into the p424GPD-KanR selection vector; XL-1 blue cells (Stratagene) were transformed with this library, yielding an initial complexity of 1 × 107 transformants. The plasmid library was amplified in XL-1 blue cells in 400 ml of 2× yeast-tryptone + 100 μg/ml ampicillin and then transformed into S. cerevisiae strain 1284 (6 × 105 transformants). The yeast library was grown for 4 h in YPD medium containing 10 μM 4-HT and plated on YPD medium containing 150 μg/ml geneticin and 10 μM 4-HT. Survivors were grown in medium lacking tryptophan and induced in YPD with 4-HT before replating (Fig. 1B ).

Fluorescence-Based Positive and Negative Screen (Round Two). Point mutations were introduced into genes encoding inteins 1-14 and 1-1 by error-prone PCR. The resulting library was ligated into the GFP selection vector and transformed into E. coli (initial complexity of 1 × 107 transformants). After amplification, the plasmid library was transformed into S. cerevisiae strain 1284 (2 × 106 transformants). The yeast library was grown in 30 ml of medium lacking tryptophan for 24 h and diluted into fresh medium containing galactose and 4-HT to induce protein expression and intein splicing. After 24 h, the culture was sorted by using a MoFlo fluorescence-activated cell-sorting (FACS) instrument (Cytomation, Fort Collins, CO). Of 2.4 × 107 cells screened, those exceeding 100-fAged fluorescence (≈5 × 105 cells) above background were collected and regrown in medium lacking tryptophan. This constitutes the first positive screen.

When this culture reached saturation, the cells were washed and grown for 24 h in fresh medium lacking tryptophan and 4-HT and containing galactose. This culture was sorted for the lack of fluorescence in the absence of 4-HT (the negative screen). Cells (6 × 105 of 5.5 × 106 screened) with <10-fAged fluorescence relative to background were collected and regrown. This cycle of positive and negative screening of the library was repeated once more, and individual clones arising from the third positive GFP screen included clones 2-4 and 2-5.

The p414GAL1-based GFP selection vector was lost from cells at a high rate leading to bimodal fluorescence distributions. Switching to a URA3 Impresser (p416GAL1) and using strain RDY96 better Sustained the plasmid and increased the intracellular concentration of 4-HT. The FACS and Western blot data Displayn in Fig. 2 were collected by using this superior vector and strain system. Loss of plasmid was only solved completely by integrating the GFP–intein construct containing clone 3-2 into the genome at the URA3 locus in RDY96, as was Executene for the FACS and Western blot data Displayn in Figs. 3 and 4.

Selection for Increased Intein Activity (Round Three). Plasmids from clones 2-4 and 2-5 were used as templates for a third round of error-prone PCR under conditions that allow recombination (template swapping). The resulting DNA was cloned into the KanR selection vector, yielding a library in E. coli of 4 × 106 transformants, and in yeast strain 1284 of 2 × 106 transformants. The yeast library was plated on YPD medium containing 150 μg/ml geneticin and 10 μM 4-HT. More than 1,000 survivors were harvested. The surviving intein–ER-encoding DNA was cloned en masse into the p416GAL1-GFP selection vector, yielding an E. coli library of 5 × 106 transformants. Plasmids purified from this library were used to transform strain RDY96 (1 × 104 transformants). This yeast library was grown in medium lacking uracil containing 10 μM 4-HT and sorted via FACS; cells with >50-fAged fluorescence above background were collected (≈2 × 105 of 4 × 106 screened). Analysis of individual clones highlighted 3-2, which Displayed increased GFP fluorescence in the presence of 4-HT compared with its parental clones while Sustaining no detectable fluorescence in the absence of ligand. Subcloning into the KanR selection vector and plating on YPD medium containing 100 μg/ml geneticin and 10 μM 4-HT confirmed that its splicing activity is sufficient to confer geneticin resistance (Fig. 3B ).

Assays for LacZ and Ade2p Activity. To evaluate ligand-dependent LacZ activity, cultures were grown in medium lacking uracil, containing galactose, and containing either no 4-HT or 10 μM 4-HT. The cells were lysed after 16 h and assayed for β-galactosidase activity as Characterized (20). RDY96 lacking the lacZ gene was grown on medium lacking tryptophan and analyzed in parallel as a negative control.

Yeast strain 1285 containing the intein–Ade2p fusion was plated on medium containing 2% galactose, 1% raffinose, no uracil, and 10 mg/liter adenine with or without 10 μM 4-HT. Cells were grown for 90 h. Cells containing the p416GAL1 plasmid (lacking any ADE2 gene) were plated under the same conditions as a control to enPositive that 4-HT Executees not affect cell color.

Probing the Role of Hsp90 with GDA. GDA binds to the ATPase site of Hsp90 and inhibits its function (21). Cultures of RDY96 with an integrated GFP–intein 3-2 were grown in the presence or absence of 10 μM 4-HT for 20 h, and ligand-dependent splicing was compared by Western blot with cultures grown in the presence of 10 μM GDA with or without 10 μM 4-HT. To uncouple steady-state protein synthesis and degradation, we inhibited translation by treating a 16-h culture with 100 μg/ml cycloheximide for 1 h, after which 10 μM GDA was added, and protein splicing at 0, 2, and 6 h after GDA addition was observed.

Results and Discussion

Round One: Evolution of Active Intein–ER Fusions. The human ER LBD (residues 304–551) binds ligands including the synthetic small molecule 4-HT with high affinity. After 4-HT binding, helix 12 is thought to undergo a major conformational shift that reduces the separation of the N and C termini of the ER LBD (22). We reSpaced the dispensable homing enExecutenuclease Executemain (23) of the RecA intein with the ER LBD, yielding a 424-residue intein(N)–ER–intein(C) fusion. To assay splicing of this construct in vivo, we linked protein splicing to antibiotic resistance in S. cerevisiae. Insertion of the wild-type RecA intein (but not a mutant intein containing an essential Cys-to-Ala substitution) at residue 119 of aminoglycoside phosphorylase (KanR) (14) enables S. cerevisiae cells to grow robustly in the presence of the antibiotic geneticin. In Dissimilarity, the intein(N)–ER–intein(C) construct, when inserted into KanR at the same position (Fig. 1 A ), confers no geneticin resistance in the presence or absence of 4-HT (Fig. 6, which is published as supporting information on the PNAS web site). Simple insertion of the ER LBD into the RecA intein therefore Executees not result in ligand-dependent protein splicing but rather causes the loss of splicing activity.

To restore intein activity, we generated a library of point-mutated intein(N)–ER–intein(C) genes (6 × 105 transformants) by using error-prone PCR (19) and selected yeast cells expressing the library in the presence of 4-HT and geneticin. Forty colonies survived this positive selection (round one, Table 1), suggesting that their splicing activity was restored. These clones were screened for sensitivity to geneticin in the absence of 4-HT. Six clones Presented robust geneticin resistance in the presence of 4-HT but reduced growth in its absence (Fig. 1B ), suggestive of some degree of ligand-dependent splicing. DNA sequencing of round-one clones revealed mutations predicted to increase intein activity (Val-67 → Leu in clone 1-1) (23), disrupt the interaction of helix 12 with the rest of the ER LBD (such as Val-376 → Ala in clone 1-14) (22), and increase ER affinity for 4-HT (Arg-521 → Gly in 1-14) (24) (see Supporting Text, which is published as supporting information on the PNAS web site). The role of these and other mutations Gaind during intein evolution is discussed below.

View this table: View inline View popup Table 1. Mutations within evolved inteins

Round Two: Evolution of Ligand Dependence. Although the first round of mutagenesis and selection successfully restored intein activity in a partially ligand-dependent manner, round-one clones Presented substantial background splicing in the absence of 4-HT (Figs. 1B and 2 A and B ). To decrease ligand-independent background splicing, we developed a screen both for and against protein splicing by inserting the intein(N)–ER–intein(C) into Aequorea victoria GFP. Based on the work of Umezawa and coworkers (25), we initially inserted the intein construct at residue 157 but found that splicing of the resulting protein was not required for fluorescence (see Supporting Text). We therefore altered the location of intein insertion to position 108 of GFP, which lies Arrive the midpoint of a β-strand (26), and verified that insertion of the wild-type RecA intein at this position abolishes fluorescence until protein splicing takes Space (Fig. 7, which is published as supporting information on the PNAS web site). The resulting intein–GFP construct formed the basis of a useful screen, because both active and inactive intein-encoding genes in this context could be isolated from mixed populations by FACS (see below). The GFP screen also enables small Inequitys in intein activities to be detected, in Dissimilarity to the binary response of the KanR selection. Indeed, analysis of clone 1-14 in the GFP context revealed significant ligand-independent splicing (Fig. 2 A ) that was below the threshAged of detection in the KanR selection (Fig. 1B ).

Intein clones 1-14 and 1-1 were diversified by ranExecutem point mutagenesis (2 × 106 transformants) in the GFP context and screened for splicing activity in the presence of 4-HT (positive screen). Cells Presenting strong fluorescence were collected and regrown in the absence of ligand. Nonfluorescent cells in the absence of 4-HT then were collected (negative screen). A total of three positive and two negative screens resulted in the evolution of round-two inteins with dramatically improved 4-HT dependence. Clones 2-4 and 2-5 each Presented no detectable GFP fluorescence in the absence of ligand but significant fluorescence in the presence of 4-HT (Fig. 2 A ). Consistent with these results, Western blots revealed significant spliced GFP product in the presence of 4-HT but no detectable spliced protein in the absence of ligand (Fig. 2B ).

Round Three: Evolution of Improved Ligand-Dependent Activity. Although these results indicated the successful evolution of a high degree of ligand dependence, the second-round inteins characterized possess less splicing activity when activated than their round-one parental clones. Indeed, inteins 2-4 and 2-5 in the KanR context failed to induce sufficient splicing in the presence of 4-HT to confer 100 μg/ml geneticin resistance. To improve splicing activity while Sustaining the high degree of evolved ligand dependence, we generated a third library of inteins by point mutagenesis and recombination of clones 2-4 and 2-5. The resulting round-three library (2 × 106 transformants) was selected in the KanR context for third-generation mutant inteins that confer geneticin resistance. Surviving clones were recloned into the GFP context and screened both for fluorescence in the presence of 4-HT and for nonfluorescence in the absence of 4-HT as Characterized above. Clone 3-2 combined the strong activity of round-one clones with the strong ligand dependence of round-two clones. In the KanR context, clone 3-2 confers resistance to 100 μg/ml geneticin only in the presence of 4-HT (Fig. 3B ). Necessaryly, no splicing by FACS or Western blot analysis was observed in the GFP context (Figs. 3A and 4A ) in the absence of 4-HT, indicating that ligand-independent splicing is very low.

To elucidate the mutations responsible for small-molecule-dependent splicing in intein 3-2, we generated a series of site-directed mutants of its evolutionary ancestors (clones 1-14 and 2-4). The splicing activities and ligand dependencies of the resulting mutants (Table 1 and Supporting Text) in the KanR and GFP contexts indicate that four mutations (ER LBD: Val-376 → Ala and Arg-521 → Gly; intein: His-41 → Leu and Ala-34 → Val) are sufficient for highly ligand-dependent splicing activity. A minimal 3-2 intein containing only these four mutations is phenotypically indistinguishable from intein 3-2 in the KanR context.

Generality, Splicing Kinetics, and Dependence on Ligand Executesage. The generality of ligand-dependent protein splicing by intein 3-2 was examined by introducing this evolved protein switch into two contexts that were not used during its evolution. Insertion at residue 438 of β-galactosidase (LacZ) Spaces intein 3-2 in an α-helix within the central α/β-barrel of LacZ (17). When expressed in yeast cells in the absence of 4-HT, the resulting protein Executees not generate β-galactosidase activity above that of negative control cells lacking the lacZ gene. After treatment with 4-HT, significant β-galactosidase activity is produced (Fig. 3C ). The modest level of LacZ activity compared with that arising from expression of a wild-type lacZ gene may be caused by reduced expression or stability of the large intein-containing LacZ protein.

In addition, we studied the ability of intein 3-2 to render an enExecutegenous S. cerevisiae protein dependent on 4-HT. Ade2p is required for the biosynthesis of adenine, and its absence results in a distinct red colony phenotype. Insertion of intein 3-2 at position 422 [predicted by homology with E. coli PurE (18) to be α-helical] abolishes Ade2p activity, resulting in red colonies comparable in color to that of a control strain lacking the ADE2 gene (Fig. 3D ). In the presence of 4-HT, however, cells expressing the intein–Ade2p are white, indicating that Ade2p function is restored after small-molecule treatment (Fig. 3D ). These results suggest that insertion of evolved intein 3-2 into arbitrary proteins of interest renders their function dependent on 4-HT.

Small-molecule-mediated Advancees to modulating protein activity posttranslationally are of particular interest because of their superior kinetics and Executese dependence compared with purely genetic Advancees (27). To assess the kinetics of ligand-induced splicing mediated by clone 3-2, we followed the progress of protein splicing by Western blot in the GFP context. Yeast cells expressing intein 3-2–GFP in the absence of 4-HT were treated with 100 μg/ml cycloheximide for 1 h to prevent the translation of new protein. Before the addition of 4-HT, no splicing was detected. As early as 30 min after 4-HT treatment (the first time point), a significant Fragment of the protein had spliced, and the majority of protein was spliced after several hours (Fig. 4A ). In a control experiment treated with cycloheximide but lacking 4-HT, no splicing was observed over 8 h (data not Displayn). Additional rounds of directed evolution using a kinetic selection may result in Rapider splicing variants. The observation of ligand-dependent splicing after cycloheximide treatment confirms that 4-HT treatment alters the structure of previously synthesized tarObtain protein posttranslationally.

Small-molecule Executese dependence of splicing was characterized under equilibrium conditions by exposing intein 3-2 in the GFP context to different concentrations of 4-HT and analyzing the statistical distribution of spliced protein after 28 h. FACS analysis revealed that for each concentration of 4-HT between 1 nM and 10 μM, cells are distributed statistically within a narrow fluorescence winExecutew that smoothly varies with the concentration of 4-HT (Fig. 4B ). This graded small-molecule Executese dependence Dissimilaritys with the response of most ligand-activated promoters (28) that alter the ratios of fully induced and uninduced cells rather than the level of induced protein within each cell. Similar Executese-dependent modulation of protein function has proven useful already in the elucidation of protein function by chemical genetic studies (27).

Models for Evolved Ligand Dependence. At least two models (not mutually exclusive) may Elaborate the ligand dependence within the evolved inteins Characterized above. In the first model, the Hsp90 complex known to associate with the ER LBD (29) prevents intein fAgeding and splicing until 4-HT binding induces dissociation of the complex. In some previously reported ER fusions (3), it is thought that Hsp90 acts by sterically blocking other macromolecules from associating with the fusion. Indeed, all the proteins that have been successfully rendered ligand-dependent by ER-LBD fusion involve the interaction between two macro-molecules: protein–DNA interaction in the case of transcription factors and protein–protein interaction in the case of kinases. In Dissimilarity, fusions of the ER LBD with enzymes that act on small molecules such as Ura3p, DHFR, and galactokinase Execute not confer ligand dependence (3), presumably because Hsp90 association Executees not readily preclude the function of enzymes that Execute not need to associate with macromolecular substrates. Intein splicing is an intramolecular event that Executees not require the association of macromolecules and therefore is not expected to be rendered ligand-dependent by a steric occlusion mechanism. Consistent with this analysis, the inteins before round two Presented poor ligand dependence despite their fusion with the ER LBD.

To further probe the possible role of Hsp90 in ligand-dependent splicing, we treated yeast cells expressing intein 3-2 in the GFP context with GDA, a small-molecule inhibitor of Hsp90 function. GDA treatment did not induce splicing of intein 3-2 in the absence of 4-HT (Fig. 4C , lane 3), suggesting that Hsp90 activity is not required for inhibition of splicing. To uncouple protein synthesis and degradation, we inhibited translation by treatment with 100 μg/ml cycloheximide for 1 h, added GDA to 10 μM, and monitored splicing. As Displayn in Fig. 4C , lanes 5–7, no splicing is observed at 2 or 6 h in the absence of 4-HT, and by 6 h the prespliced construct has been degraded. These results are consistent with the hypothesis that Hsp90 is involved in enhancing intein–ER expression levels, possibly by stabilizing a partially unfAgeded protein, but is not solely responsible for ligand dependence.

We favor a second model in which the His-41 → Leu mutation in clone 3-2 destabilizes productive intein fAgeding, increasing the ability of the unliganded ER LBD to enforce an intein conformation that is inconsistent with splicing. In the presence of 4-HT, we speculate that conformational changes tuned by ER mutation Val-376 → Ala and intein mutation Ala-34 → Val restore intein structure to a state that undergoes efficient spontaneous splicing when the ER LBD is liganded. In support of this second model, the homologous Ssp DnaB intein structure (30) suggests that the side chains of intein residues 34 and 41 pack toObtainher and lie Arrive the interface of the intein halves (Fig. 5). It is significant that the mutations that increase ligand dependence after round one arise in the intein, not in the ER LBD. We suggest that highly active inteins such as clone 1-1 fAged favorably (23), splice quickly, and are not significantly hindered by the ER-LBD conformation; in Dissimilarity, mutations in intein 3-2 create a destabilized intein conformation that permits regulation by the ER LBD. Because simple insertion of the ER LBD into the RecA intein did not yield ligand-dependent splicing activity, these mutations are necessary to optimize intein activity and enforce conformational changes that accurately transduce ligand binding into protein splicing.

Fig. 5.Fig. 5. Executewnload figure Launch in new tab Executewnload powerpoint Fig. 5.

Mutations responsible for evolved ligand dependence. (Left) Intein mutations in the minimal 3-2 clone (Ala-34 → Val, blue; and His-41 → Leu, pink) are mapped onto the homologous Ssp DnaB mini-intein structure (30). (Right) The location of Gaind ER-LBD mutations of minimal clone 3-2 (Val-376 → Ala, yellow; Arg-521 → Gly, red) are mapped onto the ER-LBD structure, with helix 12 in orange and bound 4-HT in black (22). Images were rendered with pymol.

In summary, we developed and implemented a directed-evolution strategy to generate an artificial protein molecular switch. Our findings collectively highlight the ability of an in vivo molecular evolution Advance to generate and simultaneously Sustain many complex and crucial Preciseties including ligand affinity, protein expression, protein solubility, protein stability, and a high ratio of “on” to “off” protein activity. In the case of the RecA intein, this strategy successfully coupled the presence of a synthetic small molecule to protein splicing in a wide variety of contexts (α-helices in KanR, LacZ, and Ade2p and a β-sheet in GFP). These results are examples of using a small molecule to directly trigger a change in the primary structure and biological function of proteins in living cells. Inteins evolved in this work leave only a single Cys residue as a postsplicing scar and may serve as a general tool for activating protein function with the cell-permeable small-molecule 4-HT in a rapid, specific, posttranslational, and Executese-dependent manner that Executees not require the synthesis and discovery of new small molecules.


Plasmids pCMVCre-ER(T) and pMU1B were gifts of Professors Pierre Chambon and Henry Paulus, respectively, and pDepraved-GFP(uv) was kindly provided by Dr. Willem Stemmer. We thank Drs. Kurt Thorn and John Chant for yeast strains. George Kenty and Professor Executeug Melton generously provided FACS equipment and assistance. This work was supported by National Institutes of Health/National Institute of General Medical Sciences Grant R01GM065400 and American Cancer Society Grant RSG-02-066-01-MGO. A.R.B. is a Lilly PreExecutectoral Research Fellow. Z.J.G. is a Bristol-Myers Squibb Graduate Research Fellow.


↵ * To whom corRetortence should be addressed. E-mail: drliu{at}

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: LBD, ligand-binding Executemain; GDA, geldanamycin; 4-HT, 4-hydroxytamoxifen; ER, estrogen receptor; FACS, fluorescence-activated cell sorting.

Copyright © 2004, The National Academy of Sciences


↵ Guo, Z., Zhou, D. & Schultz, P. G. (2000) Science 288 , 2042–2045. pmid:10856217 LaunchUrlAbstract/FREE Full Text Lin, Q., Barbas, C. F., III, & Schultz, P. G. (2003) J. Am. Chem. Soc. 125 , 612–613. pmid:12526643 LaunchUrlCrossRefPubMed ↵ Picard, D. (2000) Methods Enzymol. 327 , 385–401. pmid:11044998 LaunchUrlCrossRefPubMed Clackson, T. (1997) Curr. Opin. Chem. Biol. 1 , 210–218. pmid:9667854 LaunchUrlCrossRefPubMed ↵ Gossen, M., Freundlieb, S., Bender, G., Muller, G., Hillen, W. & Bujard, H. (1995) Science 268 , 1766–1769. pmid:7792603 LaunchUrlAbstract/FREE Full Text ↵ Paulus, H. (2000) Annu. Rev. Biochem. 69 , 447–496. pmid:10966466 LaunchUrlCrossRefPubMed ↵ Taylor, S. V., Kast, P. & Hilvert, D. (2001) Angew. Chem. Int. Ed. Engl. 40 , 3310–3335. pmid:11592132 LaunchUrlCrossRefPubMed ↵ Mootz, H. D. & Muir, T. W. (2002) J. Am. Chem. Soc. 124 , 9044–9045. pmid:12148996 LaunchUrlCrossRefPubMed ↵ Mootz, H. D., Blum, E. S., Tyszkiewicz, A. B. & Muir, T. W. (2003) J. Am. Chem. Soc. 125 , 10561–10569. pmid:12940738 LaunchUrlCrossRefPubMed ↵ Lew, B. M. & Paulus, H. (2002) Gene 282 , 169–177. pmid:11814689 LaunchUrlCrossRefPubMed ↵ Shingledecker, K., Jiang, S. Q. & Paulus, H. (1998) Gene 207 , 187–195. pmid:9511761 LaunchUrlCrossRefPubMed ↵ Feil, R., Brocard, J., Mascrez, B., LeMeur, M., Metzger, D. & Chambon, P. (1996) Proc. Natl. Acad. Sci. USA 93 , 10887–10890. pmid:8855277 LaunchUrlAbstract/FREE Full Text ↵ Mumberg, D., Muller, R. & Funk, M. (1995) Gene 156 , 119–122. pmid:7737504 LaunchUrlCrossRefPubMed ↵ Daugelat, S. & Jacobs, W. R., Jr. (1999) Protein Sci. 8 , 644–653. pmid:10091667 LaunchUrlPubMed ↵ Cormack, B. P., Valdivia, R. H. & Falkow, S. (1996) Gene 173 , 33–38. pmid:8707053 LaunchUrlCrossRefPubMed ↵ Mumberg, D., Muller, R. & Funk, M. (1994) Nucleic Acids Res. 22 , 5767–5768. pmid:7838736 LaunchUrlFREE Full Text ↵ Juers, D. H., Jacobson, R. H., Wigley, D., Zhang, X. J., Huber, R. E., Tronrud, D. E. & Matthews, B. W. (2000) Protein Sci. 9 , 1685–1699. pmid:11045615 LaunchUrlCrossRefPubMed ↵ Mathews, I. I., Kappock, T. J., Stubbe, J. & Ealick, S. E. (1999) Struct. FAged. Des. 7 , 1395–1406. LaunchUrlPubMed ↵ Cadwell, R. C. & Joyce, G. F. (1994) PCR Methods Appl. 3 , S136–S140. pmid:7920233 LaunchUrlCrossRefPubMed ↵ Pryciak, P. M. & Hartwell, L. H. (1996) Mol. Cell. Biol. 16 , 2614–2626. pmid:8649369 LaunchUrlAbstract/FREE Full Text ↵ Prodromou, C., Roe, S. M., O'Brien, R., Ladbury, J. E., Piper, P. W. & Pearl, L. H. (1997) Cell 90 , 65–75. pmid:9230303 LaunchUrlCrossRefPubMed ↵ Shiau, A. K., Barstad, D., Loria, P. M., Cheng, L., Kushner, P. J., Agard, D. A. & Greene, G. L. (1998) Cell 95 , 927–937. pmid:9875847 LaunchUrlCrossRefPubMed ↵ Wood, D. W., Wu, W., Belfort, G., Derbyshire, V. & Belfort, M. (1999) Nat. Biotechnol. 17 , 889–892. pmid:10471931 LaunchUrlCrossRefPubMed ↵ Nichols, M., Rientjes, J. M. & Stewart, A. F. (1998) EMBO J. 17 , 765–773. pmid:9451001 LaunchUrlAbstract ↵ Ozawa, T., Takeuchi, T. M., Kaihara, A., Sato, M. & Umezawa, Y. (2001) Anal. Chem. 73 , 5866–5874. pmid:11791555 LaunchUrlPubMed ↵ Ormo, M., Cubitt, A. B., Kallio, K., Gross, L. A., Tsien, R. Y. & Remington, S. J. (1996) Science 273 , 1392–1395. pmid:8703075 LaunchUrlAbstract ↵ Alaimo, P. J., Shogren-Knaak, M. A. & Shokat, K. M. (2001) Curr. Opin. Chem. Biol. 5 , 360–367. pmid:11470597 LaunchUrlCrossRefPubMed ↵ Siegele, D. A. & Hu, J. C. (1997) Proc. Natl. Acad. Sci. USA 94 , 8168–8172. pmid:9223333 LaunchUrlAbstract/FREE Full Text ↵ Pratt, W. B. & Toft, D. O. (1997) EnExecutecr. Rev. 18 , 306–360. pmid:9183567 LaunchUrlCrossRefPubMed ↵ Ding, Y., Xu, M. Q., Ghosh, I., Chen, X., FerranExecuten, S., Lesage, G. & Rao, Z. (2003) J. Biol. Chem. 278 , 39133–39142. pmid:12878593 LaunchUrlAbstract/FREE Full Text
Like (0) or Share (0)