An Antiproliferative Genetic Screening Identifies a Peptide Aptamer That Targets Calcineurin and Up-regulates Its Activity*

Peptide aptamers are combinatorial recognition molecules that consist of a constant scaffold protein displaying a doubly constrained variable peptide loop. They bind specifically target proteins and interfere with their function. We have built a peptide aptamer library in a lentiviral expression system to isolate aptamers that inhibit cell proliferation in vitro. Using one of the isolated aptamers (R5G42) as a bait protein, we have performed yeast two-hybrid screening of cDNA libraries and identified calcineurin A as a target protein candidate. R5G42 bound calcineurin A in vitro and stimulated its phosphatase activity. When expressed transiently in human cells, R5G42 induced the dephosphorylation of BAD. We have identified an antiproliferative peptide aptamer that binds calcineurin and stimulates its activity. The use of this ligand may help elucidate the still elusive structural mechanisms of activation and inhibition of calcineurin. Our work illustrates the power of phenotypic screening of combinatorial protein libraries to interrogate the proteome and chart molecular regulatory networks.

Peptide aptamers are combinatorial recognition molecules that consist of a constant scaffold protein displaying a doubly constrained variable peptide loop. They bind specifically target proteins and interfere with their function. We have built a peptide aptamer library in a lentiviral expression system to isolate aptamers that inhibit cell proliferation in vitro. Using one of the isolated aptamers (R5G42) as a bait protein, we have performed yeast twohybrid screening of cDNA libraries and identified calcineurin A as a target protein candidate. R5G42 bound calcineurin A in vitro and stimulated its phosphatase activity. When expressed transiently in human cells, R5G42 induced the dephosphorylation of BAD. We have identified an antiproliferative peptide aptamer that binds calcineurin and stimulates its activity. The use of this ligand may help elucidate the still elusive structural mechanisms of activation and inhibition of calcineurin. Our work illustrates the power of phenotypic screening of combinatorial protein libraries to interrogate the proteome and chart molecular regulatory networks.

Molecular & Cellular Proteomics 6:451-459, 2007.
In absence of classical genetics, the deciphering of mammalian regulatory networks rests mostly on the reverse genetics methodology and particularly on the use of transdominant negative agents such as dominant negative alleles (1), antibodies (2), nucleic acid aptamers (3), peptide aptamers (4), antisense or small interfering RNA (5), and small molecule inhibitors when available (6). In most applications, these agents are designed or selected to specifically target a protein, and they are then introduced into cellular or animal models to assess the phenotypic consequences of the targeted perturbation they exert. Another approach consists of constructing large libraries of transdominant agents in retroviral vectors and performing genetic selections or screening to isolate library members that confer given phenotypes. Libraries of antisense cDNAs (7), random fragments of cDNAs (8), ribozymes (9), combinatorial peptides (10), and short hairpin RNAs (11) have been used successfully to interrogate proteomes and identify new members of mammalian regulatory pathways.
A straightforward selection of transdominant negative agents that inhibit cell proliferation is an oxymoron. Elaborate experimental schemes have thus been developed and used successfully to identify cytostatic random cDNA fragments (12) and random linear peptides terminally fused to GFP 1 (13). In both cases, a counterselection against dividing cells has been devised and, in the latter case, coupled to a positive screening for cells that do not divide and thus maintain a fluorescent vital dye. Although different antiproliferative linear peptides have been isolated, their mechanism of action has not been elucidated so far (13).
We set out to isolate another kind of combinatorial protein reagents for their ability to inhibit tumor cell proliferation with the hope of identifying proteins playing an unexpected role in this biological phenomenon. Peptide aptamers are man-made combinatorial protein reagents that bind target proteins and can interfere with their function in living cells and organisms (Ref. 14; for a review, see Ref. 4). They consist of conformationally constrained random sequence peptide loops displayed by a scaffold protein. They bind their cognate targets with a strong affinity and, usually, a high specificity, which allows them to discriminate between closely related members within a protein family (14) or even between different allelic variants of a given protein (15). So far, peptide aptamers have been mostly selected through yeast two-hybrid screening experiments for their ability to bind a given target protein. In fewer instances, peptide aptamers have been selected for their ability to confer selectable phenotypes to yeast (16,17) and bacteria (18). Peptide aptamers selected in yeast have been used successfully to identify their cognate target proteins by two-hybrid screening.
Here we have constructed a peptide aptamer library in a simian immunodeficiency virus (SIV)-derived gene expression system. We have performed an iterative genetic screening to isolate peptide aptamers that inhibit tumor cell proliferation. We have identified the catalytic subunit of the calcium-activated protein phosphatase calcineurin as a target of one of the isolated aptamers. We have shown that this aptamer up-regulates the phosphatase activity of calcineurin in vitro and in cultured cells. Our work has identified an antiproliferative molecule that binds and stimulates calcineurin through a seemingly original mechanism.

EXPERIMENTAL PROCEDURES
Cell Culture-We maintained all mammalian cells in a 5% CO 2 atmosphere at 37°C in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% (v/v) fetal calf serum and 100 g/ml penicillin-streptomycin.
Construction of Lentiviral Vectors-All the lentivectors were derived from pR4SA-EFS-GFP-W (19). We first digested this vector with HindIII, thus eliminating EGFP, woodchuck hepatitis post-transcriptional regulatory element, and EcoRI sites, to create pVRV1. We blunted the remaining EcoRI site upstream of the cytomegalovirus promoter and religated the vector to create pVRV2. We digested pVRV2 with BamHI and HindIII, and we ligated the hybridized oligodeoxynucleotides 5Ј-GATCGCTAAGCGAATTCCTCGAGGCGCGCGT-CGACCAGGATCC-3Ј and 5Ј-AGCTTGGATCCTGGTCGACGCGCG-CCTCGAGGAATTCGCTTAGC-3Ј to create pVRV3, which bears a multiple cloning sequence. We constructed pVRV4 by inserting an IRES-EGFPf (farnesylated enhanced GFP) coding sequence in pVRV3. This was done by a multiplex ligation between SalI/BamHIcut pVRV3, a SalI/NcoI-cut encephalomyocarditis virus IRES cassette (from pIRES2-EGFP, Clontech), and an NcoI/BamHI-cut EGFPf coding sequence (from pEGFP-F, Takara Bio). We then PCR-amplified an HA-tagged human thioredoxin (HTRX) fragment from pJMX-HTRX 2 using the oligonucleotides 5Ј-GCGGCTAAGCCATGTACCCTTAT-GATGTGCCAG-3Ј and 5Ј-GGAGACTTGACCAAACCTCTG-3Ј, and we ligated this fragment into BlpI/XhoI-cut pVRV4. The resulting plasmid, pVRV6, directs the bicistronic expression of an HA-tagged human thioredoxin (with a modified active site) and of EGFP carrying a farnesylation sequence to anchor the marker protein to plasma membranes.
Construction of the Peptide Aptamer Expression Library-We constructed pBK1, a library of peptide aptamers bearing 10 amino acids within the active site of HA-tagged human thioredoxin. We annealed the oligonucleotides 5Ј-TGGGCCGAGTGGAGCGGTCCG(NNS) 9 NN-CGGACCGAGCAAGATGATCGCCCC-3Ј (N, A or C or G or T; S, C or G) and 5Ј-GGGGCGATCATCTTGCTCGGTCCG-3Ј, and we produced duplexes using the Klenow DNA polymerase. We ligated the AvaII-cut duplexes into CpoI-cut pVRV6. We transformed the ligation product into ElectroTen Blue competent bacteria (Stratagene), and we obtained 8.5 ϫ 10 9 transformants.
Viral Vector Production-We produced lentiviral particles by transfecting into 293T cells the following plasmids: (i) pVRV6, pVRV12 (pVRV6 directing the expression of p21 Cip1 ), pBK1, or any aptamer sublibrary; (ii) helper pSIV15, directing the expression of gag and pol (20); (iii) FbmoSalf, directing the expression of a murine ecotropic envelope (19); and (iv) pRev (20). In some experiments, plasmids iii and iv were replaced by the G-rev plasmid (20), directing the expression of Rev and the vesicular stomatitis virus-G pantropic envelope. We collected and filtered lentivirus-containing supernatants 48 h post-transfection through a 0.45-m filter. We determined viral titers by infecting XC or HeLa cells and counting GFP-positive cells with a cytometer (FACScan, BD Biosciences). We routinely infected from 40 to 100% cells.
Screening of Antiproliferative Peptide Aptamers-We plated XC cells 24 h before infection (2 ϫ 10 5 cells/well, 6-well plates, six plates). To infect the cells, we added a medium containing a viral supernatant and 6 g/ml Polybrene. Three days later, we collected the cells, washed them with PBS, stained 5 ϫ 10 5 cells/ml with 10 M Cell-Tracker TM Orange CMTMR (5-(and-6)-(((4-chloromethyl)benzoyl) amino)tetramethylrhodamine; Invitrogen) in PBS at 37°C for 30 min, and incubated the cells in culture medium for another 30 min at 37°C. We then plated the cells onto 10-cm dishes (10 6 cells/dish). After 72 h, we collected the cells, and we sorted the highest percentile of CMTMR fluorescent cells using a FACS Vantage flow cytometer (BD Biosciences). We pooled the sorted cells, and we extracted their genomic DNA using a Wizard Genomic DNA purification kit (Promega). We PCR-amplified aptamer coding genes using the oligonucleotides 5Ј-AACCGGTGCCTAGAGAAGGT-3Ј and 5Ј-AGACCCCT-AGGAATGCTCGT-3Ј. We cloned the EcoRI/XhoI-digested products into EcoRI/XhoI-cut pVRV6 to create successive sublibraries of peptide aptamers named pCMTMR 1 to 7.
Two-hybrid Screening of R5G42-interacting Proteins-We digested pVRV6-R5G42 with EcoRI and XhoI and ligated the fragment into EcoRI/XhoI-cut pGILDA (Clontech) to create pGILDA-R5G42, a plasmid directing the galactose-inducible expression of a LexA-R5G42 fusion protein. We transformed MB226␣ pSH18-34 yeast (21) with pGILDA-R5G42 and MB210a yeast (21) with human fetal brain and human testes cDNA libraries constructed in pJG4-5. We performed the yeast-two hybrid screening of both libraries essentially as described previously (21) using 4 ϫ 10 8 and 2.4 ϫ 10 8 colony-forming units from the brain and testes libraries, respectively. We estimated the mating efficiency at 50 and 58% and the number of diploid exconjugants at 0.2 ϫ 10 8 and 1 ϫ 10 8 for the brain and testis cDNA library-transformed yeast, respectively. We induced the expression of the bait and the libraries at 30°C for 5 h from 10% of the diploids. We collected the yeast and plated them onto 10 Ura Ϫ His Ϫ Trp Ϫ Leu Ϫ galactose/raffinose plates for 5 days. We replica-plated onto 10 Ura Ϫ His Ϫ Trp Ϫ Ade Ϫ 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal) galactose/raffinose plates. We picked 60 clones from the brain and 48 clones from the testes library that grew in the absence of leucine and adenine and that displayed a ␤-galactosidase activity. Library plasmids were recovered and retransformed into EGY48␣. The interaction phenotypes were confirmed by a mating assay with EGY42a transformed with pGILDA-R5G42. We then sequenced the library cDNAs from most reconfirmed clones.
The 5Ј-end of RH6 (uppercase) matches part of the HA epitope tag, and the 5Ј-ends of RH4, RH3, and RH7 (uppercase) match the 5Ј-extremity of the alcohol dehydrogenase terminator. We performed the PCRs using pCMV-SPORT6-CnA␤ (see below) as a template. We constructed CNA␤Cter ⌬1, ⌬2, and CaM by combining oligonucleotides RH6/RH4, RH6/RH3, and RH6/RH7, respectively. We co-transformed MB210a with the PCR products and EcoRI/XhoI-cut pJG4-5. We retrieved the prey plasmids from the transformants (21), and we checked the homologous recombination products by sequencing. We also transformed MB210a with positive and negative controls of interaction. We co-transformed TB50␣ with pSH18-34T (a plasmid bearing a high sensitivity lacZ reporter gene) and pGILDA directing the expression of LexA, LexA-R5G42, LexA-R7G44, and LexA-R5G52. We performed the yeast two-hybrid mating assays as described previously (21).
In Vitro Binding Assay-We first digested pVRV6-aptamer plasmids with EcoRI and XhoI and ligated the fragments into EcoRI/XhoI-cut pGEX4T1. We expressed GST-aptamer fusions in a BL-21(DE3) Escherichia coli strain. We diluted 1:100 overnight cultures and let them grow at 37°C to reach an A 600 of 0.6 -0.8. We induced the expression of fusion proteins by adding 1 mM isopropyl 1-thio-␤-D-galactopyranoside and incubating overnight at 20°C with vigorous shaking. We collected the bacteria and resuspended them into a lysis buffer (50 mM Tris, pH 8, 100 mM NaCl, 1 mM DTT) containing 1 mg/ml lysozyme. We froze and thawed the cells three times and sonicated them on ice. We centrifuged the lysates at 13,000 ϫ g for 30 min, and we collected the soluble fractions. We immobilized equal amounts of GST-aptamer fusion proteins on 100 l of glutathione-Sepharose 4B beads (Amersham Biosciences) at room temperature for 20 min. We washed the beads three times with lysis buffer. We incubated the beads with 1 or 3 g of bovine brain purified calcineurin (Upstate) for 1 h at 4°C. We then washed the beads five times with lysis buffer, and we eluted the bound protein by boiling samples for 10 min in the presence of electrophoresis loading buffer. We resolved the samples by SDS-PAGE, transferred the proteins to nitrocellulose membrane, and detected calcineurin by Western blotting using an anti-calcineurin pan A antibody (1:1000, Chemicon International). We revealed the blot using a horseradish peroxidase-linked rabbit antiserum and an ECL kit (PerkinElmer Life Sciences).
Cell Proliferation Assay-To stably express peptide aptamers in mammalian cells, we used the episomal eukaryotic expression vector pCEP4 that bears a cytomegalovirus promoter and a hygromycin selection marker (Invitrogen). We PCR-amplified aptamer coding sequences using the oligonucleotides 5Ј-GCAAGCTAGCATGTACCCT-TATGATGTGCCA-3Ј that hybridized to the HA coding sequence and 5Ј-CGTTGCGGCCGCTTAGACTAATTCATTAATGGT-3Ј that contained a stop codon. We digested the PCR products with NheI and NotI and ligated them into NheI/NotI-cut pCEP4 to create pEAaptamer plasmids. We plated 3 ϫ 10 5 cells/well in 6-well plates and transfected the cells 24 h later using jetPEI (Qbiogen), 3.7 g of pEA-aptamer plasmids, and 0.3 g of pEGFP-C1 (Clontech) to monitor transfection. We added hygromycin (Invitrogen) at 200 g/ml 2 days later, and we cultured the cells for 2 weeks, renewing the medium twice a week. We then rinsed the cells in PBS, and we fixed and stained them by incubating for 30 min in crystal violet (0.05% crystal violet, 20% ethanol, 0.37% formaldehyde). We removed excess crystal violet by washing with water.
In Vitro Phosphatase Assay-We first produced GST-aptamer fusion proteins as described above. For this experiment, we eluted GST-aptamer fusion proteins from glutathione-Sepharose beads using 20 mM reduced L-glutathione (Sigma), and we dialyzed the eluates overnight against a phosphatase buffer (50 mM Tris-HCl, pH 7.4, 0.1 mM CaCl 2 ). We measured the phosphatase activity of calcineurin using para-nitrophenyl phosphate (pNPP) (Sigma) as substrate in a final volume of 100 l. The sample solution contained 50 mM Tris-HCl (pH 7.4), 0.1 mM CaCl 2 , 1 mM NiSO 4 , 0.15 mg/ml BSA (Sigma), 0.1 M calcineurin (Upstate). We added purified calmodulin (Upstate) and GST-aptamer fusion proteins at different concentrations (see figure legends). After a 15-min preincubation at 37°C, we started the reactions by adding 4.1 mM pNPP, and we incubated the mixtures at 37°C for 20 min. We measured the nitrophenylate product at 405 nm using an Envision plate reader (PerkinElmer Life Sciences). We subtracted the background level that we determined using a reaction mixture lacking calcineurin.
Monitoring of BAD Phosphorylation-We cloned the peptide aptamer coding genes into pPEAt (a pCEP4-based vector that bears a tetracycline-inducible promoter and a hygromycin resistance gene) as described above (see "Cell Proliferation Assay").
We seeded 4 ϫ 10 5 HeLa-Tet cells/well in a 6-well plate 24 h before transfection. We transfected with jetPEI (Qbiogen), 1 g of pEBG-mBad (a plasmid directing the expression of the murine BAD protein; Cell Signaling Technology), 0.5 g of pCMV-SPORT6-CnA␤ and pCMV-SPORT6-CnB (plasmids directing the expression of human calcineurin A␤ and B; RZPD), and 1 g of pPEAt-R5G42, -R5G52, or -R7G44. After an overnight incubation of the transfection mixture, we washed the cells once with culture medium, and we added fresh medium with or without 0.5 M FK506 (Calbiochem). We collected the cells 24 h later, washed them twice in PBS, and lysed them for 20 min in ice-cold lysis buffer (20 mM Tris, pH 7.4, 150 mM NaCl, 2 mM EDTA, 1% Nonidet P-40, protease inhibitor mixture Complete EDTA-free (Roche Applied Science)). We centrifuged the lysates to remove cellular debris, and we quantified the protein content using the mi-croBCA protein assay kit (Pierce). We resolved 50 g of the lysates by 4 -12% SDS-PAGE, transferred the proteins to nitrocellulose membranes, and blotted with anti-phospho-BAD (Ser-112), anti-phospho-BAD (Ser-136), and anti-BAD antibodies (Cell Signaling Technology). We revealed the blots using the ECL system (PerkinElmer Life Sciences).

RESULTS
Peptide Aptamer Libraries and Screening Strategy-To construct our peptide aptamer libraries, we used an SIVderived lentiviral expression vector directing the constitutive expression of bicistrons (transgenes and a GFP marker) under the control of an EF1␣ promoter (see "Experimental Procedures"). We first built 12 low complexity peptide aptamer libraries, combining two scaffolds (human thioredoxin or an E. coli thioredoxin, whose coding sequence harbors codons optimized for expression in mammalian cells), two epitope tags (HA or His 6 ), and three variable region lengths (16, 10, or 7 amino acids). We performed pilot experiments to determine which library yielded the highest expression level of peptide aptamers upon transduction of XC cells with viral particles. We observed that the best combination was the HA-tagged, human thioredoxin displaying random peptide loops of 10 amino acids (data not shown), and we constructed accordingly pBK1, a high complexity peptide aptamer library (Fig. 1a).
To isolate library members that inhibit tumor cell proliferation, we made use of the fluorescent vital dye CMTMR, which cells incorporate and dilute as they proceed through division cycles. Those cells that do not divide maintain a high fluorescence level and can thus be sorted by flow cytometry. Because of a significant background of cells that do not grow or proliferate more slowly independent of the expression of peptide aptamers, multiple screening rounds were necessary to isolate peptide aptamers that exert an antiproliferative effect. We thus constructed a peptide aptamer sublibrary from the highest percentile of CMTMR-positive cells obtained after each screening iteration, and we submitted each sublibrary to a subsequent screening round (Fig. 1b).
Isolation of Antiproliferative Peptide Aptamers-We used rat XC cells, derived from a Rous sarcoma virus-induced sarcoma, enabling us to use viral particles harboring a murine ecotropic envelope. We performed seven screening iterations before isolating and characterizing individual peptide aptamers. We determined the antiproliferative activity of the subli-braries both in XC cells and in human HeLa cells. As shown in Fig. 2a, the mean fluorescence intensity of both cell lines increases gradually with the number of screening iterations, thereby indicating a progressive enrichment of antiproliferative peptide aptamers within the sublibraries.
We picked and sequenced 100 clones from the R5 and R7 sublibraries obtained from the fifth and seventh screening iteration, respectively. More than 40% of the peptide aptamers isolated after the seventh screening iteration corresponded to a single library member, named R5G42. The occurrence of this aptamer was already significant after the fifth FIG. 1. Design of the peptide aptamer library and of the antiproliferative screening. a, schematic representation of the pBK1 peptide aptamer library. This SIV-derived expression system directs the expression of two cistrons coding for an EGFPf transduction marker and HA-tagged peptide aptamers consisting of a 10-amino acid variable region inserted within the active site of human thioredoxin. b, workflow of the screening for antiproliferative peptide aptamers. Rat XC cells are transduced with pBK1 and labeled with CMTMR. The highest percentile of fluorescent cells is then isolated by flow cytometry, and the peptide aptamer coding sequences are amplified by PCR from genomic DNA to construct sublibraries. The sublibraries are used in successive iterations of this process (detailed under "Experimental Procedures"). aa, amino acids. EFS, elongation factor-1 short; cPPT, central polypurine tract; RRE, Rev responsive element; SA, splice acceptor.

FIG. 2.
Antiproliferative effect of peptide aptamers. a, progressive enrichment of peptide aptamer sublibraries in antiproliferative peptide aptamers through screening iterations. XC or HeLa cells were transduced with pBK1 or different R"n" sublibraries obtained after n screening iterations and were labeled with CMTMR. The mean fluorescence intensity increases with the number of screening iterations, indicating a progressive enrichment in peptide aptamers exerting an antiproliferative effect. b, colony formation assays. HeLa or MCF-7 cells were transfected with plasmids directing the stable expression of the Cdk inhibitor p21, a library of peptide aptamers from pBK1 (AptaLib), HTRX, and peptide aptamers R7G44, R5G42, and R5G52. The cells were cultured for 2 weeks, and the colonies were stained with crystal violet. iteration but was not detectable after the fourth iteration. Three other peptide aptamers that showed a lower occurrence were also isolated (R7G11, R7G44, and R5G52) ( Table I).
We wished to establish the antiproliferative activity of these peptide aptamers using alternative cellular models and a nonretroviral vector to individually express each aptamer. We cloned the aptamer coding genes into a vector bearing a hygromycin selection marker. We also subcloned the pBK1 library into this vector to create "AptaLib." We continuously expressed the aptamers, the empty thioredoxin scaffold, or AptaLib in HeLa and MCF-7 cells for 2 weeks, and we stained the cells that grew. Aptamers R5G42 and R5G52 significantly inhibited the proliferation of both cell lines as compared with AptaLib and human thioredoxin. Aptamer R7G44, similarly to other aptamers (not shown), did not exert any antiproliferative effect (Fig. 2b). These aptamers may originate from the remaining background of slowly proliferating cells during the seventh screening iteration independently of the expressed aptamers. Surprisingly aptamer R7G11 did not inhibit cell proliferation in this assay (not shown) despite showing a high occurrence in the seventh sublibrary (Table I). This could be due to the fact that the CMTMR assay is more sensitive in detecting modest antiproliferative effects than the colony formation assay or that some peptide aptamers somehow enhance the CMTMR labeling of their host cells. From all these results, we decided to focus on peptide aptamer R5G42 and to identify its target protein.
Identification of Calcineurin A as a Target Protein-We performed two yeast two-hybrid screening experiments against a LexA-R5G42 bait protein using a human testis and a human fetal brain cDNA library. We obtained 29 and 42 reconfirmed clones, respectively. We disregarded those clones that either showed a barely detectable two-hybrid interaction phenotype, cross-interacted with control aptamers, or corresponded to hypothetical proteins (Table II). We thus retained two candidates. The highest occurring clone, from both libraries, corresponded to the NS5A-TP2 protein recently discovered through a systematic search for genes that are transactivated by the non-structural NS5A protein from hepatitis C virus (22). No biological knowledge is currently available for this protein. The other remaining target candidate was calcineurin A (CNA), for which two different isoforms (␤ and ␥) were selected from the testes library (Fig. 3a). We decided to focus our work on CNA.
We first set out to map the R5G42 binding site on CNA. The CNA clones selected in the yeast two-hybrid experiments corresponded to the carboxyl-terminal regions of the ␤ and ␥ isoforms, encompassing the calmodulin-binding domain and the autoinhibitory domain (Fig. 3b). Among the three truncations constructed from the CNA␤ selected clone (Fig. 3b), only CNA␤⌬1 retained its yeast two-hybrid interaction phenotype with R5G42 ( Fig. 3a and data not shown). These results indicate that the R5G42 binding site on CNA lies between the amino terminus of the calmodulin-binding domain and the carboxyl terminus of the autoinhibitory domain and is not circumscribed to the calmodulin (CaM)-binding domain. This yeast two-hybrid mating assay also supports the specificity of interaction between R5G42 and CNA as R5G42 did not show an interaction phenotype with two unrelated bait proteins (RAS and FKBP12) and as R7G44 and R5G52 did not show an interaction phenotype with CNA. R5G52, however, did not show an interaction phenotype with peptide aptamer RG22, which interacts with LexA in the context of most (but not all) LexA fusion proteins. The LexA-R5G52 bait protein may thus not be properly expressed and/or folded in this yeast twohybrid setting.
To confirm the interaction between R5G42 and CNA, we performed an in vitro binding assay between recombinant purified GST-aptamer fusion proteins, coupled to a glutathione-Sepharose matrix, and purified CNA. The GST-R5G42

Results of the yeast two-hybrid screening against R5G42
The table lists the different clones selected from the brain and testes libraries and sequenced. Bold numbers correspond to strong, specific two-hybrid interaction phenotypes. Plain numbers correspond to weak, specific interaction phenotypes. Numbers in italics correspond to non-specific clones, which show two-hybrid interaction phenotypes with other peptide aptamers. PH, pleckstrin homology; SH2, Src homology 2.

Screening of Antiproliferative Peptide Aptamers
solid phase readily captured CNA as opposed to a GST-R7G44 control (Fig. 3c).
Modulation of Calcineurin Activity-We next explored the ability of R5G42 to modulate the enzymatic activity of its target protein. To this end, we first performed an in vitro phosphatase assay using purified CNA and pNPP as a substrate. As shown in Fig. 4a, the addition of purified CaM is required to activate CNA. The addition of recombinant purified GST-R5G42 did not result in an inhibition or an exacerbation of CaM-activated CNA phosphatase activity (not shown). However, the addition of high concentrations of GST-R5G42 activated CNA phosphatase activity in the absence of CaM to a level comparable to that observed using CaM. The addition of equal amounts of the control aptamer R7G44 did not produce a significant effect. This experiment indicates that R5G42, like CaM, binds and activates CNA phosphatase activity in vitro.
We set out to confirm this finding in human cells. BAD is a key proapoptotic protein whose activity is tightly regulated by its phosphorylation status, which is controlled by the balanced activity of several protein kinases and calcineurin. Therefore, the phosphatase activity of calcineurin in cells can be monitored by examining BAD phosphorylation. We transfected HeLa cells with plasmids directing the expression of BAD, CNA␤, calcineurin B, and either R5G42, R5G52, or R7G44. We observed that expression of R5G42 decreased the phosphorylation of BAD on serine 136 without affecting the phosphorylation on serine 112 (Fig. 4b). To demonstrate that this effect was caused by an up-regulation of calcineurin activity, we performed the same experiments in the presence of FK506, a well known inhibitor of calcineurin. The R5G42induced dephosphorylation of BAD on serine 136 was no longer observed in the presence of FK506 (Fig. 4b). DISCUSSION We have built and used a lentiviral peptide aptamer library to isolate aptamers that inhibit cell proliferation in vitro. We have determined the identity of the target proteins of one of the isolated peptide aptamers by performing yeast two-hybrid screening experiments. We have retained NS5A-TP2 and CNA as two strong target candidates. No biological information is currently available for NS5A-TP2 except that its coding gene is transactivated by the non-structural NS5A protein from hepatitis C virus (22). In contrast, calcineurin (of which CNA is the catalytic subunit) is a well studied protein phosphatase that plays a key role in coupling Ca 2ϩ signaling to cellular responses (for a review, see Ref. 23). The demonstration that calcineurin was the target of the immunosuppressants cyclosporin A and FK506 has sparked a considerable interest in this protein and has greatly facilitated the elucidation of its function especially in T cell activation. Despite numerous studies, the role of calcineurin in cell proliferation FIG. 3. Interaction between peptide aptamer R5G42 and calcineurin A. a, yeast two-hybrid mating assay. TB50␣ yeast were cotransformed with pSH18-34T (bearing a lacZ reporter gene) and plasmids directing the expression of LexA alone or in fusion with peptide aptamers R5G42, R7G44, or R5G52. MB210a yeast were transformed with the selected cDNA library plasmids directing the expression of CNA␤, CNA␥, and NS5A-TP2 truncated proteins. To obtain negative controls, MB210a yeast were also transformed with the empty prey plasmid (pJG4-5) and with pJG4-5 directing the expression of RAS and FKBP12 prey proteins. To obtain a positive control, MB210a yeast were transformed with pJG4-5 directing the expression of RG22 peptide aptamer prey fusion protein that interacts with LexA in the context of most LexA fusion proteins. b, schematic representation of the CNA clones selected through the yeast two-hybrid screening and of the truncations performed on CNA␤. c, affinity capture assay. Comparable amounts of GST-R7G44 or GST-R5G42 recombinant fusion proteins were coupled to glutathione-Sepharose beads. Purified calcineurin was added onto the beads, and the captured molecules were revealed by a Western blot experiment using an anti-calcineurin antibody. AAs, amino acids; CNB, calcineurin B; AI, autoinhibitory domain; Cter, carboxyl terminus. remains less clear. Cyclosporin A has been shown to inhibit the proliferation of various cells but at concentrations exceeding that required to observe an inhibition of T cell activation. FK506, although a more potent immunosuppressant than cyclosporin A, shows a weaker antiproliferative activity (for a review, see Ref. 24). These observations suggest that the antiproliferative activity of these immunophilins may be caused by the modulation of other target protein(s). Contrary to the hypothesis that calcineurin positively regulates cell proliferation, calcineurin has been shown to induce apoptosis through different mechanisms including the dephosphorylation of BAD, a proapoptotic Bcl-2 family member (25).
Here we have shown that an antiproliferative peptide aptamer (R5G42) bound CNA and activated its phosphatase activity in vitro. Consistent with the in vitro results, the transient expression of R5G42 in human cells induced the dephosphorylation of BAD on serine 136, which was totally reversed by FK506. The expression of R5G52, another antiproliferative peptide aptamer, did not affect BAD phosphorylation levels. Altogether these results indicate that BAD dephosphorylation is specifically caused by the activation of CNA by R5G42 as opposed to being an indirect consequence of an antiproliferative activity. The antiproliferative effect of R5G42 could stem from a calcineurin-mediated induction of apoptosis, which would only occur upon prolonged expression of the peptide aptamer. Indeed we have failed to show that transient expression of R5G42 caused conspicuous signs of apoptosis or cell cycle arrest (not shown). Importantly we cannot rule out the possibility that the targeting of NS5A-TP2 (or another unidentified target protein) also contributes to the antiproliferative effect of R5G42. NS5A-TP2 contains a conserved HD domain found in many phosphatases. The use of the R5G42 peptide aptamer may help elucidate the function of this protein, which may play a role in the control of cell proliferation.
The structural mechanisms of the activation and the inhibition of calcineurin by, respectively, calmodulin and immunophilin-immunosuppressant complexes remain poorly understood (26). Here we describe a new CNA ligand that activates CNA phosphatase activity through a potentially original mechanism because its binding site is located between the CaM-binding domain and the autoinhibitory domain but does not appear to be circumscribed to the CaM-binding domain. Structural studies of the CNA-R5G42 complex will be needed to elucidate the activation mechanism, which may help clarify the still elusive physiological activation and inhibition mechanisms of calcineurin.
This study describes the first phenotypic selection of peptide aptamers in mammalian cells. It also describes the first identification of a functional perturbation of a protein targeted by combinatorial protein molecules isolated from an antiproliferative screening. A number of arguments strongly support the choice of peptide aptamers to perform various phenotypic screening or selections with the goal of interrogating proteomes to identify target proteins involved in the underlying regulatory networks. First, proof of concept has been obtained in yeast where peptide aptamers were selected for their ability to overcome the cell cycle arrest induced by a mating pheromone and where target proteins were identified by yeast two-hybrid screening (16,17). Second, peptide aptamers can target many different kinds of intracellular proteins such as kinases, phosphatases, receptors, adaptor proteins, transcription factors, chaperones, etc. involved in many regulatory pathways (for a review, see Ref. been shown to decorate their target proteins by binding to many different surfaces involved in different functions (27). For this reason, peptide aptamers can induce a broader range of perturbations on protein function than other reverse genetics methods such as gene knock-out or the use of transdominant negative alleles. 2 Last, the double constraint imposed on the variable regions reduces the conformational freedom and yields typically high binding affinities for the target proteins, thereby facilitating their identification by different methods.
Our work illustrates the particularities of using combinatorial protein molecules for phenotypic screening of transdominant reagents. The main limitation lies in the tedious target identification step. When using nucleic acid molecules (cDNA fragments, antisense molecules, or short hairpin RNAs), the identity of the target proteins is immediately unveiled by sequencing the isolated library members. In contrast, selected combinatorial protein molecules must be used as probes to determine the identity of their targets by performing yeast two-hybrid cDNA screening (10,16,17) or affinity capture experiments followed by mass spectrometry (28). However, combinatorial protein molecules, and particularly peptide aptamers, present a considerable advantage above nucleic acid molecules. Whereas the latter can only inhibit the function of their target proteins (by a dominant negative effect or by reducing expression levels), the former can cause more diverse perturbations on the function of their targets, including an activation as observed in this study. Therefore, the use of combinatorial protein molecules for phenotypic screening or selections allows a more extensive probing of proteomes, thus enhancing the chances to identify different target proteins whose perturbations cause a given phenotype. Another significant advantage of using peptide aptamers lies in their application for drug discovery. Once their target proteins are identified, peptide aptamers can guide the identification of small molecule mimics that bind the same molecular surfaces on the targets and induce the same biological effects (27). We anticipate that the use of retroviral libraries of peptide aptamers for phenotypic screening or selections will aid the unraveling of molecular regulatory networks that control major biological processes and will impact positively therapeutic research by facilitating the discovery of new targets and small molecule drugs.