A CDC6 Protein-binding Peptide Selected Using a Bacterial Two-hybrid-like System Is a Cell Cycle Inhibitor*

Peptides or small molecules able to modulate protein-protein interactions hold promise as tools with which to probe and manipulate biological pathways. An important issue in this nascent field is to evaluate different methods with which to search libraries for molecules that modulate the function of specific target proteins. One strategy is to screen libraries for molecules that bind specifically to a protein known to be critical in the pathway of interest, with the expectation that the molecules isolated will recognize regions of the target protein important for its function and thereby exhibit biological activity. Here, a peptide library was screened using a two-hybrid-like system for molecules able to bind human CDC6 protein (CDC6p), required for the initiation of DNA replication in eukaryotic cells. From a collection of over a million peptides, a single species that exhibited good affinity and specificity for binding CDC6p was obtained. When expressed in human cells, the peptide inhibited cell cycle progression and exhibited other properties expected of a CDC6p inhibitor. This approach, which does not require detailed knowledge of the mechanism of action of a protein target, may be generally useful for isolating peptides capable of manipulating biological pathways.

Pharmacologically active compounds are essential tools for probing the cell and molecular biology of organisms that are not amenable to genetic manipulation, including humans. The availability of such reagents is limited by the number of natural products with specific, well characterized activities. In an effort to circumvent this limitation, many laboratories have turned to combinatorial libraries and high throughput screens to identify compounds with which to manipulate biological pathways. Critical issues in such studies include the type of library employed (peptides, synthetic small molecules, etc.); the diversity of the library; and the sensitivity, specificity, and throughput capacity of the screen used to identify molecules with the desired properties.
Functional screens, in which one looks for a phenotypic effect of library-derived compounds, have the advantage that one does not need to define a specific molecular target. This strategy has been applied both to synthetic, encoded combinatorial libraries (1)(2)(3) and to peptide libraries made genetically (4,5). On the other hand, rapid progress in fundamental studies of cellular signaling and regulation has provided many interesting protein targets, the functional modification of which by small molecules would have biological or even clinical importance. In particular, as the critical role of protein-protein interactions in almost any pathway of interest has become apparent, the development of methods to identify compounds able to disrupt or promote these interactions has become the focus of greater interest. However, with some notable exceptions (6), small synthetic molecules have not proven useful for this purpose. The small molecules used in most approaches to this goal are related to enzyme inhibitors, which bind in small, recessed pockets of a protein and are poorly suited to recognize more expansive, relatively flat surfaces typical of protein interaction regions. Newer libraries of macrocyclic, natural product-like species may enhance the probability of success (3,(7)(8)(9). Peptides are well suited for this role because they can easily adopt extended conformations or simple secondary structures. Although peptides are not favored by the pharmaceutical industry because of their nonoptimal pharmacological properties, they hold considerable promise as research tools with which to manipulate protein-protein interactions and therefore biological pathways. They can be expressed at high levels in cultured cells in the form of fusion proteins or introduced into cells by conjugation to "delivery peptides" able to cross cell membranes (10 -13). Finally, powerful tools have been developed by which peptide libraries can be screened for binding a particular target molecule. They include phage display (14,15), the two-hybrid system (16,17), and several related approaches (18,19). A potential drawback of using binding assays to screen peptide libraries is that there is no guarantee that the peptides isolated will bind a functionally important surface of the target and affect its activity. Fortunately, sites of macromolecular interactions often appear to be favored locations for binding, possibly because they are open and available within the overall topology of native proteins (20).
Here, we report the application of a high throughput bacterial screen to the discovery of a peptide inhibitor of cell cycle progression in human cells. Human CDC6 protein (CDC6p) 1 (21) is a key component in assembly of the replication complex prior to the initiation of DNA replication (22). CDC6p is re-cruited by the origin recognition complex (ORC), which in turn promotes loading of the MCM proteins on chromatin (23). Immunodepletion of CDC6p in living cells arrests the cell cycle in G 1 , indicating the essential role of CDC6p in human cell proliferation (24). These results suggest that compounds able to block the interaction of CDC6p with other replication proteins should be efficient cell cycle inhibitors. We report here the isolation of a CDC6p-binding peptide using a bacterial twohybrid-like system. This peptide binds specifically to CDC6p with a K D of approximately 10 Ϫ7 M and is capable of reducing CDC6p loading on chromatin in G 1 phase and blocking entry of human cells into S phase.
These results are of general interest because no particular activity of CDC6p was targeted, nor did the assay specifically seek to disrupt a particular protein-protein interaction. Indeed, the detailed mechanism of action of CDC6p is not yet understood to a point that would permit the rational design of functional screens for a CDC6p-targeted inhibitor. These findings suggest that useful reagents for functional manipulation of important pathways of cell regulation can be obtained from simple binding studies, even in the absence of detailed knowledge of the mechanisms of action of the target protein.

EXPERIMENTAL PROCEDURES
Library Construction-In order to detect the peptide encoded by genomic DNA fragment in later experiments, a T7 S10 epitope tag was introduced downstream of the cI repressor DNA binding domain (DBD) in vector pJH391 (25) by a pair of oligonucleotides as a HindIII/BamHI fragment. The resulting plasmid was named pJH391S. Chicken genomic DNA (CLONTECH) was completely digested with Sau3AI and cloned en masse into BamHI-digested and dephosphorylated pJH391S. The ligation products were transformed into Escherichia coli DH12S supercompetent cells (Life Technologies, Inc.), and cells were grown on one hundred 150-mm Luria broth plates supplemented with 50 g/ml carbenicillin at 37°C overnight. Analysis of a large number of colonies picked from these plates indicated that there are approximately 5-6 ϫ 10 6 peptide-expressing plasmids in the library.
Bacterial Two-hybrid Screening-The full-length CDC6p gene was amplified by polymerase chain reaction and cloned as a SalI/BamHI fragment into pJH391. A 2.4-kilobase EcoRI and EcoRV fragment containing the cI DBD and CDC6p was released from the resulting plasmid and cloned into EcoRI-and PvuII-digested pACYC184 containing a tetracycline resistance gene. The final construct, pCDC6, was transformed into kanamycin-resistant E. coli JH372(Kn r ) for later experiments. The purified library plasmids were transformed into strain JH372 (25) containing pCDC6 via electroporation. After a 30-min recovery, the cells were harvested by low speed centrifugation and resuspended in NZC broth supplemented with carbenicillin (50 g/ml), kanamycin (10 g/ml), and tetracycline (20 g/ml). After the cells were grew at 37°C for 60 min, phage K54H80 and K54, strains deleted for wild type cI (25), were added to the culture at a multiplicity of infection of 50. The cells were shaken for another 3 h at 37°C and then collected by centrifugation, resuspended, and grown as single colonies on Luria broth plates supplemented with the antibiotics carbenicillin, kanamycin and tetracycline. The surviving colonies were picked for further testing.
Validation of the "Hits"-Plasmid DNA was prepared from each colony and retransformed into JH372. Six colonies resistant to Cb/Kn or Cb/Kn/Tet were selected from each transformation and assayed for phage resistance. In brief, bacteria were grown in 150 l of NZC broth in a microtiter plate well at 37°C overnight. 20 l of bacterial culture was mixed with 20 l of phage solution (5 ϫ 10 8 phage) and held at room temperature for 30 min. 60 l of melted top agar (55°C) was added to the mixture and immediately plated on a Luria broth plate supplemented with Kn and Cb or with Kn, Cb, and Tet. Colonies that contained only the library plasmid but that were resistant to phage infection were considered to express self-associating peptides. In cases in which phage resistance was dependent on both the library-encoding and the target-encoding plasmids, a ␤-galactosidase assay was performed using the Repressor-regulated, integrated lacZ reporter gene in JH372 to confirm Repressor reconstitution. ␤-Galactosidase activity was measured as units/min normalized to cell density (A 600 ).
In Vitro Binding Assays-The genomic DNA fragments in candidate clones were amplified by polymerase chain reaction and cloned into the NcoI and EcoRI sites of pGEX-Cs. The resulting recombinant plasmids were transformed into E. coli BL21. Production of fusion protein was induced by adding IPTG (0.2 mM) at A 600 0.4 followed by shaking at 37°C for 3 h. The cells were harvested after centrifugation, resuspended in 1ϫ PBS buffer (138 mM NaCl, 2 mM KCl, 10 mM Na 2 PO 4 , and 2 mM K 2 PO 4 ) containing protease inhibitor mixture (Roche Molecular Biochemicals) and lysed by sonication. Fusion protein was purified over glutathione-Sepharose 4B beads according to the manufacturer's instructions (Amersham Pharmacia Biotech).
Overexpression of CDC6p in HeLa cells was achieved by transient transfection of HeLa cells with pGCN.SCH42FL in which HA-tagged human CDC6p was under the control of the CMV promoter. 24 h posttransfection, cells were washed two times by ice-cold PBS buffer. 0.5 ml of lysis buffer (20 mM Tris⅐Cl, pH 7.5, 0.2% Nonidet P-40, 100 mM NaCl, 20% glycerol, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, plus protease inhibitor mixture) was added to each 100 mm plate. The cells were scraped off the plates and left on ice for 30 min to allow for complete lysis. The supernatant was collected after centrifugation at 4°C for 10 min and used for the in vitro binding assay.
Equal volumes of cell lysate containing human CDC6p was added to glutathione-Sepharose 4B beads bearing equal amounts of the various GST-peptide fusion proteins. The mixtures were tumbled at 4°C for 2 h and beads were washed with ice-cold 1ϫ PBS twice, 10 min tumbling at 4°C for each wash. The proteins retained on the beads were denatured and subjected to SDS-PAGE and Western blotting. Mouse monoclonal antibody against the HA epitope was used for detection of HA-CDC6p protein, mouse monoclonal against GST was used for blotting GST fusion protein. Goat anti-mouse IgG horseradish peroxidase conjugate was used as the secondary antibody. The bands were visualized by SuperSignal West Pico chemiluminescent solution (Pierce).
Scrambled Mutants of CDC6-BP35-Scrambled mutants of CDC6-BP35 have identical amino acids composition but different sequence. To achieved this, five pairs of oligonucleotides encoding the new peptide were synthesized for each mutant. These oligos were annealed, ligated, and cloned into NcoI/EcoRI-digested pGEX-CS. The constructs were confirmed by DNA sequencing. The peptide sequences of the scrambled mutants were SDKFYFCVKRSIKYFITHCKIKQGKLKIIKSCKLK for CDC6-BP35S1 and KSLCKSKIKGYFKDFVKYQIKICTKIKSIFKC-RHL for CDC6-BP35S2.
Estimation of Dissociation Constants-The experiment was performed essentially as described by Zhang et al. (26). Different amounts of the GST-CDC6-BP35 fusion protein were added to the same amount of extract made from HeLa cells transfected with the CDC6p expression plasmid to determine the lowest amount of beads that could be employed. Then, this fixed amount of bead-bound fusion protein (5 ng) was mixed with HeLa cell extract containing different amounts of CDC6 protein (5,10,20,30,50, and 60 ng); the final volume was 100 l in 1ϫ lysis buffer. The mixtures were incubated at 4°C for 2 h and washed with ice-cold PBS twice. The amount of CDC6p retained was determined by SDS-PAGE/Western blot analysis of the bead-bound material. The amount of HA-tagged CDC6p in the HeLa cell lysate was estimated by quantitative Western blot using purified HA-PLP as standard (provided by L. Sun, University of Texas Southwestern Medical Center, Dallas, TX). The data were plotted as described by Freifelder (27), using the equation The slope of the line obtained in a double r plot provides the K D .
FACS Analysis-HeLa cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum at 37°C, 5% CO 2 incubator and transfected with either pEGFP, pEGFP-CDC6-BP35, or pEGFP-C2 using LipofectAMINE PLUS TM (Life Technologies, Inc.). Cells were trypsinized from plates 24 h later, washed with ice-cold PBS, and fixed with 0.4% paraformaldehyde at 4°C for 10 min. After being washed with ice-cold PBS, cells were permeabilized with 0.1% Triton X-100 at 4°C for another 10 min. Finally, cells were washed with ice-cold PBS twice and treated with 20 g/ml propidium iodide and 2 mg/ml RNase for more than 30 min before the FACS assay. CellQuest was used for the FACS data acquisition and analyses.
Cell Synchronization-HeLa cells were transfected with pEGFP or pEGFP-CDC6-BP35. 6 h after transfection, cells were cultured in growth medium containing thymidine (2 mM) for 18 h or aphidicolin (5 g/ml) for 24 h. HeLa cells were released from the block at G 1 /S boundary by washing twice in growth medium. At different time points after release, cells were harvested and processed for FACS assay as above.
Chromatin Isolation-HeLa cells were transfected with pEGFP, pEGFP-C2, or pEGFP-CDC6-BP35. 48 h after transfection, cells were stained with Hoechst (10 g/ml) at 37°C for 10 min and harvested for sorting. One million green fluorescent cells in G 1 phase expressing either GFP or GFP fusion protein were sorted. A chromatin-containing fraction was isolated from each group of sorted cells. In brief, cells were suspended in Buffer A (10 mM Hepes, pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 , 0.34 M sucrose, 10% glycerol, 1 mM dithiothreitol, and protease inhibitor). Triton X-100 was added to the mixture to a final concentration of 0.1%. After 5 min of gentle lysis on ice, nuclei were collected after spinning and washed with Buffer A once. The nuclei were lysed with Buffer B (3 mM EDTA, 0.2 mM EGTA, 1 mM dithiothreitol, and protease inhibitor), and the chromatin-containing fraction was collected after spinning and washed with Buffer B once. The same amount of each sample (equivalent to 0.3 million cells) was used for SDS-PAGE analysis. Rabbit monoclonal antibodies against ORC2 and MCM3 and mouse monoclonal antibody against CDC6p were used in immunoblots.

RESULTS
Isolation of a CDC6p-binding Peptide Using a Bacterial Twohybrid-like System-To isolate a CDC6p-binding peptide from a library, we employed a genetic assay based on reconstitution of the activity of a Repressor fragment that includes the entire DBD but lacks the C-terminal dimerization domain (19,25,29). Two Repressor DBD fusion proteins are introduced into bacteria, one fused to the target (in this case CDC6p) and the other fused to a peptide library. Binding of the library-encoded peptide to the target protein reconstitutes Repressor dimerization and high affinity binding to operator sequences (Fig. 1), thereby rendering cells that contain the appropriate peptide immune to infection (26, 30 -32).
A library of approximately five to six million peptides constructed from chicken genomic DNA fragments (26) was screened against full-length CDC6p fused to the Repressor DBD. 72 colonies survived the phage challenge. The peptide-encoding plasmids from these colonies were isolated and subjected to further analysis. A major source of false positives in this assay is from self-associating peptides that support homooligomerization of the Repressor DBD and result in phage immunity even in the absence of the target protein (33). Most of the hits were found to be self-associating peptides and were discarded. A few others proved to be false positives of unknown mechanism. In other words, retransformation of fresh cells with the peptide-encoding plasmid in the presence or absence of target-encoding plasmid failed to produce phage resistance. These were also discarded. Seven clones appeared to contain true CDC6p-binding peptides, as judged by high resistance to phage and low expression of an integrated lacZ reporter gene containing a single high affinity operator site in the promoter only when both the peptide-and target-encoding plasmids were present ( Fig. 2A).
In order to test the specificity of this interaction, the plasmids encoding the putative CDC6p-binding peptides were transformed into cells containing a plasmid encoding the Repressor DBD fused to a protein or protein fragment unrelated to CDC6p. These controls included epitopes from the human pro-insulin-like growth factor I (26), pro-IL-1␤ and full-length FKBP-12 (34). The output of the Repressor-controlled lacZ reporter gene served to monitor the level of interaction of the peptides with these alternative targets in vivo. Only one of the seven peptides, designated CDC6-BP35, was found to be specific for CDC6p by these criteria (Fig. 2B). Phage resistance assays were also performed (data not shown) and were consistent with the reporter gene results. DNA sequencing of the peptide-encoding region of the BP35-encoding plasmid revealed that the sequence of the single specific hit obtained was FSI-VFISFTCQKKKKKKKKKCYLIKCRLYSGDIHI. Because the linker regions of the peptide-encoding and target-encoding plasmids are not identical, the "nonspecific" binding of some of the peptides tested could reflect associations with an epitope between the DBD and the target protein. For example, BP50 and BP36 support repression with all three control targets (Fig.  2B), consistent with the idea that they recognize some conserved feature of all three of these constructs. However, BP48, BP98, BP38, and BP15 support repression in the presence of some, but not all, targets (Fig. 2B). Although we do not understand the molecular basis of these results completely, one possibility is that they do recognize an epitope in the linker of the target fusion protein, but depending on the nature of the Cterminal fragment, this epitope may or may not be accessible. In any case, all further efforts were focused on CDC6-BP35.
CDC6-BP35 Binds CDC6p Protein in Vitro-To confirm the interaction of peptide CDC6-BP35 and CDC6p, further experiments were performed in vitro using a GST fusion protein (GST-CDC6-BP35) and a crude extract prepared from HeLa cells 24 h after transfection with an expression vector encoding CDC6p tagged with HA. As shown in Fig. 3, GST-CDC6-BP35 retained approximately 10% of the HA-CDC6p present in the extract, as determined by Western blotting using a highly specific monoclonal antibody raised against the HA epitope. When the same experiment was repeated using control GST fusion proteins containing peptides selected at random out of the library (C2 and C54), no binding of CDC6p was observed (Fig. 3).
To test whether CDC6-BP35 binding to CDC6p is specific, three other mammalian proteins, MNF␣, PR48, and Dral, were overexpressed in HeLa cells, and lysates containing these proteins were prepared 24 h after transfection. Similar in vitro binding assays showed that the GST-CDC6-BP35 fusion protein retained CDC6p but not MNF␣, PR48, or Dral (Fig. 4), demonstrating specificity in the interaction between CDC6- As another test of binding specificity, a "scrambled" derivative of CDC6-BP35 was created, CDC6-BP35S1, in which the identical amino acid composition was retained, but the order of residues was altered randomly. This control was deemed particularly important because CDC6-BP35 is a highly basic peptide and relatively nonspecific binding to CDC6p through ionic contacts was a concern. As shown in Fig. 5, GST-CDC6-BP35S1 did not bind CDC6p in a manner detectable in this assay, even at concentrations 20-fold higher than that required for binding of CDC6p to GST-CDC6-BP35. An identical result was obtained with the another scrambled peptide CDC6-BP35S2 (data not shown). We conclude that binding of CDC6-BP35 to CDC6p is dependent on the primary sequence of the peptide and not solely on the charged nature of peptide.
The CDC6-BP35⅐CDC6p interaction was examined in somewhat more detail. The complex is stable over a 20-min period in PBS containing up to 350 mM sodium chloride (data not shown). This also argues against an entirely ionic interaction. The equilibrium dissociation constant (K D ) of the GST-CDC6-BP35/ CDC6p complex under the test conditions was approximately 10 Ϫ7 M, as determined by the titration experiment (27) shown in Fig. 6.
CDC6-BP35 Blocks the Cell Cycle When Overexpressed in HeLa Cells-One of major aims of this study was to determine whether a peptide selected simply on the basis of its ability to bind CDC6p could perturb the cell cycle. Because CDC6p is known to be critical for the initiation of DNA replication, a functional effect of the peptide most likely would be observed at the G 1 /S boundary. To test this possibility, an expression plasmid was constructed that encoded a fusion of the CDC6-BP35 to the C terminus of enhanced green fluorescent protein (EGFP) under control of the CMV promoter. Human HeLa cells were harvested 24 h after transfection with this construct, or control DNAs, and incubated with propidium iodide. The cell population was subjected to FACS analysis in order to segregate cells expressing the fusion protein, and DNA content was measured in this subpopulation of cells. Within an asynchronously growing cell population, 2n, 4n, or intermediate levels of DNA identify cells in G 1 , G 2 /M, or S phases of the cell cycle, respectively. Conditions that block entry into S phase increase the fraction of cells with a 2n content of DNA, with corresponding declines in the fraction of cells with a DNA content greater than 2n.
Cells transfected with plasmids that express EGFP alone or EGFP fused to a control peptide that does not interact with CDC6p (C2, see Fig. 3) exhibited a population distribution of about 61-64 Ϯ 4 -5% in G 1 , 16 Ϯ 3% in G 2 , and 18 -19 Ϯ 3-4% in G 2 /M 24 h after transfection. These values that are typical for asynchronously growing HeLa cells under standard culture conditions. In contrast, cells transfected with a plasmid expressing EGFP fused to CDC6-BP35 exhibited a markedly dif-

FIG. 2. Reporter gene-based tests for peptide-CDC6p binding in bacteria.
A, analysis of target plasmid dependence of repression. Library plasmids (see Fig. 1A) from the survivors of the phage challenge assay were isolated and transformed into fresh JH372 cells, which harbor an integrated Repressor-controlled lacZ gene, that either did or did not also harbor the target plasmid. The ␤-galactosidase activity of the transformants was measured. pCDC6 is the target plasmid; pJH391S is the vector used for peptide library construction; pJH370 encodes a DBD-GCN4 leucine zipper fusion protein and serves as a positive control; pJH379 encodes the DBD alone and serves as a negative control. Peptides that repressed transcription even in the absence of the target plasmid were scored as self-associating peptides and discarded. Peptides that did not repress transcription whether pCDC6 was present or absent were also discarded. B, test of the specificity of the peptide for CDC6p. Each of the seven peptides that passed the "homodimer" test in A was tested against DBD-targets other than CDC6p using the reporter gene assay. Only one, CDC6-BP35, appeared to be selective for CDC6p. Each assay was repeated three times. Error bars represent S.D. All results were normalized to the level of lacZ expression observed when only the Repressor DBD was expressed (pJH379).  2, 4, and 5). Lane 1 is the input control of CDC6p (40 ng). Mouse monoclonal anti-HA serum was used to detect HA-CDC6p. Goat anti-mouse IgG horseradish peroxidase conjugate was used as secondary antibody. The sequences of CDC6-BP35 and the C2 and C54 control peptides deduced from DNA sequencing are shown. C2 and C54 are peptides in the library that did not interact with CDC6p in the genetic screen. ferent ratio. About 83 Ϯ 5% of the cells were in G 1 , 5 Ϯ 2% in S 1 , and 7 Ϯ 2% in G 2 /M (Fig. 7A). Similar differences were also observed 30 and 36 h after transfection (data not shown). Expression of the CDC6-BP35 GFP fusion seemed lower than expression of the two control proteins as visualized by fluorescence microscopy (Fig. 7B). Furthermore, more pEGFP-CDC6-BP35-transfected cells were dead 24 h after transfection than was the case in the two control populations.
To examine the effect of the CDC6-binding peptide on cell cycle progression more carefully, cells transfected with pEGFP or pEGFP-CDC6-BP35 were synchronized with either aphidicolin or thymidine 6 h after transfection. The block was later released, and the procession of the two cell populations through the cell cycle was followed by FACS as a function of time. As shown in Fig. 8, the results in the control population and the cells transfected with the CDC6p-binding peptide were dramatically different. The fraction of the control population in S phase increased greatly after 3 h, as expected, and after 6 h, approximately 55% of the cells were in G 1 , 17% were in S 1 , and 25% were in G 2 /M. In contrast, the population transfected with pEGFP-CDC6-BP35 maintained a constant ratio of cells in G 1 (79 -82%), S 1 (7-11%), and G 2 /M (7-10%) throughout the time course. This result is consistent with these cells being prevented from continuing through the cell cycle after release of the block. In other words, most of these cells appear to be unable to initiate DNA replication.

CDC6-BP35 Inhibits CDC6p Binding to Chromatin in G 1
Cells-Although the mechanism of action of human CDC6 protein in cell cycle progression is not understood in detail, studies in yeast cells indicate that the CDC6 protein binds to the ORC and then recruits MCM proteins to the prereplication complex in late G 1 phase (23). Given the propensity of library-derived molecules to recognize native interaction sites in intact target proteins (see Introduction), we hypothesized that overexpression of CDC6-BP35 might disrupt the interaction of CDC6p with other proteins, such as the ORC or MCM, factors critical for initiation of DNA replication. Therefore, HeLa cells were transfected with pEGFP, pEGFP-C2, or pEGFP-CDC6-BP35. One million fluorescent cells in G 1 phase were sorted. A chromatin-containing fraction was isolated biochemically from each collected sample. The chromatin-associated proteins were separated by SDS-PAGE, and the levels of ORC2, MCM3, and CDC6p in this fraction were assessed by Western blotting with the appropriate antibodies. As shown in Fig. 9, the amount of ORC2 protein associated with chromatin was indistinguishable in all three samples. However, the amount of CDC6p and MCM3 was reduced to about 50% in cells transfected with pEGFP-CDC6-BP35 relative to the two controls. This observation is consistent with the idea that CDC6-BP35 inhibits CDC6p function and, specifically, represses its loading onto the chromatin, which would be predicted to block the initiation of DNA replication and cell cycle progression.

DISCUSSION
The major conclusion of this study is that a peptide selected to simply bind to a given target protein can have interesting biological activities consistent with modulating the function of the target protein. A more complicated functional screen was unnecessary to obtain the active compound. A peptide selected in bacteria to bind full-length human CDC6p exhibits the characteristics expected of an inhibitor of cell cycle progression in human cells (Figs. 7 and 8). A specific assay for loading of CDC6p on chromatin reveals that the presence of the CDC6-BP35 peptide reduces CDC6p-chromatin association by about 50% compared with controls ( Fig. 9), so it seems that the peptide interferes with some process important for CDC6p loading onto chromatin. Consistent with this interpretation is the fact that MCM3 association with chromatin is also inhibited by approximately the same amount, and this event is thought to be CDC6p-dependent. Quantitatively, the effect of CDC6-BP35 on CDC6p-chromatin association (Fig. 9) is more modest than the strong block to cell cycle progression (Fig. 8).
Although we do not understand this completely, a number of simple models can be proposed. For example, the peptide may have an effect on CDC6p function over and above the inhibition of chromatin loading that accounts for the strong block to cell cycle progression. Alternatively, because CDC6p must operate at many thousands of replication origins in a cell, it may be that a 2-fold inhibition of chromatin loading is sufficient to provide a strong overall block to the global initiation of DNA replication. Advances in characterizing the specific biochemical functions of CDC6p will be required before such specific hypotheses can be tested rigorously, but the peptide identified in this study may prove useful in this process.
These results are of general interest because the selection was conducted in a mechanistically naïve fashion. Peptides are often employed to block protein-protein interactions or to inhibit enzymatic activities of target proteins for which detailed knowledge of the interacting domains is already available. We did not demand disruption of a specific CDC6p function or require binding to a particular domain known to be critical for protein-protein interactions. Indeed, it would have been difficult to do so with confidence, given the nascent state of work in unraveling the mechanisms of human CDC6p function. Thus, a novel feature of this study is not that a peptide can bind to a protein and inhibit its function; this is well established. Rather, it is that one can carry out a simple, unbiased binding assay using a protein target about which one has little mechanistic knowledge and have a reasonable expectation of identifying peptides able to modulate the function of that protein. Even though only a single specific CDC6p-binding peptide was isolated in the selection reported here, it proved to have the biological activity expected of a CDC6p inhibitor, thereby supporting the notion that functionally important surfaces of proteins are preferential sites for binding in library screening experiments. Of course, it is important to point out that although the biological data are consistent with the peptide blocking CDC6p activity, this cannot be proven without the definitive evidence of interaction between this peptide and CDC6p in vivo.
The Repressor reconstitution assay used here is one of many different kinds of two-hybrid-like systems that can be used for screening peptide libraries genetically. We have adopted the Repressor system for general use, because in re- HeLa cells were transfected with different EGFP fusion protein-expressing plasmids and were processed for FACS analysis 24 h after transfection. CellQuest was used to analyze the data. Cells in the first peak are in G 1 phase, with two copies of each chromosome (2n). Cells in the second peak are in G 2 /M phase, with four copies of each chromosome (4n). Cells with an intermediate DNA content, i.e. between the 2n and 4n peaks, are in S phase. The EGFP fusion protein that includes CDC6-BP35 blocks entry into S phase. In contrast, neither native EGFP nor fusion proteins containing peptide selected randomly from the original library, encoded within pEGFP-C2, exert demonstrable effects on cell cycle progression. Data are presented from four independent experiments and expressed as the mean plus the S.D. B, fluorescence micrographs of HeLa cells transfected with plasmids encoding EGFP or EGFP fusion proteins allow an estimation of fusion protein expression levels. The bar represents 65 m. lated work we are interested in targeting specific epitopes on proteins (26,35). An advantageous feature of this E. coli-based system relative to yeast-or mammalian cell-based assays is the ability to handle larger libraries due to the high transformation efficiency (19). Although in vitro binding systems, such as phage display, can accommodate even larger libraries and permit manipulation of binding conditions to favor high affinity interactions, peptides identified by in vitro screens, in which the target protein is (usually) the only polypeptide present may lack the specificity required of biologically useful peptides. On the other hand, only one specific CDC6p-binding peptide was isolated in this study from a library estimated to include more than 1 million peptides. Thus, it might be that binding peptides will be isolated for some targets but not others and a technique such as phage display will have to be considered.
Another point of potential variation in the use of this general approach to the isolation of biologically active peptides is the context in which the library is displayed. Prior to our work, Brent and co-workers (36) developed an elegant "peptide aptamer" system in which a library of peptides is displayed in the context of a surface loop of thioredoxin. Using the appropriate fusion proteins, the aptamer library is then screened in a yeast two-hybrid experiment for peptides that recognize a target protein of interest. These have generally been enzymes such as protein kinases and the aptamers obtained indeed prove to be potent inhibitors of the target enzymes in their native cellular environment (37)(38)(39). A particularly impressive feature of thioredoxin-based aptamers are their high binding affinities (K D values are often in the range of 10 Ϫ9 M), whereas the CDC6-BP35⅐CDC6p dissociation constant was approximately 10 Ϫ7 M. Tight binding is probably due to the fact that constraining the peptide into the surface loop makes it "cyclic peptide-like" in the sense that it has significantly less conformational freedom than is the case for a linear peptide. Of course, this means that the selected peptides may not work well outside the context of thioredoxin (37) or at least may exhibit much lower affinities, but fortunately, these thioredoxin-based aptamers express well in a variety of cell types. Our method uses unconstrained peptides displayed as a Cterminal extension on a monomeric protein. Thus, peptides selected using this technology should be highly "portable" in terms of retaining function when presented as a free peptide or following fusion to a variety of other proteins, and our experience to date confirms this prediction (26) (this study). 2 Aptamers are likely to be the reagents of choice when high affinities are required and the use of the modified thioredoxin protein does not present a problem. A recent report from Benkovic and co-workers (28) of an intein-based method by which libraries of cyclic peptide can be produced in E. coli is also of interest. If this technology could somehow be married to a suitable twohybrid-like assay, it might combine the best facets of free peptides and thioredoxin-based aptamers. In any case, the 2 W. Zhu, R. S. Williams, and T. Kodadek, unpublished results. . work of Brent and co-workers (36) and now this study indicate that genetic protein binding assays using a suitable target appear to be an excellent strategy for the isolation of biologically active peptides. In the future, aptamers and peptides selected by these means are likely to play an increasingly important role as tools with which to manipulate and study a variety of biological pathways.