INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

In the past 10 years, random peptide libraries, either chemically synthesized or displayed on phage, have been widely used as a source of protein binding molecules and applied to the identification of peptides able to bind specifically to several types of proteins, including antibodies, receptors, cytokines, and chaperones (21). Peptide libraries have also been used to isolate proteases substrates (16, 25). For this application, libraries can only be used in vitro with purified enzymes, and consequently the selection of substrates may be biased by the experimental conditions. This aspect could be particularly relevant for intracellular proteases of eukaryotic cells, whose activity often depends on the particular microenvironment in which each enzyme is located.

The development of genetic libraries in eukaryotic viruses represents a strategy to overcome the limitations of the in vitro approach. Recently, Buchholz and colleagues described the first example of in vivo selection of protease cleavage sites in mammalian cells (4). Using a retroviral display library, these authors identified a number of cleaved sequences with a common motif characteristic for proteases belonging to the family of proprotein convertase. The experimental strategy employed did not allow a specific protease to be chosen, and individual sequences were selected based on their ability to be cleaved by unknown cellular proteases.

We report a novel type of peptide library based on recombinant Sindbis virus (SBV) and its application to the identification of substrates of the serine protease of the hepatitis C virus (HCV) (10). SBV, the prototype of the alphavirus genus, is a positive-stranded RNA virus which propagates lytically in mammalian cells (20). With the advent of readily available infectious cDNA, it became possible to use SBV to express foreign genes in cultured cells and experimental animals (11, 19). The SBV genomic RNA contains two genes: one is located at the 5' end, spans two-thirds of the genome, and codes for a precursor protein which is processed to form the replication machinery; the other occupies the rest of the genome, is translated from the 26S subgenomic RNA, and encodes the precursor of all structural proteins. The 26S region is dispensable for replication and can be manipulated to express foreign genes (11, 19, 20).

HCV is a human pathogen which affects about 2% of the world population and for which no effective therapy is available (10). Its genomic RNA encodes a precursor polyprotein of about 3,000 amino acids which is processed by cellular and viral proteases in at least 10 individual proteins, in the order C, E1, E2, p7, NS2, NS3, NS4A, NS4B, NS5A, and NS5B. The HCV serine protease (NS3-4Ap) is a heterodimeric enzyme comprising the N-terminal domain of the NS3 protein and the central domain of the NS4A protein and cleaves the nonstructural portion of the viral polyprotein at four junctions: NS3/NS4A cleavage is the first event and occurs only intramolecularly (cis), followed by cleavages at the NS5A/NS5B, NS4A/NS4B, and NS4B/NS5A sites, which can also be intermolecular (trans) (1, 5, 9, 26, 28). Being responsible for the release of the components of the virus replication machinery, this enzyme is considered a pivotal player in the virus life cycle and is thus a candidate target for developing HCV therapies (3). This protease has been extensively characterized at both the biochemical and structural level (12, 14, 18, 23, 24) and its specificity has been defined by identification and mutagenesis of its natural cleavage sites (2, 9, 13, 18, 27, 29). Thus, it is an ideal enzyme for validating the use of SBV-based libraries as a tool for identifying protease substrates.

We recently described the generation of stable SBV-HCV chimeric viruses whose propagation depends on the activity of NS3-4Ap (6, 7). In these viruses, the release of the SBV capsid-autoprotease protein (C) from the structural polyprotein precursor was transformed from an event dependent on the autoproteolytic activity of C into an event dependent on the activity of NS3-4Ap (Fig. 1A). NS3-4Ap was fused at the N terminus of the inactivated SBV capsid-autoprotease (C), and specific cleavage sites were engineered between these two proteins (NS4A/C) and between C and the SBV PE2 glycoprotein (C/PE2). These modifications made NS3-4Ap-mediated proteolysis necessary for correct processing of the SBV structural polyprotein and production of viable viral particles (6, 7).

Having made SBV propagation dependent on the proteolytic activity of NS3-4Ap, we implemented a strategy for selection of specific cleavage sites. Random libraries of cleavage sites were generated at either the NS4A/C or the C/PE2 junctions of the Mut5 virus genome (6). Assuming that only clones containing sequences cleaved by NS3-4Ap generate an infectious virus able to propagate lytically, we screened libraries in mammalian cells and selected clones able to form lysis plaques.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Construction of plasmid libraries. Plasmids pSIN.Mut6 (libraries A, B, and C) and pSIN.Mut7 (library D) were used as vectors for construction of libraries. Both plasmids are derivative of pSIN.Mut5 (6) and were modified by the insertion of two unique restriction sites. In pSIN.Mut6 the recognition sequences for Bst1107I and BstEII were added 5' and 3', respectively, of the region coding for the C/PE2 junction. In pSIN.Mut7 the recognition sequences for FseI and XhoI were added 5' and 3', respectively, of the region coding for the NS4A/C junction. In both vectors an irrelevant nucleotide sequence was introduced between the two restriction sites, resulting in a frameshift mutation.

Library screening. BHK cells (American Type Culture Collection) were grown in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum (FCS). Plasmid libraries were linearized with the appropriate restriction enzyme and transcribed in vitro with the SP6 mMessage mMachine kit (Ambion). Transcription mixtures were desalted through G50 spun columns equilibrated in phosphate-buffered saline (PBS) and electroporated in BHK cells as described previously (6). Electroporated cells were mixed with a threefold excess of nonelectroporated cells and plated in 6-cm-diameter tissue culture dishes at a density of 4 × 10⁴ cells per cm². Five hours posttransfection, cells were overlaid with 0.9% low-melting-point agarose in DMEM-5% FCS. After 3 days at 30°C, lysis plaques were visualized either by staining with neutral red or by immunostaining with an anti-SBV E2 antibody. Electroporation efficiency was monitored by immunofluorescence with the same antibody. The frequency of infectious clones was calculated by dividing the number of plaques by the number of transfected cells. Plaques were picked, diluted with 4 ml of DMEM-10% FCS, and used to infect BHK cells (10⁵ per well, plated in six-well tissue culture plates) at 30°C. Media containing progeny viruses were collected 4 to 6 days postinfection and used as inocula for the next passage. For the second passage, duplicate samples of BHK cells, plated as above, were infected with 2 ml of each medium. After 4 days at 30°C cells were processed for RNA extraction or immunoblot analysis.

Sequence analysis of viral RNAs. RNA was extracted from infected cells as described previously (7). The regions of the viral RNAs corresponding to the C/PE2 junctions were retrotranscribed using the antisense oligonucleotide SB13 (5'-CTTCTTTTGCTTCTGCCAG-3') and the Moloney murine leukemia virus reverse transcriptase (Gibco BRL) and amplified by PCR using oligonucleotides SB32 (5'-GCAGGTCGTCCGATCATGGATAACTCC-3') and SB51 (5'-CAATATGGCATTGAGCAGGG-3'). cDNAs corresponding to the NS3/C junctions were synthesized as above except that oligonucleotides SB148 (5'-GTTGAGGTCTAGTTGCCTGTCCAATGACTAGG-3') and HCV147 (5'-GGCATGCATGTCGGCTAGTATGGCTAGAATTAG-3') were used for reverse transcription and PCR. The nucleotide sequence of each cDNA was determined by automated and/or manual sequencing of the purified PCR products. Deduced amino acid sequences were aligned using the Genetics Computer Group software.

Recloning of the nucleotide sequence coding for the C/PE2 junctions of selected viruses. For cloning and analysis of the C/PE2 coding region from selected viruses, viral RNAs were retrotranscribed and amplified by PCR as above. Amplified cDNAs were digested with the Bst1107I and BstEII and ligated with the pSIN.Mut6 vector DNA digested with the same endonucleases. The resulting plasmids were given the pSIN prefix followed by the virus name (pSIN.B1, etc.).

In vitro translation and cleavage of substrate proteins. Template cDNAs for C/PE2 and C proteins were obtained by reverse transcription of viral RNAs with primer SB73 (5'-GCTTGATGGCGGTGCAACCAGT-3') and PCR amplification. cDNAs for C/PE2 proteins (amino acids 106 to 533 of SBV structural polyprotein) were amplified with oligonucleotides SB44 (5'-GTCGCCGCACTTGCACTC-3') and SB63 (5'-ACCTAATACGACTCACTATAGGAAGAAGCAACCTGCAAAAC-3', containing the promoter sequence for the T7 RNA polymerase [underlined]). cDNAs for C proteins (amino acids 106 to 264 of SBV structural polyprotein) were amplified with oligonucleotides SB63 and HCV125 (5'-GGTGACCAGATATGACATTTAYCA-3', containing an antistop codon adjacent to the degenerated anticodon for cysteine). PCR products were transcribed in vitro with the T7 mMessage mMachine kit (Ambion), and RNAs were translated with a rabbit reticulocyte system in the presence of [³⁵S]methionine (7). Translation of RNAs coding for C proteins was stopped by the addition of 10 volumes of sample buffer (0.3 M Tris [pH 8.8], 2.5% sodium dodecyl sulfate [SDS], 100 mM dithiothreitol [DTT], 1 M sucrose, 0.01% Bromophenol blue), and labeled proteins were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) on 12% polyacrylamide gels and by autoradiography.

Antibodies and immunological techniques. Anti-SBV E2 and anti-SBV C rabbit antisera and anti-NS3 rat monoclonal immunoglobulin G were used for immunoblot, immunoprecipitation, and plaque immunostaining as described (7). For pulse-chase labeling of viral proteins, BHK cells were infected with the indicated viruses at a multiplicity of infection of 5 for 8 h at 37°C, labeled with ³⁵S-labeled amino acids (Easytag; Amersham) for 5 min at 37°C, and then chased for the indicated time with DMEM supplemented with a 10% FCS, 10× amino acids, and cycloheximide (50 µg/ml). Cell monolayers were washed once with PBS, lysed under denaturing conditions, and immunoprecipitated with the anti-SBV C rabbit antiserum as described previously (7).

Peptides and HPLC protease assays. N-acetyl hexa- or decapeptides with the sequence indicated in Table 3 were synthesized by 9-fluorenylmethoxy carbonyl/tertiary butyl chemistry on a Millipore 9050 Plus synthesizer and purified by high-performance liquid chromatography (HPLC). The concentration of stock peptide aliquots was determined by quantitative amino acid analysis. Cleavage assays were performed in 60 µl of 50 mM HEPES (pH 7.5), 15% glycerol, 0.05% Triton X-100, 10 mM DTT, 80 µM Pep4A, and 150 mM NaCl where indicated. Pep4A was preincubated for 10 min at 23°C with a purified recombinant NS3 protease domain (24), and the reactions were started by adding the substrate. Cleavage of decapeptide substrates and inhibition by hexapeptides were determined by HPLC, and kinetic parameters were calculated as described previously (22, 27).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Construction and characterisation of the libraries. The genome of the Mut5 chimeric virus was used as a backbone for the construction of libraries. This virus encodes an NS3-4Ap single chain protease fused to the amino terminus of the inactivated SBV C protein (Fig. 1A). Maturation of its structural polyprotein requires two NS3-4Ap-mediated cleavages: one, at the NS4A/C junction, is presumably a cis cleavage while the other, at the C/PE2 junction, can most likely also happen in trans. Although virus propagation is partially temperature dependent, Mut5 forms small but clearly detectable lysis plaques at 30°C (6).

View larger version (24K): [in this window] [in a new window]   FIG. 1.   Experimental outline. (A) Schematic representation of SBV and Mut5 virus polyprotein processing. SBV structural polyprotein is processed co- and posttranslationally to yield the nucleocapsid protein (C) and three membrane proteins (PE2, 6K, and E1), which form the viral envelope. C is an autoprotease and cleaves itself from the nascent precursor, which is then directed to the endoplasmic reticulum by a signal sequence located at the N terminus of PE2. PE2, 6K, and E1 are released from the precursor by signal peptidase. Processing of the Mut5 structural precursor depends on the activity of a fusion protein comprising the protease domain of NS3 and the entire NS4A (NS3-4A). NS3-4A is fused at the N terminus of C, which is inactivated by substitution of its catalytic serine (CSA), and two specific cleavage sites are engineered at the N and C termini of C. Empty, gray, and black arrows and bars indicate the cleavage sites of C, NS3-4A, and signal peptidase, respectively. (B) Schematic diagram of the structural polyprotein encoded by libraries of chimeric SBV viruses. The amino acidic sequences of the NS4A/C and C/PE2 junctions are indicated in the single-letter code. X, any amino acid or stop codon; Z, any amino acid but cysteine. The position (P6 to P4') of each residue with respect the scissile bond () is indicated. (C) Selection scheme. See Results for details.

Library screening. Libraries were screened by transfecting the in vitro-transcribed RNAs in BHK cells and isolating viruses able to form lysis plaques (Fig. 1C). In each experiment, 1.6 × 10⁷ BHK cells were transfected by electroporation with each RNA. Transfection efficiency varied between 40 and 80%. Three days after transfection, lysis plaques were picked and isolated viruses were amplified by passaging on BHK cells. RNA was extracted from infected cells, and the cDNA coding for the degenerated cleavage site was amplified by PCR and sequenced. Media containing infectious viruses were used for a second round of amplification on BHK cells to monitor processing of viral proteins and to confirm sequence.

4

4

5

5

6

6

2

Characterization of selected cleavage sites. In vivo processing of the polyproteins of selected viruses was initially investigated by immunoblotting with antibodies directed to the proteins flanking the degenerated junctions. With three exceptions described below, cells infected with viruses selected from libraries A and B produced PE2 and C proteins comigrating with those produced by the Mut5 virus (Fig. 2A and data not shown), demonstrating that the C/PE2 junctions of selected viruses were cleaved, presumably at the predicted cysteine-serine motif. Cells infected with clones B15, B19, and B81 produced correct-size PE2 proteins (not shown) and C proteins migrating slightly faster than that of Mut5 (Fig. 2A). Migration of these C proteins is probably due to the sequence of the degenerated region rather than to cleavage(s) occurring at an alternative or additional site(s) in the C protein. In fact, in vitro-synthesized C proteinsbearing the sequences present in clones B15, B19, and B81 and truncated after the cysteine residue in position P1also displayed anomalous migration (Fig. 2B). Similarly, cells infected with viruses selected from library D produced C proteins comigrating with those produced by the Mut5 virus (Fig. 2A and data not shown), indicating that also in this case the selected NS4A/C junctions were presumably cleaved at the predicted cysteine/threonine-alanine motif.

View larger version (39K): [in this window] [in a new window]   FIG. 2.   In vivo processing of the polyprotein of selected viruses. (A) BHK cells were infected with the indicated viruses, and cell lysates were analyzed by immunoblot with the anti-SBV C antibody by SDS-10% PAGE. (B) C proteins, comprising amino acids 106 to 264 of SBV C and containing the P5-P1 sequence of the indicated viruses, were translated in vitro and analyzed by SDS-12% PAGE and autoradiography. (C) BHK cells were mock infected (ctrl) or infected with the indicated recloned viruses, pulse-labeled with 35S-labeled amino acids for 5 min at 37°C, and then chased for the time (in minutes) indicated above each lane. The C protein was immunoprecipitated and analyzed by SDS-12% PAGE and autoradiography. Positions of molecular mass standards (in kilodaltons), C, and C proteins are indicated. For clarity, only a representative subset of samples is shown.

View larger version (53K): [in this window] [in a new window]   FIG. 3.   Comparison of the plaque phenotype of selected and recloned viruses. Shown are plaques produced in BHK cells after infection with the selected (B5, B15, and B80) and recloned (pSIN.B5, pSIN.B15, pSIN.B80, and Mut5) viruses. Plaques were revealed by immunostaining with the rabbit anti-SBV E2 serum after 3 days of incubation at 30°C.

View larger version (29K): [in this window] [in a new window]   FIG. 4.   In vitro processing of the polyprotein of viruses selected from libraries A and B. In vitro-translated C-PE2 proteins, comprising amino acids 106 to 533 of SBV structural polyprotein and containing the C/PE2 cleavage sites of the indicated viruses, were incubated without () or with increasing amounts of full-length NS3-4Ap (final concentrations, 5, 20, and 80 nM, as symbolized by the triangles above the gel) and analyzed by SDS-12% PAGE and autoradiography. Positions of molecular mass standards (in kilodaltons), C-PE2, C and PE2 proteins are indicated. For clarity, only a representative subset of samples is shown.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

This study describes a new viral display system for in vivo use and its application to defining the substrate specificity of the HCV serine protease.

N-terminal sequencing of the NS4A-to-NS5B proteins and comparison of HCV strains initially delineated a consensus for the NS3-4Ap cleavage (9, 18), which was then refined by mutational analysis in transfected cells (2, 13) and in vitro studies with synthetic peptides (27, 29). Overall, these studies indicated different requirements for cleavage at the cis and trans sites. The cis-cleaved NS3/NS4A junction of HCV is the only natural cleavage site with threonine in P1 and tolerates mutations at almost all positions, indicating that processing at this site is primarily determined by polyprotein folding (2, 13). In contrast, a cysteine residue at the P1 position is absolutely required for trans cleavage, and additional residues are crucial for efficient processing (2, 13, 27, 29). An acidic residue is present at the P6 position of all cleavage sites, and its substitution is detrimental for cleavage both in vivo and in vitro. Also, P3 and P4' residues contribute to efficient substrate recognition, with preference shown for residues present in the natural cleavage sites. Efficient in vitro cleavage requires a peptide substrate of at least 10 residues spanning P6 to P4', suggesting that ground-state substrate binding to NS3-4Ap is mediated by multiple, weak interactions involving distal residues (P6, P3, P1', and P4') and main-chain interactions between P5 and P2, while the P1 residue dictates the efficiency with which the substrate proceeds to the transition state.

Our results demonstrated not only that the SBV system could be used to depict a comprehensive picture of the NS3-4Ap specificity but also that it was sophisticated enough to reproduce the different requisites for cis and trans cleavage observed for the HCV polyprotein. Specificity of the protease at the cis-cleaved NS4A/C junction was highly degenerated, as indicated by the fact that 5% of the clones present in library D yielded infectious viruses and that sequences found in selected viruses did not show an obvious consensus, except for the presence of cysteine or threonine residues in the P1 position. These findings were in perfect agreement with the poor specificity observed for cis cleavage in the context of the HCV polyprotein (2, 13). Conversely, the amino acid sequences of the selected C/PE2 junctions delineated a trans-cleavage consensus entirely consistent with all previous data. The P1 residue of all selected viruses was cysteine, and we were not able to isolate and propagate viruses from the library lacking cysteine residues, indicating the absolute requirement for this residue in P1. The only four residues found at the P3 position (valine, isoleucine, glutamic acid, and threonine) were the same found at this position of HCV trans-cleavage sites. Although an acidic residue was present in the constant part of our libraries, the fact that it was in the putative P6 position in all but one selected virus underlined its importance. Comparison of libraries differing only for the P2'-P4' sequence emphasized the importance of the P' part of the substrate: the frequency of infectious viruses was higher in library B, which contained an HCV sequence, than in library A, which contained an SBV sequence. In addition, most viruses isolated from the latter library could not be propagated, suggesting that C/PE2 junctions of these viruses were cleaved inefficiently.

All selected sequences were cleaved by NS3-4Ap in vivo and in vitro with an efficiency comparable to that of the natural cis- or trans-cleavage sites of HCV, demonstrating the reliability of the SBV library system. Interestingly, in vivo analysis by pulse-chase labeling indicated that four of the sequences selected from libraries A and B were processed faster than the natural NS5A/B substrate, albeit the corresponding peptides did not show enhanced kinetic parameters in all the in vitro assays. This apparent discrepancy could be rationalized considering that in vitro determinations of kinetic parameters (Table 3) are biased by the experimental conditions, which are unavoidably arbitrary and affect each substrate differently. In addition, the context in which the different substrates were analyzed, short peptides in vitro and larger precursor proteins in vivo, could also account for the different results obtained in vitro and in vivo. Therefore, we feel that comparison of substrates with in vitro assays based on peptide substrates, albeit fundamental for obtaining quantitative data, must be considered valid only for defined experimental conditions and should, whenever possible, be validated by in vivo experiments with full-length proteins.

Our results also showed a correlation between the propagation phenotype of selected viruses and the kinetics of in vivo processing of the corresponding cleavage sites, indicating that the plaque phenotype of the selected viruses can be utilized to identify sequences which are cleaved more efficiently in vivo. This correlation is probably not absolute, since it is possible that the sequence of the cleavage site affects the virus phenotype also by influencing the structure and/or function of the cleaved C protein. Despite this limitation, we have shown that the possibility of ranking selected viruses on the basis of their plaque phenotype was very useful for identifying a manageable number of substrates whose processing efficiency could be estimated both in vivo and in vitro by biochemical and kinetic methods.

Similarly to other proteases, NS3-4Ap is competitively inhibited by peptides corresponding to the N-terminal cleavage products (22). We have exploited this feature to demonstrate that selected sequences inhibit the enzyme in vitro with potencies comparable to those of peptides derived from the natural cleavage sites. Because of their size and peptidic nature, these inhibitors do not enter efficiently into cells and cannot be tested in vivo. Nonetheless, the identification of new sequences efficiently recognized by the enzyme in vivo provides useful information for the development of synthetic compounds which effectively inhibit NS3-4Ap and eventually could be used to cure the disease caused by HCV.

SBV libraries represent a new tool for applying a combinatorial approach to the definition of protease specificity in the intracellular environment. Their most important property is that they are screened in vivo, thus avoiding all problems connected with defining the appropriate in vitro conditions, as is the case with phage or chemical libraries. This aspect is obviously a clear advantage when analyzing intracellular enzymes. In this regard, SBV libraries are similar to retroviral libraries (4), the main difference being that, in the present design, SBV libraries allow the target enzyme to be chosen while retroviral libraries do not. We believe that SBV libraries can be easily adapted to study other intracellular proteases. The HCV serine protease was engineered to cleave itself off a polyprotein precursor containing additional cleavage sites, which would not be feasible for other enzymes unable to cleave at their C terminus. However, the target enzyme does not need to be fused to the SBV structural polyprotein, and alternative expression strategies could be used, as SBV vectors are an extremely flexible experimental system. Duplication of the internal promoter of the viral replicase would allow expression of the protease and of the structural polyprotein, containing a randomized cleavage site at the C/PE2 junction, from two different mRNAs (11). Alternatively, the protease and the polyprotein could be expressed by a bicistronic mRNA obtained by the insertion of an internal ribosome entry site.

An important feature of SBV libraries is that they do not require a purified enzyme but only require the cognate cDNA. In theory, it is conceivable to study proteins which are predicted to be proteolytic enzymes on the basis of their sequence and for which no biochemical information is available. While phage-displayed and chemical peptide libraries can be used for many different proteases, a new SBV library has to be constructed for each enzyme. However, since the SBV genome can be easily manipulated at the cDNA level, constructing libraries is not a limitation and in fact offers the opportunity to redesign the scaffold of the libraries to meet particular requirements of the target enzyme. An additional advantage of SBV libraries is that infectious viruses form lysis plaques and are thus easily isolated and characterized with a single round of selection.

The complexity of our SBV libraries ranged from 1 × 10⁶ to 10 × 10⁶ clones. Considering that 50% of the clones either did not contain the degenerated sequence or contained a stop codon, and assuming some redundancy, these libraries approached the theoretical diversity predicted for a library with five degenerated positions (3.2 × 10⁶ possible protein sequences). These figures are similar to those reported for phage- or retrovirally displayed libraries (4, 16). Certainly there is still room for improvement: one useful expedient to increase diversity and reduce redundancy is to use a codon-based mutagenesis strategy (17), which also allows the assembly of libraries with selectively degenerated sequences, such as our library C.