INTRODUCTION

Poliovirus gene expression relies on the synthesis of a large polypeptide precursor that is proteolytically cleaved to generate the mature viral proteins. Polyprotein processing is accomplished by two virus-encoded proteases: 2A^pro1 and 3C^pro (1, 2). The poliovirus endopeptidase 2A^pro is a polypeptide of 149 amino acid residues in which the catalytic triad is formed by His²⁰, Asp³⁸, and Cys¹⁰⁹ (3, 4). The fact that cysteine instead of serine forms part of the active site of 2A^pro has been taken as evidence that this protease is an evolutionary intermediate between serine and cysteine proteases (5). However, sequence similarities between 2A^pro and trypsin-like serine proteases have conclusively classified 2A^pro as belonging to the serine protease group (6, 7).

Both proteases 2A^pro and 3C^pro cleave substrates in cis and trans (1, 2). cis cleavage by 2A^pro occurs once the protease has been synthesized as a precursor on the nascent polypeptide chain still bound to ribosomes (8). Thus, 2A^pro activity cleaves the Tyr-Gly peptide bond between P1 and 2A, releasing P1, the precursor of the capsid proteins, from the rest of the nascent polyprotein (1). The second proteolytic cleavage effected by 2A^pro on a viral protein precursor lies in 3CD, where a Tyr-Gly dipeptide is hydrolyzed, giving rise to 3C and 3D, two products of unknown function (9). Apparently, the generation of 3C and 3D is not necessary for poliovirus growth in culture cells, since a poliovirus mutant unable to produce 3C and 3D grows as well as wild type poliovirus does (10). In addition to the two cleavages performed by 2A^pro on poliovirus protein precursors, this protease also cleaves cellular proteins (2). A number of cellular polypeptides are degraded after poliovirus infection, although the proteases responsible for these cleavages have not yet been identified (11). The best documented cleavage of a cellular protein by 2A^pro is that of eIF-4G (also known as p220), a component of the translation initiation factor eIF-4F (12). It has not been formally proven that poliovirus 2A directly cleaves p220, but the fact that coxsackievirus 2A and rhinovirus 2A directly degrade p220 (13-15) makes it unlikely that a cellular protease activated by 2A^pro is responsible for p220 cleavage (16, 17). The proteolytic cleavage of the dipeptide present in the initiation factor eIF-4G generates the N-terminal eIF-4G moiety involved in eIF-4E interaction, whereas the C-terminal eIF-4G moiety, which binds eIF-4A, participates in the translation of picornavirus RNA (18). Translation of newly synthesized mRNAs is halted by p220 cleavage (19), but protein synthesis on mRNAs already engaged in the protein-synthesizing machinery continues (20, 21). On the other hand, uncapped picornavirus RNA does not require the cap binding factor eIF-4E but does depend on the C-terminal eIF-4G moiety and eIF-4A to become engaged in translation (18).

Poliovirus 2A^pro is a multifunctional enzyme. Apart from proteolytic cleavage of viral protein precursors and eIF-4G, 2A^pro can also enhance the translation of poliovirus RNA (22). The activation of protein synthesis by 2A^pro requires only the poliovirus internal ribosomal entry site (IRES) sequences, since translation of reporter genes placed under the IRES is stimulated by this protease (22-24). The exact molecular basis by which 2A^pro stimulates translation remains unknown, but the simple generation of eIF-4G cleavage products is not sufficient for this enhancement to take place (24). The finding that poliovirus mutants in the IRES region compensate for this defect with mutations in the 2A^pro sequence (25) suggested an interaction between 2A^pro and the IRES, but direct evidence for such an interaction is still lacking. 2A^pro not only enhances poliovirus RNA translation but also participates in viral RNA replication by a still unknown mechanism (26, 27). Unlike aphtoviruses and hepatitis A virus, in which the L and 2A^pro genes, respectively, are not necessary for viral infectivity (28, 29), the poliovirus 2A^pro is required for virus RNA replication and viability (30, 31). Certainly, the generation and analysis of additional poliovirus 2A^pro variants should provide more insights into the different activities of this multifunctional protease that regulates viral and cellular gene expression.

Inducible expression of poliovirus nonstructural proteins in the yeast Saccharomyces cerevisiae showed that two of them, namely 2A^pro and 2BC, inhibited yeast growth (32-34). Recently, a similar effect was found for rhinovirus 2A^pro (35). This finding allows the use of yeast cells as a genetic system in which to select 2A^pro variants that permit S. cerevisiae growth. We now report the generation and characterization of several 2A^pro variants selected by this method. In principle, this approach can be applied to analyze the structure-function relationship of any other viral or nonviral protein that interferes with yeast cell division.

EXPERIMENTAL PROCEDURES

Transformation of yeast by the lithium acetate procedure was performed as described previously (36). Yeast growth and induction of UAS_GAL-CYC promoter was performed as described (34). The different media used in this work (YNB. Glu, YNB. Gal, and YNB. LGal) have been defined previously (34).

The plasmid pEMBL.2A (32) was randomly mutated using hydroxylamine as described (36). This mutated DNA was used to transform the S. cerevisiae strain W303-1B (Mat , Ade2, His3, Leu2, Trp1, Ura3) to obtain 100-200 colonies in each YNB.Glu plate. These plates were replicated in YNB.Gal medium and incubated for 48 h either at 30 or 37 °C. Colonies able to growth in YNB.Gal medium were considered potential mutants in the inhibitory activity of 2A. These clones were amplified, and their capacity to grow at different temperatures and to express the 2A^pro poliovirus protein in liquid medium was tested. Mutant plasmids were isolated from yeast cells using standard procedures (36), and Escherichia coli DH5 cells were transformed with these lysates. The locations of the different mutations were determined by DNA sequencing following the dideoxy method (37).

E. coli DH5 (37) was used for the construction of all expression plasmids described. To introduce the different mutations in plasmid pT7XLD (38) containing the infective clone of poliovirus, a subcassette with the 3235-3953 sequence of poliovirus was constructed (pSub.2A). The mutants 2, 3, 4, 5, 8, and 9 were cloned in this plasmid using the PstI and Asp718 sites; the mutants 6, 7, 10, 12, 13, 15, and 21 were cloned using the NcoI sites. Mutant 17 was cloned using the PstI and NcoI (partial digestion) sites. The various pSub.2A* vectors were digested with BstEI and cloned in pT7XLD digested with this same enzyme to obtain the different pT7XLD(2A*) mutants. The sequence of each plasmid was verified by partial DNA sequencing of the 2A region by the dideoxy method (37). The different pTM1-2A* plasmids were constructed by digestion of pTM1-2A (WT) (39) and all the pEMBL-2A* plasmids with PstI and SalI.

The -2A, -2C, and -3C sera were obtained by inoculation of the fusion proteins MBP.2A, MBP.2C, or MBP.3C in rabbits. The antiserum against p220 (eIF-4G) protein has been previously described (39). Yeast extracts for Western blot analysis were prepared as indicated (40). For standard immunoreactions, sera dilutions of 1:1000 in phosphate-buffered saline-Tween 20 (0.05%) were used. Immunoblots were carried out as described elsewhere (34).

All plasmids were linearized and transcribed with phage T7 RNA polymerase as described (24). HeLa S10 cell extract treated with micrococcal nuclease was used for in vitro translation as described previously (41)

For DNA and RNA transfections we used the protocol described by Aldabe et al. (42) and Ventoso and Carrasco (24), respectively. To obtain poliovirus after RNA electroporations, about 1 × 10⁷ cells were transferred to a 0.4-cm cuvette and 5 µg of in vitro synthesized full-length poliovirus RNA was added. Electroporation was performed by a single pulse at 250 V and 960 microfarads using a Bio-Rad Gene Pulser apparatus with the pulse-controller unit set at maximum resistance. Half of these cells were transferred to a 6-cm culture dish, and 4 h later the cells were overlaid with 0.45% agarose and incubated at 32.5, 37, or 39.5 °C for several days before staining with crystal violet. The remaining cells were diluted with nonelectroporated HeLa cells and were overlaid or not to pick plaques or recover the supernatant. WT and viable mutant polioviruses were isolated by picking several plaques visualized after staining with neutral red. To verify the presence of the mutated 2A^pro, the sequence was amplified from the extracted viral RNA by reverse transcriptase polymerase chain reaction and sequenced by the dideoxy method (37).

Maintenance of HeLa and infection with WT or 2A mutant polioviruses, analysis of viral proteins, transactivation assay, measurement of viral RNA synthesis, and dot blot analysis were carried out as described previously (24).

RESULTS

The finding that expression of 2A^pro was inhibitory to S. cerevisiae cell growth (32) opened the possibility of isolating 2A^pro variants devoid of this inhibitory capacity. To this end, plasmid pEMBL containing the poliovirus 2A^pro sequences was mutagenized with hydroxylamine. Yeast cells were transformed with the mutated pEMBL-2A, and colonies that grew in glucose agar were replicated in plates containing galactose to induce the synthesis of 2A^pro. To assay for thermosensitivity of the yeast clones obtained, the galactose plates were incubated at two different temperatures, 30 and 37 °C. Upon galactose induction 22 colonies of yeast cells that grew at 37 °C were obtained, indicating that poliovirus 2A^pro was not induced or was noninhibitory for these yeast clones (Table I). Six of these clones presented a ts phenotype, since their growth in liquid (results not shown) and solid media was selectively arrested at 30 °C (Fig. 1A). This result gives the percentage of yeast clones that grow in galactose after mutagenesis of 1.2% (Table I). Plasmid DNA was obtained from these clones, and the 2A sequences were determined. These results are summarized in Tables I and II. Thus, 15 different point mutations in the 2A^pro sequence have been identified. In addition, two nonsense mutants at codon 24 arose. Finally, in two other clones (numbers 1 and 11) the mutation that conferred the ability of yeast cells to grow in galactose could not be identified. Notably, none of these 2A^pro variants contained a mutation in the catalytic triad of the protease.

Table I. Hydroxylamine mutagenesis of pEMBL.2A Transformant colonies in Glu plates/Gal replica plates   pEMBLyex4 + hydroxylamine 5000/5000   pEMBL.2A 5000/0   pEMBL.2A + hydroxylamine 1800/22  1.2% Thermosensitivity   Colonies on replica plates at 30 °C: 16   Colonies on replica plates at 37 °C: 22   Clones with temperature-sensitive   phenotype 27% Summary of the mutations   Undetermined mutations: 2 (mutants 1 and 11)   Nonsense mutations: 1 (mutants 14 and 16)   Point mutations: 15 (3 of them appear twice)   Transitions G to A: 8   Transitions C to T: 6   Others     1 change C to A     1 double change dimer GG to AA

Table II. Poliovirus 2Apro mutants obtained in S. cerevisiae pb, mutated nucleotide; aas, amino acid change; ts, thermosensitivity; 2A, 2A expression in yeasts. pb, mutated nucleotide; aas, amino acid change; ts, thermosensitivity; 2A, 2A expression in yeasts. Mutant pb aas ts 2A 1 ND*a ND   + 2 G   A G60R   + 3 C   T S66F + + 4 C   T L39F   + 5 C   T N18K   + 6 G   A D135N + + 7 C   A V119M + + 8 G   A A22T   + 9 G   A C55Y   + 10 C   T S134L   + 11 *2b - +   12 G   A G110S   + 13 G   A G126D + + 14 C   T Q24Stop     15 G   A G121E   + 16 C   T Q24Stop     17 GG   AA G100N   + 18 C   T S134L   + 19 G   Ac V9M   + 20 G   A G60R   + 21 C   T A125V + + 22 G   A G121E   + 2A WT - -   + a The plasmid pEMBL-2A* was not recovered in bacteria. b Recombinant plasmid, 1 kilobase bigger than WT. No substitutions in 2A sequence. c Substitution close to N terminus of the protein; mutant not subcloned.

To test if the ability of the yeast clones to grow in galactose medium was due to the absence of 2A^pro synthesis, cell extracts were obtained, separated by SDS-polyacrylamide gel electrophoresis, and immunoblotted with anti-2A polyclonal antibodies (Fig. 1B). The majority of the clones expressed 2A^pro upon galactose induction at higher levels than in yeast cells bearing the control plasmid pEMBL-2A. Only three clones (numbers 11, 14, and 16) did not synthesize any polypeptide that reacted with anti-2A antibodies. This is logical in the case of clones 14 and 16, which encode a truncated 2A protein of only 23 amino acids, whereas clone 11 still contains the 2A^pro sequence but has lost its inducibility. This finding suggests that the various mutated 2A^pro enzymes that are expressed in these clones have lost their capacity to inhibit yeast growth. Therefore, yeast cells represent an amenable genetic system in which to easily obtain poliovirus 2A^pro variants.

The next step in our studies was to reconstitute the entire poliovirus genomes bearing the different 2A^pro variants. To this end, a subcassette containing the 3235-3953 poliovirus sequence was engineered. The 2A^pro sequences obtained from the variant pEMBL-2A* isolates were initially cloned in this subcassette, and finally the different sequences were passed to plasmid pT7XLD containing the full-length poliovirus genome. Thus, 14 poliovirus genomes bearing different mutations in 2A^pro were reconstructed. Poliovirus RNA was obtained from these mutants by in vitro transcription with T7 RNA polymerase. These RNAs were translated in a HeLa cell-free system, and the proteins synthesized were analyzed by SDS-polyacrylamide gel electrophoresis. Fig. 2A reveals that some of the mutants, namely numbers 3, 6, 7, 8, and, to a lesser extent, 12, show almost no P1-P2 precursor but produce significant amounts of 2A. The other mutants show various amounts of P1-P2 precursor and mature 2A^pro. No P1 is detected in some of the mutants, such as 2, 5, 9, 13, and 15, but instead a protein band of higher molecular weight is detected that could correspond to P1-2A. Therefore, mutants in which P1 is absent should have a defect in the ability of 2A^pro to cleave in cis the dipeptide between P1 and 2A^pro.

C

To further characterize the enzymatic capacities of the different mutated 2A^pro enzymes, eIF-4G cleavage was examined by transfection of the different pT7XLD(2A*) plasmids and infection with a recombinant vaccinia virus that expresses the T7 RNA polymerase in HeLa cells. Fig. 2B shows that almost no cleavage of eIF-4G is detected with mutants 5 and 15, little cleavage of this initiation factor occurred with mutants 2 and 17, and the rest of the mutants clearly accomplished eIF-4G cleavage. Since mutants 5 and 15 do not show any sign of synthesis of mature 2A^pro, it is possible that the 2A^pro present in the P1-2A precursor is very inefficient to cleave eIF-4G. Alternatively, it is possible that 2A^pro as such is devoid of eIF-4G cleavage activity. To investigate which of these two possibilities is correct, all the individual 2A^pro variants were cloned in pTM1 and their activities analyzed (shown below).

The formation of lytic plaques upon transfection of HeLa cells with the different poliovirus RNA variants was examined (Fig. 3). Poliovirus plaques were obtained with mutants 3, 6, 7, 8, and 12 after 3 days of incubation, although these plaques have a reduced size compared with WT poliovirus RNA. Minute plaques arise with mutant 10 and extra-minute plaques with mutant 4 after 5 days of incubation, whereas no recovery of virus was possible from the rest of the mutants. Production of batches of mutants 4 and 10 without loss of their phenotype has not been possible. The attempts to recover reconstituted virus by electroporation of pT7XLD(2A*) plasmids have been done at least twice, with the plates incubated at 32.5, 37, and 39 °C. We were also unable to recover reconstituted virus from cells transfected by the Lipofectin procedure, although we successfully recovered mutants 3, 6, 7, 8, and 12 by this approach.

To determine at the molecular level the defect that prevents poliovirus mutants 3, 6, 7, 8, and 12 from forming large plaques, several analyses were carried out. Initially, the capacity of these variant polioviruses to synthesize proteins was investigated. Fig. 4A shows that all five poliovirus mutants examined show defects in their capacity to synthesize proteins at control levels, although they all shut down host translation efficiently. To further characterize the level of poliovirus protein synthesis in cells infected with these polioviruses, cell extracts corresponding to 2 and 6 hpi were immunoblotted with an antibody against poliovirus protein 2C (Fig. 4B). Clearly, all five poliovirus variants synthesize lesser amounts of proteins 2BC and 2C compared with cells infected with control poliovirus. The reduced levels of poliovirus proteins produced by these five mutants may be due to synthesis of less viral RNA or to a lower capacity of the viral RNA to be translated. A poliovirus 2A^pro mutant was described recently that showed defects in translation of the viral RNA, whereas genome replication was much less affected (24). Therefore, the amount of poliovirus RNA was examined at various times after infection with the five poliovirus mutants (Fig. 4C). Less viral RNA was present at 5 hpi in all five variants compared with WT poliovirus, whereas this amount was similar in all of them by 8 hpi. However, the cytopathic effect of WT poliovirus at this late time was much higher than with the rest of the mutated polioviruses (results not shown). These results indicate that viral genome replication is slower in the mutant polioviruses but reaches levels at 8 hpi similar to those present in WT-infected cells at 5 hpi. Despite this similarity, viral protein synthesis by the variant poliovirus is much lower than that observed with WT poliovirus.

Poliovirus 2A^pro has the capacity to activate the translation of mRNAs that contain the poliovirus 5 untranslated region (22). Therefore, we next examined the ability of the different poliovirus variants to enhance the translation of a reporter mRNA that bears the poliovirus 5 untranslated region before the luciferase sequence. Fig. 5A shows that transfection of poliovirus leader-luciferase mRNA in poliovirus-infected HeLa cells results in almost a 6-fold stimulation of luciferase synthesis compared with mock-infected cells. This activation is observed even when the infected cells are treated with guanidine, indicating that poliovirus genome replication is not necessary for this phenomenon to occur. The five poliovirus mutants assayed are defective to varying degrees at enhancing the translation of the leader-luciferase mRNA. Thus, transactivation by mutant 3 is about 1.2 times that of control uninfected cells, whereas mutant 6 enhances leader-luciferase translation by about 2.5 times. Our conclusion from these results is that all five poliovirus mutants are hampered in their ability to enhance leader-luciferase mRNA translation.

Importantly, analysis of p220 cleavage in HeLa cells infected with the different poliovirus variants shows that all of them cleave p220 as efficiently as control poliovirus (Fig. 5B). This finding supports the idea that p220 cleavage is not sufficient to transactivate the poliovirus leader region. In addition, the presence of a WT poliovirus 2A^pro is necessary for this phenomenon to take place.

The capacity of the different viable mutants to generate 3C was also evaluated. The poliovirus precursor 3CD undergoes alternative cleavages: either it is processed by 3C^pro to yield 3C^pro plus 3D, or 3CD is cleaved by 2A^pro to produce 3C and 3D. The function, if any, of these alternative cleavage products of 3CD remains obscure, since a poliovirus unable to generate 3C and 3D replicates well in culture cells (10). Fig. 6A shows that generation of 3C is greatly diminished in cells infected with any of the five poliovirus variants assayed. Some 3C is clearly visible, particularly in mutants 7 and 8, but the proportion of 3C to 3CD or to 3C^pro in the five mutants is very small compared with that in HeLa cells infected with WT poliovirus.

Some of these 2A^pro variants, such as 3, 6, and 7, showed a temperature-sensitive phenotype to induce the inhibition of yeast growth (Fig. 1A). Therefore, it was of interest to determine if this characteristic was preserved in the reconstituted polioviruses. For this purpose, eIF-4G cleavage (results not shown), generation of 3C (Fig. 6A), and virus plaque formation (Fig. 6B) were assayed at different temperatures. None of the poliovirus variants studied showed a ts phenotype in any of these assays compared with the WT poliovirus.

Therefore, the five polioviruses mutated in 2A^pro showed defects in viral protein synthesis, transactivation of leader-luciferase translation, and 3CD cleavage to generate 3C. However, these poliovirus variants retained their capacity to cleave eIF-4G. A summary of these results is shown in Table III.

Table III. Summary of the activities of the different poliovirus 2Apro variants 2A*, different 2A variants; aas, amino acid change; PP, plaque phenotype (wt, wild type plaque size; sp, small plaque; mp, minute plaque; emp, extra-minute plaque; nv, nonviable); ts, thermosensitivity; cis, cis cleavage in cell-free system after translation; p220,  p220 cleavage; 3CD, 3CD cleavage; R, viral replication; Ta, transactivation activity; , no activity; ++, maximal activity; ND, not determined. 2A*, different 2A variants; aas, amino acid change; PP, plaque phenotype (wt, wild type plaque size; sp, small plaque; mp, minute plaque; emp, extra-minute plaque; nv, nonviable); ts, thermosensitivity; cis, cis cleavage in cell-free system after translation; p220,  p220 cleavage; 3CD, 3CD cleavage; R, viral replication; Ta, transactivation activity; , no activity; ++, maximal activity; ND, not determined. Mutant aas Viruses/pT7XLD(2A*) pTM1-2A* PP ts cis p220 3CD R Ta p220 3CD Ta 2 G60R nv     + ND ND ND +   + 3 S66F sp   ++ ++ +/ + +/ ++ +/ + 4 L39F emp   + ++ ND ND ND ++   +/ 5 N18K nv       ND ND ND +/   + 6 D135N sp   ++ ++ +/ + + ++ +/ + 7 V119M sp   ++ ++ +/ + +/ ++ +/ + 8 A22T sp   ++ ++ +/ + +/ ++ +/ + 9 C55Y nv     ++ ND ND ND +   + 10 S134L mp   + ++ ND ND ND ++ +/ +/ 12 G110S sp   ++ ++ +/ + + ++ +/ + 13 G126D nv     + ND ND ND ++   +/ 15 G121E nv       ND ND ND +/   +/ 17 G100N nv   +/ + ND ND ND +   +/ 21 A125V nv   +/ ++ ND ND ND +   +/ 2A WT   wt   ++ ++ ++ ++ ++ ++ ++ ++

To extend the analysis of the 2A^pro activities not only to the mutated 2A^pro that rendered infectious polioviruses but also to the rest of the 2A^pro variants obtained in this work, the 14 2A^pro sequences were cloned in plasmid pTM1 bearing a phage T7 promoter. Synthesis of 2A^pro is achieved upon transfection and infection with a recombinant vaccinia virus that synthesizes the T7 RNA polymerase. Expression of the 2A^pro gene occurs even when viral replication is inhibited by Ara-C (39). [³⁵S]Methionine labeling under these conditions detected no vaccinia virus proteins, whereas 2A^pro was synthesized at high levels (Fig. 7A). In fact, all 14 2A^pro variants were synthesized to higher levels than WT poliovirus 2A^pro in this system. The various degrees of 2A^pro synthesis may correlate, at least in part, with their toxicity for the transfected cells. Thus, some mutated 2A^pro variants, such as 2, 4, 5, 9, and 15, are synthesized to very high levels compared with those that yield viable polioviruses (3, 6, 7, 8, and 12). Analysis of eIF-4G cleavage in cells transfected by the 14 2A^pro variants shows that all of them possess the ability to cleave eIF-4G, although some of them still leave significant amounts of intact eIF-4G (Fig. 7B). Densitometric analyses to quantitate the proportion of cleavage products versus intact eIF-4G indicate that mutants 2, 5, and 15 are the least active (<5% of the activity of WT 2A^pro).

The capacity of 2A^pro to cleave 3CD in trans was assayed by double transfection of cells with pTM1 bearing the different 2A^pro- and pTM1-expressing poliovirus 3CD. The proteolytic activity of 2A^pro on 3CD becomes apparent by the generation of 3C that is detected by immunoblot assays with a polyclonal anti-3C antiserum (Fig. 7C). Notably, only the 2A^pro variants able to reconstitute viable poliovirus generate 3C, albeit at lower levels than WT 2A^pro. None of the other 2A^pro mutants were able to produce even minimal amounts of 3C that could be detected after longer exposures of the x-ray film.

Finally, the transactivation capacity of the 2A^pro variants on a leader-luciferase mRNA were assayed. To this end, both the RNA encoding 2A^pro and the leader-luciferase mRNA were transfected into HeLa cells by the Lipofectin method (24). As occurred with the 3CD cleavage activity, all 14 2A^pro variants showed defects in transactivating the synthesis of luciferase (Fig. 7D). Two of them, showing a higher transactivation capacity (mutants 3 and 6), were also able to produce small plaque polioviruses. However, the rest of the 2A^pro mutants able to reconstitute polioviruses (mutants 7, 8, and 12) had transactivation activities similar to another 2A^pro variant incapable of producing viable polioviruses (mutant 2). Our conclusion from these experiments is that the capacity of the various 2A^pro to cleave eIF-4G or 3CD differs. Moreover, some 2A^pro variants produce significant amounts of eIF-4G cleavage products but are devoid of transactivation activity. This result agrees with the idea that the simple cleavage of eIF-4G is not sufficient to enhance the translation of a mRNA bearing the poliovirus leader sequence. Finally, there is a good correlation between the ability of 2A^pro to generate 3C (and 3D) and its capacity to produce viable polioviruses. A summary of these results is shown in Table III.

DISCUSSION

The isolation and characterization of gene variants is of primary importance for understanding poliovirus molecular biology (31). Three main strategies have been used to isolate poliovirus variants: selection of spontaneous mutants with a given phenotype (plaque size, drug resistance, etc.), random mutagenesis followed by selection, and site-directed mutagenesis (31). Random mutagenesis of a given poliovirus gene combined with a system for assaying a particular function facilitates the isolation of mutant viruses with phenotypes of interest. Some systems have been described to detect rhinovirus 2A^pro and coxsackievirus 3C^pro variants using E. coli and S. cerevisiae, respectively (43, 44). Both assays are based on the so-called "proteinase-trapping" technique in which hybrid variants between the picornavirus protease and B-galactosidase devoid of cis cleavage activity are selected. However, the mutated proteins thus selected have not yet been analyzed in mammalian cells or in reconstituted viruses. Other functional analyses of human viral proteins in yeast cells have been reported recently (34, 35, 45-49), but again, those studies were restricted to yeast cells.

The versatile and simple genetic system described in this work for selecting poliovirus 2A^pro mutants unable to block yeast growth will provide new 2A^pro variants that will offer further insight into the structure-activity relationship of this protease and provide more details about its functioning during the poliovirus lytic cycle. The random mutagenesis method of the 2A^pro gene used in this work yields approximately 1% of 2A^pro mutants; about 25% of those have a ts phenotype in yeast cells. Notably, most of the 2A^pro variants obtained contain point mutations, allowing the characterization of new regions in this poliovirus protease involved in substrate recognition. The isolation of many more 2A^pro variants using this approach would provide not only a more detailed characterization of the domains in 2A^pro but also the possibility of reconstituting ts poliovirus mutants. However, thus far, our attempts in this direction have been unsuccessful. Another still unexplored possibility for this approach is to screen for second site revertants of some of the 2A^pro variants. This would provide additional data on the interactions of different regions of the 2A^pro.

Although poliovirus 2A^pro blocks S. cerevisiae growth strongly, the exact target of its activity remains unknown (32). Although gene expression in yeast cells is clearly affected, the target recognized by 2A^pro is probably unrelated to mammalian eIF-4G (35). It is possible that eIF-4G and the yeast target of 2A^pro share common structural motifs that make them both substrates for 2A^pro. Certainly, the 2A^pro variants isolated from yeast cells show defects in substrate recognition and cleavage when tested in mammalian cells. This finding validates the use of yeast cells as an assay to isolate 2A^pro variants with defects in poliovirus growth and 2A^pro activity in human cells. One limitation of the system described for generating different 2A^pro variants is that it relies on the same selective pressure: the toxicity of 2A^pro for yeast cells. However, the different 2A^pro mutants obtained can be classified in different groups: some generate viable polioviruses, others still retain full eIF-4G activity, and a few are deficient in all the activities assayed. In particular, all the mutant viruses obtained are deficient in 3CD cleavage and transactivation of the poliovirus IRES region. Most likely, the basis of the genetic selection in yeast is related to these phenomena in molecular terms. In conclusion, these findings show that the use of this selection system would provide a wide range of 2A^pro mutants. Moreover, it is possible that varying the conditions of yeast growth and the composition of the medium would also influence the selective pressure and the types of 2A^pro variants isolated.

Poliovirus 2A^pro is not only involved in cis and trans cleavage of poliovirus protein precursors but also degrades cellular substrates (2). In addition, 2A^pro is a translational activator of mRNAs bearing the poliovirus IRES sequence (22) and is involved in viral RNA replication (26, 27). Some of these 2A^pro activities appear to be associated in the protease variants analyzed. Thus, the five viable poliovirus mutants analyzed in this work (namely 3, 6, 7, 8, and 12) show a small plaque phenotype, defects in the transactivation activity, and reduced viral protein synthesis. The 2A^pro variants 4 and 10 produce viruses with the minute plaque phenotype, show defects in cis cleavage, and accumulate P1-P2 and P1-2A precursors (see Fig. 2A); however, they cleave eIF-4G as efficiently as the other five viable polioviruses. All viable polioviruses are able to cleave P1-2A efficiently but cleave 3CD very ineffectively. Curiously, all of the 2A^pro variants that did not give rise to viable viruses were unable to cleave 3CD. This finding is puzzling, since no role has yet been assigned to 3C and 3D in the poliovirus replication cycle. Moreover, the mutation of this alternative cleavage site in 3CD has no consequences for virus growth (10). Our finding, however, could suggest either that the 3CD cleavage products or the 2A^pro proteolytic capacity responsible for this cleavage is essential for virus viability. The fact that the 3CD cleavage site recognized by 2A^pro is maintained after passage in cultured cells, coupled with our present findings, could foster additional work on the potential function of 3C and 3D.

Notably, all the 2A^pro variants obtained in this work cleaved eIF-4G, although to varying extents. The mutants that affected eIF-4G less were those that processed P1-2A less efficiently. On the other hand, P1-2A and 3CD are left almost intact by some 2A^pro mutants, whereas eIF-4G is still hydrolyzed to some extent. These findings indicate that 2A^pro has a region involved in substrate recognition manifested not only in yeast cells but also in other substrates of mammalian cells. Mutations in this region do not have exactly the same consequences for recognition and cleavage of different substrates. It would be of interest in the future to isolate 2A^pro mutants totally devoid of eIF-4G cleavage activity.

The finding that some 2A^pro variants do not transactivate the IRES of leader-luciferase mRNA but still cleave eIF-4G indicates that generation of the C terminus of eIF-4G alone does not suffice for the stimulation of translation of poliovirus mRNA to occur. These results add support to the previous findings from our group that transactivation requires not only eIF-4G cleavage but also intact 2A^pro (24). All the 2A^pro variants generated are deficient in proteolytic and transactivation capacity. This could suggest that for transactivation to occur the level of 2A^pro proteolytic activity needs to pass a threshold to cleave a given cellular substrate. The possibility that this poliovirus protease directly participates in the translation of poliovirus RNA is also an attractive possibility. Further work to elucidate the exact mechanism by which some picornavirus proteases enhance translation is necessary.

The three-dimensional structure of picornaviral 2A^pro has been modeled based on that of -lytic proteinase (3, 6). The three-dimensional structure of 2A^pro reported by Yu and Lloyd (3) was based on the structure of smaller bacterial serine proteases that in principle lack some of the functions assigned to picornavirus 2A^pro, such as the transactivation activity and participation in RNA replication. Until this theoretical modeling is confirmed by direct determination of the three-dimensional structure, mutational analysis of 2A^pro represents a good approach through which to gain insight into its structure-function relationships. Using the three-dimensional model of 2A^pro, Macadam et al. (25) localized most of the mutations observed in their study in two clusters near the surface of the protein and away from the active site, suggesting a direct interaction between these regions of 2A^pro and the 5 noncoding sequence of poliovirus RNA.

The substitutions of the 2A^pro variants that we have found cluster near the catalytic triad. Most of the mutations concentrate in the putative mouth where the substrate binds. The schematic depiction of the different poliovirus 2A^pro variants in the three-dimensional model of 2A^pro is given in Fig. 8. Four of five 2A^pro mutants that show a ts phenotype in yeast cells are grouped in a straight line between amino acids 119 and 136. Conceivably, this region of the protease undergoes conformational changes induced by temperature that influence substrate recognition in yeast cells. The two mutants with stronger defects in cleavage in cis and proteolysis of eIF-4G map close to the catalytic site: mutant 5 in the putative active site and mutant 15 in the middle of the molecule. Both substitutions involve charged amino acids, suggesting that these two residues play a crucial role in the 2A^pro activity. The three additional 2A^pro mutants lacking cis cleavage capacity but with some activity on eIF-4G are also located in a straight line that overlaps part of the ts region defined above, further strengthening the idea that this region of 2A^pro participates in substrate recognition. Therefore, the analysis of multiple random mutants obtained by different genetic assays (2A revertants for mutations in the 5 noncoding region (25) or noncytotoxic 2A mutants in yeast (as shown in this study)) could allow the identification of the different functional domains present in the poliovirus 2A^pro.