Regulation of the Sequential Processing of Semliki Forest Virus Replicase Polyprotein*

The replication of most positive-strand RNA viruses and retroviruses is regulated by proteolytic processing. Alphavirus replicase proteins are synthesized as a polyprotein, called P1234, which is cleaved into nsP1, nsP2, nsP3, and nsP4 by the carboxyl-terminal protease domain of nsP2. The cleavage intermediate P123 (cid:1) nsP4 synthesizes minus-strand copies of the viral RNA genome, whereas the completely processed complex is required for plus-strand synthesis. To understand the mechanisms responsible for this sequential proteolysis, we analyzed in vitro translated Semliki Forest virus polyproteins containing noncleavable processing sites or various deletions. Processing of each of the three sites in vitro required a different type of activity. Site 3/4 was cleaved in trans by nsP2, its carboxyl-terminal frag-ment Pro39, and by all polyprotein proteases. Site 1/2 was cleaved in cis with a half-life of about 20–30 min. S ite 2/3 was cleaved rapidly in trans but only after release of nsP1 from the polyprotein exposing an “activa-tor” sequence present in the amino terminus of nsP2. Deletion of amino-terminal amino acids of nsP2 or addition of extra

Alphaviruses are a group of positive-strand RNA viruses with a worldwide distribution, spread by arthropods between mammalian and avian hosts. Some alphaviruses are capable of causing encephalitis in humans or domestic animals, whereas others may cause fever and arthritis (1). Recently, alphaviruses pathogenic for fish have been discovered (2), indicating that the biological diversity and significance of alphaviruses is greater than previously anticipated. The alphaviruses best studied at the molecular level are Semliki Forest virus (SFV) 1 and Sindbis virus. Because these viruses are capable of efficient replication in a wide variety of cultured cells including neurons, they have been developed into versatile vector systems for protein expression (3,4).
The alphavirus genome is ϳ11.5 kb and is expressed in the form of two large precursors, or polyproteins (reviewed in Ref. 5). The structural proteins (capsid and three-envelope proteins) are translated as a polyprotein from a separate subgenomic 26 S mRNA, which is a copy of the 3Ј-end of the genome arising during the RNA replication process. On the other hand, the nonstructural polyprotein, designated P1234, is translated directly from the viral genomic RNA. It is processed into its individual components, the nonstructural proteins nsP1-nsP4, through a highly specific thiol protease activity present in the carboxyl-terminal portion of nsP2 (6 -9). The nsPs possess enzymatic and other functions needed for virus RNA replication. NsP4 is the catalytic RNA-dependent RNA polymerase subunit, and nsP3 is an evolutionarily conserved protein of unknown function (5). The amino-terminal domain of nsP2 has NTPase, RNA helicase, and RNA triphosphatase activities (10 -12). NsP1 is a guanine-7-methyltransferase and guanylyltransferase (5). The last three activities are needed in the capping of viral mRNAs.
In addition to the final products, some of the processing intermediates of P1234 have distinct and indispensable functions during the replication process. It has been demonstrated that the complex of P123 with nsP4 produces minus-strand RNA on the parental plus-strand template (13)(14)(15)(16)(17). The minusstrand synthesizing activity is unstable and disappears within 15 min after addition of protein synthesis inhibitors. This is because processing of P123 into its individual subunits changes the specialization of the polymerase complex toward the synthesis of full-length and subgenomic plus-sense RNAs on the minus-strand template. The polyprotein precursor is also indispensable for proper assembly, targeting, and membrane association of the replication complex (18). Given such a rapid timetable and exact requirements, the processing of P1234 must be coordinated precisely. During the early hours of alphavirus infection, nsP4 is the first protein to appear in its mature form after the synthesis of P1234, even though it is the last to be translated (19 -21). Late in alphavirus infection, new replication complexes are no longer generated, because P1234 is apparently cleaved too rapidly to allow their assembly.
To understand the regulation of alphavirus RNA synthesis and polyprotein processing, we have recently expressed and * This work was supported by the Academy of Finland (Grants 8397 and 201687), Biocentrum Helsinki, and Helsinki University Research Funds, as well as the Estonian Science Foundation (Grant 5055). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ An International Senior Research Fellow of the Wellcome Trust. ʈ To whom correspondence should be addressed. Tel.: 358-9-19159400; Fax: 358-9-19159560; E-mail: leevi.kaariainen@helsinki.fi. purified the recombinant carboxyl-terminal protease domain of SFV nsP2 (9). This protease, designated as Pro39 based on its molecular weight, was highly specific, and processed the SFV nonstructural polyprotein at the three authentic sites. The site 3/4 (junction between nsP3 and nsP4) was cleaved very efficiently. However, cleavage of site 1/2 required much higher concentration of Pro39, and cleavage of site 2/3 was detectable only by sensitive mass spectrometric methods (9). As the halflives of these processing intermediates in the infected cell are short, and as the 2/3 cleavage predominates late in infection, we have sought factors that would regulate and enhance these cleavages. The results indicate that the cleavage of site 1/2 takes place in cis and must precede that of the 2/3 site. The amino-terminal sequence of nsP2, liberated by the 1/2 cleavage, acts as an activator for the 2/3 cleavage. These results provide an insight into the mechanisms, based on the control of polyprotein processing, that regulate viral RNA synthesis. Although the pathway itself may be unique for alphaviruses, the strategy of using proteolytic processing to activate replication components by releasing them from polyprotein precursor is common for many positive sense RNA viruses and reversetranscribing viruses.

EXPERIMENTAL PROCEDURES
In vitro Translation and Protease Activity Assays-Coupled in vitro transcription-translation was carried out with the T7 TNT rabbit reticulocyte lysate system (Promega) according to manufacturer's instructions. The reactions were supplemented with 15 Ci of [ 35 S]methionine (Ͼ1000 Ci/mmol; Amersham Biosciences) and 1 g of plasmid DNA/25 l of reaction, and they were incubated for 1 h at 30°C unless otherwise indicated. The reaction products were separated by SDS-PAGE in 10% gels and visualized by autoradiography using phosphorimaging (Fuji BAS-1500).
To test the proteolytic activity of nsP2-containing polyproteins in trans, substrate polyproteins with a mutation in the protease active site (P12 CA , P2 CA 3, P12 CA 3, P12 CA 34) were used. The substrates were synthesized by in vitro translation as described above in the presence of radioactive methionine. The protease-containing polyproteins were synthesized in the presence of 20 M unlabeled methionine and without [ 35 S]methionine. Protein synthesis was stopped by adding 1 mM cycloheximide, and the protease-containing polyproteins were mixed with the substrate polyproteins at 3 to 1 ratio and incubated for an additional 1 h at 30°C.
Mutagenesis of Protease Recognition Sites-Site-directed mutagenesis was used to change the protease cleavage sites in SFV nonstructural polyproteins into inactive mutant forms (9). To change the site between nsP1 and nsP2 from HAGA2GVVE to HAEA2GVVE, the site between nsP2 and nsP3 from TAGS2APSY to TAES2APSY, and the site between nsP3 and nsP4 from RAGA2YIFS to RAEV2YIFS, specific primers containing the desired mutations were used (bolded residues highlight the mutations made at the cleavage site). 2 The templates were clones P1234, P123, P23, and P12. All mutations were verified by sequencing, and the mutated clones were designated as P1 ∧ 234, P12 ∧ 34, P123 ∧ 4, P1 ∧ 23, P12 ∧ 3, P1 ∧ 2, and P2 ∧ 3 ( ∧ indicates the mutated site). To introduce combinations of several mutations into P1234 and P123, additional rounds of mutagenesis were performed.
Deletion Constructs-Serial deletions in the nsP1 region were introduced into P1 ∧ 23 polyprotein using a single reverse primer and a set of forward primers. PCR products were digested by NcoI, blunted with Klenow polymerase, and digested with SacI and cloned into BamHI (blunted with Klenow polymerase) and SacI-linearized plasmid P1 ∧ 23. Recombinant clones were verified by sequencing and designated as ⌬273, ⌬373, and ⌬473, respectively (Fig. 6A). Deletion ⌬373 was also introduced into wild type polyprotein P123 and designated ⌬A (Fig. 4A).
The construct containing a 20-aa deletion at the amino terminus of P23, designated P(⌬20)23 (Fig. 6A), was made using a primer at the new amino terminus and an internal primer within nsP2. The PCR product was digested by NcoI, blunted with Klenow polymerase and digested with NheI, and cloned into BamHI (blunted with Klenow polymerase) and NheI-linearized plasmid P23.
To introduce serial deletions into the helicase domain of nsP2 in P123, the plasmid P123 was used as a template and a primer matching the start of the protease domain (AUG at genomic position 3068 -3070) was used as a constant reverse primer together with a set of forward primers. The products of mutagenesis reaction were cloned and verified by sequencing. Clones containing the correct deletions were designated as ⌬398, ⌬338, ⌬158, and ⌬88, respectively (Fig. 5A). The longest deletion in the helicase domain (⌬438) was designated ⌬B (Fig. 4A).
Clones containing deletions in the region immediately upstream of the cleavage point were constructed using P1 ∧ 23 (with the mutated protease site sequence) as template and a primer starting from the first base coding for nsP2 as one of the primers together with a set of primers. The products of mutagenesis reaction were cloned and verified by sequencing. Other deletions in the region of cleavage site between nsP1 and nsP2 were constructed using P1 ∧ 23 as a template and a primer corresponding to the Ϫ55 aa with respect to the cleavage site as a constant primer. Correct clones were designated as shown in Fig. 6C.
Expression and Purification of nsP2 Variants-To obtain nsP2 protein with an authentic amino terminus (designated here as nsP2) the coding sequence of nsP2 was PCR amplified using plasmid P123 as template. The PCR product was digested with NdeI and XhoI, cloned into pET32b vector (Novagen), and verified by sequencing. The gene encoded an extra sequence, LEHHHHHH, at the carboxyl terminus. Amino-terminally tagged ExtA-nsP2 protein contained the sequence MNTIHHHHHHNTSSATM at the amino terminus (10). The proteins were expressed in Escherichia coli BL21(DE3) and purified with successive chromatographies on SP-Sepharose FF (Amersham Biosciences), nickel-nitrilotriacetic acid (Qiagen), and heparin-Sepharose FF (Amersham Biosciences). Details of the purification procedure will be described elsewhere. The concentration of the purified proteins was determined with a Bio-Rad protein assay kit with bovine serum albumin as a standard. The truncated protease domain of Pro39 was expressed and purified as described previously (9). To assay for protease activity, 2 l of in vitro synthesized [ 35 S]polyprotein substrate was mixed with 5 or 0.5 pmol of the proteases in a total volume of 25 l in 20 mM HEPES, pH 7.5, 133 mM NaCl, and 1 mM 1,4-dithioreithol and incubated for 1 h at 30°C.

In Vitro Synthesis of Polyproteins Mutated at the Cleavage
Sites-We have shown previously that the purified recombinant protease domain of SFV nsP2, Pro39, efficiently cleaves site 3/4 but poorly cleaves sites 1/2 and 2/3 in the SFV nonstructural polyprotein. This was shown with purified substrates in which the individual cleavage site sequences were joined to thioredoxin carrier as a fusion protein (9). We concluded that Pro39 was sufficient to cleave efficiently the 3/4 site but that the optimal cleavage at sites 1/2 and 2/3 might require additional factor(s). To carry out this study, we used in vitro translation to produce the large precursor proteins in the rabbit reticulocyte lysate system. Numerous previous studies have shown that wild type alphavirus polyproteins are efficiently synthesized and cleaved at all sites during in vitro translation (Refs. 7 and 22 and references therein), in contrast to the activity of Pro39. Thus, factors influencing the cleavage of the 1/2 and 2/3 sites may reside outside of the core protease domain.
We first produced a series of polyprotein mutants containing noncleavable proteolytic sites, marked with the symbol ∧ to indicate that the proteins remain obligatorily connected (e.g. P1 ∧ 23, P1 ∧ 234, P12 ∧ 3, etc.). The 1/2 and 2/3 sites were inactivated by changing the penultimate residue preceding the cleavage site (Gly to Glu). The 3/4 site had an additional mutation of the residue preceding the cleavage site (Ala to Val). These mutations are known to abolish the proteolysis catalyzed by recombinant SFV protease Pro39 (9).
To follow their autoproteolytic processing, the mutant pro-teins were synthesized in vitro in the presence of [ 35 S]methionine and analyzed by SDS-PAGE. The identity of the cleavage products was verified by immunoprecipitation with monospecific antisera (Fig. 1). As expected, the mutated sites were noncleavable in this system; P123 ∧ 4 gave rise to nsP1, nsP2, and P34 (Fig. 1A, lanes 1-5), whereas P12 ∧ 3 ∧ 4 yielded nsP1 and P234 (lanes 6 -10). In these experiments nsP1 was seen as a double band, due to aberrant initiation at a downstream methionine (Fig. 1A, lanes 2 and 7; see Ref. 22). Polyproteins containing mutations in all of the cleavage sites (P1 ∧ 2, P1 ∧ 2 ∧ 3, and P1 ∧ 2 ∧ 3 ∧ 4) were not processed, as shown for P1 ∧ 2 ∧ 3 ∧ 4 ( Fig.  1A, lanes [11][12][13][14][15]. Interestingly, polyproteins containing the noncleavable 1/2 site were not processed at the 2/3 site either (Fig. 1B); P1 ∧ 23 was not processed at all (lanes 1-4), whereas P1 ∧ 234 yielded P123 and nsP4 (lanes 6 -9). Mutations at the 3/4 and 2/3 sites did not prevent processing at the other remaining sites (Fig. 1A, and data not shown). According to these experiments, cleavages at the 1/2 and 3/4 sites are independent of the other sites, whereas proteolysis at the 2/3 site takes place only after the 1/2 cleavage. Therefore, given the generally rapid proteolysis of the 3/4 site, the proteolysis of the SFV polyprotein appears to proceed in the order 3/4, 1/2, and finally 2/3, yielding the intermediates P123 ϩ nsP4 and nsP1 ϩ P23 ϩ nsP4, respectively. Activity of Polyprotein Proteases in Trans-To better understand the regulation of the order of processing steps described above, we studied the ability of all of the putative polyprotein proteases to process the different sites in trans. To this end we produced radioactively labeled polyprotein substrates by in vitro translation, in which the nsP2 protease was inactivated by substituting the critical Cys 478 residue with Ala (P12 CA , P2 CA 3, P12 CA 3, and P12 CA 34). These proteins cannot process themselves but are substrates for different proteases provided in trans (22). After 60 min of incubation, protein synthesis was stopped by the addition of 1 mM cycloheximide followed by the addition of unlabeled protease from another translation mixture ( Fig. 2A). As the cleavable polyprotein proteases are also processed autocatalytically, they release proteolytically active intermediates and finally nsP2. For instance, in the example shown schematically in Fig. 2A, P123 could generate proteo-lytically active P12, P23, and nsP2 species. To obtain uncleaved polyprotein proteases we used proteins with cleavage site mutations similar to those shown in Fig. 1.
When P12 CA 3 was used as a substrate for different proteases, it was processed in trans at the 2/3 site but not at the 1/2 site, yielding P12 and nsP3 (Fig. 2B, lanes 1-9). The incubation of P12 CA 3 with nsP2 alone, synthesized in vitro, gave the same pattern; the 2/3 but not the 1/2 site was processed (lanes 10 -12). As nsP1 was efficiently released from wild type polyproteins ( Fig. 1), this suggests that the 1/2 site is normally cleaved only in cis. This result was repeated when P12 CA was used as a substrate; it was not processed in trans by any of the polyprotein proteases or nsP2 (data not shown). When proteases with noncleavable 2/3 site, for example P12 ∧ 34 or P2 ∧ 3, were used to cleave P12 CA 3, quantitative cleavage at site 2/3 was achieved ( Fig 2B, lanes 13-18), whereas proteases with a noncleavable 1/2 site (P1 ∧ 2, P1 ∧ 23, P1 ∧ 2 ∧ 3) were unable to process P12 CA 3 at all (Fig. 2C, lanes 1-6). These results indicate that nsP1 in the polyprotein protease somehow prevents the in trans cleavage at the 2/3 site of the polyprotein substrates. In contrast, all polyprotein proteases with a noncleavable 1/2 site, such as P1 ∧ 23 ∧ 4 and P1 ∧ 2, were able to cleave the 3/4 site of P34 (Fig.  2D, lanes 1-5), and P12 CA 34 substrates (not shown). 3/4 is the only site within the SFV nonstructural polyprotein that is processed efficiently also by the carboxyl-terminal protease domain of nsP2, Pro39 (9). Thus, we conclude that the protease can only process the 2/3 site after the removal of nsP1, whereas the 3/4 site can be processed by all polyprotein proteases, native nsP2 as well as its truncated protease domain, Pro39.
Kinetics of the Synthesis and Processing of P12 and P23 in Vitro-The time course of synthesis and processing of SFV nonstructural polyproteins P12 and P23 was studied in the cell-free translation system. A detailed time-course analysis showed that full-sized P12 was first detected after 25 min of incubation at 30°C and that the first cleavage products, nsP2 and nsP1, were detectable somewhat later, starting at 40 min. Some full-sized P23 was also detectable after 30 min of translation, accompanied by its cleavage products, nsP3 and nsP2. The latter was present as a double band. There was no increase in the amount of P23 upon longer incubation, suggesting that it was cleaved rapidly (data not shown).
To study processing further, the synthesis of P12 was allowed to continue for 30 min; thereafter elongation was inhibited by the addition of cycloheximide, and the incubation was continued (Fig. 3A). Part of the incubation mixture was diluted with fresh lysate in a 1 to 5 ratio (Fig. 3B) and another part in a 1 to 50 ratio (Fig. 3C) followed by incubation for up to 120 min in the presence of the inhibitor. In both diluted samples the amount of P12 continued to decrease during the observation period at about the same rate as in the undiluted sample. Simultaneously the amount of nsP2 and nsP1 increased. This result shows that the processing of P12 is an intramolecular reaction, i.e. it takes place in cis. We estimated the half-life of P12 in the diluted samples, shown in Fig. 3, B and C, by quantitating the radioactivity in P12 and nsP2 bands by phosphorimaging. Based on values between 30 and 60 min of incubation, the half-life was 25-30 min.
A dilution experiment was carried out also for the translation reaction of P23. After 30 min (Fig. 3D), protein synthesis was allowed to continue after a 5-fold (Fig. 3E) and 50-fold dilution (Fig. 3F) with new lysate in the presence of the same concentration of [ 35 S]methionine as in the original reaction mixture. It was necessary in this case to allow further protein synthesis, to produce more precursor P23, as it otherwise would have been difficult to detect. As expected, there was an increase of radioactivity during the incubation. In the undiluted mixture, mature nsP2 and nsP3 accumulated, whereas very little precursor was visible (Fig. 3D). However, the sample that was diluted 50-fold showed an increase of P23, indicating that cleavage of the polyprotein was stopped by the dilution of the reaction mixture (Fig. 3F). This result is in agreement with the idea that cleavage at site 2/3 is an intermolecular event, i.e. it takes place in trans.
Localization of Sequences Affecting Processing of Sites 1/2 and 2/3-To understand the influence of regions outside the core protease domain on the polyprotein processing at 1/2 and 2/3 sites, we made deletions in P123 as shown in Fig. 4A. From the amino terminus of nsP1, 373 residues were deleted giving construct P123⌬A. An internal deletion of 438 aa, starting from residue 21 within the helicase domain of nsP2, resulted in construct P123⌬B (Fig. 4A). The deletion constructs were analyzed by in vitro translation in the presence of [ 35 S]methionine as before, and the cleavage products were identified by immunoprecipitation. The wild type P123 served as controls as well as respective constructs P1 ∧ 23, P1 ∧ 23⌬A, and P1 ∧ 23⌬ (Fig. 4B, lane 2), which were unable to process the two sites (Fig. 4B,  lanes 1, 3, and 8, respectively).
Construct P123⌬A was processed almost completely yielding full-sized nsP2 and nsP3 (Fig. 4B, lanes 6 and 7, respectively) with some 106-kDa intermediate precipitating with antibodies against nsP1 and nsP2 (lanes 5 and 6). Construct P123⌬B was processed efficiently because no 153-kDa precursor, seen in the control (Fig. 4B, lane 8), was observed (lanes 9 -12). However, only nsP1 was released from the truncated polyprotein (Fig.  4B, lanes 9 and 10), whereas the ϳ90-kDa protein was precipitated with antisera against nsP2 and nsP3, representing Pro39-nsP3 fusion protein (lanes 11 and 12). The combination of deletions ⌬A and ⌬B in the same molecule reproduced these results ( Fig. 4A and data not shown). We conclude from these experiments that the in cis cleavage of site 1/2 took place with all constructs, indicating that 164 carboxyl-terminal residues from nsP1 and 20 residues from the amino terminus of nsP2 were sufficient for this cleavage, together with Pro39. In contrast, deletion ⌬B in nsP2 prevented the cleavage at site 2/3.
To localize the sequences necessary for site 2/3 cleavage at the amino terminus of nsP2, a series of internal deletions was generated in the helicase domain of nsP2 within the P123 polyprotein (Fig. 5A). All of the deletions ended at amino acid position 458 (the beginning of the proteolytic domain), started at aa residues 60, 120, 300, and 370 of nsP2, and were designated as ⌬398, ⌬338, ⌬158, and ⌬88. When these polyproteins were synthesized in vitro in the presence of [ 35 S]methionine, they were all processed at the 1/2 site releasing nsP1 (Fig. 5B,  lanes 1-5) and thus confirming that only a short nsP2 sequence is sufficient for the in cis cleavage at this site. The cleavage at site 2/3 could be monitored by the release of nsP3 and by the decreasing molecular weight ladder created by nsP2-derived products (Fig. 5B, bands marked by asterisks in lanes 1-4). Deletions ⌬88, ⌬158, and ⌬338 were still processed, whereas ⌬398 yielded very little if any nsP3 (Fig. 5B, compare lanes 4  and 5). We conclude that processing of site 2/3 requires, in addition to the carboxyl-terminal Pro39 protease domain, also amino-terminal sequences from nsP2, possibly up to residue 120.
Mapping of Sequences Influencing Cleavage of Site 2/3-The mutation 1 ∧ 2 inhibiting the cleavage at this site also prevented the processing of site 2/3 (Figs. 1B and 2C), suggesting that nsP1, or some part of it, inhibited the in trans protease activity of the rest of the polyprotein. To study the inhibitory effect of nsP1 on the 2/3 processing, we engineered P1 ∧ 23 polyproteins containing progressive amino-terminal deletions in nsP1 (Fig.  6A). The processing of these proteins was studied by in vitro translation as previously. Truncated proteins lacking 273 aa (⌬273), 373 aa (⌬373), or 473 aa (⌬473) had the phenotype of the full-length P1 ∧ 23, as no cleavage was observed at 2/3 site (Fig. 6B, lanes 2-4). This indicated that the first 473 aa of nsP1 were not involved in the suppression of the 2/3 site proteolysis and that the 64 carboxyl-terminal aa of nsP1 were sufficient for the "inhibitory effect." P23 by itself was processed very rapidly, as shown previously (Fig. 3E). However, complete deletion of nsP1, together with a deletion of 20 residues from the amino terminus of nsP2, also prevented the cleavage at 2/3 site (Fig.  6B, lane 5), suggesting that the amino terminus of nsP2 might be important for processing of 2/3.
A series of internal deletions at the junction region between nsP1 and nsP2 (Fig. 6C) revealed that all deletions involving the amino-terminal residues of nsP2 resulted in polyprotein proteases being unable to cleave the site 2/3 (Fig.  6D, lanes 7-9). Very inefficient cleavage could be seen with internal deletions ending close to or exactly at the carboxyl terminus of nsP1 (Fig. 6D, lanes 3-6), including deletion of those 64 aa (lane 5) that previously were sufficient for full inhibition. Similar results were obtained in strict transcleavage assays using separately synthesized, labeled P2 CA 3 as substrate (not shown). Altogether these results suggested that there were no specific "inhibitory sequences" in nsP1. Rather, the important effect was the release of the amino terminus of nsP2.
Role of the Amino Terminus of nsP2 in the Processing of Site 2/3-Next we studied the role of the amino-terminal sequences in the cleavage of the 2/3 site using labeled P2 CA 3 as a substrate and unlabeled nsP2 constructs with modifications at or close to the amino terminus as proteases. Both were translated in vitro before combining for the assay. Active P23 protease (Fig. 7A, lane 2) and its inactive variant, P2 CA 3 (lane 1), served as controls. Authentic nsP2 cleaved the substrate efficiently even in a 1 to 10 dilution (Fig. 7A, lanes 3 and 4), whereas undiluted ExtA-nsP2 with 17 extra amino-terminal residues was able to cleave only some of the substrate (lanes 5 and 6). Amino-terminal extensions in nsP2, derived from nsP1 sequences ("native" extension ExtB with cleavage site mutation or extension ExtC with unrelated nsP1 sequences) also efficiently inhibited site 2/3 cleavage (lanes 7-10). Thus, all additional sequences at the amino terminus of nsP2 led to the inhibition of 2/3 cleavage. Deletion of 8 residues from the amino terminus of nsP2 protease gave similar inhibition (not shown). The amino terminus of nsP2 was required only in the protease but not as part of the substrate, because both P123 and P23 cleaved the site 2/3 of P(⌬ 20)2 CA 3 substrate, which had a deletion of 20 residues at the amino terminus of nsP2 (not shown).
Finally, we produced nsP2, ExtA-nsP2, and Pro39 in E. coli and purified them (Fig. 7B). Amino-terminal sequencing confirmed that nsP2 had the authentic sequence and that ExtA-nsP2 had the 17 extra residues as expected from the construct (10). Radioactively labeled P2 CA 3 was used as a substrate for all three proteins (Fig. 7C). 0.5 pmol of nsP2 with an authentic amino terminus cleaved the substrate (Fig.  7C, lane 3), whereas 5 pmol of ExtA-nsP2 was only sufficient for partial cleavage (lanes 5 and 6), and no cleavage was detected with Pro39 at these concentrations. All three proteases cleaved labeled P34 substrate at the lower dose of 0.5 pmol (Fig. 7D, lanes 2-4), but they cleaved labeled P12 CA very poorly even at the higher concentration (Fig. 7D, lanes  7-9), in accordance with results showing that P12 is normally cleaved in cis (Fig. 3, A-C). Thus the free, authentic amino terminus of nsP2 is necessary to obtain an effective cleavage of the site 2/3, but its presence or absence does not influence the cleavage at site 3/4.

DISCUSSION
The RNA synthesis of alphaviruses takes place on the surface of specific cytoplasmic vacuoles, which are modified endosomes and lysosomes. The surface of the vacuoles consists of small invaginations or spherules, which seem to be the actual sites of RNA replication (5,23,24). The virus-encoded components of the late, stable, RNA-polymerase complex, synthesizing exclusively positive-stranded RNAs, consist of four nonstructural proteins, nsP1-nsP4, which are derived from a large polyprotein, P1234. Early in infection the synthesis of complementary RNA strands is catalyzed by partial cleavage products of the nonstructural polyprotein, P123 ϩ nsP4 (Fig. 8A) (15,16,25). In this study we have analyzed the proteolytic processing pathway of the ns polyprotein, which regulates the synthesis of the short-lived complementary RNA, as well as membrane association and targeting of the replication complex (18). FIG. 4. Deletion mapping of P123 sequences affecting 1/2 and 2/3 site cleavages. A, deletions in nsP1 (aa 1-373, termed ⌬A) and nsP2 (aa 21-457, termed ⌬B) are indicated in the scheme at the top. Processing at sites 1/2 and 2/3 is marked by a ϩ or a Ϫ sign. Corresponding deletion constructs with the cleavage site 1/2 mutation (P1 ∧ 23) were used as controls. B, these constructs were translated in vitro as described in the Fig. 1 legend, and the products were either analyzed directly or subjected to immunoprecipitation with antisera against nsP1, nsP2, or nsP3 as indicated. Lane 2 shows the processing products of P123 wild type protein and lane 1 the full-length precursor generated by translation of P1 ∧ 23.

Three Activities Involved in the Processing of SFV ns Polyprotein
Site 3/4 Is Cleaved by All nsP2 Proteases-Several previous studies have suggested that the 3/4 site between nsP3 and nsP4 is cleaved first in the processing pathway of P1234 (19,20). NsP4 was the first mature protein to appear when the processing of synchronously initiated SFV ns-polyproteins was studied in infected cells. The purified recombinant protease domain, Pro39, cleaves site 3/4 efficiently in P1234 and P34 substrates as well as in thioredoxin-tagged 3/4 junction sequence (9). In the present work we have shown that all stable polyprotein proteases were able to cleave the substrate polyproteins readily at the 3/4 site (Fig. 2D). Thus, we conclude that this site is processed by a "core activity" of the SFV-specific protease, present in the Pro39 domain of nsP2 and in all polyproteins in which it is contained. Ample evidence demonstrates that this cleavage can take place in trans. However, the 3/4 site cleavage may also take place in cis (7). This might happen preferentially in the early stages of virus infection when the concentration of polyproteins is low.
Site 1/2 Is Cleaved in Cis-Based on coexpression of SFV ns polyproteins in insect cells, we predicted that cleavage at site1/2 takes place in cis rather than in trans (22). Here, we have shown that none of the polyprotein proteases can cleave in vitro site 1/2 in trans of any of the polyprotein substrates (P12 CA 34, P12 CA 3, or P12 CA ) (Fig. 2). Moreover, direct evidence that the 1/2 site cleavage occurs in cis was obtained from the observation that dilutions did not prevent 1/2 site processing (Fig. 3, B and C). Thus, we conclude that this site in SFV nonstructural polyprotein is cleaved in cis. For Sindbis virus polyproteins, some trans-cleavage at this site was observed in in vitro translation experiments, but it was very  inefficient (21). The requirement for cis-cleavage also explains why recombinant proteases cleave this site very poorly in trans (Fig. 7D) and why a huge excess of recombinant Pro39 over the substrate was needed to achieve complete cleavage of the 1/2 site (9).
Deletion analysis revealed that 164 aa from the carboxyl terminus of nsP1, in addition to 20 residues from the amino terminus of nsP2, together with the core protease domain Pro39, were sufficient for site 1/2 cleavage (Fig. 4). As the 20 aa from nsP2 are not sufficient to support the site 2/3 cleavage, these two sites require different factors in addition to the Pro39. These results are supported by previous deletion mapping studies carried out for Sindbis virus nonstructural polyprotein. It was shown that an internal deletion of 172 aa in nsP1, which left only eight authentic carboxyl-terminal residues preceding cleavage site 1/2, allowed the processing at this site (7). Altogether, this suggests that a very short stretch of sequence is sufficient for cis cleavage at the 1/2 site.
Special Requirements of Cleavage at 2/3 Site-The dilution experiment strongly suggested that the cleavage of site 2/3 is an intermolecular reaction and cannot take place in cis. (Fig. 3, E and F) When different polyprotein intermediates were synthesized alone in vitro, no processing of the site 2/3 took place if the cleavage of 1/2 site was prevented by mutation, as in FIG. 7. Requirement of the native amino terminus of nsP2 for the processing of the 2/3 site. A, the in vitro produced proteases, nsP2 with authentic amino terminus, and several proteases with modified amino terminus: ExtA-nsP2 containing a 17-aa extension (MNTI-HHHHHHNTSSATM-nsP2), ExtB-nsP2 containing 21 aa from the nsP1 carboxyl terminus with the 1/2 site mutated (MAPAETGVVDVD-VEELEYHAEA-nsP2), and ExtC-nsP2 with 21 aa from the nsP1 sequence 461-482 region (MAIPVRSRIKMLLAKKTKRELI-nsP2) were incubated with radioactively labeled P2 CA 3 substrate either directly or after a 1 to 10 dilution. Processing was monitored by the appearance of the cleavage products nsP3 and nsP2, the latter migrating as a double band apparently because of initiation at a downstream methionine. B-D, processing of in vitro produced substrates by recombinant proteases isolated from E. coli. B, SDS-PAGE analysis of purified proteases nsP2 with authentic amino terminus, ExtA-nsP2, and Pro39. Lane M shows a molecular mass marker in kDa. C, [ 35 S]P2 CA 3 substrate was incubated with 0.5 or 5 pmol of the purified proteases for 1 h at 30°C. Cleavage of P2 CA 3 was monitored as described in Fig. 6 legend. D, proteolytic reactions carried out as in C but with substrate [ 35  Once nsP1 has been released, P23 is cleaved rapidly in trans. B, in the late processing pathway, free nsP2 first cleaves site 2/3 giving rise to shortlived precursors P12 and P34 (19,20,35,36). P12 is cleaved in cis and P34 in trans as shown in this article. NsP1 is directed to the plasma membrane, nsP2 to the nucleus, nsP3 to cytoplasmic aggregates, and nsP4 to the proteasomes (5,10,18,22,24,37,38). polyproteins P1 ∧ 234, P1 ∧ 23, and P1 ∧ 23 ∧ 4. In contrast, processing of site 2/3 took place in polyproteins P1234, P123, and P123 ∧ 4 in which the cleavage of 1/2 site takes place (Fig. 1). Similar results have been reported for Sindbis virus nonstructural polyprotein (26). Thus, the release of the amino terminus of nsP2 seemed to be necessary for the further processing of the ns polyprotein P123. The fact that the isolated protease domain Pro39 cleaved the site 2/3 extremely inefficiently (9), in contrast to authentic nsP2 (Fig. 2B, lanes 10 -12), suggested that nsP2 sequences beyond the protease domain would be required as a "cofactor" for this cleavage. Successive deletion of sequences within the amino-terminal helicase domain suggested that the putative cofactor would be located within the first 120 amino-terminal residues of nsP2 (Fig. 5). Deletion of 5-20 amino-terminal amino acids of nsP2 protease (Fig. 6, B and D) or extra amino acids joined to the amino terminus of nsP2 (Fig.  7A) prevented the cleavage of site 2/3, indicating that the exact amino terminus was one of the requirements of site 2/3 cleavage. This observation was supported by experiments with purified nsP2 containing an authentic amino terminus as well as ExtA-nsP2 with 17 extra residues at its amino terminus (Fig.  7). Thus, site 2/3 cleavage also requires, in addition to 20 -120 residues from the helicase domain, an exact amino terminus, which is released when nsP1 is cleaved from the polyprotein.
Activation of Viral Proteases-We have shown that authentic nsP2 or polyprotein intermediates from which the amino terminus of nsP2 domain has been released through cleavage of nsP1 are highly active for the cleavage of the 2/3 site, whereas Pro39 is extremely inefficient in cleaving it. These results indicated that the amino-terminal helicase domain of nsP2 must harbor an activator or cofactor for the site 2/3 cleavage. It seems that the exact amino terminus plays a vital role in the activation process. Notably, the activation of nsP2 is restricted toward only one of the cleavage sites (Fig. 7, C and D). The activator sequence at the amino terminus of nsP2 may interact with the protease domain to alter its conformation. On the other hand, it is possible that the activator interacts with the 2/3 site in the substrate to render it more accessible for cleavage.
Other viral proteases also undergo activation by various means. The best understood example is hepatitis C virus protease (27). The protease is located in the amino-terminal part (aa 1-181) of the NS3 protein, which consists of 631 aa residues. The remainder of NS3 forms an NTPase/RNA helicase domain (28). The amino terminus of NS3 is cleaved from a large polyprotein by NS2 protease. The carboxyl-terminal polyprotein consists of NS3, NS4A, (54 aa), NS4B, NS5A, and NS5B, the last being the catalytic subunit of the RNA-polymerase complex. NS3 is responsible for all the downstream cleavages starting from the junction NS3/4A. The isolated protease domain of 181 residues can cleave the NS5A/5B site in trans, resembling the situation of 3/4 cleavage in the alphavirus nonstructural polyprotein. Extensive structural, genetic, and biochemical studies revealed that a cofactor was needed for the proper cleavage of the other sites in the polyprotein (29 -32). The cofactor was present within NS4A protein (residues 21-39), which interacted with NS3 (31,33). In analogy with our SFV results, site NS3/4A is processed in cis (compare with nsP1/nsP2). The cofactor derived from NS4A associates with the protease domain of NS3, which makes it competent to cleave efficiently sites NS4A/4B and NS4B/5A, resembling the cleavage between nsP2 and nsP3 in SFV polyprotein processing. The flavivirus NS3 Pro requires the NS2B protein as a cofactor instead of NS4A. Closer mapping has shown that the activator resides within a 40-aa region (27). An important similarity between hepatitis C virus and flaviviruses is that both NS4A and NS2B are associated with membranes, and through interaction with the otherwise soluble NS3, they attach the proteolytic events and RNA synthesis to cytoplasmic membranes.
Because alphavirus nsP1 is membrane-bound (34), the cis cleavage of site 1/2 is also a membrane-associated process. We propose that the slow cleavage of site 1/2 enables the folding of the polyprotein so that protein-protein interactions between the components can take place before proteolytic processing (18). We further propose that the active membrane-associated replication complexes can form only through this polyprotein pathway, which includes as an obligatory step the synthesis of complementary RNA strand. Thereafter proteolytic processing in situ converts it to the stable polymerase, which synthesizes plus-strand RNAs continuously. Later in infection, when the concentration of free nsP2 is high, cleavage of site 2/3 is favored, as shown by the appearance of short-lived precursors P12 and P34 in infected cells (Fig. 8B) (35,36). At this point the possibility of creating functional replication complexes is lost. P12 is cleaved in cis giving rise to nsP1, which is targeted to the plasma membrane (37), and nsP2, which is transported to the nucleus (38). P34 is cleaved by free nsP2 resulting in the aggregation of nsP3 in the cytoplasm (18), whereas nsP4 is rapidly degraded by the ubiquitin pathway ( Fig. 8B) (20,39).