Identification of a Novel Function of the AlphavirusCapping Apparatus

Both genomic and subgenomic RNAs of theAlphavirus have m7G(5′)ppp(5′)N (cap0 structure) at their 5′ end. Previously it has been shown thatAlphavirus-specific nonstructural protein Nsp1 has guanine-7N-methyltransferase and guanylyltransferase activities needed in the synthesis of the cap structure. During normal cap synthesis the 5′ γ-phosphate of the nascent viral RNA chain is removed by a specific RNA 5′-triphosphatase before condensation with GMP, delivered by the guanylyltransferase. Using a novel RNA triphosphatase assay, we show here that nonstructural protein Nsp2 (799 amino acids) of Semliki Forest virus specifically cleaves the γ,β-triphosphate bond at the 5′ end of RNA. The same activity was demonstrated for Nsp2 of Sindbis virus, as well as for the amino-terminal fragment of Semliki Forest virus Nsp2-N (residues 1–470). The carboxyl-terminal part of Semliki Forest virus Nsp2-C (residues 471–799) had no RNA triphosphatase activity. Replacement of Lys-192 by Asn in the nucleotide-binding site completely abolished RNA triphosphatase and nucleoside triphosphatase activities of Semliki Forest virus Nsp2 and Nsp2-N. Here we provide biochemical characterization of the newly found function of Nsp2 and discuss the unique properties of the entire Alphavirus-capping apparatus.

Semliki Forest virus (SFV) 1 is member of the Alphavirus genus of the Togaviridae family. SFV has a positive-stranded 42 S RNA genome of 11.5 kilobases. RNA replication of SFV takes place in the cytoplasm and is catalyzed by the viral RNA-dependent RNA polymerase, which contains the virusspecific proteins Nsp1-4. These are cleavage products of a large (2342 aa) nonstructural polyprotein P1234. In the RNA polymerase complex all Nsps are in close association with each other (1,2). The parental 42 S RNA is copied to complementary minus-strands, which in turn are used as templates for the synthesis of new 42 S RNA plus-strands and subgenomic 26 S mRNAs. During plus-strand synthesis, the 5Ј ends of the 42 S and 26 S RNAs become modified with covalently attached m 7 GpppA (the cap0 structure) (3)(4)(5). Capping of the RNAs is believed to be obligatory also for the replication of Alphavirus, since a point mutation specifically destroying the guanylyltransferase activity of Nsp1 is lethal for the virus (6).
The functions of Alphavirus Nsps in replication have been studied using various genetic and biochemical approaches (1,2). Nsp4 is the catalytic component of the RNA polymerase (7,8), whereas the functions of the phosphoprotein Nsp3 (9) are poorly defined (10,11). Nsp2 has several distinct functions. It has nucleoside triphosphatase (NTPase) activity at its aminoterminal half (12), which is vital for the virus replication (13). Nsp2 has RNA helicase activity, which utilizes NTP hydrolysis as the energy source (14). The carboxyl-terminal part of the protein is a papain-like protease responsible for the autocatalytic cleavages of the nonstructural polyprotein (2,15,16). The carboxyl-terminal part has a nuclear localization sequence, which is responsible for sequestering of about half of the molecules to the nucleus during infection (17,18). Furthermore, Nsp2 regulates transcription of the subgenomic 26 S RNA (Ref. 19 and references therein). Here we show that Nsp2 has yet an additional activity required for capping of the virus mRNAs.
Capping of cellular mRNAs occurs in the nucleus and comprises four different reactions. RNA 5Ј-triphosphatase removes the ␥-phosphate from the 5Ј end of the nascent RNA molecule (pppRNA 3 ppRNA). Guanylyltransferase reacts with a GTP molecule to form a covalent complex with GMP, which is then transferred from guanylyltransferase to the 5Ј end of RNA to form G(5Ј)ppp(5Ј)NpRNA. Methylation by guanine-7N-methyltransferase yields an RNA molecule with the cap0 structure (m 7 GpppNpRNA). Further methylation by nucleoside-2Ј-Omethyltransferase of the riboses of the penultimate and the adjacent nucleotides yields cap1 and cap2 structures, respectively (20,21).
Unlike cellular mRNAs, the capping of Alphavirus RNAs takes place in the cytoplasm and is carried out by reactions that differ from the nuclear reactions as follows. (i) Nsp1 catalyzes transfer of the methyl group from S-adenosylmethionine to GTP to yield m 7 GTP (methyltransferase reaction), and (ii) a covalent guanylate complex Nsp1-m 7 GMP is formed (22)(23)(24)(25).
Here, we show that Nsp2 of SFV and Sindbis virus possess RNA triphosphatase activity, which is localized to the aminoterminal half of the protein. Comparison with other capping enzymes suggests that the Alphavirus-capping apparatus is a unique complex. nal part of Nsp2 (aa 470 -799), and Sindbis virus Nsp2 were expressed in Escherichia coli BL21(DE3) (Stratagene) using pHAT plasmids encoding amino-terminal His 6 tags (26). The proteins were purified using Ni 2ϩ affinity chromatography as described (12). Protein concentration was determined using the Bradford assay (27) and Bio-Rad reagents. The purity of proteins was verified by SDS-PAGE in 10% gels.
RNA Synthesis-Two different plasmids were used to produce RNA substrates: pGEM3Zf(ϩ) purchased from Promega, and pUC18⍀. Omega (⍀) leader (1-68 nucleotides) of tobacco mosaic virus was first amplified from infectious cDNA clone tobacco mosaic virus 304 (a kind gift by Dr. Kirsi Lehto, University of Turku) using cloned Pfu polymerase (Stratagene) and oligonucleotides 5Ј-CTGAATTCATTTAGGTGA-CACTATAGAATTTTTACAACAATTACCA-3Ј and 5Ј-GGTAGCTGTCT-GTGTCTAGAATATTGTAATTGTAAATAGT-3Ј as upstream and downstream primers, respectively. The upstream primer was designed to contain the SP6 promoter sequence (underlined), and the downstream primer included the SspI site (underlined). The polymerase chain reaction fragment digested with EcoRI and XbaI was then gelpurified and ligated with the EcoRI/XbaI cut vector pUC18 (New England Biolabs). RNAs were prepared by in vitro transcription of pGEM3Zf(ϩ) cut with EcoRI (RNA1) or pUC18⍀ cut with SspI (RNA2). After incubation at 37°C for 1 h, the transcription products were extracted with phenol/chloroform (1:1) and purified by PAGE in 10% gels containing 7 M urea. Radioactive bands were excised, and the RNA was eluted from the gel with a buffer containing 20 mM Tris-HCl, pH 8.0, 300 mM ammonium acetate, and 1 mM EDTA for 3 h at room temperature and precipitated with ethanol. To produce RNAs with lower specific radioactivity, transcription was carried out in 50 l of 120 mM HEPES-KOH, pH 7.5, buffer containing 24 mM MgCl 2 , 1 mM spermidine, 20 mM dithiothreitol, 5 mM each NTP, 50 Ci of either [␥-32 P]GTP or [␣-32 P]GTP, 40 units of RNasin, 80 units of SP6 RNA polymerase, and 5 to 10 g of a DNA substrate ( 28). The mixtures were incubated at 37°C for 2 h, and the reaction was stopped by the addition of 1 unit of DNase RQ (Promega) per 1 g of input DNA template. Incubation was continued for a further 15 min at 37°C. The resultant RNA samples were extracted with phenol/chloroform (1:1) and chloroform, precipitated with 3 M LiCl, and dissolved in sterile water, followed by additional purification of the RNA preparations in Sephadex G25 spin columns (Amersham Pharmacia Biotech). RNA concentrations were measured by absorbance at 260 nm. The purity of the RNAs was controlled PAGE in 10% gel containing 7 M urea.
RNA Triphosphatase Assay-Two different methods were used to assay RNA 5Ј-triphosphatase activity. The first assay was based on the liberation of the ␥-phosphate group from the ␥ -32 P-labeled RNA substrates. In this case, the RNA triphosphatase reaction mixture (40 l) contained 20 mM Tris-HCl, pH 8.0, 1 mM MgCl 2 , 5 mM KCl, 150 mM NaCl, 2 mM dithiothreitol, and 0.2-5 mol of a [␥-32 P]RNA substrate and 0.1-1 pmol of the enzyme. Reactions were carried out for 20 min at 25°C. The reaction products were separated by PAGE as above. The radioactive bands were visualized by autoradiography. Alternatively, the release of radioactive phosphate was measured by phosphoimaging (BAS-1500; Fuji) as the decrease of the radioactivity of the [␥-32 P]RNA band. Identical untreated electrophoresed samples served as controls in each assay.
In the second assay, the terminal GTP at the 5Ј end of the substrate RNA was selectively labeled at the ␣-phosphate position during transcription. The RNA triphosphatase reaction was carried out as above and stopped by phenol/chloroform extraction. The RNA was precipitated with ethanol and dissolved in TE buffer (10 mM Tris HCl, pH 8.0, and 1 mM EDTA). The RNA (100 g/ml) was treated with RNase T1 (100 units/ml) (Roche Molecular Biochemicals) at 30°C to release the 5Ј-terminal guanylate. Aliquots from the Rnase-treated samples were spotted onto polyethyleneimine (PEI)-cellulose thin layer chromatography (TLC) plates (Merck) and developed with 1 M LiCl. Spots of labeled nucleotides were visualized using autoradiography or phosphoimaging analysis.

RESULTS
Nsp2 Is an RNA Triphosphatase-To test possible RNA triphosphatase activity of the SFV nonstructural proteins, Nsp1, Nsp2, Nsp3 and Nsp4 were expressed in E. coli and purified using metal-chelating chromatography as described before (12). Assay mixtures containing serial dilutions of the different Nsps were incubated for 30 min at 25°C in the presence of a 64-nucleotide-long RNA substrate (RNA1). The 5Јterminal G of the RNA was labeled with [␥-32 P]GTP during the transcription. Reaction products were separated by PAGE under denaturing conditions, and the RNA triphosphatase activity was measured as a decrease in the intensity of the labeled RNA band (Fig. 1). Among the four Nsps tested, only Nsp2 was able to remove the label from the substrate RNA1 ( Fig. 1, Nsp2). The assay was also performed with the Nsp2 protein containing a single amino acid change (K192N) in the nucleotide-binding site (Fig. 2). This mutation has been shown to destroy the NTPase activity of Nsp2 (12). The mutant protein (K192N) had no detectable triphosphatase activity when used within the same concentration range as the wild type Nsp2 (Fig. 1, K192N).
To find out which part of the multifunctional Nsp2 harbors the RNA triphosphatase activity, we expressed and purified two halves of the Nsp2 molecule separately. The amino-terminal half Nsp2-N (residues 1-470) had similar RNA triphosphatase activity as the complete Nsp2 (Fig. 1, Nsp2-N), whereas the carboxyl-terminal half Nsp2-C (470 -799 aa) had no activity (Fig. 1, Nsp2-C). As expected, the K192N mutation in the Nsp2-N fragment rendered it inactive (not shown). Finally, it was interesting to know whether Nsp2 of another Alphavirus also possessed RNA triphosphatase activity. To this end, we expressed in E. coli and purified Sindbis virus Nsp2 (Nsp2SIN) following the same protocol as for the SFV Nsp2. Also, SIN Nsp2 could remove the ␥-labeled phosphate from RNA1 and from [␥-32 P]GTP (see Fig. 3).
Nsp2 Hydrolyzes only the ␥,␤-Triphosphate Bond at the 5Ј End of RNA-Control experiments were devised to exclude the possibility that the Nsp2 preparation contained nonspecific phosphatase or nuclease activities. In these experiments, we used a 70-nucleotide-long RNA molecule representing tobacco mosaic virus ⍀ leader RNA (RNA2), which has only a single guanylate residue at the 5Ј end. During transcription, the 5Ј . These RNAs were treated with different amounts of Nsp2, and the products were analyzed by PAGE. No decrease in the radioactivity was observed for RNA preparations labeled at the ␣-position (Fig. 4A), whereas reduction of radioactivity was clearly seen in Nsp2-treated RNA preparations labeled at the ␥-position (Fig. 4B). Thus, we conclude that the Nsp2 preparation did not contain any nonspecific nuclease activity, and that Nsp2 released ␥-phosphate but not ␣-phosphate from the 5Ј end of RNA2.
The following experiment was carried out to ensure that Nsp2 cleaves only ␥-phosphate but not ␤-phosphate group. RNA2 labeled at the ␣-position was first exposed to Nsp2. A similar untreated RNA2 preparation served as a control. Both RNA samples were thereafter digested with ribonuclease T1, which can cleave from RNA2 only the 5Ј-terminal guanylate residue. The digestion mixtures were analyzed by thin layer chromatography on PEI plates using unlabeled GMP, GDP, and GTP as markers. The untreated control RNA2 preparation yielded only radioactive GTP, whereas RNA2 that had been exposed to Nsp2 yielded radioactive GDP (Fig. 5). Thus, we conclude that Nsp2 is a genuine RNA triphosphatase that cleaves only the bond between ␥ and ␤ phosphates at the 5Ј end of the substrate RNA molecule.
Enzymatic Properties of Nsp2 Triphosphatase of SFV -The RNA triphosphatase activity was optimized for Nsp2-N fragment with respect to divalent cation concentration. For comparison, the optimization was also carried out for the NTPase activity of Nsp2-N using ␥-labeled GTP as the substrate. The optimal MgCl 2 concentration for the GTPase reaction was 5 mM, and only a slight decrease of activity was observed with higher concentrations (Fig. 6A, open circles). Interestingly, there was a sharp maximum at 0.1 mM for MnCl 2 (Fig. 6A, closed symbols). For RNA triphosphatase, the optimal Mg 2ϩ concentration was 1.0 -2.0 mM, whereas a sharp maximum of activity was obtained with a Mn 2ϩ concentration of 0.1 mM (Fig. 6B). Our results concerning NaCl dependence indicate that monovalent cations stimulate the RNA triphosphatase activity but within a wide concentration range (Fig. 7A). The RNA triphosphatase activity had a wide pH range with an optimum at pH 7-8 (Fig. 7B). To test the effect of temperature, the rates of hydrolysis of [␥-32 P]triphosphate-terminated RNA were determined at 20°C, 25°C, 37°C, and 42°C. The maximum rate of hydrolysis was achieved at 25°C, whereas incubation at 42°C apparently resulted in inactivation of the enzyme (Fig. 7C).
Knowing the optimal conditions for the RNA triphosphatase and NTPase, the apparent K m values were estimated for both reactions. The reaction initial velocity, measured within the first 20 min of incubation, was plotted as a function of the substrate concentration according to the equations by Lineweaver and Burk (Fig. 8). The RNA triphosphatase reaction catalyzed by Nsp2-N reached its maximum at 0.055 M/min, and the apparent K m was 2.99 M (Fig. 8A). K cat of 5.5 min Ϫ1 , determined for the RNA triphosphatase reaction, indicated that the ␥-phosphate from 5.5 molecules of RNA was hydrolyzed/min. A much lower apparent substrate affinity (K m value of 90 M) was obtained for the NTPase activity of Nsp2 (Fig.  8B). The maximal velocity (2.3 M/min) and the kinetic constant K cat (230 min Ϫ1 ) for the Nsp2-N-catalyzed NTPase reac- tion were also calculated from the double-reciprocal Lineweaver-Burk plot. These data indicate that Nsp2-N of SFV had about a 30-fold higher affinity for RNA than for the nucleoside triphosphate. DISCUSSION The role of the nonstructural protein Nsp1 in the capping of the viral RNAs has been well established. First, it was shown that Nsp1 has guanine-7N-methyltransferase activity (29). However, the properties of this enzyme were different from those of previously known capping enzymes. Nsp1 could not methylate an unmethylated cap at the 5Ј end of RNA. Instead, it catalyzed the methylation GTP, dGTP, and GpppG, but not of GpppA, which is the 5Ј dinucleotide of Alphavirus 42 S and 26 S RNAs (22,30). This led us to discover a novel guanylyltransferase activity of Nsp1, which unlike other capping enzymes, forms a covalent complex with m 7 GMP instead of GMP (23,24). As neither of these reactions takes place in uninfected host cells, they are potential targets for development of antiviral drugs (30).
Since removal of ␥-phosphate from the nascent 5Ј end of RNA is an essential step in the capping of mRNAs, we started to look for RNA triphosphatase activity among the virus-specific components of the RNA polymerase. Here we show that of the four nonstructural proteins of SFV, only Nsp2 had RNA triphosphatase activity, which was confined to the amino-terminal half of the protein together with the previously discovered nucleoside triphosphatase activity (12,13). To ensure that the RNA triphosphatase and NTPase activities are properties of the Alphavirus family, we produced and also assayed Nsp2 of Sindbis virus. Both activities, indistinguishable from those of SFV Nsp2, were associated with the Nsp2 of Sindbis virus (Fig.  1, Nsp2SIN, and Fig. 2). Taking into consideration the significant amino acid sequence homology among Alphavirus Nsps, one could suggest that Nsp2 proteins of all Alphavirus have both RNA triphosphatase and NTPase activities.
All viral RNA triphosphatases discovered so far have been shown to have also NTPase activity (Table I). The K m values of the RNA triphosphatase and NTPase activities of vaccinia virus D1 capping protein differ considerably (1 M and 800 M, respectively). Competition between RNA and ATP substrates revealed that ATP inhibited the RNA triphosphatase reaction, indicating that the hydrolysis of the ␥-phosphate of both RNA and ATP takes place in the same reaction center (31). Similar results have been reported for the RNA triphosphatase/NTPase of Reovirus 1 protein (32). Our preliminary competition experiments with SFV Nsp2-N with GTP and RNA substrates showed that the RNA triphosphatase activity was increased rather than decreased in the presence of GTP, strongly suggesting that the NTPase and RNA triphosphatase activities have different reaction centers (data not shown). Mutation in the NTP-binding site (K192N) abolished both activities, suggesting a close connection between them. Different reaction centers for NTPase and RNA triphosphatase have been sug- gested also for the Flavivirus West Nile NS3 protein (68 kDa) (38), which also has RNA helicase activity like Alphavirus Nsp2 (37,39). Different reaction centers may be necessary for NS3 and Nsp2, since RNA helicases utilize the energy released from the hydrolysis of NTPs (14,40). In contrast to the RNA triphosphatase activity, which is needed only in the modification of the 5Ј end of the nascent RNA, RNA helicase activity is required throughout the RNA replication cycle. This may mean that different Nsp2 (and possibly NS3) molecules are involved in these two processes. One part of Nsp2 molecules would be tightly associated with Nsp1 in the capping enzyme complex, and the other part could operate alone, perhaps as a homooligomer (40), in the RNA helicase function. This scenario might gain support from the observation that these two enzymatic activities of Nsp2 differ in respect to the optimal NaCl concentration. RNA triphosphatase is active in a wide concentration range (50 to 300 mM), whereas higher than 100 mM concentrations of NaCl were inhibitory for the RNA helicase function (14).
Triphosphatases have been classified as metal-dependent and metal-independent enzymes according to the influence of divalent cations (41). The RNA triphosphatases from mammals, brine shrimp, and West Nile virus are active in the absence of divalent cations, and addition of them in fact inhibits triphosphatase activity (33,42,43). Interestingly, a mammalian RNA triphosphatase has been shown to have a mechanism for phosphoryl group removal similar to that of proteintyrosine phosphatases. This class of enzymes does not require metal ions (Ref. 44 and references therein). Activities of yeast (41), Reovirus (32), vaccinia virus (21,45), and baculovirus LEF-4 (35,36) RNA triphosphatases are dependent on divalent cations (Table I). In this respect, Alphavirus RNA triphosphatase is similar to the other metal-dependent enzymes. However, no significant homology could be found for the aminoterminal part of Nsp2 with viral or cellular RNA triphosphatases.
Four functions of Alphavirus Nsp2, RNA triphosphatase, NTPase, RNA helicase, and protease activities, are the same as those of Flavivirus NS3. However, there is essentially no homology between NS3 and Nsp2 at the level of amino acid sequences (47). Moreover, Nsp2 is dependent on divalent cations, whereas NS3 is stimulated by EDTA. In addition, the RNA triphosphatase activity of Nsp2 is located at the amino terminus, whereas in NS3 it is in the carboxyl-terminal region of the molecule (37). Thus, our results imply that the Alphavirus RNA triphosphatase is unique among the RNA triphosphatases and exemplifies a novel divalent cation-dependent RNA triphosphatase, which is activated by both Mn 2ϩ and Mg 2ϩ ions.
Recently the three-dimensional structure of the divalent cation dependent RNA triphosphatase Cet1 of Saccharomyces cerevisiae has been solved at high resolution. It reveals a tunnel structure consisting of an eight-stranded antiparallel ␤-barrel into which the 5Ј end of the RNA protrudes. The authors (48) propose a mechanism for the specific hydrolysis of the ␥-phosphate as well as for the binding of Cet1 with the guanylyltransferase subunit (Ceg1). Essential glutamate residues, conserved in baculovirus LEF 4 and vaccinia virus D1 RNA triphosphatases, are involved in the metal binding of Cet1, suggesting structural similarity between these enzymes. Secondary structure prediction for the amino-terminal part (1-360 aa) of Nsp2 suggests that it consists mostly of ␤-strands similarly to Cet1. This region has also clusters of glutamate residues that could have analogous functions to Glu-305, Glu-307, Glu-492, Glu-494, and Glu-496 of Cet 1 (48 -50).
The composition of the capping apparatus differs between different species. The tight association of RNA triphosphatase and guanylyltransferase activities in the same protein seems to be a characteristic feature for most eukaryotic and viral mRNA capping systems. The best characterized vaccinia virus capping enzyme is a complex of two subunits D1 (95 kDa) and D12 (33 kDa). The RNA triphosphatase and guanylyltransferase reactions are carried out by the amino-terminal part of D1, whereas the carboxyl-terminal part together with D12 is responsible for the methyltransferase activity (21). The S. cerevisiae capping apparatus consists of three proteins with RNA triphosphatase (Cet1, 80 kDa), guanylyltransferase (Ceg1, 52 kDa), and methyltransferase (Abd1, 50 kDa) activities (33,34,51). Metazoan species possess a two-component capping apparatus consisting of a bifunctional RNA triphosphatase/guanylyltransferase polypeptide (Hce1, 66 kDa) and a separate methyltransferase polypeptide (Hcm1, 52 kDa). The RNA triphosphatase activity is not accompanied by NTPase activity, unlike other known enzymes (41,52). In the capping reactions catalyzed by such enzyme complexes, the order of reactions is RNA triphosphatase, then guanylyltransferase, and finally methyltransferase, which methylates the unmethylated cap at the 5Ј end of the nascent RNA molecule. According to these results, the Alphavirus capping apparatus consists of two replicase proteins, Nsp2, having the RNA triphosphatase activity, and Nsp1, with combined methyltransferase and guanylyltransferase activities. The guanylyltransferase and methyltransferase reactions take place in reverse order, as compared with capping of cellular and most other viral mRNAs.
The Alphavirus capping apparatus is part of an RNA replicase complex in the cytoplasm of Alphavirus-infected cells. The complex is tightly bound to the surface of modified endosomes and lysosomes (53)(54)(55) due to interaction of Nsp1 with membrane lipids. This interaction is mediated by an amphipathic peptide in the middle of the Nsp1 molecule (46) and strengthened by palmitate chains linked to the carboxyl-terminal cysteine residues 418 -420 (25). Attachment of the RNA polymerase complex to the membranes implies that during the polymerization of the nascent RNA molecule the template RNA must move through the fixed polymerase complex. One of the first events during this process is the capping of the nascent RNA molecule.