The N-Terminal Half of the Brome Mosaic Virus 1A Protein Has RNA Capping-Associated Activities: Specificity for GTP and S-Adenosylmethionine

Virology 259, 200–210 (1999) Article ID viro.1999.9763, available online at http://www.idealibrary.com on

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector The N-Terminal Half of the Brome Mosaic Virus 1a Protein Has RNA Capping-Associated Activities: Specificity for GTP and S-Adenosylmethionine

F. Kong, K. Sivakumaran, and C. Kao1

Department of Biology, Indiana University, Bloomington, Indiana 47405 Received February 8, 1999; returned to author for revision March 29, 1999; accepted April 14, 1999

The N-terminal half of the brome mosaic virus (BMV) 1a replication-associated protein contains sequence motifs found in RNA methyltransferases. We demonstrate that recombinant BMV methyltransferase-like (MT) domain expressed in Esche- richia coli forms an adduct with a guanine nucleotide in a reaction that requires S-adenosylmethionine (AdoMet) and divalent cations. Moieties in GTP and AdoMet required for adduct formation were determined using a competition assay and chemical analogues. In the guanine nucleotide the ribose 2Ј hydroxyl, the triphosphates, the base C6 keto group, and possibly the N1 imine are required. In AdoMet, the methyl group and the ability to transfer a methyl group to guanine nucleotide were demonstrated to be required for adduct formation. The effects of methyltransferase inhibitors on viral RNA synthesis was determined using an in vitro RNA synthesis assay. These results are consistent with the previously reported activities of alphaviral nsP1 methyltransferase protein and identify the chemical moieties required for the BMV methyltransferase activity. © 1999 Academic Press

INTRODUCTION Capping-associated guanylyltransferase and methyltransferase activities have not yet been reported for BMV. The translation and stability of mRNA are enhanced by We have been characterizing the activities of the BMV Ј Ј the addition of a 7-methyl-GMP cap through a 5 -5 replicase for RNA synthesis and viral protein–protein linkage to a transcript (for review, see Shuman and interactions. Using the yeast two-hybrid system, O’Reilly Schwer 1995). Capping of cellular mRNAs occurs in the et al. (1997) reported that 1a residues 1–516 contain a nucleus. However, many viruses replicate in the cyto- domain sufficient for interaction with the corresponding plasm and hence may code for their own capping activ- N terminus of another 1a subunit. Computer-predicted ities. Several viral guanylyltransferase and methyltrans- secondary structures within these 516 residues revealed ferases have been identified and characterized, includ- a high degree of similarity in several alphaviral species ing those from Sindbis virus (Mi and Stollar, 1991), (O’Reilly et al., 1998). Furthermore these secondary struc- Semliki Forest virus (Laakkonen et al., 1994); reovirus tures are similar to the known structures of the DNA (Mao and Joklik, 1991), rotavirus (Pizarro et al., 1991), methyltransferase, HhaI (O’Reilly et al., 1998). In this Bluetongue virus (Le Blois et al., 1992), vaccinia virus communication, we expressed the N-terminal 516 amino (Martin and Moss, 1975, 1976), and others. Viral and acids of the BMV 1a protein in E. coli and assay for a cellular capping enzymes have mechanistically distinct classic guanylyltranserase activity, the formation of a activities. For example, the alphaviral nsP1 capping pro- covalent intermediate with the guanine nucleotide. After tein methylates GTP to form 7-methyl-GTP (m7GTP) prior establishing this activity, we analyzed the chemical moi- to its transfer to viral RNA, whereas the cellular capping eties in GTP and the methyl donor, S-adenosylmethi- activity transfers the guanine nucleotide to the RNA prior onine (AdoMet) that are required for the formation of a to the methyltransfer reaction. protein-guanine nucleotide intermediate. Brome mosaic virus (BMV) is a plant-infecting member of the alphavirus-like superfamily (Koonin, 1993). The RESULTS BMV RNA genome consists of three capped RNAs. The longest RNA, RNA1 codes for the 1a protein, which has G-1a and G-MT expression sequence homologies to RNA capping and helicase ac- To facilitate the purification of the 1a MT-like domain, we tivities in its N- and C-terminal halves, respectively (Ha- fused the cDNA coding for the full-length or the N-terminal seloff et al., 1984; Ahlquist, 1992; Rozanov et al., 1992). 516 residues of 1a to the GST-coding sequence. The expected fusion protein should be of 136 kDa for full-length 1a (G-1a), and 75 kDa for the N-terminal MT domain (G-MT). 1 To whom reprint requests should be addressed. Fax: (812) 855- After enrichment in batch through glutathione beads, bands 6705. E-mail: [email protected]. of the expected mass were the predominant products in

FIG. 1. Expression and activity of the BMV capping domain. (A) Coomassie Blue-stained 10% PAGE–SDS containing the supernatant of a 15,000 g (S15)-clarified bacterial lysate and the eluant from glutathione-agarose column. Sizes of the molecular mass markers in kilodaltons (kDa) are indicated on the left of the gel. The identities of the enriched proteins, glutathione S-transferase (GST), and GST fused to the full-length 1a (G-1a) and the N-terminal 516 residues of the 1a protein (G-MT) are indicated on the right. Lane numbers are on the bottom of the autoradiograph. (B) Ability of the BMV capping domain to form a covalent adduct with guanine nucleotide. The S15 and eluant fractions in (A) were incubated with buffer B, whose 32 components include 4 ␮Ci [␣- P]GTP, 100 ␮M AdoMet, and 2 mM MgCl2. The samples were analyzed by SDS–PAGE and autoradiography. The expected positions of the three recombinant proteins are indicated on the left. gels stained with Coomassie Blue (Fig. 1A). In Western protein are essential for binding guanine nucleotide. blots, these products reacted against serum specific for Furthermore the BMV methyltransferase-like domain is GST (F. Kong, data not shown). In the preparations contain- required for guanine nucleotide binding because GST ing G-MT, some bands lower in molecular weight than 75 alone was unable to form an adduct with radiolabeled kDa were also recognized by anti-GST, indicating that they guanine nucleotide (Fig. 1B, lane 4). G-MT that had been contain truncations of the C-terminal BMV MT domain (data treated with thrombin to separate the BMV MT from the not shown). GST also bound GTP (F. Kong, data not shown). However, The nsP1 proteins of Sindbis (SIN) and Semliki Forest the cleavage reaction did not work efficiently, and we virus (SFV) can form a covalent intermediate with a meth- routinely assayed for guanine nucleotide binding with ylated GMP that will survive denaturing protein gel elec- the fusion protein. trophoresis (Laakkonen et al., 1994; Ahola and Ka¨a¨ri- G-1a did not form adduct with radioactive guanine a¨inen, 1995). To determine whether G-1a and G-MT can nucleotide significantly above background levels despite form a adduct with a guanine nucleotide, we incubated the fact that it was present in higher abundance than clarified lysate S15 and the glutathione-enriched fractions G-MT in the S15 fractions (Fig. 1B, lanes 2 and 5). Low- for 30 min with [␣-32P]GTP, AdoMet, and magnesium. level binding to radiolabeled guanine nucleotide was After stopping the reaction by the addition of Laemmli observed in some experiments, but all at much reduced sample buffer containing 1% SDS, the sample was level on a molar basis in comparison with G-MT. Several heated to 90°C and then electrophoresed in a denaturing independently generated G-1a preparations behaved in polyacrylamide gel. A parallel gel containing the same a similar manner, suggesting that the lack of guanine protein samples was stained with Coomassie Blue to nucleotide binding is not due to an unexpected mutation visualize the proteins. Autoradiography revealed a band in 1a. As shall be presented later, 1a present in the BMV identical in position to G-MT (Fig. 1B, lanes 3 and 6). replicase extracted from infected plants does have gua- Faintly labeled products seen in the S15 fractions were nine nucleotide-binding activity. absent in the preparations containing the more pure GST-enriched fractions containing G-MT. In these reac- Requirements for the formation of the G-MT-guanine tions (Fig. 1B, lane 6), two bands of Ͻ75 kDa also formed nucleotide covalent adduct an adduct with guanine nucleotide (Fig. 1B, lane 6). Because this truncated protein contains an N-terminal To define the requirements for the G-MT-guanine nu- GST as determined by Western blot analysis (F. Kong, cleotide adduct, we examined the effects of different data not shown), not all of the first 516 residues of the 1a components in the binding reaction (Fig. 2). In the pres- 202 KONG, SIVAKUMARAN, AND KAO

are similar to those seen with the alphaviral nsP1 protein (Ahola and Ka¨a¨ria¨inen, 1995).

Guanine nucleotide moieties required to interact with G-MT G-MT could bind [␣-32P]-GTP but not [␣-[32P]ATP] (Fig. 3A, lanes 2 and 5), indicating that it has specificity for guanine nucleotide. To identify the chemical moieties in GTP required for specific recognition, we added an increasing amount of different nucleotides competitors to a reaction containing constant amounts of M-GT and [␣-32P]GTP. The effect on adduct formation was then quantified after gel electrophoresis. With unlabeled GTP as the competitor, the amount of radiolabeled adduct was reduced in a concentration-dependent manner (Fig. 3B). In contrast, ATP, CTP, UTP, dATP, dCTP, and dTTP did not significantly reduce the amount of the G-MT-guanine nucleotide adduct at even a 500-fold molar excess of the FIG. 2. Requirement for the formation of a covalent guanine nucleotide-G-MT adduct. G-MT enriched from the glutathione column were radiolabeled GTP (Fig. 3B and data not shown). In fact, incubated with [␣-32P]GTP along with the components whose presence UTP and CTP stimulated adduct formation by over two- and absence are indicated above the autoradiogram with plus and fold (Fig. 3B). While the results with UTP and CTP are minus signs, respectively. The concentrations used for each compo- reproducible, we do not have an adequate explanation ␮ nent were as follows: AdoMet, when added, was to 100 M final for it. It is possible that the pyrimidine nucleotides inhibit concentration; Mg2ϩ,Mn2ϩ, and EDTA, when added to the reaction, were to 2 mM final concentration; Ca2ϩ was added to 2 and 10 mM in a reaction that indirectly affects the availability of GTP to lanes 7 and 8, respectively. The reactions were stopped by the addition bind G-MT. Unlabeled ITP (lacking the C2 amino group of of Laemmli sample buffer (Laemmli, 1970) and analyzed by 10% PAGE– the guanine base) competed with the [␣-32P-GTP] nearly SDS and autoradiography. The reactions in lanes 1 and 2 contained as well as unlabeled GTP (Fig. 3B). These results sug- conditions optimal for adduct formation (see Materials and Methods) gest that there are specific moieties in guanine required and the average amount of guanine nucleotide bound in these reactions is set at 100%. The expected electrophoretic position of G-MT is for recognition by G-MT. indicated on the right of the autoradiograph. Quantifications of the To determine the contributions of the nonbase moi- amount of guanine nucleotide bound are indicated below the auto- eties, we tested the effects of adding mono-, di-, and radiogram. triphosphate forms of guanine nucleotides. All three in- hibited adduct formation to a degree with GTP being the most efficient competitor, followed by GDP and then ence of 100 ␮M AdoMet and 2 mM MgCl , maximum 2 GMP. These results suggest that the ␥ and ␤ phosphates [␣-32P]GTP incorporation was observed after a 40-min in GTP contribute to adduct formation. Deoxy-GTP was a incubation at 22°C. In reactions lacking exogenously poor competitor, reducing adduct formation by only 40% supplied AdoMet, guanine nucleotide binding was at when present at a 500-fold molar excess relative to 13% relative to the standard reactions (Fig. 2, lane 3). In [␣-32P]GTP (Fig. 3C). This result suggests that the ribose reactions lacking exogenously added divalent metal, 2Ј hydroxyl is required for efficient G-MT-guanine nucle- guanine nucleotide binding was reduced to 14% (Fig. 2, otide interaction. A summary of the putative recognition lane 4). These residual activities may be due to contam- ϩ sites in a guanine nucleotide is presented in Fig. 3D. inating AdoMet and Mg2 that copurified with the recombinant proteins. Consistent with this, the addition of the AdoMet moieties required for G-MT-guanine metal chelator EDTA eliminated the low levels of guanine nucleotide adduct formation nucleotide binding (Fig. 2, lane 5). The divalent metal Mn2ϩ could efficiently substitute for Mg2ϩ in the adduct AdoMet (Fig. 4A) is required for G-MT to efficiently formation, whereas Ca2ϩ was a less effective substitute adduct guanine nucleotide, most likely due to the need even at fivefold higher concentration than Mg2ϩ. Last, the for GTP to be methylated prior to the formation of a presence of detergents such as deoxycholate and Triton covalent bond with G- MT. In testing different salts of X-100 and pyrophosphate at a 1- to 10-M excess of GTP AdoMet, we found that both AdoMet toluene-4-sulfonate were inhibitory to guanine nucleotide adduct formation. (TS) and AdoMet (IS) stimulated adduct formation be- All of these additions resulted in G-MT-guanine nucleo- tween G-MT and guanine nucleotide, with the sulfonate tide adduct formation at to Ͻ30% of control reactions (F. form of AdoMet being 12-fold better than the Iodo form Kong, data not shown). All of these characteristics of (Fig. 4B, lanes 3 and 4 and 7 and 8). AdoMet is a labile guanine nucleotide binding by the BMV MT-like domain molecule and AdoMet (TS) was stated by the manufac- RNA CAPPING-ASSOCIATED ACTIVITIES 203

FIG. 3. Specificity of the G-MT for guanine nucleotide. (A) Autoradiogram of nucleotide adduct formation in the presence of [␣-32P]GTP or [␣-32P]ATP and the three recombinant proteins, GST, G-MT, and G-1a. The positions of the three proteins used in the reactions, as determined by staining with silver of a parallel denaturing protein gel, are shown on the right. (B) Effect of increasing amounts of ribonucleotides on the formation of guanine nucleotide-G-MT adduct. Each reaction was performed with 15 pmol [␣-32P]GTP, and the amount of competitor nucleotide used relative to the amount of [␣-32P]GTP is shown on the horizontal axis. The amount of guanine nucleotide-G-MT adduct in the absence of any competitor was set as 100%. The identity of the competitor used is indicated on the right of the autoradiogram. (C) Competition assay with nucleotides containing one, two, or three phosphates and with dGTP. The amount of adduct formation of GTP and G-MT under standard reaction conditions was set at 100%. (D) Structure of a guanine nucleotide. The chemical moiety highlighted in bold are the ones in GTP that affect adduct formation with the BMV capping domain. turer (Sigma Inc.) to be more stable than the AdoMet(IS), vation that AdoHcy is a potent inhibitor of MT activity perhaps accounting for the observed differences. (Laakkonin et al., 1994). Finally, AMP, methionine, and a We next determined whether the transfer of a methyl combination of the two were unable to induce guanine group from AdoMet is required for guanine nucleotide nucleotide binding by G-MT (Fig. 4B, lanes 13–15 and binding. We tested three analogues of AdoMet (Fig. 4A): data not shown). AdoEth, which has an ethyl group attached to the sulfur, Although AdoEth, AdoHcy, and AdoCys were unable to AdoHcy, which lacks the methyl group of AdoMet, and efficiently stimulate guanine nucleotide binding, we de- AdoCys, which has a cysteine instead of a methionine. termined whether they could have an inhibitory effect by None of the three analogues were able to replace competing with AdoMet. The three analogues were AdoMet in inducing guanine nucleotide-binding by G-MT added to a standard reaction at 100 ␮M, 1 mM, and 10 (Fig. 4B). In fact, AdoHcy and AdoCys may have inhibitory mM (AdoMet was present at 100 ␮M). AdoHcy at 100 ␮M effects (compare Fig. 4B, lanes 9–12 with lanes 1 and 2). reduced guanine nucleotide binding to Ͼ25% of a reac- This latter observation is consistent with previous obser- tion lacking a competitor; although the other two ana- 204 KONG, SIVAKUMARAN, AND KAO

FIG. 4. Effect of AdoMet analogues on the guanine nucleotide-enzyme adduct formation. (A) Guanine nucleotide-enzyme adduct formation assay containing ␣[32P]GTP were performed in the presence of different SAM analogues whose chemical structures are shown. (B) Autoradiogram of the adducts formed between G-MT and guanine nucleotide in the presence of different AdoMet analogues. Each of the reactions were performed in to independent reactions. Where used, AdoMet and its analogues were added at 100 ␮M final concentration. The amount of adduct formed in the presence of the iodo and the TS-salt forms of AdoMet likely reflects a difference in the stability of AdoMet in these two salts. Quantitation of adduct formation are shown below the autoradiogram. (C) Effect of increasing concentrations of AdoMet analogues on G-MT-guanine nucleotide adduct formation. The amount of adducts formed in a standard reaction lacking AdoMet analogues was set at 100%. The identities of the AdoMet analogues is indicated to the right of the graph. logues were also inhibitory on a molar basis, they were that R-group size affects recognition. However, because less effective. All of the inhibitory effects were more AdoHcy is an effective competitor despite the lack of a pronounced when the analogues were present at 1 mM. methyl group, moieties other than the methyl group con- AdoCys, with one fewer carbon than AdoHcy in the tribute to AdoMet recognition by the BMV methyltrans- amino acid R group, was a poorer competitor, suggesting ferase domain. RNA CAPPING-ASSOCIATED ACTIVITIES 205

FIG. 5. Transfer of a methyl group to the guanine nucleotide-G-MT adduct. GST- and G-MT used in these reactions were both enriched by the glutathione resin. The proteins were incubated with buffer B lacking GTP and containing 5 ␮M Ci [methyl-3H]AdoMet substituting for AdoMet. Where added (as denoted with a ϩ), GTP was to 10 and 100 ␮M GTP in lanes 3 and 4, respectively. The reactions were stopped by the addition of SDS-PAGE sample buffer and analyzed by SDS–PAGE and fluorography. (B) Amount of 14C AdoMet signal incorporated into the band containing G-MT in the presence of different nucleotide substrates. The results are the average of two independent reactions whose values varied Ͻ20%. Radiolabeled products separated by SDS–PAGE were quantified with a Phosphorimager. The relative amount of signal in each reaction was adjusted relative the reaction containing GTP.

Methyltransferase activity of G-MT added nucleotide or containing ATP, and CTP did not result in labeling above background level (Fig. 5B). How- We used AdoMet containing a 3H-labeled methyl ever, labeling in the presence of ITP was at 81% relative group to determine whether the methyl moiety is trans- to GTP (percentages are the average of two independent ferred to the guanine nucleotide. GST or G-MT were reactions), consistent with ITP being a good inhibitor of incubated with 3H-AdoMet in the presence or absence of guanine nucleotide binding (Fig. 3B). Reactions contain- 100 ␮M unlabeled GTP. After the reaction, the extract ing GDP, GMP, and dGTP were labeled from 29 to 16% in was electrophoresed in SDS–PAGE and fluorographed comparison with reactions containing GTP. The results (Fig. 5A). G-MT was labeled in the presence of GTP, confirm our previous conclusions that the phosphates, whereas GST was not (Fig. 5A, lanes 3 and 4). Filter- the 2Ј hydroxyl, specific guanine moieties such as the C6 binding experiments showed that in the absence of GTP, keto group, but not the C2 amine are required for adduct AdoMet is not bound by the G-MT above background formation between the MT and guanine nucleotide. levels (Kong, data not shown), thus the guanine nucleotide-G-MT adduct cannot form in the absence of AdoMet. Guanine nucleotide binding by the BMV replicase However, this result does not indicate whether the labeled MT seen in Fig. 5A contains only the methylated The 1a protein is a component of the BMV RNA repli- guanine nucleotide or a trimolecular complex that in- case (Kao and Ahlquist, 1992; Quadt et al., 1995). Our cludes AdoMet. Based on the results from alphaviral previous result showed that recombinant G-1a was un- capping proteins, the latter possibility is less likely (i.e., able to efficiently bind guanine nucleotide (Fig. 1B, lanes the AdoHcy is released after AdoMet donates the methyl 2 and 5). BMV replicase enriched from infected barley group) (Laakkonen et al., 1994; Ahola and Ka¨a¨ria¨inen was incubated in the presence of GTP, CTP, ATP, or UTP, 1995). The reactions also contain a lower molecular all radiolabeled in the ␣-phosphate position. In denatur- weight band radiolabeled by 3H-AdoMet that does not ing gel electrophoresis, a band identical in electro- correspond to the molecular mass of either G-MT or GST. phoretic mobility to in vitro translated and 35S-Met-la- Due to difficulty in detecting the tritium-labeled MT- beled 1a was observed in the reaction containing guanine nucleotide adduct, we used 14C-Me-AdoMet to [32P]GTP (Fig. 6A, lanes 1–3). In contrast, the labeled examine features in the acceptor nucleotide required for band was not detected in the presence of the other the transfer of the methyl group. Reactions without radiolabeled nucleotides (Fig. 6A, lanes 4–6). Also, prod- 206 KONG, SIVAKUMARAN, AND KAO

FIG. 6. Guanine nucleotide binding by the 1a in the BMV replicase. (A) 1a in highly enriched BMV replicase can bind guanine nucleotide. The ␣-32P-radiolabeled nucleotide (each at ϳ400 Ci/mmol and 0.3 ␮M) used in each reaction is indicated above the autoradiogram. Lanes marked with the letter M contain in vitro translated and 35S-met-labeled 1a and 2a proteins. The bands corresponding to 1a and 2a are indicated on the left. The presence or absence of 10 ␮l of highly enriched RdRp is indicated by a ϩ or a Ϫ, respectively. (B) Adduct formation between 1a and guanine nucleotide is not efficiently competed by molar excess of ATP. Lanes 1 and 7 contains in vitro translated 1a and 2a proteins. ATP was not added (lanes 2 and 3), or added at 25 and 50 M excess of GTP (Lanes 4 and 5). (C) The addition of AdoMet analogues had slight effects on 1a-guanine nucleotide adduct formation while exogenous AdoMet increased adduct formation with GTP by 10-fold. The identities of the AdoMet or analogues added (each at 100 ␮M) are indicated on top of the autoradiogram. (D) Quantification of the effects of AdoMet analogues on adduct formation between 1a and [␣-32P-GTP]. The amount of analogue added in molar excess of AdoMet is indicated on the horizontal axis. ucts of the size consistent with the 96-kDa BMV 2a cleotide binding. These results indicate that 1a present protein were not labeled with [32P]GTP. The formation of in the BMV replicase recognize guanine nucleotide and the 1a and guanine nucleotide adduct was further exam- AdoMet in a manner similar to G-MT. ined with reactions performed in the presence of unlabeled ATP. ATP at 25- and 50-fold molar excess of radio- Inhibitors of guanine nucleotide binding does not labeled GTP reduced the formation of the 1a-guanine affect RNA synthesis by the BMV replicase in vitro nucleotide adduct to 67 and 56%, respectively (Fig. 6B, lanes 4 and 5). This result indicates that 1a binds pref- The BMV replicase can direct accurate and species- erentially to GTP. While the adduct was observed in the specific initiation of RNA synthesis (Miller et al., 1986; absence of exogenously provided AdoMet, exogenously Sun et al., 1996). We have developed RNA synthesis provided AdoMet (TS) increased adduct formation by assays using short RNAs derived from each of the three 10-fold (compare Fig. 6C, lane 6 with 1 and 2), demon- classes of BMV RNA promoters used during infection strating that AdoMet is present in the replicase prepa- (Sun and Kao, 1997; Adkins et al., 1998; Chapman and rations in limiting abundance. Exogenously supplied Kao, 1999). The three classes of BMV RNAs and their AdoHcy, AdoCys, and AdoEth were unable to substitute promoters are: (1) genomic minus-strand RNAs pro- for AdoMet in guanine nucleotide binding. In fact, addi- duced by the tRNA-like promoters present at the 3Ј end tion of 100 ␮M AdoHcy noticeably reduced guanine nu- of the virion RNAs, (2) subgenomic plus-strand RNA RNA CAPPING-ASSOCIATED ACTIVITIES 207

Novobiocin reduced the in vitro synthesis of minus- strand BMV RNA by twofold when present at ca.70␮M. When present in a guanine nucleotide binding reaction, novobiocin at concentrations Յ350 mM caused only a minimal decrease in guanine nucleotide binding. These results indicate that RNA synthesis in vitro does not require the conditions needed for efficient methyltransferase activity.

DISCUSSION RNA capping is required for the successful infection of numerous RNA viruses, including members of the alphaviral superfamily. However, the alphaviral capping activity differs from cellular capping activities in that the guanine nucleotide must be methylated at the N7 position before forming a phosphoamide bond with the enzyme (Laakkonen et al., 1994; Aloha and Kaä¨riaïnen, 1995). We also observed that the ability to transfer a methyl group from AdoMet to a guanine nucleotide is required for the FIG. 7. Effect of AdoMet, AdoCys, or AdoHcy on RNA synthesis in presumed covalent intermediate formed during guanyl- vitro. (A) AdoMet, AdoHCy, and AdoCys were added to RNA synthesis reactions at 100 ␮M, and the amount of RNA synthesis was quantified. transfer (Fig. 5B). The fact that GMP is a poor substrate The amount of synthesis from each class of BMV RNA template lacking for methyltransferase indicates that GTP is most likely AdoMet or AdoMet analogues was set at 100%. Genomic RNA synthe- methylated prior to phosphoamide bond formation. De- sis was from RNA named B2(-)26 (Sivakumaran and Kao, submitted), spite mechanistic differences, all of the capping en- subgenomic RNA synthesis was from RNA named –20/13 (Siegel et al., zymes must recognize three substrates, GTP, AdoMet, 1997), and genomic (Ϫ)-strand RNA was from the 3Ј 200 nts of BMV RNA3 (Chapman and Kao, 1999). (B) Effect of increasing concentrations and the nascent RNA. In this report, we demonstrated of novobiocin on GTP-binding by the BMV 1a protein (black circles) and that a highly enriched recombinant BMV capping on the synthesis of (Ϫ)-strand RNAs (open boxes). The template used polypeptide has specificity for a guanine nucleotide in a for RNA synthesis was total BMV RNA purified from BMV virions. reaction that requires a methyl group donated by AdoMet. The chemical moieties in GTP and AdoMet that initiated from an promoter internal to minus-strand RNA3, contribute to specific recognition by the BMV methyl- and (3) genomic plus-strand RNAs that initiated from the transferase were determined using chemical analogues. 3Ј ends of BMV minus-strand RNAs. The genomic and We also determined that the formation of a presumed subgenomic plus-strand RNAs are capped in vivo while covalent adduct with guanine nucleotide does not ap- the minus-strand RNAs is not (Ahlquist, 1992) (Fig. 7). pear to be required for viral RNA synthesis in vitro. These We next tested whether reagents that affected capping results advance the characterization of the activities of a activities affect RNA synthesis. The three classes of RNA capping enzyme from the alphavirus-like superfamily and promoters were tested in RNA synthesis reactions con- allow comparisons with the activities of nonalphaviral taining AdoMet, which should stimulate guanine nucle- capping enzymes. otide adduct formation several folds (Fig. 4B), or either Comparison of guanine nucleotide recognition by AdoCys and AdoHcy, which should reduced the guanine several capping enzymes nucleotide adduct significantly (Fig. 4C). The addition AdoMet, AdoHcy, and AdoCys to 100 ␮M final concen- The specificity for GTP has been examined for a num- tration had no discernible effects on the synthesis of ber of viral RNA capping enzymes, including those from RNA products initiated from any of the three classes of tobacco mosaic virus (Dunigan and Zaitlin, 1990), SFV promoters (Fig. 4C) despite a previously demonstrated (Laakkonen et al., 1994), and vaccinia virus (Martin and reduction on G-MT-guanine nucleotide adduct formation Moss, 1976). In addition, abundant information on sub- (Fig. 4B). However, the addition of AdoMet, AdoCys, and strate recognition is available for the mammalian guanyl- AdoHcy to 1 mM final concentration led in each case to yltransferases (Venkatesan and Moss, 1982) and inferred reduced RNA synthesis (K. Sivakumaran, data not from the algal PBCV virus capping protein, whose struc- shown). However, because both AdoMet and the com- ture was solved to 2.5 angstrom (Hakansson et al., 1997). petitors of AdoMet decreased RNA synthesis, it appears The various capping enzymes all appear to require the that the effects may not be biologically relevant. phosphates of GTP. The order of preference for guanine We had previously determined that several com- nucleotide is GTP Ͼ GDP Ͼ GMP, suggesting that the ␤ pounds inhibit RNA synthesis in vitro (Sun et al., 1996). and ␥ phosphates play a role in the recognition by the 208 KONG, SIVAKUMARAN, AND KAO capping enzymes of several viruses, including TMV exist both at the level of protein structure and the re- (Dunigan and Zaitlin, 1990), SFV (Aloha and Kaä¨riaïnen, quirements for selected residues. 1995), rotavirus (Pizarro et al., 1991), and vaccinia virus (Martin and Moss, 1976). In contrast, the recognition of Methyltransferase activity and the BMV replicase the ribose hydroxyls differs in various capping enzymes. The ability of 1a in the BMV replicase preparations to The BMV capping domain prefers GTP to dGTP. How- bind a guanine nucleotide differs from the results we ever, the TMV p126, SFV, rotavirus, and vaccinia capping observed with the recombinant G-1a (Figs. 1A and 7A). enzymes do not appear to distinguish between ribose-, Several possibilities may explain this difference. First, C2Ј deoxy-, and C2Ј,3Ј-dideoxy forms of GTP (Martin and the GST domain may prevent the proper folding of the Moss, 1976; Venkatesan and Moss, 1982; Dunigan and recombinant 1a protein. Second, O’Reilly et al. (1998) Zaitlin, 1990; Pizarro et al., 1991; Ahola and Kaä¨riaïnen, have previously observed that in the absence of the 1995). In fact, the SFV nsP1 and vaccinia capping en- polymerase-like 2a, strong protein–protein interaction zymes have been reported to prefer dGTP to GTP as a can take place between capping-like and helicase-like substrate for methyl transfer (Martin and Moss, 1976; domains of 1a. Perhaps this interaction renders 1a inca- Laakkonen et al., 1994). Perhaps different spatial con- pable of binding radiolabeled guanine nucleotide. We straints exist in the catalytic pockets of different capping favor the second possibility because GST fused to the enzymes to account for differences in the recognition of methyltransferase domain was not inhibitory for the rec- the ribose 2Ј hydroxyl. ognition of guanine nucleotide. Furthermore the addition The guanine base allows the capping enzymes to of other protein sequences N terminal to the 1a or cap- distinguish GTP from ATP (Martin and Moss, 1976; Ven- ping domain did not affect the interaction between pro- katesan and Moss, 1982; Cross, 1983; Dunigan and Zait- teins in the two-hybrid system (O’Reilly et al., 1997). lin, 1990). Guanine and adenine differ at the C6 position Guanine nucleotide binding does not appear to be where guanine contains a keto group, whereas adenine required for the synthesis of any of the three classes of contains an amino group. Thus the capping enzymes BMV RNAs in vitro. The addition of AdoMet and AdoHcy likely recognize the guanine C6 keto. For the BMV cap- did not affect the synthesis of BMV RNAs. Finally, novo- ping activity, we observed that ITP was an effective biocin, which decreases RNA synthesis in vitro, did not competitor against GTP, indicating that the C2 amino have an obvious effect on guanine nucleotide binding by group (lacking in inosine) is likely to be less important for the 1a protein. It remains possible that guanine nucleo- recognition. In addition, because both inosine and gua- tide adduct formation at levels lower than can be de- nine have a N1 imine, the contribution of this moiety in tected in our assays may be sufficient for RNA synthesis. the interaction between the BMV methyltransferase and At present it is not known at what stage in BMV RNA the guanine nucleotide interaction cannot be dismissed. synthesis does capping process take place. Another Similarly, the role of the N7 group in guanine needs to be interesting question for future studies is what determines examined further. Last, we note that Venkatesan and that plus-strand RNAs are capped but not minus-strand Moss (1982) found ITP to be a poor substrate for methyl RNAs. The production of a recombinant BMV capping transfer, reinforcing the idea that capping enzymes rec- protein coupled with an in vitro RNA synthesis assay ognize guanine nucleotide in different ways. would allow us to address some of these mechanistic Anther difference between the alphaviral and other questions. cellular and vaccinia capping activities is that the alpha- Guanine nucleotide binding and the transfer of a viral capping enzymes lack obvious similarities to the six methyl moiety to GTP are but two of the four activities of sequence motifs found in other capping enzymes (Shu- expected of an alphaviral capping enzyme. The remain- man et al., 1994; Malone et al., 1995; Shuman and ing two activities are an RNA triphosphatase activity and Schwer, 1995; Wang et al., 1997). These observations transfer of the guanine nucleotide to mRNA. Our at- suggest either that the rapid evolution of RNA viruses tempts to transfer a radiolabeled guanine nucleotide to eliminated or replaced many of the residues involved in RNA were unsuccessful. Similar attempts with the SFV substrate recognition and/or the positive-strand viral nsP1 protein were also unsuccessful (Laakkonen et al., RNA capping enzymes arose independently of other co- 1994). Further work on the triphosphatase activity awaits valent nucleotidyl transferases. In support of the greater more purified preparations of the enzyme. flexibility in the requirements of the capping domain, Koonin (1993) determined that the putative capping domain within the N terminus of the flavivirus NS5 polypep- MATERIALS AND METHODS tide is more divergent in sequence than the C-terminal Materials domain encoding the RNA-dependent RNA polymerase. Within the divergent capping domain, several residues All radiolabeled compounds and the fluorography re- putatively involved in AdoMet binding are conserved. agent Amplify were purchased from Amersham. S-ad- Evolutionary selection for methyltransferase activity may enosyl-L-methionine iodo salt (AdoMet[IS]), S-adenosyl- RNA CAPPING-ASSOCIATED ACTIVITIES 209

L-homocysteine (AdoHcy), S-adenosyl-L-cysteine (Ado- Methyltransferase assay Cys), S-adenosyl-L-ethionine (AdoEth), glutathione and MT assays were performed as a 15-␮l final reaction glutathione-Sepharose 4B, and protease inhibitors were containing buffer A, 0.2 ␮g fusion protein, 1.5 ␮Ci of obtained from Sigma (St. Louis, MO). AdoMet toluene- 3 S-Adenosyl-L-[methyl- H]methionine (500 ␮Ci/mmol) or sulfonate salt (AdoMet[TS]) was purchased from Fluka 14 0.175 ␮Ci S-Adenosyl-L-[methyl- C]methionine (59 ␮Ci/ Biochemicals (Switzerland). mmol) for 30 min at RT. The samples were then electrophoresed in a 10% SDS–PAGE followed by fluorography Expression of recombinant 1a and MT proteins with the reagent Amplify. DNA fragments encoding full-length 1a and N-terminal residues 1–516 were generated by PCR. Both PCR frag- Viral RNA synthesis in vitro ments were flanked by a 5Ј BamHI and a 3Ј EcoRI restriction site. After restriction digestion, the DNAs were BMV replicase used in some of the experiments were ligated in frame to the glutathione S-transferase (GST)- prepared from infected barley as described previously coding sequence in pGEX-2T (Pharmacia), creating plas- (Sun et al., 1996). RNA synthesis assays consisted of ␮ mids that will express G-1a and G-MT. Protein expres- 40- l reactions containing 20 mM sodium glutamate (pH sion was induced in E. coli strain BL21DE3 gown at 8.2), 4 mM MgCl2, 12 mM dithiothreitol, 0.5% (v/v) Triton ␮ ␮ ␮ 30°C. When the cultures reached A600 of 0.6, isopropyl X-100, 2 mM MnCl2, 200 M ATP, 500 M GTP, 200 M ␣ 32 thio-␤-D galactopyranoside (IPTG) was added to a final UTP, 242 nM [ - P]CTP (400 Ci/mmole, 10 mCi/ml, Am- ␮ concentration of 1 mM. Cells were harvested by centrif- ersham), 1.0 pmole of template RNA, and 5–10 l RdRp. ugation 3–4 h after induction with IPTG addition and Reactions were incubated 90 min at 30°C, and the reac- sometimes stored at Ϫ80°C until use. tions were extracted with phenol/chloroform (1:1 v/v) and ␮ Cells expressing GST, G-1a, and G-MT were sus- precipitated with 3 volumes of ethanol and 10 gof pended in 10 volumes of ice-cold buffer A [50 mM Tris–Cl glycogen (Sambrook et al., 1989). ␤ (pH 8), 150 mM NaCl, 10 mM MgCl2,5mM -mercapto- ethanol, 10 mM phenylmethylsulfonyl fluoride, 500 ng/ml ACKNOWLEDGMENTS leupeptin, and 350 ng/ml pepstatin A], then lysed by two passes through a French Press (Spectronic Instruments, We thank Stewart Shuman and the IU Cereal Killers, especially Erin O’Reilly and Matt Chapman, for helpful discussions of this work and V. Rochester, NY) at 18,000 psi followed by a 30-s sonica- Stollar for critiques of the manuscript. Funding was provided by the tion with a Branson microtip set at 50% intensity (Fisher, Indiana University Biology Department and grants from the National Pittsburgh, PA). The lysate was clarified by centrifugation Science Foundation (MCB9507344) and USDA (9702126). at 15,000 g for 30 min, resulting in a supernatant fraction named S15. G-MT and G1a were then enriched by the REFERENCES addition of a one tenth volume of glutathione–Sepharose 4B beads equilibrated with buffer A. After a 1.5-h incuba- Adkins, S., Stawicki, S. Faurote, G., Siegel, R., and Kao, C. (1998). tion at 4°C, the beads were washed thrice with 5 vol- Mechanistic analysis of RNA synthesis RNA-dependent RNA polymerase from two promoters reveals similarities to DNA-dependent umes of buffer A and the bound proteins eluted with RNA polymerase. RNA 4, 455–470. buffer containing 50 mM Tris–Cl, pH 7.5, 10% glycerol, Ahlquist, P. (1992). Bromovirus RNA replication and transcription. Curr. and 15 mM glutathione. The eluant was used directly for Opin. Genet. Dev. 2, 71–76. guanine nucleotide binding and MT activity assays or Ahlquist, P., Strauss, E. G., Rice, C. M., Strauss, J. H., Haseloff, J., and made into aliquots for storage at Ϫ80°C. Zimmern, D. (1985). Sindbis virus proteins nsP1 and nsP2 contain homology to nonstructural proteins from several RNA plant viruses. J. Virol. 53, 536–542. Guanine nucleotide binding assay Ahola, T., and R. Kaä¨riaïnen. (1995). Reaction in alphavirus RNA cap- ␮ ping: Formation of a covalent complex of nonstructural protein nsP1 Guanine nucleotide binding was performed in a 10 l with 7-methyl-GMP. Proc. Natl. Acad. Sci. USA 92, 507–511. reaction containing buffer B (final concentrations: 50 mM Ahola, T., Laakkonen, P., Vihinen, H., and R. Kaä¨riaïnen. (1997). Critical Tris–Cl, pH 7.5, 2 mM MgCl2, 10 mM KCl, and 5 mM DTT), residues of Semliki forest virus RNA capping enzyme involved in 0.2 ␮g recombinant protein, 4 ␮Ci of [␣-32P]GTP and 100 methyltransferase and guanylyltransferase-like activities. J. Virol. 71, ␮M AdoMet for 30 min at RT. Three ␮lof5ϫ SDS–PAGE 392–397. Chapman, M., and Kao, C. (1999). Minimum tRNA-like promoters capa- sample buffer was added, and the samples were sub- ble of directing accurate initiation of (Ϫ)-strand RNA synthesis. J. jected to electrophoresis in a denaturing 10% polyacryl- Mol. Biol. 286, 709–720. amide gel (SDS–PAGE) (Laemmli, 1970). For the nucleo- Cross, R. (1983). Identification of a guanine-7 methyltransferase in tide binding competition assays, specified amounts of Semliki Forest virus (SFV) infected cell extract. Virology 130, 452– each competitor were added to the reaction mixture 463. Dunigan, D., and Zaitlin, M. (1990). Capping of tobacco mosaic virus before the addition of protein sample. Quantification was RNA. Analysis of viral-coded guanylyltransferase-like activity. J. Biol. performed using a Phosphorimager (Molecular Dynam- Chem. 265, 7779–7786. ics). Hakansson, K., Doherty, A. J., Shuman, S., and Wigley, D. B. (1997). X ray 210 KONG, SIVAKUMARAN, AND KAO

crystallography reveals a large conformational change during guanyl Pizarro, J. L., Sandino, A. M., Pizarro, J. M., Fernandez, J., and Spencer, transfer by mRNA capping enzymes. Cell 89, 545–553. E. (1991). Characterization of rotavirus guanylyltransferase activity Haseloff, J., Goelet P., Zimmern D., Ahlquist, P., Dasgupta, R., and associated with polypeptide VP3. J. Gen. Virol. 72, 325–332. Kaesberg, P. (1984). Striking similarities in amino acid sequence Quadt, R., Ishikawa, M. Janda, M., and Ahlquist, P. (1995). Formation of among nonstructural proteins encoded by RNA viruses that have brome mosaic virus RNA-dependent RNA polymerase in yeast re- dissimilar genomic organization. Proc. Natl. Acad. Sci. USA 81, 4358– quires coexpression of viral proteins and RNA. Proc. Natl. Acad. Sci. 4362. USA 92, 4892–4896. Kao, C. C., and Ahlquist, P. (1992). Identification of the domains required Rozanov, M. N., Koonin, E. V., and Gorbalenya, A. E. (1992). Conserva- for direct interaction of the helicase-like and polymerase-like RNA tion of the putative methyltransferase domain: A hallmark of the replication proteins of brome mosaic virus. J. Virol. 66, 7293–7302. ‘‘Sindbis-like’’ supergroup of positive-strand RNA viruses. J. Gen. Koonin, E. V. (1993). Computer-assisted identification of a putative Virol. 73, 2129–2134. methyltransferase domain in NS5 protein of flaviviruses and L2 Sambrook, J., Fritsch, E. F., and Maniatis, T. (1989). “Molecular Cloning: protein of reovirus. J. Gen. Virol. 74, 733–740. A Laboratory Manual.” Cold Spring Harbor Laboratory Press, Cold Laakkonen, P., Hyvonen, M. Peranen, J., and Kaariainen, L. (1994). Spring Harbor, NY. Expression of Semliki Forest virus nsp1-specific methyltransferase in insect cells and in Escherichia coli. J. Virol. 69, 7418–7425. Schluckebier, G., O’Gara, M., Saenger, W., and Cheng, X. (1995). Uni- Laemmli, U. K. (1970). Cleavage of structural proteins during the as- versal catalytic domain structure of AdoMet-dependent methyltrans- sembly of the head of bacteriophage T4. Nature 277, 680–685. ferases. J. Mol. Biol. 247, 16–20. Le Blois, H., French, T., Mertens, P. P. C., Burroughs, J. N., and Roy, P. Shuman, S., Liu, Y., and B. Schwer, B. (1994). Covalent catalysis in (1992). The expressed VP4 protein of bluetongue virus binds GTP nucleotidyl transfer reactions: Essential motifs in Saccharomyces and is the candidate guanylyl transferase of the virus. Virology 189, cerevisiae RNA capping enzyme are conserved in Schizosaccharo- 757–761. myces pombe and viral capping enzymes and among polynucleotide Malone, T., Blumenthal, R. M., and Cheng, C. (1995). Structure-guided ligases. Proc. Natl. Acad. Sci. USA 91, 12046–12050. analysis reveals nine sequence motifs conserved among DNA amino- Shuman, S., and Schwer, B. (1995). RNA capping enzyme and DNA methyl-transferases, and suggests a catalytic mechanism. J. Mol. Biol. ligase: A superfamily of covalent nucleotidyl transferases. Mol. Mi- 253, 618–632. crobiol. 17, 405–410. Mao, Z., and Joklik, W. K. (1991). Isolation and enzymatic characteriza- Siegel, R., Adkins, S., and Kao, C. (1997). Sequence-specific recognition tion of protein L2, the reovirus guanylyltransferase. Virology 185, of a subgenomic promoter by a viral RNA polymerase. Proc. Natl. 377–386. Acad. Sci. USA 94, 11238–11243. Martin, S. A., and Moss, B. (1975). Modification of RNA by mRNA Sivakumaran, K., and Kao, C. Initiation of genomic positive strand RNA guanylyltransferase and mRNA (guanine-7-)methyltransferase from synthesis from DNA and RNA templates by a viral RNA-dependent vaccinia virions. J. Biol. Chem. 250, 9330–9335. RNA polymerase. J. Virol., in press. Martin, S. A., and Moss, B. (1976). mRNA guanylyltransferase and Sun, J. H., Adkins, S., Faurote, G., and Kao, C. (1996). Initiation of mRNA (Guanine 7)-methyltransferase from vaccinia virions: Donor (Ϫ)-strand RNA synthesis catalyzed by the BMV RNA-dependent and substrate specificities. J. Biol. Chem. 251, 7313–7321. RNA polymerase: Synthesis of oligonucleotides. Virology 226, 1–12. Mi, S. and, Stollar, V. (1991). Expression of Sindbis virus nsP1 and Sun, J. S., and Kao, C. (1997). RNA synthesis by the brome mosaic virus methyltransferase activity in Escherichia coli. Virology 184, 423–427. RNA-dependent RNA polymerase: Transition from initiation to elon- Miller, W. A., Bujarski, J. J., Dreher, T. W., and Hall, T. C. (1986). Minus- strand initiation by brome mosaic virus replicase within the 3Ј -tRNA- gation. Virology 233, 63–73. like structure of native and modified RNA templates. J. Mol. Biol. 187, Venkatesan, S., and Moss. B. (1982). Eukaryotic mRNA capping en- 537–546. zyme-guanylate covalent intermediate. Proc. Natl. Acad. Sci. USA 79, O’Reilly E. K., Paul, J. D., and Kao, C. (1997). Analysis of the interaction 340–344. of viral RNA replication proteins by using the yeast two-hybrid assay. Wang, H-L., O’Rear, J., and Stollar, V. (1996). Mutagenesis of the sindbis J. Virol. 71, 7526–7532. virus nsP1 protein: Effects on methyltransferase activity and viral O’Reilly, E., Wang, Z., French, R., and Kao, C. (1998). Mapping of the infectivity. Virology 217, 527–531. protein-protein interaction domains within the brome mosaic virus 1a Wang, S., Deng, L., Ho, C. K., and Shuman, S. (1997). Phylogeny of protein. J. Virol. 72, 7160–7169. mRNA capping enzymes. Proc. Natl. Acad. Sci. USA. 94, 9573–9578.