Domain structure of the vaccinia virus mRNA capping enzyme. Expression in Escherichia coli of a subdomain possessing the RNA 5'-triphosphatase and guanylyltransferase activities and a kinetic comparison to the full-size enzyme.

The RNA 5′-triphosphatase, nucleoside triphosphate phosphohydrolase, and guanylyltransferase activities of the vaccinia virus mRNA capping enzyme were previously localized to an NH-terminal 60-kDa domain of the D1R subunit. Measurement of the relative ATPase and guanylyltransferase activities remaining in D1R carboxyl-terminal deletion variants expressed in Escherichia coli BL21(DE3)plysS localizes the carboxyl terminus of the active domain to between amino acids 520 and 545. Failure to obtain a deletion mutant with the loss of one activity indicates that the catalysis of both reactions requires a common domain structure. Based on these results, a truncated D1R protein terminating at amino acid 545 was expressed in E. coli and purified to homogeneity. D1R was found to be kinetically equivalent to the holoenzyme in regard to ATPase, RNA 5′-triphosphatase, and guanylyltransferase activities. Measurement of RNA binding by mobility shift and UV photo-cross-linking analyses also demonstrates the ability of this domain to bind RNA independent of the methyltransferase domain, comprised of the carboxyl terminus of D1R from amino acids 498-844 and the entire D12L subunit. RNA binding to D1R is substantially weaker than binding to either the methyltransferase domain or the holoenzyme. Binding is inhibited by 5′-OH RNA and to a lesser extent by DNA oligonucleotides in a concentration dependent manner which correlates with the inhibition of RNA 5′-triphosphatase activity by these same oligonucleotides. We conclude that D1R represents a functionally independent domain of the mRNA capping enzyme, fully competent in substrate binding and catalysis at both the triphosphatase and guanylyltransferase active sites.

The RNA 5-triphosphatase, nucleoside triphosphate phosphohydrolase, and guanylyltransferase activities of the vaccinia virus mRNA capping enzyme were previously localized to an NH 2 -terminal 60-kDa domain of the D1R subunit. Measurement of the relative ATPase and guanylyltransferase activities remaining in D1R carboxyl-terminal deletion variants expressed in Escherichia coli BL21(DE3)plysS localizes the carboxyl terminus of the active domain to between amino acids 520 and 545. Failure to obtain a deletion mutant with the loss of one activity indicates that the catalysis of both reactions requires a common domain structure. Based on these results, a truncated D1R protein terminating at amino acid 545 was expressed in E. coli and purified to homogeneity. D1R 1-545 was found to be kinetically equivalent to the holoenzyme in regard to ATPase, RNA 5-triphosphatase, and guanylyltransferase activities. Measurement of RNA binding by mobility shift and UV photocross-linking analyses also demonstrates the ability of this domain to bind RNA independent of the methyltransferase domain, comprised of the carboxyl terminus of D1R from amino acids 498 -844 and the entire D12L subunit. RNA binding to D1R 1-545 is substantially weaker than binding to either the methyltransferase domain or the holoenzyme. Binding is inhibited by 5-OH RNA and to a lesser extent by DNA oligonucleotides in a concentration dependent manner which correlates with the inhibition of RNA 5-triphosphatase activity by these same oligonucleotides. We conclude that D1R 1-545 represents a functionally independent domain of the mRNA capping enzyme, fully competent in substrate binding and catalysis at both the triphosphatase and guanylyltransferase active sites.
Vaccinia virus, a member of the Poxviridae family, is a double-stranded DNA containing virus that carries out its replication entirely within the cytoplasm of the infected host cell (for a review, see Ref. 1). This life cycle necessitates that most enzymes required for gene expression and DNA replication are encoded by the virus. Many of these enzymes are packaged into the virion and are active in the expression of early viral genes, immediately upon infection. These enzymes include a multisubunit RNA polymerase (2,3), early gene transcription initiation (4,5), and termination factors (6), an mRNA capping enzyme (7-9) and a poly(A) polymerase (10,11).
Capping of vaccinia virus mRNA proceeds co-transcriptionally via a sequential multistep pathway (7) Reactions I-IV are carried out by two vaccinia enzymes present in the virion core. The heterodimeric mRNA capping enzyme catalyzes the first three reactions of this pathway: RNA 5Јtriphosphatase (12,13), guanylyltransferase (7,8), and (guanine-7-)methyltransferase (7)(8)(9). The product of these reactions is the covalent modification of the mRNA 5Ј terminus referred to as the cap 0 structure. Formation of cap I, the product of Reaction IV, requires an mRNA (nucleoside-2Ј-O-)methyltransferase activity encoded by a separate viral enzyme (14). The vaccinia virus mRNA capping enzyme is comprised of two subunits of 97 and 33-kDa molecular mass (7), the products of genes D1R (15) and D12L (16), respectively. In addition to catalyzing the reactions of cap 0 formation, the enzyme also possesses a general nucleoside triphosphate phosphohydrolase activity (12,28,37) and is required for both the termination of early gene transcription (6) and the initiation of intermediate gene expression (17). Both the intact enzyme (18,20,23) and the methyltransferase subdomain (24,40) have been expressed in Escherichia coli, allowing a preliminary localization of the enzyme active sites. The RNA 5Ј-triphosphatase, nucleoside triphosphate phosphohydrolase, and guanylyltransferase activities map to the NH 2 -terminal 60-kDa domain of the D1R subunit (19,21,23), while the (guanine-7-)methyltransferase activity resides in a domain comprised of the carboxyl terminus of D1R from amino acids 545-844 together with D12L (22,23,40). Although the AdoMet and GTP binding sites have been localized to two regions in the carboxyl terminus of D1R (25,26), full methyltransferase activity requires the association of the D12L subunit (23,40). The catalytic independence of the methyltransferase domain has been demonstrated (24).
The amino-terminal 60-kDa fragment of the D1R subunit appears to be sufficient for both guanylyltransferase and phosphohydrolase activities. This portion of D1R contains the lysine residue at position 260 (27,38) required for the formation of the covalent enzyme ϳ GMP intermediate of the guanylyltransferase reaction (29,30,41). The question arises whether the phosphohydrolase and guanylyltransferase activities in the 60-kDa NH 2 -terminal D1R portion of the capping enzyme may be functionally separable into two independent domains. The domain structure of the vaccinia virus enzyme is analogous to the capping enzymes purified from rat liver (42) and brine shrimp (43) in which a single protein possesses both guanylyltransferase and RNA 5Ј-triphosphatase activities. In the latter enzyme, the guanylyltransferase and RNA 5Ј-triphosphatase activities have been fractionated by limited tryptic digestion into 44-and 20-kDa proteolysis products, respectively, separable by gel filtration chromatography (43). Full capping activity, i.e. the ability to employ triphosphate-ended RNA as a substrate for guanylylation, could be reconstituted in vitro by combining the two functional subdomains. In the yeast mRNA capping enzyme, the separation of the RNA 5Ј-triphosphatase and guanylyltransferase active site is more distinct as the two activities are present in separate subunits, encoded by different genes (31,32). These latter two examples argue that the existence of two catalytic subdomains in vaccinia D1R 60-kDa domain is possible.
In this report, we describe in greater detail the structurefunction relationship of the D1R 60-kDa subdomain. Expression in E. coli of carboxyl-terminal deletions in the D1R gene demonstrates a shared functional domain for the RNA 5Јtriphosphatase and guanylyltransferase catalytic sites. A kinetic characterization of the purified D1R 1-545 domain demonstrates a functional equivalence to the holoenzyme in regard to RNA 5Ј-triphosphatase, ATPase, and guanylyltransferase activity. This domain also possesses the ability to bind triphosphorylated RNA independent of the methyltransferase domain demonstrating that the binding of RNA to this domain alone is sufficient for catalysis at the RNA 5Ј-triphosphatase and guanylyltransferase active sites.

Expression Plasmids
Construction of Plasmids Directing the Synthesis of Carboxyl-terminal Deletions of D1R-The structures of the expression plasmids pET3a D1R-63, pET8c D1Nco Ϫ and pET3a D1R/D12L are described elsewhere (23). To generate a family of plasmids capable of expressing a set of carboxyl-terminal deletions of D1R, pET3a D1R-63 was linearized at a single NcoI site at position 2234 and the DNA digested with Bal-31 exonuclease for varying times at 37°C, followed by S1 nuclease treatment. The resulting products were end-filled with the Klenow fragment of DNA polymerase I, ligated, and used to transform E. coli TB1, HMS174(DE3), and BL21(DE3)plysS (33) for the synthesis of a family of nested D1R carboxyl-terminal deletions. The carboxyl terminus of each protein product was determined by DNA sequence analysis of each pET3a D1R⌬COO Ϫ plasmid using primers complementary to the D1R coding region.
pET8c D1R 1-545 -In order to express a truncated D1R protein with a defined terminus at amino acid position 545, the plasmid pET8c D1R 1-545 was constructed (Fig. 1). A DNA fragment (term 168) corresponding to a portion of the D1R sequence from position 1501 to 1737 (including the EcoRI site at position 1546) was amplified by PCR. The 3Ј end of this fragment was modified by introducing two translation stop codons in frame beginning at position 1738 of D1R (34) using a leftward primer (5ЈϾ GGG AGA TCT TTA TTA ATC ATT ATT GGC GTA TTG AT Ͻ 3Ј) containing a BglII restriction site at its 5Ј end for ease of subcloning. The rightward primer was complimentary to D1R sequence beginning at position 1501. The term 168 PCR fragment was subcloned into the D1R expression plasmid pET8c D1Nco Ϫ by ligation of three DNA fragments including: the EcoRI to PstI fragment of the parental plasmid pET8c D1Nco Ϫ , the PstI and BamHI fragment of pET8c, and the EcoRI to BglII term 168 fragment. The ligation products were used to transform E. coli HMS174(DE3), BL21(DE3), and BL21(DE3)plysS.

RNA Synthesis
Templates for RNA Synthesis-The large-scale in vitro synthesis of specific oligoribonucleotides is described elsewhere (24). Briefly, RNA was synthesized by the transcription of linearized plasmid DNA templates by T7 RNA polymerase. To produce a transcript 23 nucleotides in length, the following plasmid was constructed. An oligonucleotide with the sequence AATTAATATATTTG was synthesized and annealed to an oligonucleotide of the complimentary sequence. The resultant doublestranded oligonucleotide, yielding AATT 5Ј overhangs, was cloned into the EcoRI site of pGEM3Zf(ϩ). Transcription of this plasmid linearized at the EcoRI site yielded a 23-mer with the sequence pppGGGC-GAAUUAAUAUAUUUGAAUU. Transcription of pGEM3Zf(ϩ) linearized at the HindIII restriction site yielded RNA 60 nucleotides in length, corresponding to the multiple cloning site sequence.
RNA Synthesis and Purification-For the synthesis of ␥-32 P-labeled RNA employed as a substrate for RNA 5Ј-triphosphatase assays, transcription reactions contained 400 g of linearized plasmid DNA, 1 mM ATP, CTP, and UTP, 0.5 mM GTP, 1 mCi of [␥-32 P]GTP (DuPont NEN, 3000 Ci/mmol) and 300 -500 units of T7 RNA polymerase per 1 ml of reaction. Transcription reactions were carried out for 2 h at 37°C, at which time 100 units of RNase-free DNase 1 (Promega) was added (0.25 unit/g of DNA) and incubated for an additional 30 min at 37°C. The reaction was quenched, and protein was extracted by adding an equal volume of 24:24:1 phenol-STE:chloroform:isoamyl alcohol. The RNA was precipitated with three volumes of 95% ethanol. Transcription products were separated by electrophoresis on 10% acrylamide, 8 M urea gels and excised from the wet gels using an autoradiograph as a guide. RNA was eluted from the gel, precipitated with ethanol, dried, and resuspended in 250 l of diethylpyrocarbonate-treated water. The sample was desalted by applying to a 10-ml Sephadex G-25 column. Peak fractions were identified by liquid scintillation counting, and the samples were pooled and concentrated by centrifugation in a Speed-Vac concentrator (Savant Instruments). Internally labeled, high specific activity [ 32 P]UMP RNA (100 -200 cpm/fmol) was synthesized for mobility shift and photo-cross-linking assays by including 1 mCi of [␣-32 P]UTP per 1 ml of transcription reaction and lowering the concentration of UTP to 0.1 mM; all other ribonucleoside triphosphates were at the standard 1 mM concentration. For the synthesis of low specific activity RNA for guanylyltransferase assays and competition of ATPase and ATP cross-linking assays (37), 5 Ci of [␣-32 P]NTP was included in the transcription reaction, thus facilitating the detection of RNA during purification.

Enzyme Purification
Synthesis and Purification of D1R Subdomains-Transformed BL21(DE3)plysS were induced with IPTG under low temperature conditions as described previously (23), except cells were induced with 200 M IPTG in the presence of 2% ethanol. Cultures were shaken overnight at 15-18°C, after which the cells were harvested by centrifugation and stored at Ϫ70°C until needed.
D1R⌬COO Ϫ -The partial purification of D1R carboxyl-terminal deletion mutants follows the purification scheme described for D1R  and as depicted in Fig. 3 (see "Results") through the DEAE-cellulose step. BL21(DE3)plysS pET3a D1R⌬COO Ϫ cells harvested from 1-liter inductions were frozen and subsequently thawed in lysis buffer (2 ml/g of cells), treated with DNase I (1 mg/10 g cells) to reduce viscosity and centrifuged at 100,000 ϫ g for 60 min at 4°C. Ammonium sulfate was added to 50% saturation, the precipitate collected by centrifugation, resuspended in 3 ml of buffer A (25 mM Tris-HCl, pH 8.0, 1 mM EDTA, 10 mM ␤-mercaptoethanol, 10% glycerol (w/v), and 50 mM NaCl), and dialyzed against 4 liters of buffer A with one change of buffer. DEAEcellulose (10 ml, Whatman DE52), preequilibrated with buffer A, was added to the ammonium sulfate dialysate, and the two components were mixed by gentle rocking at 4°C for 1.5 h. The DEAE-cellulose resin was pelleted by low speed centrifugation, and the supernatant fractions were assayed for protein composition by SDS-PAGE on 12.5% acrylamide gels and by Western blot analysis, using polyclonal antisera raised against the full-size D1R protein. ATPase activity and EϳGMP formation were assayed as described below. A supernatant fraction from BL21(DE3)plysS prepared in an identical manner was used as a control sample.
Purification of D1R 1-545 - Fig. 3 depicts the scheme employed for the purification of D1R 1-545 . The initial steps, from lysis through the 100,000 ϫ g centrifugation step, were conducted as described for purification of the co-expressed capping enzyme subunits (23) with the following modifications. Typically 250 g of BL21(DE3)plysS pET8c D1R 1-545 were thawed in 500 ml of lysis buffer containing 25 mM Tris-HCl, pH 8.0, 1 mM EDTA, 10% glycerol (w/v), 10 mM ␤-mercaptoethanol, 50 mM NaCl, 0.1% sodium deoxycholic acid, 0.01% Nonidet P-40, and 1 mM phenylmethylsulfonyl fluoride. Following the preparation of the polyethyleneimine (PEI) supernatant and high speed centrifugation, ammonium sulfate was added to the S 100 fraction to a final concentration of 60%, stirred for 30 min at 4°C, and centrifuged at 10,000 ϫ g. The pellet was resuspended in a minimal volume of buffer A and dialyzed overnight against 4 liters of buffer A, with three changes. The dialysate was clarified by low speed centrifugation, conductivity was measured, and the sample was adjusted to the conductivity of buffer A. The supernatant was applied to a 4 cm ϫ 20-cm DEAE-cellulose (Whatman DE52) column at a flow rate of 1 ml/min. The flow-through fractions were pooled, concentrated by 60% ammonium sulfate precipitation, and dialyzed against 4 liters of buffer A lacking NaCl, with three changes of buffer. The dialysate was loaded onto a 2 cm ϫ 10-cm heparin agarose (Sigma) column at a flow rate of 0.5 ml/min. The enzyme was eluted from the column by employing a 150-ml linear gradient of 0 -0.5 M NaCl. The fractions were assayed for EϳGMP formation, ATPase activity, absorbance at 280 nm, and SDS-PAGE analysis. D1R 1-545 elutes from heparin agarose at approximately 0.1 M NaCl. Peak fractions were pooled, dialyzed against buffer A, and applied to a 1.5 cm ϫ 10-cm Q Sepharose column (Pharmacia Biotech Inc.) preequilibrated with buffer A. Enzyme was eluted by applying a 100-ml linear gradient of 0.05-0.5 M NaCl. Enzyme activities were measured as in previous chromatographic steps; D1R 1-545 elutes at approximately 0.15 M NaCl. The fractions were pooled and dialyzed overnight against 4 liters low phosphate hydroxyapatite (HT) buffer [10 mM NaHPO 4 , pH 6.8, 50 mM NaCl, 10% glycerol (w/v)]. The dialyzed sample was loaded on to a 1.5 cm ϫ 6-cm hydroxyapatite (Bio-Rad Bio-Gel HT) column. The sample was eluted with a linear phosphate gradient (100 ml) from 10 mM to 500 mM NaHPO 4 . Peak fractions were determined by A 280 and by measuring EϳGMP activities. D1R 1-545 elutes at approximately 0.15 M NaHPO 4 . The pooled fractions were concentrated by ultrafiltration in an Amicon concentrator and the HT buffer was exchanged with buffer A. The concentrated sample was then applied to a 2.5 cm ϫ 48-cm Sephacryl 100 HR (Pharmacia) column at a flow rate of 0.2 ml/min. Fractions (2 ml) were assayed for ATPase, EϳGMP formation, and protein composition by SDS-PAGE analysis. Peak fractions were stored individually at Ϫ70°C.
Determination of Protein Concentration-Protein concentrations were routinely determined by measuring their absorbance at 280 nm and using an estimated E 280 1% of 8.1 and 7.4 for the full-sized capping enzyme and the D1R 1-545 subdomain, respectively, estimated from amino acid composition using the GCG program Peptidesort (Madison, WI). Values obtained by this method were in close agreement with concentrations determined by the method of Bradford.

Enzyme Assays
ATPase-A standard assay contained 50 mM Tris-HCl, pH 8.0, 20 mM MgCl 2 , 10 mM ␤-mercaptoethanol, 5 mM ATP, 0.5-1 Ci of [␥-32 P]ATP (DuPont NEN, 6000 Ci/mmol) and enzyme in a 20-l reaction volume. Enzyme was added last to a prechilled reaction mixture, shifted to 37°C for 5 min, and quenched by adding EDTA to a final concentration of 50 mM. A 2-l aliquot was spotted onto PEI-cellulose thin layer chromatography plates (Alltech) which were developed in 0.75 M LiCl, 1 M acetic acid, adjusted to pH 3.4 (39). 32 P i spots were visualized by autoradiography, and the radiolabeled spots were excised from the TLC plate and quantified by Cerenkov counting. The assay was linear up to 20 min at enzyme concentrations between 50 and 400 nM; 100 nM was chosen as the standard enzyme concentration in kinetic assays. RNasefree bovine serum albumin (at a concentration of 1 mg/ml) was included in all enzyme dilutions.
RNA 5Ј-Triphosphatase-A standard assay was similar to the radiochemical ATPase assay with the following differences: each assay contained 2 mM MgCl 2 and 20 M ␥-32 P-labeled RNA (Ͼ2000 cpm/pmol), in a 10-l reaction volume. Reactions were incubated for 2 min at 37°C and quenched with 2.5 l of 25 mM EDTA, and 2 l were spotted onto PEI-cellulose TLC plates. Plates were developed in 0.75 M KH 2 PO 4 , pH 3.4 (13). The assay was linear for up to 5 min, at enzyme concentrations between 10 and 50 nM. A 10 nM enzyme concentration was typically used in kinetic analyses and the assay was quantified as described above for the ATPase assay. In competition experiments measuring the effect of different oligonucleotides on RNA 5Ј-triphosphatase activity, a 5 M ␥-32 P-labeled 23-mer RNA substrate concentration was utilized. RNA 5Ј-triphosphatase activity was measured in the presence of varying concentrations of oligonucleotides which included diphosphorylated RNA, 5Ј-OH RNA (Oligos, Etc.), and single-stranded DNA, all 23 nucleotides in length and of the same sequence as the triphosphorylated RNA substrate (with uracil replaced by thymine in DNA).
Guanylyltransferase-The guanylylation of RNA comprises two half reactions: the formation of a guanylylated enzyme intermediate (EϳGMP formation) and the subsequent transfer of GMP to the 5Ј terminus of diphosphorylated RNA (the transguanylylation reaction) (7,8). Measurement of EϳGMP formation was as described elsewhere (27). The transguanylylation of RNA was measured in a standard reaction containing 50 mM Tris-HCl, pH 8.0, 5 mM dithiothreitol, 2 mM MgCl 2 , 100 M GTP, 2 Ci of [␣-32 P]GTP (800 Ci/mmol), triphosphorylated RNA 23-mer, and 50 nM enzyme, in a 10-l reaction volume. For RNA saturation experiments, the GTP concentration was held constant at 100 M and RNA was varied up to 20 M; for GTP saturation experiments, the RNA concentration was held constant at 5 M, and the GTP concentration was varied up to 100 M. Reactions were carried out at 37°C for 4 min and were quenched with 10 l of 2 ϫ RNA sample buffer (80% formamide, 20 mM EDTA, 0.05% bromphenol blue, and 0.05% xylene cylanol). The samples were heated for 5 min at 95°C and then loaded onto a 10% acrylamide gel (19:1 acrylamide:bisacrylamide) containing 8 M urea. The gels were run at 150 volts in 1 ϫ TBE buffer (89 mM Tris-HCl, pH 8.3, 89 mM boric acid, 2.5 mM EDTA), fixed for 10 min in 5% acetic acid, 5% methanol, dried, and visualized by autoradiography. 32 P-Radiolabeled bands corresponding to guanylylated RNA were excised and quantified by liquid scintillation counting.
RNA Binding-RNA binding was measured by a gel mobility shift assay. In this analysis, varying amounts of enzyme were preincubated on ice for 15 min in a 20-l reaction containing 50 mM Tris-HCl, pH 8.0, 10 mM ␤-mercaptoethanol, 5% glycerol (w/v), and 20 fmol of [␣-32 P]UMP RNA (200 cpm/fmol). The assay mixture was then loaded onto a native acrylamide gel (6% acrylamide with 19:1 acrylamide:bisacrylamide, 5% glycerol, 0.25 ϫ TBE) preequilibrated at 4°C and run at 200 V. Gels were dried and visualized by autoradiography. The bound RNA was determined either by quantifying the disappearance of free (unshifted) RNA or alternatively as a percentage of the input RNA determined by scintillation counting of both the bound and unbound species.
RNA Photo-cross-linking-The direct photo-cross-linking of RNA to enzyme was carried out as described for ATP (37) except 30-l reactions included 1 M [␣-32 P]UMP RNA 23-mer (ϳ100 cpm/fmol) used as crosslinking reagent, 3 pmol of D1R 1-545 , and varying concentrations of either 5Ј-OH RNA or DNA 23-mer oligonucleotides. The samples were treated with 5 g of RNase A for 30 min at 37°C prior to adding 0.3 volume of 4 ϫ SDS sample buffer. Typically, less than 4% cross-linking of RNA to the enzyme was achieved. (19,21,23,24) has enabled a partial mapping of the catalytic domains for each of the mRNA cap formation activities. The RNA 5Ј-triphosphatase, ATPase, and guanylyltransferase activities reside in a 60-kDa amino-terminal subdomain of the large subunit, D1R. The remaining portion of the heterodimeric enzyme, comprised of the carboxyl terminus of D1R from amino acids 545-844 in addition to the entire small subunit (D12L), constitutes the methyltransferase domain (24,40). In an attempt to define the minimal primary sequence of D1R required for phosphohydrolase activity (37) and to ascertain whether this activity could be separated from the guanylyltransferase active site, carboxyl-terminal deletions in the D1R gene were constructed and expressed in E. coli. BL21(DE3)plysS pET3a D1R ⌬COO Ϫ transformants were induced with IPTG at low temperature, and the truncated D1R proteins were partially purified. These samples were assayed for ATPase and EϳGMP formation, and the relative D1R concentration was determined by Western blot analysis. Each activity was measured relative to the negative control sample derived from E. coli BL21(DE3)plysS grown and induced under the same conditions.The results obtained for four D1R deletion mutants are summarized in Fig. 2. Fig. 2A presents a crude map of the deletion mutants, indicating their respective lengths. Also shown is lysine 260, the residue to which GMP is covalently linked via a phosphoamide bond during guanylyltransferase catalysis (27,38). Extracts prepared from deletions 168, 154, and 153 exhibited comparable ATPase (Fig. 2B) and guanylyltransferase (Fig. 2C) activities. In contrast, mutant 152 showed no detectable EϳGMP formation and exhibited a level of ATPase activity that was indistinguishable from the control. Western blot analysis demonstrated that comparable levels of each mutant protein were present in the induced extracts (Fig. 2D). Deletions shorter than 520 lacked both ATPase and guanylyltransferase activities (data not shown). These results indicate the carboxyl-terminal boundary of the D1R domain lies between amino acids 520 and 545. Failure to obtain a mutant defective in one but not the other activity also suggests the possibility of a spatial proximity of the ATPase and guanylyltransferase active sites.

D1R Carboxyl-terminal Deletion Analysis-Previous work
Expression of D1R 1-545 -Based on the results in Fig. 2, the T7 expression plasmid pET8c D1R 1-545 , Fig. 1, was constructed to direct the synthesis of a truncated D1R protein that terminates at amino acid 545. BL21(DE3)plysS pET8c D1R 1-545 cultures were induced with IPTG as described. The synthesis of an approximately 62-kDa D1R protein was confirmed by SDS-PAGE and Western immunoblot analysis. As was the case for the full-size capping enzyme, heterologously expressed D1R 1-545 resided largely in the insoluble inclusion body fraction. Partially purified soluble enzyme possessed both ATPase activity and the ability to form a covalent EϳGMP intermediate (data not shown).
Enzyme Purification- Fig. 3 summarizes the scheme utilized for the purification of D1R 1-545 . The initial steps (lysis through DEAE-cellulose chromatography) follow the protocol developed for the purification of the full-sized capping enzyme (23). Typically 250 g of cells were harvested from a large scale induction of BL21(DE3)plysS pET8c D1R 1-545 . Complete lysis was facilitated by the endogenous activity of T7 endolysin expressed from T7 gene 3.5 in plysS (35). The 60% ammonium sulfate precipitate was subjected to sequential chromatographic steps (Fig. 3A), and the protein composition was determined by SDS-PAGE, Fig. 3B. Fractions containing D1R 1-545 were identified by measurement of ATPase activity and EϳGMP formation. Following purification by hydroxyapatite chromatography, D1R 1-545 was greater than 75% pure. Throughout the purification, ATPase and guanylyltransferase activities co-eluted with the D1R 1-545 protein. A summary of the purification is listed in Table I. A typical total yield from this purification scheme following resolution by Sephacryl S 100 HR gel filtration chromatography was approximately 12 mg of protein. Fraction 80 was estimated to be greater than 90% pure (Fig. 3B) and enriched approximately 15-fold for ATPase activity relative to the activity of the DEAE-cellulose flow-through fraction.
Kinetic Comparison of D1R  to the Holoenzyme-Preliminary observations indicated that the D1R 1-545 subdomain behaves similarly to the heterodimeric capping enzyme with regard to ATPase and guanylyltransferase activities. Prior to using this subdomain for structure-function studies, a more thorough test of kinetic equivalence to the full-size capping enzyme was undertaken. Both D1R 1-545 and mRNA capping enzyme exhibited comparable substrate saturation kinetics for both RNA 5Ј-triphosphatase (Fig. 4A) and ATPase (Fig. 4B). To assess the relative guanylyltransferase activity, both half-reactions, the formation of the covalent EϳGMP intermediate and the transguanylylation of 23-mer RNA, were measured. In comparison to the intact capping enzyme, D1R 1-545 formed EϳGMP with the same stoichiometry (Fig. 5A) and was equally competent in transferring GMP to acceptor RNA (Fig. 5B). Moreover, the rate of guanylylated RNA formation exhibited the same RNA (Fig. 5C) and GTP (Fig. 5D) concentration dependences. A comparative kinetic summary for ATPase, RNA 5Ј-triphosphatase, and guanylyltransferase activities is presented in Table II. From these analyses, we conclude that the phosphohydrolase and guanylyltransferase kinetic activities in the D1R subdomain are not influenced by the absence of the methyltransferase domain.
RNA Binding-Prior results (24,36) have demonstrated that RNA binding to the mRNA capping enzyme is complex. Both RNA mobility shift analyses and UV photo-cross-linking studies indicate that at least two RNA binding sites exist within separate regions of the large (D1R) subunit corresponding to the NH 2 -terminal 60-kDa domain and the carboxyl terminus of D1R from amino acids 498 -844 which contains the methyltransferase active site. Minimal RNA binding to the small FIG. 4. RNA 5-triphosphatase and ATPase substrate saturation kinetics comparing D1R 1-545 to the intact mRNA capping enzyme. Substrate saturation kinetics were measured for both RNA 5Ј-triphosphatase (A) and ATPase (B) activities, comparing the D1R 1-545 subdomain (E) and the holoenzyme (q). The data presented represents an average of three different experiments, each completed in duplicate. A, RNA 5Ј-triphosphatase velocities were measured in a 10-l reaction volume for 2 min at 37°C, quenched with EDTA and assayed for the release of 32 P i by thin layer chromatography on PEIcellulose plates. B, ATPase velocities were measured in a similar radiochemical assay with the following modifications: each assay was carried out in a 20-l reaction volume for 5 min at 37°C, quenched by the addition of EDTA, and assayed for the release of radiolabeled phosphate as described for the RNA 5Ј-triphosphatase. D12L subunit was observed (24). To further characterize RNA binding to the D1R 1-545 subdomain and to compare its behavior to the methyltransferase domain, we again employed both RNA mobility shift analyses and direct UV photo-cross-linkage of RNA to D1R 1-545 , the methyltransferase domain, or the intact capping enzyme. Mobility Shift Assays-Binding of 23-mer RNA to the intact capping enzyme yields two products (Fig. 6A), a faster migrating species at low enzyme concentrations and an electrophoretically retarded complex at high enzyme levels. The appearance of a slower migrating RNA-protein complex at higher enzyme concentrations concomitant with a reciprocal disappearance of the faster migrating species is consistent with up to two molecules of enzyme bound per 23-mer RNA. Similar results were also obtained with the methyltransferase domain, Fig. 6B. The S 0.5 observed for the holoenzyme and the methyltransferase domain of 35 nM and 95 nM, respectively (Table III), demonstrate similar RNA binding affinities. D1R 1-545 also exhibits two shifted RNA-protein complexes with the slower electrophoretic species predominating at higher levels of D1R 1-545 , Fig. 6C. In comparison to the holoenzyme, however, the D1R 1-545 subdomain binds triphosphorylated 23-mer RNA with an approximately 20-fold weaker affinity (720 nM), Fig.  6D. Taken collectively, these data demonstrate that the methyltransferase domain accounts for a major component of total RNA binding in the intact capping enzyme.
Identical RNA binding saturation results were also obtained when RNA of a different length (e.g. a 60-mer) or RNA possessing a diphosphorylated 5Ј end was employed (data not shown). This fact held true for both the D1R 1-545 and the methyltransferase domains. These results indicate that maximal RNA binding is achieved at modest chain length of 23 nucleotides and that this binding does not require the terminal 5Ј␥-phosphate, at least when measured under the conditions of this assay. This observation agrees with those of Luo and Shuman (36), indicating capped RNA, triphosphorylated RNA, and RNA lacking a phosphorylated 5Ј end all interacted with the intact capping enzyme in a similar manner.
RNA UV Photo-cross-linking-RNA binding was investigated further by directly UV cross-linking radiolabeled RNA to D1R 1-545 . In a preliminary analysis, D1R 1-545 cross-linked RNA with an approximately 2% efficiency at saturating RNA concentrations (1-2 M). The specificity of RNA cross-linking for binding at the RNA 5Ј-triphosphatase active site in D1R 1-545 was tested by measuring the degree of inhibition of both RNA cross-linking and RNA 5Ј-triphosphatase activity in the presence of competing oligonucleotides, Fig. 7. The competing single-stranded DNA and RNA 23-mers possessed the same sequence as the radiolabeled RNA, but lacked a triphosphorylated 5Ј terminus. In these experiments, cross-linked samples were treated with RNase A following UV irradiation, both to remove RNA not covalently associated with the enzyme and to hydrolyze susceptible regions of RNA bound to the protein, thus permitting the migration of the modified protein as a single component during gel electrophoresis. Omission of RNase treatment resulted in a diffuse radiolabeled band (Fig.  7A, lane 1). 5Ј-OH RNA inhibited RNA cross-linking by 50% at approximately 15 M while 23-mer DNA oligonucleotide demonstrated only weak inhibition of RNA cross-linking to D1R 1-545 (Fig. 7B). These data correlate with the relative ability of the same oligonucleotides to inhibit RNA 5Ј-triphosphatase activity (Fig. 7C). In this analysis, diphosphorylated 23-mer RNA exhibited 50% inhibition at less than 10 M, similar to the degree of inhibition observed when cold triphosphorylated RNA was included in the assay to demonstrate a substrate dilution effect. These results indicate that the  diphosphorylated RNA product binds to D1R 1-545 with an affinity similar to the triphosphorylated substrate. In contrast, 5Ј-OH RNA inhibited RNA 5Ј-triphosphatase activity less effectively, with 50% inhibition occurring at 40 M, indicating that RNA lacking a phosphorylated terminus does not bind as well to D1R 1-545 as either tri-or diphosphorylated RNA. Singlestranded DNA did not effectively inhibit RNA 5Ј-triphosphatase activity, demonstrating that the observed binding is restricted to RNA. Although these oligonucleotides inhibited RNA photo-linkage and RNA 5Ј-triphosphatase activity to different extents, the relative pattern of inhibition by 5Ј-OH RNA and single-stranded DNA was consistent in the two experiments. This observation indicates that the RNA binding to D1R 1-545 , measured here by UV photo-linkage, is catalytically productive in RNA 5Ј-triphosphatase activity.

DISCUSSION
The vaccinia virus mRNA capping enzyme catalyzes three sequential reactions in the mRNA cap formation pathway (7-   (20,23), and a separate heterodimeric domain comprised of the carboxyl-terminal portion of D1R from amino acids 545-844 together with the small (D12L) subunit, possessing the (guanine-7-)methyltransferase active site (23,24,40). Therefore, a simple model predicting a single active site for concerted catalysis is impossible; at least two active sites must exist. We favor a model depicting three separate active sites, one for each step in the cap formation pathway. The three site model requires that the RNA 5Ј-triphosphatase and guanylyltransferase activities are catalyzed at separate locations, raising the possibility that the first two active sites may reside in two independent domains. As a first test of this hypothesis, carboxyl-terminal deletions in the D1R gene were constructed, expressed in E. coli, and ATPase and guanylyltransferase activities measured. The results of these measurements confirm that the D1R 60K subdomain is the minimal functional unit for both activities and delimit the carboxyl terminus of this active domain to between amino acids 520 and 545. The failure of this approach to generate deletion mutants exhibiting only ATPase or guanylyltransferase activity may indicate a close conformational linkage of the phosphohydrolase and guanylyltransferase active sites. In a related experiment, we observed that the amino-terminal deletion mutant D1R 184 -545 lacked any measurable ATPase activity or the ability to form the covalent EϳGMP intermediate (data not shown). These results further support the assignment of the ATPase/guanylyltransferase domain structure to include amino acids 1 to approximately 545. Although the physical separation of the two active sites has not yet been achieved, a kinetic argument for separate RNA 5Јtriphosphatase and guanylyltransferase active sites can be made (37).
The assignment of the RNA 5Ј-triphosphatase and guanylyltransferase activities to the D1R 1-545 subdomain potentially simplifies further structure-function analyses. In order to purify enough of this domain, a truncated D1R protein with a defined carboxyl terminus at amino acid 545 was cloned and expressed in E. coli. Purification of soluble enzyme was achieved by large-scale induction of B21(DE3)plysS D1R 1-545 under low temperature conditions followed by sequential chromatographic fractionation of the lysate. The purification of D1R 1-545 builds on the protocol for the purification of the fullsized capping enzyme. Unlike the intact enzyme, however, D1R 1-545 binds poorly to phosphocellulose or hydroxyapatite. This property mimics the chromatographic behavior of the D1R 60-kDa domain proteolysis product observed during the purification of the co-expressed subunits (20,23). As expected, ATPase and EϳGMP activities co-eluted with the D1R protein throughout purification. A yield of 25 mg of soluble D1R protein was typically achieved from the lysis of 250 g of cells.
A kinetic comparison of the D1R 1-545 subdomain to the fullsized capping enzyme demonstrated that the D1R 1-545 domain possessed comparable ATPase and RNA 5Ј-triphosphatase substrate saturation kinetics. Likewise, this subdomain was fully comparable to the holoenzyme in guanylyltransferase kinetics, both in the stoichiometry of EϳGMP formation and in the transguanylylation of RNA. Since the kinetic behavior of D1R 1-545 is equivalent to the intact capping enzyme, the pres-ence of the methyltransferase domain in the holoenzyme must not influence the turnover of RNA at either the RNA 5Ј-triphosphatase or guanylyltransferase active site. A corollary to this fact states that the structural elements which specify RNA binding for the RNA 5Ј-triphosphatase and guanylyltransferase active sites must reside within D1R  .
To investigate the relationship of RNA binding to capping enzyme function, we employed mobility shift and UV photocross-linking analyses as measures of RNA binding to both subdomains and to the full-size capping enzyme. Each subdomain of the mRNA capping enzyme binds to 23-mer triphosphorylated RNA in a manner indicating that up to two molecules of enzyme bind per RNA molecule, at high enzyme concentrations. The affinity of the methyltransferase domain for RNA is nearly 8-fold greater than D1R 1-545 and 3-fold weaker than the intact capping enzyme. This result indicates that the methyltransferase domain possesses the major RNA binding component of the holoenzyme.
Although the binding of triphosphorylated RNA to the D1R 1-545 subdomain is apparently weaker than binding to the methyltransferase domain, it represents catalytically productive binding at the RNA 5Ј-triphosphatase active center. This conclusion is based on the fact that photo-linkage of triphosphorylated RNA is inhibited by 5Ј-OH RNA and to a much lesser extent, single-stranded DNA in a concentration dependent manner, consistent with the relative inhibition of RNA 5Ј-triphosphatase activity by these same oligonucleotides. The relationship of this inhibition to the binding of RNA at the guanylyltransferase active site is not addressed per se by these experiments. We cannot differentiate between separate RNA binding sites for each active site or a single site for the binding of diphosphorylated RNA shared by the two catalytic sites.
In considering these data in the fuller context of a single round of RNA cap formation, it is surprising that that the binding of RNA to the methyltransferase domain is roughly 8-fold stronger than binding to D1R 1-545 , yet the presence of this domain exerts no influence on the RNA substrate saturation kinetics for either the RNA 5Ј-triphosphatase or guanylyltransferase activities. This fact suggests that the binding of RNA to either active site in D1R 1-545 must not be a ratelimiting step in the reaction sequence. Alternately, RNA binding at the methyltransferase active site is unrelated to binding at either catalytic site in the D1R 1-545 subdomain and perhaps reflects a binding function separate from capping at the 5Ј terminus.
A weak but specific association of RNA with D1R 1-545 does fit with the three active site model for mRNA cap formation. In this model, the RNA 5Ј terminus is processed sequentially at each site and in a fashion which is kinetically autonomous from the next step. It is sensible that the nascent RNA binds to the first active site (i.e. the RNA 5Ј-triphosphatase active site) with the greatest specificity but not necessarily the highest affinity. Subsequent to catalysis, the product must move to the next active site; a site exhibiting less specificity for the 5Ј-triphosphorylated RNA but a greater affinity for RNA would promote this transfer.