The Vaccinia Virus mRNA (Guanine-w-)-methyltransferase Requires Both Subunits of the mRNA Capping Enzyme for Activity*

Plasmid vectors capable of expressing the large and small subunits of the vaccinia virus mRNA capping enzyme were constructed and used to transform Esch- erichia coli. Conditions for the induction of the dimeric enzyme or the individual subunits in a soluble form were identified, and the capping enzyme was purified to near homogeneity. Proteolysis of the capping enzyme in bacteria yields a 60-kDa product shown pre- viously to possess the mRNA triphosphatase and guanyltransferase activities (Shuman, s. J. Biol. Chem. 266,11960-11966) was isolated and shown by amino acid sequence analysis to be derived from the NH2 terminus of D1R. The individual subunits lacked methyltransferase activity when assayed alone. How-ever, mixing the D1R and D12L subunits permitted reconstitution of the methyltransferase activity, and this appearance in activity accompanied the associa- tion of the subunits. In contrast, mixing the D12L subunit with the D1R-6OK proteolytic fragment failed to yield methyltransferase activity or result in a physical association of the two proteins. These results dem- onstrate that the methyltransferase active site requires the presence of the D12L subunit with the carboxyl- terminal portion of the DlR subunit. Furthermore, since the mRNA triphosphatase and guanyltransferase active sites reside in the NHz-terminal domain of the D1R subunit, and the methyltransferase

Plasmid vectors capable of expressing the large and small subunits of the vaccinia virus mRNA capping enzyme were constructed and used to transform Escherichia coli. Conditions for the induction of the dimeric enzyme or the individual subunits in a soluble form were identified, and the capping enzyme was purified to near homogeneity. Proteolysis of the capping enzyme in bacteria yields a 60-kDa product shown previously to possess the mRNA triphosphatase and guanyltransferase activities (Shuman, s. (1990) J. Biol.
Chem. 266,[11960][11961][11962][11963][11964][11965][11966] was isolated and shown by amino acid sequence analysis to be derived from the NH2 terminus of D1R. The individual subunits lacked methyltransferase activity when assayed alone. However, mixing the D1R and D12L subunits permitted reconstitution of the methyltransferase activity, and this appearance in activity accompanied the association of the subunits. In contrast, mixing the D12L subunit with the D1R-6OK proteolytic fragment failed to yield methyltransferase activity or result in a physical association of the two proteins. These results demonstrate that the methyltransferase active site requires the presence of the D12L subunit with the carboxylterminal portion of the DlR subunit. Furthermore, since the mRNA triphosphatase and guanyltransferase active sites reside in the NHz-terminal domain of the D1R subunit, and the methyltransferase activity is found in the carboxyl-terminal portion of this subunit and D12L, there must be at least two separate active sites in this enzyme.
Vaccinia virus, the prototypic poxvirus, possesses a doublestranded DNA genome of 191,686 base pairs (1) capable of encoding approximately 200 proteins. Poxviruses replicate within the cytoplasm of the infected cell and must encode the enzymes necessary for viral gene expression and DNA replication (for review, see Ref. 2). Encapsidated within the virion core are the enzymes required for early gene expression: a multisubunit RNA polymerase ( 3 , 4 ) , early gene transcription initiation (5, 6) and termination factors ( 7 ) , poly(A) polymerase (8,9), and the mRNA capping enzyme (10). The availability of these enzymes and the ability to manipulate their coding sequences allow for a unique opportunity to study the mechanism of transcriptional regulation and mRNA processing outside the confines of the nucleus.
Identification of the functional domains required for each of the activities in the cap formation pathway would provide insight into the mechanism of cap synthesis. The formation of a GMP intermediate during the guanyltransferase reaction permitted this activity to be mapped to a 59-kDa proteolytic fragment of the large subunit (17-19). mRNA triphosphatase activity was shown to map to this domain as well (19). Methyltransferase activity, however, did not fractionate with the guanyltransferase activity (17, 19), demonstrating the potential for separate functional units. Since antisera specific for the small subunit revealed that D12L protein co-sedimented with methyltransferase activity, it was proposed that the small subunit possessed the methyltransferase active site and that the tight association of the subunits linked the functional domains. The localization of the large and small subunits to genes D1R (20) and D12L (21) of the vaccinia Hind111 D restriction fragment (22) allowed for overexpression of the subunits in Escherichia coli (18,19). However, the large scale purification of the capping enzyme in a soluble ribonuclease-free form is essential for resolution of the domain structure of the enzyme and for conducting an investigation of both early gene transcription termination and intermediate gene transcription initiation.
This report describes the expression and purification of the vaccinia capping enzyme from recombinant E. coli. Various induction conditions were tested in order to increase the yield The abbreviations used are: AdoMet, S-adenosyl-L-methionine; IPTG, isopropyl-1-thio-6-D galactopyranoside; AdoHcy, S-adenosyl-L-homocysteine; PEI, polyethyleneimine; CAPS, 3-(cyclohexylamino)-1-propanesulfonic acid; SDS, sodium dodecyl sulfate; PAGE, polyacrylamide gel electrophoresis. of soluble protein including decreased growth temperature, addition of ethanol, and choice of BL21(DE3) pLysS. Capping enzyme was purified to near homogeneity and shown to be ribonuclease free. Using either partially purified D1R or D12L, we demonstrated, by mixing experiments, that the methyltransferase activity requires both subunits. Furthermore, we showed that the D12L subunit must associate with the carboxyl terminus of D1R to reconstitute activity.

EXPERIMENTAL PROCEDURES
Expression Plasmids pET3a DI2L-8"Gene D12L lies between 14,350 and 13,487 in the vaccinia HindIII D fragment (22). Plasmid 722, which contains sequence from 12,838 to 16,060 (22), was digested with HpaII at 14,323 and BamHI at 12,838 and the 1,435-base pair fragment isolated. The 5' region of D12L from 14,350 to 14,323 was synthesized as a doublestranded DNA linker possessing an NdeI site at the left end corresponding to the ATG of D12L and an HpaII site at the right end. A three way ligation into pET3a (23), opened at the NdeI and BamHI sites, resulted in the insertion of the entire D12L coding sequence downstream from the bacteriophage T7 gene 10 promoter.
pET3a DIR-63"Gene D1R lies between 103 and 2,637 in the HindIII D fragment (22). The left end of gene D1R was reconstructed by isolating a fragment containing a portion of the 5' end of gene DlR, from the FokI site at 118 to the BamHI site at 514 from plasmid 774a, which contains HindIII D fragment sequence from 1 to 514. A double-stranded DNA linker fragment was synthesized corresponding to the sequence from 103 to 118, containing an NdeI site at the appropriate ATG and a FokI site at the right end. The vector pET3a was cleaved with NdeI and BamHI, and a three way ligation was carried out yielding the plasmid pET3a Dl' which contains the correct 5' end of gene D1R linked to the bacteriophage T7 gene 10 promoter. The bulk of gene D1R was isolated from a derivative of the plasmid 813 which contains sequence from 1 to 4778 (18). A SmBI site at position 3,714 in 813 was converted to a BglII site yielding the plasmid 813Bg. The gene D1R containing fragment was isolated from 813Bg after digestion with XbaI at 363 and BglII at 3714, and the fragment was inserted into pET3a Dl' which was first cleaved with BamHI in the polylinker and then partially digested with XbaI at 363 in the D1R gene. This ligation reconstructed the entire D1R coding sequence downstream from the bacteriophage T7 promoter. pET3a DlRID12L-3-A plasmid encoding both D12L and D1R was constructed by isolating a BglIIIBamHI fragment from pET3a D12L-8 and inserting it into the BglII site in pET3a D1R-63. In this construct both genes lie in tandem but possess their own bacteriophage T7 RNA polymerase promoter and ribosome binding site. Similar constructs were independently generated by Guo and Moss (18) and Shuman (19).
pET3a DIR AC00-217-pET3a D1R-63 was opened at the single NcoI site at 2234 and the fragment subjected to Ba131 digestion for varying times, followed by S1 nuclease treatment. The resulting fragments were end-filled with Klenow fragment of DNA polymerase I, ligated, and used to transform E. coli TB1 and subsequently HMS174(DE3). A family of nested deletions of the carboxyl terminus of D1R was generated that were capable of expressing truncated versions of D1R. One such clone, HMS174(DE3) pET3a D1R ACOO-217, upon induction with IPTG at 37 "C, yields an insoluble protein of approximately 27 kDa.
Expression pET3a D12L-8, pET3a D1R-63, and pET3a DlR/D12L-3 were used to transform E. coli strains HMS174(DE3), HMS174(DE3) pLysS, BL21(DE3), and BL21(DE3) pLysS (23). In order to increase the yield of soluble protein, various induction conditions were tested based on protocols described by Lin et al. (24) and Steczko et al. (25). 2 X Y T media (1.6% tryptone, 1% yeast extract, and 0.5% NaCl) containing ampicillin (50 mg/liter), and when appropriate, choramphenicol (34 mg/liter), was inoculated from a fresh overnight culture and grown at 37 "C to an A m of 0.6. Cultures were chilled in an ice water bath for 15-30 min and induced by adding varying amounts of IPTG, in the presence of 0-4% ethanol, and allowed to shake at 15-20 "C for 24-48 h. The cells were harvested, and the pellets were stored at -70 "C. For large scale preparation of the capping enzyme, the cultures were induced at 400 p~ IPTG and 2.5% ethanol and the cells grown at 15-20 'C for 20 h.

Immunological Techniques
Antisera were raised in rabbits against fusion proteins containing the NH2 terminus of the E. coli (26). The fusion protein contains amino acids 147-621 of D1R. The D12L fusion gene was constructed as described (21), resulting in the synthesis of a fusion protein containing the carboxyl 97% of the D12L protein. In order to raise antisera directed against the carboxyl terminus of DlR, a DNA fragment from the BamHI site at 1959 to a PstI site at 4178 was isolated from plasmid 791 (22) and inserted into PATH10 digested with BamHI and PstI. The resulting fusion protein contains the carboxyl-terminal 225 amino acids of D1R. The pET3a D1R AC00-217-derived protein was used as an antigen for the production of antisera against the NHz terminus of D1R. Immunoblot analysis was carried out as described in the supplier's instructions (Bio-Rad and Millipore). mRNA Guanyltransferase-The formation of the enzyme-GMP intermediate was followed using assays as described (19). Each contained 50 mM Tris-HC1, pH 8.0, 10 mM P-mercaptoethanol, 5 rnM MgCl,, 1-2 pCi of [cY-~'P]GTP (650 Ci/mmol), and column fractions. Reactions were incubated for 10 min at 37 "C, 2 X SDS gel sample buffer (28) was added, and the reactions were boiled for 5 min. The samples were subjected to SDS-PAGE, autoradiographed, and quantified either by densitometry of the autoradiograph or by Cerenkov counting of sections of the gel-containing D1R protein.
Ribonuclease Assay-A 294-base radiolabeled RNA was synthesized from a linear template using bacteriophage T7 RNA polymerase and [cY-~'P]GTP (1 Ci/mmol), phenol-extracted, ethanol-precipitated, and washed. Assays contained 50 mM Tris-HCI, pH 8.0, 10 mM 0mercaptoethanol, 5 mM MgC12,200 pmol of radiolabeled RNA (60,000 trichloroacetic acid-precipitable counts), and 2 pl of enzyme fraction in a 10-pl reaction volume. Reactions were incubated at 37 "C, and at varying times an aliquot was removed and placed in gel sample buffer containing 15% formaldehyde. The samples were applied to a 4% acrylamide, 8 M urea gel, separated by electrophoresis, dried, and autoradiographed.
After elimination of the viscosity, the lysate was homogenized for 1 min at 0 "C in a Waring blender, and insoluble material was removed by centrifugation at 10,000 X g for 15 min. The pellet was resuspended, rehomogenized, and recentrifuged. The supernatants were pooled, and polyethyleneimine was added to a final concentration of 0.05%, stirred for 15 min at 4 "C, and precipitate removed by centrifugation as above. The resulting supernatant was centrifuged at 100,000 X g for 1 h. Ammonium sulfate was added to the SI00 fraction to 45% saturation, stirred for 30 min at 4 "C, and the sample centrifuged at 10,000 x g. The pellet was resuspended in a minimal volume of buffer A (25 mM Tris-HC1, pH 8.0, 1 mM EDTA, 10 mM p- mercaptoethanol, 10% glycerol (w/v), and 50 mM NaCI) and dialyzed overnight against 3 X 4 liters of buffer A. The dialysate was centrifuged a t 10,000 X g and the supernatant applied to a 5 X 25-cm DEAE-cellulose (Whatman DE-52) column equilibrated with buffer A. The flow-through containing the capping enzyme was applied to a -)-methyltransferaye 2.5 X 20-cm phosphocellulose (Whatman P-11) column equilihrated with buffer A and developed with a linear gradient of 0.05-1 M NaCl (200 X 200 ml) in huffer A. The location of the capping enzyme was determined by methyltransferase activity, puanyltransferase activity, and SDS-PAGE analysis. Pooled samples were concentrated by Amicon ultrafiltration and diluted with H T buffer (10 mM NaHPO,, pH 6.8, 10% glycerol, 50 mM NaCI, and 10 mM 13-mercaptoethanol) to achieve a 5-fold reduction in NaCl concentration. The sample was applied to a 2.5 X 10-cm hydroxylapatite (Rio-Rad Rio-Gel HT) column which was developed with a linear gradient of 10-500 mM NaHPO, in H T buffer (100 ml X 100 ml). Capping enzyme was assayed as described above, and fractions containing the capping enzyme were pooled, concentrated, and applied to a 2.5 X 50-cm Sephacryl S-200 High Resolution (HR) column equilihrated with buffer A. Methyltransferase, ATPase, and puanyltransferase activities were assayed and fractions containing activity were stored at DIR Expressed Alone"RLZl(DE3) pLysS pETBa D1R-63 was induced to overexpress the large subunit of the capping enzyme with 400 PM IPTG at 22 "C for 48 h. Purification of 10 g of cells was as described above for the co-expressed suhunits through the ammonium sulfate precipitation step. The resuspended pellet was applied to a 4 X 10-cm Sephadex G25 column equilibrated with huffer A and the desalted protein stored a t -20 "C. 11121, Expressed Alone"RL21(DE3) pLysS pET3a DIZL-8 was induced as above and 25 g of cells were lysed. The supernatant of the 10,000 X , q centrifugation step was centrifuged at 100.000 X g. Nucleic acids were partially depleted from the S100 fraction hy addition of protamine sulfate to 0.5% and centrifugation a t 10.000 X x. The resulting supernatant was made 20% saturated with ammonium sulfate, stirred for 30 min a t 4 "C, and the precipitate removed by centrifugation. The supernatant was brought to 6.5% saturation with ammonium sulfate, the precipitate collected, resuspended in huffer -70 "C. A, and dialyzed overnight against the same buffer (2 X 4 liters). The dialysate was applied to a 4 X 20-cm DEAE-cellulose (Whatman DE-52) column and the location of D12L in the flow-through determined by immunoblot analysis. DZR-60K-BL21(DE3) pLysS pET3a DlR/D12L-3 was induced by addition of 400 p~ IPTG and grown for 48 h at 22 "C. Lysis of 241 g of cells and purification through the dialysis of the ammonium sulfate pellet was as described above for the co-expressed subunits. The dialysate was centrifuged at 10,000 X g, and the supernatant applied to a 4 X 20-cm DEAE-cellulose (Whatman-52) column equilibrated in buffer A. The flow-through was applied to a 2.5 X 20-cm phosphocellulose (Whatman P11) column developed with a 0.05-1 M NaCl (200 X 200 ml) gradient. Fractions containing the capping enzyme were pooled and dialyzed overnight against buffer A containing 0.025 M NaCl (2 X 4 liters). The samples were applied to a 1.5 X 10-cm single-stranded deoxyribonucleic acid cellulose (Sigma) column and eluted with a 0.025-1 M NaCl(100 X 100 ml) gradient. Fractions containing the capping enzyme were pooled, dialyzed overnight against buffer A containing 0.025 M NaCI, and applied to a 1.5 X 10cm heparin-Sepharose CL-GB (Pharmacia LKB Biotechnology Inc.) column and developed with a 0.025-1 M NaCl (75 X 75 ml) gradient. D1R-6OK elutes in fractions ahead of the full size capping enzyme.

Peptide Sequencing
A fraction containing D1R-GOK, the proteolytic fragment of DlR, was concentrated, separated by SDS-PAGE, electroblotted in CAPS buffer onto ProBlott membrane (Applied Biosystems), and stained with Coomassie Brilliant Blue. The NH2-terminal amino acid sequence of D1R-6OK was determined by the University Peptide Sequencing Facility (SUNY Buffalo).

Reconstitution of Methyltransferme Activity
Varying amounts of either the partially purified D1R or the proteolytic fragment, D1R-GOK, were mixed with D12L and allowed to incubate on ice. After 30 min, methyltransferase assays were performed as described above.

Subunit Association
Partially purified D12L and D1R were mixed and after 30 min on ice, and the sample was applied to a 1.5 X 40-cm Sephacryl S-200 (Pharmacia) column. The elution profile of D12L was determined by immunoblot analysis and the profile of D1R determined by detection of the enzyme-GMP intermediate. DlR, D12L, partially purified capping enzyme, as well as molecular weight markers were applied separately to the column to determine their elution profiles.

RESULTS
Induction-The plasmids, pET3a D1R-63, pET3a D12L-8, and pET3a DlR/D12L-3, which express the DlR, D12L, and both D1R and D12L proteins, respectively, were used to transform E. coli strains HMS174(DE3), HMS174(DE3) pLysS, BL21(DE3), and BL21(DE3) pLysS and induced to overexpress the subunits of the vaccinia virus capping enzyme. Consistent with the results of Guo and Moss (18) and Shuman (19), addition of 400 PM IPTG a t 37 "C resulted in the synthesis of insoluble protein. Alternative induction conditions were tested in an attempt to increase the yield of soluble capping enzyme, including lowering the induction temperature (24), altering the concentration of IPTG, the addition of ethanol (25), and varying the time course of induction. Enzyme was partially purified through the dialyzed ammonium sulfate fraction and assayed for methyltransferase activity.   Varying the concentration of IPTG from 0 to 400 p~ at 15-20 "C resulted in a 70-fold increase in methyltransferase activity (Table IA). The inclusion of 2% ethanol to samples containing 50 p~ IPTG produced a %fold increase (Table  IB). In a separate experiment, the addition of 2% ethanol to cells induced with either 200 or 400 p~ IPTG generated a further 1.5-2-fold increase in soluble activity (not shown). Cells grown for 24 or 48 h did not differ in methyltransferase activity. Increasing the percentage of ethanol above 2.5% decreased the yield due to cell lysis (Table IB). HMS174(DE3) and HMS174(DE3) pLysS transformants when induced at 37 "C produced higher yields of total D1R protein than the BL21(DE3) transformed strains. However, when a low temperature induction was attempted with these strains to increase the yield of soluble protein, the cells lysed before they could be harvested. A combination of low growth temperature, induction with 200-400 p~ IPTG in the presence of 2-2.5% ethanol, increased induction time, and choice of strain BL21(DE3) pLysS all contributed in part to a significant increase in soluble, induced capping enzyme. For the D12L subunit alone, high levels of soluble protein are produced by decreasing induction temperature. Unfortunately, the D1R subunit remained primarily insoluble under all induction conditions tested, and the soluble protein present was highly susceptible to proteolysis in the absence of D12L.
Enzyme Purification-Capping enzyme was purified from >-methyltransferme BL21(DE3) pLysS pET3a DlR/D12L-3 induced by the addition of 400 p~ IPTG in the presence of 2.5% ethanol for 24 h at 15-20 "C. Four-hundred grams of induced cells were lysed, treated with DNase I, and insoluble material removed by centrifugation at 10,000 x g for 15 min. The cleavage of D1R into a prominent 60-kDa peptide observed by Guo and Moss (18) and Shuman (19) is greatly inhibited by the presence of phenylmethylsulfonyl fluoride in the lysis buffer; other protease inhibitors were not tested. Following centrifugation, PEI was added to 0.05%, and the nucleic acid precipitate was removed by centrifugation. The PEI supernatant was centrifuged at 100,000 x g and the SI00 fraction treated with ammonium sulfate to 45% saturation. The 45% ammonium sulfate fraction was dialyzed and applied to DEAE-cellulose. The flow-through fraction, containing the capping enzyme, was applied to a phosphocellulose column. The methyltransferase and guanyltransferase activities co-elute at 0.17 M NaCl, as reported previously (10, 18,19). Fractions from the phosphocellulose column were pooled, applied to hydroxylapatite, and eluted with a NaHP04 gradient (Fig. 1, A and B ) . The elution profile of this column revealed the separation of two peaks of guanyltransferase activity. The full size D1R (Fig. 1B) co-eluted with D12L (not shown) and methyltransferase activity at approximately 300 mM NaHP04. D1R-GOK, the proteolytic product of DlR, was well resolved from the full size capping enzyme, eluting at 170 mM NaHP04 (Fig. 1B). Methyltransferase activity appears to peak two or three fractions ahead of the major guanyltransferase activity. This may be due to a partial separation of two methyltransferase components present in the activity peak (see below).
Fractions 59-69, which contained the full size capping enzyme, were pooled, concentrated, and applied to a Sephacryl S200 HR sizing column. The resulting profile exhibited two methyltransferase activity peaks (Fig. 1C). ATPase and guanyltransferase activities (Fig. 1, C and D) co-eluted with the first methyltransferase activity peak. SDS-PAGE analysis (Fig. 2 A ) revealed the presence of both the D1R subunit and the D12L subunit, demonstrating that the first peak contains full size capping enzyme. An analysis of the second broader methyltransferase component revealed a lack of ATPase activity (Fig. IC) and very low guanyltransferase activity which consisted of a minor amount of D1R and its proteolytic product D1R-6OK (Fig. 1D). Coomassie Brilliant Blue staining of SDS-PAGE gels revealed the presence of D12L in both methyltransferase peaks (Fig. 2 A ) . In addition to D12L, the second methyltransferase component contains a family of slower migrating bands of 30-40 kDa as seen in the Coomassie Brilliant Blue-stained gel. In order to determine the relationship of this family of proteins to the D1R subunit of the capping enzyme, antisera directed against either the NH2 terminus or carboxyl terminus of D1R was generated. Immunoblot analysis revealed that the non-D12L proteins present in the second methyltransferase peak were derived from the carboxyl terminus of D1R (Fig. 2C). NHp-terminal proteolytic fragments found preferentially in the first activity peak (Fig. 2B) are due to nicking of the large subunit in such a way that the products remain in the enzyme complex and are released after denaturation in SDS. These results demonstrate that the methyltransferase domain is separable from the mRNA triphosphatase, guanyltransferase, and ATPase active sites and that this domain must lie either within the D12L subunit alone, the carboxyl terminus of DlR, or, alternatively, be shared between these two proteins (17, 30).
Capping enzyme produced in the E. coli expression system was purified to near homogeneity through the scheme described above (Table I1 and Fig. 3). DEAE-cellulose chromatography resulted in a 13-fold purification of the 45% am- monium sulfate pellet. Methyltransferase activity appeared to increase as endogenous AdoMet and potential inhibitors were removed by this resin. The phosphocellulose column resulted in a 3.5-fold increase in the specific activity of guanyltransferase and a 5-fold increase in methyltransferase specific activity. There was a significant decrease in total activity as compared with the DEAE-cellulose pool, since, based on SDS-PAGE analysis, only 60% of the phosphocellulose column peak was pooled and carried through the next step. The elution of the capping enzyme from the hydroxylapatite column resulted in a significant increase in purity of the enzyme leaving a few minor contaminating proteins (Fig. 3). The high NaHP04 required to elute the enzyme inhibits ATPase activity and partially inhibits the guanyltransferase. The Sepha-cry1 S200 HR column fractionated the hydroxylapatite pool into two general families of proteins ( Figs. 2A and 3). Fractions from this column were not pooled, and aliquots were stored at -70 "C. Fractions 50 and 58 were assayed for methyltransferase activity. The specific activity did not significantly increase following elution from this column, since few contaminating proteins are removed. ATPase and enzyme-GMP formation activity were detected in fraction 50 but not fraction 58 (Fig. 1, C and D). Minor contamination observed in Fig. 3, lune 50, are proteolytic products of the D1R protein (Fig. 2C). Approximately 23% of the capping enzyme in fraction 50 is in an enzyme-GMP complex under standard assay conditions based on a protein concentration determined by the Bio-Rad Protein Assay.
Since ribonuclease contamination in the capping enzyme preparation would make detailed analysis of enzymatic activity difficult, the hydroxylapatite pool as well as fractions 50 and 58 from the Sephacryl S200 HR column were tested for ribonuclease activity. Enzyme was added to a 294-base 32Plabeled RNA under assay conditions that mimicked the guanyltransferase reaction, and samples were removed at various times and analyzed by gel electrophoresis. Over the time course of the assay, ribonuclease activity was not detected in the hydroxylapatite pool or in the subsequent Sephacryl S200 HR fractions (Fig. 4). .
The localization of the guanyltransferase and mRNA triphosphatase to D1R was independently determined by Guo and Moss (18) and Shuman (17,19). Shuman and Morham (31), through the use of carboxyl-terminal deletions, further mapped these activities to the amino two-thirds of the large subunit. The mRNA triphosphatase must also map to this fragment, since the guanyltransferase assays contained triphosphate terminated poly(A) RNA as a substrate. We found that the purified proteolytic product, D1R-GOK, contains the ATPase and GTPase activities as well (not shown). The proteolytic fragment, D1R-GOK, was separated from the capping enzyme on both hydroxylapatite (Fig. 1B) and heparinagarose (Fig. 5A). In order to localize the 60-kDa proteolytic product within DlR, a fraction from a heparin-agarose column containing D1R-6OK was concentrated, separated on SDS-PAGE, transferred to a ProBlott membrane, and the NH2-terminal sequence determined. The sequence, MDANVVSSST, corresponds to the first 10 amino acids of DlR, demonstrating that the guanyltransferase, mRNA triphosphatase, and ATPase domains lie within the NHn-ter-minal60 kDa of the large subunit.
Reconstitution of Methyltransferase Activity-In order to determine which subunit of the capping enzyme is required for methyltransferase activity, a plasmid encoding each subunit was introduced into BL21(DE3) pLysS and induced at 22 "C. D12L was fractionated to approximately 20% purity (Fig. 5A) and appeared stable throughout the purification, unlike the subunit induced at 37 "C in BL21(DE3) (18). Cells induced to synthesize D1R did not yield high quantities of soluble enzyme. Furthermore, D1R was more susceptible to proteolysis in the absence of D12L and so proved difficult to purify to any great extent (Fig 5A). The proteolytic fragment of DlR, D1R-GOK, was separated from the full size capping enzyme by passage over a heparin-agarose column and is approximately 10% pure (Fig. 5A). Each extract was assayed for the ability to form an enzyme-GMP complex (Fig. 5B). As expected, fractions containing D12L alone lacked guanyltransferase activity, whereas samples containing D1R formed an enzyme-GMP complex. Fractions 35 and 37 derived from mRNA (Guanine-N-)-methyltransferuse Sephacryl S200 Chromatography the heparin-agarose column lacked full size capping enzyme but contained active D1R-6OK. Comparing partially purified D1R expressed alone and D1R-6OK fractions, there is a t least 20 times less guanyltransferase activity in the D1R preparation. However, impurities in the D1R preparation may have decreased the enzyme-GMP formation activity of the fraction.
Methyltransferase assays performed using DlR, D1R-GOK, or D12L alone lack methyltransferase activity. In order to determine if methyltransferase activity could be reconstituted by mixing the subunits, assays containing varying amounts of D1R and D12L were performed (Fig. 5C). As the level of D12L was increased from 0.0012 to 0.12 pg, at different D1R levels, an increase in the incorporation of methyl groups into "GTP was seen. This increase was dependent on the presence of GTP (Fig. 5C), and ATP could not substitute as as an acceptor (not shown). Therefore, both the D1R and D12L subunits are required for the vaccinia virus methyltransferase activity. In order to determine if reconstitution of methyltransferase activity required the carboxyl terminus of the D1R subunit, the addition of D12L to the NH2 terminus of DlR, the D1R-6OK fragment, was tested. Mixing experiments using aliquots of D12L and D1R-6OK were performed (Fig.  50). Methyltransferase activity was not detected after mixing D1R-6OK with D12L, demonstrating that the COOH-terminal region of D1R is required to reconstitute methyltransferase activity. These results are consistent with the presence of D1R carboxyl-terminal fragments in the second methyltransferase peak of the Sephacryl S200 HR column (Fig. 2C) and confirm the observations of Cong and Shuman (30).
In order to determine if a physical association of the two subunits is required for reconstitution of methyltransferase activity, the partially purified subunits were mixed and applied to a Sephacryl S200 sizing column (Fig. 6). Co-expressed capping enzyme subunits yielded a profile in which guanyltransferase and methyltransferase co-elute. An extract containing D1R in the absence of D12L eluted primarily as a high molecular weight aggregate with a trailing shoulder of enzyme-GMP complex formation activity. When extracts containing D12L were applied to the column, the small subunit peaked in fractions 31 and 32 as determined by immunoblot analysis (not shown). Extracts of DlR, containing its proteolytic fragment, D1R-GOK, and D12L expressed separately, mixed, and applied to the column resulted in a shift in the elution profile of both D1R and D12L. A fraction of the D1R aggregate dissociated and the enzyme-GMP complex formation now eluted in the region of the native capping enzyme. A portion of D12L eluted as a higher molecular weight species at the position of the co-expressed capping enzyme. Methyltransferase activity was now detectable and eluted similarly to the native capping enzyme (Fig. 6B). Therefore, the D1R and D12L subunits must associate to reconstitute methlytransferase activity. The elution of the D1R-6OK fragment was unchanged by the presence of D12L (Fig. 6C), providing further evidence that the NH2-terminal domain of D1R does not bind to the D12L subunit.

DISCUSSION
In order to purify enough of the vaccinia virus mRNA capping enzyme to carry out physical studies, it was necessary to overexpress the subunits in soluble form well above the levels found in virions or in virus-infected cells. Expression vectors were constructed which direct the synthesis of capping enzyme in E. coli, and induction conditions were tested in order to maximize the yield of soluble protein. Both a reduction in induction temperature and choice of BL21(DE3) pLysS increased solubility significantly, perhaps due to slowing the rate of protein synthesis, thus enhancing correct protein folding. In contrast to inductions carried out at 37 "C (18, 19), addition of IPTG to 400 p M at 15-20 "C resulted in a 70-fold increase in soluble protein. Ethanol, thought to induce heat shock chaperone proteins (25), increased solubility of the capping enzyme a further 2-3-fold.
A protocol for the rapid purification of the capping enzyme that resulted in nearly homogeneous protein was developed. Purification through the hydroxylapatite column yielded en-zyme that was 90% pure and ribonuclease-free. Chromatography of the hydroxylapatite pool on Sephacryl S200 HR resolved two peaks of methyltransferase activity. The first peak was shown by SDS-PAGE analysis to contain full size capping enzyme, exhibiting both the ATPase and guanyltransferase activities. The second peak contained the small subunit, D12L, in addition to a family of proteins shown by immunoblot analysis to be carboxyl-terminal fragments of the large subunit D1R. The number of proteolytic fragments in this methyltransferase fraction varied between preparations of enzyme.
A 60-kDa proteolytic fragment of D1R formed in E. coli is capable of catalyzing the mRNA triphosphatase and guanyltransferase reactions (18,19). Shuman and Morham (31) localized this fragment through deletion analysis to the amino-terminal two-thirds of D1R. We sequenced the NH2terminal portion of the 60-kDa fragment and demonstrated that it was derived from the NH2 terminus of the large subunit. Additionally, we localized the ATPase/GTPase activity to this 60-kDa domain (not shown).
The partially purified D1R and D12L subunits individually lack methyltransferase activity. Through mixing experiments both subunits were shown to be required to reconstitute methyltransferase activity. In order to determine if association of the subunits was necessary for activity, a mixture of D1R and D12L was applied to a Sephacryl S200 column. Methyltransferase activity was found only in fractions which contained the associated subunits. In contrast, when D1R-60K, the NH2-terminal fragment of DlR, was mixed with D12L it both failed to reconstitute activity and physically associate with the small subunit. These results demonstrate that the reconstitution of the methyltransferase activity requires a physical association between the D12L subunit and the carboxyl-terminal domain of the D1R subunit. This model is supported by the co-elution of carboxyl-terminal fragments of D1R in the second peak of methyltransferase observed after Sephacryl S200 HR chromatography. Furthermore, the results of Cong and Shuman (30) are in full agreement with these conclusions.
A model of the functional domain structure of the mRNA capping enzyme based on prior experiments (17, 19, 31) and these results is presented (Fig. 7). The carboxyl terminus of D1R associates with D12L forming a dimeric protein complex capable of catalyzing the first three reactions in the capping pathway. In E. coli, a protease preferentially cleaves D1R releasing an NHz-terminal 60-kDa domain and a family of 30-40-kDa COOH-terminal fragments. The NH2-terminal domain of DlR, D1R-GOK, contains the GTPase, ATPase, mRNA triphosphatase, and guanyltransferase active sites. The D1R COOH-terminal domain associated with the D12L subunit contains the methyltransferase active site.
The presence of multiple domains possessing enzyme activities required for cap formation establishes the presence of at least two active sites required for mRNA capping and at least three different sites capable of binding GTP for the GTPase, guanyltransferase, and methyltransferase activities. The arrangement of the mRNA triphosphatase and guanyltransferase activities within a single complex active site or multiple sites is currently being investigated. The methyltransferase activity is clearly carried out at a different location. We are currently attempting to identify the subunits involved in AdoMet and GTP binding at this site in order to determine the role of each subunit in methyltransferase activity.
This arrangement raises several fundamental questions related to the number of mRNA binding sites on the enzyme and the need to move the 5' end of the mRNA product of the guanyltransferase reaction to the methyltransferase active site. Currently, we are trying to further localize the functional domains of the catalytic activities of this remarkable enzyme as well as understand its role in both viral early gene transcription termination and intermediate gene expression.