Dinucleotide Priming of Transcription Mediated by RNA Polymerase

Mammalian RNA polymerase I1 was shown to utilize dinucleoside monophosphates for priming of promoter specific RNAs. In a reconstituted system containing purified polymerase and HeLa cell fractions, dinucle- otides were incorporated by complementarity with template sequences at the in vivo cap sites of the adenovirus major late and adenovirus early region IV promoters. Incorporation was shown by label transfer experiments and by determining the size of 5”terminal RNase T1-resistant oligonucleotides. All 16 dinucleotides were tested for priming of RNA chains at the major late promoter. RNA polymerase I1 initiated with various primers over a contiguous region of 9 bases, centered around the in vivo initiation site. We suggest that the polymerase drifts or oscillates over this region. Using a dinucleotide challenge protocol, the rate of initiation at the major late promoter was measured following preincubation of the template DNA with RNA polymerase I1 and factors. Initiation with ATP was 90% complete within the 1st min after addition of nucleotide triphosphates. Stimulation of transcription by dinucleotides was not observed, due to this rapid initiation. The 5”hydroxyl terminus of dinucleotide- primed RNAs remained unmodified. Although transcripts initiated with were rapidly capped in whole cell extracts, ATP-primed RNA synthesized in the reconstituted system retained free 5”terminal phosphates. Thus, capping was not essential for syn- thesis of long runoff RNAs. single-stranded M13 XEll DNA. S’-proximal 250 nucleotides of EIV RNA hybridize to this M13 probe. hybridization, free 3’ tail of the runoff was digested RNase T1, also hydrolyzed much of the nonspecific background RNA. were collected by filtration through nitrocellulose membranes, 250-nucleotide truncated RNA resolved on an 8% sequencing-type early region PM CTP, UTP, lane lane UTP lane CTP reproducibly M13 annealed

Mammalian RNA polymerase I1 was shown to utilize dinucleoside monophosphates for priming of promoter specific RNAs. In a reconstituted system containing purified polymerase and HeLa cell fractions, dinucleotides were incorporated by complementarity with template sequences at the in vivo cap sites of the adenovirus major late and adenovirus early region IV promoters. Incorporation was shown by label transfer experiments and by determining the size of 5"terminal RNase T1-resistant oligonucleotides. All 16 dinucleotides were tested for priming of RNA chains at the major late promoter. RNA polymerase I1 initiated with various primers over a contiguous region of 9 bases, centered around the in vivo initiation site. We suggest that the polymerase drifts or oscillates over this region. Using a dinucleotide challenge protocol, the rate of initiation at the major late promoter was measured following preincubation of the template DNA with RNA polymerase I1 and factors. Initiation with ATP was 90% complete within the 1st min after addition of nucleotide triphosphates. Stimulation of transcription by dinucleotides was not observed, due to this rapid initiation. The 5"hydroxyl terminus of dinucleotideprimed RNAs remained unmodified. Although transcripts initiated with ATP were rapidly capped in whole cell extracts, ATP-primed RNA synthesized in the reconstituted system retained free 5"terminal phosphates. Thus, capping was not essential for synthesis of long runoff RNAs.
In recent years, transcription mediated by eukaryotic RNA polymerase I1 has been faithfully reproduced in vitro (1,2).
The primary criterion for this reaction has been the formation of a correct 5' end, as determined by comparison with the sequence of i n vivo RNA. Analysis of mutants has shown that many of the same sequences promote initiation of transcription both i n vivo and i n vitro (3)(4)(5)(6)(7)(8). However, little is known of the details of the reaction. Several factors are required in addition to purified RNA polymerase I1 to obtain accurate transcription (9)(10)(11)(12), suggesting that the complete reaction pathway is complex.
We have begun an analysis of the transcription reaction, Recipient of National Institutes of Health Program Project Grant Pol-CA26717. T o whom reprint requests should be addressed. using a reconstituted system consisting of purified RNA polymerase I1 and partially purified HeLa cell protein fractions (12). The transcriptional activities in these fractions each titrate linearly with protein concentration and yield single peaks of activity after sedimentation through sucrose gradients. Employing a preincubation-pulse-chase protocol, several steps in the transcription reaction have been resolved (13). Components of the transcriptional apparatus can associate with the DNA to form template-specific "activated complexes during preincubation in the absence of nucleotides.
During the pulse, these complexes rapidly incorporate radioactive nucleotides into RNA chains. In addition, the activated complexes are resistant to a challenge by inhibitory concentrations of DNA. Stable protein-DNA complexes have also been observed in template competition experiments with eukaryotic RNA polymerases I1 and I11 (14)(15)(16).
A variety of physical and kinetic studies of Escherichia coli RNA polymerase has revealed a multistep pathway leading from free enzyme to initiation (17). Several intermediates occur following binding of RNA polymerase to template DNA (18,19); the final "open" complex appears to involve invasion of the double helix by the enzyme (20)(21)(22). The mechanism of initiation by E. coli RNA polymerase has been studied by the use of special primers in place of nucleoside triphosphates. The bacterial RNA polymerase can utilize a variety of short oligonucleotides to start RNA transcripts (23,24). In particular, dinucleotides specifically stimulate initiation at promoter sites having complementary sequences; different dinucleotides can be used to alter the 5' termini of transcripts from a given promoter (25-28).
Eukaryotic RNA polymerases I and I11 have recently been shown to use dinucleotides for RNA chain initiation (29,30).' As yet, incorporation of dinucleotides by RNA polymerase I1 at promoter sites has not been reported. However, the apparent stuttering of this enzyme during transcription of the polyoma early region suggests that RNA polymerase I1 can carry out transcription with oligonucleotide primers (31).
Transcription by RNA polymerase 11 differs from transcription by the other eukaryotic RNA polymerases in that the RNA polymerase I1 transcripts are modified by the addition of a cap (32). A majority of the promoter-specific RNAs synthesized in soluble cell extracts are also capped (1,2). RNAs synthesized in isolated cytoplasmic polyhedrosis virions are similarly capped; in this system, transcription and capping appear to be mechanistically coupled (33). The rapid kinetics of capping heterogeneous nuclear RNAs i n vivo (34) together with the observation that RNA polymerase I1 transcription is inhibited i n vitro by S-adenosylhomocysteine (35) have led to the speculation that cap formation is obligatory for transcription.
We describe here the use of dinucleoside monophosphates to probe transcription initiation by RNA polymerase 11.

MATERIALS AND METHODS
Unlabeled and a-32P-labeled high performance liquid chromatography-purified nucleoside triphosphates were purchased from ICN and New England Nuclear, respectively. Dinucleotides were from Sigma, ICN, or Boehringer Mannheim. The identities of all dinucleotides were confirmed by thin layer chromatography on polyethylenimine-cellulose, both in 0.1 N acetic acid and in 0.16 M LiCl (36). RNases T1, T2, and U2 were purchased from Calbiochem, RNase A was from Sigma, and proteinase K was from Boehringer Mannheim. Nitrocellulose filters, with 25-mm diameter and 0.45-p porosity, were from Schleicher and Schuell. E. coli strains infected with phage M13 mplO or M13 m p l l were a kind gift of J. Vieira and J. Messing (Dept. of Biochemistry, University of Minnesota). End-labeled marker tRNA was a gift of Harold Drabkin.
Transcriptional Proteins-Glycerol gradient-purified calf thymus RNA polymerase I1 (37), HeLa whole cell extracts (2)  M I 3 Probes-The XhoI-Hind111 fragment of pFLBH (-261 to +197 of the major late promoter) was inserted between the Sal1 and HindIII sites of both M13 mplO and mpll. Phage M13 X H l l (the m p l l recombinant) contains the strand complementary to mfior late promoter RNA, while M13 XHlO (the mplO recombinant) contains the opposite strand. For hybridization to in vitro synthesized RNA, recombinant phage DNA was purified by CsCl gradient ultracentrifugation of protease-digested phage particles. This density gradient step was important for reproducible hybridization by different phage preparation^.^ The EcoRI-XmaI fragment of pHinI5 (-331 to +250 of the EIV promoter) was inserted between the EcoRI and XmaI (SrnaI) sites of both M13 mplO and mpll. The resulting recombinants (M13 XElO and M13 XE11) were prepared as phage DNAs. M13 X E l l hybridizes to EIV RNA, while M13 XElO contains the opposite DNA strand.
In Vitro Transcriptions-The three-stage protocol and reaction conditions are described in detail in the preceding paper (13). Briefly, RNA polymerase I1 (20 units GTP (3000 Ci/mmol). In some cases, the pulse phase also contained one of the 16 dinucleotides, at 2 mM final concentration. After a short pulse of 3 or 4 min, 5 pl of chase nucleotides were added to give final concentrations of 1 mM for each of the four NTPs. Concentrations of glycerol and salts were maintained through each phase of the reaction. The chases were in all cases adequate for complete elongation of runoff transcripts.
Analysls of RNA-Reactions were stopped and the nucleic acids urea sequencing-type gels. Long runoffs were resuspended in 10 pl of 50 mM Hepes-NaOH, pH 7.0, 1 mM EDTA for M13 analysis. M13 selection was performed according to Hansen and Sharp4 (see Fig.  1B). To each reaction, 0.2 pg of recombinant M13 phage DNA was added, and NaCl was added to 0.75 M. Hybridization mixtures were heated to 70-80 "C for 5 min and then incubated at 50 "C for 2-5 b.
Hybridizations were stopped with 200 p1 of cold quench buffer (10 mM Hepes-NaOH, pH 7.5, 1 mM EDTA, 0.2 M NaCl). Hybrids were treated with 5 units of RNase T1, a single-strand RNA-specific endonuclease, for 30 min at 30 "C; the T1 was subsequently destroyed by digestion with 50 pg of proteinase K for 60 min a t 30 "C.
The hybrids were filtered slowly (about 1 ml/min) through nitrocellulose filters. Filters were washed with 5-10 ml of quench buffer. The bound nucleic acids were eluted by boiling for 5 min in 1.5 ml of 2 mM EDTA, pH 7.0, plus 25 pg of carrier tRNA, and cooled rapidly in ice water. The nucleic acids were precipitated twice from ethanol, and resuspended in 80% formamide, 0.05 M Tris-borate, pH 8.3, 1.25 mM EDTA for direct acrylamide gel electrophoresis.
For preparation of T1 oligonucleotides, the M13-selected RNA dissolved in 2 mM EDTA, pH 7.0, was heated for 5 min at 90 "C, quick chilled in ice water, and digested with 5-10 units of RNase T1 for 2 h a t 30 "C. 50 pl of water freshly saturated with diethyl pyrocarbonate were added to inactivate the RNase, and reactions were lyophilized three times with 50-pl water washes. The resulting oligonucleotides were resuspended in 80% formamide, 0.05 M Tris-borate, pH 8.3, 1.25 mM EDTA and electrophoresed on 0.2-mm thick, 15% polyacrylamide-urea gels.
For RNase U2 digestion, RNAs dissolved in 2 mM EDTA, pH 7.0, were digested with RNase T1 as above. After the T1 incubation, sodium citrat.e, pH 3.5, was added to 20 mM. After another round of heating and rapid cooling, 0.02 unit of RNase U2 was added for 30 min a t 30 "C. The digestion products were lyophilized, dissolved, and electrophoresed as described above. Decapping was performed by periodate oxidation and ,&elimination with lysine (as described; see Ref. 41) before MI3 selection. Phosphatase treatments were performed after M13 selection, with 16 units of bacterial alkaline phosphatase (Worthington) for 45 min at 55 "C in 10 mM Tris, pH 7.9, 1 mM EDTA, 0.05% sodium dodecyl sulfate. For unknown reasons, late promoter RNA was highly resistant to phosphatase at 30 "C.
For label transfer analysis, oligonucleotides from complete T1 digestion of preincubation-pulse-chase RNAs were gel purified, and digested either with RNase T2 (1 unit, 37 "C overnight in 20 mM ammonium acetate, pH 4.5) or with RNase A (0.4 pg, 37 "C for 60 min in 10 mM Tris, pH 7.4, 1 mM EDTA). The digestion products were resolved by chromatography on cellulose thin layers in one or two dimensions (42).
To confirm that the decapping reaction had worked, [LY-~'P]UTPlabeled late promoter RNA was synthesized by a whole cell extract in a simple 3-h reaction. The 5'-terminal41-nucleotide fragment was isolated as described (43). Aliquots were removed for decapping and/ or phosphatase treatment. The products of complete RNase T2 digestion were resolved by two-dimensional thin layer chromatography. From untreated, decapped, and decap + phosphatase reactions, spots were observed with the mobilities expected for GpppAmCp, pppAmCp, and AmCp, respectively (2'-0-methylation protected the A from T2 cleavage). A spot possibly consisting of m'AmCp was also observed.
Quantitation-Autoradiograms obtained with response-linearized (preflashed) film were quantitated by densitometry and planimetry (12). For estimates of absolute transcriptional efficiency, T1 oligonucleotide bands were excised from acrylamide gels and counted for Cerenkov radiation. The number of RNA chains initiated in the pulse was calculated from the specific activity of the labeled nucleotide.

RESULTS
Transcription reactions were performed using a reconsti-  3 ) . When recombinant phage M13 XE10, containing the noncoding DNA strand, was annealed to transcription products, no labeled material was protected from ribonuclease digestion. The 237-nucleotide marker was derived from HinfI-cleaved, end-laheled SV40 DNA. The 200 nucleotide marker was a truncated runoff RNA from the major late promoter. formation of "activated" complexes. Radioactive nucleotides were then added for a brief pulse, during which elongation proceeded only a short distance. An ensuing chase with excess unlabeled nucleotides allowed the completion of long runoff RNAs.
T o examine initiation by the activated complexes, pulse nucleotide concentrations were varied. The template for these experiments contained the Ad5 early region IV promoter (see Fig. 2R). Previous work with this promoter has shown that the site of initiation in oitro exhibits heterogeneity over a seven-nucleotide region in the whole cell extract, thereby accurately reproducing the heterogeneity observed in vivo (44). Moreover, the relative amounts of the various termini depend on the nucleotide concentrations in the reaction. Thus, high UTP concentrations favor the use of U initiation sites (residues -6 to -l ) , while high ATP concentrations favor initiations a t A (residue + l ) .
Analogous results were obtained with the reconstituted system using the three-stage protocol. EIV runoff RNAs were synthesized under conditions of either high ATP or high UTP concentration in the pulse. These RNAs were analyzed for differences at the 5' end by the method of Hansen and Sharp: For this analysis, the long runoff RNAs were hybridized to a single strand M13 recombinant containing a segment of adenovirus complementary to the first 250 bases of the EIV transcripts (see Fig. 2A and "Materials and Methods"). The runoffs were truncated by ribonuclease digestion, leaving a 250-nucleotide promoter-proximal RNA, which was selected by filtration through nitrocellulose. This procedure yields RNAs with common 3' termini so that length variations must reflect differences at the 5' end.
On high resolution gels, the high [UTP] reaction product migrated more slowly than the high [ATP] reaction product (Fig. 2C, lunes 2 and 3), suggesting a shift to upstream initiation sites. Direct analysis of short runoff RNAs gave similar results (data not shown). Since all labeled transcripts were subjected to this shift, these transcripts must be initiated during the pulse phase.
Initiation with Dinucleotides ut the EIV Promoter-At low nucleotide concentrations, E. coli RNA polymerase will utilize dinucleotides to initiate RNA chains, incorporating the dinucleotides according to base-pairing rules (24)(25)(26)(27)(28). To test whether RNA polymerase I1 behaved similarly, 5' shift experiments were performed using dinucleotides with the EIV promoter. When UpU was added to a transcription pulse, the resulting RNA migrated more slowly than the control RNA initiated a t low nucleotide concentrations (Fig. 2C, lanes 1  and 7). This UpU RNA co-migrated with RNAs initiated at high [UTP] (compare lunes 3 and 7), indicating that initiation in the presence of the dinucleotide had occurred in the stretch of T residues upstream of +l. A similar upstream shift was observed with CpU (Fig. 2C, lune 6 ) , as expected from the sequence of the EIV promoter. This shift in mobility was specific to UpU and CpU, as neither ApC nor CpA visibly affected the migration of the EIV RNA (Fig. 2C, lanes 4 and  5 ) . A shift of one nucleotide (expected if CpA priming occurred) would probably not be detected with this gel system. However, the fact that addition of either ApC or CpA stimulated labeling of the EIV RNA in a short pulse indicates that these dinucleotides probably did prime transcription at the EIV promoter.
Initiation with Dinucleotides ut the Major Late Promoter-The MLP is more efficient in a whole cell extract than is the EIV promoter (44). In further contrast to EIV, initiation at the MLP occurs in vivo and in vitro at a single A residue (nucleotide 6039 of adenovirus type 2 ) . Analysis of initiation at the MLP is facilitated by the presence of a large RNase TI-resistant oligonucleotide containing the 5' terminus (45, 46). This oligonucleotide was analyzed directly using the following procedure. The MLP was transcribed using the threestage protocol described above.
The short pulse with low concentrations of the radioactive nucleotide allowed incorporation of label only near the 5' terminus. After completion of chains during the chase, RNA was hybridized to a singlestranded M13 recombinant spanning the MLP (M13 XH11, containing sequences from -261 to +197 of the MLP). Remaining single-stranded RNA was digested with ribonuclease and the RNA/DNA hybrid was selected on nitrocellulose filters. RNA was released from the hybrid and digested to completion with RNase T1. The labeled oligonucleotide products yielded a simple pattern when resolved by electrophoresis in a 15% acrylamide, 8 M urea gel (Fig. 3A, lane I ) . Five major bands, designated A-E, were observed. Bands A , B , C, and D migrated in the size range of 1 to 6 bases relative to a size marker, while band E migrated in the range of 9-13 bases. Tentative assignment of these oligonucleotides is shown in Fig. 3B. Since the label was [cY-~'P]GTP and each oligomer contained only one G residue (at the 3' end), the intensity of each band was proportional to the amount of the corresponding oligomer synthesized during the pulse. The gradient of band intensities ran E-D-B- (A,C). This agreed perfectly with the assignments based on size. Long exposure revealed a faint T1 product running at 16 nucleotides. This longer oligomer corresponded to the 16-mer ending at position +62. Its low intensity suggests that very few polymerase molecules reached that position during the pulse.
In contrast to the results obtained with the EIV promoter, varying nucleoside triphosphate concentrations had no effect on the site of initiation of the MLP. Concentrations of CTP or UTP as high as 1 mM did not result in initiation at C and T residues adjacent to +1 (data not shown).
Examination of the MLP sequence (Fig. 3B) suggested the dinucleotides ApC and CpA as good candidates for priming RNA synthesis. Addition of either dinucleotide to the pulse resulted in a shift of the putative 5"terminal oligonucleotide (Fig. 3A, lanes 2 and 3). The internal oligomers (A-D) migrated identically in each reaction. As predicted from the sequence, the CpA band (El2) migrated more slowly than the ApC band (Ell). The ApC band, in turn, migrated more slowly than the control band E, which presumably resulted from initiation with ATP at position +l. A possible reason for the faster migration of band E was the presence of phosphates at the 5' end of this oligomer. The effect of phosphatase treatment on the mobility of these oligonucleotides was therefore investigated.
RNAs primed without dinucleotides (control), or with either ApC or CpA, were treated with phosphatase before complete RNase T1 digestion. In this manner, only 5'-terminal phosphates should be susceptible to hydrolysis. As expected, phosphatase digestion did not affect the migration of internal T1 oligomers (Fig. 3C, bands A-D). The ApC band (E1,) and CpA band ( E l z ) were likewise not affected (lanes 6 and l o ) , whereas band E synthesized in the control reaction was affected by phosphatase (lane 14); after treatment, it comigrated with band Ell. Thus, 5"terminal phosphates were present on ATP-primed but not on ApC-and CpA-primed RNA.
The phosphatase sensitivity demonstrated that the control RNA was not capped. In contrast, previous results have shown that RNA synthesized in the whole cell extract is efficiently capped (2). The latter results were confirmed using RNA prepared in a whole cell extract by the three-stage protocol. This RNA was analyzed by M13 selection followed by complete RNase T1 digestion. Internal oligomer bunds A-D were identical with those of the reconstituted system (Fig. 3C, lunes  1 and 13). However, the largest whole cell extract T1 oligomer (Ec) migrated more slowly than the corresponding band E from the reconstituted system. The mobility of E, was not affected by phosphatase treatment (lane 2). Chemical decapping of the whole cell extract RNA prior to T1 cleavage altered the mobility of band E, such that it co-migrated with band E (lune 3). Decapped E, was also rendered phosphatase sensitive; the combined decap + phosphatase treatment altered the mobility of band E, such that it co-migrated with ApC band E l , (compare lanes 4 and 5 ) . Decapping had no effect on RNAs made in the reconstituted system (lanes 7, 11, and 16).
The following structures, with initiating nucleotides underlined, were postulated for the 5'-terminal oligonucleotides.
In the whole cell extract:

E, (ATP-initiated): GpppACUCUCUUCCGp
In the reconstituted system: These T1 oligonucleotides were further examined by digestion with RNase U2, which cleaves after A and G residues. As expected, RNase U2 digestion converted each of the four oligomers, E,, E , Ell, and El2 to a single 10-mer (data not shown). The sensitivity of the capped oligomer (E,) to RNase U2 indicated that none of the capped A was 2'-0-methylated. The extent of methylation at the 7-position of guanine in the cap was not ascertained.
Definitive proof of dinucleotide incorporation was obtained by the following label transfer experiments. RNA was made with control nucleotides, or with added ApC or CpA in the RNAs were hybridized to single-stranded M13 X H l l DNA, containing coding sequences between -261 and +197 of the MLP. The hybrids were RNase-treated and filter-selected as in Fig. 2 A ) . C, sensitivity of late promoter RNAs to decapping and phosphatase. Long runoff RNAs synthesized with the three-stage protocol as in A were hyhridized to M13 XH11, truncated with RNase T1, and filter purified. Aliquots were removed and treated with decapping and/or phosphatase reagents as under "Materials and Methods." Products in lanes 1-4 were made in a whole cell extract; those in lanes 5-16 were made in the reconstituted system. The standard pulse nucleotides received either no additions (lanes 1-4 and 13-16), ApC (lanes 5 4 , or CpA (lanes [9][10][11][12]. Samples were treated with phosphatase (lanes 2, 6, 10, and 14), decapped (lanes 3 , 7, 11, and 15), or decapped followed by phosphatase treatment (lanes 4, 8, 12, and 16). After these treatments, the RNAs were digested to completion with RNase T1 and electrophoresed. Hands A-I) refer to oligonucleotides whose sequence is in H. Rands E, E,,, E,,, and E, refer to 5'-terminal oligonucleotides synthesized under the different pulse conditions. pulse. Each reaction contained a single labeled nucleoside triphosphate ( [w:"P]GTP, CTP, or UTP). The 5"terminal oligomers were gel purified and further digested with RNase A (pyrimidine specific) or T2 (nucleotide nonspecific). The digestion products were analyzed by thin layer chromatography. Table I shows the products expected and observed in each case.
The [w:"P]CTP reactions provided the critical test. This nucleotide should transfer label to A residues only in the control and CpA reactions. In the ApC reaction, the 3'phosphate of the A residue should be provided by the dinucleotide and thus should not be labeled. These were the observed results. As expected, label was also transferred from [tu-'"P] CTP to Up, Cp, and Gp. [w:'"P]GTP and [(u-''?P]UTP transferred label to the anticipated nucleotides (Table I).
It has been suggested that the requirement for an RNA Polymerase I factor can be obviated by the use of a dinucleotide primer (29). This was tested in the RNA polymerase I1 transcription system with the adenovirus major late promoter. The dinucleotide CpA did not substitute for any of the HeLa cell fractions [AB], [CB], or [DR] in stimulating accurate transcription mediated by RNA polymerase I1 (data not shown).
A Region Accessible for Dinucleotide Priming-The extent of sequences available for priming at the MLP was tested using all 16 dinucleotides. In parallel reactions, each with a different dinucleotide added to the pulse, a variety of patterns was generated. All 16 reactions yielded the same pattern of internal T1 oligomers (Fig. 4, bands A-D). The dinucleotides (+A, ApU, ApG, GpU, GpG, CpG, UpG, and UpU ( Fig. 4

TABLE I Secondary analysis of 5"terminal oligonucleotides
Late promoter runoff RNAs were synthesized using the three-stage protocol. Each sample was scaled to five standard reaction volumes. The preincubation contained I'stI-cleaved pFLBH at 20 pg/ml. Pulses contained 100 pCi of [n-'"P]CTP, UTP, or GTP (1 p~) plus the other three unlabeled nucleoside triphosphates at 30 p~. With each radioactive nucleotide, a reaction was done with these standard pulse conditions, or with the standard pulse plus ApC or CpA. From each reaction, the 5"terminal RNase T1 oligomer was gel-purified following MI3 selection, and was further digested with RNase T2 or A. Products of CTP labeling were separated in two dimensions as described under "Materials and Methods;" products of GTP and UTP labeling were separated only in the first dimension. "Loss of 5"terminal phosphates in this sample apparently resulted from a phosphatase contaminant. gomer (lanes 1-8). In each case, some of' the ATP-initiated RNA (band E ) was observed. As previously shown, ApC and CpA gave an 11-mer and a 12-mer, respectively, corresponding to initiation at positions +1 and -1. CpC, CpU, and UpC each yielded novel bands which were interpretable as resulting from dinucleotide incorporation according to the DNA sequence: bands of length 15 (CpC), 14 (CpU), and 13 (UpC) were observed. Also in accord with the sequence, CpU gave bands at 10 and 8 nucleotides, and UpC gave a 9-mer. These oligomers ranging from 8 to 15 bases long corresponded to initiations at each position from -4 (CpC) to +4 (CpU). No initiations further upstream or downstream were detected.

RNA Polymerase I I Dinucleotide Priming
Several bands could not be explained by strict complementarity between dinucleotide and DNA. These included 11mers generated with CpC (Fig. 4, lane l ) , UpC (lane 3), and UpA (lane 8), 12-mers generated with ApA (lane 7) and UpA (lane 8), and a 10-mer generated with GpC (lane 6). One possible explanation for these bands was that the dinucleotides were incorporated using complementarity limited to the second residue. For example, CpC and UpC might generate 11-mers by incorporation in the same position as ApC. Further digestions with RNase U2 supported this "misincorporation" hypothesis: the 11-mers made with CpC and UpC were U2-resistant (Fig. 5, lanes 5-8). As expected, U2 treatment converted the ApC-initiated 11-mer to a 10-mer by cleavage at the A residue (lanes 1 and 2). The T1 resistant GpC 10-mer was similarly explained by incorporation of GpC in the same position as ApC. In this case, however, the complete T1 digestion released the 5"terminal Gp leaving a 10-mer. This 10-mer was RNase U2 resistant as expected. The aberrant UpA and ApA oligomers were all converted to were synthesized in the three-stage protocol with 2 mM ApC, GpC, CpC, or UpC in the pulse. After hybridization to M13 XH11, digestion of hybrids with RNase, and filter selection, RNA was eluted and digested to completion with RNase T1 (lanes 2 (ApC), 4 (GpC), 6 (CpC), and 8 (UpC)). Aliquots from each reaction were removed and further digested with RNase U2 (lanes I (ApC), 3 (GpC), 5 (CpC), and 7 (UpC)). RNase U2 digestion was complete as monitored by the digestion of all 5-mers (which have the sequence CAUCG) to 3-mers (UCG). U2 also cleaved the CpC 15-mer. the UpC 13-mer, the ApC 11-mer, and all ATP 11mers. The UpC 9-mer should have no A residue and was U2-resistant. The CpC and UpC 11-mers were likewise U2-resistant, proving that these 11-mers did not contain a purine at position +1. A diffuse band migrating near 13 nucleotides was observed in all U2 digestions. lo-mers by RNase U2 treatment (data not shown). This suggests that UpA and ApA generated 12-mers by incorporation in the same position as CpA. The UpA 11-mer is difficult to explain but could be due to contamination in this dinucleotide.
Dinucleotide Titrations-Addition of dinucleotides in no case stimulated overall transcription from the major late promoter. To determine the efficiency of dinucleotide utilization relative to standard pulse triphosphates, ApC and CpU were individually titrated in the pulse phase. Complete T1 digestion products were resolved on gels, and the label in various 5"terminal oligomers was measured by densitometry. As the concentration of each dinucleotide was increased, the amount of corresponding dinucleotide-primed transcripts in- GTP), 50% of the total initiations were dinucleotide-primed at about 60 PM ApC (data not shown). The CpU initiations a t -3 and +2 increased concomitantly with CpU concentration; a t about 200 pM CpU, the sum of initiations at these two sites was 50% of the total (data not shown).
Rate of Initiation-The ability to quantitatively shift 5' termini using high concentrations of dinucleotide in the pulse suggested an experiment to analyze the actual rate of chain initiation. This "dinucleotide challenge" experiment took advantage of the different mobilities of CpA-and ATP-initiated RNAs. Transcriptional factors and polymerase were preincubated with template DNA. Pulse nucleotides were added for 4 min, followed by the usual chase. The dinucleotide CpA was added in excess, either with the pulse nucleotides, or at some time during the pulse, or with the chase. The CpA and ATP initiations were resolved by complete T1 digestion of After preincubation of transcriptional proteins with a late promoter template, 4-min pulses were conducted. ATP, CTP, and UTP were present during the pulse, each at 30 p M (lanes [1][2][3][4] or 300 pM (lanes 5-8) with 20 pCi of [w:"P]C,TP at 1 p~. Pulses were followed by 3-min chases with 1 mM unlabeled nucleotides. The dinucleotide CpA was added together with pulse nucleotides (lanes I and 5 ) , after 1 min of pulse (lanes 2 and 6 ) , after 3 min of pulse (lanes 3 and 7), or with chase nucleotides (lnnrs 4 and 8). RNAs were analyzed as in Fig. 3A. The different 5'terminal oligomers are indicated hy arrows.
At 2 mM, CpA competed effectively for initiation when added at the beginning of the pulse (lane 1 ) . However, when CpA was added after 1 min of pulse, 90% of observed chains had pppA at their 5' terminus; thus, most of the observed chains had already initiated with ATP (lane 2). By 3 min of pulse, almost all chains were ATP-initiated (lane 3 ) . In a control reaction, the addition of CpA to the chase had no effect on the 5' oligomer (lane 4). Similar results were obtained when the ATP, CTP, and UTP concentrations were 300 P M in the pulse, 10 times higher than previously (lanes [5][6][7][8]. In a separate experiment, it was shown that the incorporation of label into ATP-initiated 11-mers reached a plateau after 2-4 min of the pulse (data not shown). This independently confirmed that initiation occurred synchronously during the standard pulse.
The preincubation-pulse-chase protocol thus functions as a one-round assay for activated complexes. Using this assay, the number of templates and RNA polymerase molecules productively involved in transcription was readily determined by quantitating the amount of radioactivity in 5"terminal RNase T1-resistant oligonucleotides. 0.1% of templates were transcribed, and 0.1% of the RNA polymerase I1 molecules functioned in the promoter-specific reaction. The low fraction of participating polymerase molecules did not reflect a poor preparation, as the specific activity of the enzyme was close to the maximum reported values. Kadesch and Chamberlin (38) have directly measured the fraction of active molecules in a similar preparation of RNA polymerase 11, using a nonspecific phage DNA template. From 15-25% of the enzyme molecules were active in this assay. The fraction of templates used in the specific runoff assay accorded with previous estimates for the efficiency of RNA polymerase I1 systems (1,2).

These earlier measurements
left open the possibility that some templates were transcribed in multiple rounds; the oneround assay does not suffer from this ambiguity. DISCUSSION We have shown that RNA polymerase I1 can utilize dinucleoside monophosphates to initiate faithful transcription at the early region IV and major late promoters of adenovirus. This polymerase therefore shares the ability of E. coli RNA polymerase and eukaryotic RNA polymerases I and 111 to incorporate dinucleotides by sequence complementarity at in vivo initiation sites.
The major late and EIV promoters seemed ideal for the study of transcription initiation, as these two promoters represent extremes: initiation in vivo occurs at a unique position at the major late promoter but is distributed over several adjacent nucleotides of EIV (41, 46). Initiation in vitro at the major late promoter was examined in detail by testing the effects of all 16 dinucleotides on the positions of 5' termini of transcripts. Dinucleotides complementary to sequences from -4 (CpC) to +4 (CpU) primed RNA synthesis. In contrast, initiation a t sites other than +1 was never observed by varying nucleoside triphosphate concentrations (data not shown). This suggests that initiation by dinucleotides and by nucleoside triphosphates may proceed by different mechanisms.
The homogeneity of initiation at the major late promoter could be due to the occurrence of an A residue at a mechanistically preferred position (see Baker and Ziff (41) for a similar discussion). In vivo, RNA polymerase I1 initiates most frequently with purines, and with ATP preferentially to GTP.
This could reflect preferential binding of ATP to the initiation site of the polymerase. The experiments presented here suggest that RNA polymerase I1 also has a strong preference for site +1 of the major late promoter, independent of the initiating nucleotide. The efficiency of dinucleotide priming was greatest a t positions near +1, as shown by the ability of different dinucleotides to compete with ATP for initiation. Moreover, the novel phenomenon of misincorporation, mediated by sequence complementarity between the second residue of a dinucleotide and the template, was observed only at positions +1 and +2. Dinucleotides might be envisioned to prime the elongation reaction, thereby bypassing initiation events. This would obviate ATP preference, and identify positions on the template available to the elongation site of the polymerase. A Region Accessible for Dinucleotide Priming-The major late promoter results show that a contiguous region of at least 8-9 bases is accessible for dinucleotide priming, centered around the in vivo cap site (see Fig. 7). This corresponds to a length of about 30 A of DNA, or almost a full turn of the helix. I t is difficult to picture how a static RNA polymerase aligned at a single base could initiate transcripts over this distance. Such a static complex would have to be capable of elongating RNA primers from eight different sites on the enzyme. An alternative and more likely possibility is that the polymerase drifts or oscillates over 8-9 bases around a preferred site of initiation. In this case, it would be possible to align a single site, for instance the elongation site of the enzyme, with any of eight different positions on the template. This plasticity of initiation is apparently distinctive of RNA polymerase I1 as compared to E. coli polymerase and eukaryotic RNA polymerases I and I11 (26,29,30,47).' A requirement for hydrolysis of the 8-7 bond of ATP has been suggested for initiation of transcription by RNA polymerase I1 (48). This cofactor dependence could be related to oscillation of t - Rate of Initiation by the Activated Complex-In the accompanying paper, we describe the formation of an "activated" complex by preincubation of template DNA, RNA polymerase 11, and factors. This complex was defined by its ability to rapidly begin transcription when presented with nucleoside triphosphates. The activated complex might therefore initiate directly, or might undergo one or more transitions to form the structure immediately preceding initiation. The time required for activated complexes to initiate RNA chains with ATP at the major late promoter was measured using a dinucleotide challenge experiment. Ninety per cent of the observed chains initiated within the 1st min of the pulse. It is worth noting that accumulation of activated complexes occurs over 1-h time interval, but that these complexes initiate quite rapidly. The rate-limiting steps in transcription of the major late promoter by the reconstituted system can thus occur in the absence of nucleotides.

C G T C C T C A C T C T C T T C C G
Initiation with ATP was sufficiently fast that stimulation of label incorporation by dinucleotides was not observed in a 4-min pulse. However, stimulation of incorporation by dinucleotides was observed at the EIV promoter. Such stimulation of incorporation was also observed when high concentrations of the initiating nucleoside triphosphate were used. This suggests that initiation at EIV is slower than at the major late promoter.
The Role of Capping-In the experiments presented here, a reconstituted transcription system was used. Dinucleoside monophosphates also prime transcription by polymerase I1 in a whole cell extract (data not shown). Late promoter RNAs primed with dinucleotides had no obvious modifications of the 5"hydroxyl terminus, either in the reconstituted system (Fig. 3C) or in the whole cell extract (data not shown). Thus, the presence of a cap is not required for and does not enhance transcription by RNA polymerase 11. Moreover, RNA initiated with ATP in the reconstituted system has unprotected phosphates at the 5' terminus. We have already noted that the bulk of guanyltransferase activity (assayed as described in Ref. 49) is recovered in a fraction not included in the reconstituted system (12). It thus seems highly unlikely that the ATP-initiated RNAs in the reconstituted system were ever capped, although transcripts were efficiently capped in the whole cell extract. Recent experiments with a nuclear wash extract have shown that initiation and some elongation precede capping (50). Together these results indicate that capping plays no direct role in the transcription reaction.