Template Recognition and Ribonucleotide Specificity of the DNA Primase of Bacteriophage T7

The 63-kDa gene 4 DNA primase of phage T7 catalyzes the synthesis of oligoribonucleotides on single-stranded DNA templates. At the sequence, 5 (cid:42) -GTC-3 (cid:42) , the primase synthesizes the dinucleotide pppAC; the cytidine resi- due of the recognition sequence is cryptic. Only tetrari-bonucleotides function as primers, but the specificity for the third and fourth position is not as stringent with a preference of CMP > AMP >> UMP > GMP. The pre- dominant recognition sites on M13 DNA are 5 (cid:42) -(G/T)G-GTC-3 (cid:42) and 5 (cid:42) -GTGTC-3 (cid:42) . Synthesis is usually limited to tetranucleotides, but T7 primase can synthesize longer oligoribonucleotides on templates containing long stretches of guanosine residues 5 (cid:42) to the recognition sequence. The specificity beyond the first two positions of the primer increases as the length of the template on the 3 (cid:42) -side of 5 (cid:42) -GTC-3 (cid:42) increases. On an oligonucleotide having 20 3 (cid:42) -flanking cytidine residues GMP is incorporated at the third position; incorporation is reduced 4-fold when the flanking sequence reaches 65 residues, and little is incorporated on M13 templates. The pres- ence of the 56-kDa gene 4 helicase decreases the incorporation of GMP on long templates. We propose that pausing is required for the incorporation of less pre-ferred nucleotides and that pausing is decreased by the ability of the primase to translocate Oligoribonucleotide synthesis catalyzed by primase was determined radioactively labeled oligoribonucleotide standard reaction contained 40 m M Tris-Cl (pH m M MgCl 2 , 10 m M dithiothreitol, 50 (cid:109) g/ml bovine serum albumin, 50 m M potassium glu-tamate, m M dTTP, m M rNTPs, n M M13 ssDNA or M oligonucleotides, primase unless otherwise indicated. incubation of m M separated electrophoresis or

Template Recognition and Ribonucleotide Specificity of the DNA Primase of Bacteriophage T7* (Received for publication, October 11, 1996, and in revised form, December 10, 1996)

From the Department of Biological Chemistry and Molecular Pharmacology, Harvard University Medical School, Boston, Massachusetts 02115
The 63-kDa gene 4 DNA primase of phage T7 catalyzes the synthesis of oligoribonucleotides on single-stranded DNA templates. At the sequence, 5-GTC-3, the primase synthesizes the dinucleotide pppAC; the cytidine residue of the recognition sequence is cryptic. Only tetraribonucleotides function as primers, but the specificity for the third and fourth position is not as stringent with a preference of CMP > AMP > > UMP > GMP. The predominant recognition sites on M13 DNA are 5-(G/T)G-GTC-3 and 5-GTGTC-3. Synthesis is usually limited to tetranucleotides, but T7 primase can synthesize longer oligoribonucleotides on templates containing long stretches of guanosine residues 5 to the recognition sequence. The specificity beyond the first two positions of the primer increases as the length of the template on the 3-side of 5-GTC-3 increases. On an oligonucleotide having 20 3-flanking cytidine residues GMP is incorporated at the third position; incorporation is reduced 4-fold when the flanking sequence reaches 65 residues, and little is incorporated on M13 templates. The presence of the 56-kDa gene 4 helicase decreases the incorporation of GMP on long templates. We propose that pausing is required for the incorporation of less preferred nucleotides and that pausing is decreased by the ability of the primase to translocate 5 to 3 on templates having long 3-flanking sequences.
Gene 4 of bacteriophage T7 encodes two co-linear proteins, a 56-and a 63-kDa protein (1). In phage-infected cells, the two proteins are expressed in approximately equal amounts from two in-frame translation initiation sites 189 bases apart on the gene 4 transcript. Thus, the 56-kDa gene 4 protein lacks the 63 N-terminal amino acid residues of the 63-kDa gene 4 protein.
The 56-kDa protein (helicase) binds to single-stranded DNA (ssDNA) 1 in the presence of a nucleoside triphosphate (dNTP) and then, in a reaction coupled to the hydrolysis of the dNTP, translocates 5Ј to 3Ј along the ssDNA (2,3). Upon encountering a duplex region of DNA, its continued translocation unwinds the duplex (2).
The 63-amino acid N-terminal domain of the 63-kDa gene 4 protein (primase) contains a Cys 4 zinc-binding motif (4,5), a domain that enables the primase to synthesize oligoribonucle-otides in a template mediated reaction. Although the zincbinding motif is essential for primer synthesis (5), it alone cannot account for the ability of the protein to recognize specific sequences on ssDNA (6). Gene 4 protein is essential for phage growth and physically interacts with both the gene 5 DNA polymerase and the gene 2.5 ssDNA binding protein to mediate leading and lagging strand synthesis at the T7 replication fork (7)(8)(9)(10)(11)(12). The 63-kDa gene 4 protein has helicase activity as well as primase activity, and it alone can support the growth of T7 bacteriophage in vivo (13,14).
DNA primases are responsible for the synthesis of oligoribonucleotides that serve as primers for DNA polymerases on the lagging strand of the replication fork (15). The gene 4 primase of phage T7 catalyzes the template-directed synthesis of di-, tri-, tetra-, and pentaribonucleotides on ssDNA (16 -19). However, both in vivo and in vitro only tetra-and pentaribonucleotides can serve as functional primers for T7 DNA polymerase (17,20). Studies employing synthetic oligonucleotide templates have demonstrated that the trinucleotide sequence 5Ј-GTC-3Ј is both necessary and sufficient for the protein-DNA interaction required to initiate oligoribonucleotide synthesis (19). The 3Ј-cytidine is required for recognition but is not copied into the primer. Although limited synthesis of pppAA dinucleotide has been observed to occur at the trinucleotide sequence 5Ј-TTC-3Ј under certain conditions, the pppAA dinucleotide is not extended further (6,21). Analysis of the oligoribonucleotide primers attached to the 5Ј-end of Okazaki fragments both in vivo and in vitro has shown that the T7 primase synthesizes predominantly pppAC(C/A)(C/A) tetraribonucleotides; UMP and GMP are not found in the primers (17,18,20).
Gene 61 primase of bacteriophage T4 and DnaG primase of Escherichia coli also contain a single Cys 4 type zinc motif at their N termini (5,22). Both the T4 and E. coli primases recognize trinucleotide sequences, 5Ј-G(T/C)T-3Ј and 5Ј-CTG-3Ј, respectively (23)(24)(25)(26)(27), and the 3Ј-nucleotide of the sequence is not copied into the primer. In contrast to these similarities among the three primases, the T4 and E. coli primases differ from the T7 primase in that the template sequence 5Ј to the trinucleotide recognition sequence is less stringent both in regard to sequence and to the length that is copied (23,25,26,28). As a result the oligoribonucleotides found at the 5Ј-termini of Okazaki fragments synthesized by the T4 and E. coli primases are ppp(A/G)C(N) 3 or pppAG(N) 9 -11 , respectively (28 -30). In this report we examine in detail the specificity of nucleotide selection by the T7 primase.
The T4 and E. coli primase systems also differ from that of T7 in one other striking manner. Whereas the T7 63-kDa gene 4 protein catalyzes both helicase and primase activities, the T4 and E. coli helicase and primase activities reside within separate polypeptides (28,(31)(32)(33)(34). As a result the T7 primase is faced with the dilemma that upon reaching a primase recognition site it must either pause in its 5Ј to 3Ј translocation and synthesize a primer in the opposite direction or continue its role as a helicase. In this report we present evidence that the inherent translocation activity of the T7 primase plays a role in determining which potential sites in the template are used for oligoribonucleotide synthesis.

EXPERIMENTAL PROCEDURES
Enzymes and Biochemicals-The T7 63-kDa gene 4 protein used in these studies has Met at position 64 replaced by Gly but this protein has all of the catalytic properties of the wild-type protein (35,36). Wild-type T7 56-kDa gene 4 protein was purified by B. Beauchamp (Harvard Medical School) as described previously (19). T7 gene 5 DNA polymerase was purified as described previously (11) and kindly provided by S. Tabor (Harvard Medical School). T7 gene 2.5 ssDNA binding protein was purified by D. Kong (Harvard Medical School) as described previously (37). T4 polynucleotide kinase, T4 DNA ligase, other enzymes, and nucleotides were purchased from Amersham Corp.
DNA Templates-M13mp18 ssDNA was purified as described elsewhere (38). Oligonucleotide templates for assay of oligoribonucleotide synthesis by T7 DNA primase were chemically synthesized by C. Dahl. (Harvard Medical School). The nucleotide sequences of the DNA templates are presented in Table I. The six oligonucleotides labeled HR3Ј-5 to HR3Ј-65 were prepared by annealing 5Ј-AAAAAAAAAAGACGC-3Ј, 5Ј-CCCCCCCCCCCCCCCGCGTCTTTTT-3Ј, and poly(dT) in a ratio of 1:1:1. The poly(dT) used in separate reactions varied from 5 to 65 nucleotides in length. After ligation the products were separated on 10% urea-gel, and the DNA fragments corresponding to 25,30,35,45,65, and 85 nucleotides (nt) in length were isolated.
DNA Synthesis and Primer Synthesis at a Minimal Replication Fork-Formation of a minimal replication fork was carried out as follows. One nanomole of RF-1 oligonucleotide (71 nt) was phosphorylated using T4 polynucleotide kinase and mixed with 1 nmol of RF-3 (40 nt) in 100 mM Tris-Cl (pH 7.5). The mixture was heated at 95°C for 5 min and then slowly cooled to anneal, followed by ligation of the oligonucleotides by incubation overnight at 16°C using T4 ligase. The 71-nt closed circular DNA was isolated on a 10% urea-gel and then annealed with RF-2 oligonucleotide (75 nt) at a molar ratio of 1:3 at 65°C followed by cooling overnight. Reactions using the minimal replication fork (20 l) contained 40 mM Tris-Cl (pH 7.5), 10 mM MgCl 2 , 10 mM dithiothreitol, 50 g/ml bovine serum albumin, 50 mM potassium glutamate, 0.7 mM dTTP, 0.2 mM d(A,G,C)TPs, 0.1 mM [␣-32 P]CTP, 0.1 mM ATP or (A,G,U)TPs, 50 nM gene 5 DNA polymerase, 500 nM gene 2.5 protein, and 5 nM (hexamer) primase. The reactions were started by adding the replication fork to 25 nM. After incubation at 37°C for 5 min, the reactions were stopped by the addition of 5 l of 0.2 M EDTA. A portion of the reaction mixture (5 l) was mixed with 15 l of sequencing dye and applied directly to a 25% urea-gel for measurement of oligoribonucleotide synthesis. The remainder of the reaction (20 l) was deproteinized, and the DNA was precipitated with ethanol, digested with BamHI, and loaded onto a 6% urea-gel. After electrophoresis the DNA fragments containing ribonucleotide primer were visualized by autoradiography.

RESULTS
The primers used by T7 DNA polymerase both in vivo and in vitro are exclusively tetra-or pentaribonucleotides (18,20). However, in vitro both di-and triribonucleotides are synthesized by the 63-kDa gene 4 primase (16,19). We have previously shown that the minimal recognition sequence for the T7 primase is 5Ј-GTC-3Ј and at this site the gene 4 protein catalyzes the synthesis of the dinucleotide pppAC unless the proper nucleotides are present in the template for extension to tri-and tetranucleotides (19). Analysis of the sequence of the primers attached to the 5Ј-termini of Okazaki fragments synthesized in vivo and in vitro (18,20) as well as the characterization of the nucleotides incorporated by the primase on synthetic oligonucleotide templates (6,19,21) revealed that pppAC is the only dinucleotide that can be extended to tetra-or pentanucleotide primers. The dinucleotides pppAA and very few pppGN are synthesized on M13 ssDNA but the pppAA is not extended to the trinucleotide (21). Although the nucleotides found at the first and second position of the functional primer are invariant, the specificity of subsequent nucleotides in the primer as well as the factors controlling the length of the primers have not been examined rigorously. Previous studies have demonstrated that although there is a strong preference for CMP and AMP to be incorporated after the first two nucleotides, the stringency for nucleotide incorporation is less at these positions (6,21).
In order to determine the frequency with which each nucleotide is incorporated into primer, we have carried out oligonucleotide synthesis in reactions containing either M13 ssDNA or synthetic oligonucleotide templates and one of each of the four [␣-32 P]NTPs in the presence of the other three unlabeled rNTPs. Using the 7249-nt M13mp18 ssDNA and much shorter synthetic oligonucleotides, we demonstrated that nucleotide specificity is influenced by the length of the template. In addition, short oligonucleotide template enabled the examination of nucleotide incorporation at specific sequences. Fig. 1, A and B, when M13 ssDNA was used as a template, incorporation of GMP and UMP by gene 4 primase after the second position in the oligoribonucleotide was only 6.7 and 13.8%, respectively, compared with that of CMP. AMP was incorporated as well as CMP. In agreement with earlier studies pppAC and pppAA dinucleotides were synthesized, but only pppAC was extended (Fig. 1A). No significant amount of dinucleotide containing GMP or UMP was observed. As described above GTP can be used as the first nucleotide of the primer, albeit at very low frequency. Consequently, a portion of radioactivity observed with [␣-32 P]GTP is most likely due to incorporation of GTP as the first nucleotide. In addition to pppACC and pppACA trinucleotides in the middle two lanes, pppACU trinucleotide was also observed in the last three lanes. However, the pppACU trinucleotides were less efficiently extended to tetranucleotide primers. These results demonstrate that the gene 4 primase prefers CMP and AMP for incorporation after the second position of the primer. Although UMP can be incorporated, the resulting trinucleotides are less likely to be further extended.

Ribonucleotide Specificity at Third and Subsequent Positions of Primers Synthesized on M13 ssDNA Templates-As shown in
Although the length of the majority of primer synthesized by the gene 4 protein are limited to five nucleotides, further extension can occur as evidenced by the presence of a small amount of hexamer and heptamer seen in Fig. 1A. Since the gene 4 primase incorporates GMP and UMP poorly, this property may play a role in limiting the length of the oligoribonucleotide synthesized by the primase starting from a basic recognition site. In fact, inspection of the M13 ssDNA sequences reveals that the longest stretch of guanosine or thymidine residues following the recognition sequence 5Ј-GTC-3Ј is 5Ј-TGGGGGTC-3Ј, a site that yields the heptanucleotide pppAC-CCCCCA observed in Fig. 1A. In order to determine whether the T7 primase has an inherent property limiting the length of primers synthesized to four or five nucleotides or if the length of the primer is, at least in part, determined by the DNA sequence, we constructed a template containing a repeat of 19 guanosine residues on the 5Ј-side of the primase recognition site. On this template, T7 primase synthesizes long primers, up to 22 nt in length (Fig. 2). The presence of an additional nucleotide suggests that the primase adds a one nucleotide overhang upon reaching the 5Ј-end of the template strand. The amount of oligoribonucleotide synthesized increased linearly, up to 60 min, and the proportional amount of short and long primers was roughly constant over this period.
Ribonucleotide Specificity at Third and Subsequent Positions of Primers Synthesized on Synthetic Oligonucleotide Templates-As shown in the previous section, gene 4 primase incorporates predominantly CMP and AMP after the second position in primers synthesized on M13 ssDNA. In order to compare more precisely the ability of the T7 primase to incorporate each of the four nucleotides at positions beyond the initial dinucleotides we have examined primer synthesis on synthetic oligonucleotides of known sequence. The templates shown in Fig. 3 varies in length from 25 to 80 nt, and they all contain the basic primase recognition site 5Ј-GTC-3Ј with variation in the sequence occurring on the 3Ј-side of the recognition sequence (see Table I for complete sequences). Oligoribonucleotide synthesis reactions were carried out on the various templates shown in Fig. 3 in the presence of ATP, GTP, UTP, and [␣-32 P]CTP. To our surprise when these oligonucleotides were used as templates, GMP and UMP could be readily incorporated at the third and subsequent positions (Fig. 3A). The total amount of oligoribonucleotides having lengths greater than two residues is presented in Fig. 3B for each template. GMP (template 8) and UMP (template 9) can be incorporated into the third position of the primer when a cytidine or thymidine residue follow the basic recognition site in the template. Furthermore, these nucleotides can also be incorporated into the fourth position (templates 5 and 7, respectively). Incorporation of UMP is relatively low, but the amount incorporated is still comparable to that of CMP incorporated (compare template 1 and 9, Fig. 3, A and B). The relatively low synthesis of oligonucleotides on templates 1 and 2 is most likely due to length of the sequence 3Ј to the recognition sequence 5Ј-GTC-3Ј on these templates (see below). Interestingly, in the case of template 8, once two GMP residues have been incorporated into the third and fourth positions, the oligoribonucleotide is efficiently extended as evidenced by the accumulation of primers greater than four in length. Thus the presence of guanosine on the 3Ј-terminus of the primer does not inhibit its further extension by the primase. Two different primase recognition sites 5Ј-AGGTC-3Ј and 5Ј-ATGTC-3Ј are present in template 7, necessitating the incorporation of a uridine residue at the fourth position of the primer. Tetranucleotides are synthesized with equal efficiency on both.
We have examined the ability of the T7 primase to use templates containing adenosine or cytidine in the fourth position of the template sequence, 5Ј-G(A/C)GTC-3Ј, in the presence of a template containing the sequence, 5Ј-GGGTC-3Ј as a competitor. As shown in Fig. 4, all three templates alone were approximately equal in their ability to incorporate CMP, GMP, or UMP at the third position of the oligoribonucleotides. Interestingly, the oligoribonucleotides containing either GMP or UMP at the third position were far more efficiently extended to the 5Ј-end of the polyguanosine-containing template than those incorporating CMP at this position (see "Discussion"). When each of the templates containing adenosine or cytidine at the fourth position of the recognition sequence were mixed with an equal amount of template containing guanosine at this position, the amount of primer synthesis was reduced by half. Since the reactions were performed under conditions of template excess, a reduction in the incorporation of GMP and UMP by 2-fold indicates that the adenosine or cytidine containing templates are used as well as the guanosine containing template.
Fidelity of T7 DNA Primase Varies According to Position in Template-The three extensively studied prokaryotic DNA primases, E. coli DnaG protein, T4 gene 61 protein, and T7 gene 4 protein, all recognize a basic trinucleotide sequence in the template (23)(24)(25)(26)(27). Consequently, primers synthesized by all these primases have essentially invariant nucleotides in the first two positions. Furthermore, like the T7 primase described above, both the E. coli (25,26,30,39) and T4 (23,24) primases relax this stringency after the first two nucleotides are incorporated. These properties suggest that the mechanism by which the first two nucleotides are incorporated differ from that involved in subsequent incorporation. Furthermore, formation of the first phosphodiester bond may well require that the primases bind tightly to both the first and the second nucleotide, thus dictating a specific nucleotide at each binding site. However, as extension of the dinucleotide progresses, the primase then interacts with the growing oligoribonucleotide annealed to the template and the next incoming nucleotide. The kinetics and specificity for the two steps would be expected to differ significantly. In fact, Swart and Griep (39) have shown FIG. 1. NMP incorporation into oligoribonucleotides by gene 4 primase on M13 ssDNA. Oligonucleotide synthesis assay using an M13 ssDNA template was performed as described under "Experimental Procedures" in reaction mixtures containing 0.1 mM of one of four [␣-32 P]rNTPs and 0.1 mM of the other three rNTPs. A, the reaction products were resolved by electrophoresis on a 25% urea-gel, and the products were identified by autoradiography. The location of each oligoribonucleotide species is indicated. B, plot of the relative incorporation of each ␣-32 P-labeled ribonucleotide into oligonucleotides having a length greater than two residues. Incorporation of CMP is taken as 100%.
that synthesis of dinucleotides is a rate-limiting step for total oligoribonucleotide synthesis by the E. coli primase.
If the low stringency incorporation of DNA primases at positions after the second nucleotide is brought about by the difference in the stability of the primase/template/nucleotide complex, it is also possible that the fidelity of the enzyme at a given site in the primer varies with its relative distance from the basic recognition site 5Ј-GTC-3Ј. The high fidelity of T7 primase at the first and the second position of the primers is apparent from the fact that ATP and CMP are the only nucleotides found at the first and second positions of the primer, respectively, on M13 ssDNA, although there are rare exceptions as described in the Introduction. We have examined oligoribonucleotide synthesis on five templates in which each of the guanosines between the fourth and the eighth position of the recognition sequence 5Ј-GGGGGGTC-3Ј (indicated by boldface letters) is replaced by an adenosine. The products of oligoribonucleotide synthesis on each template were identified by gel electrophoresis (Fig. 5A). When the oligoribonucleotides were labeled with [␣-32 P]UTP in the presence of all four NTPs, UMP was incorporated at the position corresponding to the adenosine in all five templates (last five lanes in Fig. 5A). When the oligoribonucleotides were labeled with [␣- 32  following the 5Ј-GTC-3Ј recognition site. Further, when UTP and GTP were removed from the reaction, CMP was again not misincorporated at the third position (lane 6), but was readily incorporated after the third position (lanes 7-10). Thus at the third position of the primer there is no misincorporation of CMP opposite adenosine (lanes 1 and 6) or UMP opposite guanosine (lane 12), whereas at the fourth position CMP opposite adenosine (lanes 2 and 7) is misincorporated, and UMP opposite guanosine (lane 13) is not. After the fourth position misincorporation of CMP is frequent, whereas UMP is not.
In order to confirm that T7 gene 4 primase loses fidelity in proportion to its distance from the basic recognition site 5Ј-GTC-3Ј, we examined the frequency of misincorporation of CMP at a position nine residues from the recognition site. On the template 5Ј-G 10 NG 8 GTCN 10 -3Ј in which each of the four nucleotides are located at the N9 position, CMP was misincorporated opposite adenosine (Fig. 5B, lane 2), thymidine (lane 3), and cytidine (lane 4). Compared to the correct incorporation of CMP opposite to guanosine (lane 1), CMP was misincorporated opposite adenosine, thymidine, and cytidine (lanes 2-4) FIG. 2. Oligoribonucleotide synthesis by gene 4 primase using the template, 5-G 19 GTCN 10 -3. The oligonucleotide synthesis assay was performed as described under "Experimental Procedures" using the template 5Ј-G 19 GTCN 10 -3Ј. The reaction mixtures contained 100 nM template, 0.1 mM r(A,G,U)TPs, 0.1 mM [␣-32 P]rCTP and 60 nM (monomer) primase. 10-l aliquots were removed at the indicated times, and the reaction products were resolved on a 15% urea-gel and visualized by autoradiography. The lengths of the oligoribonucleotides are indicated.

FIG. 3. NTP incorporation into oligoribonucleotides by gene 4 primase on synthetic oligonucleotide templates.
A, oligonucleotide synthesis using 60 nM T7 gene 4 primase (monomer) and 100 nM of the indicated templates having the minimal trinucleotide primase recognition sequence was carried out as described under "Experimental Procedures." The reaction products were resolved on a 25% urea-gel and visualized by autoradiography. The nucleotide sequences for each template containing the basic recognition site 5Ј-CTG-3Ј are presented in Table I. The nucleotide incorporated at the 3Ј-termini of each oligoribonucleotide is indicated above the bands. The band indicated by an asterisk (lane 5) is an oligoribonucleotide containing a misincorporated CMP opposite cytosine. B, quantification of oligoribonucleotides greater than two in length synthesized on various templates. The quantity of oligonucleotides synthesized were normalized to the amounts of CMP per oligomer. approximately 50, 30, and 80%, respectively. In the absence of UTP and GTP, CMP was misincorporated more frequently opposite thymidine, and cytidine (last four lanes in Fig. 5B).

Both in the absence and presence of UTP and GTP, the misincorporation of CMP opposite thymidine and cytidine (lanes 3 and 4, lanes 7 and 8) resulted in less extension beyond the misincorporated nucleotide.
Effect of Helicase Activity on Template Usage-The results presented above (Figs. 1 and 3) show that the T7 primase relaxes its stringency on oligonucleotide templates as compared to an M13 template. For example, oligonucleotides containing the recognition sequence 5Ј-CCGTC-3Ј are efficiently used as template for tetranucleotide synthesis whereas this sequence is seldom used on M13 ssDNA. One difference between the two types of templates is the greater length of DNA surrounding the basic trinucleotide recognition site 5Ј-GTC-3Ј in M13 ssDNA. In the case of the 7249-nt M13mp18 ssDNA molecule, the gene 4 protein is likely to search for recognition sites by using its inherent helicase activity to track along the DNA 5Ј to 3Ј (35), whereas in the case of oligonucleotide templates, random diffusion to the sites may play a more important role. Consequently, at those sites where CTP or ATP would not normally be used, the rapid movement of the protein through these sites by translocation would not allow sufficient time for incorporation of a nucleotide such as GMP.
In order to examine the effect of flanking sequences on nucleotide usage, we designed six templates each having a different length of DNA on the 3Ј-side of the potential recognition site 5Ј-GCGTC-3Ј (Fig. 6A). Oligoribonucleotide synthesis was then measured in reactions containing the 63-kDa gene 4 protein that has both primase and helicase activity and in reactions containing equal amounts of the 56-and 63-kDa gene 4 protein. The 56-kDa gene 4 protein has only helicase activity. As shown in Fig. 6, B and C, the synthesis of oligoribonucleotides containing GMP increased as the length of the 3Ј-flanking sequence approached 15 nucleotides. Most likely this effect of length on primer synthesis reflects an optimal size for binding to the recognition site since there was a similar effect on dinucleotide synthesis. Flanking sequences greater than 15, however, resulted in a decrease in the synthesis of oligoribonucle-  1-GGGCT  5Ј-GGGTTCCAAG ACCTTAGGAT CCAGAGCCCA GCATTTGGAC TTATGGGGGG GGGGGGGGGG GTCTCCGAAG CTTGGGGGG  2-GGGCT  5Ј-GGGGGGGGGG GGGGGGGGGG TCATTCTTGG ACCT  3-TGGTC  5Ј-TGGCGATTCG CAGTTTATAC CGATTCAGGT ACGTTAGGTA TCCATTGGTC TCCTAGGCTT AACGCCACGG  4-GGGTC  5Ј-CCTTCTCAAC TCAGGGTCAC ATATGACA  5-CGGTC  5Ј-AACTGCCGCG GCGGTCAATT TACGGAGCA  6-TTGTC  5Ј-CGCATACGTA GCATGCGAAT TCTGCGCAGT TGTCACAAGG Table I. Oligoribonucleotide species are indicated by each panel. otides containing GMP, the effect being proportionate to length. In contrast, dinucleotide synthesis is not affected to the same extent. This effect is accentuated by the presence of additional helicase augmented by the presence of 56-kDa gene 4 protein.
Use of Primers Containing a Guanosine Residue at a Replication Fork-Tetranucleotides synthesized at the sites 5Ј-(G/ T)GGTC-3Ј and 5Ј-GTGTC-3Ј are known to serve as primers for lagging strand synthesis both in vivo (20) and in vitro (17,18). In as much as the T7 primase can also synthesize oligoribonucleotides containing a guanosine residue at recognition sites such as 5Ј-TCGTC-3Ј, it was of interest to determine if these sites are used during simultaneous synthesis of leading and lagging strands at a replication fork and, if so, whether they are used as primers by T7 DNA polymerase. To address this question we have used a preformed replication fork consisting of a single-stranded circular DNA molecule of 71 nucleotides to which a 71-nt oligonucleotide whose 3Ј-36-nt terminus is complementary to the circular DNA has been annealed. The resulting preformed replication fork depicted in Fig. 7A consists of a partially duplex circle with a 5Ј-ssDNA tail of 35 nt. This 5Ј-single stranded terminus is required for the loading of the gene 4 protein (40). The preformed replication fork also contains two gene 4 protein recognition sites, 5Ј-GGGTC-3Ј and 5Ј-TCGTC-3Ј that would give rise to primer synthesis on the linear, lagging strand. On such a preformed replication fork, leading strand synthesis can be achieved by T7 DNA polymerase, T7 gene 4 protein, and the four dNTPs (40,41). Upon addition of the four NTPs, primer synthesis by the 63-kDa gene 4 protein and their extension by T7 DNA polymerase should result in lagging strand synthesis. T7 gene 2.5 protein stimulates lagging strand synthesis (10).
As shown in Fig. 7B, under conditions that give rise to leading and lagging strand synthesis oligoribonucleotides are synthesized by the gene 4 primase as measured by the incorporation of [␣-32 P]CMP into di-, tri-, and tetranucleotides. However, the only products were pppAC, pppACC, and pppACCC. Since the pppACC was efficiently extended to pppACCC under these conditions, less pppACC accumulation was observed in Fig. 7B. No oligonucleotide-containing guanosine, such as pppACG or pppACGA was observed, suggesting that the 5Ј-TCGTC-3Ј site was not used for primer synthesis during simultaneous DNA synthesis.
However, in this experiment a significant portion of the oligoribonucleotides synthesized were used as primers by T7 DNA polymerase on the lagging strand. As a result, any primers extended would not be observed in Fig. 7B. In order to determine if any oligoribonucleotides containing GMP were used as primers, the product DNA-containing primers labeled with [␣-32 P]CMP were digested with BamHI. With BamHI digestion, Okazaki fragments synthesized from the 5Ј-GGGTC-3Ј or 5Ј-TCGTC-3Ј site resulted in a 60-or 49-nt fragment, respectively (Fig. 7A). As shown in Fig. 7C, in addition to the 60-nt fragment, a small percentage of the 49-nt fragment whose 5Ј-oligoribonucleotide contains GMP at the third position of the primer was found. Therefore, the 5Ј-TCGTC-3Ј site is used at a low frequency, and most of the primers arising from the 5Ј-TCGTC-3Ј site are extended by T7 DNA polymerase during synthesis of leading and lagging strands at the replication fork.

FIG. 5. Misincorporation of CMP by gene 4 primase.
Misincorporation of CMP was analyzed in the oligonucleotide synthesis assay as described under "Experimental Procedures." A, the reactions were carried out using the five templates, G 4 AG 0 GTC, G 3 AG 1 GTC, G 2 AG 2 GTC, G 1 AG 3 GTC, and G 0 AG 4 GTC (Table I), in the presence of either 0.1 mM rNTPs or (A,C)TPs, as indicated. The [␣-32 P]NTPs used are indicated at the top of each panel. B, oligonucleotide synthesis reactions were carried out using the templates G 10 GG 8 GTC, G 10 AG 8 GTC, G 10 TG 8 GTC, and G 10 CG 8 GTC (Table I)

DISCUSSION
The DNA primase of bacteriophage T7, the product of the viral gene 4, catalyzes the synthesis of oligoribonucleotides for use as primers by T7 DNA polymerase (16,17). Although the sequence of the oligoribonucleotides is dictated by that of the single-stranded template (18), neither the initiation of primer synthesis nor the length of the primer is random. In fact, the majority of the primers found at the 5Ј-termini of Okazaki fragments synthesized in vivo are tetra-or pentanucleotides of three predominant sequences. In this report we address some of the parameters that dictate the sites of primer synthesis.
In confirmation of earlier studies (19) we find that T7 gene 4 primase is stringent in its requirement for a trinucleotide recognition site, 5Ј-GTC-3Ј. On either single-stranded M13 DNA or synthetic oligonucleotides, all oligoribonucleotides synthesized by the T7 gene 4 primase contain pppAC at their 5Јtermini (16 -19); the 3Ј-terminal cytidine is cryptic in that it is not copied into the primer. Furthermore, the trinucleotide sequence 5Ј-GTC-3Ј is sufficient for recognition by the T7 primase, and templates containing this sequence support the synthesis of the dinucleotide pppAC (19). The dinucleotide synthesized at the trinucleotide recognition site, however, are not used by T7 DNA polymerase as primers but instead must first be extended to tetra-or pentanucleotides (17,18,20).
The precise mechanism by which the T7 primase and other prokaryotic primases recognize a trinucleotide sequence on ssDNA is at present unknown. A major determinant is undoubtedly the Cys 4 type zinc motif found in all DNA primases (5,22). In the case of phage T7 a single amino acid substitution of Ser for each of the four conserved Cys residues renders the primase inactive (5). The zinc motif is not the sole determinant for recognition, however, since chimeric DNA primases in which the zinc motif of the T7 primase has been replaced with those from the phage T4 and E. coli primases do not recognize the trinucleotide recognition site of any of the three primases, although they do still recognize a trinucleotide sequence (6).
Since the average length of Okazaki fragments (lagging strand DNA fragments synthesized from gene 4 primers) is 1000 -2000 nucleotides, it is clear that parameters other than the basic trinucleotide recognition site affect the location of primer synthesis (10, 41). 2 One important determinant is the identity of the nucleotides 5Ј to the trinucleotide recognition site. For example, both in vivo (20) and in vitro (18) studies show that the majority of primers are synthesized from three sites, 5Ј-GGGTC-3Ј, 5Ј-TGGTC-3Ј, and 5Ј-GTGTC-3Ј. In these studies a number of other minor sites were observed. These minor alternative sites always contained the basic trinucleotide recognition sequence at their 3Ј-termini, departing from the three predominant sites only with regard to the two nucleotides on the 5Ј-side of the trinucleotide recognition site. Furthermore, the percentage of oligoribonucleotides that actually function as primers for Okazaki fragment synthesis in vitro is quite low even in the presence of excess amounts of T7 DNA polymerase. 3 An interesting observation that arose during the determination of nucleotide specificity was the ability of the T7 primase to synthesize oligoribonucleotides efficiently on synthetic oligonucleotides containing the sequence 5Ј-NNGTC-3Ј, while on M13 DNA there is a clear preference for the recognition sites 5Ј-GGGTC-3Ј, 5Ј-TGGTC-3Ј, and 5Ј-GTGTC-3Ј. The rather loose specificity observed with the oligonucleotide templates is comparable to that observed with the phage T4 (23,24,44) and E. coli (25,26) primases on all templates. In several other aspects, however, these three primases are similar: (i) they contain a FIG. 6. Effect of template length and helicase activity on template usage. A, scheme of DNA templates used. All templates were constructed as described under "Experimental Procedures." The nucleotide sequence of each template is given in Table I. B, oligonucleotide synthesis assays on templates having 3Ј-flanking sequences of various lengths, as shown in A. The reactions contained either 60 nM (monomer) 63-kDa gene 4 protein alone or a 1:1 mixture of 63-and 56-kDa gene 4 proteins. Other conditions were as described under "Experimental Procedures." The length of the 3Ј-flanking sequences from the 3Ј-cryptic cytosine is shown above each lane and the oligoribonucleotide species are indicated on the left. C, amount of dinucleotides or oligoribonucleotides synthesized on each template plotted as a function of the length of the 3Ј-flanking sequences. The quantity of oligonucleotide synthesized has been normalized to the number of CMP residues per oligomer. Cys 4 type zinc motif (5,22); (ii) they require association with a helicase to achieve maximal activity (28,(32)(33)(34)(45)(46)(47); (iii) they recognize a basic trinucleotide sequence in which the 3Ј-nucleotide is cryptic (23)(24)(25)(26)(27); and (iv) they synthesize relatively short oligoribonucleotides that serve as primers for their respective DNA polymerases (23, 28 -31, 44, 48, 49). On the other hand, they do differ in one fundamental manner. Although all three primases require a helicase to translocate 5Ј to 3Ј in order to reach primase recognition sites, only the T7 63-kDa gene 4 primase has an inherent helicase activity (1,13). Both the T4 gene 61 helicase and the E. coli DnaB helicase are encoded by separate genes, and hence their respective primases must subsequently associate with them to benefit from their translocation activities. As a consequence the gene 4 primase catalyzes the synthesis of oligoribonucleotides at a considerably faster rate (0.8 oligoribonucleotide/s/hexamer of protein on M13 ssDNA) than do either the T4 gene 61 protein (0.83 ϫ 10 Ϫ2 oligoribonucleotide/s/monomer) (24) in the presence of helicase or the E. coli DnaG protein (0.89 ϫ 10 Ϫ4 oligoribonucleotide/s/monomer) (39) in the absence of helicase.
We propose that the translocation activity of T7 primase is critical for the acquisition of primase sites on large DNA molecules such as M13 ssDNA and that it also accounts for the greater specificity for the primase sites 5Ј-GGGTC-3Ј, 5Ј-TG-GTC-3Ј, and 5Ј-GTGTC-3Ј on M13 DNA. In the case of M13 ssDNA, primase sites are reached via the translocation activity of the gene 4 protein, whereas sites on the considerably shorter oligonucleotides are most likely accessed by random collision with the template. In the former case we suggest that there is a competition between the rapid translocation activity (300 nt/s) and the synthesis of oligoribonucleotides (0.8 oligoribonucleotide/s) at a primase recognition site. In this regard it is important to recall that the gene 4 protein translocates 5Ј to 3Ј on the strand to which it is bound but upon reaching a primase site polymerization of nucleotides must proceed in the opposite direction, 3Ј to 5Ј with regard to the template strand. As discussed below the mechanism of formation of the first phosphodiester bond to yield pppAC may well differ from that involved in the formation of subsequent phosphodiester bonds. Consequently, when the protein is in a translocation mode it may not linger for a sufficient time to extend the dinucleotide in those cases where a kinetically less favorable base pair is involved, namely the incorporation of GMP or UMP. The fact that the incorporation of GMP at the primase site 5Ј-GCGTC-3Ј decreases once the 5Ј-flanking region exceeds a length of 20 nucleotides supports this interpretation. Below this critical length the helicase domain does not have a sufficient stretch to engage its 5Ј to 3Ј translocation activity thus providing the gene 4 protein with sufficient time to incorporate an unfavorable nucleotide.
Likewise, we believe that our finding (Fig. 7) that the incorporation of GMP at the site 5Ј-TCGTC-3Ј always results in the synthesis of a functional tetraribonucleotide primer, pppACGA, when primer synthesis is coupled to DNA synthesis is consistent with the above interpretation. The intermediate pppACG is not observed under these conditions but pppACGA is found at the 5Ј-termini of a significant number of Okazaki fragments. This finding is compatible with a model in which the gene 4 protein incorporates GMP at this site only when its translocation movement has slowed, an event that also facilitates the transfer of the resulting oligoribonucleotide to the DNA polymerase for use as a primer.
Studies with E. coli DnaG primase indicate that the rate determining step in primer synthesis is the formation of the first phosphodiester bond to generate a dinucleotide (39), thus suggesting that this reaction proceeds by a different mechanism than the subsequent extension of the dinucleotide. Crosslinking studies have shown that the DnaG primase and ␤-sub- The products were separated by electrophoresis on a 25% urea-gel and subjected to autoradiography. The identities of the oligoribonucleotide species are indicated. C, DNA synthesis on the minimal replication fork is performed as described under "Experimental Procedures." The reaction was terminated by the addition of 0.1% SDS and phenol extraction. After ethanol precipitation, half of the products were loaded directly onto a 6% urea-gel (lanes 1 and 2). The remaining reaction products were digested with BamHI and then separated by 6% urea-gel (lanes 3 and 4). After electrophoresis, DNA fragments containing ribonucleotide primer labeled with [␣-32 P]CMP were visualized by autoradiography. Lane M, HpaII digest of pUC19 plasmid labeled with [␣-32 P]dCMP. unit of E. coli RNA polymerase bind ATP or GTP, the first nucleotide, and maintain contact with it during the extension reaction of the next several nucleotides (50,51). At some point however, RNA polymerases must release the contact between extending nucleotides and the starting nucleotide and switch to a stable elongation mode. In those instances where an efficient transition between the two modes does not occur, abortive transcripts are released (52)(53)(54). This model of polymerization readily explains our observation that misincorporation of nucleotides at a primase recognition site is rare within the first three nucleotides but occurs more frequently at the fourth and fifth nucleotide position (Fig. 4). We propose that the lack of misincorporation in the first three positions arises from a more stable interaction of the primase with the growing chain through simultaneous contacts with the starting nucleotide and the 3Ј-end. The model of polymerization can also explain the observation in Fig. 4 that oligoribonucleotides with GMP or UMP at the third position of the primer were more efficiently extended to the 5Ј-terminus of the templates than were those with CMP at this position. Incorporation of GMP or UMP at this position may require a switch from the elongation mode to the stable phase and therefore the incorporation of GMP or UMP leads to a complex that now readily polymerize nucleotide to the end of the template.
DNA primases and RNA polymerases lack a proof reading exonuclease characteristic of that found in many DNA polymerases. RNA polymerases, however, have a much higher fidelity of incorporation than do DNA primases (43). Since each event of primer synthesis involves a competition between the opposing movements of the associated helicase translocation activity and the polymerization of ribonucleotides, it seems reasonable that DNA primases have evolved such that they sacrifice fidelity for speed of primer synthesis. Furthermore, the resulting primers are eventually removed from the Okazaki fragments by RNase H activities in preparation for ligation.