Initiation of Bidirectional Replication at the Chromosomal Origin Is Directed by the Interaction between Helicase and Primase*

Several protein-protein interactions have been shown to be critical for proper replication fork function inEscherichia coli. These include interactions between the polymerase and the helicase, the helicase and the primase, and the primase and the polymerase. We have studied the influence of these interactions on proper initiation at oriC by using mutant primases defective in their interaction with the helicase and DNA polymerase III holoenzyme lacking the τ subunit so that it will not interact with the helicase. We show here that accurate initiation of bidirectional DNA replication from oriC is dependent on proper placement of the primers for leading strand synthesis and is thus governed primarily by the interaction between the helicase and primase.

synthesized on the lagging strand template is governed by an interaction between DnaG and DnaB (12).
We have shown previously that primase concentration affected the mode of DNA replication in oriC replication systems dramatically (14). Coordinated leading and lagging strand synthesis and bidirectional initiation of replication at oriC occurred only at high concentrations of primase. We suggested that this reflected a requirement for primase to act during a "replication fork assembly" step that occurred subsequent to the loading of DnaB to the DNA, but prior to nascent chain elongation by pol III. The molecular basis of this step, however, was unclear. The high concentrations of primase involved implied the need to ensure occupancy of a binding site at a specified time, but binding could have been mediated by DnaB (12,13), DnaA (15), or a subunit of the HE (11).
In this report we have continued our investigation of this effect. The role of these protein-protein interactions during initiation at oriC was investigated by using mutant primases defective in their interaction with DnaB and HE lacking the subunit. We show that proper bidirectional initiation depended only on accurate placement of the primers for leading strand synthesis that, in turn, depended only on the interaction between DnaB and DnaG.

MATERIALS AND METHODS
Replication Proteins and Plasmid DNAs-E. coli DNA replication proteins have been described previously. The preparations of mutant primases were described by Tougu and Marians (16). The subunits of pol III were generous gifts of Charles McHenry (University of Colorado, Denver, CO), and pol III HE was reconstituted as described previously (8).
An oriC plasmid, pBROTB535 type I, was prepared according to Marians et al. (17) with slight modifications. 2 Briefly, pBROTB535 type I DNA (14) was prepared from E. coli K38 tus::kan by the alkaline lysis procedure (18). The plasmid DNA was then banded in CsCl and then purified by sedimentation through 5-20% sucrose gradients containing 1 M NaCl. Form I DNA was dialyzed against TEN buffer extensively, concentrated by extraction with sec-butanol, dialyzed again, and then ethanol-precipitated.
oriC DNA Replication-Standard oriC DNA replication reaction mixtures (12.5 l) were as described by Hiasa and Marians (14). Tus protein, mutant primases, and reconstituted pol III HEs (RHEs) were as indicated in the figure legends. Gel electrophoretic analyses of replication products were performed as described previously (15).
Pulse-Chase Analysis of Replication Products-Pulse-chase analysis of oriC DNA replication products was performed according to Hiasa et al. (19) except that pol III* and DnaN were replaced with RHEs. Any other changes in reaction conditions are indicated in the figure legends.

RESULTS
The Interaction between Helicase and Primase Modulates the Mode of oriC DNA Replication-In our previous studies (14), we found that varying the primase concentration had a striking effect on the DNA products generated during oriC DNA replication in vitro. At low concentrations of primase, the dominate mode of replication was asymmetric, where synthesis of each strand was initiated at random positions both at and outside of oriC and synthesized continuously. This led to rolling circle DNA replication at longer incubation times. As a result, the majority of the products formed were either unit length or greater in size. In contrast, at high concentrations of primase, rolling circle type DNA replication was completely inhibited and the replication products appeared as two populations, one half unit length and the other centered about 400 nucleotides (nt). These populations were shown to be nascent leading and lagging strands, respectively. Under these conditions, DNA replication was initiated at and only at oriC and proceeded bidirectionally.
Primase-catalyzed primer synthesis during DNA replication in the presence of SSB requires DnaB (20). The enzyme does have a low intrinsic affinity for DNA that allows it to catalyze primer synthesis on naked single-stranded DNA, but this activity is one three-hundredth that observed in the presence of DnaB (12). In addition, it is completely inhibited in the presence of SSB (12). Thus, during oriC replication, access of primase to the DNA is directed by its interaction with DnaB.
To determine whether the effect of varying the primase concentration on the mode of oriC DNA replication was a result of the interaction between primase and DnaB, oriC DNA replication was reconstituted with purified replication proteins using either the wild-type or mutant primases, DnaG K580A and DnaG Q576A. In previous studies, we identified a C-terminal domain of primase that was required for functional interaction with DnaB (12). Subsequently, we defined the region on primase involved as the last 16 C-terminal amino acids (16). Mutant primases lacking this domain are unable to prime DNA synthesis in any DNA replication reaction requiring DnaB (13,16). However, using the specialized bacteriophage G4 origin, where primase binds to a region of secondary structure on the phage DNA in the presence of SSB to synthesize a primer (21,22), it could be demonstrated that the ability of the mutant proteins to catalyze oligoribonucleotide synthesis was completely unaffected (13,16).
The K580A and Q576A mutant proteins contain amino acid substitutions that were constructed in the C-terminal interaction domain (13,16). Using a rolling circle DNA replication system, where we showed previously that the size of Okazaki fragments was directed by the primase-DnaB interaction and varied inversely with primase concentration (23)(24)(25), DnaG K580A exhibits a slightly lower affinity for DnaB, whereas that of DnaG Q576A is severely reduced, as measured by the variation in Okazaki fragment size as a function of primase concentration (13). This has been confirmed by the demonstration that the strength of the physical interaction between primase and DnaB is reduced in the presence of the Q576A mutant primase (15).
Variation of the concentration of either the wild-type or DnaG K580A primase in the oriC replication system produced an identical pattern of DNA products (Fig. 1). On the other hand, about 15-fold higher concentrations of DnaG Q576A were required to observe similar patterns of DNA products (Fig. 1). In each case, however, the same shift in replication products from those characteristic of rolling circle replication to those characteristic of bidirectional replication was observed inversely correlated to primase concentration.
To confirm the mode of replication extent using high concentrations of the mutant primases, a direct test, which we had developed previously (14), was applied. This test utilizes a blocking template that carries oriC and two TerB sequences that are located 2 and 3 kb in opposite directions away from oriC, with roughly 1 kb of sequence between them. The TerB sequences in the blocking template are oriented so as to exclude passage of replication forks between them when they are bound by Tus. Thus, this template allows for a rapid determination of the mode of DNA replication. If DNA replication from oriC is bidirectional, denaturing gel analysis of the DNA products generated in the presence of Tus protein will show two distinct leading strands, 2 and 3 kb in length, and a population of small Okazaki fragments. In contrast, if DNA replication is continuous from random positions, the product analysis will show a maximum length of 5 kb with a smear of smaller products.
Replication reaction mixtures containing the blocking template were therefore incubated in the presence of Tus, and either low or high concentrations of both the wild-type and mutant primases and the DNA products were analyzed by electrophoresis through denaturing alkaline agarose gels (Fig.  2). As expected, at high concentrations of primases, two prominent bands corresponding to 2 and 3 kb and a discrete Okazaki fragment population were observed in all cases. On the other hand, one band corresponding to 5 kb with an associated smear of smaller products was observed in all cases when the primase concentration was low. In the absence of Tus, the pattern of DNA products observed are identical to the ones displayed in Fig. 1.
Thus, the same effect on the mode of DNA replication was observed for both the wild-type and mutant enzymes. In addition, the same qualitative dependence on primase concentration was also observed, although higher concentrations of DnaG Q576A were required compared with wild type. About 15-fold higher concentrations of DnaGQ576A were required to give the pattern typical of saturation of primase concentration in the reaction. This is identical to the shift in concentration of DnaGQ576A required to saturate the primase-directed variation in Okazaki fragment size during rolling circle DNA replication and thus parallels the reduction in affinity of the mutant primase for DnaB (13,16). We therefore conclude that the mode of replication from oriC is directed by the primase-helicase interaction. The question that we addressed next was whether this reflected a requirement for the first leading strand primer to be placed close enough to DnaB that when it was bound to the DNA pol III HE, formation of the -DnaB interaction was facilitated. To do so, we first investigated the requirement for the subunit of the HE during oriC DNA replication.
Replication Forks Formed in the Absence of Can Support oriC DNA Replication in Vitro-The DNA products from oriC replication reactions incubated in the presence of high concentrations of wild-type primase and HE reconstituted either in the presence (RHE (ϩ)) or absence of (RHE (Ϫ)) were analyzed by agarose gel electrophoresis (Fig. 3). Under these conditions, both RHE (ϩ) and RHE (Ϫ) were capable of supporting oriC DNA replication in vitro, although the efficiency of the RHE (Ϫ)-supported oriC DNA replication was slightly (10 -20%) lower than that of the RHE (ϩ)-supported reaction. Analysis by native agarose gel electrophoresis showed that the major products were highly linked DNA dimers and late Cairn's-type replication intermediates (Fig. 3A). These were the DNA products expected because these reactions contained DNA gyrase, which can not decatenate the linked daughter molecules (26), as the only topoisomerase present. Two distinct populations of DNA products were observed when replication products were analyzed through denaturing alkaline agarose gels (Fig. 3B). The larger population, which centered around half-unit length, represented nascent leading strands, and the shorter population represented nascent lagging strands. The lagging strands generated in the RHE (Ϫ)-supported reaction were slightly shorter than those generated in the RHE (ϩ)supported reaction. This may be because of the uncoupling of the pol III cores at the fork and/or the absence of the HE-DnaB interaction.
The subunit of the pol III HE interacts directly with DnaB and this interaction is required for rapid replication fork progression (8). Thus, the oriC pulse-chase protocol was employed to compare the rate of progression of replication forks formed in either the presence or absence of (Fig. 4). Early replication intermediates were formed and 32 P-labeled by incubating the replication system in the presence of [␣-32 P]dATP but in the absence of any topoisomerase. Under these conditions, replication forks form at oriC on the supercoiled template and elongation can proceed in the absence of a topoisomerase until positive supercoils accumulate. At this point (the early intermediate, where the nascent leading strands are about 600 -800 nt in length), continuation of the elongation phase of the reaction requires release of the accumulated topological strain (19). In the experiment shown in Fig. 4, the paused replication forks were released by linearizing the DNA template with the SmaI restriction endonuclease, which cuts the template once, just counterclockwise from oriC, at the same time as the nucleotide label was chased. Rapid elongation of the nascent DNA in the early replication intermediates to full-length products was observed when RHE (ϩ) was used. Within 30 s after the addition of SmaI, full-length DNA products were observed, indicating that the rate of replication fork progression was at least 180 nt/s. Because digestion of the DNA template by SmaI took some time (10 -15 s), the actual rate of replication forks was estimated to be 270 -360 nt/s. In contrast, replication forks formed with RHE (Ϫ) progressed slowly. It took 2 min to observe the appearance of full-length DNA products. The rate of replication fork progression under these conditions was calculated to be 50 -70 nt/s. Thus, similar to what has been reported in the rolling circle replication system (8), the interaction between and DnaB was required for rapid progression of replication forks formed as a result of initiation at oriC.
Even in the Absence of , the Helicase-Primase Interaction Governs the Mode of Replication from oriC-To determine whether the -DnaB interaction had any effect on the mode of replication from oriC, we asked whether the primase concentration-dependent switch in the mode of replication was still observed in the absence of . This proved to be the case. Replication products generated by replication forks formed with either RHE (ϩ) or RHE (Ϫ) responded in an identical fashion to variation of the primase concentration (Fig. 5). In each case, as the concentration of primase was increased, the mode of replication changed from rolling circle to what appeared to be bidirectional.
To confirm that initiation of replication was bidirectional in the absence of , we examined replication on the blocking template in the presence and absence of Tus. In both the presence and absence of , the addition of Tus generated two distinct bands corresponding to 2 and 3 kb in length (Fig. 6), indicating that DNA replication was initiated at and only at oriC and proceeded bidirectionally in each case. These results show that an interaction between DnaB and the pol III HE was not required for proper initiation of bidirectional DNA replication from oriC.

DISCUSSION
Initiation of DNA replication at chromosomal origins is a highly ordered process that requires identification of the origin sequence, localized denaturation of the origin region, and assembly of two replication forks that will subsequently replicate the chromosome, simultaneously synthesizing the nascent leading and lagging strands. In E. coli, the high degree of coordination demanded is specified by a sequential series of interactions between replication proteins.
The initial protein to act in this cascade is DnaA, which recognizes, binds to, and locally denatures the origin (6,27). These events require extensive oligomerization of DnaA at the origin. The manner in which activities are assigned to individual protomers (or groups of them) in this DnaA aggregate is not yet appreciated. The single-stranded DNA in the resultant denaturation bubble is presumably rapidly coated with SSB.
The key to subsequent replication fork assembly is the introduction of the replication fork helicase, DnaB, to the DNA. Avoidance of the promiscuous introduction of this protein to any single-stranded DNA in the cell comes about because DnaB cannot bind to SSB-coated DNA. This creates a demand for another specific interaction, this time between DnaA and DnaB (7), that allows DnaB to bind the denaturation bubble.
This point in the temporal sequence of events during initiation is crucial for the determination of the mode of replication. DnaB is a motor protein that can rapidly move away from the origin in an ATP-dependent manner. As we have shown previously for replication of small plasmid templates (14), if it does so before replication forks are established, synthesis of each strand becomes continuous and the location of initiation becomes random. We determined that a high concentration of primase was required to ensure that replication was, in fact, bidirectional, initiating at or very near oriC.
Replication fork formation is driven by the obligatory devel- opment of an interaction between the subunit of the pol III HE and DnaB that acts to couple DNA synthesis to unwinding of the parental template (8). We sought to determine in this report whether the required high concentration of primase reflected a requirement for primer placement to facilitate establishment of the -DnaB interaction or a requirement to mark the origin region and the site of initiation of leading strand synthesis immediately upon localized denaturation, an event that would only be dependent on the DnaG-DnaB interaction. To address this question, we examined the mode of DNA replication in the presence of pol III HEs reconstituted both with and without the subunit and mutant primases that had altered interactions with DnaB.
pol III HE reconstituted without the subunit could support oriC DNA replication in vitro. On its face, the observation that uncoupled replication forks could support DNA replication seems surprising. However, there is precedent in the monoand dipolymerase systems for SV40 DNA replication in vitro (28). HE reconstituted in the absence of presumably synthesizes leading strands in a distributive manner, just as DNA polymerase ␣ does in the monopolymerase system. However, it is probable that this observation is a manifestation of reconstituting DNA replication in vitro. The efficiency gain generated by the increased speed of replication fork progression as a result of the -DnaB interaction (8) is likely to be required in vivo where a large amount of DNA must be duplicated in a short time. This time constraint is not operative in the in vitro system. Nevertheless, the ability to observe -independent oriC replication allowed us to determine that it was the DnaG-DnaB interaction that was governing the mode of replication initiation.
An identical switch from asymmetric, randomly initiated replication to bidirectional replication in inverse relationship to the primase concentration was observed when oriC DNA replication was performed using HE reconstituted in either the presence or absence of , demonstrating that the lack of a DnaB-pol III HE interaction had no effect on oriC-specific initiation of bidirectional DNA replication that was dependent on the DnaB-DnaG interaction. This conclusion is also supported by the observation that mutant primases, having as their only defect an altered interaction with DnaB, behaved identically to the wild type with the only difference being a shift to the right in the concentration dependence of the switch in replication mode. This also argues against the possibility that it is an interaction between DnaA and DnaG that locates the primase to the origin region. If this were the case, the shift in concentration of primase required to alter the mode of replication when the mutant enzymes were used would not have been observed.
The dominance of the DnaB-DnaG interaction in directing the mode of replication from oriC suggests that it is the location of the first primers synthesized that is the governing feature. These are the primers for leading strand synthesis. How this results in essentially assuring that bidirectional initiation takes place is not clear. It could be that, even if only one replication fork initiates, the second leading strand primer serves as a roadblock to conversion to rolling-circle DNA synthesis. Another intriguing, although highly speculative, possibility is that this reflects a requirement for the formation of a dimeric replisome at the origin, resulting in the coupling of the two replication forks, as has been proposed for initiation at the SV40 origin (29).