Chapter 50

Replication Fork Propagation

KENNETH J. MARIANS

INTRODUCTION

Duplication of the 4.4-Mb Escherichia coli chromosome requires the coordinated action of 25 to 30 different gene products. The majority of these proteins participate, either directly or indirectly, in the synthesis of nascent DNA during the propagation of the replication fork. In E. coli, the fork proceeds at nearly 1,000 nucleotides (nt) per s in a manner so accurate that mistakes (i.e., misincorporated nucleotides) are made at a frequency of less than 1 in every 10⁹ polymerization events. The focus of this chapter will be a description of our current knowledge as to how the various enzymatic components fit together to form this remarkable protein machine that has been termed the replisome (62).

Historical information describing the isolation of various enzymes and the development of the original in vitro replication systems can be found in the chapter by McMacken et al. (93) in the first edition of this compendium. In addition, there are a number of recent reviews (14, 83, 88) and texts (62) that have DNA replication as their focus.

Recent studies show clearly that the initiation of DNA replication at oriC and the manner in which this process is linked with the cell cycle at least equal replication fork progression in enzymatic complexity. This aspect of the replication process, however, is covered in chapter 98. And whereas our knowledge of the final stages of DNA replication is less complete than our knowledge of initiation and replication fork propagation, it is increasing rapidly. This information is summarized in chapter 100.

REPLICATION FORK OPERATIONS

Once formed, there are four basic components to a replication fork (Fig. 1): the DNA polymerase required for nascent strand synthesis, the DNA helicase required to unwind the parental duplex DNA, the primase required to initiate Okazaki fragment synthesis, and the single-stranded DNA (ssDNA)-binding protein (SSB) required to coat exposed template ssDNA. A fundamental asymmetry is introduced to the enzymatic requirements for DNA synthesis at the fork because of the antiparallel nature of the template strands and the fact that DNA polymerases only synthesize DNA in the 5'→3' direction. Thus, one strand—the leading strand—can be made in a continuous fashion. The leading-strand DNA polymerase, if processive enough, need be introduced to the replication fork only once. On the other hand, the other strand—the lagging strand—is made discontinuously in small pieces roughly 2 kb in length (Okazaki fragments). Thus, a replication fork appears to require two distinct DNA polymerases, one highly processive and one moderately processive. In E. coli, the difference in the degree of processivity required is on the order of 1,000-fold, yet the same enzyme, the DNA polymerase III holoenzyme (Pol III HE), synthesizes both the nascent leading and lagging strands. How is the same catalytic site made to possess such distinctive properties? The solution to this problem, i.e., what makes the polymerase processive, is one of the most elegant in biology. Hypotheses detailing how processivity may be limited on the lagging strand will also be described.

Fig. 1Schematic of a replication fork and the enzymatic activities required for propagation.

Source:

DNA polymerases cannot initiate DNA synthesis de novo. Unique enzymes, the primases, have therefore evolved to synthesize short ribonucleotide primers that are then utilized by the lagging-strand polymerase for Okazaki fragment synthesis. In E. coli, a new Okazaki fragment must be initiated once every 1 or 2 s. To ensure that the primase has ready access to the template when and where it is needed, its association with the replication fork is mediated by the DNA helicase acting on the lagging-strand template to unwind the parental duplex in the 5'→3' direction. This mobile complex of helicase and primase has been termed a primosome (62, 83). In E. coli, the components of the primosome at the replication fork are still in question. It is clear that the primosome must include DnaB, the replication fork helicase (68), and DnaG, the primase (19), and requires DnaC for function. Evidence will be considered that the primosome may also contain, or at least require for assembly, four other proteins, DnaT, PriA, PriB, and PriC.

Initiation, synthesis, and completion of an Okazaki fragment occur very rapidly. This creates an additional layer of complexity at the replication fork. Does the same DNA polymerase complex synthesize each and every Okazaki fragment made by the replication fork? This would require mechanisms to cycle the polymerase from completed Okazaki fragment to new primer terminus very rapidly and to ensure that when the polymerase releases the template after completion of the Okazaki fragment, it remains associated with the replication fork. Or is a new polymerase recruited from the pool in the cytoplasm for each new Okazaki fragment? At least in the case of the E. coli fork, it now seems clear that the former scenario applies.

Finally, the RNA primers used to initiate synthesis of the Okazaki fragments must be removed, the gap between consecutive Okazaki fragments must be repaired, and the nicks must be sealed to yield a continuous nascent lagging strand. This is accomplished by the combined action of the 5'→3' exonuclease of DNA polymerase I (Pol I), RNase H1, and DNA ligase.

PROTEINS INVOLVED IN REPLICATION FORK PROPAGATION

The DNA Pol III HE

Structure.

The DNA Pol III HE (147) is the replicative polymerase (41). Cells mutated in dnaE, which encodes the α subunit of the HE (138), are temperature sensitive for growth and display an immediate-stop DNA replication phenotype (136). That is, upon a shift-up to the nonpermissive temperature, DNA synthesis, as measured by the incorporation of [³H]thymidine into acid-insoluble product, ceases immediately. This has been taken to indicate that the gene product was required for elongation of nascent chains or, in other words, for replication fork propagation. The HE is composed of 10 subunits (Table 1). It has taken 20 years of intensive enzymology and genetics to identify all the genes encoding these subunits and to demonstrate that, in fact, their copurification was not fortuitous but was a result of specific interactions between them. I will detail here only the most current information on HE composition and function.

Table 1HE subunits

Early studies showed that various subassemblies of the HE could be isolated by purification from bulk E. coli. These subassemblies differed in their processivity (35, 36) and response to the presence of SSB on the template. In ascending order of processivity they are: the core (αεθ) (89), pol III' (αεθ)₂ τ ₂ (87), and pol ΙΙΙ∗ (αεθ)₂ τ ₂ γ ₂ δδ'χψ (148). A separate complex of HE accessory proteins, the γ complex (γ ₂ δδ'χψ), could be purified free of other HE subunits (91). The HE is completed by the addition of (β ₂)₂ to pol III* (91, 148). The variation in observed processivity in vitro of these polymerase forms ranges from a few nucleotides (the core) (35) to greater than 7 kb for the HE alone on a primed ssDNA template (21, 35), to greater than 0.5 Mb for the HE when it is reconstituted into a replication fork in the presence of the φX-type primosomal proteins (95, 151). As described below, the HE subunits have different roles that can be summarized as follows: the γ complex loads β, which encircles the DNA, onto a primer terminus where it associates with the core, thereby clamping the polymerase to the DNA during elongation.

The recent identification of the genes encoding δ, δ', χ, ψ, and θ, and the ability to purify each of the 10 subunits to homogeneity from strains engineered to overproduce them, has accelerated the accrual of information on HE function tremendously.

The associations between the HE subunits have been determined by O’Donnell and his colleagues. Each of the complexes mentioned can be isolated by gel filtration. α binds ε and ε binds θ; however, θ does not bind to α (125). Of course, αε binds θ to form the core (αεθ). β is a dimer in solution (53). β ₂ binds either α or αε to form αβ ₂ or αεβ ₂ (126). Therefore, the association of β with the HE is via the α subunit. τ is also a dimer in solution (124), although higher oligomeric forms can be found at higher concentrations (M. O’Donnell, personal communication). τ ₂ will dimerize either α, αε, or αεθ to form α ₂ τ ₂, (αε)₂ τ ₂, or (αεθ)₂ τ ₂(pol ΙΙΙ') (87), respectively (124). τ ₂ will not dimerize ε (124); therefore it associates with the HE via the α subunit as well. When τ ₂ is in excess of core or αε, then αετ ₂ or αεθτ _2, respectively, will form rather than dimerization occurring. γ, which represents the amino-terminal two-thirds of τ (see below), is incapable of dimerizing α or αε (124). Thus, the carboxy-terminal third of τ is required for dimerization of the core.

The γ complex appears to assemble in a stepwise fashion (106, 154). Like τ, γ is also a dimer in solution but is prone to forming larger aggregates at high concentrations (O’Donnell, personal communication). δ and δ' associate to form a δδ' complex. δδ' can associate with γ to give a γ ₂ δδ' complex_{. γ} will not associate with either δ or δ' individually. Interestingly, τ can mirror the interaction of γ with δδ' to form τ ₂ δδ' _. This parallels the interchangeable activities of τ and γ (see below). χ and ψ associate to form χψ. γ can associate with ψ to form γ ₂ ψ, but will not associate with χ. When γ, ψ, and χ are mixed together, γ ₂ χψ will form. Thus, the γ complex is anchored by γ associating with δδ' and χψ. Once again, τ mirrors γ in these interactions.

Complete formation of pol III* is somewhat more complicated (R. Onrust, Ph.D. thesis, Cornell University, Ithaca, N.Y., 1993). The γ complex will not associate with pol III'. This appears to be a result of δ masking a site on γ that is required to bind τ. On the other hand, if γ and τ are mixed together first, a heterocomplex forms with a likely stoichiometry of γ ₂ τ ₂. If first the core and then the other members of the γ complex are now added to the γτ heterocomplex, pol III* will form, with the most likely composition being (αεθ)₂ τ ₂ γ ₂ δδ'ψχ. The addition of two β dimers then yields the complete HE.

Formation of a Processive Polymerase Complex.

Conversion of the catalytic Pol III core to a highly processive form, capable of adding at least 0.5 Mb to a primer terminus in one binding event, requires the ATP-dependent formation of an initiation complex in which the core is clamped onto the primer terminus (21, 35, 36). The mechanism of assembly of this sliding clamp and its identity have been clarified recently. Initial studies demonstrated that the presence of β was essential for initiation complex formation (one capable of subsequent processive DNA replication) and that in the absence of the core, β could be bound to the primer terminus in an activated form by the action of purified γ and elongation factor III (possibly τδ') in a reaction that required ATP hydrolysis. Subsequent addition of the core resulted in the formation of a processive complex (139).

O’Donnell (101) defined these steps more precisely, showing that preinitiation complex formation was a result of the γ complex acting to transfer, in an ATP-dependent fashion, β to the primer template. This preinitiation complex could be isolated and converted to an initiation complex by the addition of the core. Initiation complexes are stable in the absence of nucleotide polymerization as long as ATP is present (36, 53). Preinitiation complexes require the presence of both ATP and the γ complex for maximum stability (40). αε is as active and processive as αεθ when added to a preinitiation complex, indicating the θ is not required for processive polymerization (122). This is consistent with the observation that holE is not required for E. coli growth (120). However, α alone is slower and less processive than αε (122). Thus, ε, aside from carrying the 3'→5' exonuclease, must also provide a structural function.

Stukenberg et al. (126) showed that the γ complex functioned catalytically in preinitiation complex formation and could not be detected (by Western blotting [immunoblotting]) in an isolated preinitiation complex. This argues that a preinitiation complex is simply a β dimer transferred to the primer template by the ATP hydrolysis-dependent chaperonin-like action of the γ complex.

As resolution of the HE into its components proceeded and individual subunits were isolated, studies on the components of the γ complex required for initiation complex formation were undertaken. Thus, it could be shown that, in the presence of αε and β, either γδ, τδ, or τδ' was sufficient to support initiation complex formation (102). With the recent availability of highly purified HE subunits, these observations have been put into perspective.

Under physiological conditions, the active agents in initiation complex formation are likely to be either γδδ'or τδδ'. From 25- to 30-fold higher concentrations of γ are required to obtain comparable replication activity with δ alone than with δδ' (106). Similarly, 500-fold higher concentrations of τ are required to obtain comparable replication activity with either δ or δ' alone than with δδ' (106). In the absence of τ, χψ stimulates replication with αε, β, and γδδ' by 3.5-fold (154).

Initiation complex formation requires ATP hydrolysis (21, 35, 36), although some complex formation occurs in the presence of adenosine 5'-O-(3-thiotriphosphate) (ATPγS) (54). It seems likely that this ATP hydrolysis is effected by either τ or γ. Both are ssDNA-dependent ATPases, although the turnover of the τ ATPase is 54-fold greater than that of γ (106). On the other hand, the ATPase activity of the γ complex is stimulated 20-fold by primed DNA [poly(dA)-oligo(dT)] (106) whereas the τ ATPase is not, although in the presence of Mn²⁺ a similar stimulation of the τ ATPase is observed (129).

The stimulation of the γ and τ ATPases by the other accessory protein subunits mirrors both their association properties and their replication properties (106, 154). Thus, the γ ssDNA-dependent ATPase was stimulated 3-fold and 11-fold by δ and δ', respectively, but was stimulated 140-fold by δδ'. The τ ssDNA-dependent ATPase was not stimulated by either δ or δ' alone, but was stimulated eightfold by the δδ' combination. β had no effect on the γ, γδ, or γδ' ATPase, but did stimulate the γδδ'ATPase activity by 2.5-fold. Notably, the γδδ' ATPase was not stimulated by β. This may be a crucial distinction, suggesting that not all γ functions can be replaced in the HE by τ.

The γ ATPase was also stimulated threefold by ψ, but not by χ, whereas the τ ATPase was unaffected (154). Thus, based on the slight stimulation of the γ ATPase and the γδδ'replication activities by χψ, it seems that these latter two subunits may serve to stabilize the γ complex and may not participate directly in any HE function. This would be consistent with the observation that holC is dispensable for E. coli growth (S. Slater and R. Maurer, personal communication).

Up to this point, we have considered which assortment of HE subunits was required to form a processive polymerase complex. What is the nature of this extraordinary processivity? As it turns out, the manner in which an essentially nonprocessive enzyme, the core, is transformed into a superprocessive polymerase is one of the most elegant in biology. Stukenberg et al. (126) demonstrated that whereas β could be isolated by gel filtration bound to a fully replicated form II DNA, linearization of the DNA prior to gel filtration resulted in the inability to recover β in the excluded fractions, suggesting that under the latter circumstances β slid off the DNA. Additional studies showed that whereas β could diffuse along duplex DNA in either direction, it could not diffuse over ssDNA, whether it was coated with SSB or not. Clearly the nature of the sliding clamp of a processively engaged core is this ability of β to diffuse along duplex DNA. Polymerization adds a vector sense to the diffusion so that one may think of β, which would be locked topologically on the DNA, as being pulled along by the core while at the same time passively acting to lock the polymerase onto the primer template.

This view of β as a sliding topological clamp has been remarkably realized with the solution of the crystal structure of a β dimer (61). The overall structure is that of a toroid with a diameter of about 80 Å (8 nm). The protein crystallizes as a head-to-tail dimer. The diameter of the inner hole is about 35 Å, sufficient to accommodate either A or B form DNA. The twofold dimer axis is perpendicular to the face of the ring, and the thickness of the ring is about 34 Å.

Though not obvious from the amino acid sequence, a β monomer contains three repeating structural domains that are nearly superimposable. The inner hole is lined by 12 α-helices (two from each of the six repeating structural domains). When DNA is modeled into the hole in β, it becomes clear that these α-helices are what allow β to diffuse along the DNA. They are perpendicular to the local direction of the phosphate backbone; thus each pair of helices spans the major and minor grooves. Presumably, this prevents the β ring from getting caught on the double helix as it slides along it. The electrostatic charge of the surface of the hole is positive, probably contributing to the stability of the β dimer on the DNA.

Considering the β structure, it is clear that the primary task of the γ complex during preinitiation complex formation can be thought of as opening the β ring in solution, guiding it around the duplex primer-template, and reclosing the ring. Presumably, the ATP hydrolysis requirement for placing β on the primer reflects a chaperonin-like action of the γ complex to partially denature and renature β. The subsequent association on the primer of β and the core does not require the action of any other accessory protein. Recent evidence (O’Donnell, personal communication) indicates that, during formation of the initiation complex, α interacts with the face of the β ring that contains the carboxy-terminal ends of the β monomers.

Genetics of the HE.

As described above, all 10 subunits of the HE can be shown to associate with each other, supporting the argument that their isolation as a multienzyme complex was functional rather than fortuitous. By and large, the genetic phenotypes of mutations in the genes encoding the HE subunits support this.

Four subunits, α (dnaE), β (dnaN), τ (dnaX), and γ (dnaX), are encoded by genes that, when mutated, display an immediate-stop DNA replication phenotype when the mutant cells are raised to the nonpermissive temperature.

τ is the full-length product of dnaX, while γ is a shorter gene product produced by a translational frameshift that results in subsequent premature termination (18, 39, 130). The key mRNA features necessary for frameshifting are an A₆ stretch encoding two Lys residues at codons 429 and 430 that allows the ribosome to slip, when it is paused at a stable (-28 kcal) hairpin formed from codons 433–442, into the –1 reading frame. Subsequent translation in the new frame results in termination after the addition of two amino acids, resulting in a 431-amino-acid γ protein. The first 430 amino acids are identical to those of τ.

Blinkova et al. (17) have recently shown that only τ is required for E. coli growth. They engineered dnaX gene copies that produced either only τ or only γ. When present on plasmids, the τ-only dnaX gene was able to complement temperature-sensitive dnaX mutations whether the mutation occurred in the carboxyl-terminal one-third unique to τ or in the amino-terminal two-thirds common to both τ and γ. The γ-only dnaX gene could not complement any dnaX mutations. In addition, these authors could construct E. coli strains carrying a single copy of the τ-only dnaX gene, but not one with the γ-only dnaX gene.

Clearly this raises the question as to whether the presence of γ in the HE is possibly an artifact. A γ-like polypeptide (residues 1–498 of τ) can be generated from τ by proteolysis (74). In fact, there are no existing data that speak to the question of whether the γ that is present in HE purified as a complex from E. coli was generated by proteolysis or by translational frameshift. In addition, attempts to purify a pol III* complex lacking γ from the E. coli strain that expressed only τ failed. The τ in the complex seemed to be hypersensitive to proteolysis, converting to γ, and no multisubunit HE complex could be isolated except for the γ complex (17). As described above, the activities of τ and γ during initiation complex formation appear to be interchangeable in many respects, the one exception being β stimulation of the τδδ'and γδδ' ATPases. Resolution of this question awaits a complete and detailed understanding of the role of each subunit at the replication fork.

ε is the 3'→5' proofreading subunit of the HE (112). Thus, whereas cells mutated in dnaQ can grow, they are seriously compromised. They grow poorly, exhibit chronic induction of the SOS response, and possess an elevated rate of occurrence of spontaneous mutations (120).

Null mutants of holE (θ) have been constructed (120). These cells appeared to grow normally and showed no defect in DNA replication. Because θ and ε can form a complex, the phenotype of a double dnaQ holE null mutant was also analyzed; however, there was no evidence that a defective holE exacerbated the phenotype of a dnaQ mutation. Thus, genetically, θ appears dispensable.

On the other hand, both holA (δ) and holB (δ') are essential for normal growth (C. S. McHenry, personal communication). Finally, although a holC (χ) deletion is viable, the cells grow very slowly at 37°C, forming small colonies, and cannot grow at 42°C (Slater and Maurer, personal communication). No information is available as yet as to the essential nature of holD (ψ).

DNA Gyrase

The E. coli chromosome is a superhelical, covalently closed, double-stranded (ds) circle (117, 118, 149). The chromosome replicates bidirectionally as a θ structure (16, 108); that is, both parental template strands remain intact during replication. Therefore, because DNA replication requires the unwinding of the parental duplex, as the replication forks proceed away from oriC, to compensate for the loss of β turns, the negative supercoils are first removed and then positive supercoils are introduced. This positive superhelical pressure builds up rapidly and, before even 10% of the chromosome is replicated, reaches a level that can no longer be tolerated. Without removal of these positive supercoils, replication fork propagation will cease. In E. coli, DNA gyrase (45) provides this relief.

Gyrase is an ATP-dependent type II topoisomerase (44, 134). It is composed of two subunits, and the native form of the enzyme is GyrA₂ GyrB₂. The phenotypes of conditionally lethal mutations in gyrA and gyrB support a role for gyrase in replication fork progression in that they display an immediate-stop replication defect (38, 63). There are also other alleles of the gyrase genes that display an initiation defect (34, 37), as well as a par phenotype (121).

The gyrase topoisomerase activity (44, 134) is capable of catenating and decatenating dsDNA rings, as well as knotting and unknotting them. In the absence of ATP, gyrase will relax negatively supercoiled DNA. Unique among the known topoisomerases, only gyrase has the ability to introduce, in an ATP-dependent fashion, negative supercoils into a relaxed DNA. Gyrase functions to remove positive supercoils by converting them directly to negative ones. This comes about because of the nature of the gyrase supercoiling mechanism (20). Thus, not only does gyrase act to remove the positive supercoils generated during replication, it also serves to keep the chromosome negatively supercoiled. This is likely to help drive the replication process forward.

SSB

The ssDNA-binding protein is a tetramer of an 18.8-kDa polypeptide (111). Mutations in ssb have pleiotropic effects, resulting in reductions in DNA replication, repair, and recombination (94). SSB binds ssDNA tightly and cooperatively, with an affinity at least 1,000-fold greater than for dsDNA (78, 96). Thus, SSB can lower the melting temperature of dsDNA significantly (116).

It is likely that SSB-bound ssDNA assumes a particular structure. SSB-coated DNA formed at physiological salt concentrations appears in the electron microscope as a beads-on-a-string type structure. It has been suggested that this arises from the wrapping of the ssDNA around the SSB tetramer (31). Nuclease digestion experiments support this hypothesis, indicating a structure where 145 nt are wrapped around two SSB tetramers to form a "nucleosome" and where each nucleosome is separated by 30 nt (31). This mode of binding probably corresponds to a binding site size of 65 nt per tetramer (78). At lower salt concentrations (<10 mM), binding site size is reduced to 33 nt per tetramer (78) and the SSB-bound DNA now appears in the electron microscope as a smooth nucleoprotein filament (48).

The functionality of SSB during DNA replication fork progression probably arises from three properties of the protein, as follows. (i) Significant stretches of ssDNA, at least equal to the size of the most current nascent Okazaki fragment, exist on the lagging-strand template. SSB-coated ssDNA is refractory to the action of various nucleases (94). Thus, SSB is likely to protect the exposed lagging-strand template from nuclease attack.

(ii) SSB can stimulate both the rate and processivity of the DNA Pol III HE (35). This presumably relates, as mentioned above, to the ability of SSB to melt regions of secondary structure in the DNA that are normally inhibitory to the passage of the polymerase. However, this is apparently a problem only on the lagging-strand template.

(iii) There may be specific interactions between SSB and other enzymes at the replication fork that are required for efficient fork propagation. The SSB encoded by the F factor (SSF) (30, 60), which has DNA binding properties very similar to the cellular SSB, can only substitute for a deletion of the cellular protein in vivo when ssf is present at high copy number. Even so, there is a considerable reduction in the growth rate (107). This could be accounted for if ssf has a weaker affinity than SSB for some other replication fork enzyme. In addition, the ssb-113 mutation results in severe physiological defects (94). However, the SSB-113 protein has virtually identical DNA binding properties to the wild type (29). Thus, it is possible that this mutation reflects impaired protein-protein interactions as well. In support of this, Z. Kelman and M. O’Donnell (personal communication) have demonstrated that the χ subunit of the DNA Pol III HE can interact with SSB both in solution and when it is bound to DNA. No interaction can be detected with the SSB-113 protein.

The Primosomal Proteins

Synthesis of a primer by primase requires a protein-protein interaction with a DnaB molecule on ssDNA (128). The establishment of DnaB stably bound to SSB-coated ssDNA requires the action of additional proteins and occurs in the cell via one of two pathways: assembly of the ABC primosome (85), requiring DnaA and DnaC in addition to DnaB, or assembly of the φX-type primosome (8), requiring DnaC, DnaT, PriA, PriB, and PriC in addition to DnaB (9, 137, 141) (Table 2). Current knowledge about these proteins and the mechanism of priming at the replication fork will be reviewed in this and the next few sections.

Table 2Primosomal proteins

DnaB.

DnaB is the replication fork DNA helicase (68). Cells mutated in dnaB display a fast-stop DNA synthesis phenotype (136). The prediction made by the genetics, that DnaB was involved in replication fork propagation, was supported by the biochemical analysis of the protein.

DnaB is a 52.3-kDa polypeptide that forms a hexamer, which is the active form of the protein, in solution (109). The enzyme is an ss- and dsDNA-dependent nucleoside triphosphatase (NTPase) (7, 110). Binding to nucleic acid is stabilized by nonhydrolyzable forms of ATP, whereas it is destabilized by the nucleotide itself (7). A DNA-independent ATPase activity has also been reported (146); however, it is not clear whether this activity is the result of nucleic acid that contaminates the enzyme preparations.

As a helicase, DnaB moves 5'→3' along the DNA (68). DNA unwinding is stimulated 10- to 20-fold by the presence of a nonhybridized 3' tail on the DNA fragment that is displaced (68). This suggests that DnaB is bound to both template strands when engaged in DNA unwinding at the replication fork.

DnaB helicase activity requires ATP hydrolysis and proceeds at a minimal rate of about 50 nt/s (68). This is considerably slower than the rate of replication fork movement of 1,000 nt/s, suggesting that DnaB action is modified when it acts in coordination with the DNA Pol III HE. Indeed, studies on replication fork action (95) indicate that DnaB and the leading-strand polymerase are likely to be in protein-protein contact.

DnaB helicase activity can be modulated by other replication fork proteins. For example, the E. coli SSB inhibits the ssDNA-dependent ATPase activity of DnaB (7), suggesting that the two proteins compete for the same binding site on DNA, the phosphodiester backbone. Similarly, if SSB is present first on ssDNA, DnaB helicase activity will be inhibited (68). On the other hand, once loaded onto ssDNA, SSB stimulates the helicase, presumably by preventing reannealing of unwound DNA (68). Primase, which is attracted to the replication fork via a protein-protein interaction with DnaB (128), also stimulates the DnaB helicase activity (68).

DnaB forms a stoichiometric complex in solution with DnaC in an ATP-dependent manner (3, 58, 132, 133, 145). Formation of this complex inactivates the DnaB NTPase activity (58, 132, 145). When present in this form, the affinity of DnaB for DNA is raised fivefold (132, 133). Thus, acting in a chaperonin-like manner, DnaC will stimulate all activities that require an interaction between DnaB and DNA (133). It is likely that in the cell, all DnaB not bound to DNA is present as a DnaB-DnaC complex (3).

DnaB is also intimately involved in the priming of Okazaki fragments. Under most circumstances, the presence of DnaB on the DNA is required for DnaG to gain access to synthesize a primer. The participation of DnaB in priming reactions is discussed in more detail in a later section.

DnaC.

DnaC is a 27.9-kDa monomer (57). Cells mutated in dnaC display, depending on the allele, either the immediate-stop DNA replication phenotype (135), indicative of involvement in nascent chain elongation, or a slow-stop phenotype (136): i.e., when raised to the nonpermissive temperature, DNA synthesis does not cease until a new initiation event is required. The latter phenotype is generally regarded as indicating that the protein is required for initiation of DNA replication from oriC, although it cannot be formally distinguished from a requirement for the protein during termination of DNA replication.

DnaC activity appears to be expressed solely through its formation of a complex with DnaB and the subsequent delivery of DnaB either to the DNA directly or to a protein-DNA complex, such as in the assembly of the φX-type (137, 141) or ABC (85) primosome. DnaB-DnaC complex formation requires ATP (58, 132, 145). The ATP-binding site for complex formation is on DnaC (132). ATP hydrolysis is not required for complex formation; however, it is required to deliver DnaB to its target (132, 133). Thus, ATP hydrolysis appears to cause release of DnaB from the DnaB-DnaC complex. Just as the DnaB ATPase activity becomes inactivated upon DnaB-DnaC complex formation, DnaC activity, which is normally sensitive to inactivation by N-ethylmaleimide, becomes refractory (145).

The apparent chaperonin-like action of DnaC suggests that this protein would not be present at the replication fork, yet some alleles of dnaC show an immediate-stop phenotype. On the other hand, some evidence has emerged from studies of φX-type primosome (see below) assembly that DnaC is present in the final mobile protein-DNA complex. Interestingly, Allen and Kornberg (3) showed that strains carrying an elongation-defective dnaC mutant allele could not be complemented by the wild-type gene. The dominance of the mutant allele was interpreted as suggesting that the mutant DnaC protein remained complexed with DnaB at the replication fork, inhibiting elongation even in the presence of the wild type. In support of this, the authors demonstrated in vitro that excess DnaC reduced the rate of DnaB-catalyzed unwinding during the formation of form I* oriC plasmid DNA. (Form I* is highly unwound DNA that can be generated from oriC plasmid DNA by the combined action of DnaA, DnaB, DnaC, SSB, and DNA gyrase [12].) Thus, whereas the interpretation of the slow-stop dnaC phenotype is clear—DnaC must deliver DnaB to the DnaA-oriC complex to initiate local unwinding of the origin (13)—the participation of DnaC at the replication fork is somewhat problematic.

DnaG.

DnaG is the cellular primase (19), responsible for priming Okazaki fragment synthesis at the replication fork (136, 151). As would be expected of a protein required every 1 or 2 s during DNA replication, temperature-sensitive dnaG mutants show an immediate-stop phenotype (136). Interestingly, some dnaG mutations, which are located in the extreme COOH terminus of the protein, also display a par phenotype (49). The basis for this is yet to be understood.

DnaG is active as a 65.6-kDa monomer. Primase activity is supported by both NTPs and dNTPs, although either ATP or ADP is required to start primer synthesis in vitro (92, 140). DnaG is essentially inactive when presented with protein-free ssDNA. Recent studies have shown that at very high concentrations (1 to 2 μM), primase can synthesize primers on oligonucleotides (127) or phage DNAs (128) in the absence of other proteins. At concentrations more typical of those found intracellularly (10 to 50 nM), DnaG-catalyzed primer synthesis requires the presence of DnaB on the DNA. Thus, in a reaction termed "general priming," DnaG will, in the presence of DnaB, synthesize primers on any protein-free ssDNA that can be used to prime subsequent DNA synthesis by the DNA Pol III HE (6, 8). General priming, however, is inhibited by the presence of SSB on the DNA. More complicated assembly pathways that require more proteins, such as that leading to formation of the φX-type primosome, are required to introduce DnaB onto SSB-coated ssDNA so that it can serve as a target for DnaG.

There are, then, three aspects to the development of a site on the cellular template capable of supporting primer synthesis by DnaG. The first is the formation of a target for DnaG. The specific activity of DnaG priming activity on protein-free DNA is 300-fold higher in the presence of DnaB than in its absence (128). This can be thought of as either a measure of an increase in the affinity of DnaG for the DNA or as an indication of a required activation step to effect efficient priming. Either way, it ensures that priming in the cell will occur only where it is needed, i.e., at the replication fork. As described in a later section, DnaG is attracted to the replication fork by a protein-protein interaction with DnaB (128).

The second aspect is the generation of an SSB-free region of DNA. At the replication fork, this site is presumably lagging-strand template DNA as it exits DnaB just subsequent to unwinding of the parental duplex. If DnaG is not present, this DNA becomes coated with SSB. If DnaG is present, it seems likely that the enzyme scans the emerging ssDNA template for a suitable priming site as it emerges from DnaB. There is no evidence that DnaG can displace SSB from ssDNA. Perhaps in the DnaB-DnaG complex at the replication fork, the emerging ssDNA template passes directly to DnaG.

The third aspect required for primer synthesis is the interaction of primase with a recognition signal on the lagging-strand template. T. Okazaki and her colleagues have studied the nucleotide sequence about the 5' ends of Okazaki fragments formed in vivo (155, 156) and concluded that (i) on average the RNA primer is 10 to 12 nt long and (ii) primase requires the recognition signal 3'-GTC-5' in the template strand, with primer synthesis starting opposite the T. This is supported by studies of the nucleotide sequences required for DnaG-catalyzed primer synthesis at the bacteriophage G4 origin on intact phage DNA (51) and on short oligonucleotide templates (127). Thus, it is likely that once primase has been introduced to the DNA via its interaction with DnaB, it scans the nucleotide sequence until the next 3'-GTC-5' appears and then catalyzes primer synthesis. Because this sequence appears, on average, once every 64 nt, primer synthesis should occur within 0.1 s of the introduction of DnaG to the replication fork. This is well within the required time frame of the synthesis of an Okazaki fragment every 1 to 2 s.

DnaT.

DnaT (84) (formerly protein i [11] or replication factor X [142]) is a 19.5-kDa polypeptide that is active as a trimer (11). DnaT is required for assembly of the φX-type primosome (see below), where it is apparently involved in the transfer of DnaB from the DnaB-DnaC complex in solution to a PriA-PriB-DNA complex (9, 143). Cells mutated in dnaT are temperature sensitive for both growth and DNA synthesis and are defective in their ability to induce stable DNA replication (65).

Stable DNA replication is a RecA-dependent pathway (66) that can be induced as part of the SOS response or as a result of certain metabolic insults. It does not require protein synthesis and therefore allows replication to continue indefinitely. Stable DNA replication can be induced, for example, by a shift-up from minimal to rich medium or by exposure of the culture to nalidixic acid. Initiation of replication is presumably the result of a RecA-catalyzed strand invasion. In addition to recA and dnaT, stable DNA replication requires dnaC (66) and recBCD (80). The replication phenotypes of dnaT mutants lend support to the presence of the φX-type primosome at the replication fork.

PriA.

PriA (70, 99) (formerly protein n' [113] or replication factor Y [144]) is an 81.7-kDa polypeptide that is active as a monomer (113). It is essential for assembly of the φX-type primosome (10, 114). Cells mutated in priA grow very slowly and filament extensively (69, 100). They are also constitutively induced for the SOS response (100). The enzyme is multifunctional. It is an unusual ssDNA-dependent ATPase (or dATPase). Any protein-free ssDNA will support this activity, although it was shown that the ssDNAs of bacteriophages that required PriA for the complementary strand synthesis phase of their life cycle (e.g., φX174) were six- to eightfold better effectors than phage DNAs that did not require PriA (e.g., f1) (144). When these DNAs were coated with SSB, this difference increased to 100-fold (115).

This difference in ATPase effector activity could be attributed to a specific DNA region, originally isolated from φX174 ssDNA (114), of about 70 nt in length. These sites, now called primosome assembly sites (PAS) (82), have also been found on the ColE1 (98, 161), pBR322 (98), and F (52) plasmids, as well as on other DNAs. PAS sequences display little overall DNA sequence homology, although a consensus sequence has been proposed (86), suggesting that PriA binding is dictated by a particular secondary structure (47). Binding of PriA to a PAS is the first step in assembly of the φX-type primosome on DNA (see below).

PriA is also a 3'→5' DNA helicase (67, 71). This activity requires ATP (or dATP) hydrolysis and is stimulated 10- to 15-fold by the presence of a PAS sequence. No single-stranded tail is required on the fragment to be displaced, and, unlike many other DNA helicases, PriA helicase activity is not inhibited by the presence of SSB on the DNA. Because the point of entry of PriA to the DNA (the PAS) is known precisely, it could be shown that PriA translocated along SSB-coated ssDNA at 90 nt/s from the PAS to the fragment to be unwound and that translocation required ATP hydrolysis (73). To date, this is the only helicase that has been shown to translocate along ssDNA in an ATP-dependent manner. This raised the possibility that it is the DNA translocation function of PriA that plays a role in DNA replication. Such a model is discussed below; however, under normal circumstances it seems that only the primosome assembly function of PriA is required in the cell. Induction of the SOS response in priA null mutants could be suppressed by a gene encoding a mutant PriA protein that was no longer an ATPase or DNA helicase, but that still could function to assemble a primosome (158).

PriB and PriC.

PriB (4, 157) (formerly protein n [79]), an 11.4-kDa polypeptide active as a dimer, and PriC (157) (formerly protein n'' [79]), a 20.3-kDa monomer, are both involved in φX-type primosome assembly. PriB appears to be the second protein to add to the initial PriA-PAS-DNA complex (79). The role of PriC has yet to be elucidated. Less attention has been paid to these proteins than to their partner PriA, and thus there are no genetic data to consider. Their putative involvement in E. coli DNA replication is, at the moment, an inference based on their biochemical properties in replication systems reconstituted in vitro with purified replication proteins. On the other hand, current evidence suggests that PriB is a component of the functional φX-type primosome on the DNA (5, 72).

MECHANISMS AT THE REPLICATION FORK

The Replication Fork Primosome

Using replication forks reconstituted with the DNA Pol III HE and the φX-type primosomal proteins in a rolling-circle-type replication system in vitro, Wu et al. (151) showed that the primase acted distributively during the cycle of Okazaki fragment synthesis. That is, primase did not remain bound continuously at the fork, synthesizing a primer when called for to initiate synthesis of a new Okazaki fragment. Rather, a primase molecule became associated with the fork to synthesize a primer for a new Okazaki fragment, and then it dissociated, to be replaced by another primase molecule from solution to initiate synthesis of the next Okazaki fragment. Tougu et al. (128) proved that primase was attracted to the replication fork via a protein-protein interaction with DnaB. Thus, it is clear that primase associates at least transiently with the fork. Using the same system as Wu et al. (151), Mok and Marians (95) showed that leading-strand synthesis was processive for at least 150 kb. During this time, the replication forks maintained a high rate of DNA unwinding (700 nt/s at 30°C). Thus, it seems likely that once assembled into a replication fork with the HE, DnaB remained associated with the fork for its lifetime. Functionally, DnaB and DnaG account for all the activities required of a helicase/primase at the replication fork. Furthermore, replication forks formed using only the HE and DnaB, DnaC, and DnaG seemed indistinguishable in their properties from those formed using all the φX-type primosomal proteins (151). In addition, reconstitution of replication forks formed at oriC on plasmid DNAs in vitro required only DnaB, DnaC, and DnaG (55). Whither then PriA, PriB, PriC, and DnaT?

The φX-type primosome assembles in discrete steps at a PAS on SSB-coated ssDNA. PriA specifically recognizes and binds to the PAS (114). This activates its ssDNA-dependent ATPase (114, 144), although PriA-catalyzed ATP hydrolysis is not required for primosome assembly (158). PriB binds to the PriA-DNA complex (5, 79). DnaT then acts to transfer DnaB from a DnaB-DnaC complex in solution to the PriA-PriB-DNA complex, forming the preprimosome. The role of PriC in this process is unclear, although it has been suggested to act as a specificity factor, suppressing nonspecific complex formation (5). Addition of DnaG to the preprimosome completes the assembly of the φX-type primosome (137, 141).

Lee and Marians (72) demonstrated that the φX-type preprimosome could act as a DNA helicase in both directions along the DNA, driven in the 5'→3' direction by DnaB and in the 3'→5' direction by PriA. Furthermore, it could be demonstrated that the complete primosome could synthesize primers downstream from and on either side of a PAS in a linear DNA template. In addition, a PriA-PriB-DnaT complex isolated on a linear DNA that could act as a helicase only in the 5'→3' direction could be converted, upon the addition of DnaB and DnaC to complete preprimosome assembly, to a protein complex that maintained its 3'→5' activity and now manifested an equivalent amount of 5'→3' helicase activity. This suggested that both PriA and DnaB were present in the same protein-DNA complex. The authors proposed that a preprimosome consisted of a complex of PriA, PriB, and DnaB on a DNA. A similar conclusion was reached by Allen and Kornberg (5) based on a determination of which primosomal proteins were required for reactivation of a DNA-primosome complex that had been allowed to decay overnight on ice.

These studies therefore suggest that replication forks formed with all the φX-type primosomal proteins are likely to contain PriA. What might its function be? It has been suggested (72) that the presence on the same ssDNA of two DNA helicases with opposite directionalities may be a mechanism for efficient formation and processing of a loop in the lagging-strand template. In this sense, the DNA translocation activities of the two helicases would drive the template DNA through the protein-DNA complex, rather than the latter moving along the former. This proposal was consistent with models proposed by Alberts and his colleagues (1, 119) suggesting that the lagging-strand template was looped out around the lagging-strand polymerase. However, to date, no evidence exists from any system supporting the existence of this loop.

On the other hand, participation of DnaT and PriA in E. coli replication is indicated by the phenotypes of mutations in their genes (65, 69, 100). Both Lee and Kornberg (69) and Nurse et al. (100) have exploited the recent cloning of priA (70, 99) to construct E. coli strains deficient in PriA activity. Although such mutants could be isolated, cell viability was reduced 10- to 100-fold and the cells exhibited extreme filamentation (69, 100). Nurse et al. (100) noted that the SOS response had been induced and that filamentation could be suppressed in a sulA (sfiA, encoding a SOS-inducible inhibitor of filamentation [46]) background. Interestingly, in the priA sulA double mutant, both ColE1-type plasmids (which require the φX-type primosome for replication) and oriC-based plasmids (which do not require the φX-type primosome) could be maintained (100), whereas in the priA single mutant they could not (69).

Nurse et al. (100) proposed, to account for SOS induction in the priA strain, that whereas replication fork assembly was not dependent on the φX-type primosome pathway at oriC, completion of the chromosome in a significant fraction of cells could depend on subsequent assembly of φX-type primosome-dependent forks if the oriC forks were to stall or dissociate. This would account for the observed SOS induction whether PriA was an actual component of the φX-type primosome on the DNA or only required for its assembly. In support of PriA function being confined only to assembly of a φX-type primosome, Zavitz and Marians (158) have shown that the gene encoding an ATPase-defective mutant PriA that is not a helicase but still supports φX-type primosome assembly in vitro will, when provided in trans, suppress the filamentation phenotype of and restore cell viability to a priA strain. Allen et al. (2) have recently proposed a similar model. Thus, for now, the role of PriA, PriB, PriC, and DnaT in replication fork propagation remains problematic.

Protein-Protein Interactions at the Fork

DnaB and DnaG.

Priming of an Okazaki fragment is repeated every 1 to 2 s at least 1,000 times for each replication fork formed in E. coli. Okazaki fragment synthesis is therefore cyclical. How does the primase know when to synthesize a new primer? How does the primase find its priming site at the fork? Does the primase remain associated continuously with the fork? These questions will be discussed in the following sections.

The action of replication forks has been analyzed in the bacteriophage T4 and T7 systems and in E. coli (83). Rolling-circle synthesis with forks reconstituted with purified replication proteins is the system of choice because long duplex tails are made composed of a continuous leading strand and multiple Okazaki fragments. In the E. coli system (95, 151), the tails can be >0.5 Mbp in length; thus, because the average Okazaki fragment is 2 kb long, more than 250 cycles of Okazaki fragment synthesis can be observed over roughly a 10-min incubation time.

Using this experimental window, investigators can ask whether enzymes at the fork act processively—i.e., do they remain bound continuously through multiple cycles of Okazaki fragment synthesis?—or distributively—i.e., do they act once during a cycle of Okazaki fragment synthesis, to be replaced by a new molecule from the pool in solution for the next cycle? This is generally examined by diluting active replication forks into reaction mixtures that contain no additional template but where the concentrations of all the enzyme components except the one under investigation have been kept constant. After dilution, Okazaki fragment size will not change if the enzyme acts processively, whereas the fragments will increase in size if the enzyme acts distributively.

Using this type of dilution experiment, Wu et al. (151) demonstrated that primase acted distributively at the replication fork. This was also manifested as a decrease in the average size of Okazaki fragments as the primase concentration was increased in the reaction. The distributive action of primase was consistent with the failure of investigators to isolate an intact φX-type primosome with DnaG stably bound. How, then, is primase targeted to the replication fork?

It is now clear that DnaG is targeted to its priming site via a protein-protein interaction with a DnaB molecule already on the DNA. Tougu et al. (128) showed that partial trypsinolysis generated an NH₂-terminal 49-kDa primase fragment (p49) that was still an active primase both at the specialized bacteriophage G4 origin and on protein-free DNA at high concentration, but that was incapable of interacting with DnaB to synthesize a primer either during general priming, during φX-type primosome-catalyzed priming, or at the replication fork. The COOH-terminal 16-kDa fragment (p16), which neither bound DNA nor displayed any independent activity, inhibited DnaB- and DnaG-dependent replication reactions and, by direct competition, interfered with the ability of DnaG to interact with the replication fork. This was manifested, at a fixed concentration of DnaG, as an increase in the size of the Okazaki fragments as the concentration of p16 in the reaction was increased.

Thus, p16 defines the domain of primase required for functional interaction with DnaB. This interaction, however, is weak and transient, because a stable interaction between DnaG and DnaB in solution has yet to be detected. As described in a subsequent section, the interaction between primase and DnaB is crucial to the regulation of Okazaki fragment synthesis.

DnaB and the Leading-Strand Polymerase.

The E. coli replication fork moves at 1,000 nt/s at 37°C. In isolated strand displacement reactions, DnaB unwinds DNA at about 50 nt/s (68). Replication forks reconstituted in vitro with the rolling-circle system move at 700 nt/s at 30°C (95). Nucleotide polymerization by the DNA Pol III HE has been measured variously as 200 to 500 nt/s (88). Thus, at the replication fork, either DnaB is pushed along by the leading-strand polymerase, or its DNA unwinding rate is somehow increased. Either case suggests that there is protein-protein contact between the leading-strand polymerase and DnaB.

Additional support for this comes from in vitro studies using reconstituted replication forks, demonstrating that SSB was required only for lagging-strand synthesis (95). Leading-strand synthesis could proceed at high rates in the absence of SSB. This indicates that it is likely that the leading-strand polymerase is in close contact with DnaB. If there were sufficient room between the polymerase and the helicase so that hairpins could form in the unwound template, the rate of replication fork movement would have been reduced because of the inability of the DNA Pol III HE to pass these structures.

An observation relevant to this issue has been made with the oriC replication system. Hiasa and Marians (50) found that during oriC plasmid replication reconstituted with purified replication proteins, the primase concentration affected the assembly of the replication forks. At low primase concentrations, DnaB appeared to leave the origin and unwind the plasmid DNA before complete replication forks could form. At high primase concentrations, two replication forks formed at oriC and proceeded bidirectionally around the plasmid template, simultaneously synthesizing leading and lagging strands. The authors suggested that because primase was the only priming enzyme available, high concentrations were required to ensure that the HE, by binding to the primer, was situated close enough to DnaB to establish the protein-protein contact required to capture DnaB and establish a replication fork.

DnaG and the HE.

Zechner et al. (160) found that the RNA primers used to initiate the synthesis of Okazaki fragments in the rolling-circle system in vitro were invariably 10 to 12 nt in length. This was consistent with findings of Okazaki and her colleagues on the size of the RNA primers made in vivo (56). In contrast, primers made by the φX-type primosome on φX174 ssDNA in the absence of DNA synthesis vary from 8 to 70 nt in length. What, then, limits primer size at the replication fork?

Zechner et al. (160) demonstrated that primers synthesized by primase during either general priming, priming by the φX-type primosome, or priming at the bacteriophage G4 complementary strand origin of replication became reduced in size in the presence of the HE in the complete absence of deoxynucleotide polymerization. Under these conditions, primers synthesized by the φX-type primosome were limited to 10 to 12 nt in length.

Surprisingly, this effect was not a result of preinitiation complex formation on the primer terminus, because it could not be reproduced in the presence of only β and the γ complex. The core was the only subassembly of the HE that showed any isolated effect, but this was not as pronounced as that observed with the HE. Finally, the limitation on primer length was specific to the HE and could not be duplicated with either the T7 DNA polymerase, DNA polymerase I, or the T4 DNA polymerase holoenzyme.

Zechner et al. (160) proposed that at the replication fork, primer length limitation occurs because of the sequential interaction between primase (which appears to possess an intrinsic pause in polymerization at a primer length of 10 to 12 nt [160]) and the core synthesizing the lagging strand, followed immediately by preinitiation complex formation on the primer terminus. This series of interactions serves two purposes: (i) it acts to limit primer synthesis, and (ii) it acts as a signal to the lagging-strand polymerase to terminate Okazaki fragment synthesis (discussed in a subsequent section).

The Leading- and Lagging-Strand Polymerases.

Alberts and his colleagues (1, 119) noted early on that during the multiple cycles of Okazaki fragment synthesis at the fork, the lagging-strand polymerase must be released from the template and the nascent DNA after the completion of an Okazaki fragment. Synthesis of the next Okazaki fragment could therefore be accomplished by the same polymerase or a new one from the pool in solution. To maximize the speed and efficiency of targeting a polymerase to the new primer, Sinha et al. (119) proposed that the same polymerase could be targeted very efficiently if it remained bound at the fork through protein-protein interactions with the leading-strand polymerase while it transited from the completed Okazaki fragment to the new primer.

This dimeric polymerase hypothesis was modified to account for the multitude of HE subunits and the apparent requirement for differential processivities on the leading and lagging strand. McHenry and Johanson (90) first proposed, based on studies of initiation complex formation and dissociation, that the holoenzyme would be an asymmetric dimer, with a different assortment of subunits on the leading- and lagging-strand polymerases. Based on determination of the stoichiometries of subunits in isolated HE complexes, Maki et al. (81) made a similar proposal.

The driving force behind the asymmetric dimer hypothesis was that a differential assortment of subunits might generate two functionally distinct polymerases, one capable of almost unlimited processivity—the leading-strand polymerase—and one capable of only a moderate processivity—the lagging-strand polymerase. Recent studies show clearly that whereas the isolated HE may exist as an asymmetric dimer in solution, the mechanism that limits the processivity of the lagging-strand polymerase at the replication fork has nothing to do with a differential assortment of HE subunits (see next section).

Are the polymerases synthesizing the leading and lagging strands dimerized at the replication fork? Certainly, the solution structure of the HE would suggest that they were. However, this is indirect. In the only experiment reported that addresses this question directly, Marians and his colleagues (150) showed that in the rolling-circle replication system, Okazaki fragment size was maintained after dilution (by 60-fold) of the core. Formally, this indicates that the lagging-strand polymerase acts processively during multiple cycles of Okazaki fragment synthesis; i.e., the same core assembly synthesizes each Okazaki fragment. Because α must dissociate from one primer template (the just-completed Okazaki fragment) and move to another primer template (the newly synthesized primer), this suggests that the lagging-strand core is held at the fork by protein-protein interactions. It is likely this interaction is with the leading-strand polymerase, thus realizing the dimeric holoenzyme hypothesis. However, it cannot be ruled out that the lagging-strand polymerase is interacting with DnaB or some other primosomal protein at the fork.

Of course, the fact that the same lagging-strand polymerase synthesizes each and every Okazaki fragment at the fork increases the complexity of the mechanisms required to ensure efficient replication fork action. This adds another step to the cycle, one in which the lagging-strand polymerase must dissociate from the completed Okazaki fragment and transit to the new primer. This step has sometimes been called travel time or polymerase cycling. This step is intertwined with the mechanisms that regulate Okazaki fragment synthesis and is discussed in the next section.

Regulation of Okazaki Fragment Synthesis

The requirements for cycling from completed Okazaki fragment to new primer terminus have been studied in a model system by O’Donnell and his colleagues (101, 102, 122, 123). Two primed ssDNA circles are used that can be distinguished by size. An initiation complex is formed on one template. Replication is allowed to proceed, and the time required for the core to leave the first template (the donor) and find and replicate the second primed template (the acceptor) is measured. Because the time required to actually replicate either template is known, the time required for the core to cycle from the donor to the acceptor template can be derived from the kinetics of the reaction.

As shown previously (22), cycling to acceptor was very slow, on the order of minutes. However, if a preinitiation complex was first formed on the acceptor, cycling time could be reduced to the order of a few seconds (101). How fast must polymerase cycling be to fit into the workings of the replication fork? This is difficult to estimate because no measure exists comparing the duration of a cycle of Okazaki fragment synthesis, i.e., the time from one new primer synthesis event to another, to the actual time required for nascent strand synthesis.

In the worst case, polymerase cycling would have to be very fast, on the order of a few milliseconds. This would be the case if nascent strand synthesis and the preliminary events required for activating a primer (binding of primase to the replication fork, primer synthesis, termination of primer synthesis, and preinitiation complex formation) occurred sequentially within the Okazaki fragment cycle. Then, because the leading and lagging strands appear to be made simultaneously, with the difference in length between them corresponding only to the length of one Okazaki fragment, the bulk of the lagging-strand cycle would have to consist of nascent strand synthesis. In the best case, the primer synthesis and activation events would overlap the period of nascent strand synthesis. Here, polymerase cycling would likely be on the order of a few hundred milliseconds.

Polymerase cycling time in O’Donnell’s system is inversely related to the concentration of the activated acceptor (123). Extrapolation of this curve intersects the origin, suggesting that cycling times on the order of a few hundred milliseconds at an infinite concentration of activated acceptor are possible. At the replication fork, the donor and acceptor are presumably in very close proximity, possibly approximating the solution extrapolation. If this is the case, then known biochemical mechanisms can account for the rapidity of polymerase cycling.

How is the cycle of Okazaki fragment synthesis regulated? What is the signal for synthesis of a new primer, and how is it generated? Answers to some of these questions have emerged from the studies of Marians and his colleagues (150, 151, 152, 159, 160) on reconstituted replication forks using the rolling-circle system.

Reasoning that perturbation of the cycle of Okazaki fragment synthesis would result in alteration of the product, i.e., in the size of the Okazaki fragments, these authors identified reaction parameters and enzymes that, when varied, resulted in a change in the size of the Okazaki fragments produced. Analysis of the nature of these changes has led to a model for how Okazaki fragment synthesis is controlled.

As mentioned previously, it was determined that primase acted distributively during the Okazaki fragment cycle (151), whereas the lagging-strand core acted processively (150). It was also shown that β and at least some subunits of the γ complex acted distributively (150, 151). Thus, the preinitiation complex formed on each new primer terminus was composed of molecules recruited from solution. This suggested that the γ complex was not a permanent resident at the fork and that when the lagging-strand core transited from the just-completed Okazaki fragment to the new primer, it dissociated from the β molecule clamping it to the nascent 3' end and left it behind. Nevertheless, it was found that the number of Okazaki fragments synthesized during the reaction was in excess of the number of β molecules present, suggesting that the β molecules left behind when the lagging-strand core transited to the new primer were recycled eventually (151). The disposition of β during recycling of the HE from donor to acceptor template in O’Donnell’s system described earlier in this section has recently been shown to be identical to that described above (P. T. Stukenberg, J. Turner, and M. O’Donnell, personal communication).

Other factors that affected the size of the Okazaki fragments were the concentrations of the NTPs and dNTPs (151). These parameters, as well as all the others mentioned, were shown to affect different aspects of primer synthesis. Formally speaking, the size of an Okazaki fragment is determined by the distance on the lagging-strand template between two successful initiations. Thus, the rate with which template is generated will influence the size of the fragments. Wu et al. (151) demonstrated that the rate of movement of the E. coli fork remained constant when conditions were varied that resulted in dramatic changes in Okazaki fragment size. Further investigation showed that all factors that affected Okazaki fragment size did so by perturbing an event during the cycle of Okazaki fragment synthesis that occurred prior to actual nascent chain elongation. Variation in the concentration of NTPs or the distributively acting DnaG affected the frequency of primer synthesis, whereas variation in the concentration of dNTPs, the γ complex, or the β subunit of the DNA Pol III HE affected the efficiency of utilization of primers for the initiation of Okazaki fragment synthesis (159). Thus, it was the synthesis of a primer and its successful use for initiation of Okazaki fragment synthesis that governed Okazaki fragment size. What, then, is the signal for synthesis of a primer?

Under certain conditions, replication forks could be induced to synthesize aberrantly short Okazaki fragments that were separated on the lagging-strand template by large gaps (152). Nevertheless, under the same conditions, fragment size could be increased by a factor of 10 when the priming frequency was decreased (by decreasing the DnaG concentration). This indicated that the signal for synthesis of a new primer could not be the stalling of the lagging-strand polymerase when it encountered the penultimate Okazaki fragment; instead it most likely was the association of DnaG with the fork that acted to signal the lagging-strand polymerase to terminate Okazaki fragment synthesis, whether or not all the available template had been copied.

Wu et al. (152) therefore proposed that the cycle of Okazaki fragment synthesis was regulated by a clock-type mechanism (Fig. 2). The period of the clock was determined by the cycle of association and dissociation of DnaG with the replication fork. This model reflects a "best-case" situation for polymerase cycling where primer synthesis and activation are concurrent with nascent strand synthesis. (i) The cycle begins when DnaG associates with DnaB at the replication fork. Okazaki fragment synthesis is proceeding concurrently. (ii) Primer synthesis is initiated and (iii) an interaction between primase and the lagging-strand core occurs that serves to limit primer synthesis and (iv) triggers preinitiation complex formation (catalyzed by a new β and γ complex from solution) on the primer terminus. Concurrent with steps (iii) and (iv), Okazaki fragment synthesis terminates. (v) Under normal circumstances, no gaps are left between Okazaki fragments, and it is likely that preinitiation complex formation triggers transit of the lagging-strand core to the new primer terminus. Nascent strand synthesis then ensues.

Fig. 2Model for a cycle of Okazaki fragment synthesis at the E. coli replication fork. The arrow indicates the direction of fork movement. Parental DNA, heavy lines; nascent DNA, thin lines; primers, short black line segments; elongating Pol III core assemblies, large rectangles; ? subunit, small rectangles; primase, the circle; DnaB, the hexagon. Steps i through v are described in the text.

Source:

If, for some reason, step iv or v is delayed or takes longer, because the period of the cycle remains the same, the time available for nascent strand synthesis during the next cycle will be shortened. Under these circumstances, the primase–lagging-strand core interaction (step iii) triggers dissociation of the lagging-strand core–β assembly and transit of the polymerase even if all the available template has not been copied.

Finally, because lagging-strand template is generated continuously, increased priming frequency (step i) would result in smaller fragments, whereas decreases in the efficiency of primer utilization at either step iii, because of premature termination of primer synthesis, or step iv, because of bungled preinitiation complex formation, would result in larger fragments.

Proteins Involved in Sealing Okazaki Fragments

After the synthesis of Okazaki fragments occurs on the lagging-strand template, the RNA primers, which are 10 to 12 nt in length (56), must be removed, the gap between adjacent Okazaki fragments must be filled in, and the fragments must be joined together to yield a continuous strand.

Removal of the RNA primer is presumably effected by the combined action of RNase H1 (15) and the RNase H-like activity of the 5'→3' exonuclease of Pol I (76). It is not clear which polymerase fills in the gap between the Okazaki fragments. The polymerization function of Pol I is not required for E. coli viability; only the 5'→3' exonuclease function is required (32, 64, 104, 131). Thus, if Pol I is the enzyme that fills the gap, it can be substituted by either the HE or DNA polymerase II.

The requirement for the 5'→3' exonuclease results from a difference in its mode of action compared to RNase H1. The latter enzyme is an endonuclease and will not hydrolyze a phosphodiester bond formed between a ribonucleoside monophosphate and a deoxyribonucleoside monophosphate (15). The 5'→3' exonuclease of Pol I, on the other hand, will hydrolyze this bond (77). Thus, complete removal of the ribonucleotide primer at the 5'-end of Okazaki fragments requires the Pol I exonuclease. This is important because DNA ligase (42, 43, 105), the enzyme that seals the nick left after the primer gap is removed, will not seal an RNA-DNA junction (75).

As bespeaks their crucial contributions to sealing Okazaki fragments, mutations in DNA ligase (lig) or the 5'→3' exonuclease of Pol I (polAex) are temperature sensitive for growth and result in a massive accumulation of small DNA fragments at the nonpermissive temperature as the cells become nonviable (75, 77, 104, 131). A similar defect is not observed with cells mutated in RNase H1 (rnhA) (103).

ACKNOWLEDGMENTS

Studies from my laboratory were supported by NIH grants GM 34557 and GM 34558. I thank Mike O’Donnell for his critical reading of the manuscript.