|Home | About | Journals | Submit | Contact Us | Français|
Replisomes were originally thought to be multi-protein machines with a stabile and defined structure during replication. Discovery that replisomes repeatedly discard sliding clamps and assemble a new clamp to start each Okazaki fragment provided the first hint that the replisome structure changes during replication. Recent studies reveal that the replisome is more dynamic than ever thought possible. Replisomes can utilize many different polymerases; the helicase is regulated to travel at widely different speeds; leading and lagging strands need not always act in a coupled fashion with DNA loops; and the replication fork does not always exhibit semi-discontinuous replication. We review some of these findings here and discuss their implications for cell physiology as well as enzyme mechanism.
For more than a decade many labs, including our own, have promulgated a view of the replisome as an integrated machine at the replication fork with common features that contribute to rapid, efficient duplication of cellular genomes 1-4. Coupling of the leading- and lagging-strand polymerases to the replicative helicase provides a mechanism whereby both strands can be synthesized concurrently, with the leading strand copied continuously and the lagging strand discontinuously. In the coupled replisome, lagging strand synthesis occurs by a looping mechanism as the lagging strand polymerase hops from one clamp to another, all the while remaining in contact with the leading strand polymerase and the helicase 5, 6. In the context of the replication fork, the replicative helicase becomes highly processive, matching the speed and processivity of the polymerase and enabling replication forks to move at almost unimaginable speeds over long distances without interruption. All of these premises have been strongly validated in vitro, most recently by advances in single molecule approaches to studying replication forks, and collectively they allow us to understand how it is possible for cells to replicate their large genomes in a matter of minutes under ideal conditions.
At the same time as this tidy view of the replisome was being developed and validated, compelling evidence existed that the situation inside the cell was far more complicated and that this singular view of the replisome was incomplete. In particular, the first detailed studies of DNA synthesis in the cell made it clear that semi-discontinuous synthesis was probably the exception rather than the rule, as the replisome encounters numerous obstacles that must be overcome while traversing the DNA in vivo even under the best of circumstances 7. The essential nature of replication restart and post-replication repair indicated that forks frequently break down, or at least need help in traversing these obstacles. The discovery of numerous DNA polymerases in even the most rudimentary cell types, many of which use the same sliding clamp as the dedicated replicases, raised the question of whether one polymerase was in charge, and if not, what factors determined which polymerase would dominate the clamp in a given situation. These and other questions have led to a reconsideration of the unitary replisome, and recent advances in understanding the dynamic nature of the replication machinery are the subject of this review.
Chromosomes are duplicated by multi-protein replisome machines that display a high degree of coordination among their various “moving parts” 1-4. Most of what is known about replisome function stems from studies in E. coli and its phages, T4 and T7, the classic systems for genetic and biochemical studies of replisome action. The major components of a replisome include the helicase, which unwinds the parental duplex, two DNA polymerases that duplicate the leading and lagging strands concurrently, and a primase that initiates lagging strand Okazaki fragments. The organization of these components at the E. coli replication fork is illustrated in Figure 1.
The replicative helicases in these classic systems are ring shaped hexamers that encircle the lagging strand and use ATP to translocate along it to unwind the parental duplex. In all three systems, the leading strand DNA polymerase connects to the helicase to form a coupled polymerase-helicase for fast and coordinated DNA unwinding and synthesis. In E. coli, the polymerase-helicase connection is made through the τ subunit, which binds to DnaB helicase and to the heterotrimeric Pol III core polymerase (αεθ), thereby bridging these critical components 8. All cellular replisomes utilize a ring shaped sliding clamp that encircles DNA and tethers the polymerase to the template for highly processive synthesis. Sliding clamps are loaded onto DNA by a heteropentameric clamp loader machine that couples ATP hydrolysis to open and close the clamp around a primed site 9. The T7 phage polymerase (gp5) does not use a clamp, but instead recruits a host factor, thioredoxin, for processivity.
As the helicase unwinds the parental duplex, the ssDNA generated on the lagging strand is bound by single-strand DNA binding protein (SSB), which protects the ssDNA from nuclease attack. Each Okazaki fragment is initiated by an RNA primer synthesized by primase. E. coli primase (DnaG protein) synthesizes RNA primers of 10-12 nucleotides and must transiently interact with DnaB helicase for activity. The requirement of primase-helicase interaction for activity localizes RNA primers to the replication fork, and this generalizes to T4 (gp41/gp61). In T7, both activities are contained on the same polypeptide chain (T7 gp4).
As the replication fork advances, the lagging strand polymerase remains coupled to the replisome, resulting in a DNA loop during Okazaki fragment synthesis. The loop grows as an Okazaki fragment is extended, but upon completing the fragment the polymerase must release from the DNA to begin the next Okazaki fragment. This DNA looping mechanism is often referred to as the “trombone model” of replication since the DNA loop grows and dissolves repeatedly, like the motion of a trombone slide 6, 10.
The lagging strand polymerase is held tightly to DNA by a sliding clamp, yet it must rapidly dissociate from DNA upon completing each Okazaki fragment. This dilemma is resolved by a process that triggers the polymerase to sacrifice its connection to the clamp upon completing an Okazaki fragment, leaving the clamp behind on DNA. Polymerase release can be triggered upon completing a fragment (by collision with the previous fragment) or prior to completion 11, 12. New clamps are assembled onto RNA primed sites for use by the dissociated lagging strand polymerase. This mechanism of polymerase releasing from clamps and hopping to new clamps is operative in bacteria, T4 and eukaryotes 5, 13, 14.
All cells contain multiple DNA polymerases, most of which require the clamp for activity. For example, E. coli cells contain five distinct DNA polymerases and human cells have at least ten 15-18. The different polymerases serve specialized functions. The function of some specialized DNA polymerases is to bypass template lesions produced by DNA damaging agents 16, 19, 20. Template lesions block replisome progression because high fidelity chromosomal replicases cannot extend DNA past a damaged base. Specialized Y-family DNA polymerases lack a proofreading 3’-5’ exonuclease and are particularly adept at lesion bypass 17, 21. Lesion bypass enhances cell survival, even though the bypass process typically results in a mutation. E. coli contains two Y-family DNA polymerases, Pol IV and Pol V, both of which are induced upon DNA damage. Pol II is also induced by DNA damage, although it contains a 3’-5’ proofreading exonuclease and its cellular role is less clear.
A recent study shows that E. coli Pol II and Pol IV efficiently trade places with the rapid, high fidelity Pol III in a moving replisome 22. This finding is also consistent with studies of Pol IV switching with Pol III in simpler in vitro systems 16, 23, 24. The polymerase switch proceeds through an intermediate in which two polymerases bind to the same β clamp, as originally observed in the T4 system 25. At damage-induced levels of Pol II and/or Pol IV, the specialized polymerase takes over the replication fork from Pol III and functions with the DnaB helicase, as illustrated in Figure 4 22. This observation is unexpected since Pol II and Pol IV have no known interaction with DnaB. Furthermore, the Pol IV and Pol II “alternative replisomes” advance very slowly, 1 bp/s and 10 bp/s, respectively, yet the intrinsic rate of DnaB unwinding is reported to be 30 bp/s 22. These alternative replisomes are highly stabile and continue their slow advance for several kb over the course of 20-40 minutes, indicating that DnaB is exceedingly processive in these replisomes and that the rate of DnaB unwinding adapts to match the speed of synthesis. It is very important that the replicative helicase adjusts to the rate of synthesis. If the helicase were to unwind DNA faster than the polymerase can replicate it, large regions of ssDNA would be exposed which would be vulnerable to nucleases and could inappropriately trigger cellular responses to DNA damage (i.e. the SOS response).
The fact that different polymerases flow in and out of the replisome implies that the composition of the replisome changes when specialized DNA polymerases are induced upon DNA damage. Interestingly, both Pol II and Pol IV are present in normal cells, but at 7-10 fold lower levels than their induced concentration 26-28. The roles of Pol II and Pol IV during normal cell growth are largely unknown, but they almost certainly occupy the replisome some percentage of the time.
Alternative replisomes that move very slowly may be an advantage to the cell under some circumstances. After DNA damage, the highly efficient mechanisms of template-directed excision repair require the lesion to be in the context of duplex DNA. Encounter of the fork with a lesion causes it to be unwound from its template strand and places the damaged base in the context of ssDNA, where repair requires the alternative processes of mutagenic lesion bypass or high fidelity recombinational repair. Recombinational repair is especially cumbersome as it requires disassembly of the replisome and fork collapse, followed by strand pairing, repair and reassembly of the replisome. The rapid Pol III based replisome has a much greater chance than a slow replisome of encountering a lesion before it can be repaired. Therefore, slow alternative replisomes may act as a molecular brake during conditions of DNA damage, providing extra time for lesions to be repaired while they are still within duplex DNA 22, 29.
Expression of Pol II and Pol IV in living cells slows DNA synthesis even in the absence of DNA damage 22, 29. This observation is consistent with the in vitro studies on slow alternative replisomes containing Pol II and/or Pol IV, and suggests that deliberate slowing of the fork may be one reason Pol II and Pol IV are upregulated early in the E. coli DNA damage response (SOS). It seems a contradiction to use a low fidelity enzyme like Pol IV to prevent mutations (i.e. by slowing the fork). However, a recent study indicates that UmuD binds to Pol IV during the early stages of the SOS response in a way that modulates its mutagenic effect 30.
There are many different types of helicases, and one may propose to assort them into two broad categories, active and passive, by the way they harness the energy of ATP 31-34. A helicase that unwinds DNA by an active mechanism couples ATP directly to the destabilization of base pairs. In contrast, passive unwinding takes advantage of spontaneous thermal melting of the duplex and the helicase simply uses ATP to translocate along one strand of ssDNA. In either case, helicases are presumed to function with other factors that prevent the unwound strands from reannealing 35. For replicative helicases, reannealing is prevented by the leading strand polymerase, which converts one single-strand into duplex DNA.
Ring shaped replicative helicases encircle one strand of ssDNA and translocate along it, and they melt DNA by excluding the other strand from the central channel 35, 36. Their rate of unwinding is slower than ssDNA translocation because translocation through dsDNA requires either force (active) or a wait time for thermal fraying (passive). The ring shape of hexameric replicative helicases suggests they do not bind the forked junction for active destabilization of base pairs, and thus may act by a passive mechanism. Hexameric helicases are also very slow to unwind DNA by themselves, consistent with a passive mechanism of unwinding. However, hexameric helicases become highly efficient when coupled to the replicative DNA polymerase. For example, coupling of E. coli DnaB to the Pol III replicase increases its rate of unwinding from 30 bp/s to over 500 bp/s 8. This rapid rate suggests an active component to the unwinding mechanism, and a recent study lends strong support to this view 37. It is suggested as one possibility that a ring shaped helicase may actively unwind DNA by bending the excluded single-strand on the outside perimeter of the ring as the ring pushes forward over the forked junction. Inability to accommodate the dsDNA as well as the folded back excluded strand in the center of the ring may lead to base pair destabilization, as illustrated in Fig. 2.
Alternative replisomes that contain Pol IV and Pol II move 3-30 fold slower than the intrinsic rate of DnaB unwinding 8, 22. At first glance, one may presume that Pol II and Pol IV bind to DnaB and slow it down to match the rate of the polymerase. However, neither Pol II nor Pol IV is known to bind to DnaB. Indeed, heterologous DNA polymerases have been shown to increase the speed of unwinding of the T4 helicase, indicating that helicase rate is regulated in a non-specific fashion 38.
We would like to propose here a possible mechanism by which specialized polymerases regulate helicase speed in a non-specific fashion, illustrated in Figure 3A. Assuming that the outside perimeter of a hexameric helicase interacts with the excluded single-strand to exert force for active unwinding, the DNA polymerase may disrupt this connection simply by converting the ssDNA to dsDNA (Fig. 3A). This would remove the active component of the helicase mechanism, constraining it to unwind DNA solely by the passive component. This hypothesis also suggests a role for direct interaction between the replicative polymerase and helicase. Specifically, the leading strand polymerase-helicase connection may enforce a spatial separation between the polymerase and helicase, thereby ensuring that the outside perimeter of the helicase can interact with ssDNA for active unwinding (illustrated in Figure 3B). This and numerous other possibilities await more detailed investigation.
The trombone model of DNA looping during lagging strand synthesis requires that the lagging strand polymerase remains associated with the replisome continuously during fork progression. This connection is maintained through interaction of the lagging strand polymerase with the helicase and/or the leading strand polymerase. In E. coli these connections are made through the τ subunit of the clamp loader, enabling the lagging strand polymerase to remain with the replisome while it recycles from one Okazaki fragment to the next. However, neither Pol II nor Pol IV is known to dimerize or to bind other replisome components (e.g. helicase, or clamp loader), implying that the lagging strand is filled in by independent DNA polymerase molecules acting from solution. In fact, the speed of alternative replisomes depends on the concentration of Pol II or Pol IV added to the reaction, implying that polymerase action is distributive on the leading strand as well 22. Therefore, alternative replisomes that employ specialized DNA polymerases appear to function in an uncoupled manner as illustrated in Fig. 4.
One detail of these alternative replisomes that remains unresolved is the disposition of the clamp loader after dissociation of Pol III. In principle the clamp loader could remain at the fork by continued interaction with DnaB. Alternatively, its association with the fork could be lost completely and clamps could be loaded as needed by clamp loaders recruited from solution. In this regard, it is interesting to note that E. coli produces two types of clamp loaders. The dnaX gene encodes two proteins in similar amounts, τ and γ; τ is the full-length gene product while γ is truncated by a translational frameshift. The γ subunit is composed of the N-terminal 2/3 of τ and retains the clamp loading domains but lacks the C-terminal Pol III and DnaB binding domains of τ 39. Both τ and γ are homooligomers and do not tend to form mixed γ/τ heterooligomers 40. Therefore, even when τ and γ are present together, the addition of δ, δ’, χ and ψ results in a mixture of only two clamp loaders, τ3δδ’χψ and γ3δδ’χψ, rather than clamp loaders that contain both τ and γ 40. The γ3δδ’χψ clamp loader can be isolated from E. coli cells indicating that it is a physiologically relevant form, and we propose that one of the functions of the γ3δδ’χψ clamp loader is to load clamps onto primed sites for use by alternative replisomes.
In the semi-discontinuous model of DNA replication, the leading strand is extended in a continuous fashion while the lagging strand contains discontinuities. Most of the support for this model comes from in vitro studies, and it has become widely accepted despite the fact that the earliest in vivo analyses of DNA replication provided little support for such a model 7. Recent cellular investigations support the conclusion that synthesis is discontinuous on both strands 7, 41, 42. Consistent with this, biochemical studies demonstrate that primase can prime the leading strand at a replication fork, not just the lagging strand 43. Leading strand priming is important when the leading polymerase is blocked by a template lesion. Formation of a primer on the leading strand ahead of the block site provides an escape route for the stalled leading polymerase. Specifically, the clamp loader can attach a clamp to the new RNA primer for the stalled leading polymerase to hop to, thereby enabling the replisome to continue on its way. In this fashion, the template lesion is simply left in the wake of the fork where it can be repaired later by high fidelity recombinational repair 44.
Yet another source of discontinuities on the leading strand is revealed by recent studies of replisome collisions with an in-line RNA polymerase 45. Cellular analyses indicate that head-on collisions of the replisome with RNA polymerase block replication while in-line collisions do not 46, 47. Furthermore, in-line promoters outnumber head-on promoters in bacterial genomes, suggesting an advantage to an in-line versus head-on orientation of gene transcription 48. In either case, RNA polymerase must be displaced or it will block replisome progression. Furthermore, encounters between the replisome and RNA polymerase are a near certainty given the fact that RNA polymerase is 10-20 fold slower than the replisome. The high density of promoters in bacterial DNA indicates that collisions of this sort are frequent.
A recent study of replisome encounter with an RNA polymerase transcribing the leading strand (i.e. inline collision) shows a remarkable outcome 45. Specifically, the RNA polymerase is displaced but the mRNA transcript is not. Instead the mRNA is recruited by the replisome and is extended by the leading strand polymerase (Fig. 5). At most, the process results in a pause in replication and is consistent with cellular studies indicating that in-line collisions with RNA polymerase do not block replication forks 46, 47. The mechanism by which the mRNA is recruited is proposed to involve similar events to those that occur during lagging strand replication, in which a clamp is assembled onto the RNA and the leading strand polymerase transfers to the new clamp, leaving the old clamp behind (Fig. 5). These actions create a discontinuity in the leading strand. Given the density of promoters in bacterial genomes, these in-line collisions may contribute significantly to the source of discontinuous replication on the leading strand.
So far, the only complete replisomes that have been studied in biochemical detail are bacterial or bacteriophage systems that must do their job very quickly and efficiently in order to replicate their compact genomes from a single origin in a matter of minutes. Eukaryotes, with much larger genomes, have solved the same problem by initiating replication from many origins spread throughout their chromosomes, enabling some cells to replicate genomes larger than 100 Megabases in as little as five minutes using DNA polymerases (Pol δ and Pol ε) that are much slower than those of E. coli or its phages. So is coupling of leading and lagging synthesis to DNA unwinding for rapid replication necessary for all cell types, or is the coupled replisome paradigm simply reflective of its limited origins?
The only biochemical investigations of eukaryotic replications forks made use of a viral helicase, the SV-40 T antigen 49-51. Since then, a great deal of evidence has accumulated indicating that the Mcm2-7 complex is the replicative helicase in eukaryotes, but to date a cellular eukaryotic replication fork has not been reconstituted 52, 53. Thus, it is still unknown how the eukaryotic fork works, in particular whether other proteins like GINS and Cdc45 may activate the Mcm helicase or couple leading and lagging strand synthesis at the fork. Study of bacterial replisomes suggests that both of these functions may be unnecessary, and that perhaps the Mcm complex simply requires a replicative polymerase to upregulate its motor, as shown for the T4, T7, and DnaB helicases. Another intriguing question about the eukaryotic fork emerges from the recent finding that Pol ε is the leading strand polymerase in yeast 54, 55. If this is the case, it remains to be determined how Pol δ is excluded from the leading strand given its high processivity and tight binding to PCNA. If there is such a mechanism, either enforced by Pol ε itself or by other cellular factors, it should prove to be incredibly fascinating.
The authors are grateful to funding from the NIH (GM38839). LL is a recipient of a Ruth L. Kirschstein National Research Service Award from the NIH (T32CA009673).