Chapter 13:  Genome Replication

The Meselson-Stahl experiment

Delbrück's breakage-and-reunion model was important because it stimulated experiments designed to test between the three modes of DNA replication illustrated in Figure 13.2. Radioactive isotopes had recently been introduced into molecular biology so attempts were made to use DNA labeling (Technical Note 4.1) to distinguish newly synthesized DNA from the parental polynucleotides. Each mode of replication predicts a different distribution of newly synthesized DNA, and hence of radioactive label, in the double helices resulting after two or more rounds of replication. Analysis of the radioactive contents of these molecules should therefore determine which replication scheme operates in living cells. Unfortunately, it proved impossible to obtain a clearcut result, largely because of the difficulty in measuring the precise amount of radioactivity in the DNA molecules, the analysis being complicated by the rapid decay of the ³²P isotope that was used as the label.

The breakthrough was eventually made by Matthew Meselson and Franklin Stahl who, in 1958, carried out the required experiment not with a radioactive label but with ¹⁵N, the non-radioactive ‘heavy’ isotope of nitrogen. Now it was possible to analyze the replicated double helices by density gradient centrifugation (Technical Note 2.2), because a DNA molecule labeled with ¹⁵N has a higher buoyant density than an unlabeled molecule. Meselson and Stahl (1958) started with a culture of E. coli cells that had been grown with ¹⁵NH₄Cl and whose DNA molecules therefore contained heavy nitrogen. The cells were transferred to normal medium, and samples taken after 20 minutes and 40 minutes, corresponding to one and two cell divisions, respectively. DNA was extracted from each sample and the molecules examined by density gradient centrifugation (Figure 13.3A). After one round of DNA replication, the daughter molecules synthesized in the presence of normal nitrogen formed a single band in the density gradient, indicating that each double helix was made up of equal amounts of newly synthesized and parental DNA. This result immediately enabled the conservative mode of replication to be discounted, as this predicts that there would be two bands after one round of replication (Figure 13.3B), but did not provide a distinction between Delbrück's dispersive model and the semiconservative process favored by Watson and Crick. The distinction was, however, possible when the DNA molecules resulting from two rounds of replication were examined. Now the density gradient revealed two bands of DNA, the first corresponding to a hybrid composed of equal parts of newly synthesized and old DNA, and the second corresponding to molecules made up entirely of new DNA. This result agrees with the semiconservative scheme but is incompatible with dispersive replication, the latter predicting that after two rounds of replication all molecules would be hybrids.

Initiation at the E. coli origin of replication

We know substantially more about initiation of replication in bacteria than in eukaryotes. The E. coli origin of replication is referred to as oriC. By transferring segments of DNA from the oriC region into plasmids that lack their own origins, it has been estimated that the E. coli origin of replication spans approximately 245 bp of DNA. Sequence analysis of this segment shows that it contains two short repeat motifs, one of nine nucleotides and the other of 13 nucleotides (Figure 13.8A). The nine-nucleotide repeat, five copies of which are dispersed throughout oriC, is the binding site for a protein called DnaA. With five copies of the binding sequence, it might be imagined that five copies of DnaA attach to the origin, but in fact bound DnaA proteins cooperate with unbound molecules until some 30 are associated with the origin. Attachment occurs only when the DNA is negatively supercoiled, as is the normal situation for the E. coli chromosome (Section 2.3.1).

The result of DnaA binding is that the double helix opens up (‘melts’) within the tandem array of three AT-rich, 13-nucleotide repeats located at one end of the oriC sequence (Figure 13.8B). The exact mechanism is unknown but DnaA does not appear to possess the enzymatic activity needed to break base pairs, and it is therefore assumed that the helix is melted by torsional stresses introduced by attachment of the DnaA proteins. An attractive model imagines that the DnaA proteins form a barrel-like structure around which the helix is wound. Melting the helix is promoted by HU, the most abundant of the DNA packaging proteins of E. coli (Section 2.3.1).

Melting of the helix initiates a series of events that culminates in the start of the elongation phase of replication. The first step is attachment of a complex of two proteins, DnaBC, forming the prepriming complex. DnaC has a transitory role and is released from the complex soon after it is formed, its function probably being simply to aid the attachment of DnaB. The latter is a helicase, an enzyme which can break base pairs (see Section 13.2.2). DnaB begins to increase the single-stranded region within the origin, enabling the enzymes involved in the elongation phase of genome replication to attach. This represents the end of the initiation phase of replication in E. coli as the replication forks now start to progress away from the origin and DNA copying begins.

Origins of replication in yeast have been clearly defined

The technique used to delineate the E. coli oriC sequence, involving transfer of DNA segments into a non-replicating plasmid, has also proved valuable in identifying origins of replication in the yeast Saccharomyces cerevisiae. Origins identified in this way are called autonomously replicating sequences or ARSs. A typical yeast ARS is shorter than the E. coli origin, usually less than 200 bp in length, but, like the E. coli origin, contains discrete segments with different functional roles, these ‘subdomains’ having similar sequences in different ARSs (Figure 13.9A). Four subdomains are recognized. Two of these - subdomains A and B1 - make up the origin recognition sequence, a stretch of some 40 bp in total that is the binding site for the origin recognition complex (ORC), a set of six proteins that attach to the ARS (Figure 13.9B). ORCs have been described as yeast versions of the E. coli DnaA proteins (Kelman, 2000), but this interpretation is probably not strictly correct because ORCs appear to remain attached to yeast origins throughout the cell cycle (Bell and Stillman, 1992; Diffley and Cocker, 1992). Rather than being genuine initiator proteins, it is more likely that ORCs are involved in the regulation of genome replication, acting as mediators between replication origins and the regulatory signals that coordinate the initiation of DNA replication with the cell cycle (Section 13.3; Stillman, 1996).

We must therefore look elsewhere in yeast ARSs for sequences with functions strictly equivalent to that of oriC of E. coli. This leads us to the two other conserved sequences in the typical yeast ARS, subdomains B2 and B3 (see Figure 13.9A). Our current understanding suggests that these two subdomains function in a manner similar to the E. coli origin. Subdomain B2 appears to correspond to the 13-nucleotide repeat array of the E. coli origin, being the position at which the two strands of the helix are first separated. This melting is induced by torsional stress introduced by attachment of a DNA-binding protein, ARS binding factor 1 (ABF1), which attaches to subdomain B3 (see Figure 13.9B). As in E. coli, melting of the helix within a yeast replication origin is followed by attachment of the helicase and other replication enzymes to the DNA, completing the initiation process and enabling the replication forks to begin their progress along the DNA, as described in Section 13.2.2.

Replication origins in higher eukaryotes have been less easy to identify

Attempts to identify replication origins in humans and other higher eukaryotes have, until recently, been less successful (Gilbert, 2001). Initiation regions (parts of the chromosomal DNA where replication initiates) were delineated by various biochemical methods, for example by allowing replication to initiate in the presence of labeled nucleotides, then arresting the process, purifying the newly synthesized DNA, and determining the positions of these nascent strands in the genome. These experiments suggested that there are specific regions in mammalian chromosomes where replication begins, but some researchers were doubtful whether these regions contained replication origins equivalent to yeast ARSs. One alternative hypothesis was that replication is initiated by protein structures that have specific positions in the nucleus, the chromosome initiation regions simply being those DNA segments located close to these protein structures in the three-dimensional organization of the nucleus.

Doubts about mammalian replication origins were increased by the failure of mammalian initiation regions to confer replicative ability on replication-deficient plasmids, although these experiments were not considered conclusive because it was recognized that a mammalian origin might be too long to be cloned in a plasmid or might function only when activated by distant sites in the chromosomal DNA. The breakthrough eventually came when an 8 kb segment of a human initiation region was transferred to the monkey genome, where it still directed replication despite being removed from any hypothetical protein structure in the human nucleus (Aladjem et al., 1998). Analysis of this transferred initiation region showed that there are primary sites within the region where initiation occurs at high frequency, surrounded by secondary sites, spanning the entire 8 kb region, at which replication initiates with lower frequency. The presence of discrete functional domains within the initiation region could also be demonstrated by examining the effects of deletions of parts of the region on the efficiency of replication initiation.

The demonstration that the human genome contains replication origins equivalent to those in yeast raises the question of whether mammals possess an equivalent of the yeast ORC. The answer appears to be yes, as several genes whose protein products have similar sequences to the yeast ORC proteins have been identified in higher eukaryotes and some of these have been shown to be able to replace the equivalent yeast protein in the yeast ORC (Carpenter et al., 1996). These results indicate that initiation of replication in yeast is a good model for events occurring in mammals, a conclusion that is very relevant to studies of the control of replication initiation, as we will see in Section 13.3.

The DNA polymerases of bacteria and eukaryotes

The principal chemical reaction catalyzed by a DNA polymerase is the 5′→3′ synthesis of a DNA polynucleotide, as shown in Figure 13.10. We learnt in Section 4.1.1 that some DNA polymerases combine this function with at least one exonuclease activity, which means that these enzymes can degrade polynucleotides as well as synthesize them (see Figure 4.7):

• A 3′→5′ exonuclease is possessed by many bacterial and eukaryotic template-dependent DNA polymerases (Table 13.2). This activity enables the enzyme to remove nucleotides from the 3′ end of the strand that it has just synthesized. It is looked on as a proofreading activity whose function is to correct the occasional base-pairing error that might occur during strand synthesis (see Section 14.1.1).

• A 5′→3′ exonuclease activity is less common but is possessed by some polymerases whose function in replication requires that they must be able to remove at least part of a polynucleotide that is already attached to the template strand that the polymerase is copying. This activity is utilized during the process that joins together the discontinuous DNA fragments synthesized on the lagging strand during bacterial DNA replication (see Figure 13.17).

The search for DNA polymerases began in the mid-1950s, as soon as it was realized that DNA synthesis was the key to replication of genes. It was thought that bacteria would probably have just a single DNA polymerase, and when the enzyme now called DNA polymerase I was isolated by Arthur Kornberg in 1957 there was a widespread assumption that this was the main replicating enzyme. The discovery that inactivation of the E. coli polA gene, which codes for DNA polymerase I, was not lethal (cells were still able to replicate their genomes) therefore came as something of a surprise, especially when a similar result was obtained with inactivation of polB, coding for a second enzyme, DNA polymerase II, which we now know is mainly involved in repair of damaged DNA rather than genome replication (Section 14.2.5). It was not until 1972 that the main replicating polymerase of E. coli, DNA polymerase III, was eventually isolated. Both DNA polymerases I and III are involved in genome replication, as we will see in the next section.

The properties of the two E. coli DNA polymerases involved in genome replication are described in Table 13.2. DNA polymerases I and II are single polypeptides but DNA polymerase III, befitting its role as the main replicating enzyme, is multi-subunit, with a molecular mass of approximately 900 kDa. The three main subunits, which form the core enzyme, are called α, ε and θ, with the polymerase activity specified by the α subunit and the 3′→5′ exonuclease by ε. The function of θ is not clear: it may have a purely structural role in bringing together the other two core subunits and in assembling the various accessory subunits. The latter include τ and γ, both coded by the same gene, with synthesis of γ involving translational frameshifting (Section 11.2.3), β, which acts as a ‘sliding clamp’ and holds the polymerase complex tightly to the template, δ, δ′, χ and ψ.

Eukaryotes have at least nine DNA polymerases (Hübscher et al., 2000), which in mammals are distinguished by Greek suffices (α, β, γ, δ, etc.), an unfortunate choice of nomenclature as it tempts confusion with the identically named subunits of E. coli DNA polymerase III. The main replicating enzyme is DNA polymerase δ (Table 13.2), which has two subunits (three according to some researchers) and works in conjunction with an accessory protein called the proliferating cell nuclear antigen (PCNA). PCNA is the functional equivalent of the β subunit of E. coli DNA polymerase III and holds the enzyme tightly to the template. DNA polymerase α also has an important function in DNA synthesis, being the enzyme that primes eukaryotic replication (see Figure 13.12B). DNA polymerase γ, although coded by a nuclear gene, is responsible for replicating the mitochondrial genome.

Discontinuous strand synthesis and the priming problem

The limitation that DNA polymerases can synthesize polynucleotides only in the 5′→3′ direction means that the lagging strand of the parent molecule must be copied in a discontinuous fashion, as shown in Figure 13.11. The implication of this model - that the initial products of lagging-strand replication are short segments of polynucleotide - was confirmed in 1969 when Okazaki fragments, as these segments are now called, were first isolated from E. coli (Okazaki and Okazaki, 1969). In bacteria, Okazaki fragments are 1000–2000 nucleotides in length, but in eukaryotes the equivalent fragments appear to be much shorter, perhaps less than 200 nucleotides. This is an interesting observation that might indicate that each round of discontinuous synthesis replicates the DNA associated with a single nucleosome (140 and 150 bp wound around the core particle plus 50–70 bp of linker DNA: Section 2.2.1).

The second difficulty illustrated in Figure 13.11 is the need for a primer to initiate synthesis of each new polynucleotide. It is not known for certain why DNA polymerases cannot begin synthesis on an entirely single-stranded template, but it may relate to the proofreading activity of these enzymes, which is essential for the accuracy of replication. As described in Section 14.1.1, a nucleotide that has been inserted incorrectly at the extreme 3′ end of a growing DNA strand, and hence is not base-paired to the template polynucleotide, can be removed by the 3′→5′ exonuclease activity of a DNA polymerase. This means that the 3′→5′ exonuclease activity must be more effective than the 5′→3′ polymerase activity when the 3′ nucleotide is not base-paired to the template. The implication is that the polymerase can extend a polynucleotide efficiently only if its 3′ nucleotide is base-paired, which in turn could be the reason why an entirely single-stranded template, which by definition lacks a base-paired 3′ nucleotide, cannot be used by a DNA polymerase.

Whatever the reason, priming is a necessity in DNA replication but does not present too much of a problem. Although DNA polymerases cannot deal with an entirely single-stranded template, RNA polymerases have no difficulty in this respect, so the primers for DNA replication are made of RNA. In bacteria, primers are synthesized by primase, a special RNA polymerase unrelated to the transcribing enzyme, with each primer 4–15 nucleotides in length (Frick and Richardson, 2001). Once the primer has been completed, strand synthesis is continued by DNA polymerase III (Figure 13.12A). In eukaryotes the situation is slightly more complex because the primase is tightly bound to DNA polymerase α, and cooperates with this enzyme in synthesis of the first few nucleotides of a new polynucleotide. This primase synthesizes an RNA primer of 8–12 nucleotides, and then hands over to DNA polymerase α, which extends the RNA primer by adding about 20 nucleotides of DNA. This DNA stretch often has a few ribonucleotides mixed in, but it is not clear if these are incorporated by DNA polymerase α or by intermittent activity of the primase. After completion of the RNA-DNA primer, DNA synthesis is continued by the main replicative enzyme, DNA polymerase δ (Figure 13.12B).

Priming needs to occur just once on the leading strand, within the replication origin, because once primed, the leading-strand copy is synthesized continuously until replication is completed. On the lagging strand, priming is a repeated process that must occur every time a new Okazaki fragment is initiated. In E. coli, which makes Okazaki fragments of 1000–2000 nucleotides in length, approximately 4000 priming events are needed every time the genome is replicated. In eukaryotes the Okazaki fragments are much shorter and priming is a highly repetitive event.

Events at the bacterial replication fork

Now that we have considered the complications introduced by discontinuous strand synthesis and the priming problem, we can move on to study the combination of events occurring at the replication fork during the elongation phase of genome replication.

In Section 13.2.1 we identified attachment of the DnaB helicase, followed by extension of the melted region of the replication origin, as representing the end of the initiation phase of replication in E. coli. To a large extent, the division between initiation and elongation is artificial, the two processes running seamlessly one into the other. After the helicase has bound to the origin to form the prepriming complex, the primase is recruited, resulting in the primosome, which initiates replication of the leading strand. It does this by synthesizing the RNA primer that DNA polymerase III needs in order to begin copying the template.

DnaB is the main helicase involved in genome replication in E. coli, but it is by no means the only helicase that this bacterium possesses: in fact there were eleven at the last count (van Brabant et al., 2000). The size of the collection reflects the fact that DNA unwinding is required not only during replication but also during diverse processes such as transcription, recombination and DNA repair. The mode of action of a typical helicase has not been precisely defined, but it is thought that these enzymes bind to single-stranded rather than double-stranded DNA, and migrate along the polynucleotide in either the 5′→3′ or 3′→5′ direction, depending on the specificity of the helicase. Breakage of base pairs in advance of the helicase requires energy, which is generated by hydrolysis of ATP. According to this model, a single DnaB helicase could migrate along the lagging strand (DnaB is a 5′→3′ helicase), unzipping the helix and generating the replication fork, the torsional stress generated by the unwinding activity being relieved by DNA topoisomerase action (Figure 13.13). This model is probably a good approximation of what actually happens, although it does not provide a function for the two other E. coli helicases thought to be involved in genome replication. Both of these, PriA and Rep, are 3′→5′ helicases and so could conceivably complement DnaB activity by migrating along the leading strand, but they may have lesser roles. The involvement of Rep in DNA replication might in fact be limited to participation in the rolling circle process used by λ and a few other E. coli bacteriophages (Section 13.1.3).

Single-stranded DNA is naturally ‘sticky’ and the two separated polynucleotides produced by helicase action would immediately reform base pairs after the enzyme has passed, if allowed to. The single strands are also highly susceptible to nuclease attack and are likely to be degraded if not protected in some way. To avoid these unwanted outcomes, single-strand binding proteins (SSBs) attach to the polynucleotides and prevent them from reassociating or being degraded (Figure 13.14A). The E. coli SSB is made up of four identical subunits and probably works in a similar way to the major eukaryotic SSB, called replication protein A (RPA), by enclosing the polynucleotide in a channel formed by a series of SSBs attached side by side on the strand (Figure 13.14B; Bochkarev et al., 1997). Detachment of the SSBs, which must occur when the replication complex arrives to copy the single strands, is brought about by a second set of proteins called replication mediator proteins (RMPs; Beernick and Morrical, 1999). As with helicases, SSBs have diverse roles in different processes involving DNA unwinding.

After 1000–2000 nucleotides of the leading strand have been replicated, the first round of discontinuous strand synthesis on the lagging strand can begin. The primase, which is still associated with the DnaB helicase in the primosome, makes an RNA primer which is then extended by DNA polymerase III (Figure 13.15). This is the same DNA polymerase III complex that is synthesizing the leading-strand copy, the complex comprising, in effect, two copies of the polymerase. It is not two complete copies because there is only a single γ complex, containing subunit γ in association with δ, δ′, χ and ψ. The main function of the γ complex is to interact with the β subunit (the ‘sliding clamp’) and hence control the attachment and removal of the enzyme from the template, a function that is required primarily during lagging-strand replication when the enzyme has to attach and detach repeatedly at the start and end of each Okazaki fragment. Some models of the DNA polymerase III complex place the two enzymes in opposite orientations to reflect the different directions in which DNA synthesis occurs, towards the replication fork on the leading strand and away from it on the lagging strand. It is more likely, however, that the pair of enzymes face the same direction and the lagging strand forms a loop, so that DNA synthesis can proceed in parallel as the polymerase complex moves forwards in pace with the progress of the replication fork (Figure 13.16).

The combination of the DNA polymerase III dimer and the primosome, migrating along the parent DNA and carrying out most of the replicative functions, is called the replisome. After its passage, the replication process must be completed by joining up the individual Okazaki fragments. This is not a trivial event because one member of each pair of adjacent Okazaki fragments still has its RNA primer attached at the point where ligation should take place (Figure 13.17). Table 13.2 shows us that this primer cannot be removed by DNA polymerase III, because this enzyme lacks the required 5′→3′ exonuclease activity. At this point, DNA polymerase III releases the lagging strand and its place is taken by DNA polymerase I, which does have a 5′→3′ exonuclease and so removes the primer, and usually the start of the DNA component of the Okazaki fragment as well, extending the 3′ end of the adjacent fragment into the region of the template that is exposed. The two Okazaki fragments now abut, with the terminal regions of both composed entirely of DNA. All that remains is for the missing phosphodiester bond to be put in place by a DNA ligase, linking the two fragments and completing replication of this region of the lagging strand.

The eukaryotic replication fork: variations on the bacterial theme

The elongation phase of genome replication is similar in bacteria and eukaryotes, although the details differ. The progress of the replication fork in eukaryotes is maintained by helicase activity, although which of the several eukaryotic helicases that have been identified are primarily responsible for DNA unwinding during replication has not been established. The separated polynucleotides are prevented from reattaching by single-strand binding proteins, the main one of these in eukaryotes being RPA.

We begin to encounter unique features of the eukaryotic replication process when we examine the method used to prime DNA synthesis. As described on page 397, the eukaryotic DNA polymerase α cooperates with the primase enzyme to put in place the RNA-DNA primers at the start of the leading-strand copy and at the beginning of each Okazaki fragment. However, DNA polymerase α is not capable of lengthy DNA synthesis, presumably because it lacks the stabilizing effect of a sliding clamp equivalent to the β subunit of E. coli DNA polymerase III or the PCNA accessory protein that aids the eukaryotic DNA polymerase δ. This means that although DNA polymerase α can extend the initial RNA primer with about 20 nucleotides of DNA, it must then be replaced by the main replicative enzyme, DNA polymerase δ (see Figure 13.12B).

The DNA polymerase enzymes that copy the leading and lagging strands in eukaryotes do not associate into a dimeric complex equivalent to the one formed by DNA polymerase III during replication in E. coli. Instead, the two copies of the polymerase remain separate. The function performed by the γ complex of the E. coli polymerase - controlling attachment and detachment of the enzyme from the lagging strand - appears to be carried out by a multi-subunit accessory protein called replication factor C (RFC).

As in E. coli, completion of lagging-strand synthesis requires removal of the RNA primer from each Okazaki fragment. There appears to be no eukaryotic DNA polymerase with the 5′→3′ exonuclease needed for this purpose and the process is therefore very different to that described for bacterial cells. The central player is the ‘flap endonuclease’, FEN1 (previously called MF1), which associates with the DNA polymerase δ complex at the 3′ end of an Okazaki fragment, in order to degrade the primer from the 5′ end of the adjacent fragment. Understanding exactly how this occurs is complicated by the inability of FEN1 to initiate primer degradation because it is unable to remove the ribonucleotide at the extreme 5′ end of the primer, because this ribonucleotide carries a 5′-triphosphate group which blocks FEN1 activity (Figure 13.18). Two alternative models to circumvent this problem have been proposed (Waga and Stillman, 1998):

• The first possibility is that a helicase breaks the base pairs holding the primer to the template strand, enabling the primer to be pushed aside by DNA polymerase δ as it extends the adjacent Okazaki fragment into the region thus exposed (Figure 13.19A). The flap that results can be cut off by FEN1, whose endonuclease activity can cleave the phosphodiester bond at the branch point where the displaced region attaches to the part of the fragment that is still base-paired.

• Alternatively, most of the RNA component of the primer could be removed by RNase H, which can degrade the RNA part of a base-paired RNA-DNA hybrid, but cannot cleave the phosphodiester bond between the last ribonucleotide and the first deoxyribonucleotide. However, this ribonucleotide will carry a 5′-monophosphate rather than triphosphate and so can be removed by FEN1 (Figure 13.19B).

Both schemes are made attractive by the possibility that both the RNA primer and all of the DNA originally synthesized by DNA polymerase α are removed. This is because DNA polymerase α has no 3′→5′ proofreading activity (see Table 13.2) and therefore synthesizes DNA in a relatively error-prone manner. To prevent these errors from becoming permanent features of the daughter double helix, this region of DNA might be degraded and resynthesized by DNA polymerase δ, which does have a proofreading activity and so makes a highly accurate copy of the template. At present this possibility remains speculative.

The final difference between replication in bacteria and eukaryotes is that in eukaryotes there is no equivalent of the bacterial replisome. Instead, the enzymes and proteins involved in replication form sizeable structures within the nucleus, each containing hundreds or thousands of individual replication complexes. These structures are immobile because of attachments with the nuclear matrix, so DNA molecules are threaded through the complexes as they are replicated. The structures are referred to as replication factories (Figure 13.20) and may in fact also be features of the replication process in at least some bacteria (Lemon and Grossman, 1998; Cook, 1999).

Replication of the E. coli genome terminates within a defined region

Bacterial genomes are replicated bidirectionally from a single point (see Figure 13.7), which means that the two replication forks should meet at a position diametrically opposite the origin of replication on the genome map. However, if one fork is delayed, possibly because it has to replicate extensive regions where transcription is occurring, then it might be possible for the other fork to overshoot the halfway point and continue replication on the ‘other side’ of the genome (Figure 13.21). It is not immediately apparent why this should be undesirable, the daughter molecules presumably being unaffected, but it is not allowed to happen because of the presence of terminator sequences. Seven of these have been identified in the E. coli genome (Figure 13.22A), each one acting as the recognition site for a sequence-specific DNA-binding protein called Tus.

The mode of action of Tus is quite unusual. When bound to a terminator sequence, a Tus protein allows a replication fork to pass if the fork is moving in one direction, but blocks progress if the fork is moving in the opposite direction around the genome. The directionality is set by the orientation of the Tus protein on the double helix. When approached from one direction, Tus blocks the passage of the DnaB helicase, which is responsible for progression of the replication fork, because the helicase is faced with a ‘wall’ of β-strands which it is unable to penetrate. But when approaching from the other direction, DnaB is able to disrupt the structure of the Tus protein, probably because of the effect that unwinding of the double helix has on Tus, and so is able to pass by (Figure 13.22B; Kamada et al., 1996).

The orientation of the termination sequences, and hence of the bound Tus proteins, in the E. coli genome is such that both replication forks become trapped within a relatively short region on the opposite side of the genome to the origin (see Figure 13.22A). This ensures that termination always occurs at or near the same position. Exactly what happens when the two replication forks meet is unknown, but the event is followed by disassembly of the replisomes, either spontaneously or in a controlled fashion. The result is two interlinked daughter molecules, which are separated by topoisomerase IV.

Little is known about termination of replication in eukaryotes

No sequences equivalent to bacterial terminators are known in eukaryotes, and proteins similar to Tus have not been identified. Quite possibly, replication forks meet at random positions and termination simply involves ligation of the ends of the new polynucleotides. We do know that the replication complexes do not break down, because these factories are permanent features of the nucleus (see Figure 13.20).

Rather than concentrating on the molecular events occurring when replication forks meet, attention has been focused on the difficult question of how the daughter DNA molecules produced in a eukaryotic nucleus do not become impossibly tangled up. Although DNA topoisomerases have the ability to untangle DNA molecules, it is generally assumed that tangling is kept to a minimum so that extensive breakage-and-reunion reactions, as catalyzed by topoisomerases (see Figure 13.4), can be avoided. Various models have been proposed to solve this problem (Falaschi, 2000). One of these (Cook, 1998; Wei et al., 1998) suggests that a eukaryotic genome is not randomly packed into the nucleus, but is ordered around the replication factories, which appear to be present in only limited numbers. It is envisaged that each factory replicates a single region of the DNA, maintaining the daughter molecules in a specific arrangement that avoids their entanglement. Initially, the two daughter molecules are held together by cohesin proteins, which are attached immediately after passage of the replication fork by a process that appears to involve DNA polymerase κ (Takahashi and Yanagida, 2000), an enigmatic enzyme that is essential for replication but whose only known role does not obviously require a DNA polymerase activity. The cohesins maintain the alignment of the sister chromatids until the anaphase stage of nuclear division, when they are cleaved by cutting proteins, enabling the daughter chromosomes to separate (Figure 13.23; Murray, 1999).

Telomeric DNA is synthesized by the telomerase enzyme

Most of the telomeric DNA is copied in the normal fashion during DNA replication but this is not the only way in which it can be synthesized. To compensate for the limitations of the replication process, telomeres can be extended by an independent mechanism catalyzed by the enzyme telomerase. This is an unusual enzyme in that it consists of both protein and RNA. In the human enzyme the RNA component is 450 nucleotides in length and contains near its 5′ end the sequence 5′-CUAACCCUAAC-3′, whose central region is the reverse complement of the human telomere repeat sequence 5′-TTAGGG-3′ (Feng et al., 1995). This enables telomerase to extend the telomeric DNA at the 3′ end of a polynucleotide by the copying mechanism shown in Figure 13.25, in which the telomerase RNA is used as a template for each extension step, the DNA synthesis being carried out by the protein component of the enzyme, which is a reverse transcriptase (Lingner et al., 1997). The correctness of this model is indicated by comparisons between telomere repeat sequences and the telomerase RNAs of other species (Table 13.3): in all organisms that have been looked at, the telomerase RNA contains a sequence that enables it to make copies of the repeat motif present at the organism's telomeres. An interesting feature is that in all organisms the strand synthesized by telomerase has a preponderance of G nucleotides; it is therefore referred to as the G-rich strand.

Telomerase can only synthesize this G-rich strand. It is not clear how the other polynucleotide - the C-rich strand - is extended, but it is presumed that when the G-rich strand is long enough, the primase-DNA polymerase α complex attaches at its end and initiates synthesis of complementary DNA in the normal way (Figure 13.26). This requires the use of a new RNA primer, so the C-rich strand will still be shorter than the G-rich one, but the important point is that the overall length of the chromosomal DNA has not been reduced.

Telomere length is implicated in senescence and cancer

Perhaps surprisingly, telomerase is not active in all mammalian cells. The enzyme is functional in the early embryo, but after birth is active only in the reproductive and stem cells. The latter are progenitor cells that divide continually throughout the lifetime of an organism, producing new cells to maintain organs and tissues in a functioning state. The best-studied examples are the hemopoietic stem cells of the bone marrow, which generate new blood cells.

Cells that lack telomerase activity undergo chromosome shortening every time they divide. Eventually, after many cell divisions, the chromosome ends could become so truncated that essential genes are lost, but this is unlikely to be a major cause of the defects that can occur in cells lacking telomerase activity. Instead, the critical factor is the need to maintain a protein ‘cap’ on each chromosome end, to protect these ends from the effects of the DNA repair enzymes that join together the uncapped ends that are produced by accidental breakage of a chromosome (Section 2.2.1). The proteins that form this protective cap, such as TRF2 in humans, recognize the telomere repeats as their binding sequences, and so have no attachment points after the telomeres have been deleted. If these proteins are absent then the repair enzymes can make inappropriate linkages between the ends of intact, although shortened, chromosomes; it is this that is probably the underlying cause of the disruption to the cell cycle that results from telomere shortening.

Telomere shortening will therefore lead to the termination of a cell lineage. For several years biologists have attempted to link this process with cell senescence, a phenomenon originally observed in cell cultures. All normal cell cultures have a limited lifetime: after a certain number of divisions the cells enter a senescent state in which they remain alive but cannot divide (Figure 13.27). With some mammalian cell lines, notably fibroblast cultures (connective tissue cells), senescence can be delayed by engineering the cells so that they synthesize active telomerase (Reddel, 1998). These experiments suggest a clear relationship between telomere shortening and senescence, but the exactness of the link has been questioned (Blackburn, 2000), and any extrapolation from cell senescence to aging of the organism is fraught with difficulties (Kipling and Faragher, 1999).

Not all cell lines display senescence. Cancerous cells are able to divide continuously in culture, their immortality being looked upon as analogous to tumor growth in an intact organism. With several types of cancer, this absence of senescence is associated with activation of telomerase, sometimes to the extent that telomere length is maintained through multiple cell divisions, but often in such a way that the telomeres become longer than normal because the telomerase is overactive. It is not clear if telomerase activation is a cause or an effect of cancer, although the former seems more likely because at least one type of cancer, dyskeratosis congenita, appears to result from a mutation in the gene specifying the RNA component of human telomerase (Marciniak and Guarente, 2001). The question is critical to understanding the etiology of the cancer but is less relevant to the therapeutic issue, which centers on whether telomerase could be a target for drugs designed to combat the cancer. Such a therapy could be successful even if telomerase activation is an effect of the cancer, because inactivation by drugs would induce senescence of the cancer cells and hence prevent their proliferation.

Establishment of the pre-replication complex enables genome replication to commence

Studies primarily with Saccharomyces cerevisiae have led to a model for controlling the timing of S phase which postulates that genome replication requires construction of pre-replication complexes (pre-RCs) at origins of replication, these pre-RCs being converted to post-RCs as replication proceeds. A post-RC is unable to initiate replication and so cannot accidentally re-copy a piece of the genome before mitosis has occurred (Stillman, 1996). The ORC, the complex of six proteins that is assembled onto domains A and B1 of a yeast ARS (see Figure 13.9B), was an early contender for the pre-RC but is probably not a central component because ORCs are present at origins of replication at all stages of the cell cycle. Instead, the ORC is looked on as the ‘landing pad’ on which the pre-RC is constructed.

Various types of protein have been implicated as components of the pre-RC. The first is Cdc6p, which was originally identified in yeast and subsequently shown to have homologs in higher eukaryotes. Yeast Cdc6p is synthesized at the end of G2, as the cell enters mitosis, and becomes associated with chromatin in early G1 before disappearing at the end of G1, when replication begins (Figure 13.29). The involvement of Cdc6p in the pre-RC is suggested by experiments in which its gene is repressed, which results in an absence of pre-RCs, and other experiments in which Cdc6p is over-produced, which leads to multiple genome replications in the absence of mitosis. There is also biochemical evidence for a direct interaction between Cdc6p and yeast ORCs.

A second component of the pre-RC is thought to be the group of proteins called replication licensing factors (RLFs). As with Cdc6p, the first examples of these proteins were identified in yeast (the MCM family of proteins; Tye, 1999) with homologs in higher eukaryotes discovered at a later date. RLFs become bound to chromatin towards the end of M phase and remain in place until the start of S phase, after which they are gradually removed from the DNA as it is replicated. Their removal may be the key event that converts a pre-RC into a post-RC and so prevents re-initiation of replication at an origin that has already directed a round of replication (Blow and Tada, 2000).

Regulation of pre-RC assembly

Identification of the components of the pre-RC takes us some distance towards understanding how genome replication is initiated, but still leaves open the question of how replication is coordinated with other events in the cell cycle. Cell cycle control is a complex process, mediated largely by protein kinases which phosphorylate and activate enzymes and other proteins that have specific functions during the cell cycle. The same protein kinases are present in the nucleus throughout the cell cycle, so they must themselves be subject to control. This control is exerted partly by proteins called cyclins (whose abundance varies at different stages of the cell cycle), partly by other protein kinases that activate the cyclin-dependent protein kinases, and partly by inhibitory proteins. Even before we start looking for regulators of pre-RC assembly we can anticipate that the control system will be convoluted.

A number of cyclins have been linked with activation of genome replication and prevention of pre-RC reassembly after replication has been completed (Stillman, 1996). These include the mitotic cyclins, whose main function was originally thought to be activation of mitosis but which also repress genome replication. When the effects of these cyclins are blocked by, for example, overproduction of proteins that inhibit their activity, the cell is not only incapable of entering M phase but also undergoes repeated genome replication. There are also more specific S-phase cyclins, such as Clb5p and Clb6p in S. cerevisiae, inactivation of which delays or prevents genome replication, and other mitotic cyclins that are active during G2 phase and prevent the assembly of pre-RCs in the period after genome replication and before cell division (Figure 13.30).

In addition to these cyclin-dependent control systems, genome replication is also regulated by a cyclinindependent protein kinase, Cdc7p-Dbf4p, found in organisms as diverse as yeasts and mammals. The proteins activated by this kinase have not been identified, separate lines of evidence suggesting that both RLFs and ORCs are targeted. Whatever the mechanism, Cdc7p-Dbf4p activity is a prerequisite for replication, the cyclin-dependent processes on their own being insufficient to push the cell into S phase.

Early and late replication origins

Initiation of replication does not occur at the same time at all replication origins, nor is ‘origin firing’ an entirely random process. Some parts of the genome are replicated early in S phase and some later, the pattern of replication being consistent from cell division to cell division (Fangman and Brewer, 1992). The general pattern is that actively transcribed genes and the centromere are replicated early in S phase, and non-transcribed regions of the genome later on. Early-firing origins are therefore tissue specific and reflect the pattern of gene expression occurring in a particular cell.

Understanding what determines the firing time of a replication origin is proving quite difficult. It is not simply the sequence of the origin, because transfer of a DNA segment from its normal position to another site in the same or a different chromosome can result in a change in the firing pattern of origins contained in that segment. This positional effect may be linked with chromatin organization and hence influenced by structures such as locus control regions (Section 8.1.2) that control DNA packaging. The position of the origin in the nucleus may also be important as origins that become active at similar periods within S phase appear to be clustered together, at least in mammals.

Checkpoints within S phase

The final aspect of the regulation of genome replication that we will consider is the function of the checkpoints that exist within S phase. These were first identified when it was shown that one of the responses of yeast cells to DNA damage is a slowing down and possibly a complete halting of the genome replication process (Paulovich and Hartwell, 1995). This is linked with the activation of genes whose products are involved in DNA repair (Section 14.2; Zhou and Elledge, 2000).

As with entry into S phase, cyclin-dependent kinases are implicated in the regulation of S-phase checkpoints. These kinases respond to signals from proteins associated with the replication fork. The identity of these damage-detection proteins has not yet been confirmed, although DNA polymerase ε, which has not been assigned a precise function during DNA synthesis (Table 13.2), is a particularly strong candidate because mutant yeast cells that have abnormal DNA polymerase ε enzymes do not respond to DNA damage in the same way as normal cells. Other replication-fork proteins, including components of the PCNA and the accessory protein RFC, have also been assigned roles in damage detection (Waga and Stillman, 1998). The signals from these proteins are mediated by kinases such as ATM, ATR, Chk1 and Chk2, which elicit the appropriate cellular response. The replication process can be arrested by repressing the firing of origins of replication that are usually activated at later stages in S phase (Santocanale and Diffley, 1998) or by slowing the progression of existing replication forks. If the damage is not excessive then DNA repair processes are activated (Section 14.2); alternatively the cell may be shunted into the pathway of programmed cell death called apoptosis, the death of a single somatic cell as a result of DNA damage usually being less dangerous than allowing that cell to replicate its mutated DNA and possibly give rise to a tumor or other cancerous growth. In mammals, a central player in induction of cell cycle arrest and apoptosis is the protein called p53. This is classified as a tumor-suppressor protein, because when this protein is defective, cells with damaged genomes can avoid the S-phase checkpoints and possibly proliferate into a cancer. p53 is a sequence-specific DNA-binding protein that activates a number of genes thought to be directly responsible for arrest and apoptosis, and also represses expression of others that must be switched off to facilitate these processes.