During the cell‐division cycle, chromosomal DNA must initially be precisely duplicated and then correctly segregated to daughter cells. The accuracy of these two events is maintained by two interlinked cycles: the replication licensing cycle, which ensures precise duplication of DNA, and the cohesion cycle, which ensures correct segregation. Here we provide a general overview of how these two systems are coordinated to maintain genetic stability during the cell cycle.
At the heart of the eukaryotic cell‐division cycle are the precise duplication and segregation of chromosomal DNA. These two events take place at distinct points of the cell cycle: DNA is replicated during S phase, and then the replicated DNA strands are accurately partitioned to the two daughter cells during M phase. The flawless operation of this cycle is important not only for the maintenance of viability, but also for preventing genetic instability, which, in metazoans, could potentially drive tumorigenesis. In this review, we look at the cell cycle from the viewpoint of chromosomal DNA and its associated regulators, and focus on how precise chromosome duplication and segregation occur.
Similar to other cell‐cycle events, chromosome duplication and segregation are chiefly driven by cyclin‐dependent kinases (CDKs) and a ubiquitin ligase known as the anaphase‐promoting complex or cyclosome (APC/C). Immediately after cell division, cells are in the G1 phase of the cell cycle with low CDK activity. Rising levels of S‐phase CDKs induce entry into S phase, and then the M‐phase CDKs drive entry into mitosis. During early M phase, sister chromatid pairs (each comprising one of the sister DNA strands plus associated proteins) are captured by microtubules that emanate from opposite ends of the cell. During late mitosis, the APC/C bound by its activator Cdc20 (Peters, 2002) triggers separation of sister chromatids, causing their segregation to opposite poles of the cell. The APC/C also degrades cyclins, thereby lowering CDK activity and promoting exit from M phase. The cell then pinches into two parts, each containing a full complement of chromosomal DNA.
From the perspective of the protein machinery that acts on it, chromosomal DNA is extremely long. This creates two organizational problems for the cell. First, all the DNA must be replicated, but no section should be replicated more than once. Second, the sister chromatids must be identified and each sent to a different daughter cell. Eukaryotes solve both problems by stably loading chromosomal DNA with proteins during G1 that allow the cell to distinguish replicated and unreplicated DNA and to determine which two DNA strands are sisters. This ‘marking’ allows two large‐scale processes (precise chromosome replication and segregation) to emerge from small local events.
The mark involved in DNA replication comprises the Mcm2–7 complex, which is a heterohexamer of six related ‘minichromosome maintenance’ proteins. Mcm2–7 complexes are loaded onto sites on chromosomal DNA during late mitosis and early G1, thereby ‘licensing’ these sites for use as replication origins in the upcoming S phase (Blow & Laskey, 1988; Diffley, 2004; Nishitani & Lygerou, 2004; Blow & Dutta, 2005). During S phase, replication forks are initiated only at these licensed replication origins. After initiation occurs at each origin, the Mcm2–7 complex is removed and probably moves along with the replication fork. Each replication fork terminates when it encounters another heading towards it; at this stage, the Mcm2–7 complex associated with the fork is displaced from the DNA. To prevent replicated origins from becoming re‐licensed and, hence, re‐replicated, the ability to load new Mcm2–7 complexes onto DNA is prevented during the late G1, S, G2 and early M phases. This behaviour is outlined in Fig 1.
The mark involved in the correct segregation of sister strands is cohesin, which is a complex consisting of at least four proteins that can form a ring structure (Nasmyth & Haering, 2005). At the end of M phase or during G1, the cohesin complex is loaded onto DNA. During S phase, the cohesin complex establishes a physical link (cohesion) between sister chromatids. Cohesin is required for maintenance of this cohesion, and for ensuring that sister chromatids are attached to microtubules extending from opposite spindle poles (sister kinetochore bi‐orientation) during the subsequent M phase (Tanaka et al, 2000; Sonoda et al, 2001). At the metaphase‐to‐anaphase transition, after all chromosomes bi‐orientate, a protease called separase is activated; this cleaves the cohesin ring and allows the sister chromatids to be pulled apart by the mitotic spindle (Fig 1; Uhlmann, 2003). The whole cycle of licensing and cohesin loading is then able to start again.
Licensing and cohesin loading
Crystallography and electron microscopy of the archaeal Mcm2–7 homologue indicates that it forms a hexameric ring with a positively‐charged central channel that is able to encircle double‐stranded DNA (Fletcher et al, 2003; Pape et al, 2003). The loading of Mcm2–7 onto DNA as an origin is licensed might involve the opening of the Mcm2–7 ring to allow it to encircle the DNA. This would explain the high stability of Mcm2–7 complexes on DNA. At least three other proteins—the origin recognition complex (ORC), Cdc6 and Cdt1—are required for Mcm2–7 loading and the functional licensing of DNA. Together, ORC, Cdc6, Cdt1 and Mcm2–7 are components of the pre‐replicative complex (pre‐RC) of proteins that are found at replication origins during G1.
The licensing reaction has been reconstituted on the chromatin of Xenopus spermatozoa using purified ORC, Cdc6, Cdt1 and Mcm2–7, in addition to a chromatin‐remodelling protein that is required for the initial binding of ORC to DNA (Gillespie et al, 2001). ORC binds to the DNA, and then recruits Cdc6 and Cdt1. ORC, Cdc6 and Cdt1 are required for the initial loading of Mcm2–7 onto DNA, but are not essential for the continued binding of Mcm2–7. This is consistent with ORC, Cdc6 and Cdt1 acting as a clamp loader that opens up the Mcm2–7 hexamer and closes it again around the DNA (Fig 2). ORC, Cdc6 and Cdt1 are able to load several Mcm2–7 hexamers onto DNA, and each replication origin (ORC‐binding site) typically has 10–40 Mcm2–7 hexamers associated with it by the end of G1 (Blow & Dutta, 2005).
The cohesin complex consists of four proteins: Smc1, Smc3, Scc1 (also called Mcd1 and Rad21) and Scc3 (also called Psc3 and SA1/2; Nasmyth & Haering, 2005). Each Smc molecule forms an intramolecular antiparallel coiled‐coil, which brings their globular amino (N)‐ and carboxy (C)‐terminal domains together (Anderson et al, 2002; Haering et al, 2002). Smc1 and Smc3 dimerize through their central domains to form a V‐shaped heterodimer, which is closed by Scc1 and Scc3 to form a ring structure. The cohesin complex is loaded onto chromosomes at the end of M phase in fission yeast and metazoan cells, and in late G1 in budding yeast. This loading process requires a complex consisting of Scc2 (also called Mis4 and Nipped‐B) and Scc4 (Ciosk et al, 2000; Tomonaga et al, 2000). Cohesin loading also requires the ATPase activity of Smc1 and Smc3 (Arumugam et al, 2003; Weitzer et al, 2003). Similar to licensing, this loading reaction might involve opening up the cohesin ring and closing it again around the DNA. In Xenopus egg extracts, the recruitment of Scc2 to chromosomes depends on the presence of Mcm2–7 on the chromatin, which therefore tightly links these two processes (Gillespie & Hirano, 2004; Takahashi et al, 2004). However, in budding yeast, cohesin loading takes place independently of pre‐RC proteins and, in contrast to Mcm2–7 loading, can even occur after DNA replication (although, in this case, without contributing to sister chromatid cohesion; Uhlmann & Nasmyth 1998; Haering et al, 2004). Cohesin association with chromosomes is found at preferred sites, which include the regions around centromeres and sites between convergent transcription units (Blat & Kleckner, 1999; Megee et al, 1999; Tanaka et al, 1999; Tomonaga et al, 2000; Bernard et al, 2001; Glynn et al, 2004; Lengronne et al, 2004). One possibility is that once loaded onto DNA, cohesin is redistributed to different places, perhaps by the act of transcription itself.
Downregulation of licensing in late G1
To prevent replicated DNA from becoming re‐licensed and, hence, re‐replicated during a single S phase, it is crucial that the ability to load new Mcm2–7 hexamers onto DNA ceases before entry into S phase. This can be achieved by downregulating the activity of the ORC–Cdc6–Cdt1 loading machinery without affecting Mcm2–7 that is already bound to DNA. Different organisms regulate ORC, Cdc6 and Cdt1 activity in a range of ways (Diffley, 2004; Nishitani & Lygerou, 2004; Blow & Dutta, 2005). These can be grouped into four categories (proteolysis, inhibitory phosphorylation, nuclear export and inhibition by geminin), which we briefly examine in turn below.
Proteolysis. In yeast, rising CDK levels at the end of G1 cause phosphorylation of Cdc6 (fission yeast Cdc18), which targets it for ubiquitylation and subsequent proteolysis (Drury et al, 1997; Jallepalli et al, 1997). In metazoans, Cdc6 levels remain relatively constant throughout the cell cycle. Instead, metazoan Cdt1 is ubiquitylated and degraded at the onset of S phase, partly owing to CDK‐dependent phosphorylation, and this is important in preventing re‐licensing of replicated DNA (Nishitani et al, 2001; Li et al, 2003; Zhong et al, 2003; Arias & Walter, 2005; Li & Blow, 2005). In metazoans, ORC and Cdc6 are also regulated by ubiquitylation and proteolysis, although the physiological consequences of this regulation remain uncertain.
Inhibitory phosphorylation. In yeast, the Orc2 subunit is phosphorylated by CDKs during late G1, S, G2 and mitosis, and this has a role in preventing re‐licensing of replicated DNA (Nguyen et al, 2001; Vas et al, 2001). CDK‐dependent phosphorylation of the mammalian ORC and Cdc6 probably inhibits their binding to DNA during mitosis (Li et al, 2004).
Nuclear export. In budding yeast, but not fission yeast or metazoans, CDK activity leads to the nuclear export of both Mcm2–7 (Tanaka et al, 1997; Labib et al, 1999) and Cdt1. This prevents these proteins from gaining access to their DNA substrate and so prevents re‐licensing of replicated DNA. Export of Mcm2–7 and Cdt1 is interdependent, and might be driven by phosphorylation of Mcm2–7.
Inhibition by geminin. Geminin is a small protein that binds and inhibits Cdt1 (McGarry & Kirschner, 1998; Wohlschlegel et al, 2000; Tada et al, 2001). So far, it has been described only in metazoans. Geminin is ubiquitylated by the APC/C during late mitosis and early G1. In somatic cells, this action targets it for proteolysis; however, in early embryos, ubiquitylation of geminin leads to its inactivation rather than its degradation (Hodgson et al, 2002). Levels of active geminin are high during S, G2 and mitosis, and this is important in preventing re‐licensing of replicated DNA (Quinn et al, 2001; Mihaylov et al, 2002; Melixetian et al, 2004; Zhu et al, 2004; Li & Blow, 2005).
Directly or indirectly, these inhibitory mechanisms ultimately depend on the increase in CDK activity that occurs at the end of G1 and persists until late mitosis (Diffley, 2004; Nishitani & Lygerou, 2004; Blow & Dutta, 2005). The cell cycle can, therefore, be thought of as consisting of two phases. The first phase comprises late mitosis and early G1, when CDK levels are low and licensing and cohesin loading occur. The second phase comprises late G1, S, G2 and early mitosis, when high CDK levels drive DNA replication and chromosome segregation (Fig 1).
In budding yeast, these different inhibitory pathways are redundant; to achieve re‐replication of DNA, it is necessary to stabilize Cdc6, remove inhibitory phosphorylation sites on ORC, and prevent nuclear export of Mcm2–7 and Cdt1 (Nguyen et al, 2001). In metazoans, downregulation of Cdt1 alone seems to be the most important mechanism that prevents re‐licensing of replicated DNA, as overexpression of Cdt1 or elimination of geminin can cause extensive re‐replication (Quinn et al, 2001; Mihaylov et al, 2002; Vaziri et al, 2003; Zhong et al, 2003; Melixetian et al, 2004; Zhu et al, 2004; Li & Blow, 2005). It is curious that metazoans, in which the consequences of genetic instability are greater (as it might cause cancer), do not rely on the numerous mechanisms that are used by budding yeast. One possible explanation is that this might make it harder for genetically unstable clones to arise. If geminin were only one of several redundant mechanisms, its deletion might lead to clones that occasionally re‐replicate segments of DNA; however, being essential for the prevention of re‐replication, geminin deletion is instead lethal.
Mcm2–7 proteins are essential for replication of DNA during S phase. They have sequence similarity with known DNA helicases (DNA‐unwinding enzymes) and show DNA helicase activity in vitro (Blow & Dutta, 2005). Chromatin‐immunoprecipitation experiments in yeast have shown that during S phase, Mcm2–7 proteins are displaced from replication origins and move ahead of the replication fork (Aparicio et al, 1997; Tanaka et al, 1997). If Mcm2–7 function is inhibited during S phase, the rate of DNA synthesis rapidly falls, which indicates that Mcm2–7 proteins are required for the progression of replication forks (Labib et al, 2000; Pacek & Walter, 2004; Shechter et al, 2004). If Mcm2–7 proteins form a replicative helicase that unwinds the template DNA ahead of the replication fork, this explains why they are displaced from replication origins when fork initiation occurs, thereby rendering the origin unlicensed once it has been initiated.
One question left unanswered by this model is why so many Mcm2–7 hexamers are loaded onto the DNA (10–40 per origin). One possible explanation is that Mcm2–7 can act at a distance from the replication fork, pumping the DNA into the fork and thereby untwisting the DNA (Laskey & Madine, 2003). Another possibility is that the excess Mcm2–7 proteins have some other function, perhaps being involved in cell‐cycle‐checkpoint signalling or being used only under certain conditions. Consistent with this idea is the observation that the number of Mcm2–7 complexes can be lowered to approximately two per replication origin without significantly affecting the overall replication rate (Mahbubani et al, 1997).
As DNA is replicated, Mcm2–7 proteins are displaced from it. The release of Mcm2–7 proteins from DNA presumably occurs when replication forks terminate as they encounter another such fork coming from the opposite direction. Little is known about this process, which involves the disassembly of the replication‐fork machinery. It is possible that termination is driven by the release of Mcm2–7 proteins from DNA.
Another circumstance under which Mcm2–7 proteins should be released from DNA is the passive replication of unfired replication origins (Blow & Laskey, 1988). Not all licensed replication origins initiate during each cell cycle, rather they can be passively replicated by forks emanating from adjacent replication origins. Mcm2–7 proteins must be displaced from these uninitiated origins as they are replicated to prevent the replicated origin from being initiated later in the same S phase and, hence, causing re‐replication of the surrounding DNA.
Cohesion between sister chromatids is established during S phase (Uhlmann & Nasmyth, 1998), probably as soon as the DNA is replicated. Although cohesin is required for the establishment of cohesion, the loading of cohesin onto chromosomes is not sufficient for this purpose (Nasmyth & Haering, 2005; Skibbens, 2005). The establishment of cohesion requires several other factors, including Eco1 (also called Ctf7 and Eso1), Ctf4 and a variant of replication factor C containing Ctf18 (Skibbens et al, 1999; Toth et al, 1999; Hanna et al, 2001; Mayer et al, 2001). Loss of these factors causes defects in sister chromatid cohesion but not in the association of cohesin with chromosomes. Little is known about how cohesin structurally contributes to sister chromatid cohesion, but it is interesting that several of the proteins that are required for cohesion establishment might function at the replication fork. One possibility is that the cohesin ring establishes cohesion by embracing both sister chromatids (as implied in Fig 3), although alternative models have also been proposed (Nasmyth & Haering, 2005; Skibbens, 2005).
In many eukaryotic cells, including those of vertebrates (but not budding yeast), most cohesin dissociates from chromosomes during prophase. This process is probably required for the resolution of sister chromatids, and is facilitated by the Polo‐like and Aurora B kinases (Losada et al, 2002; Sumara et al, 2002). Polo‐like kinase, which is activated as CDK levels rise at the onset of M phase, promotes cohesin dissociation by phosphorylating Scc3 (Hauf et al, 2005). After the bulk of cohesin is dissociated from the chromosome arms, cohesin is retained between sister centromeres until the onset of anaphase (Waizenegger et al, 2000). This centromeric retention of cohesin is dependent on shugosin (Mei‐S332 in Drosophila; LeBlanc et al, 1999; Salic et al, 2004; Kitajima et al, 2005; McGuinness et al, 2005). In addition, the cohesin‐loading factor Scc2 is removed from DNA as a consequence of high CDK activity in Xenopus extract (Gillespie & Hirano, 2004).
For proper chromosome segregation, kinetochores assembled onto the centromeric DNA of sister chromatids must attach to microtubules extending from opposite spindle poles before the onset of anaphase (Tanaka, 2005). This state is called sister kinetochore bi‐orientation or amphitelic attachment. If sister kinetochores incorrectly orientate on the mitotic spindle—for instance, if both attach to microtubules from the same spindle pole—then the Aurora B kinase facilitates re‐orientation of kinetochore–microtubule connections by phosphorylating kinetochore components (Cheeseman et al, 2002; Tanaka et al, 2002; Hauf et al, 2003; Lampson et al, 2004; Fig 4). This process continues until proper bi‐orientation occurs, at which stage tension is generated across the two sister kinetochores as they are pulled by microtubules to opposite spindle poles (Nicklas, 1997; Dewar et al, 2004). The presence of tension across the two kinetochores, which depends on cohesin, might prevent Aurora B from further re‐orientating the kinetochore–microtubule interaction. In the absence of cohesin, half the chromosomes fail to bi‐orientate on the spindle, although kinetochores are still able to attach to microtubules on such chromosomes (Tanaka et al, 2000; Sonoda et al, 2001). In this way, sister chromatid cohesion ensures the precise segregation of sister chromatids to the opposite sides of the cell.
At the metaphase‐to‐anaphase transition, the separation of sister chromatids is triggered by the removal of the residual (in vertebrate cells) or bulk (in budding yeast) cohesin from chromosomes (Uhlmann, 2003; Nasmyth & Haering, 2005). As the separation of sister chromatids is irreversible, a number of fail‐safe mechanisms regulate this step. A chief regulator is a protease termed separase (Fig 3), which cleaves the cohesin Scc1 and thereby opens up the cohesin‐ring structure (Uhlmann et al, 2000). Separase activity is inhibited by a protein called securin, and the activation of separase at the metaphase‐to‐anaphase transition depends on the degradation of securin by the proteasome (Cohen‐Fix et al, 1996; Funabiki et al, 1996; Ciosk et al, 1998). This is facilitated by polyubiquitylation of securin by the APC/C bound by its activator Cdc20. Moreover, at least in vertebrate cells, mitotic CDKs are able to phosphorylate separase, which leads to the inhibition of its activity during early mitosis (Stemmann et al, 2001). Therefore, separase activation also requires the degradation of M‐phase cyclins, which is also facilitated by the APC/C. In addition, Polo‐like kinases phosphorylate Scc1 during M phase, which renders it more susceptible to cleavage by separase (Alexandru et al, 2001; Hauf et al, 2005). In case kinetochores fail to attach to microtubules or to bi‐orientate on the spindle, the Mad2 protein binds and inhibits Cdc20, and thereby delays the anaphase onset until all kinetochores properly bi‐orientate on the spindle (the ‘spindle checkpoint’; Lew & Burke, 2003).
On exit from mitosis, the licensing system needs to be reactivated. This is achieved by the inactivation of geminin (in metazoans) and CDKs, which occurs as a consequence of APC/C activation. Reduction of CDK activity increases the affinity of ORC and Cdc6 for chromatin (Tada et al, 2001; Li et al, 2004), and, in budding yeast, allows Mcm2–7 and Cdt1 to re‐enter the nucleus (Labib et al, 1999). In metazoans, polyubiquitylation of geminin by the APC/C leads to its proteolysis (McGarry & Kirschner, 1998) or inactivation (Hodgson et al, 2002). As both geminin and the cyclins are substrates of the APC/C, which is also responsible for cohesin cleavage, the licensing and cohesin cycles are tightly linked. The behaviour of the licensing and cohesin systems therefore ensures the coordination of chromosome replication and segregation during the cell cycle.
Thanks to N. Perkins, P. Gillespie, A. Li and A. Woodward for comments on the manuscript. J.J.B. is supported by Cancer Research UK (CR‐UK) grant C303/A3135. T.U.T. is supported by The Wellcome Trust and CR‐UK.
Second in the Cycles review series
See other issues of EMBO reports for the complete Cycles Review Series:
- Copyright © 2005 European Molecular Biology Organization