In the past few months, several discoveries relating to the mechanism underlying transcription-coupled DNA repair (TCR) have been reported. These results make it timely to propose a hypothesis for how eukaryotic cells might deal with arrested RNA polymerase II (Pol II) complexes. In this model, the transcription-repair coupling factor Cockayne Syndrome B (or the yeast equivalent Rad26) uses DNA translocase activity to remodel the Pol II-DNA interface, possibly to push the polymerase past the obstruction or to remove it from the DNA so that repair can take place if the obstacle is a DNA lesion. However, when this action is not possible and Pol II is left irreversibly trapped on DNA, the polymerase is instead ubiquitylated and eventually removed by proteolysis.
RNA polymerase II will efficiently transcribe DNA only if it can overcome obstacles on the template strand. Many such obstacles can be imagined,including DNA-binding proteins, unusual DNA structures and nucleosomes. However, the most powerful impediment to the progression of the polymerase is likely to be DNA lesions (Svejstrup,2002b). Several types of DNA lesion are known to block transcription by RNA polymerase II in vitro as well as in vivo, and, since transcription is in effect a one-track copying system, an irreversibly trapped polymerase is unacceptable for the cell, especially if this occurs in a gene whose continuous productivity is essential for cell viability. Not surprisingly, therefore, cells have evolved efficient systems to respond to arrested RNA polymerases (Conaway et al.,2000; Svejstrup,2002a).
Hanawalt and colleagues recognized almost two decades ago that DNA lesions in the transcribed strand of an active gene are repaired much faster than those in the non-transcribed strand or in the genome overall(Bohr et al., 1985; Mellon et al., 1986; Mellon et al., 1987), but,surprisingly, the molecular mechanism underlying such `transcription-coupled'DNA repair (TCR) is still unclear. The importance of TCR is evident from the fact that patients with deficiencies in the repair of lesions in the transcribed strand of an active gene suffer from a very severe hereditary disorder, Cockayne syndrome (CS) (de Boer and Hoeijmakers, 2000). Mutations in genes encoding basal transcription factors or factors that are required for nucleotide excision repair (NER) can give rise to both CS and xeroderma pigmentosum (XP),including those in genes encoding two subunits of TFIIH (XPB and XPD), and XP group G protein (XPG). By contrast, mutations in the genes encoding CS group A(CSA) and CSB (yeast counterpart Rad26) can give rise to CS but do not result in XP (Balajee and Bohr, 2000; Conaway and Conaway, 1999; de Boer and Hoeijmakers,2000). Common among all these factors is that they have been shown to play a role in both transcription and DNA repair(Feaver et al., 1993; Lee et al., 2002b; Lee et al., 2001; Lee et al., 2002c; Schaeffer et al., 1993; Selby and Sancar, 1997a). Interestingly, the repair deficiencies arising as a consequence of these mutations do not underlie CS, since patients with the repair disorder xeroderma pigmentosum are unable to repair UV-mediated DNA damage, both in the transcribed and non-transcribed strand, yet do not have the same severe clinical features that characterize patients who suffer (or also suffer) from CS (de Boer and Hoeijmakers,2000; Lehmann,2001; Svejstrup,2002b). Symptoms of CS such as growth retardation, skeletal and retinal abnormalities and progressive neural retardation are thus likely to be caused by transcription deficiencies that may be related to DNA damage rather than directly to failure to remove DNA damage from the genome. Very revealing for the possible mechanism of TCR is the presence of ATPase motifs in the CSB protein. These motifs are similar to those found in the Snf/Swi-family of ATP-dependent chromatin-remodeling enzymes(Eisen et al., 1995).
In this Hypothesis, I discuss the results of four important recent studies (Lee et al., 2002a; Park et al., 2002; Saha et al., 2002; Woudstra et al., 2002), which are relevant to the mechanism of TCR in eukaryotes. On the basis of these studies, I propose a model for the rescue of arrested Pol II complexes in which RNA polymerases stalled at DNA lesions are not considered simply as substrates for a subform of DNA repair (i.e. TCR) but rather as a problem for transcription that can be solved in several fundamentally different ways. It is, however, important to point out that TCR is turning out to be an extremely complex phenomenon and that the model argued for here represents only one current interpretation of often conflicting or at least confusing evidence. For a more comprehensive review of the literature, the reader is referred to recent reviews of the field (Balajee and Bohr, 2000; Conaway and Conaway, 1999; de Boer and Hoeijmakers, 2000; Svejstrup,2002b).
A stalled polymerase triggers a rapid cellular response
The idea that damage-stalled Pol II is an unacceptable occurrence in cells stems from the fact that DNA damage in the transcribed strand of an active gene is preferentially repaired even though — in theory — the stalled polymerase covers the lesion and therefore should make it less accessible to the repair machinery. Several lines of evidence support the suggestion that Pol II itself triggers rapid repair. First, preferential repair of the transcribed strand of an active gene ceases immediately if Pol II transcription is stopped by the use of drugs (α-amanitin) or elevated temperature (in yeast strains expressing temperature-sensitive versions of Pol II) (Christians and Hanawalt,1992; Sweder and Hanawalt,1992). Further persuasive evidence comes from work on the rates of repair in and around active genes, mostly in yeast(Teng and Waters, 2000; Tijsterman and Brouwer, 1999; Tijsterman et al., 1999; Tijsterman et al., 1996; Tijsterman et al., 1997; Tu et al., 1997; Tu et al., 1998; Tu et al., 1996). TCR is generally observed only in the open reading frame of an active gene but not in the promoter or in the region downstream from the transcription termination sites. It has also been shown that the rate of repair in the non-transcribed strand, but not in the transcribed strand, is significantly affected by the presence of nucleosomes (Tijsterman et al., 1999; Wellinger and Thoma, 1997). Importantly, repair on the non-transcribed strand is much slower at the cores of positioned nucleosomes than in internucleosomal regions, whereas the rate of repair on the transcribed strand is largely independent of the presence and position of the same nucleosomes. However,certain lesions in the internucleosomal regions are repaired more slowly in the transcribed strand than in the non-transcribed strand if TCR is perturbed by deletion of RAD26, which encodes the yeast homologue of CSB. The simplest interpretation of these results is that NER at what are likely to be more accessible sites is actually obstructed by Pol II in the transcribed strand and that one function of Rad26 (and, by extension, CSB) is somehow to remove the polymerase so that repair can take place(Tijsterman et al., 1999).
The Swi/Snf-like ATPase activity of CSB/Rad26 and its role in removal of Pol II from DNA damage
So how might CSB/Rad26 help remove Pol II? Several possibilities should be considered. First, it might remodel the Pol-II-DNA interface to make the lesion more accessible. Second, it might release Pol II from the site of DNA damage. Third, it might promote damage bypass by Pol II. As mentioned above,CSB/Rad26 is an ATPase in the Swi/Snf-family(Eisen et al., 1995) and has a prokaryotic counterpart, transcription repair coupling factor (TRCF)(Park et al., 2002; Selby and Sancar, 1993). Significantly, in a compelling recent study of TRCF, Park et al. discovered that this protein is an ATP-dependent DNA translocase that can move RNAP forward on DNA (Park et al.,2002). Thus, TCRF can `push' a stalled polymerase whose active site is out of register with the end of the RNA up to the end of the nascent RNA so that it can continue transcription if nucleotides and a clear (DNA)path are present. A well established feature of TRCF is its ability to displace RNAP from DNA when nucleotides are absent or if an obstacle is in the way. So, bacterial TRCF is also capable of pushing RNAP off the DNA if the arrest is due to a DNA lesion (Selby and Sancar, 1993). This feature does not seem to be conserved in the eukaryotic counterpart, CSB, which is unable to displace Pol II from a damage site (Selby and Sancar,1997b). Since CSB and Rad26 are only distantly related to TRCF(except for limited homology in the ATPase domains)(Eisen et al., 1995), it would thus have been reasonable for researchers interested in eukaryotic CSB/Rad26 not to pay too much attention to the lessons learnt from bacterial TRCF. However, two pieces of evidence should be mentioned that are important for building working models of the mechanism used by the eukaryotic factors. First, CSB can promote the addition of an extra nucleotide by Pol II stalled at a DNA lesion (Selby and Sancar,1997a). Second, a recent study on the reaction mechanism of eukaryotic Swi/Snf-like enzymes reports that the Swi/Snf-like enzyme RSC, like TCRF, is a DNA translocase (Saha et al.,2002). The data from Cairns and co-workers raise the possibility that all Swi/Snf-like enzymes move proteins about on DNA by pulling or pushing the DNA. For Swi/Snf-like enzymes, such as RSC, CHRAC, ISWI and Swi/Snf itself(Cairns, 1998; Langst and Becker, 2001), the target protein is likely to be nucleosomes, but in the case of family members such as Rad54, Rad16 and CSB/Rad26 (Eisen et al., 1995), the target could be something else, although nucleosomes might also be moved, as has been demonstrated in vitro(Citterio et al., 2000). In the case of CSB/Rad26, a primary target could be Pol II. As already mentioned, the action of CSB/Rad26 could in theory allow the Pol-II—DNA interface to be remodeled so that repair can take place. Alternatively/additionally, it could sometimes allow Pol II to be pushed past the site of DNA damage or perhaps to be displaced, all depending on the lesion and the context in which it is found. The possibility that accessory factors are required for some of these events is also important in this connection. The data so far obtained on the mechanism of CSB/Rad26 action have thus been derived by the use of purified recombinant proteins (Citterio et al.,2000; Guzder et al.,1996; Selby and Sancar,1997a; Selby and Sancar,1997b), rather than the native protein, which might well exist in a complex (van Gool et al.,1997; Woudstra et al.,2002).
Proteolysis of Pol II in response to DNA damage
But what happens if the action of CSB/Rad26 on damage-stalled Pol II does not solve the problem? In light of the considerations mentioned above, this might require much more drastic action. Important in this connection, Pol II becomes ubiquitylated and degraded in response to DNA damage(Beaudenon et al., 1999; Bregman et al., 1996; Luo et al., 2001; Ratner et al., 1998), which raises the possibility that it is removed from lesion sites so that rapid repair can take place. However, enabling TCR is unlikely to be the function of Pol II proteolysis (Lommel et al.,2000; Woudstra et al.,2002). Rather, a new study suggests that ubiquitylation and Pol II degradation in response to DNA damage constitutes an alternative to DNA repair— a last resort (Woudstra et al.,2002).
Woudstra et al. purified yeast Rad26 from soluble (DNA-free) as well as salt-stable chromatin and found that it exists in different forms. The form isolated from the soluble fraction does not appear to be associated with any other protein, whereas the form extracted and purified from chromatin is associated with a novel protein, Def1. Cells lacking DEF1 have phenotypes that indicate that this protein has a role in the DNA damage response as well as in Pol II transcript elongation. Remarkably, def1Δ cells are not UV sensitive and perform TCR efficiently. However, when def1Δ is combined with mutations that reduce or abolish the ability of cells to repair DNA damage, cells become extremely sensitive to UV damage, which suggests that Def1 participates in a pathway that represents an alternative to DNA repair. This pathway turns out to be the ubiquitylation-mediated proteolysis pathway, and the target is Pol II. Thus,cells lacking DEF1 are unable to ubiquitylate and degrade Pol II in response to UV damage. Interestingly, cells lacking RAD26 behave in the opposite way: here, Pol II is very rapidly and much more completely destroyed in response to damage than is the case in wild-type cells. However,in cells lacking both DEF1 and RAD26, Pol II degradation is somewhat restored. Rad26 and Def1 might thus functionally interact in vivo to regulate the degradation of Pol II in response to DNA damage(Woudstra et al., 2002). Interestingly, work in mammalian cells has shown that ubiquitylation and Pol II degradation are impaired, not increased, in the absence of functional CSB(Bregman et al., 1996; Luo et al., 2001; McKay et al., 2001). At first glance, the function of yeast Rad26 and human CSB thus appears to differ fundamentally. However, a possible explanation for the difference could be that the function of both CSB and the presumed Def1 homologue is compromised by mutation of CSB in man, whereas Rad26 and Def1 are able to perform their functions independently in yeast. In support of this scenario, gel filtration experiments have shown that (in contrast to yeast Rad26), human CSB normally exists in a large protein complex (van Gool et al., 1997). The function of this entire complex (including Pol II degradation) might thus be compromised by CSB mutation.
An important recent study by Sharp and colleagues supports the idea that arrested Pol II is a target of ubiquitylation(Lee et al., 2002a). Here, it was shown that ubiquitylation of Pol II during transcription in vitro is significantly induced by α-amanitin, which blocks Pol II transcript elongation and also causes Pol II degradation inside cells(Nguyen et al., 1996). In agreement with the above interpretation of the data from Woudstra et al.(Woudstra et al., 2002), Pol II undergoes similar ubiquitylation on DNA containing cisplatin adducts (DNA lesions) that arrest transcription (Lee et al., 2002a). Together, the data by Woudstra et al. and Lee et al. thus suggest that Pol II is subject to ubiquitylation whenever it arrests during transcription and that this can lead to its degradation(Lee et al., 2002a; Woudstra et al., 2002).
A model for repair of damage-stalled Pol II elongation complexes
The data described above make it possible to propose a hypothesis for the events that occur when Pol II arrests during transcription, and particularly in response to transcription-impeding DNA lesions in genes(Fig. 1). Most arrests can probably be resolved by the actions of general elongation factors(Conaway et al., 2000; Svejstrup, 2002a). However,Pol II will also arrest at DNA lesions; so transcription of the gene becomes blocked. General elongation factors are unlikely to be of much help under such circumstances, and the action of, for example, Spt4/5 might actually be counter productive (Jansen et al.,2000). However, in the model, arrested Pol II also leads to recruitment of Rad26/CSB and Def1. Rad26 might have two major effects. First,it might inhibit Pol II degradation, leaving time for the stalled complex to be dealt with through other mechanisms(Woudstra et al., 2002). Second, the catalytic activity of CSB/Rad26 itself (perhaps its DNA translocase activity) could `remodel' the Pol-II-DNA interface (and possibly nucleosomes as well), and its reported interactions with TFIIH might facilitate the recruitment of repair proteins when the stall is due to DNA damage (Tantin, 1998; Tantin et al., 1997). The Pol-II-DNA remodeling might involve translocation of the DNA in the direction of the polymerase [as RSC does with nucleosomes(Saha et al., 2002)], so that the lesion bulges out and can be readily repaired. Alternatively, the protein might translocate DNA in the direction away from the polymerase [as bacterial TRCF does with RNA polymerase (Park et al., 2002)] so that the polymerase is forced forward to the end of the RNA and perhaps to the other side of a lesion (where it can continue transcription, perhaps skipping or mis-incorporating a base) or is displaced off the DNA.
According to the model, if arrested Pol II complexes cannot be resolved by general elongation factors or CSB/Rad26-related mechanisms, Pol II is now ubiquitylated in a process that is at least partially dependent on Def1. Whether there is damage or not, this clears the path for alternative pathways of resolution. Because both repair and transcript elongation is affected by proteolysis-independent functions of the 19S regulatory complex of the proteasome (Ferdous et al.,2001; Gillette et al.,2001; Russell et al.,1999), it is possible that ubiquitylation does not always result in Pol II degradation. Johnston and co-workers have thus proposed that the 19S complex might somehow manipulate the structure of the Pol II elongation complex by proteolysis-independent mechanisms to facilitate transcription(Ferdous et al., 2001). However, when all else fails, ubiquitylation of Pol II would eventually lead to its degradation by the proteasome, allowing transcription of the gene to be completed by the subsequent polymerase and repair to take place by transcription-independent pathways.
It is important to remember that the action of CSB/Rad26 and ubiquitylation factors such as Def1 is not confined to situations in which there is DNA damage. These proteins are likely to have similar roles when cells try to rescue Pol II elongation complexes that are stalled for reasons other than DNA damage (such as DNA structure or sequence context, protein blocks, etc.). In support of this idea, CSB affects transcript elongation of undamaged DNA in vitro (Selby and Sancar,1997a), and cells lacking RAD26 have transcription defects in vivo (Lee et al.,2001). def1Δ cells also have defects consistent with a role for this protein in transcript elongation in the absence of DNA damage (Woudstra et al.,2002). The possibility that cells make overall transcription more efficient by clearing Pol II `roadblocks' through proteolysis of the polymerase in the absence of DNA damage clearly deserves further attention.