The Polymerase Scanning Hypothesis. John A. Heddle* and Jason H. Bielas. Department of Biology, York University, Toronto, Ontario, Canada

Transcription

1 Environmental and Molecular Mutagenesis 45:143^149 (2005) Perspective Unifying Concept of DNA Repair: The Polymerase Scanning Hypothesis John A. Heddle* and Jason H. Bielas Department of Biology, York University, Toronto, Ontario, Canada According to a series of experiments on untransformed mouse embryonic fibroblasts, quiescent mouse cells lack global genomic repair (GGR) of premutagenic DNA damage. The gene used to assess mutation and premutagenic DNA damage was the laci transgene incorporated permanently in the DNA in a l shuttle vector. The transgene lacks mammalian transcription signals and thus is unexpressed in the cells. Although the cells conducted transcription-coupled repair (TCR) of UV damage, the transgene was not repaired over a 4- day interval. These cells are not terminally differentiated and can readily be induced to resume cellular division. In this article, we discuss the interpretation of these results and suggest a new hypothesis for DNA scanning, the mechanism by which cells discover DNA damage and initiate DNA repair. Our hypothesis, which we call the polymerase scanning hypothesis, is that GGR is initiated in very much the same way as TCR, by a polymerase complex encountering the damage. We call the two together polymersase-coupled repair (PC repair). In the case of GGR, it would be the DNA replication complex during the S-phase. This is, we suggest, the dominant mechanism of repair of DNA at low doses for untranscribed genes. Evidence contrary to this hypothesis exists, which we discuss, but it should be noted that existing hypotheses about DNA scanning and DNA repair cannot account for the results that we have obtained. Environ. Mol. Mutagen. 45: , c 2005 Wiley-Liss, Inc. Key words: transcription-coupled repair; global genomic repair; cell cycle; scanning; DNA damage; DNA adducts INTRODUCTION The repair of DNA is a complex and multifaceted process [Friedberg et al., 2004] in which a large number of different repair processes involving multiple proteins act on various types of DNA damage. For quite some time, it has seemed likely that eukaryotic cells must have some means of bringing repair enzymes to the DNA damage and surmounting the barrier to them that the histones and other proteins in chromatin present [Ehrenhofer-Murray, 2004]. The alternative, simple diffusion of enzymes, would not only require the enzymes to permeate chromatin but effectively result in a race between DNA repair and DNA replication. If the replication fork should win, the damage could be fixed irreparably into a mutation. There are several proposals [Hanawalt, 2001; Fuxreiter et al., 2002; Ehrenhofer-Murray, 2004] as to how the DNA might be scanned for DNA base damage, but no previous proposal explains the results that have recently been obtained for the timing of mutation and DNA repair in murine cells, and no other proposal avoids a race between replication and repair. ONE DNA SCANNING MECHANISM IS KNOWN: TRANSCRIPTION-COUPLED REPAIR (TCR) For those genes that are being actively transcribed, the process of transcription itself discovers the damage. For example, when the RNA polymerase, or its associated proteins, encounters a thymine dimer induced by UV, it stalls. This stalling initiates nucleotide excision repair (NER), which cuts out the dimer and about 25 adjacent Jason H. Bielas s present address is: Department of Pathology, University of Washington, Seattle, Washington. Grant sponsor: the National Cancer Institute of Canada; Grant sponsor: the Canadian Cancer Society. *Correspondence to: John A. Heddle, Department of Biology, York University, Toronto, Ontario, M3J 1P3, Canada. jheddle@yorku.ca Invited article: 25th anniversary of Environmental and Molecular Mutagenesis DOI /em Published online 25 January 2005 in Wiley InterScience ( wiley.com). c 2005 Wiley-Liss, Inc.

2 144 Heddle and Bielas bases from one strand and then replaces the excised bases by copying the complementary strand, which is still intact [Svejstrup, 2002]. During the repair, the transcription complex loses some proteins and gains others to form a repairosome, which repairs the DNA. Untranscribed DNA is also repaired, but at a much slower rate. Since the bulk of the DNA is not being transcribed, the repair of untranscribed DNA is referred to as global genomic repair (GGR). It is the mechanism by which untranscribed DNA is scanned that we consider here. SIMPLE DIFFUSION Is there really any need to propose that a DNA scanning mechanism is necessary for nontranscribed DNA or could simple diffusion do the job? Aside from the intuitive feeling that the complex structure of chromatin would make diffusion of repair enzymes to damaged DNA slow or impossible, is there any evidence that simple diffusion cannot explain the observations? As will be described in the next section, our results are not compatible with the diffusion hypothesis. They show that premutagenic DNA damage is not repaired in untranscribed DNA in quiescent cells, even though these cells have the necessary repair enzymes. TCR also provides evidence that diffusion is not the limiting factor, for the nontranscribed strand of transcribed genes is usually repaired at the global genomic rate [Mellon et al., 1987; van Hoffen et al., 1993]. Clearly, the chromatin in transcribed regions is relatively open and uncompacted, unlike many untranscribed, condensed, and pycnotic regions. If access to the DNA were a limiting factor, the nontranscribed strand would be repaired faster than the bulk of the DNA, but normally it is not [Nouspikel and Hanawalt, 2002]. This is another indication that simple diffusion of enzymes in the cell, leading to random encounters with damaged bases, is not how DNA damage is repaired. Since RNA polymerase is not transcribing this DNA, it cannot be scanning it either. Thus, the damage must be discovered in some other way. NEW OBSERVATIONS ON TIMING OF DNA REPAIR AND MUTATION Mutation Observations on quiescent Big Blue mouse embryonic fibroblasts led to the polymerase scanning hypothesis. Mutations can be measured in these cells using a laci transgene that can be recovered by in vitro packaging of the l vector into viable phage and tested in bacteria. Mutations that have arisen in the bacteria from damaged DNA give rise to mosaic plaques or are unrecoverable, depending on the nature of the DNA damage [Paashuis- Lew et al., 1997; Bielas and Heddle, 2000, 2003]. Mutations that have arisen in the mouse cells before DNA isolation and packaging give rise to completely mutant nonmosaic plaques (Fig. 1). It has long been hypothesized that DNA damage would be irreversibly fixed as mutations at the time of DNA replication, something made even more likely by the discovery of bypass polymerases. Generally, it has also been assumed that this would be the only time that mutations would be fixed, but that DNA repair would operate at all stages of the cell cycle [Koundrioukoff et al., 2004]. The results with the quiescent mouse cells fit the classic expectation for the fixation of mutations very well: in both transformed [Bielas and Heddle, 2000] and several untransformed cell strains [Bielas and Heddle, 2003], no mutations were formed during 4 days of quiescence in the absence of DNA replication. The results were the same whether the mutagen used was UV or ethylnitrosourea (ENU). When the cells were stimulated to proliferate, however, all or most of the mutations were fixed in approximately one cell cycle [Bielas and Heddle, 2000, 2003, 2004], as shown in Figure 2. It should be noted here that the laci transgene lacks mammalian promoters and a polya adenylation site and is thus unexpressed. The transgene is present in each of the 40 copies of the l vector, which comprises about 2 Mb of heteropycnotic DNA. It is therefore a measure of what ensues at untranscribed DNA in these cells. The lack of formation of mutations during quiescence in such untranscribed DNA was entirely in accord with classic concepts of mutation, but the other results obtained were not. DNA Repair The repair of premutagenic DNA damage in the laci transgene was followed in the quiescent mouse cells using two approaches. For ENU, it could be followed by the formation of mosaic plaques. When phage are treated with ENU directly and then plated on bacteria, the resultant plaques are mosaic as some of the phage arise from the undamaged DNA strand and some from the adducted strand [Paashuis-Lew et al., 1997]. When the quiescent mouse cells were treated and DNA was isolated immediately posttreatment, mosaic mutations accounted for all of the increase in mutation [Bielas and Heddle, 2000, 2003]. Some preexisting spontaneous mutations exist in the cells, so there was a background of completely mutant plaques. Again, this was not surprising. When the cells were maintained at a quiescent stage for up to 4 days, no change in the frequency of mosaic plaques was observed. This shows that the DNA damage leading to these mutations was not repaired over this interval. The result was surely surprising as the quiescent cells had about 75% of the activity of alkyltransferase as proliferating cells and much more time for repair than normal, about 3 4 cell cycles worth. A similar result was obtained for mutations in p53-deficient cells, which differed from normal cells in that they were almost entirely in the G2 phase when made quiescent (data not shown).

3 Unifying Concept of DNA Repair 145 Fig. 1. Upon replating of mutants, their origins can be distinguished. If they were generated by a DNA adduct that existed in the mouse cell, then they will be mosaic, about 50% mutant, which can be shown by replating the plaque. If, in contrast, the mutant represents a mutation that existed in the mouse cell, then both DNA strands will be mutant and the resulting plaque will consist of only mutant phage. Redrawn from Bielas and Heddle [2003]. Experiments with a mutagen that produces an entirely different spectrum of DNA damage, UV light, produced the same pattern of results. No complete mutations were formed during 4 days of quiescence, but many mutations were formed within about one cell cycle after the induction of proliferation (Fig. 2B) [Bielas and Heddle, 2004]. Again, mutations were formed only during proliferation and we infer only during DNA synthesis. Similarly, no premutagenic DNA lesions were repaired during quiescence. In this case, no UV-induced mosaic plaques were recovered, so the repair of premutagenic DNA damage was followed indirectly by inducing proliferation after different periods of quiescence. If the cells were repairing premutagenic DNA damage during quiescence, then the longer the quiescence, the fewer mutations would be formed when proliferation occurred. In fact, the mutation frequency was unaltered by up to 4 days of quiescence, showing that no such repair occurred. Comparable experiments with XPA-deficient cells showed a similar result [Bielas and Heddle, 2004] during quiescence with two exceptions. First, after the induction of proliferation, the XPA-deficient cells fixed more mutations and a resultant higher mutant frequency. Second, the XPA cells, unlike the normal cells, were killed during quiescence

4 146 Heddle and Bielas Fig. 2. Mutations of the laci transgene obtained from DNA of quiescent mouse cells treated at time 0. The cells were induced to proliferate after 4 days. A: Cells treated with ethylnitrosourea. The dotted line indicates mosaic plaques, the solid line, homogeneous plaques. B: Cells treated with UV. The dotted line represents the mutant frequency in cells held in quiescence for the days indicated but assayed after proliferation had occurred. The solid line represents homogeneous plaques. Redrawn from Bielas and Heddle [2003]. by the UV treatment. Since it would be unreasonable to attribute the greater killing of these nucleotide repair-deficient cells to damage in unexpressed genes, the difference in survival is due to a lack of TCR in the XPA cells. The almost complete survival of the normal cells shows their competence to engage in NER via TCR during quiescence. Yet no premutagenic lesions were repaired in the untranscribed transgene during this interval. We concluded that the repair-proficient cells had the enzymes necessary for NER, but could not use them in GGR. It is well known that DNA single- and double-strand breaks can be repaired during G1 and G2 [Kruger et al., 2004], although there is recent evidence that homologous rejoining occurs specifically in S in yeast [Lisby et al., 2003]. Accordingly, it has been almost universally assumed that the longer cells had before DNA synthesis, the fewer number of mutations arise, given that more DNA repair can occur preceding S. The well-known association between proliferation and cancer is often accounted for in this way; however, many other proposals have also been made [Tomlinson et al., 1996; Trosko, 2003]. Hence, it was astonishing to find that mouse cells treated while quiescent did not repair any detectable amount of the DNA damage induced by either of these two quite different mutagens until the cells were stimulated to proliferate. The failure of the mouse cells to fix mutations while quiescent is entirely in accord with the long-standing concept that mutations are created during DNA replication. Recent discoveries of error-prone bypass polymerases, polymerases that step in when the normal polymerase is stumped by a lesion, provide a molecular mechanism for this [Kannouche et al., 2003]. But the failure of the cells to repair premutagenic DNA damage while quiescent is not in accord with classic concepts. The failure of the cells to repair untranscribed DNA is not due to a lack of repair enzymes: the mouse cells have almost as much alkyltransferase when quiescent as when proliferating, and they do repair transcribed genes by NER. Their failure to make use of the time available for DNA repair is difficult to explain by current models. It is noteworthy that in all of the experiments that we have performed, repair and fixation of mutations seem to be completely synchronous with no apparent tendency for the repair of adducts to start or finish earlier than the formation of mutations. Most studies of DNA repair find that GGR is occurring simultaneously with TCR, but at a much slower rate [Hanawalt, 2001]. The results with the untranscribed transgene, however, suggest that there is heterogeneity within the cell population: all cells are performing TCR, but only cells in S are performing GGR. In that case, GGR is naturally slower and could not be complete until all cells have traversed the whole of S. PC REPAIR HYPOTHESIS Since our studies revealed that neither GGR nor mutation fixation occurs in cells that are resting in either G1 or G2, they clearly implicate S-phase as the time for both. Accordingly, we propose that GGR is coupled to replica-

5 Unifying Concept of DNA Repair 147 tion and is very similar to TCR. Together, they might be called polymerase-coupled repair, but the obvious acronym has been taken, so we propose that PC repair is a suitable acronym. Specifically, we propose here that the scanning of the DNA is carried out by polymerase complexes for all repair of damaged bases. We postulate that some repair enzymes are permanently associated with the polymerase complex and may even find the damage before the polymerase encounters it, whereas others are recruited after the complex encounters the damage. We cannot claim that our data are explicable only by a restriction of GGR to S, let alone to the replication sites. It is possible, for example, that quiescence in some way represses GGR. Nevertheless, the results are not consistent with the simple diffusion of repair enzymes to the damage. Moreover, the PC repair hypothesis explains a number of otherwise unexplained observations, eliminates the need for any other mechanisms for scanning the DNA for base damage, and suggests a means by which the scanning of the DNA is both systematic and efficient. In particular, the whole of the genome would be methodically scanned once per cell cycle concurrently with replication. This avoids a race between replication and repair that is between the fixation and the avoidance of a mutation. This could be particularly advantageous at low levels of DNA damage. REPAIR PROCESSES ASSOCIATED WITH DNA REPLICATION There are many reports of an association between proteins involved in DNA repair and those involved in replication or with replicating DNA. So long as unsynchronized cells are used, such as actively growing cultures of bacteria, yeast, or mammalian cells, it would not be apparent that only those lesions encountered by a polymerase complex are repaired. An association between alkyltransferase and DNA polymerase would be much stronger evidence for scanning by DNA polymerase than an association between the enzymes involved in excision repair, since in excision repair the DNA polymerase is required to resynthesize the excised region of the DNA. No such resynthesis is required after the action of alkyltransferase because the original base remains in place and only the alkyl group is removed. Such an association has in fact been reported [Engelbergs et al., 1998; Boldogh et al., 2001]. A similar situation exists for photolyase, an enzyme that uses visible light to split thymine dimers and return them to the original thymines without the removal of either base or nucleotide. Such associations have also been observed [Meier et al., 2002; Suter and Thoma, 2002], and it has been reported by Suter and Thoma [2002] that active ribosomal genes are repaired faster by photolyase than inactive genes. The latter was attributed to an open chromatin confirmation, but it can equally be explained by an association between the polymerase and photolyase. Their finding that DNA can be repaired in the apparent absence of extensive DNA replication is more problematic, but their measurement of DNA replication was very indirect. This interesting approach needs to be repeated with direct measurements of DNA replication. The nature of the association between the replisome and the transcription complex is of interest. Smith et al. [2001] observed in bacteria that the MutS and MutL proteins, which are involved in mismatch repair, colocalized with DNA polymerase when mismatches were induced by chemical treatment, consistent with our hypothesis. We suggest that the alkyltransferases may be a permanent part of the polymerase complexes, unlike excision repair in which the repair enzymes are brought to the DNA damage after the damage has been discovered. Evidence for such an association has been reported for E. coli [Falnes et al., 2002]. Such a mechanism would provide a simple way by which different kinds of DNA damage would be repaired by different mechanisms. If stalling of the polymerase were a common signal, there would have to be some secondary step by which irrelevant repair enzymes were excluded and the appropriate repair enzymes included. Previous studies indicate that very little repair of untranscribed genes occurs outside of the S-phase in cycling cells [Russev and Boulikas, 1992; Lommel et al., 1995]. These studies on actively cycling cells are in agreement with our studies on quiescent cells and suggest that quiescence per se is not the reason for the lack of repair of untranscribed genes. Furthermore, the finding that GGR is inactive even in early S indicates that this is not simply a matter of the phase of the cell cycle, but is related to the time of replication of the genes. Inactive genes are generally replicated late in S, the time that GGR is most active. This finding is exactly what the polymerase scanning hypothesis predicts. It may be that during replication, chromatin structure is relaxed and histones are removed as the replication complex approaches or, perhaps, by the complex itself, so that at this time the DNA might become accessible to diffusing repair enzymes. Alternately, the replication complex may carry repair enzymes to the damage or may initiate a signaling mechanism that recruits repair enzymes to the site. However, it occurs, the results in both cycling and quiescent cells suggest that GGR occurs at the time of replication. DNA REPAIR IN TERMINALLY DIFFERENTIATED CELLS It has been reported that terminally differentiated neurons repair only transcribed genes and not untranscribed DNA [Nouspikel and Hanawalt, 2002]. This result is of course exactly what one would expect if the polymerase scanning hypothesis is correct. In the neurons, the nontranscribed strand of transcribed genes is also repaired; however, this is notinaccordwitheitherpcrepairornormaltcr.itmay

6 148 Heddle and Bielas be simply the result of the release of repair enzymes from the transcribed strand in an area of open chromatin. CONTRARY EVIDENCE There is an enormous body of experimental evidence on DNA repair and some of it is not consistent with our proposal that repair of untranscribed DNA occurs only at the time of replication. It is quite clear, for example, that DNA strand breaks are repaired at all stages of the cell cycle and that chromosomal aberrations can be formed from them during G1 and G2. Hence, we exclude DNA strand breaks from our hypothesis, which is thus concerned with the repair of DNA base damage. It is likely that DNA strand breaks are discovered as a result of the loss of histones or other changes in the chromatin that alert the cellular repair mechanisms to the existence of DNA strand breaks. There is nevertheless other evidence that seems inconsistent with the polymerase scanning hypothesis. Unscheduled DNA synthesis (UDS) must, by definition, occur outside of the S-phase [Painter and Cleaver, 1969], the very time we propose that all GGR occurs. One possible explanation for this is that this repair is entirely TCR. Unfortunately, Cockayne cells that lack TCR still demonstrate UDS [van Hoffen et al., 1993], so this rationalization is inadequate. There are two other possible explanations for this. The first is that the DNA damage repaired by UDS is not premutagenic, which would render its existence undetectable in our experiments that measured mutations. This seems intuitively unlikely. A similar difficulty arises from the fascinating results revealing the differential location of TFIIH in cells after UV irradiation [Volker et al., 2001]. Normal human cells at confluence showed changes in the location of TFIIH, which were also found in CS-B cells that lack TCR, but not in cells that lack GGR. This indicates that this relocation was a part of GGR, something our hypothesis says should not be occurring in confluent cells. Some of the possible explanations for these differences in conclusions are experimental. First, our experiments utilize mouse cells, whereas many other experiments involve human cells. It has been known for some time, and recently emphasized by Hanawalt [2001], that mouse cells are relatively deficient in GGR, probably as the result of a lack of p48 expression. It might be, therefore, that mouse cells lack GGR except in S-phase but that human cells at least are able to repair untranscribed genes at other stages of the cell cycle. Certainly, if that were true, GGR could not be dependent on the DNA replication complex to discover damaged DNA and to initiate repair in human cells. Hanawalt [2001] has emphasized how important it would be if the mechanisms involved in DNA repair differ in such important ways between rodents and humans, given our reliance on rodent bioassays to identify potential human carcinogens. Second, our experiments were conducted at doses that kill few if any normal cells, whereas most experiments on DNA repair have been performed at doses that kill 99% or more of the cells. It may well be, therefore, that the DNA repair measurements are not relevant to the normal experience of cells, which is exposure to low doses. At high doses, there may be the induction of other mechanisms of repair or physiological or pathological effects, including disruption of chromatin, that permit diffusion of repair proteins to damaged sites and lead to the results obtained, such as SOS repair in bacteria. The measurement of mutations has the advantage, over that of any biochemical or cytological assay for repair, of being inclusive of all repair mechanisms that influence mutation of the locus involved, whether it is NER, base excision repair, alkyltransferases, or unknown mechanisms. Finally, some of the assays may be detecting incomplete repair or repair of nonmutagenic lesions. Whatever the reason, it is clear that normal quiescent mouse cells do not repair DNA damage that causes mutation, whether that damage is induced by UV radiation or by ENU exposure. ACKNOWLEDGMENTS J.H.B. is supported by a postgraduate scholarship from the Natural Sciences and Engineering Research Council of Canada (NSERC). REFERENCES Bielas JH, Heddle JA Proliferation is necessary for both repair and mutation in transgenic mouse cells. Proc Natl Acad Sci USA 97: Bielas JH, Heddle JA Elevated mutagenesis and decreased DNA repair at a transgene are associated with proliferation but not apoptosis in p53-deficient cells. Proc Natl Acad Sci USA 100: Bielas JH, Heddle JA Quiescent murine cells lack global genomic repair but are proficient in transcription-coupled repair. DNA Repair (Amst) 3: Boldogh I, Milligan D, Lee MS, Bassett H, Lloyd RS, McCullough AK hmyh cell cycle-dependent expression, subcellular localization and association with replication foci: evidence suggesting replication-coupled repair of adenine:8-oxoguanine mispairs. Nucl Acids Res 29: Ehrenhofer-Murray AE Chromatin dynamics at DNA replication, transcription and repair. Eur J Biochem 271: Engelbergs J, Thomale J, Galhoff A, Rajewsky MF Fast repair of O6-ethylguanine, but not O6-methylguanine, in transcribed genes prevents mutation of H-ras in rat mammary tumorigenesis induced by ethylnitrosourea in place of methylnitrosourea. Proc Natl Acad Sci USA 95: Falnes PO, Johansen RF, Seeberg E AlkB-mediated oxidative demethylation reverses DNA damage in Escherichia coli. Nature 419: Friedberg EC, McDaniel LD, Schultz RA The role of endogenous and exogenous DNA damage and mutagenesis. Curr Opin Genet Dev 14:5 10.

7 Unifying Concept of DNA Repair 149 Fuxreiter M, Luo N, Jedlovszky P, Simon I, Osman R Role of base flipping in specific recognition of damaged DNA by repair enzymes. J Mol Biol 323: Hanawalt PC Controlling the efficiency of excision repair. Mutat Res 485:3 13. Kannouche P, Fernandez de Henestrosa AR, Coull B, Vidal AE, Gray C, Zicha D, Woodgate R, Lehmann AR Localization of DNA polymerases eta and iota to the replication machinery is tightly co-ordinated in human cells. EMBO J 22: Koundrioukoff S, Polo S, Almouzni G Interplay between chromatin and cell cycle checkpoints in the context of ATR/ATM-dependent checkpoints. DNA Repair (Amst) 3: Kruger I, Rothkamm K, Lobrich M Enhanced fidelity for rejoining radiation-induced DNA double-strand breaks in the G2 phase of Chinese hamster ovary cells. Nucl Acids Res 32: Lisby M, Antunez de Mayolo A, Mortensen UH, Rothstein R Cell cycle-regulated centers of DNA double-strand break repair. Cell Cycle 2: Lommel L, Carswell-Crumpton C, Hanawalt PC Preferential repair of the transcribed DNA strand in the dihydrofolate reductase gene throughout the cell cycle in UV-irradiated human cells. Mutat Res 336: Meier A, Livingstone-Zatchej M, Thoma F Repair of active and silenced rdna in yeast: the contributions of photolyase and transcriptioncouples nucleotide excision repair. J Biol Chem 277: Mellon I, Spivak G, Hanawalt PC Selective removal of transcription-blocking DNA damage from the transcribed strand of the mammalian DHFR gene. Cell 51: Nouspikel T, Hanawalt PC DNA repair in terminally differentiated cells. DNA Repair (Amst) 1: Paashuis-Lew Y, Zhang XB, Heddle JA On the origin of spontaneous somatic mutations and sectored plaques detected in transgenic mice. Mutat Res 373: Painter RB, Cleaver JE Repair replication, unscheduled DNA synthesis, and the repair of mammalian DNA. Radiat Res 37: Russev G, Boulikas T Repair of transcriptionally active and inactive genes during S and G2 phases of the cell cycle. Eur J Biochem 204: Smith BT, Grossman AD, Walker GC Visualization of mismatch repair in bacterial cells. Mol Cell 8: Suter B, Thoma F DNA-repair by photolyase reveals dynamic properties of nucleosome positioning in vivo. J Mol Biol 319: Svejstrup JQ Mechanisms of transcription-coupled DNA repair. Nat Rev Mol Cell Biol 3: Tomlinson IP, Novelli MR, Bodmer WF The mutation rate and cancer. Proc Natl Acad Sci USA 93: Trosko JE The role of stem cells and gap junctional intercellular communication in carcinogenesis. J Biochem Mol Biol 36: van Hoffen A, Natarajan AT, Mayne LV, van Zeeland AA, Mullenders LH, Venema J Deficient repair of the transcribed strand of active genes in Cockayne s syndrome cells. Nucl Acids Res 21: Volker M, Mone MJ, Karmakar P, van Hoffen A, Schul W, Vermeulen W, Hoeijmakers JH, van Driel R, van Zeeland AA, Mullenders LH Sequential assembly of the nucleotide excision repair factors in vivo. Mol Cell 8: