Journal of Cell Science 111, 395-404 (1998) 395 Printed in Great Britain © The Company of Biologists Limited 1998 JCS3690

Cells from ERCC1-deficient mice show increased genome instability and a reduced frequency of S-phase-dependent illegitimate chromosome exchange but a normal frequency of homologous recombination

David W. Melton1,*, Ann-Marie Ketchen1, Fátima Núñez1, Stefania Bonatti-Abbondandolo2,3, Angelo Abbondandolo2,4, Shoshana Squires5 and Robert T. Johnson5 1Institute of Cell and Molecular Biology, University of Edinburgh, Edinburgh EH9 3JR, Scotland, UK 2National Institute for Research on Cancer, Genova 16132, Italy 3CNR Institute of Mutagenesis and Differentiation, Pisa 56100, Italy 4Department of Clinical and Experimental Oncology, University of Genova, Genova 16132, Italy 5Department of Zoology, University of Cambridge, Cambridge CB2 3EJ, UK *Author for correspondence (e-mail: [email protected])

Accepted 1 December 1997: published on WWW 15 January 1998

SUMMARY

The ERCC1 protein is essential for nucleotide excision light, was dramatically reduced in both mutants. In rodent repair in mammalian cells and is also believed to be cells the G1 arrest induced by ultra violet light is less involved in mitotic recombination. ERCC1-deficient mice, extensive than in human cells, with the result that with their extreme runting and polyploid hepatocyte nuclei, replication proceeds on an incompletely repaired template. have a phenotype that is more reminiscent of a cell cycle Illegitimate recombination, resulting in a high frequency of arrest/premature ageing disorder than the classic DNA chromatid exchange, is a response adopted by rodent cells repair deficiency disease, xeroderma pigmentosum. To to prevent the accumulation of DNA double strand breaks understand the role of ERCC1 and the link between adjacent to unrepaired lesion sites on replicating DNA and ERCC1-deficiency and cell cycle arrest, we have studied allow replication to proceed. Our results indicate an primary and immortalised embryonic fibroblast cultures additional role for ERCC1 in this process and we propose from ERCC1-deficient mice and a Chinese hamster ovary the following model to explain the growth arrest and early ERCC1 mutant cell line. Mutant cells from both species senescence seen in ERCC1-deficient mice. In the absence of showed the expected nucleotide excision repair deficiency, ERCC1, spontaneously occurring DNA lesions accumulate but the mouse mutant was only moderately sensitive to and the failure of the illegitimate recombination process mitomycin C, indicating that ERCC1 is not essential for the leads to the accumulation of double strand breaks following recombination-mediated repair of interstrand cross links in replication. This triggers the p53 response and the G2 cell the mouse. Mutant cells from both species had a high cycle arrest, mediated by increased expression of the cyclin- mutation frequency and the level of genomic instability was dependent kinase inhibitor p21cip1/waf1. The increased elevated in ERCC1-deficient mouse cells, both in vivo and levels of unrepaired lesions and double strand breaks lead in vitro. There was no evidence for an homologous to an increased mutation frequency and genome instability. recombination deficit in ERCC1 mutant cells from either species. However, the frequency of S-phase-dependent Key words: DNA repair, Genome instability, Recombination, illegitimate chromatid exchange, induced by ultra violet Xeroderma pigmentosum

INTRODUCTION cancer. There are seven complementation groups in classical XP and most of the genes involved have now been isolated. The nucleotide excision repair (NER) pathway has evolved NER genes have also been isolated independently by their principally to deal with u.v. light-induced DNA damage, but is ability to complement u.v. light-sensitive rodent cell lines. also active against a variety of other endogenously generated There is extensive overlap between these ERCC (excision bulky DNA lesions. Components of the NER system are repair cross-complementing) genes and XP genes (for review defective in patients with the rare autosomal recessive disease, see Wood, 1996). The NER pathway has now been xeroderma pigmentosum (XP), which is characterised by reconstructed in vitro using proteins purified from mammalian photosensitivity and a thousandfold increased incidence of skin cells (Aboussekhra et al., 1995). It consists of an initial damage 396 D. W. Melton and others recognition step, followed by an incision to either side of the deficiency, but do show the characteristic sensitivity to u.v. lesion and the removal of the damage-containing light found in XP patients. oligonucleotide. DNA repair synthesis then occurs and a DNA The other possibility to explain the ERCC1-deficient ligase reseals the strand. phenotype that we have considered, and the one favoured by NER does not proceed in isolation, but is precisely co- Weeda and co-workers, is that ERCC1, in common with many ordinated with other essential cellular processes. In addition to of the other NER proteins, may have an additional function and genome-wide NER, there is also preferential repair of the that it is the lack of this function, rather than the NER transcribed DNA strand (Mellon et al., 1987), with the XP-B deficiency, that is responsible for the phenotype. One prime and XP-D proteins being essential components of the RNA candidate for such an additional function for ERCC1 is in polymerase II transcription factor TFIIH (Schaeffer et al., mitotic recombination. This is based on two observations: the 1993). u.v. light, ionising radiation and other DNA damaging homology between ERCC1 and the S. cerevisiae RAD10 agents lead to p53 stabilisation and a cell cycle arrest, during protein, which is involved in both NER and mitotic which the damage may be repaired (Hartwell and Kastan, recombination (Fishman-Lobell and Haber, 1992) and the 1994). This G1 arrest, which induces apoptosis in some cell hypersensitivity of ERCC1 (and ERCC4)-deficient cells, but types (Clarke et al., 1993) and can be prolonged in other cell not other ERCC mutants, to the DNA cross-linking agent, types, even at low doses of ionising radiation (Di Leonardo et mitomycin C. (Interstrand cross links are believed to be al., 1994), prevents cells from initiating DNA replication on a repaired by a process involving mitotic recombination.) still-damaged template, which would lead to the fixation of To improve our understanding of the link between ERCC1- mutations and the accumulation of DNA strand breaks. deficiency and cell cycle arrest, we have carried out studies on ERCC1 was the first mammalian NER gene to be cloned primary and immortalised embryonic fibroblast cultures from (Westerveld et al., 1984). In a complex with XP-F (ERCC4), ERCC1-deficient mice and on a Chinese hamster ovary ERCC1 it makes the incision 5′ to the damage site. Although ERCC1 mutant cell line. ERCC1-deficient cell lines from both species is essential for NER in vitro, an ERCC1 defect has not been showed the expected hypersensitivity to u.v. light, but the identified in XP, or indeed in any known human disease. To mouse mutant was considerably less sensitive to mitomycin C study the consequences of ERCC1-deficiency in a whole than its hamster equivalent. Both ERCC1-deficient cell lines animal system, we used gene targeting to inactivate the ERCC1 had a mutation frequency 100-times higher than the species gene (Selfridge et al., 1992). Our ERCC1-deficient mice were controls. ERCC1-deficient mouse embryos, primary fibroblast severely runted and died before weaning with an unusual liver cultures and immortalised cell lines all had a higher level of pathology (McWhir et al., 1993). At birth, some hepatocyte genome instability than wild-type controls. Using both a nuclei were already enlarged and polyploid; by three weeks of plasmid recombination and a gene targeting assay, we were age, the frequency of cells with enlarged nuclei had increased unable to detect an homologous recombination deficit in and nuclei with both polyploid and aneuploid DNA content ERCC1-deficient cells from either species. However, the were apparent. These changes were associated with increased frequency of S-phase-dependent illegitimate chromosome levels of p53 protein in some hepatocyte nuclei. ERCC1 exchange induced by u.v. light was dramatically reduced in knockout mice and animals with a C-terminal ERCC1 both mutants. Primary fibroblasts from ERCC1-deficient mice truncation have also been generated independently (Weeda et showed the same growth rate, cell cycle distribution and p53 al., 1997). In addition to the same runting and liver pathology, response following u.v. light as control cultures, with no these animals, which survived for up to six months, also indication of the polyploid cells found in hepatocytes in vivo. showed additional changes to the skin, spleen and kidneys. We, and Weeda and co-workers, have suggested that the phenotype seen in ERCC1-deficient mice is more reminiscent MATERIALS AND METHODS of the consequences of cell cycle arrest and premature ageing than NER deficiency per se. This view is reinforced by the Cells striking similarity between the ERCC1-deficient phenotype Primary fibroblasts were isolated from individual day 10.5 or 11.5 and that reported for transgenic mice overexpressing the mouse embryos by a combination of maceration and trypsinisation. cip1/waf1 cyclin-dependent kinase inhibitor, p21 , in hepatocytes The resulting cell suspension was plated into 60 or 90 mm culture (Wu et al., 1996). p21cip1/waf1 is one of the main effectors of dishes, with cultures becoming confluent within 3 days. p53, being induced as part of the response to DNA damage, Spontaneously immortalised mouse fibroblast cell lines were isolated leading to cell cycle arrest at the G1/S checkpoint (el-Deiry et from senescing primary cultures. PF20 arose from a wild-type al., 1993). Furthermore, cells lacking (Waldman et al., 1996), embryo; PF18 arose from an HPRT-deficient embryo; PF24 arose cip1/waf1 or overexpressing, p21 (Wu et al., 1996) arrest in G2 from an embryo that was both ERCC1-deficient and HPRT-deficient. and show uncoupling of S phase and mitosis. ERCC1 genotyping of embryos and resulting cell lines was carried We have speculated that, in the obvious absence of any u.v. out as described (McWhir et al., 1993). The Chinese hamster ovary light-induced DNA damage, the cell cycle arrest in ERCC1- cell line, CHO-9, and its ERCC1 mutant derivative, 43-3B (Wood and Burki, 1982), were kindly provided by Rick Wood (ICRF, Clare Hall). deficient mice might result from the inability of these mice to use the NER pathway to repair endogenously generated Cell culture and HPRT selection (oxidative) DNA damage (McWhir et al., 1993). In the interim, Cells were grown in Glasgow-modified Eagle’s medium, XP-A (de Vries et al., 1995; Nakane et al., 1995) and XP-C supplemented with foetal calf serum (10% for primary cultures, 7% (Sands et al., 1995) knockout mice, have been generated with for other cell lines). Selection for HPRT expression was in HAT a phenotype that is not consistent with this hypothesis. They medium (Littlefield, 1964); selection against HPRT expression was in do not exhibit the growth arrest characteristic of ERCC1- medium containing 5 µg/ml 6-thioguanine (6-TG). Phenotype of ERCC1-deficient cells 397

Survival assays excised from PGK/pDWM1 as an EcoRI fragment and cloned into the Survival following u.v. (254 nm) irradiation or 24 hours exposure to EcoRI site of pBluescript II SK(+) (Stratagene), to give pBT/PGK- mitomycin C was determined as described (McWhir et al., 1993). For HPRT(RI). the u.v. survival, cells were plated at 2×104/30 mm dish. For the The two substrates for the plasmid homologous recombination mitomycin C survival, cells were plated in 24-well microtitre plates assay were derived from PBT/PGK-HPRT(RI). A 5′ deletional at 104/well. derivative, pBT/PGK-HPRT (XhoI∆), was constructed by excising the 3′ 1.9 kb of the minigene as an XhoI (site within exon 3)-EcoRI (site Cell cycle distribution at 3′ end of minigene) fragment and recloning it into pBluescript II Nuclei from primary mouse fibroblasts were prepared for flow SK (+) cut with XhoI/EcoRI. A 3′ deletional derivative, pBT/PGK- cytometry and stained with propidium iodide as described (Vindelov HPRT (ScaI-EcoRV∆), was constructed by restricting pBT/PGK- et al., 1983). DNA content was determined by FACScan (5×104 nuclei HPRT (RI) with ScaI (site within exon 8) and EcoRV (site at 3′ end analysed/sample). To identify cells in S-phase, cultures growing on of minigene), to liberate the 3′ end of the minigene, and then religating glass coverslips, or in 30 mm dishes, were pulsed for 2 hours in the plasmid. The deletional derivatives are both non-functional, but medium containing 10 µM bromodeoxyuridine (BrdU). Cells were share a 0.6 kb region of homology, extending from the XhoI site in then fixed (methacarn for coverslips, cold ethanol for dishes), exon 3 to the ScaI site in exon 8. Homologous recombination between processed and incubated with an anti-BrdU monoclonal antibody these regions of the two plasmids will generate a functional minigene. (Harlan, Sera-Lab). Detection was with a biotinylated secondary Plasmids were introduced into HPRT-deficient cultured cells by the antibody followed by a streptavidin/biotin complex/horseradish calcium phosphate method (Graham and Van der Eb, 1973), with HAT peroxidase conjugate (Dako). Positively stained (i.e. S-phase) nuclei selection for HPRT expression (Littlefield, 1964) imposed 24 hours were scored by eye following photography of the cultures. later. p53 determination Gene targeting assay Total protein lysates were made from primary mouse embryonic The basis of the HPRT gene targeting assay has been described fibroblast cultures as described (Lasko et al., 1990) and samples were (Thompson et al., 1989). The HPRT-deficient mouse fibroblasts used run on 8% SDS-PAGE gels. Gels were western blotted using a semi- here have the same defective HPRT allele, where a spontaneous dry horizontal system (Bio-Rad). p53 was detected using a p53 (Ab- deletion has removed the promoter region and exons 1 and 2. The 7 sheep polyclonal) antibody kit (Oncogene Research Products) with HPRT correction vector, pDWM102, can be used to correct this a biotinylated secondary anti-sheep antibody and a streptavidin/biotin deficiency by homologous recombination. The vector contains the 5′ complex/horseradish peroxidase conjugate (Dako). end of the HPRT gene, comprising the promoter and exons 1-3. The vector is cut with XhoI to liberate plasmid sequences prior to Chromosome preparation electroporation. The XhoI site is within a region of homology between Two days before the experiment cells were plated into dishes to ensure the vector and the target locus. Insertion of the vector into the target vigorous growth at the time of u.v. irradiation. For u.v. irradiation cells locus will restore gene function, with correctant clones being directly were scraped from the perimeter of each dish and removed with the selected in HAT medium. Random integrations of the vector will not medium, phosphate buffered saline added, and the remaining cells yield HATR survivors. Conditions for electroporation were as exposed to u.v. in situ (germicidal lamp emitting predominantly at 260 described (Thompson et al., 1989), with HAT selection added 24 nm). The cells were not further disturbed by replating after u.v. and hours later. metaphases were collected after 14, 21 and 30 hours following incubation with 0.04 µg/ml nocodozole (Sigma). Hypotonic treatment, fixation and preparation of chromosome spreads were RESULTS carried out as described by Gimenez-Abian et al. (1995), and the chromosomes were stained with Crystal Violet. Aberrations were ERCC1-deficient immortalised mouse fibroblasts are recorded in approximately 50 cells per sample by microscopic observation. hypersensitive to u.v. light but only moderately sensitive to mitomycin C Detection of micronuclei We have previously reported that primary fibroblasts, isolated Cells growing on glass coverslips were treated in mild hypotonic from day 10.5 ERCC1-deficient mouse embryos, were solution (0.81% NaCl + 0.056% KCl; Iskandar, 1979) for 10 minutes hypersensitive to u.v. light with virtually no demonstrable before being fixed in 90% methanol. Centromere-containing nucleotide excision repair, but were only moderately sensitive micronuclei were identified as described previously (Nuesse et al., to the DNA cross-linking agent, mitomycin C (McWhir et al., 1989) by treating coverslips with CREST antiserum (Antibody Inc., 1993). Similar observations have been made on primary Davis, CA, USA). FITC-conjugated goat anti-human IgC (Sigma) was used as the secondary antibody. Micronuclei and nuclei were embryonic fibroblasts from independently generated ERCC1 counterstained with propidium iodide. The numbers of CREST+ and knockout mice and mice with a C-terminal ERCC1 truncation CREST− micronuclei present (expressed/103 cells) were determined (Weeda et al., 1997). To permit extended duration assays on by direct observation under a fluorescence microscope at ×1,000 cultured ERCC1-deficient mouse cells, we isolated magnification. spontaneously immortalised lines from ERCC1-deficient (PF24) and wild-type (PF20) primary embryonic fibroblast Plasmid homologous recombination assay cultures. The immortalised ERCC1-deficient mouse line The construction and expression of the functional mouse HPRT retained the expected hypersensitivity to u.v. light, with a very minigene used for this assay, pBT/PGK-HPRT(RI), have been similar survival curve to the ERCC1-deficient Chinese hamster described (Magin et al., 1992). In the referenced publication this minigene was described as PGK/pDWM1. The 2.7 kb minigene is ovary cell line, 43-3B (see Fig. 1). Control mouse (PF20) and under the control of the mouse phosphoglycerate kinase (PGK-1) gene hamster (CHO-9) cell lines gave the same response to promoter. The minigene contains part of the HPRT 5′ untranslated mitomycin C, with a D50 value (concentration of mitomycin C region, the entire coding region and 3′ untranslated region, with the for 50% survival) of 310 ng/ml for PF20 and 325 ng/ml for coding region being interrupted by introns 7 and 8. The minigene was CHO-9 (see Fig. 2). The hamster ERCC1 mutant cell line (43- 398 D. W. Melton and others

Mouse wt. 100 100 Hamster wt. Hamster wt.

Mouse wt.

Hamster mut. 10 10 Mouse -/- Survival (%) Survival (%)

Mouse -/- Hamster mut. 1 1 0 2 4 6 8 10 12 0 100 200 300 400 500 [Mitomycin C] (ng/ml) u.v. dose (Jm−2) Fig. 1. ERCC1-deficient cells are hypersensitive to u.v. light. The u.v. Fig. 2. ERCC1-deficient mouse fibroblasts are only moderately survival curves for immortalised wild-type (PF20) and ERCC1- sensitive to mitomycin C. The survival curves following 24 hours’ deficient (PF24) mouse fibroblasts and wild-type (CHO-9) and exposure to mitomycin C are shown for immortalised wild-type ᭿ ᭹ ERCC1 mutant (43-3B) Chinese hamster ovary cells are shown. Each (PF20; ) and ERCC1-deficient (PF24; ) mouse fibroblasts and ᮀ ᭺ point is the mean of two separate determinations. ᭿, mouse wild wild-type (CHO-9; ) and ERCC1 mutant (43-3B; ) Chinese type; ᭹, mouse ERCC1 −/−; ᮀ, hamster wild type; ᭺, hamster hamster ovary cells. Each point is the mean of two separate ERCC1 mutant. determinations.

3B) was 40-fold more sensitive to mitomycin C (D50 7.5 ng/ml) prohibited the same simple comparison of HPRT mutation than the control hamster cell line, while the mouse ERCC1- frequency used on the CHO cell lines. However, the HM-1- deficient cell line (PF24) was only 8-fold more sensitive (D50 derived HPRT null allele can be corrected by gene targeting to 41 ng/ml) than its species control. restore function to the endogenous HPRT gene (see later Thus, immortalised cell lines with the predicted nucleotide section). HPRT correction was carried out in ERCC1-deficient excision repair deficiency can be isolated from ERCC1- (PF24) cells and in another immortalised mouse embryonic deficient mouse embryos for further analysis. The different fibroblast line (PF18), which was ERCC1 wild type and HPRT- response of ERCC1-deficient mouse and hamster cells to deficient. The spontaneous frequency of 6-TGR was mitomycin C suggests that an alternative (i.e. non-ERCC1- determined in independent HPRT correctants of PF18 and dependent) pathway for the repair of DNA interstrand cross- PF24 (see Table 1). The mean 6-TGR frequency for the PF24 links may operate in mouse, but not hamster cells. correctants was 180-fold higher than for their PF18 equivalents. Thus, both ERCC1-deficient mouse and hamster ERCC1-deficient mouse and Chinese hamster cells cell lines have a high spontaneous mutation frequency. have a high spontaneous mutation frequency The frequency with which mutants at the X-chromosome- ERCC1-deficient mouse fibroblasts show increased linked hypoxanthine phosphoribosyltransferase (HPRT) locus genomic instability arise in cell cultures provides an indication of the mutation rate. The high spontaneous HPRT mutation frequency found in HPRT mutants are conveniently identified by their ability to ERCC1-deficient cells could arise from local events, such as grow in the presence of the purine analogue, 6-thioguanine (6- point mutations or deletions, and also from genome-wide TG). The spontaneous frequency of 6-TGR was greater than events, such as chromosome breakage and loss. To assess the 90-fold higher in the CHO ERCC1 mutant (43-3B) than in its latter events, the frequency of micronuclei was determined in parental wild-type line (CHO-9) (see Table 1). An elevated both primary and immortalised, wild-type and ERCC1- mutation rate for ERCC1 mutant cells (Wood and Burki, 1982) deficient, mouse fibroblast cultures. Micronuclei were divided and for Chinese hamster cells where the ERCC1 gene has been into two categories: those containing whole chromosomes (or, inactivated by gene targeting (Rolig et al., 1997) have been more rarely, centromeric fragments), which stained with previously reported. CREST antibodies (directed against kinetochores), and arose The immortalised ERCC1-deficient mouse fibroblast line principally from chromosome loss; and those containing (PF24) was isolated from an HPRT-deficient embryo. This was acentric fragments, which were unstained by CREST because ERCC1 gene targeting was originally carried out in the antibodies, and arose from chromosome breakage. HPRT-deficient embryonic stem (ES) cell line, HM-1 Primary fibroblasts were isolated from day 11.5 control and (Selfridge et al., 1992), with the result that the HM-1-derived ERCC1-deficient embryos and scored for micronuclei 24 hours HPRT null allele is segregating in our ERCC1 knockout mouse later. At this time, cells had not yet begun to divide in vitro, so stock. While the HPRT deficiency was ideal for a comparison any micronuclei present will have been formed during the of homologous recombination frequency between wild-type previous in vivo cell division. The frequency of micronuclei and ERCC1-deficient mouse cell lines (see later section), it was significantly higher in the ERCC1 mutant cultures than the Phenotype of ERCC1-deficient cells 399

Table 1. The spontaneous frequency of HPRT mutations is immortalised wild-type and ERCC1-deficient cultures, with the elevated in immortalised ERCC1-deficient cells* total number of micronuclei being 3-fold higher in mutant than 6-thioguanineR control cultures. Both chromosome loss and chromosome Genotype Cell line colonies/105 cells† breakage were significantly elevated in the mutant cell line, with the elevation in chromosome breakage (5-fold) being Hamster wild type CHO-9 0.3±0.3 (n=10) Hamster ERCC1 mutant 43-3B 23.3±2.6 (n=10)[93×]‡ particularly pronounced. Mouse wild type PF18/DWM102#1 4.9±1.3 (n=10) The effect of u.v. light on the frequency of micronuclei in PF18/DWM102#2 1.7±0.5 (n=10) pre-crisis primary mouse fibroblast cultures was also PF18/DWM102#3 3.2±2.1 (n=10) determined (see Table 3). From 48-72 hours after u.v. light, the Mean 3.2 Mouse ERCC1 −/− PF24/DWM102#1 1592±152 (n=2) frequency of micronuclei was significantly elevated for both PF24/DWM102#2 61.7±4.5 (n=3) irradiated wild-type and ERCC1-deficient cultures, compared PF24/DWM102#4 198±15.1 (n=3) to non-irradiated controls. The increase in total micronuclei PF24/DWM102#5 952±48 (n=2) following u.v. light was significantly higher for ERCC1- PF24/DWM102#6 85.7±7.9 (n=5) deficient than wild-type cultures, with both chromosome loss Mean 577.9 [179×]‡ and chromosome breakage being significantly elevated. *Cultures were maintained in HAT medium to prevent the accumulation of Thus, developing ERCC1-deficient mouse embryos have a pre-existing HPRT mutants, before being transferred to non-selective medium significantly higher spontaneous incidence of genome for 8 days to allow HPRT mutants to accumulate. Cells were then plated (105 instability than control embryos. This difference becomes cells/90 mm dish) in 6-thioguanine (5 µg/ml)to determine the number of HPRT mutants. progressively more pronounced on extended primary culture †For each cell line, the number of 6-thioguanineR colonies/105 cells is and in immortalised cell lines. u.v. light also causes increased given, ± s.d.; n is the number of individual determinations. The mean genome instability in ERCC1-deficient cultures compared to frequency of 6-TGR colonies is also given for the PF18 and PF24 correctant controls. series. ‡The increase in the frequency of HPRT mutants in ERCC1-deficient cells, The homologous recombination frequency is not relative to the wild-type species control, is given in square brackets. reduced in immortalised ERCC1-deficient cells The homology between ERCC1 and the S. cerevisiae RAD10 controls (see Table 2), the difference being confined to protein, which is involved in mitotic recombination, and the CREST+ micronuclei, arising from chromosome loss. The mitomycin C sensitivity of ERCC1-deficient cell lines, has led spontaneous frequency of micronuclei was also determined to the suggestion that the phenotype of ERCC1-deficient mice after extensive in vitro cell division, but prior to the could arise from defective mitotic recombination, rather than characteristic crisis that primary cultures undergo. The defective nucleotide excision repair (Weeda et al., 1997). To frequency of micronuclei was higher in this pre-crisis material address this question, we have used two assays to compare the than in the day one cultures (see Table 2), with the frequency frequency of homologous recombination in ERCC1-deficient again being significantly higher in ERCC1-deficient cultures and control mouse and hamster cells. than in controls. In this case, both chromosome loss (CREST+ The first assay measured the ability for homologous micronuclei) and chromosome breakage (CREST− recombination between two plasmids, introduced into the same micronuclei) were significantly elevated in the mutant cultures. cell, using an HPRT selection assay. The HPRT-deficient mouse A still higher frequency of micronuclei was found in the cell lines PF18 (ERCC1 wild-type) and PF24 (ERCC1-deficient)

Table 2. The spontaneous frequency of micronucleus formation is higher in ERCC1-deficient mouse primary fibroblasts and immortalised fibroblasts* Cell Cells Total micronuclei CREST+ MN CREST− MN Genotype culture scored no. (MN/103 cells) no. (MN/103 cells) no. (MN/103 cells) Wild type Day 1 11,132 197 (17.7) 122 (11.0) 75 (6.7) ERCC1 −/− Day 1 14,698 372 (25.3) 250 (17.0) 122 (8.3) P<0.0001† P<0.0001 n.s.‡ Wild type Pre-crisis 6,161 183 (29.7) 114 (18.5) 69 (11.2) ERCC1 −/− Pre-crisis 9,922 451 (45.5) 262 (26.4) 189 (19.0) P<0.0001 P<0.0001 P<0.0001 Wild type Immortalised 4,214 157 (37.3) 122 (29.0) 35 (8.3) ERCC1 −/− Immortalised 3,598 426 (118.4) 281 (78.1) 145 (40.3) P<0.0001 P<0.0001 P<0.0001

*The total number of micronuclei, and the number of CREST+ micronuclei (indicating chromosome loss) and CREST− micronuclei (indicating chromosome breakage), were determined for primary fibroblasts isolated from 10.5 day wild-type and ERCC1-deficient embryos. The determination was made 1 day after plating (before in vitro cell division) and after in vitro cell division had commenced, on pre-crisis cultures at passage 3. The day 1 data come from 4 wild-type and 7 ERCC1-deficient embryos, the pre-crisis data are from single wild-type and ERCC1-deficient primary cultures. The determination was also made on immortalised wild-type (PF20) and ERCC1-deficient (PF24) fibroblasts. †χ2-test on frequency of micronucleus formation, ERCC1 −/− versus wild-type control. ‡n.s., not significant. 400 D. W. Melton and others

Table 3. The u.v.-induced frequency of micronucleus formation is higher in ERCC1-deficient primary fibroblasts* U.V. dose Cells Total micronuclei CREST+ MN CREST− MN Genotype (Jm−2) scored no. (MN/103 cells) no. (MN/103 cells) no. (MN/103 cells) Wild type 0 6,161 183 (29.7) 114 (18.5) 69 (11.2) Wild type 2.5 3,408 206 (60.4) 98 (28.8) 108 (31.7) ERCC1 −/− 0 9,922 451 (45.5) 262 (26.4) 189 (19.0) ERCC1 −/− 2.5 5,150 535 (103.8) 282 (54.8) 253 (49.1) P<0.0001† P<0.0001 P<0.0001

*The frequency of micronuclei was determined for control primary cultures (pre-crisis at passage 3) of wild-type (PF20) and ERCC1-deficient (PF24) mouse fibroblasts and for cultures 48-72 hours after 2.5 Jm−2 of u.v. irradiation. †χ2-test on u.v.-induced frequencies (control subtracted), ERCC1 −/− (60.4−29.7 = 30.7) versus wild type (103.8−45.5=58.3). and non-reverting HPRT-deficient derivatives of the hamster cell 9 and 43-3B because they do not contain the equivalent HPRT lines CHO-9 (ERCC1 wild-type) and 43-3B (ERCC1-deficient) mutation. The correction vector was linearised prior to gene were used. HPRT-deficient cells incorporating the functional transfer. This was essential, despite wishing to investigate the HPRT minigene, pBT/PGK-HPRT(RI), can be selected in HAT possible requirement for the endonuclease function of ERCC1 medium. Two non-functional but non-overlapping deletional in homologous recombination, in order to obtain a measurable derivatives of the minigene were used. No HATR colonies were gene targeting frequency. obtained when either minigene was introduced singly into any In Table 4, the frequency of HPRT gene targeting is of the HPRT-deficient cell lines. However, if co-introduced, expressed as the frequency of HATR colonies obtained with the homologous recombination between the plasmids would correction vector, relative to the frequency of HATR colonies reconstruct a functional minigene and HATR colonies would obtained in parallel experiments with a functional minigene. arise. The plasmids were introduced in supercoiled form with This introduces an important correction for any differences in the presumption that any ERCC-1 requirement for homologous gene transfer frequency between the cell types. The same gene recombination might involve its endonuclease function. The targeting frequency (1.6 × 10−5) was obtained for both control frequency of homologous recombination observed between and ERCC1-deficient cell types. plasmids in the four cell lines gave no indication of an Thus, by both of these assays, there was no evidence for any homologous recombination deficit in ERCC1-deficient mouse or homologous recombination deficit in ERCC1-deficient cells. hamster cells (data not shown). The second assay for homologous recombination measured ERCC1-deficient primary mouse fibroblasts have a the frequency of HPRT gene targeting in the HPRT-deficient normal cell cycle distribution, growth rate and p53 mouse cell lines PF18 (wild-type) and PF24 (ERCC1- response deficient). Both contained the HM-1-derived HPRT null allele, The runted growth and polyploid liver cell phenotype in our which is deleted for the 5′ end of the gene (Thompson et al., ERCC1-deficient mice was more reminiscent of defective cell 1989). We have previously used an HPRT correction vector, cycle progression and premature ageing than a simple pDWM102, to correct this allele in ES cells (Thompson et al., consequence of DNA repair deficiency. This impression was 1989). The same procedure was applied to PF18 and PF24. It reinforced by the very similar phenotype reported for could not be used on the HPRT-deficient derivatives of CHO- transgenic mice overexpressing the cyclin-dependent kinase inhibitor, p21cip1/waf1, in the liver (Wu et al., 1996). A study on primary embryonic fibroblasts was carried out to see if the Table 4. The frequency of homologous recombination is profound in vivo abnormality associated with ERCC1 not reduced in immortalised ERCC1-deficient mouse cells* deficiency was amenable to an in vitro analysis. Homologous recombination (gene targeting) Primary cultures isolated from day 11.5 wild-type and Genotype frequency† ERCC1-deficient mouse embryos had the same growth rate, with a generation time of ~36 hours (data not shown). Following a 2 Mouse wild type 1.6×10−5±1.2×10−5 (n=4) − − × −5 × −5 hour pulse with BrdU, to identify cells in S-phase, 42±5% of Mouse ERCC1 / 1.6 10 ±1.2 10 (n=4) cells from two independent wild-type cultures showed nuclei *HPRT-deficient cells, an HPRT minigene and an HPRT gene correction stained with an anti-BrdU antibody. The corresponding value for (gene targeting) vector were used in this study, with selection in HAT medium two independent ERCC1-deficient cultures of the same passage for HPRT expression to identify homologous recombination events. The wild- (passage 2) was 44±7%, confirming the similar cell cycle type (PF18) and ERCC1-deficient (PF24) immortalised mouse fibroblast characteristics. Following an extended 2 day pulse with BrdU, lines, which were isolated from HPRT-deficient embryos, were used. 90% of cells from the same wild-type cultures showed stained †The gene targeting frequency was determined by introducing the linearised HPRT correction vector pDWM102 into HPRT-deficient mouse nuclei. The corresponding value for the equivalent ERCC1- cells and counting the number of HATR colonies arising as a result of deficient cultures was 87%, indicating that, in both cultures, homologous recombination between the correcting vector and the most of the cells are actively cycling. chromosomal HPRT gene. This was expressed relative to the number of FACS analysis of wild-type and ERCC1-deficient cultures at HATR colonies obtained with the minigene pBT/PGK-HPRT(RI) in each cell line. For each determination, 1.5×107 cells were electroporated with 170 µg passage 2 showed the same cell cycle distribution (see Fig. 3). of DNA (linearised pDWM102 or supercoiled pBT/PGK-HPRT[RI]) and Cultures of both genotypes senesced rapidly and, by passage plated into 90 mm dishes (1×106 cells for gene targeting, 3×104 for gene 3, a discrete population of (presumably apoptotic) cells, with transfer). a sub-G1 DNA content, was present to an equal extent. Phenotype of ERCC1-deficient cells 401

ERCC1 +/+ ERCC1 −/−

Fig. 4. ERCC1-deficient primary mouse fibroblasts have normal levels of p53 protein and show a normal p53 response to u.v. light. Wild-type and ERCC1-deficient primary mouse fibroblasts (passage 2) were grown to 70% confluency and u.v. irradiated (50 Jm−2). 6 hours later, cell lysates were made and analysed by SDS-PAGE and western blotting for p53 protein levels. Control cultures were mock- irradiated, but otherwise treated as irradiated samples. Lysates from 1×105 and 5×104 cells were loaded onto the gel for each sample. The location of p53 protein is indicated by the arrow. Std., indicates the −3 Fig. 3. ERCC1-deficient primary mouse fibroblasts have a normal mobility of marker proteins (Mr×10 ) on the same gel. Some cross cell cycle distribution. Flow cytometry profiles of the frequency reactivity to higher molecular mass proteins is characteristic of this distribution of DNA content in nuclei from cultures of wild-type p53 antibody. (ERCC1 +/+) and ERCC1-deficient (ERCC1 −/−) primary mouse fibroblasts at passages 2 and 3 are shown. The positions of nuclei mutations affected another recombination pathway with DNA content corresponding to G0/G1, S and G2/M stages of the responsible for the production of S-phase-dependent cell cycle are shown for each profile. , denotes the presence of a * chromatid exchanges after u.v.-induced DNA damage. In sub-G0/G1 peak, present in passage 3 cultures of both genotypes, indicating the appearance of apoptotic cells in the culture. Note that hamster cells chromatid exchanges are a common there is no indication of the presence of polyploid cells in the consequence of u.v. irradiation (Bender et al., 1973; Darroudi ERCC1-deficient cultures. and Natarajan, 1985), or after treatment with camptothecin, the S-phase-specific cytotoxic inhibitor of DNA topoisomerase I (Degrassi et al., 1989; Ryan et al., 1994). Significantly, there was no indication in the ERCC1-deficient Through the action of DNA replication both u.v. irradiation cultures of cells with a polyploid DNA content that would and camptothecin induce the formation of DNA double strand correspond to the in vivo situation in ERCC1-deficient liver. breaks predominantly in the replicating fraction (Wang and Elevated levels of p53 protein were detected in isolated Smith, 1986; Musk et al., 1990; S. Squires, J. A. Coates and individual nuclei from ERCC1-deficient mice by R. T. Johnson, unpublished data; Avemann et al., 1988; Ryan immunohistochemistry (McWhir et al., 1993). To investigate et al., 1991, 1994), and these breaks are the likely precursors the p53 response in vitro, the level of p53 protein present in of the S-phase-dependent strand exchanges necessary for the control cultures of wild-type and ERCC1-deficient primary development of chromatid exchanges. mouse embryonic fibroblasts was determined by western The frequency of these S-phase-dependent chromatid blotting. Contrary to the in vivo situation, p53 was readily exchanges, and also the frequency of chromatid breaks, were detectable in wild-type extracts and was not elevated in determined in immortalised wild-type and ERCC1 mutant ERCC1-deficient extracts (see Fig. 4). Both cultures showed mouse and hamster cell lines (see Table 5). In agreement with the classic stabilisation of p53 response following u.v. the frequency of micronuclei, following equitoxic doses of u.v. irradiation. The kinetics of the p53 response were equivalent light, the number of chromatid breaks/cell was higher in both in the two cell types (data not shown). ERCC1 mutant cell lines than their wild-type species controls. Thus, the G2 cell cycle arrest seen in ERCC1-deficient However, the number of chromatid exchanges/cell was hepatocytes in vivo is not reproduced in untreated cultured considerably (more than 6-fold) reduced in both mutant lines, primary fibroblasts and there is no evidence for an altered p53 to the extent that the ratio of chromatid exchanges/breaks was response. over 10-fold greater for both wild-type cell lines compared to the ERCC1 mutants. The frequency of S-phase-dependent chromatid This indicates that the ERCC1 protein may be playing an exchange caused by u.v. light is reduced in important role on replicating DNA after damage, promoting the immortalised ERCC1-deficient cells illegitimate recombination reaction between S-phase Having failed to detect a homologous recombination deficit in chromosomes, and thus limiting the frequency of long-lived ERCC-1 mutant cells, we investigated whether ERCC-1 DNA breaks which give rise to chromatid breaks. 402 D. W. Melton and others

Table 5. The frequency of S-phase-dependent chromatid exchange caused by u.v. light is reduced in immortalised ERCC1- deficient cells* Total chromosome Exchanges Total chromatid Breaks Exchanges Genotype Metaphases scored exchanges /cell breaks /cell /breaks Mouse wild type 51 176 3.45 25 0.49 7.04 Mouse ERCC1 −/− 64 31 0.48 46 0.72 0.67 Hamster wild type 58 94 1.6 15 0.26 6.15 Hamster ERCC1 mutant 56 15 0.27 42 0.75 0.36

*Exponentially growing cultures of immortalised cell lines were irradiated with equitoxic doses of u.v. and the subsequent metaphases were collected and scored. For the wild-type hamster (CHO-9) and mouse (PF20) cell lines, the u.v. dose was 6.5 Jm−2; for the ERCC1-deficient hamster (43-3B) and mouse (PF24) cell lines, the u.v. dose was 1.5 Jm−2.

DISCUSSION evidence for any homologous recombination deficit in ERCC1- deficient mouse and hamster cells and none has been reported We have undertaken studies on cultured ERCC1-deficient cells by Weeda and co-workers. This does not mean that ERCC1 is to investigate whether it is a deficiency in nucleotide excision not involved in some homologous recombination pathway, repair, in homologous recombination, or in some additional merely that it is not essential for homologous recombination. function of ERCC1, that is responsible for the unusual growth Our data on mitomycin C sensitivity are compatible with this retardation/cell cycle arrest/premature ageing phenotype seen explanation. Indeed, the S. cerevisiae ERCC1 homologue, in ERCC1-deficient mice. We also sought to develop a cell RAD10, is also involved in, but not essential for mitotic culture system to study the mechanism of this cell cycle arrest. recombination (Schiestl and Prakash, 1990). The relative Both primary (McWhir et al., 1993; Weeda et al., 1997) and contribution of the NER and mitotic recombination deficits to immortalised (this report) fibroblasts derived from ERCC1- the ERCC1-deficient phenotype will be resolved when the deficient mouse embryos were hypersensitive to u.v. light, with types of lesions which are believed to accumulate in ERCC1- virtually no detectable NER. Although ERCC1 mutant Chinese deficient cells are identified directly. hamster ovary cells were hypersensitive to the DNA cross We did find evidence for a profound alteration in S-phase- linking agent, mitomycin C, ERCC1-deficient mouse cell lines dependent illegitimate recombination in ERCC1-deficient were only moderately sensitive, indicating the existence of a cells, which we take to indicate an additional function for the second, non-ERCC1-dependent, homologous recombination ERCC1 protein. Despite the arrest of the cell cycle in hamster pathway for the repair of DNA interstrand cross links. cells by u.v. (Orren et al., 1995), the induced delay is much Much of the ERCC1-deficient phenotype, both in vivo and less extensive than with human cells (Wang and Ellem, 1994; in vitro, can potentially be attributed to the lack of these two S. Squires, J. F. Gimenez-Abian, R. T. Johnson, unpublished repair functions. Spontaneous levels of endogenously data). One consequence of this difference is that replication of generated oxidative damage in mammalian cells are constantly damaged S-phase cells recovers and proceeds more strongly in producing bulky DNA lesions. In addition, metabolic by- hamster than in human cells on an incompletely repaired products, such as malondialdehyde, cause interstrand cross- template (Waters, 1979; Spivak and Hanawalt, 1992). links (Chaudhary et al., 1994). Both types of damage are However, it is likely that the nascent DNA double strand breaks normally dealt with in repair-proficient cells by NER and a that are generated at the sites of transient replication arrest in repair system involving homologous recombination. In rodent cells are rapidly resolved by illegitimate recombination ERCC1-deficient cells, replication on an unrepaired template and that this results in the production of S-phase-dependent would result in the fixation of mutations and the accumulation chromatid exchanges. Support for this data comes from results of double strand DNA breaks. The elevated HPRT mutation with the DNA topoisomerase I inhibitor, camptothecin. Here, frequency that we have observed in ERCC1-deficient cells and the DNA double strand breaks that are induced by the drug in the increased genome instability, both in vivo and in primary hamster replication forks are rapidly repaired, the cells move and immortalised ERCC1-deficient cultures are compatible out of S-phase, but now they contain abundant S-phase- with this explanation. Furthermore, the DNA damage would dependent chromatid exchanges (Ryan et al., 1994). The short- trigger the p53 response, leading to the elevated p53 levels and term gain of cycle progress is rapidly succeeded by cell death cell cycle arrest, that we observe in vivo. as the cells are unable to divide accurately. As expected, Contrary to the situation with ERCC1-deficient-mice, XP-A metaphases collected following u.v. irradiation of ERCC1- (de Vries et al., 1995; Nakane et al., 1995) and XP-C (Sands deficient cells showed increased levels of chromatid breaks. Of et al., 1995) knockout mice have proved to be a good model greater interest was the profound reduction in the level of for xeroderma pigmentosum, showing the characteristic skin chromatid exchange in the ERCC1-deficient cells, such that the sensitivity to u.v. light and a high frequency of induced skin ratio of exchanges to breaks was 10-fold greater for wild-type tumours, but with no evidence of growth retardation and mouse and hamster cells than their ERCC1-deficient premature ageing. Significantly, the XP-A and XP-C proteins equivalents. We interpret this to indicate that, in addition to its are involved in NER, but not in mitotic recombination. Thus, role in NER and homologous recombination, ERCC1 also as suggested by Weeda et al. (1997), the phenotype of ERCC1- plays a key role in promoting chromatid exchanges in rodent deficient mice seems more likely to be due to a mitotic cells, by its involvement in an illegitimate recombination recombination, rather than an NER deficit. process, which operates at u.v.-induced lesion sites, to channel However, using two different assays, we have found no the disappearance of DNA double strand breaks and thus allow Phenotype of ERCC1-deficient cells 403 replication and the cell cycle to proceed. The profound G2 mice. ERCC1 is essential for NER and is involved in, but is arrest observed in ERCC1-deficient mice, could then be not essential for homologous recombination. In addition, thought to arise as a response to the accumulation of double ERCC1’s role in recombination is used by rodents to enhance strand breaks during S-phase at the sites of unrepaired DNA a more promiscuous exchange behaviour, as part of a rodent- lesions. Further rounds (or partial rounds) of DNA replication, specific mechanism, which is adapted to the low overall level in the absence of cell division, would result in the characteristic of NER in these cells, to prevent the accumulation of double enlarged polyploid and aneuploid hepatocyte nuclei seen in our strand breaks adjacent to unrepaired lesions on replicating mice. Additional experiments to investigate the role of ERCC1 DNA. In the absence of ERCC1, the failure of the illegitimate during DNA replication are planned. recombination process leads to the accumulation of double The cell cycle arrest phenotype observed in ERCC1- strand breaks following replication. This triggers the p53 deficient hepatocytes in vivo is strikingly similar to that in response and the G2 cell cycle arrest, mediated by increased transgenic mice overexpressing the cyclin-dependent kinase expression of the cyclin-dependent kinase inhibitor inhibitor, p21cip1/waf1, in the liver (Wu et al., 1996). We have p21cip1/waf1. The increased levels of unrepaired lesions and obtained preliminary evidence (data not shown) that double strand breaks would explain the increased mutation p21cip1/waf1 mRNA levels are indeed elevated in ERCC1- frequency and increased genome instability that we have deficient liver compared to wild-type controls. We and Weeda observed. et al. (1997) have used primary embryonic fibroblast cultures to study the link between ERCC1-deficiency and the cell cycle F.N. was supported by a PhD studentship from the Darwin Trust of arrest/senescence phenotype in vitro. Our ERCC1-deficient Edinburgh. This work was supported by a project grant cultures, isolated from day 11.5 embryos, had a similar growth (SP2095/0201) from the Cancer Research Campaign to D.W.M.. rate to control cultures, with BrdU pulses indicating that both R.T.J. and S.S. were supported by the Cancer Research Campaign, of cultures had the same frequency of cells in S-phase. The cell which R.T.J. is a Research Fellow. S.B.-A. and A.A. were supported by the Italian Association for Cancer Research and the Commission cycle distribution was identical in both cultures, with both of the European Communities (Project N.EV5V-CT92-0201) and senescing rapidly on subsequent passages. Significantly, FACS Concerted Action on Genetic Disorders Associated with Defects in analysis on ERCC1-deficient cultures provided no evidence for DNA Repair. the presence of the polyploid cells that are so characteristic of ERCC1-deficient hepatocytes in vivo. Contrary to the situation in vivo, p53 protein was readily detectable by western blotting in untreated cultures of both genotypes and both showed an REFERENCES equivalent p53 stabilisation response to u.v. irradiation. The Aboussekhra, A., Biggerstaff, M., Shivji, M. K. K., Vilpo, J. A., Moncollin, high levels of p53 protein found in wild-type primary V., Podust, V. N., Protic, M., Hubscher, U., Egly, J.-M. and Wood, R. D. embryonic fibroblast cultures, compared to the in vivo situation (1995). Mammalian DNA nucleotide excision repair reconstituted with in untreated normal tissues, where p53 protein is only just purified protein components. Cell 80, 859-868. detectable, indicates the limitations of this in vitro system to Avemann, K., Knippers, R., Koller, T. and Sogo, J. M. (1988). study the role of ERCC-1 deficiency in cell cycle arrest. Camptothecin, a specific inhibitor of type I topoisomerase, induces DNA breakage at replication forks. Mol. Cell. Biol. 8, 3026-3034. Weeda et al. (1997) have reported very different results from Bender, M. A., Griggs, H. G. and Walker, P. C. (1973). Mechanism of their studies of primary fibroblast cultures, established from chromosomal aberration production 1. Aberration induction by u.v. Mutat. day 12-13 mouse embryos with the C-terminal ERCC1 Res. 20, 387-402. truncation. ERCC1-deficient cultures had a 3-fold longer Chaudhary, A. K., Nokubo, M., Ramachandra Reddy, G., Yeola, S. N., generation time than wild-type cultures, with only half as many Morrow, J. D., Blair, I. A. and Marnett, L. J. (1994). Detection of endogenous malondialdehyde-deoxyguanosine adducts in human liver. cells cycling. ERCC1-deficient cultures were also reported to Science 265, 1580-1582. senesce more rapidly and accumulate polyploid cells (although Clarke, A. R., Purdie, C. A., Harrison, D. J., Morris, R. G., Bird, C. C., no comparative FACS data of wild-type and mutant cultures Hooper, M. L. and Wyllie, A. H. (1993). Thymocyte apoptosis induced by were shown). p21cip1/waf1 protein levels were also reported to p53-dependent and independent pathways. Nature 362, 849-852. Darroudi, F. and Natarajan, A. T. (1985). Cytological characterization of be elevated in mutant cultures. It is difficult to reconcile the repair-deficient CHO cell line 43-3B I. Induction of chromosomal different results obtained. Cultures were isolated from different aberrations and sister chromatid exchanges by u.v. and its modulation with strains of ERCC1-targeted mice, but the in vivo phenotype of 3-aminobenzamide. Mutat. Res. 149, 239-247. these animals is very similar. Culture conditions and different Degrassi, F., Salvia, R. D., Tanzarella, C. and Palitti, F. (1989). Induction stage embryos used for the isolation could also have an effect. of chromosomal aberrations and SCE by camptothecin, an inhibitor of mammalian DNA topoisomerase I. Mutat. Res. 211, 125-130. Additional analysis is needed to reconcile these differences. de Vries, A., van Oostrom, C. T. M., Hofhuis, F. M. A., Dornant, P. M., However, we would point out the following in support of our Berg, R. J. W., de Gruijl, F. R. et al. (1995). Increased susceptibility to results that ERCC1-deficient cell cultures are not intrinsically ultraviolet-B and carcinogens of mice lacking the DNA excision repair gene inviable. ERCC1-deficient mouse embryos, from which XPA. Nature 377, 169-173. primary fibroblast cultures are isolated, do develop to term Di Leonardo, A., Linke, S. P., Clarkin, K. and Wahl, G. M. (1994). DNA damage triggers a prolonged p53-dependent G1 arrest and long-term (albeit with reduced viability) to produce live (but runted) pups induction of Cip1 in normal human fibroblasts. Genes Dev. 8, 2540-2551. after the normal gestation period. Immortalised ERCC1- el-Deiry, W. S., Tokino, T., Velculescu, V. E., Levy, D. B., Parsons, R., deficient mouse fibroblasts and ERCC1 mutant CHO cells Trent, J. M., Lin, D., Mercer, W. E., Kinzler, K. W. and Vogelstein, B. grow well in culture, only marginally slower than their wild- (1993). WAF1, a potential mediator of p53 tumor suppression. Cell 75, 817- 825. type controls. Fishman-Lobell, J. and Haber, J. E. (1992). Removal of nonhomologous In conclusion, we favour the following model to explain the DNA ends in double-strand break recombination: the role of the yeast growth arrest and early senescence seen in ERCC1-deficient ultraviolet repair gene RAD1. Science 258, 480-484. 404 D. W. Melton and others

Gimenez-Abian, J. F., Clarke, D. J., Mullinger, A. M., Downes, C. S. and Sands, A. T., Abuin, A., Sanchez, A., Conti, C. J. and Bradley, A. (1995). Johnson, R. T. (1995). A post-prophase topoisomerase II-dependent A high susceptibility to ultraviolet-induced carcinogenesis in mice lacking chromatid core separation step in the formation of metaphase chromosomes. XPC. Nature 377, 162-165. J. Cell Biol. 131, 1-11. Schaeffer, L., Roy, R., Humbert, S., Moncollin, V., Vermeulen, W., Graham, F. L. and Van der Eb, A. J. (1973). A new technique for the assay Hoeijmakers, J. H. J., Chambon, P. and Egly, J.-M. (1993). DNA repair of infectivity of human adenovirus 5 DNA. Virology 52, 456-467. helicase: a component of BTF2 (TFIIH) basic transcription factor. Science Hartwell, L. H. and Kastan, M. B. (1994). Cell cycle control and cancer. 260, 58-63. Science 266, 1821-1828. Schiestl, R. H. and Prakash, S. (1990). RAD10, an excision repair gene of Iskandar, O. (1979). An improved method in the detection of micronuclei in Saccharomyces cerevisiae, is also involved in the RAD1 pathway of mitotic human lymphocytes. Stain Technol. 54, 221-223. recombination. Mol. Cell. Biol. 10, 2485-2491. Lasko, D. D., Tomkinson, A. E. and Lindahl, T. (1990). Mammalian DNA Selfridge, J., Pow, A. M., McWhir, J., Magin, T. M. and Melton, D. W. ligases. Biosynthesis and intracellular localisation of DNA ligase I. J. Biol. (1992). Gene targeting using a mouse HPRT minigene/HPRT-deficient Chem. 265, 12618-12622. embryonic stem cell system: inactivation of the mouse ERCC1 gene. Littlefield, J. W. (1964). Selection of hybrids from matings of fibroblasts in Somatic Cell Mol. Gen. 18, 325-336. vitro and their presumed recombinants. Science. 145, 709. Spivak, G. C. and Hanawalt, P. C. (1992). Translesion DNA synthesis in the Magin, T. M., McEwan, C., Milne, M., Pow, A. M. Selfridge, J. and Melton, dihydrofolate reductase domain of u.v.-irradiated CHO cells. Biochemistry D. W. (1992). A position- and orientation-dependent element in the first 31, 6794-6800. intron is required for expression of the mouse HPRT gene in embryonic stem Thompson, S., Clarke, A. R., Pow, A. M., Hooper, M. L. and Melton, D. cells. Gene 122, 289-296. W. (1989). Germ line transmission and expression of a corrected HPRT gene Mellon, I. G., Spivak, G. and Hanawalt, P. C. (1987). Selective removal of produced by gene targeting in embryonic stem cells. Cell 56, 313-321. transcription-blocking DNA damage from the transcribed strand of the Vindelov, L. L., Christensen, I. J. and Nissen, N. I. (1983). A detergent- mammalian DHFR gene. Cell 51, 241-249. trypsin method for the preparation of nuclei for flow cytometric DNA McWhir, J., Selfridge, J., Harrison, D. J., Squires, S. and Melton, D. W. analysis. Cytometry 3, 323-327. (1993). Mice with DNA repair gene (ERCC1) deficiency have elevated Waldman, T., Lengauer, C., Kinzler, K. W. and Vogelstein, B. (1996). levels of p53, liver nuclear abnormalities and die before weaning. Nature Uncoupling of S-phase and mitosis induced by anticancer agents in cells Genet. 5, 217-224. lacking p21cip1/waf1. Nature 381, 713-716. Musk, S. R. R., Pillidge, L., Johnson, R. T. and Downes, C. S. (1990). Wang, T. V. and Smith, K. C. (1986). Postreplication repair in u.v.-irradiated Action of caffeine on DNA replication after u.v. irradiation in Indian human fibroblasts: formation and repair of DNA double strand breaks. muntjac : no connection with cell cycle delay. Biochim. Biophys. Acta 1052, Carcinogenesis 7, 389-392. 53-62. Wang, X.-Q. and Ellem, K. A. O. (1994). Heterogeneity in the HeLa cell Nakane, H., Takeuchi, S., Yuba, S., Saijo, M., Nakatsu, Y., Murai, H. et cycle response to UVC analyzed by the BrdUrd two-parameter method. Exp. al. (1995). High incidence of ultraviolet-B or chemical-carcinogen-induced Cell Res. 212, 176-189. skin tumours in mice lacking the xeroderma pigmentosum group A gene. Waters, R. (1979). Repair of DNA in replicated and unreplicated portions of Nature 377, 165-168. the human genome. J. Mol. Biol. 127, 117-127. Nuesse, M., Viaggi, S. and Bonatti, S. (1989). Induction of kinetochore Weeda, G., Donker, I., de Wit, J., Morreau, H., Janssens, R., Vissers, C. positive and negative micronuclei in V79 cells by the alkylating agent J., Nigg, A., van Steeg, H., Bootsma, D. and Hoeijmakers, J. H. J. (1997). diethylsulphate. Mutagenesis 4, 174-178. Disruption of mouse ERCC1 results in a novel repair syndrome with growth Orren, D. K., Petersen, L. N. and Bohr, V. A. (1995). A u.v.-responsive G2 failure, nuclear abnormalities and senescence. Curr. Biol. 7, 427-439. checkpoint in rodent cells. Mol. Cell. Biol. 15, 3722-3730. Westerveld, A., Hoeijmakers, J. H. J., van Duin M., de Wit, J., Odijk. H., Rolig, R. L., Layher, S. K., Santi, B., Adair, G. M., Gu, F., Rainbow, A. J. Pastink, A. et al. (1984). Molecular cloning of a human DNA repair gene. and Nairn, R. S. (1997). Survival, mutagenesis, and host cell reactivation Nature 310, 425-429. in a Chinese hamster ovary cell ERCC1 knock-out mutant. Mutagenesis 12, Wood, R. D. and Burki, H. J. (1982). Repair capability and the cellular age 277-283. response for killing and mutation induction after u.v. Mutat. Res. 95, 505- Ryan, A. J., Squires, S., Strutt, H. L. and Johnson, R. T. (1991). 514. Camptothecin cytotoxicity in mammalian cells is associated with the Wood, R. D. (1996). DNA repair in eukaryotes. Annu. Rev. Biochem. 65, 135- induction of persistent double strand DNA breaks in replicating DNA. Nucl. 167. Acids Res. 19, 3295-3300. Wu, H., Wade, M., Krall, L., Grisham, J., Xiong, Y. and Van Dyke, T. Ryan, A. J., Squires, S., Strutt, H. L., Evans, A. and Johnson, R. T. (1994). (1996). Targeted in vivo expression of the cyclin-dependent kinase inhibitor Different fates of camptothecin-induced replication fork-associated double p21cip1/waf1 halts hepatocyte cell cycle progression, postnatal liver strand DNA breaks in mammalian cells. Carcinogenesis 15, 823-828. development, and regeneration. Genes Dev. 10, 245-260.