Transcription-Coupled Repair Deficiency and Mutations in Human Mismatch Repair Genes

Deficiencies in mismatch repair have been linked to a common cancer predisposition syndrome in humans, hereditary nonpolyposis colorectal cancer (HNPCC), and a subset of sporadic cancers. Here, several mismatch repair-deficient tumor cell lines and HNPCC-derived lymphoblastoid cell lines were found to be deficient in an additional DNA repair process termed transcription-coupled repair (TCR). The TCR defect was corrected in a mutant cell line whose mismatch repair deficiency had been corrected by chromosome transfer. Thus, the connection between excision repair and mismatch repair previously described in Escherichia coli extends to humans. These results imply that deficiencies in TCR and exposure to carcinogens present in the environment may contribute to the etiology of tumors associated with genetic defects in mismatch repair.

moving helix-distorting lesions from cellular genomes. The general strategy appears to be similar in organisms ranging from Escherichia coli to humans. This process is complex and requires the participation of a number of different proteins (1). Its role in ameliorating the carcinogenic consequences of DNA damage has been inferred from studies of the genetic disease xeroderma pigmentosum (XP). Cells from XP patients are hypersensitive to the killing and mutagenic effects of ultraviolet light (UV) and 557 11. S. LeVay, M. P. Stryker, C. J. Shatz, J. Comp. Neurol. 179, 223 (1978); A. Antonini and M. P. Stryker, J. Neurosci. 13, 3549 (1993). 12 34, 709 (1994). J. A. Gordon and M. P. Stryker [J. Neurosci., in press] show that ocular dominance plasticity in mouse primary visual cortex is subject to the same conditions and occurs in the same way as in the cat, involving both intracortical and thalamocortical changes (1). 14. Recordings were obtained from layer 11/l1l in the binocular zone of 400-p.m-thick coronal slices of mouse primary visual cortex (C57/BL6; <6 weeks old) continuously superfused with oxygenated (95% 02, 5% C02) artificial cerebrospinal fluid (ACSF), containing (in millimolar) 119 NaCI, 2.5 KCI, 1.3 MgSO4, 1.0 NaH2PO4, 26.2 NaHCO3, 2.5 CaCl2, and 11 glucose. Extracellular pipettes (1 M NaCI, 1-3 megohm) monitored stable, half-maximal baseline field potentials evoked from layer IV by a bipolar Pt-lr electrode delivering 1 00-p.s pulses at 0.1 Hz. Five episodes of TBS were given at 10-s intervals to induce LTP before depotentiation was attempted with 900 pulses at 1 Hz. A single TBS consisted of 10 repetitions of four stimuli at 100 Hz delivered at 200-ms intervals (12). All experiments were terminated by bath application of 10 p.M 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX, Tocns, UK) and 50 p.M D-(-)-2-amino-5-phosphonovalerc acid (D-APV, Sigma) to determine the synaptic component of the field response. Similar results were obtained by measurements of synaptic slope or peak amplitude normalized to the baseline period before TBS, and renormalized to the 10 min preceding 1 -Hz stimulation to adjust for variable elapsed time post-TBS across experiments. Both LTP and depotentiation were prevented by the N-methyl-D-aspartate receptor antagonist APV, as also reported for LTD of naive synapses (12). R,S-a-MCPG or (+)-MCPG (Tocris) were dissolved in 100 mM NaOH at 50 mM, then diluted to 500 p.M in ACSF. 15 are defective in the excision repair of their DNA. Nucleotide excision repair can be coupled to transcription. In normal human cells (2), E. coli (3), and Saccharomyces cerevisiae (4), UV-induced cyclobutane pyrimidine dimers are removed more rapidly from the transcribed strands than from the nontranscribed strands of active genes. This feature of repair is generally termed transcription-coupled repair (here, referred to as TCR) (5). We have shown that mutations in two genes required for DNA mismatch repair, mutS and mutL, abolish TCR of UV photoproducts from the lac operon in E. coli (6). In contrast to nucleotide excision repair, mismatch repair corrects single-base mismatches and small insertion or deletion mispairs that can be generated during replication (7). Our results indicate an association between mismatch repair and nucleotide excision repair in E. coli and may implicate components of mismatch repair in the coupling of excision repair to transcription.
In addition, this link could have important implications for human disease. Mutations in the human homologs of the E. coli mismatch repair genes mutS and mutL (hMSH2, hMLHJ, and hPMS2) are associated with HNPCC (8,9). This disease is one of the most common cancer predisposition syndromes and exhibits an autosomal dominant mode of inheritance. Patients are at risk for early onset of colon cancer (Lynch syndrome I) or early onset of colon cancer with tumors at additional sites, including the stomach, small intestine, kidney, ureter, and ovary (Lynch syndrome II) (10,11). Normal cells from affected individuals are heterozygous, with a germline mutation in one of four mismatch repair genes (hMSH2, hMLHI, hPMSJ, or hPMS2) (8,9). Tumor cells from affected individuals have mutations in both alleles of a mismatch repair gene, and the inactivation of the second allele occurs as a consequence of a somatic mutation or a reduction to homozygosity.
In addition, mutations in hMSH2, hMLHI, and hPMS2 have been identified in sporadic tumor cell lines derived from the colon and endometrium (12)(13)(14)(15)(16). In general, both sporadic and HNPCC tumor cells with genetic defects in mismatch repair exhibit elevated spontaneous mutation rates in microsatellite sequences, and several sporadic tumor cell lines have been found to be deficient in the processing of singlebase mismatches and small insertion or deletion mispairs (12,13,16). To determine whether mutations in human mismatch repair genes also affect the removal of environmentally induced DNA damage, we examined here TCR in a collection of tumorand HNPCC-derived lymphoblastoid cell lines.
We measured the removal of cyclobutane pyrimidine dimers from individual strands of the gene encoding dihydrofolate reductase (DHFR) in UV-irradiated cells. Using strand-specific RNA probes, we examined repair in a 20-kb Kpn I restriction fragment that resides within the transcription unit of the -30-kb gene. Cells were irradiated with UV light (10 J/m2) and either lysed immediately to determine the initial frequency of pyrimidine dimers (-1.5 to 2 in each strand of the fragment) or incubated for increasing periods of time to allow repair and then lysed. Large molecular mass DNA was purified and treated with restriction enzymes. Two portions of each DNA sample were analyzed: one was treated with T4 endonuclease V to produce a single-strand DNA break at each cyclobutane pyrimidine dimer and the other was mock-treated. Samples were subjected to electrophoresis under denaturing conditions and then transferred to a membrane. The membranes were then sequentially hybridized with strand-specific RNA probes to quantify the abundance of full-length restriction fragments for each strand at each time point.
As repair occurred, fewer T4 endonuclease-sensitive sites remained in the DNA, resulting in fewer strand breaks produced by the endonuclease and more full-length fragments appearing in the enzyme-treated sample. We determined the extent of repair by calculating (using the Poisson expression) the average number of dimers per strand at each time point from the ratio of full-length restriction fragments in the enzyme-treated and untreated samples.
We first examined the mismatch repairproficient colon tumor cell line SW480 (12). As in our previous observations of a human cell line (2), here pyrimidine dimers were rapidly removed from the transcribed strand of the gene encoding DHFR, whereas repair in the nontranscribed strand was slower (Fig. 1). Comparison of the enzymetreated samples probed for the two different strands revealed more full-length fragments for the transcribed strand than for the nontranscribed strand. Quantitation of data obtained from several autoradiograms showed that within 3 hours after UV treatment, almost 50% of the dimers were removed from the transcribed strand, whereas only 20% were removed from the nontranscribed strand. This difference in the amount of repair was also observed 6 and 9 hours after exposure to UV. However, repair in both strands approached completion within 20 hours after exposure to UV. Similar results were obtained for the mismatch repair-proficient endometrial tumor cell line KLE (16,17).
We next examined TCR in several sporadic tumor cell lines that exhibit microsat-  Fig. 1. TCR in human sporadic colon and endometrial tumor cell lines. Repair was measured and DNA samples were taken at the times indicated at the top of the lanes. Strand-specific RNA probes were made with the use of pGEMO.69EH (34) as a template. Open symbols represent the transcribed strand, and closed symbols represent the nontranscribed strand. Each value plotted represents data averaged from at least two independent biological experiments and at least four autoradiograms. Treatment or no treatment with T4 endonuclease V is indicated by " +" and "-," respectively; T, transcribed strand; NT, nontranscribed strand. ellite instability, have mutations in mismatch repair genes, and are defective in the repair of DNA-containing mismatches. The LoVo cell line, which arises from a colon tumor, has deletions in both alleles of hMSH2 (13); HEC59 is an endometrial tumor cell line with mutations in both alleles of hMSH2 (13,18); and HEC 1A is an endometrial tumor cell line containing a nonsense codon in the hPMS2 gene and no detectable wild-type hPMS2 transcript (16).
Examination of representative autoradiograms revealed that there was no significant difference in the repair of the transcribed and nontranscribed strands of the gene encoding DHFR in all three mutant cell lines (Fig. 1). The kinetics and extent of repair were determined by averaging results from several experiments. In each mutant cell line, repair rates in the transcribed and nontranscribed strands of the gene were similar. In LoVo cells, there also appeared to be some slowing of repair in both strands relative to the nontranscribed strand of the repair-proficient SW480 cell line. In HEC59 SCIENCE * VOL. 272 * 26 APRIL 1996 cells, there appeared to be a the extent of repair; only 50% were removed from each str; after UV treatment. Both c( mutations in hMSH2 showed tion in the extent of repair modest reduction in overall r cision repair has been obsei strains of E. coli (6). Cell lines harboring muta match repair genes show an el spontaneous mutation and ar accumulating genetic alterati sites throughout their gen growth (19). Consequently, tI of any repair deficiency to a one locus is difficult in the ab, evidence for genetic correcti repair deficiency and microsat ity observed in the colon tu SW480 Fig. 4. TCR in lymphoblastoid cell lines P7, P2, P6, and P8. Cells were grown in suspension to a density of 1 x 1 06 cells per dish, pelleted, rep suspended at the same density in phosphatebuffered saline, irradiated with 10 J/m2 of UV p light, pelleted, and either lysed immediately or resuspended in growth medium for the specific P times shown and then lysed. Samples were then processed as described in Fig. 1.   HCT1 16, which has a homozygous mutation in the hMLHI gene (another human homolog of mutL) on chromosome 3, is corrected by transfer of a normal copy of chromosome 3 (20).
Here, we examined TCR in the parental mutant HCT1 16 cell line and the chromosome 3-transfer derivative of HCT1 16 (HCT116 3-6) (Fig. 2). There was no difference in the repair of the two strands of the DHFR gene in HCT1 16, and the kinetics and extent of repair in both strands resembled that of the repair of the nontranscribed strand of the repair-proficient SW480 cell line (Fig. 1). These results are similar to those obtained with mutL mutant strains of E. coli in that TCR was abolished and there was no obvious decrease in overall repair (6). In contrast, introduction of a HCT116 [3][4][5][6] normal copy of chromosome 3 into the hMLHl mutant cell lines restored TCR. HCT116 This is clearly shown by visual comparison LoVo of the abundance of full-length restriction fragments in samples treated with T4 endonuclease V and probed for the two strands 9 and 20 hours after UV treatment. Surprislines. Cells were ingly, repair in the nontranscribed strand )-mm dishes at was reduced relative to the extent of repair 3cells per dish, observed in the parental HCT1 16 line. This idiated with UV may be a consequence o the presence of ek), stained with only one functional misrr ch repair allele okstainalogarithmc in the corrected cell line.

theaverageaob-
The HCT1 16 and LoV cell lines were )eriments. somewhat more sensitive tr the killing effects of UV irradiation the repairproficient SW480 c .g. 3). In adreduction in dition, the chromo derivative of of the lesions HCT1 16 exhibited i resistance to and 20 hours UV relative to its par( nt. e, a result that ell lines with is consistent with the rc.ation of TCR. I some reduc-Taken together, these rest are consistent A similarly with studies that have s' wn a moderate iucleotide ex-increase in UV sensiti\ ,ty in mfd (21), rved in mutS mutS, or mutL strains of E. coli (6) and in cells from patients with Cockayne's synitions in mis-drome (22), all of which are deficient in levated rate of TCR (6,23,24). re continually Recently, HNPCC patients were identiions at many fied with a heterozygous defect in hPMS2. iomes during Surprisingly, cells of nontumor origin from he assignment these patients exhibited hypermutability in l mutation at microsatellite DNA (25). In addition, exlsence of some tracts prepared from Epstein-Barr virus-Dn. Mismatch transformed lymphoblastoid cells from ellite instabilthese patients were virtually devoid of mismor cell line match repair activity. It

IT
T that these specific mutations produce a dominant mutator phenotype. To extend our investigation to human cells of nontumor origin, we investigated TCR in the mutant lymphoblastoid cell lines P7, P2, P6, and P8 (Fig. 4). P7 was derived from a patient with familial adenomatous polyposis, which is not associated with any mismatch repair defect. No alterations in microsatellite repeat length have been detected in this cell line. P2 and P6 are from HNPCC patients and are hPMS2 heterozygotes; both cell lines are deficient in the ability to process substrates containing a single-base mismatch or a dinucleotide insertion. They also exhibit microsatellite instability. P8 is an HNPCC hMSH2 heterozygote and, unlike P2 and P6, is proficient in mismatch repair.
The kinetics and extent of repair in the transcribed strand of the gene encoding DHFR were reduced in both hPMS2 mutant cell lines (Fig. 4). Repair was also reduced in the hMSH2 heterozygous cell line P8, but more rapid repair of the transcribed strand was apparent 20 hours after UV treatment. A more extensive examination of mismatch repair-proficient heterozygous lymphoblastoid cell lines is necessary to determine the significance of the reduced nucleotide excision repair observed in the hMSH2 mutant cell line.
Our results document an association between defects in human mismatch repair genes and the loss of transcription-coupled nucleotide excision repair. All cell lines studied here with defects in TCR have been shown to be virtually devoid of mismatch repair activity in an in vitro assay of cellular extracts. In addition, TCR is restored in a mutant human cell line whose mismatch deficiency has been corrected by chromosome transfer. That the loss of TCR is a direct consequence of mutations in mismatch repair genes is also supported by the observation that TCR is absent in PMS2'/_ and MSH2-/mouse embryonic fibroblast cell lines derived from animals generated by targeted disruption of the respective genes (26)(27)(28). Thus, the connection between mismatch repair and nucleotide excision repair documented in E. coli extends to mammals. In contrast, no dramatic reduction in TCR was observed in several yeast strains with mutations in mismatch repair genes (29). Yeast also differ from E. coli and mammalian cells in that mutations in mismatch repair genes do not result in an increased tolerance to treatment with the alkylating agent N-methyl-N'-nitro-N-nitrosoguanidine (MNNG) (30). It is possible that genetic redundancy and the expression levels of different mismatch repair genes influence the biological consequences of a mutation in a mismatch repair gene.
It is thought that an increase in the spontaneous accumulation of mutations that are the result of an inability to process mispairs leads to the development of cancer associated with HNPCC. However, this theory does not explain the unique spectrum of tumors associated with HNPCC (predominantly colon) or the differences in the tumors observed in mismatch repairdefective mice (lymphomas) (26) and HNPCC patients. It is possible that the inactivation of the ability to process mispairs is not the cause of tumorigenesis in HNPCC. Nucleotide excision repair recognizes a wide spectrum of DNA lesions produced by physical and chemical agents present in the environment (31). Furthermore, TCR has been demonstrated for substrates of nucleotide excision repair and base excision repair (32). Although the enhancement of excision repair by transcription may appear subtle, if mutations in human mismatch repair genes abolish many types of TCR, then a reduction in the repair of environmentally induced DNA damage could affect the development of cancer associated with defects in mismatch repair genes.
In addition, a subtle defect in the repair of DNA damage could have a more profound impact on tumorigenesis if it is less likely to be lethal to the cell. Although predisposition to cancer is not associated with Cockayne's syndrome, these patients die on average by 12 years of age (33). In general, tumors associated with HNPCC develop later in a patient's life. This could explain the observation made by Parsons et al. (25) that although there appeared to be widespread mutations in non-neoplastic cells from several different tissues in a subset of HNPCC patients, there was no significant increase in the frequency of tumors. On the basis of these results, they suggested that an increase in spontaneous mutations may not be sufficient for tumorigenesis and exposure to environmental mutagens may have a part in the process.
Before yeast can replicate DNA, they must pass Start, which requires a cyclin-dependent protein kinase composed of a catalytic subunit (Cdc28) and one of three G1 cyclins (Clnl, -2, or -3) (1). After Start, B-type cyclin-Cdc28 kinases such as Clb5-Cdc28 and Clb6-Cdc28 must be activated to allow replication (2). Although Clb5and Clb6-Cdc28 complexes are present in G1 phase, they are initially inactive because of inhi- Sicl is targeted for proteolysis by the ubiquitin-conjugating enzyme Cdc34 (2). Thus, a cdc34 mutant arrests with a 1N DNA content because it cannot degrade Sic1, but nevertheless buds, and duplicates its spindle pole body.
It is not known how Start triggers Sicl inactivation or how replication is tied to other Start-dependent events such as budding and duplication of the spindle pole body. Is Start a single event that affects multiple pathways, or is Start a collection of events, one of which regulates Sicl proteolysis and replication?