To investigate the influence of function or activity of a DNA sequence on its repair, we have studied excision repair of a number of adducts in the non-transcribed, heterochromatic alpha DNA of monkey cells (by physically isolating the DNA) and also the removal of pyrimidine dimers in a number of genes in rodent and human cells (by an indirect assay using a dimer-specific endonuclease). In confluent cells, psoralen and aflatoxin B] (AFB1) adducts are produced in similar frequencies in alpha and in the rest of the DNA, but removal from alpha is severely deficient. Adducts of A’-acetoxy acetyl aminofluorene (NA-AAF) are formed in slightly higher frequencies in alpha, and removal is slightly deficient. The removal of thymine glycols from alpha DNA in gamma-irradiated cells is proficient, as is repair synthesis elicited by exposure to methyl methane sulphonate, dimethyl sulphate, or 254nm ultraviolet light (u.v.). Removal of AFB1 and NA-AAF adducts from alpha is enhanced by small doses of u.v. but not by X-rays or DMS. The quantum efficiency of conversion of psoralen monoadducts to crosslinks is much lower in alpha DNA. Taken together, these results suggest that the highly condensed chromatin structure of alpha hinders access of the repair system that acts on bulky adducts but not of systems for repair of specific base damage, u.v. damage may alter this chromatin structure directly or facilitate the action of some system that changes accessibility of chromatin to repair. The repair deficiencies are not observed in actively growing cells, in which chromatin structure may be less condensed due to DNA replication.

We have also demonstrated preferential excision repair of pyrimidine dimers in active genes. Dimers are efficiently removed from the essential dihydrofolate reductase (DHFR) and hydroxy- methylglutaryl CoA reductase genes in Chinese hamster ovary (CHO) cells and from the transcribed c-ahl proto-oncogene in the mouse cells. Both cell types remove few dimers from their overall genomes or from sequences distal to the DHFR gene; dimers are also removed poorly from the non-transcribed mouse c-mos gene. In human cells, dimers are removed more rapidly from the DHFR gene than from the genome as a whole. However, repair is as deficient in this gene in XP-C cells as it is in the entire genome.

These results suggest that resistance to DNA damage correlates better with repair of vital or active sequences than with overall repair levels and that mutagenic efficiency may vary according to the activity of the gene under study.

In the past several years many of us* working in the laboratory of Philip Hanawalt have focused on the study of excision repair in specific DNA sequences in cultured mammalian cells. Unrepaired damage in DNA of different functional states would be expected to have differing consequences for the cell. Mutations in active sequences or blockage of their transcription could have direct effects on cell survival and function.

Most of the DNA in a cell in a complex organism contains information unnecessary or inappropriate to its own specialized function, and is not expressed. The interference with appropriate control of expression of such information by persisting damage is also likely to play a role in carcinogenesis. Different states of chromatin into which the DNA is packaged may modulate the actions of particular damaging agents and the various repair systems to different degrees. To determine how the function or state of activity of DNA relates to its repair one must study particular sequences in known states of expression and/or chromatin configuration. We have studied in some detail a repetitive DNA species in monkey cells-called alpha DNA, a non-transcribed sequence that can be isolated in sufficient quantity for examination by a variety of well-established techniques. For non-repetitive DNA, we have begun to develop methods that utilize hybridization with specific probes. Here, I will summarize our results with alpha DNA and present information about removal of pyrimidine dimers from some specific genes.

Alpha DNA is a highly repetitive DNA (Singer, 1982), found primarily near chromosome centromeres (Segal et al. 1976) in constitutive heterochromatin, comprising 15-20% of the genome of African green monkey cells (Maio, 1971). It is generally considered to be a non-transcribed sequence (Kuff et al. 1978) and no function for it has been identified. Sequences highly homologous to alpha have been found in several other primates, but in much lower quantities than in green monkey, e.g. in human cells, an alpha-like species comprises about 1 % of the DNA (Wu & Manuelidis, 1980). Musich et al. (1977) reported that alpha chromatin was depleted in histone Hl and enriched for other chromosomal proteins. Strauss & Varshavsky (1984) described a high mobility group protein that binds to alpha DNA and suggested it aids specific nucleosomal positioning with alpha arrays. More recently this protein has been shown to bind specifically to short runs of A · T base-pairs (Solomon et al. 1986), and it may play a role outside alpha chromatin. The repeating unit of alpha DNA in green monkey cells is a 172 base-pair (bp) monomer, which has been cloned and sequenced (Rosenberg et al. 1978). Most of the monomers contain a single recognition site for the restriction endonuclease HindIII, and a clustered subset (about 10%) also contain a site for AcoRI (Wu et al. 1983). The sequence does not differ significantly from total cellular DNA in base composition and nearest neighbour frequencies.

We recently investigated the lengths of alpha arrays by digesting genomic DNA with restriction endonucleases and determining the sizes of fragments that hybridized to alpha-specific probe (Madhani et al. 19866). We determined an average array length of about 450 monomer units (about 80 kb), probably an underestimate due to unavoidable shearing of DNA during manipulation.

Alpha DNA can be isolated from cellular genomic DNA by separation of HindIII fragments on preparative 2% agarose gels. Alpha DNA and the remaining cellular DNA (which we refer to as bulk DNA) are excised and purified for further study. We have compared DNA repair in alpha to that in the bulk DNA, both by measuring repair synthesis and by quantifying the persisting lesions, using radioactively labelled adducts or specific antibodies. The results described in the following sections (summarized in Tables 1, 2) were obtained with confluent cultures.

Table 1.

Repair in alpha DNA measured as repair synthesis

Repair in alpha DNA measured as repair synthesis
Repair in alpha DNA measured as repair synthesis
Table 2.

Repair in alpha DNA measured by lesion removal

Repair in alpha DNA measured by lesion removal
Repair in alpha DNA measured by lesion removal

Ultraviolet light

We found no significant differences in the extent, time course, or dose response of repair synthesis in alpha and bulk DNAs in cells irradiated with 254 nm ultraviolet light (u.v.) (Zolan et al. 1982a). Repair was determined after doses ranging from 5 J m-2 (about 2 pyrimidine dimers per 101 bases of DNA) to 50 J m-2 at 4 and 24 h after irradiation, and over the entire time course up to 48 h following 20 J m-2. Chromatographic analysis of hydrolysed DNA showed that 50 J m-2 induced pyrimidine dimers with the same quantum yield in both alpha and bulk DNAs.

Furocoumarins

These tricyclic compounds intercalate into DNA and can covalently bind to pyrimidines upon absorption of 360 nm light. Linear furocoumarins (psoralens) can form both monoadducts and interstrand crosslinks, because each of the two ends of the ring structure (called the pyrone and furan ends) is positioned to allow formation of a cyclobutyl linkage to the 5-6 carbons of a pyrimidine, one above the plane of the intercalated psoralen molecule and one below it, on opposite DNA strands. Crosslink formation is a two-photon, sequential process: in the majority of monoadducts a thymine is linked to the furan end of the psoralen (furan-T monoadducts) and the unreacted pyrone end can then react to form a crosslink if another pyrimidine is in the appropriate position on the opposite strand. Monoadducts of angelicin, an angular furocoumarin, do not have properly positioned unreacted ends and do not normally form crosslinks. Furocoumarins provide a kind of hybrid species between u.v. irradiation-induced pyrimidine dimers and direct-acting chemical agents. Their adducts form through the same cyclobutane bridge to pyrimidines as in pyrimidine dimers. The requirement for photochemical activation provides more precise control of the amount and type of damage introduced than that for direct-acting chemicals. However, like other chemicals, their interaction with DNA is strongly modulated by chromatin structure. In addition, the monoadducts and crosslinks present fundamentally different challenges to cellular repair systems as well as qualitatively different obstacles for DNA transactions such as replication or recombination.

In contrast to results with u.v., repair synthesis in response to either angelicin or the psoralen derivative aminomethyl-trimethylpsoralen (AMT) was markedly deficient in alpha, being only about 30% of that in the bulk DNA from 4-8 h after treatment. On the basis of previous measurements of angelicin adducts in DNA (Zolan et al. 19826) and AMT crosslinkable sites (unpublished results) in human cells we estimate that about 30 adducts per 101 bases were formed in the bulk DNA under conditions used in these experiments. With radioactive AMT at 4- to 16-fold lower concentrations, we found adduct frequencies to be the same in alpha and bulk DNA, and repair synthesis in alpha was still only about 30 % of that in bulk. At 0-8 AMT adducts per 101 bases, about 28 % were removed in 24 h from bulk DNA, but only 9 % were removed from alpha. These results (Zolan et al. 1982a) were the first demonstration of heterogeneous repair among different genomic sequences DNA in a single cell.

A possible explanation for these results was that some specific furocoumarin adduct that is inherently refractory to repair is formed to a much greater degree in alpha DNA. To test this, we investigated removal of another psoralen derivative, hydroxymethyl-trimethylpsoralen (HMT), because methods for analysing HMT adducts by high-pressure liquid chromatography (HPLC) and the proportions of the various adducts formed in purified DNA had been reported (Kanne et al. 1982). The major monoadduct species, the furan-T monoadduct, exists in either of two diastereomers, depending upon whether the psoralen is 5’ or 3’ to the thymine in the DNA. These are the only forms present in large enough quantity to be candidates for the hypothesis under test; the deficient repair observed with angelicin ruled out differential crosslink formation and repair as a primary cause of the deficiency.

Repair of HMT adducts in alpha DNA was only 20 % of that in bulk, both at high adduct frequencies (repair synthesis) and much lower ones (removal of labelled adducts). Analysis by HPLC showed that the relative amounts of the two diastereomers of the furan-T adduct were the same in alpha and bulk DNA, whether irradiated in cells or as purified DNA. Using treatment conditions that gave only about 3 % of the adducts as crosslinks, we found that in 48 h all the monoadduct species were removed from bulk DNA to about the same extent, roughly 40%. These studies (Zolan et al. 1984) ruled out differential formation and repairability of different individual adducts as a cause for repair deficiency in alpha. However, they did provide evidence for an altered chromatin structure of alpha. Although we always observed similar total adduct frequencies for bulk and alpha DNAs, the ratio of crosslinks to monoadducts formed in vivo was considerably lower in alpha than in the bulk DNA. This was observed at two different overall levels of modification, but not when purified alpha and bulk DNA were reacted with HMT. This lower apparent quantum yield for conversion of monoadducts to crosslinks in alpha could be due to a more condensed chromatin structure, restricting the flexibility of the nucleosomal linker DNA, in which most of the psoralen adducts are found. Model building studies (Peckler et al. 1982) show that a large kink must occur in the DNA at the site of a crosslink; a reduction of the flexibility might therefore lower the quantum yield for crosslink formation.

Aflatoxin adducts

We also examined repair in cells treated with activated aflatoxin Bj (Leadon et al. 1983). At high levels (about 440 adducts per 101 bases for both alpha and bulk DNA, measured using radiolabelled compound) repair synthesis in alpha was only about 25 % of that in the bulk DNA. At much lower levels (about 0-3 adduct per 101 bases), analysis by HPLC showed that 95 % of the initial adducts were to the N-7 position of guanine, and that loss of these adducts was much slower in alpha DNA than in the bulk DNA, but was still appreciable. In fact the rate of loss of adducts from alpha resembled that reported for the total DNA in xeroderma pigmentosum (XP) group A cells by Leadon et al. (1981). The loss in XP cells is ascribed to spontaneous release of the AFB! moiety from some guanines and the release of some of the modified guanines to leave apurinic sites (AP) sites, rather than to enzymic repair. The residual repair synthesis in alpha might result from excision repair initiated at these AP sites, indicating that such repair is at least partly proficient in alpha. Evidence for this was provided by measurements of the sizes of the repair patches in alpha and bulk DNA (see below).

Adducts of NA-AAF

This compound makes adducts to guanine (primarily at the C-8 and N1 positions) that are not subject to spontaneous release. In cells treated with 25 /1M NA-AAF, repair synthesis in alpha was about 60 % of that in the bulk DNA, considerably more than observed for furocoumarin or aflatoxin adducts (Zolan et al. 1982a). Studies with radiolabelled compound at 0 · 3 µ M (Leadon & Hanawalt, 1984) showed that more adducts formed in alpha DNA (about 0 · 5 per 101 bases) than in the bulk DNA (0 · 32per 101bases). This reflected primarily an increase in G-C8-acetylamino- fluorene, with only a slight increase in the major adduct, G-C8-aminofluorene. The minor G-N1-AAF adduct was formed in about equal frequencies, and was not removed in 24 h from either DNA species. For NA-AAF, differences between adduct removal from alpha and bulk DNA were complex. When expressed as the fraction of initial G-C8 adducts removed, repair in alpha after 24 h was about 70% as great as that in bulk, but if considered as number of adducts removed, it was 90 %. Both G-C8 adducts contributed to these minor repair deficiencies. However the kinetics of removal differed; removal from alpha appeared to taper off rapidly after about 8 h whereas adducts were progressively removed from bulk DNA up to 48 h, the longest period studied.

Alkylation damage

Repair synthesis in response to methylating agents is thought to be initiated by sequential action of glycosylases specific for methylated bases and AP endonucleases. We exposed cells to the alkylating agents methyl methane sulphonate (MMS) (at 1 and 4mM) and dimethyl sulphate (DMS) (1 mM) and incubated them in medium for 8 h. Repair synthesis in alpha was 80 % of that in the bulk for either concentration of MMS and was 72% for DMS.

Thymine glycols

Saturation of the 5-6 double bond of thymine to give various products occurs when cells are treated with oxidizing agents, ionizing radiation or u.v. One of these products, the thymine glycol, can be produced specifically in purified DNA by treatment with osmium tetroxide. This facilitated the isolation of a monoclonal antibody to this lesion in DNA by Leadon & Hanawalt (1983), who used it to demonstrate rapid removal of most of these adducts by BS-C-1 cells. Glycosylases that act on thymine glycols have been identified in bacteria and mammalian cells, and it is assumed that removal of this lesion is mediated by glycosylase/AP endonuclease. Exposure of cells to 25krad of gamma radiation produced 0-3 reactive site per 101 bases for both bulk and alpha DNAs, 90 % of which were removed from each species within 2 h.

Repair patch sizes in alpha and bulk DNA

We measured the average tract length of DNA synthesis associated with removal of adducts in both bulk and alpha DNA for most of the damaging agents used in these studies. Patch sizes were measured from density shifts of alpha DNA (172 bp) or bulk DNA reduced in length to about 200bp by sonication (Smith et al. 1981).

The density shifts observed both in bulk and alpha DNA from cells treated with u.v., AMT and NA-AAF (Zolan et al. 1982a), were all similar, and the patch size was calculated to be about 20 nucleotides. A similar shift was observed with the bulk DNA of cells treated with AFBi. However, the alpha DNA from these cells had a markedly smaller density shift indicating a patch size of about 10 nucleotides. Measurements on total DNA of XP (group A) cells and normal human fibroblasts treated with AFBi showed that patches in XP were about half the size of those in normal cells. A patch size of about eight nucleotides was calculated for both bulk and alpha DNAs from monkey cells treated with MMS. The absolute values for patch sizes cited here are less important than the clearly observable differences noted, which suggest that the average patch resulting from the glycosylase/AP endonuclease pathway is smaller than that made by the ‘bulky adduct’ pathway, deficient in XP cells. This patch size difference was originally documented with the bromo- uracil-photolysis technique (Regan & Setlow, 1974), although absolute sizes measured by that technique for bulky adducts have generally been greater than those we have measured. The smaller patch size observed for repair in alpha DNA containing AFBi adducts lends support to the hypothesis that a considerable portion of this repair synthesis results from repair of AP sites and that bona fide enzymic repair of the original AFBi adducts in alpha is severely deficient. For furocoumarin and NA-AAF adducts, the lower repair synthesis observed in alpha appears to reflect normal repair at some lower efficiency.

Enhancement of repair in alpha

With the exception of u.v., deficient repair in alpha appeared to be confined to bulky adducts. To explain the proficient repair of u.v. photoproducts, we suggested that u.v. irradiation might itself alter the chromatin structure of alpha, or that u.v. photoproducts might be more efficient at stimulating factors that generally render chromatin accessible to repair. We therefore assessed the effect of u.v. irradiation on removal of chemical damage (Leadonet al. 1983; Leadon & Hanawalt, 1984). Cells were u.v.-irradiated and treated with small amounts of labelled AFBi, and the amount of total adducts removed from bulk and alpha DNA was determined after 8h. u.v. had no effect on removal from bulk DNA, but it increased removal from alpha up to a dose of 5 J m-2. At this point it was the same in alpha as in bulk, and was not increased by higher u.v. doses. Increasing the time between u.v. irradiation and AFBi treatment decreased the effect in a manner that suggested the increased removal is a function of the absolute number of photoproducts remaining in the DNA at the time of AFBj treatment. The enhancement was not dependent upon new protein synthesis and was not observed when DMS or gamma rays were used in place of u.v. When 5 J tn-2 u.v. and NA-AAF were combined, removal of NA-AAF adducts was also enhanced. However, in two experiments we did not observe enhancement of removal of AMT adducts by u.v.

All the studies heretofore described used confluent cultures, in which the cells enter a G0-like state and very little DNA synthesis or cell division occurs. In actively growing cells one might expect changes in chromatin structure preparatory to DNA replication or cell division to affect repair in alpha. Leadon & Hanawalt (1986) showed that for actively growing cells containing low frequencies of AFB1 adducts, removal of adducts from alpha was nearly as fast as from bulk DNA. This was due both to an increased rate of removal from alpha and a decreased rate of removal from the bulk DNA, compared to confluent cells. This same result was obtained for repair in the 12-h period beginning 3 h after subculturing cells from the confluent state, a regimen that synchronizes the cells for about one cell cycle. Interestingly, very little removal was observed for either alpha or bulk DNA during the following 12-h period that corresponds to the 5 phase in this system, and removal was maximal for both species in a final 12-h period, corresponding to late S/G2- Similar results were obtained with NA-AAF treatment. We also examined repair replication in alpha DNA in growing cells in the 8h following exposure to 20 J m-2u.v. or to high levels of HMT or AFBi. For u.v., repair was the same in alpha and bulk DNA, and repair in alpha for the two chemical agents was indeed higher than we had found for confluent cells, being about 50 % of that in bulk DNA for AFB1 and 75 % for HMT.

Conclusions from study of alpha DNA

The most striking feature of the interactions of alpha DNA with damaging agents and repair systems is its complexity. In non-dividing cells, something about the organization of alpha severely restricts the cell’s ability to remove furocoumarin and AFBi adducts, both at frequencies that elicit maximal repair response in the bulk DNA and at frequencies 500-1000 times lower. Yet, overall frequencies of these adducts were the same in alpha and bulk DNA. It appears that repair systems for these adducts are somehow excluded from interacting with the DNA although the damaging agents are not. This exclusion does not extend to repair systems for AP sites, thymine glycols and alkylation damage under the same conditions. Repair deficiency for NA-AAF adducts is variable and appears to depend on adduct frequencies, u.v. damage presents a special case. In actively dividing cells all the repair deficiencies are eliminated or greatly reduced.

Our interpretation of these results involves ideas about the degree to which the DNA in chromatin must be made accessible to the incising activity of repair systems. Glycosylases and AP endonucleases are relatively small and recognize specific shortrange aspects of DNA structure. They may be able to penetrate even compact chromatin, and recognize their substrates even when the DNA is tightly associated with chromatin proteins.

Repair pathways for the bulky adducts we studied are known to share common steps because they are all poorly repaired in XP cells. The complexity of the recognition step for such adducts is revealed by the large number of complementation groups of XP cells. Part of this complexity is probably due to a multiple subunit structure of the actual incising activity, which is assumed to recognize many different bulky adducts. Studies of the UVR-ABC nuclease of Escherichia coli suggest that such an incising activity recognizes a damage-induced change in DNA conformation rather than specific features of adducts themselves (see Sancar et al. 1985). Part of this complexity of recognition probably also reflects the need for alteration of chromatin structure prior to incision, both for physical access of the repair enzymes to the damage in compact chromatin, and at the actual lesion site to allow the incising activity to interact with DNA unconstrained by nucleosome structure.

Extracts of some XP cells are able to promote excision of pyrimidine dimers in pure DNA or heterologous chromatin (Mortelmans et al. 1976; Fujiwara & Kano, 1983), and XP-C cells appear to confine repair to certain regions of the chromatin (Mansbridge & Hanawalt, 1983; Mullenders et al. 1987). These and results described below suggest the existence of systems to make lesions accessible to repair enzymes, but little information is available on how they might function. The complexity of such systems is probably related directly to that of chromatin itself. Domains containing active or potentially active genes may require less ‘opening up’ than those containing genes not competent for transcription and domains containing DNA such as alpha. A ‘scanning’ system may operate on some domains constitutively, but extend its activity to the entire genome only after triggering by DNA damage or its repair. Pyrimidine dimers may be especially efficient at triggering such a system, or their presence in domains of compact chromatin may directly facilitate its function in them. Cycling cells may simply maintain alpha chromatin in a more accessible state, perhaps due to its location at centromeres.

We are developing indirect methods to study repair in sequences in the cell that cannot be isolated in quantity. One such method uses bacteriophage T4 endonuclease V to provide specificity for u.v.-induced pyrimidine dimers and hybridization to specific probes to quantify repair in sequences present in as few as two copies per cell. The essential features of the method are as follows: after irradiation with u.v., cells are lysed either immediately or after incubation in bromodeoxyuridine for various periods. After proteinase digestion, high molecular weight DNA is extracted, freed of RNA, and digested with an appropriate restriction endonuclease. Unreplicated DNA isolated by centrifugation in CsCl is dialysed and concentrated. For each sample, two equal portions (2-10 µ g), one of which is treated with T4 endonuclease V), are electrophoresed in parallel lanes in 0 · 4% alkaline agarose gels, transferred to a support membrane and hybridized to 1P-labelled probes specific for the sequences of interest. In DNA not treated with T4 endonuclease, bands with the expected electrophoretic mobility of the full-length restriction fragments under study are revealed by autoradiography of the membrane. In samples treated with the nuclease the amount of hybridization at the full-length position is diminished due to nicking of those fragments containing one or more pyrimidine dimers. Assuming a random production of pyrimidine dimers, their actual frequency can be calculated from the fraction of fragments with no endonuclease-sites (ESS), using the Poisson expression. The amount of hybridization at the position of intact fragments for each sample is quantified by densitometry of the autoradiograms or, in the case of sufficiently amplified sequences, by scintillation counting of excised regions of the membrane.

This method places constraints on usable fragment sizes and u.v. doses. The gel must adequately resolve the full-length restriction fragments from the products of T4 endonuclease digestion. The initial frequency of pyrimidine dimers must be great enough to result in measurable reduction of the amount of full-length fragments, but still low enough that repair results in significant re-appearance of full-length fragments, at least in positive controls. These conditions are met with 15—25 kilobase (kb) fragments and u.v. doses of 10-20 J m-2. To date, our studies have concentrated on a few well-characterized genes, using detailed genomic restriction maps to guide the choice of restriction nucleases and probes. Probes from genomic clones are advantageous because they may be used to examine sequences outside transcription units, they often hybridize to a single restriction fragment, and they usually have extensive homology to the genomic fragments under study. They do often contain repeated sequences, however, which necessitates pre-hybridization of the membranes with a preparation of non-radioactive highly repeated DNA.

The method was initially developed using Chinese hamster cells (CHO) that contain an intrachromosomal amplification of the dihydrofolate reductase (DHFR) gene, to about 100 copies (Bohr et al. 1985). This provided sufficient hybridization to permit quantification by scintillation counting. The formation of ESS as a function of u.v. dose (5-40 J m-2) was 0 ·6 per 100 kb per J m-2 both for a 14 kb Kpnl fragment that comprises the 5’ half of the DHFR transcription unit and for a 22kb HindiII fragment, most of which extends 3’ to the gene. This value is in reasonable agreement with that obtained by chromatographic assay (0 ·8), by van Zeeland et al. (1981).

To measure repair we determine the disappearance of sites sensitive to T4 endonuclease V. We assume that this reflects removal of pyrimidine dimers, although any alteration of dimers that rendered them insensitive to the nuclease would appear as removal, unless the altered site was alkali-labile. The following summarizes results obtained with this method to date (May 1986).

Chinese hamster cells

Much of our work has focused on the well-characterized DHFR gene in CHO cells. A number of investigators have generously provided us with extensive and detailed maps, cell lines, and cloned genomic fragments. In cells containing an amplification of a large region that includes the DHFR gene, we determined that repair in the DHFR gene itself was much more efficient than in the genome overall (Bohr et al. 1985). After 10 Jm-2, almost 60% of the ESS were removed from a 14 kb fragment in the gene in 8h, and about 40% were removed after 20 J m-2. We confirmed the expectation from the work of others that overall removal of dimers was poor in these CHO cells using a sensitive technique in which permeabilized cells are treated with T4 endonuclease, and their DNA is analysed on alkaline sucrose gradients (van Zeeland et al. 1981). Only about 15 % of the ESS had been removed in 24 h in cells irradiated with 5 J m-2, a dose much lower than those used to measure repair in the DHFR gene. Repair in a 22kb fragment also in the amplified unit, but located at least 40 kb from the DHFR gene, resembled repair in the genome as a whole: less than 10 % of the ESS formed by 20 J m-2 were removed after 24 h. This indicated that preferential repair was not a property of the entire amplified unit.

CHO and other rodent lines survive u.v. about as well as human cells, even though human cells remove pyrimidine dimers very efficiently. However, mutants of CHO cells that have lost their residual capacity to remove dimers are u.v.-sensitive, like human XP cells, demonstrating that the excision repair observed in CHO cells is necessary for their considerable u.v. resistance (see Thompson et al. 1987). Our results with the DHFR gene suggest that the u.v. resistance of CHO cells is achieved by selective repair of actively transcribed sequences, like the DHFR gene, that are vital to cell growth.

The use of cells in which a sequence of interest is amplified greatly facilitates the analysis, but a general method requires assay of sequences present in only one or two copies per genome. A number of technical refinements were made in the hybridization procedure to accomplish this, and we used scanning densitometry to quantify the hybridization. Repair in the parental CHO line was found to be similar to that in the amplified line (Bohr et al. 1986a). After 20 J m-2, removal of ESS was substantial in the 14 kb intragenic fragment, but not detected in the fragment distal to the gene. Removal in the genome overall following 5 J m-2 was again about 15%, confirming that the heterogeneity in repair is a property of the CHO cells and not related to sequence amplification.

More recently the region around the DHFR gene has been studied in more detail, again using the cells in which these sequences are amplified (Bohr et al. 19866). Repair was assayed at 8 and 24 h following 20 J m”1 u.v. Removal of ESS from a 14 kb Kpnl fragment whose 5’ end lies only about 8 kb from the 3’ end of the transcription unit was nearly undetectable after 8h and only 21 % after 24 h. Thus a sequence downstream from the transcription unit but very close to it exhibited the deficient repair characteristic of the overall genome. Within the gene itself, repair in a 7 kb fragment that lies at the 3 ‘end of the gene was only about half of that observed for the 14 kb fragment that encompasses the 5’ end of the gene. About 7 kb of unassayed DNA separates these two fragments. On the 5’ side of the gene, repair was also very efficient. A fragment that includes the 5’ end of the gene but also extends out about 19 kb in the 5’ direction was repaired as well as the 14 kb fragment in the 5’ end of the gene.

It has recently been reported that a transcription unit with polarity opposite to the DHFR gene is located just upstream from the DHFR 5’ control region (Mitchell et al. 1986). The regions of most efficient repair thus correspond to the control region and the 5′ ends of two transcription units. The region of efficient DNA repair in the DHFR locus is roughly 60 kb, a length similar to that calculated for some kind of structural ‘domain’ in chromatin by a number of different techniques (for review, see Pienta & Coffey, 1984). The low level of repair observed in fragments just 3 ′ to the transcription unit may reflect abrupt transition to chromatin structure characteristic of non-transcribed DNA, or it may result from transition to some other level of chromatin organization. The apparently lower repair in the 3 ′ region of the transcription unit itself may in part reflect the high u.v. doses necessary to study repair in the relatively small fragments available in that region, which place several dimers in the possibly more accessible 5 ′ region. Changes in chromatin structure along the gene or some processivity in repair may also account for this result. Detailed analysis of other genes and the surrounding sequences will be necessary to determine which if any of these features are general ones.

An experimental difficulty encountered with CHO cells is their extensive amount of replication: even after 20Jm-2 u.v., about half the DNA is replicated in 24h. After longer times most of the DNA is therefore unavailable for analysis. To examine repair in non-cycling cells we used serum starvation, and followed repair for 48 h in the 14 kb fragment in the 5 ′ half of the gene and in the 14 kb fragment just downstream from the transcription unit. Replication under these conditions was only 4 %, and the difference in repair exhibited by the two fragments was about the same as observed in growing cells. After 48 h about 80% of the ESS were removed from the fragment in the gene and less than 20% were removed from the downstream fragment. Repair at early times also resembled that in the growing cells. These results indicate that the unreplicated DNA we assay in growing cells does not represent a selected population with abnormal repair characteristics, and that the heterogeneity in repair we observe is maintained for at least 48 h.

We have also begun to study the gene for hydroxymethyl glutaryl coenzyme A reductase, the pivotal gene in cholesterol biosynthesis, which has served as model for study of end-product regulation in mammalian cells (Osborne et al. 1985). Its rate of transcription in cultured cells can be varied over a large range by altering the lipid content of the medium. The low density lipoproteins (LDLs) present in serum result in a reduction of enzyme activity to about 5% of that obtained in the absence of LDLs, mainly due to repression of the gene. Initial experiments with CHO cells growing in normal medium showed that repair of a large fragment wholly within the transcription unit is as efficient as that in the DHFR gene. Thus even moderately low activity may be sufficient to facilitate repair.

Mouse cells

Like CHO cells, cultured mouse cells exhibit only limited removal of pyrimidine dimers from their genomes overall (Yagi et al. 1984). We measured removal of ESS 24 h after a dose of 20 J m-2 in three different sequences in mouse 3T3 cells: a region upstream from the DHFR gene using a genomic probe, and the proto-oncogenes c-abl and c-mos using retroviral probes (Madhani et al. 1986a). Repair was efficient in a 20 kb intragenic fragment of the c-abl transcription unit, both in actively growing cells (60 % removed) and in contact-inhibited cells (85 % removed). This gene is known to be actively transcribed in these cells. In constrast, repair was relatively deficient in a 27 kb fragment upstream from the DHFR locus, a result similar to that observed for CHO cells. It was also deficient in fragments containing the silent c-mos locus. This locus contains a 14 kb open reading sequence identified by hybridization to retroviral probes. This sequence is hypermethylated and not represented in RNA in cultured cells, but is represented in RNA in mouse gonadal tissue and early embryos. A 15 kb fragment containing c-mos exhibited 22 % removal in confluent cells, and 9% in growing cells. A 6-2 kb fragment containing the sequence exhibited little or no removal. These results extend the observation of preferential repair of active genes to another rodent system, and suggest that genes repressed for developmental reasons are not preferentially repaired. Implications for the role of repair in carcinogen-promoted activation of proto-oncogenes are discussed by Madhani et al. (1986a).

Human cells

Most pyrimidine dimers are removed by human cells in about 24 h; thus one would expect that preferential repair of active sequences might relate to the rate of their removal. Indirect evidence for rapid repair of active sequences was reported by Mayne & Lehmann (1982), who found that the rate of RNA synthesis in irradiated cells recovers after irradiation before overall repair is complete. To examine human cells for preferential repair we have used a line containing an intrachromosomal amplification for the region containing the human DHFR gene (Mellon et al. 1986). We analysed two different fragments, a 20 kb fragment located in the centre of the 30 kb gene, and a 25 kb fragment that contains only a few hundred bp of the 5′ end of the gene and extends out in the 5’ direction. For both of these fragments, more than 60% of the dimers had been removed by 4h after 10 J m-2, a rate considerably greater than we expected for the genome as a whole, judging from our own previous work and published reports. We measured the removal of pyrimidine dimers in the overall genome of these same cells by treating permeabilized cells with T4 endonuclease V and analysing the molecular weight of the DNA on alkaline sucrose gradients. Only about 25 % had been removed in 4h.

To ensure that the repair differences were not the results of using two different techniques, we devised a method to measure removal of ESS from DNA containing the DHFR gene region by the same technique used to assay total DNA. Besides determining the molecular weight average of the (T4 endonuclease-treated and untreated) mass-labelled cellular DNA in the sucrose gradients, we transferred the DNA in the gradient fractions of duplicate gradients to nitrocellulose, and hybridized to a 1· 8 kb genomic probe, located at the 5′end of the gene. The molecular weight distribution for DNA that hybridized to the probe was identical to that for the total DNA for cells analysed immediately after irradiation. However, 4 or 8h after irradiation, the molecular weight of DHFR-containing DNA was considerably greater than that of the total DNA. This was not due to replication of the DHFR-containing sequences, which was less than 6%. In addition, using a probe for human alpha DNA, we found no such differences in molecular weights, indicating, as expected, that repair in alpha resembled that in the genome as a whole. Calculations of the amount of removal (Table 3) were consistent with the values obtained by Southern analysis. The fragments containing the sequence that hybridizes to the probe could contain variable amounts of sequences outside the region in which rapid repair occurs. This would limit the increase in molecular weight used as an indicator of repair to the size of this region, resulting in an underestimate of repair within it.

Table 3.

Removal of pyrimidine dimers (ESS) by human (6A3) cells

Removal of pyrimidine dimers (ESS) by human (6A3) cells
Removal of pyrimidine dimers (ESS) by human (6A3) cells

These results indicate that at least the DHFR region is repaired in a preferential manner in these cells: the rate of its repair is considerably greater than that of the average sequence in the cell. The rapid repair of the fragment 5′ to the gene could indicate a divergent transcription unit, as has been shown in CHO and mouse cells, or might reflect some minimum size for preferentially repaired regions. We consider it unlikely that the preferential repair is related to amplification; preliminary experiments with unamplified cells also suggest rapid repair of the DHFR gene. We are currently examining DHFR and other genes in various human fibroblasts. The technique using slot blotting is sensitive enough for analysis of single-copy genes and promises to be useful for analysing repair of several different sequences in a single experiment without the requirement for extensive DNA preparation and digestion with different restriction nucleases.

We also compared the extent of repair in the DHFR gene in normal human cells to that in repair-deficient XP-C cells (Bohr et al. 1986a). Like CHO cells, XP-C cells remove only a small fraction of u.v.-induced pyrimidine dimers from their genomes, but unlike CHO cells, they are much more sensitive to u.v. than normal human cells. Recently it was demonstrated that the repair that does take place in XP-C cells seems to be clustered in certain regions of the genome (Mansbridge & Hanawalt, 1983), and also that this repair is associated with the nuclear matrix (see Mullenders et al. 1987). Measured 24h after u.v., the repair in a 23 kb HindIII fragment that included the 5′ two-thirds of the DHFR gene was found to be nearly complete (70—90 %) in cultured normal skin fibroblasts or keratinocytes, but very low (0—20%) in the XP-C cells. This indicates that the repaired regions in XP-C do not include all active genes. Taking the results for CHO and human XP-C cells together, it appears that u.v. resistance correlates better with efficient repair of active genes than with overall repair levels.

Our study of repair in coding sequences is only in its initial stage, but has already provided important insights. The demonstration of efficient repair in several genes in rodent cells suggests that these cells achieve high u.v. resistance despite low overall repair levels by repairing those sequences necessary for cell function. Whether this selective repair is a consequence of culture in vitro and whether it occurs for lesions other than pyrimidine dimers remains to be determined. For human cells, selective repair has been detected as a more rapid rate of repair for active sequences, although additional selectivity in the extent of repair could occur, but would be difficult to detect. If responsible for the rapid recovery of RNA synthesis reported by Mayne & Lehmann (1982), this selective repair would be an important factor in the resistance to u.v. of human cells, since rapid recovery of RNA synthesis appears deficient in u.v.-sensitive Cockayne’s syndrome cells. At present we cannot adequately compare the rates of repair of active genes in human and rodent cells, because of the different u.v. doses and incubation times used.

Our results also indicate that damage in silent regions of the genome may have greater potential for engendering mutations and DNA re-arrangements than damage in or near transcriptive units.

The preferential repair of a given sequence could reflect greater accessibility, perhaps due to different chromatin condensation or to location in more accessible areas of the nucleus. The rodent cells may lack (or have lost during culture) a system present in human cells that renders the remaining DNA accessible.

Whatever its cause, the preferential DNA repair we observe mandates caution interpreting correlations, whether positive or not, between overall repair capacity and other biological parameters. Defects in preferential repair could have profound effects on such parameters without noticeably altering overall repair levels.

Research in our laboratory has been supported by grants to P. C. Hanawalt from the National Insitutes of Health of the U.S. Public Health Service, the American Cancer Society, and the U.S. Department of Energy. We are grateful to R. Schimke, R. Simoni, L. Chasin, G. Attardi and J. Hamlin for supplying us with cell lines and probes, P. Seawell and E. Wauthier in our laboratory for preparation of T4 endonuclease V, and many colleagues who have supplied expertise and materials.

Bohr
,
V. A.
,
Okumoto
,
D. S.
&
Hanawalt
,
P. C.
(
1986a
).
Survival of UV-irradiated mammalian cells correlates with efficient DNA repair in an essential gene
.
Proc. natn. Acad. Sei. U.S.A.
83
,
3830
3833
.
Bohr
,
V. B.
,
Okumoto
,
D. S.
,
Ho
,
L.
&
Hanawalt
,
P. C.
(
1986b
).
Characterization of a DNA repair domain containing the dihydrofolate reductase gene in Chinese hamster ovary cells
.
J. biol. Chetn. (in press)
.
Bohr
,
V. B.
,
Smith
,
C. A.
,
Okumoto
,
D. S.
&
Hanawalt
,
P. C.
(
1985
).
DNA repair in an active gene: removal of pyrimidine dimers from the DHFR gene of CHO cells is much more efficient than in the genome overall
.
Cell
40
,
359
369
.
Fujiwara
,
Y.
&
Kano
,
Y.
(
1983
). Characteristics of thymine dimer excision from xeroderma pigmentosum chromatin. In
Cellular Responses to DNA damage
(ed. E. C. Friedberg &
B. A.
Bridges
), pp.
215
221
.
New York
:
Alan R. Liss
.
Kanne
,
D.
,
Straub
,
K.
,
Hearst
,
J. E.
&
Rapoport
,
M.
(
1982
).
Isolation and characterization of pyrimidine-psoralen-pyrimidine photoadducts from DNA
.
J. Am. Chem. Soc.
104
,
6754
6764
.
Kuff
,
E. L.
, Ferdinand, F.-J. &
Khoury
,
G.
(
1978
).
Transcription of host-substituted simian virus 40 DNA in whole cells and extracts
.
J. Virol.
25
,
28
36
.
Leadon
,
S. A.
&
Hanawalt
,
P. C.
(
1983
).
Monoclonal antibody to DNA containing thymine glycol
.
Mutat. Res.
112
,
191
200
.
Leadon
,
S. A.
&
Hanawalt
,
P. C.
(
1984
).
Ultraviolet irradiation of monkey cells enhances the repair of DNA adducts in alpha DNA
.
Carcinogenesis
5
,
1505
1510
.
Leadon
,
S. A.
&
Hanawalt
,
P. C.
(
1986
).
Cell cycle dependent repair of damage in alpha and bulk DNA of monkey cells
.
Mutat. Res.
166
,
71
77
.
Leadon
,
S. A.
,
Tyrrell
,
R. M.
&
Cerutti
,
P. A.
(
1981
).
Excision repair of aflatoxin Bl-DNA adducts in human fibroblasts
.
Cancer Res.
41
,
5125
5129
.
Leadon
,
S. A.
,
Zolan
,
M. E.
&
Hanawalt
,
P. C.
(
1983
).
Restricted repair of aflatoxin Bl induced damage in alpha DNA of monkey cells
.
Nucl. Acids Res.
11
,
5675
5689
.
Madhani
,
H. D.
,
Bohr
,
V. A.
&
Hanawalt
,
P. C.
(
1986a
).
Differential DNA repair in transcriptionally active and inactive proto-oncogenes: c-abl and c-mos. Cell
45
,
417
422
.
Madhani
,
H. D.
,
Leadon
,
S. A.
,
Smith
,
C. A.
&
Hanawalt
,
P. C.
(
1986b
).
Alpha DNA in African Green Monkey cells is organized into extremely long tandem arrays
.
J. biol. Chem.
261
,
2314
2318
.
Maio
,
J. J.
(
1971
).
DNA strand reassociation and polyribonucleotide binding in the African green monkey, Cercopithecus acthiops.J. molec Biol.
56
,
579
595
.
Mansbridge
,
J. N.
&
Hanawalt
,
P
, C
. (
1983
). Domain limited repair of DNA in ultraviolet irradiated fibroblasts from xeroderma pigmentosum complementation group C. In
Cellular Responses toDNADamage, UCLA Symp. on Mol. and Cell. Biol. New Series, vol. 2
(ed. E. C. Friedberg &
P. C.
Hanawalt
), pp.
195
207
.
New York
:
Alan R. Liss
.
Mayne
,
L. V.
&
Lehmann
,
A. R.
(
1982
).
Failure of RNA synthesis to recover after UV- irradiation: An early defect in cells from individuals with Cockayne’s syndrome and xeroderma pigmentosum
.
Cancer Res.
42
,
1473
1478
.
Mellon
,
L
,
Bohr
,
V. B.
,
Smith
,
C. A.
&
Hanawalt
,
P. C.
(
1986
).
Preferential DNA repair of an active gene in human cells
.
Proc. natn. Acad. Sci. U.S.A. (in press)
.
Mitchell
,
P. J.
,
Carothers
,
A. M.
,
Han
,
J. H.
,
Harding
,
J. D.
,
Kas
,
E.
,
Venolia
,
L.
&
Chasin
,
L. A.
(
1986
).
Multiple transcription start sites, DNase I hypersensitive sites, and an opposite-strand axon in the 5’ region of the CHO DHFR gene
.
Molec. cell. Biol.
6
,
425
440
.
Mortelmans
,
K.
,
Friedberg
,
E. C.
,
Slor
,
H.
,
Thomas
,
G.
&
Cleaver
,
J. E.
(
1976
).
Defective thymine dimer excision by cell-free extracts of xeroderma pigmentosum cells
.
Proc. natn. Acad. Sci. U.S A.
73
,
2757
2761
.
Mullenders
,
L. H. F.
, van
Zeeland
,
A. A.
&
Nataraian
,
A. T.
(
1987
).
The localization of ultraviolet light-induced excision repair in the nucleus and the distribution of repair events in higher-order chromatin loops in mammalian cells. J’
.
Cell. Sci. Suppl.
6
,
243
262
.
Musich
,
P. R.
,
Maio
,
J. J.
&
Brown
,
F. L.
(
1977
).
Subunit structure of chromatin and the organization of eukaryotic highly repetitive DNA
.
J. molec. Biol.
117
,
657
677
.
Osborne
,
T. F.
,
Goldstein
,
J. L.
&
Brown
,
M. S.
(
1985
).
5’ end of HMG CoA reductase gene contains sequences responsible for cholesterol-mediated inhibition of transcription
.
Cell
42
,
201
212
.
Peckler
,
S.
,
Graves
,
B.
,
Kanne
,
D.
,
Rapoport
,
H.
,
Hearst
,
J. E.
&
Kim
,
S.-H.
(
1982
).
Isolation and characterization of pyrimidine-psoralen photoadducts from DNA
.
J. molec. Biol.
162
,
157
172
.
Pienta
,
K. J.
&
Coffey
,
D. S.
(
1984
).
A structural analysis of the role of the nuclear matrix and DNA loops in the organization of the nucleus and chromosome
.
J. Cell Sci. Suppl.
1
,
123
135
.
Regan
,
J. D.
&
Setlow
,
R. B.
(
1974
).
Two forms of repair in the DNA of human cells damaged by chemical carcinogens and mutagens
.
Cancer Res.
34
,
3318
3325
.
Rosenberg
,
H.
,
Singer
,
M.
&
Rosenberg
,
M.
(
1978
).
Highly reiterated sequences of simiansimiansimiansimiansimian
.
Science
200
,
394
402
.
Sancar
,
A.
,
Franklin
,
K. A.
,
Sancar
,
G.
&
Tang
,
M. S.
(
1985
).
Repair of psoralen and acetylaminofluorene DNA adducts by ABC exinuclease
.
J. molec. Biol.
184
,
725
734
.
Segal
,
S.
,
Garner
,
M.
,
Singer
,
M. F.
&
Rosenberg
,
M.
(
1976
).
In situ hybridization of repetitive monkey genomic sequences isolated from defective simian virus 40 DNA
.
Cell
9
,
247
257
.
Singer
,
M. F.
(
1982
).
Highly repeated sequences in mammalian genomes
.
I nt. Rev. Cytol.
76
,
67
112
.
Smith
,
C. A.
,
Cooper
,
P. K.
&
Hanawalt
,
P. C.
(
1981
). Measurement of repair replication by equilibrium sedimentation. In
DNA Repair: A Laboratory Manual of Research Procedures, vol. 1
(ed. E. C. Friedberg &
P. C.
Hanawalt
), pp.
289
305
.
New York
:
Marcel Dekker
.
Solomon
,
M. J.
,
Strauss
,
F.
&
Varshavsky
,
A.
(
1986
).
A mammalian high mobility group protein recognizes any stretch of six A-T base pairs in duplex DNA
.
Proc. natn. Acad. Sci. USA.
83
,
1276
1280
.
Strauss
,
F.
&
Varshavsky
,
A.
(
1984
).
A protein binds to a satellite DNA repeat at 3 specific sites that would be brought into mutual proximity by DNA folding in the nucleosome
.
Cell
37
,
889
901
.
Thompson
,
L. H.
,
Salazar
,
E. P.
,
Brookman
,
K. W.
,
Collins
,
C. C.
,
Stewart
,
S. A.
,
Busch
,
D. B.
&
Weber
,
C. A.
(
1987
).
Recent progress with the DNA repair mutants of Chinese hamster ovary cells
.
J. Cell Sci. Suppl,
6
,
97
110
.
van Zeeland
,
A. A.
,
Smith
,
C. A.
&
Hanawalt
,
P. C.
(
1981
).
Introduction of T4 endonuclease V into frozen and thawed cells
.
Mutât. Res.
82
,
173
189
.
Wu
,
J. C.
&
Manuelidis
,
L.
(
1980
).
Sequence definition and organization of a human repeated DNA
.
J. molec. Biol.
142
,
363
.
Wu
,
K. C.
,
Strauss
,
F.
&
Varshavsky
,
A.
(
1983
).
Nucleosome arrangement in green monkey alpha-satellite chromatin: superimposition of non-random and apparently random patterns
.
J. Molec. Biol.
170
,
93
117
.
Yagi
,
T.
,
Nikaido
,
O.
&
Takebe
,
H.
(
1984
).
Excision repair of mouse and human fibroblast cells, and a factor affecting the amount of UV-induced unscheduled DNA synthesis
.
Mutat. Res.
132
,
101
112
.
Zolan
,
M. E.
,
Cortopassi
,
G. A.
,
Smith
,
C. A.
&
Hanawalt
,
P. C.
(
1982a
).
Deficient repair of chemical adducts in alpha DNA of monkey cells
.
Cell
28
,
613
619
.
Zolan
,
M. E.
,
Smith
,
C. A.
,
Calvin
,
N. M.
&
Hanawalt
,
P. C.
(
1982b
)
Rearrangement of mammalian chromatin structure following excision repair
.
Nature, Lond.
299
,
462
464
.
Zolan
,
M. E.
,
Smith
,
C. A.
&
Hanawalt
,
P. C.
(
1984
).
Formation and repair of furocoumarin adducts in alpha deoxyribonucleic acid and bulk deoxyribonucleic acid of monkey cells
.
Biochemistry
23
,
63
69
.