Carcinogenesis vol. 12 no. 11 pp. 1983-1992, 1991

COMMENTARY

Gene specific DNA repair

Vilhelm A.Bohr

Laboratory of Molecular Pharmacology, National Cancer Institute, Bldg 37Rm 5C-25, NIH, Bethesda, MD 20892, USA

IntroductionThere is increasing interest in gene-specific repair studies. Thisaspect of DNA repair is currently under intense investigation andnew laboratories are entering the field. The data are coming inrapidly and we are in the midst of exciting developments.Accordingly, it is premature to build extensive models of theregulation and biological importance of gene specific repair. Thepurpose of this review is (i) to explain the methodology involvedin these studies, (ii) to clarify some established concepts, (iii)to provide an update of the status of gene and strand specificDNA repair and (iv) to point out some of the directions in whichthe field is moving.

DNA damageWhereas UV irradiation mainly introduces two lesions into theDNA, the pyrimidine dimer and the 6—4 photoproduct, mostother DNA-damaging agents introduce a spectrum of lesions.Some common lesions are schematized in Figure 1, and someexamples will be mentioned. For chemotherapeutic agents, forexample, the adducts formed range from simple methylations tocomplex bulky monoadducts to crosslinks between DNA strandsor between DNA and protein. Cisplarin is the prototypical heavymetal compound with antitumor activity, and the most frequentlesion formed is the intrastrand adduct between adjacent purines(GG or AG) (1). The interstrand crosslinks acount for only — 1 %of the total DNA lesions (2), and DNA protein crosslinks canalso occur at low frequencies. AJkylating agents include simplemonofunctional methylating agents such as methylmethane-sulfonate aid bifunctional agents such as nitrogen mustard, whichforms both monoadducts and interstrand DNA crosslinks. Thetwo predominant lesions are the alkylations of the N1 and O6

positions of guanine but many other base modifications arepossible and the spectrum varies for the individual agents (3).Bleomycin, an antitumor antibiotic, intercalates into DNA andproduces free radicals which then create both single- and double-strand breaks in DNA (4). Psoralen is a naturally occurringcompound that binds to DNA and forms two types of photo-adducts after exposure to UVA light: monoadducts and inter-strand crosslinks. PUVA (psoralen plus UVA light) has been usedto treat the malignancy mycosis fungoides and other dermato-logical disorders, notably psoriasis. Other environmentalcarcinogens cause a number of bulky adducts in DNA.

•Abbreviations: XP, xeroderma pigmentosum; BrdUrd, bromodeoxyuridine;DHFR, dflrydrofolate reductase; 4NQO, 4-nitroquinoline-l-oxide; ADA, adenosinedeaminase; NAAAF, JV-acetoxy-acetylaminofluorene.

DNA repair pathwaysThe DNA repair processes are a complex set of reactions thatin a sense act as an immune surveillance system at the DNA level,recognizing and removing a multitude of adducts from the geneticmaterial. Based on estimations in Escherichia coli, as many as30% of all genes are in one way or another involved in theseprocesses, and it is thus likely that a sizeable number of the 105

mammalian genes are involved in some aspect of thesemechanisms. In the human DNA repair-deficient disorder,xeroderma pigmentosum (XP), there are now eight documentedcomplementation groups, suggesting at least that number ofdifferent genes in the early steps of UV excision repair. In yeast,at least 50 DNA repair-deficient mutants have been identified.A number of human repair genes have been cloned (for recentreview, see 5); from the sequence of these genes it appears thatmany of the products are helicases, but complete characterizationawaits the isolation and biochemical studies of these repairenzymes.

The major forms of DNA repair include direct reversal,excision repair and post-replication repair. The most commonand most relevant form of DNA repair to this discussion isexcision repair which is the pathway of repair for a wide rangeof DNA lesions. Two basic types of excision repair have beendescribed in bacteria (6). One system, base excision repair,removes smaller types of base damage such as some lesionscaused by ionizing radiation or monofunctional alkylating agents.Another system, nucleotide excision repair, removes bulky DNAadducts of which the UV-induced pyrimidine dimer is theprototype. The basic steps involved in excision repair are shownin Figure 2. They include: (i) preincision recognition of DNAdamage, (ii) incision of the damaged DNA strand near the defect,(iii) excision of the defective site with concomitant degradationof the excised DNA strand, (iv) replication of a repair patch toreplace the excised strand with the opposite (intact) strand beingused as a template and (v) ligation of the repair patch at its 3'end. The resulting patch in nucleotide excision repair is muchlarger (20-100 bases) than in base excision repair (1 - 5 bases).

Although some of the enzymes involved in prokaryotic DNArepair have been characterized, their mammalian counterpartsare less well described. Two prokaryotic enzymes, T4endonuclease V and uvr ABC excinuclease, play roles in theexcision pathways of E. coli and have proven particularly usefulin the study of gene-specific repair. The T4 endonuclease V,which recognizes cyclobutane pyrimidine dimers, has both DNAglycosylase and AP endonuclease activities resulting in a single-stranded nick in DNA at the site of a pyrimidine dimer (Figure2A). This enzyme has been used in our assay for gene-specificrepair (7). The E.coli ABC excinuclease, encoded by the urvA,uvrB and uvrC genes, recognizes a wide range of bulky DNAadducts. These proteins act in concert to accomplish the first threesteps of excision repair: recognition, incision and excision, asoutlined in Figure 2(B). This enzyme has also been used in gene-specific repair studies because the damage formed by several

© Oxford University Press 1983

V.A.Bohr

MODIFIED BASElAlkylrfrtg Again)

BASE LOSS(Alkyiiting Agents)

PYRIMIOINE DIMER BULKY ADDUCT(UVI INAAAFI

INTERSTRAND CROSSLINK(Nitrogen Muiard. Otplaun,

Pioralenl

lNTHA<rrRAwn STRAND BHEAK a.ADDUCT (Bloomydn, X-«v) pHOTOPROOUCTICJsplMnl (UVI

Fig. 1. Lesions formed by some types of DNA damage.

BASIC MODELS FOR EXCISION REPAIR

A.

•iiITIiiii I ii

I UV-Specific Endonucleast(Glycosylase + AP Endomuclsase)

I I I I I I I I I I ' I I I II I I I I I

I

ill i ' i n

DNA Polymerase

I

IUVRABC

\ ,OH

Excision ONA Polymerate

I I I I I I I I I '• I I I I I I I I I

Ugase

I I I I I I I I I I I I I I I i I 11 I I I IT

i i i i i i i m n i m i n i i i i i i

I TTIHTITTTM

Fig. 2. Some basic steps in nucleotide excision repair. Two models are shown, a classic scheme involving the T4 endonuclease (A) and the more currentmodel involving uvr ABC excinuclease (B). The steps in repair are (i) recognition, (ii) incision, (iii) polyermization, (iv) degradation and (v) ligation.

chemotherapeutic agents, including cisplatin, mitomycin C andpsoralen, can be recognized by the enzyme (8).

Genome organization and gene-specific DNA repairIn considering the relevance of gene-specific DNA repair it isappropriate to reflect on some aspects of nuclear structure. TheDNA of the mammalian cell is highly organized and packagedinto higher-order chromatin structures. Figure 3 illustrates thebasic levels of DNA organization. The DNA in the nucleus (A)represents the overall genome and it is at this level that mostDNA repair studies have been carried out. But as shown in Figure3, higher-order structure loops of the chromosome, the nuclearmatrix (C), the nucleosomes (D), the double helix (E) and itsbase constitutents (F) all represent levels for precise regulationand control of the DNA for replication, transcription and repair.The complex packaging and 50 000-fold compression of the DNAin the nucleus plays a major role in the regulation of transcriptionand replication and is bound also to affect the regulation of DNArepair processes. In fact, evidence is mounting that the DNArepair processes are intimately linked with these other DNA-related processes, and this will in part be discussed later. Genesare juxtapositioned to the nuclear matrix (C) at times of tran-scription and replication and possibly DNA repair. After lowdoses of UV damage, the repair label is found in the nuclear

1984

matrix associated fraction, whereas after higher doses of UV thereis no such association (9). At the level of the core and linkerregions (D) there is now some information about damage andrepair (10). The inset in Figure 3 shows replicating and tran-scribing DNA sequences as they relate to DNA repair, to bediscussed in the following. Almost all of the work in DNA repairover the last decades has been done by measuring the events inthe total genome. It is now established that there is considerableintragenomic heterogeneity of the DNA repair processes and thiswill affect our thinking and experimentation in this field. Sincethe genes only constitute a small fraction of the overall genome(<1%), any changes in gene-specific repair would not bedetected using traditional assays at the level of the overall genome.

MethodologyThe prototype gene-specific repair technique was first developedto study the repair of UV damage by quantitating the removalof pyrimidine dimers in gene fragments. The process will besummarized here, but for a more complete protocol, see Bohrand Okumoto (11). The key is to generate a lesion-specific strandbreak at the site of the damage and then measure the formationand repair of the strand breaks by quantitative Southern analysis.A flow diagram of the procedure used for quantifying DNAadducts is shown in Figure 4. In the standard UV experiment,the cells are prelabeled with [3H]thymidine to allow for later

Gene-specific DNA repair

Fig. 3. Levels of organization in the cellular nucleus. (A) Nucleus; (B) minibands on chromosome; (C) higher-order loops with genes positioned close to thenuclear matrix; (D) nucleosomes; (E) primary structure of DNA. (F) base. Inset also shows the replication and transcription of DNA (relation to DNArepair).

separation in an equilibrium gradient. After irradiation with UVlight at 254 nm, the cells are allowed to repair in the presenceof bromodeoxyuridine (BrdUrd) to density label the replicatedDNA. The DNA is extracted and purified from each sample andrestricted with the appropriate restriction endonuclease to detectlater the fragment of interest. The parental DNA is then separatedfrom the semiconservatively replicated DNA by CsCl equilibriumgradient centrifugation. In this experiment, as in other repairassays, it is important to remove the replicated DNA, whichwould otherwise be measured as repaired. At this point, the DNAis treated with the T4 endonuclease V enzyme to generate single-stranded breaks at the pyrimidine dimers and the sample is runin parallel with non-treated sample on an alkaline gel, transferredto a nylon membrane and probed for the sequence of interest.A ratio of the treated and untreated DNA band intensities givesthe fraction of single strands that are free of damage (the zeroclass) and the number of lesions per fragment is calculated basedon the Poisson distribution. An example of the type of datagenerated with this approach is shown in Figure 5. The formationand repair of pyrimidine dimers is shown in the dihydrofolatereductase (DHFR) 5' gene fragment, G in Figure 6.

Alternative methods for measuring gene-specific repair havebeen reported. An antibody toward a specific adduct or towardBrdUrd which is incorporated in the repair patch can be usedto precipitate selectively DNA fragments that contain damageor, respectively, that are repaired (12). These assays require goodantibodies and a high degree of sensitivity; and to be quantitative,the antibody must detect and co-precipitate a DNA fragmentcontaining only one lesion or BrdUrd substitution. When theBrdUrd antibody is used, it has the advantage that it can detectthe repair of a number of different lesions, but has thedisadvantage that the initial level of damage cannot be determined.

Lately, polymerase chain reaction (PCR) methods have beendeveloped where the repair of small gene fragments can bemeasured. In an early report by Govan et al. (13), a very highdose of 3 kJ/m2 was used, rendering experimental conditionsnon-physiological. But there are indications that this approachcan now be used at more physiological levels by using otheragents and by using larger gene fragments for the studies. It isa promising approach, and has the advantage that various lesions

rDamage Cells

Incubate for Repairen presence o) BUdR

DMIXTMCTIONAND irarnticnoN

CROSSLINK*

HL a

CsCtgradent

ADOUCTS

DenaturatkwRearmeaflng

Neutral agarou gelSouthern Wot

Cleavage{enzymatic or chemical)

Alkaline agarose gelSouthern blot

lintDS-ss-

0

—

24 0

- +

•—

24

- +Hre

cleavagegene

Fig. 4. Methodology for gene-specific repair analysis of adducU andcrosslinks. The steps are described in the text. For adducts, the key point isto create strand breaks (cf. Table I) at sites of damage, and then resolve thesingle-stranded DNA in alkaline gels for Southern probing and analysis, cf.Figure 5. The crosslinked DNA fragments are detected as double-strandedDNA after denaturation and reanneaJing, cf. Figure 8.

can be studied using the same approach. In yet another approach,Bianchi et al. (14) studied the cutting by restriction endonucleasesat rare sites in the c-myc gene. Lesions formed at the recognitionsequence after UV damage block the cutting, and the loss andreappearance of an extra band can thus be detected by Southernanalysis. In this method, the replicated DNA is not distinguished

1985

V.A.Bohr

T4 endo24

1 8

1 6

1.4

1 2

1.0

0 8

0 6

0 4

0 2

0 0

•

- •

-

-

-

1 c 1

0 4 8

Hrs

-

• i i i

12 16 20 24

Repair

Fig. 5. Repair experiment in the hamster DHFR gene. Cells were irradiatedwith 20 J/m2 and allowed to repair for 4, 8 and 24 h. Samples wereanalyzed as indicated in Figure 4, and probed for the 14 kb 5' side of theDHFR gene, cf. Figure 6. The pyrimidine dimers per fragment werequantitated and are plotted in the graph.

DHFR

N

Fig. 6. Map of the hamster DHFR domain. G, the 14 kb region analyzed inthe coding, 5' half of the DHFR gene; N, the downstream, non-codingregion, also 14 kb.

from the repaired, and the pyrimidine dimer sites cannot bedistinguished from those of other photoproducts. However, thismethod can be used to detect a broad spectrum of DNA lesions.

Preferential DNA repairIt was initially shown that the active, essential DHFR gene wasrepaired very efficiently after UV damage in hamster cells wherethe overall genome repair was low (7). This phenomenon hasbeen called 'preferential DNA repair', and it has been observedfor a number of active genes in various mammalian cell lines(15). In human cells, the overall genome DNA repair is pro-ficient, and repair is almost complete in 24 h. The preferentialrepair of the active gene is manifested as a faster repair rate thanin the overall genome, but unlike rodent cells, all genomic regionsare proficiently repaired. These and other characteristics involvedin preferential repair are shown in Figure 7 to clarify the concept.The plot shows different levels of repair of UV-inducedpyrimidine dimers in different gene fragments in mammaliancells; it also shows the preferential repair in the active DHFRgene in hamster and human cells. It is further seen in Figure

1986

D

OT•

•

Humti

Humii

* c -myc

1 DHFBCHO DHFKHoratic«wk«:CHO er

] Or«raU

nrill

Fig. 7. Preferential DNA repair. This plot shows different aspects ofpreferential DNA repair. The time course of repair is shown for the overallgenome in hamster cells (filled circle); the hamster DHFR gene (filledtriangle); the human overall genome (filled square); the human DHFR gene(open circle); the human c-myc gene (open sqaure); and an active orinactive gene in cells from an individual with Cockayne's syndrome (opentriangle).

7 that genes can be repaired at faster rates than the DHFR gene,as exemplified by the repair of the c-myc gene in a human cellline. In addition, Figure 7 incorporates another aspect about thepreferential DNA repair: the lack of preferential repair in thehuman disease Cockayne's syndrome. This condition is associatedwith severe mental retardation, dwarfism and hypersensitivityto UV irradiation. Cells from these patients have normal overallgenome repair, but genes are repaired no more efficiently thaninactive genomic regions (16). This further illustrates theimportance of measuring DNA repair at the level of the gene:the defect in repair in this disease can only be determined bysuch studies. Of interest is also that Cockayne's cells may harbora mutation for a gene product involved in gene-specific DNArepair.

Preferential DNA repair is a general feature in many biologicalsystems. It has been observed in various mammalian cell lines,in fish, yeast (17) and in E.coli (18). Its presence in bacteriasuggests that the preferential DNA repair of active genes is notdependent upon chromatin structure. Preferential DNA repairin active genes is not limited to tissue culture cells: results showthat active genes are repaired more efficiently than inactive inmouse skin (H.Ruven, personal communication).

A DNA repair domainFrom the studies on the fine structure of DNA repair of UVdamage in the hamster DHFR gene, it appeared that the regionof preferential repair was larger than the size of the gene.Whereas the gene has a size of ~ 30 kb, the domain ofpreferential DNA repair was - 6 0 - 8 0 kb (19). Since this is thesize of the higher-order structures in chromatin, it is possiblethat DNA repair is regulated within such chromatin loops. Ina study on the repair of UV damage in the hamstermetallothionein genes, the increased repair after gene activationwas also observed in a region much larger than the gene (20).The pattern of repair thus may appear to follow some aspect ofchromatin structure that regulates the accessibility of the repairproteins. One parameter to consider is the degree of genomicdemethylation; there are very few studies on the relation of DNArepair to genomic methylation patterns, and only one study atthe gene level. It was observed that genomic demethylation couldaffect the pattern of repair in the hamster DHFR domain (21).

Geoe-spedfic DNA repair

Strand specificityAs an extension of the previously described technique (Figure4), single-stranded probes (riboprobes) can be used in place ofthe double-stranded probes, normally used in these experiments.With this approach, it was observed that the preferential DNArepair of pyrimidine dimers after UV damage in the hamsterDHFR gene was largely due to repair of the transcribed DNAstrand, whereas the non-transcribed strand was unrepaired (22).This strand bias of the repair of pyrimidine dimers has beenobserved in other mammalian genes in different laboratories andhas also been found in the yeast trp gene (10) and in the laclgene of E.coli (18). Recent results from our laboratory indicatethat the strand selectivity of the repair processes is far moredistinct for the pyrimidine dimer and possibly limited to the repairof a few lesions. For a number of other lesions that we havestudied in the hamster DHFR gene, to be discussed later, thestrand bias is much less distinct. For example, there appears tobe very little if any strand bias in the repair of 6 —4 photoproducts(23) and of the UV-mimicking agent 4-nitroquinoline-l -oxide(4NQO) (24). N1 lesions formed by the monofunctionalalkylating agent dimethylsulfate are repaired with equal efficiencyfrom the two strands (23,25), whereas we find a strand bias inthe repair of cisplatin intrastrand adducts (23). The degree ofstrand bias in the repair process may relate to the degree withwhich the lesion blocks the transcription process. While it isestablished that UV blocks transcription in vivo, it is not so clearwhether transcription can bypass other lesions. The finding thatDNA repair can be strand biased has led to the proposal thatthere may be a 'transcription coupling factor' that associates therepair enzymes with the transcription mechanism (26).

Strand bias of mutationsIt has been observed in several mammalian genes that there canbe a strand bias in the formation of mutations, and this is mostlikely due to strand-selective repair. After treatment with UVor benzo[a]pyrene, more mutations are formed in the non-transcribed strand than in the transcribed strand of the humanHPRT gene (27-29). In the hamster DHFR gene, mutationsinduced by benzo[c]phenanthrene diolepoxide are also preferen-tially located in the non-transcribed strand (30). For UV thereseems to be a reasonable correlation between the strand bias ofthe DNA repair and the strand bias of the mutations as studiedcollectively in the HPRT gene (31). One limitation is that whereasthe gene-specific repair can now be measured for both pyrimidinedimers and 6 - 4 photoproducts (see later), the type of photo-product that caused the mutation in wild-type cells cannot easilybe distinguished. In repair-deficient XP complementation groupA cells, McGregor et al. (28) reported that there is no strandbias of the repair; this supports the notion that the strand-selectiverepair process is responsible for the strand bias of the mutations.It is important, however, to consider that since these mutationsare fixed by replication, the strand bias could also arise froma bias in replication after damage. It is further possible that somesituations with strand bias of mutations arise from strand-specificrepair and others arise from replication bias. This may dependupon the location of the origin of replication in relation to thesequence studied. Although the site of the origin of replicationis known for many genes, it is further possible that this locationchanges after DNA damage. It will require further study toresolve this important relationship between the repair process andthe mutational spectrum, and the best approach would seem tobe co-ordinated studies of gene-specific and strand-specific DNArepair and of mutations in the same gene in the same biologicalsystem.

Gene-specific DNA repair and transcriptionSeveral experiments have shown that the efficiency of repair ina gene can correlate to its transcriptional activity. Gene-specificDNA repair efficiency of pyrimidine dimers can be modulatedby transcriptional induction of structural genes such as in the caseof the hamster or human metallothionein genes (32,33) andinhibitors of transcription can block the gene-specific repair (34).As with UV-induced pyrimidine dimers, gene-specific repair ofalkylation damage can also correlate with transcriptional activity:in rat cells, there was substantially more repair of N-methylpurinesin the active insulin gene than when it was not transcribed (35).In these and other studies, the transcription was assayed byNorthern analysis, and we are still awaiting repair studies wherethe rate of transcription is studied more directly by nuclear run-onexperiments.

The strand specificity of pyrimidine dimer repair discussedabove further documents the association between DNA repairprocesses and the transcription complex. This association appearsto be more evident for the strand specificity of the repair thanfor the gene-specific repair. J.Venema et al. (personalcommunication) have recently shown that gene-specific andstrand-specific DNA repair can be uncoupled: in the normalhuman adenosine deaminase (ADA) gene there is preferentialrepair and strand-specific repair after UV damage. However, ina patient with severe combined immunodeficiency diseaseharboring a mutation in the ADA gene promoter, which resultsin the gene being transcriptionally silent, there is still preferentialgene-specific repair, but no strand specificity of this repair. Invitro methods are now available (36) where gene- and strand-specific repair can be studied in cell-free systems. It is possibleto begin to identify protein fractions that contain this couplingfactor, and it may not be long before the protein(s) involved havebeen characterized. For the following discussion gene-specific(preferential) repair needs to be distinguished from strand-specificrepair. Whereas the strand specificity of the repair of pyrimidinedimers appears to be closely associated with transcription, thereare data suggesting that the gene-specific repair efficiency maynot always correlate with transcriptional activity (37). In addition,in differentiating mouse cells, it has been demonstrated that thegene-specific repair of some genes correlate with the stage ofdifferentiation rather than with gene expression (38).

Levels of gene-specific repairAs shown in Figure 7, there are significant differences in therepair efficiency in different genes. With respect to repair,different genomic regions may be categorized from the lowestto the highest repair efficiency as follows: (i) the inactive genomicDNA, (ii) the inactive genes that are turned off early indevelopment, (iii) the structural genes that are capable oftranscription at certain times in metabolism, (iv) the essentialgenes that are always active but at a low level of transcription,and finally (v) an active gene transcribed at a high rate. Thisis obviously speculative and needs to be explored.

Another consideration is the degree of local cytosinemethylation of the genome, it has been suggested that activegenomic regions are undermethylated and many genes and otherregions have been mapped by isoschizomeric restriction analysisto determine the sites of demethylation. There has been very littleexperimentation on the correlation between the degree ofmethylation and DNA repair. In a study by Ho et al. (21), wefound that the overall genomic DNA repair can increasesignificantly after treatment of cells for many generations withazacytidine (75% demethylation of the genome). Of furtherinterest was the marked change in the pattern of gene-specific

1987

V.A.Bohr

repair; in our previously discussed DNA repair domain in thehamster DHFR gene, the 3' end of the gene was now repairedas efficiently as its 5' end after extensive genomic demethylation.The important determinant here might be the chromatin structuralchanges that occur after changes in the methylation pattern ofDNA.

Specific inhibitorsOne approach to searching for the specific enzymes involved inthe preferential repair has been to use specific enzyme inhibitorsin the assay. In a comparative study, the effect of a number ofinhibitors were determined on the gene-specific repair in thehamster DHFR gene and also for the overall genome repair andreplication (39). A number of compounds that inhibit overallgenome repair such as inhibitors of topoisomerase I and II haveno effect on the gene-specific repair after UV damage. However,a combination of topoisomerase I and II inhibitors do inhibit thegene specific repair of pyrimidine dimers (T.Stevsner andV.A.Bohr, in preparation). This would suggest that onetopoisomerase can take over from the other in the gene-specificrepair process. Caffeine, which among other effects is thoughtto abrogate the DNA-damage-induced G2 arrest, also have aninhibitory effect on the gene-specific repair of pyrimidine dimers,with no effect on the overall genome repair (40). Finally, theRNA polymerase inhibitor a-amanitin also specifically inhibitsgene-specific repair after UV damage (34). These inhibitorexperiments are at this point only guideposts for models, but theycertainly suggest that the enzymology involved in gene-specificrepair has dissimilarities from that involved in the repair in theaverage genome.

Gene-specific repair of lesions other than pyrimidine dimersUsing variations of the above technique a number of lesions otherthan pyrimidine dimers have been studied at the level of the gene.Table I shows approaches used to create strand breaks at the siteof the specific lesions. For the detection of the rare, but importantUV-induced 6—4 photoproduct, a combination of photolyasetreatment and cleavage with ABC excinuclease has been utilized(41). Cisplatin intrastrand adducts and a number of carcinogens(such as 4NQO and N-acetoxy-acetylaminofluorene; NAAF) canbe detected by cleavage with ABC excinuclease. Alkylating agentscan be detected by a direct chemical reaction thus circumventingthe use of an enzyme: the strand breaks are generated by neutralheat depurination followed by alkaline hydrolysis (25,42,43). Theinterstrand crosslinks are detected and quantitated in a slightlydifferent manner as shown on the left side of Figure 4. Samplesare denatured and reannealed before electrophoresis on neutralgels. After Southern probing, two bands (Figure 4 and 8) aredetected, one representing the double-stranded (crosslinked) DNAfragment and the other the native fragment (single-stranded).

The preferential repair of a number of lesions has been studiedin the hamster DHFR domain, comparing the repair in the activegene with the repair in the downstream, non-coding region(Figure 6G and N). For the data shown in Figure 9, the gene-specific repair is probed in the DHFR gene fragment, and themembrane is stripped and reprobed for the downstream fragment,which was shown to be non-transcribed by Northern analysis(K.Wassermann and V.A.Bohr, unpublished). This latterfragment was chosen because it might reflect the repair of theoverall genome (which is 99% non-coding). Also, it is locatedwithin the DHFR domain 3' to the gene and it conveniently hasthe same size as the fragment studied in the DHFR gene. Figure9 shows the time course of repair of some important lesions. For

Table I. Strand breaks at sites of damage

Type of damage Cleavage

UV irradiationPyrimidine dimers6—4 photoproducts

CisplatinIntrastrand adductsInterstrand crosslinks

Quinoline

NAAAF

Oxidative damage

Alkylation damage

X-ray, bleomycin

N 24 8

T4 endonuclease Vphotolyase + ABC excinuclease

ABC excinucleasedenaturation + neutral gel

ABC excinuclease

ABC excinuclease

ABV excinuclease

depurination + alkaline hydrolysis

direct strandbreaks

0 24 8 0 Hr+ - - - NaOH

—d

IB —d

Fig. 8. Repair analysis of interstrand crosslinks. Denaturation, reannealingof cisplatin interstrand crosslinks on a neutral agarose gel (cf. Figure 3).Hamster cells were treated with cisplatin and the repair was analyzed in the14 kb (G) fragment of the DHFR gene (Figure 6). Non-denatured samplesare run as controls, d, double-stranded DNA (crosslinked); s, single-stranded DNA (native). The repair is visualized as the disappearance ofdouble-stranded DNA with time and the corresponding increase in native(repaired) DNA.

pyrimidine dimers (PD), it is evident that there is preferentialrepair, i.e. there is much more repair in the DHFR gene (G)than in the non-coding region (N) at both 8 and 24 h. 6 - 4photoproducts (PP) are repaired much more efficiently from thenon-coding region than are pyrimidine dimers, but there is stillpreferential repair of this lesion (41).

1988

Gene-speciflc DNA repair

Differential DNA Repair of some Lesions

1CXH

Fig. 9. Time course of repair of different lesions in the hamster DHFRdomain, the active gene (G) and the non-coding region (N) are mapped inFigure 6. PD, pyrimidine dimer; PP, 6—4 photoproduct; NF, N-acetoxy-acetylaminofluorene (NAAAF); CX, cisplatin interstrand crosslinks; CA,cisplatin intrastrand adducts; NM, nitrogen mustard; DS, dimethylsulfate.y-axis is percentage repair, and the repair periods are 8 and 24 h as shown.

CisplatinAs mentioned previously, the intrastrand adduct is by far the mostcommon lesion after treatment of cells with cisplatin. Weconsistently find some preferential repair of this lesion (Figure9, CA) (44) in the hamster cells, but it is much less distinct thanthe preferential repair of the pyrimidine dimers (44). Whereaswe find that the cisplatin interstrand crosslinks are removed fasterfrom the DHFR gene than are the intrastrand adducts, as shownin Figure 9, there is no preferential repair of the cisplatininterstrand crosslink (CX) in the hamster DHFR domain (44).

Alkylating agentsUsing the neutral depurination reaction followed by alkalinehydrolysis, we mainly detect guanine N1 alkylations. Figure 9shows that the nitrogen mustard (NM) lesions are preferentiallyrepaired in the hamster DHFR domain (42), while lesionsintroduced with dimethylsulfate (DS) are not (25,42). In addition,lesions introduced with methylnitrosurea were preferentiallyrepaired (43) in the same hamster cells. These results show thatthere are considerable differences in the mode of repair ofdifferent alkylating agents. Since these agents all predominantlycause methylpurine formation, the difference in mode of repairmay be due to other aspects than the type of lesion formed. Itmay reflect differences in the mode of reaction of the alkylatingagent with DNA (e.g. SN1 or SN2 type reaction). More likelyit may depend upon the degree of chromatin distortion that thesecompounds invoke and thereby the recognition enzymes involved;alternatively, it may reflect the degree of transcription blockadeby the adducts.

Other compoundsPsoralen plus UVA light (PUVA) initially creates monoadducts,and the DNA can then be reirradiated by UVA light to formcrosslinks. By quantitating interstrand crosslinks before and afterthe reirradiation, both interstrand crosslinks and monoadductscan be measured. Using the previously described methodology,it was demonstrated that the psoralen crosslinks are removedmuch faster from the DHFR gene than the monoadducts (45).In another study in both a hamster and a human cell line, therepair of crosslinks was much more efficient in the DHFR genethan in ribosomal genes (46). This led to the proposal thatpreferential repair might be found in polymerase II transcribed

genes (DHFR) whereas polymerase I (rDNA) transcribed regionswould be repaired less efficiently. However, further confirma-tion is needed to support this hypothesis. Bleomycin single- anddouble-strand breaks have been detected at the gene level andrepair of single-strand breaks has been quantified in the humanc-myc gene (47). Using the methodology shown in Figure 4, othertypes of crosslinks have been detected. Mitomycin-inducedcrosslinks have been quantitated in human ribosomal genes (48).

The described methodology has also been used to study DNAdamage and repair following carcinogen exposure. DNA adductsproduced by the carcinogen NAAAF were repaired with similarefficiency in the DHFR gene and in the non-coding region (49).(Figure 9, NF), and the carcinogenic 4NQO adducts are alsonot preferentially repaired in the hamster DHFR gene (24). Themechanism of repair of DNA damage induced by these two agentsis thought to be UV mimicking, and it is surprising that thereis no clearly demonstrable preferential repair.

Drug resistanceThe development of drug resistance is a common and seriousproblem, particularly when treating with cisplatin or alkylatingagents (50). For example, many patients with advanced stageovarian cancer initially respond well to therapy with cisplatin andthen subsequently become resistant to such therapy at relapse (51).Several mechanisms are thought to play a role in drug resistance,including membrane alterations of drug uptake or efflux,alterations of drug metabolism, and increased DNA repair.Lately, there is increasing evidence that DNA repair may playa more important role in drug resistance than previously thought.For example, some resistant human ovarian cancer cell lines havebeen shown to have increased rates of DNA repair whencompared with the parental cell lines (52), but in other studiesthere is no correlation. The reported studies so far on DNA repairin drug resistance have only been done at the level of the overallgenome, and it is relevant to explore gene-specific repair in thesesystems. In some recent experiments in our laboratory (W.Zhenet al., submitted), the DNA repair in cisplatin-resistant humanovarian cancer cells and in their parental cell line was studied.The repair was examined in both cell lines at the level of theoverall genome and in the DHFR gene. This was done separatelyfor the two major cisplatin lesions, the intrastrand adduct andthe interstrand crosslink. Whereas no correlation was foundbetween cisplatin resistance and overall genome repair of theselesions, there was more efficient repair of the interstrandcrosslinks in the DHFR gene in the resistant than in the parentalcell lines. This may indicate a correlation between drug resistanceand gene-specific repair, and we are investigating this possibilityfurther. It is certainly suggested that further studies be done atthe level of the gene to establish new correlations to drugresistance that were not detectable when DNA repair wasmeasured at the level of the overall genome.

Initial damage formationAn important aspect of the gene-specific repair studies is thatinherent in the experiment, the initial damage formation is alwaysassessed as a basis for the repair. For UV damage we have notobserved any differences in initial damage formation in anyspecific genomic region, but for alkylating agents such as nitrogenmustard there can be differences in initial levels of damage. Wehave found (K.Wassermann, M.Pirsel and V.A.Bohr, inpreparation) that there can be at least 2-fold differences in theformation of adducts in different regions, surprisingly, sometimeswith more adducts formed in inactive regions. For psoralen

1989

V.A.Bohr

Murine c-myc Repair Heterogeneity

5' c-myc 3' c-myc

Fig. 10. Repair heterogeneity in the c-myc locus in plasmacytoma sensitive(BALB/c) and resistant (DBA) mouse B-cells (37). DNA repair efficiencyafter 12 h was determined in mouse spleen cells irradiated with a UV doseof 10 J/m2. The repair was analyzed in the 5' flanking, non-transcribedc-myc region, in the coding region of c-myc and in two regions in theDHFR domain in both strains. It is seen that plasmacytoma reistancecorrelates with efficient repair in the 5' c-myc region, whereas repair issimilar in the other regions in the two strains.

crosslink formation, there also appears to be differences in thecrosslinking efficiency in different genomic regions (53). In thiscase, more crosslinks were formed in the active gene regionsthan in an inactive gene. More studies of this type obviously needto be done for a variety of DNA lesions. In the future we maybe able to screen for chemotherapeutics that are selective forcertain genomic regions where it would be desirable to inhibitexpression.

Gene-specific repair in human repair-deficient syndromesThere have been studies on gene- and strand-specific repair afterUV damage in cells from individuals with human repair-deficientdiseases. It is not within the scope of this review to thoroughlydiscuss those studies, but it is significant that there is now anexplanation for the high degree of UV sensitivity in cells frompatients with Cockayne's syndrome. As mentioned before, cellsfrom these individuals are UV sensitive, but repair the overallgenome normally. The deficiency appears to be a lack ofpreferential repair of the active genes in these cells (16,M.K.Evans, C.J.Link and V.A.Bohr, unpublished), as shownin Figure 7. Some gene-specific repair measurements have beenperformed on cells from XP complementation group C (XPC)patients that show low but detectable overall genome repair ofpyrimidine dimers. It has been known for some time that thesecells repair in a heterogeneous fashion after UV, and that someregions are repaired more efficiently than others (54). XPC cellsshow some preferential repair (55—57), and it has been proposedthat the residual repair in this cell line stems largely from selectiverepair of the active genes (57). A comprehensive study on thegene-specific repair in a number of XP complementation groupsis now completed in our laboratory (M.K.Evans etai, inpreparation), and other laboratories are also investigating thegene-specific repair in human disorders. Clearly, the gene-specificrepair studies are going to be important in furthering ourunderstanding of DNA repair differences in human patho-physiology as well as in diagnosing disease or screeningindividuals at risk.

Gene-specific repair of oncogenesIt was an early project to investigate the heterogeneity of DNArepair in oncogenes. In mouse cells the repair of the transcrip-tionally active c-abl gene was compared to that in the inactivec-mos protooncogene, and it appeared that the c-abl gene was4-fold more efficiently repaired than the c-mos (58). This couldmean that the inactive gene might accumulate more damage orbe more susceptible to mutation when the cell undergoes mitosis.We have further explored aspects of damage and repair in proto-oncogenes, and recent experiments in our laboratory show repairdifferences in the c-myc gene which may relate to tumorpathogenesis (37). B-cells from spleen in mice that are eitherresistant or susceptible to plasmacytomas have been studied withthe gene-specific repair technique. The repair efficiency after UVirradiation was studied in the c-myc locus as well as in the DHFRgene in mouse strains that were genetically determined to becomeresistant or sensitive to plasmacytoma formation (37) (Figure 10).Whereas the repair in the DHFR regions and in the coding c-mycregion was similar in the two strains, this was not the case inthe 5' c-myc region. We found deficient repair in this region inthe sensitive mouse strain compared to the resistant before thedevelopment of tumors. Since rearrangements occur in this region5' to c-myc, some relationship may exist between the localgenomic DNA repair phenotype and c-myc activation.Interestingly, the efficient repair in the 5' c-myc region in theresistant mouse strain did not appear to be related to increasedtranscription in this region as we found no major transcripts. Inaddition, there was no strand bias of the repair, i.e. similar repairefficiencies in the two strands in this region (37). The importanceof this 5' c-myc repair phenotype is further accentuated by theobservation that efficient repair in this region correlates toplasmacytoma resistance in inbred mouse strains (E.Beechamet al., in preparation). The possibility exists that the localizedrepair deficiency plays a key role in the development ofmalignancy by somehow allowing translocations to occur. Thisrelationship will certainly be further explored in the future. Animportant question to address in this context is whether the manymutations found in the p53 gene in human tumor cells (59) aredue to deficient DNA repair.

Bianchi et al. (14) used another approach to examine theregion-specific DNA repair in the human c-myc gene. Theystudied the cutting by a restriction endonuclease at rare sites inthe c-myc gene in human COLO320HSR cells, which carries a30- to 40-fold amplification of the c-myc gene. They concludedthat there is more efficient repair of pyrimidine dimers (totalphotoproducts were studied) in the 5' flanking region of the c-mycgene than in its coding 3' part or its 3' flanking region.

Mitochondrial DNA repairNumerous studies have demonstrated a greater accumulation ofDNA damage in mitochondrial DNA than in nuclear DNA andreasons for this accumulation have been proposed (50-62). Onefactor that may be involved in the increased vulnerability ofmitochondrial DNA is that there appears to be very limited repairof DNA damage in the mitochondrial genome. Early studies didnot detect repair of UV-induced pyrimidine dimers (63,64) orthe damage induced by exposure to Af-methyl-JV-nitro-N-nitroso-guanidine or 4NQO (65). Later studies, however, havedemonstrated repair of O^-methylguanine from mitochondrialDNA, suggesting that some repair mechanisms may be operatingin mitochondrial DNA (66). Also, the presence of a number ofrepair endonucleases in mitochondria, including some thatrecognize oxy-radical damage (67), suggested that an excision

1990

Gene-specific DNA repair

repair mechanism may be present in mitochondria. Recently,using the gene-specific DNA damage and repair approach, repairof alkali-labile sites within mitochondrial DNA has beendemonstrated in a rat cell line following exposure to the nitrosureastreptozotocin (68). These results provided the first directevidence that an excision repair mechanism may be functioningin mitochondrial DNA. We have explored the repair of a numberof different DNA lesions in the mitochondria] of hamster cells(S.P.LeDoux et al., submitted). There appears to be efficientrepair of methylation damage following exposure tomethylnitrosourea and dimethysulfate, but nitrogen mustardlesions are not repaired. There is no repair of pyrimidine dimersin mitochondrial DNA after exposure to UV light. However,cisplatin interstrand crosslinks introduced into mitochondrialDNA are repaired efficiently. It thus appears that mitochondriahave repair capacity but can only repair certain lesions. We areinvestigating whether this corresponds to repair pathway, i.e.whether mitochondria have base excision repair capability butnot nucleotide excision repair capability. The findings raise manyother questions which are now being actively investigated. Itseems, for example, important to determine if repair enzymesare transported into the mitochondria.

ConclusionsIt is now possible to determine the damage formation and repairof various types of DNA-damaging agents in individual genes.The results obtained so far clearly show that such studies provideus with new and important information that relates DNA repaircapability to other biological parameters as well as aspects ofmalignancy. These relationships often do not appear when DNArepair is studied with more traditional approaches where theevents are averaged over the entire genome. The recentlydiscovered phenomenon of preferential repair of active genes afterUV damage has provided the framework for understanding thefine structure of DNA repair. There is strong evidence for alinkage between strand specificity of repair and transcriptionalactivity in a given gene. Preferential repair of a gene can alsobe linked with transcription, but recent studies suggest that thisis not the whole story. We are now able to investigate the gene-specific damage and repair for a number of carcinogens andchemotherapeutics, and these studies will hopefully provideinsight into the etiology and treatment of malignancies.

AcknowledgementsI thank Charles Link, Jr and Miroslav Pirsel for help with the preparation ofFigures 1, 3 and 4, and members of my laboratory group for helpful comments.

References1. Pkx>y,A.C.M., van Dijk,M. and Lohman,P.H.M. (1984) Induction and repair

of DNA cross-links in Chinese hamster ovary cells treated with variousplatinum coordination compounds in relation to platinum binding to DNA,cytotoxicity, mutagenicity, and antitumor activity. Cancer Res., 44,2043-2051.

2. Eastman.A. (1987) The formation, isolation and characterization of DNAadducts produced by anticancer platinum complexes. Pharmacol. Ther., 34,155-166.

3. Singer.B. (1975) The chemical effects of nucleic acid alkylation and theirrelation to mutagenesis and carcinogenesis. Prog. Nucleic Acids Res. Mol.BioL. 15,219-284.

4. Urade.M., Sugi.M. and Matsuya.T. (1988) Further characterization ofbleomycin-resistant HeLa cells and analysis of resistance mechanism, CancerRes., 79, 491-500.

5. HoeijmakersJ.H. and Bootsma.D. (1990) Molecular genetics of excisionrepair. Cancer Celts, 2, 311-320.

6. van Houten.B. (1990) Nucleotide excision repair in E.coli. MicrobiologicalRev., 54, 18-51.

7. Bohr.V.A., Smith,C.A., Okumoto.D.S. and Hanawalt,P.C. (1985) DNArepair in an active gene: removal of pyrimidine dimers from the DHFR geneof CHO cells is much more efficient than the genome overall. Cell, 40,359-369.

8. Thomas.D.C, Morton.A.G., Bohr.V.A. and Sancar.A. (1988) Generalmethod for quantifying base adducts in specific mammalian genes. Proc. Natl.Acad. ScL USA, 85, 3723-3727.

9. Mullenders.L.H.F., van KesterenAC., Bussman.C.J.M., van Zeeland,A.A.and Natarajan,A.T. (1986) Distribution of u.v.-induced repair events in higher-order chromatin loops in human and hamster fibroblasts. Carcinogenesis, 7,995-1002.

10. Smerdon.M.J. and Thoma,F. (1990) Site-specific DNA repair at thenucleosome level in a yeast minichromosome. Cell, 61, 675—684.

11. Bohr.V.A. and Okumoto.D.S. (1988) Analysis of frequency of pyrimidinedimers in specific genomic sequences. In Friedberg.E.C. and Hanawalt.P.C.(eds), DNA Repair: A Laboratory Manual of Research Procedures. MarcelDekker, New York, Vol. 3, pp. 347-366.

12. Leadon.S.A. (1988) Immunological probes for lesions and repair patches inDNA. In Friedberg.E.C. and Hanawalt,P.C. (eds), DNA Repair: A LaboratoryManual of Research Procedures. Marcel Dekker, New York, Vol. 3, pp.311-327.

13. Govan.HJ.m., Valles-Ayoub-Y. and BraunJ. (1990) Fine-mapping of DNAdamage and repair in specific genomic segments. Nucleic Acids Res., 18,3823-3830.

14. Bianchi.N.O., Bianchi.M.S., Alitalo.K. and Chapelle.A.D.L. (1991) UVdamage and repair in the domain of the human c-myc oncogene. DNA CellBioL, 10, 125-132.

15. Bohr.V.A., Evans.M.K. and Fomace.A.J.M (1989) DNA repair and itspathogenetic implications. Lab. Invest., 61, 143 — 161.

16. VenemaJ., Mullenders.L.H.F., Natarajan.A.T., Van Zeeland,A.A. andMayne.L.V. (1990) The genetic defect in Cockayne syndrome is associatedwith a defect in repair of UV-induced DNA damage in transcriptionally activeDNA. Proc. Natl. Acad. Sd. USA, 87, 4704-4711.

17. Terlleth.C, van Sluis.C.A. and van de Putte.P. (1989) Differential repairof UV damage in Saccharomyces cereviciae. Nucleic Acids Res., 17,4433-4439.

18. Mellon.I. and Hanawalt.P.C. (1989) Induction of the Escherichia coli lactoseoperon selectively increases repair of its transcribed DNA strand. Nature,342, 95-98. ,

19. Bohr.V.A., Okumoto.D.S., Ho.L. and Hanawalt.P.C. (1986) Characterizationof a DNA repair domain containing the DHFR gene in CHO cells. J. BioLChem., 261, 16666-16672.

20. Okumoto.D.S. and Bohr.V.A. (1987) DNA repair in the metallothionein geneincreases with transcriptional activation. Nucleic Adds Res., 15, 10021 -10030.

21. Ho,L., Bohr.V.A. and Hanawalt.P.C. (1989) Demethylalion enhances removalof pyrimidine dimers from the overall genome and from specific DNAsequences in Chinese hamster ovary cells. Mol. Cell. Bid. , 9, 1594-1603.

22. Mellon,I., Spivak.G. and Hanawalt.P.C. (1987) Selective removal oftranscription-blocking DNA damage from the transcribed strand of themammalian DHFR gene. Cell, 51, 241 -249.

23. Bohr.V.A., Nairn.R.S., Wassermann.K., JonesJ.C, Zhen.W. and May.A.(1991) Studies of preferential DNA repair and strand specificity of DNA repairin the DHFR gene after treatment of hamster cells with UV andchemotherapeutics. Proc. Am. Assoc. Cancer Res , 32, 4.

24. Snyderwine.E.G. and Bohr.V.A. (1991) Repair of 4NQO adducts in specificgenes in mammalian cells. Proc. Natl. Assoc. Cancer Res., 32, 108.

25. Scicchitano.D.A. and Hanawalt.P.C. (1989) Repair of A'-methylpurines inspecific DNA sequences in Chinese hamster ovary cells: absence of strandspecificity in the dihydrofolate reductase gene. Proc. Natl. Acad. Sd. USA,86, 3050-3054.

26. Hanawalt.P.C. (1989) Preferential repair of damage in actively transcribedDNA sequences in vivo. Genome, 31, 605—611.

27. Vrieling.H., Van Rooijen.M.L., Groen,N.A., Zdzienicka.M.Z.,SimosJ.W.I.M., Lohman.P.H.M. and van Zeeland.A.A. (1989) DNA strandspecificity for UV induced mutations in mammalian cells. Mol. Cell. Bid.,9, 1277-1283.

28. McGregor.W.F., Chen,R.-H., Lukash.L., Maher.V.M. and McCormickJ.J.(1991) Cell cycle dependent strand bias for UV mutations in the transcribedstrand of excision-repair proficient human fibroblasts but not in repair-deficientcells. Mol. Cell. Biol. 11, 1927-1934.

29. Chen,R.-H., Maher.V.M. and McCormickJ J. (1990) Effect of excision repairby diploid human fibroblasts on the kinds and location of mutations inducedby (±)-7,8-dihydroxy-9,10-epoxy-7,8,9,10-tttrahydrobenz(a)pyrene in thecoding region of the HPRT gene. Proc. Nad. Acad. Set USA, 87, 8680-8684.

30. Carothers.A.M., MuchaJ. and Grunberger.D. (1991) DNA strand-specificmutations induced by (±)-3,4-dihydro-l,2-epoxy-l,2,3,4-tetrahydrobenzo-

1991

V.A.Bohr

(c)phenanthrcne in the DHFR gene. Proc. Natl. Acad. Sci. USA, 88,5749-5753.

31. Vrieling,H., Venema.J., Rooyen,M.-L., van Hoffen.A., Meoechini.P.,Zdienkka.M.Z., SimonsJ.W.I.M., Mullenders.L.H.F. and van Zeeland.A.A.(1991) Strand specificity for UV induced DNA repair and mutations in theChinese hamster HPRT gene. Nucleic Adds Res., 19, 2411-2415.

32. Okumoto.D.S. and Bohr.V.A. (1987) DNA repair in the metallothionein geneincreases with transcriptional activation. Nudeic Adds Res., 15, 10021 — 10029.

33. Leadon.S.A. and Snowden.M.M. (1988) Differential repair of DNA damagein the human metallothionein gene family. Mol. Cell. Biol., 8, 5331—5352.

34. Chnstians.F.C. and Hanawalt.P.C. (1990) Reduced strand specific repair ofpyrimidine dimers in alpha-amanitin treated mammalian cells. Proc. Am. Assoc.Cancer Res., 31, 2.

35. LeDoux.S.P., Patton.N.J., NelsonJ.W., Bohr.V.A. and Wilson.G.L. (1990)Preferential repair of alkali-labile sites within the active insulin gene. /. Biol.Chem.. 265, 14875-14880.

36. Selby,C.P. and Sancar.A. (1991) Gene and strand specific repair in vitro.Proc. Natl. Acad. Sci. USA, 88, 8232-8236.

37. Beecham.E., Frederic MushinskU., Shacter.E., Potter.M. and Bohr.V.A.(1991) DNA repair in the c-myc proto-oncogene locus: possible involvementin susceptibility or resistance to plasmacytoma induction in BALB/c mice.Mol. Cell. Bio!., 11, 3095-3104.

38. Bill.C.A., Grochan.B.M., Meyn.R.M., Bohr.V.A. and Tofilon.P.J. (1991)Loss of intragenomic DNA repair heterogeneity with cellular differentiation.J. Biol. Chem., 254, 155-162.

39. JonesJ.C, Stevnsner.T., Mattern.M. and Bohr.V.A. (1991) Effect of specificenzyme inhibitors on replication, total genome DNA repair and on gene-specificDNA repair after UV irradiation in CHO cells. Mutat. Res., DNA RepairRep., 254, 155-162.

40. Link.C.J., Stevnsner.T., JonesJ.C. and Bohr.V.A. (1991) Effect of inhibitorson UV light induced damage. Proc. Am. Cancer Res., 32, 2.

41. Thomas.D.C, Okumoto.D.S., Sancar.A. and Bohr.V.A. (1989) PreferentialDNA repair of (6-4) photoproducts from the CHO DHFR gene. J. Biol.Chem., 264, 18005-18010.

42. Wasserman.K., Kohn.K.W. and Bohr.V.A. (1990) Heterogeneity of nitrogenmustard-induced DNA damage and repair at the level of the gene in Chinesehamster ovarian cells. / . Biol. Chem. 265, 13906-13913.

43. LcDoux.S.P., Thangada.M., Bohr.V.A. and Wilson.G.L. (1991)Heterogeneous repair of methylnitrosurea-induced alkali-labile sites in differentDNA sequences. Cancer Res., 51, 775-779.

44. JonesJ., Zhen.W., Reed.E., Parker.R.J., Sancar.A. and Bohr.V.A. (1991)Gene specific formation and repair of cisplatin intrastrand adducts andinterstrand crosslinks in CHO cells. J. Biol. Chem., 266, 7101-7107.

45. Vos,J.-M.H. and Hanawalt.P.C. (1987) Processing of psoralen adducts inan active human gene: repair and replication of DNA containing monoadductsand interstand cross-links. Cell, 50, 789-799.

46. Vos.J.-M.H. and Wauthier.E.L. (1991) Differential introduction of DNAdamage and repair in mammalian genes transcribed by RNA polymerasesI and II. Mol. Cell. Biol., 11, 2245-2252.

47. Bianchi.N.O., Bianchi.M.S., Lopez-Larraza.D., Alityalo.K. and de laChapelle.A. (1990) Damage and repair induced by bleomycin in the domainof human amplified myc oncogenes. Cancer Res., 50, 2379—2384.

48. Matsumoto.A., Vos,J.-M.H. and Hanawalt.P.C. (1989) Repair analysis ofmitomycin C induced DNA crosslinking in ribosomal RNA genes inlymphoblastoid cells from Fanconi's anemia patients. Mutat. Res., 217,185-192.

49.Tang,M.-S., Bohr.V.A. Zhang,X.-s., PierceJ. and Hanawalt.P.C. (1989)Quantitation of aminofluorene adduct formation and repair in defined DNAsequences in mammalian cells using UVRABC nuclease, J. Biol Chem., 264,14455-14462.

50. HainsworthJ.D., Grosh.W.W. Burnett.L.S., Jones.H.W., Wolff.S.N. andGreco.F.A. (1988) Advanced in ovarian cancer: long term results with intensivecisplatin-based chemotherapy of brief duration. Ann. Intern. Med., 108,165-170.

51. BerekJ.S. and Hacker.N.F. (1990) Ovarian Cancer. In Haskell.C.M. (ed.),Cancer Treatment. W.B.Saunders, Philadelphia, pp. 718-725.

52. Lai.G., Ozol.R.F., Smyth.J.F., Young.R.C. and Hamilton.T.C. (1988)Enhanced DNA repair and resistance to cisplatinum in human ovarian cancer.Biochem. Pharmacol., 37, 4597-4600.

53. Islas.A.L, Vos J.-M.H. and Hanawalt.P.C. (1991) Differential introductionand repair of psoralen photoadducts to DNA in specific human genes. CancerRes., 51, 2867-2873.

54. Mansbridge J.N. and Hanawalt.P.C. (1983) Domain limited repair of DNAin UV irradiated fibroblasts from XP complementation group C. InFriedberg.E.C. and Bridges.B. Cellular Responses to DNA Damage, UCLASymp. Mol. Cell. Biol Alan R. Liss, New York, Vol. II, pp. 195-207.

55. Bohr.V.A., Okumoto.D.S. and Hanawalt.P.C. (1986) Survival of UV-irradiated mammalian cells correlates with efficient DNA repair in an essential

1992

gene. Proc. Natl Acad. Sci. USA, 83, 3830-3833.56. Kantor.G.J., Barsalou,L.S. and Hanawalt.P.C. (1990) Selective repair of

specific chromatin domains in UV-irradiated cells from xerodermapigmentosum complementation group C. Mutat. Res., 235, 171-180.

57. VenemaJ., van Hoffen.A., Natarajan.A.T., van Zeeland,A.A andMuDenders, L. (1990) The residual repair capacity of xeroderma pigmentosumcomplementation group C fibroblasts is highly specific for transcriptionallyactive DNA. Nucleic Acids Res., 18, 443-448.

58. Madhani.H.D., Bohr.V.A., Hanawalt.P.C. (1986) Differential DNA repairin transcripuonalry active and inactive proto-oncogenes: c-oW and c-mos. Cell,45, 417-426.

59. Hollstein.M., Sidransky.D., Vogelstein.B. and Harris.C.C. (1991) p53mutations in human cancers. Science, 253, 49-53.

60. Wunderlich.V., Schutt.M., Booger.M. and Graffi.A. (1970) Preferentialalkylation of mitochondrial deoxyribonucleic acid by A'-methyl-A'-nitrosurea.Biochem. J., 118, 99-109.

61. Backer JM. and Weistein.I.B. (1980) Mitochondrial DNA is a major cellulartarget for a dihydrodiol-epoxide derivative of benzo(a)pyrene. Science, 209,297-299.

62. Niranjan.B.G., Bhat.N.K. and Avadhani.N.G. (1981) Preferential attack ofmitochondrial DNA by aflatoxin Bi during hepatocarcinogenesis. Science,215, 73-75.

63.Clayton,D.A., Doda.J.N. and Friedburg.E.C. (1974) The absence ofpyrimidine dimer repair mechanism in mammalian mitochondria. Proc. Natl.Acad, Sci. USA, 71, 2777-2781.

64. Prakash.L. (1985) Repair of pyrimidine dimers in nuclear and mitochondrialDNA of yeast irradiated with low doses of ultraviolet light. J. Mol. Biol.,98, 781-795.

65. Miyaki.M., Yatagai.K. and Ono,T. (1977) Strand breaks of mammalianmitochondrial DNA induced by carcinogens. Chem.-Biol. Interactions, 17,321-329.

66. Meyers JCA., Saffhill.R. and O'Connor.P.J. (1988) Repair of alkylated purinesin the hepatic DNA of mitochondria and nuclei of rat. Cardnogenesis, 9,285-292.

67.Tomkinson,A.E., Bonk.R.T., Kim,J., Bartfeld.N. and Linn.S. (1990)Mammalian mitochondrial endonuclease activities specific for ultraviolet-irradiated DNA. Nucleic Adds Res., 18, 929-935.

68. Pettepher.C.C, LeDoux,S.P., Bohr.V.A. and Wilson.G.L (1990) Repairof alkali-labile sites within the mitochondrial DNA of RINr38 cells afterexposure to the nitrosourea streptozotocin. / . Biol. Chem., 266, 3113—3117.