Grigory Dianov* and Tomas LindahlImperial Cancer Research Fund, Clare Hall Laboratories, South Mimms, Hertfordshire, EN6 3LO, UK.

Background: The base excision~repair pathway is themajor cellular defence mechanism against spontaneousDNA damage. The enzymes involved have been highlyconserved during evolution. Base excision~repair hasbeen reproduced previously with crude cell-free extractsof bacterial or human origin. To further our understand-ing of base excision-repair, we have attempted to recon-stitute the pathway in vitro ming purified enzymes.Results: We report here the successful reconstitution ofthe base exci~ion:'-repair pathway with five purifiedenzymes from Escherichia coli: uracil-DNA glycosylase,a representative of the DNA glycosylases that removevarious lesions from DNA; the AP endonuclease IV thatspecifically cleaves at ah;\sic sites; RecJ protein whichexcises a 5' terminal deoxyribose-phosphate residue;DNA polymerase I; and DNA ligase. The reaction

proceeds with high efficiency in the absence of additionalfactors in the reconstituted system. Four of the enzymesare absolutely required for completion of the repairreaction. An unusual feature we have discovered is thatthe pathway branches after enzymatic incision at anabasic DNA site. RecJ protein is required for the majorreaction, which involves replacement of only a singlenucleotide at the damaged site; in its absence, analternative pathway is observed, with generation oflonger repair patches by the 5' nuclease function of DNApolymerase I.Conclusions: Repair of uracil in DNA is achieved by avery short-patch excision-repair process involving fivedifferent enzymes. No additional protein factors seem tobe required. There is a minor, back-up pathway that usesreplication factors to generate longer repair patches.

Current Biology 1994,4:1069-1076

Background external DNA-damaging agents, the most frequentDNA lesions are those handled by BER, and conse-quently cellular back-up mechanisms are frequentlyemployed in this essential process. Thus, it has not beenpossible to construct an Escherichia coli mutant totallydeficient in repair of DNA abasic sites [3], implying thatcontinuous correction of spontaneous DNA lesions isrequired for viability.

Three different excision-repair pathways are universallypresent in living organisms, and constitute the maincellular strategies for removing lesions and pre-mutagenicerrors from DNA. Strand-specific mismatch repairprovides protection against occasional rare errors thatoccur during DNA replication and have escaped proof-reading mechanisms. Nucleotide excision-repair (NER)is responsible for correction of damage that causes majorhelix distortion, in particular the dipyrimidine atlductsgenerated on exposure of cells to sunlight. Baseexcision-repair (BER) is the main mechanism forremoving 'spontaneous' DNA lesions that are caused byhydrolysis, oxidation or exposure to reactive smallmolecules such as S-adenosylmethionine [1]. Some ofthese DNA lesions, including de aminated cytosineresidues and abasic sites, cause very little helix distortionbut are mutagenic if not repaired.

The HER pathway is initiated by hydrolytic removal of abase either by one of the several DNA glycosylases thatrecognize different types of lesions or by non-enzymaticbase loss. The pathway proceeds by incision at the abasicsite, generation of a gap, repair synthesis and ligation.The entire reaction has been reproduced in vitro withextracts from E. coli [4], Saccharomyces cerevisiae [5],Xenopus laevis [6-8] and human cells [4]. Marked hetero-geneity of repair patch sizes has been observed [4,8,9],indicating that there may be more than one route ofexcision and gap-filling. Using radiolabelled precursorsfor DNA synthesis and detailed re~triction enzymecleavage, the length of the repaired tract after excision ofan uracil residue has been found to be limited to a singlenucleotide in most repair events, in both E. coli andhuman cells, but longer tracts have also been observed[4]. Similarly, most of the repair tracts generated atapurinic sites by fractionated X. laevis cell extracts werevery short and dependent on DNA polymerase 13,though longer patches of about four nucleotides werealso produced in a proliferating cell nuclear antigen(PCNA)-dependent process [8]. In the present work, wehave investigated the HER process by employing five

In general, there is little overlap between the threeexcis;ion-repair pathways, both with regard to theenzymes involved and the structural alterations in DNAthat they recognize. For this reason, there are no cellularback-up mechanisms for the activities that initiate eitherthe mismatch correction or the NER pathway. Loss ofthese functions therefore has drastic consequences, suchas hypermutation and increased cancer frequency in man[2]. On the other hand, duplication of functions andlimited overlap with other forms of excision-repair arecharacteristic of the HER pathway. This may reflect thefact that, in the absence of exposure to large doses of

Correspondence to: Tomas Lindahl. 'Present address: Department of Pathology, University of Texas Southwestern Medical Center, Dallas,Texas 75235, USA.

@ Current Biology 1994, Vol 4 No 12 1069

1070 Current Biology 1994, Vol 4 No 12

pu~ified enzymes from E. coli instead of cell extracts toreconstitute the pathway.

Results

Requirement for a DNA glycosylase in the in vitro reactionUracil residues in DNA can be generated by cytosinedeamination and are repaired efficiendy in vivo by a BERprocess initiated by uracil-DNA glycosylase [10].Double-stranded oligonucleotides containing a centrallyplaced G. U base pair, surroundea by appropriate restric-tion enzyme sites, have been employed as substrates for invitro reactions [4]. In previous work with E. coli extracts,the dUMP residue was replaced efficiendy by a dCMPresidue. No incorporation of newly synthesized materialoccurred on the 5' side of the non-conventional residue,whereas small amounts of exchanged residues wereobserved on the 3' side. These results were independentof the base composition of the oligonucleotide and thesequence adjacent to the uracil residue.

Fig. 1. Nucleotide sequences of double-stranded oligonucleotides containing acentrally placed G.U (a) or G.C (b)base-pair. The arrow indicates thecleavage site of the restriction endo-nuclease Hpall, which is blocked byreplacement of a C with a U residuewithin the enzyme's recognition site.Internucleotide bonds close to terminiwere synthesized as phosphorothioatebonds (shown as .,,) to suppress exo-nuclease activity and turnover of 3'-

~~¥~":~" [ '~"'" $;~ \

]' ~ ~ CfC.s- I

non-specific degradation and turnover of the terminalresidues of the oligonucleotides, the terminal residueswere joined by phosphorothioate bonds (Fig. 1). Repairsynthesis of substrates was assessed in an [a-32P]dCTP-containing reaction mixture, using denaturing gel elec-trophoresis and autoradiography for analysis of theoligonucleotide product. Incorporation of radioactivematerial was completely dependent on the presence of adUMP residue in the double-stranded oligonucleotidesubstrate (Fig. 2a). An extract of an ung strain, deficient inuracil-DNA glycosylase, showed no detectable repair ofthe uracil-containing oligonucleotide, whereas an iso-genic ung+ strain was clearly repair-proficient (Fig. 2b).

7f.R..I't:.O I e\l\do ICompar~n of the~o repair-proficient strains NH5033 (sbcB, recB, endA) and W3110 indicates that neitherexonuclease I, RecBCD nuclease nor endonuclease I areinvolved in the repair process. When the double-strandedsubstrate was recovered from the reaction mixture andanalyzed by Hpall cleavage prior to denaturing gelelectrophoresis, 80-90 % of the incorporated 32p-dCMPwas found at the previous position of the dUMP residuerather than on its 3' side, as estimated from the amountof incorporation in the 12-nucleotide-long fragment

We have used a similar strategy in introductory experi-ments with cell extracts, but to minimize problems of

Fig. 2. BER of uracil-containing oligonu-cleotides by E. coli cell extracts of differentstrains. (a) DNA repair synthesis in areaction mixture containing la-32p] dCTPand an extract of E. coli NH 5033 (recB,sbcB, endA) with the control (lane 1) oruracil-containing oligonucleotide (lane 2).Oligonucleotides were analyzed by dena-turing gel electrophoresis as described inMaterials and methods. Sizes of oligonu-cleotides were determined by comparisonwith markers of known length. (b) Repairof the uracil-containing oligonucleotide bycell extracts of E. coli NH5033 (recB sbcBendA ung+, lane 1), BD10 (ung, lane 2)and its isogenic parental strain W311 0(ung+, lane 3). (c) Hpall-cleavage analysisof the uracil-containing oligonucleotideafter DNA repair synthesis by the NH5033extract. lane 1, oligonucleotide nottreated with Hpall; lane 2, oligonucleotidetreated with Hpall.

RESEARCH PAPER 10,Dianov and Lindahl

to the 18-mer (Fig. 2c). In agreement withresults [4], no detectable incorporation of

the 5' side of the dUMP(data not shown). The presence of a U ratherresidue within the Hpall recognition site makes

oligonucleotide refractory to restriction enzyme-As >95% of the oligonucleotide substrate had

a form susceptible to HpaII cleavage,in vitro repair reaction had proceeded virtually to

.(Fig. 2c). These data with oligonucleotidesterminal phosphorothioate bonds are in

agreement with previous results [4] and demon-more clearly the, absolute dependence of the

on the ung gene product.

was the Rec] protein [13], originally described as a5' ~3' exonuclease active on single-stranded DNA [14].DNA polymerase 1 and DNA ligase are required for sub-sequent repair synthesis and rejoining to complete therepair reaction.

Using low concentrations of the five E. coli enzymeS -uracil-DNA giycosylase, endonuclease IV; Rec] protein,DNA polymerase I and DNA ligase ->90% replace-ment of the dUMP residue with dCMP was achievedwith a double-stranded DNA substrate (Fig. 3). Thesubstrate for this experiment had the same sequence asthe oligonucleotide in Fig. la, but lacked terminal phos-phorothioate bonds and contained a 5'_32p residue in theupper strand. At the end of the repair reaction, theoligonucleotide duplex was recovered, deproteinized,divided into two aliquots, one of which was treated withHpall, and subject to denaturing gel electrophoresis andautoradiography. Repair was detected as conversion ofthe oligonucleotide to a full-length form susceptible toHpall digestion (Fig. 3, lanes 4 and 9). In the experi-ments with cell extracts (Fig. 2), -2 pmoles of uracil-containing oligonucleotide duplex were repaired in 20minutes at 37 °C, so in the reconstituted repair systemthe lowest concentrations of each enzyme t~at wouldallow a similar level of excision-repair were determinedand used.

or apyrimidinic site in DNA, generated by aglycosylase, is attacked by an AP endonuclease. E.

such enzymes, exonuclease III and endonu-IV, which have overlapping specificities and areto be involved in the repair of abasic sites [3]. (In

of its name, the main function and activity ofIII is as an AP endonuclease.) Both enzymes

cut DNA by a hydrolytic reaction on the 5' side of thebase-free residue [11]. In the present reconstitutionexperiments, we have used endonuclease IV as the APendonuclease, as it lacks an associated exonucleaseactivity, but equivalent results were obtained in reconsti-tution mixtures using low concentrations of exonucleaseIII (data not shown).

Fig. 3. Efficiency of BER by the reconsti-tuted repair system. After incubation ofa S'-32P-labelled (top strand) oligonu-cleotide duplex (sequence in Fig. 1 a)with reagent en~ymes as indicated (top),one-half of the material was treated withHpall and the other left uncut prior toanalysis by gel electrophoresis andautoradiography (bottom).

30

1211

...' .After the incision by AP endonuclease, the 5' dRpresidue must be removed from the incised abasic site togenerate a single nucleotide gap. The 5' ~3' exonucleasefunction of DNA polymerase I is unable to catalyze thisreaction [12]. In a survey of E. coli excision activities, theonly en~y'me found to remove a 5' terminal sugar-phosphate residue as free dRp in a hydrolytic reaction,

When all five enzymes were present, full-length oligo-nucleotide was regenerated and was cleaved by Hpall toyield a 12-mer, demonstrating replacement of uracil bycytosine and completion of the BER reaction. If uracil-DNA glycosylase was excluded from the reactionmixture, no repair occurred, as judged from the resistanceof the oligonucleotide to Hpall digestion (Fig. 3, lane 6).This result is in agreement with the experiment using anung- cell extract (Fig. 2b). Similar data were obtained~hen t~ AP endonuclease was excluded from the

Current Biology 1994, Vol 4 No 121072

polymerase I, the product of the reconstituted repairreaction was analyzed by Hpall cleavage and two differentconcentrations of DNA polymerase I were used inparallel reaction mixtures. At a relatively high concentra-tion of DNA polymerase I (0.5 unit), a larger amount ofrepair synthesis was observed in the absence than in thepresence of Rec] protein (Fig. 5a). The increased repairsynthesis in the absence of Rec] protein was almostentirely due to incorporation at sites on the 3' side of thedUMP residue rather than at this residue, as seen from thedistribution of incorporated radioactive material betweenthe two Hpa II fragments (Fig. 5a). This is consistent withthe increased strand displacement, removal of dRp as partof a small oligonucleotide by DNA polymerase I, andlonger repair patches generated in the absence o.f Rec]protein. At a 10-fold lower concentration of DNA poly-merase I (0.05 unit), the results were different in thatRec] protein now promoted DNA repair synthesis (Fig.5b, lanes 1 and 3). In the presence ofRec] protein, theamount of repair synthesis was similar with low and highconcentrations of DNA polymerase I (Fig. 5a, lane 2 andFig. 5b, lane 3) and repair occurred mainly by one-nucleotide replacement, as demonstrated by digestion ofthe product with Hpall (Figs. 5a and 5b, lane 4).

reaction mixture. In this case, reactions were stopped byNaBH4 treatment [4] to stabilize the abasic site duringsubsequent analysis. A very slight mobility shift of the 30-mer was observed, caused by conversion of an internaldUMP residue to an abasic site (Fig. 3, lanes 2 and 7). Inthe absence of DNA polymerase I, cleavage of the uracil-containing substrate occurred on the 5' side of the dUMPresidue by the concerted action of uracil-DNA glycosy-lase and endonuclease IV to generate a S'-32P-Iabelled 11-mer (Fig. 3, lane 3), but rejoining did not occur as E. coliDNA ligase shows little or no activity at an incised abasicsite. These results demonstrate that the expected reactionintermediates could be isolated, and that the crude cellextract used previously for the BER in vitro reaction couldbe efficiently replaced with purified enzymes.

When a non-radiolabelled substrate (Fig. 1a) was used inan [a-32P]dCTP-containing reaction mixture with thesame amounts of reagent enzymes as in Figure 3, replace-ment of the dUMP with a dCMP residue was visualizeddirectlY1Fig. 4, lane 6). In agreement with the data inFig. 3, no incorporation of radioactive material occurredin reaction mixtures lacking uracIl-DNA glycosylase,endonuclease IV or DNA polymerast; I (Fig. 4, lanes1-3). In the absence of DNA ligase, "a heterogeneousarray of radioactive large fragments were observed, mostlikely generated by strand displacement and gap-filling byDNA polymerase I (Fig. 4, lane 4). In the absence ofRecj protein, a larger amount of radioactive material wasincorporated into the repaired 30-mer (Fig. 4, lane 5),even more than in the fully repaired substrate from thecomplete reaction mixture (Fig. 4, lane 6; compare Fig.3, lanes 4 and 9). This effect may be dependent onexcision of dRp residues as part of small oligonucleotidesby Fol I and generation of larger repair pat~hes.

The 5' ~3' exonuclease function of DNA polymerase I isknown to be strongly but not completely inhibited bya base-free deoxyribose-phosphate residue at the 5'terminus [12], so the removal of dRp apparentlyproceeded poorly in the absence of a separate excisionfunction such as the RecJ protein. In fact, the smallamount of resynthesis observed in the absence of RecJprotein at a low DNA polymerase I concentration (Fig.5b) most likely represented removal of a small proportionof 5'-terminal dRp residues by non-enzymatic j::\-elimi-nation, rather than by the 5' ~3' exonuclease activity ofDNA polymerase I, to generate small amounts of single-nucleotide patches (Fig. 5b, lane 2). The increased effi'"ciency of generation of one-nucleotide patches in the

Role of RecJ protein in minimizing the size of repair patchesIn order to assess further the potential role of RecJprotein in counteracting extensive gap-filling by D~A

30 .

Fig. 4. Roles of individual enzymes inthe reconstituted reaction. The reactionmixtures contained the oligonucleotideduplex shown in Fig. 1 a, [a32-p] dCTPand reagent enzymes as indicated. Theanalysis was performed as in Fig. 2a.

5. Effect of Recj protein on the in.The uracil-contain,.

ing oligonucleotide duplex (Fig. la) wasincubated in a reaction mixture contain-ing the enzymes necessary to reconsti-tute the reaction. Recj protein waspresent or absent, as indicated. One halfof each mixture was treated with Hpall,as indicated, prior to analysis by gelelectrophoresis. (a) Reactions at a mod-erately high concentration (0.5 unit) ofDNA polymerase I. (b) Reactions at alow concentration (0.05 unit) of DNApolymerase I.

presence ofRec] protein (Fig. 5b, lane 4) also shows thatDNA polymerase I at a low concentration more readilyfills a one-nucleotide gap with a 5'-terminal nucleotideresidue than a gap with a remaining dRp residue at the 5'end. Thus, a single-nucleotide gap rather than an'overhang' or 'flap' structure with a single 5' displaceddRp residue is the most likely reaction intermediate inthis very short patch ex<:ision-repair process.

5' ~3' exonucleases of E. coli, are unable to release a 5'-terminal dRp residue in free form to generate a one-nucleotide gap, but enzymes such as RecJ and Fpgproteins possess the necessary activity.

In the alternative pathway, a 'patch-and~cut' strategy isused, involving strand displacement of the 5' terminuswith the dRp residue. This is followed by incision due tothe structure-specific 5' nuclease activity of DNA poly-merase I [20], which cuts at the branch point to release asingle-stranded displaced oligonucleotide. Results withE. coli [4] and X. laevis [7,8] extracts indicate that suchrepair patches are usually only two to five nucleotideslong, but occasionally much longer displaced regions aregenerated [21]. Another minor back-up pathway couldinvolve cutting on both sides of an intact or incised abasicsite by the UvrABC nuclease; a gap aboqt thirteennucleotides long would be generated, which would berepaired by the NER pathway. The UvrABC enzymecan slowly incise abasic sites [22], although it is much lessefficient than AP endonuclease, and this pathway couldbe of relevance if the dRp-excision activity is low.

A 5' terminal dRp residue at a strand interruption inDNA may be removed in a ~-elimination process ratherthan by hydrolysis. Such events are promoted efficientlyby the E. coli Fpg protein [15], and at a slower rate by avariety of basic proteins and polyamines [16,17]. Bothhydrolysis and ~-elimination events generate identicalone-nucleotide gaps in DNA, so they would seemequally useful strategies in this regard, although in thelatter process the product excised is an unsaturatedaldehyde form of dRp. We have been able to substitutethe RecJ protein with the same amount ofFpgprotein inreconstitution experiments such as that shown in Fig. 5to obtain similar results (data not shown). Thus, theseparate dRp excision function required for the one-nucleotide replacement pathway can be carried out by atleast two different enzymes with equivalent results.

In view of the size heterogeneity of repair patches,average estimates based on overall incorporation ofradioactive precursors are of limited value, and moredetailed analysis of the DNA product is required. Similarobservations of repair patch heterogeneity have beenmade with either covalently closed DNA circles [8] ordouble-stranded oligonucleotides [4] as substrates. Theoccurrence of an alternative pathway dependent onstrand displacement by DNA polymerase I explains anearly study by Verly et al. [23], in which apurinic sites inDNA could be repaired by incubation with the major E.coli AP endonuclease (exonuclease III), DNA poly-merase.I and T4 DNA ligase. This alternative pathway,which generates an 'overhang' structure of single-stranded DNA [24] that could interact with relatedDNA sequences, also offers a satisfactory explanation forthe hyper-recombinogenic phenotype of E. coli mutantsdefective in dUTPase, the 5' nuclease activity of DNApolymerase I or DNA ligase, all of which would result inincreased or longer-lasting 'overhang' reaction inter-niedi~tes [24,25].

Discussion

A general model for the BER process (Fig. 6) may beproposed on the basis of the results described above andrecent results obtained with cell extracts and partiallypurified components [4,8]. This model is similar toprevious schemes [18,19], but the existence of abranched pathway and the events associated with removalof the deoxyribose-phosphate residue at an incised abasicsite are new features. The main route, which has beenreproduced with purified enzymes in the present study,involves generation of a single nucleotide gap as reactionintermediate and appears to be a 'cut-and-patch' process.After DNA glycosylase and AP endonuclease have acted,the terminal dRp residue is excised. The AP lyaseactivity of endonuclease III, which may occasionallypromote J:3-elimination at intrinsic abasic sites, and most

Current Biology 1994, Vol 4 No 121074

Fig.

6. Scheme for the branched BERpathway. The major, very short-patchpathway, reconstituted in the presentwork with purified enzymes, is shownin the left branch. A minor pathway,resulting in replacement of severalnucleotides, is shown in the rightbranch. Newly incorporated residuesare indicated with bold, yellow:.high-lighted bases. See text for details.

5'- PT PTPTPTPTPT3'GUCGATCGGCTA

3'.L p.Lp.Lp.LP.LP.LP-5'

5'- PT PTPTPTPTPT3'G CGATCGGCTA

3'.1. p.L pi p.l.p.l.p.l.p- 5'

OH P>- '

5'-PI PTPTPTPT3'G C GATC G GCTA

3'L P -Lp-Lp-Lp-Lp-Lp- 5'

.OH

.<) /'..)../

;6

>-A" (' >-A

/' C' >-A

5'- PTPTPTPTP; OH)-J'GCCGA TCGGCT A

J'-p.Lp.Lp.Lp.LpL P ~ P-5'

5'L-P-r PTPTPTPT3'G C GATC G GCTA

3'.L P -Lp-Lp.LP.LP.LP-5'

,.i oX-

S'-PTP T PTPTP-{ PT3'GCCGA TCGGCT A

3'-LP-LPl-pl-p-L pl-p-S'

0-<- .5'-PTP--( PTPTPTPT3'

G C CGATC G G C T ,A

3'-LP .L P .Lp.Lp.Lp.Lp- 5'

5'- PT PT PT PTPTPT3'GCCGATCGGCTA

3'.L p.L p.Lp.LP.LP.LP-5'

5'-PTPTPTPTPTPT3'GCCGATCGGCTA

3'.L p.L p.L p.LP.LP.LP-5'

Ie 1994 Current Biology

properties very similar to those of E. coli endonucleaseIV; human DNA polymerase f3 can replace E. coli DNApolymerase I in vivo [28]; and, similarly, a DNA ligase-deficient E. coli strain can be complemented with thecatalytic domain of a human DNA ligase [29].Mammalian cells also contain a 50 kD activity that

The E. coli enzymes active in the reconstituted pathwayhave direct counterparts in human cells, and the mainfeatures of the reaction would be expected to be retainedin all living cells. Thus, uracil-DNA glycosylase is one ofthe most highly conserved proteins between E. coli andman [26]; human AP endonuclease [27] has biochemical

RESEARCH PAPERDianov and Lindahl

hydrolyticaily excises 5'-terminal dRp from incisedbe the counterpart of the E.

coli Rec J protein [9], but the lability of this activity has sofar prevented its extensive purification and the reconsti-tution of the pathway with the human equivalents of theE. coli enzymes. The involvement of DNA polymerase 13in the very short patch BER process [4,8,30] is ofinterest, as this enzyme lacks proofreading activity andexhibits an unusually high error rate for a DNA poly-merase, although it is efficient at filling-in small single-strand gaps in DNA [31,32]. The restriction ofresynthesis to a single nucleotide in most cases of BERminimizes the problem of error-prone DNA repairsynthesis by DNA polymerase 13 in the repair ofspon-taneous DNA damage and DNA turnover.

in equimolar amounts, as described [4]. Formation of theappropriate double-stranded molecules was verified by gelelectrophoresis under non-denaturing conditions prior to use.Double,stranded oligonucleotides, 5'-32P-labelled in one strand,were made by incubation of single-stranded oligonucleotideswith T4 polynucleotide kinase and [-y_32p]ATP, followed byannealing with a complementary strand and gel purification.

BER reactionsStandard reaction mixtures with purified DNA repair enzymes(50 ~l) contained 0.1 M Tris. HCI (pH 7.5), 50 mM KCI,5 mM MgCI2' 0.1 mM EDT A, 1 mM dithiothreitol, 0.5 mMNAD, 20 ~M each ofdATP, dGTP and TTP, 2 ~M dCTPincluding 2 ~Ci [a-32P]dCTP (3000 Ci mmol-l, Amersham),5 ng uracil-DNA glycosylase, 5 ng endonuclease IV, 50 ngRecJ protein, 0.05-0.5 units of DNA polymerase I, and 0.2units DNA ligase. The components were mixed at 0 °C andreactions were initiated by addition of 2 pmoles duplexoligonucleotide and transfer of the reaction mixture to 37°C.Thus, the ratio in reaction mixtures between dCTP moleculesand uracil residues in oligonucleotide form was 50: 1. After20 min at 37°C, reaction mixtures were chilled, and 5 ~ 3 MNaCI containing 0.1 mg ml-1 tRNA and 50 ~l of phenol-chloroform (1:1) were added. After agitation and centrifuga-tion, the aqueous phase was recovered and supplemented with3 volumes of ethanol, The precipitate was recovered by centri-fugation, washed twice with ethanol, and dried under vacuum.

Conclusions

The main pathway of the DNA base excision-repairprocess results in replacement of a single nucleotideresidue and has been reproduced at high efficiency with amixture of five different enzymes: a DNA glycosylase toexcise th~ damaged base, an AP endonuclease for incisionat the abasic site, a deoxyribophosphodiesterase forexcision of the 5' terminal deoxyribophosphate residue, aDNA polymerase for gap-filling, and a DNA ligase forrejoining. The results with the five purified enzymes aresimilar to those obtained with crude cell extracts, indi-cating that no additional enzymes or accessory proteinfactors are required for the reaction. However, a minoralternative pathway also exists and generates longer repairpatches. This route is initiated in the same way by aDNA glycosylase and an AP endonuclease, but subse-quent DNA strand displacement and resynthesis wouldbe expected to occur in a way closely related to the laterstages ofNER or lagging-strand DNA replication, and torequire similar activities. The arrangement providesincreased versatility for this major repair process, which isused to counteract spontaneous cellular DNA damage.

For restriction enzyme analysis, the duplex oligonucleotide wasredissolved in 50 ILl 10 mM Tris.HCl (pH 7.5), 10 mMMgClz, 1 mM dithioerythritol and digested with 50 unitsHpaII for 60 minutes at 37°C. Reactions were stopped byphenol-chloroform extraction in 0.3 M NaCI followed byethanol precipitation. The dried precipitate was dissolved in10 ILl 80 % formamide, 0.1 % xylene cyanol, 0.1 % bromo-phenol blue. After heating at 90 °C for 2 minutes, the materialwas loaded onto a 15 % polyacrylamide gel containing 7 Murea in 90 mM Tris-borate/2 mM EDT A (pH 8.8) and elec-trophoresed at 30 mA, followed by drying of the gel andautoradiography. Band intensities on autoradiograms werequantified with an LKB Ultrascan XL scanning laser densito-meter. Reactions containing crude cell extracts instead ofpurified enzymes were carried out in a similar reaction mixtureas described [4], and were analyzed as above.

/

Materials and methods References

1. Lindahl T: Instability and decay of the primary structure of DNA.Nature 1993, 362:709-715.

2. Cleaver IE: It was a very good year for DNA repair. Cell 1994,76:1-4.

3. Cunningham RP, 5aporito 5M, Spitzer SG, Weiss B: EndonucleaseIV (nfo) mutant of Escherichia coli. J Bacteriol 1986,168:1120-1127.

4. Oianov G, Price A, Lindahl T: Generation of single-nucleotiderepair patches following excision of uracil residues from DNA. MolCeIIBio/1992,12:1605-1612.

5. Wang Z, Wu X, Friedberg EC: DNA repair synthesis during baseexcision repair in vitro is catalyzed by DNA polymerase E and isinfluenced by DNA polymerases a and 0 in Saccharomyces cere-visiae. Mol Cell BioI 1993, 13:1051-1058.

6. Matsumoto Y, Bogenhagen OF: Repair of a synthetic abasic site inDNA in a Xenopus laevis oocyte extract. Mol Cell BioI 1989,9:3750-3757.

7. Matsumoto Y, Bogenhagen OF: Repair of a synthetic abasic siteinvolves concerted reactions of DNA synthesis followed by excisionand ligation. Mol Cell BioI 1991, 11 :4441-4447.

8. Matsumoto Y, Kim K, Bogenhagen OF: Proliferating cell nuclear

Reagent enzymesThe following E. coli enzymes were purified from enzyme-overproducing strains as described: uracil-DNA glycosylase[33], endonuclease IV [34], Rec] protein [13,14] and Fpgprotein [15]. E.coli DNA polymerase I, DNA ligase, exo-nuclease III and T4 polynucleotide kinase were obtained fromBoehringer-Mannheim. Restriction endonuclease Hpall waspurchased from New England Biolabs, Inc.

Cell extracts and oligonucleotide substratesE. coli NH5033 (recB, sbcB, endA) was obtained from S.C. West,and E. coli BDI0 (ung, thyA, deoC) and its ung+ parent strainwere from B. Weiss. Cell extracts (-10 mg ml-l protein) wereprepared by the lysozyme-EDT A me.thod [35] and stored asaliquots at -80 DC. Single-stranded oligonucleotides wereprepared on a commercial DNA synthesizer. Double-strandedoligonucleotides were made by mixing single-stranded ones

Current Biology 1994, Vol 4 No 121076

antigen-dependent abasic site repair in Xenopus laevis oocytes: analternative pathway of base excision DNA repair. Mol Cell Bioi

1994,14:6187-6197.9. Price A, Lindahl T: Enzymatic release of 5'-terminal deoxyribose

phosphate residues from damaged DNA in human cells.

Biochemistry1991,30:8631-8637.10. Duncan B, Miller JH: Mutagenic deamination of cytosine residues

in DNA. Nature 1980, 287:560-561.11. Warner HR, Demple BF, Deutsch WA, Kane CM, Linn, S:

Apurinic/apyrimidinic endonucleases in repair of pyrimidinedimers and other lesions in DNA. Proc Natl Acad Sci USA 1980,77:4602-4606.

12. Price A: Action of Escherichia coli and human 5'-t3' exonucleasefunctions at incised apurinic/apyrimidinic sites in DNA. FEBS Lett

1992,300:101-104.13. Dianov G, Sedgwick B, Daly G, Olsson M, Lovett S, Lindahl T:

Release of 5'-terminal deoxyribose-phosphate residues from incisedabasic sites in DNA by the Escherichia coli Recl protein. NucleicAcids Res 1994, 22:993-998.

14. Lovett ST, Kolodner RD: Identification and purification of a single-stranded DNA-specific exonuclease encoded by the recl gene ofEscherichia coli. Proc Natl Acad Sci USA 1989, 86:2627-2631.

15. Graves RI, Felzenswalb I, Laval I, O'Connor TR: Excision of 5'terminal deoxyribose phosphate from damaged DNA is catalyzedby the Fpg protein of Escherichia coli. J Bioi Chern 1992,267:14429-14435.

16. Lindahl T, Andersson A: Rate of chain breakage at apurinic sites indouble-stranded deoxyribonucleic acid. Biochemistry 1972,11:3618--3623.

17. Bailly V, Verly WG: Possible roles of p-elimination and o-elimina-tion reactions in the repair of DNA containing AP(apurinic/apyrimidinic) sites in mammalian cells. Biochem J 1988,253:553-559.

18. Lindahl T: New class of enzymes acting on damaged DNA. Nature1976, 259:64-66.

19. Friedberg EC: Excision repair. DNA glycosylases and AP endonucle-ases.ln DNA Repair. New York: Freeman;1985:141-149.

20. Lyamichev V, Brow MAD, Dahlberg JE: Structure-specific endonu-cleolytic cleavage of nucleic acids by eubacterial DNA poly-merases. Science 1993, 260:778-783.

21. Mosbaugh OW, Linn S: Characterization of the action ofEscherichia coli DNA polymerase I at incisions produced by repairendodeoxyribonucleases. J Bioi Chern 1982, 257:575-583.

22. Snowden A, Kow WK, van Houten B: Damage repertoire of theEscherichia coli UvrABC nuclease complex includes abasic sites,base-damage analogues, and lesions containing adjacent 5' or 3'

nicks. Biochemistry 1990,29:7251-7259.23. Verly WG, Gossard F, Crine P: In vitro repair of apurinic sites in

DNA. Proc Natl Acad Sci USA 1974, 71 :2273-2275.24. Lundquist RC, Olivera BM: Transient generation of displaced single-

stranded DNA during nick translation. Cell 1982, 31 :53-60.25. Konrad EB: Method for the isolation of Escherichia coli mutants

with enhanced recombination between chromosomal duplications.

jBacterio/1977,130:167-172.26. Olsen LC, Aasland R, Wittwer CU, Krokan HE, Helland DE:

Molecular cloning of human uracil-DNA glycosylase, a highlyconserved DNA repair enzyme. EMBO j 1989,8:3121-3125.

27. Demple B, Harrison L: Repair of oxidative damage to DNA: enzy-mology and biology. Annu Rev Biochem 1994, 63:915-948

28. Sweasy JB, Loeb LA: Mammalian DNA polymerase p can substitutefor DNA polymerase 1 during DNA replication in Escherichia coli.j Bioi Chern 1992, 267:1407-1410.

29 Kodama K, Barnes DE, Lindahl T: In vitro mutagenesis and func-tional expression in Escherichia coli of a cDNA encoding thecatalytic domain of human DNA ligase I. Nucleic Acids Res 1991,

19:6093-6099.30. Wiebauer K, Jiricny J: Mismatch-specific thymine-DNA glycosylase

and DNA polymerase p mediate the correction of G- T mispairs innuclear extracts from human cells. Proc Natl Acad Sci USA 1990,

87:5842-5845.31. Pelletier H, Sawaya MR, Kumar A, Wilson SH, Kraut J: Structures of

ternary complexes of rat DNA polymerase p, a DNA template-primer, and ddCTP. Science 1994,264:1891-1903.

32. Singhal RK, Wilson SH: Short gap-filling synthesis by DNA poly-merase p is processive. j Bioi Chern 1993, 268:15906-15911.

33. Lindahl T, Ljungquist 5, Siegert W, Nyberg B, Sperens B: DNA N-glycosidases. Properties of uracil-DNA glycosidase fromEscherichia coli. j Bioi Chern 1977, 252:3286-3294.

34 Levin JD, Johnson AW, Demple B. Homogeneous Escherichia coliendonuclease IV. j Bioi Chern 1988, 263:8066-8071.

35. Connolly B, West SC: Genetic recombination in E.coli: Hollidayjunctions made by RecA protein are resolved by fractionated cell-free extracts. Proc Natl Acad Sci USA 1990, 87:8476-8480.

Received: 19 September 1994.Accepted: 6 October 1994.