Nucleic Acids Research, Vol. 18, No. 1

Active site amino acid sequence of the bovine0 6-methylguanine-DNA methyltransferase

Bjorn Rydberg, Janet Hall* and Peter KarranImperial Cancer Research Fund, Clare Hall Laboratories, South Mimms, Herts EN6 3LD, UK

Received October 2, 1989; Revised and Accepted November 10, 1989

ABSTRACT

An 06-methylguanine-DNA methyltransferase hasbeen partially purified from calf thymus by conventionalbiochemical techniques. The enzyme was specificallyradioactively labelled at the cysteine residue of theactive site and further purified by attachment to a solidsupport. Following digestion with trypsin, a radioactivepeptide containing the active site region of the proteinwas purified by size fractionation, ion exchangechromatography and reverse phase HPLC. Thetechnique yielded an essentially homogeneousoligopeptide which was subjected to amino acidsequencing. The sequence adjacent to the acceptorcysteine residue of the bovine protein exhibits strikinghomology to the C-terminal methyl acceptor site of theE. coli Ada protein and the proposed acceptor sites ofthe E. coli Ogt and the B. subtilis Datl proteins.

INTRODUCTION

06-Methylguanine (m6-Gua) is one of the major products of thereaction of a number of methylating agents with DNA and isresponsible for the potent mutagenicity of carcinogens such as

N-methyl-N'-nitro-N-nitrosoguanidine (MNNG) and themetabolically activated form of dimethylnitrosamine (1). Thespectrum of mutations induced by this group of methylating agentsis dominated by G C to A T transition mutations (2,3) which isconsistent with the observed propensity of m6-Gua to direct theincorporation of thymine into DNA in vitro (4). In addition toits well-documented mutagenicity, m6-Gua also contributes toother biological effects of alkylating agents. A number of humanand rodent cell lines are unable to remove m6-Gua from theirDNA. These cell lines are designated Mex- (or Mer-) and are

hypersensitive to the cytotoxic, clastogenic and mutagenic actionof methylating agents (5,6). The hypersensitivity of cells whichdo not remove m6-Gua can be completely reversed byexpression of a transfected E. coli ada+ gene encoding an

m6-Gua repair function (7-9). Thus, m6-Gua is stronglyimplicated not only in the mutagenic, but also in the cytotoxicand clastogenic action of agents such as MNNG towardsmammalian cells.

In many bacteria (including E. coli (10), M. luteus (11), andB. subtilis (12)) the first line of defence against the biologicaleffects of this methylated base is repair by specific DNA

methyltransferases which demethylate the modified purine in situbut are without effect on other chemically-induced or naturallyoccurring methylated bases. The bacterial methyltransferasesinclude the inducible Ada protein, a dual functionmethyltransferase comprising two domains which act separatelyto demethylate m6-Gua or methylphosphotriesters (10), and theconstitutively expressed Ogt proteins of E. coli (13) and the Dati(14) protein of B. subtilis. A common feature of these bacterialmethyltransferases is transfer of a methyl group from the modifiedpurine base onto a particular receptor cysteine residue within themethyltransferase molecule itself; the transfer being accompaniedby an irreversible inactivation of the methyltransferase function.m6-Gua-DNA methyltransferase activities have been partially

purified from several mammalian sources (15-18) andpreliminary characterisation has indicated that they share a

number of features with their bacterial counterparts. In particular,the automethyltransfer mode of repair has apparently beenconserved and mammalian cells from a variety of sources(including human) are able to demethylate m6-Gua both in vivoand in cell-free extracts. In all cases, removal of methyl groupsfrom m6-Gua in DNA is accompanied by a stoichiometricproduction of S-methylcysteine in a protease-sensitive formindicating that a cysteine residue serves as acceptor.

Despite considerable efforts in a number of laboratories,mammalian methyltransferases have proved refractory to highyield purification and this has hampered further clarification ofthe mechanism of action of this important DNA repair enzyme.These difficulties have been partly due to excessive losses of thepartially purified enzyme during the final stages of purification.Here we report the isolation and amino acid sequence of a peptidecomprising the active site of the bovine enzyme. The derivedsequence demonstrates a remarkable homology to the m6-Gua-DNA methyltransferase active site of the E. coli Ada protein andthe putative active site sequences of the E. coli Ogt and the B.subtilis Datl proteins.

MATERIALS AND METHODSMaterialsTrypsin (Sequencing Grade) was obtained from BoehringerMannheim. Sephadex G25 Superfine and the MonoS FPLCcation exchange column were obtained from Pharmacia andUltrogel AcA54 from LKB. DE52 ion exchange cellulose and

* Present address: International Agency for Research on Cancer, 150 Cours Albert Thomas, 69372 Lyon, France

17

I.) 1990 Oxford University Press

18 Nucleic Acids Research

phosphocellulose P11 (Whatman) were prepared according to themanufacturers instructions. Single-stranded DNA cellulose wasobtained from Sigma.

Synthetic OligonucleotidesThe 21mer oligonucleotide 5'-TGATCAGTAC(m6-G)CATGA-CTAGT-3' was synthesised by the phosphoramidite method onan Applied Biosystems Model 380B DNA synthesiser and furtherpurified by HPLC. For use as a substrate for themethyltransferase, it was annealed to the oligomer 5'-ACTAGT-CATGCGTACTGATCA-3' at a concentration of 24tM in 0. 1MNaCl, lOmM TrisHCl pH 7.5, 1mM EDTA at 37°C for60min.

m6-Gua-DNA Methyltransferase Assay[3H]-methylated M. luteus DNA was prepared using [3H]-N-methyl-N-nitrosourea (MNU) (Amersham International,24Ci/mmole) and partially depurinated as described (19). Assayswere carried out in: 70mM Hepes KOH, pH 7.8/ lOmMdithiothreitol/ 1mM EDTA. To monitor the purificationprocedure, 0.1-42,1 of each column fraction was incubated inIOO/.d reaction buffer containing [3H] substrate (lOOOcpm) for20min at 37°C. Following a digestion with proteinase K(125itg/120min) at 37°C, nucleic acids were precipitated withethanol and the [3H] radioactivity in the supernatant wasdetermined by scintillation counting. When appropriate, theremoval of m6-Gua from the DNA was also monitored bypublished procedures (18).

Partial Purification of m6-Gua-DNA Methyltransferase fromCalf ThymusThe procedure is based on that reported by Hall and Karran (18).All operations were performed at 0-4°C. 1.8Kg calf thymusfreshly obtained from the slaughterhouse, was homogenised in4 1 extraction buffer (O.iM NaCl, 50mM TrisHCl pH 7.5,1mM EDTA, 0.1% f3-mercaptoethanol, 0.1% Triton X-100,1mM phenylmethylsulphonyl fluoride, and 0.5yg/ml each:leupeptin, pepstatin and chymostatin) using a Waring blendor atmaximum setting for 2 x 30sec. Extraction was for 45min at 0°C.Tissue debris was then removed by centrifugation at 3200 x gfor 30 min. To 41 supernatant was added a thick slurry of 1.21DE52 in buffer A (5OmM NaCl, 20mM Tris HCI pH 7.5, IrnMK2HPO4, 1mM EDTA, 1mM dithiothreitol, 0.1% f-mercaptoethanol). The mixture was stirred for 30min and theDE52 allowed to settle. The supernatant was decanted and theDE52 was then washed with 41 buffer A. To the two combinedsupernatants (81) was added 21 phosphocellulose P11 as a thickslurry in buffer A. The mixture was stirred for lhr, the P11 wasthen allowed to settle and the supernatant was discarded. TheP11 was washed twice with 81 buffer A and then poured intoa column (35cm x 8.5cm diam.) which was eluted with bufferA containing 0.5M NaCl at a flow rate of 200ml/h. Fractionscontaining methyltransferase activity were combined (SOOmI) anddialysed against 201 buffer B (1mM potassium phosphate pH 7.5,1mM EDTA, 0.1 % 3-mercaptoethanol, 10% glycerol) for 18h.The dialysed sample was clarified by centrifugation at 15000 x gfor 30 min and loaded on a single-stranded DNA cellulose column(12cm x4cm diam.) equilibrated with buffer B. The column waswashed with 3 column volumes of buffer B and then elutedsuccessively with with 2 column volumes each of buffer Bcontaining 0. IM NaCl, 0.3M NaCl and 1M NaCl at a flow rateof 40ml/h. The activity was eluted with 1M NaCl, although the

enzyme has typically been eluted with 0.3M NaCl with otherbatches of DNA cellulose. The pooled active fractions (25ml)were loaded onto a column of Ultrogel AcA54 (120 cm x 2.5cmdiam.) equilibrated with buffer C (0. IM NaCl, 15mM potassiumphosphate pH 7.4, 1mM EDTA, 0.1 % f-mercaptoethanol, 10%glycerol). At this NaCl concentration, the active fractions wereeluted at 1.3-1.5 times the void volume of the column (V,),which is earlier than previously reported (1.8 x VO) when bufferC containing O.5M NaCl was used. This altered elution behaviourindicates possible aggregation of the methyltransferase orinteraction with other proteins at the lower salt concentration.The active fractions were pooled and concentrated using anAmicon ultrafiltration cell equipped with a Diaflo YM10membrane. Total yield was about 800 pmole active enzyme(0.5% recovery) with a specific activity of 300 units/mg protein(500-fold purification). Incubation of the enzyme with[3H]-labelled substrate followed by SDS-Page electrophoresisand fluorography showed a major radioactive product of about24kD (and a minor labelled species at about 27kD, probablyresulting from incomplete denaturation of the labelled protein)in good agreement with previous estimates of the molecular massof the protein (17).

Labelling the Enzyme and Binding to a Solid SupportAn estimated total of 600pmole partly purified enzyme in 6mlbuffer C was dialysed into reaction buffer by ultrafiltration usinga Diaflo YMIO filter and the volume adjusted to 20ml. A 2mlaliquot of this preparation was incubated for 30min at 37°C with106 cpm of [3H]-labelled DNA substrate in a glass test-tube,188ml was similarly incubated with 1nmole m6-Gua-containingoligonucleotide. The two reaction mixtures were then combined.2.4g siliconized glass wool, prepared by immersion in Repelcote(BDH) followed by several washes in distilled H20, was thenadded and the methylated enzyme was allowed to bind to theglass wool by incubation for a further 30 min at 37°C. The glasswool was then washed 3 times in assay buffer and twice indistilled H20 and allowed to drain without drying.

Trypsination and Size FractionationThe washed siliconized glass wool containing the adsorbedradiolabelled methyltransferase was immersed in 20ml lOmMNH4HCO3 pH 8.1, 1mM CaCl2, 0.05% Tween 20 containing0.l,ug/ml trypsin and incubated at 20°C for 16 hours. Thetrypzinised sample was concentrated by evaporation undervacuum to a final volume of 1.1 ml and precipitated material wasremoved by centrifugation. In order to separate intact trypsin fromthe shorter peptides, the sample was applied to a column(38cm x lcm diam) of Sephadex G25 (superfine) equilibrated with1OmM NH4HCO3 pH 8.1. A symmetrical radioactive peakwhich eluted at 1.3 x the void volume was collected andevaporated to dryness in a vacuum desiccator. This peakcontained approximately lOOpmole methylated peptide asestimated from its [3H] content. This step also effectivelyremoved very short peptides.

Ion Exchange Chromatography (FPLC) of PeptidesThe Sephadex G25-purified dried sample was dissolved in 0.5mlbuffer D (20mM 2[N-morpholino]ethanesulfonic acid-NaOH pH6.0, 0.02% Tween 20) and subjected to FPLC using a MonoScation exchanger (5cm x\O.5cm diam). The flow rate wasO.Sml/min with a gradient from buffer D to E (buffer Dcontaining 0.4M NaCI) over 60 min. The main radioactive peak

Nucleic Acids Research 19

appeared at a NaCl concentration of 0. 12M with a minor peakat 0. 14M. The fractions corresponding to the main peak ofmethylated peptide (about 35 pmole) were further purified byHPLC.

HPLCThe sample was prepared for HPLC by adsorption to a SEP-PAK C-18 cartridge (Waters) equilibrated with 50mM NH4AcpH 7.0, washed with 20% acetonitrile and the peptide eluted with40% acetonitrile in the same buffer. The sample was thenlyophilised, dissolved in 1011 70% formic acid, adjusted to 1001Ain Buffer F (1 % acetonitrile, 0.08% trifluoroacetic acid (TFA))and applied to an Aquapore ODS C-18 column (22cmx0.21cmdiam., Brownlee Labs) equilibrated in buffer F. The column waseluted at a flow rate of 0.4ml/min with a gradient of 0-60%buffer G (90% acetonitrile in 0.06% TFA) over 70min. Imlfractions were collected and the two fractions containingradioactive material were retained and separately lyophilised.

Amino Acid SequencingThe samples were dissolved in 30,11 0.1% TFA, adsorbed toPolyprene-treated glass discs and sequenced using a ABI 477Apeptide sequencer. A portion of the sample from each Edmandegradation cycle was saved for radioactivity determination.

RESULTSPurification of an Oligopeptide Containing theMethyltransferase Active SiteIn preliminary experiments (data not shown) to purify themethylated form of the methyltransferase, we obtained extremelypoor yields of the protein using a variety of experimentalapproaches. Although the methylated form of the protein doesnot bind to glass, it exhibits an exaggerated tendency to adhereto hydrophobic surfaces even in the presence of surface activenon-ionic detergents. Since this tendency to interact withhydrophobic surfaces- in particular with siliconized glass- wasnot shared by the contaminating proteins in the partially purifiedmaterial, we used this property to effect a further purification.A quantitative recovery of the intact methylated protein after itsadsorbtion to a siliconized glass support could be achieved onlyby the use of high concentrations of ionic detergent (1% SDS),the presence of which severely limited the possibilities for furtherpurification. In contrast, radioactivity from the specificallylabelled methylated protein could be recovered in >90% yieldfollowing a digestion with trypsin in the presence of the non-ionic detergent Tween 20. This observation formed the basis ofthe purification of a tryptic peptide containing the active cysteineresidue of the methyltransferase.Approximately 60pmole of partially (500-fold) purified

methyltransferase was incubated with [3H]-substrate DNAcontaining 70 pmole [3H]m6-Gua and 540pmole enzyme wasincubated with an excess of non-radioactive m6-Gua-containingdouble-stranded oligonucleotide. The reaction mixtures werecombined and supplemented with siliconized glass wool to whichthe methylated enzyme was allowed to adsorb. Non-adsorbedproteins were removed, along with the substrate DNA andoligonucleotide, by extensive washing with assay buffer followedby water. In preliminary experiments, SDS-PAGE and proteindetermination were used to monitor the composition of the startingmaterial and the material which bound to glass wool and couldbe removed by 1 % SDS. This analysis indicated that >90% ofthe contaminating proteins were removed in this step.

The radioactive protein was removed from the glass wool bytrypsin digestion. The glass wool bearing the methylated proteinwas submerged in a solution containing 0.05% Tween 20 and0.1lAg/ml trypsin and the mixture incubated at 20°C overnight.The resulting tryptic digest was applied to a column of SephadexG25 to separate the trypsin from the oligopeptides. Trypsin elutedin the void volume (VO) of the column whereas the radioactivematerial was included and eluted as a symmetrical peak atapproximately 1.3 x Vo. This material was retained and furtherfractionated by FPLC using a MonoS cation exchange column.Figure 1 shows the elution pattern from the MonoS column. Themajority of the radioactive material eluted as a sharp peak at aNaCl concentration of 0. 12M with a small shoulder appearingat around 0.14M NaCl.

Fractions from the major radioactive peak eluting at 0.12MNaCl were pooled and further purified by reverse phase HPLCusing an Aquapore ODS C-18 column. This procedure resolvedthe complex mixture of peptides into a number of peaks whichabsorbed at 220nm (Figure 2). However, all the radioactivitywas recovered in two fractions coincident with a single peak ofabsorbance. The two radioactive fractions which containedrespectively 12 and 6pmoles of methylated peptide (representingan overall yield of 3 %) were separately subjected to amino acidsequence analysis.

Amino Acid Sequence of the Active Site PeptideThe amino acid sequence of the purified peptide from the fractionfrom the HPLC column containing 12pmole is shown below:

1 (?) 2. Asn 3. Pro 4. Ile 5. Pro 6. le (Phe)7. Leu (Asp) 8. Thr (Ile) 9. Pro (Gln)

The sequence obtained from the fraction containing 6 pmole ofpeptide provided confirmation of this sequence. In the latersequencing cycles, corresponding to weaker signals, more thanone amino acid was detected indicating the presence of a lowlevel of contaminating oligopeptide. Where two amino acids weredetected, both are shown. In each case, however, the first is themore likely as judged by its relative abundance. Methylcysteineyields very poor fluorescence following derivatisation and is notdetectable under the standard conditions used for automatedsequencing. A portion of the sample from each Edmandegradation cycle was therefore retained and its[3H]-radioactivity determined separately. Figure 3 shows that no

400%-O

OU 200._*0'U- 100I

C2-)20 40 60 80

Fraction No

0.4

0.3 2CD0.2 coz

0.1

Figure 1. Purification of the Radioactively Labelled Active Site Peptide by FPLC.Separation was carried out on a MonoS cation exchange column. Elution bufferscontained: 20mM MES NaOH pH 6.0, 0.02% Tween 20 and a gradient of NaCl(0-0.4M) as shown. Fractions (0.5ml) were collected and aliquots (I0OL) weretaken for radioactivity determination. Fractions 26-28 were pooled and usedfor further purification.

20 Nucleic Acids Research

6

CII)0

E0C\JC\ja)0cCO.00U).0

5

4

3

2

20

Fraction

Figure 2. Reverse Phase HPLC Purificati(was performed on an Aquapore ODS C-18column was eluted with a linear gradientTFA in % acetonitrile/0.08% TFA. The

The absorbance baseline has been arbitra

Fractions (1 mmli) were collected and the ra(

The total radioactivity present in fraction,

All other fractions contained only backgroi

fractions were independently subjected to

150

0

50

C2.

5

Figure 3. Localization of the S_[3H]-met"Fraction 31 from the ODS C-18 column

analysis. The radioactivity in aliquots of the

cycle from the automatic peptide sequencer

counting.

radioactivity was released before ti

than 60% of the radioactivity wa:,

The remaining < 40% was eluted

strongly indicate that the accei

methyltransferase follows the prolabove sequence.

2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3

Ada Asn L Leu] Ala lie Val Ilie Pro Cys His Arg Val

Ogt Asni Flie Ser JIlie Val Val Pro Cys His Arg Val

Datl lAsni A Leu ro] IIlel Phe Val Pro Cys His Arg Val

Bovine [snj Pro~2 le Leu Thr 1Pro Cys

x

E0L

1 5

o.0VD

Figure 4. Comparisons Among the Active Sites of 06_Methylguanine-DNAMethyltransferases.Ada: The C-terminal 06_Methylguanine-DNA methyltransferase active site ofthe E. c-oli Ada protein.Ogt and Datl: The putative active sites of the E. coli Ogt and the B. subtilisDat I proteins.Bovine: The sequence of the purified calf thymus peptide.The methyl accepting cysteine is arbitrarily assigned position 10. Amino acidsexhibiting homology to the bovine sequence are shown boxed.

DISCUSSION4~~ In Figure 4 we present a comparison of the sequence of the

m m -Gua methyltransferase active site of the Ada protein ofcn E. coli along with the proposed active sites for two other related

__________________ ]methyltransferases; the constitutive Ogt protein of E. coli and40 60 the B. subtilis Dati protein. The data reveal a high degree ofiNo conservation in the protein sequence around the active sites of

these proteins. The conserved amino acids in the bacterialenzymes; the asparagine at position 2, isoleucine at position 6

on of the Labelled Peptide. Separation and proline at position 9 (the methyl acceptor cysteine is3column (22cm x0.2l1cm diam.). The arbitrarily assigned position 10) are also conserved in the bovine(0-60%) of 90% acetonitrile/0.06% ezm.I so ute neetta h mn cd nteNflow-rate was 0.4m1/min over 70min. ezm.I so ute neetta h mn cd nteNirily set at the bottom of the Figure. terminal side of the acceptor cysteine are of a predominantlydioactivity in aliquots was determined, hydrophobic nature. In this comparison, we have used theis31 and 32 is indicated by the bars. experimentally determined most probable amino acids forund radioactivity. The two radioactive positions 5-9 of the bovine enzyme. In the case of the leucineami~no acid sequence analysis. residue at position 7, this assignment is also supported by the

observation that the radioactive tryptic peptide bound to a MonoScation exchange column but not to a MonoQ anion exchangecolumn. This behaviour is not compatible with the presence ofa negatively charged amino acid, aspartic acid, at position 7.However, if as seems likely, the His-Arg doublet adjacent to theacceptor cysteine in the other three methyltransferases isconserved in the bovine enzyme, this would introduce twopositively charged amino acids into the peptide. A peptide of thissequence would still retain an overall positive charge with Aspassigned to position 7 and would be expected to exhibit theobserved behaviour on ion exchange chromatography. We note,however, that the presence of Tween 20 is necessary to preventirreversible binding of the peptide to both the anion and cationexchangers, most probably as a result of hydrophobic interactions.For this reason, we consider the assignment of Leu to position

1 0 1 5 20 7 to be most likely since it is more compatible with the(cle No hydrophobic nature of this region of the peptide. While the

assignment of isoleucine to position 8 would represent a closer

thyl Cysteine Residue in the Peptide. homology to the sequence of the E. co/i Ada protein, the overallsubecedo utmatc min aid hydrophobic nature of the amino acid sequence preceding the

fractions from each Edman degradation acceptor cysteine would nevertheless be maintained by therwas determiined by liquid scintillation presence of a threonine residue in this position.

The substantial homology between the active sites of theseproteins probably reflects a conservation of their reaction

ie 10th sequenator cycle. More mechanism. The DatlI protein and the Ogt protein exhibits recovered in the 10Oth cycle, considerable homology (approaching 50%) elsewhere in theI[in the 11th cycle.- These data primary sequence. Similarly, the carboxy-terminal domain of theptor cysteine residue of the Ada protein which contains the M6-Gua methyltransferase activeline residue at position 9 in the site is highly homologous to both the Ogt and Datl proteins

(13,14). In fact, a C-terminal fragment of the Ada protein which

1

Cy

Nucleic Acids Research 21

is similar in size to both Ogt and Datl can function efficientlyas an m6-Gua methyltransferase in its own right (20).

Despite the close resemblances among the bacterial proteinsequences and the similarity we have observed at the active siteregion of the bovine enzyme, it seems likely that such a highdegree of conservation between the prokaryotic and mammalianenzymes around the active site of the methyltransferase merelyreflects the nature of the methyltransfer reaction, in particularthe mobilisation of the methyl group, and is unlikely to extendthroughout the protein. In a recent comparison of the knownsequences of methyltransferases which catalyse the methylationof cytosine residues in DNA, Posfai et al (21) observed thata Pro Cys doublet was present in a block of relatively invariantamino acids around the putative active site of the cytosinemethyltransferases. It is of interest that, in the cytosinemethyltransferases, the amino acid preceding the ProCysdoublet by five amino acids is either Leu or Ile. Furthermore,the Pro Cys doublet is preceded by a block of hydrophobicamino acids. Both of these structural features are present in theactive site region of the m6-Gua-DNA methyltransferases.Thus it seems likely that the particular energetic requirementsof methyl group transfer will impose a strong selective pressurefor conservation of the active site region of themethyltransferases. Such selective pressure may not apply tothe remainder of the protein sequence. In support of this ideais the observation that both monoclonal antibodies and polyclonalantisera raised against the purified C-terminal domain of theE. coli Ada protein, which bears the m6-Gua DNAmethyltransferase function, do not cross-react with the humanor bovine methyltransferases on immunoblots (J.Hall,B. Sedgwick, unpublished data).An unambiguous assignment of the active cysteine residues

of a suicidal methyltransferase has so far only been made forthe two domains of the E. coli Ada protein (22,23). While theconserved sequences among the other bacterial enzymes stronglysuggest the location of their active sites, direct evidence islacking. The establishment of the active site sequence of thebovine methyltransferase reported here will, despite itsambiguities, enable an unequivocal assignment of the active siteto be made when the full sequence of the protein becomesavailable.

(1986)Carcinogenesis, 7, 2077-2080.9. Hall, J., Kataoka, H., Stephenson, C. and Karran, P. (1988) Carcinogenesis,

9, 1587-1593.10. Lindahl, T., Sedgwick, B., Sekiguchi, M. and Nakabeppu, Y. (1988) Ann.

Rev. Biochem., 57, 133-157.11. Ather, A., Ahmed, Z. and Riazuddin, S. (1984) Nucleic Acids Res., 12,

2111-2126.12. Morohoshi, F. and Munakata, N. (1987) J. Bacteriol., 169, 587-592.13. Potter, P.M., Wilkinson M. C., Fitton, J., Carr, F. J., Brennand, J., Cooper,

D. P. and Margison, G. P. (1987) Nucleic Acids Res., 15, 9177-9193.14. Morohoshi, F., Hayashi, K. and Munakata, N. (1989) Nucleic Acids Res.,

17, 6531-6543.15. Hora, J. F., Eastman, A. and Bresnick, E. (1983) Biochemistry, 22,

3759-3763.16. Pegg, A. E., Weist, L., Foote, R. S., Mitra, S. and Perry, W. (1983) J.

Biol. Chem., 258, 2327-2333.17. Harris, A. L., Karran, P. and Lindahl, T. (1983) Cancer Res., 43,

3247-3252.18. Hall, J. and Karran, P. (1986) In: Mymes, B. and Krokan, H. (eds), Repair

of DNA Lesions Introduced by N-Nitroso Compounds. NorwegianUniversity Press, pp. 73-88.

19. Karran, P., Lindahl, T. and Griffin, B. (1979) Nature, 280, 76-77.20. Demple, B., Jacobsson, A., Olsson, M., Robins, P. and Lindahl, T. (1982)

J. Bio. Chem., 257, 13776-13780.21. Posfai, J., Bhagwat, A. S., Posfai, G. and R. J. Roberts (1989) Nucleic

Acids Res., 17, 2421-2435.22. Demple, B., Sedgwick, B., Robins, P., Totty, N., Waterfield, M. and

Lindahl, T. (1985) Proc. Natl. Acad. Sci. (USA), 82, 2688-2692.23. Sedgwick, B., Robins, P., Totty, N. and Lindahl, T. (1988) J. Biol. Chem.,

263, 4430-4433.

ACKNOWLEDGEMENTSWe thank Claire Stephenson for skilled technical assistance, lainGoldsmith for oligonucleotide synthesisis and Ron Brown for theamino acid sequence analysis.

REFERENCES1. Singer, B. and Grunberger, D. (1983) Molecular Biology of Carcinogens

and Mutagens. Plenum Press, New York.2. Loechler, E., Green, C. L. and Essigmann, J. M. (1984) Proc. Natl Acad.

Sci. (USA), 81, 6271-6275.3. Hill-Perkins, M., Jones, M. D. and Karran, P. (1986) Mutat. Res.,162,

153-163.4. Abbott, P. J. and Saffhill, R. (1979) Biochim. Biophys Acta, 562, 51-61.5. Day, R. S., Ziolkowski, C. H. J., Scudiero, D. A., Meyer, S., Lubiniecki,

A. S., Girardi, A. J., Galloway, S. M. and Bynum, G. D. (1980) Nature,288, 724-727.

6. Sklar, R. and Strauss, B. (1981) Nature, 289, 417-420.7. Ishizaki, K., Tsujimura, T., Fujio, C., Yangpei, Z., Yawata, H.,

Nakabeppu, Y., Sekiguchi, M. and Ikenaga, M. (1987) Mutat. Res., 184,*121-128.

8. White, G. R. M., Ockey, C. H., Brennand, J. and Margison, G.P.