Structural studies of MJ1529, an O6-methylguanine-DNA methyltransferase

MAGNETIC RESONANCE IN CHEMISTRYMagn. Reson. Chem. 2006; 44: S71–S82Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/mrc.1823

Structural studies of MJ1529,an O6-methylguanine-DNA methyltransferase

Anne Roberts, Jeffrey G. Pelton and David E. Wemmer∗

Department of Chemistry, University of California and Physical Biosciences Division, Lawrence Berkeley National Lab, Berkeley, CA 94720-1460, USA

Received 29 November 2005; Revised 22 February 2006; Accepted 28 February 2006

The structure of an O6-methylguanine-DNA methyltransferase (MGMT) from the thermophileMethanococcus jannaschii has been determined using multinuclear multidimensional NMR spectroscopy.The structure is similar to homologs from other organisms that have been determined by crystallography,with some variation in the N-terminal domain. The C-terminal domain is more highly conserved in bothsequence and structure. Regions of the protein show broadening, reflecting conformational flexibility thatis likely related to function. Copyright 2006 John Wiley & Sons, Ltd.

KEYWORDS: NMR; protein structure; DNA methyltransferase; DNA repair

INTRODUCTION

The survival and propagation of an organism depend on theintegrity and faithful replication of its DNA. Modificationsor alterations to DNA can be caused by external sources (UVradiation or chemical agents), endogenous metabolites, andnaturally occurring errors in incorporation of nucleotidesduring DNA replication. Each may result in deleteriouseffects on the cell. To avoid such effects, a number of DNA-repair systems have evolved to prevent these potentiallyharmful modifications from being propagated during DNAreplication: base-excision repair (BER); nucleotide excisionrepair (NER); and direct reversal (DR) repair. These DNA-repair systems are highly conserved among all branches oflife. While the first two systems involve multiple enzymes tocleave, excise, and replace the damaged DNA, DR involvessingle enzymes that can repair the damaged base withoutremoving nucleotides.1 – 3

A prominent protein that functions in DR isO6-methylguanine-DNA methyltransferase (MGMT) oralkyguanine alkyltransferase (AGT). Alkylating agentssuch as N-methyl-N’-nitro-N-nitrosoguanidine (MNNG) andN-methyl-N-nitrosourea (MNU) covalently modify DNA byalkylating the N7 and O6 position of guanine, and theO4 position of thymine. Although N7-alkylated guanineis the predominant alkylation product, this modificationis relatively harmless, as it does not affect the naturalbase pairing. In contrast, O6-methylguanine (O6MG), amuch less common product, is a mutagenic, carcinogenic,and potentially cytotoxic lesion.4,5 Incorporation of themethyl group leads to a preference for pairing of themodified guanosine with thymine rather than its normalWatson–Crick partner cytosine. This greatly increases the

ŁCorrespondence to: David E. Wemmer, Department of Chemistry,MC-1460, University of California, Berkeley, CA 94720-1460, USA.E-mail: [email protected]

chance of incorporation of thymine opposite O6MG duringsubsequent replication, resulting in a GC ! AT mutation(Fig. 1).6

MGMT repairs O6MG, and to a lesser extent O4

methylthymine, by covalently transferring the methyl (oralkyl) group to the protein. A cysteine residue in the proteinacts as a nucleophile and displaces the methyl group onthe guanine, restoring the original base. Once the protein ismethylated there is no mechanism to dealkylate the protein,thus, MGMT is known as a ‘suicide’ repair protein; oneprotein molecule is used to repair one alkyl lesion.7 AlthoughMGMT protects against alkylation-induced carcinogenesis,much interest in MGMT stems from the fact that alkylatingand chloroethylating agents are used as chemotherapeutictreatments for their cytotoxicity. Although the mechanism foralkylation-induced cell death is not completely understood(it appears to involve the mismatch repair-MMR system),alkylating agents do precipitate early cell death, and areused in the treatment of some forms of cancer.8,9 By repairingO6MG and other products formed by these agents, MGMTinterferes with the cell-killing effects of treatments. Thus,effective inhibitors of MGMT are of great interest; onein particular, O6-benzylguanine (O6BG) has been used inclinical trials.9

Investigation of the kinetics of repair by MGMT revealsthat in both human MGMT (hMGMT) and Escherichiacoli AdaC (the C-terminal domain of the Ada proteinwhich has MGMT activity) repair of a single O6MG lesionwithin single or double-stranded DNA is extremely fast,105 –108 M�1s�1, approaching the diffusion limited rate.10 – 12

The repair process is facilitated by the physical interactionof the protein and DNA: incubation of the free O6MG basewith hMGMT results in slow demethylation, but the additionof nonmethylated DNA to the free base/hMGMT mixturerestores the fast kinetics.13 While there may be some subtledifferences in repair rate depending on the surroundingsequence, it is clear MGMT has little sequence specificity for

Copyright 2006 John Wiley & Sons, Ltd.

S72 A. Roberts, J. G. Pelton and D. E. Wemmer

Figure 1. Altered base pairing of O6 methylguanine that results in a GC ! AT mutation during replication.

repairing O6MG, as might be expected from a lesion thatcould occur anywhere in the genome.14 – 18 Binding constantstudies put dissociation constants (Kd� for MGMT–DNAcomplexes in the low (0.5–20) micromolar range for bothmethylated and unmethylated DNA.10,11,19 – 21 This moderateaffinity for DNA and low preference for methylated overunmethylated DNA in particular may be reflected in the lowsequence specificity of repair.

As indicated above, the most highly studied MGMTproteins are hMGMT and AdaC. Although the proteinsare equivalent in their overall function, some differencesexist in their repair and binding characteristics. For instance,hMGMT is believed to bind cooperatively to DNA, whileAdaC does not seem to.10,15 In addition, hMGMT isinactivated by free O6-benzylguanine (O6BG), while AdaCis not. There are also differences in the rate of repairfor different adducts by these proteins.12,20 Some of thesedifferences reflect changes at the amino-acid sequence level;a tryptophan in AdaC located at an equivalent position asa serine in hMGMT prevents O6BG from diffusing into theactive site and inactivating AdaC.21 Other differences mayreflect evolutionary changes in the way MGMT carries outits activity.

For example, the Ada protein is part of the adaptiveresponse system in E. coli. Ada is a two-domain protein; the10 kD N-terminal domain (AdaN) demethylates innocuousmethylphosphotriesters with a cysteine residue, and the19 kD C-terminal domain (AdaC) has MGMT activity. Theadaptive response system appears to be a unique feedbackmechanism within bacteria. Upon methylation, AdaN turnsinto a transcriptional activator of the Ada gene and othergenes involved in DNA-repair, including, AlkA, AlkB, andOGT (another protein with MGMT activity).6 Althoughthere are feedback mechanisms for the induction of MGMTexpression in human cells, only bacteria appear to have fusedMGMT and a transcriptional activator.3,9,22 The third branchof organisms, archaea, along with eukaryotes, contain thesingle domain MGMT protein. This is consistent with theidea that certain genes from archaeal sources resemble moretheir eukaryotic than prokaryotic counterparts.3,23

MGMT is found in all branches of organisms, bacteria,archaea, and eukarya, emphasizing its importance in main-taining the integrity of the genome. Comparisons betweenthe various branches are instructive about the processesthat govern protein evolution and the different character-istics that determine reactivity. Here we report the first

solution structure of an MGMT protein from the ther-mophilic organism Methanococcus jannaschii. Comparisonsof the sequence, structure, and other characteristics of thisprotein, MJ1529, are made to those of other MGMT proteinswhose structures have been solved. The structure of MJ1529reveals a two-domain protein, highly similar in structureto hMGMT, AdaC, and MGMT from Pyrococcus kodakarensis(pMGMT) particularly in the domain known to be involvedin DNA-binding, the C-terminal domain. The structure alsoreveals some areas of disorder in both the N- and C-terminaldomains, the latter of which may have implications for howMGMT proteins recognize and accommodate lesions in DNAin the active site.

RESULTS AND DISCUSSION

Although sequence homology strongly indicates that MJ1529is an MGMT protein, the activity of the protein was verifiedin two ways. First, it is known that the addition of free O6MGbase to MGMT results in irreversible methylation of theprotein.13 Complete methylation of MJ1529 by this reactionwas verified by mass spectrometry. Additionally, MJ1529repaired O6MG within double-stranded oligonucleotides, asshown by a restriction endonuclease cleavage assay (seeSection on Experimental).24

Assignment and structure calculationsThe 1H–15N HSQC spectrum of MJ1529 is shown in Fig. 2.The peaks are labeled with the one letter amino-acid code andprotein sequence number. All backbone amide resonanceswere assigned except for I24, S141, and the last seven aminoacids of the protein, which are structurally disordered. Thereare extra resonances in the HSQC that could not be assigned,some of which likely represent proteolysis products. Someresidues gave weaker cross peaks; these arise from amideprotons in residues 60–65, 93–94, and 152–156, which areclose to each other in space and are somewhat less welldefined in the protein structure. This likely indicates someconformational flexibility in this region. The amide protonof E61 appears to shift in different spectra, and an FHSQCspectrum taken with high resolution and signal-to-noiseratio reveal three weak peaks in the spectral region of E61.The amide peak of Tyr 140, a fully conserved residue inthe C-terminal domain, is also weak and side-chain protonscould not be unambiguously assigned for this residue. Anexample of data from a three-dimensional spectrum acquiredfrom 15N labeled MJ1529 is shown in Fig. 3.

Copyright 2006 John Wiley & Sons, Ltd. Magn. Reson. Chem. 2006; 44: S71–S82

NMR Structure of an O6-MeG-DNA methyltransferase S73

Figure 2. 1H–15N HSQC spectrum of MJ1529. The peaks are labeled with the one letter amino-acid code and sequence number.Gray peaks are folded. Side-chain amides are connected by lines.

All of the restraints included in structure calculations andthe statistics of the lowest energy structures are shown inTable 1. In addition to NOE assignments, JHN�H˛ couplingsconstants, dihedral angle restraints (from TALOS), hydro-gen bonds, and residual dipolar couplings were used.25 NOEassignment was aided by the automatic assignment pro-grams CYANA and ARIA. Final calculations incorporatingresidual dipolar couplings were performed using CNS.26 – 28

An ensemble of the ten lowest energy structures forMJ1529 is shown in Fig. 4. The ensemble reveals that MJ1529is a two-domain protein with an ˛/ˇ fold. The N-terminaldomain consists of a three-stranded antiparallel ˇ sheet,ˇ1 � ˇ3, a largely disordered region with some ˛-helicalcharacteristics (helix a), and helix b. Overall, the N-terminusis somewhat less well defined than the C-terminus of theprotein. The turn between ˇ1 and ˇ2 lacks long-rangecontacts, and no assignments could be unambiguously madefor residue I24, near the end of ˇ3. The C-terminal domain

consists of helix c, helix d, helix e, helix f, helix g, ˇ5 and ˇ6.Helices e and f form a helix-turn-helix (hth) motif, a well-established DNA-binding motif. Between the hth regionand the terminal helix is a ‘wing’. The wing, or helix-turn-wing motif, is a structural element also implicated in DNAbinding.29,30

Structural comparisonsThe structure of MJ1529 can be compared with the structuresof other MGMT proteins that have been solved by X-raycrystallography; AdaC (PDB entry 1sfe), hMGMT (in theunmethylated, methylated, and benzylated forms) (1eh6),and pMGMT (from the hyperthermophile P. kodakarensis-1mgt).21,32 – 34 A structure-based sequence alignment revealsthat although there is very little sequence homology inthe N-terminal domains of the proteins (even for thethermostable MJ1529 and pMGMT, which are 35% identicalover 100 residues), all the structures reveal the three-stranded



Figure 3. Strip plot showing the NOE connections between residues 102–105 in an 15N resolved 3D NOESY. The arrows in blackrepresent HN–HN NOEs, typical of ˛-helical regions. The arrows in gray show HN–aliphatic NOEs.

Table 1. Statistics for structure calculation of MJ1529

Conformational restraintsTotal NOE restraints 1025Intra-residue 207Short (ji � jj D 1) 300Medium range (1 < ji � jj < 5) 298Long range (i � jj ½ 5) 250Hydrogen-bond restraints 1053JHN�H˛ coupling constants 101Dihedral angle restraints 114� 58 56Residual dipolar couplingsHN–N 71Residual violations –Root mean square deviation fromexperimental restraints

–

Distance violations (#/rmsd in A) 0/0.022Dihedral angle (°) 0.187Coupling constants (#/Hz) 8/0.565HN–N rdcs 1.4Root mean square deviation from idealizedgeometryBonds (A) 0.0018Angles (A) 0.35Impropers (A) 0.22Backbone RMSD

Table 1. (Continued)

Residues 2–155 1.34 AOrdered structure (3–25, 48–60, 65–130,145–155)

0.83 A

DNA-binding domain (80–155) 0.77 AProcheck statistics (ordered regions 3–25,48–155)Most favored 65.0Additionally allowed 31.6Disallowed 3.4

antiparallel ˇ sheet (Figs 5 and 6). In hMGMT, a zinc atombinds the side chains in ˇ1, ˇ3, and a loop, but no othermetal ions have been observed or their presence inferred inthe other structures.21 ˇ1 in MJ1529 is shorter than in theother structures and the sheet has a less pronounced roll.The region corresponding to the sheet and the disorderedregion before the beginning of helix b has approximately tenfewer residues than the corresponding region in pMGMT,reflecting the difference in their respective sequence lengths(167 vs 174). The shortest MGMT proteins identified to dateare also from thermostable organisms.35 A putative MGMTprotein from Thermatoga maritima, identified through a Blastsearch using the MJ1529 sequence, contains just 139 residues,with the alignment beginning at the start of helix b inMGMT.36



Figure 4. (a) Lowest energy ensemble overlay of MJ1529. The N-terminal domain contains a largely disordered region betweenresidues 29 and 39; (b) Stereoview of the ordered regions (residues 3–25, 48–155) of MJ1529; (c) 90° rotation of (b) (structures weregenerated with MOLMOL31).

Figure 5. Structure-based sequence alignment of MJ1529 from Methanococcus jannaschii, MGMT from Pyrococcus kodakarensis(KOD), Homo sapiens (hMGMT), and Escherichia coli (AdaC). Sheets are shown in rectangles, helices in rounded rectangles. Thedisordered region of MJ1529 is partially represented by a squiggly line. Similar residues are highlighted in bold, completelyconserved residues are noted with an asterisk. The active-site cysteine is shown in the black highlighted box.

N-terminusThe greatest topological differences in the structures occurafter the ˇ sheet. In both AdaC and pMGMT, the ˇ-sheetis followed by an ˛-helix. No density is observed in thatportion of hMGMT crystal structure.21,32 – 34 In the structureof MJ1529, the region corresponding to this helix is poorlydefined. Only a single amide proton, I31, is significantlyprotected from solvent exchange, compared to a muchhigher percentage of protection in other areas of secondary

structure. HN–HN NOE cross peaks indicative of an ˛-helixare observed for amides 29–35 as well as an NOE from amideof M35 to the H˛ of F32, also indicative of helix formation. Theintegrity of the corresponding helix in pMGMT is maintainedby two intrahelix ion pairs (salt links between side chains ofresidues i and i C 3/4) that span the rather long, 16-residuehelix, and are speculated to contribute to the thermostabilityof the protein.34 Comparison of the sequences of the twothermophiles reveals that MJ1529 has only one potential



(a)

(b)

(c)

Figure 6. (a) Overlay of the N-terminal domains of a minimized averaged structure of MJ1529 (yellow), hMGMT(red), pMGMT (blue),and AdaC (green). Helix a in MJ1529 is largely disordered. The zinc atom of hMGMT is represented by a red sphere. (b) Overlay ofthe ˇ-sheets, colored similarly to (a). The lengths of the strands vary, and the sheet has a pronounced roll (figures made withMOLSCRIPT and Raster3d).37,38 (c) Superposition of the C-terminal domains of an average MJ1529 structure (yellow), pMGMT(blue), AdaC(green) and hMGMT (red.) The arrow indicates the active site 310 helix and the location of the cysteine.

intrahelix ion pair (R27–E30), and this occurs in the transitionfrom ˇ3 to the disordered region. In addition, this regionin MGMT is both shorter and more hydrophobic in nature.There are two phenylalanine residues, F32 and F34, for whichonly υ-protons could be assigned owing to the particularlyhigh overlap of the phenylalanine aromatic protons. Theseresidues are near I24, for which no assignments could bemade, and it would seem likely these hydrophobic residueswould interact. For those residues whose heteronuclearT2 values could be measured in this region, R28, F34,D36, G37, D38, and V39, all had elevated values (>0.3 s)indicating this region may be dynamic. Following thisregion is helix b, which is oriented similarly in AdaC,MJ1529 and pMGMT, but lies at a slightly different anglein hMGMT.

It is not clear whether the N-terminal domain servesa biological function, but data indicate it contributes tostructural stability. This is consistent with chemical shiftmapping of DNA-binding to AdaC, which reveals thatfew amide resonances shift in the N-terminal domainduring DNA-binding.19 In addition, the lack of sequenceconservation or length in the N-terminal domain (despitesimilar structural elements) also indicates it exists as a capfor the DNA-binding domain.

C-terminusThe C-terminal domains of the MGMT proteins are highlysuperimposable. This domain consists of helices d–g, ˇ5

and ˇ6. Superposition of this region of a minimized averagestructure of MJ1529 with pMGMT, AdaC, and hMGMT,shows a C˛ rmsd of 1.8 A (Fig. 6(c)). The similarity ofthese domains correlates with the higher level of sequencehomology in this portion of the protein; all 19 of theabsolutely conserved residues occur in the C-terminaldomain (Fig. 5). The sequence PCHRV, which contains theactive-site cysteine, is the signature sequence that identifiespotential MGMT proteins; all proteins that have MGMTactivity contain this stretch of residues in their C-termini.Mutagenesis studies on these fully conserved residuesimplicate them in DNA-binding or active-site integrity.40,41

As in the other MGMT structures, the active-site cysteine inMJ1529 is located in a 310 helix, buried within the protein.It was first thought that MGMT underwent a substantialconformational change in order to place the cysteine in closeenough contact with O6 MG to repair it.33 A much moreelegant explanation, however, was derived from the studiesof other DNA-repair proteins, such as glycosylases andendonucleases. It was revealed through cocrystal structuresof these proteins and DNA that nucleotides could be ‘flipped-out’ of the double helix into the active site of the protein.42

This mode of action by MGMT was recently verified bycocrystal structures of mutant and wild-type hMGMT incomplex with O6-methylguanine containing duplex DNA,or DNA containing a modified base respectively.43,44

The C-terminal domain of MJ1529 includes an hth (helicese and f), which contains fully conserved residues Y99



(helix e) and R111, A112, and V113 (helix f) (Fig. 7(a)). Inboth MJ1529 and the crystal structures, R111 at the base ofhelix f (the recognition helix) is exposed and is in a positionto interact with negatively charged DNA. In the cocrystalstructures, this arginine is inserted into the DNA basestack, replacing the space left by the extra-helical guaninebase and stabilizing the base-flipped conformation (akin toother DNA-repair enzymes that use amino acids to promotebase-flipping.45 – 47) The other small hydrophobic residues inhelix f, A112 and G114, are proposed to make sequence-independent contacts with DNA, which is illustrated withthe cocrystal structures. Interestingly, the recognition helixof MGMT binds deeply in the minor groove of DNA; in allother hth-containing proteins whose mode of DNA-bindinghas been elucidated, the recognition helix binds in the majorgroove.

A model of MJ1529 bound to DNA based on the cocrystalstructures reveals that the majority of the interactionsbetween the protein and the phosphodiester backbone arelikely mediated by small residues, explaining the lack ofsequence specificity for repair. In addition to the hydrophobicresidues in the recognition helix, N107, T108, and S109in the turn of the hth are poised to interact with theDNA backbone, as is T98 at the start of helix e (Fig. 7(a),(b)). Residues within the wing are also likely to interactwith the DNA. A surface potential plot shows the chargecomplementarity of the recognition helix and wing withDNA (Fig. 7(c)). Interaction of this region of the protein withDNA is also consistent with chemical shift mapping dataon AdaC, in which significant shifts of amide peaks upontitration with DNA occurred in helix e and helix f, and thewing residues.19,34

(a)

(b) (d)

(c)

Figure 7. (a) Ribbon diagram of the C-terminal domain of MJ1529 with conserved residues labeled with an asterisk, which arespeculated to be involved in DNA or form part of the active-site (generated with Pymol). (b) DNA-binding domain of MJ1529modeled with double-stranded DNA based on the hMGMT/DNA complex. The break in the strand marks where the base flips intothe active-site.39,47 The wing residues are likely to shift to accommodate the base. Conserved residues are marked with an asterisk.(c) Charge potential surface plot of MJ1529, positive charge is in blue, negative in red, same orientation as in (a). (d) Ribbon diagramof MJ1529. The helix-turn-helix is show in green, the wing in magenta. Conserved residues C128, H129, R130 and E154, composingpart of the active-site network, are also shown.



The space between the wing and the recognition helixcreates a pocket for the flipped-out base (Fig. 7(a)). In MJ1529and the crystal structures, conserved Y99 of helix e sits atthe base of the pocket with its hydroxyl poised to form ahydrogen bond with the N3 of guanine. The back of thehydrophobic pocket is lined with wing residues V131 andV132, and L117 from helix f, whose corresponding residues inthe cocrystal structures interact with the flipped-out base. Inthe crystal structures, the residues corresponding to invari-ant Y140 and S141 in MJ1529 stack against and form an amidehydrogen bond to the flipped base, respectively. The areaaround these residues in MJ1529 is not well defined, andrepresents the greatest difference in the backbone C-terminirmsd between MJ1529 and the crystal structures. In the crys-tal structures, Y140 is pointing in toward the hydrophobicpocket. In MJ1529, the average structure has Y140 pointingout of the pocket. It is not clear whether this representsthe true conformation of Y140; the amide cross peak of thisresidue was weak in the HSQC, and no side-chain atomscould be positively assigned. Correspondingly, none of theresonances for S141 could be assigned. While this limitedthe NOE assignments in this area, these features are alsoconsistent with the flexibility in this region. Support for flex-ibility in the wing is also present in the crystal structures.In pMGMT, the residues in the wing have higher B factorsthan other structured areas of the protein.34 In addition, com-parison of an hMGMT/DNA complex with a modified baseflipped into the active site and an hMGMT/DNA complexwith a thymine bound partially in the active site shows thatthe greatest differences between these protein structures isin the position of the last helix and wing, which the authorsattribute to an inherent flexibility that allows for binding awide range of lesions.44

The active site of MJ1529 is constructed similarly to thoseof the other structures. The side chain of invariant H129 isin proximity to invariant E154 whose side chain interactswith that of invariant R130 (Fig. 7(d)). It is proposed thatthis hydrogen-bonding network results in deprotonationof C128, generating the nucleophile. One interaction thatcould not be detected (not likely because of its absence butbecause of data limitations) is a hydrogen bond betweenthe side chains of invariant N120 and C128. It is thisinteraction that is broken by alkylation, and results in asubtle structural change that destabilizes the protein, andin the case of hMGMT, signals the protein for ubiquitin-dependent degradation.46

The overall structure of MJ1529 is very similar to thestructures of pMGMT, hMGMT, and AdaC. The largestdifferences occur in the N-terminal regions of the proteins,in which the secondary structural elements of MJ1529 areshorter than in the other protein structures, and residues29–39 are much less structured than in pMGMT and AdaC.It is interesting to note, however, that the correspondingregion in hMGMT is not present in the crystal structure,implying some flexibility in this region. Other possiblydynamic regions include residues 60–65, which have weakamide resonances, and do not superimpose well in thestructure.

The C-terminal domains are all highly superimpos-able, as expected from the level of sequence conserva-tion. In all of the structures, the active-site cysteine isburied within the protein, and invariant residues fromthe hth and wing are exposed and poised to interactwith the DNA. The greatest difference in the C-terminibetween MJ1529 and the crystal structures is at theC-terminal tip of the wing, which is not well defined inthe NMR structure and is consistent with flexibility in thisregion.

The two thermophilic proteins share many characteristicsat the sequence level. These include helix capping by prolineresidues, the pattern of i, i C 3/4 charged side-chain residuesthat are shown to form intra and interhelix ion pairsin the KOD structure, and a short deletion in the loopbetween helices e and f, all of which may contribute to thethermostability of the proteins.34

EXPERIMENTAL

Protein expression, purification, and NMR samplepreparationA clone of MJ1529 in pet21a (Novagen) was kindly providedby Hisao Yokota, and was transformed into BL21(DE3)with plasmid pACYC (Stratagene) expressing lac repressor.Transformants were grown in 2 ml cultures of LB containingampicillin and kanamyacin and used to inoculate 1 l ofLB, 1 l of M9 with 15NH4Cl (15N labeled), or 1 l ofM9 with 15NH4Cl and 13C glucose (15N/13C labeled). Onemilligram of thiamine, and trace metals were added to M9growths. Cells were grown to an o.d. of 0.6–0.8, whereupon50 mg/l of IPTG was added to induce expression. Cellswere allowed to grow for 5–6 h more and harvested bycentrifugation.

Harvested cells were resuspended in 50 mM Tris, pH7.4, 50 mM NaCl, 3 mM DTT and 200 µM PMSF, then weresonicated on ice, and the insoluble matter was spun downby high-speed centrifugation. The supernatant was placed ina 60–70 °C water bath for approximately 1/2 h to precipitatemost E. coli proteins, then the solution was again centrifugedand the supernatant was applied to an SP Sepharose column,and washed extensively with 50 mM Tris, 200 mM NaCl,until the detector absorbance returned to baseline levels. Asalt gradient from 200 to 650 mM NaCl (300 ml total) wasrun, with MJ1529 eluting at approximately 450 mM NaCl.Typical yields were ¾10 mg/l for LB and ¾8 mg/l for M9preparations. The molecular weight of the purified proteinwas verified by mass spectrometry.

After the purification, MJ1529 was dialyzed into 50 mM

sodium phosphate buffer, pH 6.2, 500 mM NaCl, 1 mM DTT,and 0.02% NaN3 and concentrated to 0.4–0.8 mM in Amiconcentrifugal filters.

For the residual dipolar coupling experiments, the pro-tein was dialyzed into a low salt buffer (25 mM sodiumphosphate, pH 6.7, 50 mM NaCl or 200 mM NaCl) beforebeing diluted with bicelle stocks.48 The resulting proteinsolutions were 150–250 µM, but transferring assignmentsfrom different salt and different pH solutions was notdifficult.



MGMT activity assayO6 methylguanine was purchased from Sigma and incubatedwith MJ1529 overnight at 37 °C. After reverse-phase HPLC,the protein was analyzed by mass spectrometry revealingthat the protein had become fully methylated. During thereaction with the free O6MG base, it was noted that proteinprecipitate formed, which agrees with the findings thatMGMT proteins are destabilized by methylation.49

To verify MGMT activity, a single O6-methylguaninelesion was incorporated within a double-stranded pieceof DNA containing multiple restriction endonuclease sites.The O6-methylguanine containing strand was fluorescentlylabeled with 6-hexachlorofluorescein (HEX). The lesion wasincorporated in the middle of a PvuII site (CAGŁCTG), whichcannot cleave if the central guanine is methylated. Thus,when the protein is incubated with the modified oligo thelesion is repaired, and reaction with PvuII results in thecleavage of the oligo, which can be quantified by running theDNA on a gel and measuring the fluorescence.24

The oligos were ordered (purified by HPLC) from Synthe-gen (Texas) and designed as follows, with the G* represent-ing O6-methylguanine: Hex-GCCCGGCCAGŁ CTGCAGCGwith CGCTGCAGCTGGCCGGGC. The dried oligos weredissolved separately in 10 mM Tris, pH 8, 33 mM NaCl, and1 mM EDTA. The oligos contained a PvuII site and an HaeIIIsite used as an internal control. The fluorescently labeledoligo and an excess of the complementary oligo were com-bined and heated at 90 °C for 2 to 3 min and then allowed toanneal at room temperature overnight. The subsequent solu-tion was aliquoted and frozen at �20 °C for further use. 10–20pmols of the annealed oligo were incubated either without orwith various concentrations of purified MJ1529 in 150 µl totalvolume of buffer containing 50 mM Tris, pH 7.4, 5% glycerol,and 1 mM DTT. The reactions were incubated from 2 to 12 hat 37 °C, and for 2 h at 55 °C. The DNA was extracted by vor-texing with an equal volume of phenol (50): chloroform (49):isoamyl alcohol (1) and keeping the aqueous layer (repeatedtwice), followed by chloroform (99): isoamyl alcohol(1) extraction. To the retained fraction 150 µl of 0.5 M NaOAcpH 5 was added, followed by 25–30 µl of 1 mg/ml tRNA, tohelp precipitation of the DNA. To this mixture, 0.3 ml of cold100% ethanol was added and the mixture was allowed to sitovernight at �20 °C. The following day the ethanol mixturewas spun in a microcentrifuge at 13 000 rpm for 4 min andthe ethanol was poured off. The pellet was dried in a speedvacuum and then redissolved in 20 µl buffer appropriate forincubation with PvuII or HaeIII. The restriction endonucleasereactions were allowed to proceed for 1–2 h, stopped with10 µl formamide (containing 20 mM EDTA), and heated at95 °C for 5 min to separate the DNA strands, and then loadedon a pre-run 20% acrylamide gel containing 7 M Urea in TBEbuffer. The gels were imaged with a Typhoon Fluoroimagerusing excitation of the HEX oligo at 535 nm and emissionat 556 nm. The assay results reveal that MJ1529 does repairO6-methylguanine within double-stranded oligonucleotides.

NMR experiments and assignmentsNMR experiments were performed on Bruker AMX-600,DRX-500, or DRX-500 equipped with a cryoprobe. Exper-iments were performed at 30 °C, unless otherwise noted.

Inter-residue connectivity was established by analysis ofHNCA,50 CBCA(CO)NH,51 15N edited NOESY-HSQC,52

HBHA(CO)NH,53 and 15N edited TOCSY-HSQC54 spectrawith Watergate for solvent suppression.52 Intra-residue con-nections were identified using 15N edited NOESY-HSQC, 15Nedited TOCSY-HSQC, and 13C edited HCCH TOCSY55 (D2O)spectra. Carbonyl carbons were assigned with an HNCOexperiment. 3JNH˛ coupling constants were determined bypeak intensity analysis of an HNHA experiment56 and con-verted to coupling restraints in the program NMRVIEW.57

Hydrogen bonds were inferred from analysis of secondarystructure indicated by HN–HN connections in the 15N editedNOESY-HSQC, and D2O- exchange experiments, in whichthe protein solution was lyophilized and redissolved in100% D2O and monitored by successive Fast 1H–15N HSQC(FHSQC) experiments for the next several hours.58 Peaks thathad a long lifetime were inferred to be involved in hydrogenbonds, and when consistent with other secondary struc-tural indications, were used as restraints during structurecalculations.

Aromatic resonances were assigned using a 1H–1HNOESY on labeled protein, 1H–1H TOCSY, 15N editedNOESY-HSQC, 4D HMQC-NOE (D2O), and aromatic1H–13C HSQC spectra (CT-HSQC, HMQC, etc.).59,60 Becauseof the high degeneracy in the chemical shifts of the pheny-lalanine aromatic protons and carbons, a 13C–15N labeled,12C, 14N-labeled phenylalanine sample was made by intro-ducing unlabeled phenylalanine to a minimal media growth(known as ‘reverse isotopic labeling’). Using this strategy,only tyrosine and histidine signals are found in the aro-matic region of spectra relying on 13C-labeling. Analogously,12C-filtered experiments will yield signals only from pheny-lalanine, without the line broadening associated from protonattachment to isotopically labeled carbons.60 2D CT-HSQC,2D-HMQC, and 3D 13C NOESY spectra were run, placingthe carrier at 125 ppm to maximize the aromatic contribu-tions to the spectra. At least one υ-proton resonance for eachphenylalanine was assigned. A 1D of the aromatic regionverified the extent of the overlap and also some broad peaks,indicating the presence of some dynamic regions of the pro-tein, consistent with the structure presented here. Two of thephenylalanines, F32 and F34, are in a largely unstructuredportion of the protein, as are the aromatic residues afterresidue 160. Stereospecific assignments of methyl groupswere generated by making a 10% labeled 13C sample andrunning a CT-HSQC experiment.61,62

NMR data were processed on Silicon Graphics Indigo2workstations using Felix 97 or NMRPipe,63 and visualizedusing the program NMRVIEW. The chemical shift index-ing in NMRVIEW was helpful for identifying secondarystructure.64 Peak lists for all spectra were generated eitherwith the peak picking algorithm incorporated in NMRVIEWor with the program PIPP.65

Structure calculationsFor the four-dimensional NOESY data from the 13C/13CHMQC-NOE66 and 15N/13C (HMQC/HSQC) NOE67, somepeaks were assigned manually. An in-house program listingpossible assignments according to chemical shifts was used



to generate possible assignments for each peak. Initialstructures were generated (3–5 A rmsd) and were usedto distance-filter probable assignments (possibilities wereeliminated if they were >8 A apart in the structures). Thisapproach led to some assignments, but the low quality of theinitial structures made this process difficult; thus, distance-filtering was also done through examination of homologousstructures. Using sequence alignment and comparison ofregions of the secondary structure, peak filtering for MJ1529was done using the structure from P. kodakarensis, the mosthomologous protein.34

As described above, initial restraints were generatedfrom manual assignment of 3D 15N-NOESY and partialassignment of 4D 13C/13C and 15N/13C NOESYs. Theserestraints were incorporated into structure calculation usingthe program DYANA.68 In addition, 3JHN�H˛ couplingconstants providing information on � dihedral angles wereused as restraints. The program TALOS was used togenerate a list of likely � and dihedral angles on thebasis of comparison of sequence triplets to a databasecontaining structure and sequence information from theRCSB.25 One hundred structures were generated, and the20 with lowest energies were analyzed for consistent NOEdistance violations. This yielded structures with an overallrmsd of ¾2.5A.

To aid NOE assignment the programs ARIA andCYANA, both of which incorporate ambiguous NOEs intostructure determination as part of an iterative assignmentprocess, were used.26 – 28 Although in principle each programmay be used de novo, i.e. without any initial structures orrestraints, it was found that providing initial structures frommanual assignments (including NOEs, coupling constants,and hydrogen bonds) greatly improved the success ofthe assignment process for both programs. After theruns, unambiguous, ambiguous, and rejected peak listswere inspected for accuracy. Although many assignmentsappeared to be correct, some accepted peaks were incorrectlyassigned, while some ambiguous assignments that wererejected on the basis of stringent criteria were reincorporatedin the structure calculation process.

Results from ARIA yielded a highly convergedC-terminus (for residues 70–140, the overall rmsd was0.9 A). However, almost no structure was detected at allby the program in the N-terminal part of the protein, andthe C-terminal helix was distorted because of improperlydetermined contacts with the N-terminus. The programCYANA, with similar initial inputs as ARIA, yielded a clearlydiscernible three-stranded ˇ-sheet in the N-terminal part ofthe protein and resulted in an overall rmsd for residues 3–25and 48–155 of 1.5 A.

Through manual and automated assignment by ARIAand CYANA, approximately 1000 NOE distance restraints(binned either automatically by the above programsor through the program NMRVIEW) were determined.Although there is some inherent ‘unstructuredness’ to theprotein, particularly in the N-terminus, residual dipolar cou-pling (rdc) data was acquired to provide more long-rangeorientational restraints.

The final residual dipolar coupling data used in struc-ture calculations of MJ1529 were derived from a sam-ple containing 25 mM NaPhos, 200 mM NaCl, and 5%DMPC : DHPC : CTAB (30 : 10 : 1 molar ratio,69 AVANTIPolar Lipids). After temperature-induced alignment, 2DIPAP-HSQC spectra were acquired at 312 K.70 Isotropicspectra were acquired at 303 K. This data resulted in 71HN–N rdc’s being incorporated into structure calculations.A modified HNCO experiment was used to measure C˛ – C’couplings, and a SEMI-CT-HSQC was acquired to extractHN–C’, and N–C’ couplings simultaneously.71 Althoughthe addition of CTAB enhanced the stability of the sampleover undoped bicelles, deterioration of the sample over timewas noticeable.

For determining the magnitude and rhombicity of thealignment tensor, a histogram of the measured dipolarcouplings was plotted, yielding an approximate Da andR by analysis of the distribution.72 This last method reliesupon a uniform distribution of couplings (and hence bondvectors) and the accuracy improves as more types of rdcdata are added. Histograms of the data yielded by study ofMJ1529 did not resemble a powder pattern. The maximumvalues of Da were different in different experiments, meaningthat there was either nonuniformity in the distribution ofcouplings, or that the alignment tensor was changing overtime because of inherent instability of the sample. Thus,rough estimates of Da and R were refined according tothe procedure of Clore, et al.73 Residual dipolar couplingswere incorporated into the program CNS as restraints withthe alignment tensor being represented as a pseudo four-atom molecule. Given the difficulty in data acquisitionand analysis, only data from one type of bond vectorswere used (HN–N) and the force constant was set toa relatively low value (0.3 kcal/Hz2). The final Da andR values used were �7.0 Hz and 0.30, respectively. Forthe final structures, the lowest energy structures from atotal of 100 calculated without residual dipolar couplingswere refined with residual dipolar couplings (50 totalstructures calculated), with the ten lowest energy takenas the final structures. The annealing protocol included2000 slow cooling torsion angle dynamics steps, and 6000additional Cartesian dynamics steps. The final structureswere deposited to the PDB, accession code 2g7h.

CONCLUSIONS

Protein/DNA interfaces are proposed to be somewhatflexible.19,74,75 This would seem to be particularly impor-tant for DNA-repair proteins, such as MGMT, that have littlesequence specificity, little preference for double over single-stranded DNA, and can accommodate different sizes andtypes of base modifications. In the case of MGMT, this flex-ibility is likely to come from the wing portion of the protein.The broadening of some amide residues within the wing mayindicate a conformational flexibility that allows the wing toadjust or move to allow the proposed extra-helical base to beexposed to C128 and other conserved residues in the activesite. This is more likely than flexibility within the hth, whichis tightly coupled and constrained by secondary structureorder.



One of the most intriguing questions that remainabout MGMT is how it locates damaged bases amongscores of unmodified bases. Studies of O6MG-containingoligonucleotides indicate that the lesion is destabilizing (asjudged by melting temperature), but otherwise causes onlyslight deformations in the backbone of DNA.76,77 While it ispossible that these modified bases are occasionally alreadyextra-helical when MGMT comes along, other proposedmethods of action by MGMT include processive movementalong DNA by cooperative interactions between proteins,and targeting of MGMT to active replication sites.20,43,77

In the cocrystal structures of hMGMT with DNA, MGMTbinding introduces a slight bend in the DNA that mayexploit already weak interactions of modified bases in theduplex.43,44 Further clarification of the repair mechanism willbe addressed by other biophysical studies.

AcknowledgementsThis work was supported by the Office of Biological and Environ-mental Research, Office of Science of the Department of Energy undercontract DE-AC03-76SF00098 to the Lawrence Berkeley NationalLaboratory. Funding for equipment was provided by the NationalScience Foundation (DMB 8609035 and BBS 8720134).

REFERENCES

1. Vassylyev DG, Morikawa K. Curr. Opin. Struct. Biol. 1997; 7: 103.2. Verdine GL, Bruner SD. Chem. Biol. 1997; 4: 329.3. Eisen JA, Hanawalt PC. Mutat. Res. 1999; 435: 171.4. Bignami M, O’Driscoll M, Aquilina G, Karran P. Mutat. Res.

2000; 462: 71.5. Kleibl K. Mutat. Res. 2002; 512: 67.6. Lindahl T, Sedgwick B, Sekiguchi M, Nakabeppu Y. Annu. Rev.

Biochem. 1988; 57: 133.7. Demple B. In Protein Methylation, Paik WS, Kim S (eds). CRC

Press: Boca Raton, Fl, 1990; 285.8. Lips J, Kaina B. Mutat. Res. 2001; 487: 59.9. Pegg AE. Mutat. Res. 2000; 462: 83.

10. Takahashi M, Sakumi K, Sekiguchi M. Biochemistry 1990; 29:3431.

11. Chan CL, Wu Z, Ciardelli T, Eastman A, Bresnick E. Arch.Biochem. Biophys. 1993; 300: 193.

12. Spratt TE, Wu JD, Levy DE, Kanugula S, Pegg AE. Biochemistry1999; 38: 6801.

13. Goodtzova K, Crone TM, Pegg AE. Biochemistry 1994; 33: 8385.14. Bender K, Federwisch M, Loggen U, Nehls P, Rajewsky MF.

Nucleic Acids Res. 1996; 24: 2087.15. Fried MG, Kanugula S, Bromberg JL, Pegg AE. Biochemistry

1996; 35: 15295.16. Liem L-K, Wong C-W, Lim A, Li BFL. J. Mol. Biol. 1993; 231: 950.17. Dolan EM, Oplinger M, Pegg AE. Carcinogenesis 1988; 9: 2139.18. Delaney JC, Essigmann JM. Biochemistry 2001; 40: 14968.19. Verdemato PE, Brannigan JA, Damblon C, Zuccotto V, Moody

PCE, Lu-yun L. Nucleic Acids Res. 2000; 28: 3710.20. Rasimas JJ, Pegg AE, Fried MG. J. Biol. Chem. 2003; 278: 7973.21. Daniels DS, Mol CD, Arvai AS, Kanugula S, Pegg AE, Tainer JA.

EMBO J. 2000; 19: 1719.22. Grombacher T, Eichhorn U, Kaina B. Oncogene 1998; 17: 845.23. Bult CJ, White O, Olsen GJ, Zhou L, Fleischmann RD,

Sutton GG, Blake JA, FitzGerald LM, Clayton RA, Gocayne JD,Kerlavage AR, Dougherty BA, Tomb JF, Adams PD, Reich CI,Overbeek R, Kirkness EF, Weinstock KG, Merrick JM, Glodek A,Scott JL, Geoghagen NS, Venter JC. Science 1996; 273: 1058.

24. Kreklau EL, Limp-Foster M, Liu N, Xu Y, Kelley MR,Erickson LC. Nucleic Acids Res. 2001; 29: 2558.

25. Cornilescu G, Delaglio F, Bax A. J. Biomol. NMR 1999; 13: 289.26. Brunger AT, Adams PD. Acta Crystallogr., Sect. D 1998; 54: 905.

27. Herrmann T, Guntert P, Wuthrich K. J. Mol. Biol. 2002; 319: 209.28. Linge JP, O’Donoghue SI, Nilges M. Methods Enzymol. 2001; 339:

71.29. Wintjens R, Rooman M. J. Mol. Biol. 1996; 262: 294.30. Vora RA, Pegg AE, Ealick SE. Proteins: Struct., Funct., Genet. 1998;

32: 3.31. Koradi R, Billeter M, Wuthrich K. J. Mol. Graph. 1996; 14: 51.32. Wibley JEA, Pegg AE, Moody PCE. Nucleic Acids Res. 2000; 28:

393.33. Moore MH, Gulbis JM, Dodson EJ, Demple B, Moody PCE.

EMBO J. 1994; 13: 1495.34. Hashimoto H, Inoue T, Nishioka M, Fujiwara S, Takagi M,

Imanaka T, Kai Y. J. Mol. Biol. 1999; 292: 707.35. Skorvaga M, Raven NDH, Margison GP. Proc. Natl. Acad. Sci.

U.S.A. 1998; 95: 6711.36. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. J. Mol.

Biol. 1990; 215: 403.37. Kraulis EA. J. Appl. Crystallogr. 1991; D50: 869.38. Merritt EA, Bacon DJ. Methods Enzymol. 1997; 277: 505.39. DeLano WL. The PyMOL Molecular Graphics System. DeLano

Scientific: San Carlos, CA, 2002.40. Pieper RO, Morgan SE, Kelley MR. Carcinogenesis 1994; 15: 1895.41. Goodtzova K, Kanugula S, Edara S, Pegg AE. Biochemistry 1998;

37: 12489.42. Roberts RJ, Cheng X. Annu. Rev. Biochem. 1998; 67: 181.43. Daniels DS, Woo TT, Luu KX, Noll DM, Clarke ND, Pegg AE,

Tainer JA. Nat. Struct. Mol. Biol. 2004; 11: 714.44. Duguid EM, Rice PA, He C. J. Mol. Biol. 2005; 350: 657.45. Daniels DS, Tainer JA. Mutat. Res. 2000; 460: 151.46. Lau A, Scharer OD, Samxon L, Verdine GL, Ellenberger T. Cell

1998; 95: 249.47. Slupphaug G, Mol CD, Kavli B, Arvai AS, Krokan HE,

Tainer JA. Nature 1996; 384: 5105.48. Ottiger M, Bax A. J. Biomol. NMR 1998; 12: 361.49. Kanugula S, Goodtzova K, Pegg AE. Biochem. J. 1998; 329: 545.50. Grzesiek S, Bax A. J. Magn. Reson. 1992; 99: 201.51. Grzesiek S, Bax A. J. Am. Chem. Soc. 1992; 114: 6291.52. Sklenar V, Piotto M, Leppik R, Saudek V. J. Magn. Reson., Ser. A

1993; 102: 241.53. Grzesiek S, Bax A. J. Biomol. NMR 1993; 3: 185.54. Cavanagh J, Rance M. J. Magn. Reson. 1992; 96: 670.55. Kay LE, Xu G-Y, Singer AU, Muhandiram DR, Forman-Kay JD.

J. Magn. Reson., Ser. B 1993; 101: 333.56. Kuboniwa H, Grzesiek S, Delaglio F, Bax A. J. Biomol. NMR 1994;

4: 871.57. Johnson BA, Blevins RA. J. Biomol. NMR 1994; 4: 603.58. Mori S, Abeygunawardana C, Johnson M, Vanzijl PCM. J. Magn.

Reson., Ser. B 1995; 108: 94.59. Santoro J, King GC. J. Magn. Reson. 1992; 97: 202.60. Vuister GW, Kim S-J, Wu C, Bax A. J. Am. Chem. Soc. 1994; 116:

9206.61. Neri D, Szyperski T, Otting G, Senn H, Wuthrich K. Biochemistry

1989; 28: 7510.62. Johnson PE, Joshi MD, Tomme P, Kilburn DG, McIntosh LP.

Biochemistry 1996; 35: 14381.63. Delaglio F, Grzesiek S, Vuister GW, Zhu G, Pfeifer J, Bax A.

J. Biomol. NMR 1995; 6: 277.64. Garrett DS, Powers R, Gronenborn AM, Clore GM. J. Magn.

Reson. 1991; 95: 214.65. Kuszewski J, Qin J, Gronenborn AM, Clore GM. J. Magn. Reson.,

Ser. B 1995; 106: 92.66. Vuister GW, Clore GM, Gronenborn AM, Powers R, Garrett DS,

Tschudin R, Bax A. J. Magn. Reson., Ser. B 1993; 101: 210.67. Kay LE, Clore GM, Bax A, Gronenborn AM. Science 1990; 249:

411.68. Guntert P, Mumenthaler C, Wuthrich K. J. Mol. Biol. 1997; 273:

283.69. Losonczi JA, Prestegard JH. J. Biomol. NMR 1998; 12: 447.70. Ottiger M, Bax A. J. Magn. Reson. 1998; 131: 373.71. Ottiger M, Bax A. J. Am. Chem. Soc. 1998; 120: 12334.



72. Clore GM, Gronenborn AM, Bax A. J. Magn. Reson. 1998; 133:216.

73. Clore GM, Gronenborn AM, Tjandra N. J. Magn. Reson. 1998;131: 159.

74. Slijper M, Boelnes R, Davis AL, Konings RNH, van derMarel GA, Van Boom JH, Kaptein R. Biochemistry 1997; 36: 249.

75. Patel D, Shapiro L, Kozlowski SA, Gaffney BL, Jones RA.Biochemistry 1986; 25: 1027.

76. Gaffney BL, Jones RA. Biochemistry 1989; 28: 5881.77. Ali RB, Teo AK-C, Oh H-K, Chuang LS-H, Ayi T-C, Li BFL. Mol.

Cell. Biol. 1998; 18: 1660.


Documents

Structural studies of MJ1529, an O6-methylguanine-DNA methyltransferase