Mario F. Fraga and Manel EstellerSpanish National CancerResearch Center (CNIO),Madrid, SpainABSTRACTEver since methylcytosine was found ingenomic DNA, this epigenetic alteration hasbecome a center of scientific attraction, es-peciallybecauseofitsrelationtogenesi-lencing in disease. There is currently a widerange of methods designed to yield quantita-tive and qualitative information on genomicDNA methylation. The earliest approacheswereconcentratedonthestudyofoveralllevels of methylcytosine, but more recent ef-forts have focused on the study of the methy-lationstatusofspecificDNAsequences.Particularly,optimizationofthemethodsbased on bisulfite modification of DNA per-mits the analysis of limited CpGs in restric-tionenzymesites(e.g.,combinedbisulfiterestriction analyses and methylation-sensi-tive single nucleotide primer extension) andthe overall characterization based on differ-ential methylation states (e.g., methylation-specific PCR, MethyLight, and methylation-sensitivesingle-strandedconformationalpolymorphism) and allows very specific pat-terns of methylation to be revealed (bisulfiteDNA sequencing). In addition, novel meth-ods designed to search for new methylcyto-sinehotspotshaveyieldedfurtherdatawithoutrequiringpriorknowledgeoftheDNA sequence. We hope this review will beavaluabletoolinselectingthebesttech-niques to address particular questions con-cerningthecytosinemethylationstatusofgenomic DNA.INTRODUCTIONMethylation of cytosines in the 5 po-sition of the pyrimidinic ring is the mostimportantepigeneticalterationineu-karyotes. In animals, methylcytosine ismainlyfoundincytosine-guanine(CpG)dinucleotides(10),whereasinplantsitismorefrequentlylocatedincytosine-anybase-guanine(CpNpG)trinucleotidesequences.Thepresenceof 5-methylcytosine in the promoter ofspecific genes alters the binding of tran-scriptional factors and other proteins toDNA in plants (26) and animals (18). Italso attracts methyl-DNA-binding pro-teins and histone deacetylases that mod-ify chromatin structure around the genetranscriptionstartsite.Bothmecha-nismsblocktranscriptionandcausegene silencing (10). Thus, methylationof C residues in genomic DNA plays akeyroleintheregulationofgeneex-pression (99).DNAmethylationresearchcanbeapproachedfromseveralstandpointsbecause there is a wide range of tech-niques available for the study of the oc-currenceandlocalizationofmethyl-cytosineinthegenome(45,82,83).Eachtechniquehasitsownpeculiari-ties, which implies that there is a suit-abletechniqueforeachspecificprob-lem.Inanycase,DNAmethylationstatuscanonlybeestablishedbythejointuseofvariousmethodologies.Here we classify the available methodsfor studying the degree of DNA methy-lation, with respect to the type of infor-mation produced (Figure 1), includingthedegreeofwholegenomicDNAmethylation, the DNA methylation sta-tus of a specific sequence, and how tofind new methylation hot spots.Levelsofmethylcytosineoccur-rence in the genomic DNA can be mea-suredbyhigh-performanceseparationtechniquesorbyenzymatic/chemicalmeans. The latter are never as sensitiveas the former, and sometimes their res-olutionisrestrictedtoendonucleasecleavage sites. Despite the drawbacks,enzymatic/chemicalapproachesarestillcommonlyusedbecause,unlikeseparationtechniques,theydonotre-quireexpensiveandcomplexequip-ment that is not always available. Whenseparation devices are available, high-performancecapillaryelectrophoresis(HPCE) may be the best choice since itisfaster,cheaper,andmoresensitivethan HPLC (33). Although in situ hy-bridization methods for studying cyto-sine methylation status sometimes giveaccurate measures of the degree of totalDNA methylation, one of the most in-teresting aspects of these approaches isthat they provide information on tissue-specificmethylationpatterns.Bymeansoflabeledanti-methylcytosineantibodies,DNAmethylationcanbemonitored in metaphase chromosomes,hetero/euchromatin,and,mostimpor-tantly, on a cell-by-cell basis within thesamesample.Thelatteralternative,whichgenerallyyieldsqualitativere-sults,isofgreatinterestincancerre-searchbecauseitcanrevealmethyla-tiondifferencesbetweennormalandtumor tissues in the same sample.Two alternatives are currently usedtostudythedegreeofDNAmethyla-tion in particular DNA sequences: non-bisulfiteandbisulfitemethods.Thefirstreliesontheuseofmethylation-sensitiverestrictionendonucleasescombinedwithSouthernblotanalysisorPCRdetection,whichsometimesmeans that results are limited to cleav-age sites. This problem can be avoidedbythebisulfitemodificationoftheDNA, which comprises a wide range ofReview632 BioTechniques Vol. 33, No. 3 (2002)DNA Methylation: A Profile of Methods and ApplicationsBioTechniques 33:632-649 (September 2002)techniquesthatallowthequantitativeandaccuratedeterminationofthemethylation status of the allele, even atthe cell population level. All bisulfite-associated methods require PCR ampli-fication of the bisulfite-modified DNA,and differences in methylcytosine pat-terns are displayed by methylation-de-pendentprimerdesign(e.g.,methyla-tion-specific PCR), in conjunction withmethylation-sensitiverestrictionen-donucleases[e.g.,combinedbisulfiterestrictionanalyses(COBRA)],ge-nomic sequencing, and other approach-es. Some of these approaches even pro-vide quantitative data about the averageproportion of methylated and unmethy-lated alleles in a population.Thus,becauseofthelargeassort-ment of methods available for evaluat-ing the degree of genomic DNA methy-lation and the wide range of data typestheyyield,thegoalofthispaperistohelpresearcherstochoosethebestmethodforaddressingspecificDNAmethylationquestions.Todothis,wecategorizedthemostimportantmeth-ods currently used by the type of infor-mation they provide. We also describeall the outstanding approaches and pro-videmoredetailedinformationaboutthe techniques we consider most signif-icant, discussing their advantages, dis-advantages, and possible artifacts.QUANTIFICATION OF OVERALL METHYLCYTOSINEIN GENOMIC DNAGlobalDNAmethylationhasbeenproposedasamolecularmarkerforavariety of biological processes such ascancer(32,76)andontogeneticdevel-opment in both plants (34) and animals(55).MethylcytosineconcentrationingenomicDNAiscurrentlyquantifiedby high-performance separation means(HPCEandHPLC)orbyenzymatic/chemical approaches.HPLC-Based MethodsRelative methylcytosine contents ofgenomicDNAcanbeanalyzedbychemical hydrolysis to obtain the totalbase composition of the genome and thesubsequent fractionation and quantifica-tion of hydrolysis products using HPLCtechnologies.DNAhydrolysiscanbecarriedoutbyincubationwithformicacid at high temperature (27). However,Catania et al. (14) suggested the use ofhydrofluoric acid for chemical hydroly-sisofDNAtopreventdeaminationofcytosine and methylcytosine, which of-ten occurs with formic acid.In any case, enzymatic hydrolysis ofthe DNA is reported to be a better alter-nativeforquantifyingthedegreeofDNAmethylation(58).The2-deoxy-mononucleosidesareobtainedusingNuclease P1, DNase I, and snake ven-omphosphodiesterase(25)ormicro-coccal nuclease and phosphodiesteraseII (61), followed by hydrolysis with al-kaline phosphatase. Resulting deoxyri-bonucleosides are subsequently separat-edbyHPLC,andthemethylcytosinelevels are quantified by comparing therelativeabsorbanceofcytosineandmethylcytosine at 254 nm in the samplewith external standards of known bases.The degree of DNA methylation of sev-eral animal (38) and plant tissues (98)has been quantified by this method, butatleast2.5gDNAaregenerallyre-quiredtoquantify5-methylcytosinewith a low standard deviation for repli-catesamples.Moreover,thismethodsuffersfromoccasionalproblemswhereby elution buffers precipitate intothe column (unpublished observation).The sensitivity of the system can beincreasedwithmassspectrometryde-tection, which has a detection limit 106timesthelimitofabsorptionspec-troscopydetectors.Annanetal.(3)used HPLC-mass spectrometry coupledvia a moving-belt interface to analyzeelectrophoreticderivativesofmethyl-cytosineand5-hydroxymethyluracilnucleobases. As little as 9.9 pg (signal-to-noise ratio, 5) and 180 fg (signal-to-noise ratio, 10) of the respective nucle-obasesweredetectedinthenegativeelectron-capturechemicalionizationmode,andlinearresponseswereob-served over a moderate dynamic range.Alternatively, Gowher et al. (44) haveReview634 BioTechniques Vol. 33, No. 3 (2002)Figure 1. Main methods available for the study of genomic DNA methylation status. MSP, methyl-ation-specificPCR;MS-SnuPE,methylation-sensitivesingle-nucleotideprimerextension;MS-SSCP,methylation-sensitive single-stranded conformational polymorphism; MS-REs, methylation-sensitive re-strictionendonucleases;LM-PCR,ligation-mediatedPCR;MS-ISH,methylation-specificinsituhy-bridization; IPEM, incomplete primer extension mixture; CPBS, competitive primer binding sites; SP-PE,solid-phase primer extension; and ERMA, enzymatic regional methylation assay.developed a highly specific and sensi-tiveassayfordetectingtraceamountsofmethylcytosineingenomicDNA.TheydegradedDNAtonucleosides,purified methylcytosine by HPLC, and,for detection by 1-D and 2-D thin-layerchromatography(TLC),radiolabeledusingdeoxynucleosidekinaseand[32P]ATP.Usingthisassay,theyshowedthatmethylcytosineoccursintheDNAofDrosophilamelanogasteratalevelof1in10002000cytosineresiduesinadultflies.Thislabelingstrategy has already been used to quan-tify methylcytosine concentration (61),but,inthiscase,itwasemployedinconjunctionwith1-Dhigh-perfor-mance TLC using alkylamino-modifiedsilica plates.HPCE-Based MethodsThedevelopmentofCEtechniqueshas given rise to an approach to researchthat has several advantages over currentmethodologies used to quantify the ex-tent of DNA methylation. From its be-ginnings, CE has proved to be extremelyhelpful in separating various DNA com-ponents,includinganumberofbaseadducts(71),andrecentlyFragaetal.(33) have reported the quantification oftherelativemethylcytosinecontentofthegenomicDNAofplantsusingaHPCEsystemtoanalyzeacid-hy-drolyzed genomic DNA. In this context,the separation and quantification of cyto-sine and methylcytosine are only possi-ble by the use of an SDS micelle system.This method is faster than HPLC (takingless than 10 min/sample), reasonably in-expensivebecauseitdoesnotrequirecontinuous running buffers, and displaysa great potential for fractionation (theo-retically up to 106plates). Nevertheless,almost no preparative analyses are possi-ble with HPCE systems because of thelow injection volumes involved.ThereleaseofbasesfromDNAbychemical means involves the productionof a complicated mixture of moleculesthat makes the detection and quantifica-tionofmethylcytosinedifficultwhenDNA is not sufficiently purified and/orconcentrated.ThisproblemcanbeavoidedbyenzymatichydrolysiswithNuclease P1 and alkaline phosphatase toproduce2-deoxymononucleosides,whicharethenfractionatedusingamodification of the previously describedHPCE method (33). Approximately onemethylcytosine in 200 cytosine residuescan be detected by this method using 1ggenomicDNA(Figure2).Toin-crease sensitivity, laser-induced fluores-cence and mass spectrometry detectorsshouldbeused.However,therearenoreferencesdescribinghowtoquantifymethylcytosine. Nevertheless, capillaryelectrophoreticseparationandonlinedetectionbymonitoringlaser-inducedfluorescence have already been appliedsuccessfully to quantify other modifiedbasesinDNA(100).Usinglaser-in-duced fluorescence monitoring systems,aconcentrationdetectionlimitofap-proximately3 10-10M(91)or1adduct/107normal nucleotides/g DNA(85) can be achieved.Analyses of Genome-WideMethylation by Chemical orEnzymatic MeansAs stated earlier, quantifying the de-gree of DNA methylation by HPLC orHPCErequiresaccesstosophisticatedequipment that is not always available.TheradioactivelabelingofCpGsitesusingthemethyl-acceptorassay(101)has been developed to address this prob-lem,but,amongthetechniquesotherdrawbacks,itcanonlymonitorCpGmethylationchanges,thusCpNpGmethylationcannotbedetected.Thismethod uses bacterial SssI DNA methyl-transferasetotransfertritium-labeledmethyl groups from S-adenosylmethio-nine (SAM) to unmethylated cytosinesReviewVol. 33, No. 3 (2002)Figure2.ResolutionofnucleosidesobtainedfromenzymatichydrolysisofgenomicDNAfrom human cancer cell line HT29 by HPCE.mC, 5-methylcytidine; A, adenosine; C, cytidine;G, guanosine; and T, thymidine.in CpG targets. The data obtained froma scintillation counter are used to calcu-late the number of methyl groups incor-poratedintheDNA.Theamountoftritium incorporated is inversely propor-tionaltothelevelofCpGmethylation(23). This method has revealed changesin the DNA methylation patterns of pa-tients who suffer from a number of dis-eases, including cancer (40).Recently, a modification of the SssIassay was reported to allow the quan-tification of the degree of DNA methy-lationinspecificDNAregions(37).This quantitative approach, termed en-zymaticregionalmethylationassay(ERMA),provideshigh-qualityinfor-mationaboutthemethylationdensityof a sequence. However, as in the previ-ously described case of the SssI assay,the acquisition of data is limited to spe-cific endonuclease cleavage sites. Themethodmergesbisulfitemodificationof DNA and PCR amplification of theregion of interest with SssI and 3H-la-beledSAMincubationusingdammethyltransferaseand 14C-labeledSAM as an internal control [primers forPCR are designed to contain two damsites (CATG)]. Under this combinationof conditions, the relative 3H/14C signalratioislinearlycorrelatedwiththemethylation density of the region of in-terest. In general, methods based on theacceptanceofmethylgroupsinvitrocan give inaccurate but reproducible re-sults.Ifthesemethodsareused,thenthey should be validated on a few sam-ples with a direct quantitative analysisfor global 5-methylcytosine levels.Combining the sodium bisulfite re-actionwiththeabilityofchloroac-etaldehyde to label DNA fluorescently,Oakeley et al. (74) described a methodfor analyzing DNA methylation in anysequence by fluorescent labeling. TheytreatedtheDNAwithchloroacetalde-hyde, which reacts quantitatively withcytosine and adenine bases to yield thefluorescent adducts ethenocytosine andethenoadenine, respectively, while ura-cil, thymine, and guanine do not under-go this reaction. Following DNA depu-rination,theextentoffluorescenceasquantifiedwithaluminescencespectrometerisproportionaltotheamountofmethylcytosineinthegenome.Thismethodhasseveralad-vantagesoverthemethyl-acceptoras-say:achemical,ratherthananenzy-maticreaction,usedquitestableandinexpensive reagents, and its measure-ments are not restricted to the methyl-cytosine located in GpC targets.Diaz-Sala et al. (22) described otherapproachesinplantsamples.Theywereabletomeasuretherelativede-gree of DNA methylation accurately bythefluorescencequantificationoftheonline slab of DNA digestion productsthatwereobtainedwithmethylation-sensitive endonucleases.In Situ Hybridization Methods forStudying Total Cytosine MethylationGlobal DNA methylation can also bequantifiedbymethylcytosine-specificantibodies(67).Theaccessibilityofmethylcytosine to specific antibodies indsDNAwasfirstreportedinthemid-1980s (2). An outstanding advantage ofthis approach is that it may be carried outonacell-by-cellbasisratherthaninaheterogeneous population. Every in situhybridization-basedtechniqueforthestudy of DNA methylation must be pre-viously validated with a control for theaccessibility of the antibody to the DNA.Classical immunoassay detection ap-proachesinvolvequantifyingtheretention of radiolabeled DNA by poly-clonalantibodiesonnitrocellulosefil-ters, immunoprecipitation, gel filtration,andvisualizationunderelectronmi-croscopy (2). Cytosine methylation canalso be detected in metaphase chromo-somes(4,70)andinchromatinusingmonoclonalantibodies(MAbs)com-binedwithfluorescencestaining(68).The incorporation of 5-bromodeoxyuri-dine in DNA with a thermal denatura-tionstepmayincreasethebindingefficiencyofanti-methylcytosineanti-bodies on metaphase chromosomes andincrease the efficiency of antibody bind-ing, as revealed by immunofluorescencestaining (70). As a result of differentialimmunofluorescencestaining,thismethodallowsthesemi-quantitativeanalysis of the antibody binding sites inmetaphase chromosomes. Furthermore,in situ hybridization with fluorescencedetectionmakesitpossibletodistin-guishdifferentmethylcytosine-richDNA sites (4). For example, secondaryconstrictions, juxtacentromeric regions,andT-bandsemitstrongfluorescence,whiletheshortarmsofacrocentricsproduce strong polymorphic signals.Quantitative in situ DNA methylationimmunofluorescenceanalysescanbeaccomplishedusinganimageanalyzerandacharge-coupleddevicecamera(97).Forthispurpose,sampleresultsmust be compared with results obtainedfrom cells with known methylation lev-els. The study of global DNA methyla-tionofhumanfemaleXchromosomesusingthismethodrevealedglobalhy-pomethylation of the late-replicating X.In addition, there is a method that allowsnotonlythemappingofstructural(banding) but also functional (methyla-tion status) features of different chromo-somedomains(6).ThisstrategyusesMAbsonmildlydenaturedchromo-somes.Analternativetofluorescencedetectionistoconnectacoloreden-zyme-dependent reaction. For instance,peroxidase-taggedsecondantibodies,with4-chloro-1-naphtholusedassub-strate with anti-5-methylcytosine MAbs,have been used to quantify the effects ofthedemethylatingagent5-azacytidineReviewontheconstitutiveheterochromatinofhuman chromosomes (21).When comparing the results of glob-alDNAmethylationevaluatedbytheradiolabeledmethylincorporationas-say (an alternative method for the quan-tificationofglobalDNAmethylationlevels) (19) with immunohistochemicalstainingofthesametissuesectionswithMAbsagainstmethylcytosine,similar contents are generally observed.Inaddition,insituhybridizationcanalso reveal differences in the methyla-tion status of cancer and normal adja-cent tissues by analyzing the same tis-sue section.Because of the high sensitivity andspecificityrequired,antibodydesignand production must be carefully con-sidered,eventhoughthesetopicsde-pend on the detection method. One ofthehighestsensitivities(inthepmolrange)canbeachievedbytheFMC9MAbinconjunctionwithanELISA(69).ThisMAbdisplaysexcellentspecificitybecauseitreactswithmethylcytidine and methylcytosine butnot with other nucleosides. When suffi-cientspecificityisachievedandnocross-reactionsaredetected,insituDNA methylation immunofluorescenceanalyses can be used to monitor signifi-cant growth processes in animals [e.g.,embryodevelopment(11)]andplants[e.g., pollen nuclei maturation (73)].NON-BISULFITE METHODS FORMAPPING METHYLCYTOSINESIN SPECIFIC DNA SEQUENCESThemostwidelyusedmethodsforstudying DNA methylation patterns ofspecificregionsofDNAwithnobasemodificationsarebasedontheuseofmethylation-sensitiveandinsensitiverestrictionendonucleases(15).Oneoftherestrictionenzymesoftheiso-schizomer pair is able to cut the DNAonlywhenitstargetisunmethylated,whereastheotherisnotsensitivetomethylatedcytosines.Themostcom-mon isoschizomers used are the HpaII/MspIpair.Ingeneral,bothcleavetheDNA at the CCGG target, but HpaII isnot able to cut when the second cytosineismethylated(CmCGG).Moreover,itwas reported that cleavage by MspI ofGGCCGGsequencesisalsoinhibitedby the methylation of the C next to theCpG (12). EcoRII/BstNI isoschizomersaremorefrequentlyusedinplants,whereas most methylcytosine is locatedat CCNGG sequences (where N is anybase other than G) (46). EcoRII recog-nizesCCNGGtargets,butitonlycleaves when unmethylated (9), where-as BstNI is insensitive to cytosine meth-ylation.Althoughthesepairsofen-zymescancleavehemimethylatedDNA, they do not distinguish betweencytosinesmethylatedatdifferentposi-tions in the pyrimidinic ring (13). How-ever,thereareseveralrestrictionen-zymes that recognize the localization ofthe methyl group (64).OnceDNAhasbeendigestedwithmethylation-sensitiveendonucleases,identification of the methylation statusofagenecanbeaccomplishedbySouthernblothybridizationorPCRprocedures (Figure 3). In the first case,digestion products are separated by gelelectrophoresis,transferredtoanitro-cellulosefilter,andhybridizedwitharadiolabeledprobe(63).WhenDNAdigestion is accomplished by methyla-tion-sensitivenucleases,obtainingaDNAfragmentthatislargerthanex-pected indicates methylation at one orbothrestrictiontargetsthatflankthehomologous region of the DNA (Figure3). An methylcytosine-positive will bedetected when at least 10% of the DNApresentthatmodificationbecauseoftheir hybridization sensitivity.When the amount of tissue is limit-ing,detectionofcytosinemethylationcanbeachievedbyPCR,whichre-quireslessthan10ngDNA,whereasSouthernblothybridizationgenerallyrequiresupto10gDNA.Further-more, a methylcytosine-positive can bedetected when only 0.1% of the DNAmolecules present such base modifica-tion(60).PCRprimersmustcorre-spond to flanking sequences of the re-striction targets so that the absence ofmethylation is revealed in the presenceof a PCR band (Figure 3) (87). For quantitative analyses, Heiskanenet al. (50) described a solid-phase quan-titativeprimerextensionmethodthatcan be adapted to a microtitration wellformat, thereby allowing the analysis ofa large series of samples. In particular,the method produces linearity in a rangefrom2%to100%ofmalignantblastsdiluted with normal leukocytes. Quanti-tative PCR can also be performed by amodification(88)ofthemethodde-scribed by Singer-Sam et al. (87). Theimprovement,termedcompetitiveprimer binding sites, includes an inter-nalstandardsharingprimer-bindingsites with the genomic template so thatdemethylationresultsinadecreaseofthe PCR products. Nevertheless, a com-parison of the samples entails the accu-ratequantificationoftheamountofDNA,whichiscommonlyadifficulttask when such small amounts of DNAarehandled.Thereisanothermethodthatconsistsofmeasuringtheconver-sion of a large amplified DNA fragmentto a shorter DNA product that correlateswith the amount of demethylation (65).Theprocedureusespairsofnon-isoschizomericenzymestocleavege-nomic DNA at closely spaced sites. Theextent of cleavage by the methylation-sensitiverestrictionenzymeisquanti-fiedbytheamplificationofthediges-tionproductswithligation-mediatedPCR and radioactive labeling. The ratioofthetwoamplifiedfragmentscorre-lates with the degree of methylation atthe restriction site.PCR methods involving prior diges-tion with CpG methylation-sensitive re-striction endonucleases can be compli-cated by the inability to achieve 100%digestionofthesensitivesites,evenReview640 BioTechniques Vol. 33, No. 3 (2002)Figure 4. Methylation-specific PCR and bisulfate sequencing. (A) Methylation status of the p16INK4aCpG island in human primary breast tumors determined by methylation-specific PCR. PBR322/Msp di-gest (New England Biolabs, Beverly, MA, USA) is shown at left as the molecular weight marker. The pres-ence of a visible PCR product in the lanes marked U indicates the presence of unmethylated genes, and thepresence of PCR product in the lanes marked M indicates the presence of methylated genes. In vitro methy-lated DNA was used as a positive control for methylation, normal lymphocytes were used as a negativecontrol for methylation, and water was used as a negative PCR control. The breast tumor BR2 is hyperme-thylated at P16, while the remaining tumors (BR1, BR3, BR4, BR5, BR6, and BR7) are unmethylated. (B)Graphical representation of the sodium bisulfite genomic sequencing of the CpG island of the APC pro-moter. Each row represents an individual cloned and sequenced allele after sodium bisulfite DNA modifi-cation. CpG sites are shown as circles and drawn to accurately reflect CpG density of the region. Whitedots are unmethylated CpGs, and black dots are methylated CpGs. IVD, in vitro methylated DNA.Figure 3. Anticipated results when studying site-specific DNA methylation by Southern blot analy-sis and PCR (two of the most important non-bisulfite-related methods) in the cases of methylationand unmethylation for a single CpG target. Black circles indicate cytosine methylation. Arrows sym-bolize PCR primers. The discontinuous lines illustrate probes for Southern blot analysis.with multiple rounds of digestion usinghighconcentrationsofenzyme.Al-though PCR can amplify even a singletargetDNAmoleculewhensufficientrounds of amplification are employed,misleading results can be obtained withsuch methods if adequate controls arenot used. This is why cleavage with arestriction isoschizomer of the methy-lation-sensitive endonuclease is a com-monly used control (95). Even thoughpartiallycleavage-resistantsiteshavebeencommonlyinterpretedasbeingpartially methylated, the importance oftheintrinsicresistancetocleavageofcertain DNA targets should be stressed(59). This resistance can be overcomebytheadditionofsite-containingoligonucleotide duplexes that drive thecleavage of the resistant sites.Non-bisulfite methods for the quan-tification of DNA methylation patternsare simple, rapid, and can be used forany known-sequence genomic DNA re-gion. These methods are extremely spe-cific, but their limitation to specific re-striction sites reduces their value.BISULFITE METHODS FORMAPPING METHYLCYTOSINESIN SPECIFIC DNA SEQUENCESAlthough isoschizomer-based meth-ods for the study of methycytosine oc-currenceatspecificDNAregionsdis-playhighsensitivity,theycannotprovidethecriticalinformationre-quired for a complete understanding oftheroleofmethylcytosineincellandmolecularbiology.Theseaspectscanbe addressed with the bisulfite modifi-cation of the DNA (17).There is an extensive range of meth-ods for the quantification of the methy-lation status of cytosines located in spe-cific DNA regions based on the sodiumbisulfite treatment (Figure 1). Bisulfiteconvertsunmethylatedcytosinetouracil, while methylated cytosine doesnot react (36). This reaction constitutesthebasisfordiscriminatingbetweenmethylatedandunmethylatedDNA.Bisulfite transformation of DNA can befollowed by several methods, includingsequencing, methylation-specific PCR,combinedbisulfiterestrictionassays,and others. In general, each modifica-tion leads to greater sensitivity and res-olutionorcontributestothedevelop-ment of a novel perspective on the re-sults.AllofthemworkwithPCR,which, among other advantages, is suit-ableforanalyzingparaffin-embeddedtissues and poorly purified DNA.BisulfitemodificationofDNAre-quires prior DNA denaturation becauseonly methylcytosines that are located insingle strands are susceptible to attack(84).ThepartialdenaturationoftheDNA is a common event that causes oneof the potential artifacts of the method.Renaturation induced by high-salt con-centrations could be an additional criti-cal problem, although bisulfite is gener-ally able to react with the DNA beforerenaturation occurs. Nevertheless, falsepositives caused by the lack of the sin-gle-strandedconformationhavebeenreported (81). Incomplete desulfonationafterbisulfitetreatmentmayalsogiverise to several problems (93). The pres-ence of residual bisulfite can avert theadequate alkalization of the solution sothat,ifthepHislowerthannine,therateofpyrimidinedesulfonationisratherslow,therebyrenderingcertainDNApolymerasesincapableofrepli-cating the template. However, the prob-lem may be avoided by measuring thepH after the alkalization step.It was originally believed that all thecytosines in densely methylated regionsinthechromosomereplicationoriginswere methylated (92), but methylationatnon-CpGresidueswaslaterdiscov-ered to be an artifact of partial bisulfiteconversion. Harrison et al. (48) attempt-edtoexplainthisobservationbysug-gesting that a methylated cytosine couldprotecttheneighboringunmethylatedoutercytosinefromcompletebisulfiteconversion and that this is a sequence-specific event. They also suggested thatthe problem could be avoided by usinganinternalstandardthatconsistsofchecking the complete bisulfite conver-Vol. 33, No. 3 (2002) BioTechniques 641sion of a non-CpG site from an in vitrosynthetic methylated oligonucleotide.As previously stated, the total con-version of cytosines to uracils is criticaltothesuccessoftheanalyses.Themaximum conversion rates of cytosineoccur at 55C (418 h) and 95C (1 h).Special attention must be paid to the in-fluence of the temperature and incuba-tion time on DNA integrity during thebisulfite treatment because, under theseconditions,84%96%oftheDNAisdegraded (47). Raizis et al. (80) also re-portedtheoccurrenceofpartialDNAdegradation during bisulfite treatment,but in their case, it was due to low pH-dependent depurination. A simple solu-tiontolimitthedegradationoftheDNAtemplateisprovidedbytheau-thors,whosuggestreducingthetimerequired to complete the bisulfite reac-tion by means of increasing the bisul-fite concentration to 5 M. Under theseconditions,optimalPCRproductionoccurs after only 4 h, thereby minimiz-ing DNA fragmentation.SequencingSequencing bisulfite-altered DNA isthe most straightforward means of de-tecting cytosine methylation. In gener-al,afterthedenaturationandbisulfitemodification,dsDNAisobtainedbyprimer extension, and the fragment ofinterestisamplifiedbyPCR(17).Methylcytosine can then be detected bystandard DNA sequencing of the PCRproducts(Figure4B).CloningPCRproducts into plasmid vectors followedby the sequencing of individual clonesis an alternative method that, althoughslower,couldprovidemethylationmaps of single DNA molecules, insteadof the average values of the methylationstatusinthepopulationofmoleculesprovidedbydirectsequencingofthePCR products. This approach has beenhelpful in the study of the DNA methy-lation of genes associated with cancersuch as APC (30) and Rb (89).Consistent quantification of cytosinemethylationbydirectsequencinghasbeenwidelyreportedtoentailthemeans of fluorescence-based detection(78).PaulandClark(75)describedamethod of automated genomic sequenc-ingwithfluorescencedetection(GENESCAN; GeneScan Australia Pty.Ltd., Bundoora, Australia) that providesa measure of the degree of methylationat a particular target by comparing cyto-sine and thymine signals that have beenlabeled with the same fluorescent dyes.This method has been used to examinethe detailed methylation state of manyCpG island-containing genes in cancer(66).Moreover,Radlinskaetal.(79)described an alternative method that al-lows the direct localization of methyl-cytosines in the primary product of thebisulfitetreatmentinsteadofthePCRamplification product. To achieve this,theyusedaprimerextensionmixturecontaining only three deoxynucleotides(dATP,dCTP,anddTTP)andlackingdGTP (IPEM) that produced an elonga-tion stop at methylcytosine points. Thepositions of the methylcytosine are rec-ognized by comparing the localizationofrunoffproductstoproductsofse-quencing reactions performed on bisul-fite-untreatedtemplateDNA.Themethod,whichgavenofalse-positiveresults at all, has a sensitivity of 47 ng(50 fmol of the methylated sites).Normally, four termination reactionsweresetuptodisplaymethylcytosineoccurrencebythesequencingmethod,butonlyoneterminationreactionisneeded when the sequence of the targetregionisknown(103).Thechoiceoftermination reaction depends on the typeofbaseconversionforwhichthese-quencing primers are designed, and theresults are obtained by comparing prod-ucts of the single termination reaction tothose of normal sequencing terminationsfrom the control unmodified DNA.Methylation-Specific PCRMethylation-specificPCRisthemost widely used technique for study-ingthemethylationofCpGislands,which are common in the promoter re-gions of many genes. Cytosines in CpGislandsareusuallyunmethylatedinnormaltissues,whereastheybecomemethylated in the promoter sequencesof genes associated with certain abnor-malcellularprocessessuchascancer(31). Methylation-specific PCR (51) isone of the most effective choices for in-vestigatingthemethylationprofileofReview642 BioTechniques Vol. 33, No. 3 (2002)Figure 5. Methylation results for methylated, unmethylated, and mixtures of methylated/unmethy-latedallelesobtainedbyseveralbisulfite-relatedmethods. MSP,methylation-specificPCR;Ms-SnuPE, methylation-sensitive single nucleotide primer extension; Ms-SSCP, methylation-sensitive sin-gle-stranded conformational polymorphism; MS-REs, methylation-sensitive restriction endonucleases;U, unmethylated reaction; M, methylated reaction; C, cytosine band; and T, thymine band. Black circlesindicate cytosine methylation.these regions (28,31,32).The differences between methylatedand unmethylated alleles that arise fromsodium bisulfite treatment are the basisof methylation-specific PCR (Figure 5)and are especially valuable in CpG is-lands because of the abundance of CpGsites.Primerdesignisacriticalandcomplexcomponentoftheprocedure(Figure5).Bisulfite-convertedDNAstrandsarenolongercomplementary,soprimerdesignmustbecustomizedfor each DNA chain. Methylation pat-ternsofallsequencesmustbedeter-minedinseparatereactions(Figure4A). To optimize the PCR amplificationstep, the following critical requirementsmust be considered when designing theprimers:(i)theannealingtemperatureof both primers must be similar and al-ways between 55C and 65C; (ii) thePCR product should be between 80 and175 bp; (iii) each primer should containatleasttwoCpGpairs;(iv)thesenseprimer should contain a CpG pair at its3-end; and (v), to avoid false positives(amplificationofunmethylated,un-modified DNA), primers should containnon-CpG cytosines.The great sensitivity of the methodallows the methylation status of smallsamples of DNA, including those fromparaffin-embeddedormicrodissectedtissues (51). However, resolution at thenucleotidelevelrequiresthesequenc-ing of the PCR products, and quantita-tivedataandidentificationofcellularheterogeneityinpopulationsarestillnot possible with this technique. If thePCR involves too many cycles of am-plification without ensuring that the re-actionisinthelinearresponserangewith respect to the template concentra-tion, then large overestimates of the ex-tentofmethylationcanbeobtainedifthesequenceisamplifiablewithboththemethylation-specificprimersandtheprimersfortheunmethylatedse-quence.Inaddition,theappropriatecontrolDNAsequencesshouldhavebeen shown first to give no PCR prod-uctwiththemethylation-specificpri-mers.Theconclusionaboutmethyla-tioncouldrefertoonlyasmallpercentage of the examined cells in thepopulationunlessthePCRisdonesemi-quantitatively,varyingthetem-plateconcentrationorcyclenumber,andtheefficiencyofthemethylation-specific primers and the primers for theunmethylated sequences has been com-pared.Non-quantitative,methylation-specificPCRresultscanbevalidatedwith a quantitative method on some ofthe samples (e.g., Southern blot analy-siswithaCpGmethylation-sensitiverestrictionendonuclease).Withprevi-ouslyunusedprimers,unmethylatedandinvitromethylatedcontrolse-quences should be tested.To avoid such problems, Nuovo et al.(72) made a remarkable advance in themethod by combining methylation-spe-cificPCRwithinsituhybridization.The modification allows for the methy-lation status of specific DNA sequencesto be visualized in individual cells. Thispowerfulapproachismainlyusedtomonitor methylation patterns in sampletissueswithcomplexmixturesoftu-mor-likeandnormalcells.Anothermethodthatwasrecentlydescribedcombinesmethylation-specificPCRanddenaturingHPLC(DHPLC),withwhichevensmallcellmosaicismsofstructurallynormaland/orabnormalchromosomes can be detected (5). Fol-lowingPCRamplification,theallelescanberesolvedfromthetwopopula-tions of PCR products by DHPLC be-causetheydifferatseveralpositionswithin the amplified sequence. Distinctclonotypicepigenotypescanalsobeisolatedandcharacterizedbydenatur-inggradientgelelectrophoresis(DGGE) (1), which allows the detectionof almost any sequence change in everyDNAregion,includingdifferentiallymethylated sequences.Anotherquantitativeapproachiscalled MethyLight (24), which uses flu-orescence-based,real-timePCRtech-nology and does not involve additionalPCRsteps.TheDNAismodifiedbythe bisulfite treatment and amplified byfluorescence-based, real-time quantita-tivePCRusinglocus-specificPCRprimersthatflankanoligonucleotideprobewitha5 fluorescencereporterdye and a 3 quencher dye. The reporteris enzymatically released during the re-action, and fluorescence, which is pro-portional to the amount of PCR productand thus to the degree of DNA methy-lation,canbesequentiallydetectedinan automated nucleotide sequencer de-vice. Fluorescence detection greatly in-creasesthesensitivityofthemethod,Vol. 33, No. 3 (2002) BioTechniques 643makingitpossibletodetectasinglemethylated allele in 105unmethylatedalleles. However, the major advantageis that sequence discrimination can beachievedatanylevelofprobehy-bridizationorPCRamplificationasaresultoftheparticularprimerdesign.The quantitative nature of the assay isbased on real-time PCR and the inclu-sionofamethylatedreferenceDNA.Thisassayallowsthecomparisonofnormal tissue samples with low levelsof methylation at the test sequences andtumors with significantly more methy-lation. It also allows the quantitative as-sessmentofDNAmethylationatspe-cificsequenceswithhighthroughput,permittingtherapidanalysisofmanyDNA samples at many different sites. Becauseoftheversatilityofthemethod, methylation-specific PCR hasbeenwidelyproposedasarapidandcost-effective clinical tool to use in theinitial evaluation of a wide range of pa-tientswithspecificallele-dependentdiseases.Forinstance,methylation-specificPCRiscurrentlyappliedtoevaluatethemethylationstatusoftheCpGislandoftheSNRPNgeneand,hence,fortherapiddiagnosisofthePrader-Willi and Angelman syndromes(57).Thisapproachalsoprovidesanaccurate assay for the determination ofX inactivation and can be carried out onDNA samples that are unsuitable for re-strictiondigestion.Incidentally,theseare the most widely used means to eval-uatethemethylationstatusinactiveandinactiveXchromosomes.Uchidaet al. (96) described a slight modifica-tion of this method, which they calledhuman androgen-receptor gene-methy-lation-specific PCR (HUMARA-MSP).ThisinvolvestheanalysisoftheXchromosome inactivation pattern by anoptimizedmethylation-specificPCRmethodthatcanidentifytheaverageproportionsofmethylatedandun-methylatedallelesthroughtheexami-nation of a polymorphic short tandemrepeat near the 5 promoter region.Methylation-specificPCRhasalsobeensuccessfullyusedtoevaluatetheresponsivenessofhumancancerpa-tientstoalkylatingagents.Byusingmethylation-specific PCR, Esteller et al.(29)foundhypermethylationoftheDNA-repair enzyme MGMT to be a keyfactorintheresistancetoalkylatingagents. This involves tumor regressionbecauseoftheincreaseoftherespon-siveness to alkylating agents caused bythe lack of MGMT. Methylation-specif-ic PCR has also been useful in the de-tection of the presence of tumoral DNAin the serum of cancer patients (41).COBRACOBRA(102)constitutesahighlyspecific approach releasing on the cre-ation or detection of a target for restric-tion endonuclease after bisulfite treat-ment.Themajoradvantageofthismethod is that it provides semi-quanti-tativedataregardingthemethylationstatusatspecificregionsinanyDNAsample (Figure 5).DNA amplification products of spe-cific loci, previously modified by bisul-fite,aredigestedwithrestrictionen-zymes that distinguish methylated fromunmethylated sequences so that the de-gree of DNA methylation is linearly cor-related with the relative amounts of di-gestedandundigestedproducts.TheBstUI case may illustrate this point; itscleavagesite(CGCG)isresistanttobisulfite modification when it is methy-lated but is transformed to TGTG whenit is unmethylated. Thus, DNA cleavageafterbisulfitetreatmentonlyoccursiftherestrictiontargetismethylatedandthe cleavage products are proportional tothedegreeofmethylationoftheana-lyzedsequence.Theaveragerelativeproportions of digestion products can bequantified by hybridization with 5-end-labeledoligonucleotidesandphospho-rimager detection. In contrast to methy-lation-specificPCR,thePCRprimersshouldnotcontainCpGpairstoavoiddiscrimination between different meth-ylatedtemplates.Ingeneral,thisap-proachcanbeusedwhenabsolutepercentages of methylated and unmeth-ylated alleles are required to guaranteethe final diagnosis. The major drawbackis that BstUI will also cut if unconverted;therefore,theuseofthisenzymecanleadtotheoverestimationofmethyla-tion.Monitoringconversionstatewithenzymes such as HpaII is needed (54).Although it is specific, quantitative,and sensitive, this approach entails com-plete bisulfite modification and cannotbe used for all DNA sequences becauseit is confined to restriction targets.Methylation-Sensitive SingleNucleotide Primer ExtensionMethylation-sensitivesinglenu-cleotideprimerextensionemploysbisulfite-PCRcombinedwithsinglenucleotide primer extension to analyzeDNA methylation status quantitativelyat a particular DNA region without us-ing restriction enzymes (Figure 5). Thisallows the analysis of almost all DNAregions(42).Inthisapproach,thein-corporation of C instead of T results inmethylationintheoriginalDNAse-quence (Figure 5).Following the bisulfite modificationof the whole genomic DNA, the regionofinterestisamplifiedbyPCRtech-niques,usingprimersspecificallyde-signedforthenewbisulfite-modifiedsequenceofDNA.PCR-generatedproducts are also amplified using inter-nal primers that terminate immediately5 of the single nucleotide to be assayed.Therelativeaverageamountofallelemethylation is quantified by monitoringtheincorporationofboth[32P]dCTPsand [32P]dTTPs using a phosphorimag-ing device. Under these conditions, theintensities of bands in the C tracks areproportionaltotheaverageamountofmethylationateachtestedCpGsite.Meanwhile, the intensity of the T tracksis linearly correlated with the percent-age of unmethylated cytosines.Similartootherquantitativebisul-fite-PCRmethods,methylation-sensi-tivesinglenucleotideextensionhasabroad range of clinical applications be-cause of its high sensitivity, specificity,andrequirementforonlysmallamounts of sample. Despite all of theseadvantages,PCRbiasandanalysesinCpG-rich regions can be a problem be-causeofthedifficultyofsuccessfullydesigningprimerslackingCpGdinu-clotides. This is a key point if one con-sidersthepriorlackofknowledgeoftheir methylation status (82).Methylation-Sensitive Single-StrandConformation Analysis Single-stranded conformational poly-morphism (SSCP) analyses were report-ed several years ago in the broad contextof detecting single-base polymorphismsasameansofresolvingPCRproductsthatdifferedbynomorethanafewReview644 BioTechniques Vol. 33, No. 3 (2002)nucleotides (77). Recently, Bianco et al.(8) adapted this technique to the study ofmethylation changes in specific DNA re-gions. They proposed combining bisul-fite modification of the DNA with SSCPin a new method that they called methy-lation-sensitive single-strand conforma-tion analysis (MS-SCCA). However, thesame method has also been called bisul-fite-PCR-SSCP (62).Bisulfite modification of DNA gen-erates accurate sequence disparities be-tween methylated and unmethylated al-leles, which can be resolved by SSCP(Figure5).MethylcytosinedetectionusingconventionalSSCPrequiresthePCRamplificationoftheDNAfrag-mentsofinterest,denaturationofthedouble-strandedproduct,followedbynon-denaturing slab gel electrophoresisanddetectionviasilverstainingbystandard methods.MS-SSCA also yields semi-quantita-tivedata.Inparticular,whenoneana-lyzesmixturesofmethylatedandun-methylated DNA of a known ratio, therelative intensities of the bands correlatewith the relative degree of methylationof the original sequence. However, de-spite the great sensitivity [e.g., success-fulmethylationanalysisonmicrodis-sectedparaffin-embeddedtissues(7)]and rapidity of the method, the resolu-tion for detecting single-base changes inaDNAfragmentusuallyrequiresatleast 10% of the population to show thealteration. Nevertheless, this is no long-erawidespreadproblembecausethemodification of DNA normally involvesseveralbasechangesinafragment,which implies that the efficiency of de-tection may, in fact, reach almost 100%.In addition to classical non-denatur-ingslabgelelectrophoresis,HPCEtechniques can be considered an alter-native approach to resolving PCR poly-morphismsproducedafterbisulfitemodificationoftheDNA(Figure1)(90).Thisoptionhasseveraladvan-tages over the classical method, such asthe simultaneous separation of alleles,useofnon-denaturinggel,andauto-matedmeansthatallowsequentialanalyses of a large number of samples.Moreover,therelativepercentageofmethylationcanbeaccuratelyquanti-fiedbycalculatingtheareaundermethylated and unmethylated peaks di-rectly on the electropherogram with anappropriate computer program (e.g., 32Karat Software;BeckmanCoulterSpainS.A.,Madrid,Spain).Underthese conditions, the correlation coeffi-cientinrecoveryexperimentsis0.9994.Althoughthisisoneofthemost powerful approaches to the quan-titativeanalysisoftheDNAmethyla-tionstatusofadefinedDNAregion,unfortunately, there are currently no re-liable preparative facilities available.OTHER METHYLATION-DEPENDENT STRATEGIES NOT BASED ON BISULFITEMethylation-specificmodificationof DNA can be achieved by the use ofhydrazineandpermanganate.Hy-drazine can react with C and T but notwith methylcytosine (16), whereas per-manganate reacts with methylcytosineand T but not with C (35). Both can re-act with ssDNA, but hydrazine can alsoreact with dsDNA.Hydrazine and permanganate-modi-fiednucleotidescanberemovedwithpiperidine and detected by a number ofsequencingmethodssuchasligation-mediatedPCR(93).Themajordraw-backofbothapproachesistheirpoorsensitivity because, even under optimalconditions, usually at least 2 g DNAare required (65). Furthermore, at least10%20% of the population should dis-play a certain alteration to be detectedby these methods (81).Lowsensitivityandresolutioninbothmethods,aswellasthefactthatsequencingmethodsareextremelyla-bor-intensive,explainwhyhydrazineand permanganate are only occasional-lyusedtocheckdataobtainedbythebisulfite-modification technique.HOW TO FIND NEWMETHYLATION HOT SPOTS Classical DNA methylation researchhelps to investigate the methylation sta-tus of cytosines that occur in known (orpartiallyknown)DNAsequences.However,alternativewaysofinvesti-gatinggenome-widemethylationbysearching for as yet unidentified spotshavebeendeveloped.Allofthesemethods rely on the distinctive proper-Vol. 33, No. 3 (2002) BioTechniques 645tiesoftheCpGislandstofindnewmethylated sequences in the genome.Therestrictionlandmarkgenomicscanning (RLGS) (49) technique is oneoftheearliestwaysreportedforgenome-wide,methylation-specificsearching (56). DNA is radioactively la-beledatmethylation-specificcleavagesites and then size-fractionated in 1-D.The products are then digested with anyrestriction endonuclease that is specificforhigh-frequencytargets.Fragmentsare then separated in 2-D, which yieldsseveral scattered methylation-related hotspots. The location and strength of a spotreveal its locus and the copy number ofthecorrespondingrestrictionsite.Thisapproach led to the discovery of severalCpG islands of which there was no pre-vious sequence knowledge (52,56).Gonzalgo et al. (43) described anoth-er suitable tool for screening the genomefor regions that display altered patternsofDNAmethylation.Themethod,termed methylation-sensitive arbitrarilyprimedPCR,isasimpleDNAfinger-printing technique that relies on arbitrar-ily primed PCR amplification, followedby digestion with restriction isoschizo-mers.Strain-specificarraysofDNAfragments are generated by PCR ampli-fication using arbitrary oligonucleotidesto prime DNA synthesis from genomicsites that accidentally or roughly match.Usually,twocyclesofPCRareper-formedunderlow-stringencycondi-tions,followedbyPCRathighstrin-gencywithspecificprimers.DNAamplified in this manner is digested withacoupleofmethylation-sensitiveiso-schizomers,andfragmentsdisplayingdifferentialmethylationpatternsarecloned and used as probes for Southernblot analysis to corroborate the differen-tial methylation of such DNA regions.Another approach is called CpG is-land amplification (94). DNA is digest-edwithrestrictionisoschizomers,andthe restriction products are PCR-ampli-fiedafterend-adaptorligation.Al-though methylated CpG islands are se-lectively amplified, the cloning of trulyCpG-rich DNA regions is frequently alaborious task.Another original approach to isolatemethylatedCpG-richregionshasre-cently been described (86). This methodemploystheaffinitychromatography(20)ofafragmentofthemethyl-CpGbindingdomainofMeCP2topurifymethylatedCpG-richfragmentsfrommixturesobtainedbydigestionwithmethylation-specific restriction endonu-cleases. Fragments of interest are thenclonedintoa ZapIIvector(Strata-gene, La Jolla, CA, USA), and the frag-ments that are mostly rich in CpG dinu-cleotides are isolated by the segregationofpartiallymeltedmoleculesinpoly-acrylamidegelsthatcontainalineargradientofchemicaldenaturant.De-spitetheadvantagesofthisapproach,the specificity of the methylated DNAbindingcolumnneedstobeimprovedfor it to be a first-class method.ReviewUndoubtedly, one of the most effec-tivemeansofgenome-widesearchingfor CpG islands is the use of the novelCpG island arrays technology. Huang etal. (53) proposed an array-based meth-od, termed differential methylation hy-bridization, which allows the simultane-ousdeterminationofthemethylationrate of greater than 276 CpG island loci.CpG island library is obtained as previ-ously described by Cross et al. (20), andthe DNA fragments are gridded on high-density arrays. Genomic DNA from thetissues of interest is digested with MseI,whichyieldsalargenumberoffrag-mentsthatcontainintactCpGislands.Half of the subtracted DNA is then di-gestedwithmethylation-sensitiveen-donuclease BstUI, whose sequence tar-getoccursfrequentlywithinCpGislands. The digestion products are usedas templates for linker PCR. Unmethy-latedtargetsaredifferentiatedfrommethylated targets since the former arecut and no PCR products are obtained,while the latter can be amplified by link-erPCR.Resultingoligonucleotides,termed pre-treated products, are used asprobesforscreeninghypermethylatedsequences within the CpG island library.DifferentialmethylationhybridizationhasbeenappliedtothescreeningofCpG methylation in cancer using CpGarrays of 300 and 1104 targets, respec-tively.Therefore,differentialmeth-ylationhybridizationcanbeusedasasophisticatedscreeningtoolforselect-ing putative DNA fragments susceptibletoanalysisingreaterdepthbyothermore specific methods. A modificationofthismethodforthestudyofDNAmethylationincanceristhemethyla-tion-specific oligonucleotide microarray(39). After bisulfite treatment and PCRamplification,productsarearray-hy-bridized.Methylation-specificoligo-nucleotide microarrays are designed todetect methylation at specific nucleotidepositions.Quantitativedifferencescanbe obtained by fluorescence detection.SUMMARYIncontrasttogeneticmutations,DNA methylation alters gene expressionwithoutDNAbasechanges.Currently,DNA methylation can be studied using agreat variety of experimental techniques,involvingamultidisciplinaryperspec-tive on the DNA methylation status. OneapproachisthequantificationoftheoveralldegreeofDNAmethylation.This can be accomplished by high-per-formance separation techniques, by en-zymatic/chemical means, and even by insitu hybridization using antibodies anti-methylcytosine.ToanalyzetheDNAmethylation status of a particular DNAsequence in depth, methylation-sensitiverestriction endonucleases are commonlyused. In addition, once bisulfite modifi-cationoftheDNAhasbeenaccom-plished, it is possible to obtain quantita-tive or semi-quantitative data regardingallele-specific methylation. Finally, sev-eral innovative techniques have been de-velopedtoinvestigatenew,previouslyunknownmethylationhotspotswithinthe whole genome. Future steps towardautomation and multi-assay arrays willallowlargenumbersofsamplestobechecked simultaneously.ACKNOWLEDGMENTSWe thank Dr. Esteban Ballestar forhelpful discussion. This work was sup-ported by I+D+I Project SAF grant no.2001-0059andTheInternationalRettSyndrome Association.REFERENCES1.Abrams, E.S., S.E. Murdaugh, and L.S. Ler-man. 1990. Comprehensive detection of singlebasechangesinhumangenomicDNAusingdenaturinggradientgelelectrophoresisandaGC clamp. Genomics 7:463-475.2.Adouard,V.,R.Dante,A.Niveleau,E.De-lain,B.Revet,andM.Ehrlich. 1985.Theaccessibility of 5-methylcytosine to specific an-tibodiesindouble-strandedDNAofXantho-monas phage XP12. Eur. J. Biochem. 152:115-121.3.Annan,R.S.,G.M.Kresbach,R.W.Giese,and P. Vouros. 1989. Trace detection of modi-fiedDNAbasesviamoving-beltliquidchro-matography-massspectrometryusingelec-trophoric derivatization and negative chemicalionization. J. Chromatogr. 465:285-296.4.Barbin, A., C. Montpellier, N. Kokalj-Vokac,A. Gibaud, A. Niveleau, B. Malfoy, B. Dutril-laux, and C.A. Bourgeois. 1994. New sites ofmethylcytosine-richDNAdetectedonmeta-phase chromosomes. Hum. Genet. 94:684-692.5.Baumer,A.,U.Wiedemann,M.Hergers-berg,andA.Schinzel. 2001.AnovelMSP/DHPLCmethodfortheinvestigationofthemethylation status of imprinted genes enablesthe molecular detection of low cell mosaicisms.Hum. Mutat. 17:423-430.6.Bensaada, M., H. Kiefer, G. Tachdjian, J.M.Lapierre,V.Cacheux,A.Niveleau,andP.Metezeau. 1998.AlteredpatternsofDNAmethylationonchromosomesfromleukemiacelllines:identificationof5-methylcytosinesbyindirectimmunodetection.CancerGenet.Cytogenet. 103:101-109.7.Bian, Y.S., P. Yan, M.C. Osterheld, C. Fon-tolliet,andJ.Benhattar. 2001.Promotermethylation analysis on microdissected paraf-fin-embedded tissues using bisulfite treatmentand PCR-SSCP. BioTechniques 30:66-72.8.Bianco, T., D. Hussey, and A. Dobrovic. 1999.Methylation-sensitive, single-strand conforma-tionanalysis(MS-SSCA):arapidmethodtoscreen for and analyze methylation. Hum. Mu-tat. 14:289-293.9.Bigger, C.H., K. Murray, and N.E. Murray.1973. Recognition sequence of a restriction en-zyme. Nat. New Biol. 244:7-10.10.Bird,A.P. 1986.CpG-richislandsandthefunction of DNA methylation. Nature 321:209-213.11.Bourchis, D., D. Le Bourhis, D. Patin, A. Ni-veleau, P. Comizzoli, J. Renard, and E. Vie-gas-Pequignot. 2001. Delayed and incompletereprogrammingofchromosomemethylationpatterns in bovine cloned embryos. Curr. Biol.11:1542-1546.12.Busslinger,M.,E.deBoer,S.Wright,F.G.Grosveld,andR.A.Flavell. 1983.These-quence GGCmCGG is resistant to MspI cleav-age. Nucleic Acids Res. 11:3559-3569.13.Butkus, V., L. Petrauskiene, Z. Maneliene, S.Klimasauskas, V. Laucys, and A. Janulaitis.1987.CleavageofmethylatedCCCGGGse-quencescontainingeitherN4-methylcytosineor5-methylcytosinewithMspI,HpaII,SmaI,XmaI and Cfr9I restriction endonucleases. Nu-cleic Acids Res. 15:7091-7102.14.Catania,J.,B.C.Keenan,G.P.Margison,and D.S. Fairweather. 1987. Determination of5-methylcytosinebyacidhydrolysisofDNAwithhydrofluoricacid.Anal.Biochem.167:347-351.15.Cedar,H.,A.Solage,G.Glaser,andA.Razin. 1979.Directdetectionofmethylatedcytosine in DNA by use of the restriction en-zyme MspI. Nucleic Acids Res. 6:2125-2132.16.Church, G.M. and W. Gilbert. 1985. The ge-nomic sequencing technique. Prog. Clin. Biol.Res. 177:17-21.17.Clark, S.J., J. Harrison, C.L. Paul, and M.Frommer. 1994. High sensitivity mapping ofmethylatedcytosines.NucleicAcidsRes.22:2990-2997.18.Costello, J.F. and C. Plass. 2001. Methylationmatters. J. Med. Genet. 38:285-303.19.Cravo, M., R. Pinto, P. Fidalgo, P. Chaves, L.Gloria,C.Nobre-Leitao,andM.F.Costa.1996. Global DNA hypomethylation occurs inthe early stages of intestinal type gastric carci-noma. Gut 39:434-438.20.Cross, S.H., J.A. Charlton, X. Nan, and A.P.Bird. 1994. Purification of CpG islands using amethylated DNA binding column. Nat. Genet.6:236-244.21.de Capoa, A., F. Menendez, I. Poggesi, P. Gi-ancotti, C. Grappelli, M.R. Marotta, M. DiLeandro, C. Reynaud, et al. 1996. Cytologi-cal evidence for 5-azacytidine-induced demeth-Vol. 33, No. 3 (2002) BioTechniques 647ylationoftheheterochromaticregionsofhumanchromosomes.ChromosomeRes.4:271-276.22.Diaz-Sala,C.,M.Rey,A.Boronat,R.Bes-ford,andR.Rodriguez. 1995.Variationsinthe DNA methylation and polypeptide patternsof adult hazel (corylus avellana L.) associatedwith sequential in vitro subcultures. Plant CellRep. 15:218-221.23.Duthie, S.J., S. Narayanan, S. Blum, L. Pirie,andG.M.Brand.2000.Folatedeficiencyinvitro induces uracil misincorporation and DNAhypomethylation and inhibits DNA excision re-pair in immortalized normal human colon ep-ithelial cells. Nutr. Cancer 37:245-251.24.Eads, C.A., K.D. Danenberg, K. Kawakami,L.B. Saltz, C. Blake, D. Shibata, P.V. Danen-berg,andP.W.Laird. 2000.MethyLight:ahigh-throughput assay to measure DNA methy-lation. Nucleic Acids Res. 28:E32.25.Ehrlich, M., M.A. Gama-Sosa, L.H. Huang,R.M. Midgett, K.C. Kuo, R.A. McCune, andC. Gehrke. 1982. Amount and distribution of5-methylcytosine in human DNA from differ-ent types of tissues of cells. Nucleic Acids Res.10:2709-2721.26.Ehrlich, M. and K.C. Ehrlich. 1993. Effect ofDNA methylation on the binding of vertebrateand plant proteins to DNA. EXS 64:145-168.27.Eick, D., H.J. Fritz, and W. Doerfler. 1983.Quantitative determination of 5-methylcytosineinDNAbyreverse-phasehigh-performanceliquidchromatography.Anal.Biochem.135:165-171.28.Esteller, M., S.R. Hamilton, P.C. Burger, S.B.Baylin, and J.G. Herman. 1999. InactivationoftheDNArepairgeneO6-methylguanine-DNA methyltransferase by promoter hyperme-thylation is a common event in primary humanneoplasia. Cancer Res. 59:793-797.29.Esteller, M., J. Garcia-Foncillas, E. Andion,S.N. Goodman, O.F. Hidalgo, V. Vanaclocha,S.B. Baylin, and J.G. Herman. 2000. Inacti-vation of the DNA-repair gene MGMT and theclinicalresponseofgliomastoalkylatingagents. N. Engl. J. Med. 343:1350-1354.30.Esteller,M.,A.Sparks,M.Toyota,M.Sanchez-Cespedes, G. Capella, M.A. Peina-do,S.Gonzalez,G.Tarafa,etal. 2000.Analysisofadenomatouspolyposiscoli pro-moter hypermethylation in human cancer. Can-cer Res. 60:4366-4371.31.Esteller, M., P.G. Corn, S.B. Baylin, and J.G.Herman. 2001. A gene hypermethylation pro-fileofhumancancer.CancerRes.61:3225-3229.32.Esteller,M.,M.F.Fraga,M.Zhou-Guo,J.Garcia-Foncillas, A.K. Godwin, J. Trojan, I.Hedenfalk,C.Vaurs-Barrihre,etal. 2001.AberrantCpG-islandmethylationprofileand5-methylcytosineDNAcontentinhereditaryhuman cancer mimics sporadic tumorigenesis.Hum. Mol. Genet. 10:3001-3007.33.Fraga, M.F., R. Rodriguez, and M.J. Canal.2000.RapidquantificationofDNAmethyla-tionbyhighperformancecapillaryelec-trophoresis. Electrophoresis 21:2990-2994.34.Fraga, M.F., M.L. Centeno, A.E. Valds, P.Moncalen, M.J. Canal, and R. Rodriguez.2000.GenomicDNAmethylationandpoly-amines levels as key processes in plant aging:applications for clonal multiplication of maturePinus radiata trees, p. 495-506. In E. Ritter andS. Espinel (Eds.), Applications of Biotechnolo-gy to Forest Genetics. A.F. Albundia, Vitoria.35.Fritzsche, E., H. Hayatsu, G.L. Igloi, S. Iida,and H. Kossel. 1987. The use of permanganateas a sequencing reagent for identification of 5-methylcytosineresiduesinDNA.NucleicAcids Res. 15:5517-5528.36.Furuichi, Y., Y. Wataya, H. Hayatsu, and T.Ukita. 1970. Chemical modification of tRNA-Tyr-yeast with bisulfite. A new method to mod-ifyisopentenyladenosineresidue.Biochem.Biophys. Res. Commun. 41:1185-1191.37.Galm, O., M. Rountree, E. Bachman, K.-W.Jair, S.B. Baylin, and J. Herman. 2001. En-zymaticregionalmethylationassay:anovelmethod to quantify regional CpG methylationdensity. Genome Res. 12:153-157.38.Gama-Sosa,M.A.,R.M.Midgett,V.A.Slagel, S. Githens, K.C. Kuo, C.W. Gehrke,and M. Ehrlich. 1983. Tissue-specific differ-encesinDNAmethylationinvariousmam-mals. Biochim. Biophys. Acta 740:212-219.39.Gitan, R.S., H. Shi, C.M. Chen, P.S. Yan, andT.H. Huang. 2002. Methylation-specific oligo-nucleotidemicroarray:anewpotentialforhigh-throughput methylation analysis. GenomeRes. 12:158-164.40.Gloria,L.,M.Cravo,A.Pinto,L.S.deSousa,P.Chaves,C.N.Leitao,M.Quina,F.C. Mira, et al. 1996. DNA hypomethylationandproliferativeactivityareincreasedintherectal mucosa of patients with long-standing ul-cerative colitis. Cancer 78:2300-2306.41.Goessl,C.,M.Muller,R.Heicappell,H.Krause, and K. Miller. 2001. DNA-based de-tection of prostate cancer in blood, urine, andejaculates. Ann. NY Acad. Sci. 945:51-58.42.Gonzalgo, M.L. and P.A. Jones. 1997. Rapidquantitation of methylation differences at spe-cificsitesusingmethylation-sensitivesinglenucleotide primer extension (MS-SNuPE). Nu-cleic Acids Res. 25:2529-2531.43.Gonzalgo, M.L., G. Liang, C.H. Spruck, III,J.M.Zingg,W.M.Rideout,III,andP.A.Jones. 1997. Identification and characterizationof differentially methylated regions of genomicDNAbymethylation-sensitivearbitrarilyprimed PCR. Cancer Res. 57:594-599.44.Gowher,H.,O.Leismann,andA.Jeltsch.2000. DNA of Drosophila melanogaster con-tains5-methylcytosine.EMBOJ.19:6918-6923.45.Grigg, G. and S. Clark. 1994. Sequencing 5-methylcytosineresiduesingenomicDNA.Bioessays 16:431-436.46.Gruenbaum, Y., T. Naveh-Many, H. Cedar,andA.Razin. 1981.SequencespecificityofmethylationinhigherplantDNA.Nature292:860-862.47.Grunau,C.,S.J.Clark,andA.Rosenthal.2001. Bisulfite genomic sequencing: systemat-icinvestigationofcriticalexperimentalpara-meters. Nucleic Acids Res. 29:E65.48.Harrison,J.,C.Stirzaker,andS.J.Clark.1998.CytosinesadjacenttomethylatedCpGsites can be partially resistant to conversion ingenomic bisulfite sequencing leading to methy-lation artifacts. Anal. Biochem. 264:129-132.49.Hatada, I., Y. Hayashizaki, S. Hirotsune, H.Komatsubara,andT.Mukai. 1991.Age-nomic scanning method for higher organismsusing restriction sites as landmarks. Proc. Natl.Acad. Sci. USA 88:9523-9527.50.Heiskanen, M., A.C. Syvanen, H. Siitari, S.Laine, and A. Palotie. 1994. A novel methodto quantitate methylation of specific genomicregions. PCR Methods Appl. 4:26-30.51.Herman, J.G., J.R. Graff, S. Myohanen, B.D.Nelkin, and S.B. Baylin. 1996. Methylation-specific PCR: a novel PCR assay for methyla-tionstatusofCpGislands.Proc.Natl.Acad.Sci. USA 93:9821-9826.52.Hirotsune,S.,K.Hirose,H.Kataoka,J.Kuromitsu, Y. Furuichi, M. Muramatsu, Y.Matsuda,andY.Hayashizaki. 1994.Spotmapping on the standard profile of restrictionlandmark genomic scanning (RLGS) of sortedchromosome 20 using methylation-insensitiveenzyme. Genomics 24:593-596.53.Huang,T.H.,M.R.Perry,andD.E.Laux.1999. Methylation profiling of CpG islands inhuman breast cancer cells. Hum. Mol. Genet.8:459-470.54.Jackson, P., D. Millar, E. Kingsley, G. Yard-ley, K. Ow, S. Clark, and P.J. Russell. 2000.MethylationofaCpGislandwithinthepro-moter region of the KAI1 metastasis suppressorgene is not responsible for down-regulation ofKAI1 expression in invasive cancers or cancercell lines. Cancer Lett. 157:169-176.55.Jones, P.L. and A.P. Wolffe. 1999. Relation-shipsbetweenchromatinorganizationandDNA methylation in determining gene expres-sion. Semin. Cancer Biol. 9:339-347.56.Kawai, J., S. Hirotsune, K. Hirose, S. Fushi-ki, S. Watanabe, and Y. Hayashizaki. 1993.MethylationprofilesofgenomicDNAofmouse developmental brain detected by restric-tionlandmarkgenomicscanning(RLGS)method. Nucleic Acids Res. 21:5604-5608.57.Kosaki, K., M.J. McGinniss, A.N. Veraksa,W.J. McGinnis, and K.L. Jones. 1997. Prad-er-WilliandAngelmansyndromes:diagnosiswithabisulfite-treatedmethylation-specificPCR method. Am. J. Med. Genet. 73:308-313.58.Kuo, K.C., R.A. McCune, C.W. Gehrke, R.Midgett, and M. Ehrlich. 1980. Quantitativereversed-phase high performance liquid chro-matographic determination of major and modi-fieddeoxyribonucleosidesinDNA.NucleicAcids Res. 8:4763-4776.59.Kupper, D., M. Reuter, A. Meisel, and D.H.Kruger. 1997. Reliable detection of DNA CpGmethylationprofilesbytheisoschizomersMspI/HpaII using oligonucleotide stimulators.BioTechniques 23:843-847.60.Kutueva, L.I., V.V. Ashapkin, and B.F. Van-yushin. 1996. The methylation pattern of a cy-tosine DNA-methyltransferase gene in Arabi-dopsis thaliana plants. Biochem. Mol. Biol. Int.40:347-353.61.Leonard,S.A.,S.C.Wong,andJ.W.Nyce.1993. Quantitation of 5-methylcytosine by one-dimensional high-performance thin-layer chro-matography. J. Chromatogr. 645:189-192.62.Maekawa, M., K. Sugano, H. Kashiwabara,M. Ushiama, S. Fujita, M. Yoshimori, and T.Kakizoe. 1999. DNA methylation analysis us-ingbisulfitetreatmentandPCR-single-strandconformation polymorphism in colorectal can-cer showing microsatellite instability. Biochem.Biophys. Res. Commun. 262:671-676.63.Maniatis, T., E.F. Fritsch, and J. Sambrook.Review648 BioTechniques Vol. 33, No. 3 (2002)1982. Molecular Cloning: A Laboratory Manu-al. CSH Laboratory Press, Cold Spring Harbor,NY.64.McClelland, M., M. Nelson, and E. Raschke.1994. Effect of site-specific modification on re-striction endonucleases and DNA modificationmethyltransferases.NucleicAcidsRes.22:3640-3659.65.McGrew,M.J.andN.Rosenthal. 1993.Quantitation of genomic methylation using lig-ation-mediatedPCR.BioTechniques15:722-729.66.Melki,J.R.,P.C.Vincent,andS.J.Clark.1999.ConcurrentDNAhypermethylationofmultiplegenesinacutemyeloidleukemia.Cancer Res. 59:3730-3740.67.Miller,O.J.,W.Schnedl,J.Allen,andB.F.Erlanger. 1974. 5-methylcytosine localized inmammalian constitutive heterochromatin. Na-ture 251:636-637.68.Miniou,P.,M.Jeanpierre,V.Blanquet,V.Sibella, D. Bonneau, C. Herbelin, A. Fischer,A. Niveleau, et al. 1994. Abnormal methyla-tionpatterninconstitutiveandfacultative(Xinactive chromosome) heterochromatin of ICFpatients. Hum. Mol. Genet. 3:2093-2102.69.Mizugaki,M.,K.Itoh,T.Yamaguchi,S.Ishiwata,T.Hishinuma,S.Nozaki,andN.Ishida. 1996. Preparation of a monoclonal anti-bodyspecificfor5-methyl-2-deoxycytidineanditsapplicationforthedetectionofDNAmethylation levels in human peripheral bloodcells. Biol. Pharm. Bull. 19:1537-1540.70.Montpellier,C.,C.A.Burgeois,N.Kokalj-Vokac,M.Muleris,A.Niveleau,C.Rey-naud,A.Gibaud,B.Malfoy,etal. 1994.Detectionofmethylcytosine-richheterochro-matin on banded chromosomes. Application tocells with various status of DNA methylation.Cancer Genet. Cytogenet. 78:87-93.71.Norwood,C.B.,E.Jackim,andS.Cheer.1993. DNA adduct research with capillary elec-trophoresis. Anal. Biochem. 213:194-199.72.Nuovo, G.J., T.W. Plaia, S.A. Belinsky, S.B.Baylin, and J.G. Herman. 1999. In situ detec-tion of the hypermethylation-induced inactiva-tion of the p16 gene as an early event in onco-genesis.Proc.Natl.Acad.Sci.USA96:12754-12759.73.Oakeley, E.J., A. Podesta, and J.P. Jost. 1997.DevelopmentalchangesinDNAmethylationof the two tobacco pollen nuclei during matura-tion.Proc.Natl.Acad.Sci.USA94:11721-11725.74.Oakeley, E.J., F. Schmitt, and J.P. Jost. 1999.Quantification of 5-methylcytosine in DNA bythechloroacetaldehydereaction.BioTech-niques 27:744-752.75.Paul,C.L.andS.J.Clark. 1996.Cytosinemethylation:quantitationbyautomatedge-nomic sequencing and GENESCAN analysis.BioTechniques 21:126-133.76.Piyathilake,C.J.,A.R.Frost,W.C.Bell,D.Oelschlager,H.Weiss,G.L.Johanning,A.Niveleau, D.C. Heimburger, et al. 2001. Al-tered global methylation of DNA: an epigeneticdifferenceinsusceptibilityforlungcancerisassociatedwithitsprogression.Hum.Pathol.32:856-862.77.Poduslo,S.E.,M.Dean,U.Kolch,andS.J.OBrien. 1991.Detectinghigh-resolutionpolymorphisms in human coding loci by com-biningPCRandsingle-strandconformationpolymorphism (SSCP) analysis. Am. J. Hum.Genet. 49:106-111.78.Quivy,J.P.andP.B.Becker. 1994.Directdideoxy sequencing of genomic DNA by liga-tion-mediatedPCR.BioTechniques16:238-241.79.Radlinska,M.andK.Skowronek. 1998.Novel procedure for the detection of 5-methyl-cytosine. Acta Microbiol. Pol. 47:327-334.80.Raizis, A.M., F. Schmitt, and J.P. Jost. 1995.Abisulfitemethodof5-methylcytosinemap-pingthatminimizestemplatedegradation.Anal. Biochem. 226:161-166.81.Rein, T., D.A. Natale, U. Gartner, M. Nigge-mann,M.L.DePamphilis,andH.Zorbas.1997. Absence of an unusual densely methy-lated island at the hamster dhfr ori-. J. Biol.Chem. 272:10021-10029.82.Rein, T., M.L. DePamphilis, and H. Zorbas.1998. Identifying 5-methylcytosine and relatedmodifications in DNA genomes. Nucleic AcidsRes. 26:2255-2264.83.Saluz,H.andJ.P.Jost. 1993.Majortech-niques to study DNA methylation. EXS 64:11-26.84.Shapiro,R.,V.DiFate,andM.Welcher.1974. Deamination of cytosine derivatives bybisulfite.Mechanismofthereaction.J.Am.Chem. Soc. 96:206-212.85.Sharma, M., R. Jain, E. Ionescu, and H.K.Slocum. 1995. Capillary electrophoretic sepa-ration and laser-induced fluorescence detectionof the major DNA adducts of cisplatin and car-boplatin. Anal. Biochem. 228:307-311.86.Shiraishi,M.,Y.H.Chuu,andT.Sekiya.1999. Isolation of DNA fragments associatedwith methylated CpG islands in human adeno-carcinomasofthelungusingamethylatedDNA binding column and denaturing gradientgel electrophoresis. Proc. Natl. Acad. Sci. USA96:2913-2918.87.Singer-Sam,J.,M.Grant,J.M.LeBon,K.Okuyama, V. Chapman, M. Monk, and A.D.Riggs. 1990. Use of a HpaII-polymerase chainreaction assay to study DNA methylation in thePgk-1CpGislandofmouseembryosatthetime of X-chromosome inactivation. Mol. Cell.Biol. 10:4987-4989.88.Singer-Sam, J., J.M. LeBon, R.L. Tanguay,and A.D. Riggs. 1990. A quantitative HpaII-PCRassaytomeasuremethylationofDNAfromasmallnumberofcells.NucleicAcidsRes. 18:687.89.Stirzaker,C.,D.S.Millar,C.L.Paul,P.M.Warnecke,J.Harrison,P.C.Vincent,M.Frommer,andS.J.Clark. 1997.ExtensiveDNA methylation spanning the Rb promoter inretinoblastoma tumors. Cancer Res. 57:2229-2237.90.Suzuki, H., F. Itoh, M. Toyota, T. Kikuchi, H.Kakiuchi,Y.Hinoda,andK.Imai. 2000.Quantitative DNA methylation analysis by flu-orescentpolymerasechainreactionsingle-strandconformationpolymorphismusinganautomatedDNAsequencer.Electrophoresis21:904-908.91.Tan,W.G.,T.J.Carnelley,P.Murphy,H.Wang, J. Lee, S. Barker, M. Weinfeld, andX.C. Le. 2001. Detection of DNA adducts ofbenzo[a]pyreneusingimmunoelectrophoresiswithlaser-inducedfluorescence.AnalysisofA549 cells. J. Chromatogr. A 924:377-386.92.Tasheva, E.S. and D.J. Roufa. 1994. Denselymethylated DNA islands in mammalian chro-mosomalreplicationorigins.Mol.Cell.Biol.14:5636-5644.93.Thomassin, H., E.J. Oakeley, and T. Grange.1999.Identificationof5-methylcytosineincomplex genomes. Methods 19:465-475.94.Toyota, M., C. Ho, N. Ahuja, K.W. Jair, Q.Li, M. Ohe-Toyota, S.B. Baylin, and J.P. Issa.1999. Identification of differentially methylat-ed sequences in colorectal cancer by methylat-edCpGislandamplification.CancerRes.59:2307-2312.95.Turk,P.W.andS.A.Weitzman. 1995.Freeradical DNA adduct 8-OH-deoxyguanosine af-fects activity of HpaII and MspI restriction en-donucleases. Free Radic. Res. 23:255-258.96.Uchida, T., H. Ohashi, E. Aoki, Y. Nakahara,T.Hotta,T.Murate,H.Saito,andT.Ki-noshita. 2000. Clonality analysis by methyla-tion-specific PCR for the human androgen-re-ceptorgene(HUMARA-MSP).Leukemia14:207-212.97.Veilleux,C.,J.Bernardino,A.Gibaud,A.Niveleau, B. Malfoy, B. Dutrillaux, and C.A.Bourgeois. 1995. [Changes in methylation oftumor cells: a new in situ quantitative approachon interphase nuclei and chromosomes]. Bull.Cancer 82:939-945.98.Wagner, I. and I. Capesius. 1981. Determina-tionof5-methylcytosinefromplantDNAbyhigh-performanceliquidchromatography.Biochim. Biophys. Acta 654:52-56.99.Wolffe,A.P.,P.L.Jones,andP.A.Wade.1999.DNAdemethylation.Proc.Natl.Acad.Sci. USA 96:5894-5896.100.Worth,C.C.,O.J.Schmitz,H.C.Kliem,and M. Wiessler. 2000. Synthesis of fluores-cently labeled alkylated DNA adduct standardsandseparationbycapillaryelectrophoresis.Electrophoresis 21:2086-2091.101.Wu, J., J.P. Issa, J. Herman, D.E. Bassett,Jr., B.D. Nelkin, and S.B. Baylin. 1993. Ex-pressionofanexogenouseukaryoticDNAmethyltransferase gene induces transformationof NIH 3T3 cells. Proc. Natl. Acad. Sci. USA90:8891-8895.102.Xiong, Z. and P.W. Laird. 1997. COBRA: asensitive and quantitative DNA methylation as-say. Nucleic Acids Res. 25:2532-2534.103.Zhou, Y., J.M. Magill, R.J. Newton, and C.Magill. 1997. Use of a single sequencing ter-mination reaction to distinguish between cyto-sine and 5-methylcytosine in bisulfite-modifiedDNA. BioTechniques 22:850-854.Address correspondence to:Dr. Manel EstellerCancer Epigenetics LaboratoryProgram of Molecular PathologyCentro Nacional de Investigaciones Oncolgicas (CNIO)C/Melchor Fernndez Almagro, no. 3. E-28029,Madrid, Spaine-mail: mesteller@cnio.esVol. 33, No. 3 (2002) BioTechniques 649