Severalstudiesfromthelastthreedecadeshaveclearly demon-stratedthatchangesinDNAmethylationpatternsassociatewith chronologicalagingacrossnearlytheentirehumanlifespan(Alisch etal.,2012;Belletal.,2012;Bocklandtetal.,2011;Boksetal.,2009;
Bollatietal.,2009;Christensenetal.,2009;Florathetal.,2013;
Garagnanietal.,2012;Gentilinietal.,2013;Hannumetal.,2013;
Heynetal.,2012;McClayetal.,2014;Rakyanetal.,2010).This phenomenon,commonly defined as“epigenetic drift”,accounts fortheevidencethatepigeneticsimilaritiesbetweenyoung
indi-viduals are lost over time leading to divergent methylomes in elderlypopulation(Boksetal.,2009;Fragaetal.,2005;Heynetal., 2012).DNAmethylationdriftisanon-directionalchangeof methy-lomeasitinvolvesbothhypermethylationandhypomethylation eventsandithasbeenassociatedwiththeprogressive accumu-lationof epigenetic damagedue toenvironmentalfactorsorto spontaneousstochastic errorsin theprocess of transmissionof theepigenetic information.Thisphenomenon leadstobasically unpredictabledifferences in the methylome among aging indi-viduals.Theseobservationsseemtodenythepossibilitythatthe driftmayreflectaprogrammedchangeofthemethylationcode.
However,partofmethylationchangesthatareobservedwithage involvespecificregionsofthegenomeandaredirectional.Infact, severalstudiesindicatetheexistenceof aging-associated differ-entiallymethylatedregions(a-DMRs),clustersofconsecutiveCpG siteswhichexhibitchangeovertime inthesamedirection.The existenceofa-DMRsindicatesthatpartofmethylationchangesis notstochastic,butinsteadtheycouldbeassociatedwithbiological mechanismscloselyinvolvedintheagingprocessorlongevity.
Most of the initial analyses in aging have been focused on candidateloci withpotentialrelevanceforage-related diseases.
Collectively,thesestudiesshowedthataging, similartocancer, associateswithgradualbutprofoundchangesinDNAmethylation where epigenomeis markedby genome-widehypomethylation togetherwithsite-specifichypermethylationpreferentially occur-ringatCGIpromoters.Notably,althoughthefunctionalimportance of both events in aging progression and outcome remains to beascertained,theseobservationssupportthenotionthatthese changesmayconcomitantlycontributetodevelopmentof aging-associateddiseases.
5.1. DNAhypomethylationevents
GlobaldeclineofgenomicCpGmethylationisthepredominant eventin aging.Thischangeis widespreadasit typicallyoccurs atrepetitivesequencesdispersedthroughoutthegenomesuchas
Fig.4.Age-associatedchangesofDNAmethylome.SchematicrepresentationofDNAmethylationpatternsinyoung(A)vsolder(B)people.ThetoplinerepresentsDNA containingthreegenesandthreeclassesofinterspersedrepeats.Arrowsindicatetranscriptionstartsites,withbluecolorindicatingactivetranscription,whileredcolor indicatesrepressedtranscription.Exonsareshownindarkblue.EachbottomlinerepresentsthemethylationstateofDNAasdetectedforasingleindividual.White circles,unmethylatedCpGs;blackcircles,methylatedCpGs.Hypomethylatedandhypermethylatedaging-associatedDMRsarehighlightedinblueandred,respectively.(For interpretationofthereferencestocolorinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)
SINEs(i.e.,Aluelements)andLTR(i.e.,HERV-K)(Bollatietal.,2009;
Christensenetal.,2009;JintaridthandMutirangura,2010). How-ever,age-relatedlossofmethylationdoesnotaffectrepetitiveDNA equally.HypomethylationofAluandHERV-Ksequencesoccursat differentageswhileitdoesnotaffectLINE-1repeatssignificantly (Fig.4)(Bollatietal.,2009;JintaridthandMutirangura,2010).It isreasonabletoassumethatthiseventaccountsfortheincreased genomeinstabilityobservedintheelderly(VijgandDolle,2007).
However,progressiveage-dependentlossofmethylationalso con-cernsspecificgenepromoters,includingITGALandIL17RCwhose demethylation-dependenttranscriptionalactivationhasbeen pro-posedtotriggerautoimmuneresponses(Weietal.,2012;Zhang etal.,2002).
More recently, the application of novel next-generation sequencing technologies for genome-wide assessment of DNA methylation levels in aging research haspermitted the confir-mationof theseearlier observations. Studiesinvestigating DNA methylationonthewholehavebeenparticularlyinstructivefor DNA hypomethylationevents(Heyn et al.,2012; McClayet al., 2014).Infact,bywhole-genomebisulfitesequencing(WGBS),Heyn etal.(2012)comparedtheDNAmethylationstateofmorethan90%
ofallCpGspresentinthegenomebetweennewbornand nonage-narian/centenariansamples.AsignificantlossofmethylatedCpGs wasfoundinthecentenarianvsnewbornDNAs.Thiswasobserved for allchromosomes and concernedall genomiccompartments suchaspromoters,exonic,intronicandintergenicregions.Most ofthesechangeswerefocalandtheagedgenomewasthusless homogeneouslymethylatedwithrespecttothenewbornaccording totheage-dependentepigeneticdrift.
However, a part of methylation changes involved several neighboringCpGs.Amorerefinedsearchofdomainscontaining consecutiveCpGunitsshowingunidirectionalmethylationchanges actuallyrevealedthatabout2.2%oftotalgenomicCGsiteswere locatedinregionsthatweredifferentiallymethylatedinthe
cente-narianwithrespecttothenewborn.Inagreementwithprevious observations,most oftheseDMRswere locatedinintronic and intergenic regionsand commonly colocalizedwithinterspersed repetitiveDNAelements.Interestingly,DMRsoftencorresponded tolaminaassociateddomains(LADs),wherecancer-specific meth-ylation changes had been observed (Berman et al., 2011).The functionalsignificanceofLADs’hypomethylationincancerandin agingremainstobedefined.However,afar-reachingimpactofthis eventontheepigeneticregulationofgenomecanbeenvisaged.
Infact,theLADsdefinelargeheterochromatincompartmentsof thegenomewhichembedkeydevelopmentalandcell-type spe-cificgenesthataremaintainedinanepigeneticallyrepressedstate (Guelenetal.,2008;Harretal.,2015;Peric-Hupkesetal.,2010;
Reddyetal.,2008).Hence,aneventofhypomethylationofthese regionsmaycauseorreflectafault ofthemechanismsthatare centraltothedevelopmentandconservationofnormalstatesof differentiationandtissue-specificpatternsofgeneexpression.
Concerningregulatoryorcodingsequences,suchaspromoters and exons, onlyabout 20%of theDMRscolocalized withthese regions and more than 80% of them underwent hypomethyla-tionin theelderly.Thisevent,however, waslargelydependent on promoter CpG content being much more common in CpG-poor/tissue-specific gene promoters than in CGI/housekeeping genepromoters.Significantly,someoftheseDMRscolocalizedwith promotersofgenesinvolvedintheagingprocesssuchasSirtuins andIGFsignalingpathwaycomponents(Heynetal.,2012).
Recently,this findinghasbroadlybeenvalidatedin acohort of morethan 700individualsaged 25–92years. In fact,alsoin thiscase,agingwaspredominantlyfoundassociatedwith meth-ylationdeficit.About60%ofage-associatedDMRsactuallyshowed age-relatedhypomethylationandcolocalizedwithbindingsitesfor chromatinregulatoryproteins,suchasCTCFand Polycomb pro-teins,orspecifichistonemodificationstypicallyassociated with active chromatin (McClay et al., 2014). These findings indicate
thatage-relatedhypomethylationmayleadtoglobalchromatin changesofpotentialfunctionalrelevancefortranscriptional regu-lation.
5.2. DNAhypermethylationevents
Besides extensive genome-wide hypomethylation, aging involvesaprogressivegainofDNAmethylationthatmarkstheloss ofexpressionofspecificgenes.Infact,themajorityofage-related hypermethylated sites corresponds to CGI-promoters, where methylation often correlates withsuppression of transcription.
Thisaspectis ofparticular interestasit conferstomethylation changes,which accompany aging,the featuresof an epigenetic reprogramming.
Drivenbytheaccidentalobservationthatcancer-related hyper-methylationoftheCGI-promoterofestrogenreceptor(ER)gene arisesasadirectfunctionofageinnormaltissues(Issaetal.,1994), themajorityofinitialinvestigationsintoCGImethylationinaging focusedonspecificgenesselectedonthebasisoftheir involve-mentinthepathogenesisofcancerorotherage-relateddiseases.
Thisstrategyhighlightedthatagingviolates,inatime-dependent manner,theunmethylatedstateofCGIsofgenesinvolvedintumor suppression(p16INK4A(Soetal.,2006),CHD1(Wakietal.,2003), RASSF1(Wakietal.,2003),LOX(Soetal.,2006),RUNX3(Soetal., 2006),N33(Ahujaetal.,1998)andTIG1(Soetal.,2006)),genome stabilityandrepair (hTERT(Silvaetal.,2008),MLH1(Nakagawa etal.,2001),MGMT(Matsubayashietal.,2005)andOGG(Madrigano etal., 2012)),metabolism (e.g., COX7A1 (Ronn etal., 2008) and CRAT(Madriganoetal.,2012)),differentiationandgrowth(MYOD1 (Ahuja et al.,1998), c-Fos (Choiet al., 1996), IGF-2 (Issa et al., 1996)), regulationofimmune response(INFG(Madrigano etal., 2012)), coagulation (F3 (Madrigano et al., 2012)) and connec-tivetissue homeostasis(CollagenI(Takatsu etal.,1999)).Taken together, these studies also suggested that physiological aging couldpredisposeonetoage-relatedpathologicalphenotypesvia amethylation-relatedgenesilencing.Itisinterestingtoobserve thathypermethylationinage-relateddiseasesaffectsessentially allCpGsites withina CGI, whereas themethylation pattern in physiologically-agednormalindividuals ispartialand heteroge-neous.Therefore,partialmethylationchangesaccumulatinginthe aginggenome maypredispose one tothedevelopmentof age-relateddisease,inwhichthemethylationchangesarethereafter aggravatedbyamethylationspreadingmechanism(Wongetal., 1999).Examplesofsuchaprecursor-productrelationshipofaging andage-relateddiseaseshavebeenproposedforprostate,bladder, coloncancersandAlzheimer’sdisease(Florletal.,2004;Neuhausen etal.,2006;Shenetal.,2005;Siegmundetal.,2007).
Later genome-wide investigations further confirmed that age-associated hypermethylation happens preferentially at CGI promoters(Alischetal.,2012;Belletal.,2012;Bocklandtetal., 2011;Christensenetal.,2009;Florathetal.,2013;Garagnanietal., 2012;Gentilinietal.,2013;Hannumetal.,2013;Heynetal.,2012;
McClayetal.,2014;Rakyanetal.,2010;XuandTaylor,2014).Of note,thischangedoesnotequallyimpactCGIshoreregions,which insteadundergosimilarproportionsofhyper-and hypomethyla-tionevents(Florathetal.,2013;McClayetal.,2014).
Althoughtheuseofarray-basedtechniquesfocusingmainlyon genepromotersandCGIsindicatedhypermethylationasprevalent eventinaging(Alischetal.,2012;Belletal.,2012;Bocklandtetal., 2011;Boksetal.,2009;Christensenetal.,2009;Florathetal.,2013;
Hannumetal.,2013;Rakyanetal.,2010;XuandTaylor,2014), age-associatedhypermethylationseemsaphenomenon ofrelatively lowmagnitude.Fewerthana hundredCpGshaveactuallybeen observedtomethylatewithageoveraboutatotalamountof37,000 CGIsinthehaploidhumangenome,confirmingthatCGI hyper-methylationinagingisrelativelyuncommon.Thisclearlyemerges
fromthestudiesthatinterrogatedthegenomeasawhole(Heyn etal.,2012;McClayetal.,2014).Forexample,astudybasedon WGBSshowedthatonlyabout13%ofageassociatedDMRswere hypermethylatedincentenarianscomparedtonewborns,thus sug-gestingthathypomethylationispredominantinaging(Heynetal., 2012).
Despitethis, aging-associatedDMRsaremore likelytobein CGIsundergoingmethylationwithage(McClayetal.,2014). Consis-tently,hypermethylationfrequentlyoccursatgeneswithpotential relevance for age-related phenotypes/diseases including genes involvedindevelopment(protocadherins,homeoboxgenes)and signaling(MAPKpathways’members,ryanodinereceptors) asso-ciatedwithcancer,longevity,senescenceandneurodegeneration (Belletal.,2012;Hannumetal.,2013;McClayetal.,2014;Rakyan et al., 2010; Xu and Taylor, 2014). It is interesting toobserve that,althoughmostanalyseshavebeenperformedonbloodcells, hypermethylatedaging-relatedDMRsseemtobelargelyshared by multiple tissues (Horvath et al., 2012; Rakyan et al., 2010;
Teschendorffetal.,2010).ThissupportstheviewthatDNA meth-ylationchangesdonotoccurrandomlyinthecontextofthehuman genomebutaredirectedtoregionssharingcommonfeatures.
Notably,age-associatedhypermethylatedDMRsin differenti-atedtissuesoftenoverlapwithpromotersofgenesthatinstemcells havebivalentchromatinmarks(H3K4me3andH3K27me3)andare targetofthepolycombrepressivecomplex2(PRC2)(Hannumetal., 2013;Heynetal.,2012;Rakyanetal.,2010;Teschendorffetal., 2010;XuandTaylor,2014).Manyofthesegenesencode transcrip-tionfactorsnecessaryfordifferentiationandarealreadytargetof epigeneticderegulationinstemcells duringaging,likely under-lyingtheobserveddeclineofstemcellfunction(Beermanetal., 2013;Borketal.,2010;BrackandRando,2007).Thiswouldsuggest thattheage-associatedmethylationdefectsobservedin differen-tiatedtissuesmayreflectchangesoccurringintheagedstemcell population.Moreover,thesamestemcell-likebivalentmarksand PRC2occupancyhavebeenrecognizedtopredisposetumor sup-pressorgenepromoterstoDNAhypermethylationincancer.Infact, thereissignificantcorrespondencebetweenthosegenes under-goinghypermethylationinagingandgenesundergoingthesame eventincancer(Teschendorffetal.,2010)andincancer-associated conditionssuchasobesity,inflammationandsmokingaddiction (Issa,2011;Issaetal.,2001;Selamatetal.,2012;Suzukietal.,2009;
Xuetal.,2013).Theseobservationslinkstemcellagingtocancer riskandsuggestthatagingmayelicitanepigeneticswitchfromless stablehistone-basedgenerepressioninstemcellstopermanent DNAmethylation-basedgene repressionincancercells (Xuand Taylor,2014).Takingcancerasanexample,thisepigenetic mech-anismcouldbeamodelexplaininghowtheagingprocesscould predisposeonetoage-specificphenotypes/diseases.
Allinall,itisevident thatsomespecificDNAregions direc-tionallyundergoDNAmethylationchangesacrossagedindividuals probablyas aconsequence ofsharedchromatinfeatures. How-ever,thesesite-specificeventsco-existwiththeepigeneticdrift accordingtowhichdeviationofinter-individualgenomic methyl-ationpatternsoccursovertime(Fig.4).Onepossibleexplanation isthatagingmightprimarilyintroduceageneraldisorderinthe methylationpatternswhichcouldbethenfollowedbytheselection andclonalexpansionofthosecellsbearingmethylationdefectsin specificregionsofthegenomewherethesedefectswouldbe toler-atedandgivethecellsadvantagesofsurvivalorproliferation(Issa, 2014).
6. 5hmCandaging
TheidentificationofthenewDNAepigeneticmark5hmCopens newperspectivesforthestudyof epigeneticreprogrammingin
aging.OurknowledgeofDNAmethylationpatternsinboth phys-iologicalandpathologicalconditionsindeedneedsarevaluation afterthediscoveryof5hmC.Thisisduetothefactthatconventional bisulfitesequencingmethod,whichhasbeengenerallyconsidered thegoldstandardmethodforDNAmethylationanalysis,isnotable todiscriminatebetween5mCand5hmC.Therefore,studiesbased onconventionalbisulfitemodificationactuallymaskthe contribu-tionof5hmC(Huangetal.,2010).
Information regarding 5hmC in aging process is currently limited.Onlyfewreportshaveaddressedthisissuefocusing atten-tion on 5hmC changes in aged mice. Mouse cerebellum and hippocampusshowanincreaseof5hmClevelswithagingwhich canbepreventedbycaloricrestriction,awell-knownphenomenon associated withlongevity(Chen etal., 2012b; Chouliarasetal., 2012; Szulwachet al.,2011).Anincrease in 5hmC signalswas observedingenesactivatedinoldmicewithrespecttoyoungones demonstratingthat5hmCisacquiredindevelopmentallyactivated genes(Szulwachetal.,2011).Age-dependentincreaseof5hmCin mousehippocampuswasalsoobservedonthe5-LOXgene,whose expressionisknowntoincreaseduringaging(Chenetal.,2012b).
Theinterestfor5hmCisnowgrowingenormouslyconsidering itsinvolvement not onlyin physiological states, suchas devel-opmentand aging,butalsoinpathologicalconditionsincluding cancer,autoimmuneandneurodegenerativedisorders(Calabrese etal.,2014;Chenetal.,2012b;Chengetal.,2014;Pfeiferetal., 2014;Putirietal.,2014;Villar-Menendezetal.,2013).