Age-dependent expression of DNMT1 and DNMT3B in PBMCs from a large European population enrolled in the MARK-AGE study

Age-dependen exp ession o DNMT1 and DNMT3B in PBMCs om a la ge Eu opean popula ion en olled in he MARK-AGE s udy Fabio Cicca one, 1,2,†,a Ma co Mala ol a, 3,a Robe a Calab ese, 1,2 Tiziana Guas a ie o, 1,2 Ma ia Giulia Bacalini, 4,5 Anna Reale, 1 Claudio F anceschi, 4,5 Mi iam Cap i, 4,5 An i He onen, 6 Mikko Hu me, 6 Bea ix G ubeck-Loebens ein, 7 Be nha d Kolle , 8 J€ u gen Be nha d , 9 Ch is iane Schӧn, 9 P. Eline Slagboom, 10 Oli ie Toussain , 11 Ewa Siko a, 12 E s a hios S. Gonos, 13 Nicolle B eusing, 14 Tilman G une, 15 Eug ene Jansen, 16 Ma ijn Doll e, 16 Ma  ıa Mo eno-Villanue a, 17 Thilo Sindlinge , 17 Alexande B€ u kle, 17 Michele Zampie i 1,2,b and Paola Caia a 1,2,b 1 Facul y o Pha macy and Medicine, Depa men o Cellula Bio echnologies and Hema ology, Sapienza Uni e si y o Rome, Rome 00161, I aly 2 Pas eu Ins i u e-Fondazione Cenci Bologne i, Rome 00161, I aly 3 Na ional Ins i u e o Heal h and Science on Aging (INRCA), Nu i ion and Ageing Cen e, Scien i ic and Technological Resea ch A ea, 60100 Ancona, I aly 4 Depa men o Expe imen al, Diagnos ic and Special y Medicine, Alma Ma e S udio um-Uni e si y o Bologna, Bologna 40126, I aly. 5 CIG-In e depa men al Cen e ‘L. Gal ani’, Alma Ma e S udio um, Uni e si y o Bologna, 40126 Bologna, I aly 6 The School o Medicine, The Uni e si y o Tampe e, 33014 Tampe e, Finland 7 Resea ch Ins i u e o Biomedical Aging Resea ch, Uni e si € a Innsb uck, Rennweg, 10, 6020 Innsb uck, Aus ia 8 Depa men o In e nal Medicine, Uni e si y Teaching Hospi al Hall in Ti ol, Milse s . 10, 6060 Hall in Ti ol, Aus ia 9 BioTeSys GmbH, 73728 Esslingen, Ge many 10 Depa men o Molecula Epidemiology, Leiden Uni e si y Medical Cen e, Leiden, The Ne he lands 11 Resea ch Uni on Cellula Biology, Uni e si y o Namu , Rue de B uxelles, 61, Namu B-5000, Belgium 12 Labo a o y o he Molecula Bases o Ageing, Nencki Ins i u e o Expe imen al Biology, Polish Academy o Sciences, 3 Pas eu S ee , 02-093 Wa saw, Poland 13 Na ional Hellenic Resea ch Founda ion, Ins i u e o Biology, Medicinal Chemis y and Bio echnology, A hens, G eece 14 Ins i u e o Nu i ional Medicine (180c), Uni e si y o Hohenheim, F uwi hs aße 12, 70599 S u ga , Ge many 15 Ge man Ins i u e o Human Nu i ion Po sdam-Rehb uecke (DI E), A hu - Scheune -Allee 114-116, 14558 Nu he al, Ge many 16 Cen e o Heal h P o ec ion, Na ional Ins i u e o Public Heal h and he En i onmen , PO Box 1, 3720 BA Bil ho en, The Ne he lands 17 Molecula Toxicology G oup, Depa men o Biology, Uni e si y o Kons anz, 78457 Kons anz, Ge many Summa y Aging is associa ed wi h al e a ions in he con en and pa e ns o DNA me hyla ion i ually h oughou he en i e human li espan. Reasons o hese a ia ions a e no well unde s ood. Howe e , se e al lines o e idence sugges ha he epigene ic ins abili y in aging may be aced back o he al e a ion o he exp ession o DNA me hyl ans e ases. He e, he associa ion o he exp ession o DNA me hyl ans e ases DNMT1 and DNMT3B wi h age has been analysed in he con ex o he MARK-AGE s udy, a la ge-scale c oss-sec ional s udy o he Eu opean gene al popula ion. Using pe iphe al blood mononuclea cells, we assessed he a ia ion o DNMT1 and DNMT3B gene exp ession in mo e han wo housand age-s a i ied women and men (35– 75 yea s) ec ui ed ac oss eigh Eu opean coun ies. Signi ican age- ela ed changes we e de ec ed o bo h ansc ip s. The le el o DNMT1 g adually d opped wi h aging bu his was only obse ed up o he age o 64 yea s. By con as , he exp ession o DNMT3B dec eased linea ly wi h inc easing age and his associ- a ion was pa icula ly e iden in emales. We nex a emp ed o ace he age- ela ed changes o bo h ansc ip s o he in luence o di e en a iables ha ha e an impac on changes o hei exp ession in he popula ion, including demog aphics, die a y and heal h habi s, and clinical pa ame e s. Ou esul s indica e ha age a ec s he exp ession o DNMT1 and DNMT3B as an almos independen a iable in espec o all o he a iables e alua ed. Key wo ds: aging; DNA me hyla ion; DNMT1;DNMT3B. In oduc ion Epigene ic p ocesses a e a molecula in e ace ha media es he in e ac ion be ween genome and en i onmen du ing he en i e li espan o o ganisms. Abe an epigene ic signalling, including DNA me hyla ion de ec s, plays a c ucial ole in aging (Zampie i e al., 2015). The unde s anding o he mechanisms behind hese e en s is an impo an esea ch opic as i can e eal he molecula mechanisms con ibu ing o age-associa ed physiological decline and disease. DNA me hyla ion is a modi ica ion o he genome ha occu s a e DNA eplica ion by he a achmen o a me hyl g oup o he cy osine o a CpG dinucleo ide. In mammals, 5-me hylcy osine (5mC) ep e- sen s an epigene ic modi ica ion o he genome ha ma ks ansc ip- ionally ep essed domains and se es as a he i able signal su icien o es o e silen ch oma in ollowing DNA eplica ion (Bi d, 2002). In he human genome, almos 80% o he CpGs a e me hyla ed in non andom ashion wi h he emaining unme hyla ed esidues p e - e en ially es ic ed o CpG islands (CGIs) wi hin gene p omo e s (Dea on & Bi d, 2011). This bimodal pa e n is es ablished du ing Co espondence Paola Caia a, Depa men o Cellula Bio echnologies and Hema ology, Sapienza Uni e si y o Rome, Viale Regina Elena 324, 00161-Rome, I aly. Tel.: +390649976530; ax: +390644231961; e-mail: [email p o ec ed] a Equal con ibu ion. b Sha ed senio au ho ship. † P esen add ess: Depa men o Biology, Uni e si y o Rome ‘To Ve ga a’, Via della Rice ca Scien i ica 1, 00133 Rome, I aly Accep ed o publica ion 21 Ma ch 2016 ª2016 The Au ho s. Aging Cell published by he Ana omical Socie y and John Wiley & Sons L d. This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s use, dis ibu ion and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed. 755 Aging Cell (2016) 15, pp755–765 Doi: 10.1111/acel.12485 Aging Cell de elopmen and di e en ia ion when issue-speci ic changes in DNA me hyla ion shape he epigene ic pa e ns o each indi idual cell ype. This accomplishes ansc ip ional ep ession o epe i i e DNA sequences and allows housekeeping genes o be exp essed, bu i also p o ides long- e m s abili y o he issue-speci ic gene exp ession p o ile o soma ic cells. DNA me hyla ion pa e ns a e dynamic s a es balanced by me hyla- ion and deme hyla ion p ocesses. The ‘main enance me hyl ans e ase’ DNMT1 mainly main ains he me hyla ion pa e ns ac oss eplica ion cycles while he de no o DNMT3A and DNMT3B enzymes mainly in oduce me hyl g oups on o DNA a si es ha had been unme hyla ed (Ju kowska e al., 2011). Rega ding DNA deme hyla ion, bo h passi e and ac i e p ocesses ha e been p oposed. Passi e deme hyla ion includes e e y hing ha nega i ely a ec s DNA me hyl ans e ase ac ion, hus leading o he loss o 5mC ma ks on he genome du ing DNA eplica ion. Howe e , e idence has been accumula ed indica ing he exis ence o eplica ion- independen ac i e DNA deme hyla ion in ol ing 5-hyd oxyme hylcy- osine o ma ion and DNA epai mechanisms (Sch€ ubele , 2015). Se e al s udies ha e clea ly demons a ed ha dis inc DNA me hy- la ion changes highly co ela e wi h aging h oughou he en i e li espan o humans and mice. Collec i ely, hese s udies ha e shown ha aging, simila o cance , is associa ed wi h g adual bu p o ound changes in DNA me hyla ion whe e he epigenome is ma ked by global hypome hy- la ion oge he wi h an opposi e p ocess o ocal hype me hyla ion p e e en ially a CGI p omo e s (Issa, 2014). These phenomena e ode he no mal genomic me hyla ion pa e ns leading o di e gen me hylomes in he no mal popula ion as a unc ion o ch onological age. Howe e , di ec ional changes in speci ic egions in aged indi iduals occu in he con ex o age-associa ed inc ease o DNA me hyla ion en opy and his p ocess is commonly de ined as ‘epigene ic d i ’ (F aga e al., 2005). Al hough he p esence o an abe an DNA me hyla ion signalling and i s ou come in aging p og ession a e well desc ibed, he unde lying molecula mechanisms a e a om being unde s ood. T ansc ip ional con ol o DNA me hyl ans e ases (DNMTs) can be esponsible o changes in p o ein le el o enzyma ic ac i i y bo h in physiological and in pa hological condi ions (Tsai e al., 2012; Calab ese e al., 2014). P e ious s udies indica e ha aging a ec s he exp ession o DNMTs, sugges ing his as being one o he mechanisms in ol ed in he de egula ion o he me hyla ion pa e ns obse ed in aging. Howe e , hese obse a ions de i e om e y ew s udies analysing ela i ely small sample sizes (Zhang e al., 2002; Casillas e al., 2003; Balada e al., 2008; Xiao e al., 2008; Li e al., 2010; Qian & Xu, 2014). He e, we epo esul s om an Eu opean la ge-scale c oss-sec ional s udy aimed a in es iga ing he associa ion o DNMTs exp ession wi h age as pa o he MARK-AGE p ojec . MARK-AGE is a Eu opean-wide popula ion s udy, suppo ed by he Eu opean Commission (FP7), aiming o disco e bioma ke s o aging which would se e as a eliable measu e o biological age (B€ u kle e al., 2015; Cap i e al., 2015). We measu ed he ansc ip le els o DNMT1 and DNMT3B genes and analysed hei a ia ion wi h age in pe iphe al blood mononuclea cells (PBMCs) om mo e han wo housand age-s a i ied dono s (35–75 yea s) om he gene al popula ion om eigh Eu opean coun ies (Cap i e al., 2015). Ou da a show ha age has an in luence on he exp ession o DNMT1 and DNMT3B genes in PBMCs. This in e ac ion does no signi ican ly depend on nu i ional, li es yle and clinical a iables in luencing he exp ession o DNMTs in he popula ion. Resul s Cha ac e is ics o he s udy popula ion The analysis o he ansc ip le els o DNMT1 and DNMT3B was ca ied ou on PBMCs om blood samples ob ained om dono s (2453 indi iduals) om eigh Eu opean coun ies (Table 1). The la ges g oup o samples consis ed o he RASIG (Randomly- Rec ui ed Age-S a i ied Indi iduals om he Gene al popula ion), ep esen ing indi iduals unde going a no mal aging p ocess. O sp ing o nonagena ians p e iously s udied in he GEHA s udy (Gene ics o Heal hy Ageing; F anceschi e al., 2007) and e med GO (‘GEHA o sp ing’) we e also ec ui ed as a po en ial model o ‘ e a ded aging’, based on he assump ion ha hei gene ic backg ound may p edispose hem o longe i y. As a con ol o en i onmen al ac o s and li es yle, ec ui men o GO was accompanied by he ec ui men o hei spouses, e med SGO (‘spouses o he GO’) (B€ u kle e al., 2015; Cap i e al., 2015). MARK-AGE subjec s co e ed he age ange be ween 35 and 75 yea s and we e s a i ied in o ou 10-yea age g oups. The age dis ibu ion o RASIG indi iduals was almos homogeneous, while GO and SGO indi iduals ell in o age anges >54 yea s. The composi ion o male and emale indi iduals was mainly compa able in all age g oups, albei emale ep esen a ion was a ew pe cen age poin s la ge han males. In ag eemen wi h li e a u e da a s udying subjec s in he age ange 35– 75 yea s, he body mass index (BMI) appea ed o inc ease wi h age indica ing ha he analysed popula ion was e ec i ely ep esen a i e o a physiological aging p ocess. The cha ac e is ics o he s udy popula ion e e ed o each ec ui ing cen e a e epo ed in Table S1 (Suppo ing in o ma ion). Iden i ica ion o ou lie s and dis ibu ion es s Due o posi i ely skewed alues, DNMT1 and DNMT3B exp ession da a ailed o pass he Kolmogo o –Smi no es o no mal dis ibu ion. We also iden i ied da a o DNMT1 mRNA ( h ee samples) wi h alues abo e 89 he in e qua ile ange, which we e conside ed ou lie s and excluded in all pa ame ic analyses (bu included in nonpa ame ic es s, o example Spea man co ela ion and K uskal–Wallis, which a e no sensi i e o ew ou lie s). A e emo al o ou lie s, he DNMT1 exp ession da a s ill ailed o pass no mali y es s. Hence, we analysed Q-Q plo s o he log- ans o med and no - ans o med a iables o DNMT1 and DNMT3B o es ablish he dis ibu ion ha bes ep esen ed ou da a. On he basis o ou analysis, he dis ibu ions ha be e ep esen ed da a we e gamma dis ibu ion o DNMT1 and no mal dis ibu ion o log- ans o med DNMT3B da a (Fig. S1, Suppo ing in o ma ion). Howe e , since he op imal dis ibu ion was no iden i ied, we pe o med bo h pa ame ic and nonpa ame ic es s o e i y he obus ness o esul s. Exp ession le els o DNMT1 and DNMT3B ansc ip s wi h espec o age and demog aphics In he RASIG samples, we iden i ied nonlinea bu signi ican changes o DNMT1 in age g oups by pa ame ic and nonpa ame ic es s. DNMT1 ansc ip dec eased g adually below 64 yea s o age, ansien ly achie ing a signi ican educ ion in he 55–64 s. he 35–44 age g oup be o e e u ning o le els compa able o he unde -54 in indi iduals o e he age o 65 yea s. Rega ding DNMT3B, i s exp ession appea ed o DNMT1 and DNMT3B exp ession in aging, F. Cicca one e al. 756 ª2016 The Au ho s. Aging Cell published by he Ana omical Socie y and John Wiley & Sons L d. dec ease linea ly wi h age, wi h a signi ican educ ion in he wo oldes age classes compa ed o he i s one (Table 2). G aphical ep esen a ion o DNMT1 and DNMT3B da a desc ibes be e he minimal, bu signi ican age- ela ed changes obse ed in he samples. In pa icula , DNMT3B displayed a signi ican linea dec ease wi h age, while DNMT1 displayed a U-shaped bimodal pa e n. This U-shaped bimodal pa e n o DNMT1 mRNA a ia ion wi h age was less e iden when he samples we e analysed as a whole (RASIG oge he wi h GO and SGO samples) due o a lack o signi icance o he eco e y o ansc ip le els in he oldes age g oup compa ed o younge age g oups. Rega ding DNMT3B, i s educ ion wi h inc easing age was independen o he inclusion o GO and SGO in he analysis (Fig. S2). Associa ion be ween age and DNMTs ansc ip le els in he RASIG popula ion was u he es ed by co ela ion analyses. DNMT3B showed a weak nega i e linea co ela ion wi h age. Howe e , his associa ion emained signi ican only in emales when he samples we e spli based on gende . The same analysis con i med ha he associa ion o DNMT1 wi h age was no linea when es ed o e he en i e age ange, which is in line wi h he U-shaped end o DNMT1 a ia ion wi h age. Consis en ly, a quad a ic ela ionship be ween DNMT1 and age showed a be e i o he da a han he linea one, which was, ins ead, signi ican and nega i e o bo h gende s when indi iduals o e he age o 65 yea s we e excluded om he analysis (Fig. 1). The le el o exp ession o bo h DNMT1 and DNMT3B was dissimila ac oss he di e en Eu opean coun ies (Table 2). Mo eo e , he exp ession le els ac oss age g oups wi hin each ec ui men cen e we e signi ican ly di e en o pa o hem p obably as a consequence o he high a iabili y, gende -speci ic e ec s and he ela i e lowe sample size o each cen e compa ed o he o e all sample (Table S2). Table 2 also shows ha bo h he DNMT1 and DNMT3B mRNA exp ession we e signi ican ly highe in emales compa ed o males. This associa ion wi h gende was e iden o DNMT1 only by he gene alized linea model (GLM) es ha includes age and coun y as co a ia es, sugges ing hese a iables as con ounding ac o s in he nonpa ame ic es . The same conclusions can be d awn o he posi i e associa ion obse ed be ween DNMT3B exp ession and BMI. In es iga ion o a iables (li es yle, die a y habi s, haema ological pa ame e s and ca dio ascula isk ac o s) ha po en ially a ec DNMT1 and DNMT3B mRNA exp ession F om all subjec s en olled in he MARK-AGE p ojec , a la ge numbe o an h opome ic and medical da a we e collec ed. In a p elimina y analysis, we e alua ed a iables ha migh ha e a po en ial impac on DNMT1 and DNMT3B exp ession, including die a y and li es yle habi s, haema ological pa ame e s, and ca dio ascula o diabe es isk ac o s. Se e al a iables we e ound o be associa ed wi h DNMTs exp ession by nonpa ame ic K uskal–Wallis (KW) es only. In ac , many o hese associa ions we e disp o ed o blun ed by GLM which included he e ec s o gende , ec ui men cen e and age as co a ia es (See Tables S3–S5). Such e idence indica es ha hose associa ions wi h DNMTs exp ession, epo ed by KW analysis, we e indi ec . Examples o such disco dance include he associa ions o DNMT1 exp ession wi h F ench ies and b own b ead consump ion o wi h glucose and homocys eine blood le els. Based on he cong uence be ween KW and GLM es s, a limi ed se o a iables appea ed o be consis en ly associa ed wi h he exp ession o DNMTs. Apa om a link be ween weekly consump ion o whi e b ead and DNMT1 mRNA le els, no associa ion wi h o he sel - epo ed die a y habi s was de ec ed (Table S3). This also conce ns consump ion o alcoholic be e ages and smoking habi s as well as ca dio ascula and diabe es isk bioma ke s (Table S4). As a as haema ological pa ame e s a e conce ned (Table S5), DNMT1 and DNMT3B mRNAs we e less exp essed wi h inc easing monocy e coun s. Ins ead, only DNMT1 was Table 1 Cha ac e is ics o he s udy popula ion by age g oups† Age g oups All 1 2 3 4 P Age ange (yea s) 35–75 35–44 45–54 55–64 65–75 N2453 383 484 807 779 Age (yea s) 57.6 10.7 39.6 2.8 49.6 2.8 59.9 2.8 69.0 3.0 <0.001 Male % (n) 47 (1145) 46 (177) 47 (226) 45 (366) 48 (376) 0.707 Smoke , ne e 52 (1264) 59 (227) 48 (233) 49 (393) 53 (401) <0.001 Fo me 34 (826) 17 (66) 33 (162) 37 (297) 39 (301) Cu en 15 (363) 23 (90) 18 (89) 14 (117) 9 (67) G oup, GO 18 (437) 0 (1) 3 (17) 25 (205) 28 (214) <0.001 RASIG 72 (1774) 99 (381) 93 (449) 60 (483) 59 (461) SGO 10 (242) 0 (1) 4 (18) 15 (119) 13 (104) BMI (kg/m 2 ) 26.3 4.5 24.9 4.6 25.6 4.4 26.7 4.5 27.0 4.3 <0.001 <25 43 (1059) 59 (226) 50 (244) 39 (317) 35 (272) 25 o <30 39 (957) 28 (109) 35 (171) 41 (331) 44 (346) <0.001 ≥30 18 (436) 12 (48) 14 (69) 20 (159) 21 (160) Finland 10 (249) 2 (9) 4 (18) 13 (109) 14 (113) I aly 20 (485) 26 (98) 22 (105) 18 (144) 18 (138) Aus ia 11 (267) 18 (69) 14 (69) 9 (69) 8 (60) G eece 13 (317) 18 (71) 16 (78) 11 (86) 10 (82) <0.001 Poland 11 (270) 12 (46) 11 (52) 12 (99) 9 (73) The Ne he lands 8 (189) 0 (0) 1 (3) 10 (79) 14 (107) Belgium 14 (350) 8 (30) 14 (69) 16 (131) 15 (120) Ge many 13 (326) 16 (60) 18 (90) 11 (90) 11 (86) †Values a e mean SD and pe cen age (numbe ), all such a iables; one missing case o BMI; P- alue: one-way ANOVA (con inuous a iables) and chi-squa e es (p e alence). De ini ion o abb e ia ions is p o ided in he Da a S2. DNMT1 and DNMT3B exp ession in aging, F. Cicca one e al. 757 ª2016 The Au ho s. Aging Cell published by he Ana omical Socie y and John Wiley & Sons L d. exp essed mo e in samples cha ac e ized by an inc eased numbe o lymphocy es and o highe lymphocy es/monocy es a io. This was e en mo e e iden when looking a he T lymphocy e (CD3+CD45+) con en , which also showed a posi i e associa ion wi h DNMT3B.DNMT1 was also less exp essed in samples cha ac e ized by an inc eased amoun o neu ophils, while DNMT3B showed an inc easing end wi h inc easing numbe o pla ele s. O no e, he ela ionship o DNMTs exp ession le els in PBMCs wi h lymphocy es and monocy es blood coun s mos ly esembles he highe exp ession o hese genes in lymphocy es sub ypes wi h espec o monocy es, as e ealed by he analysis o publicly a ailable mic oa ay exp ession da a (Su e al., 2004) (Fig. S3). Simila esul s we e ob ained by analysing he whole popula ion o samples (RASIG oge he wi h GO and SGO samples) (Table S6). Con ibu ion o selec ed a iables on age- ela ed changes o DNMT1 and DNMT3B mRNA exp ession in he RASIG popula ion GLM and eg ession analyses we e used o de e mine whe he he age- ela ed changes obse ed o DNMT1 and DNMT3B genes a e an in insic ea u e o he aging p ocess o whe he one o mo e o he a iables shown o in luence hei exp ession can jus i y hese changes. As shown in Table 3, he changes o DNMT1 exp ession in age g oups seem o depend on a combined impac o mul iple pa ame e s a he han on a speci ic one. Howe e , he age- ela ed a ia ions o DNMT1 mRNA le els emained signi ican e en when conside ing all he a iables ha signi ican ly a ec ed DNMT1 exp ession (i.e. showing a signi ican in e ac ion wi h DNMT1 exp ession as de ec ed by bo h KW and GLM es s, see Table 2 and Tables S3–S5), indica ing ha he analysed pa ame e s canno ully explain he DNMT1 ansc ip di e ences obse ed be ween age g oups. To con i m he esul , se e al addi ional models we e un wi h p og essi e inclusion o a iables ca ego ized on he basis o any signi ican associa ion wi h DNMT1 exp ession, also including only hose iden i ied by KW es (Tables S7A and B). Fu he analysis using he log- ans o med alues o DNMT1 yielded compa able da a (Table S8A and B). Simila esul s conce ned he age- ela ed a ia ion o DNMT3B ansc ip . I s linea dec ease wi h age (Table 2 and Fig. 1) allowed he in es iga ion by eg ession analysis. Taking in o accoun he gende di e ence iden i ied in Fig. 1, we pe o med a sepa a e analysis o gende . The analysis con i med he linea dec ease o DNMT3B mRNA le els in emales (bu no males). The linea dec ease o DNMT3B in he emales was no a ec ed by he inclusion o o he a iables associa ed wi h di e ences in DNMT3B mRNA le els be ween indi iduals (Table 4). Iden i ica ion o majo a iables a ec ing he measu emen o DNMT1 and DNMT3B exp ession Decision ee analysis including all a iables wi h a leas one signi ican e ec on DNMT1 and DNMT3B (see Table 2 and Tables S3–S5) was pe o med o iden i y he majo ac o s a ec ing DNMTs exp ession. We iden i ied he ec ui men cen e and he lymphocy e- o-monocy e a io Table 2 E ec o age, gende , BMI on DNMT1 and DNMT3B exp ession in he RASIG popula ion S a N DNMT1 DNMT3B Median (IQ) P(KW)†P(GLM)‡Median (IQ) P(KW)†P(GLM)‡ Age g oup (yea s) 35–44 a 381 0.134 (0.102–0.171) c 0.003 0.049 0.022 (0.017–0.029) c,d 0.004 0.030 45–54 b 449 0.129 (0.095–0.159) 0.021 (0.016–0.029) 55–64 c 483 0.117 (0.088–0.165) a,d 0.020 (0.015–0.027) a 65–75 d 461 0.129 (0.095–0.171) c 0.020 (0.015–0.026) a Cen e Finland a 80 0.125 (0.092–0.154) b <0.001 <0.001 0.024 (0.018–0.031) g <0.001 <0.001 I aly b 362 0.144 (0.113–0.183) a,c,d,g,h 0.020 (0.015–0.025) c Aus ia c 267 0.113 (0.080–0.165) b,e 0.023 (0.015–0.035) b.g G eece d 296 0.129 (0.095–0.171) b,g 0.021 (0.016–0.031) g Poland e 202 0.139 (0.105–0.177) c,g 0.022 (0.016–0.028) g The Ne he lands 0 –– Belgium g 241 0.109 (0.077–0.139) b,d,e,h 0.017 (0.014–0.024) a,c,d,e Ge many h 326 0.129 (0.102–0.159) b,g 0.020 (0.016–0.027) Gende F a 940 0.129 (0.098–0.165) 0.056 0.031 0.021 (0.016–0.029) b 0.001 0.002 M b 834 0.125 (0.092–0.165) 0.020 (0.015–0.027) a BMI classes <25 a 810 0.125 (0.095–0.165) 0.047 0.377 0.020 (0.015–0.028) 0.078 0.019 25 o <30 b 662 0.125 (0.092–0.165) 0.021 (0.016–0.027) ≥30 c 301 0.139 (0.102–0.171) 0.022 (0.016–0.028) †KW es : nonpa ame ic compa ison by he K uskal–Wallis es o DNMT1 and DNMT3B mRNAs le els (da a o wo g oup compa ison a e analysed wi h he Mann– Whi ney U- es ); da a a e epo ed as median and in e qua ile ange (IQ). Pai wise compa isons a e e e ed o he KW es and adjus ed o mul iple compa isons (compa isons wi h P<0.05 a e ma ked by he associa ed supe sc ip s). ‡GLM: compa ison by gene alized linea models o DNMT1 (gamma dis ibu ion wi h log-link unc ion model) and DNMT3B (linea model wi h log- ans o med alues and iden i y link unc ion) mRNAs le els. Fo he in es iga ion o age-g oup e ec s, he model included he e ec s o gende and ec ui men cen e as ac o s. All o he GLM models included he e ec s o gende , ec ui men cen e and age (con inuous a iable) as co a ia e (only P alues o he selec ed a iables a e shown). De ini ion o abb e ia ions is p o ided in he Da a S2. DNMT1 and DNMT3B exp ession in aging, F. Cicca one e al. 758 ª2016 The Au ho s. Aging Cell published by he Ana omical Socie y and John Wiley & Sons L d. as being he mos impo an ac o s ha a ec bo h DNMT1 (Fig. S4) and DNMT3B (Fig. S5). Mino subg oups we e iden i ied on he basis o age, WBC and CD3+CD45+cell da a. Di e ences in DNMT1 mRNA exp ession be ween GO, SGO and RASIG To in es iga e he possible in luence o gene ic ad an age expec ed o GO samples, he analysis o DNMT1 and DNMT3B exp ession was pe o med compa ing samples wi h an age be ween 54 and 75 yea s as GO and SGO we e well ep esen ed in his age in e al. The s a i ica ion by 10-yea age g oups o subjec s aged >54 yea s led o a simila di ision o GO, SGO and RASIG ac oss he age g oups, hus educing po en ial bias due o unequal di isions. As shown in Fig. 2A, he exp ession o DNMT1 was signi ican ly highe in RASIG compa ed o GO and SGO, while no di e ence was obse ed be ween GO and SGO. The compa ison o DNMT1 exp ession be ween age-s a i ied GO, SGO and RASIG samples was pe o med acco ding o gende and ec ui men cen e. Da a show ha he highe le el o DNMT1 mRNA in RASIG s. GO and SGO seemed o be a common ea u e, al hough he s a is ical signi icance was eached in some age/cen e subg oups, likely due o educed sample size (Table S9). In con as , he exp ession le els o Fig. 1 Age- ela ed changes o DNMT1 and DNMT3B mRNA le els wi h age in emales and males om RASIG popula ion. The pic u e shows a g aphical ep esen a ion o DNMT1 and DNMT3B mRNA as log- ans o med da a s. age in he RASIG sample. (A1) Do plo o log- ans o med DNMT3B da a s. age in RASIG emales; (A2) do plo o log- ans o med DNMT3B da a s. age in RASIG males; (A3) do plo o log- ans o med DNMT3B da a s. age in all RASIG samples; (B1) do plo o log- ans o med DNMT1 da a s. age in RASIG emales; (B2) do plo o log- ans o med DNMT1 da a s. age in RASIG males; (B3) do plo o log- ans o med DNMT1 da a s. age in all RASIG samples. Co ela ion coe icien s (Pea son R o log- ans o med da a and Spea man’s ho o un ans o med da a) a e no ed below each g aph. Signi icance and 95% con idence in e al o co ela ion coe icien s (wi hin b acke s) a e es ima ed by bias-co ec ed and accele a ed (BCa) boo s ap wi h s a i ied sampling (1000 samples s a i ied o coun y, and also o gende when all da a a e used). R-squa e and signi icance o he unique ele an quad a ic eg ession is no ed abo e panel B3. G aph and eg ession coe icien s a e also epo ed conside ing da a o DNMT1 in he age ange om 35 o 64 yea s in RASIG emale (panel C1), in RASIG males (panel C2) as well as in all RASIG popula ion (panel C3). DNMT1 and DNMT3B exp ession in aging, F. Cicca one e al. 759 ª2016 The Au ho s. Aging Cell published by he Ana omical Socie y and John Wiley & Sons L d. DNMT3B we e e y simila in he h ee classes o samples (Fig. 2B and Table S10). Wi h he aim o ace po en ial causes o he di e ences in DNMT1 mRNA exp ession be ween GO, SGO and RASIG, he con ibu ion o selec ed a iables was assessed by GLM analysis (Table 5). Resul s show ha he di e ences be ween he h ee classes o samples emained highly signi ican e en when se e al a iables ha associa ed wi h in e indi idual a ia ions in DNMT1 mRNA exp ession we e aken in o accoun . Addi ional models ha con i med hese esul s we e also un including all ca ego ized a iables associa ed wi h DNMT1 in Tables S3– S5 (Table S11A and 11B). The e ec o ba ch co ec ion s a egies on he associa ion o DNMT1 and DNMT3B mRNA exp ession wi h age and on di e ences in DNMT1 mRNA exp ession be ween GO, SGO and RASIG Ou expe imen al design has been concei ed o minimize he impac o ba ch e ec s by he adop ion o a alida ed housekeeping gene o no maliza ion o exp ession da a in aging (Zampie i e al., 2010) as well as he use o a calib a o sample o educe in e - un a ia ion. Clea ly, he p esence o addi ional ba ch e ec s could no be comple ely uled ou hough. In an a emp o add ess his p oblem, he age- ela ed changes in RASIG and he di e ences be ween GO, SGO and RASIG o DNMTs exp ession we e es ed agains di e se p ocedu es o ba ch e ec s adjus men by he applica ion o a dedica ed so wa e. The ollowing s a egies we e adop ed: (i) simple co ec ion o ba ch e ec s; (ii) co ec ion a emp ing o e ain di e ences be ween g oups (GO, SGO and RASIG), age and gende ; (iii) g oup-, age- and gende -sensi i e co ec ion combined wi h il e ing o e en ual ou lie s due o o e co - ec ion. As shown in he Fig. S6, he applica ion o he co ec ion p ocedu es g adually no malized he dis ibu ion o da a o bo h DNMTs. Conce ning DNMT1, co ec ion p ocedu es unco e ed a weak albei signi ican linea decline wi h age (Fig. S7). The s a i ica ion o da a o age and gende e ealed ha adjus men o ba ches mainly a ec ed he up egula ion o DNMT1 in he las age class ob ained wi h unadjus ed da a, while a dec easing end cha ac e ized bo h unco ec ed and co ec ed da a in age g oups up o 64 yea s (Fig. S8). Consis en ly, a signi ican nega i e linea co ela ion o DNMT1 exp ession wi h age was Table 3 In luence o selec ed ac o s and co a ia es on age- ela ed changes o DNMT1 exp ession† Va iables Type III Wald chi-squa e d Sig. Age g oups 8.760 3 0.033 Gende 0.826 1 0.363 Cen e 53.167 6 <0.001 Whi e b ead consump ion 6.877 2 0.032 Monocy es 0.123 1 0.726 Ra io lymphocy e o monocy e 19.989 1 <0.001 Neu ophils 0.240 1 0.624 Model: age g oups (o), gende (n), cen e (n), whi e b ead consump ion (s), monocy es (s), lymphocy es/monocy es (s), neu ophils (s), (o=o dinal a iable: n=nominal a iable; s=scale a iable). †Analysis was pe o med by GLM using gamma dis ibu ion wi h log-link unc ion and non ans o med da a o he dependen a iable: DNMT1 mRNA (s). Table 4 Reg ession analysis o DNMT3B exp ession in emales and males om RASIG† Gende Va iables Coe icien s Boo s ap o coe icien s BSE Be a Bias Sig 95% CI Model 1 F Age (yea s) 0.004 0.001 0.086 0.000 0.008 0.007; 0.001 M Age (yea s) 0.003 0.002 0.059 0.000 0.077 0.006; 0.000 Model 2 F Age (yea s) 0.004 0.001 0.084 0.000 0.009 0.007; 0.001 Lymphocy es/monocy es 0.293 0.100 0.087 0.004 0.005 0.083; 0.466 M Age (yea s) 0.002 0.002 0.042 0.000 0.310 0.005; 0.001 Lymphocy es/monocy es 0.190 0.121 0.056 0.001 0.087 0.035; 0.408 Model 3 F Age (yea s) 0.006 0.002 0.128 0.000 0.001 0.009; 0.002 Lymphocy es/monocy es 0.271 0.131 0.079 0.011 0.021 0.041; 0.525 BMI 0.481 0.111 0.167 0.007 0.001 0.280; 0.671 HDL 0.148 0.073 0.075 0.003 0.047 0.019; 0.278 Monocy es 0.045 0.036 0.048 0.001 0.068 0.102; 0.006 MHC 0.140 0.361 0.013 0.176 0.688 1.417; 0.238 Pla ele s 0.065 0.052 0.044 0.013 0.385 0.173; 0.133 HGB 0.062 0.234 0.009 0.018 0.801 0.527; 0.445 M Age (yea s) 0.002 0.002 0.038 0.000 0.302 0.005; 0.002 Lymphocy es/monocy es 0.223 0.152 0.062 0.007 0.145 0.092; 0.524 BMI 0.206 0.163 0.052 0.005 0.161 0.077; 0.507 HDL 0.038 0.088 0.018 0.001 0.690 0.127; 0.222 Monocy es 0.015 0.048 0.013 0.001 0.685 0.083; 0.065 MHC 0.128 0.190 0.025 0.250 0.359 1.594; 0.026 Pla ele s 0.087 0.059 0.055 0.000 0.109 0.037; 0.215 HGB 0.169 0.262 0.024 0.021 0.482 0.745; 0.407 †Reg ession analysis was pe o med by using log- ans o med da a o dependen and independen a iables (wi h he exclusion o age). All da a we e included as con inuous a iables. Boo s ap esul s a e based on 1000 s a i ied (by ec ui men cen e) boo s ap samples. De ini ion o abb e ia ions is p o ided in he Da a S2. DNMT1 and DNMT3B exp ession in aging, F. Cicca one e al. 760 ª2016 The Au ho s. Aging Cell published by he Ana omical Socie y and John Wiley & Sons L d. ound in he samples up o 64 yea s independen ly o co ec ion o ba ches (Fig. S9). The ba ch co ec ion p ocedu e had no ele an impac on he di e ences be ween GO, SGO and RASIG g oups e en when conside ing he majo con ounding ac o s in he analysis (Fig. S10). As DNMT3B conce ns, ba ch co ec ion con i med and s eng hened he dec emen o exp ession wi h age (Figs S11 and S12). Discussion The mechanisms esponsible o he coexis ence o global hypome hy- la ion and hype me hyla ion o speci ic sequences in he aging genome a e s ill an open esea ch ield. In his con ex , he al e ed exp ession o DNMTs has long been pos ula ed o con ibu e o epigene ic ins abili y in aging. This possibili y was mainly p omp ed by he ac ha changes o DNA me hyla ion pa e ns and o DNMTs exp ession a e well docu- men ed in aging-associa ed diseases such as cance , au oimmune diseases and Alzheime ’s disease. The esul s o pionee ing s udies, ca ied ou on cellula models o aging, ha e shown a subs an ial change o DNMTs ac i i y wi h inc easing age due o a combina ion o main enance me hyla ion de ici and inc eased de no o me hyla ion (Lopa ina e al., 2002), e en s which we e hen associa ed wi h educed DNMT1 le els oge he wi h inc eased DNMT3B, espec i ely (Casillas e al., 2003). In pa , hese ini ial obse a ions ha e been u he suppo ed by he esul s o in i o s udies in humans and mice. In ac , he le els o DNMT1 exp ession dec ease wi h age in human T lymphocy es and his is associa ed wi h hypome hyla ion o speci ic gene p omo e s (Zhang e al., 2002; Balada e al., 2008; Li e al., 2010). By con as , da a on changes o DNMT3B exp ession du ing aging a e less conco dan since, while i s inc ease was con i med in he li e o aged humans (Xiao e al., 2008), a dec ease wi h age was obse ed in human T lymphocy es (Balada e al., 2008) as well as in mouse skin (Qian & Xu, 2014). Collec i ely, hese da a indica e ha a ansc ip ional de egula ion o DNMTs would p obably accompany he aging p ocess. Howe e , he a ailable da a on humans a e de i ed om a limi ed numbe o co ela ional s udies gene ally ca ied ou on ela i ely small sample sizes. In he p esen wo k, he possible ela ionship be ween aging and exp ession o DNMT1 and DNMT3B has been es ed in he con ex o a la ge-scale popula ion-based s udy he eby p o iding, o he i s ime, a e e ence amewo k o ac o s ha a e associa ed wi h DNMTs exp ession a ia ion such as demog aphics, clinical labo a o y pa am- e e s as well as die a y and heal h habi s. Using PBMCs, one o he ew accessible issues, we b oadly assessed a ia ion o DNMT1 and DNMT3B ansc ip le els in mo e han wo housand indi iduals Fig. 2 Le els o DNMT1 and DNMT3B mRNA in GO, SGO and RASIG. The pic u e shows a g aphical ep esen a ion o DNMT1 (A) and DNMT3B (B) mRNA le el in PBMCs om GO, SGO and RASIG om he whole MARK-AGE sample abo e 54 yea s. Analysis was pe o med in subjec s abo e 54 yea s due o non ep esen a i e numbe so GO and SGO below his age. **P<0.01 om RASIG by pos hoc (LSD) o GLM analysis and by KW es pe o med wi hin each coun y. Table 5 Con ibu ion o selec ed a iables and co a ia es on g oup (GO, RASIG and SGO) ela ed changes o DNMT1 exp ession in PBMCs om popula ion aged >54 yea s† Tes s o model e ec s Sou ce Type III Wald chi-squa e d Sig. G oup (GO, SGO, RASIG) 15.707 2 <0.001 Rec ui men cen e 78.056 7 <0.001 Gende 0.041 1 0.839 Age (yea s) 4.107 1 0.043 BMI 2.565 1 0.109 Se um glucose 0.079 1 0.778 Glycosyla ed haemoglobin A1C 0.015 1 0.903 Homocys eine 4.236 1 0.040 Neu ophils 1.372 1 0.241 Lymphocy es/monocy es 19.814 1 <0.001 Monocy es 0.538 1 0.463 WBC 1.179 1 0.278 HCT 0.925 1 0.336 MCV 0.635 1 0.426 MCH 0.064 1 0.801 †Analysis was pe o med by GLM using gamma dis ibu ion wi h log-link unc ion. Dependen a iable: DNMT1 mRNA. Model: g oup (n), ec ui men cen e (n), gende (n), age (s), BMI (s), se um glucose (s), glycosyla ed haemoglobin A1C (s), homocys eine (s), neu ophils (s), lymphocy es/monocy es (s), monocy es (s), WBC (s), HCT (s), MCV (s), MCH (s); (n=nominal a iable; s=scale a iable). De ini ion o abb e ia ions is p o ided in he Da a S2. DNMT1 and DNMT3B exp ession in aging, F. Cicca one e al. 761 ª2016 The Au ho s. Aging Cell published by he Ana omical Socie y and John Wiley & Sons L d. ec ui ed ac oss he Eu opean popula ion co e ing he age ange o 35–75 yea s. Al hough no subs an ial co ela ion was obse ed be ween DNMT1 exp ession and age by linea eg ession analysis, a signi ican , bu e y small a ia ion o he exp ession o bo h ansc ip s, was ound when samples we e s a i ied in o 10-yea age g oups. In ag eemen wi h p e ious epo s (Zhang e al., 2002; Balada e al., 2008; Li e al., 2010), he le el o DNMT1 g adually d opped wi h aging. Howe e , his was obse ed up o 64 yea s, whe e i ei he appea ed o s abilize (ba ch co ec ed da a) o o e en ually aise again (unco ec ed da a) a e wa ds. Anyway, he U-shaped pa e n o he unco ec ed DNMT1 exp ession da a wi h age in humans is also suppo ed by obse a ions in a ecen publica ion, whe e he same end was obse ed o he exp ession o DNMT1 as well as o he global me hyla ion le el o he genome in mouse li e du ing aging (A ms ong e al., 2013). In e es ingly, he dec ease wi h age o he me hyla ion le els o epe i i e DNA sequences is no linea in humans ei he , bu i mainly occu s du ing ages 40–60 yea s (Jin a id h & Mu i angu a, 2010). Conce ning DNMT3B, i s ansc ip le els dec eased wi h age in ag eemen wi h p e ious obse a ions in human T lymphocy es (Balada e al., 2008). Looking a gende , he in e se co ela ion o DNMT3B exp ession wi h age was signi ican in he emale popula ion. No ably, he changes wi h age o bo h DNMTs wi hin each ec ui men cen e we e no uni o m. While his could be in pa a ibu ed o he educed sample size, a di e en ial impac o en i on- men al a iables on he ela ionship be ween exp ession o DNMTs and age ac oss di e en coun ies canno be excluded. Signi ican ly, en i onmen al a iables ha e been desc ibed o ha e an impac on genomic DNA me hyla ion pa e ns as well as on DNMTs exp ession le el (see Zampie i e al., 2015 o e iew). We nex pe o med an analysis o examine he associa ion o DNMTs exp ession wi h a la ge se o a iables including demog aphics, clinical labo a o y pa ame e s, die a y and li es yle habi s. Conce ning demog aphic a iables, an in e ac ion wi h he exp ession o DNMTs was ound o gende and BMI in addi ion o age and coun y. Gende di e ences in he exp ession o bo h DNMTs in PBMCs a e in line wi h da a p e iously ob ained in he li e whe e highe le els o bo h DNMT1 and DNMT3B we e obse ed in emales (Xiao e al., 2008). Fu he mo e, gende - ela ed di e ences o DNA me hyla ion pa e ns in PBMCs ha e ecen ly been desc ibed (Lam e al., 2012). Howe e , ou da a sugges ha gende di e ences o DNMT1 a e he consequences o a di e en p opo ion in leucocy es subse s. The posi i e associa ion o DNMT3B exp ession wi h BMI is o g ea in e es . In ac , an associa ion be ween BMI and epigene ic age accele a ion was ecen ly obse ed, especially in li e (Ho a h e al., 2014). In his con ex , ou esul s would sugges an in e es ing link be ween obesi y, al e ed DNMT3B exp ession and me hyla ion de ec s ha p edispose o disease. Consis en ly, enhanced DNMT3B exp ession was p oposed o con ibu e o de egula ed adipose issue mac ophage pola iza ion, in lamma ion and insulin esis ance in obesi y (Yang e al., 2014). Su p isingly, no signi ican associa ion be ween die a y habi s and le el o DNMTs ansc ip s was ound al hough inc easing e idence indica es DNA me hyla ion is ulne able o nu i ional in luences (Bacalini e al., 2014). This was also he case o clinical chemis y pa ame e s associa ed wi h ca dio ascula and diabe es isk. Fo some o hem, associa ion wi h DNMTs exp ession was ound o be indi ec and pa ially explained by o he demog aphic ac o s including, bu no exclusi ely, age. The lack o associa ion be ween smoking and le el o DNMTs was unexpec ed conside ing ha ciga e e smoking is one o he mos powe ul en i onmen al modi ie s o he DNA me hyla ion pa e n (B ei ling e al., 2011; Lee & Pauso a, 2013) and has been shown o de egula e he exp ession o DNMTs in b ain and lung (Sa a e al., 2008; Lin e al., 2010). This disc epancy could be explained by he ac ha his e ec could be issue-speci ic and does no conce n he PBMCs. In ac , a lack o s a is ical associa ion was epo ed be ween smoking and genomewide DNA me hyla ion a ia ion in PBMCs (Lam e al., 2012) as well as be ween smoking and epigene ic age accele a ion in mul iple issues (Ho a h e al., 2014). Finally, he exp ession o DNMTs was in luenced by he amoun o lymphocy es, monocy es and by a speci ic subse o T lymphocy es. This poin s o he composi ion o PBMCs as a con ounding ac o ha may lead o di e ences in exp ession o DNMTs be ween indi iduals. Hence, he s udy o he ela ionship be ween abno mal me hyla ion machine y and DNA me hyla ion changes in blood cells should ake in o accoun di e ences in leucocy e composi ion be ween indi iduals. In pa icula , he lymphocy e- o-monocy e a io eme ged as he main a iable a ec ing DNMT1 and DNMT3B. These da a also sugges ha DNMTs migh be mo e exp essed in lymphocy es han monocy es. In e es ingly, i seems o be he case acco ding o he compa ison be ween blood cell ypes o DNMT1 and DNMT3B ansc ip le els ob ained om published mic oa ay da a. A e he iden i ica ion o a iables po en ially in ol ed in de e mining a ia ions in he exp ession o DNMTs in ou popula ion, we hen sough o de e mine he impac o hese a iables on he di e ences in he exp ession o DNMTs be ween age g oups. The esul s indica e ha , al hough all he c i ical a iables we e aken in o accoun in he analysis, he di e ences in he exp ession o bo h DNMT1 and DNMT3B be ween he age g oups a e s ill signi ican (excluding DNMT3B in he male RASIG popula ion). This indica es ha age a ec s he exp ession o DNMTs in PBMCs as an almos independen a iable wi h espec o all o he a iables e alua ed he e. Ne e heless, ac o s such as geog aphical o igin o he samples and lymphocy es- o-monocy es a io seemed o ha e a g ea e in luence on in e indi idual di e ences in DNMTs exp ession wi h espec o age in PBMCs. On he o he hand, he weak associa ion o DNMTs exp ession wi h age se e ely limi s he possibili y o p edic ing age by measu ing DNMTs ansc ip s in PBMCs in con as o ha shown o he me hyla ion s a us o speci ic genomic loci (Ho a h, 2013). The ela ionship be ween DNMTs exp ession and aging was also e alua ed in he GO and SGO popula ions. GO ep esen s a g oup o indi iduals ha is assumed o ha e gene ic bene i s o heal hy aging compa ed o he no mal aging popula ion (RASIG) and o hei con ol o en i onmen and li es yle (SGO). In his expe imen al se ing, inhe en ad an ages o GO wi h espec o aging seem no o be ela ed o DNMTs exp ession. In ac , DNMT3B le els we e equal be ween he h ee g oups while he exp ession o DNMT1 is e en lowe in GO wi h espec o RASIG. Mo eo e , GO and SGO showed compa able exp ession o DNMT1. This ules ou he possibili y ha he di e ences in DNMT1 exp ession be ween GO and RASIG a e due o gene ics while i poin s o he exis ence o en i onmen al o li es yle ac o s ha dis inguish GO and SGO om RASIG. Howe e , di e ences in exp ession o DNMT1 be ween GO, RASIG and SGO do no seem o depend on any o he a iables which we ound o be associa ed wi h he a ia ion o i s exp ession in he analysed popula ion. This sugges s ha GO and SGO sha e a simila en i onmen al ac o whose in luence on he exp ession o DNMT1 canno be aced, a leas in any o he pa ame e s analysed he e. A specula i e hypo hesis could be ha he poin o con ac be ween GO and SGO is hei mic obiome (including bac e ia, ungi and i uses ha colonize ou o ganism). In ac , he e is e idence ha DNMT1 and DNMT3B exp ession in aging, F. Cicca one e al. 762 ª2016 The Au ho s. Aging Cell published by he Ana omical Socie y and John Wiley & Sons L d. cohabi a ion leads o mic obio a simila i ies be ween indi iduals (Song e al., 2013) and ha he cen ena ians ha e a e y di e en mic obio a composi ion om he es o he popula ion (Biagi e al., 2010; Rampelli e al., 2013). No ably, di e ences in in es inal mic obio a be ween indi iduals ha e ecen ly been associa ed wi h di e ences in he pa e n o DNA me hyla ion in he blood (Kuma e al., 2014). Finally, gi en ha he mos signi ican age- ela ed esul s o his pape o igina ed om e y small e ec sizes, we nex sough o de e mine whe he he aging e ec on DNMTs exp ession was obus o ba ch e ec s. I is expec ed ha ba ch e ec s would lead o inc eased a iabili y and dec eased powe o de ec a eal biological signal. In ac , in many cases, ba ch e ec s inc ease da a a iabili y and can be con used wi h an ou come o in e es hus leading o misleading biological conclusions. This claims he necessi y o pe o m dedica ed adjus men s especially in popula ion s udies ha equi e la ge sample sizes and ha e o be ca ied o e a long ime pe iod, as in he case o his s udy. Howe e , he applica ion o ba ch emo al ools o ou da a, which consis ed o he e ogeneous ( o gende , g oup and coun y) small ba ches, was likely o in oduce addi ional p oblems. In ac , in he case o s udy g oups being none enly dis ibu ed ac oss ba ches, algo i hms o emo e ba ch e ec s may bias g oup di e ences, bu also ca y he po en ial conce n o emo ing in ag oup biological he e ogenei y. In o de o a oid he pi all de i ing om inapp op ia e ba ch co ec ion, di e se app oaches ha e been applied. All adop ed p ocedu es con- i med he ela ionship be ween exp ession o DNMTs and age. Howe e , ba ch co ec ion impac ed bo h he pa e n and he s eng h o his associa ion. The linea decline o DNMT3B wi h age was con i med and ein o ced, as well as he di e ences in DNMT1 exp ession be ween GO, SGO and RASIG, al hough wi h a lowe s a is ical signi icance. Con e sely, emo al o ba ch e ec s con i med he decline o DNMT1 exp ession up o 65 yea s while blun ed i s up egula ion in he las age class ob ained wi h unadjus ed da a, hus disp o ing he U-shaped pa e n. Howe e , ba ch co ec ion appea ed o in oduce a possible bias as shown by he p e e en ial impac on he exp ession le els o DNMT1 in he 65–75 age class as well as by he in e sion o he mean alues be ween male and emales in he same age class he o bo h DNMTs. This sugges s ha he esul s a e co ec ion should be in e p e ed wi h cau ion. Collec i ely, esul s om his s udy con i m in la ge-scale popula ion se ing ha aging has an impac on he exp ession o DNMTs. Con e ging e idence is gi en o he linea dec ease o DNMT3B exp ession wi h age as well as o he DNMT1 up o 64 yea s. Fo highe age, da a on he end o DNMT1 we e con lic ing and a possible eading could be ha he dec ease o DNMT1 a hese ages is a enua ed. These da a would o m he basis o u u e in es iga ions aimed a es ablishing i hese changes a e causally linked o a ia ion in DNA me hyla ion pa e s and pa icipa e o he mechanism o DNA me hyla ion changes du ing aging. Expe imen al P ocedu es S udy popula ion, ec ui men , da a and blood collec ion MARK-AGE is a Eu opean-wide c oss-sec ional popula ion s udy aimed a he iden i ica ion o bioma ke s o aging (B€ u kle e al., 2015; Cap i e al., 2015). In he p esen wo k, he exp ession o DNMT1 and DNMT3B ansc ip s was analysed in PBMCs samples om a o al o 2453 dono s in he age ange o 35–75 yea s ec ui ed in eigh di e en Eu opean coun ies. De ails o he ec ui men p ocedu es and o he collec ion o an h opome ic, clinical and demog aphic da a ha e been published (Mo eno-Villanue a e al., 2015a,b). PBMCs isola ion p ocedu e has been desc ibed (Mo eno-Villanue a e al., 2015a). B ie ly, PBMCs we e isola ed om EDTA-whole blood, ob ained by phlebo omy a e o e nigh as ing, by discon inuous densi y g adien cen i uga ion in Pe coll and subsequen ly c yop e- se ed and s o ed in liquid ni ogen. Samples we e hen shipped om he a ious ec ui men cen es o he MARK-AGE Biobank loca ed a he Uni e si y o Hohenheim, S u ga , Ge many. F om he Biobank, coded sampleswe e subsequen ly sen o he Sapienza Uni e si y o Rome on d y-ice whe e hey we e s o ed in liquid ni ogen un il analysis o he DNMT1 and DNMT3B mRNA le els. RNA ex ac ion and cDNA syn hesis Samples we e hawed by incuba ion a 37°C, ollowed by d op wise addi ion o RPMI con aining 10% FCS o a inal dilu ion o 1:20. Cells we e collec ed by cen i uga ion and p ocessed o RNA ex ac ion. Isola ion o o al RNA was pe o med using RNeasy Mini Ki (Qiagen, Hilden, Ge many) acco ding o he manu ac u e ’s ins uc ions and subjec ed o DNase I diges ion using RNase- ee DNase (Qiagen, Hilden, Ge many). RNA concen a ion, pu i y and in eg i y we e e alua ed as p e iously desc ibed (Zampie i e al., 2010). Re e se ansc ip ion was ca ied ou using he Supe Sc ip VILO cDNA Syn hesis Ki (In i ogen, MA, USA) on equal amoun s o o al RNA (0.5 lg). Real- ime quan i a i e RT–PCR The exp ession o DNMT1 and DNMT3B was de e mined by quan i a i e PCR using he Taqman Gene Exp ession Assays (Applied Biosys ems, CA, USA) ollowing he manu ac u e ’s p o ocol on he iCycle IQ de ec ion sys em (Bio-Rad, He cules, CA, USA). The PCR e iciency o each gene assay was es ed using wo old se ial dilu ions ( om 50 o 3.125 ng) o cDNAs andomly chosen om among he samples. Each se o p ime s and p obe showed an e iciency o 90–100%. All calib a ion cu es exhibi ed co ela ion coe icien s >0.99. Assays we e pe o med in iplica e wi h cDNA equi alen o 30 ng o e e se ansc ibed RNA. Gene exp ession analysis was pe o med by he ela i e calib a o no malized quan i ica ion me hod using he exp ession le el o he b-glucu onidase gene (GUSB)as e e ence (Zampie i e al., 2010). An in e - un calib a ion sample was used in all pla es o co ec o he echnical a iance be ween he di e en uns and o compa e esul s om di e en pla es. The calib a o consis ed o cDNA p epa ed om HCT116 cells. The Taqman Gene Exp ession Assays IDs o each se o p ime s and p obe we e as ollows: Hs00154749_m1 (DNMT1), Hs00171876_m1 (DNMT3B), Hs99999908_m1 (GUSB). The da a ob ained we e uploaded o he MARK-AGE da abase (Mo eno- Villanue a e al., 2015b), es ablished a he Uni e si y o Kons anz (Kons anz, Ge many), whe e hey we e eco ded, cu a ed and me ged wi h he an h opome ic, clinical and demog aphic da a elemen s o dono s. S a is ical analysis See suppo ing in o ma ion. Acknowledgmen s The au ho s hank he MARK-AGE subjec s o pa icipa ing in his s udy; Ms Gud un on Sche en o excellen echnical assis ance; and all DNMT1 and DNMT3B exp ession in aging, F. Cicca one e al. 763 ª2016 The Au ho s. Aging Cell published by he Ana omical Socie y and John Wiley & Sons L d.