The Dnmt3L ADD Domain Controls Cytosine Methylation Establishment during Spermatogenesis

A icle The Dnm 3L ADD Domain Con ols Cy osine Me hyla ion Es ablishmen du ing Spe ma ogenesis G aphical Abs ac Highligh s dFull es ablishmen o CG me hyla ion du ing male game ogenesis equi es Dnm 3L ADD dNon-CG me hyla ion es ablishmen c i ically equi es Dnm 3L ADD dCo ec spe ma ogenesis and e ili y equi es Dnm 3L ADD dDe ec s a e a consequence o al e ed gene exp ession and e oelemen s exp ession Au ho s Geo gios Vlachogiannis, Chad E. Niede hu h, ..., Robe J. Schmi z, S een K.T. Ooi Co espondence [email p o ec ed] In B ie The adap o p o ein Dnm 3L plays a c ucial ole in es ablishing cy osine me hyla ion du ing male game ogenesis, bu i s mechanism o ac ion is unclea . Vlachogiannis e al. now show ha he N- e minal ADD domain is necessa y o ull me hyla ion es ablishmen , wi h CH si es being c i ically dependen on his domain. Accession Numbe s GSE58066 Vlachogiannis e al., 2015, Cell Repo s 10, 944–956 Feb ua y 17, 2015 ª2015 The Au ho s h p://dx.doi.o g/10.1016/j.cel ep.2015.01.021 Cell Repo s A icle The Dnm 3L ADD Domain Con ols Cy osine Me hyla ion Es ablishmen du ing Spe ma ogenesis Geo gios Vlachogiannis, 1 Chad E. Niede hu h, 2 Salih Tuna, 1 A hanasia S a hopoulou, 1 Keijo Vii i, 1,4 Di k G. de Rooij, 3 Richa d G. Jenne , 1 Robe J. Schmi z, 2 and S een K.T. Ooi 1, * 1 Epigene ic Signaling G oup, Depa men o Cance Biology, UCL Cance Ins i u e, Paul O’Go man Building, 72 Hun ley S ee , London WC1E 6BT, UK 2 Depa men o Gene ics, Uni e si y o Geo gia, 120 Eas G een S ee , A hens, GA 30602, USA 3 Cen e o Rep oduc i e Medicine, Academic Medical Cen e , Uni e si y o Ams e dam, 1105 AZ Ams e dam, he Ne he lands 4 P esen add ess: Paedia ic Resea ch Cen e, Uni e si y o Tampe e Medical School and Tampe e Uni e si y Hospi al, 33520 Tampe e, Finland *Co espondence: [email protected] h p://dx.doi.o g/10.1016/j.cel ep.2015.01.021 This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/licenses/by/4.0/). SUMMARY A c i ical aspec o mammalian game ogenesis is he ep og amming o genomic DNA me hyla ion. The ca aly ically inac i e adap o Dnm 3L is essen ial o ensu ing his occu s co ec ly, bu he mechanism by which i unc ions is unclea . Using gene a ge ing o enginee a single-amino-acid mu a ion, we show ha he Dnm 3L his one H3 binding domain (ADD) is necessa y o spe ma ogenesis. Genome-wide single-base- esolu ion DNA me hylome analysis o mu an ge m cells e ealed o e all educ ions in CG me hyla ion a epe i i e sequences and non-p o- mo e CpG islands. S ikingly, we also obse e an e en mo e se e e loss o non-CG me hyla ion, sug- ges ing an unexpec ed ole o he ADD in his p o- cess. These epigene ic de iciencies we e coupled wi h de ec s in spe ma ogonia, wi h mu an cells dis- playing ma ked changes in gene exp ession and eac i a ion o e o ansposons. Ou esul s demon- s a e ha he Dnm 3L ADD is necessa y o Dnm 3L unc ion and ull ep oduc i e i ness. INTRODUCTION Cy osine me hyla ion is a key ch oma in modi ica ion equi ed o no mal gene egula ion and genome s abili y (Goll and Bes o , 2005). In mammals, i is p edominan ly ound in CG dinucleo- ides, unc ioning in he silencing o e o ansposons and egu- la ion o single-copy sequences and genes subjec o pa en -o - o igin imp in ing. While a e in mos issues and di e en ia ed cells, me hylome analysis e eals ha non-CG me hyla ion is abundan in emb yonic s em cells (ESCs) (Lis e e al., 2009), ge m cells (Kobayashi e al., 2013), and he b ain (Lis e e al., 2013; Xie e al., 2012), al hough i s unc ion emains unclea . DNA me hyla ion is es ablished du ing game ogenesis, ollowing e asu e in sexually uncommi ed p imo dial ge m cells (PGCs) (Seisenbe ge e al., 2012). In males, PGCs en e ing he gonadal idge di e en ia e in o p ospe ma ogonia (PSG)/gonocy es ha a e mi o ically a es ed by 16.5 days pos coi um (dpc) (Ve gou- wen e al., 1991). Be ween 16.5 dpc and 2 days pos pa um (dpp), genome-wide de no o DNA me hyla ion is hough o be es o ed o p io le els, ensu ing he co ec ese ing o DNA me hyla ion imp in s. Following bi h, p ospe ma ogonial s em cells (PSCs) e-en e he cell cycle (Cle mon and Pe ey, 1957), mig a e o and colonize he semini e ous ubule basemen mem- b anes, and es ablish he ini ial pool o spe ma ogonial s em cells (SSCs) ha main ain spe ma ogenesis o he es o he an- imal’s li e (Kluin and de Rooij, 1981; Yoshida e al., 2006) DNA me hyla ion is es ablished by wo DNA me hyl ans- e ases, Dnm 3A and Dnm 3B (Okano e al., 1999). In addi ion, a ca aly ically inac i e adap o , Dnm 3L, is essen ial o his p o- cess o occu p ope ly du ing game ogenesis (Bou c’his and Bes o , 2004; Bou c’his e al., 2001; Ha a e al., 2002; Webs e e al., 2005). Dnm 3L-de icien male animals a e azoospe mic due o a block in meiosis a he lep o ene-zygo ene ansi ion, ma ked by non-homologous synapsis/meio ic ca as ophe. Re o ansposons a e eac i a ed due o a ailu e o es ablish me hyla ion a hese elemen s. Fu he mo e, al hough ini ially p esen , SSCs a e epo edly deple ed by 8–10 weeks o age, esul ing in mu an es es con aining semini e ous ubules comple ely de oid o ge m cells (Ha a e al., 2006). Gi en he c i ical ole o Dnm 3L in game ogenesis, much a en ion has cen e ed on unde s anding he mechanism by which i ope a es. Biochemical assays indica e Dnm 3L di ec ly associa es wi h bo h Dnm 3A and Dnm 3B; c ys allog aphic s udies using he C- e minal domains o mu ine Dnm 3A and Dnm 3L indica e his in e ac ion is media ed h ough he C- e - minal me hyl ans e ase olds (Jia e al., 2007). Dnm 3L unc ions as an allos e ic ac i a o (Chedin e al., 2002; Sue ake e al., 2004) and imp o es Dnm 3A p ocessi i y (Holz-Schie inge and Reich, 2010). DNMT3L o ms he e o e ame ic complexes wi h DNMT3A (Jia e al., 2007), o ming ilamen s coa ing DNA molecules in i o (Ju kowska e al., 2008). Dnm 3L encodes a 50-kDa p o ein and con ains an N- e minal cys eine- ich ATRX-Dnm 3-Dnm 3L (ADD) domain (Aapola e al., 2002). Using ei he human o mouse Dnm 3L p o eins, we and o he s ha e shown ha he ADD domain in e ac s wi h he N e minus o 944 Cell Repo s 10, 944–956, Feb ua y 17, 2015 ª2015 The Au ho s his one H3. Impo an ly, his in i o in e ac ion can be dis up ed by an aspa ic acid o alanine mu a ion a posi ion 90 o 124 in human o mouse Dnm 3L p o ein, espec i ely (Hu e al., 2009; Ooi e al., 2007; O ani e al., 2009; Zhang e al., 2010). This obse - a ion has gi en ise o he Dnm 3L his one ecogni ion hypo h- esis (Ooi e al., 2010), whe eby he ecogni ion o DNA me hyl- a ion a ge sequences is dependen on he abili y o Dnm 3L o bind o his one H3. As men ioned, exis ing s udies in o Dnm 3L biological unc ion ha e elied on null alleles ha comple ely elimina e exp ession o Dnm 3L p o ein. Al hough g ea ly in o ma i e, his app oach p e- en s a di ec es o whe he o no he ADD is in ol ed in Dnm 3L unc ion. In addi ion, gi en ha Dnm 3L unc ions in complex wi h Dnm 3A/Dnm 3B, i is unclea he deg ee o which he pheno ypes obse ed a e a di ec consequence o dis up ing Dnm 3L p o ein o o he p o eins dependen on Dnm 3L exp es- sion/p esence. Using a ‘‘knockin’’ app oach, we ha e gene a ed a mouse s ain wi h a single-amino-acid mu a ion o es he bio- logical impo ance o he ADD domain. RESULTS De ec i e Cy osine Me hyla ion, Pa icula ly a Non-CG Con ex s, in Dnm 3L A/A P ospe ma ogonia In ag eemen wi h ou p e ious biochemical s udies on human DNMT3L, we obse ed ha mouse Dnm 3L p o ein binds o his- one H3 and ha a D124A mu a ion wi hin he Dnm 3L ADD is su - icien o dis up his binding (Figu e 1A). To es he biological impo ance o his esidue, he D124A mu a ion (he ea e e e ed o as Dnm 3L A ) was in oduced in o he endogenous Dnm 3L locus by homologous ecombina ion in ESCs and animals we e gene a ed by subsequen b eeding (Figu es S1A–S1D). Dnm 3L A/A animals we e iable, consis en wi h he Dnm 3L / pheno ype. To examine he ex en o me hyla ion de ec s in Dnm 3L A/A ge m cells, we pe o med whole-genome bisulphi e sequencing (WGBS) on luo escence-ac i a ed cell so ing (FACS)-pu i ied PSG om 1-day-pos pa um (1-dpp) wild- ype and mu an animals. Ge m cells we e pu i ied by combining he D124 mu a ion wi h a ge m cell epo e ansgene (Oc 4-GiP), which exp esses GFP unde he con ol o he Oc 4 p omo e (Ying e al., 2002). Analysis o he sequence con ex o cy osines iden i ied as be- ing me hyla ed in 9.85 310 8 mapped eads e ealed ha as well as canonical CG me hyla ion, wild- ype PSG ha bo ed signi ican numbe s o eads con aining me hyla ed cy osines in a non-CG con ex (Figu e 1B). I is well known ha mos cy osines in a CG con ex in mammals a e ypically me hyla ed and unde go deam- ina ion o TpG, which esul s in CG dinucleo ides ep esen ing a much smalle han expec ed p opo ion o o al cy osines. Indeed, in ou da ase , mapped eads con aining CG dinucleo ides (i e- spec i e o me hyla ion s a e) cons i u e only 3.86% o he o al (Table 1). We ound ha 72% o cy osines ha showed e idence o me hyla ion occu in a non-CG con ex ; his is in line wi h ha p e iously published in emb yonic day 16.5 (E16.5) PSG (Kobaya- shi e al., 2013), which ha e no ye comple ed de no o me hyl- a ion, as well as in g owing oocy es (Shi ane e al., 2013)(49% and 66%, espec i ely). Consis en wi h his, analysis o he le el o non-CG me hyla ion om bisulphi e PCR da a a long in e - spe sed nuclea elemen -1 (LINE-1) and in acis e nal A pa icle (IAP) elemen s also suppo s hese indings o abundan non- CG me hyla ion (Figu e S1E). Mo i analysis e ealed ha mos non-CG me hyla ion occu s a TNCAG and TNCAC mo i s (Fig- u e 1C), consis en wi h p e ious obse a ions in ge m cells (Ko- bayashi e al., 2013, Shi ane e al., 2013), human ESCs (Lis e e al., 2009), and b ain (Lis e e al., 2013). We nex examined he e ec s o he D124A mu a ion on cy o- sine me hyla ion. Globally, numbe s o me hyla ed CG si es in Dnm 3L A/A PSG ela i e o ha obse ed in wild- ype we e educed (Figu e 1D). Analysis o he me hyla ion le el h oughou he genome di ided in o 100-bp bins indica ed CG me hyla ion was signi ican ly educed in Dnm 3L A/A PSG compa ed o wild- ype (Wilcoxon ank sum es , p < 2.2 310 16 )(Figu e 1E). To de e mine i any speci ic compa men was a ec ed, we compa ed weigh ed me hyla ion le els wi hin a ious genomic compa men s. The majo i y o CG dinucleo ides a e loca ed wi hin he epe i i e compa men o he genome. Analysis o CG me hyla ion a LINEs, long e minal epea s (LTRs), and sho in e spe sed nuclea elemen s e ealed ha while hese ele- men s we e p edominan ly me hyla ed in wild- ype ge m cells, hese we e hypome hyla ed in Dnm 3L A/A mu an cells (Figu es 1F and S1F). This is consis en wi h bo h me hyla ion-sensi i e Sou he n blo and bisulphi e sequencing da a examining LINE1 and IAP sequences showing elemen s ha a e bo h no mally me hyla ed and hypome hyla ed (Figu es S1G and S1H). Anal- ysis o CpG islands (CGIs) e ealed di e ences depending on con ex . As expec ed, hose associa ed wi h gene p omo e s we e la gely unme hyla ed in bo h wild- ype and mu an cells (Figu e 1F). CGIs no associa ed wi h p omo e s (so-called o phan CGIs; Illingwo h e al., 2010) ended o be me hyla ed, wi h educ ions in mu an cells. As well as epe i i e elemen s, DNA me hyla ion is also es ablished a pa e nally imp in ed di e en ially me hyla ed egions (DMRs). CG me hyla ion a pa e nally imp in ed loci ended owa d educed le els in Dnm 3L A/A ge m cells (Figu e 1G), wi h signi ican educ ions a he Gp 1/Zdb 2 and Rasg 1 DMRs (Fishe exac es alues, 6.66 310 4 and 4.73 310 5 , espec i ely). In e es ingly, he Rasg 1 DMR, which is associa ed wi h an adjacen epe i i e e- gion, showed he g ea es educ ion (49% e sus 19% me hyl- a ion) As expec ed, analysis o ma e nally imp in ed DMRs e ealed hese o be hypome hyla ed. These da a indica e ha despi e Dnm 3L p o ein being p esen , ull es ablishmen o CG me hyla ion a epea s and he pa e nally imp in ed DMRs analyzed is de ec i e in Dnm 3L A/A ge m cells. In addi ion o he global educ ion in CG me hyla ion, we also obse ed a d as ic educ ion in non-CG me hyla ion in Dnm 3L A/A PSG ela i e o wild- ype (Figu es 1D and 1E). We no e ha he le el o exp ession o componen s o he Dnm 3 complex was no signi ican ly al e ed by he D124A mu a ion (Figu e S1I), a guing ha he obse ed me hyla ion de ec s could no be a ibu ed o educ ion in he exp ession le els o hese p o eins. Analysis o non-CG me hyla ion in a ious genomic compa - men s e ealed ha simila o CG me hyla ion, his was educed h oughou all egions in Dnm 3L A/A PSG, wi h mos non-CG me hyla ion ound in non-p omo e -associa ed CGIs (Figu es 1F and S1F). Toge he , hese da a sugges ha non-CG me hyla ion is c ucially dependen on co ec ADD unc ion. Cell Repo s 10, 944–956, Feb ua y 17, 2015 ª2015 The Au ho s 945 Figu e 1. Whole-Genome Bisulphi e Sequence Analysis in P ospe ma ogonia (A) Pep ide pull-down da a showing in e ac ion be ween in i o ansc ibed and ansla ed ecombinan mouse Dnm 3L p o ein and he N e minus o his one H3 pep ide (amino acids 1–21). A ow indica es Dnm 3L p o ein. Highe bands in inpu lanes ep esen BSA, used as a blocking agen du ing pep ide binding. P o eins we e esol ed on a 7.5% polyac ylamide gel and subsequen ly s ained using Bio-Sa e Coomassie s ain (Bio-Rad). No e ha his in e ac ion is spe- ci ically los upon mu a ion o aspa ic acid in o alanine a posi ion 124 (D124A). (B) Pie cha showing dis ibu ion o me hyl cy osine in CG, CHG, and CHH con ex s in wild- ype p ospe ma ogonia (PSG). Only uniquely mapped eads we e used in his and subsequen analyses. H e e s o any base excep a guanine. (C) Weblogo plo s o bases p oximal and dis al o me hyla ed cy osines in CHG and CHH con ex s. (D) Ba g aphs compa ing ela i e no malized numbe s o me hyla ed si es in wild- ype e sus Dnm 3L A/A PSG. (E) Box-whiske plo s showing me hyla ion le els h oughou wild- ype and Dnm 3L A/A PSG genomic DNA di ided in 100-bp bins. Boxes ep esen qua iles and whiske s dema ca e maximum and minimum alues. Exac p alue is < 2.12 310 16 . (F) Hea maps indica ing le els o CG me hyla ion a di e en sequence classes indica ed. (G) Ba g aphs depic ing me hyla ion le els a pa e nal and ma e nal imp in ed di e en ially me hyla ed egions (DMRs). DMR coo dina es we e ob ained om Xie e al. (2012). 946 Cell Repo s 10, 944–956, Feb ua y 17, 2015 ª2015 The Au ho s Dnm 3L ADD Ensu es Hype me hyla ion a No mally Me hyla ed Loci in P ospe ma ogonia CG me hyla ion ou side o p omo e egions bu wi hin gene bodies has been epo ed o show a posi i e co ela ion wi h he le el o ansc ip ion (Hellman and Chess, 2007). We he e- o e conside ed whe he such an associa ion exis ed in PSG. Using a ailable RNA-sequencing da a, we examined CG me hyl- a ion a genes highly exp essed in E16.5 PSG (Seisenbe ge e al., 2012). Con a y o published da a, a e age le els o bo h CG and non-CG me hyla ion wi hin highly exp essed genes showed sligh bu signi ican educ ions compa ed o le els obse ed a weakly exp essed genes (Figu e 2A). We no e also ha a e age me hyla ion le els o bo h hese se s o genes a e lowe s ill compa ed o ha obse ed wi hin he bodies o all Re- Seq genes. This a gues ha , a leas in PSG, he e is no di ec posi i e co ela ion be ween he le els o CG me hyla ion wi hin genes and ansc ip ional ac i i y. Consis en wi h his, we obse e ha bo h CG and CH me hyl- a ion ac oss he bodies o genes in se e al g oups closely ack ha obse ed along all Re Seq genes (Figu e 2B). In o he wo ds, gene-body me hyla ion is obse ed ac oss all genes and i is gene ally educed in Dnm 3L A/A ge m cells. Ou da a indica e ha ull me hyla ion ac oss gene bodies is dependen on Dnm 3L unc ion. In addi ion, when we compa e gene-body CG me hyl- a ion o all mouse Re Seq genes in wild- ype 1-dpp PSGs wi h ha obse ed in o he issues, we ind subs an ially highe le els o me hyla ion (Figu e 2C). In line wi h all o he compa men s analyzed, CG gene-body me hyla ion a all Re Seq genes was signi ican ly educed in Dnm 3L A/A PSG compa ed o wild- ype (Figu e 2D). Genome-wide analysis o he dis ibu ion o CG me hyla ion le els indica es ha ela i e o o he issue ypes, PSG a e hype me hyla ed, which is a enua ed in Dnm 3L A/A ge m cells (Figu e 2E). Dnm 3L A/A Male Animals Ha e De ec i e Spe ma ogenesis and Reduced Fe ili y We nex sough o de e mine he e ec s o he D124A mu a ion on es es de elopmen and e ili y. Analysis o Dnm 3L A/A s ud males e ealed ha al hough hey we e capable o si ing o sp ing, hey we e sub e ile, wi h signi ican educ ions in o e all es es size (Figu e 3A) and a ious me ics o ep oduc- ion (Figu e 3B). In a ew cases, animals we e ound o be comple ely azoospe mic. His ological examina ion o mu an es es unexpec edly e ealed he p esence o semini e ous u- bules de oid o ge m cells beyond he spe ma ocy e s age, co-exis ing wi h ubules displaying all s ages o spe ma ogen- esis (Figu e 3C, see panels I and II). S aging analysis e ealed apop osis o la ge coho s o spe ma ocy es in some ubules in s age IV while no apop osis o spe ma ocy es was obse ed in o he s age IV ubules (Figu es S2A–S2F). When apop o- sis occu ed, ound and elonga ing spe ma ids we e also absen in he a ec ed a ea. These obse a ions sugges ha Dnm 3L A/A spe ma ocy es ha e a diminished chance o su i e he s age IV checkpoin . Howe e , spe ma ocy es can su i e s age IV and con inue o de elop no mally h ough o ma u e spe ma ozoa. This explains he pheno ype o ubules wi h no mal, comple e spe ma ogenesis, adjacen o ubules lacking gene a ions o spe ma ocy es and spe ma ids. We no e ha spe ma ogonia and, depending on he epi helial s age, ea ly ypes o spe ma o- cy es we e always p esen a 12.5 o 16 weeks o age (Figu es S2A–S2F). This de elopmen al block is consis en wi h p e i- ously published da a showing ha semini e ous ubules in Dnm 3L-de icien es es do no p og ess beyond s age IV (Webs e e al., 2005). The p opo ion o de ec i e ubules in di e en Dnm 3L A/A animals a ied, wi h some animals being signi ican ly a ec ed while o he s showed minimal signs o de ec i e ubules (Table 2). FACS analysis on ge m cells om mu an animals ca ying he Oc 4-GiP ansgene showed ha Dnm 3L A/A male animals had signi ican educ ions in absolu e numbe s o GFP + cells, which ansla ed o educ ions a all spe - ma ogenic s ages (Figu es 3D, S2G, and S2H). Analysis o meio ic sp eads e ealed he p esence o bo h homologous (no mal) as well as non-homologous (abno mal) synapsis e en s (Figu e 3E). Taken oge he , hese da a indica e ha he D124A mu a ion is hypomo phic and ha he ADD is necessa y o no mal Dnm 3L biological ac i i y. D124A Mu a ion Dis up s Gene Exp ession in SSCs The SSC niche is main ained by a delica e balance o signals p omo ing ei he sel - enewal o di e en ia ion. P e iously pub- lished da a examining Dnm 3L null animals showed ha spe - ma ogonia a e comple ely absen by 8–10 weeks o bi h, caused by s em cell exhaus ion (Ha a e al., 2006). As men ioned abo e, despi e he p esence o de ec i e ubules, his ological analysis indica ed ha ge m cells we e p esen beyond 8– 10 weeks o bi h. His ological analyses o es es om 1-dpp an- imals sugges ed he absence o any educ ion in he numbe s o PSG ha ini ially seed he semini e ous ubules (Figu es S2I and S2J). This is consis en wi h p e ious epo s showing no di e - ence in numbe s o PSG in Dnm 3L-de icien animals (Ha a e al., 2006). To be e unde s and he mechanisms by which he D124A mu a ion causes spe ma ogenic de ec s, we examined he ansc ip ional consequences in FACS-pu i ied spe ma ogonia. Table 1. Numbe o Mapped Reads Con aining Me hyla ed and Unme hyla ed Cy osines in Di e en Sequence Con ex s in Dnm 3L +/+ and Dnm 3L A/A 1-dpp P ospe ma ogonia Me hyla ed Unme hyla ed To al Dnm 3L +/+ Dnm 3L A/A Dnm 3L +/+ Dnm 3L A/A Dnm 3L +/+ Dnm 3L A/A CG 3.03 310 7 2.54 310 7 7.76 310 6 1.17 310 7 3.81 310 7 3.71 310 7 CHH 5.79 310 7 8.8 310 6 6.68 310 8 7.02 310 8 7.26 310 8 7.10 310 8 CHG 2.04 310 7 2.24 310 6 1.86 310 8 2.01 310 8 2.06 310 8 2.03 310 8 To al 1.09 310 8 3.64 310 7 8.61 310 8 9.14 310 8 9.70 310 8 7.26 310 8 Cell Repo s 10, 944–956, Feb ua y 17, 2015 ª2015 The Au ho s 947 (legend on nex page) 948 Cell Repo s 10, 944–956, Feb ua y 17, 2015 ª2015 The Au ho s Cells we e pu i ied om 9-dpp animals o examine he immedi- a e consequences o he D124A mu a ion in newly di e en ia ing spe ma ogonia and o assis pu i ica ion o su icien numbe s o cells o analysis. Mic oa ay analysis e ealed up- and down e- gula ion o 530 genes and 411 genes, espec i ely in Dnm 3L A/A SSCs (Figu e 4A). Among o he p ocesses, Gene On ology (GO) analysis e ealed a educ ion in he exp ession o genes in ol ed in cell a e commi men as well as hose con aining homeobox domains (Figu es 4B and S3). qRT-PCR o ma ke s o undi e - en ia ed spe ma ogonia and genes unc ionally impo an in SSCs main enance and di e en ia ion e ealed educ ions in he exp ession o a ious ma ke s/ ansc ip ional egula o s, including signi ican educ ions in Pou5 1/Oc 4,Sox3,Dazl, Ngn3, and Sohlh2 (Figu e 4C) (Phillips e al., 2010). Consis en wi h unc ions in de elopmen , we ound ha genes down egula ed in Dnm 3L A/A SSCs a e en iched o polycomb a ge genes. Me agene analysis o ESC ch oma in immunop ecipi a ion sequencing (ChIP-seq) da a e ealed an en ichmen o he polycomb ep essi e complex 2 (PRC2) sub- uni Suz12 a genes down egula ed in Dnm 3L A/A SSCs (Fig- u e 5A). Simila ly, using lis s o PRC2 a ge genes in ESCs (Ku e al., 2008), we ound ha while 10% o all genes a e associa ed wi h PRC2, his inc eases o 36% o genes ha a e down egu- la ed in Dnm 3L A/A spe ma ogonia; up egula ed genes showed only an 8% o e lap wi h Suz12 binding. Down egula ed genes a e he e o e 3.5 imes mo e likely o be a a ge o he PRC2 complex. A simila en ichmen was obse ed when examining bo h a second PRC2 componen (Ezh2) as well as he pos - ansla ional his one modi ica ion i media es (H3K27me3) (Fig- u es 5B and 5C), wi h 44% o down egula ed genes associa ed wi h he modi ica ion compa ed o 29% o up egula ed genes (27% o all genes a e associa ed wi h his modi ica ion). We no e ha in 1-dpp PSG, bo h CG and non-CG me hyla ion p o- iles ac oss gene bodies o PcG a ge genes a e e y simila o me hyla ion p o iles o non-exp essed and all Re Seq genes (Figu e 5D), a guing agains a di ec in e ac ion be ween Dnm 3L- a ge ed me hyla ion and he egula ion o hese genes. These da a sugges ha de- egula ion o genes in ol ed in spe - ma ogenesis and de elopmen con ibu e o he obse ed de- ec s in ge m cell unc ion. Re o anposons A e Ac i a ed by he D124A Mu a ion We nex sough o p o ide insigh s in o he po en ial mecha- nisms by which he D124A mu a ion migh a ec gene induc ion. The 50 egula o y egions o e o ansposons, pa icula ly IAP and LINE1 elemen s, a e known a ge s o CG me hyla ion and can possess enhance ac i i y (Schmid e al., 2012). In si u hyb idiza ion using p obes o hese elemen s e ealed ha Dnm 3L A/A mu an es es display a mosaic pa e n o exp ession, wi h some ubules showing high le els o exp ession and o he ubules showing li le o no exp ession (Figu es 6A and 6B). In e es ingly, his is consis en wi h ou his ological da a indica ing de ec s in some, bu no all, semini e ous ubules. Consis en wi h his, analysis o ou mic oa ay da a o 9-dpp spe ma ogonia indica ed ha ansc ip s con aining e o ans- posons we e also up egula ed in Dnm 3L A/A animals (Figu e 6C) We hen examined whe he he e was any ela ionship be ween he de- egula ed genes we obse ed in Dnm 3L A/A spe ma o- gonia and hei p oximi y o mapped epe i i e elemen s. P ox- imi y analysis e ealed ha on a e age, up egula ed genes ended o be close o he endogenous e o i us (ERV) class o e oelemen s, which include IAP elemen s (Figu e 6D; exac p alues a e 0.01279 [up egula ed genes] and 0.9953 [down e- gula ed genes]). Analysis o he me hyla ion s a e in 1-dpp PGS o ERVs ep esen ed on he a ay indica ed ha hese oo we e hypome hyla ed in Dnm 3L A/A compa ed o wild- ype ge m cells (Figu es 6E and 6F). This suppo s a model whe eby he Dnm 3L ADD unc ions o ensu e ull me hyla ion and silencing o hese elemen s, in u n necessa y o p e en ec opic gene exp ession du ing spe ma ogenesis. DISCUSSION The mechanism by which epigene ic in o ma ion is bo h es ab- lished and main ained is o g ea impo ance bu emains poo ly de ined, pa icula ly in e ac ions be ween di e en epigene ic p ocesses. The adap o Dnm 3L is c ucial o he es ablishmen o DNA me hyla ion du ing game ogenesis. Howe e , p e ious da a in he li e a u e we e ob ained using gene ic models lacking exp ession o Dnm 3L, and li le is known abou which o he do- mains p esen in Dnm 3L is equi ed o egula ion o epigene ic ep og amming du ing game ogenesis. We ha e now shown ha he ADD is necessa y o co ec Dnm 3L biological unc ion, po en ially h ough i s abili y o con ol non-CG me hyla ion and CG me hyla ion and ep ession o e o ansposons. To he bes o ou knowledge, ou esul s a e he i s demon- s a ion o he impo ance o he Dnm 3L ADD in game ogenesis. Ou own da a and ha o o he s ha e shown ha in i o, he D124A mu a ion esul s in loss o his one H3 binding. A ecen s udy epo ed ha he Dnm 3L ADD di ec ly in e ac s wi h he Figu e 2. CG and Non-CG Me hyla ion Is Obse ed a Simila Le els ac oss All Genes and Reduced by D124A Mu a ion (A) Box-whiske plo s o no malized CG and CH me hyla ion le els ac oss genes highly and weakly/no exp essed in E16.5 PSG. Me hyla ion le els a all Re Seq genes a e also shown. E16.5 PSG exp ession da a we e ob ained om Seisenbe ge e al. (2012). Wilcoxon ank sum es esul s compa ing highly o weakly exp essed genes we e 1.403 310 5 and 4.091 310 34 o CG and CH me hyla ion, espec i ely. (B) Me aplo s showing gene-body me hyla ion a CG and CH in he a ious gene g oups indica ed. Blue and ed lines deno e le els obse ed in wild- ype and mu an , espec i ely. No e ha le els a e simila i espec i e o gene ac i i y. (C) Me aplo o no malized CG me hyla ion le els ac oss all genes in he di e en cell ypes and issues indica ed. (D) Box-whiske plo s showing CG me hyla ion le els wi hin all Re Seq genes in wild- ype and Dnm 3 A/A PSG. Resul o Wilcoxon ank sum ex , E16.5 high e sus E16.5 low: p = 1.403 310 5 (mCG), p = 4.091 310 4 (mCH). (E) G aph showing he dis ibu ion o me hyla ion le els o each me hyla ed cy osine in he genome in he di e en issues and cell ypes indica ed. The shi o he igh (close o 1.0 o 100%) indica es ha many mo e me hylcy osines a e me hyla ed 100% o he ime based on he sequencing da a compa ed o lowe indi idual me hyla ion le els obse ed in ge m cells. The equency o me hyla ed cy osines me hyla ed a he highes le els is g ea e in wild- ype han Dnm 3L A/A ge m cells and o he soma ic issues es ed. The da a indica e ha PSG a e hype me hyla ed compa ed o di e en ia ed issues. Cell Repo s 10, 944–956, Feb ua y 17, 2015 ª2015 The Au ho s 949 PRC2 componen Ezh2 in ESCs (Ne i e al., 2013). In con as , ou own s udies ha e ailed o de ec in e ac ion be ween endog- enous Dnm 3L and Ezh2 by co-immunop ecipi a ion using an an i-Dnm 3L an ibody (Figu e S4). As i is unknown whe he he D124A mu a ion would simila ly a ec Ezh2 in e ac ion, and gi en he ex ensi e li e a u e demons a ing ADD in e ac ion Figu e 3. Dnm 3L D124A Mu a ion Causes Fe ili y De ec s (A) Image o es es emo ed om 15-week-old Dnm 3L +/+ and Dnm 3L A/A animals. (B) Biome ic da a compa ing Dnm 3L +/+ and Dnm 3L A/A animals. No e ha Dnm 3L A/A animals ha e a signi ican educ ion in es es mass (Dnm 3L +/+ : n = 19, age: 8–19 weeks; Dnm 3L A/A : n = 18, age: 8–19 weeks). Dnm 3L A/A males also p oduce signi ican ly less spe ma ozoa han age-ma ched Dnm 3L +/+ males (n = 6, age: 15–17 weeks). Two Dnm 3L A/A animals we e ound o be azoospe mic. The a e age li e sizes si ed by Dnm 3L A/A s ud males we e also educed (5.6 pups e sus 9.2 pups). Fo Dnm 3L +/+ animals, ou li e s om ou di e en s uds we e eco ded. Fo Dnm 3L A/A animals, eigh di e en li e s om eigh di e en s uds we e eco ded. Sca e plo s show mean ±SEM. Exac p alues a e 0.0374 (no malized es es mass, NMT), 0.0087 (spe m coun s), and 0.0283 (li e sizes). (C) Low- (63) and high-magni ica ion (303) images o H&E-s ained sec ions o ixed and pa a in-embedded es es om a Dnm 3L A/A male (16 weeks). No e he coincidence o no mal (I) and emp y (II) semini e ous ubules. SG, spe ma ogonia; P, pachy ene spe ma ocy e; S , ound spe ma id; eS , elonga ing spe ma id; Ap, apop o ic cell. (D) Ba g aphs depic absolu e cell numbe s (mean ±SEM) o he di e en cell popula ions indica ed measu ed by FACS. Exac p alues a e 0.0059 (GFP + ) and 0.0012 (mi o ic/4n, diploid/2n, and haploid). (E) Immuno luo escence mic og aphs o meio ic sp eads om ei he Dnm 3L +/+ o Dnm 3L A/A es es. Sp eads we e s ained using Scp1 (synap onemal com- plex 1) an ibody. No e ha Dnm 3L A/A males p oduce spe ma ocy es capable o unde going co ec ch omosomal pai ing (no mal) as well as spe ma ocy es de ec i e in he p ocess (abno mal). 950 Cell Repo s 10, 944–956, Feb ua y 17, 2015 ª2015 The Au ho s wi h his one H3, we a o he in e p e a ion ha he pheno ypes epo ed in his s udy a e a consequence o de ec i e his one H3 binding and ch oma in associa ion. Ou esul s indica e ha he D124A mu a ion is a hypomo phic allele; male Dnm 3L A/A animals a e sub e ile wi h educ ions in a ious me ics o e ili y, which is somewha unexpec ed. Based on he his one ecogni ion model, ab oga ion o his one H3 bind- ing should esul in a ailu e by Dnm 3L o co ec ly ec ui Dnm 3A/B me hyl ans e ases o all me hyla ion a ge loci. The e o e, one would p edic ha Dnm 3L A/A male animals should phenocopy Dnm 3L / animals (comple e s e ili y ins ead o sub e ili y). The animals used in he s udies epo ed ca y he D124A mu a ion on an ou b ed backg ound, which migh ac- coun o he incomple e pene ance o he epo ed pheno ype. We ha e ound ha combining he D124 mu a ion wi h a null allele (Dnm 3L A/ ) esul ed in male animals ha we e all azoospe mic, had mic ochidia, and whose es es only con ained ubules wi h de ec i e spe ma ogenesis (Figu e S5). Combined wi h ou ob- se a ions in Dnm 3L A/A animals, his suppo s a dosage e ec and ha despi e he p esence o one copy o Dnm 3L, he D124A mu a ion has c i ical consequences on game ogenesis and phenocopies comple e absence o Dnm 3L p o ein. Ou analyses o he ansc ip ional de ec s obse ed in Dnm 3 A/A undi e en ia ed spe ma ogonia p o ide insigh in o he spe ma ogenesis pheno ypes p e iously obse ed in Dnm 3L-de icien animals. Ou bioin o ma ics analyses suppo he no ion ha e o ansposons can se e as abe an con ol- ling elemen s o enhance he exp ession o nea by genes. They also unde sco e he impo ance o co ec exp ession o de el- opmen al egula o s and genes in ol ed s em cell main enance. Al hough a ecen s udy has implica ed Dnm 3L p o ein in play- ing a di ec unc ional ole in SSCs (Liao e al., 2014), ou esul s and hose o o he s do no co obo a e his inding. I has p e i- ously been epo ed ha Dnm 3L p o ein is no longe exp essed om 2 dpp (Sakai e al., 2004), al hough he e is e idence o he p esence o Dnm 3L mRNA ha canno be ansla ed (O’Dohe y e al., 2011). We ha e been unable o de ec he p esence o Dnm 3L p o ein in FACS-pu i ied spe ma ogonia (Figu e S6), which leads us o conclude ha he ansc ip ional de ec s we obse e a his s age o spe ma ogenic de elopmen a e a consequence o pe u ba ions du ing he p eceding PSG s age. Gi en he gene al impo ance o Dnm 3L in es ablishing me hyla ion a cy osines, ou inding ha non-CG me hyla ion is subs an ially a ec ed by he D124A mu a ion is s iking and unexpec ed. I is widely conside ed ha as bo h CG and non- CG me hyla ion is dependen on he Dnm 3 de no o machine y, any pe u ba ion in Dnm 3L unc ion would be p edic ed o a ec bo h ypes o me hyla ion equally. As bo h CG and non- CG me hyla ion a e comple ely obli e a ed in Dnm 3L null ani- mals, ou esul s sugges di e ences in he mechanism by which me hyla ion is es ablished in CG and non-CG con ex s, namely ha he la e is g ea ly dependen on he Dnm 3L ADD domain. Genomic non-CG me hyla ion has been desc ibed in plu ipo- en s em cells (Lis e e al., 2009), ge m cells (bo h g owing oo- cy es and PSG; Shi ane e al., 2013; Kobayashi e al., 2013), and p e on al co ical neu ons (Lis e e al., 2013). Common ea u es o hese cell ypes a e he exp ession o Dnm 3A and he ac ha hey a e non-cycling. Dnm 3L p o ein exp ession has no been documen ed ou side o plu ipo en and ge m cells; consis en wi h his, we ailed o de ec Dnm 3L in p e on al co ex om ju- enile male animals (Figu e S6B). The absence o de ec able le els o Dnm 3L combined wi h ou obse a ion ha ge m cells a e dependen on Dnm 3L-his one H3 binding o non-CG me hyla ion sugges s cell- ype-speci ic mechanisms in how his modi ica ion maybe egula ed. I has long been known in mammals ha CG me hyla ion is p edominan ly loca ed in epe i i e elemen s and imp in ed DMRs. Ou esul s show ha Dnm 3L-his one associa ion plays a gene al ole o ensu e ha hese sequences a e me hyla ed o he co ec le els. We no e ha no all pa e nally imp in ed loci a e equally a ec ed by he D124A mu a ion. The mos se e ely a ec ed DMR, Rasg 1, is conside ed o be an a ypical pa e nally imp in ed locus wi h a DMR associa ed wi h an adjacen andem epe i i e egion (Shiba a e al., 1998). Ou esul s also add o he deba e o e he unc ion o gene-body me hyla ion. We ob- se ed a gene al educ ion ac oss hese egions in Dnm 3L A/A PSG, consis en wi h educ ions in CG me hyla ion in e oele- men s ha a e p edominan ly loca ed in in ons. Howe e , ou esul s sugges ha a leas in PSG, he e is no conspicuous as- socia ion wi h highly as compa ed o weakly/non-exp essed genes; his a gues ha i he e is any co ela ion, i is sub le. This is simila o he obse a ion in hESCs, which show highe o e all CG me hyla ion compa ed o soma ic cells and he exis- ence o pa ially me hyla ed domains (Lis e e al., 2009). We p opose ha posi i e associa ions be ween ansc ip ion le els and gene-body me hyla ion may a y in a cell- ype-dependen manne . Ou esul s imply ha he genomes o ge m cells such as PSG may ha e he highes o e all le els o DNA me hyla ion o all mammalian cell ypes desc ibed which we p opose is a conse- quence o Dnm 3L exp ession. This leads o he p edic ion ha cell ypes exp essing Dnm 3L should ha e high le els o CG me hyla ion. In acco dance wi h his, we no e ha hESCs, which also exp ess Dnm 3L, ha e been epo ed o ha e high le els o CG me hyla ion (Lis e e al., 2009). Exp ession o Dnm 3L ende s PSG hype me hyla ed ela i e o genomes o o he issues, and ou esul s show his o be a de aul p ocess necessa y o ull ep oduc i e i ness. An ou s anding ques ion is how genomic egions emain e ac o y o cy osine me hyla ion, speci ically p omo e -associa ed CGIs and ma e nally imp in ed DMRs. We p opose ha u u e s udies Table 2. Analysis o he F equency o De ec i e Tubules in Tes es Sec ions om Dnm 3L +/+ and Dnm 3L A/A Animals Geno ype Age (Weeks) Numbe No mal De ec i e Dnm 3L +/+ 17 200 0 Dnm 3L A/A 18 140 9 Dnm 3L A/A 16 95 73 Dnm 3L A/A 18 175 58 Dnm 3L A/A 15 0 200 In each ans e se sec ion, 200 ubules we e coun ed. Cell Repo s 10, 944–956, Feb ua y 17, 2015 ª2015 The Au ho s 951