+ All Categories
Home > Documents > Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic...

Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic...

Date post: 30-Jun-2018
Category:
Upload: hoangkien
View: 223 times
Download: 0 times
Share this document with a friend
51
Global Epigenomic Reconfiguration During Mammalian Brain Development Ryan Lister,* Eran A. Mukamel, Joseph R. Nery, Mark Urich, Clare A. Puddifoot, Nicholas D. Johnson, Jacinta Lucero, Yun Huang, Andrew J. Dwork, Matthew D. Schultz, Miao Yu, Julian Tonti-Filippini, Holger Heyn, Shijun Hu, Joseph C. Wu, Anjana Rao, Manel Esteller, Chuan He, Fatemeh G. Haghighi, Terrence J. Sejnowski, M. Margarita Behrens,* Joseph R. Ecker* Introduction: Several lines of evidence point to a key role for dynamic epigenetic changes during brain development, maturation, and learning. DNA methylation (mC) is a stable covalent modifi- cation that persists in post-mitotic cells throughout their lifetime, defining their cellular identity. However, the methylation status at each of the ~1 billion cytosines in the genome is potentially an information-rich and flexible substrate for epigenetic modification that can be altered by cellular activity. Indeed, changes in DNA methylation have been implicated in learning and memory, as well as in age-related cognitive decline. However, little is known about the cell type–specific patterning of DNA methylation and its dynamics during mammalian brain development. Methods: We performed genome-wide single-base resolution profiling of the composition, pat- terning, cell specificity, and dynamics of DNA methylation in the frontal cortex of humans and mice throughout their lifespan (MethylC-Seq). Furthermore, we generated base-resolution maps of 5-hydroxymethylcytosine (hmC) in mammalian brains by TAB-Seq at key developmental stages, accompanied by RNA-Seq transcriptional profiling. Results: Extensive methylome reconfiguration occurs during development from fetal to young adult. In this period, coincident with synaptogenesis, highly conserved non-CG methylation (mCH) accumulates in neurons, but not glia, to become the dominant form of methylation in the human neuronal genome. We uncovered surprisingly complex features of brain cell DNA methylation at multiple scales, first by identifying intragenic methylation patterns in neurons and glia that distin- guish genes with cell type–specific activity. Second, we report a novel mCH signature that identifies genes escaping X-chromosome inactivation in neurons. Third, we find >100,000 developmentally dynamic and cell type–specific differentially CG-methylated regions that are enriched at putative regulatory regions of the genome. Finally, whole-genome detection of 5-hydroxymethylcytosine (hmC) at single-base resolution revealed that this mark is present in fetal brain cells at locations that lose CG methylation and become activated during development. CG-demethylation at these hmC-poised loci depends on Tet2 activity. Discussion: Whole-genome single-base resolution methylcytosine and hydroxymethylcytosine maps revealed profound changes during frontal cortex development in humans and mice. These results extend our knowledge of the unique role of DNA methylation in brain development and function, and offer a new framework for testing the role of the epigenome in healthy function and in pathologi- cal disruptions of neural circuits. Overall, brain cell DNA methylation has unique features that are precisely con- served, yet dynamic and cell-type specific. FIGURES IN THE FULL ARTICLE Fig. 1. Methylcytosine in mammalian frontal cortex is developmentally dynamic and abundant in CG and CH contexts. Fig. 2. mCH is positionally conserved and is the dominant form of DNA methylation in human neurons. Fig. 3. mCH is enriched in genes that escape X inactivation. Fig. 4. Cell type–specific and developmental differences in mC between mouse neurons and glia. Fig. 5. hmCG is enriched within active genomic regions in fetal and adult mouse brain. Fig. 6. Developmental and cell type–specific differential mCG. SUPPLEMENTARY MATERIALS Materials and Methods Figs. S1 to S12 Tables S1 to S5 References (63–78) mCH 53% mCG mCH 38% mCG % methylated CH sites 0 50 100 % methylated CG sites ~50 year old neurons >6 week old neurons birth Human Mouse eye-opening 55 65 0 10 20 Age (years) 75 85 0 10 Age (weeks) 0 1 0.5 1.5 CH {CA,CC,CT} CG ... ... A G T A C A C T A T G T G C The DNA methylation landscape of human and mouse neu- rons is dynamically reconfigured through development. Base-resolution analysis allowed identification of methylation in the CG and CH context (H = A, C, or T). Unlike other differentiated cell types, neurons accumulate substantial mCH during the early years of life, coinciding with the period of synaptogenesis and brain maturation. READ THE FULL ARTICLE ONLINE http://dx.doi.org/10.1126/science.1237905 Cite this article as R. Lister et al., Science 341, 1237905 (2013). DOI: 10.1126/science.1237905 The list of author affiliations is available in the full article online. *Corresponding author. E-mail: [email protected] (R.L.); [email protected] (M.M.B.); [email protected] (J.R.E.) www.sciencemag.org SCIENCE VOL 341 9 AUGUST 2013 RESEARCH ARTICLE SUMMARY 629 Published by AAAS
Transcript
Page 1: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

Global Epigenomic Reconfi guration During Mammalian Brain DevelopmentRyan Lister,* Eran A. Mukamel, Joseph R. Nery, Mark Urich, Clare A. Puddifoot, Nicholas D.

Johnson, Jacinta Lucero, Yun Huang, Andrew J. Dwork, Matthew D. Schultz, Miao Yu, Julian

Tonti-Filippini, Holger Heyn, Shijun Hu, Joseph C. Wu, Anjana Rao, Manel Esteller, Chuan He,

Fatemeh G. Haghighi, Terrence J. Sejnowski, M. Margarita Behrens,* Joseph R. Ecker*

Introduction: Several lines of evidence point to a key role for dynamic epigenetic changes during brain development, maturation, and learning. DNA methylation (mC) is a stable covalent modifi -cation that persists in post-mitotic cells throughout their lifetime, defi ning their cellular identity. However, the methylation status at each of the ~1 billion cytosines in the genome is potentially an information-rich and fl exible substrate for epigenetic modifi cation that can be altered by cellular activity. Indeed, changes in DNA methylation have been implicated in learning and memory, as well as in age-related cognitive decline. However, little is known about the cell type–specifi c patterning of DNA methylation and its dynamics during mammalian brain development.

Methods: We performed genome-wide single-base resolution profi ling of the composition, pat-terning, cell specifi city, and dynamics of DNA methylation in the frontal cortex of humans and mice throughout their lifespan (MethylC-Seq). Furthermore, we generated base-resolution maps of 5-hydroxymethylcytosine (hmC) in mammalian brains by TAB-Seq at key developmental stages, accompanied by RNA-Seq transcriptional profi ling.

Results: Extensive methylome reconfi guration occurs during development from fetal to young adult. In this period, coincident with synaptogenesis, highly conserved non-CG methylation (mCH) accumulates in neurons, but not glia, to become the dominant form of methylation in the human neuronal genome. We uncovered surprisingly complex features of brain cell DNA methylation at multiple scales, fi rst by identifying intragenic methylation patterns in neurons and glia that distin-guish genes with cell type–specifi c activity. Second, we report a novel mCH signature that identifi es genes escaping X-chromosome inactivation in neurons. Third, we fi nd >100,000 developmentally dynamic and cell type–specifi c differentially CG-methylated regions that are enriched at putative regulatory regions of the genome. Finally, whole-genome detection of 5-hydroxymethylcytosine (hmC) at single-base resolution revealed that this mark is present in fetal brain cells at locations that lose CG methylation and become activated during development. CG-demethylation at these hmC-poised loci depends on Tet2 activity.

Discussion: Whole-genome single-base resolution methylcytosine and hydroxymethylcytosine maps revealed profound changes during frontal cortex development in humans and mice. These results extend our knowledge of the unique role of DNA methylation in brain development and function, and offer a new framework for testing the role of the epigenome in healthy function and in pathologi-cal disruptions of neural circuits. Overall, brain cell DNA methylation has unique features that are precisely con-served, yet dynamic and cell-type specifi c.

FIGURES IN THE FULL ARTICLE

Fig. 1. Methylcytosine in mammalian

frontal cortex is developmentally dynamic

and abundant in CG and CH contexts.

Fig. 2. mCH is positionally conserved and

is the dominant form of DNA methylation

in human neurons.

Fig. 3. mCH is enriched in genes that escape

X inactivation.

Fig. 4. Cell type–specifi c and developmental

differences in mC between mouse neurons

and glia.

Fig. 5. hmCG is enriched within active

genomic regions in fetal and adult mouse

brain.

Fig. 6. Developmental and cell type–specifi c

differential mCG.

SUPPLEMENTARY MATERIALS

Materials and MethodsFigs. S1 to S12Tables S1 to S5References (63–78)

mCH

53%mCG

mCH

38%

mCG

% m

eth

yla

ted

CH

sit

es

0

50

100

% m

eth

yla

ted

CG

sit

es

~50 year old

neurons

>6 week old

neurons

birth

Human Mouse

eye-opening

55 650 10 20

Age (years)

75 850 10

Age (weeks)

0

1

0.5

1.5

CH{CA,CC,CT}

CG

...

...

A

G

T

A

C

A

C

T

A

T

G

T

G

C

The DNA methylation landscape of human and mouse neu-rons is dynamically reconfi gured through development. Base-resolution analysis allowed identifi cation of methylation in the CG and CH context (H = A, C, or T). Unlike other differentiated cell types, neurons accumulate substantial mCH during the early years of life, coinciding with the period of synaptogenesis and brain maturation.

READ THE FULL ARTICLE ONLINE

http://dx.doi.org/10.1126/science.1237905

Cite this article as R. Lister et al., Science 341, 1237905 (2013). DOI: 10.1126/science.1237905

The list of author affi liations is available in the full article online.*Corresponding author. E-mail: [email protected] (R.L.); [email protected] (M.M.B.); [email protected] (J.R.E.)

www.sciencemag.org SCIENCE VOL 341 9 AUGUST 2013

RESEARCH ARTICLE SUMMARY

629

Published by AAAS

Page 2: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

Global Epigenomic ReconfigurationDuring Mammalian Brain DevelopmentRyan Lister,1,2*† Eran A. Mukamel,3* Joseph R. Nery,1 Mark Urich,1 Clare A. Puddifoot,3

Nicholas D. Johnson,3 Jacinta Lucero,3 Yun Huang,4 Andrew J. Dwork,5,6 Matthew D. Schultz,1,7

Miao Yu,8 Julian Tonti-Filippini,2 Holger Heyn,9 Shijun Hu,10 Joseph C. Wu,10 Anjana Rao,4

Manel Esteller,9,11 Chuan He,8 Fatemeh G. Haghighi,5 Terrence J. Sejnowski,3,12,13

M. Margarita Behrens,3† Joseph R. Ecker1,13†

DNA methylation is implicated in mammalian brain development and plasticity underlyinglearning and memory. We report the genome-wide composition, patterning, cell specificity, anddynamics of DNA methylation at single-base resolution in human and mouse frontal cortexthroughout their lifespan. Widespread methylome reconfiguration occurs during fetal to youngadult development, coincident with synaptogenesis. During this period, highly conservednon-CG methylation (mCH) accumulates in neurons, but not glia, to become the dominantform of methylation in the human neuronal genome. Moreover, we found an mCH signaturethat identifies genes escaping X-chromosome inactivation. Last, whole-genome single-baseresolution 5-hydroxymethylcytosine (hmC) maps revealed that hmC marks fetal brain cell genomesat putative regulatory regions that are CG-demethylated and activated in the adult brain andthat CG demethylation at these hmC-poised loci depends on Tet2 activity.

Dynamic epigenetic changes have been ob-served during brain development, matura-tion, and learning (1–6). DNAmethylation

(mC) is a stable covalent modification that per-sists in postmitotic cells throughout their lifetime,defining their cellular identity. However, themeth-ylation status at each of the ~1 billion cytosines inthe genome is potentially an information-rich andflexible substrate for epigenetic modification thatcan be altered by cellular activity (7, 8). Changesin DNA methylation were implicated in learn-ing and memory (9, 10), as well as in age-related

cognitive decline (11). Mice with a postnatal de-letion of DNA methyltransferases Dnmt1 andDnmt3a in forebrain excitatory neurons, or with aglobal deletion of methyl-CpG-binding protein2 (MeCP2), show abnormal long-term neural plas-ticity and cognitive deficits (2, 12).

DNAmethylation composition and dynamicsin the mammalian brain are highly distinct. Amod-ification of mC catalyzed by the Tet family of mChydroxylase proteins, 5-hydroxymethylcytosine(hmC), accumulates in the adult brain (13–15)along with its more highly oxidized derivatives5-formylcytosine and 5-carboxylcytosine. Thesemodifications of mC were implicated as inter-mediates in an active DNA demethylation path-way (16–19). In addition, methylation in thenon-CG context (mCH, where H = A, C, or T) isalso present in the adult mouse and human brains(20, 21) but is rare or absent in other differen-tiated cell types (22, 23). Little is known aboutcell type–specific patterning of DNAmethylationand its dynamics during mammalian brain devel-opment. Here, we provide integrated empiricaldata and analysis of DNA methylation at single-base resolution, across entire genomes, with cell-type and developmental specificity. These resultsextend our knowledge of the unique role of DNAmethylation in brain development and functionand offer a new framework for testing the role ofthe epigenome in healthy function and in path-ological disruptions of neural circuits.

Accumulation of Non-CG DNA MethylationDuring Brain DevelopmentTo identify the composition and dynamics of tran-scription and methylation during mammalianbrain development, we performed transcriptomeprofiling (mRNA-Seq) and whole-genome bi-

sulfite sequencing [MethylC-Seq (24)] to com-prehensively identify sites of cytosine DNAmethylation (mC and hmC) and mRNA abun-dance at single-base resolution throughout thegenomes of mouse and human frontal cortex(table S1). DNA methylation in embryonic stem(ES) cells occurs in both the CG (mCG) andnon-CG (mCH) contexts, but mCH is largelylost upon cell differentiation (22, 23, 25, 26). Wefound that although mCH levels are negligiblein fetal cortex, abundant mCH occurs in adultfrontal cortex (Fig. 1A). mCHhas previously beenidentified throughout the genome of the adultmouse brain (20) and at several hundred genomicpositions in the human adult brain (21). Sup-porting previous studies, we found that mamma-lian brain mCH is typically depleted in expressedgenes, with genic mCH level inversely propor-tional to the abundance of the associated tran-script (Fig. 1, A and B) (20). This pattern is theopposite of that observed in ES cells (22) andsuggests that genic mCH in the brain may inhibittranscription. The absence of mCH in fetal brainsuggests that this signature for gene repressionis added to the genome at a later developmen-tal stage.

We performed MethylC-Seq on mouse andhuman frontal cortex during early postnatal, juve-nile, adolescent, and adult stages (Fig. 1C). CHmethylation level, defined as the fraction of allbase calls at CH genome reference positions thatwere methylated (denoted mCH/CH), accumu-lates in mouse and human brain during earlypostnatal development to a maximum of 1.3 to1.5% genome-wide at the end of adolescencebefore diminishing slightly during aging. mCHincreases most rapidly during the primary phaseof synaptogenesis in the developing postnatalbrain, from 2 to 4 weeks in mouse (27) and in thefirst 2 years in humans (28), followed by sloweraccumulation of mCH during later adolescence.mCH accumulation initially parallels the increasein synapse density within human middle frontalgyrus (synaptogenesis lasts from birth to 5 years),but it subsequently continues to increase duringthe period of adolescent synaptic pruning, whichin humans occurs between 5 and 16 years of age(Fig. 1C). Notably, the accumulation of mCH inmice from 1 to 4 weeks after birth coincides witha transient increase in abundance of the de novomethyltransferase Dnmt3a mRNA and protein(Fig. 1D). Analysis of the context of mCH sitesshowed that it is mainly present in the CA context(fig. S1, A to F), as previously reported for mCH(20, 22, 23, 26).

Overall, genomes in the frontal cortex arehighlymethylated.Whereas CG partially methyl-ated domains (PMDs) account for about a third ofthe genome of various differentiated human cells(22, 25), human brain genomes have negligibleCGPMDs, resembling pluripotent cellmethylomes(25) (fig. S1, G and H). Given the high spatialconcordance of CG PMDs and nuclear lamina-associated domains reported previously (29), the

RESEARCHARTICLE

1Genomic Analysis Laboratory, The Salk Institute for BiologicalStudies, La Jolla, CA 92037, USA. 2Plant Energy Biology [Aus-tralian Research Council Center of Excellence (CoE)] and Com-putational Systems Biology (Western Australia CoE), School ofChemistry and Biochemistry, The University of WesternAustralia, Perth, WA 6009, Australia. 3Computational Neuro-biology Laboratory, The Salk Institute for Biological Studies,La Jolla, CA 92037, USA. 4La Jolla Institute for Allergy andImmunology and Sanford Consortium for Regenerative Med-icine, La Jolla, CA 92037, USA. 5Department of Psychiatry,Columbia University and The New York State Psychiatric In-stitute, New York, NY 10032, USA. 6Department of Pathologyand Cell Biology, Columbia University, New York, NY 10032,USA. 7Bioinformatics Program, University of California at SanDiego, La Jolla, CA 92093, USA. 8Department of Chemistry andInstitute for Biophysical Dynamics, The University of Chicago,Chicago, IL 60637, USA. 9Cancer Epigenetics Group, CancerEpigenetics and Biology Program (PEBC), Bellvitge BiomedicalResearch Institute (IDIBELL), L’Hospitalet de Llobregat, Bar-celona 08907, Spain. 10Department of Medicine, Division ofCardiology, Stanford University School of Medicine, Stanford,CA 94305, USA. 11InstitucióCatalana de Recerca i Estudis Avançats(ICREA), Barcelona, Catalonia, Spain. 12Division of BiologicalSciences, University of California at San Diego, La Jolla, CA92037, USA. 13Howard Hughes Medical Institute, The SalkInstitute for Biological Studies, La Jolla, CA 92037, USA.

*These authors contributed equally to this work.†Corresponding author. E-mail: [email protected] (R.L.);[email protected] (M.M.B.); [email protected] (J.R.E.)

www.sciencemag.org SCIENCE VOL 341 9 AUGUST 2013 1237905-1

Page 3: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

D

mR

NA

-Seq

FP

KM

0

20

40

Pro

tein (n

orm

. to actin

)

0.5

1.0

1.5

Age (days post-conception)20 40 60 80

Dnmt1Dnmt3aDnmt3lDnmt3bDnmt3a protein

mR

NA

C

humanmouse

Juve

nile

Adoles

cent

Adult

Aged

Childh

ood

Adoles

cent

Adult

Age (days post conception)

birth birth

% m

CG

/CG

syn

apti

c d

ensi

ty†

Fetal

1 wk

2 wk

4 wk

6 wk10 wk

22 mo

% m

CH

/CH

0

0.4

0.8

1.2

1.6

0 50 100 680

Fetal

35 do

12 yr

16 yr25 yr 55 yr

64 yr

0

20

40

60

80

0

mCH/CHmCG/CGsynapses†

10,0

00

20,0

00

2 yr

5 yr

human chr2: 96,109,000 - 96,442,000

mouse chr2: 126,895,000 - 127,217,000

mCHmCG

20 kb

W

C

W

C

H1

H1

GenesASTL

DUSP2STARD7 TMEM127

SNRNP200

LOC285033CIAO1

NCAPHITPRIPL1ADRA2B

Dusp2 Stard7 Snrnp200Ciao1

Itpripl1

Adra2bGenes

25 yr

25 yrfetal

mRNA

mCG

mCH

mRNA

mCG

mCH

H125 yrfetal

fetal10 wk

10 wkfetal

10 wkfetal

20 kb

A

hmC 6 wkfetal

Ncaph

Tmem127Astl

mCG/CG

mCH/CH5hmC

(CMS-IP/input)

mRNA

Genes

Chr 12 (Mb)

G

DNaseI HS

ChIP input

113 114 115 116 117 118 119

F

0

40

0.030

1

0

Chr 12 (Mb)0 20 40 80 100 120

human

mouse

Fetal35 d2 yr5 yr

12 yr16 yr25 yr

mC

H/C

H

0

0.01

0.02

0 20 40CEN 80 100 120

mC

H/C

H 0.02

Chr 12 (Mb)

mC

G/C

G

0

0.5

1.0

64 yr

Fetal1 wk2 wk4 wk6 wk

10 wk22 mo

mC

G/C

G

0

0.5

1.0

CEN

ImmunoglobulinVH locus

Chrom

atin

accessibility, y

0

0.01

0.02

0

0.1

0.2

0.3

0.4

0.5

H

Den

sity(a.u.), z

mCH/CH (Mm FC 10 wk), x

E

Fetal

Adult (6 wk)

0 2 4 6 100

% of cytosine basecalls (mouse)

hmCG mCGCG hmCHmCH CH

96.32.9

0.55 0.20 0.052 (mCH)

94.10.017(hmCH)

1.32.90.870.91

0 (hmCH)

B

100%

100%0

100%

100%0

100%

100%0

100%

100%0

100%

100%0

110

0.01

0.02

0

0.01

0.02

0

0.01

0.02

1 15,000 20,000 15,000

mRNA rank(least abundant most abundant)

mC

H/C

HH

um

an 2

5 yr

co

rtex

mC

H/C

HM

ou

se 1

0 w

k co

rtex

mC

H/C

HH

um

an H

1 E

SC 0

Max

Density (a.u.)

Fig. 1. Methylcytosine in mammalian frontal cortex is developmentallydynamic and abundant in CG and CH contexts. (A) Browser representationof mC and mRNA transcript abundance in human and mouse frontal cortex andhuman ES cells. Chr2, chromosome 2. (B) mCH/CH within gene bodies exhibitsopposite correlation with gene expression in ES cells (ESC) and brain. Contoursshow data point density, and red line shows smoothed mCH/CH as a functionof mRNA. a.u., arbitrary units. (C) Synaptic density (for mouse, per 100 mm2; forhuman, per 100 mm3) andmC level in CG and CH contexts through developmentin mouse and human frontal cortex. †Synaptic density quantitation from De Felipe et al. (27) andHuttenlocher and Dabholkar (28). (D) DNA methyltransferase mRNA and protein abundance (mean T SEM)in mouse frontal cortex through development. FPKM, fragments per kilobase of exon per million fragmentsmapped. (E) Fraction of cytosine base calls with each modification in fetal and adult mouse frontal cortex.(F) Cortex mC level in CG and CH contexts throughout mouse and human chromosome 12 in 100-kbbins smoothed with ~1-Mb resolution. CEN, centrosome. (G) Transcript abundance, chromatin accessi-bility [8-week mouse cortex ChIP input and DNaseI hypersensitivity (HS) normalized read density], andmC levels in 5-kb bins at the mouse immunoglobulin VH locus. (H) Density (z) plot of 10-week mousefrontal cortex mCH level (x) versus 8-week mouse cortex ChIP-input normalized read density (y ) for all10-kb bins of the mouse genome.

9 AUGUST 2013 VOL 341 SCIENCE www.sciencemag.org1237905-2

RESEARCH ARTICLE

Page 4: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

paucity of CG PMDs in these brain methylomescould indicate that lamina-associated domainsare altered or much less frequent in the brain.

The adult mammalian brain contains the high-est levels of hmC that have been observed (15),accounting for about 40% of methylated CG sitesin cerebellar Purkinje cells (30). hmC accumu-lates during early postnatal brain development inmice (31, 32), becoming enriched in highly ex-pressed genes (33). Given the evidence that hmCcan be an intermediate in an active DNA demeth-ylation pathway (16, 17), high-resolution analy-sis of the genomic distribution of hmC is neededto understand its role in the control of DNAmeth-ylation dynamics through brain development.Standard bisulfite-sequencing data does not distin-

guish betweenmethylated and hydroxymethylatedsites, somethylcytosines identified byMethylC-Seqanalysis represent the sum of these two contri-butions. Therefore, we used Tet-assisted bisulfitesequencing [TAB-Seq (34)], a base-resolutiontechnique that distinguishes hmC fromC andmCgenome-wide, to profile hmC in mouse fetal andadult frontal cortex (Fig. 1A). Integration of thegenome-wide profiles of mC and hmC enabled adetailed breakdown of the methylated subset ofthe genome at these distinct developmental stages(Fig. 1E). hmC constitutes 0.20% of total cytosinebase calls in fetal cortex and increases to 0.87% inadult cortex. This modification appears to be re-stricted to the CG context, as also observed inhuman and mouse ES cells (34); after correction

for false detection, we estimated that 0.017% ofcytosine base calls were hmCHgenome-wide (99%confidence interval: 0 to 0.059%), and significanthmCH was detected at few individual sites (fig.S2, A and B). The overwhelming presence ofhmC in the CG context (99.98%) in mouse adultand fetal frontal cortex is consistent with recentfindings in human ES cells, where 99.89%of hmCis in the CG context (34). hmC was present atmany highlymethylatedCG sites (fig. S2C). There-fore, although only a small fraction of all cytosinesthroughout the genome are methylated (mCG =2.9%, mCH = 1.3%, hmC = 0.87%), mCH andhmC constitute major, and nonoverlapping, com-ponents of the methylated fraction of the ge-nome in adult frontal cortex (mCG = 57.2%,

C

0

2

4

% m

CH

R1 R2 R3 R1 R2 R3

glia

NeuN+ NeuN–

0

20

40

60

80

100

% m

CG

R1 R2 R3 R1 R2 R3

glia

NeuN+ NeuN–

Dnmt3a binding sitesRandom

A human chr5: 87,917,000 - 88,317,000

mouse chr13: 83,518,000 - 83,911,000

mCHmCG

WC

Genes MEF2C

Mef2c

R1R2R1R2R1R2R1R2

mC

Gm

CH

20 kb

20 kb

Genes

Dnmt3a ChIP-chip

R1R2R1R2R1R2R1R2

mC

Gm

CH

NeuN+

NeuN–

NeuN+

NeuN–

NeuN+

NeuN–

NeuN+

NeuN–

B

0

2

4

6

8

10

R1 R2 R3 R1 R2 R3

glia

mCHmCG

% m

C/C

Tiss

ue(2

5 yr

)

Tiss

ue(1

0 w

k)

NeuN+ NeuN+NeuN– NeuN–

R1 R2 R1 R24.

20

4.66

4.29

4.12

4.11

1.44

5.12

4.84

0.39

0.58

3.56

3.43

3.44

3.43

3.24

3.22

3.29

4.03

1.22 2.

13

2.12

2.09

0.48

0.38

0.43 0.

29

D

0.7R2R3R1R2R1R3

NeuN

+N

euN–

01-r

R2R3R1R2R1R3

NeuN+ NeuN–

mC

H

R1

R1

R2

R2 R1

R1

R2

R2

NeuN+

NeuN

+

NeuN–

NeuN

1-r00.7

mC

H

Pearson correlation0.3 1

F

E

0

0.2

0.4

0.6

0.8

1

Per

−sit

e co

rrel

atio

no

f m

C/C

(n

orm

aliz

ed)

Neu

N+

R2

Neu

N+

R3

NeuN+ R2

NeuN+ R1

0 1

1

1

0

1

CH sites

human

mouse

chr3: 36,078,990 - 36,079,55750 bp

chr2: 171,489,889 - 171,490,454

NeuN+mCH

R1R2R3

CH sitesNeuN+

mCHR1R2

50 bp

CG

CH

Autosomes

CG

CH

ChrX

vs

NeuN+

Neu

N+

vs NeuN+

NeuN+

vsNeu

N+

NeuN+

vsNeu

N+

ES

vs

H1 ES

HuES6

vs NeuN+

H1 ES

Fig. 2. mCH is positionally conserved and is the dominant formofDNAmethylation in human neurons. (A) Browser representation of mCG andmCH in NeuN+ and NeuN– cells. Human NeuN+/NeuN– samples: R1, 53-year-old female; R2, 55-year-old male. Mouse NeuN+/NeuN– samples: R1, 7-weekmales; R2, 6-week females; R3, 12-month females (not shown). (B) Percentageof methylated base calls in each sequence context throughout the genome. (C)Box and whisker plot of mCG and mCH level in neurons and glia at genomic

regions bound by Dnmt3a versus a random set. Whiskers indicate 1.5 times theinterquartile range. (D) mCH correlation between NeuN+ and NeuN– cells inmouse and human, measured in 10-kb bins. (E) Browser representation of mCHsites in neurons. Scatter plots (right) show consistent mCH/CH at all single sitesin a 20-kb window overlapping the example region (left). (F) Correlationanalysis of methylation state at single sites between neurons and ES cells inhuman and mouse. Correlation values are normalized by a simulation (62).

www.sciencemag.org SCIENCE VOL 341 9 AUGUST 2013 1237905-3

RESEARCH ARTICLE

Page 5: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

mCH = 25.6%, hmC = 17.2%). These data sug-gest that the steady-state population of hmC inthe adult brain is not an intermediate stage in thedemethylation of mCH. However, these steady-state measurements do not preclude the possibil-ity that hmCH could be rapidly turned over afterconversion from mCH, leading to negligible de-tected hmCH despite Tet-mediated demethylationat CH sites.

Protection of Inaccessible Genomic Regionsfrom de Novo MethylationmCH accumulates in parallel across most of thegenome (Fig. 1F). However, we found numerous(36 in human, 34 in mouse) noncentromeric,megabase-sized regions that do not accumulatemCH. These regions, which we termed mCHdeserts, are enriched for large gene clusters thatencode proteins involved in immunity and recep-

tors required for sensory neuron function (tableS2). One mCH desert spans the immunoglobulinVH locus, which encodes variable domains of theimmunoglobulin heavy chain that rearrange in Blymphocytes. The VH locus is transcriptionallyquiescent in the frontal cortex of 10-week-oldmice,and the chromatin state is highly inaccessible, asinferred from deoxyribonuclease I (DNaseI) hyper-sensitivity profiling (35) and chromatin immuno-precipitation (ChIP) input sequence read densitydata (36, 37) (Fig. 1G). In contrast, mCG is notdepleted in mCH deserts.

Genome-wide detection of hmC by cytosine5-methylenesulphonate immunoprecipitation(CMS-IP) (38, 39) revealed that hmC is alsostrongly depleted in the VH locus. mCH desertsare observed at other loci in the genome, in-cluding olfactory receptor gene clusters that formheterochromatic aggregates required for mono-

allelic receptor expression in olfactory sensoryneurons (40, 41). Genome-wide comparison ofmCH/CH with chromatin accessibility, as in-ferred from ChIP input read density (36, 37),for all 10-kb windows of the mouse genomerevealed two discrete groups of genomic regions(Fig. 1H). Low-accessibility regions tend to con-tainminimalmCH,whereasmore-accessible regionsof the genome show a proportional relationshipbetween genome accessibility and mCH levels.Thus, although mCG is unaffected in these re-gions, lower chromatin accessibility appears tobe highly inhibitory to deposition of mCH andhmC, potentially via inaccessibility to de novomethyltransferases and Tet mC hydroxylases. Fur-thermore, this indicates that accumulation of mCHand hmC during mammalian brain developmentoccurs via processes that are at least partly inde-pendent from methylation at CG dinucleotides.

C D

A

B

NeuN+ mCH

NeuN – mCH

human10 kb

Genes UBA1RBM10 CDK16

USP11NDUFB11

25 yrmRNA

NeuN+ mCH

NeuN – mCH

Genes

10 wkmRNA

mouse20 kb

Kdm5cIqsec2

Tspy12Gm15266Gpr173

chrX: 46,840,000 - 47,047,000

chrX: 148,510,000 - 148,852,000

0.4

0

-0.4

0.1

0.05

0

Intr

agen

ic m

CH

/CH

Fem

ale

− M

ale

Pro

mo

ter

mC

G/C

GF

emal

e −

Mal

e

0 1-2 3-4 5-6 7-8 9

Bi−allelic expression [0−9]

Inactivated Escaped 0 10

1

False detection rate

Co

rrec

t d

etec

tio

n r

ate

ChrX−escapee discriminability

CG (AUC=0.78)

CH (0.75)

Both (0.88)

Fem

ale

Neu

N+

Fem

ale

Neu

N+

Male NeuN+

Pro

mo

ter

mC

G/C

GIn

trag

enic

mC

H/C

H

R1 R2 R1 R2

R1 R2 R3 R1 R2 R3

0 0.05 0.10

0.05

0.1

0 0.5 10

0.5

1

X-inact.* (253)X-escapee* (37)X-escapee§ (33)X-escapee† (7)

Chr2 (1,472)

Fig. 3. mCH is enriched in genes that escape X inactivation. (A) Browserrepresentation showingmCH-hypermethylated female human andmouse genesthat escape X inactivation (shaded genes). (B) Box and whisker plots of genderdifferences in promoter mCG and intragenic mCH in inactivated and escapeegenes on human chrX. (C) Scatter plot of gender differences in mCG and mCH

in human chrX genes. Reported X inactivated and escapee genes: *Carrel andWillard (49); §Sharp et al. (50); †predicted escapee genes, and autosomal(Chr2) genes are indicated. (D) Discriminability analysis of genes that escapefemale X inactivation using mC data, showing correct versus false detectionrate mapped for all possible mC/C thresholds.

9 AUGUST 2013 VOL 341 SCIENCE www.sciencemag.org1237905-4

RESEARCH ARTICLE

Page 6: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

Cell Type–Specific DNA Methylation Patternsin Neurons and GliaThe diversity of neuronal and glial cells in thefrontal cortex raises the question of which fea-tures of DNA methylation are found in specificcell types. We isolated populations of nuclei byfluorescence-activated cell sorting that were high-ly enriched for neurons (NeuN+) or glia (NeuN–)from human and mouse adult frontal cortex tis-sue. An additional glial population was isolatedfrommice expressing enhanced green fluorescentprotein (eGFP) under the S100b promoter.MethylC-Seq revealed differences in the com-

position and patterning of mCG and mCH inneurons and glia (Fig. 2A). Whereas differentialmCG between neurons and glia was restricted tolocalized regions, neurons were globally enrichedfor mCH compared with glia. Indeed, we dis-covered that the level of mCH in glia is similar tothat of fetal and early postnatal cortical tissue,whereas adult neurons have the greatest frequen-cy of mCH that has been observed in mammaliancells. This indicates that the rapid developmentalincrease in mammalian brain mCH that coincideswith the period of synaptogenesis is primarilydue to mCH accumulation in neurons. Further-

more, our data show that in human neurons mCHis the dominant form of methylation in the ge-nome: It is more abundant than mCG and occursin 5% of CH and 10% of CA sites (Fig. 2B andfig. S1, A, B, and H). Of the total methylatedfraction of adult human neuronal genomes,mCH accounts for ~53%, whereas mCG con-stitutes ~47%.

Although sparse in glia, mCH enrichment oc-curs within genes that are CH-hypomethylated inneurons, such as Mef2c (Fig. 2A), a transcrip-tional activator that plays critical roles in learn-ing and memory, neuronal differentiation (42),

C mCG/CG mCH/CH MedianmRNA−Seq FPKM

NeuN

+tis

sue

NeuN

R1R2R3

2 wk4 wk6 wk

22 mo10 wk

1 wkfetal

R1glia

R2R3

-100 kb +100 kb

mCG/CG mCH/CH MedianmRNA−Seq FPKM

-100 kb +100 kb

Neu

ron

alU

p-

reg

ula

ted

Co

nst

itu

tive

ly h

igh

1 10 100

1

10

0.1

1

10

0.1

1

10

0.1

Age (wks post-conception)

gene5’ 3’ gene5’ 3’1 10 100Age (wks post-

conception)

gene5’ 3’ gene5’ 3’

Ast

rocy

teD

ow

n-

reg

ula

ted

Co

nsi

tuti

vely

low

1

10

0.1

1

10

0.1

1

10

0.1

A B

fetal-100 kb +100 kb

2 wk 10 wk NeuN+ NeuN–

mCG/CG hmCG/CG mCH/CH

6 wk fetal-100 kb +100 kb

2 wk 10 wk NeuN+ NeuN–

Con

stitu

tivel

y hi

ghN

euro

nal

Upr

egul

ated

Ast

rocy

ticD

ownr

egul

ated

Con

stitu

tivel

y lo

w

Enrichment (q<0.05)

n.s.100.1

Normalized mCG/CG1.20.8

Normalized hmCG/CG1.50.5

Normalized mCH/CH20

fetal10 w

k

Gen

es

1

25,260

mRNA−Seq FPKM10.1 10

1

5

2

987

10

3

6

4

Fig.4.Cell type–specificanddevelopmental dif-ferences inmCbetweenmouseneuronsandglia.(A) Heat-map represen-tation of 25,260 mousegenes organized in genesets identified by k-meansclustering using normal-ized genic mCG andmCHlevels in adult develop-mental and NeuN+ andNeuN– samples. Left-handplot shows mRNA abun-dance. (B) Enrichment ordepletion of each clusterfor developmental and cell-type specific gene sets.n.s., not significant (FET,FDR< 0.05). (C) mCG andmCHthroughoutgenebodyand flanking 100 kb forindicated gene sets. Tran-script abundance (mRNA-Seq FPKM) over mousedevelopment is shown forthe same gene sets. Colorscales are as in (A).

www.sciencemag.org SCIENCE VOL 341 9 AUGUST 2013 1237905-5

RESEARCH ARTICLE

Page 7: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

synaptic plasticity (43), and regulation of synapsenumber and function (44). Genome-wide surveysidentified 174 mouse genes in which glia werehypermethylated relative to neurons in the CHcontext (table S3). Unbiased gene ontology anal-ysis revealed that these glial hyper-mCH genesare highly enriched for roles in neuronal and syn-aptic development and function (table S3). Thesegenes also overlapped significantly with a set of461 genes expressed at higher levels in neuronsthan in astrocytes (13-fold higher overlap thanchance, P < 10−30, Fisher exact test, FET) (45)and 233 developmentally up-regulated genes(7.5-fold, P < 10−7, FET). These genes showhypomethylation of CG and CH in neurons andhypermethylation of CH in glia (fig. S3A), con-sistent with a potential role of mCH in transcrip-tional repression of neuronal genes in the glialgenome. Furthermore, genes associated witholigodendrocyte or epithelial function accumu-late mCH through development (fig. S3B), witholigodendrocyte up-regulated genes showing intra-genic mCH hypermethylation in neurons andhypomethylation in glia, whereas epithelial genes

display mCH hypermethylation in both neuronaland glial populations. Consistent with CHmethyl-ation requiring Dnmt3a, glial hyper-mCH genesfrequently intersect areas of the genome boundby Dnmt3a in mouse postnatal neural stem cells(46). Dnmt3a-binding regions are greatly enrichedfor mCH, particularly in glia, whereas mCG isnot enriched in Dnmt3a-binding regions in gliaand is depleted in neurons (Fig. 2C). Thus, thereis an association with Dnmt3a binding sites spe-cific to mCH and not mCG, suggesting partialindependence between these two marks.

mCH Position Is Highly ConservedWe examinedwhether the position of DNAmeth-ylation is stochastic or precisely controlled atdifferent genomic scales. The level of mCH in10-kbwindows throughout the genomewas high-ly reproducible between independent samples ofthe same cell type, with lower, but substantial,correlation between cell types (Fig. 2D). Closerinspection revealed consistency between themeth-ylation level at individual mCH sites in neuronsfrom different individuals in both mice and hu-

mans (Fig. 2E). At single-base resolution (fig. S4),perfect correlation between individuals would notbe observed even if the true methylation levelwere identical at each site because of the stochas-tic effect of a finite number of sequenced reads.To correct for this, we normalized the observedcorrelation by that of simulated data sets with thesame coverage per site as each of our experi-mental samples but with identical methylationlevels (Fig. 2F and fig. S4). To assess statisticalsignificance, we used a permutation test, whichcompared the data correlation with the correla-tion after randomly shuffling the relative posi-tions of CH sites in each sample (fig. S4). Thisrevealed that autosomal CG and CH sites havenearly identical methylation levels in neuronalpopulations isolated from different individuals ofthe same species. Observed differences could beexplained by stochastic sampling rather than trueindividual variation. Unexpectedly, normalized per-site correlation is higher for mCH than mCGbetween neuronal populations isolated from thefrontal cortices of different human individuals,and mouse neuronal mCG and mCH per-site

CFetal6 wk

Fetal6 wk

Fetal6 wk

Fetal

IntragenichmCG/CG

6 wk

Fetal6 wk

Fetal6 wk

Neuronal

Upregulated

Astrocyte

Downregulated

Constitutivelylow

Constitutivelyhigh

0.8 1.2NormalizedhmCG/CG

flan

k n

orm

aliz

ed h

mC

G/C

G

-100 kb +100 kb

0 0.1 0.2

B

CG

Is

TES

± 2

kb

Adu

lt D

HS

Adu

lt en

h

TSS

± 2

kb

Intra

geni

c

Feta

l DH

S

Feta

l enh

A

0

hmCG hmCH

Auto

som

esC

hrX

Auto

som

esC

hrX

0.1

0.05

0.15

hm

C/C

(6

wk)

hm

CG

/CG

D

Fetal

mCG/CG

Gen

es

1

16,077

hmCG/CG

6 wk Fetal 6 wk Fetal

mCH/CH

6 wk-100 kb +100 kb

mRNA-Seq

2 w

k4

wk

6 w

k

22 m

o10

wk

1 w

kfe

tal

mRNA−Seq FPKM10.1 10

-100 kb +100 kb

0.5 1.5flank normalized

mCH/CH

0.8 1.2flank normalized

mCG/CG

0.8 1.2flank normalized

hmCG/CG

Fetal 6 wk

0

0.05

0.1

0.15

0.2

Fig. 5. hmCG is en-riched within activegenomic regions in fe-tal and adult mousebrain. (A) hmCG levelin 6-week mouse frontalcortex for autosomes andChrX. (B) Median hmCGlevel within genomic fea-tures (error bars 32ndto 68th percentile). enh,enhancer. (C) Median nor-malized hmCG throughoutgene body and flanking100 kb for indicated genesets. Bars show absolutehmCG/CGlevelswithingenebodies for each class. (D)mC and hmC throughoutgene body and flanking100 kb for each. Tran-script abundance (mRNA-Seq FPKM) during mousedevelopment is also shown(left).

9 AUGUST 2013 VOL 341 SCIENCE www.sciencemag.org1237905-6

RESEARCH ARTICLE

Page 8: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

correlations are equivalent. Per-site correlation be-tween two human ES cell lines (H1 and HUES6)is also high (>0.8) for both mCG and mCH.

The high interindividual correlation of mCHat the kilobase and single-site scales indicates thatmethylation of CH positions, particularly inmam-malian neurons, is a highly controlled process. Itis not consistent with a stochastic event that takesplace at any available CH position in a partic-ular genomic region that accumulates mCH. Com-parison of mCH between human and mouseneurons at conserved exonic CH positions re-vealed a low but significant interspecies corre-lation (Fig. 2F; P < 0.005, shuffle test), possiblyindicating conservation of the cellular processesthat precisely target or restrict mCH at these po-sitions. Last, per-site mCG and mCH correlationbetween human ES cells and neurons is signif-icantly lower, likely because of differences in theprocesses governing methylation of particular ge-nomic features in the distinct cell types, for ex-ample, enrichment and depletion of mCH in highlytranscribed genes in ES cells and neurons, respec-tively (Fig. 1, A and B).

The precise conservation of mCH positionmay be partly caused by the physical configu-ration of DNA within nucleosomes. Consistentwith this, neuronal mCH patterns contain robustperiodic components at the scale of nucleosomespacing [~170 base pairs (bp), fig. S5A] and theDNA helix coil length (~10.5 bp, fig. S5B). Suchperiodic components may arise from sequence-dependent constraints on mCH position, whichwould be the same in every neuronal cell. Al-ternatively, epigenetic heterogeneity within thepopulation of NeuN+ nuclei in our sample maylead to stronger correlation for CH sites locatedon the same physical chromosome, compared withthe correlation between the same locations on chro-mosomes from different cells. To test this, we mea-sured the cross-correlation within individual reads,revealing a contribution of within-chromosomecorrelation to the periodic methylation pattern(fig. S5C).

Gender-Specific DNA Methylation Patternson the X ChromosomeInterindividual correlation of mCG and mCH onchromosome X (ChrX) is frequently lower thanon autosomes (Fig. 2F), prompting a closer analy-sis of ChrX mC patterns. ChrX mCG and mCHlevels were generally lower in females comparedwith males, presumably because of the effect ofChrX inactivation (fig. S5, F and G) (47, 48).However, a subset of genes in both humans andmice have significantly greater intragenic mCHlevels in females compared with males (Fig. 3A).Inspection of these genes revealed that most werepreviously found to escape inactivation in humanfemales (X-escapees), displaying biallelic expres-sion (49) and a reduction in promoter mCG hy-permethylation, a DNA methylation signature ofinactivated alleles (50). Quantification of humangender differences in neuronal DNAmethylationfor ChrX genes previously characterized as show-

ing biallelic expression (49) revealed that femaleshave reduced promoter mCG and a large increasein intragenic mCH but not intragenic mCG (Fig.3, B and C, and fig. S5, D and E). The sequencecomposition of mCH is very similar in the wholegenome, within autosomal gene bodies, and with-in X-chromosome inactivated and escapee genebodies (fig. S5H). Analysis of gender-specificmethylation in additional human cell types re-vealed that female promoter mCG hypomethyla-tion is observed at X-escapee genes in glia andhuman embryonic stem cells (fig. S6). IntragenicmCH hypermethylation of X-escapees was alsoobserved in female glia, albeit to a lesser extentthan in neurons, but was not present in ES cells.Thus, X-escapee mCH hypermethylation may bea feature that is specific to neural cell types. Al-though both promoter CG hypomethylation andintragenic CH hypermethylation provide signif-icant information for discriminating X-escapees[Fig. 3D, discriminability index (area under thecurve, AUC) = 0.75 and 0.78, respectively], com-bining bothmCG andmCHmeasurements boostsdiscriminability (AUC = 0.88). By using thisintragenic mCH hypermethylation signature, weidentified seven new putative X-escapee genes(table S4). On the basis of these data, we hypothe-size that intragenic CH hypermethylation inneurons may play a compensatory role in genesthat fail to acquire repressive CG hypermethyla-tion in the promoter region, restoring equal geneexpression between male and female cells (51).

Distinct Genic DNA Methylation StatesDemarcate Functionally Relevant Gene ClustersDNA methylation within promoter regions andin gene bodies is implicated in regulation of geneexpression (22, 52), suggesting that the preciselyconserved, cell type–specific DNA methylationpatterns may be related to specific neuronal andglial cellular processes. We therefore used an un-biased approach to classify patterns of mCG andmCH within each annotated gene body and inflanking regions extending 100 kb up- or down-stream. After normalizing themethylation patternaround each autosomal gene by the local baselinemCG or mCH level in each adult neuronal orglial sample, we combined these features into alarge datamatrix containing 4200 individual DNAmethylation measurements for each gene [sevensamples, two contexts (CG and CH), 300 1-kbbins within and around each gene]. Using prin-cipal component (PC) analysis, we extracted fivemethylation features (PCs) that together accountfor 46% of the total data set variance (fig. S7A).Gene sets with specific neuronal or astrocyticexpression, aswell as ChrX genes, segregatewithinPC space (fig. S6B). We then used k-means clus-tering to classify all genes into 15 clusters on thebasis of their mCG and mCH patterns (Fig. 4Aand fig. S8). Several dominant patterns of DNAmethylation and transcript abundance and dynam-ics between developmental and cellular states areevident. A cluster of genes that progressively losesgene-body mCG and mCH through development

contains constitutively highly expressed genes thatare strongly enriched for neuronal function anddepleted for astrocyte-specific roles (Fig. 4A, box1). These genes show intragenic mCG enrich-ment in glia and depletion in neurons (box 2),indicating that glial gene body mCG resemblesthat of the neural precursor cells that predominatethe fetal brain. This indicates that the loss ofmCG in brain tissue during development is dueto CGhypomethylation inmature neurons. Theseconstitutively highly expressed genes enriched forneuronal function also show extensive intragenicmCH hypomethylation in neurons in contrast toglia (box 3), and they are enriched for hmCG(box4)as previously described (32, 33). Genes that arenot as highly transcribed, but that are associatedwith neuronal function and are developmentallyup-regulated, also show intragenicmCG andmCHhypomethylation in neurons but not glia (box 5).For these gene sets, mCG and mCH enrichmentor depletion is precisely localized to transcribedregions, suggesting that this modification of genicmC is tightly coupled to transcription. Notably,the bodies of constitutively high genes that arenot enriched for neuronal function (box 6) do notshowmarked fetal/glial mCG enrichment or neu-ronal mCH depletion, indicating that this differ-ential methylation is specific for genes enrichedfor neuronal function and not simply an associ-ation with particular levels of transcriptional ac-tivity. Genes associated with astrocyte functionshow an opposite pattern to genes associated withneuronal function: a progressive increase in intra-genic mCG andmCH in frontal cortex tissue overdevelopment, neuronal mCG and mCH hyper-methylation, and glial mCG andmCH hypometh-ylation (Fig. 4A, boxes 7 to 9). Last, genes withconstitutively low expression do not show devel-opmental or cell type–specific DNA methylationpatterns (box10), demonstrating that dynamicDNAmethylation in genes is highly associated withdifferential transcriptional activity in mammalianbrain development and neural cell specialization.

Each of the gene clusters identified in ourunbiased analysis was significantly enriched ordepleted for cell type–specific function [neuronalor astrocytic genes (45)] or particular expressionpatterns (constitutively high or low expression,developmentally up- or down-regulated) (Fig. 4B).Profiling the median mCG and mCH of geneswithin each of these categories allows direct com-parison of developmental and cell type–specificDNAmethylation in mouse (Fig. 4C) and human(fig. S3C). This analysis recapitulates many of theconclusions of the unbiased clustering (Fig. 4A).

The inverse relationship observed betweengenic mCH level and transcriptional activity isconsistent with a model whereby intragenic accu-mulation ofmCH impedes transcriptional activity.Alternatively, the process of transcription couldinterfere with mCH de novo methylation or in-duce active mCH demethylation, although theseare not consistent with the DNMT3A-dependentintragenic mCH in human embryonic stem cellsthat is positively correlated with gene expression

www.sciencemag.org SCIENCE VOL 341 9 AUGUST 2013 1237905-7

RESEARCH ARTICLE

Page 9: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

(73)

A

R1

R2

R3

R1

R2

R3

2 w

k4

wk

6 w

k

22 m

o10

wk

1 w

kfe

tal

6 w

kfe

tal

glia R1

R2

2 yr

5 yr

12 y

r

25 y

r64

yr

16 y

r

35 d

ofe

tal

H1

ES

R1

R2

NeuN+ NeuN- NeuN+ NeuN-hmC

CG

dif

fere

nti

ally

met

hyl

ated

reg

ion

s (C

G-D

MR

s)

Neu

N+

57,4

64N

euN

−10

,787

Neu

N+

58,4

45N

euN

−16

,139

hyp

er-m

CG

hyp

o-m

CG

CG

-DM

Rs

hyp

er-m

CG

hyp

o-m

CG

Neu

N+

138,

346

Neu

N−

38,6

49N

euN

+48

,745

Neu

N−

42,0

59

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

Med

ian

mC

G/C

G

mCG/CG10

mCG/CG10

hmCG/CG0.250

NeuN+ hyper mCGNeuN− hyper mCGNeuN+ hypo mCGNeuN− hypo mCG

C

Fol

d en

richm

ent/d

eple

tionn=

72,7

06

n=35

,391

n=57

,464

n=16

,139

n=58

,445

n=10

,787

n=25

0,93

2

not sig.(p>0.001)

0.25

1

4

16

B

0 30 60

CGIsPromoter (TSS ±2kb)

TES ±2kbIntragenicAdult DHSFetal DHS

Adult DHS (unique)Fetal DHS (unique)

Adult enhancersFetal enhancers

% of CG−DMRsper region

Fetal

> Adu

lt

NeuN+

hype

r mCG

NeuN+

hypo

mCG

All DM

Rs

Adult >

Fet

al

NeuN−

hypo

mCG

NeuN−

hype

r mCG

Adult > FetalFetal > Adult

Distance (kb) Distance (kb)

1

2

3

00.01

0.02

0.8

1.0

1.2

2

4

3x10-7

1x10-7

0

0.5

1.0

0

0.1

0.2

-2 0 +2

1.5

2.5

fetal adult fetal adult

0

0.01

1.0

1.5

-2 0 +2

1.5

2.0

3x10-7

1x10-7

0

1.0

0

0.1

0.2NeuN+

tissu

e

mC

hmC

hmC

mC

DNaseI

H3K4me1

H3K27ac

NeuN–

R1R2R3

2 wk4 wk6 wk

22 mo10 wk

1 wkfetal

6 wkfetal

fetal

fetal

hmC

G =

31%

hmC

G =

13%

hmC

G =

17%

hmC

G =

0%

mC

G =

92%

mC

G =

11%

mC

G=

100%

mC

G =

75%

WC

WC

adult

adultfetal

adultfetal

adult

R1

glia

R2R3

H3K27ac

mCH / CH

hmCG / CG(norm)

H3K4me1

DNaseI HS

mCG / CG

hmCG / CG

hmCDMRs

fetal > 6 wk6 wk > fetal

111,738,000 - 111,743,200chr9

mCHmCGF GCG-DMRs

H

D E

14

2

10

1057

19

47

2431

8

10

Adult DHS (147)

CG-DMRsAdult > Fetal

(36)

CG-DMRsFetal > Adult

Fetal DHS (103)

28 7

6110 10

112

0

0.1

Adult uniqueenhancers (20)

CG-DMRsAdult > Fetal (35)

CG-DMRsFetal > Adult (73)

Fetal uniqueenhancers (19)

I

−0.2 0 0.2 −0.2 0 0.20

0.08

0.04

Adult > Fetal

Difference in mCG/CG vs. WT

Fra

ctio

n o

f D

MR

s

Fetal > Adult Control: WT vs. WT

WT vs.Tet2 –/–

% o

f C

G−D

MR

s(F

DR

=0.0

5)

CG-DMRs: Fetal > Adult (20,483)

CG-DMRs:Adult > Fetal (10,870)

0

5

10

15

20

3.6%0.13%

19.7%

0.06%

mCG/CG increased

in Tet2-/-

mCG/CG decreased

in Tet2-/-

mCG/CG increased

in Tet2-/-

mCG/CG decreased

in Tet2-/-

tissue tissue

Fig.6.Developmentaland cell type–specificdifferential mCG. (A)Heat map of absolutemCG level in CG-DMRsidentified between neu-rons and glia and overdevelopment in mouse(left) and human (right).(B) Fraction of all CG-DMRs located in distinctgenomic features inmouse.(C) Enrichment or deple-tion of distinct cell type–specificanddevelopmentalCG-DMR sets within ge-nomic features from (B).(D and E) Intersection ofdevelopmentally dynam-icCG-DMRsand(D)DNaseIhypersensitive sites or (E)enhancers inmousebrainin thousands. (F) Browserrepresentation of mousedevelopmentally dynam-ic CG-DMRs and quanti-ficationof local enrichmentof chromatin modifica-tions, genome accessibil-ity, andmC. (G) TAB-Seqreadsshowingfetal-specifichmCG in the Fetal>AdultCG-DMR in mouse. (H)Proportion ofmouse de-velopmental CG-DMRswhere mCG/CG is sig-nificantly increased ordecreased in Tet2 knock-out mice. (I) Distributionof mCG level differencebetween wild-type (WT)and Tet2mutant atmouseCG-DMRs. Significantlydifferent DMRs are indi-cated by coloration.

9 AUGUST 2013 VOL 341 SCIENCE www.sciencemag.org1237905-8

RESEARCH ARTICLE

Page 10: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

(22, 23). Overall, glial mCG and mCH patternsclosely resemble those of the fetal and the earlypostnatal brain, indicating that DNA methylationin early mammalian brain developmental stagesmay be a default state that largely persists throughto maturity in glial cells, whereas neuronal dif-ferentiation and maturation involve extensive re-configuration of the DNA methylome that ishighly associated with cell type–specific changesin transcriptional activity.

hmCG Is Enriched Within Active GenomicRegions in Fetal and Adult Mouse BrainOur base-resolution analysis of hmC using TAB-Seq revealed that intragenic and global hmCGlevels are largely equivalent between chromosomes,whereas hmCG/CG is 22% lower on the maleChrX, consistent with previous reports from en-richment based detection of hmC (32, 33) (Fig. 5A).Analysis of hmCG levels in different genomicregions revealed that, although adult hmCG/CGis similar across transcriptional end sites and intra-genic, DNaseI-hypersensitive (DHS), and enhancerregions, the fetal frontal cortex shows a relativeenrichment of hmCG in DHS regions and en-hancers, in particular enhancer regions that areunique to the fetal developmental stage (Fig. 5B).The inverse pattern can be observed for adultmCG levels, which are lower in DHS regions andenhancers (fig. S9, A and B), suggesting that re-gions of relatively high hmCG levels in the fetalbrain show relatively lowmCG levels in the adultbrain. Analysis of intragenic hmCG enrichmentrelative to flanking genomic regions, for cell type–specific or developmentally dynamic gene sets(Fig. 5C), showed that neuronal and astrocytegene bodies that are highly enriched with hmCGin adult are also highly enriched at the fetal stage.

Thus, despite lower absolute levels of intra-genic hmCG in the fetal stage, the adult patternsof hmCG enrichment at these cell type–specificgenes are already forming in utero. Constitutivelylowly expressed genes show intragenic depletionof hmCG, in contrast to constitutively highlytranscribed genes, which show localized enrich-ment of hmCG throughout part or all of the genebody.Developmentally down-regulated genes showenrichment of hmCG in the fetal frontal cortex butnot in adults, indicating that reduced transcriptionis accompanied by a loss of hmCG enrichment.Overall, transcriptional activity is associated withintragenic hmCG enrichment, as reported (33),with in utero establishment of adult hmCG pat-terns for cell type–specific genes and loss of hmCenrichment associated with developmentally cou-pled transcriptional down-regulation.

Measurement of mC and hmC in all genes infetal and adult mouse frontal cortex indicated thatboth mCG and hmCG are depleted at promotersand in gene bodies of lowly expressed genes,whereas hmCG is enriched throughout the genebodies of more highly transcribed genes (Fig. 5D).The most highly expressed genes in the adultfrontal cortex show intragenicmCGhypomethyl-ation (Figs. 4 and 5D) but still retain high intra-

genic hmCG. Ranking all genes by transcriptabundance, it is evident that the highest meanintragenic hmCG levels, which occur in the mosthighly transcribed genes, correspond to hmCG/CG~0.25 andmCG/CG~0.5 (fig. S10D). Frontalcortex development is accompanied by increasedenrichment of hmCG at intragenic regions thatare already hyper-hydroxymethylated at the fetalstage (Fig. 5D and fig. S10), demonstrating thatadult patterns of genic hmC are already evident inthe immature fetal brain.

CG Differentially Methylated RegionsEnriched in Regulatory RegionsBecause differences in genic mCGwere observedover development and between neuronal and glialcell populations (Fig. 4), we scanned the human andmouse methylomes to comprehensively identifyCG differentially methylated regions (CG-DMRs)throughout the genome. CG-DMRs were identi-fied between fetal and adult frontal cortex, neu-rons and glia, and combined into four sets: neuronaland glial hyper- and hypo-methylated CG-DMRs.In total, 267,799 human and 142,835 mouseCG-DMRs were identified (median lengths: formouse, 473 bp; human, 533 bp), revealing sev-eral predominant dynamics in mCG during braindevelopment and cellular specialization (Fig. 6Aand fig. S11). Neuronal CG-DMRs are the mostnumerous in both mice and humans, because ofthe very distinct mCG patterns that emerge duringneuronal differentiation and maturation. At thesesites, CG methylation in adult neurons is distinctcompared with those in glial and/or fetal and ear-ly postnatal development frontal cortex tissue sam-ples. Neuronal hypermethylated CG-DMRs alsoshowmCHhypermethylation (fig. S11). Inmouse,mCG/CG within neuronal hypomethylatedCG-DMRs declines to a stable level by 1 weekafter birth. In contrast, neuronal hypermethylatedCG-DMRs do not begin to change until 1 weekafter birth, after which they accumulate mCGuntil 2 to 4 weeks of age. These data indicate thatincreases in neuronal mCG occur during synap-togenesis after most decreases in neuronal mCGhave already occurred. Furthermore, we foundthat hydroxymethylation in the adult cortex ishighest in CG-DMRs that show neuronal hyper-methylation and is depleted from CG-DMRs thatdisplay neuronal hypomethylation (Fig. 6A). Thissuggests that hmCG may be most abundant inneurons, rather than glial cells, in the frontal cortex.

Analysis of the genomic features in whichCG-DMRs are located revealed that although halfare found within gene bodies, they are not com-mon within promoters and transcriptional startand end regions. Instead, they are disproportion-ately located at DHS regions and enhancers uniqueto fetal or adult brain (Fig. 6B). Closer inspectionof the enrichment and depletion of these CG-DMRsrevealed that fetal enhancers andDHS sites uniqueto the fetal brain are enriched for hypermethyla-tion in adult brain but not in the fetal brain (Fig.6C). In contrast, adult enhancers and unique adultDHS sites are highly associated with CG hyper-

methylation in fetal stages but are not associatedwith hypermethylated CG-DMRs in the adultbrain and in neurons. Thus, developmentally dy-namic enhancers and DHS sites in frontal cortexhave dynamic CG methylation that is depletedwhere chromatin accessibility and regulatory el-ement activity increase, consistent with a range ofhuman cell lines (53).

To characterize gene functions associated withthe CG-DMRs, we analyzed the association be-tween proximal genes (transcriptional start sitewithin 5 kb of the DMR) and cell type–specific ordevelopmentally dynamic gene sets (fig. S12A).We observed an inverse relationship betweenmeth-ylation and gene function. Genes associated withneuronal function and up-regulation during devel-opment are enriched for promoter hypermethyl-ation in glia and hypomethylation in neurons,whereas genes down-regulated during brain de-velopment and those related to astrocyte functionare enriched for promoter hypermethylation in neu-rons and hypomethylation in glia. Genes that areconstitutively expressed at either high or low lev-els are not associated with promoter/transcriptionstart site CG-DMRs, indicating that dynamic CGmethylation is highly associated with changes intranscriptional activity and cell type–specific tran-scriptional regulation.

Because the majority of all developmentallydynamic CG-DMRs are associated with DHSsites, we examined the directional relationshipsbetween dynamic mCG and DNA accessibilitystates over development (Fig. 6D). Notably, DHSsites unique to fetal frontal cortex overlap with28% of CG-DMRs that gain methylation throughdevelopment (Adult>Fetal). However, these sitesonly overlap 7.3% of CG-DMRs that lose mCGduring development (Fetal>Adult). Similarly, DHSsites unique to adult frontal cortex rarely overlapAdult>Fetal CG-DMRs. A similar analysis of de-velopmentally dynamic enhancers active in onlyone of the developmental stages (37) (Fig. 6E)showed that enhancer activation is associatedwith mCG hypomethylation of the enhancer,whereas enhancer inactivation is associated withenhancer mCG hypermethylation.

This inverse relationship between genome ac-cessibility and mCG level at putative functionalregions of the genome suggests that nuclear fac-tors that bind the region and increase accessibil-ity may cause localized reduction in mCG, aspreviously reported for a small number of DNAbinding proteins (54). Alternatively, mCG hy-permethylation may cause reduced genome ac-cessibility by direct inhibition of DNA-proteininteractions or induction of chromatin compac-tion, with loss of mCG enabling increased chro-matin accessibility and genome interaction withDNA binding factors.

Discrete regions that show increased or de-creased CGmethylation through development areassociated with specific local chromatin modifi-cations. We found that CG-methylated regions ofthe fetal frontal cortex that become hypomethyl-ated in the adult (Fig. 6F, Fetal>Adult) gain

www.sciencemag.org SCIENCE VOL 341 9 AUGUST 2013 1237905-9

RESEARCH ARTICLE

Page 11: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

localized histone modifications characteristic ofactive enhancers (H3K4me1 and H3K27ac) andincreased DNaseI hypersensitivity in the adult. Inaddition, these regions have reduced accumula-tion of mCH (Fig. 6F), consistent with an overalldecrease in mC linked to increased genome ac-cessibility and enhancer activity. In contrast, ge-nomic regions that gainmCGduring development(Adult>Fetal) lose localizedenrichment ofH3K4me1,H3K27ac, and DNaseI hypersensitivity and showincreasedmCH.These changes indicate inactivationof these genomic regions through brain develop-ment and suggest that this inactivation is associatedwith increased local mCG.

A Hydroxymethylation Signature ofDevelopmentally Activated RegionsAdult>Fetal CG-DMRs show broad low-levelhmCG enrichment flanking the CG-DMR and alocalized depletion of hmC at the center in bothfetal and adult genomes (Fig. 6F), and the ab-solute abundance of hmC is several-fold lower infetal compared with adult frontal cortex (Fig. 1E).This suggests that althoughAdult>FetalCG-DMRsgain mCG through development, they tend tobe refractory to conversion to hmC, potentiallybecause of lower accessibility to the Tet hydrox-ylases. In contrast, Fetal>Adult CG-DMRs havea local enrichment of both mCG and hmCG inthe fetal cortex that becomes a local depletion inthe adult. Two enrichment-based genome-widehmC profiling techniques, CMS-IP (38) and biotin-glucosyl tagging (31), confirmed the localizedenrichment of hmC at Fetal>Adult CG-DMRs(fig. S12B). The localized enrichment of hmC atthese inaccessible and quiescent genomic regions,which lose mCG and hmCG later in develop-ment, indicates that theymay be premodified withhmCG in the fetal stage to create a dormant statethat is poised for subsequent demethylation andactivation at a later developmental stage. Closerinspection of base-resolution hmC data revealedthat 4% of the hmCG bases that have significant-ly higher hmC levels in fetal compared with adult[false discovery rate (FDR) 0.05] directly overlapwith Fetal>Adult CG-DMRs, far exceeding thenumber expected by chance (0.5%). This indi-cates that despite lower global levels of hmC inthe fetal brain, developmentally demethylated CG-DMRs are enriched for hmCG bases that aremore highly hydroxymethylated in fetal than inadult brain (Fig. 6G). The localized enrichmentof mCG at these CG-DMRs in the fetal cortex indi-cates that CG-demethylation has not yet taken place.

If fetal hmCG is poised at dormant genomicregions in order to facilitate active DNA demeth-ylation at later developmental stages, then the Tethydroxylase enzymes that catalyze conversion ofmC to hmC should be necessary for mCG hypo-methylation in the adult frontal cortex at theseregions. To test this, we performed MethylC-Seqof genomicDNA from frontal cortex tissue of adultTet2−/−mice. Adult>Fetal CG-DMRs, which gainmCG through development, are largely unaffectedin Tet2−/− compared with wild-type adult mice

(Fig. 6, H and I; 3.6% hypermethylated, FET,FDR 0.05). By contrast, a substantial fraction ofFetal>Adult CG-DMRs are hypermethylated inTet2−/− (19.7%) versus wild type. The mutantshows a small but significant increase in mCG atFetal>Adult CG-DMRs (Fig. 6I and fig. S12C)(Tet2−/−: 7.9% T 4.6%, P < 10−11, Wilcoxonsigned rank test). The partial effect of the mutationonCGmethylation is not unexpected given thatall three Tet genes are expressed in the brain(fig. S9C) and may exhibit some functional re-dundancy. Additionally, a genome-wide searchidentified 14,340CG-DMRs hypermethylated inTet2−/− relative to wild type (6 weeks, 10 weeks,and 22 months), >fourfold more numerous thanhypomethylated CG-DMRs (3099). This furtherindicates a role for Tet2 in mediating mCG de-methylation during brain development.

DiscussionThe essential role of frontal cortex in behaviorand cognition requires the coordinated interac-tion, via electrical and chemical signaling, of mul-tiple neuronal cell types and a diverse populationof glial cells. Individual brain cells have uniqueroles within circuits that are defined by their lo-cation and pattern of connections as well as bytheir molecular identity. The development andmaturation of the brain’s physical structure andthe refinement of the molecular identities of neu-rons and glial cells occur in parallel in a finelyorchestrated process that starts early during theembryonic period and continues, in humans, wellinto the third decade of life (55, 56). An earlypostnatal burst of synaptogenesis is followed byactivity-dependent pruning of excess synapsesduring adolescence (28, 57, 58). This processforms the basis for experience-dependent plastic-ity and learning in children and young adults (59),and its disruption leads to behavioral alterationsand neuropsychiatric disorders (60). During thisperiod, profound transcriptional changes lead tothe appearance of adult electrophysiological char-acteristics in neocortical neurons.

Our study suggests a key role of DNA meth-ylation in brain development and function. First,CHmethylation accumulates significantly in neu-rons through early childhood and adolescence,becoming the dominant form of DNA methyla-tion in mature human neurons. This shows thatthe period of synaptogenesis, during which theneural circuit matures, is accompanied by a par-allel process of large-scale reconfiguration of theneuronal epigenome. Indeed, central nervous sys-tem deletion of Dnmt3a during late gestation in-duces motor deficits, and animals die prematurely(61). However, mice with a postnatal deletionrestricted to the pyramidal cell population (com-plete recombination around 1 month old) do notshow overt behavioral or transcriptional altera-tions (2). Our data suggest that expression ofDnmt3a specifically around the second postnatalweek may be critical for establishing a normalbrain DNA methylation profile and allowinghealthy brain development.

Second, the precise positioning of mCG andmCH marks, which are conserved between indi-viduals and across humans and mice, is consist-ent with a functional role.Whether this is the case,or whether the conserved patterns are instead areflection of conserved nucleosome position orchromatin structure, requires further investigation.Third, the relationship between DNAmethylationpatterns and the function of neuron- or astrocyte-specific gene sets suggests a role for DNA meth-ylation in distinguishing these two broad classesof cortical cells. DNA methylation could there-fore play a key role in sculpting more-specificcellular identities. If this is the case, we expectthat purified subpopulations will reveal high spec-ificity of methylation at specific sites for partic-ular cell types. Thus, the observation that mostCH sites with nonzero methylation are methyl-ated in ~20 to 25% of sampled cells (fig. S1H)could be explained by the heterogeneity of thesebrain circuits rather than by stochastic methyla-tion within each cell. These conclusions obtainedfrom our genome-wide, base-resolution, cell type–specific DNAmethylomes for brain cells throughkey stages of development are the first steps towardunraveling the genetic program and experience-dependent epigenetic modifications leading to afully differentiated nervous system.

References and Notes1. E. Borrelli, E. J. Nestler, C. D. Allis, P. Sassone-Corsi,

Decoding the epigenetic language of neuronal plasticity.Neuron 60, 961–974 (2008). doi: 10.1016/j.neuron.2008.10.012; pmid: 19109904

2. J. Feng et al., Dnmt1 and Dnmt3a maintain DNAmethylation and regulate synaptic function in adultforebrain neurons. Nat. Neurosci. 13, 423–430 (2010).doi: 10.1038/nn.2514; pmid: 20228804

3. C. A. Miller, J. D. Sweatt, Covalent modification of DNAregulates memory formation. Neuron 53, 857–869(2007). doi: 10.1016/j.neuron.2007.02.022;pmid: 17359920

4. S. Numata et al., DNA methylation signatures indevelopment and aging of the human prefrontal cortex.Am. J. Hum. Genet. 90, 260–272 (2012). doi: 10.1016/j.ajhg.2011.12.020; pmid: 22305529

5. M. Suderman et al., Conserved epigenetic sensitivity toearly life experience in the rat and human hippocampus.Proc. Natl. Acad. Sci. U.S.A. 109 (suppl. 2), 17266–17272(2012). doi: 10.1073/pnas.1121260109; pmid: 23045659

6. K. D. Siegmund et al., DNA methylation in the humancerebral cortex is dynamically regulated throughout thelife span and involves differentiated neurons. PLoS ONE2, e895 (2007). doi: 10.1371/journal.pone.0000895;pmid: 17878930

7. K. Martinowich et al., DNA methylation-related chromatinremodeling in activity-dependent BDNF gene regulation.Science 302, 890–893 (2003). doi: 10.1126/science.1090842; pmid: 14593184

8. J. U. Guo et al., Neuronal activity modifies the DNAmethylation landscape in the adult brain. Nat. Neurosci.14, 1345–1351 (2011). doi: 10.1038/nn.2900;pmid: 21874013

9. I. B. Zovkic, M. C. Guzman-Karlsson, J. D. Sweatt,Epigenetic regulation of memory formation andmaintenance. Learn. Mem. 20, 61–74 (2013).doi: 10.1101/lm.026575.112; pmid: 23322554

10. J. M. Levenson et al., Evidence that DNA (cytosine-5)methyltransferase regulates synaptic plasticity in thehippocampus. J. Biol. Chem. 281, 15763–15773 (2006).doi: 10.1074/jbc.M511767200; pmid: 16606618

11. A. M. M. Oliveira, T. J. Hemstedt, H. Bading, Rescue ofaging-associated decline in Dnmt3a2 expression restores

9 AUGUST 2013 VOL 341 SCIENCE www.sciencemag.org1237905-10

RESEARCH ARTICLE

Page 12: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

cognitive abilities. Nat. Neurosci. 15, 1111–1113 (2012).doi: 10.1038/nn.3151; pmid: 22751036

12. P. Moretti et al., Learning and memory and synapticplasticity are impaired in a mouse model of Rettsyndrome. J. Neurosci. 26, 319–327 (2006).doi: 10.1523/JNEUROSCI.2623-05.2006; pmid: 16399702

13. N. W. Penn, R. Suwalski, C. O’Riley, K. Bojanowski,R. Yura, The presence of 5-hydroxymethylcytosine inanimal deoxyribonucleic acid. Biochem. J. 126, 781–790(1972). pmid: 4538516

14. N. W. Penn, Modification of brain deoxyribonucleic acidbase content with maturation in normal andmalnourished rats. Biochem. J. 155, 709–712 (1976).pmid: 949331

15. D. Globisch et al., Tissue distribution of 5-hydroxymethylcytosineand search for active demethylation intermediates. PLoS ONE5, e15367 (2010). doi: 10.1371/journal.pone.0015367;pmid: 21203455

16. J. U. Guo, Y. Su, C. Zhong, G.-L. Ming, H. Song,Hydroxylation of 5-methylcytosine by TET1 promotesactive DNA demethylation in the adult brain. Cell 145,423–434 (2011). doi: 10.1016/j.cell.2011.03.022;pmid: 21496894

17. Y.-F. He et al., Tet-mediated formation of 5-carboxylcytosine and its excision by TDG in mammalianDNA. Science 333, 1303–1307 (2011). doi: 10.1126/science.1210944; pmid: 21817016

18. C.-X. Song et al., Genome-wide profiling of 5-formylcytosine reveals its roles in epigenetic priming. Cell153, 678–691 (2013). doi: 10.1016/j.cell.2013.04.001;pmid: 23602153

19. L. Shen et al., Genome-wide analysis reveals TET- andTDG-dependent 5-methylcytosine oxidation dynamics. Cell153, 692–706 (2013). doi: 10.1016/j.cell.2013.04.002;pmid: 23602152

20. W. Xie et al., Base-resolution analyses of sequence andparent-of-origin dependent DNA methylation in themouse genome. Cell 148, 816–831 (2012).doi: 10.1016/j.cell.2011.12.035; pmid: 22341451

21. K. E. Varley et al., Dynamic DNA methylation acrossdiverse human cell lines and tissues. Genome Res. 23,555–567 (2013). doi: 10.1101/gr.147942.112;pmid: 23325432

22. R. Lister et al., Human DNA methylomes at baseresolution show widespread epigenomic differences. Nature462, 315–322 (2009). doi: 10.1038/nature08514;pmid: 19829295

23. M. J. Ziller et al., Genomic distribution and inter-samplevariation of non-CpG methylation across human cell types.PLoS Genet. 7, e1002389 (2011). doi: 10.1371/journal.pgen.1002389; pmid: 22174693

24. R. Lister et al., Highly integrated single-base resolutionmaps of the epigenome in Arabidopsis. Cell 133,523–536 (2008). doi: 10.1016/j.cell.2008.03.029;pmid: 18423832

25. R. Lister et al., Hotspots of aberrant epigenomicreprogramming in human induced pluripotent stem cells.Nature 471, 68–73 (2011). doi: 10.1038/nature09798;pmid: 21289626

26. B. H. Ramsahoye et al., Non-CpG methylation isprevalent in embryonic stem cells and may be mediatedby DNA methyltransferase 3a. Proc. Natl. Acad. Sci.U.S.A. 97, 5237–5242 (2000). doi: 10.1073/pnas.97.10.5237; pmid: 10805783

27. J. De Felipe, P. Marco, A. Fairén, E. G. Jones, Inhibitorysynaptogenesis in mouse somatosensory cortex. Cereb.Cortex 7, 619–634 (1997). doi: 10.1093/cercor/7.7.619;pmid: 9373018

28. P. R. Huttenlocher, A. S. Dabholkar, Regional differencesin synaptogenesis in human cerebral cortex. J. Comp.Neurol. 387, 167–178 (1997). doi: 10.1002/(SICI)1096-9861(19971020)387:2<167::AID-CNE1>3.0.CO;2-Z;pmid: 9336221

29. B. P. Berman et al., Regions of focal DNAhypermethylation and long-range hypomethylation incolorectal cancer coincide with nuclear lamina-associateddomains. Nat. Genet. 44, 40–46 (2012). doi: 10.1038/ng.969;pmid: 22120008

30. S. Kriaucionis, N. Heintz, The nuclear DNA base5-hydroxymethylcytosine is present in Purkinje neurons

and the brain. Science 324, 929–930 (2009);10.1126/science.1169786. doi: 10.1126/science.1169786; pmid: 19372393

31. C.-X. Song et al., Selective chemical labeling reveals thegenome-wide distribution of 5-hydroxymethylcytosine.Nat. Biotechnol. 29, 68–72 (2011). doi: 10.1038/nbt.1732; pmid: 21151123

32. K. E. Szulwach et al., 5-hmC-mediated epigeneticdynamics during postnatal neurodevelopment and aging.Nat. Neurosci. 14, 1607–1616 (2011). doi: 10.1038/nn.2959; pmid: 22037496

33. M. Mellén, P. Ayata, S. Dewell, S. Kriaucionis, N. Heintz,MeCP2 binds to 5hmC enriched within active genes andaccessible chromatin in the nervous system. Cell 151,1417–1430 (2012). doi: 10.1016/j.cell.2012.11.022;pmid: 23260135

34. M. Yu et al., Base-resolution analysis of 5-hydroxymethylcytosinein the mammalian genome. Cell 149, 1368–1380 (2012).doi: 10.1016/j.cell.2012.04.027; pmid: 22608086

35. J. A. Stamatoyannopoulos et al., An encyclopedia ofmouse DNA elements (Mouse ENCODE). Genome Biol.13, 418 (2012). doi: 10.1186/gb-2012-13-8-418;pmid: 22889292

36. R. K. Auerbach et al., Mapping accessible chromatinregions using Sono-Seq. Proc. Natl. Acad. Sci. U.S.A.106, 14926–14931 (2009). doi: 10.1073/pnas.0905443106; pmid: 19706456

37. Y. Shen et al., A map of the cis-regulatory sequences inthe mouse genome. Nature 488, 116–120 (2012).doi: 10.1038/nature11243; pmid: 22763441

38. W. A. Pastor et al., Genome-wide mapping of5-hydroxymethylcytosine in embryonic stem cells.Nature 473, 394–397 (2011). doi: 10.1038/nature10102;pmid: 21552279

39. Y. Huang, W. A. Pastor, J. A. Zepeda-Martínez, A. Rao,The anti-CMS technique for genome-wide mapping of5-hydroxymethylcytosine. Nat. Protoc. 7, 1897–1908(2012). doi: 10.1038/nprot.2012.103; pmid: 23018193

40. A. Magklara et al., An epigenetic signature formonoallelic olfactory receptor expression. Cell 145,555–570 (2011). doi: 10.1016/j.cell.2011.03.040;pmid: 21529909

41. E. J. Clowney et al., Nuclear aggregation of olfactoryreceptor genes governs their monogenic expression. Cell151, 724–737 (2012). doi: 10.1016/j.cell.2012.09.043;pmid: 23141535

42. H. Li et al., Transcription factor MEF2C influencesneural stem/progenitor cell differentiation andmaturation in vivo. Proc. Natl. Acad. Sci. U.S.A. 105,9397–9402 (2008). doi: 10.1073/pnas.0802876105;pmid: 18599437

43. M. W. Akhtar et al., In vivo analysis of MEF2 transcriptionfactors in synapse regulation and neuronal survival.PLoS ONE 7, e34863 (2012). doi: 10.1371/journal.pone.0034863; pmid: 22496871

44. A. C. Barbosa et al., MEF2C, a transcription factor thatfacilitates learning and memory by negative regulation ofsynapse numbers and function. Proc. Natl. Acad.Sci. U.S.A. 105, 9391–9396 (2008). doi: 10.1073/pnas.0802679105; pmid: 18599438

45. J. D. Cahoy et al., A transcriptome database for astrocytes,neurons, and oligodendrocytes: a new resource forunderstanding brain development and function. J. Neurosci.28, 264–278 (2008). doi: 10.1523/JNEUROSCI.4178-07.2008; pmid: 18171944

46. H. Wu et al., Dnmt3a-dependent nonpromoter DNAmethylation facilitates transcription of neurogenic genes.Science 329, 444–448 (2010). doi: 10.1126/science.1190485; pmid: 20651149

47. T. Mohandas, R. S. Sparkes, L. J. Shapiro, Reactivation ofan inactive human X chromosome: Evidence for Xinactivation by DNA methylation. Science 211, 393–396(1981). doi: 10.1126/science.6164095; pmid: 6164095

48. K. Plath, S. Mlynarczyk-Evans, D. A. Nusinow, B. Panning,Xist RNA and the mechanism of X chromosomeinactivation. Annu. Rev. Genet. 36, 233–278 (2002).doi: 10.1146/annurev.genet.36.042902.092433;pmid: 12429693

49. L. Carrel, H. F. Willard, X-inactivation profile revealsextensive variability in X-linked gene expression in

females. Nature 434, 400–404 (2005). doi: 10.1038/nature03479; pmid: 15772666

50. A. J. Sharp et al., DNA methylation profiles of humanactive and inactive X chromosomes. Genome Res. 21,1592–1600 (2011). doi: 10.1101/gr.112680.110;pmid: 21862626

51. C. M. Johnston et al., Large-scale population studyof human cell lines indicates that dosagecompensation is virtually complete. PLoS Genet.4, e9 (2008). doi: 10.1371/journal.pgen.0040009;pmid: 18208332

52. A. K. Maunakea et al., Conserved role of intragenic DNAmethylation in regulating alternative promoters. Nature466, 253–257 (2010). doi: 10.1038/nature09165;pmid: 20613842

53. R. E. Thurman et al., The accessible chromatin landscapeof the human genome. Nature 489, 75–82 (2012).doi: 10.1038/nature11232; pmid: 22955617

54. M. B. Stadler et al., DNA-binding factors shape themouse methylome at distal regulatory regions. Nature480, 490–495 (2011). pmid: 22170606

55. Z. Petanjek et al., Extraordinary neoteny ofsynaptic spines in the human prefrontal cortex.Proc. Natl. Acad. Sci. U.S.A. 108, 13281–13286(2011). doi: 10.1073/pnas.1105108108pmid: 21788513

56. B. Kolb et al., Experience and the developing prefrontalcortex. Proc. Natl. Acad. Sci. U.S.A. 109(suppl. 2), 17186–17193 (2012). doi: 10.1073/pnas.1121251109; pmid: 23045653

57. H. T. Chugani, A critical period of brain development:Studies of cerebral glucose utilization with PET.Prev. Med. 27, 184–188 (1998). doi: 10.1006/pmed.1998.0274; pmid: 9578992

58. P. Levitt, Structural and functional maturation of thedeveloping primate brain. J. Pediatr. 143 (suppl.),S35–S45 (2003). doi: 10.1067/S0022-3476(03)00400-1;pmid: 14597912

59. M. V. Johnston, Plasticity in the developing brain:implications for rehabilitation. Dev. Disabil. Res. Rev. 15,94–101 (2009). doi: 10.1002/ddrr.64; pmid: 19489084

60. P. J. Uhlhaas, W. Singer, The development of neuralsynchrony and large-scale cortical networks duringadolescence: Relevance for the pathophysiology ofschizophrenia and neurodevelopmental hypothesis.Schizophr. Bull. 37, 514–523 (2011). doi: 10.1093/schbul/sbr034; pmid: 21505118

61. S. Nguyen, K. Meletis, D. Fu, S. Jhaveri, R. Jaenisch,Ablation of de novo DNA methyltransferaseDnmt3a in the nervous system leads toneuromuscular defects and shortened lifespan.Dev. Dyn. 236, 1663–1676 (2007). doi: 10.1002/dvdy.21176; pmid: 17477386

62. Materials and methods are available as supplementarymaterials on Science Online.

Acknowledgments: We thank W. F. Loomis for criticalreading of this manuscript; J. Chambers for technicalassistance with animal breeding; D. Chambers for technicalassistance with cell sorting; M. Lutz and T. Berggren forprovision of the HUES6 cells; F. Yue, G. Hon, and B. Renfor providing mapped mouse ChIP-Seq data and assistancewith TAB-Seq; and A. Nimmerjahn for providing theS100b-eGFP mice. Human brain tissue was obtained fromthe National Institute of Child Health and Human DevelopmentBrain and Tissue Bank for Developmental Disorders at theUniversity of Maryland, Baltimore, Maryland, the IDIBELLBiobank, which is part of the BrainNet Europe Bank funded bythe European Commission (LSHM-CT-2004-503039). We thankthe Stamatoyannopoulos laboratory (University of Washington)and the Mouse ENCODE Consortium for generating andproviding access to the DNaseI hypersensitivity data sets (35).This work was supported by a National Institutes of MentalHealth grant to M.M.B. and J.R.E. (MH094670); the HowardHughes Medical Institute to T.J.S. and J.R.E.; the Gordon andBetty Moore Foundation (GMBF3034) to J.R.E.; NIH grantHG006827 to C.H.; NIH RO1 grants AI44432, HD065812, andCA151535; grant RM-01729 from the California Institute ofRegenerative Medicine; and Translational Research grant

www.sciencemag.org SCIENCE VOL 341 9 AUGUST 2013 1237905-11

RESEARCH ARTICLE

Page 13: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

TRP 6187-12 from the Leukemia and Lymphoma Society(to A.R.). R.L. was supported by an Australian Research CouncilFuture Fellowship. Support was also provided by the Governmentof Western Australia through funding for the WesternAustralia CoE for Computational Systems Biology. E.A.M. wassupported by the Center for Theoretical Biological Physics,University of California San Diego, and NIH (National Instituteof Neurological Diseases and Stroke grant K99NS080911).W.A.P. was supported by a predoctoral graduate research

fellowship from the NSF. Analyzed data sets can be accessedat http://neomorph.salk.edu/brain_methylomes. Sequencedata can be downloaded from National Center forBiotechnology Information GEO (GSE47966). Tet2 mutantmice are available under a material transfer agreement fromthe La Jolla Institute for Allergy and Immunology. Correspondenceand requests for materials should be addressed to R.L.([email protected]), M.M.B. ([email protected]), andJ.R.E. ([email protected]).

Supplementary Materialswww.sciencemag.org/content/341/6146/1237905/suppl/DC1Materials and MethodsFigs. S1 to S12Tables S1 to S5References (63–78)

15 March 2013; accepted 7 June 201310.1126/science.1237905

9 AUGUST 2013 VOL 341 SCIENCE www.sciencemag.org1237905-12

RESEARCH ARTICLE

Page 14: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

www.sciencemag.org/cgi/content/full/science.1237905/DC1

Supplementary Materials for

Global Epigenomic Reconfiguration During Mammalian Brain Development

Ryan Lister,* Eran A. Mukamel, Joseph R. Nery, Mark Urich, Clare A. Puddifoot, Nicholas D. Johnson, Jacinta Lucero, Yun Huang, Andrew J. Dwork, Matthew D.

Schultz, Miao Yu, Julian Tonti-Filippini, Holger Heyn, Shijun Hu, Joseph C. Wu, Anjana Rao, Manel Esteller, Chuan He, Fatemeh G. Haghighi, Terrence J. Sejnowski, M.

Margarita Behrens,* Joseph R. Ecker*

*Corresponding author. E-mail: [email protected] (R.L.); [email protected] (M.M.B.); [email protected] (J.R.E.)

Published 4 July 2013 on Science Express

DOI: 10.1126/science.1237905

This PDF file includes:

Materials and Methods Figs. S1 to S12 Table S1 to S5 captions References

Other Supplementary Material for this manuscript includes the following: available at www.sciencemag.org/cgi/content/full/science.1237905/DC1

Tables S1 to S5

Page 15: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

2

Materials and Methods Animals The mouse lines C57Bl/6 and S100b-eGFP (B6;D2-Tg(S100B-EGFP)1Wjt/J), both from Jackson Laboratories, ME were bred and maintained in our animal facility in 12 h light/dark cycles with food ad libitum. To produce the S100b-eGFP animals used for FACS, animals were crossed to C57BL/6 to produce heterozygotes for eGFP expression. Tet2-/- mice were produced by targeted disruption of the Tet2 gene as described previously (63). Animals were weaned in groups of 3-4 per cage, and used at the postnatal time-points indicated. Three animals obtained from separate litters were processed together for DNA and RNA isolations. All protocols were approved by the Salk Institute's Institutional Animal Care and Use Committee (IACUC). Human samples Human brain tissue was obtained from the NICHD Brain and Tissue Bank for Developmental disorders at the University of Maryland, Baltimore, MD, the IDIBELL Biobank (Barcelona, Spain), which is part of the BrainNet Europe Bank (Munich, Germany) funded by the European Commission (LSHM-CT-2004-503039). Procedures were approved by the Institutional Review Board of the Salk Institute.

Tissue production Animals were anesthetized with isoflurane and decapitated. Brains were quickly extracted and placed in dissection media (64). The frontal cortex was obtained by slicing the adult brains coronally at Bregma 1 mm, and dissecting the frontal cortical tissue, being careful to avoid contamination with olfactory or putamen regions, under a dissection microscope. Meninges were then dissected out from the frontal cortex, which was rapidly frozen on dry ice or minced and processed for nuclei isolation and FACS. Brain dissections in younger animals occurred at a similar coordinate (i.e., when the genu of the corpus callosum unites between the two hemispheres). For the embryonic cortex, the front third of the cortical plate was dissected under a microscope. Protein isolation and Western blot Tissue dissected as above was disrupted in RIPA buffer (Hepes 50 mM, pH 8.0; 1% NP-40; 0.7% Deoxycholate; 0.5 M LiCl; complete protease inhibitors) with the addition 0.5% SDS, by incubation on ice for 30 min followed by centrifugation at 15,000 x g to separate nucleic acids. Protein was determined using a commercial kit (BCA, Pierce, Rockford, IL)). Fifty micrograms of protein for each timepoint were separated in 7.5% SDS-PAGE gels and transferred onto nitrocellulose membranes. Dnmt3a was detected by incubation with anti-Dnmt3a antibody (1:300, Imgenex Cat# IMG-268A, San Diego, CA) followed by HRP conjugated secondary antibodies and chemiluminescence (Cell Signaling, Danvers, MA) in four independent experiments.

Nuclei isolation

Page 16: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

3

Mouse: Frontal cortices from four animals for each time-point were processed as described (65) with the following modification: isolated nuclei from 4 animals were incubated for 1 h at 4°C in 1 ml of PBS + 2% horse serum containing a 1:1000 dilution of AlexaFluor488 conjugated anti-NeuN monoclonal antibody (Millipore Cat# MAB377X, Temecula, CA) before FACS. Human: production of nuclei followed a similar procedure as for mice. Freshly frozen samples of middle frontal gyrus were obtained from the NICHD Brain and Tissue Bank for Developmental disorders at the University of Maryland. Sample demographics are described in Table S5. Dissociation of frontal cortex tissue for FACS Frontal cortices from S100b-eGFP mice at 7-8 week old, obtained as above were minced and placed in oxygen bubbled dissection media containing 1 mg/ml freshly prepared pronase. After incubation at 37°C for 30 min, the tissue was washed thoroughly with the same media, resuspended in 5 ml PBS containing 1 % horse serum and gently triturated using a 5 ml pipette. The solution was then filtered through a 70 µm nylon mesh and subjected to FACS.

FACS Cells (S100b-eGFP) or nuclei (NeuN) were sorted using a FacsVantage SE DiVacell sorter (BD Biosciences, San Jose, CA) using an 80 µm tip. Singlet cells were gated based on forward scatter pulse height and pulse width characteristics. eGFP positive cells or AlexaFluor488 labeled nuclei were discriminated from autofluorescence by plotting green fluorescence versus orange fluorescence, and bright expressors were chosen for sorting. Sorted cells or nuclei were centrifuged and pellets rapidly frozen in dry ice until processed for nucleic acid isolation.

MethylC-Seq MethylC-Seq library generation, read mapping, processing, and analysis were performed as described previously (25), aligning reads to the mouse mm9 and human hg18 reference genomes, except a previous filter that excluded reads containing >3 cytosine bases in the CH context was not applied in this study. Library amplification was performed with either PfuTurboCx Hotstart DNA polymerase (Agilent Technologies, Santa Clara, CA) or KAPA HiFi HotStart Uracil+ DNA polymerase (Kapa Biosystems, Woburn, MA). mRNA-Seq mRNA-Seq libraries were generated from total RNA with polyA+ selection of mRNA using the TruSeq RNA Sample Prep Kit v2 (Illumina, San Diego, CA). Strand-specific libraries were constructed using a dUTP methodology as described previously (66), while non-strand-specific libraries were constructed with the TruSeq RNA Sample prep Kit vs as per manufacturer’s instructions. TopHat2 and Cufflinks2 packages (67, 68) were used to map sequence reads to the mouse mm9 and human hg18 reference genomes and quantitate differential gene expression. Enrichment-based detection of 5-hydroxymethylcytosine (hmC)

Page 17: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

4

Cytosine 5-methylenesulphonate immunoprecipitation (CMS-IP): mouse E13 fetal cortex or 10 wk frontal cortex genomic DNA fragments (100 - 200 bp) were first end repaired, 3’ adenylated, ligated to methylated Illumina adaptor oligonucleotides and bisulfite converted as per the MethylC-Seq protocol (25). Bisulfite conversion of hmC to cytosine 5-methylenesulphonate (CMS) was followed by immunoprecipitation of gDNA fragments containing CMS with a specific antiserum to CMS. Bisulfite converted gDNA was first denatured for 10 minutes at 95°C (0.4 M NaOH, 10 mM EDTA), neutralized by addition of cold 2 M Ammonium Acetate pH 7.0, incubated with anti-CMS antiserum in 1x IP buffer (10 mM sodium phosphate pH 7.0, 140 mM NaCl, 0.05% Triton X-100) for 2 h to overnight at 4°C, and then precipitated with Protein G beads. Precipitated DNA was washed 3 times with 1x IP buffer and eluted with Proteinase K, then purified with Phenol Chloroform. Following immunoprecipitation, sequencing libraries were amplified by 8 cycles of PCR and DNA sequencing was performed using a HiSeq 2000 Genome Analyzer (Illumina) as described previously (38). Biotin-glucosyl tagging and enrichment of hmC (biotin-gmC): the Hydroxymethyl Collector Kit (Active Motif, Carlsbad, CA) was used to selectively tag hmC bases within mouse E13 fetal cortex or 6 wk frontal cortex genomic DNA fragments (100 - 200 bp) with glucose and biotin moieties to form biotin-N3-5-gmC (biotin-gmC), as described previously (31). Genomic DNA fragments containing biotin-gmC were then enriched and purified by high-affinity capture using streptavidin magnetic beads. Enriched DNA containing biotin-gmC was used to generate Illumina DNA sequencing libraries with the TruSeq DNA Sample Prep Kit v2 and sequenced using a HiSeq 2000 Genome Analyzer (Illumina). TAB-Seq Performed as described in Yu et al. 2012 (34). Conversion and protection control sequences, read filtering, and quantitation of hmC protection and mC non-conversion are detailed below (Correction for non-conversion and protection).

Data Analysis Methods Bisulfite non-conversion rate estimation for MethylC-Seq To estimate the bisulfite non-conversion frequency, the frequency of all cytosine basecalls at reference cytosine positions in the lambda genome (unmethylated spike in control) was normalized by the total number of base-calls at reference positions in the lambda genome. This was performed for all cytosines as well as specific sequence contexts (CG, CH, CA, CC, CT, see Table S1). Identification of CG Differentially Methylated Regions (CG-DMRs) To identify DMRs, we used a two-step process. The first step involved performing a root mean square test, as outlined previously (69), on each individual CG. For this test, we constructed a contingency table where the rows indicated a particular sample and the columns indicated the number of reads that supported a methylated cytosine or an unmethylated cytosine at this position in a given sample. The p-values were simulated

Page 18: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

5

using 5000 permutations. For each permutation, a new contingency table was generated by randomly assigning reads to cells with a probability equal to the product of the row marginal and column marginal divided by the total number of reads squared. To speed up this process, if a p-value returned 50 permutations with a statistic greater than or equal to the original test statistic, we stopped running permutations, thus using adaptive permutation testing. To determine a p-value cutoff that would control the false discovery rate (FDR) at our desired rate (see details for each DMR analysis, below), we used the procedure outlined previously (70). Briefly, this method first generates a histogram of the p-values and calculates the expected number of p-values to fall in a particular bin under the null. This expected count is computed by multiplying the width of the bin by the current estimate for the number of true null hypotheses (m0), which is initialized to the number of tests performed. It then looks for the first bin where the expected number of p-values is greater than or equal to the observed value, starting from the most significant bin and working its way towards the least significant. The differences between the expected and observed counts in all the bins up to this point are summed, and a new estimate of m0 is generated by subtracting this sum from the current total number of tests. This procedure was iterated until convergence, which we defined as a change in the m0 estimate less than or equal to the FDR. With this m0 estimate, we were able to estimate the FDR of a given p-value by multiplying the p-value by the m0 estimate (the expected number of positives at that cutoff under the null hypothesis) and dividing that product by the total number of significant tests we detected at that p-value cutoff. We chose the largest p-value cutoff that still satisfied our FDR requirement. Once this p-value cutoff was chosen, for mCG-DMR analysis, significant sites were combined into blocks if they were within 500 bases of one another and had methylation changes in the same direction, for example if at both sites sample A was hypermethylated and sample B was hypomethylated. Furthermore, blocks that contained fewer than 4 differentially methylated sites were discarded. Finally, CG-DMR blocks were filtered based on a requirement for a minimum number of samples to all show the same significant differential methylation patterns. These sample comparison details for the CG-DMR sets described in this study are described below.

- Human fetal vs adult frontal cortex CG-DMRs were identified (FDR = 0.05) where fetal frontal cortex was differentially methylated compared to both 16 yr and 25 yr frontal cortex samples.

- Human NeuN+/NeuN- vs fetal frontal cortex CG-DMRs were identified (FDR = 0.05) between NeuN+/NeuN- samples and fetal frontal cortex, requiring both human NeuN+/NeuN- samples (53 yr female R1, 55 yr male R2) to be differentially methylated compared to fetal.

- Human NeuN+ vs NeuN- CG-DMRs were identified (FDR = 0.05) between NeuN+ and NeuN- samples, requiring both human individuals (53 yr female, 55 yr male) to be differentially methylated between cell types.

- Mouse fetal vs adult frontal cortex CG-DMRs were identified (FDR = 0.01) where fetal frontal cortex was differentially methylated compared to both 6 wk and 10 wk frontal cortex samples.

- Mouse NeuN+/NeuN- vs fetal frontal cortex CG-DMRs were identified (FDR = 0.01) between NeuN+/NeuN- samples and fetal frontal cortex, requiring !2 of 3 mouse

Page 19: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

6

NeuN+/NeuN- samples (7 wk male R1, 6 wk female R2, 12 mo female R3) to be differentially methylated compared to fetal.

- Mouse NeuN+ vs NeuN- CG-DMRs were identified (FDR = 0.01) between NeuN+ and NeuN- samples, requiring !2 of 3 mouse samples (7 wk male R1, 6 wk female R2, 12 mo female R3) to be differentially methylated between cell types.

- Mouse Tet2-/- vs wild-type CG-DMRs were identified (FDR = 0.01) where Tet2-/- was differentially methylated compared to ! 2 of the 6 wk, 10 wk and 22 mo wild-type adult frontal cortex samples.

- Mouse fetal hmCG vs 6 wk hmCG CG-DMRs were identified (FDR = 0.05) as above except individual significantly differentially hydroxymethylated bases were identified and no block joining was performed.

DNaseI hypersensitivity Data generated by the Mouse ENCODE Consortium (35) was mapped to the mouse mm9 reference genome with Bowtie (71) and peaks of read enrichment were identified using MACS (72). ChIP-Seq and enhancers Mouse E14.5 embryonic brain predicted enhancer regions, adult cortex predicted enhancer regions, and sequence reads mapped to mm9 for H3K4me1, H3K27ac and input were from Shen et al. 2012 (37). Coordinates of predicted enhancer regions that overlapped with DNaseI hypersensitive regions were adjusted to center the enhancers on the midpoint of DNaseI hypersensitive sites. Dnmt3a binding sites Liftover was used to convert the coordinates of Dnmt3a binding sites identified in mouse postnatal neuronal stem cells (46) to the mm9 reference. IP detection of hmC CMS-IP reads were mapped to the mouse mm9 genome as described previously for MethylC-Seq sequencing data (25). Biotin-gmC reads were mapped to the mouse mm9 genome using Bowtie2 (71). Estimation of methylation level (mC/C) with correction for non-conversion and protection The majority of our analyses of DNA methylation are based on the level (mC/C, mCG/CG, or mCH/CH), which is an estimate of the fraction of cytosines in the sequenced population which are methylated. To estimate mC/C, we computed the fraction of all MethylC-Seq base-calls at cytosine reference positions that were cytosine (protected from bisulfite conversion). We then corrected these estimates for known experimental biases. Both the whole-genome bisulfite sequencing and TAB-Seq techniques used in this study are subject to a low error rate, r, due to the failure of the chemical conversion of unmethylated cytosine to uracil. Such errors leads to false positive detections of mC. In addition, TAB-Seq suffers from a small rate of error due to non-protection, s, which is the probability that a hydroxymethylated cytosine is converted

Page 20: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

7

by the bisulfite reaction and leads to a false negative (missed call). These rates were calibrated as described previously (34), summarized as follows.

TAB-Seq cytosine non-conversion and hmC protection rates: M.SssI-converted pUC19 DNA (NEB), fully methylated (mC) at all CG cytosines, was spiked into the mouse genomic DNA sample at 0.5% (w/w) prior to library preparation (34). The frequency of C basecalls at pUC19 CG reference positions was calculated from reads uniquely mapped to the pUC19 reference, following removal of clonal reads and reads containing >3 cytosine basecalls in the CH context (34). The cytosine non-conversion rate represents the aggregate frequency of failed mC conversion by Tet1 and failure of bisulfite conversion.

hmC non-protection rate: The 38-48 kb region of the unmethylated lambda phage genome (Promega, Madison, WI) was amplified using ZymoTaq DNA polymerase in 5 separate, nonoverlapping 2 kb PCR amplicons in the presence of dATP, dGTP, dTTP and d5hmCTP (Zymo Research), as described previously (34). PCR products purified by agarose-gel electrophoresis were spiked into the mouse genomic DNA sample at 0.5% (w/w) prior to library preparation. The frequency of C basecalls at CG reference positions (protection frequency) was calculated from all reads uniquely mapped to the 38-48 kb region of the lambda genome, only considering CG reference positions where the closest neighboring reference C is at least 4 bases away (e.g. hmCGNNhmC), in order to reflect the subset of bases in the similar hmC context as mammalian genomes (34). Finally, given the commercial source of 5hmdCTP used to synthesize the control contains ~5% dCTP, which does not get protected, the protection frequency was increased by 5%. Thus, the non-protection frequency (1 - protection frequency) represents the rate at which hmC bases failed to be glucosylated and protected from Tet1 conversion and subsequent bisulfite conversion.

We corrected the observed number of converted and protected reads (a, c respectively) to take these rates into account. If the true rate of methylation or hydroxymethylation at a given site is p, the probability of observing (c, a) is Binom[c|a+c,q], where ! ! ! !! ! ! !! ! ! is the probability of a read being sequenced as cytosine. The maximum likelihood estimate of p is therefore ! ! !!!

!!!!! !!!!!!!!!!!!!!! . When ! ! ! we set ! ! !! In all of our samples, !! ! ! !. For some analyses,

we therefore used an approximation valid for low non-conversion rate, ! ! !! ! !!!!! ! !.

Supplementary Text Figure-specific data analysis details: Fig. 1A, 2A, 2E and Fig. 3A (browser representations) The AnnoJ browser (24) shows tracks of hg18 and mm9 UCSC gene annotations, mRNA-Seq read density, and DNA methylation on the Watson and Crick strands (W, C). DNA methylation at each CG or CH site is represented as a vertical tick (green or blue, respectively) whose height corresponds to the fraction of all reads mapped to that site

Page 21: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

8

which were protected from bisulfite conversion (mC/C, range from 0 to 1). Regions which are homologous in human and mouse are indicated by light-blue shading.

Fig. 1B CH methylation level (mCH/CH) for each gene as a function of the rank of mRNA transcript abundance (mRNA-Seq fragments per kilobase of mapped read (FPKM)). We analyzed all genes which are expressed (FPKM > 0.01), whose length > 1 kb, mCH/CH < 0.03, and we omitted genes with no MethylC-Seq coverage for CG or CH reference positions. Note: mCG/CG and mCH/CH are defined as the total number of mC basecalls in each context, normalized by the total number of basecalls in the same context (both methylated and unmethylated), corrected by subtraction of context specific non-conversion rates (see above). Fig. 1E We calculated the fraction of all cytosine basecalls in each context, considering only sites covered by at least 10 reads in the mouse MethylC-Seq and TAB-Seq samples (fetal and 6wk). The proportion of basecalls was corrected for each sample’s non-conversion rate as described above. Fig. 1F For all 100 kb contiguous windows of mouse and human chromosome 12, mCG/CG and mCH/CH were calculated, with subtraction of mCG or mCH non-conversion values. Fig. 1G For all 5 kb contiguous windows of the mouse genome in chr12:112,250,000 - 119,500,000, the following mouse data was summarised:

- mRNA-Seq (10 wk frontal cortex): fraction of all reads located in window. - ChIP-input reads (8 wk cortex): fraction of all reads located in window. - DnaseI HS (8 wk cortex): fraction of all reads located in window. - mCG/CG and mCH/CH: fraction of basecalls in window that are methylated in

each context. - hmC CMS-IP (10 wk frontal cortex): [fraction of all CMS-IP reads located in

window] / [fraction of all CMS-input reads located in window]. - Note: mRNA, ChIP input and DNaseI HS tracks use arbitrary units proportional to

read density. Fig. 1H Chromatin accessibility vs CH methylation level (mCH/CH). For all 10 kb contiguous windows of the mouse genome, we calculated mCH/CH level from 10 wk mouse frontal cortex MethylC-Seq and estimated chromatin accessibility as ChIP-Seq input read density (reads per 10 kb per million total reads) for 8 wk mouse cortex (37). We omitted any 10 kb regions where ChIP-seq input read density = 0 or mCH/CH = 0. Density of bins is shown for mCH/CH " 0.02 and ChIP-seq input read density " 0.5. Fig. 2B: as per Fig. 1E. Only sites that were covered by at least one read in all samples were included, and correction for non-conversion was applied.

Page 22: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

9

Fig. 2C For each mouse NeuN+, NeuN- and glia sample, mCG/CG and mCH/CH were calculated for all Dnmt3a binding regions (46) (n = 67,050) and a set of 100,000 regions randomly distributed throughout the genome, each the size of the mean length of Dnmt3a binding regions (555 bp). Data is presented in a box and whisker plot with outliers not shown. Fig. 2D Heatmaps show Pearson correlation of mCH/CH within bins of size 10kb throughout the human and mouse genome. The median total coverage in each bin (i.e., the number of CH sites times the number of reads overlapping that site) was >24,000 for all samples. This provides confidence that the estimated mCH/CH was not severely affected by stochastic sampling of reads. Finally, we applied hierarchical clustering to all the data sets with complete linkage and distance given by one minus the sample correlation. Fig. 2F and S4 We developed statistical procedures to analyze the base-resolution, or per-site, correlation of methylation level between samples within the same species (intra-species) and across species (inter-species comparison of mouse and human). For this analysis, we focused on samples with substantial mCH, including NeuN+ in mouse (R1, R2, R3) and human (R1, R2), as well as human stem cell lines (H1, HUES6). Intra-species correlation: To reduce the effect of stochastic sampling, we set a minimum coverage threshold of 10 reads per site. Any site which had fewer than 10 reads in any of the samples was excluded. To account for slight differences in the bisulfite non-conversion between samples, we corrected the methylation level (mC/C) using the procedure described above. We then calculated the Pearson correlation of the corrected mC/C across all single sites between each pair of samples. These values defined the correlation values for “individual replicates,” shown as open circles in Fig. S4. In some cases more than two samples were available for comparison, so we computed a group average correlation coefficient by summing each of the pairwise covariances of each independent pair of samples, then normalizing by the appropriate sum of variances. For the three mouse NeuN+ replicates, let the three replicates be m1, m2 and m3. The average correlation, which determines the bar height in Fig. 2F and S4, is then:

!!"# ! !!!"!!" ! !!"!!" ! !!"!!"!!!!!"!!" ! !!"!!" ! !!"!!"!

Inter-species correlation: To compare methylation patterns between mouse and human, we first identified homologous CH sites. We restricted analysis to CH sites of high coverage in all three mouse NeuN+ samples (!20 reads). For each of these sites, we used the UCSC liftOver utility to determine whether a unique homologue site exists in the human (73). We then filtered these homologous sites to retain only those which were CH sites in both species, and which had a minimum coverage in all human NeuN+ or ESC samples (!10 reads).

We identified 66,489 autosomal and 3,323 ChrX sites for comparison of human and mouse NeuN+; for comparison of mouse NeuN+ with human H1 ESC we used slightly

Page 23: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

10

more (88,771 autosomal, 5,120 ChrX). The correlation coefficients for individual pairwise comparisons of replicates were then calculated as for the intra-species correlations (above).

For the human-mouse NeuN+ comparison, we combined 6 total comparisons (human replicates r1={h1,h2} vs. mouse replicates r2={m1,m2,m3}) to form an average correlation coefficient as follows:

!!"# !!!!!!!!!!!

!!!!!!!!!! !!"!!"!"!"

This procedure was also used to compute an average correlation for comparison of

the two embryonic stem cell lines with the human and mouse NeuN+ replicates. To assess the magnitude and significance of the correlation values we observed, we

carried out two control analyses for each of the above correlation measurements. Simulation of full correlation to assess correlation magnitude: Because our estimate of methylation level at each site is based on the observed frequency of methylated reads (mC/C), this measurement suffers from stochastic noise arising from the random sampling of reads from the population of DNA fragments. Such noise lowers the correlation coefficient we observe between each pair of samples. To correct for this, we performed a simulation to determine how much the correlation coefficient is reduced by sampling noise. We did this by simulating pairs of samples with identical methylation patterns, but independent sampling noise. At each site, we combined the number of observed methylated and unmethylated reads from two experimental samples (!!! !!! !!! !!, respectively) and determined the total empirical frequency of methylated reads: ! ! !!! ! !!!!!!! ! !! ! !! ! !!!! We then simulated random samples from a binomial distribution with probability of success f and total trials !!! ! !!! or !!! ! !!!, respectively. Finally, we computed the correlation coefficient using these simulated samples. Fig. S4 shows that these simulated correlation coefficients were ~0.8 for most comparisons. Since the true correlation of the simulated samples, which would be observed in the limit of very high coverage, is 1, in Fig. 2F we normalized the observed correlation of the original data by the simulated correlation. Fig. S4 shows the correlation of the data, without normalization Shuffle control to assess correlation significance: A second control analysis was conducted to compare the observed correlation with what would be expected in case the true correlation was 0. The finite size of the data set leads to a non-zero observed correlation, particularly for inter-species comparisons that cannot take advantage of large populations of homologous sites. In addition, long length-scale correlations related to genomic regions such as CG-islands may induce correlation that is not specific to a particular site (Fig. S5A). For each comparison of a pair of samples, we performed a set of 200 shuffle controls in which the methylation level at a given site in sample 1 was compared with the methylation level at a randomly chosen site in sample 2, constrained to be <1 Mb distant. After shuffling sites, we computed the correlation coefficient exactly as for the unshuffled data.

Page 24: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

11

Fig. 3: Analysis of gender-specific methylation patterns on X-chromosome We compared the brain DNA methylation patterns in our human female and male NeuN+ samples with two published surveys of human X-linked genes. The first of these studies measured expression from the active and inactive ChrX alleles using a human/rat somatic cell hybrid assay (49). The authors of that study assigned each surveyed gene a score in the range 0-9 corresponding to the number of individual hybrid lines in which expression from the inactivated ChrX was detected. A score of 0 thus corresponds to complete inactivation, whereas 9 corresponds to complete escape from X-inactivation. We remapped these genes to the reference used in this study (hg18) using the UCSC liftOver utility (73). We also compared our data with a second survey that used an immunoprecipitation based method (MeDIP) to profile DNA methylation from peripheral blood in normal human females (karyotype 46, XX) as well as Turner syndrome patients (45, X) (50). That study reported that inactivated genes feature hypermethylation of promoter region CG islands, whereas escapee genes show low levels of promoter methylation. These two studies using complementary techniques each provide a list of putative X-inactivated and X-escapee genes, which we used to compare with the mCG and mCH patterns in our neuronal samples. To avoid noise and bias, we excluded short genes (<2 kb in length) as well as any gene with low coverage (<50% of the mean read density for the entire ChrX). Fig. 3B and S4D For each of the genes assayed in (49), we examined the total mCG/CG within the promoter region, defined to be a 2 kb region centered on the TSS, and the total mCG/CG or mCH/CH within the gene body. The box plot shows the difference between female and male methylation level for genes ranked according to the X-inactivation status index. For each box, the central line is the group median, the box edges are the 25th and 75th percentiles, and the whiskers extend to the maximum and minimum values (including all “outliers”). Fig. 3C and S5E Scatter plots of human female vs. male NeuN+ mCG/CG within the promoter region (TSS±1kb) of each gene, or intragenic mCG/CG or mCH/CH. Fig. 3D Receiver operating characteristic (ROC) analysis was used to assess how well DNA methylation patterns allow discrimination of X-escapee genes. For this analysis, we used only those genes whose expression was measured by (49). We defined escapee genes as those which were expressed from both the active and inactivated ChrX in all of the cell lines tested (score = 9). Genes with score 0-8 were defined as non-escapee. We tested discriminability using the female-male difference in promoter region mCG/CG, which was previously shown to correlate with X-inactivation status (50), as well as the difference in neuronal intragenic mCH/CH. In addition, we tested a linear combination of four different features (female-male promoter region mCG/CG and mCH/CH, female-male intragenic mCG/CH and mCH/CH) with coefficients determined by linear regression against the X-inactivation score (49). For each of these three measures of

Page 25: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

12

DNA methylation, we created an ROC curve by plotting the fraction of escapee genes that are correctly discriminated, as well as the fraction of false detections, at each value of a discrimination threshold. The area under the ROC curve (AUC) is a statistical measure of discriminability, which ranges from 0.5 when little or no discrimination information is present to 1 for perfect discriminability. Fig. 4A, S7: Unbiased clustering of gene sets based on DNA methylation profiles To identify sets of genes that share similar DNA methylation patterns in an unbiased fashion, we developed a two-stage procedure that first represents all annotated genes within a low dimensional feature space, and then groups nearby genes in this space to create gene clusters.

The starting point is a large data array with one entry for each gene in each of 7 samples (6 wk, 7 wk and 12 mo NeuN+ and NeuN-, and glia (S100b+)) and in each sequence context (CG and CH). For each entry, we profiled the methylation level (mCG/CG and mCH/CH) in bins of size 1 kb starting 100 kb upstream of the TSS and ending 100 kb downstream of the transcription end site (TES). To compare genes with different lengths, we divided each gene body into 10 non-overlapping bins of equal size extending from the TSS to the TES. We then linearly interpolated the gene-body mC/C data at 100 evenly spaced bins within the gene body in order to give roughly equal weight to the gene-body and flanking methylation data. The absolute mCG/CG and mCH/CH level in each sample for each gene was normalized by the median value in the distal flanking region (50-100 kb upstream of TSS or downstream of TES). Normalized mC/C values were then log-transformed. Combining all of the normalized methylation data points, we obtained a matrix of 4,200 features for each of 25,260 genes. Any bins with missing data due to insufficient coverage in one of the samples (5.9% of the total) were replaced with the median value of the entire data set. We performed singular value decomposition on this data matrix to identify the linear combinations of methylation features that account for the largest fraction of the total data variance [principal components (PCs), Fig. S7]. We retained the top 5 PCs as a low-dimensional representation of robust genomic methylation features, accounting for 46% of the total data variance. Although the remaining PCs contain a great deal of meaningful data which may be useful for more refined analysis of DNA methylation features, our analysis focused on this reduced space.

Next, we used k-means clustering to estimate gene sets with highly similar within-set methylation patterns. We chose to extract k=15 clusters to capture a diverse range of methylation features, while still allowing visualization and statistical enrichment analysis of functional association for each gene set. We repeated the clustering procedure 10 times using random initialization of the cluster centers, choosing as the final estimate the run with the smallest within-cluster sum of distances from each point to the cluster centroid. To display the heatmaps of mRNA-Seq and mC/C patterns for each of 25,260 genes, we smoothed and downsampled the genes 40-fold to allow representation of genome-scale features. Fig. 4B: Enrichment analysis of functional gene categories To relate DNA methylation profiles with cell-type specific gene functions and with specific patterns of developmental regulation, we defined six mouse gene sets, as follows:

Page 26: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

13

1. Neuronal genes (461) which were reported to have >3-fold enrichment in mouse neurons (45). To increase the specificity of this list, we also required that genes were homologous to rat genes which are highly expressed in neuronal somata (74). 2. Astrocytic genes (2,618), reported to have >20-fold enriched expression in astrocytes (45). 3. Constitutively high genes (60) which did not appear in the neuronal or astrocytic gene set; whose transcripts were highly abundant (mRNA-Seq FPKM ! 10) in brain tissue samples from seven developmental stages (fetal, 1, 2, 4, 6 and 10 wk, and 22 mo); and which had "20% variation between the maximum and minimum abundance in these samples. 4. Constitutively low genes (5,278) which did not appear in the neuronal or astrocytic gene set; whose transcript abundance remained below a threshold (FPKM " 0.5); and which had "20% variation between the maximum and minimum abundance in these samples. 5. Upregulated genes (233) which are either neuronal or astrocytic genes and which have >10-fold increased expression at 2 wk compared with fetal. 6. Down-regulated genes (207) which are either neuronal or astrocytic genes and which have >5-fold lower expression at 2 wk compared with fetal.

For each functional gene set, we computed the overlap with each of the 15 gene clusters identified by k-means (Fig. 4A). We compared this number with the expectation for random gene sets with the same size to compute a fold-enrichment value, and we assessed statistical significance using Fisher’s exact test. We applied a Benjamini-Hochberg procedure (75) to control the false discovery rate < 0.05. Fig. 4C For each gene set, we computed the median of the flank-normalized mC/C profiles in each developmental and cell-type specific sample. The mRNA-Seq expression data (right-hand column) show the median expression level over all genes in each set. Fig. 5A Total hmCG/CG and hmCH/CH are shown for autosomes and ChrX in 6 wk cortex. Methylation values have been corrected for the biases introduced by non-conversion and non-protection (see above). Fig. 5B We profiled the total hmCG/CG within the following regions: CG islands (UCSC), promoter regions (TSS±2kb), gene ends (TES±2kb), gene bodies (TSS-TES), DHS (35) identified in adult or fetal brain, DHS identified exclusively in adult or fetal, and enhancers identified in adult or fetal brain (37). The plot shows the median over all the individual regions of each type, and the error bars indicate the 32-68th percentile range corresponding to ~1 standard deviation. Fig. 5C Heatmaps show the median profile of flank-normalized hmCG/CG around six categories of genes. Methods are the same as Fig. 4C.

Page 27: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

14

Fig. 5D Heatmaps showing the profile of mC/C and hmC/C for each of 16,077 genes expressed in adult mouse cortex (mRNA-Seq FPKM > 0.1 in 6 wk tissue). Genes are ranked by their expression in adult cortex and the flank-normalized mC/C or hmC/C pattern around each gene is displayed, as described in the methods for Fig. 4A (above). Fig. 6A Starting with six distinct categories of CG-DMRs for each species (described above), we excluded those on the X and Y chromosomes and organized the remaining DMRs into 4 largely non-redundant sets comprising most of the specific CG-DMR types. These were defined as follows: NeuN+ hyper-mCG: This set includes CG-DMRs where mCG/CG is larger in NeuN+ compared with NeuN– and Fetal; larger than Fetal only; or larger than NeuN– only NeuN– hyper-mCG: This set includes CG-DMRs where mCG/CG is larger in NeuN– compared with NeuN+, and those in which mCG/CG is larger in NeuN– than in fetal. NeuN+ hypo-mCG: This set includes CG-DMRs where mCG/CG is smaller in NeuN+ compared with NeuN– and Fetal; or smaller compared with Fetal only. NeuN– hypo-mCG: This set includes CG-DMRs where mCG/CG is smaller in NeuN– compared with NeuN+ and Fetal, or Fetal only.

We also created two additional DMR sets (heatmaps not shown): Fetal hyper-mCG: This set includes CG-DMRs where mCG/CG is larger in fetal compared with adult NeuN+ and/or NeuN–. Adult hyper-mCG: This set includes CG-DMRs where mCG/CG is larger in adult NeuN+ and/or NeuN– compared with fetal.

To display heatmaps of the methylation level in each CG-DMR, we first sorted the DMRs within each set by the difference between NeuN+ and NeuN– mCG/CG, averaged over replicates. We then smoothed and down-sampled the ordered DMR methylation patterns by 10 DMRs. The line plots below each heatmap show the median methylation level for each CG-DMR category, while the shaded regions show the 32nd-68th percentiles. Fig. 6B To assess the genomic distribution of CG-DMRs, we computed the fraction of all CG-DMRs (the union of the six sets listed above) that overlap the genomic regions defined above (see Fig. 5B and associated methods). Fig. 6C We compared the number of CG-DMRs of each type overlapping each region, with the expected number that would be observed if both sets were randomly distributed. Suppose

Page 28: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

15

there are N CG-DMRs and R regions, and let the total genomic length of the CG-DMRs be !! and the total genomic length of the regions be !!. Then the expectation value of the number of overlapping regions is !! ! !!!! ! !!!!!!, where G is the total length of the genome. In the limit !!! !! ! !, the number of observed overlapping regions will be Poisson distributed with mean !!. To assess the significance of enrichment or depletion, we therefore compared the actual number of observed overlaps with the Poisson cumulative distribution function. No multiple-comparison correction was applied to this analysis. Fig. 6D,E For each category of developmental CG-DMRs (Fetal > Adult or Adult > Fetal), we calculated the fraction that overlap a fetal or adult DHS site or a fetal- or adult-specific enhancer. The Venn diagrams represent these overlaps, and the relative sizes and positions of the circles are chosen to approximate the true proportions as closely as possible. Exact proportionality of the Venn diagrams with the true overlaps is sometimes not geometrically feasible. Fig. 6F For each CG-DMR, data was analyzed in contiguous 100 bp bins from 2.5 kb upstream to 2.5 kb downstream of the central position of the CG-DMR. H3K4me1, H3K27ac and CMS-IP read density profiles were normalized to their respective input control samples. The DNaseI HS read density profile uses arbitrary units proportional to read density. Fig. 6H To test the hypothesis that Tet2 is involved in active demethylation during development of brain cells, we analyzed differences between mCG/CG in adult wild-type mice (6 wk) and Tet2-/- knockout mice (63). At each developmental CG-DMR, we counted the total methylated and unmethylated reads within a window of size 1 kb centered on the midpoint of the DMR. We then used Fisher’s exact test to assess the significance of differences between the 6 wk wild-type sample and each of the other four samples. We applied a Benjamini-Krieger-Yekutieli procedure to control FDR<0.05 (MATLAB procedure fdr_bky by David M. Groppe) (76). The bars in Fig. 6H show the proportion of developmental CG-DMRs that had higher or lower methylation in the Tet2-/- mice compared to the 6 wk wild-type. Fig. 6I Histograms show the distribution of the difference in mCG/CG within 1 kb windows centered on each developmental CG-DMR between adult WT (6 wk) and two groups: control (4 wk, 10 wk and 22 mo WT), as well as the Tet2-/- knockout. Filled bars show the magnitude of differences at CG-DMRs that were significantly altered (see statistical procedure above, Fig. 6H). Fig. S1A-D The fraction of all basecalls methylated in each context was calculated as follows (where CN=CC, CA, CT or CG):

Page 29: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

16

Total mCG basecalls (corrected for CG BS non-conv.): mCGbasecalls-corr = #mCGbasecalls - #CGbasecalls* mCGBS-non-conv Total mCA basecalls (corrected for CA BS non-conv.): mCAbasecalls-corr = #mCAbasecalls - #CAbasecalls* mCABS-non-conv Total mCC basecalls (corrected for CC BS non-conv.): mCCbasecalls-corr = #mCCbasecalls - #CCbasecalls* mCCBS-non-conv Total mCT basecalls (corrected for CT BS non-conv.): mCTbasecalls-corr = #mCTbasecalls - #CTbasecalls* mCTBS-non-conv Fraction of all C basecalls that are mCG: mCGfraction = mCGbasecalls-corr / #CNbasecalls Fraction of all C basecalls that are mCA: mCAfraction = mCAbasecalls-corr / #CNbasecalls Fraction of all C basecalls that are mCC: mCCfraction = mCCbasecalls-corr / #CNbasecalls Fraction of all C basecalls that are mCT: mCTfraction = mCTbasecalls-corr / #CNbasecalls Fraction of all C basecalls that are mCH: mCHfraction = mCAfraction + mCCfraction + mCTfraction Fig. S1G: Partially methylated domains were identified as described previously (22). Fig. S2A,B Overall, before correcting for the rate of false positives introduced by bisulfite non-conversion and failure of Tet1 oxidation, we measured 0.5262% non-converted cytosines in TAB-Seq reads in the CH context in adult (6wk) mouse cortex. Our empirical control measurement of non-conversion in unmethylated pUC19 DNA in the CH context yielded a rate of 0.5245%, with a 99% confidence interval spanning 0.467 - 0.587% (based on fit of a binomial distribution using the Clopper-Pearson method; MATLAB function “binofit”). We thus estimate the global hmCH/CH to be 0.017%, with a 99% confidence interval 0 - 0.059%.

To test whether individual CH sites contain statistically significant hmCH, we developed a statistical model based on the null hypothesis that CH sites are composed of 1.4% mCH, 98.6% CH, and 0% hmCH. The non-conversion rate for unmethylated CH sites was assumed to be 0.52%, whereas the sum of bisulfite non-conversion and Tet1 non-oxidation for mC sites was set at 2.08% based on measurement of pUC19 DNA (Table S1). Our model assumes that the number of non-converted reads will follow a mixture of two binomial distributions with these two rates. Based on this model we calculated the p-value associated with the reads observed at each CH site across the genome. We detect only 50 significant sites [FDR=0.01, Benjamini-Hochberg-Yekutieli procedure (76)]. These sites were marginally significant, with hmCH/CH ~ 0.5. Fig. S2C Our base-resolution measurement of hmC (TAB-Seq) allowed us to examine the relationship between the hydroxymethylation level (hmCG/CG) and the total methylation level (mC/C as measured in bisulfite sequencing, which is the sum of the frequency of hydroxymethylated and methylated sites). We restricted analysis to sites covered by at least 10 reads in the TAB-Seq data set. We sorted each of these sites into ten equally spaced bins according to the total level of methylation observed in bisulfite sequencing (mC/C) of the same sample. Within each bin, we plot the proportion of sites that have statistically significant 5hmC/C (uncorrected p<0.01, Fisher’s exact test). Significance

Page 30: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

17

was assessed by comparing the number of hmC reads with the cumulative binomial distribution with mean parameter given by the bisulfite non-conversion rate. Fig. S3A and Table S2 For each of the three adult mouse brain cell-type specific methylomes (R1, R2, R3) we used Fisher’s exact test to assess whether intragenic mCH/CH is larger in glia (NeuN–) vs. neurons (NeuN+) for each gene. A final list of 174 glial hyper-mCH genes was obtained by requiring a significant effect (FDR=0.05, Benjamini-Hochberg) in all three comparisons. Fig. S3B We profiled mC/C around genes reported to be enriched in oligodendrocytes (45) and in brain epithelial cells (78). The top oligodendrocyte genes were those with at least 20-fold enrichment (45). Fig. S3C We repeated our analysis of methylation patterns around functional gene categories in human brain samples by mapping each mouse gene to a human ortholog using the gene-oriented ortholog database (GOOD) (77). Fig. S8: Methods as in Fig. 4.

Page 31: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

!"#$%&'$%!(")#$%&'()&$*+,-.,/01$'*$23$45675/$)&&$016'85/,$()8,0)&&8"$0)6,9'+5:,;$(1$8,-.,/0,$0'/6,<6"$*'+$,)07$7.2)/$)/;$2'.8,$8)2=&,>$!*"+#$?+,-.,/01$'*$23$5/$/,.+'/)&$8)2=&,8$45675/$,)07$3@A$0'/6,<6>$!,#[email protected])/$/,.+'/$'+$!!#$'+$9&5)&$0,&&$8,-.,/0,$0'/8,+B)65'/$)6$7597&1C2,671&)6,;$)/;$./2,671&)6,;$3@$='8565'/8>$D;,'9+)28$87'4$8,-.,/0,$0'/8,+B)65'/$)/;$0'2='8565'/$)6$='8565'/8$ *&)/E5/9$ 67,$2'86$7597&1$2,671&)6,;$3@$856,8$ !6'=$ +'4#"$./2,671&)6,;$856,8$!25;;&,#"$'+$)&&$3@$='8565'/8$!('66'2#>$!-#$F'6)&$&,/967$'*$=)+65)&&1$2,671&)6,;$;'2)5/8$5/$7.2)/$9,/'2,8>$!.#$G+'='+65'/$'*$2,671&)6,;$()8,8$5/$3%$'+$3@$0'/6,<6$67)6$)+,$2,671&)6,;$)6$)$95B,/$&,B,&$5/$7.2)/$0'+6,<$)/;$/,.+'/$8)2=&,8>$$

% m

CG

/CG

0

30

60

90

% m

CH

/CH

0

1

2

3

4

5

6

% mCC/CC 02

% m

CA/

CA

0

2

4

6

8

10

12

A

C D

Hs 25yr frontal cortexHs 53yr NeuN+Hs 55yr NeuN+

Frac

tion

of m

CG

site

s

0

0.1

0.2

0.3

0.4

0.5

0.6

20 40 60 80 100

Frac

tion

of m

CH

site

s

0

0.1

0.2

0.3

0.4

Methylation level20 40 60 80 100

mCG

mCH

FE HG

Neu

N+

53 y

r N

euN

– 53

yr

Neu

N+

55 y

r N

euN

– 55

yr

2 yr

5 yr

12 y

r25

yr

55 y

r

64 y

r

16 y

r

35 d

ofe

tal

H1

ESiP

CS-

19.1

1IM

R90

fibr

obla

stfo

resk

in fi

brob

last

Tota

l PM

D le

ngth

(bp)

0

2

4

6

8

1012x108

IMR9

0

fore

skin

fibro

blast

H1 E

SCiP

SC-1

9.11

feta

l bra

in25

yr b

rain

Neu

N+

53 y

r N

euN

– 53

yr

Neu

N+

55 y

r N

euN

– 55

yr

2 yr

5 yr

12 y

r25

yr

55 y

r

64 y

r

16 y

r

35 d

ofe

tal

H1

ESiP

CS-

19.1

1IM

R90

fibr

obla

stfo

resk

in fi

brob

last

% m

CG

/CG

0

20

40

60

80

% m

CH

/CH

0

0.5

1.0

1.5

2.0

2.5

2 w

k4

wk

6 w

k22

mo

10 w

k

1 w

kfe

tal

Neu

N+

7wk

Neu

N–

7wk

Neu

N+

6wk

Neu

N–

6wk

Neu

N+

12m

o N

euN

– 12

mo

glia

S10

0b+

Tet2

-/-

% mCT/CT0

2

4

% m

CA/

CA

0

1

2

3

4

5

% mCT/CT0

1

% mCC/CC01

% m

C/C

0

2

4

6

8

10

B

% m

C/C

01

2

3

4

56

2 w

k4

wk

6 w

k22

mo

10 w

k

1 w

kfe

tal

Neu

N+

7wk

Neu

N–

7wk

Neu

N+

6wk

Neu

N–

6wk

Neu

N+

12m

o N

euN

– 12

mo

glia

S10

0b+

Tet2

-/-

0

0.1

0.2

0.3

0

0.1

0.2

0.3

CAC CAT

CAA CTA

CC

C

CC

A

CTC

CAG CTT

CTG CC

T

CC

G

Hum

an N

euN

+ 53

yr

Mou

se N

euN

+ 7w

k%

mC

HN

/CH

N

% m

CH

N/C

HN

CAC

CAG CAA

CTG

CC

C

CC

A

CTC CAT

CTT CTA

CC

G

CC

T

Con

serv

atio

n (b

its)

Unm

ethy

late

dm

C/C

< 0

.01

All

cyto

sine

s

Hs NeuN+

0

1

2

Position Position0 2

0

1

2

Hyp

er-

met

hyla

ted

0

1

2

0

1

2

0

1

2

0 20

1

2

HI

Page 32: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

100 101 102 10310

10

10

10

A

B

0

1 50 10 108

10

100

n)

C

0 0.5 10

0.5

1

(p<0

.01,

unc

orre

cted

) Fetal6 wk

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

EF

Page 33: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

NeuN

+tis

sue

tissu

eNe

uN–

NeuN

+Ne

uN–

R1R2R3

2 wk4 wk6 wk

22 mo10 wk

1 wkfetal

R1glia

R2R3G

lial h

yper

-mC

H (1

74)

R1R2

2 yr5 yr

12 yr

25 yr64 yr

16 yr

35 dofetal

H1 ESC

R1R2

Neu

rona

lU

p-re

gula

ted

Ast

rocy

teD

own-

regu

late

dC

onsi

tutiv

ely

low

Con

stitu

tivel

y hi

gh

median

median

Birth 10 100

1

10

0.1

1

10

0.1

1

10

0.1

1

10

0.1

Birth 10 100

gene5! 3! gene5! 3!

gene5! 3! gene5! 3!

gene5! 3! gene5! 3!

-100 kb +100 kb -100 kb +100 kbA

C

Normalized mCG/CG0.8 1 1.2

0Normalized mCH/CH

1 2

Birth 10 100 <0.1

1

10

<0.1

1

10

<0.1

1

10

<0.1

1

10

<0.1

1

10

<0.1

1

10

0 1mCG/CG

0 0.06mCH/CH

mCG/CG mCH/CH

-100 kb +100 kb -100 kb +100 kb

mCG/CG mCH/CH

-100 kb +100 kb -100 kb +100 kb

mCG/CG mCH/CH median

BNe

uN+

tissu

eNe

uN–

R1R2R3

2 wk4 wk6 wk

22 mo10 wk

1 wkfetal

R1glia

R2R3O

ligod

endr

ocyt

es (5

40)

Top

olig

o-de

ndro

cyte

s(4

2)

Epith

elia

l(1

31)

!"#$%&'%!"#$% !"&% '()*+,(*+'% ,-.-% /*01% 2.0% 342.56.,% 788% 5/$% 3*)% !*+9-% ,-.-9% '(2'% 9(*:% ;(<% ,4624% !"&%(1=-)!-'(142'6*.%2.0%;)<%*46,*0-.0)*>1'-%2.0%-=6'(-4624%,-.-9?%@)2.9>)6='%2/+.02.>-%;ABCDE-F%GHIJ<%*K-)%0-K-4*=!-.'%69%9(*:.%3*)%'(-%92!-%,-.-%9-'9?%;*<%!"#$%!"&%'()*+,(*+'%,-.-%/*01%2.0%342.56.,%788%5/$%3*)%6.06>2'-0%,-.-%9-'9%6.%(+!2.?%@)2.9>)6='%2/+.02.>-%;ABCDE-F%GHIJ<%*K-)%0-K-4*=!-.'%69%9(*:.%3*)%'(-%92!-%,-.-%9-'9?% L8

Page 34: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

CGCH

CGCH

0

0.2

0.4

0.6

0.8

1

vs

!NeuN+

!NeuN+

vs NeuN+

NeuN+

vs NeuN+

NeuN+

vs NeuN+

ES

vs

H1 ES

HuES6

vs NeuN+

H1 ES

!"#$%&'%"#$%&'(!)'%%#*$+,'(!)'#--,),#(+!'-!.#+/0*$+,'(!&+$+#!$+!&,(1*#!&,+#&!2#+3##(!(#4%'(&!$(5!67!)#**&!,(!/4.$(!$(5!.'4&#8!9$+$!$%#!+/#!&$.#!$&!,(!:,18!;:<!24+!&/'3(!3,+/!('!('%.$*,=$+,'(!20!+/#!&,.4*$+#5!-4**!)'%%#*$+,'(!>$*4#8!7/4--*#5! 5$+$! $%#! %#?%#&#(+#5! 20! +/#! .#5,$(! '>#%! ;@@! %$(5'.! &/4--*#&! $(5! #%%'%! 2$%&! &/'3! +/#! ABCA+/!?#%)#(+,*#&8!

;D

Page 35: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

A

B C

D

E

F

H

G

0.20

-0.2

Intra

geni

c m

CG

/CG

0 1-2 3-4 5-6 7-8 9

Intragenic mCG/CG

Fem

ale N

euN+

0 50 100

0.08

0.09

0.1

0.11

mC

H/C

H

Lag (bp)

0 200 400 600 800

0.08

0.09

0.1

0.11m

CH

/CH

Lag (bp)

Same strandOpposite strand

0 25 50 750

0.1

0.2

Lag (bp)

mC

H/C

H

ShuffledAcross readsWithin-read

Autosomes

Fem

ale

mC

G/C

G

0 0.5 10

0.5

1

Fem

ale

mC

H/C

H

0 0.1 0.20

0.1

0.2

Chromosome XFe

mal

e m

CG

/CG

Fem

ale

mC

H/C

H

0 0.5 10

0.5

1

0 0.1 0.20

0.1

0.2

0.3

1 100 10,000Number of 1kb bins Number of 1kb bins

1 10 100 1,000

0 0.5 10

0.2

0.4

0.6

0.8

1

Inactivated Escaped

X-inact.* (253)X-escapee* (37)X-escapee§ (33)X-escapee† (7)

Chr2 (1472)

0

0.5

1

Whole

geno

me

Human female53 yr NeuN+

Human male55 yr NeuN+

Chr2 ge

nes

Whole

geno

me

Chr2 ge

nes

Frac

tion

of m

CH

mCAmCTmCC

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

Page 36: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

Prom

oter

mC

G/C

G

A B C

Intr

agen

ic m

CH

/CH

0 0.05 0.10

0.05

0.1

0 0.5 10

0.5

1

X-inact.* (253)X-escapee* (37)X-escapee§ (33)X-escapee† (7)

Chr2 (1472)

Intr

agen

ic m

CG

/CG

0 0.5 10

0.2

0.4

0.6

0.8

1

0 0.5 10

0.2

0.4

0.6

0.8

1

Male MaleMale0 0.5 1

0

0.2

0.4

0.6

0.8

1

0 0.02 0.040

0.01

0.02

0.03

0.04

0.05

Human ESC(H1 male; HUES6 female)

Human adult NeuN–(55 yr male; 53 yr female)

Human adult NeuN+(55 yr male; 53 yr female)

0 0.5 10

0.2

0.4

0.6

0.8

1

0 0.5 10

0.2

0.4

0.6

0.8

1

0 0.02 0.040

0.01

0.02

0.03

0.04

0.05

Fem

ale

Fem

ale

Fem

ale

!"#$%&'%!"#$"%&'(")*+*),-"./012.*3#,(2.."%#',2)%3'',)"11,.0("'4,521",6'4,+"-21",(%3-3."%,-7!87!,9.3(,%3:;<,="#*),-7>87>,9-*$$1",%3:;,2#$,="#*),-7!87!,9?3..3-,%3:;,*#,/@-2#,2$@1.,+%3#.21,)3%."A,#"@%3#',9(;<,=1*2,9);<,2#$,"-?%03#*),'."-,)"11,1*#"',9*;4,B*++"%"#.,'0-?31',)3%%"'(3#$,.3,7/%C,="#"',%"(3%."$,.3,?",*#2).*62."$,3%,.3,"')2(",*#2).*62.*3#,9D,72%%"1,!"#$%&E,F,G/2%(,!"#$%4;<,(%"$*)."$,"')2("",="#"',9H<,./*','.@$0;<,2#$,[email protected]'3-21,97/%I;,="#"'4,J"@%3#',2#$,=1*2<,?@.,#3.,"-?%03#*),'."-,)"11'<,2%",/0("%-"./012."$,'(")*+*)2110,*#,+"-21",)/%C&"')2("",="#"'4,

IK

Page 37: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

100 101 102 1030

0.2

0.4

0.6

0.8

1

Principal component (PC) rank

Frac

tion

of to

tal d

ata

varia

nceA

B

PC2

PC1

PC3

PC2

PC4

PC3

ChrX genes (1175)Neuronal genes (61)Astrocytic genes (50)

Density of genes (a.u.)0 Max

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

4F

Page 38: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

A B

fetal-100 kb +100 kb

2 wk 10 wk NeuN+ NeuN–

mCG/CG hmCG/CG mCH/CH

6 wk fetal-100 kb +100 kb

2 wk 10 wk NeuN+ NeuN–

Con

stitu

tivel

y hi

ghN

euro

nal

Upr

egul

ated

Astro

cytic

Dow

nreg

ulat

edC

onst

itutiv

ely

low

Enrichment (q<0.05)

n.s.100.1

fetal10 w

k

Gen

es

1

25,260

10.1 10

mCG/CG0 1

hmCG/CG0 0.25

mCH/CH0 0.04

!

"

#

$%&

!'

(

)

*

!"#$%&'%!(")#$%&'$()*)$+,-.$/012$3$),'$4&-56$&','$50*&$)6$)74-89*'$!966-,.)80:'(#$4;)8'$+-,$.<=<2$

>?

Page 39: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

mCG/CG

Fe

tal7 w

k Neu

N+6 wk

Fetal

7 wk N

euN+6 w

k

0 1mCH/CH0 1

hmCH/CH0 0.2

hmCG/CG0 0.2

CGIs

Intragenic

Adult enhFetal enh

CGIs

Intragenic

Adult enhFetal enh

0

0.5

1

CG

Is

Adul

t enh

Intra

geni

c

Feta

l enh

CG

Is

Adul

t enh

Intra

geni

c

Feta

l enh

(h)m

CG

/CG

0

0.005

0.01

0.015

(h)m

CH

/CH

6 wk mC

Fetal hmC6 wk hmC

A

B

C

10

100

101

feta

l

6 w

k10

wk

5 yr

16 y

r

53 y

r

123

!"#$%&'%!("#$%&'(#)*+,*+-#)*.,*.-#/)*+,*+#'01#/)*.,*.#20#3'4/#530%)24#6352%07#8'(93:#/';3#<330#4%6634&31#

=%6# &/3# :')>(3?:>342=24# 6'&3# %=# <2:9(=2&3# 0%0?4%0;36:2%07# !)"#@312'0# /)*#'01#)*# (3;3(# 20# 12==3630&# 530%)3#

=3'&963:#!366%6#<'6:#AB?CD&/#>36430&2(3"7#/)*#1'&'#'63#19>(24'&31#=6%)#E257#FG7#!*"#$6'0:462>&#'<901'043#196205#

13;3(%>)30&#=%6#&/3#&/633#!"##=')2(H#5303:#)3':9631#<H#)IJK?L3M7#

BC

Page 40: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

Genic mCH/CH

Gen

ic h

mC

G/C

G

0 0.02 0.040

0.1

0.2

0.3

Genic mCG/CG(hmCG/CG subtracted)

Gen

ic h

mC

G/C

G

0 0.5 10

0.1

0.2

0.3

Num

ber o

f gen

es1

10

100

NormalizedmCG/CG

0.8 1.2

% mCG/CG0 1

% hmCG/CG0 25

% mCH/CH0 3

NormalizedhmCG/CG

0.8 1.2

NormalizedmCH/CH

0.8 1.2

0.1 1 10 100 10001

20,184

FPKM

Gen

e

-100 kb +100 kb -100 kb +100 kbA B C

D E

!"#$%&'(%!"#$%&'()*&+,-"%.""(,/0,$(1,*/0,$(1,2"(","3+4"))&'(5,6)7,84$()94&+%,$-:(1$(9",;'4,"$9*,';,<=>?@A,2"("),&(,B, .C,/':)", ;4'(%$#, 9'4%"3, .&%*, "3+4"))&'(, #"D"#, EFGHI=5?5, 6*7, J-)'#:%", #"D"#, ';, /"%*K#$%&'(, 6/0LM0L>,/0NM0N7,$(1,*K14'3K/"%*K#$%&'(,6*/0LM0L7,.&%*&(,"$9*,2"(",$(1,&(,;#$(C&(2,?==,C-,4"2&'()5,6+7,/0,$(1,*/0,#"D"#,('4/$#&O"1,-K,%*",;#$(C&(2,4"2&'(5,6,>-7,8'%$#,&(%4$2"(&9,/0M0,&),9'/+$4"1,.&%*,*/0LM0L5,

<P

Page 41: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

R1

R2

R3

2 w

k4

wk

6 w

k

22 m

o10

wk

1 w

kfe

tal

6 w

kfe

tal

R1

glia R2

R3

R1

R2

2 yr

5 yr

12 y

r

25 y

r64

yr

16 y

r

35 d

ofe

tal

H1

ESR1

R2

tissue NeuN+ NeuN- tissue NeuN+ NeuN-

R1

R2

2 yr

5 yr

12 y

r

25 y

r64

yr

16 y

r

35 d

ofe

tal

H1

ESR1

R2

tissue NeuN+ NeuN-hmC

R1

R2

R3

2 w

k4

wk

6 w

k

22 m

o10

wk

1 w

kfe

tal

6 w

kfe

tal

R1

glia R2

R3

tissue NeuN+ NeuN- hmC

CG

-DM

Rs

Neu

N+

5746

410

787

Neu

N+

5844

516

139

hype

r-m

CG

hypo

-mC

G

CG

-DM

Rs

CG

-DM

Rs

CG

-DM

Rs

Neu

N+

5746

410

787

Neu

N+

5844

516

139

hype

r-m

CG

hypo

-mC

G

Neu

N+

1383

4638

649

Neu

N+

4874

542

059

hype

r-m

CG

hypo

-mC

G

Neu

N+

1383

4638

649

Neu

N+

4874

542

059

hype

r-m

CG

hypo

-mC

G

mCG/CG10

hmCG/CG0.250

mCH/CH, hmCH/CH0.070

mCH/CH0.090

mCG/CG10

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

Med

ian

mC

G/C

G

Med

ian

mC

G/C

G

0

0.02

0.04

0.06

0.08

0

0.02

0.04

0.06

0.08

Med

ian

mC

H/C

H

Med

ian

mC

H/C

H

NeuN+ hyper mCG

NeuN+ hypo mCG

NeuN+ hyper mCG

NeuN+ hypo mCG

!"#$%&''!"#$%#&'( )*( %+,( #-.( %+!( #$( +,/012'( 3."-$3*3".( 4"$5""-( -"67)-'8( 9:3#8( #-.( *7)-$#:( ;)7$"<( $=7)69=(.">":)&%"-$(3-(%)6'"(#-.(=6%#-?(

@A

Page 42: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

C

A B

Neu

N+

Neu

N+

hyper-mCG

hypo-mCG

NeuronalUpregulated

AstrocyteDownregulated

Consitutively low

Constitutively highGene set

Promoter (TSS±5kb)CG-DMRs Adult > Fetal

10x10-8

6x10-8

8

2

3

Fetal > Adultfetal adult fetal adult

hmC(biotin-gluc)10

14x10-8

6x10-8

Distance (kb)-2 -1 0 +1 +2

Distance (kb)-2 -1 0 +1 +2

hmC(CMS-IP)

2

3

4

CG-DMRs

WT (6 wk)

Feta

l

0 1

1

0 1

1

0 1

1

WT (6 wk)

WT (6 wk)

WT (6 wk)

WT (6 wk)

WT (6 wk)

Feta

l

0 1

1

WT

(4 w

k,10

wk,

22 m

o)W

T (4

wk,

10 w

k,22

mo)

0 1

1

0 1

1

Frac

tion

of D

MR

s(p

<0.0

1)0

2

4 x 10-3Adu

lt >

Feta

lFe

tal >

Adu

lt

Tet2

-/-Te

t2-/-

Fold enrichment0.25

n.s.1 4

!"#$%&'(%!"#$%&'(")*+(,-()$."/.0/1234567/$"/8)/)&(/)#8"7%#$,)$."8-/7)8#)/7$)(/.0/%(--3)9,(/7,(%$0$%/8"+/+(:(-.,'(")8--9/+9"8'$%/ ;("(7/ $"/ '.<7(=/ >)?/ @<8")$0$%8)$."/ .0/ -.%8-/ ("#$%&'(")/ .0/ &'1/ 8)/ +(:(-.,'(")8--9/ +9"8'$%/ 1234567/A9/&'1/("#$%&'(")/'()&.+.-.;$(7=/>*?/4("7$)9/'8,7/7&.B$";/8--/+(:(-.,'(")8-/1234567/>-(0)3&8"+/,-.)7?C/.#/."-9/)&.7(/B&$%&/&8:(/7$;"$0$%8")-9/'.#(/.#/-(77/'12*12/%.',8#(+/B$)&/8/B$-+3)9,(/8+<-)/A#8$"/>D/BE?=/F&(/!"#$%&%/'<)8")/'$%(/&8:(/ $"%#(87(+/'12*12/%.',8#(+/B$)&/GFC/,8#)$%<-8#-9/ 0.#/H()8-IJ+<-)/1234567C/$"+$%8)$";/F()3+(,("+(")/+('()&9-8)$."/8)/)&(7(/7$)(7/+<#$";/+(:(-.,'(")=/

KL

Page 43: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

30

Table S1 (separate file) Sample information and sequencing metrics.

Table S2 (separate file) mCH deserts in human and mouse and the genes located within these domains.

Table S3 (separate file) Glial mCH-hypermethylated genes.

Table S4 (separate file) Gender-specific methylation on human X chromosome.

Table S5 (separate file) Sample demographics.

Page 44: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

References and Notes

1. E. Borrelli, E. J. Nestler, C. D. Allis, P. Sassone-Corsi, Decoding the epigenetic language of neuronal plasticity. Neuron 60, 961–974 (2008). doi:10.1016/j.neuron.2008.10.012 Medline

2. J. Feng, Y. Zhou, S. L. Campbell, T. Le, E. Li, J. D. Sweatt, A. J. Silva, G. Fan, Dnmt1 and Dnmt3a maintain DNA methylation and regulate synaptic function in adult forebrain neurons. Nat. Neurosci. 13, 423–430 (2010). doi:10.1038/nn.2514 Medline

3. C. A. Miller, J. D. Sweatt, Covalent modification of DNA regulates memory formation. Neuron 53, 857–869 (2007). doi:10.1016/j.neuron.2007.02.022 Medline

4. S. Numata, T. Ye, T. M. Hyde, X. Guitart-Navarro, R. Tao, M. Wininger, C. Colantuoni, D. R. Weinberger, J. E. Kleinman, B. K. Lipska, DNA methylation signatures in development and aging of the human prefrontal cortex. Am. J. Hum. Genet. 90, 260–272 (2012). doi:10.1016/j.ajhg.2011.12.020 Medline

5. M. Suderman, P. O. McGowan, A. Sasaki, T. C. Huang, M. T. Hallett, M. J. Meaney, G. Turecki, M. Szyf, Conserved epigenetic sensitivity to early life experience in the rat and human hippocampus. Proc. Natl. Acad. Sci. U.S.A. 109, (Suppl 2), 17266–17272 (2012). doi:10.1073/pnas.1121260109 Medline

6. K. D. Siegmund, C. M. Connor, M. Campan, T. I. Long, D. J. Weisenberger, D. Biniszkiewicz, R. Jaenisch, P. W. Laird, S. Akbarian, DNA methylation in the human cerebral cortex is dynamically regulated throughout the life span and involves differentiated neurons. PLoS ONE 2, e895 (2007). doi:10.1371/journal.pone.0000895 Medline

7. K. Martinowich, D. Hattori, H. Wu, S. Fouse, F. He, Y. Hu, G. Fan, Y. E. Sun, DNA methylation-related chromatin remodeling in activity-dependent BDNF gene regulation. Science 302, 890–893 (2003). doi:10.1126/science.1090842 Medline

8. J. U. Guo, D. K. Ma, H. Mo, M. P. Ball, M. H. Jang, M. A. Bonaguidi, J. A. Balazer, H. L. Eaves, B. Xie, E. Ford, K. Zhang, G. L. Ming, Y. Gao, H. Song, Neuronal activity modifies the DNA methylation landscape in the adult brain. Nat. Neurosci. 14, 1345–1351 (2011). doi:10.1038/nn.2900 Medline

9. I. B. Zovkic, M. C. Guzman-Karlsson, J. D. Sweatt, Epigenetic regulation of memory formation and maintenance. Learn. Mem. 20, 61–74 (2013). doi:10.1101/lm.026575.112 Medline

10. J. M. Levenson, T. L. Roth, F. D. Lubin, C. A. Miller, I. C. Huang, P. Desai, L. M. Malone, J. D. Sweatt, Evidence that DNA (cytosine-5) methyltransferase regulates synaptic plasticity in the hippocampus. J. Biol. Chem. 281, 15763–15773 (2006). doi:10.1074/jbc.M511767200 Medline

11. A. M. M. Oliveira, T. J. Hemstedt, H. Bading, Rescue of aging-associated decline in Dnmt3a2 expression restores cognitive abilities. Nat. Neurosci. 15, 1111–1113 (2012). doi:10.1038/nn.3151 Medline

Page 45: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

12. P. Moretti, J. M. Levenson, F. Battaglia, R. Atkinson, R. Teague, B. Antalffy, D. Armstrong, O. Arancio, J. D. Sweatt, H. Y. Zoghbi, Learning and memory and synaptic plasticity are impaired in a mouse model of Rett syndrome. J. Neurosci. 26, 319–327 (2006). doi:10.1523/JNEUROSCI.2623-05.2006 Medline

13. N. W. Penn, R. Suwalski, C. O’Riley, K. Bojanowski, R. Yura, The presence of 5-hydroxymethylcytosine in animal deoxyribonucleic acid. Biochem. J. 126, 781–790 (1972). Medline

14. N. W. Penn, Modification of brain deoxyribonucleic acid base content with maturation in normal and malnourished rats. Biochem. J. 155, 709–712 (1976). Medline

15. D. Globisch, M. Münzel, M. Müller, S. Michalakis, M. Wagner, S. Koch, T. Brückl, M. Biel, T. Carell, Tissue distribution of 5-hydroxymethylcytosine and search for active demethylation intermediates. PLoS ONE 5, e15367 (2010). doi:10.1371/journal.pone.0015367 Medline

16. J. U. Guo, Y. Su, C. Zhong, G.-L. Ming, H. Song, Hydroxylation of 5-methylcytosine by TET1 promotes active DNA demethylation in the adult brain. Cell 145, 423–434 (2011). doi:10.1016/j.cell.2011.03.022 Medline

17. Y.-F. He, B. Z. Li, Z. Li, P. Liu, Y. Wang, Q. Tang, J. Ding, Y. Jia, Z. Chen, L. Li, Y. Sun, X. Li, Q. Dai, C. X. Song, K. Zhang, C. He, G. L. Xu, Tet-mediated formation of 5-carboxylcytosine and its excision by TDG in mammalian DNA. Science 333, 1303–1307 (2011). doi:10.1126/science.1210944 Medline

18. C.-X. Song, K. E. Szulwach, Q. Dai, Y. Fu, S. Q. Mao, L. Lin, C. Street, Y. Li, M. Poidevin, H. Wu, J. Gao, P. Liu, L. Li, G. L. Xu, P. Jin, C. He, Genome-wide profiling of 5-formylcytosine reveals its roles in epigenetic priming. Cell 153, 678–691 (2013). doi:10.1016/j.cell.2013.04.001 Medline

19. L. Shen, H. Wu, D. Diep, S. Yamaguchi, A. C. D’Alessio, H. L. Fung, K. Zhang, Y. Zhang, Genome-wide analysis reveals TET- and TDG-dependent 5-methylcytosine oxidation dynamics. Cell 153, 692–706 (2013). doi:10.1016/j.cell.2013.04.002 Medline

20. W. Xie, C. L. Barr, A. Kim, F. Yue, A. Y. Lee, J. Eubanks, E. L. Dempster, B. Ren, Base-resolution analyses of sequence and parent-of-origin dependent DNA methylation in the mouse genome. Cell 148, 816–831 (2012). doi:10.1016/j.cell.2011.12.035 Medline

21. K. E. Varley, J. Gertz, K. M. Bowling, S. L. Parker, T. E. Reddy, F. Pauli-Behn, M. K. Cross, B. A. Williams, J. A. Stamatoyannopoulos, G. E. Crawford, D. M. Absher, B. J. Wold, R. M. Myers, Dynamic DNA methylation across diverse human cell lines and tissues. Genome Res. 23, 555–567 (2013). doi:10.1101/gr.147942.112 Medline

22. R. Lister, M. Pelizzola, R. H. Dowen, R. D. Hawkins, G. Hon, J. Tonti-Filippini, J. R. Nery, L. Lee, Z. Ye, Q. M. Ngo, L. Edsall, J. Antosiewicz-Bourget, R. Stewart, V. Ruotti, A. H. Millar, J. A. Thomson, B. Ren, J. R. Ecker, Human DNA methylomes at base resolution show widespread epigenomic differences. Nature

Page 46: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

462, 315–322 (2009). doi:10.1038/nature08514 Medline

23. M. J. Ziller, F. Müller, J. Liao, Y. Zhang, H. Gu, C. Bock, P. Boyle, C. B. Epstein, B. E. Bernstein, T. Lengauer, A. Gnirke, A. Meissner, Genomic distribution and inter-sample variation of non-CpG methylation across human cell types. PLoS Genet. 7, e1002389 (2011). doi:10.1371/journal.pgen.1002389 Medline

24. R. Lister, R. C. O’Malley, J. Tonti-Filippini, B. D. Gregory, C. C. Berry, A. H. Millar, J. R. Ecker, Highly integrated single-base resolution maps of the epigenome in Arabidopsis. Cell 133, 523–536 (2008). doi:10.1016/j.cell.2008.03.029 Medline

25. R. Lister, M. Pelizzola, Y. S. Kida, R. D. Hawkins, J. R. Nery, G. Hon, J. Antosiewicz-Bourget, R. O’Malley, R. Castanon, S. Klugman, M. Downes, R. Yu, R. Stewart, B. Ren, J. A. Thomson, R. M. Evans, J. R. Ecker, Hotspots of aberrant epigenomic reprogramming in human induced pluripotent stem cells. Nature 471, 68–73 (2011). doi:10.1038/nature09798 Medline

26. B. H. Ramsahoye, D. Biniszkiewicz, F. Lyko, V. Clark, A. P. Bird, R. Jaenisch, Non-CpG methylation is prevalent in embryonic stem cells and may be mediated by DNA methyltransferase 3a. Proc. Natl. Acad. Sci. U.S.A. 97, 5237–5242 (2000). doi:10.1073/pnas.97.10.5237 Medline

27. J. De Felipe, P. Marco, A. Fairén, E. G. Jones, Inhibitory synaptogenesis in mouse somatosensory cortex. Cereb. Cortex 7, 619–634 (1997). doi:10.1093/cercor/7.7.619 Medline

28. P. R. Huttenlocher, A. S. Dabholkar, Regional differences in synaptogenesis in human cerebral cortex. J. Comp. Neurol. 387, 167–178 (1997). doi:10.1002/(SICI)1096-9861(19971020)387:2<167::AID-CNE1>3.0.CO;2-Z Medline

29. B. P. Berman, D. J. Weisenberger, J. F. Aman, T. Hinoue, Z. Ramjan, Y. Liu, H. Noushmehr, C. P. Lange, C. M. van Dijk, R. A. Tollenaar, D. Van Den Berg, P. W. Laird, Regions of focal DNA hypermethylation and long-range hypomethylation in colorectal cancer coincide with nuclear lamina-associated domains. Nat. Genet. 44, 40–46 (2012). doi:10.1038/ng.969 Medline

30. S. Kriaucionis, N. Heintz, The nuclear DNA base 5-hydroxymethylcytosine is present in Purkinje neurons and the brain. Science 324, 929–930 (2009). doi:10.1126/science.1169786 Medline

31. C.-X. Song, K. E. Szulwach, Y. Fu, Q. Dai, C. Yi, X. Li, Y. Li, C. H. Chen, W. Zhang, X. Jian, J. Wang, L. Zhang, T. J. Looney, B. Zhang, L. A. Godley, L. M. Hicks, B. T. Lahn, P. Jin, C. He, Selective chemical labeling reveals the genome-wide distribution of 5-hydroxymethylcytosine. Nat. Biotechnol. 29, 68–72 (2011). doi:10.1038/nbt.1732 Medline

32. K. E. Szulwach, X. Li, Y. Li, C. X. Song, H. Wu, Q. Dai, H. Irier, A. K. Upadhyay, M. Gearing, A. I. Levey, A. Vasanthakumar, L. A. Godley, Q. Chang, X. Cheng, C. He, P. Jin, 5-hmC-mediated epigenetic dynamics during postnatal neurodevelopment and aging. Nat. Neurosci. 14, 1607–1616 (2011).

Page 47: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

doi:10.1038/nn.2959 Medline

33. M. Mellén, P. Ayata, S. Dewell, S. Kriaucionis, N. Heintz, MeCP2 binds to 5hmC enriched within active genes and accessible chromatin in the nervous system. Cell 151, 1417–1430 (2012). doi:10.1016/j.cell.2012.11.022 Medline

34. M. Yu, G. C. Hon, K. E. Szulwach, C. X. Song, L. Zhang, A. Kim, X. Li, Q. Dai, Y. Shen, B. Park, J. H. Min, P. Jin, B. Ren, C. He, Base-resolution analysis of 5-hydroxymethylcytosine in the mammalian genome. Cell 149, 1368–1380 (2012). doi:10.1016/j.cell.2012.04.027 Medline

35. J. A. Stamatoyannopoulos, M. Snyder, R. Hardison, B. Ren, T. Gingeras, D. M. Gilbert, M. Groudine, M. Bender, R. Kaul, T. Canfield, E. Giste, A. Johnson, M. Zhang, G. Balasundaram, R. Byron, V. Roach, P. J. Sabo, R. Sandstrom, A. S. Stehling, R. E. Thurman, S. M. Weissman, P. Cayting, M. Hariharan, J. Lian, Y. Cheng, S. G. Landt, Z. Ma, B. J. Wold, J. Dekker, G. E. Crawford, C. A. Keller, W. Wu, C. Morrissey, S. A. Kumar, T. Mishra, D. Jain, M. Byrska-Bishop, D. Blankenberg, B. R. Lajoie, G. Jain, A. Sanyal, K.-B. Chen, O. Denas, J. Taylor, G. A. Blobel, M. J. Weiss, M. Pimkin, W. Deng, G. K. Marinov, B. A. Williams, K. I. Fisher-Aylor, G. Desalvo, A. Kiralusha, D. Trout, H. Amrhein, A. Mortazavi, L. Edsall, D. McCleary, S. Kuan, Y. Shen, F. Yue, Z. Ye, C. A. Davis, C. Zaleski, S. Jha, C. Xue, A. Dobin, W. Lin, M. Fastuca, H. Wang, R. Guigo, S. Djebali, J. Lagarde, T. Ryba, T. Sasaki, V. S. Malladi, M. S. Cline, V. M. Kirkup, K. Learned, K. R. Rosenbloom, W. J. Kent, E. A. Feingold, P. J. Good, M. Pazin, R. F. Lowdon, L. B. Adams, An encyclopedia of mouse DNA elements (Mouse ENCODE). Genome Biol. 13, 418 (2012). doi:10.1186/gb-2012-13-8-418 Medline

36. R. K. Auerbach, G. Euskirchen, J. Rozowsky, N. Lamarre-Vincent, Z. Moqtaderi, P. Lefrançois, K. Struhl, M. Gerstein, M. Snyder, Mapping accessible chromatin regions using Sono-Seq. Proc. Natl. Acad. Sci. U.S.A. 106, 14926–14931 (2009). doi:10.1073/pnas.0905443106 Medline

37. Y. Shen, F. Yue, D. F. McCleary, Z. Ye, L. Edsall, S. Kuan, U. Wagner, J. Dixon, L. Lee, V. V. Lobanenkov, B. Ren, A map of the cis-regulatory sequences in the mouse genome. Nature 488, 116–120 (2012). doi:10.1038/nature11243 Medline

38. W. A. Pastor, U. J. Pape, Y. Huang, H. R. Henderson, R. Lister, M. Ko, E. M. McLoughlin, Y. Brudno, S. Mahapatra, P. Kapranov, M. Tahiliani, G. Q. Daley, X. S. Liu, J. R. Ecker, P. M. Milos, S. Agarwal, A. Rao, Genome-wide mapping of 5-hydroxymethylcytosine in embryonic stem cells. Nature 473, 394–397 (2011). doi:10.1038/nature10102 Medline

39. Y. Huang, W. A. Pastor, J. A. Zepeda-Martínez, A. Rao, The anti-CMS technique for genome-wide mapping of 5-hydroxymethylcytosine. Nat. Protoc. 7, 1897–1908 (2012). doi:10.1038/nprot.2012.103 Medline

40. A. Magklara, A. Yen, B. M. Colquitt, E. J. Clowney, W. Allen, E. Markenscoff-Papadimitriou, Z. A. Evans, P. Kheradpour, G. Mountoufaris, C. Carey, G. Barnea, M. Kellis, S. Lomvardas, An epigenetic signature for monoallelic olfactory receptor expression. Cell 145, 555–570 (2011).

Page 48: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

doi:10.1016/j.cell.2011.03.040 Medline

41. E. J. Clowney, M. A. LeGros, C. P. Mosley, F. G. Clowney, E. C. Markenskoff-Papadimitriou, M. Myllys, G. Barnea, C. A. Larabell, S. Lomvardas, Nuclear aggregation of olfactory receptor genes governs their monogenic expression. Cell 151, 724–737 (2012). doi:10.1016/j.cell.2012.09.043 Medline

42. H. Li, J. C. Radford, M. J. Ragusa, K. L. Shea, S. R. McKercher, J. D. Zaremba, W. Soussou, Z. Nie, Y. J. Kang, N. Nakanishi, S. Okamoto, A. J. Roberts, J. J. Schwarz, S. A. Lipton, Transcription factor MEF2C influences neural stem/progenitor cell differentiation and maturation in vivo. Proc. Natl. Acad. Sci. U.S.A. 105, 9397–9402 (2008). doi:10.1073/pnas.0802876105 Medline

43. M. W. Akhtar, M. S. Kim, M. Adachi, M. J. Morris, X. Qi, J. A. Richardson, R. Bassel-Duby, E. N. Olson, E. T. Kavalali, L. M. Monteggia, In vivo analysis of MEF2 transcription factors in synapse regulation and neuronal survival. PLoS ONE 7, e34863 (2012). doi:10.1371/journal.pone.0034863 Medline

44. A. C. Barbosa, M. S. Kim, M. Ertunc, M. Adachi, E. D. Nelson, J. McAnally, J. A. Richardson, E. T. Kavalali, L. M. Monteggia, R. Bassel-Duby, E. N. Olson, MEF2C, a transcription factor that facilitates learning and memory by negative regulation of synapse numbers and function. Proc. Natl. Acad. Sci. U.S.A. 105, 9391–9396 (2008). doi:10.1073/pnas.0802679105 Medline

45. J. D. Cahoy, B. Emery, A. Kaushal, L. C. Foo, J. L. Zamanian, K. S. Christopherson, Y. Xing, J. L. Lubischer, P. A. Krieg, S. A. Krupenko, W. J. Thompson, B. A. Barres, A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function. J. Neurosci. 28, 264–278 (2008). doi:10.1523/JNEUROSCI.4178-07.2008 Medline

46. H. Wu, V. Coskun, J. Tao, W. Xie, W. Ge, K. Yoshikawa, E. Li, Y. Zhang, Y. E. Sun, Dnmt3a-dependent nonpromoter DNA methylation facilitates transcription of neurogenic genes. Science 329, 444–448 (2010). doi:10.1126/science.1190485 Medline

47. T. Mohandas, R. S. Sparkes, L. J. Shapiro, Reactivation of an inactive human X chromosome: evidence for X inactivation by DNA methylation. Science 211, 393–396 (1981). doi:10.1126/science.6164095 Medline

48. K. Plath, S. Mlynarczyk-Evans, D. A. Nusinow, B. Panning, Xist RNA and the mechanism of X chromosome inactivation. Annu. Rev. Genet. 36, 233–278 (2002). doi:10.1146/annurev.genet.36.042902.092433 Medline

49. L. Carrel, H. F. Willard, X-inactivation profile reveals extensive variability in X-linked gene expression in females. Nature 434, 400–404 (2005). doi:10.1038/nature03479 Medline

50. A. J. Sharp, E. Stathaki, E. Migliavacca, M. Brahmachary, S. B. Montgomery, Y. Dupre, S. E. Antonarakis, DNA methylation profiles of human active and inactive X chromosomes. Genome Res. 21, 1592–1600 (2011). doi:10.1101/gr.112680.110 Medline

51. C. M. Johnston, F. L. Lovell, D. A. Leongamornlert, B. E. Stranger, E. T.

Page 49: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

Dermitzakis, M. T. Ross, Large-scale population study of human cell lines indicates that dosage compensation is virtually complete. PLoS Genet. 4, e9 (2008). doi:10.1371/journal.pgen.0040009 Medline

52. A. K. Maunakea, R. P. Nagarajan, M. Bilenky, T. J. Ballinger, C. D’Souza, S. D. Fouse, B. E. Johnson, C. Hong, C. Nielsen, Y. Zhao, G. Turecki, A. Delaney, R. Varhol, N. Thiessen, K. Shchors, V. M. Heine, D. H. Rowitch, X. Xing, C. Fiore, M. Schillebeeckx, S. J. Jones, D. Haussler, M. A. Marra, M. Hirst, T. Wang, J. F. Costello, Conserved role of intragenic DNA methylation in regulating alternative promoters. Nature 466, 253–257 (2010). doi:10.1038/nature09165 Medline

53. R. E. Thurman, E. Rynes, R. Humbert, J. Vierstra, M. T. Maurano, E. Haugen, N. C. Sheffield, A. B. Stergachis, H. Wang, B. Vernot, K. Garg, S. John, R. Sandstrom, D. Bates, L. Boatman, T. K. Canfield, M. Diegel, D. Dunn, A. K. Ebersol, T. Frum, E. Giste, A. K. Johnson, E. M. Johnson, T. Kutyavin, B. Lajoie, B. K. Lee, K. Lee, D. London, D. Lotakis, S. Neph, F. Neri, E. D. Nguyen, H. Qu, A. P. Reynolds, V. Roach, A. Safi, M. E. Sanchez, A. Sanyal, A. Shafer, J. M. Simon, L. Song, S. Vong, M. Weaver, Y. Yan, Z. Zhang, Z. Zhang, B. Lenhard, M. Tewari, M. O. Dorschner, R. S. Hansen, P. A. Navas, G. Stamatoyannopoulos, V. R. Iyer, J. D. Lieb, S. R. Sunyaev, J. M. Akey, P. J. Sabo, R. Kaul, T. S. Furey, J. Dekker, G. E. Crawford, J. A. Stamatoyannopoulos, The accessible chromatin landscape of the human genome. Nature 489, 75–82 (2012). doi:10.1038/nature11232 Medline

54. M. B. Stadler, R. Murr, L. Burger, R. Ivanek, F. Lienert, A. Schöler, E. van Nimwegen, C. Wirbelauer, E. J. Oakeley, D. Gaidatzis, V. K. Tiwari, D. Schübeler, DNA-binding factors shape the mouse methylome at distal regulatory regions. Nature 480, 490–495 (2011). Medline

55. Z. Petanjek, M. Judaš, G. Šimic, M. R. Rasin, H. B. Uylings, P. Rakic, I. Kostovic, Extraordinary neoteny of synaptic spines in the human prefrontal cortex. Proc. Natl. Acad. Sci. U.S.A. 108, 13281–13286 (2011). doi:10.1073/pnas.1105108108 Medline

56. B. Kolb, R. Mychasiuk, A. Muhammad, Y. Li, D. O. Frost, R. Gibb, Experience and the developing prefrontal cortex. Proc. Natl. Acad. Sci. U.S.A. 109, (Suppl 2), 17186–17193 (2012). doi:10.1073/pnas.1121251109 Medline

57. H. T. Chugani, A critical period of brain development: studies of cerebral glucose utilization with PET. Prev. Med. 27, 184–188 (1998). doi:10.1006/pmed.1998.0274 Medline

58. P. Levitt, Structural and functional maturation of the developing primate brain. J. Pediatr. 143, (Suppl), S35–S45 (2003). doi:10.1067/S0022-3476(03)00400-1 Medline

59. M. V. Johnston, Plasticity in the developing brain: implications for rehabilitation. Dev. Disabil. Res. Rev. 15, 94–101 (2009). doi:10.1002/ddrr.64 Medline

60. P. J. Uhlhaas, W. Singer, The development of neural synchrony and large-scale cortical networks during adolescence: relevance for the pathophysiology of schizophrenia and neurodevelopmental hypothesis. Schizophr. Bull. 37, 514–523

Page 50: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

(2011). doi:10.1093/schbul/sbr034 Medline

61. S. Nguyen, K. Meletis, D. Fu, S. Jhaveri, R. Jaenisch, Ablation of de novo DNA methyltransferase Dnmt3a in the nervous system leads to neuromuscular defects and shortened lifespan. Dev. Dyn. 236, 1663–1676 (2007). doi:10.1002/dvdy.21176 Medline

62. Materials and methods are available as supplementary material on Science Online.

63. M. Ko, H. S. Bandukwala, J. An, E. D. Lamperti, E. C. Thompson, R. Hastie, A. Tsangaratou, K. Rajewsky, S. B. Koralov, A. Rao, Ten-Eleven-Translocation 2 (TET2) negatively regulates homeostasis and differentiation of hematopoietic stem cells in mice. Proc. Natl. Acad. Sci. U.S.A. 108, 14566–14571 (2011). doi:10.1073/pnas.1112317108 Medline

64. J. W. Kinney, C. N. Davis, I. Tabarean, B. Conti, T. Bartfai, M. M. Behrens, A specific role for NR2A-containing NMDA receptors in the maintenance of parvalbumin and GAD67 immunoreactivity in cultured interneurons. J. Neurosci. 26, 1604–1615 (2006). doi:10.1523/JNEUROSCI.4722-05.2006 Medline

65. Y. Jiang, A. Matevossian, H.-S. Huang, J. Straubhaar, S. Akbarian, Isolation of neuronal chromatin from brain tissue. BMC Neurosci. 9, 42 (2008). doi:10.1186/1471-2202-9-42 Medline

66. S. Zhong, J. G. Joung, Y. Zheng, Y. R. Chen, B. Liu, Y. Shao, J. Z. Xiang, Z. Fei, J. J. Giovannoni, High-throughput illumina strand-specific RNA sequencing library preparation. Cold Spring Harb. Protoc. 2011, 940–949 (2011). doi:10.1101/pdb.prot5652 Medline

67. C. Trapnell, D. G. Hendrickson, M. Sauvageau, L. Goff, J. L. Rinn, L. Pachter, Differential analysis of gene regulation at transcript resolution with RNA-seq. Nat. Biotechnol. 31, 46–53 (2013). doi:10.1038/nbt.2450 Medline

68. C. Trapnell, A. Roberts, L. Goff, G. Pertea, D. Kim, D. R. Kelley, H. Pimentel, S. L. Salzberg, J. L. Rinn, L. Pachter, Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and Cufflinks. Nat. Protoc. 7, 562–578 (2012). doi:10.1038/nprot.2012.016 Medline

69. W. Perkins, M. Tygert, R. Ward, Computing the confidence levels for a root-mean-square test of goodness-of-fit. Appl. Math. Comput. 217, 9072–9084 (2011). doi:10.1016/j.amc.2011.03.124

70. T. Bancroft, C. Du, D. Nettleton, Estimation of false discovery rate using sequential permutation p-values. Biometrics 69, 1–7 (2013). doi:10.1111/j.1541-0420.2012.01825.x Medline

71. B. Langmead, C. Trapnell, M. Pop, S. L. Salzberg, Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 10, R25 (2009). doi:10.1186/gb-2009-10-3-r25 Medline

72. Y. Zhang, T. Liu, C. A. Meyer, J. Eeckhoute, D. S. Johnson, B. E. Bernstein, C. Nusbaum, R. M. Myers, M. Brown, W. Li, X. S. Liu, Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, R137 (2008). doi:10.1186/gb-2008-9-9-r137

Page 51: Global epigenomic reconfiguration during mammalian …papers.cnl.salk.edu/PDFs/Global epigenomic reconfiguration during...Global Epigenomic Reconfi guration During Mammalian Brain

Medline

73. A. S. Hinrichs, D. Karolchik, R. Baertsch, G. P. Barber, G. Bejerano, H. Clawson, M. Diekhans, T. S. Furey, R. A. Harte, F. Hsu, J. Hillman-Jackson, R. M. Kuhn, J. S. Pedersen, A. Pohl, B. J. Raney, K. R. Rosenbloom, A. Siepel, K. E. Smith, C. W. Sugnet, A. Sultan-Qurraie, D. J. Thomas, H. Trumbower, R. J. Weber, M. Weirauch, A. S. Zweig, D. Haussler, W. J. Kent, The UCSC Genome Browser Database: update 2006. Nucleic Acids Res. 34, D590–D598 (2006). doi:10.1093/nar/gkj144 Medline

74. I. J. Cajigas, G. Tushev, T. J. Will, S. tom Dieck, N. Fuerst, E. M. Schuman, The local transcriptome in the synaptic neuropil revealed by deep sequencing and high-resolution imaging. Neuron 74, 453–466 (2012). doi:10.1016/j.neuron.2012.02.036 Medline

75. Y. Benjamini, Y. Hochberg, J. R. Stat. Soc., B 57, 289 (1995).

76. Y. Benjamini, D. Yekutieli, The control of the false discovery rate in multiple testing under dependency. Ann. Stat. 29, 1165–1188 (2001). doi:10.1214/aos/1013699998

77. M.-R. Ho, C.-H. Chen, W.-C. Lin, Gene-oriented ortholog database: a functional comparison platform for orthologous loci. Database (Oxford) 2010, baq002 (2010). doi:10.1093/database/baq002 Medline

78. R. Daneman, L. Zhou, D. Agalliu, J. D. Cahoy, A. Kaushal, B. A. Barres, The mouse blood-brain barrier transcriptome: a new resource for understanding the development and function of brain endothelial cells. PLoS ONE 5, e13741 (2010). doi:10.1371/journal.pone.0013741 Medline


Recommended