Molecular Cell Article - Harvard Universityds.dfci.harvard.edu/~gcyuan/mypaper/das; 2b2c; mol...

Molecular Cell

Article

Distinct and Combinatorial Functionsof Jmjd2b/Kdm4b and Jmjd2c/Kdm4cin Mouse Embryonic Stem Cell IdentityPartha Pratim Das,1,10 Zhen Shao,1,10 Semir Beyaz,1 Eftychia Apostolou,3,4 Luca Pinello,5 Alejandro De Los Angeles,1

Kassandra O’Brien,1 Jennifer Marino Atsma,1,9 Yuko Fujiwara,1 Minh Nguyen,1 Damir Ljuboja,1,8 Guoji Guo,1

Andrew Woo,7 Guo-Cheng Yuan,5 Tamer Onder,1,6 George Daley,1,2,3 Konrad Hochedlinger,2,3,4 Jonghwan Kim,1,8

and Stuart H. Orkin1,2,*1Department of Pediatric Oncology, Dana-Farber Cancer Institute and Division of Hematology/Oncology, Boston Children’s Hospital,

Harvard Stem Cell Institute, Harvard Medical School, Boston, MA 02115, USA2Howard Hughes Medical Institute, Boston, MA 02115, USA3Department of Stem Cell and Regenerative Biology, Harvard University and Harvard Medical School, 7 Divinity Avenue, Cambridge,MA 02138, USA4Center for Regenerative Medicine, Massachusetts General Hospital Cancer Center, Boston, MA 02114, USA5Department of Biostatistics and Computational Biology, Dana-Farber Cancer Institute, Harvard School of Public Health, Boston,

MA 02115, USA6School of Medicine, Koc University, Rumelifeneri Yolu, Sariyer 34450, Istanbul, Turkey7Western Australian Institute for Medical Research, Royal Perth Hospital and School of Medicine and Pharmacology, The University of

Western Australia, Nedlands, WA 6009, Australia8Present address: Section of Molecular Cell and Developmental Biology, Institute for Cellular and Molecular Biology, University of Texasat Austin, Austin, TX 78712, USA9Present address: Abbott Bioresearch Center, 100 Research Drive, Worcester, MA 01605, USA10These authors contributed equally to this work*Correspondence: [email protected]

http://dx.doi.org/10.1016/j.molcel.2013.11.011

SUMMARY

Self-renewal and pluripotency of embryonic stemcells (ESCs) are established by multiple regulatorypathways operating at several levels. The roles of his-tone demethylases (HDMs) in these programs areincompletely defined. We conducted a functionalRNAi screen for HDMs and identified five potentialHDMs essential for mouse ESC identity. In-depthanalyses demonstrate that the closely related HDMsJmjd2b and Jmjd2c are necessary for self-renewalof ESCsand inducedpluripotent stemcell generation.Genome-wide occupancy studies reveal that Jmjd2bunique, Jmjd2cunique, and Jmjd2b-Jmjd2ccommontarget sites belong to functionally separable Core,Polycomb repressive complex (PRC), and Myc regu-latory modules, respectively. Jmjd2b and Nanog actthrough an interconnected regulatory loop, whereasJmjd2c assists PRC2 in transcriptional repression.Thus, twoHDMsof the same subclass exhibit distinctand combinatorial functions in control of the ESCstate. Such complexity of HDM function reveals anaspect of multilayered transcriptional control.

INTRODUCTION

Embryonic stem cells (ESCs) are capable of indefinite self-

renewal and differentiation into all lineages. Somatic cell reprog-

32 Molecular Cell 53, 32–48, January 9, 2014 ª2014 Elsevier Inc.

ramming to induced pluripotent stem cells (iPSCs) by defined

factors has greatly improved prospects for cellular therapies

(Takahashi and Yamanaka, 2006; Cherry and Daley, 2012).

Althoughmuch has been learned, the components that establish

and maintain ESC identity are incompletely defined.

ESC identity is maintained by activation of ESC-specific genes

and repression of lineage-specific developmental genes. This

balance of gene expression is maintained through crosstalk

between essential transcription factors (TFs) and chromatin reg-

ulators (Orkin and Hochedlinger, 2011; Young, 2011). Extensive

studies of protein-protein and protein-DNA interactions have re-

vealed distinct ESC regulatory modules, termed Core, Myc, and

Polycomb, that are essential for the entire ESC regulatory

network (Chen et al., 2008; Kim et al., 2010). The Core module

is composed of canonical ESC factors (Oct4, Sox2, and Nanog)

and their associated partners, which positively regulate ESC-

specific genes and repress lineage-specific genes (Kim et al.,

2010; Young, 2011). The Myc module, consisting of cMyc and

associated factors, is also transcriptionally active. However,

the Myc module is functionally separable from the Core module

and is activated earlier than the Core module during iPSC gener-

ation at the partial iPSC stage (Sridharan et al., 2009; Soufi et al.,

2012). The Polycomb repressive complex (PRC) module,

composed of PRC1 and PRC2 components, functions in repres-

sion of lineage-specific genes (Boyer et al., 2006; Kim et al.,

2010; Margueron and Reinberg, 2011).

Various approaches have implicated the role of chromatin reg-

ulators in self-renewal of ESCs (Fazzio et al., 2008; Hu et al.,

2009; Kagey et al., 2010). Histone demethylases (HDMs) are his-

tone-modifying enzymes, which have an opposing biochemical

mailto:[email protected]


http://crossmark.crossref.org/dialog/?doi=10.1016/j.molcel.2013.11.011&domain=pdf

Molecular Cell

Functions of Jmjd2b and Jmjd2c in Mouse ESCs

function to histone methyltransferases (HMTs). HDMs are

required for normal development and are implicated in patho-

logic states including cancers (Pedersen and Helin, 2010).

HDMs are divided into two broad classes, FAD-dependent

amine oxidases (Lsd1/Kdm1) and Fe(II)- and a-ketoglutarate-

dependent JmjC domain-containing HDMs (Mosammaparast

and Shi, 2010). The JmjC domain-containing HDMs have a

conserved catalytic triad (H, D/E, and H), which catalyzes lysine

demethylation of histones through an oxidative reaction that

requires Fe(II) and a-ketoglutarate as cofactors. JmjC domain-

containing HDMs are further subclassified based on their

sequence homology, domains, and substrate specificity (Agger

et al., 2008; Pedersen and Helin, 2010). Studies have proposed

roles for Jmjd1a, Jmjd2c, and Jarid1b/Kdm5b in ESC self-

renewal (Loh et al., 2007; Xie et al., 2011). However, roles of other

HDMs in ESCs remain unknown.

Here, we conducted a functional RNAi screen against all

annotated HDMs and identified five candidate HDMs essential

for mouse ESC (mESC) identity. To gain mechanistic insight,

we chose two closely related HDMs, Jmjd2b/Kdm4b and

Jmjd2c/Kdm4c, belonging to the same subclass (HDMs for

H3K9me2/me3 and H3K36me2/me3) for in-depth analysis. In

addition to their requirement in ESCs, both HDMs are required

for efficient somatic cell reprogramming. Although depletion of

either HDM generates a similar differentiation phenotype, chro-

matin occupancy studies reveal both unique and common target

sites. Jmjd2b unique, Jmjd2b-Jmjd2c common, and Jmjd2c

unique targets partition to the Core, Myc, and PRC regulatory

modules of the overall ESC network, respectively. Specifically,

we show that Jmjd2b and Nanog act through an interconnected

regulatory loop, whereas Jmjd2c assists PRC2 in full repression

at poised and repressed target genes. The dedicated and

combinatorial relationships between these two related HDMs

reveal an unsuspected level of complexity in how HDMs partic-

ipate in transcriptional control.

RESULTS

Functional RNAi Screens Reveal Candidate HDMs formESC IdentityMost HDMs are expressed in mESCs (Figure S1A available

online). To identify HDMs required for maintenance of the ESC

state, we performed a functional RNAi screen. We used five

different shRNA lentiviral constructs to knock down each of 20

HDMs in mESCs. The screen was scored in terms of alterations

of ‘‘ESC growth phenotype’’ and ‘‘colony morphology’’ (Fig-

ure 1A; Figure S1B; Table S1). Normally, ESCs grow as spherical

three-dimensional colonies. Upon depletion of a number of the

candidate HDMs, cells exhibited flattened morphology and

grew as a monolayer with reduced cell-cell contacts (Fig-

ure S1B), which we term a ‘‘differentiation’’ phenotype. A spec-

trum of differentiation phenotypes from mild to severe was

observed (Figure S1B; Table S1). For each candidate HDM, we

scored the phenotype from at least the two best individual

shRNAs. Two secondary screens were performed for validation.

All three screens reproduced the original phenotypes (Table S1).

Our functional RNAi screen identified five candidate HDMs,

namely Jmjd1a (Kdm3a), Jmjd2b (Kdm4b), Jmjd2c (Kdm4c),

Utx (Kdm6a), and Jmjd6. Among these, Jmjd1a (Kdm3a) and

Jmjd2c (Kdm4c) were identified previously as having a role in

self-renewal of mouse ESCs (Loh et al., 2007). Jmjd1a, Jmjd2b,

Jmjd2c, and Jmjd6 knockdown showed moderate to severe dif-

ferentiation phenotypes, whereas knockdown of Utx showed a

more subtle phenotype (Figure S1B; Table S1). Further analyses

were performed for the candidate HDMs. Alkaline phosphatase

staining typical of pluripotent ESCs was reduced upon knock-

down of candidate HDMs compared to controls (Anti-GFP and

Empty) (Figure 1B). Pluripotency marker SSEA-1 expression

was also reduced significantly upon knockdown (Figure 1C).

We validated each by western blotting (Figure S1C), and noted

general correlation of knockdown efficiency with the extent of

differentiation. mRNA expression and protein levels of all five

candidate HDMs were reduced upon ESC differentiation (Fig-

ures 1D and 1E). Additionally, these candidate HDMswere highly

expressed in mESCs compared to mouse embryonic fibroblasts

(Figure S1D), consistent with crucial roles in mESC identity.

Both Jmjd2b/Kdm4b and Jmjd2c/Kdm4c Are Requiredfor ESC Identity and Efficient Somatic CellReprogramming/iPSC GenerationWe selected two closely related candidate HDMs of the same

subclass, Jmjd2b and Jmjd2c, for in-depth study (Agger et al.,

2008). As depletion of either leads to an apparently similar differ-

entiation phenotype, we suspected that their comparison might

uncover insights into their nonredundant roles and possibly their

relationship to each other.

We first determined that Jmjd2b and Jmjd2c restore the

normal ESC growth and colony morphology in Jmjd2b-deficient

and Jmjd2c-deficient ESCs (Figure S1E). Indeed, full-length

Jmjd2b and Jmjd2c restored the normal ESC growth phenotype,

ESC colony morphology, and also SSEA-1 expression (Figures

S1F and S1G). However, catalytic domain (JmjC)-containing

Jmjd2b and Jmjd2c (1–333 aa) and catalytic mutants (JmjC

domain-containing catalytic triad [H, D/E, and H]) of Jmjd2b

and Jmjd2c achieved only partial rescue (Figures S1F and

S1G), suggesting that the full length of Jmjd2b and Jmjd2c,

as well as their catalytic residues, is required for self-renewal

of mESCs.

Next, we tested the role of Jmjd2b and Jmjd2c in somatic re-

programming. Depletion of Jmjd2b and Jmjd2c showed reduced

numbers of iPSC colonies compared to control knockdown, as

scored by colony morphology and SSEA-1-positive colonies

(Figure 1F), which is reminiscent of the effects of knockdown

of all H3K9 demethylases (Chen et al., 2013). Taken as a whole,

the functional assay demonstrates that Jmjd2b and Jmjd2c are

required for efficient somatic reprogramming induced by Oct4-

Sox2-Kl4-Myc and establishment of the ESC-like state.

Jmjd2b/Kdm4b and Jmjd2c/Kdm4c Are Required forSelf-Renewal of mESCsWe next conducted gene expression profiling upon depletion of

Jmjd2b and Jmjd2c compared to control (Anti-GFP) knockdown.

Gene expression profiling revealed that 1,787 genes were differ-

entially expressed by >2-fold upon depletion of Jmjd2b and

Jmjd2c (Figure 2A, left; Table S3). The gene cluster down-

regulated upon knockdown displayed higher expression in

Molecular Cell 53, 32–48, January 9, 2014 ª2014 Elsevier Inc. 33

shRNA plasmids against HDMs

Lentivirus production in HEK 293T cells

Infection to J1 mES cells

J1 mES cells 24 well

Puromycin selection Split cells after 2 days

Evaluate differentiation phenotype of mES cells

Check the knock down level

A

B C

D

24hrs

Differentiation (-LIF, +RA)

Day 0 2 6 8 10

anti-Actin

anti-Jmjd1a

anti-Jmjd2b

anti-Jmjd2c

anti-Utx

anti-Jmjd6

anti-Oct4

anti-Nanog

OSKM+Empty shRNA

OSKM+Jmjd2b-1 shRNA

OSKM+Jmjd2c-1 shRNA

E

Jmjd2c-2Jmjd2c-1 Utx-2 Utx-4

Jmjd6-2 Jmjd6-4 Oct4-1 Nanog-4

Jmjd1a-2 Jmjd1a-3 Jmjd2b-1 Jmjd2b-4 Anti-GFP

Empty

0

100

200

300

400

500

Num

ber o

f SS

EA

-1+

colo

nies

OSKM+E

mpt

y shR

NA

OSKM+J

mjd2

b-1s

hRNA

OSKM+J

mjd2

c-1

shRNA

******

SSEA-1

Anti-G

FP

Jmjd1

a-2

Jmjd1

a-3

Jmjd2

b-1

Jmjd2

b-4

Jmjd2

c-1

Jmjd2

c-2

Utx-2

Utx-4

Jmjd6

-2

Jmjd6

-4

Oct4-1

Nanog

-40

50

100

150

Mea

n F

luor

esce

nt In

tens

ity (

%)

F

***

*** ***

* * * *

***

** ***

*** ***

Undiff

5d_d

iff

7d_d

iff

11d_

diff

Undiff

5d_d

iff

7d_d

iff

11d_

diff

Undiff

5d_d

iff

7d_d

iff

11d_

diff

Undiff

5d_d

iff

7d_d

iff

11d_

diff

Undiff

5d_d

iff

7d_d

iff

11d_

diff

Undiff

5d_d

iff

7d_d

iff

11d_

diff

Undiff

5d_d

iff

7d_d

iff

11d_

diff

Undiff

5d_d

iff

7d_d

iff

11d_

diff

0.00000.00020.00040.00060.00080.0010

0.51.01.5

50010001500

Jmjd

1a

Jmjd

2b

Jmjd

2c

Jmjd

6

Gat

a6

Utx

Oct

4

Gat

a4

Fol

d ch

ange

diff

/und

iff m

RN

A le

vel

** * *** *** *

**

***

5days_diff 7days_diff 11days_diff

Differentiation (-LIF, +RA) Undiff

***

*

(legend on next page)

Molecular Cell



Molecular Cell


undifferentiated ESCs compared to differentiated ESCs; con-

versely, the gene cluster upregulated upon knockdown showed

reduced expression in undifferentiated ESCs (Figure 2A, right;

Figure S2A).

Expression of individual ESC-specific and lineage-specific

genes from Jmjd2b and Jmjd2c depleted cells revealed

reduced ESC-specific gene expression and enhanced expres-

sion of differentiation genes (Figures 2C and 2D; Figures S2B

and S2C). We observed modest downregulation of ESC-specific

genes, including Nanog, Esrrb, Klf4, and Tbx3, in Jmjd2b- or

Jmjd2c-depleted cells. In addition, specific knockdown of

Jmjd2b and Jmjd2c did not affect expression of other HDMs

of the same subclass. We observed upregulation of several

lineage-specific genes upon depletion of Jmjd2b or Jmjd2c,

such as Brachyury (T), Pitx2, Fgf8, Wnt, and Fgf5 for mesoendo-

derm/mesoderm, Otx2, Nestin, Pax6, Fabp7, and Zic1 for ecto-

derm/neuroectoderm, andCdx2 for trophoectoderm (Figures 2C

and 2D; Figures S2B and S2C).

Gene ontology (GO) and Ingenuity Pathway Analysis (IPA) of

the upregulated genes revealed significant enrichment for

several developmental processes and related signaling path-

ways, whereas the downregulated genes correlated with

several metabolic pathways, including glycolysis and gluconeo-

genesis (Figures S2D and S2E; Table S5). Furthermore, gene

set enrichment analyses (GSEAs) demonstrated that Jmjd2b

and Jmjd2c depletion significantly repressed the ‘‘undifferenti-

ated’’ ESC state and enhanced the ‘‘differentiation’’ state (Fig-

ures 2E and 2F; Figure S2F; Table S4), including changes in

several developmental signaling pathway gene sets (Fig-

ure S2G). In these analyses, we observed a similar overall

pattern of global gene expression upon depletion of Jmjd2b

or Jmjd2c. However, the total number of differentially ex-

pressed genes differed between Jmjd2b and Jmjd2c knock-

down cells (Figure 2B), raising the possibility that Jmjd2b and

Jmjd2c function through common and distinct mechanisms in

self-renewal of ESCs.

Differential Distribution of Genome-wide Targets ofJmjd2b/Kdm4b and Jmjd2c/Kdm4c in mESCsTo dissect mechanisms by which Jmjd2b and Jmjd2c function,

we determined their genome-wide DNA chromatin occupancy

using chromatin immunoprecipitation sequencing (ChIP-seq).

Due to the lack of suitable antibodies, we generated biotinylated

versions of Jmjd2b and Jmjd2c in mESCs (Figures S3A and S3B)

and performed in vivo biotinylation-mediated ChIP-seq (Bio-

Figure 1. Functional RNAi Screens Reveal Candidate HDMs for mESC

(A) Schematic diagram representing the outline of the RNAi screen.

(B) Alkaline phosphatase staining of mESCs upon knockdown of candidate HDM

(C) SSEA-1 staining of mESCs upon knockdown of candidate HDMs. SSEA-1-po

percentage of mean fluorescent intensity of candidate HDMs is represented. Dat

***p < 0.0001, **p < 0.001, *p < 0.01.

(D) Real-time PCR analyses of candidate HDMs at different time points during ES

levels are shown in differentiated ESCs relative to undifferentiated ESCs for each

using t test; ***p < 0.0001, **p < 0.001, *p < 0.01.

(E) Western blot analyses of candidate HDMs at different time points during ESC

(F) SSEA-1 staining of iPSC colonies from Empty (control) and Jmjd2b and Jmjd2

calculated using t test; ***p < 0.0001.

See also Figure S1.

ChIP-seq) (Kim et al., 2008, 2009, 2010); 16,807 and 18,025

significantly enriched binding peaks were identified for Jmjd2b

and Jmjd2c, respectively (Table S6). Despite their close relation-

ship, we observed differential genome-wide distribution of

Jmjd2b and Jmjd2c peaks. Jmjd2b peaks are distributed to

the promoter (�40%), intergenic (�32%), and gene body regions

(�28%), whereas Jmjd2c mostly occupies promoter regions

(�70%) compared to �20% binding at intergenic and �11% at

gene body regions (Figure 3A). Jmjd2b and Jmjd2c share

�40% common peaks/targets (Figure 3B). We further classified

targets as ‘‘Jmjd2b-Jmjd2c common,’’ ‘‘Jmjd2b unique,’’ and

‘‘Jmjd2c unique’’ (Figure 3B). Interestingly, the majority of

Jmjd2b-Jmjd2c common and Jmjd2c unique targets are distrib-

uted at promoter regions (�77% and �65%, respectively),

whereas Jmjd2b unique targets predominantly localized to

gene body regions (�42%) and intergenic regions (�48%), and

only �10% to promoter regions (Figure 3B).

Next, we compared the occupancy of Jmjd2b and Jmjd2c

with four metagenes, groups of genes with correlated expres-

sion (high, moderate, low, and very low) in mESCs. Further ana-

lyses showed that the majority of Jmjd2b-Jmjd2c common

targets and Jmjd2b unique targets are occupied at promoters

and distal regions, respectively, and correspond to highly ex-

pressed genes, whereas the majority of Jmjd2c unique targets

are occupied at the promoters of moderately, as well as lowly,

expressed genes (Figure S3C). These initial observations reveal

distinct distributions and functions of Jmjd2b and Jmjd2c in

mESCs.

Jmjd2b/Kdm4b Unique, Jmjd2c/Kdm4c Unique, andJmjd2b/Kdm4b-Jmjd2c/Kdm4c Common TargetRegions Belong to Different Regulatory Modules ofmESCsTo explore the significance of Jmjd2b-Jmjd2c common, Jmjd2b

unique, and Jmjd2c unique targets, we correlated occu-

pancy maps with histone marks (H3K4me1/3, H3K9me2/me3,

H3K27me3, H3K36me2/me3, and H3K27ac), components of

the PRC2 complex (Ezh2 and Suz12), ESC-related TFs (Oct4,

Nanog, Sox2, and Klf4), and associated transcription factors

and chromatin regulators (cMyc, E2F1, p300, Med1, and

Med12). As the majority of the Jmjd2b unique targets mapped

to ‘‘nonpromoter’’ regions (Figure 3B), we generated ChIP-seq

intensity heat maps around the summit of all three peak sets,

which were further classified into ‘‘promoter’’ and ‘‘distal’’ re-

gions. We found that the three target classes associated with

Identity

s. The scale bar represents 20 mm.

sitive cells were quantified through fluorescence-activated cell sorting, and the

a are represented as mean ± SEM; n = 3; p values were calculated using t test;

C differentiation. mRNA expression was normalized by actin, and expression

gene. Data are represented as mean ± SEM; n = 3; p values were calculated

differentiation.

c knockdown cells. Data are represented as mean ± SEM; n = 3; p values were


D

C

B

154

30699

Jmjd2c KD_up Jmjd2b KD_up

23982

583

Jmjd2c KD_down Jmjd2b KD_down

Downregulated genes

Upregulated genes

A

Tbx

3

Klf4

Esr

rb

Nan

og

Zfp

42

Sta

t3

Sox

2

Pou

5f1

Spa

rcl1

Sox

7

Pdg

fra

Pth

1r

Hnf

4a

Sox

17

Gat

a6

Gat

a4

Col

4a2

Lam

b1-1

Fox

a2

Sm

ad4

Sm

ad2

Nod

al

Lefty

1

Nkx

1-2

Pitx

2

Fgf

8

Fgf

5 T

Msx

2

Srf

Wnt

3

Sox

18

Pax

6

Otx

2

Krt

18

Nes

Zic

1

Fab

p7

Gat

a3

Cdx

2

Gap

dh

Act

b

1

0.1

0.01

10

100

1000

0.1

1

10

100

Jmjd

2c/K

dm4c

Jmjd

2b/K

dm4b

Jmjd

2a/K

dm4a

Jmjd

2d/K

dm4d

Tbx

3

Klf4

Esr

rb

Nan

og

Zfp

42

Pou

5f1

Sox

2

Sta

t3

Pdg

fra

Pth

1r

Hnf

4a

Gat

a6

Spa

rc

Gat

a4

Sox

17

Sox

7

Col

4a2

Lam

b1-1

Fox

a2

Nod

al

Sm

ad2

Lefty

1

Fgf

8

Pitx

2

Fgf

5

Srf

Wnt

3

Nes

Sox

18

Pax

6

Otx

2

Tubb

3

Olig

2

Zic

1

Olig

1

Gli1

Fab

p7

Olig

3

Gap

dh

Act

b

Kdm4 family members

ESC-specific genes Endoderm Mesoendoderm Mesoderm

Ectoderm Neuroectoderm House keeping genes

Kdm4 family members

ESC-specific genes Endoderm Mesoendoderm Mesoderm

Ectoderm Neuroecto-derm

Trophoecto-derm

House keeping genes

Fol

d ch

ange

Jm

jd2b

KD

/Ant

i-GF

P K

D

m

RN

A le

vel (

log

scal

e)

Fol

d ch

ange

Jm

jd2c

KD

/Ant

i-GF

P K

D

m

RN

A le

vel (

log

scal

e)

Jmjd

2c/K

dm4c

Jmjd

2b/K

dm4b

Jmjd

2a/K

dm4a

Jmjd

2d/K

dm4d

p-value =1E-256

p-value =2E-174

Anti-GFP vs Jmjd2b KD

Enr

ichm

ent s

core

Anti-GFP vs Jmjd2b KD

mES undiff. upregulated genes 6 day diff. upregulated genes

NES= -3.27p-value<0.01

NES= 2.67p-value<0.01

Enr

ichm

ent s

core

Anti-GFP vs Jmjd2c KD Anti-GFP vs Jmjd2c KD

NES= 2.16p-value<0.01 NES= -2.64

p-value<0.01

mES undiff. upregulated genes 6 day diff. upregulated genesE F

−2.5

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5A

nti-G

FP

KD

_rep

1 2per_D

K P

FG-itn

A3per_

DK

PF

G-itnA

Jmjd

2b K

D_r

ep1

Jmjd

2b K

D_r

ep2

Jmjd

2b K

D_r

ep3

Jmjd

2c K

D_r

ep1

Jmjd

2c K

D_r

ep2

-1 -0.50 0.51

Diff d6 / Undiff ESCs(log2 scale)


Molecular Cell



Molecular Cell


different histone marks and transcription factors. Jmjd2c unique

targets strongly overlapped with bivalent marks (both

H3K4me3 and H3K27me3 marks), and also with components

of the PRC2 complex (Ezh2 and Suz12) at promoter and distal

target loci (Figure 3C; Figure S3E). Approximately 80% of Ezh2

and Suz12 targets overlapped with Jmjd2c targets genome-

wide (Figure 3D). Additionally, H3K36me3 and RNA Pol II occu-

pancy was reduced at Jmjd2c unique promoter and distal

target loci (Figure 3C; Figure S3E), suggesting that the majority

of Jmjd2c unique targets are poised or repressed. In contrast,

Jmjd2b-Jmjd2c common and Jmjd2b unique targets showed

reduced occupancy of H3K27me3, Ezh2, and Suz12 and gain

of H3K4me3, H3K36me3, and RNA Pol II occupancy both at

promoters and distal target loci (Figure 3C; Figure S3E), indi-

cating that these targets are transcriptionally active. Notably,

we found that binding of ESC-Core transcription factors Nanog

and Sox2 but not Oct4 was strongly biased toward Jmjd2b

unique targets rather than the other two peak sets at both pro-

moter and distal regions (Figures 3E and 3F). This was further

supported by peak overlap analysis (Figure 3G). Nonetheless,

H3K27ac, H3K4me1, p300, mediator (Med1, Med12), and

Klf4 binding sites overlapped with both Jmjd2b unique and

Jmjd2b-Jmjd2c common targets, and targets of cMyc and its

associated factor E2f1 showed significant overlap with

Jmjd2b-Jmjd2c common target genes at promoter and distal

target loci (Figures 3E–3G).

For a global view of Jmjd2b unique, Jmjd2c unique, and

Jmjd2b-Jmjd2c common targets along with transcription factors

and histone marks, we collated genome-wide targets. Hierarchi-

cal clustering tree and heat map representation of occupancy

correlation revealed three distinct ESC regulatory modules:

Polycomb, Myc, and Core, as described previously (Kim et al.,

2010). Jmjd2c unique targets segregated to the Polycomb mod-

ule, whereas Jmjd2b unique and Jmjd2b-Jmjd2c common target

genes correlated to the Core and Myc modules, respectively,

which were interconnected by targets of mediators Med1 and

Med12 (Figure 3H).

To determine whether genes targeted by Jmjd2b and Jmjd2c

correlate with different functions as predicted by the correlation

of common targets with distinct modules, we performed GO

analysis. Jmjd2b-Jmjd2c common target genes were highly

enriched for transcriptional regulation, DNA binding, and cell-

cycle-related terms, whereas Jmjd2b and Jmjd2c unique targets

related to nucleosome assembly and several development pro-

cesses, respectively (Figure S3F). In addition, IPA demonstrated

that Jmjd2b-Jmjd2c common target genes were closely related

to DNA damage and many cancer-related pathways, such as

Figure 2. Jmjd2b/Kdm4b and Jmjd2c/Kdm4c Are Essential for mESC

(A) Hierarchical clustering of gene expression profiles fromAnti-GFP (control) and

shRNAs)depletedcells. Theaverage foldchange fromtwodifferent shRNAswasca

redcurveshows the relativeexpressionchangesof thesecluster genesbetweenun

(B) Venn diagrams representing overlapping up- or downregulated genes betwe

(C and D) Gene expression analyses for ESC-specific genes and lineage-specific

were used as internal controls. Data shown are averaged from three biological rep

using ANOVA; p < 0.01.

(E and F) GSEA of a differentially expressed gene set of Jmjd2b and Jmjd2c dep

See also Figure S2.

p53, PI3K-AKT, and ESC pluripotency (Figure S3G), consistent

with the Myc module (Kim et al., 2010). Meanwhile, Jmjd2c

unique target genes were mainly associated with several

neuronal signaling pathways (Figure S3G), compatible with

association of the PRC2 module with development. Taken

together, these data provide evidence that Jmjd2b and Jmjd2c

act individually and combinatorially through different regulatory

cassettes in mESCs.

Jmjd2b/Kdm4b and Jmjd2c/Kdm4c Interact with KeyComponents of Different Regulatory ModulesNuclear factors that exhibit common targets often physically

associate (Kim et al., 2010). Therefore, we examined physical

interactions between Jmjd2b and Jmjd2c with core compo-

nents of each regulatory module. Coimmunoprecipitation (co-

IP) with Ezh2 antibody revealed an association of Jmjd2c and

components of the PRC2 complex (Ezh2, Suz12, Jarid2) (Fig-

ure 3I). In contrast, other Kdm4 family members, Jmjd2a,

Jmjd2b, and Jmjd2d, did not coimmunoprecipitate with Ezh2.

Conversely, co-IP with Jmjd2c antibody confirmed interaction

between Jmjd2c and members of the PRC2 complex, but not

with key components (Nanog and Sox2) of other regulatory

modules (Figure 3I). The interaction between Jmjd2c and

PRC2 components is specific, as Jmjd2c does not interact de-

tectably with the PRC1 subunit Bmi1. Jmjd2c does not interact

with Jmjd2b (Figure 3I). We further examined whether stability of

the PRC2 complex is affected upon depletion of Jmjd2c, and

the converse. The level of Ezh2 protein was unaffected upon

depletion of Jmjd2c; moreover, the level of Jmjd2c protein

was unaltered upon Ezh2 depletion (Figure S3H). Finally, co-IP

of Jmjd2b revealed association with Oct4, Nanog, and cMyc,

but not with Ezh2 and Suz12 of the PRC2 complex (Figure 3J).

These data strengthen the functional relationship between

Jmjd2b and Jmjd2c with their central components of different

ESC regulatory modules.

To explore molecular changes upon depletion of Jmjd2b

and Jmjd2c, we examined global histone marks, including

H3K9me2/me3 and H3K36me2/me3, the putative substrates of

the Jmjd2b and Jmjd2c demethylases (Cloos et al., 2006; Agger

et al., 2008). We did not observe significant global changes,

including H3K9me2/me3 and H3K36me2/me3 (Figures S3I and

S3J). To examine potential redundant function of Jmjd2b and

Jmjd2c for these histone modifications, knockdown of both

Jmjd2b and Jmjd2c was performed (Figure S3K). An increased

level of global H3K9me3 was observed (Figure S3L). This finding

implies redundant or overlapping roles among the members of

the same subclass.

Identity

Jmjd2b (Jmjd2b-1 and Jmjd2b-4 shRNAs) and Jmjd2c (Jmjd2c-1 and Jmjd2c-2

lculated. Theheatmap represents the zscoreof expressionvalues,whereas the

differentiatedanddifferentiatedESCssmoothedbyamovingwindowof size50.

en Jmjd2b and Jmjd2c depleted cells.

differentiation genes from Jmjd2b and Jmjd2c depleted cells. Gapdh and actin

licates. Data are represented as mean ± SEM; n = 3; p values were calculated

leted cells. NES, normalized enrichment score; p, nominal p value.


B

PromoterExonIntronIntergenic region


32.8%

18% 9.2%

40%

69.4%

19.3%

4%

7.3%

A

8732

670711022

Jmjd2b Jmjd2c

Jmjd2b/Jmjd2c

Jmjd2b unique (~60%) Jmjd2c unique (~60%)Jmjd2b & Jmjd2c common (~40%)

All Peaks

Ezh2Jarid2Suz12

BivalentJmjd2c_unique

H3K27me3_OnlyJmjd2b_Jmjd2c Common

H3K4me3_OnlyH3K27ac_Promoter

p300_PromoterE2f1

ZfxcMycnMyc

Med12Med1

Klf4Tcfcp2l1

Jmjd2b_uniqueH3K4me1

P300_Distal

NanogSox2Oct4

Smad1Stat3

−0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5

H3K27ac_Distal


10.3%

47.8%

12.5%

29.5%

22.3%

4.3%

8.6%64.8%

14.5%3.5%5.2%

76.8%

Peak overlap

Jmjd2b & Jmjd2c distal target loci

Jmjd2b−Jmjd2c common

Jmjd2b unique

Jmjd2c unique


Jmjd2b unique

Jmjd2c unique

Jmjd2b Jmjd2c

Ezh2

Jmjd2b Jmjd2c

Suz12

Jmjd2b

Jmjd2c Nanog

cMyc

Jmjd2b Jmjd2c

C

D

E F

G

H I

Polycomb module

Core module

Myc module

Jmjd2b genome-wide peaks (16087) Jmjd2c genome-wide peaks (18025)

Genome-wide peak distribution

6143

845

8699 7704

3262 1052

33 2728

60298523 5548

170 4507

1435

5659

10466 17708

6096 3052

556 901 8513 10247

467

5427

219 775 1564

J

Inpu

t

Ezh2:

IP

Anti-Jmjd2c/Kdm4c

Anti-Suz12

Anti-Ezh2Ig

G

Anti-Jmjd2b/Kdm4b

Anti-Jarid2

Anti-Jmjd2a/Kdm4a

Anti-Jmjd2d/Kdm4d

Inpu

t

Jmjd2

c:IP

IgG

Anti-Ezh2

Anti-Jmjd2b

Anti-Bmi1

Anti-Suz12

Anti-Jarid2

Anti-Oct4

Anti-Nanog

Anti-Sox2

Anti-Jmjd2b/Kdm4b

Anti-Myc

Anti-Gcn5

Inpu

t

IgG

Jmjd2

b : I

P

Anti-Ezh2

Anti-Suz12

Anti-Nanog

Anti-Sox2

Ezh2 Suz12 Pol2 InputH3K36me2 H3K36me3H3K9me2 H3K9me3H3K4me3 H3K27me3Jmjd2cJmjd2b

5

4.5

4

3.5

3

2.5

2

1.5

1

Jmjd2b & Jmjd2c promoter target loci


Jmjd2b unique

Jmjd2c unique

6

5.5

5

4.5

4

3.5

3

2.5

2

1.5

1

H3K27ac H3K4me1 Nanog Oct4 Sox2 P300 Med1 Med12 Tet1 Klf4 cMyc E2f1

Jmjd2b & Jmjd2c promoter target loci6

5.5

5

4.5

4

3.5

3

2.5

2

1.5

1H3K27ac H3K4me1 Nanog Oct4 Sox2 P300 Med1 Med12 Tet1 Klf4 cMyc E2f1


Molecular Cell



Molecular Cell


Jmjd2b/Kdm4b and Nanog Act through anInterconnected Regulatory LoopTo interrogate potential molecular mechanisms of Jmjd2b

action, we generated ChIP-seq data sets of Nanog, Oct4, and

cMyc, as well as histone marks H3K4me3, H3K27me3,

H3K36me3, H3K9me3, and H3K27ac from both control (Anti-

GFP) and Jmjd2b-depleted ESCs. We calculated the log2 ratio

(M value) of ChIP-seq intensities between the two cell types

using MAnorm, a quantitative comparison algorithm (Shao

et al., 2012). The vast majority of Nanog binding peaks were

reduced >2-fold upon Jmjd2b knockdown (corresponding to

M value < �1) (Figure 4A; Table S7). As Jmjd2b unique targets

predominantly occupy distal regions, we assessed H3K27ac

occupancy at distal regions from control and Jmjd2b depleted

cells. Global occupancy of H3K27ac was unchanged at both

promoter and distal regions upon depletion of Jmjd2b (Fig-

ure 4B). Further analyses showed that global Nanog occupancy

was not dependent on Jmjd2b occupancy (Figure 4C), although

global Nanog occupancy was reduced significantly upon deple-

tion of Jmjd2b, suggesting other mechanisms by which Jmjd2b

regulates Nanog binding.

To investigate how Jmjd2b impinges on the Nanog regulatory

pathway, we examined expression of Nanog upon Jmjd2b

knockdown and observed significant reduction of Nanog, but

not Oct4, expression (Figures 4D and 4E). Subsequently, we

examined binding of Jmjd2b, Jmjd2c, Nanog, Oct4, mediators

(Med1 and Med12), and H3K27ac specifically at the Nanog

locus, and observed loss of Nanog binding selectively at its

enhancer locus upon Jmjd2b knockdown (Figure 4F). Similarly,

reduction or loss of Nanog binding was also found at regulatory

regions of several of its downstream target genes, including both

ESC-specific (Klf4 and Tbx3) and lineage-specific genes (Lefty1

and Otx2) upon knockdown of Jmjd2b (Figure 4G; Figures S4A–

S4C). Next, we determined whether Nanog regulates Jmjd2b

expression; surprisingly, we found that expression of Jmjd2b

was also reduced upon depletion of Nanog (Figure 4H) and

that occupancy of Nanog was reduced at the Jmjd2b locus in

Jmjd2b knockdown cells (Figure 4I), suggesting that Jmjd2b

andNanog control each other through a regulatory loop. Further-

more, we compared global gene expression changes upon

knockdown of Jmjd2b, Jmjd2c, Oct4, and Nanog. We observed

Figure 3. Genome-wide Mapping of Jmjd2b/Kdm4b and Jmjd2c/Kdm4

(A) Pie charts showing genome-wide peak distribution of Jmjd2b and Jmjd2c.

(B) Venn diagram showing the number of Jmjd2b and Jmjd2c overlapped and u

unique, and Jmjd2b-Jmjd2c common peaks.

(C and E) Heat maps showing the ChIP intensities of selected histone marks, TF

the ±5 kb regions around the peak summit of Jmjd2b unique, Jmjd2c unique, an

(D and G) Venn diagram showing the peak overlaps between Jmjd2b, Jmjd2c, a

(F) Heat map showing the ChIP intensities of histonemarks, TFs, and chromatin re

unique, and Jmjd2b-Jmjd2c common target loci at distal regions.

(H) Correlation map of binding loci showing the degree of co-occupancy between

defined by the combination of Jmjd2b and Jmjd2c as described in (B). Here, the co

is derived from hierarchical clustering.

(I) Coimmunoprecipitation using antibodies against Ezh2, Jmjd2c, and IgG. End

mESCs, and their interacting partners were analyzed by western blot using antib

(J) Jmjd2b antibody was used to immunoprecipitate endogenous Jmjd2b from nu

blot using antibodies. The asterisk indicates unspecific interactions.

See also Figure S3.

the highest correlation of global gene expression changes

between Jmjd2b and Nanog knockdown samples (Figure 4J).

Genes differentially expressed upon knockdown of Nanog signif-

icantly overlapped with differentially expressed genes following

Jmjd2b depletion (Figure 4K), indicating that their regulatory

pathways are tightly interconnected. Further analyses revealed

that the combined binding targets of Jmjd2b and Nanog

highly correlated with Jmjd2b-depleted downregulated genes,

whereas Nanog-only binding targets were correlated with

Jmjd2b-depleted upregulated genes (Figure 4L). These findings

are consistent with the prior observation that Nanog acts combi-

natorially in target gene activation but binds alone or with few

other proteins at repressed or nonexpressed targets (Kim

et al., 2008). Taken together, our data argue that Jmjd2b and

Nanog act through highly interconnected regulatory pathways.

To test this further, we performed rescue experiments with full-

length Jmjd2b, the catalytic domain (JmjC) of Jmjd2b (1–333 aa),

a catalyticmutant (HTE>ATA) of Jmjd2b, and full-length Nanog in

Jmjd2b-deficient ESCs (Figures S1F, S4L and S4M). Full-length

Jmjd2b fully rescued Nanog expression, and vice versa, in

Jmjd2b-deficient ESCs, whereas the catalytic domain (JmjC)-

containing Jmjd2b (1–333 aa) and the catalytic mutant

(HTE>ATA) of Jmjd2b only partially rescued Nanog expression

(Figures S4L and S4M), which is correlated with the restored

ESC growth phenotype and SSEA-1 expression (Figure S1F).

Our findings fail to replicate a prior report that Jmjd2c binds at

the Nanog promoter and regulates Nanog expression through

H3K9me3 levels (Loh et al., 2007). We observed Jmjd2b occu-

pancy at the Nanog locus, but did not detect a change in

H3K9me3 at this locus upon depletion of Jmjd2b (Figure S4D).

Moreover, we observed significant reduction of Nanog expres-

sion and occupancy at target genes (Nanog, Klf4, Otx2) upon

depletion of Jmjd2b but not upon depletion of Jmjd2c (Fig-

ure S4E), suggesting Jmjd2b-Nanog interconnected regulation.

Jmjd2c/Kdm4c Assists PRC2 in Repression in mESCsJmjd2c associates with the PRC2 complex, and targets of

Jmjd2c in chromatin correlate with the PRC2 regulatory module.

To gain insight into these relationships, we performed ChIP-

seq of Oct4, Nanog, Ezh2, cMyc, H3K4me3, H3K27me3,

H3K36me3, and H3K9me3 in both control (Anti-GFP) and

c in mESCs by ChIP-Seq

nique peaks. Pie charts represent the distribution of Jmjd2b unique, Jmjd2c

s, and chromatin regulators including components of PRC2 and RNA Pol II at

d Jmjd2b-Jmjd2c common promoter target loci.

nd the other factors Ezh2, Suz12, Nanog, and cMyc.

gulators at the ±5 kb regions around the peak summit of Jmjd2b unique, Jmjd2c

selected histone modifications, TFs, chromatin regulators, and three peak sets

lor scale represents the Pearson correlation coefficient, and the clustering tree

ogenous Ezh2 and Jmjd2c were immunoprecipitated from nuclear extracts of

odies as shown on the right. Asterisks indicate unspecific interactions.

clear extracts of mESCs, and its interacting partners were analyzed by western


Anti-G

FP_kd

Jmjd2

b-1_

kd

Anti-Actin

Anti-Jmjd2b

Anti-Oct4

Anti-Nanog

Jmjd2b Nanog Oct4

M value = log2(Jmjd2b kd / Anti-GFP kd)

Promoter H3K27ac

Distal H3K27ac

B

D E

F

G

H


−20 −15 −10 −5 0 5 10

−20 −15 −10 −5 0 5 10

Nanog with Jmjd2b targets

Nanog without Jmjd2b targets


Anti-GFP_kd Jmjd2b-1_ kd

Anti-GFP_kd Jmjd2b-1_ kd C

log2 ratio of Jmjd2c Kd vs control

−0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5

Nanog kd

Jmjd2b kd

712

163

110

Down-regulated genes

695

158

288 Nanog kd

Jmjd2b kd

Up-regulated genes

Anti-G

FP_kd

Jmjd2

b-4_

kd

Anti-Oct4

Anti-Nanog

Anti-Jmjd2b

Anti-Actin

Anti-G

FP_kd

Jmjd2

c-1_

kd

Anti-Oct4

Anti-Nanog

Anti-Jmjd2c

Anti-Actin

1 0.66 0.25 0.01

0.66 1 0.36 0.18

0.25 0.36 1 0.44

0.01 0.18 0.44 1

I

J

−20 −15 −10 −5 0 5 10

Nanog

Oct4

cMyc

H3K27ac

H3K4me3

H3K27me3

H3K9me3

H3K36me3

Anti-GFP_kd Jmjd2b-1_ kd A

K L

p-value =7E-74

p-value =2E-94

Klf4

Nanog_Anti-GFP_kd

Nanog_Jmjd2b-1_kd

Oct4_Anti-GFP_kd

Oct4_Jmjd2b-1_kd

Med1

Med12

H3K27ac_Anti-GFP_kd

H3K27ac_Jmjd2b-1_kd

H3K4me3

RNA Pol II

BirA

Jmjd2b-Fb

020

020

015

015

025

25

012

012

50

90

0

0

0

0150

15

Nanog Jmjd2b Jmjd2cOct4

Kdm4b/ Jmjd2b

010

0

10

08.5

08.5

035

035

014

010

025

0

60

Nanog_Anti-GFP_kd

Nanog_Jmjd2b-1_kd

Oct4_Anti-GFP_kd

Oct4_Jmjd2b-1_kd

Med1

Med12

H3K27ac_Anti-GFP_kd

H3K27ac_Jmjd2b-1_kd

H3K4me3

RNA Pol II

Jmjd2b-Fb 0700

70

BirA

Pearson Correlation Coefficient

log2 ratio of Jmjd2b Kd vs control

log2 ratio of Nanog Kd vs control

log2 ratio of Oct4 Kd vs control 0 2 4 6 8 10 12

Jmjd2b & Nanog common target genes

Nanog unique target genes

Jmjd2b unique target genes

−log10(P−value)

Overlap with Jmjd2b KD FC2 downregulated GenesOverlap with Jmjd2b KD FC2 upregulated Genes

Jmjd2

b-1_

kd

Anti-G

FP

Anti-G

FP

Anti-G

FP0.0

0.5

1.0

1.5

Fol

dch

ange

(mR

NA

leve

l)

Jmjd2

b-1_

kd

Jmjd2

b-1_

kd

Anti-G

FP_kd

Nanog

_kd

Nanog

_kd

Nanog

_kd

Nanog

_kd

0.0

0.5

1.0

1.5

Fold

chan

ge(m

RN

Ale

vel)

Anti-G

FP_kd

Anti-G

FP_kd

Anti-G

FP_kd

**

*

** *

***

*

0

10

010

060

060

040

040

016

Nanog

030

030

0

750

65

0

20

Nanog_Anti-GFP_kd

Nanog_Jmjd2b-1_kd

Oct4_Anti-GFP_kd

Oct4_Jmjd2b-1_kd

Med1

Med12

H3K27ac_Anti-GFP_kd

H3K27ac_Jmjd2b-1_kd

H3K4me3

RNA Pol II

BirA

Jmjd2b-Fb


Molecular Cell



Molecular Cell


Jmjd2c knockdown mESCs. Quantitative comparison indicated

that none of these factors demonstrated a significant global

change of binding intensity (Figure 5A; Table S7), suggesting

that the predominant effect of Jmjd2c depletion is restricted to

a subset of genes.

Genes that displayed increased expression upon Jmjd2c

knockdown were composed principally of developmentally

regulated genes (Figures S2D and S2E). Because this class is

highly enriched in bivalent marks (and the presence of Ezh2) in

ESCs, we compared the binding of Ezh2/H3K27me3 and Jmjd2c

at the promoter region of these genes. We observed highly sig-

nificant co-occupancy of Jmjd2c and Ezh2 at these Jmjd2c-

depleted upregulated genes (Figure 5B). Yet neither Jmjd2c

nor Ezh2 alone significantly occupies these upregulated genes,

suggesting that Jmjd2c and Ezh2 act combinatorially in gene

repression and that Jmjd2c exerts a direct gene regulatory influ-

ence through occupancy of these promoters (Figure 5B). More-

over, Jmjd2c knockdown did not significantly alter occupancy of

Ezh2/H3K27me3 at Jmjd2c-depleted upregulated genes (Fig-

ure 5C; Figure S5A), indicating that Ezh2 alone is insufficient to

repress these upregulated genes and requires Jmjd2c for full

repression. Further study of several developmental genes,

Pitx2, Fgf5, Wnt3, T, and Olig3 that were strongly upregulated

in Jmjd2c knockdown cells, showed that indeed these gene

loci are occupied by both Jmjd2c and Ezh2. However, the bind-

ing of Ezh2 and the presence of H3K27me3 were largely

‘‘invariant’’ at these loci upon depletion of Jmjd2c (Figures 5D

and 5F; Figures S5B and S5C). Thus, although recruitment of

PRC2/Ezh2 and H3K27me3 deposition are independent of

Jmjd2c, full repression associated with PRC2 at these targets

requires Jmjd2c.

Jmjd2c/Kdm4c and PRC2 Block Transcription at PRC2Target GenesBecause the H3K27me3 mark is largely unchanged at upregu-

lated genes upon Jmjd2c depletion (Figures 5C–5F; Figure S5A),

we speculated that Jmjd2c might affect other aspects of PRC2-

mediated repression. Recent studies indicate that bivalent

genes bound by PRC2 exhibit RNA Pol II pausing at proximal

Figure 4. Jmjd2b/Kdm4b and Nanog Act through an Interconnected R

(A) Box plots of the log2 fold changes of binding intensities between Anti-GFP (c

H3K4me3, H3K27me3, H3K36me3, and H3K9me3. Red crosses correspond to

(B) Box plot of the log2 fold changes of H3K27ac marks at its promoter and dista

(C) Box plot of the log2 fold changes of Nanog binding intensities with or without

(D) mRNA expression levels of Jmjd2b, Nanog, and Oct4 from Anti-GFP (control)

represented as mean ± SEM; n = 3; p values were calculated using t test; **p < 0

(E) Western blot analyses of Nanog and Oct4 from Jmjd2b (using two individual

(F and G) Genomic tracks of ChIP intensities of factor and histone mark binding

(H)mRNA expression levels of Jmjd2b, Jmjd2c, Oct4, and Nanog fromNanog KD a

Data are represented as mean ± SEM; n = 3; p values were calculated using t te

(I) Genomic tracks of ChIP intensities of factor and histone mark binding at the Kdm

(J) Clustering patterns of global gene expression changes by log2 ratio in Jmjd2b, J

Colors and numbers represent the Pearson correlation coefficient.

(K) Venn diagram of overlapped up- or downregulated genes in Jmjd2b and Nan

(L) p values represent enrichment of overlap between up- or downregulated gene

and target genes occupied by either Jmjd2b or Nanog or both inmESCs (see Supp

target genes). The height of the bars represents a�log10-transformed p value, de

p value of 0.01, above which overlap will be considered as significantly enriched

See also Figure S4.

promoters with reduction of productive transcriptional elonga-

tion (Min et al., 2011), thereby providing another mechanism of

repression. Through co-IP, we observed that both Ezh2 and

Jmjd2c interact with pausing factors (NelfA and Spt5), as well

as with components of the core transcriptionmachinery involved

in initiation and elongation, including unphosphorylated RNA Pol

II (RNAP) (initiation), RNA Pol II Ser-5p (RNAP-S5p) (initiation),

RNA Pol II Ser-2p (RNAP-S2p) (elongation), Cdk9 (initiation),

Ctr9 (component of Paf1c elongation complex), and cMyc (initi-

ation; involved in pause release) (Peterlin and Price, 2006; Rahl

et al., 2010; Nechaev and Adelman, 2011) (Figure S6A). Further-

more, full-length Jmjd2c, as well as its catalytic domain (1–333

aa), interacted with Ezh2, RNAP, and Cdk9 (Figure S6B).

To gain further insights into themechanistic role of Jmjd2c and

Ezh2 in RNA Pol II pausing, we examined three sets of genes

defined by promoter occupancy of Jmjd2b, Jmjd2c, H3K4me3,

and H3K27me3: Jmjd2b-Jmjd2c common active genes

(H3K4me3-only), Jmjd2c unique active genes (H3K4me3-only),

and all H3K27me3 repressed genes (including both bivalent

and H3K27me3-only genes, most of which are PRC2 target

genes and overlap with Jmjd2c target genes) (Figure 6A; Fig-

ure S3D). Jmjd2b-Jmjd2c common active genes showed

strong binding of transcription initiation factors, including

RNAP, RNAP-S5p, NelfA, Spt5, and cMyc at the transcription

start site (TSS), as well as strong occupancy of factors

associated with elongation, including RNAP-S2p, Ctr9, Spt5,

H3K36me3, and H3K79me2 at gene bodies (Figure 6A). Binding

of these transcription initiation and elongation factors at Jmjd2c

unique active genes was reduced compared to Jmjd2b-Jmjd2c

common active genes. In addition, repressed genes bound by

Jmjd2c and PRC2 (H3K27me3) showed dramatic loss of binding

of these factors at the TSS and gene bodies (Figure 6A), indi-

cating that they are indeed poised/repressed in mESCs. Further-

more, comparing the expression levels among these three

sets of genes confirmed that they exhibit significantly different

levels of expression, which are themselves correlated with occu-

pancy of transcription initiation and elongation factors (Fig-

ure 6B). Specifically, we observed that combined PRC2/Jmjd2c

target genes are occupied by RNAP, RNAP-S5p, NelfA, and

egulatory Loop

ontrol) and Jmjd2b knockdown (KD) ESCs for Nanog, Oct4, cMyc, H3K27ac,

outliers.

l targets from Anti-GFP (control) and Jmjd2b KD cells.

Jmjd2b occupancy in Anti-GFP (control) and Jmjd2b KD cells.

and Jmjd2b KD cells. Transcript levels were normalized using Gapdh. Data are

.001, *p < 0.01.

shRNAs) and Jmjd2c and Anti-GFP (control) depleted cells.

at Nanog and Klf4 loci.

nd Anti-GFPKD (control) cells. Transcript levels were normalized usingGapdh.

st; **p < 0.001, *p < 0.01.

4b/Jmjd2b locus. Jmjd2b shows no enrichment of its binding at its own locus.

mjd2c, Nanog, andOct4 depleted cells relative to their corresponding controls.

og knockdown cells.

s in Jmjd2b knockdown cells compared to Anti-GFP (control) knockdown cells

lemental Experimental Procedures for the definition of Jmjd2b andNanogChIP

rived from right-tailed Fisher’s exact test, and the dotted line corresponds to a

.


A

B C

D

−20 −15 −10 −5 0 5 10

Nanog

Oct4

Ezh2

cMyc

H3K4me3

H3K27me3

H3K9me3

H3K36me3

Jmjd2c-1_kd Anti-GFP_kd

M value = log2(Jmjd2c kd / Anti-GFP kd)

E

F

Pitx2

0

10

0

10

015

0

15

015

0

15

0

10

010

0

3

03

0

8

08

Ezh2_Anti-GFP_kdEzh2_Jmjd2c-1_ kd

H3K27me3_Anti-GFP_kdH3K27me3_Jmjd2c-1_ kd


BirAJmjd2c-FB



Fgf5

0

10

0

10

010

010

010

010

08

08

0

3

03

0

5

05




BirAJmjd2c-FB



Wnt3

0

10

0

10

010

0

10

015

015

0

10

010

010

0

10

010

010




BirAJmjd2c-FB



Jmjd2c and Ezh2 common target genes

Jmjd2c unique target genes

Ezh2 unique target genes

Genes occupied by Ezh2 in both Anti-GFP kd and Jmjd2c kd cells

Genes occupied by Ezh2 only in Anti-GFP kd cells

Genes occupied by Ezh2 only in Jmjd2c kd cells

0 5 10 15 20

−log10(P−value)

Overlap with Jmjd2c kd downregulated GenesOverlap with Jmjd2c kd upregulated Genes

0 2 4 6 8 10 12 14 16 18

−log10(P−value)

Overlap with Jmjd2c kd downregulated GenesOverlap with Jmjd2c kd upregulated Genes

Figure 5. Jmjd2c/Kdm4c Assists PRC2 in Repression in mESCs

(A) Box plots of the log2 fold change of binding intensities between Anti-GFP (control) and Jmjd2c depleted cells for Nanog, Oct4, Ezh2, cMyc, H3K4me3,

H3K27me3, H3K36me3, and H3K9me3. Red crosses correspond to outliers.

(B) p values represent enrichment of overlap between up- or downregulated genes from Jmjd2c knockdown cells compared to Anti-GFP (control) knockdown

cells and target genes occupied by either Jmjd2c or Ezh2 or both in mESCs at promoters. The height of the bars represents�log10-transformed p value derived

from right-tailed Fisher’s exact test, and the dotted line corresponds to a p value of 0.01.

(legend continued on next page)

Molecular Cell



Molecular Cell


Spt5 at the promoter regions but not with elongation factors

(RNAP-S2p, Ctr9, H3K36me3, and H3K79me2) at gene bodies,

suggesting that these poised/repressed genes are probably

paused with RNA Pol II (Figure 6A) (Nechaev and Adelman,

2011). We observed that cMyc occupancy directly correlated

with transcription of two sets of active genes but not with

PRC2 target genes (Figure 6A), consistent with a role for cMyc

in active transcription through RNA Pol II pause release at

actively transcribed genes in mESCs (Rahl et al., 2010). These

data point to a role for Jmjd2c in transcriptional repression,

most likely through blocking of RNA Pol II recruitment and/or

pausing of RNA Pol II at Jmjd2c-occupied PRC2 target genes.

Next, we examined occupancy of selected factors and histone

marks involved in transcription initiation and elongation in

Jmjd2c and control (Anti-GFP) knockdown cells. Global occu-

pancy of these factors and histone marks did not differ signifi-

cantly between these cells (Figure 5A; Figure S6C; Table S7).

We then used a quantitative comparisonmethod to discover fac-

tors whose binding is altered significantly at Jmjd2c-depleted

up- and downregulated genes (see Supplemental Experimental

Procedures for detailed methods). Jmjd2c depleted upregu-

lated genes reveal significantly higher binding of RNAP, RNAP-

S5p, RNAP-S2p, Ctr9, and H3K36me3 mark (a mark associated

with transcription elongation), whereas Jmjd2c-depleted down-

regulated genes showed reduced occupancy of these factors

and histone marks (Figure 6C; Figure S6D). Moreover, we

observed increased and decreased occupancy of transcription

initiation and elongation factors at specific upregulated and

downregulated gene loci, respectively, upon Jmjd2c depletion

(Figures 6D–6I; Figures S6E–S6H), indicating correlation be-

tween occupancy of the transcription machinery and gene

expression. We demonstrate that Jmjd2c and Ezh2 co-occupy

significantly at Jmjd2c-depleted upregulated genes compared

to downregulated genes (Figure 5B) and that these upregulated

genes display increased binding of transcription machinery fac-

tors and histone marks upon depletion of Jmjd2c, suggesting

that Jmjd2c directly regulates these upregulated genes. On the

other hand, no correlation was observed between Jmjd2c occu-

pancy and Jmjd2c-depleted downregulated genes (Figure 5B),

indicative of indirect regulation of these genes through Jmjd2c.

Taken together, these data indicate that Jmjd2c is required in

ESCs for full repression by PRC2 at poised/repressed develop-

mental genes.

DISCUSSION

Jmjd2b/Kdm4b and Jmjd2c/Kdm4c Belong to DistinctRegulatory Modules in mESCsDepletion of either Jmjd2b or Jmjd2c leads to ostensibly similar

differentiation phenotypes and comparable overall global gene

expression patterns in ESCs (Figures 1B and 2; Figure S1B).

However, global chromatin occupancy studies of Jmjd2b and

(C) p values represent enrichment of overlap between up- or downregulated gen

cells and the promoter target genes of Ezh2 in these two cell types. The p value

(D–F) Genomic tracks of ChIP-seq intensities of Jmjd2c, Ezh2, and histone mark

loci upon Jmjd2c depletion (Pitx2, Fgf5, and Wnt3).

See also Figure S5.

Jmjd2c demonstrate that each occupies numerous unique and

common targets and correlates with functionally distinct regula-

tory modules. Jmjd2b unique, Jmjd2b-Jmjd2c common, and

Jmjd2c unique targets partition to the Core, Myc, and PRC reg-

ulatory modules of the overall ESC network, respectively (Fig-

ure 3H). Thus, these closely related HDMs subserve specialized

targets, in part relating to their combinatorial partitioning. Meta-

gene analyses provide further support that occupancy of Jmjd2b

and Jmjd2c unique and common targets is correlated with the

pattern of gene expression in mESCs (Figure S3C). Protein-

protein interaction studies corroborate that Jmjd2b and Jmjd2c

physically interact with distinct regulators, specifically Jmjd2b

with factors of the Core pluripotency network and Jmjd2c with

components of the PRC2 complex (Figures 3I and 3J). Therefore,

we speculate that combinatorial binding patterns of HDMs

across the genome will be critical for understanding regulatory

programs in cell-type-specific transcriptional regulation.

Jmjd2b/Kdm4b, Jmjd2c/Kdm4c, and HistoneModificationsCombinations of histone modifications play a crucial role in gene

regulation (Zhou et al., 2011). Jmjd2b and Jmjd2c are classified

as H3K9me2/3 and H3K36me2/3 HDMs (Agger et al., 2008;

Cloos et al., 2006). As H3K36me2/me3 and H3K9me2/3 are

indicative of divergent transcriptional outcomes, it is possible

that Jmjd2b and Jmjd2c function differently in a cell-type-

specific and context-dependent manner. A relatively low level

of H3K9me3 correlates with the highly active and open ESC

and iPSC genome (Meshorer and Misteli, 2006; Soufi et al.,

2012). Recent studies also demonstrate that depletion of all

H3K9 demethylases, including Jmjd2b and Jmjd2c, significantly

inhibits reprogramming of pre-iPSCs to iPSCs. Conversely,

depletion of the H3K9 methyltransferase Setdb1 leads to highly

efficient reprogramming, indicating that removal of H3K9me3 is

essential for establishment of the ESC/iPSC state (Chen et al.,

2013). In this study, we did not observe significant changes in

‘‘global’’ histone levels or occupancy, including H3K9me2/3

and H3K36me2/3 marks, upon depletion of either Jmjd2b or

Jmjd2c (Figures 4A and 5A; Figures S3I and S3J). However, dou-

ble knockdown of Jmjd2b and Jmjd2c led to an increased level

of global H3K9me3 (Figures S3K and S3L), suggesting a redun-

dant function in ESCs. We also monitored the effects of Jmjd2b

and Jmjd2c depletion on histone marks (H3K4me3, H3K27me3,

H3K9me3, and H3K36me3) at several ‘‘individual’’ loci. For

Jmjd2b-Jmjd2c common active genes, we observed reduction

of H3K4me3 and H3K36me3 occupancy at several active gene

loci, such as Klf4, Tbx3, and Esrrb, upon depletion of either

Jmjd2b or Jmjd2c (Figure 7; Figures S4F–S4K and S7A–S7C).

In the context of Jmjd2b-Nanog interconnected regulation, we

did not observe any significant change of H3K9me3 at theNanog

locus and its downstream target genes upon depletion of

Jmjd2b. However, we observed occupancy changes at several

es from Jmjd2c knockdown cells compared to Anti-GFP (control) knockdown

represents the same as in (B).

s in both Anti-GFP (control) and Jmjd2c KD cells at selected upregulated gene


Jmjd2b Jmjd2c cMyc Ezh2 K4me3 K27me3 K79me2 K36me3 NelfA Spt5 Ctr9 RNAP RNAP-S5pRNAP-S2p

Jmjd2b/2c Common Active Genes(+H3K4me3 -H3K27me3)

All H3K27me3 Repressed Genes

3.5

2.5

1.5

1

2

3

4

Jmjd2c Unique Active Genes(+H3K4me3 -H3K27me3)

Jmjd2b & Jmjd2c promoter target genes

Jmjd2b/2c Common Active Genes

Jmjd2c Unique Active Genes

All K27me3 Repressed Genes

Exp

ress

ion

Val

ue

B C

(both bivalent & H3K27me3 only)

A

Name Up-regulated genes (p-value)

Down-regulated genes (p-value)

Significance of differential binding of factors and histone marks at Jmjd2c up and down regulated genes

0

500

1000

1500

2000

2500

3000

3500

4000

P=4E-58

P<1E-100

0

5

05

08.5

08.5

0

6

0

6

05

0

5

03.5

03.5

Pax6

RNA Pol II_Anti-GFP_kd

RNA Pol II_Jmjd2c-1_kd

RNA Pol II Ser5p_Anti-GFP_kd

RNA Pol II Ser5p_Jmjd2c-1_kd



H3K36me3_Anti-GFP_kd

H3K36me3_Jmjd2c-1_kd

Ctr9_Anti-GFP_kd

Ctr9_Jmjd2c-1_kd

0

0.113

-0.734

0.694

0

2.288

0

1.552

0

3.391

Pax6

RNA Pol II_M value

RNA Pol II Ser5p_M value


H3K36me3_M value

Ctr9_M value

G

0

0.556

-0.387

1.108

0

1.053

0

1.552

0

3.072

Pitx2

RNA Pol II_M value



H3K36me3_M value

Ctr9_M value

0

8.5

08.5

012

012

05

05

03

03

03.5

0

3.5

Pitx2





RNA Pol II Ser2p_Anti-GFP_kdRNA Pol II Ser2p_Jmjd2c-1_kd



Ctr9_Anti-GFP_kd

Ctr9_Jmjd2c-1_kd

F

D E

RNA Pol II 1.7E 06 1.0E 06RNA Pol II Ser5p 0.001 4.4E 09RNA Pol II Ser2p 0.005 0.0017Ctr9 0.0002 NSSpt5 NS 3E 18H3K36me3 3.7E-05 4.5E 20Ezh2 NS 0.00017H3K27me3 NS NSH3K4me3 NS NScMyc NS NSH3K9me3 NS NS

0

45

045

025

025

05

05

020

020

07.5

07.5

Klf4



RNA Pol II Ser5p_Anti-GFP_kdRNA Pol II Ser5p_Jmjd2c-1_kd





Ctr9_Anti-GFP_kd

Ctr9_Jmjd2c-1_kd

-0.6880.023

-0.660

-6.3840

-0.6310

Klf4

RNA Pol II_M value



H3K36me3_M value

H I


Molecular Cell



Molecular Cell


discrete loci for H3K4me3 (e.g., Nanog, Klf4, Tbx3, and Esrrb

promoters) and H3K36me3 (e.g., Klf4, Tbx3, and Otx2 gene

bodies) (Figure 7; Figures S4F–S4K). Interestingly, we found

that some of the Jmjd2b-Jmjd2c common active genes are the

downstream targets of Nanog, suggesting a fraction of common

targets is regulated through the Jmjd2b-Nanog regulatory loop

(Figures 4G and 7; Figures S4A–S4C and S7A–S7C). On the

other hand, with respect to the role of Jmjd2c in PRC2-mediated

repression, we observed that the upregulated genes upon

Jmjd2c depletion are significantly co-occupied by Jmjd2c and

Ezh2, and that Jmjd2c assists PRC2 in full repression at these

poised/repressed genes (e.g., Fgf5, Pitx2, Pax6) through tran-

scription repression (Figure 5B). Moreover, these upregulated

genes showed significantly increased binding of the transcrip-

tional machinery (RNAP, RNAP-S5p, RNAP-S2p, Ctr9) and his-

tone marks, especially H3K36me3 (an elongation mark), but

not other histone marks upon depletion of Jmjd2c (Figures 6C–

6I and 7; Figures S6D–S6H), indicative of a link between

Jmjd2c-PRC2 and transcription elongation. This supports a pre-

vious observation in which Polycomb-like-PRC2 recognizes

H3K36me3 and silences active chromatin regions for mainte-

nance of ESC pluripotency (Cai et al., 2013).

Our study reveals unexpected outcomes of the histone deme-

thylase function of Jmjd2b and Jmjd2c and its substrate speci-

ficity. However, rescue studies demonstrate that catalytic

activity of Jmjd2b and Jmjd2c is indeed required for self-renewal

of ESCs. Moreover, our findings indicate a complex interplay

between various histone modifications and/or that Jmjd2b and

Jmjd2c may have additional roles beyond histone targets.

Recent studies have indicated that Jmjd2b is an integral part

of the MLL2 complex that deposits the H3K4me3 mark, and

H3K9me3 is a prerequisite of H3K4me3 deposition for ERa-acti-

vated gene transcription (Shi et al., 2011), suggesting crosstalk

between histone modifications.

Jmjd2b/Kdm4b and Jmjd2c/Kdm4cAct throughDistinctand Combinatorial Pathways in mESCsAs a core pluripotency factor, Nanog functions together with

other Core transcription factors (such as Oct4 and Sox2) to posi-

tively regulate their expression through autoregulatory loops, as

well as occupy and activate ESC-specific genes and repress

lineage-specific genes to maintain the ESC state (Young,

2011). We demonstrate that Jmjd2b is involved in the regulation

of Nanog expression and occupancy of its downstream target

genes. Moreover, Nanog regulates Jmjd2b expression that is

Figure 6. Jmjd2c/Kdm4c and PRC2 Block Transcription at PRC2 Targ

(A) Heat map showing the ChIP-seq densities of selected factors, histone marks,

at the ±5 kb regions around the TSS of promoters of Jmjd2b-Jmjd2c common a

(B) Box plot of gene expression values of Jmjd2b-Jmjd2c common active genes

(C) Significance of binding changes of profiled factors at differentially expressed

calculated by using Kolmogorov-Smirnov test to compare the global distribution

factor at its peak loci associated with up- or downregulated genes relative to thos

are marked in red or green; otherwise they are labeled as nonsignificant (NS).

(D, F, and H) Genomic tracks represent ChIP intensities of transcription initiation

Jmjd2c depleted upregulated (Pitx2, Pax6) and downregulated (Klf4) gene loci.

(E, G, and I) Genomic tracks represent the log2 fold change of binding intensities

(F), and (H), which is referred by M values derived by the MAnorm model. For fu

See also Figure S6.

important for maintenance of Nanog expression through a feed-

back loop (Figures 4A–4I and 7).

In contrast, Jmjd2c physically interactswith components of the

PRC2 repressive complex, and its targets correlate with the PRC

regulatorymodule (Figure 3). Furthermore, PRC2 requires Jmjd2c

for full repressive function (Figure 5). Although the molecular de-

tails of PRC2-mediated gene repression remain incomplete, pre-

vious studies implicate several mechanisms in the control of tran-

scriptional output including the recruitment block of RNAPol II, as

well as promoter-proximal pausing of RNA Pol II at PRC2 target

genes (Boyer et al., 2006; Lee et al., 2006; Min et al., 2011). Our

data suggest that Jmjd2c may influence PRC2 activities for both

RNA Pol II recruitment and pausing at PRC2/Jmjd2c targets (Fig-

ures 6A–6C; Figures S6A–S6D). High co-occupancy of Jmjd2c-

Ezh2 at Jmjd2c-depleted upregulated genes (Figure 5B) and

increased distribution of RNAP, RNAP-S5p, RNAP-S2p, Ctr9,

and H3K36me3 at these Jmjd2c/PRC2-targeted upregulated

genes in Jmjd2c-depleted cells (Figures 6C–6I; Figures S6D–

S6H) argue persuasively for the role of Jmjd2c in transcriptional

repression. Therefore, Jmjd2c is required for the full repressive

functionof PRC2at PRC2/Jmjd2c common target genes inESCs.

Global Jmjd2b and Jmjd2c targets are differentially distrib-

uted. However, Jmjd2b and Jmjd2c share �40% common

targets (Figure 3B). These common targets are associated with

active genes and belong to Myc modules (Figure 3). Jmjd2b

and Jmjd2c do not physically interact with each other, but both

Jmjd2b and Jmjd2c can physically interact with cMyc (Figure 3J;

Figure S6A). Nevertheless, occupancy of cMyc is not signifi-

cantly changed globally or locally upon depletion of either

Jmjd2b or Jmjd2c (Figures 4A and 5A; Figure S7). Interestingly,

we found that some of the Jmjd2b-Jmjd2c common active

genes (Klf4, Tbx3, and Esrrb) are downstream targets of Nanog

and its occupancy was reduced at these loci upon depletion of

Jmjd2b, suggesting a fraction of common targets is regulated

through the Jmjd2b-Nanog regulatory loop.We observed reduc-

tion of H3K4me3 and H3K36me3 at these common target genes

upon depletion of either Jmjd2b or Jmjd2c (Figures S4 and S7)

suggesting, indeed, that Jmjd2b and Jmjd2c combinatorially

regulate these active gene loci.

We present a model in Figure 7 that summarizes our findings

regarding how Jmjd2b and Jmjd2c exert both distinct and

combinatorial functions in ESC identity (Figure 7). Our study sug-

gests that exploration of combinatorial roles of HDMs with other

chromatin regulators, transcription factors, and histone marks

will elucidate multilayered regulatory networks in ESC biology.

et Genes

and a different modified form of RNA Pol II (RNAP, RNAP-S5p, and RNAP-S2p)

ctive genes, Jmjd2c unique active genes, and all H3K27me3 target genes.

, Jmjd2c unique active genes, and all H3K27me3 target genes.

genes in Jmjd2c KD cells relative to Anti-GFP (control) KD cells. p values were

of log2 fold changes of ChIP intensities (M values inferred by MAnorm) of each

e with nondifferentially expressed genes as control. Significant p values (<0.01)

and elongation factor binding in Jmjd2c KD and Anti-GFP (control) KD cells at

at each peak locus between selected pairs of ChIP-seq data sets shown in (D),

rther description, please see the figure legends of Figures S6F and S6H.


Jmjd2c/Kdm4c unique target genes (Poised/ Repressed Genes)

Jmjd2c depletion

ESC state Differentiated state

Fgf5, Pitx2, Pax6

RNAp

Ctr9RNAP-S5p

Jmjd2c RNAP-S2p

H3K36

me3

H3K4m

e3

H3K27

me3 RNAp Ctr9

RNAP-S5p RNAP-S2pH3K

36m

e3

H3K4m

e3

H3K27

me3

H3K36

me3

H3K4m

e3

Nanog

Oct4Med Smc

Jmjd2b

Sox2

Jmjd2cMyc RNAp

P300

Jmjd2b depletion

H3K36

me3

H3K4m

e3

Nanog

Oct4Med SmcSox2

Jmjd2c

H3K27ac

H3K4me1

Myc RNAp

P300

NanogNanog

H3K36

me3

H3K4m

e3

Nanog

Oct4Med Smc

Jmjd2b

Sox2RNAp

P300

Jmjd2b depletion H3K

36m

e3

H3K4m

e3

Nanog

Oct4Med Smc

Sox2RNAp

P300Nanog

Nanog

E2F1

Oct4Sox2

Nanog

Oct4Sox2

Klf4, Esrrb, Tbx3

Nanog

Nanog

Jmjd2b/Kdm4b unique target genes (Active Genes)

Jmjd2b/Kdm4b & Jmjd2c/Kdm4c common target genes (Active Genes)

E2F1

EEDSUZ12

EZH2

EEDSUZ12

EZH2

NanogNanog

NanogNanog

Nanog

Nanog target genes (Klf4, Esrrb, Tbx3, Jmjd2b Otx2, Lefty1) Nanog target genes

H3K36

me3

H3K4m

e3

Nanog

Oct4Med SmcSox2

H3K27ac

Myc RNAp

P300

E2F1Jmjd2b

Jmjd2c depletion

H3K4me1

H3K27ac

H3K4me1

H3K27ac

H3K4me1

H3K27ac

H3K4me1

Nanog

Nanog

RNAp

RNAP-S5p

Ctr9

RNAP-S2p

Jmjd2b

Figure 7. The Working Model of Jmjd2b/Kdm4b and Jmjd2c/Kdm4c in mESCsJmjd2b is generally associated with active genes, whereas Jmjd2c is associated with poised/repressive genes. However, we found that most of the active genes

are occupied by both Jmjd2b and Jmjd2c. We propose a model in which Jmjd2b and Jmjd2c work distinctly and combinatorially. Blue thick and thin lines

correspond to higher and lower levels of transcription, respectively. The black dotted line represents lower binding of Nanog to its target genes upon depletion of

Jmjd2b.

See also Figure S7.

Molecular Cell


EXPERIMENTAL PROCEDURES

ESC Lines and Culture

Mouse ESC lines were maintained in ES medium as documented in Supple-

mental Experimental Procedures.


Lentiviral Production and RNAi Screen

Individual shRNA constructs were used for lentiviral production. We used five

different shRNA lentiviral constructs to knock down each of 20 histone deme-

thylases in mESCs. mESCs were seeded in 48-well plates and infected with in-

dividual shRNA lentiviruses, and the screen was scored in terms of alterations

Molecular Cell


of ESC colony morphology. A detailed protocol of lentiviral production and

infection of mESCs is available in Supplemental Experimental Procedures.

Gene Expression Analysis

mRNA profiling of knocked down samples was performed using Affymetrix

mouse genome 430 2.0 arrays. See Supplemental Experimental Procedures.

ChIP

Bio-ChIP and native antibody ChIP were performed as described elsewhere

(Kim et al., 2008, 2009). Detailed procedures and lists of antibodies are avail-

able in Supplemental Experimental Procedures.

ChIP Sequencing and Library Generation

Purified ChIP DNA was used to prepare Illumina multiplexed sequencing

libraries. A New England BioLabs Next Generation Sequencing kit was used

to prepare the libraries.

ChIP-Seq Data Analysis

Peaks were called using MACS, a peak-finding algorithm to identify regions of

ChIP-seq enrichment over background (Zhang et al., 2008). Detailed analyses

are available in Supplemental Experimental Procedures.

Coimmunoprecipitation

Nuclear extracts were prepared from J1 mouse ESCs (Kim et al., 2010) and

immunoprecipitated using specific antibodies.

ACCESSION NUMBERS

The cDNA microarray and ChIP-seq data have been deposited in the Gene

Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) under accession

number GSE43231.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures,

seven figures, and seven tables and can be found with this article online at

http://dx.doi.org/10.1016/j.molcel.2013.11.011.

ACKNOWLEDGMENTS

We thank Dan Bauer, Jian Xu for critical reading of the manuscript, and David

Hendrix for discussions. We also thank Marc Kerenyi for reagents. We thank

F. Abderazzaq and R. Rubio at the Center for Computational Biology

sequencing facility at the Dana-Farber Cancer Institute for Illumina HiSeq

2000 sequencing. This work was supported by funding from NIH grant HLBI

U01HL100001. S.H.O. is an Investigator of the Howard Hughes Medical

Institute.

Received: May 10, 2013

Revised: September 16, 2013

Accepted: November 13, 2013

Published: December 19, 2013

REFERENCES

Agger, K., Christensen, J., Cloos, P.A., and Helin, K. (2008). The emerging

functions of histone demethylases. Curr. Opin. Genet. Dev. 18, 159–168.

Boyer, L.A., Plath, K., Zeitlinger, J., Brambrink, T., Medeiros, L.A., Lee, T.I.,

Levine, S.S., Wernig, M., Tajonar, A., Ray, M.K., et al. (2006). Polycomb com-

plexes repress developmental regulators in murine embryonic stem cells.

Nature 441, 349–353.

Cai, L., Rothbart, S.B., Lu, R., Xu, B., Chen, W.-Y., Tripathy, A., Rockowitz, S.,

Zheng, D., Patel, D.J., Allis, C.D., et al. (2013). AnH3K36methylation-engaging

Tudor motif of polycomb-like proteins mediates PRC2 complex targeting. Mol.

Cell 49, 571–582.

Chen, X., Xu, H., Yuan, P., Fang, F., Huss,M., Vega, V.B.,Wong, E., Orlov, Y.L.,

Zhang, W., Jiang, J., et al. (2008). Integration of external signaling pathways

with the core transcriptional network in embryonic stem cells. Cell 133,

1106–1117.

Chen, J., Liu, H., Liu, J., Qi, J., Wei, B., Yang, J., Liang, H., Chen, Y., Chen, J.,

Wu, Y., et al. (2013). H3K9 methylation is a barrier during somatic cell reprog-

ramming into iPSCs. Nat. Genet. 45, 34–42.

Cherry, A.B.C., and Daley, G.Q. (2012). Reprogramming cellular identity for

regenerative medicine. Cell 148, 1110–1122.

Cloos, P.A.C., Christensen, J., Agger, K., Maiolica, A., Rappsilber, J., Antal, T.,

Hansen, K.H., and Helin, K. (2006). The putative oncogene GASC1 demethy-

lates tri- and dimethylated lysine 9 on histone H3. Nature 442, 307–311.

Fazzio, T.G., Huff, J.T., and Panning, B. (2008). An RNAi screen of chromatin

proteins identifies Tip60-p400 as a regulator of embryonic stem cell identity.

Cell 134, 162–174.

Hu, G., Kim, J., Xu, Q., Leng, Y., Orkin, S.H., and Elledge, S.J. (2009). A

genome-wide RNAi screen identifies a new transcriptional module required

for self-renewal. Genes Dev. 23, 837–848.

Kagey, M.H., Newman, J.J., Bilodeau, S., Zhan, Y., Orlando, D.A., van

Berkum, N.L., Ebmeier, C.C., Goossens, J., Rahl, P.B., Levine, S.S., et al.

(2010). Mediator and cohesin connect gene expression and chromatin archi-

tecture. Nature 467, 430–435.

Kim, J., Chu, J., Shen, X., Wang, J., and Orkin, S.H. (2008). An extended tran-

scriptional network for pluripotency of embryonic stem cells. Cell 132, 1049–

1061.

Kim, J., Cantor, A.B., Orkin, S.H., and Wang, J. (2009). Use of in vivo bio-

tinylation to study protein-protein and protein-DNA interactions in mouse

embryonic stem cells. Nat. Protoc. 4, 506–517.

Kim, J., Woo, A.J., Chu, J., Snow, J.W., Fujiwara, Y., Kim, C.G., Cantor, A.B.,

and Orkin, S.H. (2010). A Myc network accounts for similarities between

embryonic stem and cancer cell transcription programs. Cell 143, 313–324.

Lee, T.I., Jenner, R.G., Boyer, L.A., Guenther, M.G., Levine, S.S., Kumar, R.M.,

Chevalier, B., Johnstone, S.E., Cole, M.F., Isono, K.-I., et al. (2006). Control of

developmental regulators by Polycomb in human embryonic stem cells. Cell

125, 301–313.

Loh, Y.H., Zhang, W., Chen, X., George, J., and Ng, H.H. (2007). Jmjd1a and

Jmjd2c histone H3 Lys 9 demethylases regulate self-renewal in embryonic

stem cells. Genes Dev. 21, 2545–2557.

Margueron, R., and Reinberg, D. (2011). The Polycomb complex PRC2 and its

mark in life. Nature 469, 343–349.

Meshorer, E., and Misteli, T. (2006). Chromatin in pluripotent embryonic stem

cells and differentiation. Nat. Rev. Mol. Cell Biol. 7, 540–546.

Min, I.M., Waterfall, J.J., Core, L.J., Munroe, R.J., Schimenti, J., and Lis, J.T.

(2011). Regulating RNA polymerase pausing and transcription elongation in

embryonic stem cells. Genes Dev. 25, 742–754.

Mosammaparast, N., and Shi, Y. (2010). Reversal of histone methylation:

biochemical and molecular mechanisms of histone demethylases. Annu.

Rev. Biochem. 79, 155–179.

Nechaev, S., and Adelman, K. (2011). Pol II waiting in the starting gates: regu-

lating the transition from transcription initiation into productive elongation.

Biochim. Biophys. Acta 1809, 34–45.

Orkin, S.H., and Hochedlinger, K. (2011). Chromatin connections to pluripo-

tency and cellular reprogramming. Cell 145, 835–850.

Pedersen, M.T., and Helin, K. (2010). Histone demethylases in development

and disease. Trends Cell Biol. 20, 662–671.

Peterlin, B.M., and Price, D.H. (2006). Controlling the elongation phase of tran-

scription with P-TEFb. Mol. Cell 23, 297–305.

Rahl, P.B., Lin, C.Y., Seila, A.C., Flynn, R.A., McCuine, S., Burge, C.B., Sharp,

P.A., and Young, R.A. (2010). c-Myc regulates transcriptional pause release.

Cell 141, 432–445.


http://www.ncbi.nlm.nih.gov/geo


http://cccb.dfci.harvard.edu/cccb/sequencing

Molecular Cell


Shao, Z., Zhang, Y., Yuan, G.-C., Orkin, S.H., and Waxman, D.J. (2012).

MAnorm: a robust model for quantitative comparison of ChIP-seq data sets.

Genome Biol. 13, R16.

Shi, L., Sun, L., Li, Q., Liang, J., Yu, W., Yi, X., Yang, X., Li, Y., Han, X., Zhang,

Y., et al. (2011). Histone demethylase JMJD2B coordinates H3K4/H3K9

methylation and promotes hormonally responsive breast carcinogenesis.

Proc. Natl. Acad. Sci. USA 108, 7541–7546.

Soufi, A., Donahue, G., and Zaret, K.S. (2012). Facilitators and impediments of

the pluripotency reprogramming factors’ initial engagement with the genome.

Cell 151, 994–1004.

Sridharan, R., Tchieu, J., Mason, M.J., Yachechko, R., Kuoy, E., Horvath, S.,

Zhou, Q., and Plath, K. (2009). Role of the murine reprogramming factors in

the induction of pluripotency. Cell 136, 364–377.


Takahashi, K., and Yamanaka, S. (2006). Induction of pluripotent stem cells

from mouse embryonic and adult fibroblast cultures by defined factors. Cell

126, 663–676.

Xie, L., Pelz, C., Wang, W., Bashar, A., Varlamova, O., Shadle, S., and Impey,

S. (2011). KDM5B regulates embryonic stem cell self-renewal and represses

cryptic intragenic transcription. EMBO J. 30, 1473–1484.

Young, R.A. (2011). Control of the embryonic stem cell state. Cell 144,

940–954.

Zhang, Y., Liu, T., Meyer, C.A., Eeckhoute, J., Johnson, D.S., Bernstein, B.E.,

Nusbaum, C., Myers, R.M., Brown, M., Li, W., and Liu, X.S. (2008). Model-

based analysis of ChIP-seq (MACS). Genome Biol. 9, R137.

Zhou, V.W., Goren, A., and Bernstein, B.E. (2011). Charting histone modifica-

tions and the functional organization ofmammalian genomes. Nat. Rev. Genet.

12, 7–18.

Molecular Cell Article - Harvard Universityds.dfci.harvard.edu/~gcyuan/mypaper/das; 2b2c; mol...

Documents

Transcript of Molecular Cell Article - Harvard Universityds.dfci.harvard.edu/~gcyuan/mypaper/das; 2b2c; mol...