cell reports article reports article global changes in the mammary epigenome are induced by hormonal...
TRANSCRIPT
Cell Reports
Article
Global Changes in the Mammary EpigenomeAre Induced by Hormonal Cuesand Coordinated by Ezh2Bhupinder Pal,1,3 Toula Bouras,1,3,8 Wei Shi,2,5,8 Francois Vaillant,1,3 Julie M. Sheridan,1,3 Naiyang Fu,1,3 Kelsey Breslin,1
Kun Jiang,1 Matthew E. Ritchie,2,3 Matthew Young,2 Geoffrey J. Lindeman,1,4,7,8 Gordon K. Smyth,2,6,8
and Jane E. Visvader1,3,*1ACRF Stem Cells and Cancer Division2Bioinformatics Division
The Walter and Eliza Hall Institute of Medical Research, 1G Royal Parade, Parkville, VIC 3052, Australia3Department of Medical Biology4Department of Medicine5Department of Computing and Information Systems6Department of Mathematics and Statistics
The University of Melbourne, Parkville, VIC 3050, Australia7Department of Medical Oncology, The Royal Melbourne Hospital, Grattan Street, Parkville, VIC 3050, Australia8These authors contributed equally to this work*Correspondence: [email protected]
http://dx.doi.org/10.1016/j.celrep.2012.12.020
SUMMARY
The mammary epithelium is a dynamic, highlyhormone-responsive tissue. To explore chromatinmodifications underlying its lineage specification andhormone responsiveness, we determined genome-wide histonemethylation profiles ofmammary epithe-lial subpopulations in different states. The markeddifferences in H3K27 trimethylation between subpop-ulations in the adult gland suggest that epithelialcell-fate decisions are orchestrated by polycomb-complex-mediated repression.Remarkably, themam-maryepigenomeunderwenthighly specificchanges indifferent hormonal contexts, with a profound changebeing observed in the global H3K27me3 map ofluminal cells during pregnancy. We therefore exam-ined the role of the key H3K27 methyltransferaseEzh2 in mammary physiology. Its expression andphosphorylation coincided with H3K27me3 modifica-tions and peaked during pregnancy, driven in partby progesterone. Targeted deletion of Ezh2 impairedalveologenesis during pregnancy, preventing lacta-tion, and drastically reduced stem/progenitor cellnumbers. Taken together, these findings reveal thatEzh2 couples hormonal stimuli to epigenetic changesthat underpin progenitor activity, lineage specificity,and alveolar expansion in the mammary gland.
INTRODUCTION
The mammary gland, which comprises a branching ductal epi-
thelial network embedded in an adipose-rich stromal matrix,
C
is remarkably adaptive to physiological requirements and
undergoes dramatic morphological changes during puberty
and pregnancy. At birth, it manifests as a rudimentary branched
structure, but ductal elongation and branching commence with
puberty and pregnancy provokes the rapid expansion of alveolar
units that differentiate into milk-secretory cells prior to parturi-
tion. The steroid hormones estrogen and progesterone exert
pivotal roles during mammary development via their cognate
receptors, the estrogen receptor (ER) and progesterone receptor
(PR) (Brisken and O’Malley, 2010). ER is essential for ductal
morphogenesis in puberty (Mallepell et al., 2006; Mueller et al.,
2002), whereas PR governs ductal side-branching and alveolar
development during pregnancy (Brisken et al., 1998; Lydon
et al., 1995; Mulac-Jericevic et al., 2003).
The mammary epithelium can be divided into two primary line-
ages: themyoepithelial lineage constitutes the outer layer of cells
that contact the basement membrane, whereas the luminal
lineage comprises both ductal and alveolar cells. Adult stem
cells prospectively isolated from the mouse mammary gland
display the requisite stem cell properties of multilineage differen-
tiation and self-renewal (Shackleton et al., 2006; Stingl et al.,
2006). A recent study has added a new layer of complexity to
the prevailingmodel of themammary epithelial hierarchy through
the identification of unipotent cells that contribute to homeo-
stasis of the gland (Van Keymeulen et al., 2011). It is not yet clear
whether these correspond to stem or progenitor cells and what
their relationship is to the prospectively isolated epithelial sub-
sets. In the mouse, mammary stem cells (MaSCs) have been
shown to be highly responsive to steroid hormones (Asselin-
Labat et al., 2010; Joshi et al., 2010), while progesterone
augmented the number of human bipotent stem-like cells in
cellular assays (Graham et al., 2009). MaSCs lie at the apex of
the hierarchy and give rise to progenitors and mature cells
through progressive restriction. Two distinct types of luminal
progenitor cells have been isolated from the mouse mammary
ell Reports 3, 411–426, February 21, 2013 ª2013 The Authors 411
CA
B
50%
Score4 8 12
H3K4me3 H3K27me3H3K9me2
100%-5 TSS +5 -5 TSS +5 -5 TSS +5
MaSC/basal subset
H3K4me3 and H3K27me3H3K27me3 onlyH3K4me3 onlyNo H3K4me3/H3K27me3
MaS
C/b
asal
LP ML
1.0
0.8
0.6
0.4
0.2
0
LP vs MaSC/basal
DE genes
TSS Body
K4me3
K27me3
TrilJag2Tbx2Fjx1Wnt10aId4Il17bDlk2Bmp7Igfbp3Vwa2Dll1Elovl4Kirrel3Lama3Dpysl3Ntrk32810032GO3RikGnai1Arhgap24Rtn1ArcCol14a1Ccdc106Wif1Psd2Ism1Nrg1Dkk3Hs3st3a1Lama1Mtap91500009L16RikSema6dLhfpLgals7Apobec1Cacna1gBcat1Sept3Abcg5Fkbp10Sema5aCol23a1Ank2Ugt1a6aArt4TaglnVcanSema3cArsjCtgf1600014C10RikKirrelCd70Antxr1PappaCol17a1Diras2Kcnmb1Sostdc11500015O10RikAardC030048H21RikSh3gl2Moxd1Camk4Tnfrsf11b4732456N10RikCol17a1Krt5Adamts18Irx4PdpnWnt4Upk3aEsrrbCrymCapslLtkScinHomer2Capn6Smarca1Slc38a5Tceal5Ric3MgpUgt1a10Speer4aEgr4Crispld2Sox11MdkAdamts7Zcchc18Hs6st2Drp2Krt15Serpinb11
TSS Body
2
0
-2
1
-1Col
or K
ey
ML vs LP subset
DE genes
K4me3
Tnfsf11Bmp3Slc7a2Defb45Slc16a5Clca3Zbtb8aCapsiRundc3aAdra1aWtipFasIfit2Flrt3MmdVldlrE130203B14RikNtng1Mob3bCcdc129NdnStac2MkxBtbd11Pdlim3EgfrCst3Hey1PkibSnta1Rftn2HexbDock9Crispld2AmtnPtprz1EdnraCcrl2Bcl2St6galnac4Col9a1Gcnt1Nat8Chi3l12610528A11RikAlp1Srgap3Hsd11b1MgllGng7Cyp2d22Il4i1Slc13a2CckSfrp2AbpbCdhr1Cyp24a1Tspan8Cd177Anxa8Egln3Gprc5bPaqr6Cd44Celf2Ncam1Adam23Pi16Zic4Efhd1Mfap2Ramp1Cited2Rb1Cdc42ep3Maml2Slc4a4Nedd9Serpinf1Fam49aEndod1Ephx1Ppap2b1300014l06RikPmp22Chn2Arhgef6Plb1Cdr2Slitrk4PdgfraPaplnPgm5Stard8Fbn1Pisd-ps2Gsg1lH2-OaCol5a1
K27me3
Figure 1. Histone Methylation Profiles of Mammary Epithelial Subpopulations in the ‘‘Steady State’’ and Their Correlation with Gene
Expression Changes
(A) Genome-wide heat map showing the pattern of H3K4me3, H3K9me2, and H3K27me3marks in MaSC-enriched cells from 5 kb upstream to 5 kb downstream
of the TSS of each gene. Rows correspond to genes clustered by coverage pattern. All 26,310 genes in themm9 genome are shown. The first three columns show
(legend continued on next page)
412 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
gland, but committed basal progenitor cells remain elusive (As-
selin-Labat et al., 2007, 2011; Sleeman et al., 2007).
Although a number of transcription factors and pathways have
been implicated in controlling specific steps along the mammary
differentiation hierarchy (reviewed in Visvader, 2009), the role of
the epigenome in regulating cell-fate decisions and differentia-
tion within this epithelial compartment remains unclear. In other
systems, there is substantial evidence that histone methylation
governs lineage-specific developmental programs and that its
deregulation leads to oncogenesis (Bracken and Helin, 2009;
Sauvageau and Sauvageau, 2010). It is thought that histone
modifications establish discrete domains of active and inactive
chromatin to effect gene expression. Histone lysine methylation
can serve as either an active or repressive mark: trithorax-medi-
ated methylation of lysine 4 on histone H3 within nucleosomes is
associated with activated gene expression, while methylation of
lysine 27 by polycomb group (PcG) proteins is linked with gene
repression and chromatin condensation (Margueron and Rein-
berg, 2011). Ezh2, a member of the PcG family, is a histone
methyltransferase that forms the catalytic component of the
polycomb repressive complex PRC2. This complex silences
lineage specification genes to regulate the maintenance and
differentiation of embryonic and adult stem cells (reviewed in
Margueron and Reinberg, 2011). In embryonic stem cells, where
genome-wide histone methylation patterns have been exten-
sively studied, key developmental genes often exhibit both
repressive H3K27me3 marks and activating H3K4me3 marks
(Bernstein et al., 2006). This bivalent modification has been
proposed tomaintain these genes ‘‘poised’’ for subsequent acti-
vation or repression upon lineage specification. In vivo mapping
studies indicate that PcG-dependent H3K27me3 selectively
marks genes in the epidermal lineages and controls gene ex-
pression changes during the differentiation of skin stem cells
(Lien et al., 2011).
In this report, we examine the contribution of epigenetic mech-
anisms to regulation of the lineage hierarchy in the steady-state
mammary gland and in response to different hormonal milieu.
We determined genome-wide histone methylation profiles of
the MaSC-enriched, luminal progenitor, and mature luminal
subsets. Correlating the global H3K4me3 and H3K27me3 modi-
fication maps with gene expression signatures indicated that
the epigenome has an important role in directing cell-fate
changes from the basal to luminal cell lineage. Moreover, the
mammary epigenome was found to be highly sensitive to dif-
ferent hormonal environments. H3K27 trimethylation of chro-
matin emerged as a key mediator of gene expression changes
during pregnancy, concomitant with high levels of Ezh2, appar-
coverage depth on a linear color scale, with the x axis showing distance from the T
In this case, the x axis shows the scale from 4 (nonexpressed) to 12 (maximum
ordered by expression level: the fifth gene group shows little histone marking o
H3K4me3 and increasing H3K27me3 levels. Genes are sorted by expression wit
(B) Segmented bar graphs showing genome-wide percentage of genes with histon
marks in the TSS region are shown, but virtually identical data were obtained for
(C) Heatmaps of gene expression and histone modification changes as cells res
mature luminal (ML) cells. Columns give log2-fold changes for differential gene e
H3K27me3 marking across the broad gene, respectively, for the 200 most differ
See also Figure S1 and Table S1.
C
ently activated by phosphorylation. Targeted deletion of Ezh2 in
the mammary epithelium dramatically reduced both ductal and
alveolar morphogenesis. The expression of Ezh2 and its phos-
phorylation appear to be coordinated through progesterone,
a key pregnancy hormone. Thus, hormonally driven expansion
of the alveolar compartment is programmed, at least in part,
by Ezh2-mediated changes in chromatin modification. Given
the critical importance of both progesterone and EZH2 to breast
cancer, these data implicate progesterone-induced global
changes in chromatin structure in the genesis of this disease.
RESULTS
Histone Methylation Landscapes of Mammary EpithelialSubpopulations in the ‘‘Steady State’’To explore the relevance of histone modification to the regula-
tion of gene expression along the mammary differentiation hier-
archy, chromatin immunoprecipitation sequencing (ChIP-seq)
was performed on distinct epithelial populations that have
been prospectively isolated from the mouse mammary gland.
These correspond to MaSC-enriched (CD29hiCD24+), com-
mitted luminal progenitor (CD29loCD24+CD61+), and mature
luminal cells (CD29loCD24+CD61�) (Asselin-Labat et al., 2007;Shackleton et al., 2006), all readily isolated from FVB/N glands.
The MaSC-enriched population also contains mature myoepi-
thelial cells and likely basal progenitor cells and is referred to
as the MaSC/basal subset from here on. To avoid changes
that are known to occur in epithelial cells during cell culture,
freshly sorted subsets (approximately 250,000 cells) isolated
from young adult females were used for ChIP. No preamplifica-
tion step was incorporated prior to library preparation in order
to avoid potential bias. High-resolution genome-wide maps
were determined for H3K4me3, H3K27me3, and H3K9me2
modifications. Between 17 and 40 million DNA fragments were
sequenced for each ChIP or input sample using 35 bp paired-
end reads (Table S1).
The overall pattern of histone methylation marking was exam-
ined using heatmaps (Figure 1A and Figure S1A) and density
plots (Figure S1B) of fragment coverage. H3K4me3 occupancy
typically peaked sharply around the transcriptional start site
(TSS) of each gene, whereas the repressive H3K9me2 and
H3K27me3 marks were more evenly spread over the promoter
region and gene body, although with weaker peaks and troughs
still discernible around the TSS. In the MaSC/basal subset,
genes could be clustered into five broad groups by the level
of H3K4me3 marking, from very high to very low (Figure 1A).
Apart from the fifth cluster, lacking any of the marks, the other
SS. The far right column shows log2-normalized expression for the same genes.
expression). Genes are clustered into five groups by their histone pattern and
r expression, while the other four groups correspond roughly to decreasing
hin each cluster, so the right panel appears as an increasing curve.
e methylationmarks in each epithelial subset (FDR < 0.05). Percent H3K27me3
H3K27me3 marks across the gene body plus the TSS.
trict from the MaSC/basal to luminal progenitor (LP) cells and from the LP to
xpression (DE), H3K4me3 marking, H3K27me3 marking in the TSS region, and
entially expressed genes.
ell Reports 3, 411–426, February 21, 2013 ª2013 The Authors 413
C
A
ckit
Trp63Krt5Snai2
Elf5 Hey2
Fol
d ch
ange
vs In
put
MaSC
ML
LP
ML
LP
TSS TSS TSS
TSS TSS TSS
Fol
d ch
ange
vs In
put
H3K4me3 H3K27me3
1.8
1.0
0.2
4.0
2.0
1.2
0.8
0.4
1.2
0.8
0.4
1.8
1.2
0.6
6.0
4.0
2.0
MaSC LP ML
1.0
3.0
0
0.6
1.4
0 0
000
MaSC
MaSC LP ML MaSC LP ML
MaSC LP ML MaSC LP ML MaSC LP ML
ML vs LP subsetLP vs MaSC/basal subset
2
1
0
-1
-2
2
1
0
-1
-2-6 -4 0 2 4-2 -4 0 2-2
3
-3
2
1
0
-1
-2
-6 -4 0 2 4-2
3
-3
2
1
0
-1
-2
-4 0 2-2
Log
fold
cha
nge
Expression log fold change
H3K
4me3
H3K
27m
e3
Log
fold
cha
nge
Expression log fold change
H3K
4me3
H3K
27m
e3
B
(legend on next page)
414 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
four clusters showed an inverse pattern of H3K4me3 and
H3K27me3 marking. H3K9me2 occupancy was relatively inde-
pendent of H3K4me3 and H3K27me3 status. The luminal pro-
genitor and mature luminal cell populations exhibited similar
patterns (Figure S1A).
H3K4me3 and H3K27me3 but Not H3K9me2 Correlatewith Overall Gene ExpressionNext, we related the histone methylation patterns to our previ-
ously published microarray gene expression profiles for the
same cell subpopulations (Lim et al., 2010). Expression level is
shown in the right-most panel of each heatmap, with genes
sorted from high to low expression within each cluster (Figures
1A and S1A). While all the clusters contain genes with a range
of expression levels, average expression strongly increased
with H3K4me3 coverage and decreased with H3K27me3
coverage (Figure 1A). This held for all cell subpopulations (Fig-
ure S1A). While H3K27me3 marks were distributed across the
whole promoter region and gene body, gene expression was
especially sensitive to coverage just upstream and downstream
of the TSS, with low expressed genes showing a trough in
H3K27me3 coverage around the TSS relative to the gene body
and highly expressed genes showing a downstream peak (Fig-
ure S1B). Unlike the other two histone marks, H3K9me2 showed
no overall correlation with expression and was equally associ-
ated with both high and low expression.
To assess enrichment for each histone mark statistically,
we recorded the number of fragments mapping within 3 kb
upstream to 2 kb downstream of the TSS of each gene. For
H3K27me3, the number of fragments mapping to a broad region
comprising the TSS and the entire genomic span of the gene
were recorded. For each cell sample and each histone mark,
enrichment was assessed using a statistical model that treats
the input fragment count profile as representative of nonenriched
genes. Most genes were significantly marked by at least one of
the three modifications (Figure 1B). Interestingly, the number of
genes showing enriched H3K27me3 occupancy at the TSS
increased upon luminal lineage specification (21%, 32%, and
34% in the MaSC/basal, luminal progenitor, and mature luminal
subsets, respectively). Moreover, bivalent marking at the TSS
increased upon basal to luminal cell commitment (10% and
22% in the MaSC/basal and luminal progenitor subsets). As
the MaSC subset remains heterogeneous, it is not clear whether
bivalency plays a role in these adult stem cells.
Expression Changes Correlate with HistoneMethylationChanges during Lineage RestrictionGenes with increased expression in the luminal progenitor rela-
tive to the MaSC/basal subset tended to show increased
H3K4me3 and decreased H3K27me3 marking, whereas genes
Figure 2. Histone Methylation Profiles across Key MaSC-Enriched and
(A and B) Scatter plots show that expression changes for the 500 most DE gen
correlated with H3K27me3 (bottom panels) (p < 10�6).
(C) Read coverage graphs for H3K4me3 (red) and H3K27me3 (blue) in each epith
luminal progenitor-specific genes. Y-axes show fragments per million on the sca
samples; error bars show SEM.
See also Figure S1 and Table S1.
C
with decreased expression showed the opposite epigenetic
changes (Figures 1C and 2A). The same was true of genes differ-
entially expressed in the mature luminal versus the luminal
progenitor subset (Figures 1C and 2B). This shows that histone
methylation is dynamically associated with gene expression
during lineage restriction and is likely to be a key mediator of
expression changes that direct basal to luminal cell-fate switch-
ing and luminal maturation. Illustrative read coverage graphs of
H3K4me3 and H3K27me3 patterns across genes characteristic
of the MaSC/basal (snai2, cytokeratin 5, and Trp63) and luminal
progenitor subsets (c-kit, Elf-5, and Hey2) are shown in Fig-
ure 2C. Histone modifications in the TSS regions were verified
for a number of candidates using ChIP combined with quantita-
tive RT-PCR (Figure 2C).
Hormonal Changes Drive Global Mammary EpigeneticAlterationsSince hormone deprivation and pregnancy drastically alter
MaSC numbers and the gene expression profiles of different
epithelial subtypes (Asselin-Labat et al., 2010), we investigated
whether histone methylation is a regulator of these changes.
ChIP-seq profiles were generated of H3K4me3 and H3K27me3
marks in the MaSC/basal and luminal populations from the
glands of ovariectomized or midpregnant (12.5 days) mice as
well as from control virgin mice. The total luminal subset
(CD29loCD24+) was used, given that the CD61+ progenitor pop-
ulation declines precipitously during pregnancy (Asselin-Labat
et al., 2007). Expression changes in ovariectomized mice were
limited to a few dozen genes (15 differentially expressed genes
in the MaSC/basal subset and 82 genes in luminal cells at
a 5% false discovery rate [FDR]) and were only modestly
correlated with epigenetic changes (Figure 3A). Nevertheless,
changes in the expression of specific genes, such as Arf and
cyclin D2, in the MaSC/basal population correlated with
changes in H3K27me3 occupancy at their TSS (Figure S2A).
Cell cycle genes were significantly enriched among differentially
H3K27me3-marked genes in the luminal population upon ovari-
ectomy (p = 0.02), consistent with cell cycle regulators driving
proliferation in response to steroid hormones.
A more global function for H3K27 trimethylation was revealed
in the hormonal milieu of pregnancy that is largely governed by
progesterone and prolactin. Most noticeably, a strong inverse
relationship was apparent between H3K27me3 marking and
expression changes in the luminal cells of pregnant mice (Fig-
ure 3B). Key luminal genes required for differentiation and
concomitant milk production and upregulated in pregnancy
(Elf-5, Wap, and Csn2) showed strongly decreased H3K27me3
modifications that were confirmed by ChIP-quantitative RT-
PCR (qRT-PCR) analysis (Figure 3C). Conversely, luminal genes
expressed in the steady-state gland but downregulated during
Luminal Progenitor Genes
es are directly correlated with H3K4me3 changes (top panels) and inversely
elial subset 10 kb ± of the TSS region of three MaSC/basal-specific and three
le 0–10. Bar plots show ChIP qRT-PCR data for three independent biological
ell Reports 3, 411–426, February 21, 2013 ª2013 The Authors 415
Log
fold
cha
nge
MaSC/ basal Lum
Expression log fold change
OVX vs control pregnant vs virgin
H3K
4me3
H3K
27m
e3
3
10
-2
0 4-6
-4 0 2 4-2 6
-4 0 2 4-2 6
MaSC/ basal Lum
32
0
-2-3
0 5-5
-5
Expression log fold change
1
-1
-4-8 -2 2
2
-1
-3-4-4
3
10
-2
2
-1
-3-4
0 4-6 -4-8 -2 2
32
0
-2-3
0 5-5
1
-1
-4
32
0
-2-3
1
-1
-4
32
0
-2-3
1
-1
-4
32
0
-2-3
1
-1
-4
0 5
0 5-5
32
0
-2-3
1
-1
-4
FoxA1
Wnt7b
Hey1
Fol
d ch
ange
vs
Inpu
t
1.51.0
0.50
3.0
2.5
2.0
1.5
1.0
0.5
0
3.02.5
2.0
0.4
0
1.8
1.2
0.8
virgin
12.5 dP
virgin
12.5 dP
Wap
Elf5
Csn2
Fol
d ch
ange
vs
Inpu
t F
old
chan
ge v
s In
put
Fol
d ch
ange
vs
Inpu
t
virgin 12.5 dP
virgin 12.5 dP
virgin 12.5 dP
virgin
12.5 dP
1.2
0.6
1.2
0.4
0.8
1.2
0.4
0.8
1.6
0
virgin
12.5
0
0
BA
DC
virgin
12.5
virgin
12.5
virgin
12.5
virgin
12.5
virgin
12.5
Fol
d ch
ange
vs
Inpu
tF
old
chan
ge v
s In
put
virgin 12.5 dP
virgin 12.5 dP
virgin 12.5 dP
Figure 3. The Mammary Epithelial Epigenome Is Influenced by Hormonal Status
(A and B) Scatter plots of expression versus epigenetic log2-fold changes in the mammary epithelial subsets of (A) ovariectomized mice and (B) pregnant mice.
Increased H3K27me3 marking strongly mediates decreased expression in luminal cells from 12.5 day pregnant glands (p < 10�6). Other correlations were also
significant (p < 0.05), except H3K27me3 modifications in the MaSC subset at midpregnancy.
(C) Derepression of milk genes and Elf-5 in the luminal subset of pregnant glands at 12.5 days. The left panel shows read coverage of H3K27me3 marks around
the TSS of each gene. The right panel shows ChIP-qRT-PCR confirmation (n = 3; error bars show SEM).
(D) Repression of luminal commitment genes (Wnt7b, Foxa1, and Hey1) in the luminal subset during pregnancy correlates with increased H3K27me3 marks. The
left panel shows read coverage for H3K4me3 (red) and H3K27me3 (blue) around the TSS of each gene. The right panel shows ChIP-qRT-PCR confirmation (n = 3;
error bars show SEM).
See also Figure S2, and Tables S1 and S2.
416 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
pregnancy, such as wnt7b, and the luminal commitment genes
Foxa1 (Bernardo et al., 2010) and Hey1 (Bouras et al., 2008)
had abundant H3K27me3 modifications at 12.5 days of
pregnancy, as validated by ChIP-qRT-PCR (Figure 3D). These
changes likely reflect a shift in gene expression within the
emerging alveolar cells toward the highly specialized function
of milk production. In general, genes repressed in pregnancy
with a concomitant increase in H3K27me3 marking were en-
riched for the mammary morphogenesis and developmental
gene categories. Genes upregulated during pregnancy and
with reduced H3K27me3 modifications showed enrichment for
lipid biosynthesis and lipid catabolism (Table S2), commensu-
rate with the changed mammary function during pregnancy.
Although H3K27me3 was not consistently correlated with ex-
pression changes in the MaSC/basal subset of pregnant mice, it
nevertheless appeared to play an important role for specific
genes. In particular, it was associated with derepression of a
number of genes that are normally expressed only in the luminal
lineage. This observation is of particular interest given the
dramatic expansion of theMaSC pool and its altered gene signa-
ture during pregnancy (Asselin-Labat et al., 2010). Quantitative
RT-PCR confirmed expression of the milk protein genes Wap
and Csn2 and the luminal progenitor transcription factor Elf-5,
all of which are normally restricted to luminal subpopulations
(Lim et al., 2010; Figure S2B). Compatible with their derepres-
sion, each of these genes showed diminished H3K27me3 marks
at their TSS during pregnancy, as confirmed by ChIP-qRT-PCR
analysis (Figure S2C). Thus, lineage-priming may occur in the
expanded stem cell population during pregnancy prior to com-
mitment along the alveolar lineage. Intriguingly, expression of
the basal-specific gene Lgr5 was extinguished in the MaSC
pool during pregnancy, accompanied by augmented H3K27 tri-
methylation (Figures S2B and S2C).
At a more global level, the total number of genes within the
luminal subset with significant (FDR< 0.05) H3K27me3modifica-
tions relative to input increased in pregnancy but decreased in
ovariectomized mice significantly (Figure S2D). In summary,
striking epigenetic changes occurred within a specific cellular
subset during pregnancy and were selectively observed for
H3K27me3 but not H3K4me3 or H3K9Ac modifications (data
not shown), which showed small changes.
Dynamic Expression of the Polycomb Group RepressorEzh2 in the Mammary GlandIn view of the marked hormone-induced changes in the global
H3K27me3 profile during pregnancy, we examined the ex-
pression of Ezh2 during mammary ontogeny. Ezh2 is the core
enzymatic subunit of PRC2 that catalyzes K27 trimethylation
on H3 (Margueron and Reinberg, 2011) and has emerged as
an important prognostic marker in breast cancer. Western blot
analysis showed that Ezh2 expression was low in virgin glands
and peaked during early to midpregnancy before declining in
late pregnancy (Figure 4A). Interestingly, the profile of total
H3K27me3-modified protein closely mirrored that of Ezh2 (Fig-
ure 4A). Immunohistochemical staining confirmed the ex-
pression of Ezh2 during mammary morphogenesis and further
revealed that it was abundant in the terminal end buds (TEBs)
of the developing pubertal gland, with lower levels visible in the
C
nuclei of myoepithelial and luminal cells of mature ducts (Fig-
ure 4B). Ezh2 staining was most intense in the ducts and alveoli
during pregnancy (Figure 4B) and declined to low levels in
lactating and involuting glands (Figure 4B).
Ezh2 Deficiency Delays Mammary Morphogenesisduring PubertyTo investigate the physiological role of Ezh2 in the mammary
gland, we conditionally targeted the Ezh2 locus using cre recom-
binase driven by the mouse mammary tumor virus (MMTV)
promoter. Fluorescence-activated cell sorting (FACS) analysis
of reporter mice demonstrated that the MMTV promoter is active
in both luminal and basal mammary epithelial cells (Figure S3A)
but not in the stroma (<0.05% cells). Immunohistochemistry
and western blot analysis confirmed efficient deletion of Ezh2
in MMTV-cre;Ezh2f/f glands (Figures 4A and 4C). Its deletion
markedly impaired elongation and branching of the mammary
epithelial tree during puberty (n = 6), resulting inmarkedly smaller
ductal trees in young adult mice (n = 14) relative to control age-
matched glands (n > 40) from either MMTV-cre;Ezh2f/+, Ezh2f/f,
or Ezh2f/+ littermates or MMTV-cre mice (Figures 4D and S3B).
The same phenotype was observed on a mixed C57Bl6/FVB/N
or pure FVB/N background. MMTV-cre transgenic females
behaved like wild-typemice and could support large-sized litters
through normal lactation, thus differing from the cre strains
recently described (Robinson and Hennighausen, 2011). The
delay in morphogenesis was also associated with decreased
filling of the fat pad in Ezh2-deficient mice (Figure 4E) and the
persistence of TEBs (Figure S3B). By 12 weeks of age, ductal
morphogenesis of most Ezh2-deficient mammary glands was
partially rescued, presumably through recurrent hormonal stim-
ulation (Figure S3B). Cell fate appeared unaffected by loss of
Ezh2, as assessed by immunostaining for luminal, myoepithelial,
and alveolar markers (K18, K14, p63, and Npt2b) (Figures S3C
and S3D; data not shown).
Ezh2Controls the Activity ofMammary Stem/ProgenitorCellsA potential role for Ezh2 in regulating MaSC function was next
addressed using the mammary fat pad reconstitution assay.
Limiting dilution assays of freshly sorted CD29hiCD24+ cells re-
vealed a 14-fold decrease in the frequency of mammary repopu-
lating cells in MMTV-cre;Ezh2f/f compared to control mammary
glands (Table 1). This observation indicates that Ezh2 has amajor
role in either stem or descendant progenitor cells. Further delin-
eation of the hierarchy will be required to distinguish between
these possibilities.
The effect of Ezh2 loss on epithelial proliferation was next
assessed by in vivo bromodeoxyuridine (BrdU) labeling. Prolifer-
ative activity was significantly impaired in the TEBs of MMTV-
cre;Ezh2f/f mammary glands compared to MMTV-cre;Ezh2f/+
and Ezh2f/+ control glands (Figures 4F and S3E). Furthermore,
both the MaSC/basal and luminal populations from Ezh2-defi-
cient glands had little clonogenic activity in vitro on fibroblast
feeders, with only small colonies visible (Figure 4G). These
data point to a critical role for Ezh2 in regulating the activity
of multiple progenitor cell types in the mammary gland. It
seems likely that the developmental defect in MMTV-cre;Ezh2f/f
ell Reports 3, 411–426, February 21, 2013 ª2013 The Authors 417
A
CB
4 w
eeks
8 w
eeks
6.5
days
12
.5 d
ays
16.5
day
s
pregnancy
H3K27me3
Tubulin
4 w
eeks
8 w
eeks
6.5
days
12
.5 d
ays
16.5
day
s
Tubulin
Ezh2
D
F
Ezh2
f/+
cKO
18.5
day
s
virgin pregnancy
18.5
day
s 2
dL
virgin
6 weeks virgin 6.5 dP
12.5 dP 1 dL 4 dI
8 weeks virgin
MM
TV
-cre
; Ezh
2f/f
Ezh2
Ezh
2f/+
CD29hi CD29lo
MM
TV
-cre
; E
zh2f/f
Ezh
2f/+
Col
onie
s pe
r 10
0 ce
lls
CD29lo CD29hi
10
0
35
30
20
25
15
5
* *
Ezh2f/+
MMTV-cre; Ezh2f/f
MMTV-cre; Ezh2f/fMMTV-cre; Ezh2f/+Ezh2f/+
G
MMTV-cre; Ezh2f/fEzh2f/+ MMTV-cre; Ezh2f/+
E
Ezh2f/+
MMTV-cre; Ezh2f/f
Fat
pad
fi lli
ng (
%)
6 7 8
weeks
0
100
60
80
40
20
(legend on next page)
418 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
Table 1. Limiting Dilution Analysis of theMammary Repopulating
Frequency of CD29hiCD24+ Cells Isolated from YoungMMTV-cre;
Ezh2f/f or Control Glands
Number of CD29hiCD24+ Cells
Injected per Mammary Fat Pad
Number of Positive Outgrowthsa
Ezh2f/+ MMTV-cre; Ezh2f/f
100 5/8
200 2/5 0/6
400 11/16 0/10
800 4/7 1/17
1,600 5/6 1/5
3,200 9/12
6,000 2/7
Repopulating frequency 1/460 1/6,419
(95% confidence interval): (1/722–1/293) (1/11,188–1/3,683)
p value <0.00001
CD29hiCD24+ cells frommammary glands of 8- to 9-week-old mice were
injected into the cleared mammary fat pads of 3-week-old nonobese dia-
betic-severe combined immunodeficient female recipients. Data are
pooled from three independent experiments collected 8 weeks post-
transplantation. The repopulation frequency was calculated using limiting
dilution analysis as described (Hu and Smyth, 2009).aShown as number of outgrowths per number of injected cleared
mammary fat pads.
mammary glands manifests in puberty, because this stage
requires large numbers of progenitor cells to orchestrate ductal
growth (Asselin-Labat et al., 2007).
Loss of Ezh2 Profoundly Affects the Expression of CellCycle and Epidermal GenesTo identify potential downstream effectors of Ezh2, the genome-
wide transcriptional profiles of the MaSC/basal and luminal pop-
ulations in their steady state were determined following ex vivo
cre-mediated excision of Ezh2. Similar to freshly sorted cells
(Figure 4G), the clonogenic capacity of these Ezh2-deficient
MaSC/basal and luminal populations was dramatically reduced
compared to control cultures (Figure S4A). Gene ontology anal-
ysis of the top 500 differentially expressed genes revealed a sig-
nificant association with cell cycle, DNA replication, and DNA
Figure 4. Ezh2 Is Required for Normal Mammary Gland Development
(A) Western blot analysis of mammary gland lysates for expression of Ezh2 and H3
provided the controls.
(B) Immunohistochemical staining of mammary gland tissue sections from FVB
(6 weeks), mature ducts in 8-week-old mice; alveoli in early pregnancy (6.5 d
bars, 50 mm.
(C) Immunohistochemical staining of TEBs for Ezh2 expression in MMTV–cre; Ez
(D) Whole-mounts of mammary glands from 7-week-old virgin MMTV–cre; Ezh2f/f
lymph node in the inguinal gland is marked by a white arrow. Scale bars, 2.0 mm
(E) Extent of fat pad filling was estimated in virgin mice at 6, 7, and 8 weeks of a
(F) Immunohistochemical staining of terminal end buds for BrdU incorporation, s
glands compared to those from Ezh2f/+ and MMTV–cre; Ezh2f/+ mice. Scale bar
(G) Colony-forming capacity of sorted MaSC-enriched (CD29hiCD24+; labeled C
fibroblast feeders from 8-week-old MMTV–cre; Ezh2f/f mice compared to Ezh2f
CD61+ luminal progenitors that could be isolated from the smaller targeted gland
quantitation of the colony forming capacity of the CD29loCD24+ and CD29hiCD2
represent mean ± SD of three independent experiments, with eight replicates fo
See also Figure S3.
C
repair (Figure S4B). Of the top ranked genes, three potent cell
cycle inhibitors were derepressed: Cdkn1c (p57), Cdkn2a
(Ink4a/Arf), and Cdkn1a (p21). Consistent with Arf being an
important target of Ezh2, Arf transcript levels were considerably
lower in the luminal population from pregnant glands than in
other subsets and Arfwas derepressed in Ezh2-deficient luminal
cells at midpregnancy (Figure S4C). Gene expression changes in
Ezh2-deficient cells were inversely correlated with changes in
pregnancy (Asselin-Labat et al., 2010), with 67% (52 of 78) of
differentially expressed genes in the MaSC/basal subset and
91% (146 of 160) in the luminal subset showing changes in oppo-
site directions.
Interestingly, one of the top derepressed gene sets in the
MaSC/basal population, other than those related to cell cycle
regulation, was keratinocyte differentiation (Figure S4B). The
expression of genes within the epidermal differentiation complex
on mouse chromosome 3, previously shown to be a target of
Ezh2 repression in skin (Ezhkova et al., 2009), was activated in
the MaSC-enriched subset from Ezh2-deficient glands. This is
consistent with previous reports of misexpression of nonlineage
genes associated with Ezh2 deletion (see Discussion). To inves-
tigate a potential relationship between the gene expression sig-
natures of Ezh2-deficient cells and metaplastic breast cancers,
the molecular profiles of the different breast cancer subtypes
were interrogated with the Ezh2-deficient MaSC/basal cell
signature. Intriguingly, this signature was found to bemost highly
represented in the claudin-low subgroup based on signature ex-
pression scores (Figure S4D). The claudin-low subtype exhibits
metaplastic features, expresses lower levels of Ezh2 than the
other subtypes, and can even display epidermal traits (Keller
et al., 2012), suggesting that they have undergone metaplasia
as a result of aberrant differentiation.
Ezh2 Deficiency Leads to Reduced AlveolarDevelopment and Failure of LactationWe next examined the effect of Ezh2 deficiency on pregnancy
and lactation. Although alveoli formed in MMTV-cre;Ezh2f/f
mammary glands during pregnancy (Figure 5A), they were fewer
and more disorganized than those in Ezh2f/+, Ezh2f/f, or MMTV-
cre mammary glands (Figure 5B; data not shown). This pheno-
type appeared most obvious from midpregnancy (n = 9), where
K27me3 protein during development. Ezh2 cKO tissue (12.5 dP) and antitubulin
/N females for Ezh2 expression, representing TEBs that characterize puberty
P) and midpregnancy (12.5 dP); lactation (1 dL); and involution (4 dI). Scale
h2f/f mice compared to littermate Ezh2f/+ glands. Scale bars, 50 mm.
mice compared to glands from Ezh2f/+ and MMTV-cre; Ezh2f/+ littermates. The
.
ge using ImageJ software, with four to nine mice for each time point.
howing reduced numbers of proliferating cells in MMTV–cre; Ezh2f/f mammary
s, 50 mm. Isotype control antibody panels are shown in insets.
D29hi) and luminal cells (CD29loCD24+; labeled CD29lo) grown on irradiated/+ littermates. The same results were obtained for 6-week-old mice. The few
s also had reduced clonogenic activity (data not shown). Histogram showing
4+ subpopulations (200 and 300 cells were plated per well, respectively). Data
r each. *t test p < 0.0001 compared to Ezh2f/+ controls for both subsets.
ell Reports 3, 411–426, February 21, 2013 ª2013 The Authors 419
A
B
16.5
dP
12.5
dP
MMTV-cre; Ezh2f/f
16.5
dP
Ezh2f/+
E
MMTV-cre; Ezh2f/fMMTV-cre
C
D
Tubulin
Ezh2
H3K27me3
MMTV-cre;Ezh2f/f Ezh2f/+
Cdk
n1c
Fol
d ch
ange
vs
Inpu
t
0
1.5
1.0
0.5
2.0
MMTV-cre; Ezh2f/f
Ezh2f/+
Fol
d ch
ange
vs
Inpu
tC
amk2
n1
0
3.5
2.5
1.5
0.5
MMTV-cre; Ezh2f/f
Ezh2f/+ Fol
d ch
ange
vs
Inpu
tW
nt7b
0
2.5
1.5
1.0
0.5
2.0
MMTV-cre; Ezh2f/f
Ezh2f/+
Cdk
n2a
Fol
d ch
ange
vs
Inpu
t
MMTV-cre; Ezh2f/f
Ezh2f/+0
4
3
2
1
Ezh2f/+
12.5
dP
MMTV-cre; Ezh2f/fMMTV-cre; Ezh2f/+ Ezh2f/+
Figure 5. Ezh2 Deficiency Leads to Abnormal Alveolar Development
(A) Whole-mounts of mammary glands from MMTV–cre; Ezh2f/f mice (right panel) compared to those from Ezh2f/+ and MMTV–cre; Ezh2f/+ mice at day 12.5 of
pregnancy show retardation of ductal growth. Mice were mated at 7 weeks of age. No gross abnormalities in the alveolar units were evident in early pregnancy
(6.5 dP; data not shown). Scale bars, 4.0 mm.
(B) H&E sections of mammary glands from MMTV–cre;Ezh2f/f, Ezh2f/+ littermates, and MMTV-cre mice at days 12.5 and 16.5 of pregnancy. Scale bars, 50 mm.
(C) Western blot analysis of mammary glands from MMTV-cre; Ezh2f/f and Ezh2f/+ littermate control mice for expression levels of Ezh2, H3K27me3 protein, and
tubulin.
(D) Derepression of cell cycle genes in pregnant glands lacking Ezh2. ChIP-qRT-PCR for H3K27me3 marks across Cdkn2a/Arf, Cdkn1c (p57), Wnt7b, and
Camk2n1 in the expanding luminal population from day 12.5 pregnant mice. Histograms show the mean of two independent samples with at least two technical
replicates for each.
(E) Immunostaining for milk protein of MMTV-cre; Ezh2f/f tissue sections (right panel) compared to an Ezh2f/+ littermate control at 16.5 days of pregnancy (left
panel). Scale bars, 100 mm.
See also Figures S3 and S4.
ductal elongation of the mammary tree remained stunted in 50%
of Ezh2-deficient glands (Figure 5A). Heterozygotes seemed to
have an intermediate phenotype, with less dense but apparently
normal alveoli (Figures 5A and S3F). Notably, progenitor cell
activity was severely compromised in all three epithelial subsets
isolated frommidpregnant Ezh2-deficient mammary glands (Fig-
ure S5). Western blot analysis confirmed loss of Ezh2 in these
glands and showed a pronounced decrease in H3K27me3
protein in Ezh2-deficient glands in either the pregnant (Figure 5C)
or virgin state (Figure S5D). The residual level of H3K27me3
protein in targeted glands may reflect low levels of Ezh2 de-
tectable in MMTV-cre;Ezh2f/f mammary glands but also sug-
gests that other methylases contribute to H3K27 trimethylation.
Importantly, ChIP combined with qRT-PCR confirmed that the
420 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
cell cycle inhibitors Arf and p57 (Cdkn1c) were derepressed in
primary Ezh2-deficient epithelial cells at midpregnancy (Fig-
ure 5D), as well as other genes abundantly marked by
H3K27me3 at this time-point, including Wnt7b and Camk2n1.
Overall, these data indicate that Ezh2 plays an important role
in the methylation of H3K27 in mammary epithelial cells.
Milk production was readily detectable in MMTV-cre;Ezh2f/f
glands at 16.5 days of pregnancy, suggesting that Ezh2 does
not affect alveolar differentiation (Figure 5E). However, lactation
was severely compromised (n = 4), resulting in all pups dying
within 2 or 3 days of birth. Concordantly, Ezh2-deficient mam-
mary glands showed grossly abnormal morphology at day two
of lactation, relative to control glands that underwent normal
lactation (Figure S3G). This phenotype is likely to result from
a decrease in alveologenesis rather than differentiation, reflect-
ing the critical role of Ezh2 in regulating progenitor cell activity.
Progesterone-Mediated Regulation of Ezh2 ExpressionWe explored whether posttranscriptional control mechanisms
contributed to augmented Ezh2 expression during pregnancy,
since only a modest increase in Ezh2 transcript levels was evi-
dent in epithelial subsets from midpregnant glands (Asselin-
Labat et al., 2010). Ezh2 has been shown to be phosphorylated
on multiple residues, including serine 21 and threonine residues
345 and 487 (reviewed in Caretti et al., 2011). Western blot
analysis using phosphospecific antibodies revealed a striking
increase in the level of phosphorylated Ezh2 on Thr487 in early
tomidpregnancy (Figure 6A). Little or no change in the phosphor-
ylation of Ser21 or Thr345 was detected (data not shown). Inter-
estingly, expression of the cell cycle-dependent kinases Cdk1
and Cdk2, which have been demonstrated to phosphorylate
Ezh2 on threonine residues, mimicked that of Ezh2 and phos-
pho-Ezh2 (Figure 6A). Hence, it is plausible that phosphorylation
of Thr487, perhaps byCdk1 or Cdk2, activates or stabilizes Ezh2.
To investigate a more direct role for pregnancy hormones in
regulating Ezh2 and its phosphorylation, we treated mice in vivo
with progesterone or the peptide hormone prolactin that con-
tribute to the formation and differentiation of alveoli during
pregnancy (Oakes et al., 2008). It is noteworthy that PR levels
follow a similar pattern to that of Ezh2 and phospho-Ezh2, in
part reflecting epithelial content (Figure 6B). Analysis of mam-
mary glands following hormonal treatment showed that phos-
phorylated Ezh2 (Thr487) increased with progesterone but not
prolactin (Figure 6C). At 72 hr after progesterone treatment,
phospho-Ezh2 and Cdk1 were both substantially elevated (Fig-
ure 6D). Moreover, knockdown of PR in T-47D breast cancer
cells was accompanied by a decrease in Ezh2 and phospho-
Ezh2 levels (Figure 6E), suggesting that both transcriptional
and posttranslational mechanisms contribute.
To further examine the effects of progesterone on Ezh2 ex-
pression in different cellular compartments, we isolated discrete
subtypes of luminal progenitor cells that differ in their hormone
receptor status through the use of anti-CD49b and Sca-1 anti-
bodies (Li et al., 2009; Figure S6). Quantitative RT-PCR analysis
showed that the hormone receptor (HR)+ and HR� progenitor
populations contained cells highly enriched for PR+ and PR�
cells (Figure 6F). Interestingly, progesterone strongly induced
Ezh2 in the PR� subset, indicating a paracrine mode of stimula-
tion (Figure 6G). Conversely, Ezh1 expression was not induced
by progesterone (data not shown). As Rankl is a known target
of PR, we evaluated the expression of Rankl and the Rank
receptor in the three luminal subsets. Rankl was most highly
induced by progesterone in mature PR-positive ductal cells
(CD24+CD29loCD49b�Sca-1+), whereas Rank was abundantly
expressed in PR� luminal progenitor cells, suggesting that the
Rankl/Rank axis is an important paracrine mediator of proges-
terone-induced Ezh2 expression.
DISCUSSION
The epigenome is presumed to play a critical role in adult stem
cells and their progressive commitment to differentiated cells.
C
Indeed, elucidation of the genome-wide histone methylation
profiles of distinct mammary epithelial subtypes revealed that
H3K27me3-mediated epigenetic silencing is a key determinant
of gene expression during lineage restriction in the steady-state
mammary gland. Moreover, we found that pregnancy hormones
triggered striking genome-wide changes in H3K27me3 chro-
matin modifications in the expanding alveolar cell population.
Our elucidation of the physiological role of Ezh2 in controlling
mammary progenitor activity and alveolar development provides
direct evidence that this histone methylase has a critical role in
coordinating changes in the epigenome with regulatory gene
networks in response to hormonal stimuli.
The H3K4 and H3K27 trimethylated ‘‘landscapes’’ of the
mammary epithelium are highly dynamic in the steady-state
gland. Chromatin analyses combined with expression profiling
showed that these histone modifications tightly correlated with
transcriptional activity. In particular, H3K27me3 modifications
increased profoundly upon restriction of the MaSC/basal popu-
lation to the luminal lineage, suggesting that it contributes to
gene expression changes during lineage restriction. Histone
and DNA methylation patterns have been addressed in human
breast epithelial cells (CD44+ and CD24+), but these are not
directly comparable to the three cellular subsets defined in the
mouse mammary gland (Maruyama et al., 2011). Pertinently,
in another ectodermal lineage, the transition from activated
skin stem cells to committed transit amplifying cells involves
K27me3-PcG-mediated repression of stem cell genes and dere-
pression of PcG-silenced regulators (Lien et al., 2011). Given that
the MaSC/basal subset is heterogeneous at the cellular level,
although remarkably similar at the gene expression level (Stingl
et al., 2006; unpublished data), it remains unclear whether the
observed epigenetic changes in this population reflect those
occurring in MaSCs. The proportion of genes exhibiting bivalent
marks at their TSS also increased with restriction along the hier-
archy, with lowest levels in theMaSC/basal cell subset. It is inter-
esting to note that hair follicle stem cells contain few bivalent
marks (Lien et al., 2011).
Steroid hormones profoundly influence the mammary epige-
nome. Modulation of hormone levels via ovariectomy, preg-
nancy, or treatment with antiestrogen inhibitors strongly affects
both MaSCs and the luminal cell compartment (Asselin-Labat
et al., 2010). Hormone deprivation only modestly changed
H3K4me3 and H3K27me3 marking in both the stem cell-en-
riched and luminal populations, implying that other epigenetic
mechanisms may contribute to changes in gene expression.
However, the changes apparent in cell cycle regulators may
suffice to drive MaSCs into a more quiescent state, as indicated
by molecular profiling and cellular analyses (Asselin-Labat et al.,
2010). On the other hand, a striking role for H3K27me3
modifications was revealed during pregnancy, specifically in
the expanding alveolar luminal subset. In that subset, there
was a dramatic increase in the number of genes marked by
H3K27me3. Moreover, Ezh2 and H3K27-trimethylated protein
levels mirrored each other during mammary gland develop-
ment, with highest levels in early to midpregnancy, returning
to low levels in the late stages of pregnancy. In parallel, a
decrease in H3K27me3 marks has been observed during the
differentiation of keratinocytes (Sen et al., 2008), neuronal cells
ell Reports 3, 411–426, February 21, 2013 ª2013 The Authors 421
4 w
eeks
8 w
eeks
6.5
dP12
.5 d
P16
.5 d
P
P-Ezh2T487
Cdk1
Cdk2
Tubulin
Ezh2
cKO 4 w
eeks
8 w
eeks
6.5
dP12
.5 d
P16
.5 d
P18
.5 d
P
PR
Tubulin
Ezh2
Ezh2
P-Ezh2T487
Tubulin
Pg (72 h)vehicle
D
Prlvehicle
Cdk1
P-Ezh2T487
Tubulin
P-Ezh2T487
Tubulin
Pgvehicle
CBA
E
Ezh2
siR
NA
PR s
iRN
A
RIS
C-fr
ee
Ezh2
Tubulin
PR
P-Ezh2T487
H
luminal cell expansion
breast cancer
gene repression
progesterone
Ezh2
P-Ezh2 (T487)
&
CDK1/2
PRC2
steady state
Ezh2
K27
me3 me m3 e3me m3 e3
GOil Pg
F
Ezh
2 (
rela
tive
to 1
8S r
RN
A)
0
3.5
0.5
2.5
1.5
4.5X10-2
ML LP(PR+)
LP(PR–)
Ran
k (
rela
tive
to 1
8S r
RN
A)
0
1.2X10-5
0.8
0.4
0.2
0.6
1.0
ML LP(PR+)
LP(PR–)
Ran
kL (
rela
tive
to 1
8S r
RN
A)
0
4
2
1
5
6
3
7
ML LP(PR+)
LP(PR–)
PR
(re
lativ
e to
18S
rR
NA
)
0
2.5X10-2
2.0
1.0
0.5
1.5
ML LP(HR+)
LP(HR–)
Ezh2
Ezh2
K27 K27 K27 K27P P
Figure 6. Pregnancy Hormones Induce Phosphorylation of Ezh2 and Expression of Cdk-1/2, with Progesterone as an Important Mediator
(A) Western blot analysis of whole gland cell lysates for total Ezh2, phospho-Ezh2(Thr 487), Cdk1, and Cdk2 during mammary gland development. dP, days
pregnancy; cKO corresponds to MMTV-cre;Ezh2f/f at 12.5 dP. Antitubulin provided the loading control.
(legend continued on next page)
422 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
(Mikkelsen et al., 2007), and skeletal muscle (Caretti et al.,
2004).
Our findings indicate that upregulation of Ezh2 in pregnancy is
a key event that coordinates changes in the chromatin land-
scape with modulation of gene expression. The critical role of
Ezh2 in governing the activity of progenitor cells appears to
underlie its requirement for driving expansion of the alveolar
cell compartment. Notably, Ezh2 deficiency in the mammary
gland led to a global decrease in H3K27me3 levels in pregnancy,
indicating that it is a key methyl-transferase for H3K27. Ezh1
and/or demethylases may also contribute to the overall level of
H3K27me3 protein in the mammary epithelia. It is noteworthy
that targeted deletion of Eed, which is required for the function
of both Ezh1 and Ezh2, yielded a similar phenotype to that
following Ezh2 ablation (S. Orkin, K.B., and J.E.V., unpublished
data).
Ezh2 is also essential for the activity of basal and luminal
progenitor cells in the mammary gland during puberty. In Ezh2-
deficient mice, mammary morphogenesis was severely compro-
mised but could be effectively rescued through recurrent estrus
cycling by 15 weeks of age. Notably, mammary repopulating
activity was reduced 14-fold upon loss of Ezh2. Whether this
reflects a defect in the stem cell itself or in bipotent or committed
progenitor cells is uncertain. Pertinently, Ezh2-deficient progen-
itor cells in the MaSC/basal and luminal subsets had almost no
clonogenic activity in vitro. The cell cycle GO groups were
substantially affected by Ezh2 loss, consistent with Ezh2 func-
tioning as a silencer of cell cycle inhibitors in progenitor cells
and thus preventing exit from the cell cycle. Indeed, targeted
disruption of Ezh2 has demonstrated that it is required for the
proliferation of myogenic and skin basal progenitors, in part
due to repression of the Ink4a/Arf locus in the skin (Ezhkova
et al., 2009) and Ink4a in muscle stem cells (Juan et al., 2011).
The Ink4a/Arf locus is also likely to be a key mediator of Ezh2-
dependent cell cycle regulation in mammary progenitor cells,
given the reciprocal expression of Ezh2 and Arf. As a regulator
of proliferation, Ezh2 also plays an essential role in lymphopoie-
sis (Su et al., 2003), adipogenesis (Wang et al., 2010), postnatal
cardiac homeostasis (Delgado-Olguın et al., 2012), and hair
follicle morphogenesis in adult skin (Ezhkova et al., 2011). In
the latter case, however, targeted deletion of both Ezh1 and
Ezh2 was required to perturb hair follicle homeostasis.
Cell-fate decisions and differentiation in the mammary epithe-
lium do not appear to require Ezh2. Its absence did not affect
(B) Western blot analysis of PR and Ezh2 expression in whole mammary gland ly
(C) Western blot analysis of phospho-Ezh2(Thr 487) expression in whole gland lys
for 16 hr. No induction by prolactin was observed 24 hr posttreatment (data not
(D) Western blot analysis of total Ezh2, phospho-Ezh2(Thr 487), and Cdk1 express
(E) Knockdown of PR in human T-47D cells using small interfering RNAs (siRNAs)
Western blot analysis of PR, Ezh2, phospho-Ezh2(T487), and tubulin expression
(F) Quantitative RT-PCR analysis of PR expression in steady-state populations
Sca-1+) and hormone receptor-negative luminal progenitor (CD24+CD29loCD49
CD29, CD24, Sca1, and CD49b, as in Figure S6. n = 3 independent experiments
(G) Quantitative RT-PCR analysis of Ezh2, Rank, and Rankl expression in mature l
vehicle for 48 hr (n = 3; error bars show SEM). Note the induction of Ezh2 in the
(H) Schematic model of the role of progesterone in coordinating changes in the e
during pregnancy. Continual exposure to progesterone is hypothesized to result
See also Figures S5 and S6.
C
expression of mammary lineage markers, including alveolar
markers, and those required for milk synthesis. Although Ezh2-
deficient glands could produce milk, lactation failed due to the
low density of alveolar units. Furthermore, Ezh2 levels decline
precipitously in late pregnancy and lactation when terminal
differentiation occurs. Indeed, downregulation of Ezh2 prior to
lactation (when progesterone levels fall) may serve as a switch
to mediate derepression of genes required for lactogenesis (Ru-
dolph et al., 2003), including lipid biosynthesis genes, as sug-
gested by our bioinformatic analysis. The role of Ezh2 is thus
distinct from that of the polycomb family member Bmi1, which
represses alveolar differentiation in the adult mammary gland
(Pietersen et al., 2008a). However, polycomb-mediated repres-
sion of broad lineage determination programs can prevent the
activation of ‘‘extraneous’’ differentiation networks. Expression
profiling of Ezh2-deficient epithelial cells revealed inappropriate
expression of genes within the epidermal differentiation complex
(EDC), which is normally expressed in keratinocytes. The EDC
comprises six late-differentiation genes, including involucrin.
Interestingly, increased numbers of involucrin-positive luminal
cells appeared in Ezh2-deficient glands compared to control
glands (data not shown). Thus, Ezh2 normally represses expres-
sion of EDC genes in the mammary epithelium. In parallel, non-
skin lineage genes are also silenced in hair stem cells by
H3K27me3 (Ezhkova et al., 2011; Lien et al., 2011), and unsched-
uled expression of nonmyogenic lineage genes has been ob-
served in Ezh2-null skeletal muscle (Juan et al., 2011).
Ezh2 phosphorylation may be an important mechanism
by which hormones influence mammary epithelial expansion
during pregnancy. Phosphorylation of Ezh2 on threonine resi-
due 487 and levels of the cell cycle kinases Cdk1 and Cdk2
increased markedly during pregnancy, mimicking the pattern
for H3K27me3 protein. Both Cdk1 and Cdk2 can phosphorylate
Ezh2, and Cdk1 is positively associated with cell proliferation
(Malumbres and Barbacid, 2009). Furthermore, Cdk-mediated
phosphorylation of Ezh2 during cell cycle progression has
emerged as an important mechanism regulating Ezh2 function.
Thr345-phosphorylated Ezh2 appears to orchestrate epigenetic
silencing by promoting the recruitment of PRC2 to Ezh2 targets
(Chen et al., 2010; Kaneko et al., 2010). However, the role of Ezh2
phosphorylation on Thr487 remains unclear, since Thr487 phos-
phorylation impaired Ezh2 methyltransferase activity in one
study (Wei et al., 2011) but did not correlate with Ezh2 loading
at target genes in two other studies (Kaneko et al., 2010; Wu
sates from mice at different developmental time points.
ates from mice treated in vivo with progesterone (Pg), prolactin (Prl), or vehicle
shown). Tubulin provided a loading control.
ion in whole gland lysates frommice treated in vivo with Pg or vehicle for 72 hr.
. Cells were transfected with siGenome Smartpools for RISC-free, PR, or Ezh2.
was performed 60 hr following transfection.
of ML, hormone receptor-positive luminal progenitor (CD24+CD29loCD49b+
b+Sca-1�). Lineage-negative cells were fractionated using a combination of
; error bars show SEM.
uminal and progenitor subsets isolated from mice treated with progesterone or
PR� LP subset.
pigenome with gene expression changes in the alveolar luminal compartment
in epigenetic changes that culminate in breast cancer.
ell Reports 3, 411–426, February 21, 2013 ª2013 The Authors 423
and Zhang, 2011). Our data suggest that Thr487 phosphorylation
is more likely to promote Ezh2 methyltransferase activity in the
mammary gland, since K27-trimethylated H3, total Ezh2, and
phospho-Ezh2 levels were tightly correlated in pregnancy.
Progesterone appears to regulate Ezh2 through different
mechanisms. Acute treatment of mice with progesterone but
not prolactin led to a striking elevation in total Ezh2 and phos-
phorylated protein, suggesting that progesterone influences
the epigenetic ‘‘landscape’’ in pregnancy through modulating
the level of Ezh2 and its activity. Notably, the upregulation of
Ezh2 transcript levels occurred predominantly in the PR-nega-
tive luminal progenitor subset, implying a paracrine regulatory
mechanism. The Rank-Rankl signaling axis emerged as a likely
paracrine mediator, as the progesterone-target Ranklwas highly
induced in mature luminal cells, while Rank was most abundant
in PR-negative progenitors, in which Ezh2 expression was pro-
foundly stimulated. The Rank pathway has also been implicated
in paracrine signaling toMaSCs (Asselin-Labat et al., 2010; Joshi
et al., 2010). The dynamic expression pattern of phosphorylated
(T487) Ezh2 protein during mammary ontogeny suggests that
posttranslational mechanisms also regulate Ezh2 activity. In-
deed, it is tempting to speculate that Cdk1 (or 2)-mediated phos-
phorylation of Ezh2 is an important mechanism for integrating
cues from progesterone with chromatin modifications in order
to regulate the balance between proliferation and differentiation
in the developing gland (Figure 6H). Progesterone may also
directly influence cell cycle progression by receptor-mediated
recruitment of the cyclin A/Cdk2 complex to progesterone-
responsive promoters (Narayanan et al., 2005). Interestingly,
phosphorylation-mediated repression by Ezh2 occurs in a cell
stage-specific manner during the differentiation of satellite cells
to myotubes in response to signals from regenerating muscle
(Palacios et al., 2010).
Overexpression of EZH2 is a marker of poor prognosis and
metastatic disease in many solid carcinomas, including breast
(Bachmann et al., 2006; Bracken et al., 2003; Kleer et al., 2003;
Pietersen et al., 2008b; Raaphorst et al., 2003) and prostate
(Varambally et al., 2008), both of which are hormone-associated
malignancies. Notably, BRCA1-associated basal-like breast
tumors are selectively dependent on high levels of EZH2, thus
rendering them sensitive to the small molecule inhibitor DZNep
(Puppe et al., 2009). Moreover, EZH2 promotes the expansion
of breast tumor-initiating cells, possibly through the repression
of DNA damage repair pathways (Chang et al., 2011; Zeidler
et al., 2005). EZH2 overexpression is most frequently observed
in basal-like breast cancers. In these cancers, aberrant EZH2
expression may occur early in the oncogenic process and
within luminal progenitor cells that are exposed to sustained
progesterone signaling before they transition to a hormone
receptor-independent state. Interestingly, in utero exposure to
diethylstilbestrol may also influence breast cancer risk through
Ezh2 (Doherty et al., 2010). Our findings that Ezh2 coordinates
signals from progesterone with global changes in the mammary
epigenome have important implications for cancer, given the
impact of ovarian hormones on breast tumorigenesis. Specifi-
cally, the data suggest that sustained hormone exposure may
initiate oncogenesis through dysregulated histone methylation
of chromatin.
424 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
EXPERIMENTAL PROCEDURES
Mouse Strains and Hormone Treatment
Floxed Ezh2 mice were a kind gift from Dr. A. Tarakovsky (Su et al., 2003), and
the MMTV–cre (line A) mice were a gift from K.-U. Wagner (Eppley Institute,
Omaha, NE). MMTV-cre mice were maintained as a pure strain on a FVB/N
background. Floxed or deleted Ezh2 mice were analyzed on either a FVB/N
or mixed FVB/N C57/Bl6 background. For in vivo hormonal treatment,
8-week-old FVB/N female mice were injected intrascapularly daily with 1 mg
progesterone in 200 ml peanut oil or vehicle alone for either 16 or 72 hr or
prolactin (12.5 mg/kg) intraperitoneally for 16 hr. Prolactin (gift from the
National Hormone and Peptide Program, NIDDK) was dissolved at 1 mg/ml
in 0.1% BSA in PBS. The inguinal mammary glands (minus lymph node)
were harvested and snap-frozen. All animal experiments conform to regulatory
standards and were approved by the Walter and Eliza Hall Institute Animal
Ethics Committee.
Mammary Cell Preparation, Cell Sorting, and Transplantation
Mammary epithelial cell suspensions from female mice and flow cytometry
were performed as described (Shackleton et al., 2006). Cells sorted by flow
cytometry were manually counted and transplanted at limiting dilution as
described. Whole-mounting of glands, BrdU-labeling, immunohistochemistry,
and western blotting are described in Extended Experimental Procedures.
Cell Culture and Retroviral-Mediated Infection
For primary cell culture, freshly sorted mammary cells were plated on irradi-
ated fibroblast feeder layers (3,000 Rads) on collagen-coated, six-well plates
and infected as described (Bouras et al., 2008).
ChIP Sample Preparation and Sequencing
Freshly sorted cells were crosslinked with 1% paraformaldehyde and the ChIP
assay performed according to themanufacturer’s protocol (Millipore #17-371).
Briefly, cells were lysed and the Diagenode Bioruptor used for chromatin
shearing to a size range of 200 to 400 bp. Sheared chromatin was diluted
and incubated at 4�C overnight with antibodies against Histone H3 trimethyl
Lys4 (Millipore #07-473), Histone H3 trimethyl Lys27 (Millipore #07-449),
Histone H3 dimethyl Lys9 (Abcam #ab1220), or mouse isotype control (Milli-
pore #12-371). Immune complexes were handled as per the manufacturer’s
protocol.
Libraries for paired-end sequencing were prepared by GeneWorks
(Adelaide, Australia) using 10 ng ChIP DNA according to the Illumina ChIP-
seq Sample Preparation protocol, revision A, with the following modifications.
Paired-end adapters were substituted for standard genomic adapters and
library amplification (18 cycles) performed using primers PE 1.0 and 2.0.
Size selection for 300 bp DNA was performed prior to amplification. Final
libraries were analyzed using the Agilent Bioanalyzer High Sensitivity DNA
Kit and sequenced on the Illumina Genome Analyzer IIx running SCS 2.8,
generating 35 base paired-end reads using the CASAVA 1.7 analysis pipeline.
The statistical analyses are described in the Extended Experimental Proce-
dures. Sequence data have been submitted to the Gene Expression Omnibus
under accession number GSE43212.
ACCESSION NUMBERS
The GEO database accession number for the ChIP-seq data is GSE43212.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, two
tables, and six figures and can be found with this article online at http://dx.
doi.org/10.1016/j.celrep.2012.12.020.
LICENSING INFORMATION
This is an open-access article distributed under the terms of the Creative
Commons Attribution-NonCommercial-No Derivative Works License, which
permits non-commercial use, distribution, and reproduction in any medium,
provided the original author and source are credited.
ACKNOWLEDGMENTS
We are very grateful to A. Tarakovsky for providing floxed Ezh2 mice and J.
Adams for critical review of the manuscript. We also thank GeneWorks for
library preparation and sequencing of ChIP samples; D. Wu for help with bio-
informatic analysis; E. Nolan, T. McLennan, B. Capaldo, and T.Ward for expert
help; C. Clarke, K. Simpson, D. Reinberg, G. McArthur, and B. Sarcevic for
reagents; and the animal, FACS, and histology facilities at WEHI. We thank
C. Clarke, J. Carroll, and S. Clark for discussions. This work was supported
by the Australian National Health and Medical Research Council (NHMRC)
#461224, #461221, #637307, #637308, #1016701, and Australia Fellowship
(to J.E.V.); NHMRC #490037 (to G.K.S.); the WEHI Genomics Fund; NHMRC
IRIISS; the Victorian State Government through VCA funding of the Victorian
Breast Cancer Research Consortium and Operational Infrastructure Support;
and the ACRF. B.P. and T.B. were supported by NHMRC Biomedical and
National Breast Cancer Foundation (Australia) Fellowships, respectively.
Received: May 29, 2012
Revised: December 20, 2012
Accepted: December 28, 2012
Published: January 31, 2013
REFERENCES
Asselin-Labat, M.L., Sutherland, K.D., Barker, H., Thomas, R., Shackleton, M.,
Forrest, N.C., Hartley, L., Robb, L., Grosveld, F.G., van der Wees, J., et al.
(2007). Gata-3 is an essential regulator of mammary-gland morphogenesis
and luminal-cell differentiation. Nat. Cell Biol. 9, 201–209.
Asselin-Labat, M.L., Vaillant, F., Sheridan, J.M., Pal, B., Wu, D., Simpson, E.R.,
Yasuda, H., Smyth, G.K., Martin, T.J., Lindeman, G.J., and Visvader, J.E.
(2010). Control of mammary stem cell function by steroid hormone signalling.
Nature 465, 798–802.
Asselin-Labat, M.L., Sutherland, K.D., Vaillant, F., Gyorki, D.E., Wu, D., Hol-
royd, S., Breslin, K., Ward, T., Shi, W., Bath, M.L., et al. (2011). Gata-3 nega-
tively regulates the tumor-initiating capacity of mammary luminal progenitor
cells and targets the putative tumor suppressor caspase-14. Mol. Cell. Biol.
31, 4609–4622.
Bachmann, I.M., Halvorsen, O.J., Collett, K., Stefansson, I.M., Straume, O.,
Haukaas, S.A., Salvesen, H.B., Otte, A.P., and Akslen, L.A. (2006). EZH2
expression is associated with high proliferation rate and aggressive tumor
subgroups in cutaneousmelanoma and cancers of the endometrium, prostate,
and breast. J. Clin. Oncol. 24, 268–273.
Bernardo, G.M., Lozada, K.L., Miedler, J.D., Harburg, G., Hewitt, S.C., Mosley,
J.D., Godwin, A.K., Korach, K.S., Visvader, J.E., Kaestner, K.H., et al. (2010).
FOXA1 is an essential determinant of ERalpha expression and mammary
ductal morphogenesis. Development 137, 2045–2054.
Bernstein, B.E., Mikkelsen, T.S., Xie, X., Kamal, M., Huebert, D.J., Cuff, J., Fry,
B., Meissner, A., Wernig, M., Plath, K., et al. (2006). A bivalent chromatin
structure marks key developmental genes in embryonic stem cells. Cell 125,
315–326.
Bouras, T., Pal, B., Vaillant, F., Harburg, G., Asselin-Labat, M.L., Oakes, S.R.,
Lindeman, G.J., and Visvader, J.E. (2008). Notch signaling regulatesmammary
stem cell function and luminal cell-fate commitment. Cell Stem Cell 3,
429–441.
Bracken, A.P., and Helin, K. (2009). Polycomb group proteins: navigators of
lineage pathways led astray in cancer. Nat. Rev. Cancer 9, 773–784.
Bracken, A.P., Pasini, D., Capra, M., Prosperini, E., Colli, E., and Helin, K.
(2003). EZH2 is downstream of the pRB-E2F pathway, essential for prolifera-
tion and amplified in cancer. EMBO J. 22, 5323–5335.
Brisken, C., and O’Malley, B. (2010). Hormone action in the mammary gland.
Cold Spring Harb. Perspect. Biol. 2, a003178.
C
Brisken, C., Park, S., Vass, T., Lydon, J.P., O’Malley, B.W., andWeinberg, R.A.
(1998). A paracrine role for the epithelial progesterone receptor in mammary
gland development. Proc. Natl. Acad. Sci. USA 95, 5076–5081.
Caretti, G., Di Padova, M., Micales, B., Lyons, G.E., and Sartorelli, V. (2004).
The Polycomb Ezh2 methyltransferase regulates muscle gene expression
and skeletal muscle differentiation. Genes Dev. 18, 2627–2638.
Caretti, G., Palacios, D., Sartorelli, V., and Puri, P.L. (2011). Phosphoryl-EZH-
ion. Cell Stem Cell 8, 262–265.
Chang, C.J., Yang, J.Y., Xia, W., Chen, C.T., Xie, X., Chao, C.H., Woodward,
W.A., Hsu, J.M., Hortobagyi, G.N., and Hung, M.C. (2011). EZH2 promotes
expansion of breast tumor initiating cells through activation of RAF1-b-catenin
signaling. Cancer Cell 19, 86–100.
Chen, S., Bohrer, L.R., Rai, A.N., Pan, Y., Gan, L., Zhou, X., Bagchi, A., Simon,
J.A., and Huang, H. (2010). Cyclin-dependent kinases regulate epigenetic
gene silencing through phosphorylation of EZH2. Nat. Cell Biol. 12, 1108–
1114.
Delgado-Olguın, P., Huang, Y., Li, X., Christodoulou, D., Seidman, C.E., Seid-
man, J.G., Tarakhovsky, A., and Bruneau, B.G. (2012). Epigenetic repression
of cardiac progenitor gene expression by Ezh2 is required for postnatal cardiac
homeostasis. Nat. Genet. 44, 343–347.
Doherty, L.F., Bromer, J.G., Zhou, Y., Aldad, T.S., and Taylor, H.S. (2010). In
utero exposure to diethylstilbestrol (DES) or bisphenol-A (BPA) increases
EZH2 expression in the mammary gland: an epigenetic mechanism linking
endocrine disruptors to breast cancer. Horm. Cancer 1, 146–155.
Ezhkova, E., Pasolli, H.A., Parker, J.S., Stokes, N., Su, I.H., Hannon, G., Tara-
khovsky, A., and Fuchs, E. (2009). Ezh2 orchestrates gene expression for the
stepwise differentiation of tissue-specific stem cells. Cell 136, 1122–1135.
Ezhkova, E., Lien, W.H., Stokes, N., Pasolli, H.A., Silva, J.M., and Fuchs, E.
(2011). EZH1 and EZH2 cogovern histone H3K27 trimethylation and are essen-
tial for hair follicle homeostasis and wound repair. Genes Dev. 25, 485–498.
Graham, J.D., Mote, P.A., Salagame, U., van Dijk, J.H., Balleine, R.L., Huscht-
scha, L.I., Reddel, R.R., and Clarke, C.L. (2009). DNA replication licensing and
progenitor numbers are increased by progesterone in normal human breast.
Endocrinology 150, 3318–3326.
Hu, Y., and Smyth, G.K. (2009). ELDA: extreme limiting dilution analysis for
comparing depleted and enriched populations in stem cell and other assays.
J. Immunol. Methods 347, 70–78.
Joshi, P.A., Jackson, H.W., Beristain, A.G., Di Grappa, M.A., Mote, P.A.,
Clarke, C.L., Stingl, J.,Waterhouse, P.D., and Khokha, R. (2010). Progesterone
induces adult mammary stem cell expansion. Nature 465, 803–807.
Juan, A.H., Derfoul, A., Feng, X., Ryall, J.G., Dell’Orso, S., Pasut, A., Zare, H.,
Simone, J.M., Rudnicki, M.A., and Sartorelli, V. (2011). Polycomb EZH2
controls self-renewal and safeguards the transcriptional identity of skeletal
muscle stem cells. Genes Dev. 25, 789–794.
Kaneko, S., Li, G., Son, J., Xu, C.F., Margueron, R., Neubert, T.A., and Rein-
berg, D. (2010). Phosphorylation of the PRC2 component Ezh2 is cell cycle-
regulated and up-regulates its binding to ncRNA. Genes Dev. 24, 2615–2620.
Keller, P.J., Arendt, L.M., Skibinski, A., Logvinenko, T., Klebba, I., Dong, S.,
Smith, A.E., Prat, A., Perou, C.M., Gilmore, H., et al. (2012). Defining the cellular
precursors to human breast cancer. Proc. Natl. Acad. Sci. USA 109, 2772–
2777.
Kleer, C.G., Cao, Q., Varambally, S., Shen, R., Ota, I., Tomlins, S.A., Ghosh, D.,
Sewalt, R.G., Otte, A.P., Hayes, D.F., et al. (2003). EZH2 is a marker of aggres-
sive breast cancer and promotes neoplastic transformation of breast epithelial
cells. Proc. Natl. Acad. Sci. USA 100, 11606–11611.
Li, W., Ferguson, B.J., Khaled, W.T., Tevendale, M., Stingl, J., Poli, V., Rich, T.,
Salomoni, P., and Watson, C.J. (2009). PML depletion disrupts normal
mammary gland development and skews the composition of the mammary
luminal cell progenitor pool. Proc. Natl. Acad. Sci. USA 106, 4725–4730.
Lien, W.H., Guo, X., Polak, L., Lawton, L.N., Young, R.A., Zheng, D., and
Fuchs, E. (2011). Genome-wide maps of histone modifications unwind in vivo
chromatin states of the hair follicle lineage. Cell Stem Cell 9, 219–232.
ell Reports 3, 411–426, February 21, 2013 ª2013 The Authors 425
Lim, E., Wu, D., Pal, B., Bouras, T., Asselin-Labat, M.L., Vaillant, F., Yagita, H.,
Lindeman, G.J., Smyth, G.K., and Visvader, J.E. (2010). Transcriptome anal-
yses of mouse and human mammary cell subpopulations reveal multiple
conserved genes and pathways. Breast Cancer Res. 12, R21.
Lydon, J.P., DeMayo, F.J., Funk, C.R., Mani, S.K., Hughes, A.R., Montgomery,
C.A., Jr., Shyamala, G., Conneely, O.M., and O’Malley, B.W. (1995). Mice lack-
ing progesterone receptor exhibit pleiotropic reproductive abnormalities.
Genes Dev. 9, 2266–2278.
Mallepell, S., Krust, A., Chambon, P., and Brisken, C. (2006). Paracrine
signaling through the epithelial estrogen receptor alpha is required for prolifer-
ation and morphogenesis in the mammary gland. Proc. Natl. Acad. Sci. USA
103, 2196–2201.
Malumbres, M., and Barbacid, M. (2009). Cell cycle, CDKs and cancer:
a changing paradigm. Nat. Rev. Cancer 9, 153–166.
Margueron, R., and Reinberg, D. (2011). The Polycomb complex PRC2 and its
mark in life. Nature 469, 343–349.
Maruyama, R., Choudhury, S., Kowalczyk, A., Bessarabova, M., Beresford-
Smith, B., Conway, T., Kaspi, A., Wu, Z., Nikolskaya, T., Merino, V.F., et al.
(2011). Epigenetic regulation of cell type-specific expression patterns in the
human mammary epithelium. PLoS Genet. 7, e1001369.
Mikkelsen, T.S., Ku, M., Jaffe, D.B., Issac, B., Lieberman, E., Giannoukos, G.,
Alvarez, P., Brockman, W., Kim, T.K., Koche, R.P., et al. (2007). Genome-wide
maps of chromatin state in pluripotent and lineage-committed cells. Nature
448, 553–560.
Mueller, S.O., Clark, J.A., Myers, P.H., and Korach, K.S. (2002). Mammary
gland development in adult mice requires epithelial and stromal estrogen
receptor alpha. Endocrinology 143, 2357–2365.
Mulac-Jericevic, B., Lydon, J.P., DeMayo, F.J., and Conneely, O.M. (2003).
Defective mammary gland morphogenesis in mice lacking the progesterone
receptor B isoform. Proc. Natl. Acad. Sci. USA 100, 9744–9749.
Narayanan, R., Adigun, A.A., Edwards, D.P., and Weigel, N.L. (2005). Cyclin-
dependent kinase activity is required for progesterone receptor function: novel
role for cyclin A/Cdk2 as a progesterone receptor coactivator. Mol. Cell. Biol.
25, 264–277.
Nicol, J.W., Helt, G.A., Blanchard, S.G., Jr., Raja, A., and Loraine, A.E. (2009).
The Integrated Genome Browser: free software for distribution and exploration
of genome-scale datasets. Bioinformatics 25, 2730–2731.
Oakes, S.R., Rogers, R.L., Naylor, M.J., and Ormandy, C.J. (2008). Prolactin
regulation of mammary gland development. J. Mammary Gland Biol.
Neoplasia 13, 13–28.
Palacios, D., Mozzetta, C., Consalvi, S., Caretti, G., Saccone, V., Proserpio, V.,
Marquez, V.E., Valente, S., Mai, A., Forcales, S.V., et al. (2010). TNF/p38a/pol-
ycomb signaling to Pax7 locus in satellite cells links inflammation to the epige-
netic control of muscle regeneration. Cell Stem Cell 7, 455–469.
Pietersen, A.M., Evers, B., Prasad, A.A., Tanger, E., Cornelissen-Steijger, P.,
Jonkers, J., and van Lohuizen, M. (2008a). Bmi1 regulates stem cells and
proliferation and differentiation of committed cells in mammary epithelium.
Curr. Biol. 18, 1094–1099.
Pietersen, A.M., Horlings, H.M., Hauptmann, M., Langerød, A., Ajouaou, A.,
Cornelissen-Steijger, P., Wessels, L.F., Jonkers, J., van de Vijver, M.J., and
van Lohuizen, M. (2008b). EZH2 and BMI1 inversely correlate with prognosis
and TP53 mutation in breast cancer. Breast Cancer Res. 10, R109.
Puppe, J., Drost, R., Liu, X., Joosse, S.A., Evers, B., Cornelissen-Steijger, P.,
Nederlof, P., Yu, Q., Jonkers, J., van Lohuizen, M., and Pietersen, A.M. (2009).
BRCA1-deficient mammary tumor cells are dependent on EZH2 expression
and sensitive to PolycombRepressive Complex 2-inhibitor 3-deazaneplanocin
A. Breast Cancer Res. 11, R63.
426 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
Raaphorst, F.M., Meijer, C.J., Fieret, E., Blokzijl, T., Mommers, E., Buerger, H.,
Packeisen, J., Sewalt, R.A., Otte, A.P., and van Diest, P.J. (2003). Poorly differ-
entiated breast carcinoma is associated with increased expression of the
human polycomb group EZH2 gene. Neoplasia 5, 481–488.
Robinson, G.W., and Hennighausen, L. (2011). MMTV-Cre transgenes can
adversely affect lactation: considerations for conditional gene deletion in
mammary tissue. Anal. Biochem. 412, 92–95.
Rudolph, M.C., McManaman, J.L., Hunter, L., Phang, T., and Neville, M.C.
(2003). Functional development of the mammary gland: use of expression
profiling and trajectory clustering to reveal changes in gene expression during
pregnancy, lactation, and involution. J. Mammary Gland Biol. Neoplasia 8,
287–307.
Sauvageau, M., and Sauvageau, G. (2010). Polycomb group proteins: multi-
faceted regulators of somatic stem cells and cancer. Cell Stem Cell 7,
299–313.
Sen, G.L.,Webster, D.E., Barragan, D.I., Chang, H.Y., and Khavari, P.A. (2008).
Control of differentiation in a self-renewing mammalian tissue by the histone
demethylase JMJD3. Genes Dev. 22, 1865–1870.
Shackleton, M., Vaillant, F., Simpson, K.J., Stingl, J., Smyth, G.K., Asselin-
Labat, M.L., Wu, L., Lindeman, G.J., and Visvader, J.E. (2006). Generation of
a functional mammary gland from a single stem cell. Nature 439, 84–88.
Sleeman, K.E., Kendrick, H., Robertson, D., Isacke, C.M., Ashworth, A., and
Smalley, M.J. (2007). Dissociation of estrogen receptor expression and in vivo
stem cell activity in the mammary gland. J. Cell Biol. 176, 19–26.
Statham, A.L., Strbenac, D., Coolen, M.W., Stirzaker, C., Clark, S.J., and Rob-
inson, M.D. (2010). Repitools: an R package for the analysis of enrichment-
based epigenomic data. Bioinformatics 26, 1662–1663.
Stingl, J., Eirew, P., Ricketson, I., Shackleton, M., Vaillant, F., Choi, D., Li, H.I.,
and Eaves, C.J. (2006). Purification and unique properties of mammary epithe-
lial stem cells. Nature 439, 993–997.
Su, I.H., Basavaraj, A., Krutchinsky, A.N., Hobert, O., Ullrich, A., Chait, B.T.,
and Tarakhovsky, A. (2003). Ezh2 controls B cell development through histone
H3 methylation and Igh rearrangement. Nat. Immunol. 4, 124–131.
Van Keymeulen, A., Rocha, A.S., Ousset, M., Beck, B., Bouvencourt, G., Rock,
J., Sharma, N., Dekoninck, S., and Blanpain, C. (2011). Distinct stem cells
contribute to mammary gland development and maintenance. Nature 479,
189–193.
Varambally, S., Cao, Q., Mani, R.S., Shankar, S., Wang, X., Ateeq, B., Laxman,
B., Cao, X., Jing, X., Ramnarayanan, K., et al. (2008). Genomic loss of micro-
RNA-101 leads to overexpression of histone methyltransferase EZH2 in
cancer. Science 322, 1695–1699.
Visvader, J.E. (2009). Keeping abreast of themammary epithelial hierarchy and
breast tumorigenesis. Genes Dev. 23, 2563–2577.
Wang, L., Jin, Q., Lee, J.E., Su, I.H., and Ge, K. (2010). Histone H3K27 meth-
yltransferase Ezh2 represses Wnt genes to facilitate adipogenesis. Proc. Natl.
Acad. Sci. USA 107, 7317–7322.
Wei, Y., Chen, Y.H., Li, L.Y., Lang, J., Yeh, S.P., Shi, B., Yang, C.C., Yang, J.Y.,
Lin, C.Y., Lai, C.C., and Hung, M.C. (2011). CDK1-dependent phosphorylation
of EZH2 suppresses methylation of H3K27 and promotes osteogenic differen-
tiation of human mesenchymal stem cells. Nat. Cell Biol. 13, 87–94.
Wu, S.C., and Zhang, Y. (2011). Cyclin-dependent kinase 1 (CDK1)-mediated
phosphorylation of enhancer of zeste 2 (Ezh2) regulates its stability. J. Biol.
Chem. 286, 28511–28519.
Zeidler, M., Varambally, S., Cao, Q., Chinnaiyan, A.M., Ferguson, D.O., Mer-
ajver, S.D., and Kleer, C.G. (2005). The Polycomb group protein EZH2 impairs
DNA repair in breast epithelial cells. Neoplasia 7, 1011–1019.
Supplemental Information
EXTENDED EXPERIMENTAL PROCEDURES
Mouse StrainsAll animal experiments were conducted using mice bred at and maintained in our animal facility according to institutional and the
Melbourne Health Research Directorate Animal Ethics Committee guidelines. Ezh2 mice were on a pure FVB/N or mixed C57Bl/6
and FVB/N background; the same results were obtained for each. Adult female mice were subjected to timed pregnancies, scored
by the presence of vaginal plugs and confirmed by examination of embryos on collection of mammary glands. Mice were genotyped
using the primers listed below.
Mammary Cell Preparation and Cell SortingAntibodies against mouse antigens were purchased from BD PharMingen (San Diego, CA) unless otherwise specified, and included
CD24-PE, biotinylated CD31, CD45 and Ter119, CD29–FITC (Chemicon, Temecula, CA), CD61-APC and streptavidin–APC–Cy7,
CD14-biotin (eBioscience, San Diego, CA). For the luminal progenitor fractionation experiments, CD49b-FITC, Sca1-APC, CD45-
PECy-7 and CD31-PECy-7 were from eBioscience, and CD29-APC-Cy7 from Biolegend (San Diego, CA). Single cell suspensions
were sorted on a FACSAria or FACSDiva (BD PharMingen).
Histology and Whole MountingFor histological examination of mouse mammary glands, tissues were fixed in 4% paraformaldehyde overnight and embedded in
paraffin. Sections (5 mm) were prepared and stained with hematoxylin and eosin (H&E). For whole-mount analysis, mammary glands
were harvested and fixed in Carnoy’s solution (six parts 100% ethanol, three parts chloroform, one part glacial acetic acid) and
stained with hematoxylin.
Bromodeoxyuridine ImmunodetectionMicewere injected with BrdUCell Labeling Reagent (0.5mg/10 g body weight, AmershamBiosciences) 1 hr prior to tissue collection.
Tissueswere fixed in 4%paraformaldehyde and embedded in paraffin. For immunohistochemical detection of BrdU-labeled cells, rat
anti-BrdU (Becton Dickinson) and biotinylated rabbit anti-rat IgG antibody (Dako) were used, followed by HRP-conjugated strepta-
vidin (Dako, LSAB2).
Cell Culture and Retroviral-Mediated InfectionFor primary cell culture, freshly sorted mammary epithelial cells were plated on irradiated fibroblast feeders on collagen-coated
6-well plates and infected as described (Bouras et al., 2008). Colony assays in 2D have been described (Shackleton et al., 2006).
Primary cells following transduction with cre-MIG (MSCV-cre-IRES-GFP) or empty retrovirus were manually counted after sorting
for GFP and replated at 200 or 300 cells per well in a 24-well plate containing a feeder layer of irradiated fibroblasts. After 7 days,
cultures were fixed with ice-cold acetone/methanol and stained with Giemsa for enumeration.
For siRNA-mediated knockdown of PR in T-47D cells, the following siGENOME SMARTpools (Dharmacon) were used: Ezh2 (hu)
M-004218-03-0005, RISC-free D-001220-01 and progesterone receptor (hu) M-003433-01-0005. Transfections were performed as
described by the manufacturer (Dharmacon) using Dharmafect-1 and cells harvested for analysis after 60 hr.
ImmunohistochemistryParaffin-embedded sections (5 mm) were dewaxed in xylene and rehydrated through an alcohol series, blocked with 3% hydrogen
peroxide, and subjected to antigen retrieval by boiling in 10 mM citrate buffer pH 6.0 for 30 s at pressure using a DAKO pressure
cooker. The mouse-on-mouse (MOM) kit (Vector) was used for mousemonoclonal antibodies. Immunostaining with other antibodies
was performed using the streptavidin-biotin peroxidase detection system as per the manufacturer’s instructions (ABC reagent,
Vector Laboratories). 3,3-diaminobenzidine was used as substrate (DAKO). In all cases, an isotype-matched control IgG was
used as a negative control. The following antibodies were used: anti-SMA (Sigma), anti-milk (Accurate Chemical and Scientific),
anti-p63 (BD PharMingen), anti-keratin 18 (Progen Biotechnik), anti-keratin 14 (Covance), anti-BrdU (Becton Dickinson), anti-cas-
pase-3, anti-Ezh2 (Lake Placid Biologicals, Active Motif). Secondary antibodies were biotin-conjugated anti-rabbit IgG and anti-
mouse IgG (Vector Laboratories). The Npt2b antibody was a kind gift of Dr Juerg Biber from the University of Zurich, Switzerland.
Western Blot AnalysisWhole mammary gland lysates were prepared in RIPA buffer (50 mM Tris pH7.4, 150 mM NaCl, 0.25% Na-deoxycholate, 1% NP40,
1mMPMSF, 1X Roche complete mini protease inhibitor cocktail, 1X Roche PhosSTOP phosphatase inhibitor cocktail). The following
antibodies were used for western blot analysis: anti-H3K27me3 (Millipore #07-449); anti-Cdk1 (Cell Signaling #9112); anti-Cdk2: (Cell
Signaling #2546); anti-Ezh2 (BD Biosciences #612666); anti-phospho-Ezh2-Thr487 (Abcam #ab109398); anti-phospho-Ezh2-Ser21
(Abcam #ab84989); anti-tubulin (Sigma #T9026); anti-progesterone receptor Ab-7 was kindly provided by C. Clarke.
ChIP-seq Read MappingReads were mapped to the mouse reference genome mm9 using the Subread aligner (http://sourceforge.net/projects/subread).
A fragment was judged to be successfully mapped if the paired ends mapped to locations between 50 and 500 bp apart.
Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors S1
Genome-wide heatmaps of methylation mark coverage were drawn using Repitools (Statham et al., 2010). Density plots were
drawn using in-house scripts written in R (http://www.r-project.org). Coverage graphs were generated for genes of interest using
the Integrated Genome Browser (Nicol et al., 2009).
ChIP-seq Statistical AnalysisFragment counts were formed for each gene. Fragments were counted if the center of the fragment was contained in the TSS or
broad region for that gene. Statistical analysis of the count data was performed using edgeR package (Robinson et al., 2010) of
the Bioconductor software project (Gentleman et al., 2004). The Benjamini-Hochberg method was used to control the FDR. Genes
were called as significantly enriched for each histone mark using the normalizeChIPtoInput function, which normalizes the ChIP
counts to input and evaluates enrichment using a negative binomial statistical model. Log2-fold-changes in histone mark coverage
between cell populations or between conditions were computed using the predFC function. This function adjusted the ChIP counts
for input by fitting a negative binomial log-linear generalized linear model (McCarthy et al., 2012), using a prior count of 0.5 per sample
to avoid unreliably large log-fold-changes that might otherwise arise from zero or small counts. The negative binomial dispersion was
set to 0.01 for all calculations, allowing for a degree of biological variation typical of mouse experiments (McCarthy et al., 2012). Gene
Ontology analysis used the DAVID tool (Huang da et al., 2009). Gene set enrichment used the mean-rank gene set enrichment test
(Michaud et al., 2008).
RNA Extraction and Quantitative RT-PCR AnalysisTotal RNA was isolated from primary mammary epithelial subpopulations with the RNeasy Micro kit (QIAGEN). Reverse transcription
by using oligo(dT) primer and Moloney murine leukemia virus reverse transcriptase (Invitrogen) was according to the manufacturer’s
protocol. Quantitative RT-PCR was carried out by using a Rotorgene RG-6000 (Corbett Research) and SensiMix (dT) DNA kit
(Quantace) under the following conditions: 10 min at 95�C followed by 35 cycles consisting of 15 s at 95�C, 20 s at 62�C, and20 s at 72�C. Gene expression was determined with the Rotor-Gene software (version 1.7). The primer sequences are listed below.
Microarray AnalysisMicroarray expression data for steady-state epithelial cell subsets, pregnant and ovariectomized mice were analyzed as described
previously (Asselin-Labat et al., 2010; Lim et al., 2010). For Ezh2-deficient profiles, cell subsets were sorted from three independent
mouse pools, and up to 250 ng of RNA per sample was labeled, amplified and hybridized to Illumina Mouse-WG6 V2 Expression
BeadChips according to Illumina standard protocols at the Australian Genome Research Facility (Melbourne, Australia). Summary
probe profiles were exported from GenomeStudio and was analyzed using the limma software package (Smyth, 2004). Expression
values were normexp background adjusted and quantile normalized using control probes (Shi et al., 2010). The data have been
uploaded into GSE38203. Probes were filtered if not expressed (detection p-value > 0.05 across all arrays) or poorly annotated
(Barbosa-Morais et al., 2010). Differential expression between between Ezh2-deficient and control cells was assessed using linear
models and empirical Bayes moderated t-statistics (Smyth, 2004). Reliability was improved using array quality weights (Ritchie et al.,
2006), and each mouse pool was treated as a random effect to allow for dependence between samples from the same pool
(Smyth et al., 2005). Gene ontology enrichment for biological process (BP) terms was carried out on the top 500 genes from each
contrast using the GOstats package (Falcon and Gentleman, 2007). Focused gene set testing using Ezh2 signatures obtained
from Kamminga et al. (2006) (34 genes matched by gene symbols) and Ezhkova et al. (2009) (65 genes matched by gene symbols)
were performed using the roast method (Wu et al., 2010). Microarray probes werematched to ChIP-seq profiles by gene symbol. The
probe with highest average expression data was chosen to represent each gene.
Genes with FDR < 0.05 and at least 50% fold-change were selected as Ezh2-deficient signature genes and compared with the
human breast cancer data set from Herschkowitz et al. (2007). Where multiple probes were present for the same gene on a given
platform, the probe with the highest average expression level was kept for further analysis. Genes were matched between platforms
usingGene symbols and a signature score calculated for each sample as in Lim et al. (2010). TheEzh2-deficient signature geneswere
also compared across tumor subtypes using roast gene set tests (Wu et al., 2010) with Ezh2–deficient log-fold changes as gene
weights.
Oligonucleotides Used in the StudyGenotyping oligonucleotides (50 to 30)
Ezh2
Fwd 1: TTATTCATAGAGCCACCTGG
Fwd 2: ACGAAACAGCTCCAGATTCAGGG
Rev: CTGCTCTGAATGGCAACTCC
ChIP-q-RT-PCR oligonucleotides
c-Kit
Fwd: TCAGGGGTGCCACGATCCGT
Rev: TAGTCGGGATTGCCGGGCGA
S2 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
Lgr5
Fwd: TATCAAAGCCTCAAAGAAATC
Rev: GTGTTCTTCCAAGGTGTCTGA
Snai2
Fwd: TGTCATCAGCCGGTGGACTTCCT
Rev: TCCAAGGACCCAGGGGTTGTG
Hey2
Fwd: TTGGCTTGCCCAGAAGCACCT
Rev: TGCGCAGCTCAGCCTGTTTAG
Krt5
Fwd: GGCCTGTGACCTGTGAGGGACA
Rev: TCTCCTCCAGAGGTTGCCCCA
DNp63
Fwd: GAGTCCCGCCCCTCATGCCT
Rev: CTGAGAGCCTTGCGCTGCGA
Elf5
Fwd: GAGCCCCGACACCCCCTTTCA
Rev: TACAGTCCGCTGGTGCTGGGA
Wap
Fwd: AGCCACACCCGGTAGTAAGGTG
Rev: CTGGAGGTGGCCCTCGCTCA
Csn2
Fwd: ACAAGGCAGCAATTCAGAAGCTGGT
Rev: CCCCGGTCCTCTCACTTGGC
FoxA1
Fwd: CTAGCGCCACCCAGCGGTC
Rev: GTCGGTGCTCGCTTACCGGG
Wnt7b
Fwd: AGGCTGGGCTAACAGAGACCCC
Rev: GGAAGGACCTGGGTGCCCGA
Hey1
Fwd: CACAGCTCGCTTCGCTCCTGT
Rev: TGGCGTCAAGGGAGGCAGGT
Cdkn2a
Fwd: CAGCCGGTAAGAAGGGTTCACCT
Rev: GCTACCCGATAGCAAGCACTAGGA
Cdkn1c
Fwd: GCTGGCCCTAAGACCCTCTA
Rev: ATGGGCCCAACTTGTGTCTC
Camk2n1
Fwd: ACCTACGGGTAGAGACCCAG
Rev: GGGCTTTACCTTCAGTTGCC
Ccnd2
Fwd: GCCTCGGCCACGCAGGAAAA
Rev: ACGCTCCGCGCAGACACCTA
Quantitative RT-PCR oligonucleotides
Elf5
Fwd: CCCTGAATACTGGACCAAGC
Rev: GCTGCCTCAATGAACTCCTC
Wap
Fwd: TGC CTC ATC AGC CTC GTT CTT G
Rev: CTG GAG CAT TCT ATC TTC ATT GGG
Csn2
Fwd: AAA GGA CTT GAC AGC CAT GAA
Rev: TAG CCT GGA GCA CAT CCT CT
Lgr5
Fwd: CCA ATG GAA TAA AGA CGA CGG CAA CA
Rev: GGG CCT TCA GGT CTT CCT CAA AGT CA
Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors S3
Arf
Fwd: CGCAGGTTCTTGGTCACTGTGAGG
Rev: TGCCCATCATCATCACCTGGTCC
Rankl
Fwd: TGTACTTTCGAGCGCAGATG
Rev: CCACAATGTGTTGCAGTTCC
Rank
Fwd: ACACCCTGCCTCCTGGGCTT
Rev: AAGCCTGGGCCTCCTTGGGT
PR
Fwd: GCTTGCATGATCTTGTGAAACAGC
Rev: GGAAATTCCACAGCCAGTGTCC
Ezh2
Fwd: GCAATTTAGAAAACGGAAATGC
Rev: GTACAAAACACTTTGCAGCTGG
18S rRNA
Fwd: GTAACCCGTTGAACCCCATT
Rev: CCATCCAATCGGTAGTAGCG
SUPPLEMENTAL REFERENCES
Barbosa-Morais, N.L., Dunning, M.J., Samarajiwa, S.A., Darot, J.F., Ritchie, M.E., Lynch, A.G., and Tavare, S. (2010). A re-annotation pipeline for Illumina
BeadArrays: improving the interpretation of gene expression data. Nucleic Acids Res. 38, e17.
Falcon, S., and Gentleman, R. (2007). Using GOstats to test gene lists for GO term association. Bioinformatics 23, 257–258.
Gentleman, R.C., Carey, V.J., Bates, D.M., Bolstad, B., Dettling, M., Dudoit, S., Ellis, B., Gautier, L., Ge, Y., Gentry, J., et al. (2004). Bioconductor: open software
development for computational biology and bioinformatics. Genome Biol. 5, R80.
Herschkowitz, J.I., Simin, K., Weigman, V.J., Mikaelian, I., Usary, J., Hu, Z., Rasmussen, K.E., Jones, L.P., Assefnia, S., Chandrasekharan, S., et al. (2007).
Identification of conserved gene expression features between murine mammary carcinoma models and human breast tumors. Genome Biol. 8, R76.
Huang da, W., Sherman, B.T., and Lempicki, R.A. (2009). Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat.
Protoc. 4, 44–57.
Kamminga, L.M., Bystrykh, L.V., de Boer, A., Houwer, S., Douma, J., Weersing, E., Dontje, B., and de Haan, G. (2006). The Polycomb group gene Ezh2 prevents
hematopoietic stem cell exhaustion. Blood 107, 2170–2179.
McCarthy, D.J., Chen, Y., and Smyth, G.K. (2012). Differential expression analysis of multifactor RNA-Seq experiments with respect to biological variation.
Nucleic Acids Res. 40, 4288–4297.
Michaud, J., Simpson, K.M., Escher, R., Buchet-Poyau, K., Beissbarth, T., Carmichael, C., Ritchie, M.E., Schutz, F., Cannon, P., Liu, M., et al. (2008). Integrative
analysis of RUNX1 downstream pathways and target genes. BMC Genomics 9, 363.
Ritchie, M.E., Diyagama, D., Neilson, J., van Laar, R., Dobrovic, A., Holloway, A., and Smyth, G.K. (2006). Empirical array quality weights in the analysis of
microarray data. BMC Bioinformatics 7, 261.
Robinson, M.D., McCarthy, D.J., and Smyth, G.K. (2010). edgeR: a Bioconductor package for differential expression analysis of digital gene expression data.
Bioinformatics 26, 139–140.
Shi, W., Oshlack, A., and Smyth, G.K. (2010). Optimizing the noise versus bias trade-off for Illumina whole genome expression BeadChips. Nucleic Acids Res. 38,
e204.
Smyth, G.K. (2004). Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat. Appl. Genet. Mol. Biol. 3,
Article 3.
Smyth, G.K., Michaud, J., and Scott, H.S. (2005). Use of within-array replicate spots for assessing differential expression in microarray experiments.
Bioinformatics 21, 2067–2075.
Vikstrom, I., Carotta, S., Luthje, K., Peperzak, V., Jost, P.J., Glaser, S., Busslinger, M., Bouillet, P., Strasser, A., Nutt, S.L., and Tarlinton, D.M. (2010). Mcl-1 is
essential for germinal center formation and B cell memory. Science 330, 1095–1099.
Wu, D., Lim, E., Vaillant, F., Asselin-Labat, M.L., Visvader, J.E., and Smyth, G.K. (2010). ROAST: rotation gene set tests for complex microarray experiments.
Bioinformatics 26, 2176–2182.
S4 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
H3K4me3 H3K27me3H3K9me2
100%
50%
Expression
-5 TSS +5
Score
Luminal Progenitor (LP) cells Mature Luminal (ML) cells
4 8 12 4 8 12H3K4me3 H3K27me3H3K9me2
-5 TSS +5 -5 TSS +5 -5 TSS +5 -5 TSS +5 -5 TSS +5
1.0
0.5
0
1.5
TSS 5kb 10kb-5kb-10kb TSS 5kb 10kb-5kb-10kb TSS 5kb 10kb-5kb-10kb
0.10
0.05
0
0.150.10
0.05
0
0.150.20
High expression
MaSC/basal
Luminal progenitor
TSS 5kb 10kb-5kb-10kb TSS 5kb 10kb-5kb-10kb TSS 5kb 10kb-5kb-10kb
Mature luminal
Low expression
TSS 5kb 10kb-5kb-10kb TSS 5kb 10kb-5kb-10kb TSS 5kb 10kb-5kb-10kb
0.04
0.02
0.06
0.10
0.08
0.12
0
0.04
0.02
0.06
0.10
0.08
0.12
0
1.0
0.5
0
1.5
0.04
0.02
0.06
0.10
0.08
0.12
0
1.0
0.5
0
1.5
0.04
0.02
0.06
0.10
0.08
0.12
0
0.14
Ave
rage
cov
erag
eA
vera
ge c
over
age
Ave
rage
cov
erag
e
H3K4me3 H3K9me2 H3K27me3
H3K4me3 H3K9me2 H3K27me3
H3K4me3 H3K9me2 H3K27me3
100%
50%
Score
Expression
A
B
Figure S1. Genome-Wide Heatmaps of Methylation Profiles for Luminal Progenitor and Mature Luminal Subpopulations in the ‘‘Steady
State’’, Related to Figures 1 and 2
(A) Heat maps for luminal progenitor (left) and mature luminal cells (right) show distribution of H3K4me3, H3K9me2, and H3K27me3 marks ± 5 kb of the TSS of
each gene. Genes are clustered into groups according to their histonemethylation profiles and ordered within groups by expression level. Far right column shows
log2-expression. Other columns show percentage coverage.
(B) Density plots of average histone methylation coverage stratified by gene expression. The x-axes show distance from TSS in base pairs. The y-axis shows
average number of covering reads at each base pair. For each cell subset, genes are stratified into four equal-size groups by expression level, and the average
coverage is shown for each group. Lines are red to green from high to low expression. Plots are shown for each epigenetic mark in each epithelial cell subset.
H3K4me3 marks peak sharply around the TSS and directly correlate with gene expression, whereas H3K27me3 shows a broad pattern and reverse association
with expression.
Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors S5
Wap
Elf5 Csn2
Lgr5
Fol
d ch
ange
vs
Inpu
t
virgin 12.5 dP
virgin 12.5dp
virgin 12.5 dP
virgin 12.5 dP
virgin
12.5 dP
1.0
0.5
0.75
1.2
0.4
0.8
1.6
2.0
0.5
1.0
2.5
1.5virgin
12.5 dP
virgin
12.5 dP
virgin
12.5 dP
Fol
d ch
ange
vs
Inpu
t F
old
chan
ge v
s In
put
0.25
Fol
d ch
ange
vs
Inpu
t
1.0
0.5
0.75
0
0.25
0
0
0
Csn2Elf5
MaSC Luminal MaSC Luminal
Wap
MaSC Luminal
Lgr5
MaSC Luminal
0.35
0.05
0.15
0.45
0.25
0.014
0.002
0.006
0.018
0.010
0.35X101.0X102
0.05
0.25
0.2
0.4
0.6
0.8
0.15
00 0 0
A
B
8 wk virgin 12.5 dP
Exp
ress
ion
rela
tive
to 1
8S r
RN
A
control
Ovx
Ovx
Ccnd2
Cdkn2a
Ovx
control Ovx
4
2
3
1
0
1.6
0.8
1.2
0.4
0
control
control
TSS
TSS
Fol
d ch
ange
vs
Inpu
t F
old
chan
ge v
s In
put
C
OVXvirgin 12.5 dP1.0
0.8
0.6
0.4
0.2
0
MaSC/basal Luminal
1.0
0.8
0.6
0.4
0.2
0
OVXvirgin 12.5 dP
H3K4me3 and H3K27me3H3K27me3 onlyH3K4me3 onlyNo H3K4me3/H3K27me3
D
Figure S2. Genome-Wide Histone Modification Changes in Mammary Epithelial Cells from Pregnant and Ovariectomized Mice, Related to
Figure 3
(A) Read coverage maps and corresponding ChIP-qRT-PCR for H3K27me3 marks on Ccnd2 and Cdkn2a (Ink4a/arf) in the MaSC/basal population. n = 3
independent biological samples for each, error bars represent SEM.
(B) Pregnancy induces derepression of luminal genes in the MaSC-enriched subset invoking ‘‘lineage-priming’’. Quantitative RT-PCR was performed to
determine the levels of b-casein (Csn2), Wap, Elf-5, and Lgr5 mRNA in the MaSC/basal and luminal subsets isolated from virgin (8 weeks old) and 12.5 day
pregnant (12.5 dP) glands of FVB/N mice (n = 3, error bars represent SEM).
(C) Read coverage maps and corresponding ChIP-qRT-PCR for H3K27me3 marks on three luminal genes (Elf5, Csn2, Wap) and the MaSC/basal-specific
gene Lgr5 in the MaSC/basal population. Data are shown for virgin and 12.5 day pregnant (12.5 dP) mice. Coverage graphs show fragments per million on the
scale 0–10. ChIP-qRT-PCR was performed on three independent biological samples; error bars represent SEM.
(D) Genome-wide proportions of genes with histone methylation marks in pregnant or ovariectomized mice. Segmented bar graphs show the percentages of all
genes significantly marked (FDR < 0.05) with H3K4me3 and/or H3K27me3 for MaSC/basal or luminal subsets from virgin, ovariectomized, or pregnant mice.
S6 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
A
E
C
K18
Ezh2f/+ MMTV-cre; Ezh2f/f
p63
D
MM
TV
-cre
; Ezh
2f/f
6 wk 8 wk
8 wk
Ezh
2f/+
lactation (8 d)
F
Brd
U p
ositi
ve c
ells
per
TE
B
4
0
14
12
8
10
6
2
16Ezh2f/+
MMTV-cre; Ezh2f/f
GMMTV-cre; Ezh2f/f
2 dL
MMTV-cre
Ezh
2f/+
12 wk 6 wk
MM
TV
-cre
; Ezh
2f/f
B
MM
TV
-cre
; Ezh
2f/+
16.5 dP12.5 dP
hCD
4
FSC-A FSC-A
FSC-A FSC-A
Lum MaSC/basal
MM
TV
-cre
; Mcl
-1/h
CD
4 (f
/+)
Mcl
-1/h
CD
4 (f
/+) 1.74% 0.20%
83.5% 99.0%
hCD
496.9% 99.3%
12.7% 0.91%
Figure S3. Delayed Mammary Morphogenesis in Ezh2-Deficient Mice, Related to Figures 4 and 5
(A) Efficient MMTV-cre-mediated deletion in luminal and basal mammary epithelial cells based on hCD4 reporter mice. FloxedMcl-1-hCD4 reporter mice in which
human CD4 surface expression is activated upon cre-mediated excision of Mcl-1, thus serving as a reporter of Mcl-1 deletion (Vikstrom et al., 2010). Mcl-1 is
expressed throughout luminal and myoepithelial cells in the mammary gland (unpublished data). MMTV-cre induced effective deletion of the floxed Mcl-1-hCD4
reporter in the luminal (83%) andMaSC/basal (99%) subsets, respectively. MMTV-cre mediated deletion in GTRosa26 reporter mice confirmed activity of MMTV
in mammary epithelial cells but not stroma (data not shown).
(B) Whole-mounted mammary glands from virgin MMTV–cre; Ezh2f/f mice compared to Ezh2f/+ littermate controls. Stunted development was evident in
Ezh2-deficient glands at 6 weeks but generally not at 12 weeks of age. Scale bars, 2.0 mm.
(C) Cell fate appears unchanged in Ezh2-deficient mammary glands. Immunostaining for lineage markers in mammary gland tissue sections from virgin 8-week
old MMTV–cre; Ezh2f/f mice compared to Ezh2f/+ controls. Scale bars, 50 mm.
(D) Npt2b immunostaining of sections fromMMTV–cre; Ezh2f/f glands at 6 and 8weeks of age did not reveal premature alveolar differentiation in the ducts or TEBs
(top panels). A small minority of cells (<2%) were positive for Npt2B staining (middle right panel). Immunostaining of mammary gland sections from Ezh2f/+ mice
for Npt2b at 8 days of lactation (bottom left panel) serves as a positive control. Red = Npt2b, Green = E-cadherin, Blue = DAPI. Scale bars, 50 mm.
(E) Histogram showing the number of BrdU-positive cells per TEB in pubertal mammary glands from MMTV–cre; Ezh2f/f mice compared to Ezh2f/+ mice (n = 3
mice; error bars represent SEM).
(F) H&E sections of mammary glands from MMTV–cre; Ezh2f/+ mice at days 12.5 and 16.5 of pregnancy. Scale bars: 50 mm.
(G) H&E sections of mammary glands at day two of lactation from MMTV–cre; Ezh2f/f and MMTV–cre mice. Scale bars: 100 mm.
Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors S7
A
B
CD
29hi
CD
24+
crecontrolCD29hi CD29lo controls
WT cKOcre - -+ +
WTΔ/Δ
C
GOBPID
GO:0007049GO:0022402GO:0000278GO:0022403GO:0051301GO:0000087GO:0000280GO:0007067GO:0048285GO:0000279GO:0016043GO:0019751GO:0050896GO:0006974GO:0051716GO:0006261GO:0006260GO:0006259GO:0006270GO:0014902GO:0031424GO:0033554GO:0006996GO:0006020
P-value
2.1E-111.3E-101.6E-097.4E-098.0E-098.0E-098.0E-098.0E-091.3E-087.7E-083.3E-055.4E-051.0E-041.7E-042.3E-042.4E-042.6E-042.9E-044.2E-044.2E-044.2E-047.4E-047.5E-048.7E-04
GO Term
cell cyclecell cycle processmitotic cell cyclecell cycle phasecell divisionM phase of mitotic cell cyclenuclear divisionmitosisorganelle fi ssionM phasecellular component organizationpolyol metabolic processresponse to stimulusresponse to DNA damage stimuluscellular response to stimulusDNA-dependent DNA replicationDNA replicationDNA metabolic processDNA replication initiationmyotube differentiationkeratinizationcellular response to stressorganelle organizationinositol metabolic process
GOBPID
GO:0022403GO:0022402GO:0007049GO:0000278GO:0000279GO:0000087GO:0000280GO:0007067GO:0048285GO:0051301GO:0006996GO:0006259GO:0006260GO:0016043GO:0071103GO:0006974GO:0006281GO:0051276GO:0006323GO:0007017GO:0006261GO:0033554GO:0006270GO:0030261
P-value
3.50E-341.49E-331.81E-317.35E-318.94E-309.35E-289.35E-289.35E-282.97E-274.25E-241.41E-203.81E-198.78E-161.33E-125.55E-114.00E-101.00E-093.45E-094.58E-096.55E-081.45E-072.27E-077.52E-077.52E-07
GO Term
cell cycle phasecell cycle processcell cyclemitotic cell cycleM phaseM phase of mitotic cell cyclenuclear divisionmitosisorganelle fi ssioncell divisionorganelle organizationDNA metabolic processDNA replicationcellular component organizationDNA conformation changeresponse to DNA damage stimulusDNA repairchromosome organizationDNA packagingmicrotubule-based processDNA-dependent DNA replicationcellular response to stressDNA replication initiationchromosomal condensation
MaSC/basal Luminal
MMTV-cre;Ezh2f/f
Ezh2f/+
Luminal (12.5 dP)
0
1.2X103
1.0
0.8
0.6
0.2
0.4
0
4.5X103
3.5
2.5
1.5
0.5
MaSC/basalLuminal
virgin 12.5 dP
(re
lativ
e to
18S
rR
NA
)
Arf
virgin 12.5 dP
Ezh
2-de
fi cie
nt s
igna
ture
(M
aSC
/bas
al)
Lum
A
clau
din-
low
HER
2
Lum
B
Basa
l
D
Figure S4. Derepression of Cell Cycle Genes in Ezh2-Deficient Mammary Glands, Related to Figure 5
(A) MaSC/basal (CD29hiCD24+, denoted CD29hi) and luminal (CD29loCD24+, denoted CD29lo) populations fromEzh2f/f micewere transducedwith cre-expressing
or empty control retrovirus in 2D cultures and harvested after 72 hr. PCR analysis of genomic DNA confirmed excision. The top and bottom bands represent the
wild-type and floxed alleles respectively. Ex vivo cre-mediated excision of Ezh2 in the MaSC/basal (and luminal subsets) severely impaired clonogenic capacity.
Shown here are data for the MaSC/basal (CD29hi) population.
(B) Microarray analysis of the MaSC/basal (CD29hi) and luminal (CD29lo) subsets from control (undeleted) and Ezh2-deficient cells following ex vivo excision
revealed significant enrichment of gene signatures related to cell cycle regulation. Comparative analyses of the Ezh2 signature genes fromKamminga et al. (2006)
and Ezhkova et al. (2009) with our gene expression profiles showed that the signatures were significantly differentially expressed in the basal and luminal subsets:
1) for upregulated genes in the MaSC/basal and luminal subsets, p = 0.0009 and 0.0003 respectively (Kamminga et al., 2006) and 2) for up or downregulated
genes in the MaSC/basal and luminal subsets, p = 0.0009 and 0.0099 respectively.
(C) Quantitative RT-PCR analysis of Arf expression in cellular subsets from virgin versus pregnant glands (left panel). qRT-PCR showed that Arf is derepressed in
Ezh2-deficient luminal cells isolated from 12.5 day pregnant mice (right panel) (n = 3 independent experiments; error bars represent SEM).
(D) Ezh2 transcriptional signature by breast cancer tumor subtype. Box plots show the aggregate gene expression score in each tumor for genes associated
with Ezh2-deficiency in the MaSC/basal subset. The Ezh2-deficient expression score is highest in the claudin-low subtype and lowest in the basal and luminal
B subtypes. Gene set testing (Wu et al., 2010) confirms that these comparisons are statistically significant (p = 0.0002 for claudin-low versus basal subtypes and
p = 0.0027 for claudin-low versus luminal B).
S8 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors
A
MM
TV
-cre
; Ezh
2f/f
CD14+ CD14–CD29hi
Col
onie
s pe
r 10
0 ce
lls
10
0
40
30
20
CD14+
CD29lo
CD24+
CD29hi
CD24+
Ezh2f/+
MMTV-cre; Ezh2f/f
B
C
MMTV-cre; Ezh2f/f
CD
24
CD29
CD
24
CD14
CD
24
CD29
CD
24
MMTV-cre; Ezh2f/f
CD14
CD
24
CD29
CD
24
CD14
CD
24
CD29
CD
24
CD14
virg
inpr
egna
nt
CD14_
CD29lo
CD24+
Ezh
2f/+
Ezh2f/+
Ezh2f/+
H3K27Me3
Tubulin
4 w
k
8 w
k6.
5 d
12.5
d16
.5 d
6 w
k
Ezh2
f/+
Virgin Pregnancy
cKO
12.5 dP
D
8 w
k
6 w
k
Ezh2f/+
8 w
k8
wk
H3K27me3
Tubulin
Virgin
MMTV-cre;Ezh2f/f
Figure S5. Ezh2 Deficiency Dramatically Reduces Progenitor Activity among the Epithelial Subsets during Pregnancy, Related to Figure 6
(A) CD24/CD29 and CD24/CD14 flow cytometric plots of lineage-negative mammary epithelial cells (CD45�CD31�Ter119�) from MMTV–cre; Ezh2f/f mice and
Ezh2+/f littermate controls in either the virgin or pregnant (day 12.5) states. CD14 was used to subdivide the luminal population since CD61 is rapidly down-
regulated during pregnancy. In virgin glands, CD14 expression enriches for luminal progenitor activity whereas the CD14� subset is enriched for mature luminal
cells (Asselin-Labat et al., 2011). In pregnancy, the spectrum of CD14 expression is broader, with progenitor activity detectable in all subsets, thus suggesting
a continuum of alveolar and ductal progenitor cells. Ezh2-deficiency leads to diminution of all progenitor activity.
(B) Colony forming capacity of freshly sorted CD29hiCD24+MaSC/basal (denoted CD29hi), CD29loCD24+CD14+ (denoted CD14+), and CD29loCD24+CD14�
(denoted CD14�) luminal subfractions from mid-pregnant MMTV–cre; Ezh2f/f mice or Ezh2f/+ mice, grown on irradiated fibroblast feeders.
(C) Histogram showing the colony forming capacity of each subpopulation. Data represent the mean of two independent biological experiments with eight
replicates for each.
(D) Global diminution of H3K27me3 protein in Ezh2-deficient mammary epithelial cells: Western blot analyses of total H3K27me3 protein and tubulin in lysates
from Ezh2-deficient (MMTV-cre; Ezh2f/f) versus littermate control mammary glands (Ezh2f/+) (upper panel). Loss of trimethylated H3K27 protein in Ezh2-deficient
glands at 12.5 days pregnancy (lower panel, right lane).
Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors S9
FSC-H
FS
C-W
SSC-HS
SC
-WFSC-A
PI
FSC-A
Line
age
Sca
1-A
PC
CD49b-FITC
CD
24-P
E
CD29-APC-Cy7
Sca
1-A
PC
CD49b-FITCC
D24
-PE
CD29-APC-Cy7
Oil Pg
Mat Lum Lum Prog (PR+)
Lum Prog (PR–)
Figure S6. Isolation of Hormone Receptor-Positive and Negative Luminal Progenitor Populations from Young Adult Mammary Glands,
Related to Figure 6
CD24/CD29 and CD49b/Sca-1 flow cytometric plots of lineage-negative mammary epithelial cells (CD45�CD31�) from 8 week-old mice. CD49b/Sca-1 flow
cytometric plots are shown for lineage-negative CD29loCD24+ cells derived from mice treated with progesterone or vehicle (oil) for 48 hr. The three luminal
subsets used for qRT-PCR analysis are depicted: mature luminal, PR+, and PR� luminal progenitor cells.
S10 Cell Reports 3, 411–426, February 21, 2013 ª2013 The Authors