The transcriptional coactivator TAZ
regulates mesenchymal differentiation
in malignant glioma
Krishna P.L. Bhat,1,6,11Katrina L. Salazar,1,6Veerakumar Balasubramaniyan,2,5,6Khalida Wani,1
Lindsey Heathcock,1Faith Hollingsworth,1Johanna D. James,1Joy Gumin,3Kristin L. Diefes,1
Se Hoon Kim,1,7Alice Turski,1Yasaman Azodi,1Yuhui Yang,3Tiffany Doucette,3Howard Colman,2,8,9
Erik P. Sulman,4Frederick F. Lang,3Ganesh Rao,3Sjef Copray,5Brian D. Vaillant,2,10
and Kenneth D. Aldape1
1Department of Pathology,2Department of Neuro-Oncology,3Department of Neurosurgery,4Department of Radiation Oncology,
The University of Texas, M.D. Anderson Cancer Center, Houston, Texas 77030, USA;5Department of Neuroscience, University
Medical Center Groningen, 9713 AV Groningen, The Netherlands
Recent molecular classification of glioblastoma (GBM) has shown that patients with a mesenchymal (MES) gene
expression signature exhibit poor overall survival and treatment resistance. Using regulatory network analysis of
available expression microarray data sets of GBM, including The Cancer Genome Atlas (TCGA), we identified the
transcriptional coactivator with PDZ-binding motif (TAZ), to be highly associated with the MES network. TAZ
expression was lower in proneural (PN) GBMs and lower-grade gliomas, which correlated with CpG island
hypermethylation of the TAZ promoter compared with MES GBMs. Silencing of TAZ in MES glioma stem cells
(GSCs) decreased expression of MES markers, invasion, self-renewal, and tumor formation. Conversely, over-
expression of TAZ in PN GSCs as well as murine neural stem cells (NSCs) induced MES marker expression and
aberrant osteoblastic and chondrocytic differentiation in a TEAD-dependent fashion. Using chromatin immuno-
precipitation (ChIP), we show that TAZ is directly recruited to a majority of MES gene promoters in a complex
with TEAD2. The coexpression of TAZ, but not a mutated form of TAZ that lacks TEAD binding, with platelet-
derived growth factor-B (PDGF-B) resulted in high-grade tumors with MES features in a murine model of glioma.
Our studies uncover a direct role for TAZ and TEAD in driving the MES differentiation of malignant glioma.
[Keywords: HIPPO; TAZ; TEAD; glioma; mesenchymal]
Supplemental material is available for this article.
Received August 16, 2011; revised version accepted November 9, 2011.
Glioblastoma (GBM) is the most common and aggres-
sive form of glioma responsible for nearly 60% of malig-
nant primary brain tumors (Furnari et al. 2007; Huse and
Holland. 2010). Although all GBMs share similar histo-
logical features, such as microvascular proliferation and
pseudopalisading necrosis, patients present with differ-
ential treatment response and survival rates that can be
predicted based on molecular determinants (Burton et al.
2002; Freije et al. 2004; Hegi et al. 2005; Nigro et al. 2005;
Phillips et al. 2006; Verhaak et al. 2010). In particular,
recent work has shown that patients whose tumors have
a signature enriched in genes associated with neural de-
velopment (proneural [PN]) have better survival com-
pared with those that have signatures resembling the
mesenchyme (mesenchymal [MES]) (Phillips et al. 2006;
Colman et al. 2010). Although the genetic abnormalities
associated with these signatures are not fully understood,
studies by The Cancer Genome Atlas (TCGA) consortium
point to alterations of Neurofibromatosis-1 (NF1) and
platelet-derived growth factor receptor-A (PDGFRA)/
isocitrate dehydrogenase 1 (IDH1) as defining features of
MES and PN subtypes, respectively (Verhaak et al. 2010).
Gene regulatory network analyses identified the tran-
scription factors (TFs) signal transducer and activator of
transcription 3 (STAT3) and CCAAT enhancer-binding
protein-b (C/EBP-b) as synergistic initiators and master
regulators of the MES signature in glioma (Carro et al.
2010). Silencing these TFs collapsed the MES network
and led to improved survival in mice implanted with
6These authors contributed equally to this work.
Present addresses:7Department of Pathology, College of Medicine, Yonsei
University, Seoul 120-752, Korea;8Department of Neurosurgery, Univer-
sity of Utah, Salt Lake City, UT 84132, USA;9Huntsman Cancer Institute,
Salt Lake City, UT 84132, USA;10The Methodist Hospital Neurological
Institute, Houston, TX 77030, USA.
Article is online at http://www.genesdev.org/cgi/doi/10.1101/gad.176800.111.
2594GENES & DEVELOPMENT 25:2594–2609 ? 2011 by Cold Spring Harbor Laboratory Press ISSN 0890-9369/11; www.genesdev.org
tumor cells lacking these TFs. While this study repre-
sents the first comprehensive mapping of the transcrip-
tional control of the MES signature, additional TFs could
be uncovered by subjecting broader data sets such as
TCGA to master regulatory analysis.
MES differentiation is a well-studied phenomenon in
a variety of solid tumors, including breast carcinoma
(Kalluri and Weinberg 2009; Thiery et al. 2009). Epithelial-
to-MES transition (EMT) is a dynamic and reversible
cellular event in which cells lose epithelial features, such
as polarity and intercellular junctions, and acquire MES
characteristics, leading to increased migration and in-
vasion (Kalluri and Weinberg 2009; Thiery et al. 2009).
Cancer cells that have undergone EMT exhibit enhanced
metastatic potential and acquire stem cell-like properties
(Mani et al. 2008; Gupta et al. 2009). These events are pre-
dominantly orchestrated by TFs (snail homolog 1 [SNAI1],
SNAI2, twist homolog 1 [TWIST1] forkhead box C2
[FOXC2], zinc finger E-box-binding homeobox-1 [ZEB-1],
and ZEB-2) or microRNAs (miR-141, the miR-200 family,
and miR-205), many of which are integral components of
early embryonic development and pathological condi-
tions that also require EMT (Gregory et al. 2008; Park
et al. 2008; Thiery et al. 2009). Until recently, however,
the presence of a MES component in tumors such as GBM
had not been demonstrated (Phillips et al. 2006; de Groot
et al. 2008; Verhaak et al.2010).If the proposed cell of origin
of these tumors is a neural stem cell (NSC) or oligoden-
drocyte precursor cell (OPC) (Alcantara Llaguno et al. 2009;
Ward et al. 2009; Liu et al. 2011; Sugiarto et al. 2011), it is
intriguing how most GBMs exhibit MES characteristics.
Furthermore, it has been shown that PN GBMs tend to
One hypothesis is that aberrant activation of TFs occurring
during GBM progression and/or recurrence can trigger a
global MES shift. This is a well-established phenomenon
in breast cancer progression, where overexpression of
EMT inducers, SNAI1, TWIST1, or FOXC2 is sufficient
to cause metastasis to distant organs (Yang et al. 2004;
Manietal.2007;Kudo-Saito etal.2009). Thus,TFs playing
a role in EMT in other tumor types could be candidate
master regulators of MES differentiation in GBM.
Transcriptional coactivator with PDZ-binding motif
(TAZ), also known by gene name WW domain-containing
tor that plays pivotal roles in EMT, cell growth, and organ
development (Hong and Yaffe 2006; Zhao et al. 2010).
TAZ functions by transactivating numerous TFs, includ-
ing runt-related transcription factor 2 (RUNX2), paired
box-3 (PAX3), PAX8, transcription termination factor-1
(TTF-1), T-box 5 (TBX5), mothers against decapentaplegic
homologs (SMADs), and TEA domain family members
(TEAD) (Hong et al. 2005; Murakami et al. 2005, 2006;
Varelas etal. 2008; DiPalma et al.2009; Zhanget al.2009).
The regulation of TAZ and its paralog, Yes-associated
protein (YAP), occurs primarily via HIPPO tumor sup-
pressor pathway, many components of which are con-
served from flies to humans (Zhao et al. 2008a; Pan 2010;
Halder and Johnson 2011). Autophosphorylation of mam-
malian sterile 20-like-1 (MST-1) and MST-2 (the verte-
brate homolog for HIPPO) in complex with its regulatory
protein, salvador-1 (SAV1), triggers phosphorylation and
activation of large tumor suppressor homolog-1 (LATS-1)
and LATS-2, which in turn phosphorylate TAZ on four
serine residues (S66, S89, S117, and S311) (Zhao et al.
2008a; Pan 2010; Halder and Johnson 2011). This kinase
cascade causes cytoplasmic sequestration of phospho-
TAZ by 14-3-3-e. Phosphorylation of TAZ by LATS on
Ser 311 also primes subsequent phosphorylation by casein
kinase I (CKId/e), which induces recruitment of SCFb-TRCP
E3 ligase, leading to the ubiquitination and degradation of
TAZ (Liu et al. 2010). Recently discovered membrane
sequestration of TAZ/YAP by PDZ domain-containing
tight junction proteins adds another complex layer of con-
trol of these crucial transcription coactivators (Duning
et al. 2010; Remue et al. 2010; Varelas et al. 2010b; Chan
et al. 2011; Zhao et al. 2011). Upon translocation to the
nucleus, TAZ induces cell proliferation, migration, in-
vasion, and EMT (Lei et al. 2008; Zhang et al. 2009).
Silencing TEAD or even preventing the TAZ–TEAD in-
teraction blocks the ability of TAZ to promote cell pro-
liferation and EMT, implying TEAD as a major facilitator
of TAZ functions. Thus, restraining TAZ interaction with
TEAD may be an important step in preventing aberrant
events central to cancer progression.
The current study was specifically initiated to identify
candidate TFs/cofactors that are causally associated with
the MES signature. We employ gene network analyses of
TCGA data sets, mimic glioma-associated alterations
in patient-derived glioma stem cells (GSCs) and murine
models of glioma, and show that reprogramming of other-
wise PN GSCs to a MES state can be achieved by over-
expression of a single transcriptional coactivator.
Identification of TAZ association with the MES
network, grade IV gliomas, and its epigenetic control
We subjected microarray expression data sets from TCGA
(n = 385) to the previously described Algorithm for the
to identify novel TFs/cofactors associated with the MES
gene signature (Margolin et al. 2006). For the initial screen,
we used an expanded list of transcriptional regulators that
included cofactors associated with transcription (Supple-
mental Table S1A) that were not part of a previous study
(Carro et al. 2010). We next generated a subnetwork con-
taining known GBM MES genes (Supplemental Table
S1B) as defined by two independent groups (Phillips et al.
2006; Verhaak et al. 2010)and the associatedTFs/cofactors
as predicted by ARACNE. Henceforth, we refer to the two
data sets simply as Phillips and Verhaak. We subsequently
filtered for the higher-order ‘‘hub’’ (i.e., most connected)
TFs/cofactors, which represent critical regulators in a
scale-free network model. Our analyses revealed ;70%
overlap with previously identified candidates, including
STAT3, C/EBP-b, andother TFs thathad strong correlation
with the MES network as either activators or repressors
(Supplemental Table S1C; Carro et al. 2010). The unique
TAZ and mesenchymal malignant glioma
GENES & DEVELOPMENT2595
transcriptional regulators that were identified to be
positively correlated with the MES signature in our study
included v-maf musculoaponeurotic fibrosarcoma onco-
gene homolog B (MAFB), hematopoietic cell-specific Lyn
substrate 1 (HCLS1), and the HIPPO pathway transcrip-
tion cofactors TAZ and YAP (Supplemental Table S1C).
While a role for MAFB and HCLS1 in inducing MES
differentiation cannot be ruled out, TAZ and YAP have
previously been demonstrated to induce EMT in other
cancer types (Lei et al. 2008; Zhao et al. 2008b). Further-
more, for reasons that will become obvious, we chose to
characterize TAZ over YAP. TAZ inferred MES network
targets were relatively nonoverlapping with those of
STAT3 and C/EBP-b (Fig. 1A,B; Supplemental Fig. S1A,B;
Supplemental Table S1D). Intrigued by this, we were
curious what other types of functions were predicted to
be regulated by TAZ in the overall ARACNE-generated
GBM regulatory network. Thus, we generated a list of
genes predicted to be regulated by TAZ and analyzed this
using the Database for Annotation, Visualization, and
Integrated Discovery tool (DAVID), which distills a list of
genes to biologically meaningful activities (Dennis et al.
2003; Huang da et al. 2009). This analysis further sup-
ported TAZ as playing a role in MES activities (e.g., wound
response) and immunologic functions (Fig. 1C). In concor-
dance with this, the expression of TAZ in the TCGA data
set was strongly correlated with a MES metagene score
generated from the union of the Phillips and Verhaak MES
genes (R2= 0.48) (Supplemental Fig. S1C). Thus, in silico
analysis uncovered a positive association of TAZ with the
Since methylation of CpG islands of gene promoters can
influence their expression (Deaton and Bird 2011), we
analyzed the methylation status of TAZ and other com-
ponents of the HIPPO pathway (Supplemental Fig. S1D)
from TCGA data sets. The CpG island of TAZ (Fig. 1D,E;
Supplemental Fig. S1E) was dramatically hypermeth-
ylated in the PN subgroup compared with MES tumors.
YAP, LATS2, and MST1 also appeared methylated in
PN tumors, albeit to a less significant extent (Fig. 1D;
Supplemental Fig. S1F). If the TCGA tumors are stratified
by Verhaak subtypes, the TAZ promoter is hypermeth-
ylated only in the PN group, but no obvious differences
were seen in the YAP promoter methylation across all
four subtypes (Supplemental Fig. S1G,H). We next com-
pared TAZ methylation across grades and found higher
methylation frequencies in grade II and grade III gliomas
(86% and 75%, respectively) compared with only 30% in
grade IV tumors (Pearson’s x2test, P = 2.2?16) (Fig. 1F).
Since lower-grade gliomas are typically PN in nature (Li
et al. 2009; Cooper et al. 2010), whereas GBMs tend to be
both PN and MES, we view the association of TAZ
methylation with lower grade as reflective of the gene
expression signature rather than differences in grade.
Consequently, TAZ expression was lower in grades II/III
when compared with grade IV gliomas (P < 0.001) (Fig.
1G), as well as in long-term (>1 year) versus short-term
survivors (P < 0.001) (Fig. 1H). Consistent with gene
expression patterns, TAZ protein was higher in grade IV
when compared with grade II/III tumors (Supplemental
Fig. S1I). A similar pattern emerged for MES marker
fibronectin1 (FN1), YAP, and TEAD4, but not other TEAD
family members (Supplemental Fig. 1I). Interestingly,
LATS2 and MST1, whose promoters were methylated in
the PN subgroup, showed higher expression in grade IV
gliomas, but activation of these kinases was not seen, as
evidenced by weak induction of p-MST1 in all grades of
glioma and p-LATS1/2 in only one out of the eight GBMs
tested (Supplemental Fig. 1I). This, in turn, could facili-
tate TAZ localization to the nucleus. All other HIPPO
pathway components failed to exhibit notable differences
between grades (Supplemental Fig. 1I). Because FN1 is also
associated with endothelial cells (Martinez et al. 1994),
and to delineate whether the increased FN1 expression
was a consequence of MES shift or increased vasculature,
we performed immunohistochemistry (IHC) on paraffin-
embedded sections. A membranous pattern of staining on
tumor cells was observed in grade IV tumors, but not in
grade II, where only endothelial cells expressed FN1
(Supplemental Fig. S1J).
We next analyzed TAZ mRNA expression and its cor-
relation with survival in the four subtypes of GBM in the
TCGA data sets using a Cox proportional hazards model.
In the Verhaak-called subtypes (see the Supplemental
Material), overall survival was predicted by TAZ expres-
sion only in the PN group (Supplemental Table S2). The
reason for this is unclear, although this could be partially
attributed to the PN group having the widest range of
TAZ expression, and so its influence is more readily seen.
On the contrary, when using calculated Verhaak subtypes
(see the Supplemental Material), TAZ expression pre-
dicted overall survival in both the PN and MES groups
(Supplemental Table S2). Like the PN group as deter-
mined by either method, the calculated MES group had
a wider spread of TAZ expression.
To assess the clinical impact of subcellular localization
of TAZ, we tested 187 gliomas for the expression of TAZ
by IHC. TAZ was readily detected in the nucleus in
a large subset of the cases that were predominantly grade
IV, whereas most grade II gliomas failed to express TAZ
(representative images shown in Fig. 1I). In cases where
TAZ was expressed in both nucleus and cytoplasm
(score = 2), patients showed significantly reduced sur-
vival (P = 0.007) (Fig. 1J) compared with cases that had
either one or the other (score = 1) or neither (score = 0).
This is not entirely surprising, since TAZ has been shown
to promote Wnt/b-catenin signaling in the cytoplasm
(Varelas et al. 2010a), raising the possibility that cyto-
plasmic TAZ could still promote gliomagenesis by alter-
nate mechanisms. However, nuclear TAZ expression
strictly correlated with high CD44 cases, arguing that
nuclear TAZ aligns with MES subtypes of tumors (Sup-
plemental Fig. S1K). To rule out TAZ expression by
nontumorigenic cell types that are part of the microen-
vironment, we costained TAZ with a monoclonal anti-
body (J1-31) that specifically detects reactive astrocytes
(Ridet et al. 1997). Tumor areas that showed nuclear TAZ
expression lacked J1-31 staining and vice versa (Supple-
mental Fig. S1L). Taken together, these data imply that
TAZ is intricately connected to the MES network and is
Bhat et al.
2596 GENES & DEVELOPMENT
from the ARACNE analysis limited to the union of Phillips and Verhaak MES target genes (n = 281). Nodes are color-coded to show the
membership of a given gene to a regulatory network. (B) Venn diagram of gene overlap of the TAZ, C/EBP-b, and STAT3 networks. The
number in parentheses is the total number of genes in the ARACNE network for the given TF. (C) Bar graph depicting enrichment of
genes possessing MES properties in the initial ARACNE network. Functional categories showing significant enrichment based on log10
of the P-value (line) and counts (bars) are shown. (D) Methylation status of TAZ and YAP in 62 PN (blue) or 147 MES (red) GBMs from
TCGA data set (Illumina Infinium platform). The black bar is the mean of the methylation b-score. Two-sample t-test between the
groups was performed to assess statistical significance. (E) Correlation of TAZ expression with TAZ methylation status. Two-hundred-
nine GBMs from TCGA data set with both expression data (Affymetrix platform) and methylation data (Illumina Infinium platform) are
plotted. GBMs were color-coded as either MES (red) or PN (blue) based on the composite metagenes, as defined by Phillips et al. (2006)
and Verhaak et al. (2010). (F) Bar graphs showing the frequency of methylation on TAZ CpG sites across various grades of glioma. Red
bars indicate unmethylation and blue bars indicate methylation. Pearson’s x2test was used to estimate significance. (G) TAZ expression
in multi-institutional microarray data sets (;800). Bar graphs indicate normalized mean probe intensity of TAZ across grades. (H) Bar
graphs indicate normalized mean probe intensity of TAZ across survivors. Long-term survivors (LTS) were defined as those who lived
longer than 2 years, and those who survived less than that were short-term survivors (STS). Student’s t-test values are shown for
significance. (I) Representative IHC images (403) of TAZ expression. (J) Kaplan-Meier survival analysis of TAZ expression based on the
IHC staining pattern. Tumors lacking TAZ expression were scored 0, and staining in the nucleus or cytoplasm was scored 1, whereas
tumors showing both nuclear and cytoplasmic staining were scored 2. Log rank test values are shown.
Association of TAZ to the MES network, the MES subclass of gliomas, and its epigenetic control. (A) Subnetwork generated
TAZ and mesenchymal malignant glioma
GENES & DEVELOPMENT 2597
regulated by promoter methylation and that its nuclear
expression correlates with the MES signature, higher
grade, and worse overall survival in gliomas.
TAZ is required for MES transition and aggressive
gliomagenesis in GSCs
Studies in our laboratory have shown a dichotomous
nature of patient-derived GSCs—based on their gene ex-
pression signatures—that mirror the GBM subtypes of
PN and MES (KPL Bhat and KD Aldape, unpubl.). We
tested theexpression of TAZ inthese GSCs and compared
them against TAZ promoter methylation status. TAZ
expression was undetectable or significantly lower in
PN GSCs (GSC7–11, GSC8–11, GSC11,andGSC23) when
compared with those deemed MES (GSC6–27 and GSC20)
(Fig. 2A). A similar pattern was seen for MES marker
CD44 (Fig. 2A; Schieker et al. 2004; Schieker et al. 2007).
YAP expression did not necessarily correlate with the PN
or MES status of the GSCs (Fig. 2A). For example, a PN
line (GSC23) had levels of YAP expression equal to
GSC20, a MES line (Fig. 2A). Thus, TAZ, but not YAP,
expression parallels the MES nature of GSCs, similar to
clinical observations. Moreover, using cell fractionation,
we show that nuclear TAZ was undetectable in PN GSCs
compared with MES GSCs that exhibited dramatically
higher levels of TAZ in the nucleus (Supplemental Fig.
S2A). Cytosolic TAZ was also higher in the MES GSCs,
but to only modest levels compared with PN GSCs. This
could be attributed to increased amounts of active MST1
(phosphorylated form) in MES GSCs. Based on these con-
siderations, and given that YAP is not a predictor of
survival in the TCGA data set (Supplemental Fig. S2B),
we chose TAZ for further characterization. Bisulfite
sequencing of the TAZ promoter revealed hypermeth-
ylation in GSC7–11 and GSC8–11 and a converse hypo-
methylation in GSC6–27 and GSC20 (Fig. 2B). Treatment
with the demethylation agent 5-aza-29-deoxycytidine
(DAC) induced TAZ expression by greater than twofold
in GSC8–11 (Fig. 2C), whereas it failed to have an effect in
GSC20 (data not shown). Interestingly, expression of
CD44 and CTGF was also increased upon DAC treatment
(Fig. 2C), implying that induction of TAZ parallels ex-
pression of MES genes. Our findings uncover a previously
unknown epigenetic regulation of TAZ in gliomas.
To directly test whether TAZ is required for the expres-
sion of MES markers, we performed transient knockdown
of TAZ using siRNA. Silencing TAZ for 72 h led to dra-
matic reduction of its protein levels with concomitant
decreases in CD44 and FN1 (Fig. 2D). Encouraged by this
markers, invasion, self-renewal, and tumor initia-
tion in GSC20. (A) Western analyses of basal YAP,
TAZ, and CD44 levels in GSCs. (B) DNA methyla-
tion analysis by bisulfite conversion followed by
sequencing of cloned PCR products. (Filled circle)
Methylated CG pair; (white circle) unmethylated
CG pair. Each column corresponds to 11 CpG sites
proximal to the transcription start site of TAZ
(shown on top). Each row represents the methylation
status of a clone. (C) Real-time qPCR of TAZ, CD44,
and CTGF expression in GSC8–11 after treatment
with 5 mM DAC for 72 h. Fold change before and
after treatment was used for comparison of tran-
script levels and assessed using a t-test. (D) Western
analysis of TAZ, FN1, and CD44 after transient
knockdown of TAZ in GSC2 and GSC20. Cells were
cultured in laminin and poly-L-ornithine-coated
plates and transfected with siRNA for 48 h prior to
immunoblotting. (E) Western analysis of stable TAZ
knockdown clones showing reduction of TAZ but
not YAP levels. (F) Invasion assay of TAZ stable
knockdown clones in GSC20. The invasion effi-
ciency of the nontargeting controls was set to
100% for comparison. P-values were generated using
a t-test for significance. (G) Neurosphere assay of
TAZ knockdown clones. Bar graphs indicate per-
centage of neurosphere formation, and P-values are
shown for significance. (H) Representative hematox-
ylin and eosin-stained brain images from SCID mice
injected intracranially with the TAZ stable knock-
down clones. The bottom row shows higher magni-
fication (403) of tumors on the top row (103). (I)
Kaplan-Meier analysis of tumor-free progression for
the TAZ knockdown clones compared with the
TAZ is required for expression of MES
Bhat et al.
2598GENES & DEVELOPMENT
finding, we sought to analyze the phenotypic consequence
of stable silencing of TAZ using shRNA. Western analysis
of two independent GSC20 clones targeting nonoverlap-
ping regions of the TAZ transcript showed a significant
control, but YAP levels were unaltered (Fig. 2E). In line
with previous reports, silencing TAZ caused reduced in-
vasion across a chemotactic gradient compared with con-
trol GSC20 cells (Fig. 2F). Furthermore, self-renewal of
GSC20 as measured by neurosophere formation was also
reduced in both knockdown clones when compared with
the nontargeting control (Fig. 2G). These data imply that
TAZ is required for the invasive ability and self-renewal of
GSC20. We then compared the tumor-initiating capacity
of these clones using an orthotopic intracranial model
using severe combined immunodeficiency (SCID) mice.
All of the mice injected with control GSC20 (n = 5) de-
veloped tumors characterized by pseudopalisading necro-
sis and microvascular proliferation (Fig. 2H). In contrast,
none of the mice injected with shTAZ-1 and only two out
of five of the mice injected with shTAZ-2 formed tumors,
2H,I). Histologically, the tumors that did form in shTAZ-2
CD44 as detected by IHC (Supplemental Fig. S2C). In sum,
TAZ expression is regulated by CpG island methylation in
GSCs and is required for expression of MES markers,
invasion, self-renewal, and tumor initiation in GSCs.
The TAZ–TEAD complex mediates MES transition
in GSCs and murine primary NSCs
Next, we asked whether overexpressing TAZ is sufficient
to induce MES transition in otherwise PN GSCs. TAZ
interacts with numerous TFs, but TEAD and RUNX2
play prominent roles in TAZ-mediated EMT and osteo-
genic differentiation, respectively (Hong et al. 2005; Zhang
et al. 2009), and thus represent candidate TFs that could
mediate TAZ-induced MES differentiation. To test this
hypothesis, we used previously characterized mutants of
TAZ (Lei et al. 2008; Zhang et al. 2009). Four serine
residues (S66, S89, S117, and S311) substituted to alanine
results in unphosphorylated TAZ that is constitutively
nuclear (termed 4SA). An additional point mutation in
the TEAD-binding domain (S51 to alanine) results in TAZ
that lacks TEAD binding (4SA-S51A). We transduced
a retrovirus-expressing Flag-tagged vector, 4SA, or 4SA-
S51A, into GSC11 and generated stable clones (Fig. 3A).
Consistent with previous reports, immunofluorescence
(IF) analysis revealed predominant nuclear localization of
transition in GSCs. (A) Western analysis of GSC11
clones overexpressing vector, 4SA, or 4SA-S51A show-
ing Flag expression. Retrovirus encoding pBabe vector
or TAZ mutants was used for transduction of GSC11.
Stable clones were generated by culturing transduced
cells in puromycin. (B) Immunoprecipitation of TAZ
using Flag antibody followed by Western analysis of
TAZ, TEAD4, and RUNX2 in 4SA- and 4SA-S51A-
expressing cells. (C) Western analyses of MES pro-
teins in GSC11 clones expressing vector, 4SA, and
4SA-S51A. (D) Invasion assay of GSC11 clones across
Matrigel-coated transwell plates. Bar graphs indicate
percentage of invasion. A t-test was used to compare
statistical differences. (E) GSC11 clones were sub-
jected to EdU incorporation assay; cells were plated
on laminin/poly-L-ornithine-coated chamber slides,
fixed, and counterstained with DAPI; and Alexa
488-positive cells were quantified using a fluores-
cent microscope in 10 independent fields. Bar graphs
indicate mean values. A t-test was used to assess
statistical significance. (NS) Not significant. (F)
Western analyses of cell cycle proteins in GSC11
clones. (G) Western analyses of MES proteins after
stable knockdown of TEADs. Two independent
stable knockdown pools (shTEAD-a and shTEAD-b)
were generated using pGipZ lentivirus. Both clones
appearedto silence TEAD2
tively. (H) Osteoblast induction of GSC11 clones.
Cells were cultured in differentiation medium for
3–4 wk, fixed in 70% ethanol, stained with Alizarin
Red for assessing intercellular calcium deposition,
and photographed. (I) Chondrocyte induction of
vector control and 4SA. Cells were grown as pel-
The TAZ–TEAD interaction mediates MES
lets in chondrocyte medium for 4 wk, fixed in buffered 10% formalin, and embedded in paraffin. Five-micron were slide-
mounted and stained for glycosaminoglycans with Safranin O.
TAZ and mesenchymal malignant glioma
GENES & DEVELOPMENT2599
4SA, whereas 4SA–S51A showed both cytoplasmic and
nuclearexpression(data not shown).Immunoprecipitation
using Flag antibody showed specific interaction of 4SA
with both TEAD4 and RUNX2 (Fig. 3B). 4SA was also
found to interact with TEAD2, suggesting that TAZ can
bind to other TEAD family TFs with equal efficiency in
GSCs (data not shown). On the contrary, the 4SA-S51A
mutant retained binding only to RUNX2, but not TEAD4.
Immunoblotting showed dramatic induction of MES
markers—namely,connective tissuegrowthfactor (CTGF),
CD44, caveolin 2 (CAV2), and FN1—in 4SA, but not 4SA-
S51A (Fig. 3C). The increase in MES markers was also
observed in GSC7–11 and GSC8–11 that lack TAZ ex-
pression (Supplemental Fig. 3). Functionally, GSC11 over-
expressing 4SA showed increased invasion, whereas 4SA-
S51A showed efficiency similar to vector controls (Fig.
3D). Previous studies have shown that Yorkie (Yki), the
Drosophila homolog of TAZ/YAP, promotes cell prolifer-
ation via activation of cyclin E (Huang et al. 2005). To test
whether this effect was conserved in glial tumors, we
assessed cell proliferation rates using EdU incorporation
assay, but found no significant differences between the
GSC11 clones (Fig. 3E). In addition, the expression of cell
cycle regulatory proteins p-CDK1, p-CDK4, Cyclin A,
Cyclin B1, and Cyclin E did not significantly differ be-
tween the groups (Fig. 3F). Thus, at least in GSCs, the
primary event in response to TAZ activation and signal-
ing via TEAD appears to be MES differentiation, rather
than enhanced proliferation as observed in other cell
types (Lei et al. 2008). Stable knockdown of TEAD2 and
TEAD4 in 4SA also resulted in decreased MES protein
expression, providing further evidence that TEAD is an
important player in TAZ functions (Fig. 3G).
Previous studies have shown that TAZ is highly ex-
pressed in MES stem cells (MSCs) and promotes osteo-
genic differentiation via interaction with RUNX2 (Hong
et al. 2005). Because GSC11 expressing 4SA-S51A retains
binding to RUNX2, but lacks MES properties, we asked
whether TAZ–TEAD interaction mediates osteogenic
differentiation of GSCs. GSC11 clones were cultured in
osteogenesis induction medium for 30 d and stained for
alizarin red to assess calcium chelation (Meloan et al.
1972). No staining was observed in vector control or 4SA-
S51A, whereas 4SA-overexpressing cells showed dra-
matic red coloration (Fig. 3H). Cartilage-specific proteo-
glycan expression, as determined by Safranin O staining
(Rosenberg1971), wasalsoinducedin 4SAwhen compared
with vector transfectedcells (Fig.3I). 4SA-S51A-expressing
cells failed to form a detectable pellet in chondrogenic
medium and hence could not be assessed in this assay.
The induction of bone and cartilage marker expression by
TAZ in GSCs is consistent with a robust MES transition
occurring via a TAZ–TEAD interaction in these cells.
Because GSCs contain numerous genetic alterations
that could cooperate with TAZ to induce MES differen-
tiation, we asked whether TAZ can reprogram primary
cells of neural origin into MES lineages. To address this,
we isolated NSCs from embryonic day 14 (E14) mouse
embryonic brain telencephalon and cultured them as
neurospheres. Immunoblotting revealed very low levels
of basal TAZ expression, but TEAD2 and TEAD4 where
highly expressed in NSCs at levels comparable with
GSC20 (Supplemental Fig. S4A). We generated wild-type
and mutant TAZ-overexpressing stable clones of NSCs
by retroviral transduction. 4SA-expressing NSCs appeared
larger, showed fibroblast-like morphology upon attach-
ment (Fig. 4A), and grew at slower rates (data not shown)
than the other NSC clones, suggesting a growth inhibitory
effect of constitutive TAZ activation. However, basal
levels of nestin expression were not altered significantly
among the NSC clones under self-renewing conditions
tiation in murine NSCs. (A) NSCs were isolated from E14
mouse brains and transduced with vector, wild-type, or mutant
TAZ, and stable pools were generated using puromycin. (Top
panel) Phase-contrast images of neurospheres of corresponding
NSC stable pools. (Middle panel) Phase-contrast images of NSCs
after 1 d of differentiation in mitogen-free medium. (Bottom
panel) Dissociated NSC clones were stained for NSC marker
nestin (red), and the nuclei were stained with Hoeschst dye.
Bars, 100 mm. (B) Differentiation of TAZ-overexpressing NSCs
in 5% serum-containing medium. The top panel shows dual
staining for GFAP (green) and FN1 (red). The bottom panel shows
SMA (green) staining in these clones. Bars, 100 mm. (C) Quanti-
fication of expression of various markers in NSCs. For each
experiment, FN1-, GFAP-, or SMA-positive cells were counted in
10 randomized microscopic fields. Bars indicate the mean value
(6SD) for at least two to three independent experiments. A t-test
was used to assess statistical significance. (*) P < 0.05; (**) P <
0.005; (NS) not significant.
The TAZ–TEAD interaction mediates MES differen-
Bhat et al.
2600 GENES & DEVELOPMENT
(Fig. 4A). In response to environmental cues such as
withdrawal of growth factors or presence of serum, NSCs
differentiate into neurons, astrocytes, and oligodendro-
cytes (Gage et al. 1995). To test whether TAZ interferes
with these processes, we cultured NSC clones in the
presence of serum for 6 d and analyzed expression of
various proteins by IF. The astrocytic marker glial fibril-
lary acidic protein (GFAP) (Takizawa et al. 2001) was
expressed in >60% of the cells transduced with a vector,
whereas TAZ- or 4SA-expressing cells showed signifi-
cantly lower percentages (10%–20%) (Fig. 4B,C). Con-
versely, TAZ and 4SA clones showed dramatic induction
of MES markers FN1 and smooth muscle-specific a-actin
(SMA)(Nakajima et al.1997)—insomecases,uptoan80%
of MES markers was abolished and astrocytic differentia-
tion was restored in clones expressing 4SA-S51A. Similar
results were obtained in independent differentiation ex-
periments upon withdrawal of growth factors (Supplemen-
tal Fig. S4B). These experiments suggest that the TAZ–
TEAD interaction causes aberrant MES reprogramming
of primary NSCs at the expense of normal differentiation
toward neural lineages.
Global MES differentiation is induced by TAZ–TEAD
via direct recruitment to target promoters
We next asked whether the TAZ/TEAD complex directly
activates a global MES transcriptional program by target
promoter occupancy. First, to rule out a plausible hierar-
chical transcription module between TAZ, STAT3, and
C/EBP-b, and therefore indirect activation of MES genes
by TFs (such as STAT3 and C/EBP-b) that could signal
downstream from TAZ, we silenced these TFs individu-
ally in GSC20 and tested their protein levels by Western
analysis. Despite significant reduction of all three TFs,
none of them individually altered the expression of the
other two TFs, indicating a lack of transcriptional in-
terdependency between TAZ, STAT3, and CEBP-b (Sup-
plemental Fig. S5A). Hypothetically, TAZ may also cause
MES differentiation via CTGF, a direct downstream target
of TAZ (Zhang et al. 2009) that by itself has been proven
to induce EMT in other cell types (Gore-Hyer et al. 2002;
Burns et al. 2006). To rule out this possibility, we silenced
CTGF in GSC11 overexpressing 4SA, but found no alter-
ation in the expression of MES markers (Supplemental Fig.
S5B). These data are consistent with the hypothesis that
TAZ is a direct inducer of MES differentiation in GSCs.
To test this, we subjected GSC11 clones to microarray
analyses and found 975 genes that were significantly (>1.5
fold) up-regulated in 4SA versus vector control (Fig. 5A;
expression returned to vector control levels in 61% of the
genes in the 4SA-S51A group. Quantitative real-time PCR
(qRT–PCR) confirmed the induction of several microarray-
4SA, but not control or 4SA-S51A, cells (Supplemental Fig.
S5D). Functional pathway analyses using DAVID revealed
that the genes induced by 4SA showed significant similar-
ities to the MES subclass (Fig. 5B), while gene set enrich-
ment analyses (GSEA) (Mootha et al. 2003; Subramanian
et al. 2005) against Phillips and Verhaak MES genes
showed significant enrichment with the two independent
groups (4SA vs. Verhaak normalized enrichement score =
2.27; 4SAvs. Phillips normalized enrichment score = 1.84)
(Fig. 5C).Interestingly, genes that were significantly down-
regulated by overexpression of 4SA showed a PN charac-
teristic (Supplemental Fig. S5E; Supplemental Table S3B).
These data point to the important role of TAZ as master
modulator of the PN-to-MES transition in GBM.
Next, to address whether target genes inferred by micro-
array were indeed direct targets of TAZ–TEAD, we per-
formed chromatin immunoprecipitation (ChIP) coupled
with qPCR (ChIP-qPCR). We designed primers around
putative TEAD consensus sequences 3 kb proximal to the
transcriptional start site. All genes had at least one TEAD
consensussequence,and a 100- to200-base-pair(bp)region
surrounding this region was chosen for amplification. We
tested 15 microarray-inferred targets, including CTGF
(Supplemental Fig. S5F), which has been previously
annotated as a direct target of TAZ and YAP (Zhao et al.
2008b; Zhang et al. 2009). E2F7, a gene in the proliferative
subclass of GBM (de Bruin et al. 2003; Phillips et al. 2006),
was chosen for comparison against occupancy of the MES
targets by TAZ/TEAD. Strikingly, all 15 of the MES gene
promoters were bound by TAZ in 4SA cells, whereas
S51A showed significantly lower enrichment, implying
that TAZ is recruited to MES gene promoters via in-
teraction with TEAD (Fig. 5D; Supplemental Fig. S5F). Of
note, no differential enrichment of TAZ on the E2F7
promoter was seen between 4SA and 4SA-S51A. On the
contrary, TEAD2 was recruited to MES promoters in both
4SA- and 4SA-S51A-expressing cells (data not shown),
implying that the binding of TAZ to TEAD—and in turn
to target promoters—is a critical determinant of MES
gene induction. Thus, biochemical studies prove that
TAZ and TEAD2 are recruited directly to a majority of
The TAZ–TEAD pathway decreases survival
and increases tumor grade in the RCAS/N-tva
Although GSCs isolated from patients are valuable tools
to study pathway alterations, they present two major tech-
nical limitations. First, by using GSCs, one cannot address
the transforming potential of candidate oncogenes; that is,
if TAZ-mediated MES transition is an oncogenic event, it
cannot be addressed using GSCs. Second, tumor-initiating
capacity in GSCs is assessed in immunocompromised re-
cipient mice that form tumors, lacking immunological
features of clinical tumors. To address these limitations
and determine whether TAZ behaves as an oncogene, we
used the well-characterized RCAS/N-tva model, in which
the expression of retrovirus-encoded genes can be directed
to express in neural progenitor cells (NPCs) (Uhrbom and
Holland 2001; Begemann et al. 2002). Chicken fibroblast
(DF1) cells transfected with the RCAS vector were directly
injected into the brains of N-tva mice as previously
described. Because this model allows for testing of specific
TAZ and mesenchymal malignant glioma
GENES & DEVELOPMENT2601
top 50 induced genes in 4SA-expressing GSC11 compared with control. Plots are the log2Affymetrix expression values. (B) Bar graph
showing the top 10 gene ontology (GO) terms associated with the twofold or greater 4SA-induced genes, as ranked by P-value. Shown is the
number of genes that are common between the GO term’s gene set and the 4SA gene set. The line is the log10of the P-value as determined
by DAVID functional analysis. (C) GSEA analysis of 4SA-mediated global gene expression. A ranked list of all of the 4SA-induced genes on
the HGU133a2 microarray chip was compared against Phillips (left) or Verhaak (right) MES genes. (D) ChIP-qPCR analysis of TAZ
recruitment at selected target gene promoters in GSC11 overexpressing 4SA or S51A. Bar graphs represent mean percentage of input
occupancy by TAZ. A t-test was used to determine statistical significance. (*) P < 0.05; (**) P < 0.005; (NS) not significant.
Global MES differentiation induced by TAZ–TEAD via direct recruitment to target promoters. (A) Heat map showing ranked
Bhat et al.
2602 GENES & DEVELOPMENT
gene alteration combinations in an otherwise normal
background, PDGF-B was overexpressed with WT-TAZ,
4SA, and 4SA-S51A to test the effects on gliomagenesis.
Moreover, PDGF-driven tumors exhibit predominantly
PN characteristics (Lei et al. 2011), thus providing us with
an ideal model to test TAZ effects in vivo. Consistent with
in grade II gliomas with a median survival of ;11 wk
(Fig. 6A; Dai et al.2001). Overexpression of TAZ or 4SA
alone had no impact, since mice lived longer than 90 d
Kaplan-Meier survival analysis of PDGF-B-, WT-TAZ+PDGF-B-, 4SA+PDGF-B-, or 4SA-S51A+PDGF-B-injected mice. (B) Stacked bar
graph showing WHO glioma grades within each group. (C) Representative images of hematoxylin and eosin-stained slides of brains
isolated from mice injected with PDGF-B, WT-TAZ+PDGF-B, 4SA+PDGF-B, or 4SA-S51A+PDGF-B. Note that necrosis is observed only
in WT-TAZ+PDGF-B and 4SA+PDGF-B mouse tumors. (D) Real-time qPCR analyses of gene expression in RCAS mouse tumors. The
gene expression value of PDGF control was normalized to 1, and the relative expression of CD44, CTGF, and FN1 is shown.
TAZ enhances tumor grade, reduces survival, and promotes MES differentiation in the RCAS/N-tva mouse model. (A)
TAZ and mesenchymal malignant glioma
GENES & DEVELOPMENT 2603
post-implantation with no detectable tumors, suggesting
that aberrant expression or activation of TAZ is insuffi-
cient by itself to cause transformation in NPCs (data not
shown). On the other hand, when WT-TAZ or 4SA was
coexpressed with PDGF-B, survival was dramatically
reduced to <5 wk (Fig. 6A). A majority of the tumors
from WT-TAZ or 4SA+PDGF-B were grade III (45%–50%)
or grade IV (38%–41%), while those from PDGF-B alone
were predominantly grade II (76%) (Fig. 6B,C). Although
4SA mice exhibit survival rates similar to the wild-type
TAZ, 4SA+PDGF-B tumors were highly angiogenic and
showed increased vascular proliferation (data not shown).
Tumors resulting from 4SA-S51A coexpression with
PDGF-B were predominantly grade II (67%) and were in-
distinguishable from those of PDGF-B alone (Fig. 6B,C).
Last, we analyzed the expression of FN1, CD44, CTGF,
and ACTG2 in tumors that arose from these mice to un-
derstand the spectrum of MES differentiation induced
by TAZ. The expression of FN1, CD44, and CTGF was
dramatically induced in both TAZ- and 4SA-driven tu-
mors (Fig. 6D). ACTG2 was not induced in any of the
groups compared with PDGF (data not shown). We next
analyzed four grade II, two grade III, and one grade IV
gliomas from the 4SA-S51A+PDGF-B tumors. Interest-
ingly, expression levels of FN1, CD44, and CTGF were
modestly increasedwith highergrade(PDGF+4SA-S51A-6
and 4SA-S51A-7). Thus, in rare cases, it appears that TAZ
can induce high-grade MES tumors in a TEAD-indepen-
dent fashion. However, the overall expression of MES
markers in 4SA-S51A+PDGF tumors was significantly
lower compared with that stemming from 4SA+PDGF
overexpression (Fig. 6D). Thus, in line with our clinical
observations and in vitro findings, overexpression of TAZ
in conjunction with PDGF-B increases tumor grade and
induces predominant MES differentiation in murine
NPCs via interaction with TEAD.
Our study demonstrates that the HIPPO pathway tran-
scriptional coactivator TAZ is integral to the MES phe-
notype in glioma. First, we show that TAZ is epigeneti-
cally silenced in lower-grade gliomas as well as PN GBMs
when compared with MES tumors. Second, manipulating
TAZ expression in GSCs as well as murine NSCs affects
expression of MES genes. Third, we demonstrate direct
promoter co-occupancy of TAZ and TEAD2 in a majority
ofMES gene targets. Finally, weshowthat TAZ cooperates
with PDGF-B to induce high-grade MES gliomas in the
RCAS/N-tva model and that TEAD interaction is required
for all of these processes. The results presented here
implicate TAZ as a third key modulator of MES transition
in glioma in addition to STAT3 and C/EBP-b.
Identification of TAZ as a key regulator
of the MES network
Previous expression profiling efforts in GBM had failed to
identify distinct patterns of gene expression owing pri-
marily to fewer sample numbers. Large-scale studies by
multiple groups eventually identified three (PN, MES,
and proliferative) or four (PN, MES, neural, and classical)
subtypes of tumors (Phillips et al. 2006; Verhaak et al.
2010). Of these, it is becoming evident that two robust,
mutually exclusivegene expressionpatternscanreadily be
found across multiple data sets (PN and MES) (Huse et al.
2011). In the study by Phillips et al. (2006), a Proliferative
subtype was identified in addition to the PN and MES
groups. However, the Proliferative group may simply peel
off tumors from the other two groups that have increased
cell cycling, as suggested by the analysis of Huse et al.
(2011). It may be reasonable to think of proliferation as an
independent tumor axis, along with the mutually exclu-
sive PN/MES axis. Likewise, the additional Neural and
Classical groups that Verhaak et al. (2010) described seem
to reclassify tumors that would be distributed equally
between the PN and MES groups if these were the only
classifications. Even though the active signaling pathways
are different among the four Verhaak groups, the clinical
relevance of adding the additional two Verhaak groups is
not readily apparent. The MES subgroup, however, is of
high interest to us, since these patients exhibit worse
subtypes (Pelloski et al. 2005; Phillips et al. 2006; Colman
et al. 2010). Identifying molecular drivers of the MES
signature is a first step in effectively treating patients with
this subclass of tumors, and a beginning for personalized
medicine for GBM. Although various methods of genome-
wide analyses of regulatory networks have been described,
ARACNE is unique in that it identifies direct transcrip-
tional interactions without a priori assumptions (Margolin
et al. 2006). Using this method, the MES gene expression
program was shown to be controlled in part by STAT3 and
Carro et al. (2010) study and ours have several key differ-
ences that may account for us finding additional master
regulators of the MES network. First, we used a larger data
set (386 vs. 176 expression profiles) with a different micro-
array platform (HU-133A vs. HT-HU133). Second, our
et al. (2010) used a combination of grade III and grade IV
astrocytomas. Third, we used a more comprehensive list of
hub markers, which included transcriptional coregulators,
TAZbeing one of thoseabsentfrom the prior work’s TFlist.
Epigenetic regulation of TAZ is important
for its activity
Out of all of the HIPPO pathway components that had
annotated CpG islands, only the TAZ promoter showed
dramatic methylation differences between PN and MES
subclasses of GBMs in the TCGA data sets. We further
confirmed that TAZ is epigenetically silenced in lower-
grade gliomas, PN GBMs, and PN GSCs. The paradoxical
methylation of a tumor-enhancing gene like TAZ in
gliomas is unconventional, since frequent epigenetic in-
activation is typically associated with tumor suppressors
(Palii and Robertson 2007). However, TAZ CpG methyl-
ation is concordant with a clinically distinct subclass of
gliomas that is constitutively methylated in a cassette of
genes (termed glioma CpG island methylator phenotype
Bhat et al.
2604GENES & DEVELOPMENT
[G-CIMP]) (Noushmehr et al. 2010). A closer analysis of
a few G-CIMP methylated genes from this study (CHI3L1,
MMP9, LGALS3, PDPN, etc.) reveals that they are pre-
dominantly MES in nature (Hiratsuka et al. 2002; Wicki
et al. 2006; Zhao et al. 2009). Perhaps methylation of TAZ
and other G-CIMP genes restrains their expression in PN
GBMs and lower-grade gliomas, which in turn results in
favorable clinical response. In fact, the improved outcome
in the PN subclass of tumors could be accounted for by a
G-CIMP-positive subset that included methylated TAZ. As
an aside, we noted that three of the 16 ARACNE-identified
TFs in the MES network are part of the G-CIMP signature
identified TAZ target genes were G-CIMP (Supplemental
for tight regulation of its expression and, in turn, its activity.
Taken together with previous studies, TAZ now appears to
be regulated at multiple levels, including promoter methyl-
ation (this study), subcellular localization, and phosphory-
lation. Indeed, our studies affirm that epigenetic regulation
and cellular partitioning of TAZ play a major role in re-
gulating its functions in high-grade gliomas and GSCs.
however, that DAC did not induce TAZ levels to those
seen in GSC20, a MES GSC. Therefore, additional mech-
anisms may act to induce TAZ in MES GSCs.
MES transition is induced by the TAZ/TEAD complex
An important finding of our study is that TAZ alone can
cause MES reprogramming in murine NSCs and human
GBM-derived GSCs and that the TAZ–TEAD interaction
is required for this process. Previous studies have shown
ous TFs (Yagi et al. 1999; Vassilev et al. 2001; Basu et al.
2005, 2006; Varelas et al. 2008; Di Palma et al. 2009; Zhang
etal.2009).Of particularinterestisthe associationofTAZ/
YAP with the TEAD family of TFs, as elucidated by the
Guan group (Zhao et al. 2008b, Zhang et al. 2009). The
TAZ/YAP interaction with TEAD promotes cell prolif-
eration, EMT, invasion, and cellular transformation. In
addition, YAP has been shown to regulate epidermal
stem cell proliferation and expansion in mice via TEAD
(Schlegelmilch et al. 2011). Other recent studies impli-
by transforming growth factor-b (TGF-b) (Varelas et al.
2008). Although TGF-b has been shown to promote GSC
self-renewal and tumor-initiating properties (Penuelas
et al. 2009; Anido et al. 2010), we did not observe induc-
tion of MES genes by TGF-b in our GSC studies (data not
shown). This observation, coupled with previous reports
highlighting the importance of TEAD in mediating TAZ-
induced EMT, led us to pursue TEAD as a candidate TF
for TAZ-induced MES transition in GSCs. Indeed, TEAD
interaction was sufficient and required for MES transition
induced by TAZ, while RUNX2 was dispensable for these
We aimed at delineating TAZ and YAP functions in this
study in the context of glial tumors. While a role for YAP
in inducing MES differentiation cannot be formally ruled
out, we found YAP to be highly expressed in PN GSCs at
levels comparable with MES. Thus, YAP may not be
required for maintenance of theMES phenotype. Previous
studies point to a more predominant role for YAP in
embryonic stem cell self-renewal and proliferation (Lian
et al. 2010), and in the chicken neural tube, YAP over-
expression caused expansion of neural progenitor num-
bers, whereas loss of function promotes apoptosis and
premature neuronal differentiation (Cao et al. 2008). On
the contrary, we found that TAZ induces MES markers in
murine NSCs, with a parallel loss of glial/neuronal dif-
ferentiation, indicating that the predominant effect of
TAZ in NSCs is to promote MES differentiation. In fact,
NSCs overexpressing 4SA invariably exited the cell cycle
after few rounds of proliferation (data not shown), arguing
against a role for TAZ in inducing proliferation in these
cells. Similarly, in mice, deletion of YAP or TEAD1/2
induces embryonic lethality (Morin-Kensicki et al. 2006;
Sawada et al. 2008), whereas mice lacking TAZ are viable
and develop renal cysts, polycystic kidney disease, and
minor skeletal abnormalities (Hossain et al. 2007; Makita
et al. 2008). Thus, although some properties of TAZ and
YAP, including TEAD binding, appear to be shared, gain
and loss of function of these cofactors elicit differential
responses in a spatiotemporal manner.
While we show that TAZ induces MES high-grade
gliomas in nestin-positive NPCs using the PDGF-driven
RCAS mouse model, it is currently unclear whether TAZ
can reprogram other cell types. The exact cell of origin for
GBM is debatable, and the concept is still evolving. Early
studies showed that Ink4a/Arf?/?NSCs or astrocytes with
constitutively active EGFR induce high-grade gliomas
(Bachoo et al. 2002). Subsequently, Parada and colleagues
in NSCs, but not in the adult nonneurogenic brain, caused
gliomagenesis (Alcantara Llaguno et al. 2009). On the
contrary, recent studies pointtoOPCs, butnot other NSC-
derived lineages or NSCs themselves, as the cell of origin
of gliomas (Liu et al. 2011; Sugiarto et al. 2011). It remains
unknown whether TAZ can reprogram every cell typethat
is proposed to be the cell of originof glioma. Future studies
wherein TAZ is expressed in the Ctv-a mouse, in which
tumor formation can be examined in OPCs (Lindberg et al.
2009). could help address this important issue.
Ourinsilico, invitro,andinvivostudiesofTAZ inglioma
have important therapeutic implications. Patients whose
tumors express MES genes present a considerable clinical
challenge, since they show worse survival and radiation
resistance (Phillips et al. 2006). We identified TAZ as an
additional key transcriptional activator of the MES phe-
notype, which therefore presents another molecular tar-
get. Furthermore, we developed a mouse model using
the RCAS/N-tva system and showed that TAZ in com-
bination with PDGFB can drive formation of high-grade
TAZ and mesenchymal malignant glioma
GENES & DEVELOPMENT2605
MES gliomas and therefore could be a useful preclinical
model. This, to our knowledge, remains the first demon-
stration of TAZ as a tumor promoter in an in vivo setting.
Identifying inhibitors of TAZ expression or TAZ–TEAD
interaction will be the next challenge. With numerous
clinical trials using STAT3 inhibitors for solid tumors
currently ongoing, combination therapy with a TAZ in-
hibitor could be a viable strategy for treating the aggres-
sive MES subclass of GBMs.
Materials and methods
The ARACNE algorithm, as implemented in the geWorkBench
suite (https://cabig.nci.nih.gov/tools/geWorkbench), was per-
formed on the TCGA Affymetrix expression data set (down-
loaded April 28, 2011; n = 385), which was processed using a
custom CDF and RMA normalization using R and Bioconductor
(http://www.R-project.org). The resulting matrix of data was pro-
cessed with the following ARACNE settings: P-value of 0.01,
adaptive partitioning, and data processing inequality (DPI) tol-
set of TFs was generated from the TRANSFAC Web site and
genes with the term ‘‘transcription factor’’ in their parent gene
ontology (GO) term, which includes transcription cofactors. The
initial resulting ARACNE network was limited to GBM MES
genes by combining the gene lists from Phillips et al. (2006) and
Verhaak etal. (2010) and then selectingthe immediately adjacent
hub genes. Target genes that were identified to be associated
with TAZ by ARACNE analysis were processed using the
DAVID Web tool (http://david.abcc.ncifcrf.gov) to obtain associ-
ation of these genes with specific GO functional categories.
Default processing was done, except the analysis was limited to
DAVID’s GO biological process FAT (GOTERM_BP_FAT) cate-
gory. The resulting GO terms were ranked from smallest to
largest P-values after removing GO terms that had 10 or fewer
genes overlapping with the query gene list. To determine
whether a TCGA GBM was PN or MES, first a metagene score
for MES or PN was generated using a union of the respective
Phillips and Verhaak gene sets. The two metagene scores were
then compared, with class being assigned based on the greater
metagene score. Tumors that had both Illumina Infinium meth-
ylation data and Affymetrix gene expression data were sub-
sequently analyzed for correlation of TAZ methlation with
TAZ expression and GBM subtype.
Cell culture and transfection
GSCs were isolated from patients undergoing surgery at M.D.
Anderson Cancer Center (MDACC) and grown in neural basal
medium (Dulbecco’s Modified Eagle Medium [DMEM]/F12 50/
50; Cellgro) supplemented with B27 (Invitrogen), 20 ng/mL EGF
(Chemicon),and 20 ng/mL FGF (Akron-Biotech). Phoenix Ampho
(for GSCs) or Phoenix Eco (for mouse NSCs) was obtained from
a commercial source (Orbigen), grown in DMEM–F12 plus 10%
fetal bovine serum (FBS), and transfected with 16 mg of DNA
(pBABE vector containing wild-type TAZ, 4SA, 4SA-S51A) using
Lipofectamine 2000 (Invitrogen) or Fugene HD (Roche) according
to the manufacturer’s instructions. Viral supernatants were
collected 48 h after transfection, filtered, and immediately used
for transduction of GSCs. For lentivial transduction, pGIPZ
vectors expressing shRNA against two independent mRNA
regions of TAZ were used. Viral particles were generated using
the Trans-lentiviral packing system (Open Biosystems). Cells
were plated on laminin (BD Biosciences)-coated plates, and then
transduced with viral supernatant plus polybrene (Chemicon).
The pBABE-4SA-S51A plasmid was made using the QuikChange
II XL Site-Directed Mutatgenesis kit (Stratagene 200521) and
primers designed to introduce serine-to-alanine mutation (59-
TTCTTCCG-39). Both pBABE and pGIPZ transduced GSCs were
selected in neurobasal medium containing puromycin (4 mg/mL)
for generation of stable clones.
Intracranial mouse injections
SCID mice were bolted as previously described (Lal et al. 2000)
and injected with cells (10 K/mL) in serum-free medium. Five
mice wereinjected foreachgroup. Mice were sacrificedoncethey
showed neurological symptoms or appeared moribund. Brains
were immediately fixed in 4% paraformaldehyde and processed
for hematoxylin and eosin staining or IHC.
NSCs were isolated from the telecephalon of C57/BL6 mouse
embryos at E14 and cultured as described (Sher et al. 2008).
(Sigma-Aldrich), and cultured in neurobasal medium supple-
mented with 2% B27, 20 ng/mL EGF, 20 ng/mL bFGF, 1%
GlutaMAX (GIBCO), 100 mg/mL primocin, and 5 mg/mL hep-
arin (Sigma-Aldrich). After two passages, NSCs were used for
viral transduction as described previously. For differentiation,
stable pools of NSCs were dissociated and plated in poly-d-lysine
and laminin-coated 12-mm glass coverslips in NBM supple-
mented with 5% FBS or after withdrawal of bFGF and EGF from
the proliferation medium for 6 d and subsequently processed for
IF. The following antibodies were used to identify undifferenti-
ated NSCs and differentiated cell types: nestin, FN1 (Millipore),
GFAP (DAKO), and SMA (Sigma-Aldrich).
Fifteen micrograms of cRNA was used in the hybridizations to
U133A 2.0 human GeneChip expression arrays done according
to the specifications of the manufacturer (Affymetrix). Intensity
data were obtained from array images, and data were analyzed
using R Suite as previously described.
ChIP assay was performed after cross-linking cells using form-
aldehyde. DNA was sonicated using an Ultrasonic Processor
(GE130, Sorvall) at two cycles of six pulses each at 50% power
with a 1-min interval between cycles. Sonicated DNA was then
centrifuged at 13,5000 rpm at 4°C. Supernatant from 100,000
cells was used for each ChIP assay using MAGnify ChIP system
(Invitrogen). Two micrograms of mouse IgG, TAZ (BD Biosci-
ences), or TEAD2 (Novus) was used per ChIP. Immunoprecipi-
tated DNA was analyzed by Syber Green PCR, and Ct values
were used to calculate the percentage of input enrichment.
Primer sequences are shown in Supplemental Table S5.
The RCAS model for somatic gene transfer has been previously
described in detail (Uhrbom and Holland 2001; Begemann et al.
Bhat et al.
2606GENES & DEVELOPMENT
2002). Briefly, DF1 transfected cells were directly injected in-
tracranially on both sides (10,000 cells per microliter) into 2-d-
old pups. Mice were sacrificed when neurological symptoms
were present (i.e., hydrocephalus, seizures, inactivity, and/or
ataxia). Mice were sacrificed after 90-d post-injection if they did
not show symptoms. Brains were fixed in formalin and embed-
ded in paraffin blocks for further analyses.
We thank the generosity of Dr. Kun-Lian Guan (University of
California at San Diego) for donating the pBABE plasmids. This
research was supported by funding from the American Brain
Tumor Association basic research fellowship, Odyssey Special
Fellowship, Caroline Ross Endowment Fellowship, and MDACC
Brain Tumor SPORE developmental research project grant (to
K.P.L.B.); the Center for Clinical and Translational Sciences T32
grant (to K.S.); the NINDS grant (NS070928 to G.R.); the Brain
Tumor Funders’ Collaborative, the V Foundation, Rose Founda-
tion, National Brain Tumor Society basic research award, and
SPORE grant P50CA127001 from NIH/NCI (to K.A.); Ben and
Cathy Ivy Foundation Research Award (to F.F.L., K.A., and E.P.S);
and SPORE Animal Core grant (to F.F.L.).
Alcantara Llaguno S, Chen J, Kwon CH, Jackson EL, Li Y, Burns
DK, Alvarez-Buylla A, Parada LF. 2009. Malignant astro-
cytomas originate from neural stem/progenitor cells in
a somatic tumor suppressor mouse model. Cancer Cell 15:
Anido J, Saez-Borderias A, Gonzalez-Junca A, Rodon L, Folch G,
Carmona MA, Prieto-Sanchez RM, Barba I, Martinez-Saez E,
Prudkin L, et al. 2010. TGF-b receptor inhibitors target the
CD44(high)/Id1(high) glioma-initiating cell population in
human glioblastoma. Cancer Cell 18: 655–668.
Bachoo RM, Maher EA, Ligon KL, Sharpless NE, Chan SS, You MJ,
Tang Y, DeFrances J, Stover E, Weissleder R, et al. 2002. Epider-
mal growth factor receptor and Ink4a/Arf: Convergent mech-
anisms governing terminal differentiation and transformation
along the neural stem cell to astrocyte axis. Cancer Cell 1:
Basu S, Totty NF, Irwin MS, Sudol M, Downward J. 2003. Akt
phosphorylates the Yes-associated protein, YAP, to induce in-
teraction with 14-3-3 and attenuation of p73-mediated apo-
ptosis. Mol Cell 11: 11–23.
Begemann M, Fuller GN, Holland EC. 2002. Genetic modeling
of glioma formation in mice. Brain Pathol 12: 117–132.
Burns WC, Twigg SM, Forbes JM, Pete J, Tikellis C, Thallas-
Bonke V, Thomas MC, Cooper ME, Kantharidis P. 2006.
Connective tissue growth factor plays an important role in
advanced glycation end product-induced tubular epithelial-
to-mesenchymal transition: Implications for diabetic renal
disease. J Am Soc Nephrol 17: 2484–2494.
Burton EC, Lamborn KR, Feuerstein BG, Prados M, Scott J,
Forsyth P, Passe S, Jenkins RB, Aldape KD. 2002. Genetic
aberrations defined by comparative genomic hybridization
distinguish long-term from typical survivors of glioblastoma.
Cancer Res 62: 6205–6210.
Cao X, Pfaff SL, Gage FH. 2008. YAP regulates neural progenitor
cell number via the TEA domain transcription factor. Genes
Dev 22: 3320–3334.
Carro MS, Lim WK, Alvarez MJ, Bollo RJ, Zhao X, Snyder EY,
Sulman EP, Anne SL, Doetsch F, Colman H, et al. 2010. The
transcriptional network for mesenchymal transformation of
brain tumours. Nature 463: 318–325.
Chan SW, Lim CJ, Chong YF, Pobbati AV, Huang C, Hong W.
2011. Hippo pathway-independent restriction of TAZ and
YAP by angiomotin. J Biol Chem 286: 7018–7026.
Colman H, Zhang L, Sulman EP, McDonald JM, Shooshtari NL,
Rivera A, Popoff S, Nutt CL, Louis DN, Cairncross JG, et al.
2010. A multigene predictor of outcome in glioblastoma.
Neuro Ooncol 12: 49–57.
Cooper LA, Gutman DA, Long Q, Johnson BA, Cholleti SR, Kurc
T, Saltz JH, Brat DJ, Moreno CS. 2010. The proneural molec-
ular signature is enriched in oligodendrogliomas and predicts
improved survival among diffuse gliomas. PLoS ONE 5:
e12548. doi: 10.1371/journal.pone.002548.
Dai C, Celestino JC, Okada Y, Louis DN, Fuller GN, Holland
EC. 2001. PDGF autocrine stimulation dedifferentiates cul-
tured astrocytes and induces oligodendrogliomas and oli-
goastrocytomas from neural progenitors and astrocytes in
vivo. Genes Dev 15: 1913–1925.
Deaton AM, Bird A. 2011. CpG islands and the regulation of
transcription. Genes Dev 25: 1010–1022.
de Bruin A, Maiti B, Jakoi L, Timmers C, Buerki R, Leone G.
2003. Identification and characterization of E2F7, a novel
mammalian E2F family member capable of blocking cellular
proliferation. J Biol Chem 278: 42041–42049.
de Groot JF, Gilbert MR, Aldape K, Hess KR, Hanna TA, Ictech
S, Groves MD, Conrad C, Colman H, Puduvalli VK, et al.
2008. Phase II study of carboplatin and erlotinib (Tarceva,
OSI-774) in patients with recurrent glioblastoma. J Neuro-
oncol 90: 89–97.
Dennis G Jr, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC,
Lempicki RA. 2003. DAVID: Database for Annotation, Visu-
alization, and Integrated Discovery. Genome Biol 4: 3. doi:
Di Palma T, D’Andrea B, Liguori GL, Liguoro A, de Cristofaro T,
Del Prete D, Pappalardo A, Mascia A, Zannini M. 2009. TAZ is
a coactivator for Pax8 and TTF-1, two transcription factors
involved in thyroid differentiation. Exp Cell Res 315: 162–175.
Duning K, Rosenbusch D, Schluter MA, Tian Y, Kunzelmann K,
Meyer N, Schulze U, Markoff A, Pavenstadt H, Weide T.
2010. Polycystin-2 activity is controlled by transcriptional
coactivator with PDZ binding motif and PALS1-associated
tight junction protein. J Biol Chem 285: 33584–33588.
Freije WA, Castro-Vargas FE, Fang Z, Horvath S, Cloughesy T,
Liau LM, Mischel PS, Nelson SF. 2004. Gene expression
profiling of gliomas strongly predicts survival. Cancer Res
Furnari FB, Fenton T, Bachoo RM, Mukasa A, Stommel JM,
Stegh A, Hahn WC, Ligon KL, Louis DN, Brennan C, et al.
2007. Malignant astrocytic glioma: Genetics, biology, and
paths to treatment. Genes Dev 21: 2683–2710.
Gage FH, Ray J, Fisher LJ. 1995. Isolation, characterization, and use
of stem cells from the CNS. Annu Rev Neurosci 18: 159–192.
Gore-Hyer E, Shegogue D, Markiewicz M, Lo S, Hazen-Martin
D, Greene EL, Grotendorst G, Trojanowska M. 2002. TGF-b
and CTGF have overlapping and distinct fibrogenic effects on
human renal cells. Am J Physiol Renal Physiol 283: F707–
F716. doi: 10.1152/ajprenal.00007.2002.
Gregory PA, Bert AG, Paterson EL, Barry SC, Tsykin A, Farshid G,
Vadas MA, Khew-Goodall Y, Goodall GJ. 2008. The miR-200
family and miR-205 regulate epithelial to mesenchymal transi-
tion by targeting ZEB1 and SIP1. Nat Cell Biol 10: 593–601.
Gupta PB, Chaffer CL, Weinberg RA. 2009. Cancer stem cells:
Mirage or reality? Nat Med 15: 1010–1012.
Halder G, Johnson RL. 2011. Hippo signaling: Growth control
and beyond. Development 138: 9–22.
Hegi ME, Diserens AC, Gorlia T, Hamou MF, de Tribolet N,
Weller M, Kros JM, Hainfellner JA, Mason W, Mariani L, et al.
TAZ and mesenchymal malignant glioma
GENES & DEVELOPMENT 2607
2005. MGMT gene silencing and benefit from temozolomide
in glioblastoma. N Engl J Med 352: 997–1003.
Hiratsuka S, Nakamura K, Iwai S, Murakami M, Itoh T, Kijima H,
Shipley JM, Senior RM, Shibuya M. 2002. MMP9 induction by
vascular endothelial growth factor receptor-1 is involved in
lung-specific metastasis. Cancer Cell 2: 289–300.
Hong JH, Yaffe MB. 2006. TAZ: A b-catenin-like molecule that
regulates mesenchymal stem cell differentiation. Cell Cycle
Hong JH, Hwang ES, McManus MT, Amsterdam A, Tian Y,
Kalmukova R, Mueller E, Benjamin T, Spiegelman BM, Sharp
PA, et al. 2005. TAZ, a transcriptional modulator of mesen-
chymal stem cell differentiation. Science 309: 1074–1078.
Hossain Z, Ali SM, Ko HL, Xu J, Ng CP, Guo K, Qi Z, Ponniah S,
Hong W, Hunziker W. 2007. Glomerulocystic kidney disease
in mice with a targeted inactivation of Wwtr1. Proc Natl
Acad Sci 104: 1631–1636.
Huang J, Wu S, Barrera J, Matthews K, Pan D. 2005. The Hippo
signaling pathway coordinately regulates cell proliferation
and apoptosis by inactivating Yorkie, the Drosophila Homo-
log of YAP. Cell 122: 421–434.
Huang DW, Sherman BT, Lempicki RA. 2009. Systematic and
integrative analysis of large gene lists using DAVID bioin-
formatics resources. Nat Protoc 4: 44–57.
Huse JT, Holland EC. 2010. Targeting brain cancer: Advances in
the molecular pathology of malignant glioma and medullo-
blastoma. Nat Rev Cancer 10: 319–331.
Huse JT,PhillipsHS, Brennan CW.2011.Molecularsubclassification
of diffuse gliomas: Seeing order in the chaos. Glia 59: 1190–1199.
Kalluri R, Weinberg RA. 2009. The basics of epithelial–
mesenchymal transition. J Clin Invest 119: 1420–1428.
Komuro A, Nagai M, Navin NE, Sudol M. 2003. WW domain-
containing protein YAP associates with ErbB-4 and acts as
a co-transcriptional activator for the carboxyl-terminal frag-
ment of ErbB-4 that translocates to the nucleus. J Biol Chem
Kudo-Saito C, Shirako H, Takeuchi T, Kawakami Y. 2009.
Cancer metastasis is accelerated through immunosuppres-
sion during Snail-induced EMT of cancer cells. Cancer Cell
Lal S, Lacroix M, Tofilon P, Fuller GN, Sawaya R, Lang FF. 2000.
An implantable guide-screw system for brain tumor studies
in small animals. J Neurosurg 92: 326–333.
Lei QY, Zhang H, Zhao B, Zha ZY, Bai F, Pei XH, Zhao S, Xiong
Y, Guan KL. 2008. TAZ promotes cell proliferation and
epithelial–mesenchymal transition and is inhibited by the
hippo pathway. Mol Cell Biol 28: 2426–2436.
Lei L, Sonabend AM, Guarnieri P, Soderquist C, Ludwig T,
Rosenfeld S, Bruce JN, Canoll P. 2011. Glioblastoma models
reveal the connection between adult glial progenitors and
the proneural phenotype. PLoS ONE 6: e20041. doi: 10.1371/
Li A, Walling J, Ahn S, Kotliarov Y, Su Q, Quezado M,
Oberholtzer JC, Park J, Zenklusen JC, Fine HA. 2009. Un-
supervised analysis of transcriptomic profiles reveals six
glioma subtypes. Cancer Res 69: 2091–2099.
Lian I, Kim J, Okazawa H, Zhao J, Zhao B, Yu J, Chinnaiyan A,
Israel MA, Goldstein LS, Abujarour R, et al. 2010. The role of
YAP transcription coactivator in regulating stem cell self-
renewal and differentiation. Genes Dev 24: 1106–1118.
Lindberg N, Kastemar M, Olofsson T, Smits A, Uhrbom L. 2009.
Oligodendrocyte progenitor cells can act as cell of origin for
experimental glioma. Oncogene 28: 2266–2275.
Liu CY, Zha ZY, Zhou X, Zhang H, Huang W, Zhao D, Li T,
Chan SW, Lim CJ, Hong W, et al. 2010. The hippo tumor
pathway promotes TAZ degradation by phosphorylating
a phosphodegron and recruiting the SCFb-TrCP E3 ligase.
J Biol Chem 285: 37159–37169.
Liu C, Sage JC, Miller MR, Verhaak RG, Hippenmeyer S, Vogel
H, Foreman O, Bronson RT, Nishiyama A, Luo L, et al. 2011.
Mosaic analysis with double markers reveals tumor cell of
origin in glioma. Cell 146: 209–221.
Makita R, Uchijima Y, Nishiyama K, Amano T, Chen Q,
Takeuchi T, Mitani A, Nagase T, Yatomi Y, Aburatani H,
et al. 2008. Multiple renal cysts, urinary concentration
defects, and pulmonary emphysematous changes in mice
lacking TAZ. Am J Physiol Renal Physiol 294: F542–F553.
Mani SA, Yang J, Brooks M, Schwaninger G, Zhou A, Miura N,
Kutok JL, Hartwell K, Richardson AL, Weinberg RA. 2007.
Mesenchyme Forkhead 1 (FOXC2) plays a key role in
metastasis and is associated with aggressive basal-like breast
cancers. Proc Natl Acad Sci 104: 10069–10074.
Mani SA, Guo W, Liao MJ, Eaton EN, Ayyanan A, Zhou AY,
Brooks M, Reinhard F, Zhang CC, Shipitsin M, et al. 2008.
The epithelial–mesenchymal transition generates cells with
properties of stem cells. Cell 133: 704–715.
Margolin AA, Wang K, Lim WK, Kustagi M, Nemenman I,
Califano A. 2006. Reverse engineering cellular networks.
Nat Protoc 1: 662–671.
Martinez J, Chalupowicz DG, Roush RK, Sheth A, Barsigian C.
1994. Transglutaminase-mediated processing of fibronectin
by endothelial cell monolayers. Biochemistry 33: 2538–
Meloan SN, Puchtler H, Valentine LS. 1972. Alkaline and acid
alizarin red S stains for alkali-soluble and alkali-insoluble
calcium deposits. Arch Pathol 93: 190–197.
Mootha VK, Lindgren CM, Eriksson KF, Subramanian A, Sihag
S, Lehar J, Puigserver P, Carlsson E, Ridderstrale M, Laurila E,
et al. 2003. PGC-1a-responsive genes involved in oxidative
phosphorylation are coordinately downregulated in human
diabetes. Nat Genet 34: 267–273.
Morin-Kensicki EM, Boone BN, Howell M, Stonebraker JR, Teed
J, Alb JG, Magnuson TR, O’Neal W, Milgram SL. 2006.
Defects in yolk sac vasculogenesis, chorioallantoic fusion,
and embryonic axis elongation in mice with targeted disrup-
tion of Yap65. Mol Cell Biol 26: 77–87.
Murakami M, Nakagawa M, Olson EN, Nakagawa O. 2005.
AWW domain protein TAZ is a critical coactivator for TBX5,
a transcription factor implicated in Holt-Oram syndrome.
Proc Natl Acad Sci 102: 18034–18039.
Murakami M, Tominaga J, Makita R, Uchijima Y, Kurihara Y,
Nakagawa O, Asano T, Kurihara H. 2006. Transcriptional
activity of Pax3 is co-activated by TAZ. Biochem Biophys
Res Commun 339: 533–539.
Nakajima Y, Mironov V, Yamagishi T, Nakamura H, Markwald
RR. 1997. Expression of smooth muscle a-actin in mesen-
chymal cells during formation of avian endocardial cushion
tissue: A role for transforming growth factor b3. Dev Dyn
Nigro JM, Misra A, Zhang L, Smirnov I, Colman H, Griffin C,
Ozburn N, Chen M, Pan E, Koul D, et al. 2005. Integrated
array-comparative genomic hybridization and expression
array profiles identify clinically relevant molecular subtypes
of glioblastoma. Cancer Res 65: 1678–1686.
Noushmehr H, Weisenberger DJ, Diefes K, Phillips HS, Pujara K,
Berman BP, Pan F, Pelloski CE, Sulman EP, Bhat KP, et al.
2010. Identification of a CpG island methylator phenotype
that defines a distinct subgroup of glioma. Cancer Cell 17:
Palii SS, Robertson KD. 2007. Epigenetic control of tumor
suppression. Crit Rev Eukaryot Gene Expr 17: 295–316.
Bhat et al.
2608 GENES & DEVELOPMENT