miR-141 and miR-200a act on ovarian tumorigenesis by controlling oxidative stress response.

Abstract

Although there is evidence that redox regulation has an essential role in malignancies, its impact on tumor prognosis remains unclear. Here we show crosstalk between oxidative stress and the miR-200 family of microRNAs that affects tumorigenesis and chemosensitivity. miR-141 and miR-200a target p38α and modulate the oxidative stress response. Enhanced expression of these microRNAs mimics p38α deficiency and increases tumor growth in mouse models, but it also improves the response to chemotherapeutic agents. High-grade human ovarian adenocarcinomas that accumulate miR-200a have low concentrations of p38α and an associated oxidative stress signature. The miR200a-dependent stress signature correlates with improved survival of patients in response to treatment. Therefore, the role of miR-200a in stress could be a predictive marker for clinical outcome in ovarian cancer. In addition, although oxidative stress promotes tumor growth, it also sensitizes tumors to treatment, which could account for the limited success of antioxidants in clinical trials.

Full text

Page 1

ARTICLES
NATURE MEDICINE VOLUME 17 | NUMBER 12 | DECEMBER 2011
1627
Epithelial ovarian cancer is the most lethal form of gynecologic malig-
nancy. The clinical prognosis factors for this type of cancer are based
on the state of the disease at diagnosis, including histological subtype,
grade, stage and extent of residual disease after surgery. Despite the
use of treatment strategies that combine surgery and chemotherapy,
patients with epithelial ovarian cancer often relapse and eventually
die from their disease1. In recent years, several studies have analyzed
large-scale transcription profiling to identify differentially expressed
genes according to tumor status, histological subtypes and metastatic
spread2. However, the molecular biology of ovarian cancer is still not
completely understood, making the development of more effective
therapies difficult. Therefore, there is a pressing need to explore the
biological basis of ovarian cancer to search for early diagnostic classi-
fiers that can reliably stratify patients for therapeutic intervention.
Accumulation of reactive oxygen species (ROS) in tumor cells dam-
ages their cellular components and alters various processes, including
gene expression, proliferation and genomic stability3–9. Of the many
adaptive mechanisms that modulate gene expression in response to
stress, the p38? mitogen-activated protein kinase (MAPK) family
acts as a sensor of oxidative stress, and its redox-sensing function is
essential in the control of tumor development10,11. In contrast to other
MAPKs such as extracellular signal-regulated kinase (ERK) or cJun
NH2-terminal kinase (JNK), which can promote either proliferation
or survival, p38? often suppresses tumorigenesis by blocking prolif-
eration or promoting apoptosis12–15. In addition, miRNA expression
can be altered by distinct stress conditions such as radiation, oxidative
stress or hypoxia16–21. In this regard, miRNAs are essential regulators
of the stress response across multiple species.
Here we identify a new function for the miR-200 family members
(miR-200s) in oxidative stress and ovarian tumorigenesis. These miRNAs
were previously shown to modulate cellular motility and control
‘stemness’ and apoptosis22–33. We now show that two members of
the miR-200 family (miR-141 and miR-200a) inhibit p38? and have
an essential role in redox sensing. In mouse models, accumulation of
these miRNAs mimics p38? deficiency and promotes malignancy.
Human ovarian adenocarcinomas characterized by high miR-200a
expression show low amounts of p38? protein and have an associated
oxidative stress signature. Notably, this signature is correlated with
longer progression-free survival (PFS) and overall survival. Consistent
with these findings, although overexpression of miR-200a or miR-141
promotes tumorigenesis under untreated conditions, it increases
tumor-cell death and slows tumor growth under treatment with pacli-
taxel, a chemotherapeutic drug known to increase ROS34–36. Thus,
the miR-200–dependent stress response could have a dual function
in tumors: although this response increases tumor growth, it could
also enhance sensitivity to chemotherapy. In addition, we identify a
new function for miR-200s in the stress response, which could be a
predictive marker for clinical outcome in ovarian cancers.
RESULTS
Expression of miR-200s is stimulated by oxidative stress
Using a microarray analysis, we identified a set of 36 miRNAs
whose expression was changed after exposure of fibroblasts to acute oxi-
dative stress (Fig. 1a). Twenty of these miRNAs were upregulated and
16 were downregulated by H2O2 treatment over time. In particular, the
expression of the miR-200s was stimulated by oxidative stress (Fig. 1a,b).
1Stress and Cancer Laboratory, Institut Curie, Paris, France. 2Institut National de la Santé et de la Recherche Médicale, U830, Paris, France. 3Institut Curie,
Department of Pathology, Paris, France. 4Institut Curie, Functional Genomic Platform, Paris, France. 5Institut Curie, Department of Medical Oncology, Paris, France.
Correspondence should be addressed to F.M.-G. (fatima.mechta-grigoriou@curie.fr).
Received 12 January; accepted 14 September; published online 20 November 2011; doi:10.1038/nm.2512
miR-141 and miR-200a act on ovarian tumorigenesis
by controlling oxidative stress response
Bogdan Mateescu1,2, Luciana Batista1,2, Melissa Cardon1,2, Tina Gruosso1,2, Yvan de Feraudy1,2, Odette Mariani3,
André Nicolas3, Jean-Philippe Meyniel4, Paul Cottu5, Xavier Sastre-Garau3 & Fatima Mechta-Grigoriou1,2
Although there is evidence that redox regulation has an essential role in malignancies, its impact on tumor prognosis remains
unclear. Here we show crosstalk between oxidative stress and the miR-200 family of microRNAs that affects tumorigenesis and
chemosensitivity. miR-141 and miR-200a target p38a and modulate the oxidative stress response. Enhanced expression of
these microRNAs mimics p38a deficiency and increases tumor growth in mouse models, but it also improves the response to
chemotherapeutic agents. High-grade human ovarian adenocarcinomas that accumulate miR-200a have low concentrations of
p38a and an associated oxidative stress signature. The miR200a-dependent stress signature correlates with improved survival of
patients in response to treatment. Therefore, the role of miR-200a in stress could be a predictive marker for clinical outcome in
ovarian cancer. In addition, although oxidative stress promotes tumor growth, it also sensitizes tumors to treatment, which could
account for the limited success of antioxidants in clinical trials.
© 2011 Nature America, Inc. All rights reserved.

Page 2

ARTICLES
1628
VOLUME 17 | NUMBER 12 | DECEMBER 2011 NATURE MEDICINE
miRNAs from each locus (Supplementary Fig. 1a,b) had the same kinet-
ics after induction by stress (Fig. 1b). Overall, expression of the miR-200s
was induced within 1 h of treatment, reached its maximum between
2 and 3 h after treatment and was maintained at later time points. The
miR-200s were upregulated with similar kinetics in fibroblasts and
epithelial cells from mouse and human (Fig. 1b). In cell lines already
expressing high basal concentrations of the miR-200s, we did not detect
upregulation (Supplementary Fig. 1c,d). Upregulation of the miR-200s
was specific to H2O2 exposure, and we did not observe upregulation
with other stimuli or stressors (Supplementary Fig. 1e), further showing
these miRNAs might be involved in response to oxidative stress.
The miR-200s have been shown to regulate the mesenchymal-to-
epithelial transition (MET) through modulation of the E-cadherin
transcriptional repressor zinc finger E-box binding homeobox 1
(ZEB1)22–27,32,33. Accordingly, treatment of fibroblasts or colon
carcinoma cells with H2O2 reduced the amount of ZEB1 protein and
enhanced E-cadherin transcription but was not sufficient to induce
a stable MET, as H2O2 hydrolysis prevented long-lasting miR-200
accumulation (Fig. 1c, Supplementary Fig. 1f and data not shown).
This suggests that the regulation of miR-200s by stress, among other
mechanisms, could be involved in the control of E-cadherin expres-
sion. When overexpressed in fibroblasts, the miR-200s stabilized
E-cadherin expression by targeting its transcriptional repressor, ZEB1
(Fig. 1d). Although all the miR-200s were overexpressed to the same
extent, stabilization of E-cadherin was more efficient when miR-200b
or miR-200c was overexpressed than when miR-141 or miR-200a was
overexpressed. Accordingly, we detected miRNA-induced clustering
of epithelial cells only with miR-200b or miR-200c and not with miR-
141 or miR-200a (Fig. 1e). This further suggests that the miR-200s
have different cellular functions, which led us to investigate their
specific role in the oxidative stress response.
miR-141 and miR-200a inhibit p38a
To determine the function of miR-200 in oxidative stress, we compared
the ROS-dependent signaling pathway of cells overexpressing miR-141
or miR-200a to cells overexpressing a control miRNA when exposed to
H2O2 (h)
H2O2
–
H2O2
–
– 0.5 1 2 424
miR-200a
miR-141
miR-223
miR-122
miR-215
miR-192
miR-142-3p
miR-682
miR-187
Log2 fold change
Log2 fold change
Log2 fold change
Log2 fold change
CDH1 mRNA
(Log2 fold change)
CDH1 mRNA
(Log2 fold change)
Log2 fold change
Log2 fold change
Log2 fold change
Log2 fold change
8
abc
d
e
Mouse fibroblasts
CT26
Mouse cell lines
miR-141
miR-200c
miR-200a
miR-200b
NMuMG
6
4
2
–2
12
10
8
6
4
2
0
–2
Kras-transformed
fibroblasts
Hras-transformed
fibroblasts
3
2
1
0
–1
5
BT-549SKOV3
4
3
2
1
0
–1
3
2
1
–1
0
Log2 fold change
Log2 fold change
3
2
1
0
–1
3
4
2
1
0
–1
12
10
8
6
4
2
0
–2
0
8
10
6
4
2
–2
–1
0
4
3
2
1
0
5
6
2468
2468
2468
Time of
H2O2 treatment (h)
Time of
H2O2 treatment (h)
2468
2468
24682468
Time of
H2O2 treatment (h)
Time of
H2O2 treatment (h)
20
Time of H2O2 treatment (h)
468
20
Time of H2O2 treatment (h)
468
Mouse fibroblasts
ZEB1
p-JUN
JUN
170 kDa
43 kDa
34 kDa
170 kDa
E-cadherin
Control– 141 200a 200b
miRNA
200c
135 kDa
170 kDa
34 kDa
ZEB1
GAPDH
– ControlmiR-141
miR-200amiR-200b miR-200c
43 kDa
34 kDa
GAPDH
5
4
3
2
1
0
5
6
4
3
2
1
0
ZEB1
p-JUN
JUN
GAPDH
++
CT26
Time of
H2O2 treatment (h)
Time of
H2O2 treatment (h)
Time of
H2O2 treatment (h)
Human cell lines
MDA-MB-435SMDA-MB-436
293T
Time of
H2O2 treatment (h)
2468246
Time of
H2O2 treatment (h)
Time of
H2O2 treatment (h)
miR-142-5p
miR-709
miR-194
miR-679
miR-676
miR-592
miR-489
miR-615-3p
miR-691
miR-212
miR-411*
miR-673-5p
miR-547
miR-494
miR-335-5p
miR-181c
miR-677
miR-703
miR-501-3p
miR-384-3p
miR-302c
miR-539
miR-101a
miR-696
miR-704
miR-497
miR-467b*
–3.0 –1.8
1.8
–0.6
0.63.0
8
8
8246
Figure 1 Expression of the miR-200s is stimulated by oxidative stress. (a) Unsupervised hierarchical clustering of miRNA microarray expression data from
mouse fibroblasts after exposure to H2O2 for various durations. These miRNAs were selected on the basis of their differential expression (P ? 0.05) at
early (0.5–1 h), middle (2–4 h) or late (8–24 h) time points of treatment, when compared to untreated fibroblasts. The values shown are the means of the
normalized expressions from two independent experiments. Upregulated and downregulated miRNAs after stress are shown in red and blue, respectively.
(b) qRT-PCR of the miR-200s after a time course of H2O2 treatment in mouse cells (immortalized fibroblasts, colon carcinoma (CT26), mammary gland
epithelial cells (NMuMG) and Kras- and Hras-transformed fibroblasts) and human cell lines (melanoma cells (MDA-MB-435S), kidney cells (293T), breast
adenocarcinoma (MDA-MB-436 and BT-549) and ovarian adenocarcinoma (SKOV3)). Data are the means of fold changes (normalized to untreated and
expressed as log2) ? s.d. (c) Top, western blot of ZEB1 protein in mouse fibroblasts and carcinoma cells (CT26) after 3 h of H2O2 exposure, at which
point the miR-200s reached their maximum expression. Jun protein, its phosphorylated isoform (p-Jun) and glyceraldehyde-3-phosphate dehydrogenase
(GAPDH) were used as internal controls for H2O2 treatment and protein loading, respectively. Bottom, E-cadherin mRNA levels in fibroblasts and CT26
cells after H2O2 exposure are shown. (d,e) Shown are the amounts of E-cadherin and ZEB1 protein (d) and the cellular morphology (e) in untransfected
fibroblasts (−) or at 3 d after transfection with a control miRNA or the miR-200s. All transfected miRNAs are expressed to the same range (with a qRT-PCR
cycle threshold (Ct) of 13.4 ? 0.7 Ct on average). Scale bars, 15 ?m. All experiments were performed in triplicate, and error bars represent means ? s.e.m.
(unless otherwise specified) from at least three independent experiments.
© 2011 Nature America, Inc. All rights reserved.

Page 3

ARTICLES
NATURE MEDICINE VOLUME 17 | NUMBER 12 | DECEMBER 2011
1629
H2O2. Microarray experiments and further analyses using Ingenuity
Pathway Analysis (IPA) software indicated that miR-141 or miR-200a
altered MAPK response to H2O2 by modulating the p38? and JNK
pathways (Fig. 2a). Indeed, expression of mitogen-activated protein
kinase 14 (Mapk14) (encoding p38?) was significantly downregulated
by miR-141 or miR-200a overexpression, as evaluated by microarrays
and confirmed by quantitative RT-PCR (qRT-PCR) (Fig. 2b). The
expression of Mapk14 was unchanged by overexpression of miR-200c
or miR-200b (Fig. 2b), suggesting that only miR-141 and miR-200a
are involved in the regulation of the MAPK pathway.
miR-141 and miR-200a have high sequence homology
(Supplementary Fig. 1a), suggesting they target the same proteins.
Overexpression of miR-141 or miR-200a severely reduced the total
amount of p38? protein under either basal or stressed conditions
(Fig. 2c and Supplementary Fig. 2b). This reduction prevented the
expected accumulation of the phosphorylated form of p38? and led
to the subsequent decreased phosphorylation of MAPK-activated
protein kinase 2 (MAPKAPK2), one of the major downstream effec-
tors of p38? (Fig. 2c). Furthermore, overexpression of miR-141 or
miR-200a was associated with the constitutive activation of the JNK
pathway (Fig. 2c), as was previously seen after Mapk14 inactivation12.
In cells expressing miR-141 or miR-200a, as in Mapk14−/− cells, we
observed an earlier and higher phosphorylation rate of JNK1, JNK2
and JNK targets, for example, the proto-oncogene Jun (Fig. 2c).
We confirmed that overexpression of miR-141 or miR-200a led to
p38? downregulation in various mouse and human cell lines, whereas
overexpression of miR-200c or miR-200b had no effect (Fig. 2d).
These results indicate specific control of p38? by miR-141 or
miR-200a in the basal state or under stressed conditions in human
and mouse cells. Accordingly, cell lines with high endogenous con-
centrations of miR-141 or miR-200a had lower amounts of p38?
protein than those lines characterized by low expression of miR-141 or
miR-200a (Fig. 2e). The MAPK14 mRNA level was significantly
inversely correlated with miR-200a expression (Fig. 2f) when evalu-
ated using the US National Cancer Institute’s NCI60 database, which
contains a panel of 60 diverse human cancer cell lines37. In addition,
miR-200a was the sixth-ranked miRNA among 422 miRNAs with
respect to an inverse correlation with MAPK14. Reciprocally, MAPK14
was ranked 3,616th among the 41,078 probe sets for miR-200a.
Finally, we found that the 3? untranslated region of MAPK14 was
directly targeted by miR-141 or miR-200a, and we identified
the genuine binding site of these miRNAs in humans and mice
(Supplementary Fig. 3). These data show that miR-141 and miR-200a
are key direct regulators of p38?.
miR-141 and miR-200a promote tumorigenesis in mouse models
Because miR-141 and miR-200a have homologous sequences, we
tested whether the downregulation of p38? by miRNA could affect
1.5
Chemical or toxicant
Kinase
Complex or group
Ras
ROS
ac
d
ef
b
JNK
1/2
c-Raf
MEK5
MKK
4/7
MKK
3/6
MEKK
ASK1TAK1
PERK
ER stress
p38
MAPK
GSK3?
PKC
AKT
Pl3K
ERK5
MEK
1/2
ERK
1/2
Direct interaction
Indirect interaction
Acts on
Inhibits and
acts on
1.2
0.8
0.4
0
Trans
MAPK14 mRNA
MAPK14 mRNA
0.5
Trans
Control
*
**
***
miR-141
miR-200c
Control
miR-141
JNK pathwayp38 pathway
Control
p-p38?
p38?
p-MAPKAPK2
MAPKAPK2
GAPDH
p38?
Mouse fibroblasts
High miR-141
or miR-200a
43 kDa
miR-200a (log2)
6
R = –0.29
P = 0.023
4
2
–2
–2–1012
–4
MAPK14 mRNA (log2)
0
GAPDH
p38?
Ponceau staining
Fibroblasts
MDA-MB-231MDA-MB-435S
SKOV3 SW480
HCT116HT29
SKBR3
MCF7
IGROV
SHIM
NMuMGMDA-MB-435SSKOV3
43 kDa
37 kDa
Control
mR-200amR-200bmR-200c mR-141
Control
mR-200amR-200bmR-200cmR-141
Control
mR-200a mR-200bmR-200c mR-141
Control
mR-200amR-200b mR-200c mR-141
55 kDa
44 kDa
55 kDa
44 kDa
37 kDa
43 kDa
43 kDa
p-JNK2
p-JNK1
54 kDa
46 kDa
54 kDa
46 kDa
48 kDa
42 kDa
37 kDa
p-JUN
JUN
GAPDH
JNK2
JNK1
– 0.5 1 2 – 0.5 1 2 – 0.5 1 2 – 0.5 1 2– 0.5 1 2– 0.5 1 2
H2O2 (h)H2O2 (h)H2O2 (h)H2O2 (h)H2O2 (h)H2O2 (h)
miR-141miR-200a ControlmiR-141miR-200a
miR-200c
miR-200a
miR-200b
0
1.0
Probe intensity
(fold change)
qRT-PCR (fold change)
Low miR-141
or miR-200a
Figure 2 Balance of the p38 and JNK pathways by miR-141 and miR-200a. (a) Scheme of the ROS signaling pathway showing mRNAs affected by mir-
141 overexpression in mouse fibroblasts stimulated by H2O2. Significantly upregulated and downregulated mRNAs are shown in red and blue, respectively
(P ? 0.05). The fold changes are normalized to cells overexpressing control miRNA. (b) MAPK14 mRNA level evaluated by microarray (left) or qRT-PCR
(right) in mouse fibroblasts 3 d after transfection with a transfection reagent (trans), control miRNA (control) or a specific miRNA, as indicated. Data are
the means of fold changes (normalized to control) ? s.e.m. *P ? 0.05, **P ? 0.01, ***P ? 0.001 by Student’s t test. (c) Western blots showing MAPK-
related proteins in mouse fibroblasts transfected with specific miRNA, as noted, under basal conditions (−) or after H2O2 exposure. P-, phospho-.
(d) Western blots showing the amount of p38? protein in mouse (fibroblasts and NMuMG) and human (MDA-MB-435S and SKOV3) cell lines after miRNA
transfection. (e) Western blot showing the amount of p38? protein in diverse human cancer cell lines. A cell line belongs to the ‘high’ class of miR-141
or miR-200a if at least one of these mRNA’s content is above the average miR-141 or miR-200a level from all the analyzed cell lines; a cell line belongs
to the ‘low’ class if both miR-141 and miR-200a content is below the average. Using this rule, cells belonging to the ‘high miR-141 or miR-200a’ class
express at least one of the miRNAs at a fivefold higher level than their average levels. (f) Correlation plot extracted from NCI60 microarray data sets
showing that MAPK14 mRNA level is inversely correlated with endogenous miR-200a levels (R = −0.29, P = 0.023 by Pearson’s test).
© 2011 Nature America, Inc. All rights reserved.

Page 4

ARTICLES
1630
VOLUME 17 | NUMBER 12 | DECEMBER 2011 NATURE MEDICINE
transformation in Kras-transformed fibroblasts overexpressing
miR-141. We confirmed the accumulation of miR-141 in these sta-
ble clones (Supplementary Fig. 4a). When plated in soft agar, cells
overexpressing miR-141 showed a markedly enhanced plating effi-
ciency and growth rate compared to controls (Fig. 3a), indicating
that miR-141 facilitates cell growth without substrate attachment. In
addition, overexpression of miR-141 significantly increased the tumor
size in xenografted nude mice (Fig. 3b). Analyzed when they reached
the same size, tumors overexpressing miR-141 had a lower amount
of p38? protein than control tumors (Fig. 3c). Furthermore, these
tumors had a higher mitotic index and a greater number of large blood
vessels compared to controls (Fig. 3c), as was previously shown for
p38?-deficient tumors in a Ras-dependent context10.
Because we detected marked effects for miR-141 and miR-200a
in human ovarian carcinomas (see below), we assessed their effect
on xenografted ovarian tumors. Ovarian cells stably overexpressing
miR-141 or miR-200a (Supplementary
Fig. 4b) gave rise to larger tumors when com-
pared to control tumors (Fig. 3d). Tumors
expressing miR-141 or miR-200a appeared
earlier and grew faster than control tumors
(Fig. 3e). Histological analyses confirmed
that tumors overexpressing miR-141 or miR-
200a had low amounts of p38? protein and
were associated with a high mitotic index
(Fig. 3f). Therefore, these data suggest that miRNA-dependent down-
regulation of p38? has a positive role in tumorigenesis.
miR-200a controls p38? and stress response in ovarian cancers
Because the miR-200s have been shown to accumulate in aggres-
sive human ovarian adenocarcinomas38–44, we investigated whether
p38? is a predictive marker for these tumors. We analyzed a large
set of human ovarian adenocarcinomas that included mostly serous-
subtype and high-grade tumors. Clinical details and subject information
are given in Supplementary Table 1 and Supplementary Table 2. We
observed that there was no correlation between MAPK14 mRNA level
and the amount of p38? protein (Fig. 4a and Supplementary Table 2).
Notably, the miR-200a expression rate was significantly inversely
correlated with the amount of p38? protein (Spearman correlation
coefficient (R) = −0.37, P = 6 × 10−3) (Fig. 4b and Supplementary
Table 3). We observed this same tendency with miR-141, although
Ras + control
a
b
d
f
e
c
800
600
Tumor volume (mm3)
CD31
p38?
P-Ser10-H3
Tumor volume (mm3)
Probability
P-Ser10-H3
p38?
Tumor volume (mm3)
400
200
0
control
Ras + miR-141
Ras + control
Ras + miR-141
NS
1.5
**
***
***
***
***
***
***
***
***
***
***
**
**
**
*
*
*
**
**
60
40
20
0
1.0
0.5
Ras + controlRas + miR-141
SmallMediumLarge
Ras +
control
Ras +
miR-141
Ras +
Control
800
1.0
700
600
500
400
300
200
100
Control
miR-141
miR-200a
0.8
0.6
0.4
0.2
miR-141 vs control: P = 0.02
miR-200a vs control: P = 0.09
0
01020
Time (d)
0 10
Time (d)
205152530 40
600
400
200
0
miR-141
miR-200a
Control miR-141 miR-200a
Ovarian cancer cells
Ras +
miR-141
0
Mitosis mm–2: 81 ? 15
Mitosis mm–2: 163 ? 9
Mitosis mm–2: 187 ? 20
Clonogenicity (fold change)
Colony size (%)
Control
miR-141
miR-200a
Mitosis mm–2: 100 ? 19
Vessels mm–2: 1.3 ? 0.8
Vessels mm–2: 6.6 ? 3.1
Mitosis mm–2: 178 ? 6.2
Figure 3 miR-141 and miR-200a enhance
tumorigenesis in mouse models. (a) Left,
representative dishes and individual colonies
from Kras-transformed fibroblasts stably
transfected with a control miRNA (Ras
+ control) or miR-141 (Ras + miR-141).
Middle, relative number of visible colonies
in soft agar formed from control and miR-
141–overexpressing clones. Right, percentage
of visible colonies arranged according to their
size: small, ? 2.2 mm2; 2.2 mm2 < medium ?
4.5 mm2; large, > 4.5 mm2. Data are means ?
s.e.m. (b) Tumor volumes 9 d after xenografting
of control or miR-141–overexpressing Ras-
transformed fibroblasts. Values are means ?
s.e.m. n = 10 in each group. (c) Representative
immunohistochemistry of each type of tumor
using specific antibodies. Indicated in each
category are the numbers of mitotic cells per
mm2 (mm−2) (P = 0.032 by Student’s t test)
and large (width >15 ?m) blood vessels per
mm2 (P = 0.039 by Student’s t test). (d) Tumor
volumes 20 d after xenografting of control
and miR-141− or miR-200a–overexpressing
ovarian cancer cells. n ? 17 in each group.
(e) Left, Kaplan-Meir tumor-free survival
curves from control and miR-141− or miR-
200a–overexpressing mice (log-rank test).
Right, tumor growth curves (Student’s t test).
Data are means ? s.e.m. (f) Representative
immunohistochemistry of control and miR-
141− or miR-200a–overexpressing ovarian
tumors using specific antibodies. Indicated in
each category are the numbers of mitotic cells
per mm2. Control compared to miR-141, P =
0.0008 and control compared to miR-200a,
P = 0.0017 by Student’s t test. *P ? 0.05,
**P ? 0.01, ***P ? 0.001. NS, not significant
(Student’s t test). Scale bars, 50 ?m.
© 2011 Nature America, Inc. All rights reserved.

Page 5

ARTICLES
NATURE MEDICINE VOLUME 17 | NUMBER 12 | DECEMBER 2011
1631
the correlation was not significant (R = −0.25,
P = 0.07). Tumors with high miR-200a expres-
sion showed faint p38? staining, and, con-
versely, tumors with low miR-200a expression
showed bright p38? epithelial staining (Fig. 4c). Thus, in high-grade
ovarian adenocarcinomas, regulation of p38? occurs, at least in part,
in a miR-200a–dependent manner.
We next identified transcriptomic signatures associated with miR-
200a in these tumors. Consistent with the function of miR-200 in MET,
genes involved in MET-related pathways were significantly negatively
correlated with miR-200a expression (Fig. 4d and Supplementary
Table 4). Notably, the only relevant pathway significantly positively
correlated with miR-200a expression was involved in oxidative stress
response (Fig. 4d and Supplementary Table 4), indicating that the
link between oxidative stress and miR-200a we uncovered in vitro is
also relevant in a human-tumor context.
miR-200a-dependent signatures predict subject survival
We defined ‘stress’ and ‘fibrosis’ signatures by the genes that were
positively and negatively correlated with miR-200a, respectively. The
use of this dual signature allowed us to build a hierarchical cluster
of ovarian tumors and to stratify subjects into two groups (Fig. 4e).
Subjects classified as having the stress pattern (stress high, fibrosis low)
had longer PFS and overall survival than those who had the fibrosis
pattern (stress low, fibrosis high) (Fig. 4f). In contrast, using each
single signature to stratify subjects did not allow to observe a dif-
ferential in subject survival (data not shown), suggesting that both
stress and fibrosis processes are required for fully recapitulating
miR-200–dependent survival. Overall survival was significantly dif-
ferent according to the level of p38?-targeting miR-200s (miR-141
or miR-200a), and PFS tended toward the same observation (Fig. 4f).
Notably, we validated the stress and fibrosis signatures in another
set of individuals with ovarian cancer (from the Australian Ovarian
Cancer Study (AOCS))45 (Fig. 4g) and saw better prognoses for sub-
jects with the stress pattern (Fig. 4h).
Although overexpression of miR-141 or miR-200a increased
tumor growth in mice (Fig. 3da–f), subjects with the stress pattern
(high miR-200a expression) had better prognoses than subjects with
the fibrosis pattern (low miR-200a expression) (Fig. 4e–h). This
apparent paradox might be explained by different tumor aggressive-
ness or varying responses to treatment. We observed no correlation
between signature and tumor grade (Table 1), probably because
4
Low miR-200a
High miR-200a
R = –0.018
P = NS
a
b
f
g
h
cd
e
R = –0.363
P = 0.0094
3
MAPK14 mRNA (fold)
2
1
0
4
3
miR-200a (fold)
Probability
p38?
p38?
2
1
0
1.0
0.8
P = 0.014
P = 0.2
P = 0.05
P = 0.035
0.6
0.4
0.2
0
Probability
1.0
0.8
0.6
0.4
0.2
0
1.0
0.8
0.6
0.4
0.2
0
1.0
0.8
Positively correlated genes with miR-200a
Oxidative stress response
Negatively correlated genes with miR-200a
0
0.5
Fibrosis
FAK signaling
ILK signaling
TREM1 signaling
Cdc42 signaling
Acute myoloid leukemia signaling
Actin cytoskeleton signaling
1.0
0.8
P = 0.0013
P = 0.031
0.6
Probability
0.4
0.2
0
1.0
0.8
0.6
0.4
0.2
0
050100
Stress
Stress
PFSOS
Fibrosis
Fibrosis
Time (months)
150
050 100
Time (months)
Oxidative stress response Fibrosis
2
0
–2
150200
Dentritic cell maturation
1.0
–log P
2.0
2.51.5
0
0.5
1.0
–log P
2.0
2.51.5
0.6
0.4
0.2
0
0 50100200 250150
012
p38? IHC score
345
012
p38? IHC score
Time (months)
050 100
High miR-200s + miR-141
Low miR-200s + miR-141
High miR-200s + miR-141
Oxidative stress response Fibrosis
Low miR-200s + miR-141
200 250 150
Time (months)
050 100200 250150
Time (months)
050 100200 250
EPHX1
CCR5
CCL5
LY96
CD40
PRKCH
TXN
MYL9
TIMP2
PDGFRB
FN1
IGF1
ACTA2
COL1A1 COL3A1
MYH9
TNFRSF11B
AKT1
PlK3R3
NQO1
PRDX1
AKR1A1
GSR
DNAJA4
SOD1
STIP1
NRAS
UBE2K
GSK3B
FGFR2
MYH10
DNAJC10
MAP2K6
MGST1
PRKCZ
DNAJC11 DNAJC16
IGF1
MYL9
ACTA2
COL3A1COL1A1
FN1
PDGFRB
TIMP2
MYH9
CCL5
CCR5
LY96
CD40
FGFR2
DNAJC16DNAJC11
DNAJA4
TXN
PRKCZ
NQO1
AKR1A1
PRDX1
GSR
EPHX1
PRKCH
HSP90AA1
PIK3R3
AKT1
STIP1
UBE2K
NRAS
DNAJC10
GSK3B
MAP2K6
MGST1
SOD1
MYH10
TNFRSF11B
Fibrosis up, stress down
Fibrosis down, stress up
150
Time (months)
PFS
Stress
Stress
Fibrosis
Fibrosis
OS
PFS OS
345
Fibrosis up, stress down
Fibrosis down, stress up
2
0
–2
Figure 4 miR-200a and oxidative stress
response predict good prognosis in patients with
ovarian cancer. (a,b) Scatter diagrams showing
the p38? immunohistochemistry (IHC) score
relative to MAPK14 mRNA (a) and miR-200a
levels (b) in ovarian tumors (Spearman test).
(c) Representative views of p38? immunostaining
in tumors with low or high miRNA-200a levels.
(d) Cellular pathways positively or negatively
correlated with miR-200a levels in ovarian
tumors. An unsupervised comparative analysis
was done using a Fisher’s exact test. P values
were adjusted using a Benjamini-Hochberg
multiple-testing correction. (e) An unsupervised
hierarchical clustering of the stress and fibrosis
signatures from the microarray ovarian cancer
data set from the Institut Curie. Each row
represents a tumor, and each column represents
a gene. Red and blue indicate gene expression
in a single tumor above and below the mean,
respectively. The color saturation indicates the
magnitude of the deviation from the mean.
Genes cluster together according to the stress
and fibrosis signatures (shown at the bottom of
the matrix). The dendrogram of samples (on the
left of the matrix) allows for the classification
of subjects according to these signatures. Gene
symbols and names are listed in Supplementary
Table 4. (f) Kaplan-Meier curves of PFS and
overall survival (OS) according to the two
signatures (n = 51 for the stress signature, and
n = 56 for the fibrosis signature) or to miR-
200a or miR-141 expression (n = 46 for high
expression and n = 45 for low expression).
(g) Unsupervised hierarchical clustering of the
signatures from the AOCS microarray data set.
(h) Kaplan-Meier curves of PFS and overall
survival of subjects from the AOCS data set
according to the stress (n = 158) and fibrosis
(n = 128) signatures. A log-rank test was used
for the Kaplan-Meier curves. Scale bars, 50 ?m.
© 2011 Nature America, Inc. All rights reserved.

Page 6

ARTICLES
1632
VOLUME 17 | NUMBER 12 | DECEMBER 2011 NATURE MEDICINE
there is a bias toward high-grade tumors in the cohorts from the
Institut Curie and the AOCS (Supplementary Table 1). The fibrosis
pattern was associated with partial debulking and high-stage tumors
(Table 1). In addition, when considering tumor residual volume after
the first round of treatment (as evaluated by plasma concentrations
of cancer antigen 125, also called carbohydrate antigen 125, CA-
125), we observed that the CA-125 concentrations were significantly
lower in subjects with the stress pattern compared to those with
the fibrosis pattern (Table 1). Similarly, physicians considered the
clinical response, defined by variation in tumor mass after treat-
ment, to be ‘complete’ (with significant reduction of tumor mass) in a
large proportion of subjects with the stress pattern (Table 1). Finally,
a multivariate analysis that included the miR-200a–dependent
signatures (stress or fibrosis), age, tumor grade, histology and
chemotherapy type showed that the signatures had an independent
association with PFS (Supplementary Table 5). These observations
suggest that the stress and fibrosis signatures defined by miR-200a
may act on a patient’s response to treatment.
We next investigated the effect of paclitaxel, a ROS-producing chemo-
therapeutic agent, on ovarian cancer cells overexpressing miRNA.
Overexpression of miR-141 or miR-200a enhanced cell apoptosis
after treatment with paclitaxel, whereas it protected cells against death
in untreated conditions (Fig. 5a). As expected, the same results were
obtained by inactivation of p38? (Fig. 5a). Moreover, we prevented
the miR-141- and miR-200a–dependent effects on apoptosis by using
antioxidants or inhibitors specific to miR-141 or miR-200a, further
confirming the role of ROS, miR-141 and miR-200a in the sensitivity of
cancer cells to paclitaxel (Fig. 5a). Notably, we observed similar effects
in vivo. Whereas overexpression of miR-141 or miR-200a significantly
increased tumor growth in untreated conditions (Figs. 5b and 3d–f),
mice with ovarian tumors overexpressing miR-141 or miR-200a were
significantly more sensitive to paclitaxel than control mice (Fig. 5b–d).
Therefore, expression of miR-141 or miR-200a promotes tumor growth
but also increases sensitivity to chemotherapy. Our data prompted us
to establish a proposed model, as described in Figure 5e.
DISCUSSION
Ovarian cancers are the most lethal gynecologic malignancies, and
determining the molecular mechanisms involved in their development
is a pressing need. The miR-200s were previously shown to accumulate
in ovarian cancer38–41,46,47. Here we identify an miR-200a–dependent
dual signature involved in oxidative stress and fibrosis that has
predictive value. The fibrosis signature, defined by genes negatively
correlated with miR-200a expression, corroborates its previously
identified function on ZEB1 and ZEB2. Indeed, by inhibiting ZEB1
and ZEB2, ectopic expression of the miR-200s causes upregulation of
E-cadherin and reduces cell motility22–27,33,41. The stress signature,
defined by genes correlated with miR-200a expression, is the novel
feature of our study. The fibrosis and stress signatures predict survival
only when both are taken into account. Stratifying subjects with the
fibrosis signature alone did not predict survival, further indicating
that the stress component is a key effector of miR-200a and miR-141
in ovarian tumors. The miR-200s were shown to be highly expressed
in localized tumors and downregulated in metastases, defining a
two-stage model of miR-200 expression48–50. Our data are consistent
with this model in two ways: we observed (i) a stage of high miR-200
expression, which provided a selective advantage to cancer cells but
was associated with high sensitivity to treatment, in part by modulat-
ing the stress response, followed by (ii) a dissemination stage during
which miR-200 expression was lost, further facilitating the spread of
cancer cells and conferring resistance to chemotherapeutic agents.
Consistent with our signatures, several studies have shown that high
miR-200s expression is linked to a favorable prognosis40,41,51, with
downregulation of the miR-200s being associated with relapse in
patients with ovarian cancer52. The segregation of high-grade ovarian
carcinomas into miR-200a–dependent stress and fibrosis subgroups is
associated with short- and long-term prognoses, specifically in regard
to clinical response and disease progression. Our findings have broad
implications for applications in clinical research. One could specu-
late that tumors with the stress pattern may be best treated with an
optimal combination of ROS-producing chemotherapy and maximal
surgery. In contrast, tumors with the fibrosis pattern, characterized by
a low debulking efficiency and reduced treatment response, should
be subjected to alternative approaches based on further molecular
characterization. Future clinical trials may consider our results to
generate prospective confirmation data. Such approaches could
greatly enhance the multidisciplinary strategy currently applied in
ovarian carcinomas.
In this paper we show a new mechanism of action for miR-141 and
miR-200a in oxidative stress response by identifying p38? MAPK as
one of their relevant targets. Alignments of MAPK14 3? untranslated
region sequences from various species indicate that Mpk2, the p38?
homolog in Drosophila, is predicted to be a target of miR-8, the miR-141
and miR-200a ortholog53–55. Because miR-8 regulates the osmotic
stress response, the crosstalk between miR-200, p38? and stress could
be conserved through evolution. p38? MAPK was previously shown
to reduce tumorigenesis by acting on cell proliferation, survival and
stress response14,15. miRNA-dependent downregulation of p38? mim-
ics the phenotype observed in p38? deficiency10. In addition, inacti-
vation of p38?, either pharmacologically or genetically, is associated
Table 1 Characteristics of subjects and tumor samples associated with the stress and fibrosis signatures
Grade
123I
StageDebulking status
Optimal Suboptimal
Biological response
Log CA-125
1.52 ? 0.10
1.92 ? 0.11
P = 0.010
Clinical response
Complete IncompleteIIIIIIV
Institut CurieStress
Fibrosis
P
Stress
Fibrosis
P
4
3
17
17
NS
56
41
NS
30
36
15
6
7
3
26
33
325
13
26
43
28
23
14
26 14
P = 0.004
16
2
P = 0.002
P = 0.008
95
65
P = 0.014
P = 0.059
AOCS9 82
82
16
8
108
109
8 29
411014
Association of the stress and fibrosis signatures with clinical characteristics and remission status of patients in the Institut Curie and AOCS cohorts, when available. The biological
response was evaluated by CA-125 after the first round of treatment. The clinical response was evaluated by the evolution of the mass of the tumor, determined by monitoring
patients through their chemotherapeutic treatment; treatment was considered incomplete in patients with no or partial response to treatment. Debulking status was defined as
optimal for tumor residues ?1 cm in diameter after resection and as suboptimal for tumor residues >1 cm in diameter. Data are means of log values ? s.e.m. The P values were
determined using ?2 (for grade, stage and clinical response), Fischer exact (for debulking status) or Student’s t (for CA-125) tests. NS, not significant.
© 2011 Nature America, Inc. All rights reserved.

Page 7

ARTICLES
NATURE MEDICINE VOLUME 17 | NUMBER 12 | DECEMBER 2011
1633
with ROS accumulation and the subsequent
stimulation of antioxidant defenses mediated
in part by Nrf2 (refs. 10,56). Nrf2-dependent
genes are also markedly upregulated in ovarian
tumors with high miR-200a expression and
subsequent low amounts of p38?. miR-200a
overexpression, as well as p38? inactivation,
sensitizes ovarian cancer cells to paclitaxel, which is prevented with
the use of antioxidants, further confirming the essential role of ROS
in sensitivity to paclitaxel34–36,57. Recently published results show-
ing that miR-200b, miR-200c and miR-429 target class III ?-tubulin
provided a molecular clue into the increased paclitaxel sensitivity
in ovarian tumors overexpressing miR-200 (refs. 42,58). Our study
provides new findings about other miR-200 family members, specifi-
cally miR-200a and miR-141, that increase paclitaxel sensitivity in an
ROS-dependent manner by targeting p38?. To our knowledge, this
is the first time that the effects of miR-141 and miR-200a have been
shown on tumor growth and chemosensitivity using mouse models
of ovarian cancer that corroborated human clinical data. Our results
show a new function for miR-200 miRNAs and strengthen their direct
implications in the antineoplastic effect of paclitaxel.
Our data indicate that oxidative stress has a dual role: it increases
the growth rate of tumors and further promotes tumorigenesis, but it
also potentiates the effect of ROS-producing therapeutic agents, for
example, paclitaxel. These data may also provide new insights into
clinical trials showing the lack of an obvious positive effect of anti-
oxidant supplementation at pharmacological doses in the treatment
of cancer. Conversely, it has been shown that cancer cells are very
sensitive to ROS-producing agents, including piperlongumine59. The
oxidative stress signature identified here could therefore be used as a
predictive marker of response to treatment and survival. In conclusion,
we identified a new reciprocal interaction between the miR-200s
and the oxidative stress response that affects human ovarian carcino-
genesis and prognosis.
METHODS
Methods and any associated references are available in the online
version of the paper at http://www.nature.com/naturemedicine/.
Accession codes. The Institut Curie’s ovarian cancer microarray
data set is freely accessible in the Gene Expression Omibus under
a
b
c
e
d
Tumor growth inhibition (%)
25
20
15
10
5
Apoptotic cells (%)
0
Control
Untreated
***
*
*
miR-141
miR-200a
si-p38
25
20
15
10
5
0
+ paclitaxel
***
***
***
Control
miR-141
miR-200a
si-p38
25
20
15
10
5
0
Control
+ Paclitaxel
+ anti-miRNA
miR-141
miR-200a
25
20
15
10
5
0
Control
+ Paclitaxel + NAC
miR-141
miR-200a
si-p38
–100
–150
–200
***
*
ControlmiR-141
miR-200a
Control
0
–50
50
–100
–150
–200
–250
–300
0
–50
50
–100
–150
–200
–250
–300
miR-141
Tumor growth inhibition (%)
0
–50
0
–50
50
–100
–150
–200
–250
–300
miR-200a
Tumor volume (mm3)
Control
Control + paclitaxel
600
500
400
300
200
100
0
0246
Time (d)
810 12 14
600
500
***
***
***
***
***
***
***
***
***
**
**
**
**
400
300
200
100
0
0246
Time (d)
8 10 12 14
miR-141
miR-141 + paclitaxel
***
***
**
**
**
**
**
**
**
**
**
**
**
600
500
400
300
200
100
0
0246
Time (d)
8 10 12 14
miR-200a
miR-200a + paclitaxel
Normal cellsCancerous cells
Step 1 : high miR-200
Cancerous cells
Step 2 : low miR-200
miR-200bc
miR-200a or
miR-141
miR-200bc
miR-200a or
miR-141
miR-200bc
miR-200a or
miR-141
p38?
p38?
ROSROS
Paclitaxel sensitive Paclitaxel resistant
Oxidative
stress
Oxidative
stress
Fibrosis
Fibrosis
ROS
Fibrosis
Oxidative
stress
P
P
p38?
P
S
I
G
N
A
T
U
R
E
S
S
I
G
N
A
T
U
R
E
S
S
I
G
N
A
T
U
R
E
S
Figure 5 miR-141 and miR-200a enhance the
response to chemotherapeutic reagents.
(a) Histograms showing the percentage of
apoptotic (annexin V positive and DAPI
negative) SKOV3 cells transfected with miR-141,
miR-200a or siRNA-p38 (si-p38) with
or without treatment. NAC, N-acetylcysteine.
(b) Growth curves of control, miR-141− or miR-
200a–overexpressing tumors under untreated
conditions (solid lines) or after treatment with
paclitaxel (dashed lines). (c) Percentage of
tumor growth inhibition after 8 d of treatment
with paclitaxel. n ? 11 tumors in each category
for each condition (treated or untreated).
(d) Percentage of tumor growth inhibition per
tumor after 8 d of treatment with paclitaxel.
(e) In normal cells (left), the reciprocal crosstalk
between miR-141 or miR-200a, p38? and ROS
allows for a mutual balance of each partner,
further ensuring the integrity of the MAPK stress
response pathways and low concentrations of ROS.
In cancer cells, two successive stages can be
considered, as has been previously stated48–50.
At the early stage (middle), one can postulate
that concentrations of ROS rise because, at
least in part, of oncogenic transformation and
metabolic alterations. High concentrations of
ROS promote the accumulation of the miR-
200s, which, in turn, reduce p38?. Chronically
elevated ROS and miR-200s drive a persistent
oxidative stress response, which is associated
with an improved sensitivity to paclitaxel.
At the advanced tumor stage (right), the
downregulation of the miR-200s directly boosts
p38? and ZEB1 or ZEB2 activity, thus restoring
the normal stress response and inducing
an epithelial-mesenchymal transition–like
phenotype, which could be part of the fibrosis
signature. Data are means ? s.e.m. *P ? 0.05,
**P ? 0.01, ***P ? 0.001 (Student’s t test).
© 2011 Nature America, Inc. All rights reserved.

Page 8

ARTICLES
1634
VOLUME 17 | NUMBER 12 | DECEMBER 2011 NATURE MEDICINE
the accession number GSE26193. H2O2-induced miRNA in the mouse
fibroblast microarray data and data from fibroblasts overexpressing
miR-141 or miR-200a are freely accessible in the Gene Expression
Omnibus with the following accession numbers: GSE26194
and GSE26113.
Note: Supplementary information is available on the Nature Medicine website.
ACKNOWLEDGMENTS
We thank O. Delattre and S. Chanock for fruitful discussions and comments on the
manuscript. We acknowledge S. Alran and B. Baranger (Surgery Department of
Institut Curie) and the Biological Resource Center of Institut Curie for providing
human ovarian tumors and B. Hasselain and D. Hajage for advice regarding our
statistical analyses. We thank the members of the animal facility and the flow-
cytometry platform of Institut Curie for their expertise. The experimental work
was supported by grants from Institut National de la Santé et de la Recherche
Médicale, the Institut Curie, the Ligue Nationale Contre le Cancer, the Institut
National du Cancer and the Association pour la Recherche Contre le Cancer. B.M.
was supported by a post-doctoral fellowship from the INSERM Avenir program
and the Association pour la Recherche Contre le Cancer.
AUTHOR CONTRIBUTIONS
B.M. and F.M.-G. participated in the conception and design of the experiments.
B.M., L.B., M.C., T.G. and Y.d.F. performed the experiments. X.S.-G. selected
the human ovarian cancers after adapted characterization, and J.-P.M. provided
transcriptome data. P.C. provided the associated clinical data from the subjects.
O.M. and A.N. provided technical assistance and expertise in the ovarian tumor
sample preparation. J.-P.M., B.M., L.B. and M.C. contributed to the statistical
analyses of the data. F.M.-G. wrote the paper with suggestions and comments from
all authors.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
Published online at http://www.nature.com/naturemedicine/.
Reprints and permissions information is available online at http://www.nature.com/
reprints/index.html.
1. Hennessy, B.T., Coleman, R.L. & Markman, M. Ovarian cancer. Lancet 374,
1371–1382 (2009).
2. The Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian
carcinoma. Nature 474, 609–615 (2011).
3. Allen, R.G. & Tresini, M. Oxidative stress and gene regulation. Free Radic. Biol.
Med. 28, 463–499 (2000).
4. Gerald, D. et al. JunD reduces tumor angiogenesis by protecting cells from oxidative
stress. Cell 118, 781–794 (2004).
5. Pouysségur, J. & Mechta-Grigoriou, F. Redox regulation of the hypoxia-inducible
factor. Biol. Chem. 387, 1337–1346 (2006).
6. Laurent, G. et al. Oxidative stress contributes to aging by enhancing pancreatic
angiogenesis and insulin signaling. Cell Metab. 7, 113–124 (2008).
7. Weinberg, F. & Chandel, N.S. Reactive oxygen species–dependent signaling regulates
cancer. Cell. Mol. Life Sci. 66, 3663–3673 (2009).
8. Pani, G., Galeotti, T. & Chiarugi, P. Metastasis: cancer cell?s escape from oxidative
stress. Cancer Metastasis Rev. 29, 351–378 (2010).
9. Toullec, A. et al. Oxidative stress promotes myofibroblast differentiation and tumour
spreading. EMBO Mol. Med. 2, 211–230 (2010).
10. Dolado, I. et al. p38? MAP kinase as a sensor of reactive oxygen species in
tumorigenesis. Cancer Cell 11, 191–205 (2007).
11. Kennedy, N.J., Cellurale, C. & Davis, R.J. A radical role for p38 MAPK in tumor
initiation. Cancer Cell 11, 101–103 (2007).
12. Hui, L. et al. p38? suppresses normal and cancer cell proliferation by antagonizing
the JNK-c-Jun pathway. Nat. Genet. 39, 741–749 (2007).
13. Wada, T. et al. Antagonistic control of cell fates by JNK and p38-MAPK signaling.
Cell Death Differ. 15, 89–93 (2008).
14. Wagner, E.F. & Nebreda, A.R. Signal integration by JNK and p38 MAPK pathways
in cancer development. Nat. Rev. Cancer 9, 537–549 (2009).
15. Kim, E.K. & Choi, E.J. Pathological roles of MAPK signaling pathways in human
diseases. Biochim. Biophys. Acta 1802, 396–405 (2010).
16. Marsit, C.J., Eddy, K. & Kelsey, K.T. MicroRNA responses to cellular stress. Cancer
Res. 66, 10843–10848 (2006).
17. Ivan, M., Harris, A.L., Martelli, F. & Kulshreshtha, R. Hypoxia response and microRNAs:
no longer two separate worlds. J. Cell Mol. Med. 12, 1426–1431 (2008).
18. Lin, Y. et al. Involvement of microRNAs in hydrogen peroxide–mediated gene
regulation and cellular injury response in vascular smooth muscle cells. J. Biol.
Chem. 284, 7903–7913 (2009).
19. Simone, N.L. et al. Ionizing radiation–induced oxidative stress alters miRNA
expression. PLoS ONE 4, e6377 (2009).
20. Wang, Z. et al. Profiles of oxidative stress–related microRNA and mRNA expression
in auditory cells. Brain Res. 1346, 14–25 (2010).
21. Magenta, A. et al. miR-200c is up-regulated by oxidative stress and induces
endothelial cell apoptosis and senescence via ZEB1 inhibition. Cell Death Differ. 18,
1628–1639 (2011).
22. Hurteau, G.J., Carlson, J.A., Spivack, S.D. & Brock, G.J. Overexpression of the
microRNA hsa-miR-200c leads to reduced expression of transcription factor 8 and
increased expression of E-cadherin. Cancer Res. 67, 7972–7976 (2007).
23. Bracken, C.P. et al. A double-negative feedback loop between ZEB1–SIP1 and the
microRNA-200 family regulates epithelial-mesenchymal transition. Cancer Res. 68,
7846–7854 (2008).
24. Burk, U. et al. A reciprocal repression between ZEB1 and members of the
miR-200 family promotes EMT and invasion in cancer cells. EMBO Rep. 9,
582–589 (2008).
25. Gregory, P.A. et al. The miR-200 family and miR-205 regulate epithelial to
mesenchymal transition by targeting ZEB1 and SIP1. Nat. Cell Biol. 10, 593–601
(2008).
26. Korpal, M., Lee, E.S., Hu, G. & Kang, Y. The miR-200 family inhibits epithelial-
mesenchymal transition and cancer cell migration by direct targeting of E-cadherin
transcriptional repressors ZEB1 and ZEB2. J. Biol. Chem. 283, 14910–14914
(2008).
27. Park, S.M., Gaur, A.B., Lengyel, E. & Peter, M.E. The miR-200 family determines
the epithelial phenotype of cancer cells by targeting the E-cadherin repressors ZEB1
and ZEB2. Genes Dev. 22, 894–907 (2008).
28. Shimono, Y. et al. Downregulation of miRNA-200c links breast cancer stem cells
with normal stem cells. Cell 138, 592–603 (2009).
29. Wellner, U. et al. The EMT-activator ZEB1 promotes tumorigenicity by repressing
stemness-inhibiting microRNAs. Nat. Cell Biol. 11, 1487–1495 (2009).
30. Iliopoulos, D. et al. Loss of miR-200 inhibition of Suz12 leads to polycomb-
mediated repression required for the formation and maintenance of cancer stem
cells. Mol. Cell 39, 761–772 (2010).
31. Schickel, R., Park, S.M., Murmann, A.E. & Peter, M.E. miR-200c regulates
induction of apoptosis through CD95 by targeting FAP-1. Mol. Cell 38, 908–915
(2010).
32. Chang, C.J. et al. p53 regulates epithelial-mesenchymal transition and stem
cell properties through modulating miRNAs. Nat. Cell Biol. 13, 317–323
(2011).
33. Kim, T. et al. p53 regulates epithelial-mesenchymal transition through microRNAs
targeting ZEB1 and ZEB2. J. Exp. Med. 208, 875–883 (2011).
34. Ramanathan, B. et al. Resistance to paclitaxel is proportional to cellular total
antioxidant capacity. Cancer Res. 65, 8455–8460 (2005).
35. Alexandre, J., Hu, Y., Lu, W., Pelicano, H. & Huang, P. Novel action of paclitaxel
against cancer cells: bystander effect mediated by reactive oxygen species. Cancer
Res. 67, 3512–3517 (2007).
36. Konstantinopoulos, P.A. et al. Carboplatin-induced gene expression changes in vitro are
prognostic of survival in epithelial ovarian cancer. BMC Med. Genomics 1, 59
(2008).
37. Liu, H. et al. mRNA and microRNA expression profiles of the NCI-60 integrated
with drug activities. Mol. Cancer Ther. 9, 1080–1091 (2010).
38. Iorio, M.V. et al. MicroRNA signatures in human ovarian cancer. Cancer Res. 67,
8699–8707 (2007).
39. Nam, E.J. et al. MicroRNA expression profiles in serous ovarian carcinoma.
Clin. Cancer Res. 14, 2690–2695 (2008).
40. Hu, X. et al. A miR-200 microRNA cluster as prognostic marker in advanced ovarian
cancer. Gynecol. Oncol. 114, 457–464 (2009).
41. Bendoraite, A. et al. Regulation of miR-200 family microRNAs and ZEB transcription
factors in ovarian cancer: evidence supporting a mesothelial-to-epithelial transition.
Gynecol. Oncol. 116, 117–125 (2010).
42. Cochrane, D.R., Howe, E.N., Spoelstra, N.S. & Richer, J.K. Loss of miR-200c: a
marker of aggressiveness and chemoresistance in female reproductive cancers.
J. Oncol. 2010, 821717 (2010).
43. Mezzanzanica, D., Bagnoli, M., De Cecco, L., Valeri, B. & Canevari, S. Role of
microRNAs in ovarian cancer pathogenesis and potential clinical implications.
Int. J. Biochem. Cell Biol. 42, 1262–1272 (2010).
44. van Jaarsveld, M.T., Helleman, J., Berns, E.M. & Wiemer, E.A. MicroRNAs in ovarian
cancer biology and therapy resistance. Int. J. Biochem. Cell Biol. 42, 1282–1290
(2010).
45. Tothill, R.W. et al. Novel molecular subtypes of serous and endometrioid ovarian
cancer linked to clinical outcome. Clin. Cancer Res. 14, 5198–5208 (2008).
46. Yang, H. et al. MicroRNA expression profiling in human ovarian cancer: miR-214
induces cell survival and cisplatin resistance by targeting PTEN. Cancer Res. 68,
425–433 (2008).
47. Wyman, S.K. et al. Repertoire of microRNAs in epithelial ovarian cancer as determined
by next generation sequencing of small RNA cDNA libraries. PLoS ONE 4,
e5311 (2009).
48. Olson, P. et al. MicroRNA dynamics in the stages of tumorigenesis correlate with
hallmark capabilities of cancer. Genes Dev. 23, 2152–2165 (2009).
49. Gibbons, D.L. et al. Contextual extracellular cues promote tumor cell EMT and
metastasis by regulating miR-200 family expression. Genes Dev. 23, 2140–2151
(2009).
© 2011 Nature America, Inc. All rights reserved.

Page 9

ARTICLES
NATURE MEDICINE VOLUME 17 | NUMBER 12 | DECEMBER 2011
1635
50. Wiklund, E.D. et al. Coordinated epigenetic repression of the miR-200 family and
miR-205 in invasive bladder cancer. Int. J. Cancer 128, 1327–1334 (2011).
51. Eitan, R. et al. Tumor microRNA expression patterns associated with resistance to
platinum based chemotherapy and survival in ovarian cancer patients. Gynecol.
Oncol. 114, 253–259 (2009).
52. Marchini, S. et al. Association between miR-200c and the survival of patients with
stage I epithelial ovarian cancer: a retrospective study of two independent tumour
tissue collections. Lancet Oncol. 12, 273–285 (2011).
53. Karres, J.S., Hilgers, V., Carrera, I., Treisman, J. & Cohen, S.M. The conserved
microRNA miR-8 tunes atrophin levels to prevent neurodegeneration in Drosophila.
Cell 131, 136–145 (2007).
54. Hyun, S. et al. Conserved MicroRNA miR-8/miR-200 and its target USH/FOG2
control growth by regulating PI3K. Cell 139, 1096–1108 (2009).
55. Flynt, A.S. et al. miR-8 microRNAs regulate the response to osmotic stress in
zebrafish embryos. J. Cell Biol. 185, 115–127 (2009).
56. Naidu, S., Vijayan, V., Santoso, S., Kietzmann, T. & Immenschuh, S. Inhibition and
genetic deficiency of p38 MAPK up-regulates heme oxygenase-1 gene expression
via Nrf2. J. Immunol. 182, 7048–7057 (2009).
57. Helleman, J., Smid, M., Jansen, M.P., van der Burg, M.E. & Berns, E.M. Pathway
analysis of gene lists associated with platinum-based chemotherapy resistance in
ovarian cancer: the big picture. Gynecol. Oncol. 117, 170–176 (2010).
58. Leskelä, S. et al. The miR-200 family controls ?-tubulin III expression and is
associated with paclitaxel-based treatment response and progression-free survival
in ovarian cancer patients. Endocr. Relat. Cancer 18, 85–95 (2011).
59. Raj, L. et al. Selective killing of cancer cells by a small molecule targeting the
stress response to ROS. Nature 475, 231–234 (2011).
© 2011 Nature America, Inc. All rights reserved.

Page 10

NATURE MEDICINE
doi:10.1038/nm.2512
ONLINE METHODS
Ovarian cancer sample collection. Ovarian tumors were obtained from a cohort
of patients treated at the Institut Curie between 1989 and 2005. The median age
of the patients was 57.8 years (with a range of 31–86 years). Ovarian carcinomas
were classified according to the World Health Organization histological clas-
sification of gynecological tumors. Pathological analysis identified 82 serous
tumors, 8 mucinous tumors, 8 endometrioid tumors, 6 clear-cell carcinomas,
2 carcinosarcomas and 1 malignant Brenner tumor. One-hundred subjects were
classified as having a high (grade 2 and 3) histological grade, and seven subjects
were classified as having a low (grade 1) histological grade. Thirty-one subjects
(29%) were considered to be in the early stages (International Federation of
Gynecology and Obstetrics (FIGO) I–IIc) and 76 subjects (71%) were considered
to be in an advanced stage (III or IV) of disease. Patients were treated with a
combination of surgery and chemotherapy, the latter of which included alkylat-
ing or alkylating-like agents with or without taxane as a first-line treatment. The
study was approved by the Institutional Review Board and Ethics committee of
the Institut Curie. Before inclusion in the study, patients were informed that
their biological samples could be used for research purposes and that they had
the right to refuse if they so wished. Analysis of tumor samples was performed
according to the relevant national law on the protection of people taking part
in biomedical research.
qRT-PCR of mRNA and miRNA in ovarian tumors and cell lines. For the
mRNA analysis, 1 ?g of total RNA was reverse transcribed using an iScript Reverse
Transcription Kit (Bio-Rad), and qRT-PCR was performed using Power SYBR
Green PCR Master Mix (Applied Biosystems) on a Chromo4 System (Bio-Rad).
Data were analyzed using an Opticon Monitor (Bio-Rad) and normalized
to GAPDH mRNA (cell lines) or to the average of hypoxanthine guanine
phosphoribosyl transferase (HPRT), Ribosomal protein large P0 (RPLP0),
glyceraldhyde-3-phosphate dehydrogenase (GAPDH) and ?-2 microglobu-
lin (B2M) mRNA levels (ovarian tumors) (see Supplementary Table 6 for
sequences). Mature miRNA levels were quantified using TaqMan miRNA Assays
(Applied Biosystems) and normalized to the average of U6 small nucleolar
RNA and miR-16 (cell lines) or to the average of U6B, U6, RNU24, RNU49 and
RNU48 small RNA levels (ovarian tumors); fold changes were calculated using
the 2−??Ct method. Relative miR-200a and miR-141 levels were quantified from
RNA from a subset of 82 ovarian tumors.
Ovarian microarray data sets and gene expression profiling. Ovarian tumor
samples were analyzed on a Human Genome U133 Plus 2.0 array (Affymetrix)
according to the manufacturer’s procedures60. Transcriptome data were normal-
ized using the Guanine Cytosine Robust Multiarray Analysis (GCRMA) algo-
rithm. Only probes with a log intensity value greater than 3.5 (log2) in at least
80% of the samples were kept for further analysis. The microarray normalization
and analysis were performed using Partek Genomic Suite software (Partek). The
dataset from the Institute Curie, from which we determined miR-200a expres-
sion values, the correlation between mature miR-200a levels and the detected
probe sets, was calculated using a Pearson correlation coefficient. The probe
sets, identified as either positively or negatively correlated with miR-200a, were
submitted to analysis using IPA software. Enriched canonical pathways were
selected using a Fisher exact test, and adjusted P values (P ? 0.05) were calcu-
lated using a Benjamini-Hochberg multiple-testing correction. For the probe
sets that were correlated with miR-200a, the most significant Ingenuity canoni-
cal pathway was “Nrf2-mediated oxidative stress response,” and for probe sets
anticorrelated with miR-200a, the most significant Ingenuity canonical pathway
was “Hepatic fibrosis–hepatic stellate cell activation.” Genes composing the stress
and fibrosis signatures are listed in Supplementary Table 4.
Immunohistochemistry from human ovarian carcinomas. A tissue micro-
array from 56 ovarian serous adenocarcinomas was created using two cores of
tumor tissue per subject (each 1 mm in diameter) hybridized simultaneously.
Sections of paraffin-embedded tissue (3 ?m) were stained using the streptavidin-
peroxidase protocol (Benchmark Immunostainer, Ventana) with specific anti-
bodies recognizing p38? (1:50 dilution; 9218, Cell Signaling). For quantification,
two sections from distinct areas of each tumor were evaluated independently
by two different investigators. An immunohistochemistry score (ranging from
0 to 4) was calculated as follows: (staining intensity × percentage of positively
labeled cells)/100.
miRNA microarray data from mouse fibroblasts and fibroblasts overexpress-
ing miR-141 or miR-200a. The methods used for this analysis are described in
the Supplementary Methods.
Xenograft experiments. The cell lines and antibodies used in these experiments
are described in the Supplementary Methods. Graft experiments were per-
formed by subcutaneous injection of 2 × 105 fibroblasts or 3 × 106 exponentially
growing SKOV3 cells into 6-week-old nude female mice. The mice were checked
daily for tumor growth using calipers. Tumor volume was determined using the
following equation: 0.5 × (length × width2). When tumors reached a volume of
75 mm3, paclitaxel treatment was triggered using a single intraperitoneal injec-
tion of paclitaxel (Hospira) at 30 mg per kg of body weight. The percentage of
tumor growth inhibition was calculated using the following formula: 100 − 100 ×
(tumor volume from untreated mice/tumor volume from paclitaxel-treated
mice). The Institut Curie’s ethical committee approved all mouse experiments.
Apoptosis assays. Apoptosis was monitored by annexin V (561012, BD
Biosciences) and DAPI (1 ?g ml−1) staining according to the manufacturer’s
instructions. Where indicated, cells were treated with 20 nM paclitaxel for 48 h
and with 4 mM NAC daily. Apoptotic cells were defined as the population that was
positive for annexin and negative for DAPI. FACS experiments were conducted
on an LSRII cytometer, and data were analyzed using FlowJo 9.1 software.
Statistical analyses. All experiments were performed in triplicate, and data
shown are means ? s.e.m. (unless otherwise specified) from at least three inde-
pendent experiments. Differences were considered to be statistically significant
at values of P ? 0.05 by Student’s t test or Mann-Whitney test. Single, double and
triple asterisks indicate statistically significant differences: *P ? 0.05; **P ? 0.01;
***P ? 0.001. All survival analyses were carried out using Kaplan-Meier method
and log-rank test in R. Univariate or multivariate Cox proportional hazards
regression was conducted with SPSS 19.0 software using the enter method.
60. Meyniel, J.P. et al. A genomic and transcriptomic approach for a differential
diagnosis between primary and secondary ovarian carcinomas in patients with a
previous history of breast cancer. BMC Cancer 10, 222 (2010).