The Journal of Clinical Investigation http://www.jci.org
ZEB1 drives prometastatic actin
cytoskeletal remodeling by
downregulating miR-34a expression
Young-Ho Ahn,1 Don L. Gibbons,1,2 Deepavali Chakravarti,3 Chad J. Creighton,4
Zain H. Rizvi,1 Henry P. Adams,5 Alexander Pertsemlidis,6 Philip A. Gregory,7,8
Josephine A. Wright,7,8 Gregory J. Goodall,7,8,9 Elsa R. Flores,3 and Jonathan M. Kurie1
1Department of Thoracic/Head and Neck Medical Oncology, 2Department of Molecular and Cellular Oncology, and
3Department of Biochemistry and Molecular Biology, University of Texas MD Anderson Cancer Center, Houston, Texas, USA. 4Dan L. Duncan Cancer Center,
Baylor College of Medicine, Houston, Texas, USA. 5Department of Genetics, University of Texas MD Anderson Cancer Center, Houston, Texas, USA.
6Greehey Children’s Cancer Research Institute, Department of Pediatrics, and Department of Cell and Structural Biology,
UT Health Science Center at San Antonio, San Antonio, Texas, USA. 7Centre for Cancer Biology, SA Pathology, Adelaide, South Australia, Australia.
8Discipline of Medicine and 9School of Molecular and Biomedical Science, University of Adelaide, Adelaide, South Australia, Australia.
Metastatic cancer is extremely difficult to treat, and the presence of metastases greatly reduces a cancer
patient’s likelihood of long-term survival. The ZEB1 transcriptional repressor promotes metastasis through
downregulation of microRNAs (miRs) that are strong inducers of epithelial differentiation and inhibitors
of stem cell factors. Given that each miR can target multiple genes with diverse functions, we posited that the
prometastatic network controlled by ZEB1 extends beyond these processes. We tested this hypothesis using
a mouse model of human lung adenocarcinoma metastasis driven by ZEB1, human lung carcinoma cells,
and human breast carcinoma cells. Transcriptional profiling studies revealed that ZEB1 controls the expres-
sion of numerous oncogenic and tumor-suppressive miRs, including miR-34a. Ectopic expression of miR-34a
decreased tumor cell invasion and metastasis, inhibited the formation of promigratory cytoskeletal struc-
tures, suppressed activation of the RHO GTPase family, and regulated a gene expression signature enriched
in cytoskeletal functions and predictive of outcome in human lung adenocarcinomas. We identified several
miR-34a target genes, including Arhgap1, which encodes a RHO GTPase activating protein that was required
for tumor cell invasion. These findings demonstrate that ZEB1 drives prometastatic actin cytoskeletal remod-
eling by downregulating miR-34a expression and provide a compelling rationale to develop miR-34a as a
therapeutic agent in lung cancer patients.
Metastasis currently represents a tipping point in a cancer patient’s
likelihood of achieving long-term survival. Metastatic deposits
cannot be eradicated with any standard treatment options and are
the most common cause of death from epithelial malignancies (1).
Understanding the basic biological processes that govern metasta-
sis represents a critical barrier to attaining long-term survival for
patients afflicted with epithelial cancer.
In one working hypothesis, metastasis is initiated by a popu-
lation of tumor cells that undergo epithelial-to-mesenchymal
transition (EMT) in response to extracellular cues, leading to
loss of polarized features, detachment from neighboring cells,
increased motility, and invasion into surrounding matrix (2).
Context-dependent cellular plasticity has been identified in
side populations of established human cancer cell lines and
in cell lines derived from mouse models of human epithelial
cancers; these cells are marked by increased expression of alde-
hyde dehydrogenase or prominin-1 (CD133) (3, 4). The unique
biological features of these cells and their dearth within pri-
mary tumors has led to the belief that they originate from rare
populations of pluripotent cells or have uncommon gain-of-
function somatic mutations (5).
EMT is driven by several families of transcriptional repressors
(ZEB, SNAIL, and basic helix-loop-helix factors) (6). ZEB factors
contain 2 widely separated clusters of zinc fingers that bind to
paired CAGGTA/G E-box–like promoter elements. They induce
EMT by downregulating the expression of epithelial genes, includ-
ing E-cadherin (7, 8). During development, ZEB expression is
upregulated in cells that undergo EMT and migrate by extracellular
signals such as TGF-β and Notch ligands (9, 10). Beyond its physi-
ological roles, ZEB1 is overexpressed in many human cancers (e.g.,
prostate, colon, breast, and pancreatic) and has been implicated
in metastasis and cellular events thought to precede it, including
reduced expression of basement membrane components and induc-
tion of EMT (9). Increased ZEB1 levels correlate with poor progno-
sis in a variety of epithelial tumor types (11). In tumor cells, ZEB1
represses the expression of certain microRNAs (miRs) — includ-
ing miR-183, miR-203, and miR-200 family members (miR-200a,
miR-200b, miR-200c, miR-141, and miR-429) — that function not
only as strong inducers of epithelial differentiation, but also as
inhibitors of stem cell properties through repression of the stem
cell factors SOX2, BMI1, and KLF4 (9). Reciprocally, miR-200
family members directly target the ZEB1 3′–untranslated region
(3′-UTR); hence, ZEB1 and miR-200 are interconnected through a
double-negative feedback loop (12–15). The relevance of these find-
ings to metastasis is supported by findings in a mouse model of
human lung adenocarcinoma driven by expression of Trp53R172HΔG
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. doi:10.1172/JCI63608.
2 The Journal of Clinical Investigation http://www.jci.org
and KrasG12D alleles (KP mice) that recapitulates features of poor-
prognosis human lung adenocarcinomas, including overlapping
oncogenic driver mutations, distribution of metastases, and gene
expression signatures (16, 17). Metastatic tumor cell lines derived
from these mice (KP cells) have high basal ZEB1 expression, form
polarized epithelial spheres in 3D cultures, and undergo ZEB1-
dependent EMT in response to specific extracellular cues that can
be reversed by ectopic expression of the miR-200b/a/429 cluster,
whereas nonmetastatic KP cells have low basal ZEB1 expression
and do not form spheres or undergo EMT (17). Thus, ZEB1 plays
a key role in determining the metastatic fate of epithelial cancers.
Given that each miR downregulated by ZEB1 has the capacity to
target multiple genes with diverse functions, we here posited that
the scope of prometastatic biological processes controlled by ZEB1
extends beyond EMT and stem-ness, testing this hypothesis in KP
cells and human lung and breast carcinoma cells. Our findings
showed that ZEB1 controlled an unexpectedly large miR network
implicated in diverse cellular functions, activated the RHO family
of GTPases, and enhanced the formation of promigratory cyto-
skeletal structures by downregulating miR-34a. We elucidated a
gene expression signature regulated by miR-34a that was enriched
in cytoskeletal functions and prognostic in human lung adeno-
carcinomas and revealed several miR-34a target genes, including
Arhgap1, a RHO GTPase-activating protein (RHOGAP). Together
with the known therapeutic effects of miR-34a on prostate tumors
(18), these findings provide a strong rationale for developing
miR-34a as a therapeutic agent to inhibit tumor growth and
metastasis in lung cancer patients.
ZEB1 regulates the expression of diverse miRs. Microarray-based inter-
rogation of global miR expression was carried out on a nonmeta-
static KP cell line (393P) that undergoes EMT and gains invasive
and metastatic capabilities after ectopic ZEB1 expression (referred
to herein as 393P_ZEB1 cells; GEO accession no. GSE38386). Zeb1
levels in 393P_ZEB1 cells were 4-fold higher than in 393P_vector
cells and were similar to endogenous ZEB1 levels in human lung
cancer cell lines (Supplemental Figure 1; supplemental material
available online with this article; doi:10.1172/JCI63608DS1). Using
393P_vector as the reference, we found that 46 miRs were differen-
tially expressed in 393P_ZEB1 cells: 27 downregulated (fold change
less than 0.5, P < 0.01) and 19 upregulated (fold change greater than
2.0, P < 0.01; Table 1 and Figure 1, A and B). Of the differentially
expressed miRs, 19 were clustered within 7 genomic loci that are
transcribed and processed together (Table 1). Quantitative RT-PCR
(Q-PCR) analysis confirmed up- or downregulation of 14 of 16 miRs
sampled, including 6 known direct transcriptional targets of ZEB1
(miR-200a, miR-200b, miR-200c, miR-141, miR-429, and miR-203)
and 8 other miRs not known to be regulated by ZEB1 (Figure 1C).
Several of the miRs regulated by ZEB1 function as oncogenes (e.g.,
miR-181b, miR-181d, and miR-10a) or tumor suppressors (e.g., miR-34a,
miR-210, miR-326, miR-193a, miR-370, miR-206, miR-126, and
miR-203), or lack reported roles in cancer development (e.g., miR-331,
miR-605, miR-470, miR-581, and miR-351) (19–27). miR-34a was
of particular interest as a candidate downstream mediator of ZEB1,
given its tumor-suppressing activity in various models (18, 27) and
prominent downregulation by ZEB1 (0.0015-fold; Table 1). In a panel
of KP cell lines that have different basal Zeb1 levels, miR-34a levels cor-
related negatively with Zeb1 (R = –0.78; P = 1.9 × 10–9, 1-tailed Spear-
man rank correlation test) and positively with miR-200c (R = 0.78;
miRs regulated by ZEB1
Fold change P
1.68 × 10–4
1.97 × 10–6
8.82 × 10–5
2.96 × 10–5
2.03 × 10–4
3.03 × 10–4
1.36 × 10–4
3.06 × 10–5
1.35 × 10–4
5.06 × 10–3
8.66 × 10–4
1.34 × 10–3
1.69 × 10–3
3.11 × 10–5
2.36 × 10–4
6.26 × 10–3
9.35 × 10–3
6.82 × 10–3
3.24 × 10–3
6.83 × 10–3
8.74 × 10–3
8.49 × 10–3
8.82 × 10–4
8.26 × 10–3
1.68 × 10–3
2.00 × 10–3
1.66 × 10–3
3.53 × 10–5
2.45 × 10–3
6.87 × 10–3
8.59 × 10–3
5.64 × 10–6
4.22 × 10–3
5.87 × 10–3
1.85 × 10–3
8.78 × 10–3
1.63 × 10–4
6.81 × 10–3
7.13 × 10–4
1.86 × 10–4
8.03 × 10–3
2.86 × 10–3
2.86 × 10–3
9.37 × 10–3
1.01 × 10–3
7.32 × 10–3
393P_ZEB1 and 393P_vector cells subjected to global miR profiling
(Figure 1, A and B) to identify upregulated (>2-fold) or downregulated
(<0.5-fold) miRs. A P value (2-tailed Student’s t test) less than 0.01
was considered significant. AGenomic cluster with chromosomal loci
Chr6:30,115,918–30,119,737. BGenomic cluster with chromosomal loci
Chr6:124,667,932–124,668,408. CGenomic cluster with chromosomal
loci Chr4:155,428,014–155,429,859. DGenomic cluster with chromosom-
al loci Chr2:10,397,969–10,437,548. EGenomic cluster with chromosomal
loci ChrX:50,402,580–50,407,231. FGenomic cluster with chromosomal
loci Chr1:139,863,032–139,863,295. GGenomic cluster with chromo-
somal loci Chr8:86,702,615–86,702,860.
The Journal of Clinical Investigation http://www.jci.org
P = 2.9 × 10–9; Figure 2, A and B); similar findings were observed in a
panel of 39 human lung cancer cell lines (Figure 2C).
Although located on a separate chromosome from miR-34a,
expression of the miR-34b/c cluster is frequently coregulated with
miR-34a (27, 28). However, TaqMan PCR assays confirmed the evi-
dence from our microarray studies (Figure 1) that ZEB1 downregu-
lated the expression of miR-34a, but not that of miR-34b or miR-34c
(Supplemental Figure 2). Furthermore, miR-34a levels did not corre-
late with miR-34b or miR-34c in the panel of human lung cancer cell
lines (Figure 2C). Collectively, these findings suggest that miR-34a
expression is regulated through mechanisms different from those
controlling the miR-34b/c cluster in the human and murine lung
adenocarcinoma cells examined in this study.
ZEB1 downregulates miR-34a through ΔNp63. Although miR-34a is a
known transcriptional target of p53 (27, 29), KP cells have the same
germline Trp53 mutation, and the activity of p53 reporter constructs
ZEB1 regulates the expression of numerous
miRs. (A) Volcano plot depiction of findings from
microarray analysis showing the miRs differentially
expressed in 393P_vector and 393P_ZEB1 cells.
The –log10 of P values (y axis) is plotted against
the log2 of fold change between 2 groups (x axis).
The size of the circle for each probe is proportional
to the miR detection rate for the entire experiment.
Each symbol is color coded according to average
expression of the probe across the 2 groups (scale
at right). Dotted lines delineate the cutoffs for miRs
significantly downregulated (left) or upregulated
(right) in 393P_ZEB1 cells. (B) Heat map depic-
tion of miRs differentially expressed in 393P_ZEB1
cells (ZEB1), using 393P_vector cells (vec) as ref-
erence. (C) Taqman microRNA assays (Q-PCR)
to confirm miRs differentially expressed by
microarray. Data are mean ± SD (n = 3 samples).
P values are indicated (2-tailed Student’s t test).
4 The Journal of Clinical Investigation http://www.jci.org
was not different between 2 KP cell lines with high and low miR-34a
levels (Figure 3A), which indicates that differential p53 activity does
not contribute to relative miR-34a levels in KP cells. As an alternative
intermediary, we turned our attention to ΔN isoforms of p63 (ΔNp63),
which lack the transactivation domain and are transcriptional
targets of ZEB1 in 393P cells (Supplemental Figure 3 and ref. 30).
ΔNp63 acts in a dominant-negative fashion against p53, TAp63, and
TAp73, but has been shown to function as a transcriptional activa-
tor of specific genes such as T and T2 (31). Murine embryonic fibro-
blasts (MEFs) deficient in all p63 isoforms had reduced miR-34a
levels (Figure 3B), which suggests that p63 positively regulates miR-34a
expression. On the basis of these findings in MEFs, we sought
to determine whether ΔNp63 serves as an intermediate in ZEB1-
induced miR-34a repression in KP cells. Basal levels of ΔNp63 tran-
scriptional target genes T and T2 were higher in low-ZEB1 cell line
393P than they were in high-ZEB1 cell line 344SQ, and introduc-
tion of ZEB1 into 393P cells decreased T and T2 levels (Figure 3C),
which demonstrated that ZEB1 repressed ΔNp63 transcriptional
activity. Ectopic ZEB1 expression downregulated ΔNp63 mRNA
levels in 393P cells (Figure 3D), and siRNA-mediated knockdown of
ZEB1 upregulated ΔNp63 mRNA levels in 344SQ cells (Figure 3E).
The activity of a ΔNp63 promoter fragment (–1,128 to +109) was
repressed by exogenous ZEB1, but not SNAI1 or TWIST1 (Figure 3F).
Promoter deletion studies demonstrated that ZEB1 repressed the
activity of a minimal ΔNp63 promoter fragment (–482 to +109) con-
taining 2 putative ZEB1-binding sites (E-boxes; Figure 3G), and site-
directed mutagenesis of the most proximal E-box element abrogated
ZEB1-induced repression of the ΔNp63 promoter (Figure 3H).
We next examined whether ΔNp63 regulates miR-34a expression
and acts directly on the miR-34a gene promoter in KP cells. Ectopic
ΔNp63β expression in 344SQ cells upregulated miR-34a, but not
miR-34b or miR-34c (Supplemental Figure 4), and increased the
activity of a WT promoter, but not a mutant miR-34a promoter
lacking a p53-binding site (Figure 3I). Binding of endogenous p63
to the miR-34a promoter was examined by performing ChIP assays
using an antibody that recognizes all p63 isoforms. In 344SQ cells,
total p63 binding roughly reflected ΔNp63 binding, because ΔN
isoforms were 67.0-fold more abundantly expressed than were TA
isoforms (Supplemental Figure 5A). Murine keratinocytes were
included as a positive control. The percentage of total p63 that
bound to the p53/p63 site and to a nonspecific site in the miR-34a
promoter was 0.19% and 0.06%, respectively (P = 0.08; Supplemen-
tal Figure 5B). To more specifically examine ΔNp63 binding to that
site, ChIP assays were performed on 393P cells stably transfected
with Myc-tagged ΔNp63. Using an anti-Myc antibody, we found the
percentage of ectopic ΔNp63 that bound to the p53/p63 site and to
a nonspecific site in the miR-34a promoter to be 0.39% and 0.17%,
respectively (P = 0.002; Figure 3J). We concluded that ΔNp63 serves
as an intermediate in ZEB1-induced miR-34a repression in KP cells.
miR-34a abrogates tumor cell invasion and metastasis and induces tran-
scriptional changes that are prognostic in human lung adenocarcinomas.
To examine the biological role of miR-34a repression, miR-34a was
miR-34a levels correlate tightly with the ZEB1/miR-200 axis. (A and B) Expression analysis of a panel of KP cell lines. (A) Log-scale cluster plots
of normalized miR-200c and Zeb1 mRNA levels in 13 KP cell lines. Data are mean ± SD (n = 3 samples). 393P_ZEB1 and 393P_vector cells are
included as controls. (B) Heat map representation of gene expression in A. (C) Expression analysis of a panel of human lung cancer cell lines.
EMT marker expression in these 39 human cell lines was reported previously (17). Correlation (R and P, 1-tailed Spearman’s rank correlation
test) is indicated for each gene relative to miR-34a.
The Journal of Clinical Investigation http://www.jci.org
6 The Journal of Clinical Investigation http://www.jci.org
exogenously expressed under the control of a doxycycline-inducible
promoter in 344SQ cells, which have high ZEB1 and low miR-34a
expression (Figure 4A). Grown in monolayer, 344SQ_miR-34a
cells proliferated at a rate similar to that of 344SQ_vector cells and
demonstrated no biochemical evidence of apoptosis (Figure 4B
and Supplemental Figure 6A). However, when grown in suspen-
sion, 344SQ_miR34a cells did not proliferate and demonstrated
evidence of apoptosis (Supplemental Figure 6, B and C), which
suggests that miR-34a enhanced the susceptibility of 344SQ cells
to anoikis. 344SQ_miR34a cells exhibited reduced migration and
invasion in Boyden chambers (Figure 4, C and D) and generated
flank tumors in syngeneic mice that were smaller and metastasized
to the lung less frequently (Figure 4E). Conversely, transfection of
miR-34a hairpin inhibitor into 393P cells, which have low ZEB1
and high miR-34a expression, induced a 84% decrease in endog-
enous miR-34a levels and 1.6- and 1.5-fold increases in cell migra-
tion and invasion, respectively (Supplemental Figure 7, A–C). In
MDA-MB-231 human breast cancer cells and H1299 human lung
cancer cells, which have high basal ZEB1 expression and undergo
EMT in response to TGF-β (13, 32), overexpression of miR-34a
antagonized migration and invasion (Figure 4F and Supplemen-
tal Figure 8, A–C), which was not a consequence of apoptosis or
reduced proliferation (Supplemental Figure 8, D and E). There was
no evidence of EMT reversal on the basis of expression of epithelial
and mesenchymal markers in 344SQ cells and MDA-MB-231 cells
(Figure 4, G and H). Thus, miR-34a downregulation was required
for ZEB1-induced metastatic properties and apparently mediated
these actions through EMT-independent mechanisms.
To gain insight into the biologic processes induced by miR-34a
repression, microarray-based interrogation was carried out on RNA
from tumor samples (344SQ_miR-34a and 344SQ_vector), which
revealed a total of 805 genes that were differentially expressed (fold
change greater than 1.5, P < 0.01; Supplemental Figure 9; GEO
accession no. GSE38341). Q-PCR analysis of the same RNA samples
confirmed differential expression of 22 of 24 genes sampled (Sup-
plemental Figure 10). The 512 downregulated genes were enriched
in, among other Gene Ontology terms mitosis (P = 3.80 × 10–24, Fisher
exact test), cell cycle (P = 3.40 × 10–23), mRNA processing (P = 4.72 × 10–10),
and cytoskeleton (P = 3.04 × 10–7), whereas the 293 upregulated genes
were enriched in protein dimerization activity (P = 7.55 × 10–4) and
glutathione transferase activity (P = 1.12 × 10–3). Human lung adenocar-
cinomas for which both gene expression and clinical outcome data
are publicly available (33–35) were scored based on the presence or
absence of this 805-gene signature (high or low t score, respectively),
as described previously (36); the absence of this expression signature
in primary lung tumors correlated with poor prognosis in 3 inde-
pendent cohorts of lung cancer patients (Figure 4I). We conclude
that the genes regulated by miR-34a are functionally diverse and
have prognostic value in lung adenocarcinoma patients.
miR-34a inhibits promigratory cytoskeletal processes and attenuates
RHO GTPase activity. Given the prominent effect of miR-34a on cell
migration and invasion in Boyden chambers and the enrichment
of the miR-34a transcriptome in cytoskeletal functions, we next
carried out studies in 3D cultures to visualize tumor cell polariza-
tion and invasion. In Matrigel, 344SQ cells form hollow, polarized
spheres with apical localization of ZO-1 and basolateral localiza-
tion of α6 integrin; these structures lose apical-basal polarity,
become hyperproliferative, and form invasive membrane protru-
sions after TGF-β treatment (17). 344SQ_miR-34a and 344SQ_
vector cells formed hollow polarized spheres in the absence of
TGF-β (Figure 5, A and C) and became aberrantly polarized and
filled in their central cores after TGF-β treatment (Figure 5,
B and D). However, TGF-β–treated 344SQ_vector structures
formed invasive projections (Figure 5B), whereas 344SQ_miR-34a
structures remained spherical in morphology (Figure 5D), despite
upregulation of mesenchymal markers in both (Figure 5E). Thus,
miR-34a abrogated TGF-β–induced formation of invasive cellular
protrusions in spite of biochemical evidence of EMT.
We next carried out scratch injury assays to examine whether miR-34a
regulates the formation of focal adhesions and promigratory mem-
branous protrusions. At the leading edge, 344SQ_vector cells formed
lamellipodia with prominent filopodial extensions, whereas 344SQ_
miR-34a cells formed no filopodia (Figure 6A), generated more focal
adhesions per surface area (mean, 232 versus 158; Figure 6B), and
exhibited increased cell cross-sectional area (1,330 μm2 versus 410 μm2;
Figure 6C). Consistent with these findings, 344SQ_miR-34a cells
anchored more avidly to plastic (Figure 6D) and formed fewer col-
onies in soft agar (Figure 6E). In the larger panel of KP cell lines,
anchorage avidity correlated positively with endogenous miR-34a
levels (R = 0.6770; P = 0.0055, 1-tailed Pearson correlation test;
Figure 6F). In MDA-MB-231 cells, exogenous miR-34a decreased cor-
tical F-actin–containing filopodial extensions and increased cell size
without reversing mesenchymal features on the basis of ZO-1 local-
ization, which was predominantly cytoplasmic (Figure 6G). In con-
trast, exogenous miR-200b in MDA-MB-231 cells decreased filopodial
extensions and shifted ZO-1 to a membranous distribution typical of
a mesenchymal-to-epithelial transition (Figure 6G). Given the central
role of RHO GTPases in actin cytoskeletal remodeling (37, 38), we quan-
tified activated (GTP-bound) RHO family members after treatment
with promigratory cytokines, which revealed that ectopic miR-34a
expression in 344SQ cells attenuated EGF- or TGF-β–induced
increases in GTP-bound CDC42, RAC1, and RHO (Figure 6, H and I).
Thus, miR-34a increased cellular adhesions, attenuated cytokine-
induced RHO GTPase activation, and inhibited formation of mem-
brane protrusions and migration in response to external stimuli.
ZEB1 downregulates miR-34a levels through ΔNp63. (A) Luciferase
reporter assays on KP cells cotransfected with p53 or empty expres-
sion vectors and reporters containing promoters from Mdm2, 14-3-3σ,
or consensus p53-binding sites (Generic). Results are expressed rela-
tive to empty vector transfectants, set at 1.0. (B) Q-PCR analysis of
miR-34a in WT or Trp63-null (p63–/–) MEFs. (C) Q-PCR values rela-
tive to 344SQ cells (left) and 393P_ZEB1 cells (right), set at 1.0. (D
and E) Q-PCR values relative to 393P_vector cells (D) and control
siRNA-transfected 344SQ cells (E), set at 1.0. (F) Location of putative
E-boxes in the ΔNp63 promoter are shown above ΔNp63 promoter
luciferase assays on cotransfected 393P cells. (G) ΔNp63 promoter
reporter constructs are shown above luciferase activities in cotransfect-
ed 393P cells, expressed relative to basal reporter activity (set at 1.0).
(H) ΔNp63 promoter constructs are WT or lack 1 (Mut-a and Mut-b)
or 2 (Mut-a/b) proximal E-boxes. Luciferase activities in cotransfected
393P cells are also shown, expressed relative to basal reporter activity
(set at 1.0). (I) Luciferase reporter constructs that do (Mut) or do not
(WT) contain a site-directed mutation in a p53-binding site. Lucifer-
ase activities in cotransfected 393P cells are also shown, expressed
relative to basal reporter (GL3) activity (set at 1.0). (J) DNA amount at
specific (p53/p63) and nonspecific (NS) sites of the miR-34a promoter
precipitated from transiently transfected 393P cells by anti-Myc anti-
body, corrected for nonspecific binding activity by IgG precipitation and
expressed as percent p63 bound. P values were determined by 2-tailed
Student’s t test. Data are mean ± SD (n = 3).
The Journal of Clinical Investigation http://www.jci.org
miR-34a regulates multiple biological properties of tumor cells. (A–D) 344SQ_miR-34a cells and 344SQ_vector cells were cultured in the pres-
ence or absence of doxycycline (Dox). (A) Q-PCR analysis of miR-34a levels. (B) Cell numbers in monolayer. Migrating (C) and invading (D)
cells in Boyden chambers were photographed and counted. Scale bars: 100 μm. (E) Primary tumor weight and total lung metastases from flank
tumors in syngeneic mice (mean ± SD, n = 5). P values were determined by 2-tailed Student’s t test. (F) MDA-MB-231 cells were transiently
transfected with a random sequence miR precursor molecule control or with pre–miR-34a precursor. Shown are Q-PCR analysis of miR-34a
levels, expressed relative to control transfectants (set at 1.0), and migration and invasion assays in Boyden chambers. (G and H) Q-PCR analysis
of epithelial (Cdh1 and Scrib) and mesenchymal (Cdh2 and Vim) markers and their transcriptional regulators (Zeb1, Zeb2, Snai1, Snai2, and
Twist1) in 344SQ_vector and 344SQ_miR-34a cells (G) and in MDA-MB-231 cells transiently transfected with pre-miR control or pre–miR-34a
precursor (H). Results are expressed relative to control transfectants (set at 1.0). Data are mean ± SD (n = 3). *P < 0.01. (I) Kaplan-Meier analysis
of 3 independent cohorts of lung cancer patients (33–35), comparing the differences in risk between tumors with high (>0) or low (<0) scores
(36), reflecting the presence or absence, respectively, of overlap with the murine miR-34a signature. P values from log-rank (differences between
arms) and univariate Cox (gene signature score as a continuous variable) tests are shown.
8 The Journal of Clinical Investigation http://www.jci.org
Arhgap1 is a miR-34a target gene required for the regulation of RHO
GTPase activity and tumor cell invasion. We next examined whether
miR-34a inhibits RHO GTPases at the level of GTP loading or
upstream at the level of focal adhesion kinase (FAK), which is auto-
phosphorylated at Tyr397 and initiates RHO GTPase activation
after recruitment to ligand-bound EGFR (39). Arguing against
the latter possibility, EGF-induced FAK-Tyr397 phosphorylation
was unchanged by exogenous miR-34a, as determined by Western
blot analysis (data not shown). At the level of GTP loading, RHO
GTPases cycle from inactive (GDP-bound) to active (GTP-bound)
states by binding to guanine nucleotide exchange factors (GEFs),
and in the reverse direction by binding to GAPs (37). To examine
whether miR-34a directly targets GEFs or GAPs, we used a predic-
tion algorithm (TargetScan; http://www.targetscan.org) to scan
the genome for putative miR-34a binding sites and discovered sites
in the 3′-UTR of a RHOGAP (Arhgap1) and a CDC42 downstream
effector (Cdc42se1) (Figure 7A). Reporter assays were carried out to
determine whether miR-34a binds directly to these 3′-UTRs as well
as those of other predicted miR-34a target genes identified in our
analysis that are not directly involved in RHO GTPase regulation,
but are potentially important in tumorigenesis (Figure 7B). Rela-
tive to its effect on a negative con-
trol 3′-UTR (Flt1), cotransfection of
miR-34a precursors repressed
3′-UTRs of Arhgap1 by 59%, Satb2 (a
regulator of chromatin structure; ref.
40) by 47%, Lef1 (a regulator of Wnt
signaling; ref. 41) by 44%, and Hnf4a
(a known miR-34a target; ref. 42) by
37%, but did not affect other report-
ers (Figure 7B). Site-directed muta-
genesis of 2 predicted binding sites
in the Arhgap1 3′-UTR with differing
PCt values (0.71 and <0.1) revealed
that miR-34a suppressed Arhgap1
3′-UTR reporter activity through the
conserved binding site, but not the
nonconserved site (Figure 7C).
In the KP cell line panel, ARHGAP1
and miR-34a levels correlated nega-
tively (R = –0.7692; P = 0.0021, 1-tailed
Spearman rank correlation test; Fig-
ure 7D), and exogenous miR-34a
decreased ARHGAP1 levels in 344SQ
cells (Figure 7E). Conversely, trans-
fection of miR-34a hairpin inhibitor
into 393P cells induced a 2.0-fold
increase in ARHGAP1 expression
(Supplemental Figure 11A). In
MDA-MB-231 and H1299 cells,
ectopic miR-34a downregulated
ARHGAP1 (Figure 7F and Supple-
mental Figure 11B). To determine
whether ARHGAP1 downregulation
recapitulates the effects of miR-34a,
ARHGAP1 was depleted by introduc-
tion of shRNAs (Figure 8A), which
increased basal levels of GTP-bound
CDC42 and RAC1 (Figure 8B),
attenuated EGF-induced levels of
GTP-bound CDC42 and RAC1 (Figure 8B), reduced invasive pro-
jections on tumor spheres after TGF-β treatment (Figure 8C), and
increased cell adherence (Figure 8D).
The anti-invasive effect of Arhgap1 shRNA was paradoxical, given
that RHO is required for proinvasive actin cytoskeletal remodeling.
We performed experiments to confirm this finding and to deter-
mine whether ARHGAP1 is required for miR-34a–induced pheno-
typic features. ARHGAP1 expression was reconstituted in 344SQ_
miR-34a cells through stable transfection of an Arhgap1 cDNA
expression vector (Figure 8E), and the 344SQ_miR-34a/ARHGAP1
double transfectants were compared with 344SQ_miR-34a/vector
cells from the standpoint of basal and cytokine-induced RHO
activity and invasive activity. ARHGAP1 reconstitution repressed
basal GTP-bound CDC42 and rescued TGF-β–induced sphere
invasion in Matrigel (Figure 8, F and G), which suggests that
miR-34a represses TGF-β–induced tumor cell invasion by down-
regulating ARHGAP1. However, ARHGAP1 reconstitution did
not restore RHO activation in response to EGF or TGF-β treat-
ment (Figure 8F) or rescue tumor growth or metastasis (data not
shown), which suggests that these phenotypic effects of miR-34a
are mediated through other target genes.
miR-34a blocks invasion, but does not reverse EMT. (A–D) miR-34a repressed TGF-β–induced inva-
sion in 3D Matrigel cultures. 344SQ_vector cells formed polarized epithelial spheres (A) that became
hyperproliferative and invasive in the presence of TGF-β (B). 344SQ_miR-34a cells formed polarized
epithelial spheres (C) that did not become invasive in the presence of TGF-β (D). Shown are light (left)
and fluorescent (right) microscopic images of structures formed after 10 days in Matrigel containing
doxycycline in the presence or absence of TGF-β (10 ng/ml). Blue, Topro-3; red, anti–α6 integrin;
green, anti–ZO-1. Scale bars: 200 μm (light); 50 μm (fluorescent). (E) miR-34a did not abrogate TGF-β–
induced EMT. Q-PCR analysis of epithelial markers (Cdh1, Scrib, and Crb3) and mesenchymal mark-
ers (Cdh2 and Vim) and their transcriptional regulators (Zeb1, Zeb2, Snai1, and Snai2) in 344SQ_vec-
tor and 344SQ_miR-34a cells after 10 days in Matrigel cultures containing doxycycline in the presence
or absence of TGF-β. Results are expressed relative to empty vector transfectants treated without
TGF-β (set at 1.0). Data are mean ± SD (n = 3).
The Journal of Clinical Investigation http://www.jci.org
10 The Journal of Clinical Investigation http://www.jci.org
miR network that targets RHO GTPases and their associated GEFs
and GAPs (53). Examples include RHOA (miR-31, miR-133, and
miR-155), RHOC (miR-138 and miR-10b), CDC42 (miR-29), TIAM1
(miR-10b), and ARHGDIA (miR-151) (53). Here we showed that
miR-34a inhibited cytokine-induced RHO family GTPase activa-
tion and discovered that a RHOGAP, Arhgap1, was a miR-34a tar-
get gene. ARHGAP1 reconstitution in miR-34a–overexpressing
cells did not rescue RHO activation in response to EGF or TGF-β
treatment, which was expected, given that RHOGAPs inhibit RHO
GTPase activity. However, TGF-β–induced invasion was abrogated
in metastasis-prone tumor cells by ARHGAP1 depletion and was
rescued in miR-34a–overexpressing cells by ARHGAP1 reconstitu-
tion. The proinvasive effect of ARHGAP1 was paradoxical, given
that RHO GTPase activity stimulates the formation of actin cyto-
skeletal structures that drive cell migration. Although the mecha-
nism is unclear, ARHGAP1 binds to a number of proteins other
than RHO family GTPases — including BNIP2, SRC, UBC, and
PIK3R1 (BioGRID; http://thebiogrid.org) — that regulate diverse
biological processes and may have contributed to the proinvasive
effect of ARHGAP1 through RHO GTPase–independent mecha-
nisms. Collectively, these findings suggest that ARHGAP1 mediates
some, but not all, of the biological effects of miR-34a (Figure 8H).
We discovered that ZEB1 regulated a larger number of miRs
than had previously been reported (12, 13, 15). This multiplicity
was due in part to 19 miRs clustered within 7 genomic loci that
are transcribed and processed together. ZEB1 downregulated cer-
tain miRs and upregulated others, which could be related either
to the capacity of ZEB1 to function as a transcriptional repressor
or activator (54–56) or to indirect regulation of miRs by ZEB1.
In support of the latter possibility, we found that ZEB1 indirect-
ly repressed miR-34a through ΔNp63. The reported biological
functions of the 46 miRs were diverse, encompassing hypoxic
response (miR-210), cell differentiation (miR-326), proliferation
(e.g., miR-224, miR-206, miR-542-3p, and miR-126), apoptosis
(miR-96, miR-193a, and miR-181a), and migration (miR-206,
miR-503, and miR-181b), among other functions (Supplemental
Table 1), which indicates that ZEB1 might control a number of
biological processes by regulating the expression of these miRs.
The p63 transcription factor family plays a central role in the
regulation of embryonic development, normal adult tissue homeo-
stasis, and malignancy (57). The tumor-suppressive properties of
TAp63 are exerted through the upregulation of a wide variety of
miRs, including let-7, miR-15/16a, miR-145, miR-129, miR-26,
miR-30, and miR-146a (57). Senescence in keratinocytes is activated
through ΔNp63-induced downregulation of miR-138, miR-181a,
miR-181b, and miR-130b (58). TAp63 is also a transcriptional acti-
vator of Dicer, an endoribonuclease required for miR biogenesis
(29). The findings presented here build on this growing body of
evidence that miRs are central mediators of the diverse biological
actions of p63 by showing that miR-34a was upregulated by ΔNp63
and was a potent tumor suppressor in a Kras/Trp53-driven lung
adenocarcinoma model. Furthermore, our finding that ΔNp63
served as a downstream mediator of ZEB1 completes a feedback
circuit initiated by p63, which transcriptionally activates the
miR-200b/a/429 cluster (59) and, in turn, directly targets ZEB1 (9,
13, 14), thereby relieving the ZEB1-induced repression of ΔNp63
shown here. There are numerous other p63/miR circuits, includ-
ing one involving miR-193-5p, which targets p63 and is directly
repressed by p63 (60). Thus, miR homeostasis is tightly regulated
through multiple mechanisms involving p63 and ZEB1.
Studies using experimental tumor models have established a
strong link between high levels of EMT activators and loss of cell
polarity, reduced expression of basement membrane components,
and increased propensity for metastasis (43–47). The discovery
that EMT activators endow epithelial tumor cells with pluripo-
tency led to the current belief that metastatic propensity is directly
related to plasticity in response to extracellular cues (12, 48–50).
Here, positing that the scope of prometastatic biological processes
controlled by ZEB1 extends beyond EMT and stem-ness, we dis-
covered that ZEB1 drove promigratory cytoskeletal processes and
metastasis by downregulating the expression of miR-34a. Exoge-
nous miR-34a decreased tumor cell invasion and metastasis, inhib-
ited the formation of promigratory cytoskeletal structures, sup-
pressed activation of the RHO GTPase family, and regulated a gene
expression signature that was enriched in cytoskeletal functions
and prognostic in human lung adenocarcinomas. Biological repro-
gramming of this magnitude supports a central role for miR-34a
in metastasis regulation by ZEB1.
RHO family members play key parts in the regulation of actin
cytoskeletal remodeling and tumorigenesis. RAC1 is required for
the development of primary lung adenocarcinomas in mice that
express mutant KRAS (51). The activities of RHO, RAC1, and
CDC42 are coordinated to regulate membrane protrusions and
cell-matrix adhesions at the leading edge of migrating cells to con-
trol forward movement (52). Effectors of RHO GTPases include
RHO-associated protein kinase, focal adhesions, and membrane
protrusions, which together mediate cell adhesion to extracellular
matrix, link matrix attachments to intracellular signaling path-
ways, and drive actomyosin contractility and cell locomotion (52).
Beyond these roles, a large body of evidence implicates RAC1 in
the assembly, disassembly, and maintenance of adherens junctions
and tight junctions, which play a central role in the regulation of
apical-basal polarity (52). Tightly regulating these processes is a
miR-34a regulates actin cytoskeletal remodeling and RHO family
GTPase activity. (A–C) Cells imaged at leading edge of scratch-wound-
ed confluent cultures. (A) Filopodia (arrows) formed in 344SQ_vector,
but not 344SQ_miR-34a, cells. Brackets denote lamellipodia. (B) Focal
adhesions in anti-vinculin–stained cultures, counted per defined sur-
face area of confluent cells (circles) using ImageJ. Data are mean ± SD
(n = 10). Red, phalloidin; green, vinculin. (C) Cells were outlined (white
lines), and their surface areas were measured using ImageJ. Data
are mean ± SD (n = 20 [344SQ_vector]; 12 [344SQ_miR-34a]). (D)
Attached cells were quantified 1, 2, or 3 hours after seeding by opti-
cal densitometry (595 nm) of cells stained with crystal violet. Data are
mean ± SD (n = 3). *P < 0.01. (E) Cells seeded in soft agar were stained
with nitrotetrazolium blue 3 weeks after seeding, and colonies larger
than 100 μm in diameter were counted. Data are mean ± SD (n = 3).
(F) Attached cells were quantified by optical densitometry 3 hours after
seeding and expressed relative to miR-34a levels from Figure 2A. Cor-
relation (R and P, 1-tailed Pearson’s correlation test) is indicated. (G)
MDA-MB-231 cells transiently transfected with control, miR-34a, or
miR-200b precursors and imaged under fluorescence (blue, DAPI; red,
phalloidin; green, anti–ZO-1). As a comparison, miR-200b–transfected
cells demonstrated mesenchymal-to-epithelial transition. (H and I)
Western blot analysis of GTP-bound (CDC42-GTP) and total (CDC42)
RHO family GTPases. Phospho-ERK1/2 (H) and phospho-SMAD3 (I)
were included as positive controls for EGF- and TGF-β–induced signal-
ing, respectively. Scale bars: 50 μm (A, B, and G); 100 μm (C); 500 μm
(E). See complete unedited blots in the supplemental material.
The Journal of Clinical Investigation http://www.jci.org
miRs locally or systemically to the tumor tissue where they regulate
their target genes (62, 63). Physical and chemical moieties of the
particles that facilitate the targeted distribution and the controlled
and sustained release of miRs are under clinical investigation (64).
External moieties, such as aptamers and ligands that enhance miR
uptake by cancer cells, are being developed to direct the particles to a
particular tissue (65, 66). Moreover, efforts are underway to initiate
clinical trials that deliver miRs into patients with advanced cancer.
Antibodies and plasmid constructs. Antibodies against ERK, phospho-ERK,
SMAD3, phospho-SMAD3, PARP, cleaved caspase-3 (Cell Signaling Tech-
nology), ACTIN (Sigma-Aldrich), p63 (Abcam), and ARHGAP1 were pur-
chased (Santa Cruz Biotechnologies). Doxycycline (Sigma-Aldrich), EGF
(Invitrogen), and TGF-β (Calbiochem) were purchased. Human SNAI1
cDNA (catalog no. 16218), murine Twist1 cDNA (catalog no. 1783; Add-
gene), murine Arhgap1 cDNA, and murine Arhgap1 shRNA (Origene) were
purchased. To construct the miR-34a overexpression vector, a tet opera-
tor–H1 promoter fusion (tH1) was cloned into the XmnI and BamHI sites
of pENTR2B (Invitrogen). A 487-nt fragment containing miR-34a (∼200 nt
either side of the mature miR) was amplified from human cDNA by PCR
and directionally cloned into the BamHI and EcoRI sites of pENTR2B-tH1.
The pLV711G lentiviral expression construct contains a Gateway cassette
(Invitrogen) and the regulatory protein (T-REx) under the control of the
The evidence presented here that miR-34a is a potent repressor
of tumor growth and metastasis in a mouse model of human lung
cancer bolsters evidence from other mouse models that miR-34a is
a promising therapeutic agent. Delivery of miR-34a oligomers sys-
temically by tail vein inhibits tumor growth in mice bearing lung
adenocarcinomas, suppresses metastasis to the lung and other
organs, and prolongs the survival of mice bearing orthotopic human
prostate carcinomas (61). The mechanisms by which miR-34a exerts
its therapeutic effects are tumor cell type specific. For example, in
the lung adenocarcinoma metastasis model shown here, miR-34a
downregulation enhanced promigratory cytoskeletal processes,
but was not required for stem cell features, based on formation of
polarized epithelial spheres, whereas it targets the stem cell marker
CD44 in prostate cancer cells and represses stem-ness in prostate,
glioblastoma, pancreatic, and gastric cancer cells (18, 49). The dis-
tinct mechanisms by which miRs exert tumor suppressor functions
in a given tumor type might be leveraged to create combinatorial
treatment approaches. In metastatic KP cells, the miR-200 family
members and miR-34a are all sharply downregulated, and ectopic
expression of the miR-200b/a/429 cluster locks KP cells into an epi-
thelial state and abrogates metastasis (17). Thus, combined delivery
of miR-34a and miR-200 family members might be complemen-
tary in these cells. Safe and efficient approaches using lipid-based
nanoparticles (neutral or charged) have been developed that deliver
Arhgap1 is a miR-34a target gene. (A) 3′-UTRs analyzed. Shown are size (right) and location of predicted miR-34a binding sites. Poorly conserved
sites were defined on the basis of criteria established by TargetScan. mut1 and mut2, mutations generated in 2 putative miR-34a binding sites in the
Arhgap1 3′-UTR. Flt1 3′-UTR is a negative control. (B and C) miR-34a–induced repression of 3′-UTR reporters. Luciferase assays on 344SQ cells
transiently cotransfected with the indicated 3′-UTR reporters and control or miR-34a precursors. Data are mean ± SD (n = 3). *P < 0.001; #P < 0.01.
(D) Correlation of miR-34a levels in KP cell lines (from Figure 2A), with densitometric analysis of ARHGAP1 protein expression by Western analysis.
ACTIN was used as a loading control. Correlation (R and P, 1-tailed Spearman rank correlation test) is indicated. (E) miR-34a downregulated ARHGAP1.
Western blot analysis of ARHGAP1 in 344SQ_vector and 344SQ_miR-34a cells. Densitometric analysis (numbers below ARHGAP1 gel) is shown
relative to 344SQ_vector cells (set at 1.0). (F) Q-PCR analysis of ARHGAP1 mRNA levels in MDA-MB-231 cells 72 hours after transient transfection
with pre-miR negative control or pre–miR-34a precursors. Data are mean ± SD (n = 3). P values were determined by 2-tailed Student’s t test. See
complete unedited blots in the supplemental material.
12 The Journal of Clinical Investigation http://www.jci.org
later. For migration and invasion assays, 1 × 105 cells were cultured in the
upper wells of Transwell and Matrigel chambers, respectively (BD Biosciences),
and allowed to migrate toward 10% FBS in the bottom wells. After 16 hours
of incubation, migrating or invading cells were stained with 0.1% crystal
violet, photographed, and counted. Cellular proliferation was measured in
anchorage-dependent and -independent conditions by counting cells seeded
onto high- or low-adherence plates, respectively (Greiner Bio-One), using the
Countess automated cell counter (Invitrogen). For immunocytochemistry,
cells were cultured on collagen-coated coverslips and then stained with DAPI
(Sigma-Aldrich), Alexa Fluor 568–conjugated phalloidin (Invitrogen), and
anti-vinculin (Millipore) antibody. Cells were cultured in 3D Matrigel cultures
(BD Biosciences) and stained with immunofluorescently tagged antibodies, as
described previously (17). A Zeiss LSM 510 confocal microscope was used to
EF-1a promoter (67). Gateway technology (Invitrogen) was used for the
transfer of miR-34a into the pLV711G vector to create a single lentiviral
vector enabling doxycycline-responsive expression of miR-34a.
Cell culture studies. Murine (307P, 344LN, 344P, 344SQ, 393LN, 393P, 412P,
531LN1, 531LN2, 531LN3, 531P1, 531P2, and 713P) and human lung can-
cer cells (H2009 and H1299) were cultured in RPMI 1640 (Mediatech) with
10% FBS (Sigma-Aldrich) in the presence or absence of 1 μg/ml doxycycline
(Sigma-Aldrich). WT and p63–/– murine embryonic fibroblasts (MEFs) were
maintained in DMEM (Mediatech) with 10% FBS. MDA-MB-231 cells were
cultured in Leibovitz’s L-15 (Mediatech) with 10% FBS. Cells were transfected
using Dharmafect-DUO (Dharmacon). For soft agar assays, 5 × 104 cells (in
0.3% agar) were seeded into 6-well plates layered with 0.8% agar, and colonies
were stained with 0.5 mg/ml nitrotetrazolium blue (Sigma-Aldrich) 21 days
ARHGAP1 mediates specific phenotypic effects of miR-34a. (A) Western blot analysis of ARHGAP1 in 344SQ cells stably transfected with
scrambled shRNA (scr) or 1 of 4 Arhgap1 shRNAs (shA, shB, shC, or shD). ACTIN was used as loading control. (B) Western blot analysis of
GTP-bound (-GTP) and total RHO family GTPases. (C) Spheres formed by scrambled shRNA– and Arhgap1 shB–transfected 344SQ cells after
10 days in 3D Matrigel cultures in the presence or absence of TGF-β (10 ng/ml). Scale bars: 100 μm. (D) Attached cells were quantified by optical
densitometry 3 hours after seeding. Data are mean ± SD (n = 3). P values were determined by 2-tailed Student’s t test. (E) Western blot analysis
of 344SQ_miR-34a cells stably transfected with Flag-tagged ARHGAP1 cDNA or empty vector using antibodies against ARHGAP1 (top), Flag
(middle), or ACTIN as a loading control (bottom). Arrows at top denote locations of endogenous (ARHGAP1) and ectopic (Flag-ARHGAP1)
ARHGAP1. (F) Western blot analysis of GTP-bound and total CDC42 in 344SQ_miR34a/vector and 344SQ_miR34a/ARHGAP1 cells treated
for 10 minutes with or without EGF or TGF-β. (G) 344SQ_miR34a/vector and 344SQ_miR34a/ARHGAP1 cells were cultured for 10 days in
Matrigel in the presence or absence of TGF-β and photographed under phase-contrast microscopy. TGF-β induced loss of lumen formation in
both transfectants, but invasive projections formed only in 344SQ_miR34a/ARHGAP1 cells. Scale bars: 50 μm. (H) Proposed model illustrating
the ARHGAP1-dependent and -independent phenotypic effects of miR-34a. See complete unedited blots in the supplemental material.
The Journal of Clinical Investigation http://www.jci.org
(68). p63-DNA complexes were diluted 10-fold in ChIP dilution buffer and
incubated overnight at 4°C with 2 μg anti–pan-p63 antibody (4A4; Abcam)
or 2 μg IgG. Resulting chromatin was resuspended in 300 μl double-dis-
tilled H2O. The percent DNA bound was calculated as (2ΔCt × 2.5)p63 antibody −
(2ΔCt × 2.5)IgG, where ΔCt is Ctinput chromatin − Ctsample.
Western blotting. Cells were lysed in 50 mM Tris-HCl (pH 7.4), 150 mM
NaCl, 1 mM EDTA, 1% Triton X-100, and protease/phosphatase inhibi-
tors (Sigma-Aldrich). Cell lysates were separated by SDS-PAGE, transferred
onto PVDF membrane, and then incubated with primary antibodies and
HRP-conjugated secondary antibodies. Protein bands were visualized with
Pierce ECL Western Blotting substrate (Thermo). RAC1, CDC42, and RHO
activation assay kits (Upstate) were used for GTPase assays.
Animal husbandry. Syngeneic (129/Sv) mice (n = 5 per group) were injected
subcutaneously in the right flank with 344SQ cells (1 × 106 per mouse) that
had been stably transfected with miR-34a or empty vectors. Mice were treat-
ed with doxycycline (2 mg/ml) in drinking water (2% sucrose), monitored
daily for tumor growth, sacrificed at 6 weeks, and necropsied to isolate pri-
mary tumors and sites of metastasis, which were confirmed histologically
by analysis of hematoxylin and eosin–stained, formalin-fixed tissues.
Affymetrix gene expression profiling. Total RNA was extracted from primary
tumors from mice injected with 344SQ_vector and 344SQ_miR-34a cells
using RiboPure kit (Ambion), and then hybridized to Affymetrix GeneChip
Mouse Genome 430 2.0 array (Asuragen). Data processing and determi-
nation of differentially expressed genes were carried out essentially as
described previously (69). Transcriptomic data sets were deposited in GEO
(accession no. GSE38341).
Statistics. With the exception of mRNA and miR profiling, data were ana-
lyzed using 2-tailed Student’s t test and Spearman rank correlation test. A
P value less than 0.05 was considered significant.
Study approval. All mouse studies were approved by the IACUC at the Uni-
versity of Texas MD Anderson Cancer Center. Mice received standards of
care and were euthanized according to the standards set forth by the IACUC.
This work was supported by R01 CA157450 (to J.M. Kurie). J.M. Kurie
is the Elza and Ina A. Shackelford Endowed Professor in Lung Can-
cer Research. D.L. Gibbons was supported by NCI K08 CA151651,
an International Association for the Study of Lung Cancer Fellow
Grant, and received financial support from Dr. Waun Ki Hong (MD
Anderson Cancer Center). C.J. Creighton was supported by P30
CA125123. D. Chakravarti was funded by a CPRIT training grant
(RP101502). Z.H. Rizvi was supported by HHMI-Medical Research
Fellows Program. A. Pertsemlidis was funded by R01 CA129632 and
P50 CA70907 (the UT Southwestern/MD Anderson Cancer Center
Lung Specialized Program of Research Excellence). E.R. Flores was
funded by R01CA134796 and is a Leukemia and Lymphoma of
America Scholar. G.J. Goodall was supported by NHMRC project
grant 1008327. We thank Suraya Roslan for technical assistance.
Received for publication February 28, 2012, and accepted in
revised form June 14, 2012.
Address correspondence to: Jonathan M. Kurie, MD Anderson Cancer
Center, Box 432, Department of Thoracic/Head and Neck Medical
Oncology, 1515 Holcombe Blvd., Houston, Texas 77030, USA. Phone:
713.792.6363; Fax: 713.796.8655; E-mail: firstname.lastname@example.org.
capture fluorescent images of Matrigel cultures. Fluorescence-stained slides
depicting filopodia, lamellipodia, and focal adhesions were imaged on a Nikon
Ti microscope equipped with a CoolSnap HQ2 camera (Photometrics). Image
stacks of 100-nm sections were then deconvoled with Autoquant (Media
Cybernetics). For attachment assays, 1 × 105 cells were seeded on 24-well plates
and incubated for 1–3 hours. After washing with PBS twice, attached cells were
stained with 0.1% crystal violet, and optical density was measured at 595 nm.
miR expression profiling. Total RNA was isolated from 393P_vector and
393P_ZEB1 cells, profiled using the DiscovArray platform by Asuragen,
and analyzed as described previously (17). The platform included all but
91 of the 1,086 probes in the Sanger miRBase version 9.2 covering humans,
rats, and mice and an additional 12,894 exploratory probes covering mul-
tiple other species. Transcriptomic data sets were deposited in GEO (acces-
sion no. GSE38386). Heat maps were generated using Cluster and Tree-
View software (http://rana.lbl.gov/EisenSoftware.htm).
Quantitative RT-PCR (Q-PCR). Total RNA was isolated from the cells
using TRIzol (Invitrogen) according to the manufacturer’s protocol. To
analyze mRNA levels, Q-PCR assays were performed after reverse transcrip-
tion with Superscript III reverse transcriptase (Invitrogen) using a SYBR-
Green–based system (Applied Biosystems). mRNA levels were normalized
on the basis of mRNA for ribosomal protein L32 (Rpl32). See Supplemental
Table 2 for primer sequences. miR levels were quantified using Taqman
microRNA assays (Applied Biosystems) according to the manufacturer’s
protocol and normalized on the basis of snoRNA-135.
Luciferase reporter assays. Cells were seeded on 24-well plates (1 × 105 cells/
well) 1 day before transfection, and then transfected with 500 ng luciferase
reporter plasmids and 50 ng hRL-control vector. After 48 hours, lucifer-
ase activity was measured using Dual-Luciferase Reporter Assay System
(Promega). A murine miR-34a promoter region was isolated by PCR from
TC-1 murine ES cell genomic DNA and ligated into the pGL3-basic vector
(Promega). p53-reporter plasmids (Mdm2, 14-3-3σ, and generic promoters)
were gifts from M.-H. Lee (University of Texas MD Anderson Cancer Center,
Houston, Texas, USA), and the ΔNp63 and TAp63 promoter reporters were
gifts from I. Shachar (Weizmann Institute of Science, Rehovot, Israel). For
the 3′-UTR assay, murine 3′-UTRs were amplified by PCR from genomic
DNA and ligated into pCI-neo-hRL vector (13). 3′-UTR reporters (500 ng)
and pGL3-control (50 ng; Promega) were cotransfected into 344SQ cells
seeded on 24-well plates (1 × 105 cells/well) in the presence or absence of
pre–miR-34a precursor (5 nM; Ambion). A PCR-based site-directed muta-
genesis strategy was carried out to generate mutant constructs.
ChIP assays. 393P cells were transiently transfected with Myc-ΔNp63β
or empty vector. Cells were cross-linked with 1% formaldehyde and then
incubated in lysis buffer (50 mM Tris-HCl, pH 8.1; 1% SDS; 10 mM EDTA;
and protease inhibitor cocktail) on ice for 10 min. After sonication (Cole-
Parmer GEX-130 sonicator; 50% power, pulse on for 10 s, pulse off for 10 s,
20 cycles), samples were immunoprecipitated with anti-Myc tag antibody
(Millipore) or anti-mouse IgG (Santa Cruz). DNA was eluted and purified
with PCR purification kit (Qiagen), and quantitative PCR was carried out
with specific primers to amplify the p53/63-binding region of the miR-34a
promoter (forward, 5′-CAGCCTGGAGGAGGATCGA-3′; reverse,
5′-TCCCAAAGCCCCCAATCT-3′) or a nonspecific region within exon 2 as
a negative control (forward, 5′-AAGCGGGTTTCAAGTGCATCTCAG-3′;
reverse, 5′-TCAGGCTACTAAACCAGTTGCCCT-3′). To detect binding of
endogenous p63 to miR-34a promoter elements, cellular proteins from
393P cells and murine primary keratinocytes were cross-linked to DNA
using 1% formaldehyde, and chromatin was prepared as described earlier
1. Mehlen P, Puisieux A. Metastasis: a question of life
or death. Nat Rev Cancer. 2006;6(6):449–458.
2. Brabletz T, Jung A, Spaderna S, Hlubek F, Kirchner
T. Opinion: migrating cancer stem cells - an inte-
grated concept of malignant tumour progression.
Nat Rev Cancer. 2005;5(9):744–749.
3. Singh SK, et al. Identification of human
brain tumour initiating cells. Nature. 2004;
4. Ginestier C, et al. ALDH1 is a marker of normal
and malignant human mammary stem cells and a
predictor of poor clinical outcome. Cell Stem Cell.
14 The Journal of Clinical Investigation http://www.jci.org
5. Nguyen LV, Vanner R, Dirks P, Eaves CJ. Cancer
stem cells: an evolving concept. Nat Rev Cancer. 2012;
6. Kalluri R, Weinberg RA. The basics of epitheli-
al-mesenchymal transition. J Clin Invest. 2009;
7. Sanchez-Tillo E, et al. ZEB1 represses E-cadherin
and induces an EMT by recruiting the SWI/SNF
chromatin-remodeling protein BRG1. Oncogene.
8. Aghdassi A, et al. Recruitment of histone deacety-
lases HDAC1 and HDAC2 by the transcriptional
repressor ZEB1 downregulates E-cadherin expres-
sion in pancreatic cancer. Gut. 2012;61(3):439–448.
9. Brabletz S, Brabletz T. The ZEB/miR-200 feedback
loop—a motor of cellular plasticity in development
and cancer? EMBO Rep. 2010;11(9):670–677.
10. Brabletz S, et al. The ZEB1/miR-200 feedback loop
controls Notch signalling in cancer cells. EMBO J.
11. Browne G, Sayan AE, Tulchinsky E. ZEB proteins
link cell motility with cell cycle control and cell sur-
vival in cancer. Cell Cycle. 2010;9(5):886–891.
12. Wellner U, et al. The EMT-activator ZEB1 promotes
tumorigenicity by repressing stemness-inhibiting
microRNAs. Nat Cell Biol. 2009;11(12):1487–1495.
13. Gregory PA, et al. The miR-200 family and miR-205
regulate epithelial to mesenchymal transition
by targeting ZEB1 and SIP1. Nat Cell Biol. 2008;
14. 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. 2008;
15. Bracken CP, et al. A double-negative feedback loop
between ZEB1-SIP1 and the microRNA-200 family
regulates epithelial-mesenchymal transition. Cancer
16. Zheng S, El-Naggar AK, Kim ES, Kurie JM, Lozano
G. A genetic mouse model for metastatic lung can-
cer with gender differences in survival. Oncogene.
17. Gibbons DL, et al. Contextual extracellular cues
promote tumor cell EMT and metastasis by regu-
lating miR-200 family expression. Genes Dev. 2009;
18. Liu C, et al. The microRNA miR-34a inhibits pros-
tate cancer stem cells and metastasis by directly
repressing CD44. Nat Med. 2011;17(2):211–215.
19. Wang B, et al. TGFbeta-mediated upregulation
of hepatic miR-181b promotes hepatocarcino-
genesis by targeting TIMP3. Oncogene. 2010;
20. Weiss FU, et al. Retinoic acid receptor antagonists
inhibit miR-10a expression and block metastatic
behavior of pancreatic cancer. Gastroenterology. 2009;
21. Meng F, Wehbe-Janek H, Henson R, Smith H, Patel
T. Epigenetic regulation of microRNA-370 by
interleukin-6 in malignant human cholangiocytes.
22. Huang X, et al. Hypoxia-inducible mir-210 regu-
lates normoxic gene expression involved in tumor
initiation. Mol Cell. 2009;35(6):856–867.
23. Ferretti E, et al. Concerted microRNA control
of Hedgehog signalling in cerebellar neuronal
progenitor and tumour cells. EMBO J. 2008;
24. Gao XN, et al. MicroRNA-193a represses c-kit
expression and functions as a methylation-silenced
tumor suppressor in acute myeloid leukemia. Onco-
25. Yan D, et al. MicroRNA-1/206 targets c-Met and
inhibits rhabdomyosarcoma development. J Biol
26. Miko E, et al. miR-126 inhibits proliferation of
small cell lung cancer cells by targeting SLC7A5.
FEBS Lett. 2011;585(8):1191–1196.
27. He L, et al. A microRNA component of the
p53 tumour suppressor network. Nature. 2007;
28. Siemens H, et al. miR-34 and SNAIL form a
double-negative feedback loop to regulate epi-
thelial-mesenchymal transitions. Cell Cycle. 2011;
29. Su X, et al. TAp63 suppresses metastasis through
coordinate regulation of Dicer and miRNAs.
30. Fontemaggi G, et al. deltaEF1 repressor controls
selectively p53 family members during differentia-
tion. Oncogene. 2005;24(49):7273–7280.
31. Cho MS, Chan IL, Flores ER. DeltaNp63 transcrip-
tionally regulates brachyury, a gene with diverse
roles in limb development, tumorigenesis and
metastasis. Cell Cycle. 2010;9(12):2434–2441.
32. Shintani Y, Maeda M, Chaika N, Johnson KR,
Wheelock MJ. Collagen I promotes epithelial-to-
mesenchymal transition in lung cancer cells via
transforming growth factor-beta signaling. Am J
Respir Cell Mol Biol. 2008;38(1):95–104.
33. Bhattacharjee A, et al. Classification of human
lung carcinomas by mRNA expression profiling
reveals distinct adenocarcinoma subclasses. Proc
Natl Acad Sci U S A. 2001;98(24):13790–13795.
34. Chitale D, et al. An integrated genomic analysis of
lung cancer reveals loss of DUSP4 in EGFR-mutant
tumors. Oncogene. 2009;28(31):2773–2783.
35. Tomida S, et al. Relapse-related molecular sig-
nature in lung adenocarcinomas identifies
patients with dismal prognosis. J Clin Oncol. 2009;
36. Gibbons DL, et al. Expression signatures of meta-
static capacity in a genetic mouse model of lung
adenocarcinoma. PLoS One. 2009;4(4):e5401.
37. Etienne-Manneville S, Hall A. Rho GTPases in cell
biology. Nature. 2002;420(6916):629–635.
38. Gimona M, Buccione R, Courtneidge SA, Linder
S. Assembly and biological role of podosomes
and invadopodia. Curr Opin Cell Biol. 2008;
39. Tomar A, Schlaepfer DD. Focal adhesion kinase:
switching between GAPs and GEFs in the regu-
lation of cell motility. Curr Opin Cell Biol. 2009;
40. Savarese F, et al. Satb1 and Satb2 regulate embry-
onic stem cell differentiation and Nanog expres-
sion. Genes Dev. 2009;23(22):2625–2638.
41. Kim NH, et al. p53 and microRNA-34 are suppres-
sors of canonical Wnt signaling. Sci Signal. 2011;
42. Takagi S, Nakajima M, Kida K, Yamaura Y, Fukami
T, Yokoi T. MicroRNAs regulate human hepatocyte
nuclear factor 4alpha, modulating the expression of
metabolic enzymes and cell cycle. J Biol Chem. 2010;
43. Spaderna S, et al. A transient, EMT-linked loss of
basement membranes indicates metastasis and
poor survival in colorectal cancer. Gastroenterology.
44. Spaderna S, Brabletz T, Opitz OG. The
miR-200 family: central player for gain and loss of
the epithelial phenotype. Gastroenterology. 2009;
45. Yang J, et al. Twist, a master regulator of morpho-
genesis, plays an essential role in tumor metastasis.
46. Mani SA, et al. Mesenchyme Forkhead 1 (FOXC2)
plays a key role in metastasis and is associated with
aggressive basal-like breast cancers. Proc Natl Acad
Sci U S A. 2007;104(24):10069–10074.
47. Peinado H, Olmeda D, Cano A. Snail, Zeb and
bHLH factors in tumour progression: an alliance
against the epithelial phenotype? Nat Rev Cancer.
48. Yang MH, et al. Bmi1 is essential in Twist1-induced
epithelial-mesenchymal transition. Nat Cell Biol.
49. Shimono Y, et al. Downregulation of miRNA-200c
links breast cancer stem cells with normal stem
cells. Cell. 2009;138(3):592–603.
50. Proia TA, et al. Genetic predisposition directs
breast cancer phenotype by dictating progenitor
cell fate. Cell Stem Cell. 2011;8(2):149–163.
51. Kissil JL, et al. Requirement for Rac1 in a K-ras
induced lung cancer in the mouse. Cancer Res.
52. Mack NA, Whalley HJ, Castillo-Lluva S, Malliri A.
The diverse roles of Rac signaling in tumorigenesis.
Cell Cycle. 2011;10(10):1571–1581.
53. Valastyan S, Weinberg RA. Roles for microRNAs in
the regulation of cell adhesion molecules. J Cell Sci.
54. Postigo AA, Depp JL, Taylor JJ, Kroll KL. Regula-
tion of Smad signaling through a differential
recruitment of coactivators and corepressors by
ZEB proteins. EMBO J. 2003;22(10):2453–2462.
55. Nishimura G, et al. DeltaEF1 mediates TGF-beta
signaling in vascular smooth muscle cell differen-
tiation. Dev Cell. 2006;11(1):93–104.
56. Eger A, et al. DeltaEF1 is a transcriptional repressor
of E-cadherin and regulates epithelial plasticity in
breast cancer cells. Oncogene. 2005;24(14):2375–2385.
57. Boominathan L. The guardians of the genome
(p53, TA-p73, and TA-p63) are regulators of tumor
suppressor miRNAs network. Cancer Metastasis Rev.
58. Rivetti d V, et al. p63-microRNA feedback in
keratinocyte senescence. Proc Natl Acad Sci U S A.
59. Knouf EC, et al. An integrative genomic approach
identifies p73 and p63 as activators of miR-200
microRNA family transcription. Nucleic Acids Res.
60. Ory B, et al. A microRNA-dependent program
controls p53-independent survival and chemosen-
sitivity in human and murine squamous cell carci-
noma. J Clin Invest. 2011;121(2):809–820.
61. Kasinski AL, Slack FJ. Epigenetics and genetics.
MicroRNAs en route to the clinic: progress in vali-
dating and targeting microRNAs for cancer therapy.
Nat Rev Cancer. 2011;11(12):849–864.
62. Trang P, et al. Systemic delivery of tumor sup-
pressor microRNA mimics using a neutral lipid
emulsion inhibits lung tumors in mice. Mol Ther.
63. Wiggins JF, et al. Development of a lung can-
cer therapeutic based on the tumor suppressor
microRNA-34. Cancer Res. 2010;70(14):5923–5930.
64. Duncan R. The dawning era of polymer therapeutics.
Nat Rev Drug Discov. 2003;2(5):347–360.
65. Estevez MC, et al. Nanoparticle–aptamer conju-
gates for cancer cell targeting and detection. Methods
Mol Biol. 2010;624:235–248.
66. Chen Y, Zhu X, Zhang X, Liu B, Huang L. Nanopar-
ticles modified with tumor-targeting scFv deliver
siRNA and miRNA for cancer therapy. Mol Ther.
67. Brown CY, et al. Robust, reversible gene knock-
down using a single lentiviral short hairpin RNA
vector. Hum Gene Ther. 2010;21(8):1005–1017.
68. Flores ER, et al. p63 and p73 are required for
p53-dependent apoptosis in response to DNA
damage. Nature. 2002;416(6880):560–564.
69. Creighton CJ, et al. Insulin-like growth factor-I
activates gene transcription programs strongly
associated with poor breast cancer prognosis. J Clin