Wnt activity defines colon cancer stem cells and is regulated by the microenvironment.

Abstract

Despite the presence of mutations in APC or beta-catenin, which are believed to activate the Wnt signalling cascade constitutively, most colorectal cancers show cellular heterogeneity when beta-catenin localization is analysed, indicating a more complex regulation of Wnt signalling. We explored this heterogeneity with a Wnt reporter construct and observed that high Wnt activity functionally designates the colon cancer stem cell (CSC) population. In adenocarcinomas, high activity of the Wnt pathway is observed preferentially in tumour cells located close to stromal myofibroblasts, indicating that Wnt activity and cancer stemness may be regulated by extrinsic cues. In agreement with this notion, myofibroblast-secreted factors, specifically hepatocyte growth factor, activate beta-catenin-dependent transcription and subsequently CSC clonogenicity. More significantly, myofibroblast-secreted factors also restore the CSC phenotype in more differentiated tumour cells both in vitro and in vivo. We therefore propose that stemness of colon cancer cells is in part orchestrated by the microenvironment and is a much more dynamic quality than previously expected that can be defined by high Wnt activity.

Full text

Page 1

ARTICLES
Wnt activity defines colon cancer stem cells and is
regulated by the microenvironment
Louis Vermeulen1,5, Felipe De Sousa E Melo1,5, Maartje van der Heijden1, Kate Cameron1, Joan H. de Jong1,
Tijana Borovski1, Jurriaan B. Tuynman1, Matilde Todaro2, Christian Merz3, Hans Rodermond1, Martin R. Sprick1,
Kristel Kemper1, Dick J. Richel1, Giorgio Stassi2,4 and Jan Paul Medema1,6.
Despite the presence of mutations in APC or β-catenin, which are believed to activate the Wnt signalling cascade constitutively,
most colorectal cancers show cellular heterogeneity when β-catenin localization is analysed, indicating a more complex regulation
of Wnt signalling. We explored this heterogeneity with a Wnt reporter construct and observed that high Wnt activity functionally
designates the colon cancer stem cell (CSC) population. In adenocarcinomas, high activity of the Wnt pathway is observed
preferentially in tumour cells located close to stromal myofibroblasts, indicating that Wnt activity and cancer stemness may be
regulated by extrinsic cues. In agreement with this notion, myofibroblast-secreted factors, specifically hepatocyte growth factor,
activate β-catenin-dependent transcription and subsequently CSC clonogenicity. More significantly, myofibroblast-secreted
factors also restore the CSC phenotype in more differentiated tumour cells both in vitro and in vivo. We therefore propose that
stemness of colon cancer cells is in part orchestrated by the microenvironment and is a much more dynamic quality than
previously expected that can be defined by high Wnt activity.
The sequential events driving the transition of normal colonic mucosa
to adenocarcinoma have been well documented and depend critically on
alterations to Wnt signalling1–4. In normal cells the transcriptional regula-
tor β-catenin is tightly controlled by a multiprotein complex that contains
the tumour suppressor adenomatous polyposis coli (APC)2. Activation
of Frizzled receptors by Wnt ligands disrupts this complex and results in
the translocation of β-catenin to the nucleus, where it associates with the
T-cell factor/lymphoid enhancer factor (TCF/LEF) family of transcription
factors to activate specific Wnt target genes5,6. Under physiological condi-
tions, Wnt activity is crucial for intestinal stem cells and crypt homeosta-
sis7. However, Wnt signalling also has a central function in pathological
settings2. Mutation of APC or β-catenin is an early event in the transfor-
mation of colonic epithelial cells, but established colorectal cancers also
depend critically on Wnt signalling8. APC mutations generally result in
a defective β-catenin degradation complex and, as a consequence, the
accumulation of β-catenin in the nucleus and the perpetual transcrip-
tion of Wnt target genes9. Although this would imply that all tumour cells
contain active Wnt signalling, immunohistochemical studies have revealed
that colon carcinomas harbouring APC mutations do not contain nuclear
β-catenin homogeneously (Fig. 1a)10,11. This so-called β-catenin paradox
is intriguing in the light of recent observations that indicate that only a
subset of tumour cells, CSCs, are endowed with tumorigenic capacity12,
whereas most tumour cells have undergone differentiation and lost their
tumorigenic potential. CSCs have been identified by using several markers,
such as CD133 and CD166 (refs 13–16). Because Wnt signalling is a domi-
nant force preserving the normal fate of colon stem cells2, we proposed a
preponderant role for this pathway in colon CSCs as well.
Using a TCF/LEF reporter that directs the expression of enhanced
green fluorescent protein (TOP–GFP) we provide evidence that Wnt
signalling activity is a marker for colon CSCs and is regulated by the
microenvironment. Moreover, we show that differentiated cancer cells,
which have lost the capacity to form tumours and are no longer clono-
genic, can be reprogrammed to express CSC markers and regain their
tumorigenic capacity when stimulated with myofibroblast-derived fac-
tors. This suggests that cancer stemness is not a rigid feature but can be
modulated and even installed by the microenvironment.
ReSultS
Colon cancer spheroidal cultures have heterogeneous Wnt
activity levels
Primary spheroidal cultures of colon cancer cells consist of cells
expressing CSC markers and on injection induce a tumour that closely
1Laboratory for Experimental Oncology and Radiobiology (LEXOR), Center for Experimental Molecular Medicine (CEMM), Academic Medical Center (AMC), University of
Amsterdam, 1105 AZ Amsterdam, The Netherlands. 2Department of Surgical and Oncological Sciences, Cellular and Molecular Pathophysiology Laboratory, University
of Palermo, 90127 Palermo, Italy. 3APOGENIX GmbH, 69120 Heidelberg, Germany. 4Present address: Cellular and Molecular Oncology, IRCCS Fondazione Salvatore
Maugeri, 27100 Pavia, Italy.
5These authors contributed equally to this work.
6Correspondence should be addressed to J.P.M. (J.P.Medema@amc.uva.nl)
Received 22 January 2010; accepted 26 March 2010; published online 25 April 2010; DOI:10.1038/ncb2048
468
nature cell biology VOLUME 12 | NUMBER 5 | MAY 2010
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 2

ARTICLES
resembles the original human malignancy17. To investigate the rela-
tion of the β-catenin paradox (Fig. 1a) with the CSC phenotype, we
assessed Wnt signalling activity in these colon CSC cultures with a
TOP–GFP reporter or with control reporters, phosphoglycerate kinase
(PGK)–GFP or cytomegalovirus (CMV)–GFP18. To exclude variation
in lentiviral integration site and copy number between cells, the trans-
duced culture was single-cell cloned, a procedure that preserves all
characteristics of the spheroidal culture17. Strikingly, cells in these
single-cell-derived TOP–GFP cultures still showed considerable het-
erogeneity in Wnt signalling levels (about 100-fold) (Fig. 1b). These
expression differences were not observed when a constitutively active
CMV promoter was used to drive GFP expression (Fig. 1b), indi-
cating that heterogeneity is due to differences in β-catenin-driven
transcription even though that these cells carry an APC mutation.
In agreement with this notion, the variation in TOP–GFP levels was
accompanied by heterogeneity in nuclear β-catenin localization
(Fig. 1c; Supplementary Information, Fig. S1a, b). Moreover, micro-
array analysis on the highest and lowest 10% of TOP–GFP-expressing
cells revealed that stem-cell-associated Wnt target genes, including
those encoding the intestinal stem cell markers LGR5 (leucine-rich
repeat-containing G protein-coupled receptor 5) and ASCL2 (acha-
ete-scute-like 2) (refs 19, 20), were predominantly upregulated in the
TOP–GFPhigh fraction, whereas genes associated with epithelial dif-
ferentiation (those encoding mucin 2, MUC2; cytokeratin 20, KRT20
and fatty-acid binding protein 2, FABP2) were clearly enhanced in
the TOP–GFPlow fraction (Fig. 1d; see Supplementary Information,
Table S1, for a list of the most significant genes). Expression of dif-
ferentiation-associated genes in TOP–GFPlow was confirmed by quan-
titative PCR (qPCR) as well as at the protein level (Fig. 1e, f). Taken
together, these results indicates that spheroidal CSC cultures contain
heterogeneity in Wnt signalling activity, which is inversely correlated
with heterogeneity in differentiation markers.
abd
f
e
0
1
2
3
c
–3–2
log2 (fold change)
–1012
TOP–GFPlow
TOP–GFPPGK/CMV–GFPTOP–GFPhigh
TOP–GFPlow
TOP–GFPhigh
Diff.
Wnt targets
KRT20
FABP2
CGHA
MUC2
LEF1
Cyclin D1
Survivin
MYC
BMP4
CD44
AXIN2
Lgr5
Tcf-1
Ascl2
mRNA
10.0
7.5
17.5
7.5
12.5
2.5
15.0
5.0
0
10.0
5.0
LGR5
MUC2
KRT20
FABP2
Survivin
AXIN2
TOP–GFPhigh
TOP–GFPlow
Relative expression
CK20FABP2MUC2
Figure 1 Wnt heterogeneity in primary tumours and in CSC culture. (a)
Colorectal cancer patient material stained for β-catenin. Scale bar, 20 µm.
(b) The top two rows show two separate single-cell-cloned colon CSC
cultures, lentivirally transduced with TOP–GFP18, revealed by phase contrast
(left) and fluorescence microscopy (middle). Nuclei were counterstained with
4,6-diamidino-2-phenylindole (red, right). Scale bars, 20 µm. Bottom two
panels: fluorescence-activated cell sorting (FACS) profile of GFP intensity range
for a single-cell-cloned TOP–GFP-carrying colon CSC culture (green line, left)
compared with the sharp range of GFP in the CMV–GFP (blue line, right) culture
and non-single-cell-cloned PGK–GFP-transduced culture (red line, right). Non-
transduced parental culture was used as a control for GFP background intensity
(black lines). (c) Cytospins of the highest and lowest 10% TOP–GFP fractions
stained for β-catenin. Scale bars, 20 µm. (d) Microarray analysis was performed
on the highest and lowest 10% TOP–GFP fractions. The log2 fold change in
gene expression between the TOP–GFPhigh and TOP–GFPlow fraction is shown.
Each bar represents the average fold change of three different single-cell-cloned
TOP–GFP cultures. Several Wnt target genes are listed and their expression is
correlated with the TOP–GFPhigh fraction. Genes associated with differentiation
(Diff.) are upregulated in the TOP–GFPlow fraction. Error bars represent s.e.m. for
four different data sets from three different cultures. (e) Differential expression
of several Wnt target genes (left) and differentiation markers (right) between
TOP–GFPhigh and TOP–GFPlow fractions validated by qPCR. Error bars represent
s.d. (n = 3). (f) Cytospins of sorted TOP–GFPhigh and TOP–GFPlow cells stained for
differentiation markers as indicated. Scale bars, 80 µm.
nature cell biology VOLUME 12 | NUMBER 5 | MAY 2010
469
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 3

ARTICLES
Colon CSCs are characterized by high expression levels of Wnt
The variation in TOP–GFP levels not only determined gene-expres-
sion differences; it also had significant biological consequences. In
the TOP–GFPhigh fraction, clonogenic potential was superior to that of
TOP–GFPlow cells, whereas differential clonogenicity was not observed
in cultures expressing GFP driven by constitutive promoters (Fig. 2a).
Intermediate TOP–GFP-expressing cells had intermediate clonogenic-
ity and intermediate gene-expression profiles (Fig. 2a; Supplementary
Information, Fig. S1c), suggesting that Wnt signalling activity could
serve as a functional marker for tumour cell clonogenicity and cancer
stemness. Indeed, TOP–GFP-based segregation of clonogenicity is a
common principle and was observed with separate single-cell-derived
clones of the same CSC line (Fig. 2b) as well as in a series of other pri-
mary colon spheroidal cultures or several established human colon can-
cer cell lines (Fig. 2c–e; Supplementary Information, Fig. S2a). These
separate spheroidal cultures and cell lines were not cloned from a single
cell, which excludes the possibility of clonal artefacts but also limits the
segregation of clonogenicity as a result of differences between cells in
lentiviral integration. Mutation analysis of the different lines indicated
that Wnt signalling heterogeneity and its effect on clonogenicity seems
to be independent of the mutation status of p53 (compare Co100; wild-
type, WT, p53, for example) with LM5 (mutant p53)) or K-Ras (compare
Co100; mutant K-Ras with LM5; WT K-Ras, for example), or micros-
atellite stability (compare Co56, DLD-1; microsatellite unstable with
Co100 and LM5; microsatellite stable, for example). The only striking
exception is the HCT-116 cell line, which contains a β-catenin mutation.
Whether this is a consistent finding for β-catenin mutant lines remains
to be established, but recent findings have pointed out that HCT-116 is
not organized in a hierarchical fashion and may therefore not contain a
more clonogenic CSC fraction21,22. We therefore propose that the activ-
ity of the Wnt pathway defines a hierarchical divergence of CSCs and
their differentiated progeny in colorectal cancer. This is supported by
the finding that Wnt activity levels correlate with the surface expression
of previously applied markers for colon CSCs (Fig. 2f).
CSCs are defined by their ability to induce tumours that closely
resemble the original malignancy on injection into immune-deficient
mice23. We reported previously that spheroidal cultured colon cancer
cells fulfilled these criteria and therefore contain CSCs14,17. Injection of
a
1 in every
1
2
4
8
1 in every1 in every
GFPhigh
GFPintermediate
GFPlow
bc
d
ef
1 in every
512
256
128
64
32
16
2
4
8
512
256
128
64
32
16
2
4
8
128
64
32
16
2
4
8
128
64
32
16
TOP
PGK
CMV
C2
DLD1
HT29
HCT116
LM3
LM5
Co56
Co2
Co1
Co3
Apo.1
B5
A4
A6
G7
Co56
Co100
LM3LM5
CD29CD44
CD133
CD166
TOP–GFP
CD24
Figure 2 Cells with high Wnt signalling show CSC properties in vitro. (a) A
limiting dilution was performed on the high, intermediate and low TOP–GFP cell
fractions. The graph presents the clonogenic potential of each cell fraction. As
a control a CMV–GFP single-cell-derived culture and a non-single-cell-derived
PGK–GFP culture are depicted. GFPhigh, GFPintermediate and GFPlow represent the
highest, middle and lowest 10% cell fractions, respectively. See Methods for
details on limiting-dilution statistics and scheme. (b) Limiting-dilution assay
and clonogenic potential of the five independent single-cell-derived clones from
line Co100. (c) Limiting-dilution assay and clonogenic potential of the highest
and lowest 10% TOP–GFP levels of a series of TOP–GFP-expressing CSC lines
(non-single-cell-cloned). (d) A panel of haematoxylin/eosin stainings of the
corresponding primary colon cancers used to generate the spheroidal cultures
in a–c. (e) Limiting-dilution assay and clonogenic potential of the highest
and lowest 10% TOP–GFP levels of three established colon cancer cell lines.
b, c and e, white bars represent highest 10% TOP–GFP cells and black bars
represent lowest 10% TOP–GFP cells. Error bars in a–c and e represent 95%
confidence intervals. See Methods for details on limiting dilution statistics and
scheme. (f) TOP–GFP levels are associated with CSC marker expression. CD133
correlates with TOP–GFPhigh intensity. The combination of CD24/CD29 or CD44/
CD166 also shows a correlation with the TOP–GFPhigh fraction.
470
nature cell biology VOLUME 12 | NUMBER 5 | MAY 2010
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 4

ARTICLES
TOP–GFP cells into nude mice confirmed that CSCs can be identified
through their Wnt signalling activity, because the TOP–GFPhigh frac-
tion was much more effective in inducing tumours on injection into
mice (Fig. 3a). A similar segregation was observed when TOP–GFPhigh
cells (3 of 6 injections yielded tumours when 100 cells injected) and
TOP–GFPlow cells (0 of 6 injections yielded tumours) were injected into
fully immunodeficient NOD/SCID (non-obese diabetic/severe com-
bined immunodeficiency) IL-2 receptor gamma chain null (Il2rg-/-) mice,
suggesting that the TOP–GFPhigh definition of CSCs is independent of
a remnant immune response. More significantly, xenografts derived
from TOP–GFP-transduced spheroidal cultures closely recapitulated
the original malignancy (Figs 2d and 3b) but also showed a heterogene-
ous distribution of cells containing nuclear β-catenin and TOP–GFP
activity (Fig. 3c–e). The TOP–GFPhigh fraction in vivo contained both
cycling and non-cycling cells, as demonstrated by co-staining for Ki-67
(Supplementary Information, Fig. S2b). We consider this as strong evi-
dence for the maintained regulation of Wnt signalling in clonal popula-
tions of colorectal cancer cells in vivo, which is in line with the β-catenin
paradox10,11. Consistent with the idea that the clonogenic CSC potential
resides in the ‘Wnt-high’ population, we found that in vitro outgrowth
of cells isolated from such TOP–GFP-expressing xenografts is much
more effective with TOP–GFPhigh cells than in TOP–GFPlow cells (Fig. 3f).
Moreover, we also observed a strong correlation between TOP–GFP
activity and previously used colon CSC markers17 in ex vivo-derived
cells (Fig. 3g).
To prove formally that TOP–GFPhigh cells are CSCs and have self-
renewal capacity, we performed direct in vivo transfer of xenograft-derived
cells with a limiting dilution of the 10% highest, lowest and unselected
TOP–GFP cells. Whereas TOP–GFPhigh cells induced tumour growth
effectively on transfer, TOP–GFPlow cells largely failed to do so (Fig. 4a).
Using Wnt activity as a CSC marker, we even managed to generate colon
cancer xenografts starting from either one ex vivo-derived TOP–GFPhigh
cell (1 of 4 injected) (Fig. 4b) or from one spheroidal-culture-derived
TOP–GFPhigh cell (8 of 50 injected). The resulting tumours again resem-
bled the original malignancy in differentiation and also showed clear het-
erogeneity in TOP–GFP (Fig. 4c). Taken together, these results illustrate
that a subfraction of colon tumour cells are endowed with tumorigenic
potential, which can be identified on the basis of Wnt activity. At the same
time, these data point to Wnt regulatory mechanisms in colorectal cancer
cells even in the presence of APC mutations.
Myofibroblast-secreted factors enhance Wnt signalling activity
in colon cancer cells
Having established that TCF/LEF-driven transcription is regulated in
colorectal cancer and is an important determinant for CSC features, we
wished to understand the mechanisms underlying this regulation. It
d
a
b
f
c
g
12 4 6 812 16 2024
0
0.2
0.4
0.6
0.8
1.0
8
*
TOP–GFPhigh
TOP–GFPlow
Fraction sphere formation
Number of cells plated per well
High
Low
1 in every
TOP–GFP
CD44
CD133
CD24
1,024
512
256
128
64
32
16
Cells injected
100
6/6
0/6
5/6
5/6
Fraction
Total
TOP–GFPlow
TOP–GFPintermediate
TOP–GFPhigh
10
1/6
0/6
4/6
3/6
–
–
–
5,000
3/6
1,000
6/6
1/6
4/6
6/6
e
Figure 3 Cells with high Wnt signalling show CSC properties in vivo. (a) In vivo
transfer of TOP–GFP-transduced cultures. Cell numbers from the indicated
populations were injected into nude mice after FACS. The number of successful
tumour initiations after nine weeks out of six injections for each condition
is shown. The TOP–GFPhigh fraction showed the highest tumour-initiating
capacity. (b) Haematoxylin/eosin staining of xenografts shows clear evidence
of colon carcinoma morphology. (c) Xenografts show heterogeneous nuclear
β-catenin (arrow); the asterisk indicates a region without nuclear localization.
(d, e) Immunofluorescence (d) and immunohistochemistry (e) for GFP show
clear heterogeneous TOP–GFP levels. Scale bars in b–e, 50 µm. (f) TOP–GFP
xenografts were dissociated, tumour cells from the TOP–GFPhigh and TOP–GFPlow
fractions were plated for limiting dilutions and the clonogenic potential was
determined. Error bars represent 95% confidence intervals. A representative
example of two independent experiments is shown. (g) Correlation of CSC
markers and TOP–GFP levels is also seen in freshly dissociated xenografts.
a
b
c
Ex vivo single TOP–GFPhigh cell induced xenograft
Cells injected
Fraction
Total
TOP–GFPlow
TOP–GFPhigh
110
0/6
100
1/6
1,000
2/6
1/6
2/6
0/6
3/6
0/6
5/6
ND
ND
1/4
TOP–GFP
Figure 4 Ex vivo tumorigenic assay and single-cell-derived tumours. (a)
TOP–GFP xenografts were dissociated and subcutaneously injected to
determine their tumorigenicity and thus their self-renewal capacity. Cell
numbers from indicated populations were injected into nude mice after
FACS. The number of successful tumour initiations after nine weeks out
of six injections for each condition is shown. The TOP–GFPhigh fraction
showed the highest tumour-initiating capacity. The total population was
used as a control. Note that one single TOP–GFPhigh cell initiated a tumour
out of four implants (see Methods for details). ND, none detected. (b)
Xenotransplanted tumour derived from a single TOP–GFPhigh cell sorted
directly ex vivo. The inset shows a single cell; the figure shows an isolated
xenotransplant. (c) Haematoxylin/eosin staining (left), Alcian Blue staining
(middle) and anti-GFP immunohistochemistry (right) of the single-cell-
derived tumour. Scale bars, 100 µm.
nature cell biology VOLUME 12 | NUMBER 5 | MAY 2010
471
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 5

ARTICLES
was apparent from our spheroidal cultures that part of the regulation is
a tumour-cell-intrinsic feature, because single CSCs quickly reformed
cultures with heterogeneous Wnt activity. Nevertheless, in vivo observa-
tions suggest that extrinsic factors may also be instrumental in directing
Wnt activity in colorectal cancer10,11. Different cell types recruited and/or
activated within the tumour could participate in shaping and organizing
this heterogeneity. This external control emerges as a sustainable force in
malignancy progression24. Strikingly, normal intestinal stem cells have also
recently been shown to have cell-autonomous regulation of Wnt activity25,
although extrinsic cues emanating from cells within or surrounding the
crypt also contribute to normal epithelial stem cell function26. In colorectal
cancer, the β-catenin paradox has indirectly implicated the stroma, of which
myofibroblasts are an important component (Supplementary Information,
Fig. S3a), as a potential regulator of Wnt signalling10,11. To directly address
whether myofibroblasts could affect Wnt signalling we analysed co-cultures
of myofibroblasts and CSCs. We have previously shown that colon CSCs
differentiate and lose CSC markers in serum-containing medium14,17.
However, when CSCs were co-cultured with a colonic myofibroblast cell
line or even in the presence of conditioned medium derived from such
myofibroblasts (MFCM), morphological and molecular differentiation
was prevented (Fig. 5a, b; Supplementary Information, Fig. S3b, c). We also
observed a strong functional effect of MFCM on tumour cell clonogenicity,
which was markedly changed when CSCs were incubated in the presence
or absence of MFCM (more than 50-fold; Fig. 5c). Consistent with the fact
that Wnt activity is correlated with clonogenicity, stimulation with MFCM
strongly increased nuclear β-catenin localization and Wnt reporter activity
(Fig. 5f, g), which shows that myofibroblasts secrete factors regulating Wnt
activity and thereby clonogenic capacity.
In an effort to define the growth factors involved in this crosstalk,
a cytokine antibody array was performed on MFCM, which revealed
hepatocyte growth factor (HGF) as one of the abundant factors present
in MFCM (Fig. 5d; Supplementary Information, Fig. S4a). In agreement,
a significant amount of HGF was also secreted by myofibroblasts derived
from samples from primary colon cancer (Fig. 5e). HGF has been sug-
gested to have a role in colorectal cancer, and its receptor HGFR/c-
Met is associated with both Wnt signalling activity27–29 and colorectal
ab
Relative expression,
Control
HGF
MFCM
HGF + PHA
MFCM + PHA
PHA
MFCM + antibody
Antibody
1
2
3
4
5
6
0
5
10
15
20
c
d
h
e
0
50
100
150
ND
4
8
16
32
64
128
128
512
2,048
0
1
2
3
4
5
6
7
8
024816h HGF
p-β-Cat
(S552)
β-Cat
+ 1.8-fold
+ 3.6-fold
– 2.4-fold
p-Met
p-Akt
(S473)
Akt
Met
–
fg
i
CSC mediumMFCM
2% FCS
MUC2 mRNA
Relative expression,
FABP2 mRNA
CSC
FCS
MFCM
CSC
FCS
MFCM
MFCM
Control
1 in every
Control
MFCM
HGF (ng ml–1)
18Co
MF-49
MF-66
Control
HGF
MFCM
Control
Relative luciferase intensity
TOPFOP
Control
HGF
p-Gsk3β
(S9)
p-β-Cat
(T41/S45)
β-Cat
Gsk3β
Figure 5 Myofibroblasts support stem-cell properties and regulate Wnt
signalling. (a) Colon CSC cultures (left) can be differentiated by plating
the cells on tissue-culture-treated plastic plates in medium containing
serum (middle). Serum-induced morphological differentiation is prevented
in the presence of MFCM derived from the colonic cell line 18Co (right).
Scale bar, 20 µm. (b) Expression of differentiation genes Muc2 and FABP2
as determined by qPCR in CSCs and in cells induced to differentiate
in the absence or presence of MFCM as in a. Error bars represent s.d.
(n = 3). (c) Clonogenic potential determined by limiting-dilution assays in
the presence or absence of MFCM. Error bars represent 95% confidence
intervals. A representative example is shown. (d) Cytokine array detecting
a panel of 79 cytokines. Analysis of MFCM reveals a high level of HGF
production (red circle). (e) HGF production detected by enzyme-linked
immunosorbent assay in MFCM and in two primary cell lines (CRC-MF49
and CRC-MF66) isolated from colorectal cancer patients. Error bars
represent s.d. (n = 3). ND, none detectable. (f) Treatment of CSC cultures
with MFCM or HGF for 4 h induces nuclear translocation of β-catenin.
Scale bars, 20 µm. (g) Treatment of CSC cultures with MFCM or HGF for
16 h activates TOP–luciferase reporter activity, which is prevented by PHA
or specific anti-HGF neutralizing antibodies. Error bars represent s.e.m.
(n = 3); the experiment shown is representative of three independent
experiments. (h) Western blot of c-Met, PKB/Akt, GSK3β and T41/
S45 β-catenin with normal or phospho-specific antibodies with and
without stimulation with 50 ng ml−1 HGF for 2 h. Numbers indicate the
fold induction or decrease as determined by total and phospho-specific
signals. (i) Total amount of β-catenin and S552 phosphorylated β-catenin
over time after stimulation with HGF. Uncropped images of blots in h and i
are shown in Supplementary Information, Fig. S7.
472
nature cell biology VOLUME 12 | NUMBER 5 | MAY 2010
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 6

ARTICLES
cancer progression30. In addition, our CSC cultures expressed c-Met
(Supplementary Information, Fig. S4b), and stimulation with HGF or
MFCM rapidly activated receptor tyrosine phosphorylation (Fig. 5h;
Supplementary Information, Fig. S4c). It is not fully understood how
HGF modulates Wnt signalling activity, but we clearly observed an
increase in phosphorylation of protein kinase B (PKB)/Akt and gly-
cogen synthase kinase 3β (GSK3β), the latter being associated with
diminished GSK3β activity (Fig. 5h). Although the crosstalk between
PKB/Akt and the APC complex is a matter of debate31, in spheroidal
cultures of primary colon cancer the stimulation of hepatocyte growth
factor decreased the phosphorylation of Thr 41 and Ser 45 in β-catenin
(Fig. 5h), which are important regulatory sites for proteasomal degra-
dation of the transcription factor. It remains to be established whether
this is a direct effect, but β-catenin stability was induced by HGF (about
fourfold at 16 h; Fig. 5i). Moreover, a concomitant increase in phospho-
rylation of β-catenin on Ser 552, which is associated with enhanced
transcription32, is also detected at an early stage (Fig. 5h, i). The stimula-
tory effect of both MFCM and HGF on TCF/LEF transcriptional activ-
ity was inhibited with a specific c-Met inhibitor (PHA665752) and also
with an HGF-blocking antibody (Fig. 5g; Supplementary Information,
Fig. S5). We therefore conclude that HGF secreted by myofibroblasts
is modulating nuclear β-catenin activity through c-Met and thereby
affects CSC features in colorectal tumour cells.
Myofibroblast-secreted factors install a CSC phenotype in
differentiated cancer cells
The present observations support a model in which myofibroblasts cre-
ate what can tentatively be called a CSC niche. By analogy with nor-
mal stem cells, distance from the niche then determines the level of
stemness-stimulating factors and thereby differentiation, assuming that
CSCs follow a unidirectional differentiation pattern and cannot regain
CSC features when they have undergone differentiation. However, in
contrast to this proposition we observed that HGF-induced β-catenin
modulation is a dominant effect of myofibroblasts and can revert dif-
ferentiation. Treatment of TOP–GFPlow cells, which have low levels of
nuclear β-catenin and have lost their clonogenic potential (Figs 1c and
2a), with MFCM induced nuclear β-catenin localization as early as
4 h afterwards (Fig. 6a). More significantly, exposure of TOP–GFPlow
cells to MFCM induced the re-expression of CSC markers (Fig. 6b)
and β-catenin-dependent transcription (Fig. 6c) and even restored the
clonogenic potential of this population (Fig. 6d). This effect of HGF
is a rapid modulation of the TOP–GFPlow cells that occurs within 16 h
and enhances the clonogenicity by about 5–10-fold, almost to the level
of TOP–GFPhigh cells. This excluded the selective expansion of a small
population of CSCs remaining within the TOP–GFPlow cells because the
doubling time was more than 24 h, thus showing that myofibroblast-
secreted factors enhance Wnt signalling and can reinstall features of
stemness in more differentiated tumour cells.
The relevance of our findings for Wnt activity in vivo and CSC main-
tenance are supported by the observation that TOP–GFPhigh cells were
preferentially located close to myofibroblasts in xenografts (Fig. 7a). In
addition, in samples of primary colon cancer we observed a clear co-
localization between cells expressing α-smooth muscle actin (α-SMA)
and tumour cells displaying nuclear β-catenin (white arrows in Fig. 7b).
This observation was corroborated by co-immunofluorescence of
xenotransplanted tumours, which confirmed the correlation between
the position of myofibroblasts and the activity of Wnt signalling in
the neighbouring tumour cells (Fig. 7c, d). In addition, in primary
tumours a role for myofibroblast-derived HGF in Wnt regulation was
suggested, because we observed clear expression of HGF in tumour-
associated myofibroblasts in most cases (Supplementary Information,
Fig. S6a) and, as shown above, significant amounts of HGF produc-
tion by myofibroblasts isolated from primary colon cancer samples
(Fig. 5e; Supplementary Information, Fig. S6b,c). To establish for-
mally that myofibroblasts can reinstall the tumour-initiating capacity
in TOP–GFPlow cells, we performed co-injections. This revealed that
TOP–GFPlow cells, which have a very limited capacity to induce tumours
in comparison with TOP–GFPhigh cells, regained tumorigenicity and
thus CSC features when co-injected with myofibroblasts or with an
admixture of the factors that these cells secrete (MFCM) and Matrigel
(Fig. 7e). Our data therefore provide clear evidence of a role for myofi-
broblasts in installing and maintaining colon CSC fate through the
regulation of Wnt signalling.
DISCuSSIon
Wnt signalling is crucially important in maintaining stemness in nor-
mal colon stem cells and is a common pathway that is deregulated in
most colon cancers. Using Wnt signalling activity as a readout, we have
unravelled a fundamental characteristic of cancer stemness. In a similar
manner to normal intestinal stem cells26, Wnt activity is not merely a
ab
c
Low
Low + MFCM
High
d
1
2
4
8
16
32
64
128
256
512
1,024
0
10
20
30
0
100
200
300
TOP–GFPlow
Control
MFCM
5 h
––
16 h
5 h
––
16 h
Low
MFCM
HighLow
MFCM
High
1 in every
Low
High
Low + HGF
Low + MFCM
Low + HGF + PHA
Low + MFCM + PHA
Whole
Whole + PHA
TOP–GFP
Relative expression,
LGR5 mRNA
Relative expression,
Survivin mRNA
Figure 6 Myofibroblasts restore Wnt activity and clonogenic potential in
TOP–GFPlow cells. (a) Stimulation of TOP–GFPlow-sorted cells with MFCM for 4 h
induces nuclear translocation of β-catenin. Scale bars, 20 µm. (b) Induction
of Lgr5 and Survivin as determined by qPCR after 5 and 16 h in TOP–GFPlow
cells after treatment with MFCM. Error bars represent s.d. (n = 3). (c) TOP–GFP
expression after 16 h in TOP–GFPlow cells after treatment with MFCM. (d)
Limiting-dilution assay to determine the clonogenicity of TOP–GFPhigh, TOP–
GFPintermediate and TOP–GFPlow cells. Clonogenicity can be restored in TOP–GFPlow
cells with HGF or MFCM, whereas phytohaemagglutinin (PHA) blocks this effect
yet has no effect on normal clonogenicity. Error bars represent 95% confidence
intervals. See Methods for details on limiting-dilution statistics and scheme.
nature cell biology VOLUME 12 | NUMBER 5 | MAY 2010
473
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 7

ARTICLES
cell-intrinsic feature that can be used to define CSCs in a series of colon
tumours carrying a variety of mutations, but it is also regulated by extrin-
sic factors. More significantly, our data show that cells surrounding the
CSC not only maintain a high Wnt activity in CSCs but can also instruct
more differentiated tumour cells to activate the Wnt pathway and thereby
restore clonogenicity or tumorigenicity (summarized in Fig. 8). We have
excluded the possibility that this is due to the activation of a small subset
of residual CSCs in the TOP–GFPlow population: the effect is too rapid
and occurs well before any cell expansion has taken place. However, to
what extent these cells possess unlimited replicative potential, which is
a crucial hallmark in the definition of a CSC, remains to be elucidated
by repetitive isolations and subsequent injections.
Nonetheless, the potential of these cells to proliferate extensively and
form tumours that contain a similar Wnt signalling heterogeneity to
that observed with TOP–GFPhigh-induced tumours warrants the con-
clusion that the microenvironment is central to the growth of colon
cancers under these conditions. Previous observations by us and others
have suggested that brain CSCs are also supported by (perivascular)
endothelial cells33,34. Although those studies did not reveal whether
differentiated tumour cells can regain stemness, they may point to a
similar relation between CSCs and microenvironment and suggest that
the CSC phenotype may be a much more dynamic characteristic than
has previously been assumed.
In this light we note that there is an ongoing debate in the field on the
existence and incidence of CSCs within different tumours. This has in
part been explained by immune responses observed in xenotransplanta-
tion assays35–37 and by differences between tumours. Our present obser-
vations indicate that the usage of fully immune-deficient mice does
not annihilate the difference between TOP–GFPhigh and TOP–GFPlow
cells, proving that cancer stemness is not due to a different sensitivity
to immune control. However, we feel that our data, by implicating the
microenvironment as a dominant factor in CSC, can potentially shed
new light on this controversy. One could envisage that tumours such as
late-stage melanoma have become independent of the signals elicited
by the microenvironment and therefore do not show a clear-cut CSC
organization37. Alternatively, a subset of tumours could be more effec-
tive in modulating the microenvironment or could simply be more
receptive to the local environment at the site of injection. In this light
we note that our observations indicate a rapid attraction or formation
of myofibroblasts at the site of tumour formation. Combined with the
observation that differentiated tumour cells induce tumours when
co-injected with such myofibroblasts, this suggests that differentiated
tumour cells simply fail to create the right environment and therefore
fail to show tumour-initiating capacity. This conclusion would have
far-reaching consequences when discussing tumour therapy. As CSC
have been shown to be relatively resistant to therapy14,38–40, present strat-
egies are aimed at developing CSC selective therapies, thereby attacking
the tumour at its root. However, this approach would be frustrated by
plasticity of the system, which would simply lead to CSC regeneration
from the differentiated tumour compartment. In contrast, the model
we propose indicates that attention should be directed to the potential
opportunity offered by the interface of the microenvironment and the
CSC as a therapeutic target.
It has previously been shown that nuclear β-catenin localization was
predominantly observed in the invasive regions of colon carcinomas, that
is, at the leading edges10. These observations concur with our hypothesis
that myofibroblasts, which are present at high density at the tumour front,
could direct Wnt activation. This suggests a framework in which stro-
mal components, in this case HGF-producing myofibroblasts, stimulate
CSC features of cancer cell populations mainly at the tumour edges and
simultaneously promote the invasion and spread of the malignancy into
the surrounding tissue, as has also been suggested previously41. The fact
e
d
0
250
500
750
1,000
c
b
a
α-SMA
TOP–GFP
α-SMA
α-SMA–
α-SMA+
TOP–GFP
β-Catenin
α-SMA
β-Catenin
α-SMA
β-Catenin
α-SMA
β-Catenin
α-SMA
TOP–GFP
α-SMA
GFP intensity
Cells Injected
100
Line
C100.B5
C100.G7
–
–
–
–
–
+ MFCM
+ MFCM
+ MF
Condition 10
0/6
3/6
0/6
0/6
0/6
0/6
0/6
0/6
0/6
5/6
0/6
0/6
2/6
1/6
4/6
0/6
1/6
6/6
0/6
2/6
3/6
4/6
3/6
5/6
1,000
–
–
–
3/6
4/6
6/6
0/6
6/6
5,000
Fraction
TOP–GFPlow
TOP–GFPhigh
TOP–GFPlow
TOP–GFPlow
TOP–GFPlow
TOP–GFPhigh
TOP–GFPlow
TOP–GFPintermediate
Figure 7 Myofibroblasts restore tumorigenicity in TOP–GFPlow Wnt-active
cells and co-localize with highly Wnt-active cells in vivo. (a) Myofibroblasts
are present in the stroma of TOP–GFP xenografts as demonstrated by
α-SMA-positive cells and localize close to cells expressing TOP–GFP as
detected by α-GFP immunohistochemistry in consecutive slides. Scale
bars, 50 µm. (b) In human colorectal cancer specimens, cells positive
for nuclear β-catenin (brown) show an intimate relationship with α-SMA-
positive myofibroblasts (purple). The right panels show enlargements;
white and black arrowheads indicate tumour cells with and without
nuclear β-catenin, respectively. (c) Immunofluorescence co-staining for
α-SMA and GFP indicates a close relationship between TOP–GFPhigh cells
and myofibroblasts. Scale bar, 50 µm. (d) Quantification of TOP–GFP
intensity in areas of epithelial cells that are next to α-SMA cells reveals
a relationship between myofibroblasts and enhanced Wnt signalling
(Student’s t-test, P < 0.001; see Methods for details). (e) In vivo transfer
of TOP–GFP-transduced cultures. Cell numbers from the indicated
populations were injected into nude mice after FACS. The number of
successful tumour initiations after nine weeks out of six injections for
each condition is shown. Injection of TOP–GFPlow cells with myofibroblasts
(MF) or after treatment with MFCM shows enhanced tumour growth in
comparison with TOP–GFPlow cells alone. The upper part of the table is a
partial representation of Fig. 3a.
474
nature cell biology VOLUME 12 | NUMBER 5 | MAY 2010
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 8

ARTICLES
that HGF/c-Met overexpression is correlated with colon cancer progres-
sion only strengthens this notion30,42. This is not merely due to enhanced
scattering of tumour cells by HGF, because c-Met silencing also induces
almost complete regression of already established metastases in an experi-
mental model43. This indicates a constitutive need for c-Met-mediated
signals in the growth or maintenance of metastases and is consistent with
a role for HGF in CSC maintenance.
The tight relationship between CSC properties, invasion and
microenvironment is corroborated by findings that suggest that the
CSC phenotype, epithelial–mesenchymal transition (EMT), invasion
and microenvironmental interactions are closely associated in breast
cancer44. In addition, plasticity takes a central position here because
activation of the EMT-promoting factors Snail or Twist generates
CSCs from more differentiated tumour cells, which revert when these
EMT-promoting signals are not maintained44. Intriguingly, the original
studies on MDCK cell scattering, which spurred the interest in EMT45,
in essence showed that HGF was a potent EMT inducer that was in
later work related to the induction of Snail in epithelial cancer cells46.
Moreover, in normal development Wnt signals are crucially tethered to
the induction of EMT47. Taken together, these observations therefore
point to the hypothesis that colon cancer stemness is partly defined by
environmental cues, such as HGF, and can be induced in more differ-
entiated tumour cells by the microenvironment and may therefore be
closely related to EMT. This has important consequences for the way
in which we perceive CSCs as critical therapeutic targets, shifting the
attention from the CSC to its environment.
MethoDS
Methods and any associated references are available in the online version
of the paper at http://www.nature.com/naturecellbiology/
Note: Supplementary Information is available on the Nature Cell Biology website.
ACKnowLEDGEMEnTS
The authors are indebted to the work performed by Willem Bemelman and Thomas
van Gulik, surgeons at the AMC without whom the studies described here would
not have been possible. In addition, we thank Riccardo Fodde for constructive
discussions on CSCs and Wnt, and the animal care takers for their work for this
project. Finally, we thank Berend Hooibrink and Toni van Capel for assistance
with fluorescence-activated cell sorting experiments. This work was supported
by a VICI grant from the Netherlands Organisation for Scientific Research and
a Dutch Cancer Society (KWF Kankerbestrijding) grant (2009-4416) (to J.P.M.),
an Academisch Medisch Centrum (AMC) fellowship (to L.V. and F.d.S.M.), the
AMC Graduate School (to K.K.) and by the Associazione Italiana per la Ricerca sul
Cancro (AIRC) (to G.S. and M.T.).
AuTHoR ConTRiBuTionS
L.V., F.d.S.M., M.v.d.H., K.C., J.H.d.J., T.B., J.B.T., H.R., M.R.S., K.K. designed and
conducted experiments. M.T. and C.M. isolated and cultured CSC lines, D.J.R., G.S.
and J.P.M. planned and supervised the experiments. L.V., F.d.S.M and J.P.M wrote
the manuscript.
CoMPETinG FinAnCiAL inTERESTS
The authors declare no competing financial interests.
Published online at http://www.nature.com/naturecellbiology
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Fearon, E. R. & Vogelstein, B. A genetic model for colorectal tumorigenesis. Cell 61,
759–767 (1990).
2. Clevers, H. Wnt/β-catenin signaling in development and disease. Cell 127, 469–480
(2006).
3. Bienz, M. & Clevers, H. Linking colorectal cancer to Wnt signaling. Cell 103, 311–320
(2000).
4. Vogelstein, B. et al. Genetic alterations during colorectal-tumor development. N. Engl.
J. Med. 319, 525–532 (1988).
5. Tetsu, O. & McCormick, F. β-Catenin regulates expression of cyclin D1 in colon carci-
noma cells. Nature 398, 422–426 (1999).
6. He, T. C. et al. Identification of c-MYC as a target of the APC pathway. Science 281,
1509–1512 (1998).
7. Korinek, V. et al. Depletion of epithelial stem-cell compartments in the small intestine
of mice lacking Tcf-4. Nature Genet. 19, 379–383 (1998).
8. Groden, J. et al. Response of colon cancer cell lines to the introduction of APC, a
colon-specific tumor suppressor gene. Cancer Res. 55, 1531–1539 (1995).
9. van de Wetering, M. et al. The β-catenin/TCF-4 complex imposes a crypt progenitor
phenotype on colorectal cancer cells. Cell 111, 241–250 (2002).
10. Brabletz, T. et al. Variable β-catenin expression in colorectal cancers indicates tumor
progression driven by the tumor environment. Proc. Natl Acad. Sci. USA 98, 10356–
10361 (2001).
11. Fodde, R. & Brabletz, T. Wnt/β-catenin signaling in cancer stemness and malignant
behavior. Curr. Opin. Cell Biol. 19, 150–158 (2007).
12. Vermeulen, L., Sprick, M. R., Kemper, K., Stassi, G. & Medema, J. P. Cancer stem
cells—old concepts, new insights. Cell Death. Differ. 15, 947–958 (2008).
13. O’Brien, C. A., Pollett, A., Gallinger, S. & Dick, J. E. A human colon cancer cell
capable of initiating tumour growth in immunodeficient mice. Nature 445, 106–110
(2007).
14. Todaro, M. et al. Colon cancer stem cells dictate tumor growth and resist cell death by
production of interleukin-4. Cell Stem Cell 1, 389–402 (2007).
15. Dalerba, P. et al. Phenotypic characterization of human colorectal cancer stem cells.
Proc. Natl Acad. Sci. USA 104, 10158–10163 (2007).
16. Ricci-Vitiani, L. et al. Identification and expansion of human colon-cancer-initiating
cells. Nature 445, 111–115 (2007).
17. Vermeulen, L. et al. Single-cell cloning of colon cancer stem cells reveals a multi-
lineage differentiation capacity. Proc. Natl Acad. Sci. USA 105, 13427–13432
(2008).
18. Reya, T. et al. A role for Wnt signalling in self-renewal of haematopoietic stem cells.
Nature 423, 409–414 (2003).
19. Barker, N. et al. Identification of stem cells in small intestine and colon by marker gene
Lgr5. Nature 449, 1003–1007 (2007).
20. van der Flier, L. G. et al. Transcription factor achaete scute-like 2 controls intestinal
stem cell fate. Cell 136, 903–912 (2009).
21. Yeung, T. M., Gandhi, S. C., Wilding, J. L., Muschel, R. & Bodmer, W. F. Cancer stem
cells from colorectal cancer-derived cell lines. Proc. Natl Acad. Sci. USA 107, 3722–
3727 (2010).
22. Kai, K. et al. Maintenance of HCT116 colon cancer cell line conforms to a stochastic
model but not a cancer stem cell model. Cancer Sci. 100, 2275–2282 (2009).
b
a
Wnt
CSCs Differentiation
+MFCM/+HGF
Figure 8 Schematic representation of the proposed model. (a) Cells with
a high Wnt signal activity (depicted in green) reside close to stromal
myofibroblasts (red) in colorectal cancer tissue. Cells with a relatively
low Wnt activity are depicted in grey. Tumour-associated myofibroblasts
secrete factors such as HGF (black dots) that stimulate Wnt signalling.
(b) Stimulation of the Wnt signalling cascade in differentiated colon
cancer cells, characterized by low Wnt signalling activity, restores CSC
characteristics. These include expression of CSC markers and tumorigenicity.
Taken together, these results suggest a pivotal role for the environment in
determining CSC features of colon cancer cells.
nature cell biology VOLUME 12 | NUMBER 5 | MAY 2010
475
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 9

ARTICLES
23. Clarke, M. F. et al. Cancer stem cells—perspectives on current status and future direc-
tions: AACR Workshop on cancer stem cells. Cancer Res. 66, 9339–9344 (2006).
24. Karnoub, A. E. et al. Mesenchymal stem cells within tumour stroma promote breast
cancer metastasis. Nature 449, 557–563 (2007).
25. Sato, T. et al. Single Lgr5 stem cells build crypt–villus structures in vitro without a
mesenchymal niche. Nature 459, 262–265 (2009).
26. Yen, T. H. & Wright, N. A. The gastrointestinal tract stem cell niche. Stem Cell Rev. 2,
203–212 (2006).
27. Boon, E. M., van der, N. R., van de, W. M., Clevers, H. & Pals, S. T. Wnt signaling
regulates expression of the receptor tyrosine kinase met in colorectal cancer. Cancer
Res. 62, 5126–5128 (2002).
28. Rasola, A. et al. A positive feedback loop between hepatocyte growth factor receptor and
beta-catenin sustains colorectal cancer cell invasive growth. Oncogene 26, 1078–1087
(2007).
29. Brembeck, F. H. et al. Essential role of BCL9-2 in the switch between β-catenin’s
adhesive and transcriptional functions. Genes Dev. 18, 2225–2230 (2004).
30. Di Renzo, M. F. et al. Overexpression and amplification of the Met/HGF receptor gene
during the progression of colorectal cancer. Clin. Cancer Res. 1, 147–154 (1995).
31. Ng, S. S. et al. Phosphatidylinositol 3-kinase signaling does not activate the Wnt
cascade. J. Biol. Chem. 284, 35308–35313 (2009).
32. Fang, D. et al. Phosphorylation of β-catenin by AKT promotes β-catenin transcriptional
activity. J. Biol. Chem. 282, 11221–11229 (2007).
33. Borovski, T. et al. Tumor microvasculature supports proliferation and expansion of
glioma-propagating cells. Int. J. Cancer 125, 1222–1230 (2009).
34. Calabrese, C. et al. A perivascular niche for brain tumor stem cells. Cancer Cell 11,
69–82 (2007).
35. Shackleton, M., Quintana, E., Fearon, E. R. & Morrison, S. J. Heterogeneity in cancer:
cancer stem cells versus clonal evolution. Cell 138, 822–829 (2009).
36. Kelly, P. N., Dakic, A., Adams, J. M., Nutt, S. L. & Strasser, A. Tumor growth need not
be driven by rare cancer stem cells. Science 317, 337 (2007).
37. Quintana, E. et al. Efficient tumour formation by single human melanoma cells. Nature
456, 593–598 (2008).
38. Guan, Y., Gerhard, B. & Hogge, D. E. Detection, isolation, and stimulation of quiescent
primitive leukemic progenitor cells from patients with acute myeloid leukemia (AML).
Blood 101, 3142–3149 (2003).
39. Guzman, M. L. et al. Preferential induction of apoptosis for primary human leukemic
stem cells. Proc. Natl Acad. Sci. USA 99, 16220–16225 (2002).
40. Bao, S. et al. Glioma stem cells promote radioresistance by preferential activation of
the DNA damage response. Nature 444, 756–760 (2006).
41. Brabletz, T., Jung, A., Spaderna, S., Hlubek, F. & Kirchner, T. Opinion: migrating cancer
stem cells—an integrated concept of malignant tumour progression. Nature Rev. Cancer
5, 744–749 (2005).
42. Kammula, U. S. et al. Molecular co-expression of the c-Met oncogene and hepatocyte
growth factor in primary colon cancer predicts tumor stage and clinical outcome. Cancer
Lett. 248, 219–228 (2007).
43. Corso, S. et al. Silencing the MET oncogene leads to regression of experimental tumors
and metastases. Oncogene 27, 684–693 (2008).
44. Mani, S. A. et al. The epithelial–mesenchymal transition generates cells with properties
of stem cells. Cell 133, 704–715 (2008).
45. Stoker, M. & Perryman, M. An epithelial scatter factor released by embryo fibroblasts.
J. Cell Sci. 77, 209–223 (1985).
46. Grotegut, S., von, S. D., Christofori, G. & Lehembre, F. Hepatocyte growth factor induces
cell scattering through MAPK/Egr-1-mediated upregulation of Snail. EMBO J. 25,
3534–3545 (2006).
47. Moustakas, A. & Heldin, C. H. Signaling networks guiding epithelial–mesenchymal transi-
tions during embryogenesis and cancer progression. Cancer Sci. 98, 1512–1520 (2007).
476
nature cell biology VOLUME 12 | NUMBER 5 | MAY 2010
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 10

DoI: 10.1038/ncb2048
METHODS
MethoDS
Cell culture. CSCs were isolated from different patients in accordance with the
rules of the medical ethical committee of the AMC and University of Palermo and
Heidelberg, and named after their site of origin. Primary samples were named Co
and liver metastases were named LM, both followed by a number representing
the isolation. CSCs were cultured in modified neurobasal A medium containing
N2 supplement (Invitrogen), Lipid Mixture-1 (Sigma), basic fibroblast growth
factor (bFGF; 20 ng ml−1) and epidermal growth factor (EGF; 50 ng ml−1). Colon
cancer cell lines were cultured in Iscove’s modified defined medium (IMDM)
plus 10% FCS, glutamine and penicillin/streptomycin. 18Co cells were purchased
from the American Type Culture Collection and maintained in DMEM medium
supplemented with 10% FCS and 1% glutamine.
Colon CSC cultures were derived as described previously14. In brief, primary
resected human colon carcinomas were digested enzymatically for 1 h with a
mixture of collagenase (1.5 mg ml−1; Roche) and hyaluronidase (20 µg ml−1)
at 37 °C and shaken repeatedly. The dissociated sample was then filtered
(40 µm pore size) and washed in CSC medium. Erythrocytes and cell debris
were removed by Lympholyte (M; Cedarlane) centrifugation. Cells were then
washed with CSC medium and subsequently cultured. For isolation of primary
myofibroblasts, colorectal specimens were first cut into small pieces and incu-
bated for 30 min with 1 mM EDTA (Sigma) at 37 °C with repetitive shaking to
remove epithelial cells. Samples were washed with Hanks balanced salt solution
between each incubation and then digested enzymatically for 30 min with col-
lagenase (1.5 mg ml−1; Roche) at 37 °C with shaking. Cells were then washed
and plated at high density with DMEM supplemented with 10% FCS and 1%
non-essential amino acids.
Lentiviral and luciferase reporter assays. TCF/LEF reporter driving expression
of GFP (TOP–GFP) was a gift from Laurie Ailles and was described previously18.
Spheroidal cultures were transduced lentivirally with either the TOP–GFP or con-
stitutive CMV–GFP or PGK–GFP constructs. Generation of single-cell-derived
cultures was performed by FACSaria (BD Biosciences) single-cell plating in
96-well ultralow-adhesion plates (Corning) containing stem-cell medium. We
stringently gated on single, propidium iodide-negative/GFP-positive cells. After
visible spheres arose, they were transferred to ultralow-adhesion flasks (Corning)
and expanded. Luciferase assays were performed as recommended by the manu-
facturer (SA Biosciences). Cells were transiently transfected (FuGENE 6; Roche)
with a mixture of inducible TCF/LEF-responsive firefly luciferase and constitu-
tively expressed Renilla luciferase (40:1), or with a negative control containing a
mixture of non-inducible firefly luciferase and constitutively expressed Renilla
luciferase (40:1). All experiments were performed in triplicate. Cells were lysed
in luciferase reporter lysis buffer and monitored for luciferase and Renilla activity
with a Dual-Luciferase Reporter Assay System (Promega). Cells were starved for
12 h and stimulated for 12–16 h with indicated factors.
Limiting-dilution assay. Cells from different GFP intensities were deposited
at 1, 2, 4, 6, 8, 12, 16, 20 and 24 cells per well. Clonal frequency and statistical
significance were evaluated with the Extreme Limiting Dilution Analysis (ELDA)
‘limdil’ function (http://bioinf.wehi.edu.au/software/elda/index.html).
Immunohistochemistry and western blot antibodies and reagents.
Immunofluorescence and/or immunohistochemistry were performed on paraf-
fin-embedded sections or on cytospins as described17. The following antibodies
were used for immunohistochemistry and immunofluorescence: anti-β-catenin
(Cell Signaling Technology, 1:100 dilution; Transduction Labs, 1:250), anti-GFP
(Roche, 1:500), anti-α-SMA (Abcam, 1:100), anti-Vimentin (Abcam, 1:100), anti-
desmin (Neomarkers, 1:100), anti-HGF (R&D Systems, 1:100), anti-cytokeratin 20
(Genetex, 1:250), anti-FAPB2 (Abcam, 1:200) and anti-Muc2 (Abcam, 1:200). The
following antibodies were used for western blotting: anti-β-actin (Sigma, 1:1,000),
anti-phospho-Met (Tyr 1234/1235, 1:500), anti-Met (25H2, 1:500), anti-phospho-
Akt (Ser 473, 1:500), anti-Akt (1:500), anti-phospho-Gsk3(Ser 9, 1:500), anti-
Gsk3 (1:1,000), anti-phospho-β-catenin (Thr 41/Ser 45, 1:500; Ser 552, 1:500)
and anti-β-catenin (1:1,000); all except the anti-β-actin were purchased from Cell
Signaling Technology. Alcian Blue staining was performed with Alcian Blue 8GX
(Sigma). Haematoxylin/eosin staining was performed with Ehrlich haematoxylin/
eosin solution (Sigma). Western blotting was performed as described48. Full blots
are shown in Supplementary Information, Fig. S7.
Conditioned medium, enzyme-linked immunosorbent assay (ELISA) and
cytokine array, cytokines and inhibitors. A total of 7.5 × 105 18Co cells
were seeded in 75-cm2 flasks. On the next day, cells were washed twice with
PBS and incubated for 24 h with 10 ml of CSC medium without EGF and
bFGF. MFCM was then collected and cleared by centrifugation and used at
1:2 dilution.
A cytokine/chemokine array kit (Ray Biotech Inc.) was used to detect a panel
of 79 secreted cytokines and chemokines in MFCM. The manufacturer’s recom-
mended protocol was used. Human recombinant HGF (50 ng ml−1) was from Relia
Tech. Inc. A novel small molecule inhibiting c-Met phosphorylation (PHA665752)
was provided by Pfizer. A neutralizing HGF antibody (R&D Systems) was used
to deplete HGF from the MFCM. In brief, MFCM was incubated for 1 h at 37 °C
with the neutralizing antibody (10 µg ml−1) before use in the stimulation assay.
HGF secretion was quantified by quantitative ELISA according to manufacturer’s
instructions (R&D).
Flow cytometry. Flow cytometry was performed on trypsin-dissociated TOP–
GFP CSC cultures with AC133 (Miltenyi Biotec, 1:100), CD44 (BD Biosciences,
1:100), CD166 (R&D Systems, clone 105901, 1:100), CD24 (BD Biosciences,
1:100), CD29 (BD Biosciences, 1:100) and c-Met (Upstate, 1:100). Dead cells
were excluded with propidium iodide.
RNA extraction, microarray and PCR. Total RNA from cells comprising the
lowest and highest 10% TOP–GFP of spheroidal SCD cultures was extracted with
Trizol reagent (Invitrogen) in accordance with the manufacturer’s protocol. RNA
concentration was determined with NanoDrop ND-1000, and quality was deter-
mined using the RNA 6000 Nano assay on the Agilent 2100 Bioanalyzer (Agilent
Technologies). Affymetrix microarray analysis, fragmentation of RNA, label-
ling, hybridization to Human Genome U133 Plus 2.0 microarrays, and scanning
were performed in accordance with the manufacturer’s protocol (Affymetrix).
Microarray data can be viewed online (http://www.ncbi.nlm.nih.gov/geo/index.
html) under GEO accession number GSE17375. Real time RT–PCR was per-
formed with SYBR green (Abgene) in accordance with the manufacturer’s instruc-
tions on a Bio-Rad MyiQ Thermal cycler. For primers used see Supplementary
Information, Table S2.
In vivo tumour propagation. Mice experiments were performed in accordance
with the ethical committee of the AMC. For transplantation of cancer cells, 30
spheres (about 100 cells per sphere) suspended in 100 µl of PBS/BSA admixed with
Matrigel at a 1:1 ratio were injected subcutaneously into nude mice (Hsd:Athymic
Nude/Nude) (Harlan). After 3–8 weeks visible tumours arose. When the tumour
size reached 1 cm3, mice were killed and tumours were processed either for anal-
ysis or for culture in vitro. For in vivo limiting-dilution injection, TOP–GFP-
transduced cultures or established xenografts from the TOP–GFP cultures were
dissociated and 1, 10, 100, 1,000 and 5,000 cells from the 10% lowest, 10% highest
or total TOP–GFP intensities were deposited, by FACS, in a 96-well plate contain-
ing CSC medium, admixed with Matrigel, and injected as described above. For
myofibroblast co-injection experiments 50,000 18Co cells were plated in 96-well
plates; TOP–GFPlow cells were added by FACS deposition in the indicated amounts
and injected as described. For MFCM stimulation of TOP–GFPlow cells, cells were
deposited by FACS in MFCM and incubated at 37 °C for 2 h; afterwards cells and
MFCM were admixed 1:1 with Matrigel and injected. For single-cell injection
single cells were deposited from the 0.5% highest TOP–GFP cells, and single-cell
deposition was confirmed microscopically.
Co-immunostaining and quantification. Paraffin-embedded xenografts from
TOP–GFP cultures were co-stained with anti-GFP and anti-α-SMA antibody.
GFP intensities of epithelial cell regions in the direct presence or not of α-SMA-
positive cells were quantified with an inverted fluorescence microscope (Zeiss),
using the Axiovision software. Serial sections (n = 45) and multiple fields per
section were scored. Student’s t-test was used for statistical significance. Paraffin-
embedded primary human specimens were co-stained with α-SMA (Abcam) and
anti-β-catenin (Transduction Labs) and then incubated with anti-rabbit-AP/anti-
mouse-HRP (Powervison) (1:1). RED Alkaline Phosphatase substrate (Vector)
followed by DAB+ (Dako) were applied to the slides.
48. Tuynman, J. B. et al. Cyclooxygenase-2 inhibition inhibits c-Met kinase activity and
Wnt activity in colon cancer. Cancer Res. 68, 1213–1220 (2008).
nature cell biology
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 11

supplementary information
www.nature.com/naturecellbiology
1
DOI: 10.1038/ncb2048
Figure S1 TOP-GFP expression and relation with nuclear β-catenin, wnt
targets and differentiation markers. (a) Co-immunostaining of TOP-GFP
spheroid culture with GFP and β-catenin antibody demonstrates a clear
correlation between TOP-GFP levels and nuclear localization of β-catenin
(scale bar, 50µm), quantification shown in (b). (c) Microarray analysis
of a specific gene-set in high, intermediate and low TOP-GFP cell
fractions indicates gradual increase in differentiation marker expression
and a decrease in Wnt target gene expression from TOP-GFPhigh to
TOP-GFPlow populations (Each box plot represents a minimal of two
data points from separate single-cell cloned TOP-GFP CSC cultures).
Genes were picked based on significant differences observed in
Fig 1d.
c
a
MUC2
TOP-GFP high
TOP-GFP int.
TOP-GFP low
0
200
400
600
800
Expression
FABP2
TOP-GFP high
TOP-GFP int.
TOP-GFP low
0
5
10
15
20
25
Expression
KRT20
TOP-GFP high
TOP-GFP int.
TOP-GFP low
0
1000
2000
3000
Expression
LGR5
TOP-GFP high
TOP-GFP int.
TOP-GFP low
0
500
1000
1500
Expression
ASCL2
TOP-GFP high
TOP-GFP int.
TOP-GFP low
0
500
1000
1500
2000
Expression
AXIN2
TOP-GFP high
TOP-GFP int.
TOP-GFP low
0
500
1000
1500
2000
Expression
!-catenin
Top-GFP
b
1 2 3 4 5 6 7 8
Nuclear Beta-Catenin
0
1
2
3
GFP Intensity
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 12

supplementary information
2
www.nature.com/naturecellbiology
Figure S2 Limiting dilution on various clones, Ki-67 staining. (a) Limiting dilution assay and clonogenic potential of 3 independent lines derived from the same
patient as Apo.1 (Apo.1.A, Apo.1.B and Apo.1.C). Errors bars represent 95% CI. Representative examples are shown. See Methods for details on limiting dilution
assays. (b) Ki-67 co-staining with GFP in TOP-GFP xenografts. Ki-67 positivity encompasses both GFP positive and negative cells. Scale bar, 100µm.
a
*
*
Top-GFP
ki67
b
Apo.1.A
2
4
8
16
32
64
128
1 in every
TOP-GFP high
TOP-GFP Low
Apo.1.B
Apo.1.C
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 13

supplementary information
www.nature.com/naturecellbiology
3
Figure S3 Myofibroblasts prevent differentiation of colon CSCs. (a)
Immunohistochemistry for α-SMA shows myofibroblasts in the stroma
of primary human colorectal malignancies. (Scale bar, 50µm) (b) Phase
contrast pictures to show morphological differentiation of CSC. Upper
left represents a spheroid culture growing in medium containing EGF
and bFGF, upper right is after differentiation in 2% FCS. Lower left is
differentiation with 2% FCS, but plated on myofibroblasts (18Co) and
lower right is differentiation with FCS in the presence of MFCM. (Scale
bar, 20µm) (c) Immunofluorescence for FABP2 and Muc2 on cytospins
of spheroid cells (EGF/bFGF) or cells induced to differentiate with 2%
FCS in the absence (middle) or presence of MFCM (right). (Scale bar,
20µm)
a
"-SMA
FABP2
GFFCSMFCM
Muc2
c
b
EGF/bFGFFCS
FCS + 18Co
FCS + MFCM
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 14

supplementary information
4
www.nature.com/naturecellbiology
Figure S4 Myofibroblasts produce HGF and human colon CSCs express c-Met.
(a) A graph depicting the detected secreted factors in MFCM (see Methods for
details). (b) PCR showing expression of c-Met in spheroid cultured colon CSCs.
Right panel shows FACS analysis for c-Met. (red; background and blue; c-Met)
(c) Full blot of phospho-c-Met and β-actin of Co200 stimulated with 18Co
conditioned medium (MFCM) or CSC control medium for the time indicated.
b
gapdh
c-met
Co100 Co200
0 102
103
APC-A
104
105
0
20
40
60
80
100
% of Max
c-met
a
c
0’5’10’
30’
5’
CSC med
10’30’
MFCM
p-met
50
75
100
150
!-actin
75
100
50
37
25
!-actin
75
100
50
37
25
HGFMFCM
control10’
30’
+ PHA
10’30’
+ PHA
p-met
50
75
100
150
HGF
MCP-1
TIMP-1TIMP-2
GRO
IGFBP-1
Osteoprotegerin
Eotaxin
IL-8
GRO-alpha
Angiogenin
PARC
IGFBP-4
MIP-3-alpha
IL-7
MCP-2
IL-6
IL-3
0
1000
2000
3000
control
MFCM
Relative spot intensity
© 2010 Macmillan Publishers Limited. All rights reserved.

Page 15

supplementary information
www.nature.com/naturecellbiology
5
Figure S5 TOP/FOP assay on various lines and with various conditions.
Depicted are the results of TOP/FOP assays on different human colon cancer
lines including several established colorectal cancer lines (DLD1, HCT116,
HT29). LM5 is a liver metastasis derived primary CSC line. The stimulatory
effect of MFCM is HGF dependent. Error bars represent SEM (n=3), data
from at least 2 replicates is shown.
DLD1HCT116
TOP FOP
0
1
2
3
4
Control
HGF
MFCM
HGF + PHA
MFCM + PHA
PHA
Relative Luciferase
Intensity
TOPFOP
0
2
4
6
Control
HGF
MFCM
HGF + PHA
MFCM + PHA
PHA
Relative Luciferase
Intensity
HT29
TOPFOP
0
1
2
3
4
Control
HGF
MFCM
HGF + PHA
MFCM + PHA
PHA
Relative Luciferase
Intensity
LM5
TOPFOP
0
1
2
3
4
Control
HGF
MFCM
MFCM + PHA
MFCM + Antibody
Relative Luciferase
Intensity
© 2010 Macmillan Publishers Limited. All rights reserved.