BMP Signaling Promotes the Growth of Primary
Human Colon Carcinomas in vivo
Aiala Lorente-Trigos, Fre ´de ´ric Varnat, Alice Melotti, and Ariel Ruiz i Altaba*
Department of Genetic Medicine and Development, University of Geneva School of Medicine, 8242 CMU, 1 rue Michel Servet, CH-1211 Geneva, Switzerland
* Correspondence to: Ariel Ruiz i Altaba, Tel: +41-22-379-5646; Fax: +41-22-379-5962; E-mail: firstname.lastname@example.org
Human colon carcinomas (CCs) represent a growing worldwide problem. One of the pathways that has been negatively implicated in
the genesis of CCs is triggered by bone morphogenetic protein (BMP) ligands, which activate BMP receptors leading to the function
of SMAD proteins in the nucleus. BMP signaling is alteredin familial human polyposis, and micewith compromised BMP signaling in
the intestine develop tumors. Here, we have re-evaluated the presence and roles of BMP signaling in advanced sporadic human CCs,
using both primary tumors and established cell lines, and directly modulating BMP pathway activity in a cell-autonomous manner
using constitutively active and dominant-negative BMP receptor Ib forms. We find evidence for active endogenous BMP signaling in
all primary CC samples and for its role in promoting primary CC tumor growth and CC cell survival and proliferation in vivo in xeno-
grafts. In vitro, we also document autonomous and non-autonomous effects of enhanced BMP receptor activity on gap closure in
culture, suggesting possible roles in invasion. Caution should thus be exerted in trying to augment or restore its activity for thera-
peutic purposes. In contrast, we raise the possibility that blockade of BMP signaling might have beneficial effects against at least a
subset of advanced colon cancers.
Keywords: BMP, colon cancer, GLI1, human
Colon cancer incidence is growing around the globe (American
Cancer Society, Cancer Research UK; Jemal et al., 2009).
Although there is efficient treatment for early and localized
disease involving surgery, a significant number of patients pre-
senting advanced localized colon carcinomas (CCs) already have
or will develop liver metastases, and metastatic CCs remain
largely incurable. There is thus a pressing need to re-evaluate
current ideas and to find novel ways to treat CCs and metastatic
Severaltranscription and secreted factors havebeenimplicated
in the genesis of intestinal tumors in humans and mice. For
instance, activation of the WNT–TCF signaling pathway is essen-
tial for the formation of human and mouse intestinal adenomas
and most CCs harbor an activated pathway through loss of APC,
which normally acts to promote b-catenin degradation, or
through direct constitutive activation of b-catenin (Cho and
Vogelstein, 1992; Powell et al., 1992; Morin et al., 1997; Sparks
et al., 1998). Other pathways cooperate or synergize with WNT
signaling. For example, the NOTCH pathway is implicated in the
formation of adenomas in mice (van Es et al., 2005; Fre et al.,
2009), and HEDGEHOG-GLI signaling is essential in advanced
human tumors, including liver metastases (Varnat et al., 2009),
acting in parallel with WNT signaling in mice (Arimura et al.,
2009; Varnat et al., 2010a).
An additional pathway implicated in intestinal tumor formation
is that triggered by the secreted bone morphogenetic protein
(BMP) ligands. The numerous BMPs are members of the trans-
forming growth factor-beta (TGF-b) superfamily. It is estimated
that ?80% of human colorectal cancers harbor mutations affect-
ing at least one component of TGF-b superfamily signaling (Grady
et al., 1999; Blobe et al., 2000). Moreover, there is a tight link
between loss of BMPR1a or SMAD4 and juvenile polyposis, a syn-
drome characterized by the early appearance of multiple precan-
cerous masses in the intestine with increased risk to develop
colon cancers (Howe et al., 1998; Zhou et al., 2001).
BMP signaling involves a complex pathway with multiple
levels of regulation. In the core pathway, BMP ligands bind
BMPRII and BMPRI transmembrane tyrosine kinase receptors
promoting their heterodimerization and the phosphorylation of
BMPRI by BMPRII. Phosphorylated BMPRI receptors then trans-
duce the BMP signal intracellularly by phosphorylating SMAD1,
5 and 8, and it is the action of such phospho-SMADs, in
cooperation with SMAD4, that activates BMP target genes
(Lagna et al., 1996; Massague, 1996; Kretzschmar et al., 1997;
Lo ´pez-Rovira et al., 2002). However, the phenotype of SMAD4
KO mice raises the possibility of SMAD4-independent BMP func-
tions (Sirard et al., 1998; Derynck and Zhang, 2003), suggesting
September 21, 2010.
#The Author (2010). Published by Oxford University Press on behalf of Journal of
Molecular Cell Biology, IBCB, SIBS, CAS. All rights reserved.
August30, 2010. RevisedSeptember 14, 2010. Accepted
Journal of Molecular Cell Biology (2010), 2, 318–332 doi:10.1093/jmcb/mjq035
that the presence or absence of SMAD4 might have functional
The results summarized above support the widespread and
accepted notion that BMP signaling acts as a tumor suppressor
for sporadic human CCs (Hardwick et al., 2008). Indeed, immuno-
chemical analyses of BMP pathway components have suggested
the loss of BMP signaling in CCs (Kodach et al., 2008a, b).
BMP3 inactivation is a frequent event in CCs and its addition
decreases CC colony formation in vitro (Loh et al., 2008).
Moreover, exogenous BMP signaling is reported to inhibit CC
cell line growth in vitro (Lee et al., 2010). BMP signaling can
also have tumor-suppressive effects on other cells (Nishanian
et al., 2004; Piccirillo et al., 2006). To our knowledge, however,
in vivo experiments in mice supporting the tumor-suppressive
role of BMP signaling in advanced human CCs have not been
Other lines of evidence suggest a more complex scenario for
BMP signaling in CCs. For instance, BMP signaling appears to
be intact in primary CCs (Beck et al., 2006); BMPRI receptors
are expressed in all CC cells tested (Kodach et al., 2008b; Lee
et al., 2010); BMP4 expression is found at higher levels in late
CCs than in adenomas or normal colonic mucosa (Deng et al.,
2007); some CC cell lines are resistant to induced differentiation
and growth suppression by BMP4 (Nishanian et al., 2004); BMP7
can have pro-invasive activity on CC cells independent of SMAD4
(Grijelmo et al., 2007); BMP4 overexpression promotes invasive-
ness of CC cells with or without SMAD4 (Deng et al., 2007, 2009);
and BMP2 induces the epithelial-mesenchymal transition in CC
cells (Kang et al., 2009). Therefore, the role of BMP signaling in
human advanced CCs is far from clear (Thawani et al., 2010).
BMP signaling could have multiple roles in various contexts, for
instance acting initially as a tumor suppressor and later as a
tumor promoter, as proposed for the TGF-b superfamily in differ-
ent tumors (Blobe et al., 2000; Xu and Pasche, 2007).
To investigate the role of BMP signaling specifically in
advanced human CCs, we have sought to test the consequences
of autonomously altered levels of endogenous P-SMAD1,5,8, the
mediators of BMP signaling, adopting a strategy previously used
to analyze cartilage development (Zou et al., 1997) and forebrain
patterning (Gulacsi and Lillien, 2003) in vivo. Namely, we have
used constitutively active and dominant-negative forms of BMP
receptor Ib (BMPRIb) (also known as ALK6) to enhance or
reduce P-SMAD1,5,8 levels. These forms are encoded in mutant
cDNAs expressed from replication-incompetent lentiviral vectors
(LV), which insert into the genome upon infection and remain
stable, at low copy number given the low multiplicity of infection
(MOI of 1–3) used, and act cell-autonomously. With this strategy,
we should thus be able to alter P-SMAD1,5,8 levels only in CC
cells in vitro and in vivo, not directly affecting the stroma as it
would be the case in vivo using BMP ligands or secreted
Using these tools, we have studied the effect of the mutant
receptors in primary CC cells and in established CC cell lines,
both in vitro and in vivo. Here, we present evidence that BMP sig-
naling has tumor-promoting activity in vivo in primary CCs, but
not in established cell lines, and that a subset of primary
tumors require endogenous BMP signaling for normal growth
in vivo. Our data support the notion that BMP signaling has mul-
tiple roles in tumorigenesis, including an oncogenic action in
advanced CCs, making its blockade a possible therapeutic
target for advanced disease.
Human CC cells harbor active BMP signaling
To begin to characterize the status of BMP signaling in human
CCs, we tested for the expression of mRNAs encoding key
pathway elements in three advanced primary CCs and three CC
cell lines (Figure 1A; Varnat et al., 2009): CC14 is a tumor node
metastasis (TNM) stage 4 local tumor, CC36 is a TNM3 local
tumor and mCC11 is a TNM4 liver metastasis. We have also
chosen three CC cell lines that represent various kinds of CCs.
Ls174T is a near-diploid cell line with an activation mutation in
b-catenin, SW480 is a cell line with a loss of function mutation
in APC that also lacks SMAD4 and DLD1 is a microsatellite
unstable cell line harboring mutant APC. As control, we used a
sample of normal human colon.
RT–PCRanalysesof these varioussamples revealed the ubiqui-
tous presence of BMP2, BMP4, SMAD5 and SMAD1 albeit at
different levels. SMAD8 was expressed in at least half of the
samples tested, whereas the BMP inhibitors CHORDIN and
GREMLIN (Walsh et al., 2010) were only expressed in one
sample each (Figure 1A). The ubiquitous presence of BMP
ligands and of SMAD1,5,8, which mediate BMP signaling,
suggested an active pathway. Interestingly, SMAD4 was present
at high levels only in the primary tumors (Figure 1A). The three
primary tumors expressed 110-fold more SMAD4 than the three
cell lines tested on average (73-fold not counting SW480, that
is known to lack SMAD4; Zhang et al., 1996). Expression levels
in normal colon were low, likely owing to the dilution of the differ-
ent cell types in the heterogeneous sample.
To test whether BMP ligands were expressed at higher or lower
levels in the samples chosen when compared with normal human
colon, quantitative RT–PCR was used to measure mRNA levels.
These are shown as the ratios of the expression in the given
sample over the average expression in two independent
samples of normal human colon (Figure 1B). Notably, BMP2
was underexpressed and BMP4 overexpressed in all the
samples (Figure 1B, shown as the ratios of normalized expression
in CC cells over the normalized expression in normal human
colon). This raised the possibility that BMP4 signaling could
have a role in CC.
To ascertain that BMP signaling is active in all CC cells ana-
lyzed, and since BMPRIa/b expression is ubiquitous in all CC
cells analyzed (Deng et al., 2007; Kodach et al., 2008a, b; Lee
et al., 2010), we tested for the presence of phosphorylated (P-)
SMAD1,5,8, as these are present only in cells with active BMP
receptors, which are normally triggered by the presence of BMP
ligands. Western blot analyses using an antibody that recognizes
only the P-SMAD1,5,8 isoforms showed that all CC samplestested
harbor active P-SMADs, albeit at different levels (Figure 1C).
range, with CC14 at the low end and Ls174T at the high end.
BMP signaling in human colon cancerJournal of Molecular Cell Biology
Thus, data indicate that all primary CC cells and CC cell lines
appear to harbor active BMP signaling albeit at different levels.
Modulation of BMPRIb receptor activity
alters P-SMAD1,5,8 levels
Since BMP signaling is triggered by ligands activating type II and
type I BMP receptors leading to P-SMAD1,5,8 action (Figure 2A),
we have used the previously characterized constitutively active
(ca) and dominant-negative (dn) BMPRIb forms (Zou et al.,
1997) that we have cloned in GFP+(green fluorescent protein)
LV (Figure 2B). The caBMPRIb mutant harbors a Q203D substi-
ligand-BMPRII regulation, and the dnBMPRIb mutant harbors a
K231R substitution that diminishes its kinase activity (Figure 2B
and C). In addition, the lentivectors used also express GFP in a
ubiquitous manner (Figure 2B). These dominant mutant BMPRI
forms should alter endogenous P-SMAD1,5,8 levels.
Transduction of CC cells with these GFP+
incompetent lentivectors as well as with the GFP+parental
lentivector as control, followed by FACS-sorting for GFP fluor-
escence to insure a homogeneously transduced cell population,
showed the expected modulation of P-SMAD1,5,8 levels
(Figure 2D), confirming that these constructs enhance in the
case of the caBMPRIb, or diminish in the case of the
dnBMPRIb, BMP signaling. Quantification of the protein levels
normalized to HSP90 revealed that, compared with control
lentivector transduced cells, those expressing caBMPRIb or
P-SMAD1,5,8, respectively (Figure 2D).
To further verify that the caBMPRIb and dnBMPRIb constructs
had the specific effects expected on the BMP pathway, RT–
qPCR analyseswere performed
FACS-purified CC36 cells. Gene-expression values were then nor-
malized by the geometrical mean of the housekeeping genes
GAPDH and HMBS. Changes are presented as ratios of normalized
gene expression in caBMPRIb or dnBMPRIb over those in control
cells (Figure 2E). caBMPRIb induced the BMP targets SMAD6 and
ID1,2,3 by 200%–300%, whereas their levels were reduced by
dnBMPRIb by 40%–60% (Figure 2E). These manipulations thus
afford a range of modified signaling of +7.5-fold.
Modulation of BMPRIb receptor activity has
minor effects on CC cell proliferation and no
effects on apoptosis in vitro
To measure cell proliferation, we employed the nucleotide ana-
logue bromodeoxyuridine (BrdU), the incorporation of which
directly reports DNA synthesis and thus cells that were in S
phase during the pulse. BrdU incorporation was then revealed
by immunohistochemistry. Short-term (30–60 min) BrdU incor-
poration analyses of transduced cells at +70% confluence
showed minor effects, ranging from 10% to 40% changes when
compared with controls (Figure 3A). CC14, DLD1 and Ls174T
were largely unaffected (Figure 3A and data not shown),
whereas CC36 and SW480 showed a small (+12%) but signifi-
cant increase with caBMPRIb, and cell proliferation in CC36 and
mCC11 was also significantly compromised (reduced by +20%)
after signaling inhibition by dnBMPRIb (Figure 3A and B).
Similar analyses testing for the number of cleaved Caspase3
positive (Casp3+) cells, as a measure of apoptosis, failed to
reveal any significant changes with either caBMPRIb or
dnBMPRIb in relation to control (Figure 3C and D).
Figure 1 Primary CC cells and cell lines harbor active BMP signaling.
(A) Agarose gel showing RT–PCR evaluating the mRNA levels of BMP
signaling related genes in primary normal human colon, CC cells
(CC14, CC36, mCC11) and in CC cell lines (DLD1, Ls174T, SW480)
grown in vitro. BMP2 and BMP4 encode BMP ligands previously
reported to be present in CCs. CHORDIN and GREMLIN encode
secreted BMP inhibitors. SMAD1, SMAD5 and SMAD8 encode the
BMP-transducing SMADs. GAPDH levels were measured to quantify
mRNA levels. The control lanes represent reactions without cDNA.
(B) Quantitative RT–PCR showing the expression levels of BMP2
and BMP4 transcripts in primary CC cells and CC cell lines as
above. BMP2 and BMP4 mRNA levels were normalized with the
mean of the values of GAPDH and HMBS, and are shown as ratios
over the normalized levels in two normal human colon biopsies.
(C) Western blotanalysisof
SMAD1,5,8 proteins using a phospho-specific SMAD1,5,8 antibody.
Presence of these P-SMADs reveal BMP signaling activity in the
primary CC cells and cell lines as indicated. HSP90 protein levels
were measured for quantification.
Journal of Molecular Cell BiologyLorente-Trigos et al.
BMPRIb receptor modulation alters cell
behavior in a scratch assay in vitro
The small effects on cell proliferation described above suggested
a very minor role of BMP signaling in CC cells in vitro. However, to
explore this idea further, we tested for effects of altered BMP
receptor signaling using a scratch assay. In this test, cells are
grown to near confluency and then empty spaces are created
on the Petri dishes by scratching out cells with a blunt spatula,
thus making a gap between near-confluent cell populations.
Such assay is routinely used to test for altered behavior in the
interface of confluent and empty zones, assaying for the closure
of the gap left by the scratch.
CC14, CC36 and mCC11 primary CC cells were tested in the
scratch assay after transduction with control lentivectors or with
those expressing caBMPRIb or dnBMPRIb (Figure 4). The results
after 3 days of culture post-scratch showed that caBMPRIb
enhanced gap closure in all three primary CCs tested, whereas
Figure 2 Schematic diagram of BMP signaling and activity of mutant BMPRIb constructs. (A) Simplified representation of canonical BMP signal-
ing. In the absence of BMP ligands (left panel), SMAD1,5,8 exist as unphosphorylated proteins in the cytoplasm and transmembrane type I and
type II receptors do not associate. In presence of BMP ligands (right panel), BMPRIa and b receptors associate with ligand-bound BMPRII at the
cell’s membrane, triggering the phosphorylation of BMPRI receptors. This in turn results in the phosphorylation of SMAD1,5,8, which associate
with SMAD4 and translocate in the nucleus where they bind DNA-specific elements to modulate the expression of target genes. (B) Schematic
representation of the BMPRIb mutants used in this study derived from Zou et al. (1997). The constitutively active form of BMPRIb (caBMPRIb)
carries a Q203D amino acid residue substitution (upper panel). The dominant-negative form of BMPRIb (dnBMPRIb) carries a K231R amino acid
residue substitution (lower panel). The cDNAs were inserted in appropriate LV for use in CC cells. (C) Simplified representation of the mode of
action of BMPRIb mutants used in this study. The upper panel shows that the caBMPRIb kinase is constitutively active even in the absence of
BMP ligands, resulting in unregulated phosphorylation of SMAD1,5,8 leading to the permanent high activation of BMP signaling. The lower
panel shows that the kinase function of the dnBMPRIb is compromised, preventing signaling. Overexpression of such a mutant leads to com-
petition with endogenous receptors resulting in the downregulation of BMP signaling. (D) Western blot analysis of the phosphorylated
SMAD1,5,8 proteins revealing BMP signaling activity in primary CC14, CC36 and mCC11 cells and in the SW480 CC cell line upon expression
of caBMPRIb or dnBMPRIb when compared with parental control lentivector transduced cells (control). HSP90 protein levels were measured
for quantification. (E) Quantitative RT–PCR showing the in vitro expression levels of BMP pathway target genes ID1, ID2, ID3 and SMAD6 in
CC36 primary cells expressing either caBMPRIb or dnBMPRIb. Specific mRNA levels were normalized with GAPDH and HMBS and are shown
as ratios over the normalized levels in control lentivector transduced cells. ID1, ID2 and ID3 are canonical targets of BMP signaling and
SMAD6 encodes a pathway repressor that is induced transcriptionally by positive signaling and acts in a negative feedback loop.
BMP signaling in human colon cancerJournal of Molecular Cell Biology
dnBMPRIb significantly impeded normal healing in two out of
three cases, all when compared with control lentivector trans-
duced cells (Figure 4A and B). For quantification purposes, the
average gap distance observed with control cells was equated
to one (Figure 4B). The most extreme results were obtained
with CC36, in which dnBMPRIb increased the gap width by 40%
and caBMPRIb reduced it to ,10% of control width on average
(Figure 3B), closing it in some cases (Figure 3A). These results
demonstrate an involvement of endogenous BMP signaling in
gap closure. They also suggest that BMP signaling may favor
and may be required for enhanced proliferation or migration in
zones of asymmetric cell densities.
Increased cell proliferation at the gap edge
in cells expressing caBMPRIb
To test if cell proliferation could underlie the abnormal healing of
the scratched gap of cells with altered BMPR activity when com-
pared with control cells, we performed BrdU incorporation tests
using 15 min pulses before fixation and immunolabeling. These
Figure 3 Effects of BMPRIb activity modulation on CC cells in vitro. (A) Evaluation of proliferation of three primary CCs and of the SW480 cell line
in vitro upon expression of caBMPRIb or dnBMPRIb by BrdU incorporation assay, when compared with control(ctrl) lentivector transduced cells.
Histograms show BrdU+index defined as the ratio of BrdU+cells over the total number of DAPI+cells per field represented in percentages.
Asterisks denote significant changes (P , 0.05, Student’s t-test). ns: not significant. (B) Representative images of LV-caBMPRIb infection
levels in CC36 revealed by GFP expression (green; left panel), and of GFP+CC36 cells labeled with anti-BrdU antibodies (red; right panel).
(C) Evaluation of apoptosis in primary CCs upon expression of caBMPRIb or dnBMPRIb in vitro as assessed by activated-Caspase3 immunos-
taining. Histograms show activated Caspase3 positive Index (Casp3+index), which is the ratio of activated Casp3+cells over the number of
DAPI+cells per field. Asterisks denote significant changes (P , 0.05). ns: not significant. (D) Representative picture of activated Caspase3
immunolabeling (red) in vitro in mCC11 primary culture overexpressing caBMPRIb. Nuclei are counterstained with DAPI (blue). Scale bar ¼
60 mm for (B) and 50 mm for (D).
Journal of Molecular Cell Biology Lorente-Trigos et al.