Foxf2 in Intestinal Fibroblasts Reduces Numbers of Lgr5?Stem Cells and
Adenoma Formation by Inhibiting Wnt Signaling
ALI MOUSSAVI NIK,1AZADEH REYAHI,1FREDRIK PONTÉN,2and PETER CARLSSON1
1Department of Chemistry and Molecular Biology, University of Gothenburg, Gothenburg, Sweden; and2Department of Immunology, Genetics and Pathology,
Uppsala University, Uppsala, Sweden
See editorial on page 873.
BACKGROUND & AIMS: The stem cell niche at the base
of the intestinal crypts, as well as stemness and high clono-
genicity in colon cancer cells, depend on Wnt signaling to
?-catenin. Fibroblasts modulate the Wnt pathway in normal
and neoplastic epithelial cells via unclear mechanisms. We
investigated how in intestinal fibroblasts the forkhead tran-
scription factor Foxf2 controls Wnt signaling to affect num-
bers of stem cells and formation and growth of adenomas in
mice. METHODS: We created mice with different copy
numbers of Foxf2 by generating Foxf2?/?mice and a trans-
genic strain, Tg(FOXF2). Adenoma formation was investi-
gated in ApcMin/?mice, stem cells were counted in mice with
the Lgr5–enhanced green fluorescent protein knock-in allele,
proliferation was measured by incorporation of bromode-
oxyuridine, Foxf2 and Sfrp1 were localized by immunohis-
tochemistry, and signaling pathways were analyzed by quan-
titative polymerase chain reaction and immunoblot assays.
RESULTS: Epithelial ?-catenin was stabilized in Foxf2?/?
mice, resulting in increased number and size of adenomas.
Tg(FOXF2) mice, however, were partially resistant to intesti-
nal neoplasia and developed fewer and smaller adenomas;
Foxf2?/?mice developed 24-fold more tumors than Tg-
(FOXF2) mice. Epithelial cells of Foxf2?/?mice also had
higher numbers of Lgr5?stem cells and greater amounts of
crypt cell proliferation and expression of Myc (a target of
Wnt signaling) than Tg(FOXF2) mice. Expression of Sfrp1,
which encodes an extracellular inhibitor of Wnt, in fibro-
blasts increased with copy number of Foxf2. CONCLU-
SIONS: Foxf2 is a fibroblast factor that inhibits para-
crine Wnt signaling and restricts the crypt stem cell
niche in intestines of mice. Loss of Foxf2 promotes ad-
enoma formation and growth.
Keywords: Colon Cancer; Tumor Stroma; Mouse Model;
lumnar cells express Lgr5 and are constantly cycling stem
cells responsible for the continuous renewal of the intes-
tinal epithelium.2–6Stemness of Lgr5 cells requires contact
with Paneth cells, which provide the critical Wnt3 ligand
and other niche factors, such as the Notch ligand Dll4.7
n the normal intestine, Wnt signaling is critical for
maintenance of epithelial stem cells.1Crypt base co-
Early genetic events that ultimately lead to colorectal
cancer consistently involve mutations that constitutively
activate the canonical Wnt pathway.8–11In approximately
80% of colorectal tumors, this is caused by inactivation of
the APC gene, encoding a component of the ?-catenin
degradation complex. The inheritance of one defective
APC allele in individuals with familial adenomatous pol-
yposis, or the murine equivalent multiple intestinal neo-
plasia (ApcMin), dramatically increases the frequency of
APC inactivation through somatic loss of heterozygosity.
Despite the mutations in the ?-catenin degradation path-
way, colon cancer cells are heterogeneous within the tumor
with respect to levels of nuclear ?-catenin, indicating addi-
tional levels of regulation of Wnt signaling.12,13Cells with a
high level of Wnt pathway activity are most efficient in
initiating new tumors and possess characteristics of cancer
stem cells.14,15The critical role of Wnt signaling in inducing
stemness thus appears to be preserved in neoplasia but also
subject to modulation by the microenvironment; levels of
nuclear ?-catenin in colorectal tumors are influenced by
stromal fibroblasts12and, in vitro, fibroblasts can induce
Wnt pathway activity and stemness in tumor cells with
originally low clonogenicity.14This emphasizes the impor-
tance of identifying the nature of the interactions between
stromal fibroblasts and cancer cells, which is emerging as a
potentially important new target area for anticancer drugs,
fibroblast factors responsible for paracrine modulation of
which inhibit Wnt target gene expression,16,17Bmp antago-
nists,17,18and Hgf, which stabilizes ?-catenin.14
The forkhead transcription factors Foxf1 and Foxf2 are
expressed in subepithelial fibroblasts of the embryonic gut
and control epithelio-mesenchymal cross talk.19–22Mouse
development is highly sensitive to reduction in FoxF gene
dosage; compound Foxf1/f2 heterozygotes, as well as both
homozygous null mutants and a share of Foxf1 heterozy-
gotes, die at or before birth.19,21–23Perinatal mortality in
humans due to FOXF1 haploinsufficiency indicates that our
own species is equally susceptible.24During intestinal devel-
Abbreviations used in this paper: ANOVA, analysis of variance; BrdU,
bromodeoxyuridine; E, embryonic day; mRNA, messenger RNA; PCR,
polymerase chain reaction; qPCR, quantitative polymerase chain reac-
© 2013 by the AGA Institute
opment, epithelial Hedgehog signaling activates mesenchy-
mal expression of FoxF genes, which control Bmp4, Wnt5a,
and a range of genes encoding extracellular matrix pro-
teins.19Foxf2 null mutants are born with severe gut malfor-
mations, including muscular hypoplasia, anal atresia,
Hirschsprung’s disease, and tissue disintegration due to de-
ficiency of collagens and other connective tissue proteins.19
Here, we used moderate alterations in gene dosage to
show that Foxf2 expression in intestinal fibroblasts of adult
mice limits the extent of the Lgr5?stem cell niche. Mecha-
nistically, this is achieved through reduction of Wnt/?-
catenin signaling in epithelial cells and affects the initiation
frequency, as well as the growth rate, of intestinal adenomas.
Expression of the extracellular Wnt inhibitor Sfrp1 in fibro-
blasts is positively correlated with Foxf2 gene dosage and is a
candidate mediator of paracrine Wnt inhibition.
Materials and Methods
ApcMinand Lgr5-EGFP-ires-CreERT2 mice were obtained
from The Jackson Laboratory (Bar Harbor, ME). Foxf1 and Foxf2
targeted mutants have been described elsewhere.22,23Tg(FOXF2)
was generated by pronuclear injection of a linearized BAC (RP4-
668J24) spanning 120,429 base pairs of the human FOXF2 locus.
Full-length insertion was verified by polymerase chain reaction
(PCR) with primer pairs distributed over the BAC insert. Copy
number was determined by coamplification of human and mu-
rine sequences. All mutants were maintained on a C57Bl/6J
background. Experiments were approved by the Gothenburg
Animal Ethics Committee.
Quantification of Adenomas
The intestine was flat mounted with the luminal side
exposed, fixed, and stained with methylene blue. Polyps were
counted separately for the colon/rectum and the small intestine
(divided into 5 segments of equal length). Diameters were mea-
sured with digital calipers for all polyps in the colon/rectum and
the most posterior 20% of the small intestine. Wild-type litter-
mates of mutants were used as controls. No measurements
differed significantly between the controls for Foxf2?/?and Tg-
(FOXF2), and for the final analysis all controls were pooled.
Bromodeoxyuridine (BrdU) was injected intraperitoneally
exactly 1 hour before the mice were killed and perfused with 4%
formaldehyde, and the intestines were dissected and fixed over-
night. Biotin-labeled secondary antibodies, amplification (TSA Bi-
otin System Kit; NEN Life Science Products, Boston, MA), horse-
tetrahydrochloride was used for visualization of primary antibody
binding, and methylene blue was used for nuclear counter staining.
Primary antibodies are listed in Supplementary Table 1.
Total RNA was prepared from a fixed length of small
intestine from 12-week-old mice, taken at 70% of the distance
between the pylorus and cecum, from whole embryonic intestine,
or from primary intestinal fibroblasts cultured in vitro. Samples
quantitative polymerase chain reaction (qPCR) in triplicate and
normalized against the 36B4 transcript.25For primer sequences, see
Supplementary Table 1.
Quantification of Lgr5?Stem Cells
Segments of small intestine, corresponding to 60% to
80% of the distance between the pylorus and cecum, from 12-
week-old mice carrying the Lgr5-EGFP-ires-CreERT2 allele2were
fixed in 4% paraformaldehyde under pressure between filter
papers. Pieces of the flattened guts were analyzed by confocal
microscopy. From Z-stacks, 3-dimensional reconstructions were
made of each crypt separately and the wedge-shaped, EGFP?
cells at the bottom of the crypts were counted.
Analysis of variance (ANOVA) was used to test how param-
eters depended on the number of Foxf2 alleles, and 2-tailed Student
t test was used for pairwise comparisons of means. Raw data were
log transformed when required to obtain normal distribution.
Heterozygosity for FoxF Genes Increases
Intestinal Adenomas in ApcMin/?Mice
To study if FoxF gene dosage would affect intesti-
nal adenoma formation, we generated Foxf1?/?; ApcMin/?
and Foxf2?/?; ApcMin/?mice and analyzed the number and
size of intestinal polyps at the age of 12 weeks. Heterozy-
gosity for Foxf1 increased the number of polyps (P ? .08;
n ? 14 and n ? 11; Supplementary Figure 1), but loss of
one Foxf2 allele had a greater effect on both number and
size of adenomas (see the following text); from this point
on, we focused on this gene.
Generation of Transgenic Mice With an Extra
Copy of FOXF2
To investigate if an increased Foxf2 gene dosage
would mirror the heterozygosity and protect against tumors,
we generated a transgenic strain, Tg(FOXF2), which carries a
human BAC with FOXF2 located at the center of a 120-
kilobase insert and contains no other gene. The human/
murine orthologues are highly conserved (95% identical),
and a human BAC was chosen to simplify genotyping and
mapping (Supplementary Figure 2). Functionality of the
transgene was confirmed by rescue of the Foxf2 knockout.
Foxf2?/?mice normally die at birth with cleft palate and
intestinal malformations,19,23but they grew to apparently
normal adults when they carried the Tg(FOXF2) transgene
(Supplementary Figure 3). This showed that human FOXF2
can substitute for the murine orthologue and that the BAC
clone contains cis-regulatory elements capable of directing
adequate expression with correct tissue specificity in mice.
Adenoma Formation and Growth Are
Negatively Correlated With Foxf2 Gene Dosage
The number of adenomas in intestines of ApcMin/?
mice was negatively correlated (ANOVA, P ? 10?6) with
Foxf2 gene dosage; heterozygotes had 55% more, and Tg-
(FOXF2) 57% less, than wild type (Figure 1A and B and
Supplementary Figure 4). The size of small intestine adeno-
mas was also negatively correlated (ANOVA, P ? 10?22) with
1002NIK ET AL GASTROENTEROLOGY Vol. 144, No. 5
Foxf2 allele number; diameters were on average 51% larger in
heterozygotes, and 20% smaller in Tg(FOXF2), compared
with wild type (Figure 1C and D). Colorectal polyps were few
and differences between genotypes did not reach statistical
significance except for the 5.7-fold larger tumor volume in
Foxf2 heterozygotes (P ? .001; n ? 22) compared with wild
type (n ? 8; Figure 1A). The calculated combined effect of
differences in size and number of small intestine adenomas
was equivalent to a 24-fold difference in tumor burden
between Foxf2?/?and Tg(FOXF2).
Foxf2 Is Present in Nuclei of Subepithelial
Adenomas develop from the gut epithelium,
whereas (at least in embryos) Foxf2 expression is confined
Figure 1. Foxf2 inhibits intesti-
growth in ApcMinmice. (A) Ade-
nomas (arrowheads) in distal il-
eum and colon from 12-week-
old ApcMin/?mice with different
Foxf2 gene dosages. (B) Adeno-
mas per ApcMin/?mouse (aver-
age ? SEM) for different Foxf2
genotypes: Foxf2?/?, 66.2 pol-
yps/mouse (n ? 17); wild-type,
42.8 polyps/mouse (n ? 24); Tg-
(n ? 14) (ANOVA, P ? 10?6;
pairwise P ? .02, P ? 10?4, and
P ? 2 ? 10?7, respectively).
(C) Diameter of adenomas in dis-
Foxf2 genotypes (average ?
SEM of10log of diameter in milli-
meters): Foxf2?/?, 1.22 mm (n ?
216); wild-type, 0.81 mm (n ?
133); Tg(FOXF2), 0.65 mm (n ?
52) (ANOVA, P ? 10?22; pair-
wise P ? 10?14, P ? .005, and
P ? 10?16, respectively). (D) His-
tograms of polyp size for the 3
May 2013Foxf2 INHIBITS INTESTINAL Wnt SIGNALING 1003
to mesenchymal cells.19,26,27The distribution of Foxf2 in
the adult intestine has not been described because of lack
of specific antibodies, and the possibility that Foxf2 ex-
pression is activated in the epithelium at gut maturation
or during the transition to adenoma could not be ex-
cluded. With a novel antibody (HPA004763; Human Pro-
tein Atlas28), we could detect Foxf2 in the nuclei of mes-
enchymal and smooth muscle progenitor cells in
embryonic day (E)18.5 intestine (black arrowheads in Figure
2A) and confirm the specificity by absence of staining in
the corresponding tissue from Foxf2?/?embryos (Figure
2B). The staining pattern is consistent with expression of
Foxf2 in the nonepithelial cells of embryonic intestine, as
determined by in situ hybridization,19,26,27and with nu-
clear localization of the Foxf2 protein.29Analysis of small
intestine from adult mice showed that Foxf2 is confined
to the nuclei of fibroblasts of the lamina propria, whereas
all epithelial cells and the smooth muscle cells of the
muscularis externa are consistently negative (Figure 2C).
Throughout the lamina propria, Foxf2?fibroblasts (black
arrowheads in Figure 2C) occurred side by side with Foxf2?
cells (white arrowheads), but the Foxf2?cells were most
abundant in villi and along the upper parts of the crypts.
Around the crypt base and between crypts and the mus-
cular layers, Foxf2?cells were significantly less frequent
(Figure 2C, top right). The epithelial cells of adenomas
from ApcMin/?mice were negative, whereas the stroma
exhibited the same mix of Foxf2?and Foxf2?nuclei as
normal intestine (Figure 2D). The restriction of Foxf2 to
subepithelial fibroblasts thus implicates intercellular
communication between fibroblasts/stroma and epitheli-
um/adenoma in mediating the effects of Foxf2 genotype
on tumor initiation and growth.
Foxf2 Reduces Epithelial Proliferation in
To understand how Foxf2 in fibroblasts affects
adenoma initiation and growth, we searched for Foxf2-
dependent alterations in the normal intestinal epithelium.
Figure 2. Intestinal Foxf2 is restricted to nuclei of nonepithelial cells. (A and B) Immunohistochemistry with Foxf2 antibody on sections of small
Nuclear staining of mesenchymal fibroblasts and smooth muscle progenitors is absent from Foxf2?/?and represents Foxf2. Black arrowheads
in the lamina propria, with density of Foxf2?cells decreasing toward the base of crypts. (D) Foxf2 in adenoma of ApcMin/?mice. A mixture of Foxf2?
and Foxf2?fibroblasts is seen in the stroma. Diffuse staining of epithelium in all panels is nonspecific background.
1004 NIK ET AL GASTROENTEROLOGY Vol. 144, No. 5
Loss of heterozygosity in the Apc locus, leading to ade-
noma formation, is a stochastic process, the frequency of
which is assumed to be proportional to the proliferation
rate, and we therefore hypothesized that Foxf2 in fibro-
blasts reduces proliferation of epithelial cells. Quantifica-
tion of S-phase cells in the small intestine of the 3 Foxf2
genotypes by BrdU pulse labeling showed that the num-
ber of BrdU?cells per crypt was negatively correlated
(ANOVA, P ? 10?20) with Foxf2 gene dosage: 24% higher
in Foxf2?/?and 17% lower in Tg(FOXF2) compared with
wild type (Figure 3A and B). The magnitude of these
differences suggests that altered epithelial proliferation
accounts for a major share of the effect of Foxf2 gene
dosage on tumor initiation. The consequence of increased
proliferation was evident from the histology of Foxf2?/?;
at birth it was indistinguishable from wild type, but at 3
Figure 3. Foxf2 controls crypt
cell proliferation and Lgr5?stem
cell number. (A) Anti-BrdU sta-
ining of longitudinal sections
from distal ileum of 12-week-
old Foxf2?/?, wild-type, and
Tg(FOXF2) mice exposed to a
1-hour BrdU pulse. (B) BrdU?
nuclei per crypt (average ?
SEM): Foxf2?/?, 9.9; wild-type,
8.0; Tg(FOXF2), 6.7 (ANOVA,
P ? 10?20; pairwise P ? 10?9,
P ? 10?4, and P ? 10?20, re-
spectively). Each data point is
the number of BrdU?nuclei di-
vided by the number of crypts in
one microscopic field (n ? 62,
n ? 48, and n ? 41, respec-
tively). (C) Clusters of epithelial
cells in Foxf2?/?ileal villi at the
age of 12 weeks. (D) qPCR of
Lgr5 mRNA from (left) E16.5 in-
testine (average ? SEM; P ?
.007; n ? 5 wild-type, 6 mu-
tants), and (right) adult small in-
testine (average ? SEM; P ?
.08; n ? 5 per genotype). (E)
Number of Lgr5?stem cells
intestine (average ? SEM;
ANOVA, P ? 10?5; Foxf2?/?,
13.92 ? 0.32 [n ? 49]; wild-
type, 12.79 ? 0.37 [n ? 59];
Tg(FOXF2), 11.40 ? 0.32 [n ?
83]; pairwise P ? .03, P ? .004,
and P ? 10?4, respectively). (F)
Two-dimensional projections of
confocal 3-dimensional recon-
structions of ileal crypts from
mice carrying the Lgr5-EGFP-
May 2013Foxf2 INHIBITS INTESTINAL Wnt SIGNALING1005
months enlarged villi with excess epithelial cells (Figure
3C), reminiscent of E18.5 Foxf1?/?; Foxf2?/?double
Foxf2 Controls the Size of the Crypt Stem Cell
To investigate how Foxf2 controls epithelial prolif-
eration, we analyzed how Foxf2 affected intestinal stem
cells. Epithelial turnover is driven by the continuous pro-
liferation of Lgr5?stem cells, the number of which re-
mains remarkably constant over time and between
crypts.4,7Intestines from E16.5 Foxf2?/?embryos con-
tained 27% more Lgr5 messenger RNA (mRNA) than wild
type (Figure 3D; P ? .007). Adult wild-type intestine
contained 19% more Lgr5 mRNA, and Foxf2?/?33% more,
than Tg(FOXF2), but the differences were not statistically
significant (P ? .08; Figure 3D). To enable direct quanti-
fication of Lgr5?cells, we introduced the Lgr5-EGFP-ires-
CreERT2 allele2into the 3 Foxf2 genotypes. Variegated
expression of the transgene generates a mosaic pattern
with expressing crypts mixed with nonexpressing crypts,
which makes quantification of EGFP?cells by fluores-
counted the number of EGFP?cells in each positive crypt
directly by confocal microscopy. The number of Lgr5?
cells showed a strong negative correlation (ANOVA, P ?
10?5) with Foxf2 gene dosage (Figure 3E and F). Foxf2
expression in fibroblasts thus appears to limit the extent
of the stem cell niche for Lgr5?cells and thereby, indi-
rectly, the production of epithelial cells.
Foxf2 Inhibits Wnt Signaling
Canonical Wnt signaling is essential for stemness
of Lgr5?cells. Secretion of factors that modulate the Wnt
pathway is therefore one way through which fibroblasts
may control the extent of the stem cell niche. Consistent
with this hypothesis, immunostaining suggested that
Foxf2?/?epithelium contained more ?-catenin than wild
type; epithelial staining in wild type contrasted more with
adenomas than with the negative connective tissue,
whereas the opposite was true in Foxf2?/?(Figure 4A).
Western blot confirmed that Foxf2?/?intestines contain
more ?-catenin protein than wild type, subcellular frac-
tionation confirmed that the difference is attributable to
an increase in nuclear ?-catenin (Figure 4B), and qPCR
confirmed that the difference is not caused by an elevated
mRNA level (Figure 4C). These results are consistent with
stabilization of the ?-catenin protein by increased Wnt
signaling in the Foxf2 heterozygote.
Altered Wnt signaling and ?-catenin levels are expected
to influence the expression of Wnt target genes. The cell
cycle activator Myc is a key mediator of Wnt/?-catenin
signaling in intestinal epithelial cells.30–33Simultaneous
deletion of Apc and Myc rescues many of the perturbations
normally associated with Apc inactivation, and the major-
ity of Wnt target genes require Myc for their activation by
nuclear ?-catenin.31Myc expression in the small intestine
Figure 4. Foxf2 inhibits the Wnt/?-catenin pathway in intestinal epithe-
lium. (A) Anti–?-catenin immunohistochemistry of small intestine with
adenomas from ApcMin/?and ApcMin/?; Foxf2?/?mice. ?-catenin is con-
stitutively nuclear in adenomas (black arrowheads), and normal epithe-
lium (white arrowheads) stains stronger in Foxf2?/?(contrasts more with
negative mesenchyme [asterisk] than with adenoma) than in wild type.
(B) Western blot showing more nuclear ?-catenin protein in Foxf2?/?
intestine than in wild type (Foxf2?/?235% of wild type in nucleus, 96% in
cytoplasm; HDAC1 and ?-tubulin as loading controls for nuclear and
cytoplasmic, respectively). (C) mRNA encoding ?-catenin is not elevated
in Foxf2?/?intestine (qPCR; n ? 3/genotype). (D) Foxf2?/?intestine
contained 31% more Myc mRNA, and Tg(FOXF2) 62% less, than wild
type (qPCR; ANOVA, P ? .001; pairwise P ? 0.18, P ? .009, and P ?
(E) Myc Western blot of proteins from small intestine of adult wild type,
Foxf2?/?, and Tg(FOXF2) (?-actin as loading control). Quantification
(Myc/actin): Foxf2?/?178% and Tg(FOXF2) 21% of wild-type level.
1006 NIK ET AL GASTROENTEROLOGY Vol. 144, No. 5
of the 3 Foxf2 genotypes was negatively correlated with
Foxf2 gene dosage, both at the mRNA and protein levels
(Figure 4D and E). Compared with wild type, Foxf2
heterozygotes had 131% Myc mRNA and Tg(FOXF2) had
38% (ANOVA, P ? 10?3); based on Western blot (Myc/
actin), the Myc protein level was 178% of wild type in
Foxf2?/?and 21% in Tg(FOXF2). The strong correlation
between Foxf2 copy number and expression of the Wnt
target Myc is consistent with the notion that altered Wnt
signaling is responsible for the observed effects on stem
cell number and crypt cell proliferation.
Foxf2 Increases Mesenchymal Expression of
the Wnt Inhibitor Sfrp1
The mediators of Wnt inhibition by Foxf2 are
expected to be extracellular because Foxf2 is a transcrip-
tion factor present in fibroblasts. During embryonic de-
velopment, Foxf2 activates Bmp4 and inhibits mesenchy-
mal Wnt5a,19but in the adult intestine we found no
evidence for altered expression of Bmp4 or Wnt5a (Supple-
mentary Figure 5). Another potential candidate was Hgf,
which is produced by fibroblasts and has been shown to
induce stemness in colon cancer stem cells,14but no
increase in Hgf expression was detected in Foxf2?/?intes-
tine compared with wild type (data not shown). The
concentration of Wnt transcripts was either below the
detection limit (Wnt1, 3a, 7a, 7b, 8a, and 8b) or did not
change significantly with increasing Foxf2 gene dosage
(Wnt2, 2b, 3, 11, 16, 4, 5a, 5b, 6, and 9; Supplementary
Figure 5). Of the 6 genes encoding extracellular Wnt
inhibitors, Sfrp1 had the highest expression in the intes-
tine (Supplementary Figure 5), and its mRNA level was
positively correlated with Foxf2 gene dosage (Figure 5B);
wild type contained 23% and Tg(FOXF2) 56% more Sfrp1
mRNA than Foxf2?/?(ANOVA, P ? .005). Western blot
confirmed that Foxf2?/?contained less (16%) and Tg-
(FOXF2) more (154%) Sfrp1 protein than wild type (Figure
To investigate the relationship directly between Sfrp1
and Foxf2 mRNA, rather than with gene dosage, we quan-
tified intestinal Foxf2 mRNA by qPCR. The primer pair
targets a site in Foxf2 with 100% conservation between
mouse and human and thus allows detection of the Tg-
(FOXF2)-derived human transcript together with the mu-
rine. E18.5 Foxf2?/?guts were indistinguishable from neg-
ative controls, whereas Foxf2/FOXF2 mRNA in the 3 adult
Foxf2 genotypes was positively correlated with gene dos-
age; Foxf2?/?contained 40% and Tg(FOXF2) 127% of the
wild-type level (Figure 5C; ANOVA, P ? 10?3without and
2 ? 10?4with Foxf2?/?included). Sfrp1 mRNA correlated
strongly with Foxf2 mRNA in the same samples (Figure
5D; covariance P ? .01).
The changes in Sfrp1 expression could either contribute
to or be a consequence of Foxf2-dependent paracrine sig-
naling. To distinguish between these alternatives, we first
established where Sfrp1 is expressed. Immunohistochem-
istry showed that Sfrp1 is present in the lamina propria
mesenchyme (Figure 5A) and has a distribution similar to
that of Foxf2: abundant in villi and around the upper part
of crypts but sparse around the crypt base. Double im-
munostaining for Foxf2 and Sfrp1 confirmed colocaliza-
tion, and the marker collagen I verified that the express-
ing cells are fibroblasts, whereas no overlap with the
smooth muscle marker SMA was observed (Supplemen-
tary Figure 6). Separation of epithelium from mesen-
chyme confirmed that Sfrp1 mRNA and protein are
localized predominantly to the mesenchyme (Figure 5F)
and have the same tissue specificity as Foxf2 and Fn1
(encoding fibronectin). This pattern supports the no-
tion that Sfrp1 contributes to paracrine inhibition of
the Wnt pathway by fibroblasts.
Even though Sfrp1 is expressed in the same cell popu-
lation as Foxf2, the correlation between the amount of
Sfrp1 mRNA and Foxf2 might still be a secondary effect,
induced by altered signaling from epithelium to mesen-
chyme. To investigate if the changes in Sfrp1 expression
are direct, cell-autonomous responses to altered Foxf2
gene dosage, we isolated fibroblasts from Foxf2?/?, wild-
type, and Tg(FOXF2) small intestine, cultured them in
vitro in the absence of epithelial cells, and analyzed their
content of Sfrp1 mRNA. The effect of Foxf2 gene dosage
on Sfrp1 expression was of the same magnitude in this
experimental system as in the intact tissue (Figure 5B);
wild type contained 30% and Tg(FOXF2) 55% more Sfrp1
mRNA compared with Foxf2?/?(P ? .05 for Foxf2?/?
versus Tg[FOXF2]). Cultured Foxf2?/?fibroblasts from
E18.5 gut contained 40% less Sfrp1 mRNA than wild type
(Figure 5B; P ? .005). Taken together, these results indi-
cate that Foxf2 activates Sfrp1 expression in a direct,
Fibroblasts in close proximity to the intestinal
epithelium are assumed to be important for defining the
shape and extent of the stem cell niche. Wnt agonists
(either direct, such as Wnt5a,34or indirect, such as Hgf14
and the Bmp antagonist Grem118) are expressed by fibro-
blasts that surround the basal crypt epithelium, whereas
Bmps, which inhibit the Wnt pathway, are produced by
fibroblasts higher up along the crypt-villus axis.16,17,35
Extracellular Wnt inhibitors are potent antagonists of
epithelial proliferation, crypt integrity, and stem cell
survival,6,36–38as shown by addition of recombinant
proteins in vitro36or ectopic expression in vivo.6Fur-
thermore, these experiments showed plasticity, not only
in the epithelial proliferation rate but also in the num-
ber of Lgr5?cells per crypt. However, the physiologic
regulation of Wnt inhibitors, and their contribution to
shaping the stem cell niche under normal conditions,
has remained unclear. We show that Foxf2 expression in
fibroblasts correlates with repression of epithelial Wnt
signaling, that Foxf2-positive fibroblasts are wide-
spread in the villi and along the upper parts of the
May 2013 Foxf2 INHIBITS INTESTINAL Wnt SIGNALING 1007
crypts but sparse around the crypt base, and that vari-
ations in Foxf2 level have a direct effect on Wnt target
expression, stem cell number, proliferation, and both
initiation and growth of adenomas. Finally, we iden-
tify Sfrp1 as an extracellular Wnt inhibitor produced
by intestinal fibroblasts, whose expression increases
with Foxf2 expression. A schematic summary of the
suggested role of Foxf2 in the adult intestine, and
its relation to other signaling pathways, is shown in
1008 NIK ET ALGASTROENTEROLOGY Vol. 144, No. 5
Foxf2 is a transcription factor with potentially hun-
dreds of target genes, and the mechanistic links with the
Wnt pathway may be complex and multifaceted. Foxf2-
null mutants have, for example, a severe deficiency of
extracellular matrix proteins,19and association with ma-
trix is important for retention of many paracrine signal-
ing proteins. An altered matrix composition has been
suggested as a mechanism behind elevated Wnt signaling
in the gut.39
Approximately 95% of colorectal cancers exhibit epige-
netic silencing of SFRP1.40Restoration of SFRP1 expres-
sion in colon cancer cell lines increased apoptosis and
diminished ?-catenin levels and Myc expression, all con-
sistent with reduced canonical Wnt signaling.41This oc-
curred despite mutations in APC or CTNNB1,41which
suggests that Wnt ligands are important even when
?-catenin degradation is defective. The widespread silenc-
ing of SFRP genes in colorectal cancer indicates that
autocrine secretion of SFRPs is an important mechanism
that limits growth and progression of intestinal neo-
plasms. However, we found that in normal intestine Sfrp1
expression is confined to nonepithelial cells, and similar
results have been obtained with human colon and small
intestine, where SFRP1 mRNA was found to be abundant
in fibroblasts but absent from the epithelium.42Activa-
tion of SFRP genes in neoplastic cells may therefore be a
response to pathologic activation of the Wnt pathway and
a defense mechanism that stalls the growth of tumors
until silenced epigenetically or by mutation.
Foxf2 is a target of Hedgehog signaling in many organs,
including the intestine.19,26,27,43,44In the gut, Hedgehog
ligands secreted by the epithelium target the mesen-
chyme/fibroblasts and pattern the crypt-villus axis, with
highest activity in the villus.45–47Inhibition of Hedgehog
signaling resulted in epithelial overproliferation and for-
mation of ectopic crypts,48whereas constitutive activation
of the same pathway led to stem cell depletion.49In all of
these experimental situations, Wnt pathway activity was
negatively correlated with Hedgehog signaling, which sug-
gests that Hedgehog acts as a negative feedback regulator
of epithelial proliferation and restricts high-level Wnt
signaling to the crypt base. With Foxf2 as a mesenchymal
Hedgehog target, the results presented here contribute to
our understanding of how Hedgehog inhibits the Wnt
cultured fibroblasts from the same 3 genotypes (center panel), wild type contained 30%, and Tg(FOXF2) 55%, more Sfrp1 mRNA than Foxf2?/?(P ? .05 for
Foxf2?/?vs Tg[FOXF2]; n ? 14 per genotype). Cultured intestinal fibroblasts from wild-type E18.5 embryos contained 59% more Sfrp1 mRNA than cells from
were indistinguishable from negative controls (ND, not detected). Foxf2?/?contained 40% and Tg(FOXF2) 127% of the wild-type level (ANOVA, P ? 10?3with
Foxf2?/?excluded and 2 ? 10?4when included). (D) Scatter plot of10log Sfrp1 mRNA vs10log of Foxf2/FOXF2 mRNA for small intestine samples from 24
as loading control). Quantification (Sfrp1/actin): Foxf2?/?16% and Tg(FOXF2) 154% of wild-type level. (F) qPCR quantification of mRNA (left panel) for Sfrp1, 2
mesenchymal markers (Foxf2 and Fn1), and 2 epithelial (Ctnnb1 and Lgr5) markers in RNA prepared separately from epithelium (red bars) and mesenchyme-
Figure 6. (A) Schematic illustration of gradients formed by signaling
molecules of the crypt-villus axis, and the proposed mechanism for how
Foxf2 in fibroblasts limits the stem cell niche for Lgr5?cells (yellow) by
inhibition of Wnt signaling. (B) Simplified summary of paracrine signaling
between epithelium (red) and fibroblasts (blue), and a proposed mech-
anism through which Foxf2 in fibroblasts inhibits Wnt signaling in adja-
cent epithelial cells. “Wnt” is used as a generic term for different Wnt
May 2013 Foxf2 INHIBITS INTESTINAL Wnt SIGNALING1009
Note: To access the supplementary material
accompanying this article, visit the online version of
Gastroenterology at www.gastrojournal.org, and at http://
1. Korinek V, Barker N, Moerer P, et al. Depletion of epithelial stem-
cell compartments in the small intestine of mice lacking Tcf-4. Nat
2. Barker N, van Es JH, Kuipers J, et al. Identification of stem cells in
small intestine and colon by marker gene Lgr5. Nature 2007;449:
3. Sato T, Vries RG, Snippert HJ, et al. Single Lgr5 stem cells build
crypt-villus structures in vitro without a mesenchymal niche. Na-
4. Snippert HJ, van der Flier LG, Sato T, et al. Intestinal crypt homeo-
stasis results from neutral competition between symmetrically
dividing Lgr5 stem cells. Cell 2010;143:134–144.
5. van der Flier LG, Clevers H. Stem cells, self-renewal, and differen-
tiation in the intestinal epithelium. Annu Rev Physiol 2009;71:
6. Yan KS, Chia LA, Li X, et al. The intestinal stem cell markers Bmi1
and Lgr5 identify two functionally distinct populations. Proc Natl
Acad Sci U S A 2012;109:466–471.
7. Sato T, van Es JH, Snippert HJ, et al. Paneth cells constitute the
niche for Lgr5 stem cells in intestinal crypts. Nature 2011;469:
8. Fearon ER, Vogelstein B. A genetic model for colorectal tumori-
genesis. Cell 1990;61:759–767.
9. Clevers H. Wnt/beta-catenin signaling in development and dis-
ease. Cell 2006;127:469–480.
10. Bienz M, Clevers H. Linking colorectal cancer to Wnt signaling.
11. Vogelstein B, Fearon ER, Hamilton SR, et al. Genetic alterations
during colorectal-tumor development. N Engl J Med 1988;319:
12. Brabletz T, Jung A, Reu S, et al. Variable beta-catenin expression
in colorectal cancers indicates tumor progression driven by the
tumor environment. Proc Natl Acad Sci U S A 2001;98:10356–
13. Fodde R, Brabletz T. Wnt/beta-catenin signaling in cancer stem-
ness and malignant behavior. Curr Opin Cell Biol 2007;19:
14. Vermeulen L, De Sousa EMF, van der Heijden M, et al. Wnt activity
defines colon cancer stem cells and is regulated by the microen-
vironment. Nat Cell Biol 2010;12:468–476.
15. Merlos-Suarez A, Barriga FM, Jung P, et al. The intestinal stem cell
signature identifies colorectal cancer stem cells and predicts
disease relapse. Cell Stem Cell 2011;8:511–524.
16. Haramis AP, Begthel H, van den Born M, et al. De novo crypt
formation and juvenile polyposis on BMP inhibition in mouse
intestine. Science 2004;303:1684–1686.
17. He XC, Zhang J, Tong WG, et al. BMP signaling inhibits intestinal
stem cell self-renewal through suppression of Wnt-beta-catenin
signaling. Nat Genet 2004;36:1117–1121.
18. Kosinski C, Li VS, Chan AS, et al. Gene expression patterns of
human colon tops and basal crypts and BMP antagonists as
intestinal stem cell niche factors. Proc Natl Acad Sci U S A
19. Ormestad M, Astorga J, Landgren H, et al. Foxf1 and Foxf2 control
murine gut development by limiting mesenchymal Wnt signaling
and promoting extracellular matrix production. Development
20. Astorga J, Carlsson P. Hedgehog induction of murine vasculogen-
esis is mediated by Foxf1 and Bmp4. Development 2007;134:
21. Mahlapuu M, Enerbäck S, Carlsson P. Haploinsufficiency of the
forkhead gene Foxf1, a target for Sonic hedgehog signaling,
causes lung and foregut malformations. Development 2001;128:
22. Mahlapuu M, Ormestad M, Enerback S, et al. The forkhead tran-
scription factor Foxf1 is required for differentiation of extra-embry-
onic and lateral plate mesoderm. Development 2001;128:155–
23. Wang T, Tamakoshi T, Uezato T, et al. Forkhead transcription
factor Foxf2 (LUN)-deficient mice exhibit abnormal development of
secondary palate. Dev Biol 2003;259:83–94.
24. Stankiewicz P, Sen P, Bhatt SS, et al. Genomic and genic dele-
tions of the FOX gene cluster on 16q24.1 and inactivating muta-
tions of FOXF1 cause alveolar capillary dysplasia and other mal-
formations. Am J Hum Genet 2009;84:780–791.
25. Akamine R, Yamamoto T, Watanabe M, et al. Usefulness of the 5’
region of the cDNA encoding acidic ribosomal phosphoprotein P0
conserved among rats, mice, and humans as a standard probe for
gene expression analysis in different tissues and animal species.
J Biochem Biophys Methods 2007;70:481–486.
26. Aitola M, Carlsson P, Mahlapuu M, et al. Forkhead transcription
factor FoxF2 is expressed in mesodermal tissues involved in
epithelio-mesenchymal interactions. Dev Dyn 2000;18:136–149.
27. Ormestad M, Astorga J, Carlsson P. Differences in the embryonic
expression patterns of mouse Foxf1 and -2 match their distinct
mutant phenotypes. Dev Dyn 2004;229:328–333.
28. Ponten F, Jirstrom K, Uhlen M. The Human Protein Atlas—a tool
for pathology. J Pathol 2008;216:387–393.
29. Hellqvist M, Mahlapuu M, Blixt A, et al. The human forkhead
protein FREAC-2 contains two functionally redundant activation
domains and interacts with TBP and TFIIB. J Biol Chem 1998;273:
30. He TC, Sparks AB, Rago C, et al. Identification of c-MYC as a target
of the APC pathway. Science 1998;281:1509–1512.
31. Sansom OJ, Meniel VS, Muncan V, et al. Myc deletion rescues Apc
deficiency in the small intestine. Nature 2007;446:676–679.
32. Sansom OJ, Reed KR, Hayes AJ, et al. Loss of Apc in vivo imme-
diately perturbs Wnt signaling, differentiation, and migration.
Genes Dev 2004;18:1385–1390.
33. van de Wetering M, Sancho E, Verweij C, et al. The beta-catenin/
TCF-4 complex imposes a crypt progenitor phenotype on colorectal
cancer cells. Cell 2002;111:241–250.
34. Pacheco, II, Macleod RJ. CaSR stimulates secretion of Wnt5a
from colonic myofibroblasts to stimulate CDX2 and sucrase-
isomaltase using Ror2 on intestinal epithelia. Am J Physiol Gas-
trointest Liver Physiol 2008;295:G748–G759.
35. Hardwick JC, Van Den Brink GR, Bleuming SA, et al. Bone mor-
phogenetic protein 2 is expressed by, and acts upon, mature
epithelial cells in the colon. Gastroenterology 2004;126:111–
36. Ootani A, Li X, Sangiorgi E, et al. Sustained in vitro intestinal
epithelial culture within a Wnt-dependent stem cell niche. Nat Med
37. Kuhnert F, Davis CR, Wang HT, et al. Essential requirement for
Wnt signaling in proliferation of adult small intestine and colon
revealed by adenoviral expression of Dickkopf-1. Proc Natl Acad
Sci U S A 2004;101:266–271.
38. Pinto D, Gregorieff A, Begthel H, et al. Canonical Wnt signals are
essential for homeostasis of the intestinal epithelium. Genes Dev
39. Perreault N, Katz JP, Sackett SD, et al. Foxl1 controls the Wnt/
beta-catenin pathway by modulating the expression of proteogly-
cans in the gut. J Biol Chem 2001;276:43328–43333.
40. Suzuki H, Gabrielson E, Chen W, et al. A genomic screen for genes
upregulated by demethylation and histone deacetylase inhibition
in human colorectal cancer. Nat Genet 2002;31:141–149.
41. Suzuki H, Watkins DN, Jair KW, et al. Epigenetic inactivation of
SFRP genes allows constitutive WNT signaling in colorectal can-
cer. Nat Genet 2004;36:417–422.
1010NIK ET AL GASTROENTEROLOGY Vol. 144, No. 5
42. Hughes KR, Sablitzky F, Mahida YR. Expression profiling of Wnt Download full-text
family of genes in normal and inflammatory bowel disease primary
human intestinal myofibroblasts and normal human colonic crypt
epithelial cells. Inflamm Bowel Dis 2011;17:213–220.
43. Lan Y, Jiang R. Sonic hedgehog signaling regulates reciprocal
epithelial-mesenchymal interactions controlling palatal outgrowth.
44. Jeong J, Mao J, Tenzen T, et al. Hedgehog signaling in the neural
crest cells regulates the patterning and growth of facial primordia.
Genes Dev 2004;18:937–951.
45. van den Brink GR, Bleuming SA, Hardwick JC, et al. Indian
Hedgehog is an antagonist of Wnt signaling in colonic epithelial
cell differentiation. Nat Genet 2004;36:277–282.
46. Kolterud A, Grosse AS, Zacharias WJ, et al. Paracrine Hedgehog
signaling in stomach and intestine: new roles for hedgehog in
gastrointestinal patterning. Gastroenterology 2009;137:618–
47. Kosinski C, Stange DE, Xu C, et al. Indian hedgehog regulates
intestinal stem cell fate through epithelial-mesenchymal interac-
tions during development. Gastroenterology 2010;139:893–903.
48. Madison BB, Braunstein K, Kuizon E, et al. Epithelial hedgehog
signals pattern the intestinal crypt-villus axis. Development 2005;
49. van Dop WA, Uhmann A, Wijgerde M, et al. Depletion of the colonic
epithelial precursor cell compartment upon conditional activation
of the hedgehog pathway. Gastroenterology 2009;136:2195–
Author names in bold designate shared co-first authorship.
Received April 16, 2012. Accepted January 22, 2013.
Address requests for reprints to: Peter Carlsson, PhD, Department
of Chemistry and Molecular Biology, University of Gothenburg, Box
462, 405 30 Gothenburg, Sweden. e-mail: firstname.lastname@example.org;
fax: (46) 31 7863801.
The authors thank Naoyuki Miura for the Foxf2 knockout strain,
Mozhgan Ghiami for technical assistance, and the staff at the
University of Gothenburg core facility for cellular imaging for
assistance with 3-dimensional reconstruction of crypts.
Conflicts of interest
The authors disclose no conflicts.
Supported by the Swedish Cancer Foundation and the Swedish
Medical Research Council.
May 2013 Foxf2 INHIBITS INTESTINAL Wnt SIGNALING1011