Content uploaded by Runjan Chetty
Author content
All content in this area was uploaded by Runjan Chetty on Mar 02, 2015
Content may be subject to copyright.
Articles
62 VOLUME 21 | NUMBER 1 | JANUARY 2015 nAture medicine
The intestinal mucosa is covered by a self-renewing layer of epithe-
lium, making it ideal for the study of tissue-specific stem cells and
cell fate determination. Lineage-tracing experiments have helped
identify genes selectively expressed by intestinal stem cells. One of
these marker genes, the Wnt target, leucine-rich-repeat–containing
G protein–coupled receptor 5 (Lgr5) is expressed in crypt base
columnar cells (CBCs) within the crypt base stem cell niche, which
also comprises surrounding Paneth cells and intestinal subepithelial
myofibroblasts1. In homeostasis, cell fate determination is coupled to
position along the crypt-villus (vertical) axis of the epithelium. Stem
cell progeny exit the niche, initially as proliferating transit-amplifying
cells before progressively differentiating into post-mitotic specialized
cells. This is controlled by strict gradients of interacting morphogens—
soluble molecules produced by a restricted region of a tissue that
form an activity gradient away from source. The phenotypic response
of a cell is determined by its position within this concentration
gradient2. Wnt and bone morphogenetic protein (BMP) pathways
form polarized expression gradients along the epithelial vertical
axis. Stem cell division and transit-amplifying cell proliferation are
driven by high Wnt and low BMP levels in the lower half of the crypt,
whereas daughter cell differentiation and apoptosis are controlled by
low Wnt and high BMP at the luminal surface3. These gradients are
maintained partly by diffusion of ligands but also by the restricted
paracrine secretion of ligand-sequestering BMP antagonists, such as
Gremlin1, Gremlin2 and Noggin, that are exclusively derived from
subcrypt myofibroblasts and act locally within the crypt base stem cell
niche. These antagonists are thought to prevent BMP activity within
the niche, promoting intestinal stem cell function4.
Dysregulation of the homeostatic Wnt-BMP balance can promote
intestinal tumorigenesis. The conventional adenoma-carcinoma
sequence is commonly initiated by activation of Wnt signaling in the
epithelium through adenomatous polyposis coli (APC) or β-catenin
(CTNNB1) mutation5. However, disrupted BMP signaling can also
predispose to intestinal polyps and cancer6. Human juvenile polyposis
syndrome (JPS) results from inactivating germline bone morphogenetic
protein receptor 1a (BMPR1A) or Mothers against decapentaple-
gic homolog 4 (SMAD4) mutations, and epithelial expression of
Nog under the control of villin (Vil1) or fatty-acid–binding protein
(Fabp1) regulator y elements causes a JPS-like phenotype in mice7,8.
Recently, we demonstrated that human HMPS is caused by a 40-kb
1Gastrointestinal Stem Cell Biology Laboratory, Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, UK. 2Department of Systems Biology,
Columbia University Medical Center, New York, New York, USA. 3Colorectal Cancer Genetics, Centre for Digestive Diseases, Blizard Institute, Barts and the
London School of Medicine and Dentistry, Whitechapel, London, UK. 4Molecular and Population Genetics Laboratory, Wellcome Trust Centre for Human Genetics,
University of Oxford, Oxford, UK. 5Polyposis Registry, St. Mark’s Hospital, Harrow, UK. 6Department of Histopathology, University College London Hospital, London, UK.
7Laboratory Medicine Program, University Health Network and University of Toronto, Toronto, Ontario, Canada. 8Beatson Institute for Cancer Research, Garscube
Estate, Bearsden, Glasgow, UK. 9Georg-Speyer-Haus Institute for Tumor Biology and Experimental Therapy, Frankfurt, Germany. 10Cellular Pathology, Level 1,
John Radcliffe Hospital, Headington, Oxford, UK. 11Translational Gastroenterology Unit, Experimental Medicine Division, Nuffield Department of Clinical Medicine,
John Radcliffe Hospital, Headington, Oxford, UK. 12Oxford National Institute for Health Research Comprehensive Biomedical Research Centre, Wellcome Trust
Centre for Human Genetics, Oxford, UK. 13These authors contributed equally to this work. Correspondence should be addressed to I.T. (ian.tomlinson@well.ox.ac.uk)
or S.J.L. (simon.leedham@well.ox.ac.uk).
Received 21 May; accepted 17 October; published online 1 December 2014; doi:10.1038/nm.3750
Aberrant epithelial GREM1 expression initiates colonic
tumorigenesis from cells outside the stem cell niche
Hayley Davis1,13, Shazia Irshad1,13, Mukesh Bansal2, Hannah Rafferty1, Tatjana Boitsova1,3, Chiara Bardella4,
Emma Jaeger4, Annabelle Lewis4, Luke Freeman-Mills4, Francesc C Giner4, Pedro Rodenas-Cuadrado1,
Sreelakshmi Mallappa5, Susan Clark5, Huw Thomas5, Rosemary Jeffery3, Richard Poulsom3,
Manuel Rodriguez-Justo6, Marco Novelli6, Runjan Chetty7, Andrew Silver3, Owen J Sansom8, Florian R Greten9,
Lai Mun Wang10, James E East11, Ian Tomlinson4,12 & Simon J Leedham1,11
Hereditary mixed polyposis syndrome (HMPS) is characterized by the development of mixed-morphology colorectal tumors and
is caused by a 40-kb genetic duplication that results in aberrant epithelial expression of the gene encoding mesenchymal bone
morphogenetic protein antagonist, GREM1. Here we use HMPS tissue and a mouse model of the disease to show that epithelial
GREM1 disrupts homeostatic intestinal morphogen gradients, altering cell fate that is normally determined by position along the
vertical epithelial axis. This promotes the persistence and/or reacquisition of stem cell properties in Lgr5-negative progenitor cells
that have exited the stem cell niche. These cells form ectopic crypts, proliferate, accumulate somatic mutations and can initiate
intestinal neoplasia, indicating that the crypt base stem cell is not the sole cell of origin of colorectal cancer. Furthermore, we
show that epithelial expression of GREM1 also occurs in traditional serrated adenomas, sporadic premalignant lesions with a
hitherto unknown pathogenesis, and these lesions can be considered the sporadic equivalents of HMPS polyps.
npg © 2015 Nature America, Inc. All rights reserved.
Articles
nAture medicine VOLUME 21 | NUMBER 1 | JANUARY 2015 6 3
duplication upstream of the BMP antagonist GREM1, which results in
ectopic epithelial gene expression and resultant BMP signaling antag-
onism throughout the vertical axis of the intestine9 (Supplementary
Fig. 1c–e). HMPS is an autosomal dominant condition, and untreated
individuals develop colorectal cancer at a median age of 47 (ref. 10).
HMPS is named for the distinctive morphology of the polyps, with
individual lesions exhibiting mixed adenomatous cr ypts, epithelial
serration and dilated cysts (Fig. 1a).
Although it has recently been shown that scarce, post-mitotic tuft
cells can persist outside the stem cell niche11, the majority of differ-
entiated cells (enterocytes, colonocytes and goblet cells) are shed into
the lumen within 5 d. As a consequence of this rapid cell turnover,
the perpetual stem cell at the crypt base has been considered the cell
of origin of colorectal cancer (CRC)12. Here, we use a mouse model
of HMPS to show that disruption of homeostatic BMP gradients by
aberrant epithelial expression of Grem1 alters cell fate determination,
allowing cells outside the crypt base stem cell niche to act as tumor
progenitors. Furthermore, we demonstrate that this is the pathogenic
mechanism underpinning the development of human HMPS polyps
and some sporadic intestinal tumors.
RESULTS
HMPS polyps are characterized by ectopic crypt foci formation
All crypts in individuals with HMPS have epithelial GREM1 expression9,
yet the polyps are discrete, often containing mixed dysplastic
and nondysplastic areas. Histopathological review of the polyps
revealed ectopic crypt foci (ECFs) that developed orthogonally to
the crypt axis and contained actively proliferating cells (Fig. 1a–c
and Supplementary Fig. 1b). Within some polyps, we identified
dysplastic cells emerging from ECFs rather than from the crypt base
(Fig. 1b) and hypothesized that dysplasia resulted from somatic
mutations within the ECFs.
Sanger sequencing of candidate genes in 23 polyps from 14 patients
with HMPS revealed a high frequency of CRC driver mutations. We
observed mutually exclusive KRAS or BRAF mutations in 100% of
lesions and APC, predominantly p.Arg1450X, mutations in 48%. The
CpG island methylator phenotype (CIMP)13 was present in 53% of
lesions tested (Fig. 1d and Supplementary Fig. 2a). In contrast, we
found a very low frequency of known driver mutations in a cohort of JPS
polyps with germline BMPR1A mutations (Supplementary Fig. 2b).
Clonal ordering following microdissection of individual HMPS crypts
revealed clonal KRAS or BRAF mutations detected throughout entire
polyps, including ECFs and the different morphological crypt subtypes
(Supplementary Fig. 2d,e). In contrast, APC mutations were often
spatially restricted to dysplastic areas, indicating that Wnt dysregulation
was a subsequent event (Fig. 1e and Supplementary Fig. 2a,c).
A mouse model of HMPS
In order to understand more about the pathogenesis of HMPS, we
generated Vil1-Grem1 mice expressing mouse Grem1 cDNA under the
control of the intestinal epithelium–specific Vil1 promoter. Epithelial
expression of Grem1 was confirmed by in situ hybridization (Fig. 2c)
and quantitative RT-PCR (qRT-PCR), with highest levels in the proxi-
mal small bowel (Supplementary Fig. 3d). We assessed BMP signaling
using BMP ligand and target gene expression (Supplementary Fig. 3e)
and immunohistochemistry for phosphorylated Smad1, Smad5 and
Smad8. There was loss of the normal p-Smad1, p-Smad5 and p-Smad8
staining pattern throughout the vertical axis of the intestines of
Vil1-Grem1 mice (Fig. 2d and Supplementary Fig. 4b).
Transgenic animals’ small intestines were 28% longer than those
of their wild-type littermates (n = 10, P < 0.001, t-test), partly owing
to an increase in the size and cell count of the villi (P < 0.001, t-test)
(Fig. 2a and Supplementar y Fig. 4a). Although there was no change
in the proportion of Ki-67–positive proliferating cells in the crypts,
a b c
HMPS polyps
pSMAD1,5,8 Ki-67 Lysozyme
SOX9 EPHB2 CK20
HMPS polyps
Adenomatous
Serrated
Dilated
cyst
3
1 cm
10
9
4
2
1
6
5
7811
12
13
Morphology
Crypt no.
Normal Low-grade dysplasia
WT
WT WT
p.G12D
p.R1450X
WT WT
MSS MSS
CIMP+CIMP+
KRAS
BRAF, TP53
PTEN, PIK3CA
Microsatellite
instability
Methylator
phenotype
APC
1 2 3 4 5 6 7 8 9 10 11 12 13
ed
80
70
60
50
40
Mutation frequency (%)
30
20
10
0
CIMP
+
KRAS
G13D
E1322X
S1465fs3
APC
BRAF
G12D
R1450X
V600E
Figure 1 Human HMPS polyps. (a) H&E staining of an
HMPS polyp showing mixed adenomatous, serrated and
dilated cyst morphology and close up of ectopic crypts
growing orthogonally to crypt axis. (b) Dysplastic cells
(black arrowhead) emerging from an ectopic crypt rather
than from the crypt base. (c) Top, immunostaining of HMPS
polyps showing patchy loss of phosphorylated SMAD1,
SMAD5 and SMAD8 (p-SMAD1,5,8) stain, Ki-67 stain in
proliferating ectopic crypt foci cells and ectopic lysozyme
stain in dysplastic crypts. Bottom, SOX9 and EPHB2
immunostaining is increased whereas staining for the
differentiation marker CK20, is lost in the ectopic crypt
foci of HMPS polyps (n = 23 polyps for all stains).
(d) Candidate gene (epi)genetic mutation spectra in HMPS
polyps. (e) Laser-capture isolation of individual crypts across
HMPS lesions. Spatial distinction of mutant clones allowed
inference of mutation timing (see also Supplementary Fig. 2).
Scale bars are 100 µm unless otherwise stated.
npg © 2015 Nature America, Inc. All rights reserved.
Articles
64 VOLUME 21 | NUMBER 1 | JANUARY 2015 nAture medicine
there were significantly more proliferating cells on the villi of Vil1-
Grem1 animals compared to wild-type littermates (P = 0.038, t-test,
Fig. 2a,d). Analysis of epithelial cell lineages showed a decrease in the
number of goblet cells in both small intestinal (P = 0.01, t-test) and
colonic (P = 0.002, t-test) crypts. The presence of lysozyme-positive
Paneth cells on the villus (P = 0.038, t-test) of Vil1-Grem1 mice was
notable as this is a cell type normally restricted to the small intestinal
crypt base (Fig. 2a,d). In 3-month-old Vil1-Grem1 animals, Grem1-
expressing ectopic crypts were seen developing orthogonally to the
vertical axis of the widened and flattened small bowel villi (Fig. 2b,c).
Ectopic crypts budded off to become actively proliferating intravillus
lesions, which subsequently developed dysplastic features with con-
comitant loss of p16INK4A expression (Fig. 2e). By 7 months, these
lesions had progressed to a pan-intestinal polyposis (Supplementary
Fig. 4c), with a median of 183 polyps per mouse. Small intestinal
lesions had a mixed serrated, adenomatous and cystic phenotype
characteristic of the lesions seen in HMPS (Fig. 2b).
We observed epithelial Grem1 and membranous β-catenin expres-
sion in unaffected intestine and small polyps in the Vil1-Grem1 mice
but some larger polyps with more advanced dysplasia exhibited
marked downregulation of epithelial Grem1 expression, which cor-
related with foci of cytoplasmic and nuclear β-catenin staining. Sanger
sequencing demonstrated activating Ctnnb1 mutations in some of
these lesions (Fig. 2f and Supplementary Fig. 4d).
Although the total colonic length was unchanged in Vil1-Grem1
mice, the proximal colonic folds were exaggerated in comparison with
wild-type littermates. In Vil1-Grem1 mice hyperplastic-appearing
colonic crypts had an increased cell count (P = 0.03, t-test, Fig. 2a).
Crypt crowding meant that ECFs could not be easily distinguished
in the colon, but colonic dysplasia originated at the luminal surface
and progressed to form lesions that contained all three morphological
crypt phenotypes (Fig. 2b).
These data indicate that aberrant epithelial Grem1 expression results
in a progressive intestinal polyposis with small intestinal and colonic
a b Wild type
Adenomatous
Adenomatous
Serrated
Serrated
Dilated
cyst
Dilated
cyst
Small bowelColon
Vil1-Grem1
d
Wild type
p-Smad1,5,8 Ki-67 Lysozyme Sox9 EphB2 Ck20
Vil1-Grem1
c
Small bowelColon
Wild type Vil1-Grem1
Grem1 ISH
e
Ki-67 p16INK4A
H&E
f
c.122 C>T,p.T41I
Ctnnb1 mutation
Grem1 ISH
Grem1 ISH β-catenin
60
Intestinal length
P <
0.001
40
Intestine length (cm)
20
0
Small bowel
Colon
Wild type Vil1-Grem1
30
Paneth cells
P =
0.03
20
% Lysozyme+ cells
10
0
SB crypt
SB villus
Proliferating cells
P =
0.038
100
80
% Ki-67+ cells
20
40
60
0
SB crypt
SB villus
Colon crypt
140
120
Crypt/villus cell count
P < 0.001
P =
0.03
100
80
Cell number (SEM)
20
40
60
0
SB crypt
SB villus
Colon crypt
P = 0.01
Goblet cells
80
% Alcian blue+ cells
20
40
60
0
SB crypt
SB villus
Colon crypt
P <
0.002
Figure 2 Vil1-Grem1
mouse phenotype.
(a) Macroscopic and
microscopic phenotyping
of Vil1-Grem1 versus
wild-type mouse intestine.
Significant differences
between Vil1-Grem1 mice
and their wild-type
littermates were noted in
intestinal length and diameter
(n = 10 mice per group, P < 0.001, t-test),
villus (n = 50 villi per group) and colonic crypt
(n = 100 crypts per group) cell count (P = 0.001,
t-test), villus proliferating cell proportion (n = 50 villi
per group, P = 0.03, t-test), small bowel (n = 50
crypt-villus units per group, P < 0.002, t-test) and
colonic (n = 50 crypts per group, P = 0.01, t-test)
goblet cell count and proportion of lysozyme-positive
Paneth cells on the villus (n = 50 villi per group,
P = 0.038, t-test). In all cases, Student’s t-test using
two-tailed, unpaired and unequal variance was employed.
The data from each group did not significantly deviate
from a normal distribution (Shapiro-Wilk test). Data represent means ± s.e.m. (b) Top left, wild-type mouse small bowel 1 (SB1). Top middle, SB1 of
a 3-month-old Vil1-Grem1 mouse, with widened villi containing intravillus ectopic crypts (black arrowheads). Top right, dysplastic polyp formation in
a 7-month-old Vil1-Grem1 animal exhibiting mixed morphology with serrated (inset), adenomatous and dilated cyst phenotypic regions. Bottom left,
wild-type mouse colon. Bottom middle, early colonic lesion with luminal surface dysplasia distant from the crypt basal stem cell niche. Bottom right,
colonic polyps in a 7-month-old mouse with mixed crypt morphology (serrated crypts, inset). (c) In situ hybridization (ISH) for mouse Grem1 with normal
intestinal expression of Grem1 exclusively from the subcrypt myofibroblasts (black arrowheads). Aberrant epithelial expression is seen in early small
intestinal and colonic lesions from the Vil1-Grem1 mouse. (d) Immunohistochemical analysis of Vil1-Grem1 versus wild-type small intestine shows
loss of p-Smad1,5,8 throughout the crypt-villus axis (n = 10 mice), with Ki-67, Sox9 and EphB2 staining all present in the villus ectopic crypts. Ck20
differentiation marker staining was lost in villus ECFs (n = 135 polyps for all stains). (e) Dysplasia arising in intravillus ectopic crypts (black arrowhead)
with active cell proliferation correlating with p16Ink4A stain loss. (f) Grem1 ISH in Vil1-Grem1 mouse tissue shows loss of Grem1 expression in a large
polyp, which correlates with nuclear β-catenin staining, resulting from a Ctnnb1 p.T41I mutation. Scale bars are 100 µm.
npg © 2015 Nature America, Inc. All rights reserved.
Articles
nAture medicine VOLUME 21 | NUMBER 1 | JANUARY 2015 6 5
polyps containing the three characteristic
morphologies seen in human HMPS lesions.
Throughout the bowel, early lesions can be
seen developing outside the crypt basal stem
cell niche: on the luminal surface in the colon
and within ECFs that bud into the villus of
the small intestine. Downregulation of epi-
thelial expression in some advanced lesions
indicates that epithelial Grem1 is no longer
required once epithelial somatic mutation
events have occurred.
Vil1-Grem1 mice have enlargement of the progenitor cell pool
As the villus ECFs appeared to be the origin of small intestinal dysplasia
in the Vil1-Grem1 mice, we looked for increased expression of stem
cell markers on the villi of these animals. We crossed Vil1-Grem1
mice with Lgr5-EGFP-IRES-CreERT2 reporter mice1 but were unable
to detect discrete Lgr5-EGFP–positive cells outside of the crypt base
stem cell niche, with none seen in the villus ECFs (Fig. 3a). To confirm
this, we mechanically separated crypts and villi from Vil1-Grem1 and
age-matched wild-type mice and used qRT-PCR to detect stem cell
markers aberrantly expressed in the Vil1-Grem1 mouse villus com-
partment. Of the stem cell markers tested, Sox9 showed the highest
expression in villus cells (Fig. 3b), and immunostaining for Sox9 con-
firmed ectopic crypt-specific expression in both human HMPS and
Vil1-Grem1 mouse tissue (Figs. 1c and 2d).
Next, we examined the global mRNA expression profiles of
Vil1-Grem1 versus wild-type mouse crypt and villus compartments
using gene set enrichment analysis (GSEA). GSEA confirmed no sig-
nificant enrichment of Lgr5-based intestinal stem cell profiles14,15
on the villi of Vil1-Grem1 transgenic mice. In contrast, there was
enrichment of villus gene programs characterizing proliferating early
transit-amplifying cells15 (normalized enrichment score (NES) 7.35,
P < 0.001, Kolmogorov-Smirnov test) with concomitant reduced levels
of genes normally expressed in differentiating cells15 (NES −6.46;
P < 0.001). Furthermore, we also observed reduced expression of
villus genes regulating apoptosis16 (NES −2.9; P = 0.008), cellular
senescence17 (NES −2.36; P = 0.026) and autophagy18 (NES −3.89;
P < 0.001) (Fig. 3c), homeostatic processes that have all been shown
to have tumor suppressor roles in early lesions19,20.
A decreasing gradient of ephrin type-B receptor 2 (EphB2) expres-
sion from the crypt base along the vertical axis of the intestine has
been used to distinguish intestinal stem (EphB2high), progenitor
(EphB2medium) and differentiating (EphB2low/absent) cell populations in
the mouse and human intestine15. Consistent with this and our GSEA
findings, EphB2 protein was aberrantly expressed in the ECFs of both
Vil1-Grem1 transgenic mouse and human HMPS polyps (Figs. 1c and
2d). Protein expression of cytokeratin 20 (Ck20), a marker of differ-
entiated cells, was reduced or absent in the ECFs of both Vil1-Grem1
mouse and human HMPS polyps (Figs. 1c and 2d).
Taken together, these data indicate that epithelial Grem1 expres-
sion disrupts the coupling of cell fate to position along the vertical
axis of the intestine. Although cell fate–position uncoupling does not
generate an ectopic Lgr5-positive stem cell population, it does cause
expansion of an Lgr5-negative proliferating progenitor cell population
on to the villi. Concomitant downregulation of differentiation, apop-
tosis, senescence and autophagy gene programs means there is a
marked expansion of a progenitor cell pool outside of the intestinal
stem cell niche that can be defined immunohistochemically as
Sox9+EphB2+Ck20− (Figs. 1c and 2d).
Epithelial Grem1 promotes persistence of somatically mutant
villus cells
In order to test the clonogenic, tumor-forming potential of villus
cells in our Vil1-Grem1 mouse model, we used the established
in vitro enteroid technique that uses media supplemented by the
niche-derived morphogens, epidermal growth factor (E), Noggin
(N) and R-spondin (S) (ENS medium)21. Normal-appearing crypts
Apoptosis
0.2
0.1
0
NES –2.9
P = 0.008
ES –0.1
–0.2
–0.3
–0.4
–0.5
0 9 18
Gene rank (×1,000)
V-G1
villi
WT
villi
Senescence
0.3
0.2
0.1
0
–0.1
–0.2
–0.3
–0.4
NES –2.36
P = 0.026
0 9 18
Gene rank (×1,000)
V-G1
villi
WT
villi
Autophagy
NES –3.89
P < 0.001
0.2
0.1
0
–0.1
–0.2
–0.3
–0.4
–0.5
0 9 18
Gene rank (×1,000)
V-G1
villi
WT
villi
Intestinal stem cells
0.2
c
0.1
0
NES 0.5
P = 0.44
ES
–0.1
–0.2
0 9 18
Gene rank (×1,000)
V-G1
villi
WT
villi
Transit-amplifying cells
0.6
0.5
0.4
0.3
0.2
0.1
0
NES 7.35
P < 0.001
0 9 18
Gene rank (×1,000)
V-G1
villi
WT
villi
Differentiating cells
0
–0.2
–0.4
–0.6
–0.8
NES –6.46
P < 0.001
0 9 18
Gene rank (×1,000)
V-G1
villi
WT
villi
a b
Anti-GFP 100
P = 0.01
Ascl2
Lgr5
Olfm4
Bmi1
Dclk1
Cd24
Cd44
Rnf43
Sox9
Lrig1
10
1
0.1
Vil1-Grem1 (vs. wild type) villi gene
expression (fold change, log scale)
0.01
Figure 3 Gene expression analysis of separated
villus compartments in Vil1-Grem1 mice.
(a) Intestinal tissue from Lgr5-EGFP-IRES-
CreERT2;Vil1-Grem1 mice stained with anti-
GFP antibody. Lgr5-expressing EGFP-positive
cells were found in the underlying crypt bases
(black arrowheads and inset), but none were
seen in the ectopic crypts on the villi (white
arrowheads). Scale bars are 100 µm.
(b) qRT-PCR analysis of villus stem cell marker
expression, There was a significant increase in
expression of Sox9 in individual Vil1-Grem1
villi versus wild-type littermate villi (n = 10,
P = 0.01, t-test). (c) Villi from Vil1-Grem1 (V-
G1) and wild-type littermates underwent gene
expression microarray and GSEA analysis using
established gene program sets. GSEA plots
shown are for Vil1-Grem1 (n = 5) versus wild-
type (n = 6) villi. Enrichment score is calculated
using Kolomogrov-Smirnov test36. P values were
calculated using a permutation test.
npg © 2015 Nature America, Inc. All rights reserved.
Articles
66 VOLUME 21 | NUMBER 1 | JANUARY 2015 nAture medicine
from Vil1-Grem1 mice readily formed branching enteroids not only
in these standard conditions but also in the absence of Noggin sup-
plementation (ES medium). Endogenous epithelial Grem1 expres-
sion could be overcome by addition of competing BMP ligands 2, 4
and 7 to medium without Noggin (ES + rBMP2,4,7), which inhibited
Vil1-Grem1 enteroid development, indicating that the culture of
non-dysplastic crypts was dependent on a source of BMP antagonist
(Fig. 4a). In contrast, culture of dysplastic mouse adenomas (ApcMin/−)
generated nonbranching clonogenic cystic spheroids, which could be
successfully initiated and propagated in the absence of BMP antagonist
or following addition of recombinant BMP ligands (Fig. 4a). These
data further support the notion that BMP antagonism is dispensable
once Wnt-activating somatic mutations have occurred12,22.
Next, we separated and cultured villi from Vil1-Grem1 and
age-matched wild-type littermates in order to assess whether aber-
rant epithelial Grem1 expression influenced the tumor-forming
capacity of villus cells. In the presence of epidermal growth factor,
Noggin, R-spondin and recombinant Wnt3a (W) (ENSW medium),
clonogenic cystic spheroids did develop from Vil1-Grem1 mouse
villi, but these were rare events (<0.1% of villi). We reasoned that
the inefficiency of this transformation resulted from the short
half-life of Wnt3a in the medium23. To counter this, we generated
Vil1-Grem1; ApcMin/+ mice to activate the endogenous Wnt pathway.
In extracted villi, germline ApcMin/+ mutation did induce an increase
in villus Wnt target gene expression (Fig. 4b). However, epithelial
Grem1 reduced the expression of these Wnt targets in an Apc-mutant
background while increasing the expression of the progenitor
markers Sox9 and Ephb2 (Fig. 4b). These results were consistent
with our GSEA findings.
We then repeated villus culture taking care to sample from regions
without microscopically visible polyps. Villi from wild-type and
age-matched ApcMin/+ mice did not sur vive in culture. By contrast,
villi from Vil1-Grem1; ApcMin /+ animals rapidly formed clono-
genic spheroids that could be successfully propagated in long-term
culture and were unaffected by the addition of competing BMP lig-
and to the media (Fig. 4a). Somatic loss of heterozygosity of the
wild-type Apc allele was seen in all spheroids, and this was coupled
with a further increase in Wnt target gene expression (Fig. 4b–d).
ES + rBMP2,4,7
ENS
ENSW
Branching enteroid
Branching enteroid
Branching enteroid
Branching enteroid Spheroid
Spheroid
Spheroid
Spheroid
Spheroid
Spheroid
Spheroid
SpheroidSpheroidBranching enteroid
ES
c
e
d
Media
supplementation
Wild-type
crypts Wild-type villi
Mouse genotype and tissue compartment
Vil1-Grem1
crypts
Vil1-Grem1;
ApcMin/+ villi
Day 0 Day 1
Matrigel and ES media
Day 3 Day 7
ApcMin/+
villi
Vil1-Grem1 villi Vil1-Grem1; Apc
Min
villi
Apc
Min
polyps Apc
Min/+
villi
Villus spheroids
β-catenin EphB2 Sox9
23 d 30 d 40 d 47 d
Vil1-Grem1; ApcMin/+ mouse age
ApcMin allele (144 bp)
Wild-type allele (123 bp)
V-G1; ApcMin/– spheroid
V-G1; ApcMin/– spheroid
V-G1; ApcMin/+ villus
V-G1; ApcMin/+ villus
Wild type
80
60
40
20
% villi generating
spheroids
0
23 d 30 d 40 d 47 d
b
105
Lgr5
104
103
102
Villus expression fold change (log scale)
10
1
0.1
Ccnd1
103
102
10
1
0.1
Sox9
103
102
10
1
0.1
Ephb2
103
102
10
1
0.1
Ascl2
106
107
105
104
103
102
10
1
0.1
Axin1
102
10
1
0.1
0.01
Wild type Vil1-Grem1 ApcMin/+ Vil1-Grem1; ApcMin/+ Spheroids (Vil1-Grem1; ApcMin/–)
a
Figure 4 In vitro villus cell clonogenicity. (a) Heatmap showing in vitro culture of different tissue compartments taken from different genotype mice in
varying medium conditions. Blue boxes indicate successful culture of structures with the described morphology. Red boxes indicate failure to establish
tissue culture. ENS medium was considered standard conditions. Illustrated results are representative of three separate experiments. Cell culture
images show development of spheroid structures from Vil1-Grem1; ApcMin/+ villi with dissociation of villi from age matched ApcMin/+ mice. (b) qRT-PCR
analysis of effect of mouse genotype on villus Wnt target gene expression (versus wild type) showed that a single Apc hit was sufficient to increase
endogenous individual villus Wnt target gene expression, including the stem cell markers Lgr5 and Ascl2. Expression of the progenitor cell markers
Sox9 and Ephb2 was significantly increased in Vil1-Grem1 animals (n = 10 villi for each genotype, P < 0.01, t-test), and emergent spheroids had a
further increase in both Wnt target and progenitor marker genes. Data represent mean log fold change ± s.e.m. (c) Villus spheroid immunostain showing
nuclear β-catenin, membranous EphB2 and nuclear Sox9 staining (images are representative of 8 spheroids in each group). (d) Extracted nondysplastic
villi from Vil1-Grem1 (V-G1); ApcMin/+ mice entering into culture retained a residual wild-type Apc allele, whereas emergent spheroids had lost the
wild-type allele. (e) Percentage of villi transforming into clonogenic spheroids with increasing age of Vil1-Grem1/ApcMin/+ mice. n = 300 villi from three
different experiments. Data represent means ± s.e.m. Histology review showed that this increase correlated with the emergence of villus ectopic crypts
(black arrowheads). Scale bars are 100 µm.
npg © 2015 Nature America, Inc. All rights reserved.
Articles
nAture medicine VOLUME 21 | NUMBER 1 | JANUARY 2015 6 7
Spheroids formed from mice as young as 23 d old, but transforma-
tion efficiency increased with the age of the mouse, and histological
analysis showed that this coincided with the emergence of villus
ECFs (Fig. 4e).
Taken together, these results suggest that modest Wnt activation
induced by a single Apc mutation alone is insufficient for the clonogenic
growth of villus spheroids in nondysplastic ApcMin/+ villi. However, epi-
thelial Grem1-induced disruption of cell fate allows the persistence of
progenitor cells on the villus, and if Wnt-activating somatic mutations
are present within this accumulating, expanded progenitor population,
these cells are capable of generating BMP antagonist–independent
spheroid growth in the cell culture environment.
Grem1 and activated Wnt signaling act in a synergistic fashion
Crossing Vil1-Grem1 mice with ApcMin/+ mice caused a profound
exacerbation of intestinal tumorigenesis, with tumor burden in
2-month-old double-transgenic animals greater than that seen in ani-
mals with the individual parental mutations (Fig. 5a). Vil1-Grem1;
ApcMin/+ mice had to be killed at mean 57 d as opposed to >200 d for
ApcMin/+ mice and >250 d for Vil1-Grem1 animals. Notably, morpho-
logical elements of the polyps from both parental strains could be seen
in individual lesions in double-transgenic animals. In the colon, super-
ficial aberrant crypt foci progressed to lesions with luminal surface,
not crypt base dysplasia. In the small bowel, dysplasia arose from
intravillus ectopic crypts contained within nondysplastic serrated
epithelium (Fig. 5b and Supplementary Fig. 5a). Laser dissection
of these different morphological elements demonstrated somatic
loss of the wild-type Apc allele exclusively in the dysplastic crypts
within the serrated villi (Fig. 5b), indicating that somatic Apc inac-
tivation provides a selective advantage and occurs rapidly in actively
proliferating villus cells.
Grem1 knockout reduces Wnt-driven tumor progression
Next, we induced widespread, multicompartmental Grem1 knock-
out in 6-week-old mice. To do this we used a tamoxifen-inducible,
Cre-mediated recombination system driven by the chicken β-actin
promoter and cytomegalovirus early enhancer, crossed with a
homozygous Grem1-floxed mouse (Cagg-CreERT2; Grem1fl/fl mice)24.
We confirmed successful Grem1 knockout by qRT-PCR and in situ
hybridization, but Cagg-CreERT2; Grem1fl/fl animals had no consistent
change in the expression of BMP constituents, Wnt targets or stem
cell markers and developed no pathological phenotype, indicating
possible functional redundancy or buffering of BMP antagonists in
adult intestinal homeostasis (Supplementar y Fig. 6). As we and oth-
ers have seen epithelial and/or stromal upregulation of GREM1 in
some, but not all, sporadic human intestinal polyps, cancers and cell
lines25–27 (Supplementary Figs. 5d and 7a), we examined the effect of
knockout of physiological Grem1 expression on Wnt-initiated tumor
burden by crossing Cagg-CreERT2; Grem1fl/fl mice with ApcMin mice.
Grem1 knockout caused a significant reduction of ApcMin/+ mouse
polyp burden (Pinteraction < 0.002 for all regions of the bowel, linear
regression, Fig. 5c) and size (Pinterac tion < 0.001, linear regression,
Vil1-Grem1; Apc
Min/+
b
Wild-type allele (123 bp)
Wild type
Apc
Min
allele (144 bp)
Serrated Dysplastic
Apc
Min
polyp
V-G1
;
Apc
Min/+
serrated
V-G1
;
Apc
Min/–
dysplastic
d
100
P = 0.0162
80
60
40
4
Low GREM1
AMC-AJCCII-90 set, stage 2 CRC
High GREM1
6 8 10
Time (years)
Disease-free
survival (%)
0 2
8
Log expression
GREM1
The Cancer Genome
Atlas, RNA-seq
CCS2 CCS3
n = 54
n = 36
n = 104
5.09
6.02
7.01
P < 0.0001
CCS1
Sporadic CRC subtype
7
6
5
4
3
2
1
f
TSA - GREM1 ISH
e
100
Epithelial gene expression
TSA vs. normal (log scale)
No epithelial expression
BAMBI
CHRDL2
DAND5
FST
GREM1
GREM2
NOG
SOSTDC1
CHRDL1
CHRD
No epithelial expression
No epithelial expression
10
1
0.1
0.01
a
40
Apc
Min/+
alone
Vil1-Grem1; Apc
Min/+
Vil1-Grem1
30
20
10
0
SB1
Polyp count (mean 57 d)
SB2 SB3 Colon
Apc
Min/+
alone
CAGG-CreER
T2
; Grem1
fl/fl
; Apc
Min/+
c
40
30
20
10
0
Polyp count (mean 248 d)
SB1 SB2
P = 0.03
P = 0.027
SB3 Colon
Figure 5 Effect of Grem1 on conventional
Wnt-driven tumorigenesis and pathogenic
role in human sporadic TSAs. (a) Intestinal
polyp burden of Vil1-Grem1; ApcMin/+ and
parental single-transgenic strains at mean 57 d
(Vil1-Grem1 n = 5; ApcMin/+ n = 7; Vil1-Grem1; ApcMin/+ n = 8,
Pinteraction < 0.002 for all regions of the bowel, generalized linear
regression incorporating a multiplicative interaction term between
Apc mutation and Grem1 status). Polyp size was also significantly
greater in Vil1-Grem1/ApcMin/+ animals (Pinteraction < 0.001, linear
regression, data not shown). The data from each group did not
significantly deviate from a normal distribution (Shapiro-Wilk test).
Data represent means ± s.e.m. (b) Histology of Vil1-Grem1 (V-G1);
ApcMin/+ mouse polyps with central dysplastic areas and a sharp
cutoff between enclosing serrated epithelium (enlarged). Laser dissection
of the different morphological types with loss of the wild-type allele in
dysplastic epithelium (representative of 6 different dissected polyps). (c) Intestinal polyp burden of CAGG-CreERT2; Grem1; ApcMin/+ and age-matched
ApcMin/+ mice at mean 248 d (n = 4 mice for test and noninjected control groups, P = 0.027, t-test unpaired with unequal variances). The data from
each group did not significantly deviate from a normal distribution (Shapiro-Wilk test). Data represent means ± s.e.m. (d) Top, expression of GREM1
in the AMC-AJCCII-90 human CRC set and correlation with disease-free survival (differences between above- and below-median GREM1-expressing
tumors, P = 0.0162, log-rank test). Bottom, classification of the Cancer Genome Atlas RNA-seq data into three CCSs28 (correlation between CCS3-
subtype cancers and high whole-tumor GREM1 expression, P < 0.0001, ANOVA). (e) qRT-PCR measurement of known BMP antagonists from individual
fresh TSAs (22 crypts from four different lesions) compared with surrounding normal crypts. Data represent means ± s.e.m. (f) In situ hybridization for
GREM1 in archival human TSA samples showing aberrant epithelial GREM1 mRNA expression in TSA epithelium (brown dots). Scale bars are 100 µm
(representative of 6 different TSAs).
npg © 2015 Nature America, Inc. All rights reserved.
Articles
68 VOLUME 21 | NUMBER 1 | JANUARY 2015 nAture medicine
Supplementary Fig. 6d), demonstrating that knockout of stromal
and/or aberrant epithelial Grem1 expression ameliorates the develop-
ment of conventional intestinal tumorigenesis. To investigate whether
GREM1 might influence the behavior of human tumors, we analyzed
associations between GREM1 expression and survival in two publi-
cally available CRC data sets. Individuals with above-median GREM1
expression had significantly shorter disease-free survival in both
data sets (AMC-AJCCII-90 set, stage 2: log-rank P = 0.0162, Moffit-
Vanderbilt-Royal Melbourne set, stages 1–3: log-rank P = 0.0112)
(Fig. 5d and Supplementary Fig. 5e).
Collectively, these results strongly support the notion that GREM1
and Wnt signaling act in a synergistic fashion in the initiation and
progression of intestinal polyps. We propose that in both human and
mouse neoplasia resulting from aberrant GREM1 expression, there
is a strong selective pressure for somatic Wnt activation in actively
dividing ECFs, whereas in conventional Wnt-driven tumorigenesis,
stromal-derived GREM1 provides a favorable microenvironment for
the persistence of somatically mutant cells.
Epithelial GREM1 expression in sporadic traditional serrated
adenomas
On the basis of gene expression and somatic mutation profiles,
a recent study divided sporadic colorectal cancer into three main
molecularly distinct subtypes: CCS1, enriched for tumors with chro-
mosomal instability; CCS2, enriched for tumors with microsatellite
instability; and a third group, CCS3. This group was more heter-
ogenous, appeared to arise from serrated adenoma precursors and
gave rise to an aggressive subset of cancers with poor prognosis28.
To see whether whole-tumor GREM1 expression correlated with this
classification, we used The Cancer Genome Atlas RNA-seq data to
match sporadic tumor samples to the published CCS1, CCS2 and
CCS3 subtypes and found a highly significant correlation between
high GREM1 expression and CCS3-subtype tumors (P < 0.0001,
analysis of variance (ANOVA)) (Fig. 5d).
Traditional serrated adenomas (TSAs) are distal colonic polyps
with hitherto unknown pathogenesis that make up about 2% of the
lesions removed at colonoscopy. The characteristic histological feature
Normal colon
Normal small bowel
Matrigel and
ENS media
Isolated villi Spheroids
Grem1 cDNA driven
by Vil1 promoter
Vil1-Grem1; ApcMin/+
HumanMouseIn vitro
Grem1
BMP
Wnt
Sox9/Ephb2
Grem1
BMP
Wnt
Sox9/Ephb2
Grem1
BMP
Wnt
Sox9/Ephb2
Grem1
BMP
Wnt
Sox9/Ephb2
Grem1
BMP
Wnt
Sox9/Ephb2
Grem1
BMP
Wnt
Sox9/Ephb2
HMPS: chromosome 15
duplication
HMPS/TSA
Somatic mutation
HMPS
TSA
KRAS > APC
> p16 loss
p16 loss > Ctnnb1
Somatic mutation
Somatic mutation
Apc LOH
(2nd hit)
Vil1-Grem1 mouse
TSA: unknown
BRAF
KRAS
BRAF
Figure 6 Model summarizing the proposed mechanistic consequences of disrupted GREM1 morphogen gradients. Aberrant ectopic epithelial
expression of GREM1 disrupts the coupling of cell fate determination to position along the crypt-villus axis and allows persistence and expansion
of an Lgr5-negative progenitor cell pool (characterized by aberrant SOX9 and EPHB2 expression) that forms orthogonal ectopic crypt foci. Aberrant
cell proliferation in this progenitor cell population within these ECFs predisposes toward somatic (epi)mutation events and gives rise to neoplastic
transformation (inset boxes). In vitro, the persistence of somatically mutated progenitor cells in dissected villi gives rise to clonogenic tumor spheroid
growth from cells that have exited the crypt basal stem cell niche. Colored bars represent morphogen and gene expression gradients in the normal and
pathological states. Blue squares represent physiological Grem1 expression from pericryptal myofibroblasts. CBC stem cells are colored red.
npg © 2015 Nature America, Inc. All rights reserved.
Articles
nAture medicine VOLUME 21 | NUMBER 1 | JANUARY 2015 6 9
of these sporadic lesions is the development of ectopic crypt foci2 9.
We collected and dissected individual crypts from four fresh TSA
specimens. Individual crypt analysis of BMP antagonist expression
revealed a mean 87-fold upregulation of epithelial GREM1 expression
over surrounding normal mucosa (Fig. 5e), which was significantly
greater than that seen in conventional, hyperplastic and sessile
serrated adenomas (P < 0.01, t-test, Supplementary Fig. 7a). We vali-
dated epithelial GREM1 expression in a set of ten paraffin-embedded
TSAs using mRNA in situ hybridization. We detected clearly
visible epithelial GREM1 expression in a further six of ten polyps
(Fig. 5f). Immunohistochemical assessment, mutation and methylation
analysis revealed a very similar molecular phenotype and somatic
(epi)mutation pattern to HMPS polyps (Supplementary Fig. 7b,c).
We observed CIMP in 33% of TSAs, and p16INK4A promoter methyla-
tion correlated with p16INK4A protein downregulation and resultant
cell proliferation similar to the p16INK4A and Ki-67 staining observed
in Vil1-Grem1 mice (Supplementary Fig. 7d). These data suggest that
the ECFs that characterize sporadic TSAs also arise from disruption
of homeostatic morphogen gradients and that in the majority of cases
this is the consequence of aberrant epithelial GREM1 expression.
DISCUSSION
Here we use a mouse model of a human disease to demonstrate the
pathogenetic mechanism in hereditary mixed polyposis syndrome
(Fig. 6). We hypothesize that aberrant epithelial GREM1 expression
disrupts intestinal morphogen gradients, altering daughter cell fate
and promoting the persistence of Lgr5-negative progenitor cells in
ECFs distant from the crypt base. Cell proliferation within ECFs means
that progenitor cells are prone to tumor-causing somatic mutations.
Human HMPS polyps progress through KRAS or BRAF mutation and
frequent selection of a very restricted set of APC mutations. In the
mouse model, Grem1-initiated lesions advance through p16 loss and
Ctnnb1 mutation, and once somatic mutation has occurred in some
advanced lesions, epithelial Grem1 expression becomes redundant.
Ex vivo culture of transgenic mouse villi demonstrates that
persistence of somatically mutated villus cells can initiate clonogenic
growth. Using mouse models, we have also shown the exacerbation
or amelioration of conventional, Wnt-driven neoplasia initiation by
Grem1 overexpression or knockout, respectively. Lastly, we show that
epithelial expression of GREM1 also occurs in sporadic TSAs, lesions
similarly characterized by the development of aberrantly proliferating
cells in ectopic crypt foci, and that these lesions can thus be consid-
ered the sporadic equivalents of HMPS polyps.
Multiple BMP antagonists have been described, with varying levels of
intestinal expression and importance in intestinal homeostasis. There
is much greater physiological expression of both GREM1 and GREM2
than of NOG in human intestinal stroma (Supplementary Fig. 5f),
and the importance of GREM1 has been highlighted by its causative
role in HMPS9 and the association of sporadic CRC with GREM1 com-
mon allelic variants30. The association of germline-inactivating NOG
mutations with skeletal conditions such as symphalangism and tarsal-
carpal coalition31 correspondingly reflects the greater significance
of NOG’s role in bone development than in intestinal homeostasis.
Previous investigation of the pathogenesis of juvenile polyposis led
to the generation of mice expressing Xenopus nog under control of
the Vil1 promoter7 and mice expressing Xenopus nog under control
of the Fabp1 promoter8. Both models initially developed intravillus
ectopic crypts before progressing to a juvenile polyposis–like pheno-
type. The differences between small intestinal polyp morphology in
these models and in the Vil1-Grem1 mouse may reflect the Xenopus
origin of the nog, the relative importance of these different antagonists
in intestinal homeostasis or fundamental differences in the ligand
targets or biology of these BMP antagonists, which share minimal
sequence homology. Human HMPS and JPS are pathogenically and
morphologically distinct32, and the profound differences in patho-
genesis, phenotype and somatic pathway progression between these
conditions highlights how subtle alterations in the BMP signaling
cascade can have different effects on intestinal tumorigenesis.
The histogenesis of colorectal neoplasia has historically been a
subject of some debate: the classical bottom-up model favors dysplasia
arising from the crypt basal stem cell niche33, whereas the top-down
model proposes that dysplasia originates at the luminal surface and
spreads downwards34. The identification of the Lgr5-positive CBC
stem cell at the very base of the crypt1 seemed to have settled the
argument in favor of the bottom-up model. Here, we demonstrate
that pathological disruption of morphogen gradients in animal mod-
els and human inherited and sporadic disease causes a profound
change in the fate of cells situated outside of the crypt base stem and
proliferative zone, with expansion of an Lgr5-negative progenitor
cell population. These cells form ectopic crypt structures, proliferate,
accumulate somatic mutations and are capable of initiating intestinal
neoplasia, indicating that the crypt basal stem cell is not the exclusive
cell of origin of all subtypes of colorectal cancer. Conceivable
histogenic origins of this progenitor population include the vertical
expansion of a crypt base progenitor population containing variably
competent stem cells, the generation of an ectopic niche for a migrated
Lgr5-negative stem cell and dedifferentiation of post-mitotic special-
ized intestinal epithelial cells. Very recently, Schwitalla et al.35 have
shown that activated nuclear factor-κB–induced mucosal inflammation
in combination with constitutive epithelial Wnt signaling can also
promote the initiation of neoplasia from cells situated outside the
crypt base stem cell niche. Taken together, these studies highlight
the importance of the intestinal microenvironment in maintaining
stem cell homeostatic control and cell-fate determination. This work
challenges the concept of a strict unidirectional tissue organizational
hierarchy in the intestine and demonstrates that a top-down model of
tumor histogenesis may fit some subtypes of inherited, sporadic and
inflammation-associated colorectal cancers.
METHODS
Methods and any associated references are available in the online
version of the paper.
Accession codes. Gene expression data have been deposited in
the NCBI Gene Expression Omnibus with accession number
GSE62307.
Note: Any Supplementary Information and Source Data files are available in the
online version of the paper.
ACKNOWLEDGMENTS
This work was funded by Cancer Research UK Clinician Scientist Fellowship
A16581 to S.J.L. and Programme grant A16459 to I.T. Core funding to the
Wellcome Trust Centre for Human Genetics was provided by the Wellcome Trust
(090532/Z/09/Z). We thank the transgenics core and the staff of the Functional
Genomics Facility at the Wellcome Trust Centre for Human Genetics. Villin-MES-
SV40polyA plasmid was a kind gift from S. Robine (Institut Curie, Paris, France).
AUTHOR CONTRIBUTIONS
S.J.L. and I.T. conceived and designed the project. Experiments were conducted
by H.D., S.I., H.R., T.B., C.B., E.J., A.L., P.R.-C. and S.J.L. In situ hybridization was
completed by H.D., R.J., R.P. and A.S. Bioinformatic analysis carried out by
S.I., M.B., L.F.-M. and F.C.G. Pathology support, tissue provision and intellectual
npg © 2015 Nature America, Inc. All rights reserved.
Articles
70 VOLUME 21 | NUMBER 1 | JANUARY 2015 nAture medicine
input were provided by S.M., S.C., H.T., M.R.-J., M.N., R.C., L.M.W. and J.E.E.
Mouse resources were supplied by O.J.S. and F.R.G. The manuscript was written
by S.J.L. and I.T.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
Reprints and permissions information is available online at http://www.nature.com/
reprints/index.html.
1. Barker, N. et al. Identification of stem cells in small intestine and colon by marker
gene Lgr5. Nature 449, 1003–1007 (2007).
2. van den Brink, G.R. & Offerhaus, G. The morphogenetic code and colon cancer
development. Cancer Cell 11, 109–117 (2007).
3. Hardwick, J.C. et al. Bone morphogenetic protein 2 is expressed by, and acts upon,
mature epithelial cells in the colon. Gastroenterology 126, 111–121 (2004).
4. Scoville, D.H., Sato, T., He, X. & Li, L. Current view: intestinal stem cells and
signaling. Gastroenterology 134, 849–864 (2008).
5. Fearon, E.R. & Vogelstein, B. A genetic model for colorectal tumorigenesis.
Cell 61, 759–767 (1990).
6. Hardwick, J.C., Kodach, L., Offerhaus, G. & van den Brink, G. Bone morphogenetic
protein signalling in colorectal cancer. Nat. Rev. Cancer 8, 806–812
(2008).
7. Haramis, A.P. et al. De novo crypt formation and juvenile polyposis on BMP
inhibition in mouse intestine. Science 303, 1684–1686 (2004).
8. Batts, L.E., Polk, D.B., Dubois, R.N. & Kulessa, H. Bmp signaling is required for
intestinal growth and morphogenesis. Dev. Dyn. 235, 1563–1570 (2006).
9. Jaeger, E. et al. Hereditary mixed polyposis syndrome is caused by a 40-kb upstream
duplication that leads to increased and ectopic expression of the BMP antagonist
GREM1. Nat. Genet. 44, 699–703 (2012).
10. Whitelaw, S.C. et al. Clinical and molecular features of the hereditary mixed
polyposis syndrome. Gastroenterology 112, 327–334 (1997).
11. Westphalen, C.B. et al. Long-lived intestinal tuft cells serve as colon cancer-initiating
cells. J. Clin. Invest. 124, 1283–1295 (2014).
12. Barker, N. et al. Crypt stem cells as the cells-of-origin of intestinal cancer. Nature
457, 608–611 (2009).
13. Weisenberger, D.J. et al. CpG island methylator phenotype underlies sporadic
microsatellite instability and is tightly associated with BRAF mutation in colorectal
cancer. Nat. Genet. 38, 787–793 (2006).
14. Muñoz, J. et al. The Lgr5 intestinal stem cell signature: robust expression of
proposed quiescent ‘+4’ cell markers. EMBO J. 31, 3079–3091 (2012).
15. Merlos-Suárez, A. et al. The intestinal stem cell signature identifies colorectal cancer
stem cells and predicts disease relapse. Cell Stem Cell 8, 511–524 (2011).
16. Qiagen Biosciences. Curated apoptosis array http://www.sabiosciences.com/rt_pcr_
product/HTML/PAMM-012Z.html (2012).
17. Qiagen Biosciences. Curated cellular senescence array http://www.sabiosciences.
com/rt_pcr_product/HTML/PAMM-050Z.html (2012).
18. Qiagen Biosciences. Curated autophagy array http://www.sabiosciences.com/rt_pcr_
product/HTML/PAMM-084Z.html (2012).
19. Bieging, K.T., Mello, S.S. & Attardi, L.D. Unravelling mechanisms of p53-mediated
tumour suppression. Nat. Rev. Cancer 14, 359–370 (2014).
20. Brech, A., Ahlquist, T., Lothe, R.A. & Stenmark, H. Autophagy in tumour suppression
and promotion. Mol. Oncol. 3, 366–375 (2009).
21. Sato, T. et al. Single Lgr5 stem cells build crypt-villus structures in vitro without
a mesenchymal niche. Nature 459, 262–265 (2009).
22. Sato, T. et al. Long-term expansion of epithelial organoids from human colon,
adenoma, adenocarcinoma, and Barrett’s epithelium. Gastroenterology 141,
1762–1772 (2011).
23. Dhamdhere, G.R. et al. Drugging a stem cell compartment using Wnt3a protein as
a therapeutic. PLoS ONE 9, e83650 (2014).
24. Hayashi, S. & McMahon, A.P. Efficient recombination in diverse tissues by a
tamoxifen-inducible form of Cre: a tool for temporally regulated gene activation/
inactivation in the mouse. Dev. Biol. 244, 305–318 (2002).
25. Lewis, A. et al. A polymorphic enhancer near GREM1 influences bowel cancer
risk through differential CDX2 and TCF7L2 binding. Cell Rep. 8, 983–990
(2014).
26. Segditsas, S. et al. Putative direct and indirect Wnt targets identified through
consistent gene expression changes in APC-mutant intestinal adenomas from
humans and mice. Hum. Mol. Genet. 17, 3864–3875 (2008).
27. Sneddon, J.B. et al. Bone morphogenetic protein antagonist gremlin 1 is widely
expressed by cancer-associated stromal cells and can promote tumor cell
proliferation. Proc. Natl. Acad. Sci. USA 103, 14842–14847 (2006).
28. De Sousa, E. & Melo, F. et al. Poor-prognosis colon cancer is defined by a molecularly
distinct subtype and develops from serrated precursor lesions. Nat. Med. 19,
614–618 (2013).
29. Torlakovic, E.E. et al. Sessile serrated adenoma (SSA) vs. traditional serrated
adenoma (TSA). Am. J. Surg. Pathol. 32, 21–29 (2008).
30. Jaeger, E. et al. Common genetic variants at the CRAC1 (HMPS) locus on chromosome
15q13.3 influence colorectal cancer risk. Nat. Genet. 40, 26–28 (2008).
31. Potti, T.A., Petty, E.M. & Lesperance, M.M. A comprehensive review of reported
heritable noggin-associated syndromes and proposed clinical utility of one broadly
inclusive diagnostic term: NOG-related-symphalangism spectrum disorder (NOG-
SSD). Hum. Mutat. 32, 877–886 (2011).
32. Tomlinson, I., Jaeger, E., Leedham, S. & Thomas, H. Reply to “The classification
of intestinal polyposis”. Nat. Genet. 45, 2–3 (2013).
33. Preston, S.L. et al. Bottom-up histogenesis of colorectal adenomas: origin in the
monocryptal adenoma and initial expansion by crypt fission. Cancer Res. 63,
3819–3825 (2003).
34. Shih, I.M. et al. Top-down morphogenesis of colorectal tumors. Proc. Natl. Acad.
Sci. USA 98, 2640–2645 (2001).
35. Schwitalla, S. et al. Intestinal tumorigenesis initiated by dedifferentiation and
acquisition of stem cell-like properties. Cell 152, 25–38 (2013).
36. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach
for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. USA 102,
15545–15550 (2005).
npg © 2015 Nature America, Inc. All rights reserved.
nAture medicine
doi:10.1038/nm.3750
ONLINE METHODS
Generation and genotyping of Vil1-Grem1 mice. Grem1 cDNA
was amplified from normal mouse intestinal mesenchyme using the
following primers F 5′-GACGTCGACCAATGGAGAGAC-3′ and R 5′-GAC
GTCTGACGGACAAGCAA-3′, adding an AatII restriction enzyme site
at each end. The excised cDNA fragment was then cloned into the Villin-
MES-SV40polyA plasmid (kind gift from S. Robine) downstream of the
9-Kb Villin promoter (active after embryonic day E9). This plasmid was
linearized with Kpn1 digestion. Transgenic mice were derived by pronu-
clear injection in BL6/129 mice using standard methods by the Transgenic
Facility, WTCHG, Oxford University. To identify successful transmission,
tail snips were taken from mice at weaning age and amplified by PCR using
the following primers; F 5′-GAGGTCGAGGCTAAAGAAGA-3′ and R 5′-AC
CTAGTGTCGGGCATCTCC-3′. Successful transmission was confirmed by
Southern blot using standard methods. In brief, the probes were amplified
using the following primers F 5′-AGGTAGGGAGGTCGAGGCTA-3′ and
R 5′-CAACGCTCCCACAGTGTATG-3′. The probes were then radioactively
labeled by random priming with α32P dCTP using the random primed labe-
ling kit (Roche) as previously described37. 20 µg of genomic DNA was digested
to completion with SphI and MscI restriction enzymes. The digested DNA
was electrophoresed on 0.8% gels in 1× TBE overnight. After denaturation,
the gels were blotted onto Hybond N+ membrane (GE Healthcare) overnight,
UV crosslinked (Stratalinker, Stratagene) and hybridized with the radiolabeled
probes. Hybridization and washing of Southern blots was performed using
standard methods and detected a 2,079-bp region of the inserted transgene in
founder mice (Supplementary Fig. 3c). Three founder lines were established
and all developed an identical phenotype. One line was arbitrarily chosen to
be taken forward.
Mouse procedures. All procedures were carried out in accordance to Home
Office UK regulations and the Animals (Scientific Procedures) Act 1986. All mice
were housed at the animal unit at Functional Genomics Facility, Wellcome Trust
Centre for Human Genetics, Oxford University. All strains used in this study
were maintained on C57BL/6J background for ≥6 generations. To induce recom-
bination in conditional lines, mice were treated with 1 mg tamoxifen by intra-
peritoneal injection for five days. Genotyping protocols for the ApcMin/+ (ref. 38),
Cagg-CreERT2 and Gremfl/fl (ref. 39) mice have previously been reported.
Tissue preparation and histology. Both male and female mice were used
throughout the study. Mice were killed at predefined time points (as indicated
in the text) or when showing symptoms of intestinal polyps (anemia, hunch-
ing) by cervical dislocation. The intestinal tract was removed immediately and
divided into small intestine (proximal/SB1, middle/SB2 and distal/SB3) and
large intestine. The intestines were opened longitudinally using a gut preparation
apparatus40, washed in PBS and fixed overnight in 10% neutral buffered formalin
(NBF). For scoring of polyps, gut preparations were stained with 0.2% methylene
blue for 10 s and washed in PBS for 20 min, and polyps were counted/measured
using a dissecting microscope at ×3 magnification. Specimens of 10% formalin-
fixed tissue were embedded in paraffin and then sectioned at 4 µm. Fixed
specimens were embedded and H&E stained following standard protocols.
Immunohistochemistry. Formalin-fixed, paraffin-embedded tissue sections
(4 µm) were de-waxed in xylene and rehydrated through graded alcohols
to water. Endogenous peroxidase was blocked using 1.6% H2O2 for 20 min.
For antigen retrieval, sections were pressure cooked in 10 mmol/L citrate buffer
(pH 6.0) for 5 min. Sections were blocked with 10% serum for 30 min. Slides
were incubated with primary antibody for 2 h. Antibodies to the following
proteins have been used in this study: alkaline phosphatase (Abcam, ab65834,
1:50), β-catenin (BD, 610154, 1:50), Caspase 3 (R & D, AF835, 1:800), Cdkn2a/
p16Inka (mouse) (ThermoScientific, PA1-30670, 1:500), CDKN2A/p16Inka
(human) (Abcam, ab51243, 1:250), Chromogranin A (Abcam, ab15160, 1:1,250),
Cytokeratin 20 (Abcam, ab118574, 1:200), EphB2 (R and D, AF467, 1:125),
GFP (Life Technologies, A-6455, 1:500), Ki-67 (mouse) (DAKO, TEC-3, 1:125),
KI-67 (human) (DAKO, MIB-1, 1:150), Lysozyme (DAKO, EC 3.2.1.17, 1:500),
pSmad1/5/8 (Cell Signaling, 9511L, 1:50) and Sox9 (Millipore, ab5535, 1:1,000).
Appropriate secondary antibodies were applied for 1 h at room temperature.
Sections were then incubated in ABC (Vector labs) for 30 min. DAB solution
was applied for 2–5 min and development of the color reaction was monitored
microscopically. Slides were counterstained with hematoxylin, dehydrated,
cleared and then mounted.
Alcian-blue stain for goblet cells. Sections were dewaxed in xylene for 5 min
and then rehydrated through graded ethanols (100%, 90%, 70%) for 5 min each
followed by 2 min in tap H2O. Slides were then stained in alcian-blue solution
(Sigma) for 30 min, washed in running tap H2O for 2 min, and rinsed in dH2O.
Slides were then stained in nuclear fast red solution (Sigma) for 5 min and
washed in running tap H2O for 1 min. Slides were then dehydrated through
degraded alcohols for 2–5 min each, before mounting a coverslip with DPX.
Immunohistochemical quantification. For mouse phenotype quantification
analyses, we used no fewer than three animals per group (control and experi-
mental, at age 10 months). To assess overall change in epithelial cell numbers
between Vil1-Grem1 mice compared to age-matched, wild-type counterparts, we
microscopically captured images of hemotoxylin-eosin–stained 4-µm sections of
the small intestine and colon at ×10 magnification and then randomly selected
crypts and villi, counting individual cells within these compartments. An average
of 50 individual villi within the small intestine and 100 crypts were counted.
Similarly, 100 crypts of the colon were counted, giving us enough power to run
a t-test. Data are represented as total number of cells per crypt/villus. Excel was
used to calculate means and standard error. To determine the proliferative index
in experimental and control mice, a total of three mice per group were used.
Small intestinal and colon sections were immunohistochemically labeled with
Ki-67–specific antibody. Microscopic images at ×20 magnification were taken
and number of Ki-67 positive cells per total epithelial cells in randomly selected
50 crypts (colon or small intestine) and 50 villi (small intestine) were counted.
Data are represented as a percentage of Ki-67+ cells in each compartment. The
percentage of goblets cells in epithelial cells of the villus (small intestine) and
crypt (small intestine/colon) was quantified by counting Alcian blue–positive
cells in at least 50 crypts of the colon/small intestine and 50 villi from small
intestine in three mice of each genotype. Similarly, quantification of Paneth
cells was determined by counting lysozyme-positive cells over total epithelial
cells in three Vil1-Grem1 and three wild-type mice. To assess apoptosis, cleaved
Caspase 3–positive cells over total number of epithelial cells per crypt (50 crypts
in total) in small intestine/colon, and per villus (50 villi in total) in small intestine
were counted. No randomization or blinding of mouse genotypes was used. All
statistical analyses were done in Excel.
Individual crypt and villus isolation, RNA extraction and qRT-PCR. For both
mouse and human individual crypt/villus isolation, biopsies were washed with
PBS and incubated in 5 ml dissociation media (30 mM EDTA in DMEM without
Ca2+ and Mg2+, 0.5 mM DTT, 2% RNAlater (Life Technologies) for 10 min at
room temperature. The digested tissue was then transferred to PBS and shaken
vigorously for 30 s to release individual crypts and villi. Individual structures
were selected using a drawn-out glass pipette under a dissection microscope and
transferred to RLT buffer ready for subsequent RNA extraction with the RNeasy
microkit (Qiagen) according to manufacturer’s instructions. RNAs were treated
with DNase I to degrade residual DNA. When required, complementary DNA was
reverse transcribed in vitro using the High Capacity cDNA Reverse Transcription
Kit (Applied Biosystems). When using cDNA generated from individual crypts
or villi, preamplification of these cDNAs was necessary before qRT-PCR. The
TaqMan PreAmp (Applied Biosystems) kit was used following manufactur-
er’s instructions. Absolute quantification qRT-PCR was performed on the ABI
7900HT cycler (Applied Biosystems) with GAPDPH/Gapdh serving as an endog-
enous control. A list of TaqMan Gene Expression assays (Applied Biosystems)
used is available on request. The primary assumption in analyzing Real time PCR
results is that the effect of a gene can be adjusted by subtracting Ct number of
target gene from that of the reference gene (∆Ct). The ∆Ct for experimental and
control can therefore be subject to t-test, which will yield the estimation of ∆∆Ct.
In all cases the data met the normal distribution assumption of the t-test.
In situ hybridization. 4-µm sections were prepared using DEPC (Sigma)-treated
H2O. In situ hybridization was carried out using the GREM1 (312831), PPIB
npg © 2015 Nature America, Inc. All rights reserved.
nAture medicine doi:10.1038/nm.3750
(313901), Grem1 (314741) Ppib (313911) and DapB (310043) (Advanced Cell
Diagnostics) probes and the RNAscope 2.0 HD Detection Kit (Advanced Cell
Diagnostics) following manufacturer’s instructions.
Mouse Apc loss-of-heterozygosity. Apc allelic status was assessed by a PCR
assay described previously using primers F 5′-TCTCGTTCTGAGAAAGACA
GAAGCT-3′ and R 5′-TGATACTTCTTCCAAAGCTTTGGCTAT-3′ (ref. 41).
PCR products were electrophoresed on 2.5% agarose gels. Briefly, the amplifi-
cation of the ApcMin allele resulted in a 155-bp PCR product with one HindIII
site, whereas the 155-bp product for the Apc+ allele contained two HindIII sites.
HindIII digestion of PCR-amplified DNA from ApcMin/+ heterozygous tissue
generates 123-bp and 144-bp products. PCR products from tissue with LOH
displayed only one band (144-bp).
Laser capture microdissection. MembraneSlide 1.0 PEN slides (Zeiss)
were exposed to UV light for 30 min before mounting 8-µm sections.
These slides were baked at 37 °C for 30 min and then dewaxed for 5 min and
rehydrated through graded alcohols to water for 3–5 min each. The slides
were then briefly dipped in methyl green, washed in water and dried at 37 °C
for 1 h. Laser capture microdissection was performed with the laser capture
PALM system (Zeiss). DNA was extracted using the PicoPure DNA extraction
kit (Arcturus).
Sequencing. Sequencing of gDNA was carried out using the 2× Big Dye Terminator
v3.1 reagent (Applied Biosystems). Unincorporated dye terminators were removed
with the DyeEx 2.0 Spin kit (Qiagen) and the purified products were run on
the ABI 3730 DNA analyzer (Applied Biosystems). Sanger sequencing primers
of candidate genes are as follows: Ctnnb1 E2F TTCAGGTAGCATTTTCAGTT
CAC, Ctnnb1 E2R TAGCTTCCAAACACAAATGC, KRAS12/13F GTGT
GACATGTTCTAATATAGTC, KRAS12/13R GAATGGTCCTGCACCAGTAA,
BRAF F TCATAATGCTTGCTCTGATAGGA, BR AF R GGCCAAAAA
TTTAATCAGTGGA, TP53 5F TCTGTCTCCTTCCTCTTCCTACA, TP53
5R AACCAGCCCTGTCGTCTCT, TP53 6F CAGGCCTCTGATTCCTCACT,
TP53 6R T TAACCCCTCCTCCCAGAG, TP53 7F CTTGGGCCTGTGTTAT
CTCC, TP53 7R GTGTGCAGGGTGGCAAGT, TP53 8F GCCTCTTGCT
TCTCTTTTCC, TP53 8R CTTCTTGTCCTGCTTGCTT, PTEN 5F AGA
CCATAACCCACCACAGC, PTEN 5R TGGTCCTTACTTCCCCATAGA,
PTEN 6F TCATAATGCTTGCTCTGATAGGA, PTEN 6R GGCCAAAAAT
TTAATCAGTGGA, PTEN 7F TGCAGATCCTCAGTTTGTGG, PTEN 7R
GCATCTTGTTCTGTT TGTGGA, PTEN 8F CAAAATGTTTCACTTTT
GGGTAAA, PTEN 8R TTGGAGAAAAGTATCGGTTGG, PIK3CA 9F
CTGTGAATCCAGAGGGGAAA, PIK3CA 9R TTTGGCTGATCTCAGC
ATGT, PIK3CA 20F CAGCATGCCAATCTCTTCAT, PIK3CA 20R TTTT
CAGTTCAATGCATGCTG, APC 1F GGACAAAGCAGTAAAACCGAAC,
APC 1R AACTACATCTTGAAAAACATATTGGA, APC 2FGCCACTT
GCAAAGTT TCTTCT, APC 2R TGCTTCCTGTGTCGTCTGA, APC 3F
GATACTCCAATATGTTTTTCAAGATG, APC 3R GCCTGGCTGATTCT
GAAGAT, APC 4F CCCTGCAAATAGCAGAAATAAAA, APC 4R AAC
ATGAGTGGGGTCTCCTG, APC 5F CAGACTGCAGGGTTCTAGTTTATC,
APC 5R CATTCCACTGCATGGTTCAC, APC 6F CCAAAAGTGGT
GCTCAGACA, APC 6R CATGGTT TGTCCAGGGCTAT, APC 7F
TTTGAGAGTCGTTCGATTGC, APC 7R TCTCTTTTCAGCAGTAGGTG
CTT, APC 8F CATGCAGTGGAATGGTAAGT, APC 8R GCAGCATTTA
CTGCAGCTT, APC 9F CAAGCGAGAAGTACCTAAAAA, APC 9R TT
CTGTATAAATGGCTCATCG, APC 10F GGTTCTTCCAGATGCTGATA,
APC 10R CTTGGTTTTCATTTGATTCTTT, APC 11F AATTAAGAAT
AATGCCTCCAGT, APC 11R TTTACGTGATGACTTTGTTGG, APC 12F
CCAAGAGAAAGAGGCAGAAAAA, APC 12R TGATGGTAGAAGTTTGT
ACACAGG.
Determination of the CpG island methylator phenotype. 500 ng of DNA was
bisulphite treated using the EZ DNA Methlyation-Gold kit (Zymo) following
the manufacturer’s protocol. Subsequent determination of CIMP status was
performed using real-time PCR-based protocol (MethyLight) assaying a five
marker panel composed of CACNA1G, IGF2, NEUROG1, RUNX3 and SOCS1,
as described in ref. 42. Samples with ≥3 of five markers positive for methylation
are considered CIMP positive, whereas a sample with ≤2 of the five markers
positive for methylation is considered CIMP negative.
Microsatellite stability determination. A multiplex PCR (Qiagen) was performed
for the markers Bat25, Bat26 and D2S123 using the following primers Bat25 F
5′-TCGCCTCCAAGAATGTAAGT-3′, R 5′-TCTGGATTTTAACTATGGCTC,
Bat26 F 5′-TGACTACTTTTGACTTCAGCC-3′ and R 5′-TTCTTCAGTATAT
GTCAATGAAAACA-3′, D2S123 F 5′-AAACAGGATGCCTGCCTTTA-3′ and
R 5′-GGACTTTCCACCTATGGGAC-3′. The forward primers were differen-
tially dye-labeled (Bat25 and D2S123 Hex labeled, Bat26 FAM labeled). The PCR
products were run on a semi-automated ABI3100 Genetic Analyzer (Applied
Biosystems) by Sequencing Service, Zoology Department, Oxford University.
Results were analyzed using GeneMarker software (SoftGenetics). Microsatellite
instability was assigned when two or more markers showed instability.
Culture of mouse intestinal crypts. Mouse intestinal crypts were isolated and
cultured as described by Sato et al.22. In brief, crypts were isolated, resuspended
in Matrigel (BD Biosciences) and plated out in 24-well plates. The basal culture
medium (advanced Dulbecco’s modified Eagle medium/F12 supplemented
with penicillin/streptomycin, 10 mmol/L HEPES, Glutamax, 1× N2, 1× B27
(all from Invitrogen), and 1 mmol/L N-acetylcysteine (Sigma)) was overlaid
containing the following growth factors; Epidermal Growth Factor at 50 ng/ml
(Life Technologies), Noggin at 100 ng/ml (PeproTech) and R-spondin1 at
500 ng/ml (R and D) (ENS media). The medium was changed every 2 d.
To see whether addition of recombinant BMP ligands to the media could compete
with the endogenous overexpression of Grem1 and to demonstrate the necessity
of a source of BMP antagonist, Vil1-Grem1 crypt enteroids were grown in
media lacking Noggin (ES media) together with a range of rBMP ligand 2, 4 and
7 concentrations (rhBMP2, rmBmp4 and rmBmp7 (R&D Systems) 0–1,000 ng/ml
of each of the three BMP ligands). Cells cultured in media containing rBmp
concentrations of 50–1,000 ng/ml of each ligand were not viable (enterospheres
formed and were still present on day 2 but completely disaggregated by day 5).
Culture of intestinal adenomas. From symptomatic ApcMin/+ animals the intes-
tines were opened longitudinally and adenomas were scraped with a coverslip.
The first scrape of material was discarded and the subsequent scrapes were
collected into sterile PBS. The material was gently washed with PBS and then
centrifuged at 100g for 3 min. The material was then plated out in 24-well plates
in Matrigel (BD Biosciences). The wells then overlaid with Epidermal Growth
Factor at 50 ng/ml (Life Technologies), R-spondin1 at 500 ng/ml (R and D) and
Gremlin1 at 100 ng/ml (R and D) (EGS media). After spheroids had formed
(4 d of culture) Gremlin1 was subsequently withdrawn from the media
(ES media). Adenoma spheroids were comparable when maintained in both
EGS and ES media. In the following experiment, adenomas were cultured
in ES media immediately on removal from the animals and again sphe-
roids were found and were comparable with the EGS control. In summary,
recombinant Gremlin1 is not required for the expansion or maintenance
of ApcMin adenoma spheroids. This concurs with previous findings that
neither R-spondin or Noggin are required for the expansion and maintenance
of ApcMin adenoma spheroids22.
Culture of mouse intestinal villi. Vil1-Grem1 animal intestines were opened
longitudinally and the villi were scraped off with a coverslip. The first scrape
of material was discarded and the subsequent scrapes were collected into
sterile PBS. The material was gently washed with PBS and then centrifuged
at 100g for 3 min. The villi were then plated out in 24-well plates in Matrigel
(BD Biosciences) and were overlaid by standard media that was additionally
supplemented with mWnt3a at 100 ng/ml (ENSW media) (R and D Systems).
The media was changed every 2 d. To calculate the percentage efficiency of villi
transformation, the number of villi per well were counted after seeding (day 0)
and the number of spheroids present on day 4. Multiple wells were plated
out for each time point. Spheroids were cultured from Vil1-Grem1; ApcMin/+
as described above, although in addition recombinant Bone morphogenetic
proteins (as above) were added to the media in an attempt to outcom-
pete the antagonist effect of endogenous overexpression of Grem1 in this
model. Spheroids formed and grew in all of these concentrations of BMP
npg © 2015 Nature America, Inc. All rights reserved.
nAture medicine
doi:10.1038/nm.3750
ligands, showing the overexpression of Grem1 is no longer required for
the growth and maintenance of villus spheroids.
Passaging and embedding of spheroids. After the intestinal villi had been
cultured for appropriately a week and had grown into spheroid structures, they
were passaged by adding cold media to melt the Matrigel and subsequently
replated in fresh Matrigel. To collect material for embedding, the Matrigel was
melted by adding cold media; then multiple wells were combined. The cells were
fixed with 500 µl PFA 30 min at room temperature, centrifuged at 5,000 r.p.m.
and resuspended in 150 µl 2% agarose (in PBS). The paraffin-embedded cell
pellet was then processed and embedded using standard protocols.
DNA extraction. To extract DNA from crypt, villi and spheroids the QiaAmp
Micro kit (Qiagen) was used following manufacturer’s instructions.
Gene expression arrays. For GEP study, we used n = 6 controls and n = 6 experi-
mental groups to have a good estimate of the mean expression as well as to give
us sufficient power for the t-test. Raw data from Illumina gene expression arrays
(MouseWG-6_V2_0_R0_11278593_A chips) was processed after removing one
outlier sample from initial quality control (detection score of < 0.95 of the back-
ground intensity for majority of probes) using the VSN (variance-stabilization
and normalization) algorithm (Supplementary Table 1). We applied a filter by
taking a detection score of > 0.95 of the background intensity distribution for all
samples to consider a probe detectable, resulting in a total of 24,854 detectable
probes. Differentially expressed genes between experimental (Vil1-Grem1 small
intestinal crypts (n = 6 mice)/villi (n = 5 mice)) and normal (wild-type small
intestinal crypt (n = 6 mice)/villi (n = 6 mice)) were identified using Student’s
t-test by running “ttest2” command in MATLAB (Supplementary Table 2).
Gene expression data have been deposited in the NCBI Gene Expression
Omnibus with accession number GSE62307.
Gene set enrichment analysis (GSEA) was performed using Kolmogorov-
Smirnov statistics and gene shuffling permutations as described36. Genes were
ranked by computing their differential expression in the experimental versus
age-matched normal samples by the Student’s t-test method. If multiple probes
were present for a gene, probe with the highest absolute differential expression
between experimental and normal was selected. We used gene shuffling with
1,000 permutations to compute the P value for the enrichment score. A list of
gene signatures14–19 used in the enrichment analysis is given in Supplementary
Table 3a,b. If the signature was from human data set, we mapped these human
genes to their mouse orthologs using the sequence-based method available from
MGI (http://www.informatics.jax.org/).
Human colorectal cancer patient cohorts. Two publically available patient
cohorts were downloaded from the NCBI Gene Expression Omnibus (GEO)
using the R BioConductor package GEOquery. The first set comprises 90 patients
with stage 3 CRC treated in the Academic Medical Center in Amsterdam, defined
by the GEO accession number GSE33114 (AMC-AJCCII-90 set)28. The second
set contains 345 patients with colorectal cancer (CRC), defined by GEO accession
numbers GSE14333 and GSE17538 (Moffit-Vanderbilt-Royal Melbourne set)15.
Both sets are explained in detail in ref. 28.
Kaplan-Meier survival curves were generated using the R package survival.
The median expression value of GREM1 was used to segregate patients into
low- or high-GREM1 expressors. P value was calculated using the log-rank test
as informative covariates for Cox proportional hazards (such as stage) were
not present in all data sets and because the test is robust to the high degree of
right-censored data present.
RNA-seq data from patients with sporadic colorectal cancer was downloaded
from The Cancer Genome Atlas (TCGA) Data portal (http://cancergenome.
nih.gov/) and normalized using Voom in Limma package. Tumor samples were
matched to the published CRC subtypes CCS1, CCS2 and CCS3, as described
previously28. If for any given patient multiple samples were profiled, we randomly
selected one sample for further analysis. The analysis of variance (ANOVA) test
was used to compare the mean expression value of GREM1 across multiple
patients. We assessed whether GREM1 expression levels in the TCGA cohort
correlated with any of the CCS1, CCS2, CCS3 sub-types.
Ethics. Ethical approval for use of archival HMPS tissue was provided by the
Southampton and South-West Hampshire Research Ethics Committee A (REC
06/Q1702/99). Ethical approval for the collection and use of endoscopic and
archival TSA samples was obtained from the Oxfordshire Research Ethics
Committee A (REC 10/H0604/72.) Informed consent for tissue use in research
was obtained from all patients before endoscopic or surgical procedure.
All mouse experiments were completed in accordance with UK Home Office
regulations under the Animals (Scientific Procedures) Act 1986 (Animals in
Science Regulation Unit, project license PPL30/2763) and the University of
Oxford Clinical Animal Welfare and Ethical Review Body.
37. Feinberg, A.P. & Vogelstein, B. A technique for radiolabeling DNA restriction
endonuclease fragments to high specific activity. Anal. Biochem. 132, 6–13 (1983).
38. Moser, A.R., Dove, W.F., Roth, K.A. & Gordon, J.I. The Min (multiple intestinal
neoplasia) mutation: its effect on gut epithelial cell differentiation and interaction
with a modifier system. J. Cell Biol. 116, 1517–1526 (1992).
39. Gazzerro, E. et al. Conditional deletion of gremlin causes a transient increase in
bone formation and bone mass. J. Biol. Chem. 282, 31549–31557 (2007).
40. Rudling, R., Hassan, A.B., Kitau, J., Mandir, N. & Goodlad, R.A. A simple device
to rapidly prepare whole mounts of murine intestine. Cell Prolif. 39, 415–420
(2006).
41. Luongo, C., Moser, A.R., Gledhill, S. & Dove, W.F. Loss of Apc+ in intestinal
adenomas from Min mice. Cancer Res. 54, 5947–5952 (1994).
42. Weisenberger, D.J., Campan, M., Long, T.I. & Laird, P.W. Determination of the CpG
Island Methylator Phenotype (CIMP) in colorectal cancer using MethyLight. Protocol
Exchange doi:10.1038/nprot.2006.152 (30 June 2006).
npg © 2015 Nature America, Inc. All rights reserved.
