The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 2 February 2005
Transcriptional activation of integrin β6
during the epithelial-mesenchymal
transition defines a novel prognostic
indicator of aggressive colon carcinoma
Richard C. Bates,1 David I. Bellovin,1 Courtney Brown,2 Elizabeth Maynard,2 Bingyan Wu,3
Hisaaki Kawakatsu,4 Dean Sheppard,4 Peter Oettgen,2 and Arthur M. Mercurio1
1Division of Cancer Biology and Angiogenesis, Department of Pathology, and 2Division of Cardiology, Beth Israel Deaconess Medical Center and
Harvard Medical School, Boston, Massachusetts, USA. 3Department of Biostatistical Science, Dana-Farber Cancer Institute,
Boston, Massachusetts, USA. 4Lung Biology Center and Department of Medicine, UCSF, San Francisco, California, USA.
We used a spheroid model of colon carcinoma to analyze integrin dynamics as a function of the epithelial-
mesenchymal transition (EMT), a process that provides a paradigm for understanding how carcinoma cells
acquire a more aggressive phenotype. This EMT involves transcriptional activation of the β6 integrin subunit
and a consequent induction of αvβ6 expression. This integrin enhances the tumorigenic properties of colon
carcinoma, including activation of autocrine TGF-β and migration on interstitial fibronectin. Importantly, this
study validates the clinical relevance of the EMT. Kaplan-Meier analysis of β6 expression in 488 colorectal carci-
nomas revealed a striking reduction in median survival time of patients with high β6 expression. Elevated recep-
tor expression did not simply reflect increasing tumor stage, since log-rank analysis showed a more significant
impact on the survival of patients with early-stage, as opposed to late-stage, disease. Cox regression analysis con-
firmed that this integrin is an independent variable for these tumors. These findings define the αvβ6 integrin as
an important risk factor for early-stage disease and a novel therapeutic candidate for colorectal cancer.
A dynamic transition of epithelia into mesenchyme occurs dur-
ing normal morphogenetic processes such as embryonic develop-
ment, tissue remodeling, and wound repair by a process known
as the epithelial-mesenchymal transition (EMT; reviewed in refs.
1–3). Of particular interest to cancer biologists is the emerging
realization that the progression of epithelial-derived tumors
(carcinomas) may also involve spatial or temporal occurrences of
EMT (4–7), allowing carcinoma cells to acquire a more aggres-
sive phenotype. This hypothesis is supported by the observa-
tion that molecules whose expression is altered during the EMT,
such as E-cadherin, are often useful markers for prognosis. The
mechanism of the EMT is complex and involves signals from the
microenvironment, such as TGF-β, a well-characterized inducer
of the process in a variety of in vitro and in vivo settings (5, 6).
Although loss of E-cadherin may be a primal event for EMT,
alterations in other adhesion mechanisms must occur to gener-
ate a mesenchymal phenotype. In this respect, the paradigm of an
EMT is apt because an invasive carcinoma cell must acquire the
ability to interact with distinct interstitial matrices subsequent
to its transgression of the basement membrane. Accordingly, the
integrin family of adhesion molecules represents the major recep-
tors that mediate attachment to the ECM, with ligand occupancy
triggering critical intracellular signaling pathways (8, 9). Surpris-
ingly, however, little is known about how integrin expression and
function are regulated during EMT.
Here, we have used our recently characterized in vitro model
of EMT in colon carcinoma, which employs a multicellular, 3D
spheroid cell system (10), to analyze integrin dynamics as a func-
tion of EMT. We report that the EMT induces a marked increase
in the expression of the αvβ6 integrin, a receptor for fibronectin
and tenascin. This increase is dependent on the transcriptional
activation of the β6 integrin subunit mediated by Ets-1, a result
that implicates this proto-oncogene as a key regulator of EMT.
Further, the consequences of elevated αvβ6 expression are directly
linked to both the mechanism of the EMT itself and the function
of mesenchymal cells. The αvβ6 integrin promotes the activation of
autocrine TGF-β that sustains the EMT and is also required for the
migration of post-EMT cells on fibronectin. Moreover, this study
validates our EMT model as a valuable tool for the identification of
clinically relevant markers, as analysis of almost 500 colorectal car-
cinoma samples revealed that tumors with elevated αvβ6 expression
are associated with a significantly reduced survival time of patients
in comparison with tumors with no β6 expression or low β6 expres-
sion. Finally, our results define β6 expression as an independent
prognostic variable for colorectal cancer and, most significantly,
one that is predictive of outcome in early-stage disease.
Increased expression of αvβ6 integrin accompanies EMT of LIM 1863
organoids. We recently characterized a novel colon carcinoma
model for EMT, in which the well-differentiated LIM 1863 cell
line switches from a suspension culture of 3D spheroids (termed
organoids) to a migratory monolayer phenotype following expo-
sure to TGF-β (10). Further, we demonstrated that this bona
fide EMT conversion was accelerated in response to a synergistic
effect of TNF-α signaling (10). LIM 1863 organoids express 2 αv
Nonstandard abbreviations used: EMT, epithelial-mesenchymal transition.
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J. Clin. Invest. 115:339–347 (2005).
340 The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 2 February 2005
integrins, αvβ5 and αvβ6 (11), and the β5 subunit is expressed
in excess of the β6 subunit (Figure 1A). Following induction of
EMT, however, a striking increase in the surface expression of β6
occurs as assessed by surface labeling and immunoprecipitation
with an αv antibody (Figure 1A). To verify the identity of β6,
the αv immunoprecipitations were repeated and subjected to
immunoblotting with a β6 antibody (Figure 1B); this confirmed
that a dramatic induction of αvβ6 expression occurs during the
EMT. No changes in αvβ5 expression were evident (data not
shown). Flow cytometry revealed a greater–than–3-fold increase
in β6 expression following the transition (Figure 1C).
The time course of β6 induction during the EMT was evaluated
by immunoblotting (Figure 1D). Expression remained low during
the first 3 hours, with an initial increase apparent at 6–12 hours
and maximal expression seen at 24 hours. In contrast, expres-
sion of another cell surface receptor, the α6β4 integrin, remained
unchanged over the same time course (data not shown). We then
sought to determine whether this β6 upregulation occurs at the
mRNA level using real-time quantitative PCR. As shown in Fig-
ure 1E, we observed a greater–than–3-fold increase in mRNA level
for β6 in post-EMT cells as compared with organoids, a result
consistent with the flow cytometry data. Taken together, the data
Integrin αvβ6 expression increases following EMT. (A) LIM 1863 organ-
oids or cells harvested 24 hours after induction of EMT were surface
biotinylated, and the extracts were immunoprecipitated with an mAb
directed against integrin αv or control IgG. Relative molecular masses
are shown to the left in kDa. Arrows indicate the positions of αv, and
its associated chains β5 and β6. (B) Cell extracts were prepared from
organoids or from cells 24 hours after EMT, immunoprecipitated as
in A, and then immunoblotted with an mAb against the β6 integrin
subunit. Relative molecular masses are shown to the left in kDa.
(C) LIM 1863 organoids were either disaggregated into single-cell sus-
pensions (Organoid) or harvested 24 hours after induction of the EMT
(EMT), and surface expression of β6 was assessed by flow cytometry.
A greater–than–3-fold increase in β6 surface expression occurs after
the EMT. (D) Cell extracts were prepared over the time course shown
after EMT induction and immunoblotted with an anti-β6 antibody to
determine the kinetics of upregulation of the receptor. Relative molecu-
lar masses are shown to the left in kDa. (E) Integrin β6 mRNA lev-
els were quantified using real-time quantitative PCR in organoid or
EMT cultures of LIM 1863 cells. The fold change between treatments
(3.2-fold induction following the EMT) is represented graphically.
Transcriptional regulation of β6 by the
Ets-1 transcription factor. (A) Schematic
of the human integrin β6 promoter. The
transcription start site (TSS) and trans-
lation start site (ATG) are indicated.
Putative Ets-binding sites are shown
(triangles), and the 4 corresponding
sequences are listed, including loca-
tion detail. The consensus DNA-bind-
ing sequence (GGAA) is shown in bold.
(B) Transactivation of the β6 luciferase
reporter construct (–926/+208) by a
panel of Ets transcription factors com-
pared with the empty mammalian expres-
sion plasmid (PCI) in HEK293 cells. The
change in luciferase activity is expressed
as fold induction compared with PCI.
(C) Gel mobility shift assay for Ets-1
binding to putative Ets sites in the β6
promoter. In vitro–translated Ets-1 pro-
tein or control extract was used with end-
labeled oligonucleotide probes encoding
putative Ets sites 1–4, as indicated.
(D) Mutational analysis of the β6 promot-
er. Transactivation of either the wild-type
(black bars) or mutant Ets-1–binding site
–66/–63 (white bars) β6 luciferase report-
er constructs in response to increasing
doses of Ets-1 is shown.
The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 2 February 2005
indicate that upregulation of the β6 integrin subunit mRNA and
a consequent increase in αvβ6 expression occur during the EMT
of LIM 1863 colon carcinoma cells.
Transcriptional control of β6 integrin by the Ets-1 transcription factor.
Although transcriptional regulation of the β6 integrin has not
been studied, we reported recently that the expression and activ-
ity of the Ets-1 transcription factor are induced during the EMT
(12). This proto-oncogene is a member of a family of transcrip-
tion factors that regulate genes involved in development, cellular
differentiation, and cell proliferation (13). Interestingly, we iden-
tified 4 potential consensus Ets-binding sites in the first 1 kb
of the human β6 promoter (numbered 1–4, with 1 being most
proximal to the transcription start site; Figure 2A). To deter-
mine whether Ets-1 is capable of transactivating the β6 promot-
er, we performed luciferase assays with a number of Ets family
members (Figure 2B). Ets-1 displayed a robust activation of the
β6 promoter (greater than 2.5-fold), whereas none of the other
family members tested showed any inducing effect. The closely
related factor Ets-2, as well as Mef and NERF2, failed to activate
the promoter to a greater degree than the PCI control did, while
the others tested repressed activity (Figure 2B). Because all these
family members bind the same core recognition sequence, this
finding strongly argues for selectivity of Ets-1 in promoting β6
integrin transcription. To assess the relative contributions of
each of these 4 potential regulatory sites, we performed gel shift
assays. Ets-1 bound strongly to site 1, and no binding was evi-
dent for any of the other 3 more distal sequences (Figure 2C). To
substantiate this result, we performed site-directed mutagenesis
of the core recognition sequence GGAA to TTAA at this first site.
When increasing doses of wild-type Ets-1 were titrated into the
transcription assay, a clear dose-response effect of β6 promoter
activity was observed (Figure 2D). In contrast, mutagenesis of
site 1 completely abrogated any luciferase activity, independent
of dose, thus confirming the specificity and importance of this
Ets-1–binding site for regulating β6 transcription.
Elevated αvβ6 expression promotes tumorigenic functions following
EMT. To ascribe a functional role for increased αvβ6 expression
during the EMT, we focused initially on its classical role as a
fibronectin receptor (14). Although LIM 1863 organoids are non-
motile (11), they are capable of undergoing chemotaxis following
EMT (10). Indeed, using Transwells (Corning Inc.) coated with
specific matrix proteins, we observed that post-EMT cells were
significantly more chemotactic when the Transwells were coated
with fibronectin than when they were not coated or were coated
with laminin-1 (Figure 3A). This migration is mediated by αvβ6
and is not attributable to an alternative compensatory response
of the post-EMT phenotype, since it was inhibited by a function-
blocking β6 antibody (Figure 3B). These data indicate that induc-
tion of αvβ6-mediated migration on fibronectin is a consequence
of the EMT of LIM 1863 cells.
TGF-β is a central mediator of the EMT process (5, 6, 15), and
it is the exogenous application of this cytokine that drives the
EMT in the LIM 1863 spheroid model (10). Unexpectedly, we dis-
covered that LIM 1863 cells undergoing EMT exhibit autocrine
expression and secretion of TGF-β (Figure 4A). Organoids or
cytokine-treated cells were seeded for 7 hours in RPMI medium
supplemented with 5% FCS, washed 3 times with serum-free
medium to remove the endogenous TGF-β present in serum,
and then cultured for an additional 17 hours. Serum-free media
contained no TGF-β, whereas the final of the 3 washes contained
background levels derived from the FCS (Figure 4A). TGF-β pro-
duction by intact organoids was negligible compared with the
residual TGF-β levels of the controls. In contrast, post-EMT cells
secreted high levels of TGF-β, confirming the existence of auton-
omous TGF-β production in these cells.
TGF-β family members are secreted as latent complexes that
require activation in order to bind their receptors, and an essential
role for αvβ6 to bind and activate latent TGF-β has emerged (16).
Increased αvβ6 expression promotes migration on fibronectin. (A)
Chemotactic migration assay of post-EMT LIM 1863 cells for 3 days on
untreated control Transwells (Con), or Transwells coated with laminin
(Lm) or fibronectin (Fn). Data are expressed as means and SDs of 8 indi-
vidual fields randomly selected from each well, with each experiment per-
formed in triplicate. *P < 0.05. (B) LIM 1863 cells were subjected to the
chemotaxis assay as described, on untreated (Con) or fibronectin-coat-
ed Transwells as indicated. Chemotaxis on fibronectin was performed
in the absence (–) or presence of function-blocking anti-β6 monoclonal
(10D5) or isotype-matched control (IgG) antibody, at a concentration of
100 μg/ml. The anti-β6 antibody reduced chemotaxis back to the levels
of the uncoated migration control for these cells. *P < 0.05.
Autocrine TGF-β production and activation by αvβ6 in EMT cells.
(A) ELISA was performed to measure secreted TGF-β levels in serum-
free media (Media), media after washing (Wash), or overnight culture
media from organoid or EMT cultures of LIM 1863 cells (Org, EMT).
(B) Reporter and LIM 1863 cells were cultured 16–20 hours and lysed
for measurement of luciferase activity. Results are for organoids (Org)
and EMT cells (EMT). Addition of function-blocking anti-β6 monoclo-
nal (3G9, 30 μg/ml) or control antibody (Con) is shown at the bottom.
Relative luciferase activity (RLU) is the measured activity divided by
the activity of the coculture of organoid (control) cells in the presence
of function-blocking antibody. *P < 0.001.
342 The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 2 February 2005
Therefore we assessed whether the αvβ6 induced during the EMT
is capable of activating TGF-β, by coculturing intact organoids
or post-EMT cells with mink lung epithelial reporter cells sta-
bly expressing a portion of the plasminogen activator inhibitor
1 promoter, as described previously (16, 17). Coculture with the
post-EMT cells caused a significant increase in luciferase levels
compared with coculture with control (organoid) cells (Figure 4B).
Significantly, this increase was abolished by a function-blocking
mAb against β6 (3G9), which indicates that this effect is com-
pletely attributable to the increased expression of αvβ6 receptor
on these cells. Taken together, these results show that LIM 1863
cells induced to undergo an EMT not only produce an autocrine
supply of TGF but also have the capacity to process and activate
the latent form of the cytokine using the αvβ6 integrin.
EMT and αvβ6 expression in vivo. Having found that upregulation
of αvβ6 is a consequence of the EMT process in vitro, we sought
to establish a relationship among EMT, β6 expression, and tumor
progression in vivo. To achieve this, LIM 1863 xenograft tumors
were grown s.c. in nude mice, before immunohistochemical
analysis of relevant markers. As expected, these organoid-derived
tumors stained only weakly for β6 (Figure 5A). However, there was
a dramatic induction of β6 expression in localized regions of the
tumors, such as is shown in Figure 5A, and in particular on tumor
cells that were invading the stroma. Thus, tumor cells that were
migrating out from an invasive front displayed a clear increase
in receptor expression. Significantly, when a sequential tumor
section was stained for E-cadherin, the in situ tumor cells were
strongly positive, whereas the invading cells had completely lost
expression of this molecule (Figure 5B). Based on our model, we
conclude that the invading cells had undergone an EMT in vivo
(most likely the result of stromal factors), resulting in loss of the
epithelial marker E-cadherin and a concomitant upregulation of
β6. These morphological data not only support our hypothesis
that elevated β6 expression facilitates the invasion and dissemina-
tion of colon carcinoma cells but also provide direct evidence that
the EMT correlates with disease progression.
High αvβ6 expression in human colon carcinoma is a prognostic indi-
cator of poor survival. Based on our EMT data, we predicted that
the frequency of αvβ6 expression in human disease is skewed
toward late-stage, more aggressive tumors, rather than adenomas
and in situ carcinomas. To examine this postulate, we performed
immunohistochemistry on tissue arrays of malignant colorectal
tumors with known clinical outcome (Figure 6). Table 1 summa-
rizes the patient characteristics for 488 successfully stained pri-
mary carcinoma samples, of which 181 (37%) were positive for β6
expression. The median age of patients was 68 years, with a range
from 23 to 94. For subsequent analysis, the staining intensity of β6
was scored as 0, 1, or 2. Of the 488 samples, 307 (63%) were nega-
tive (score 0), 95 (19%) exhibited low expression (score 1), and 86
(18%) exhibited high expression (score 2) (Table 1).
Normal colonic mucosa did not exhibit any staining for β6
(Figure 6A). A tumor sample negative for β6 expression is shown in
Figure 6B, and, in contrast, an example of strong immunoreactivity
of β6 expression (score 2) is shown in Figure 6C. Consistent with
the hypothesis that αvβ6 expression occurs as a function of tumor
progression, high β6 expression was detected in the more poorly
differentiated infiltrating tumor cells invading through desmo-
plastic stroma (Figure 6D). Heterogeneous distribution patterns
were frequently observed (e.g., Figure 6E) that displayed an intense
upregulation and preferential localization of the integrin at the
edges of infiltrating tumor islands. A sequential serial section of
this particular tumor was used as a negative control (Figure 6F).
We then examined whether there was an association between
αvβ6 expression and patient survival. We analyzed the data by con-
sidering only disease-related death as an event, censoring deaths
unrelated to disease and the patients who were alive when they were
last seen. To determine differences in the cancer survival among
patients who differed in αvβ6 expression (scores 0–2), the log-rank
test was performed, and the corresponding Kaplan-Meier plot is
shown in Figure 7. The log-rank test indicated there was no signifi-
cant survival difference between patients with score 0 and patients
with score 1 (P = 0.99). There was, however, a significant difference
The integrin αvβ6 is a marker of EMT in vivo. Integrin β6 (A) or E-cadherin
(B) immunostaining of sequential tumor sections derived from LIM
1863 xenografts. E-cadherin expression is prominent in the tumor
tissue (T) and absent from the stroma (S). β6 Immunoreactivity is
relatively weak in the tumor masses, but it is strongly upregulated in
tumor cells invading the stroma, which are also negative for E-cadherin
expression (arrows). Scale bar: 50 μm.
Expression of αvβ6 in malignant human colon carcino-
ma. (A–E) Representative β6 immunostaining of normal
human colon (A) or malignant colon carcinoma tissue
(B–E). A negative (B) and a positive (C) tumor sample
are shown. β6 Immunostaining is shown for a separate
carcinoma sample in D, with tumor cells infiltrating the
stroma showing high expression of the receptor. Hetero-
geneous receptor expression in another primary tumor,
with intense upregulation and preferential localization
to tumor islets, is illustrated in E. (F) The corresponding
negative control for the tumor shown in E. Scale bars:
100 μm (A–C, E, and F) and 50 μm (D).
The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 2 February 2005
between patients with score 0 and patients with score 2 (P = 0.048),
and between patients with score 0 or 1 and patients with score 2
(P = 0.04). Further, the Kaplan-Meier plot showed that patients
with score 0 or 1 (i.e., either no expression or low expression of β6)
had longer survival times than patients with score 2 (high expres-
sion of β6). For this reason, we combined the patients with a score
of 0 or 1 in the data set and compared their survival with that of
the patients with score 2. The survival estimates showed a striking
difference in median survival between the 2 groups: the group with
a score of 0 or 1 averaged 16.5 years, whereas that with a score of 2
averaged less than 5 years and showed a 15% lower 5-year survival
rate (Table 2). From a biological perspective, we conclude that ele-
vated expression of αvβ6 is a prognostic indicator of poor survival
for patients with colorectal carcinoma. This key finding substanti-
ates the data predicted by the spheroid model.
The log-rank results suggested a significant interaction
between the effects of score and stage, and it was possible that the
survival effects of β6 simply correlated with the degree of inva-
sion. For this reason, we separated the expression data by stage,
either early (stages 1 and 2) or late (stages 3 and 4) (Table 2).
For patients with stage 3 or 4 tumors, high expression of β6
resulted in a minor (4–5%), not significant, reduction in both
5- and 10-year survival rates. In stark contrast, elevated β6
expression had a much more significant impact on the survival
of those patients with early-stage tumors (P = 0.0015), high-
lighted by a greater-than-28% reduction in 5-year survival com-
pared with that of patients with no expression or low expression.
Importantly, these data define β6 expression as an independent
prognostic variable for colorectal cancer that is predictive of
outcome in early-stage disease.
Based on the above findings, the proportional hazard ratio
model was used to obtain more precise estimates of the effects of
β6 expression on survival by adjusting for the effects of age, sex,
stage, and tumor type. Table 3 summarizes the χ2 P values, the
hazard ratios, and the 95% confidence interval of hazard ratio,
for all variables. Again, when score 0 and 1 patients (no expres-
sion or low expression) were compared with score 2 patients (high
expression), the P value of 0.007 revealed a significant difference
in survival. As expected, the standard prognostic indicators of
tumor stage and age at detection each showed significant effects
on survival. Significantly, the hazard ratio of 1.6 for elevated β6
expression indicated that the hazard of dying was 60% higher
in patients with score 2 than in patients with score 0 or 1. This
result confirmed the findings of the log-rank test in identifying
β6 expression as an independent variable and, importantly, sug-
gested that high β6 expression in colon tumors represented a sig-
nificant risk factor for patient survival.
Expression of αvβ6 in human colon cancer metastases. In Figure 8
we provide evidence of αvβ6 expression in human colon car-
cinoma metastases as assessed by immunohistochemistry on
lymph nodes and liver specimens obtained from patients with
colorectal cancer. Infiltrating colon carcinoma cells were easily
discernible by H&E staining of a section of lymph node from
a patient with invasive colon cancer (Figure 8A). β6 Staining of
the same tissue showed strong immunoreactivity in the tumor
cells, while the lymphatic cells of the organ were negative, as
expected (Figure 8B). To define receptor expression in more dis-
tal organs, we stained metastatic lesions in randomly selected
liver biopsies (Figure 8, C and D). The colonic metastases were
readily distinguished from the hepatic tissue by H&E staining
(Figure 8C), and these tumor cells again displayed strong β6
staining (Figure 8D; 3 of 4 cases). Negative-control tissue stain-
ing showed no immunoreactivity, as expected (data not shown).
Characteristics No. of patients Percent
Lymph node status
Other digestive organ
Score (β6 expression)
Kaplan-Meier survival analysis using a log-rank test. The samples
were grouped according to β6 expression level (score 0, negative;
score 1, low expression; score 2, high expression) and analyzed using
the log-rank test for overall survival.
344 The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 2 February 2005
The demonstration of αvβ6 expression in distal metastases sup-
ports our conclusion that αvβ6 expression in colorectal cancer
correlates with disease progression and poorer clinical outcome.
The progression to invasive and metastatic carcinoma involves
profound alterations in epithelial structure and function. We
show here that the EMT of colon carcinoma is coincident with
an increase in αvβ6 integrin expression, and the culmination
of our observations is that high expression of β6 in primary
colon carcinomas is a prognostic marker for aggressive disease
and patient mortality. This integrin is expressed primarily dur-
ing development, and high αvβ6 expression in adults is limited
only to a few epithelial tissues (18, 19). However, αvβ6 can be
reexpressed in parallel with specific morphogenetic events such
as inflammation and wound healing (19). Wound healing, in
particular, is believed to involve EMT processes, and the induc-
tion of αvβ6 implies a central role for the receptor (19–21). In
addition, several studies have shown neo-expression of αvβ6
in oral squamous cell carcinomas (19, 22, 23), although a link
between expression and patient outcome has not been estab-
lished. Nonetheless, these studies are consistent with our obser-
vation that αvβ6 expression is induced as a consequence of
EMT. Interestingly, studies on the phenotypic changes induced
by the EMT have traditionally focused on the transcriptional
repression of epithelial markers that maintain tissue polarity,
such as E-cadherin, and the functional consequences of induced
mesenchymal markers. Our discovery that an epithelial integrin
is dynamically regulated during the EMT of colon carcinoma,
and specifically one that is upregulated during development,
is consistent with the long-standing hypothesis that colorectal
carcinogenesis more accurately reflects a dedifferentiation of
epithelia to a more embryonic phenotype.
The induction of αvβ6 expression that we observed during
the EMT is dependent on the transcriptional upregulation of
the β6 integrin subunit. The transcriptional regulation of β6
had not been investigated previously to our knowledge, and it
now appears that Ets-1 may be a key transcription factor that
influences EMT. In fact, we have recently reported that Ets-1
expression is upregulated as a consequence of the EMT, and
that it is responsible for the induction of the VEGF receptor
Flt-1 (12). A functional role for Ets-1 in β6 regulation was evi-
denced by its ability to bind and transactivate the human β6
promoter. Our screening identified 4 putative consensus sites
within the first kilobase of the promoter sequence, but only
the first site appeared to be of importance. Ets-1 bound pref-
erentially to this site in gel shift assays, and, most significantly,
site-directed mutagenesis of this region completely abrogated
transactivation activity. Since Ets-1 typically exerts its effects in
cooperation with other transcriptional factors and cofactors, we
are currently seeking to identify other regulators of β6 transcrip-
tion. The role, if any, of the other 3 consensus sites also awaits
further characterization. Although they appear dispensable for
β6 regulation during the EMT process, the possibility exists that
alternate triggers, or tissue-specific factors, might act to influ-
ence β6 gene regulation via these sites.
We define 2 key functions of αvβ6 that can be linked to the biol-
ogy of aggressive colon carcinoma: activation of autocrine TGF-β,
and promotion of migration on fibronectin matrices. Firstly,
TGF-β has been proposed to act as a cell-autonomous promoter
of late-stage human tumor development because of its ability
to regulate EMT (5). In this context, the autocrine TGF-β pro-
duction we observed would provide the means to stabilize and
sustain the EMT by continuous TGF-β receptor signaling in a
cell-autonomous fashion. However, it is the ability of αvβ6 to
activate autocrine TGF-β in post-EMT cells that is most novel
with respect to colon carcinoma progression. The recogni-
tion that this integrin can bind and activate latent TGF-β (16)
identified a likely role for this receptor in contributing to the
onset of TGF-β–mediated diseases, such as inflammation and
fibrosis (24, 25). From our perspective, induced expression of
αvβ6 in colon carcinoma would provide a mechanism to locally
regulate TGF-β function in vivo, to provide a feedback loop to
perpetuate the EMT process, and, in turn, to create a tumor
microenvironment more amenable to progression. Indeed, the
morphological analysis of the xenografts (Figure 5) provides in
vivo evidence for this concept.
Survival estimates for disease-related deaths
Stage 1–2 and score 0–1
Stage 1–2 and score 2
Stage 3–4 and score 0–1
Stage 3–4 and score 2
Median survival (yr) and 95% CI
16.5 (8.2, not estimable)
4.8 (3.2, not estimable)
5.0 (3.2, not estimable)
3.8 (2.9, 6.9)
4.4 (1.8, not estimable)
5-Year survival rate (%) and 95% CI
61.7 (56.6, 66.7)
46.7 (35.5, 57.8)
77.5 (71, 84)
49.3 (35, 63.6)
45.7 (38.4, 53.1)
41.7 (22.8, 60.6)
10-Year survival rate (%) and 95% CI
53.7 (48.3, 59.0)
41.7 (30.3, 52.9)
69.6 (62.2, 77.1)
46.4 (31.9, 61)
37.8 (30.4, 45.1)
32.4 (13.9, 51)
Log-rank test suggested significant survival difference between groups: stage 1–2 vs. 3–4 (P < 0.0001), score 0–1 vs. 2 (P = 0.04), and score 0–1 vs. 2
among patients at stage 1–2 (P < 0.0015). There was no significant difference between score 0–1 and score 2 among patients at stage 3–4 (P = 0.65).
CI, confidence interval.
Multivariate analysis for disease-related deaths (Cox regression
Group (0–1 vs. 2)
Stage (1–2 vs. 3–4)
(colon vs. rectum/other)
The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 2 February 2005
Secondly, having a specific fibronectin receptor elicited as a
consequence of the EMT also has selective advantages for tumor
cells. The colonic epithelium resides upon a basement membrane,
composed primarily of laminins, collagen IV, and proteoglycans,
which provides an interface with the lamina propria beneath (26).
The structural composition of this zone differs in its high fibril-
lar collagen content, and also in enrichment with fibronectin
and tenascin, both ligands for αvβ6 (14, 27). Thus, an enhanced
migratory capacity on fibronectin would facilitate the escape and
dissemination of invading colon carcinoma cells into the lamina
propria. Moreover, αvβ6 expression and fibronectin recognition
may also be important for colonization of the liver, the primary
site for metastases derived from colorectal tumors. In addition
to the anatomical consideration that portal blood flow from
the colon goes directly to this organ, our data suggest that there
may be an additional dimension to colonization that specifically
involves αvβ6. Since there is no basement membrane under the
endothelium of liver microvessels as in other tissues, it has been
suggested that adhesion to fibronectin may be of central impor-
tance for liver metastasis, because it is abundant on the surface
of hepatocytes (28). We propose, therefore, that an increased pro-
pensity to adhere to fibronectin would promote colon carcinoma
metastasis to the liver by aiding extravasation.
The key conclusion of this study relates to the expression of αvβ6
in human colorectal cancer. We identified this receptor as a mark-
er of aggressive carcinoma, in part based on its high frequency of
expression (almost 40%) in malignant tumors. Additional compel-
ling evidence for the involvement of αvβ6 in tumor progression is
provided by our demonstration of β6 expression in both lymphatic
and hepatic metastases derived from human colorectal carcino-
mas. However, the result of greatest consequence is that analysis
of the clinical data revealed that elevated β6 expression, rather
than the simple presence or absence of the receptor, is linked to
significantly reduced survival times. Even more striking was the
discovery that its value as a prognostic marker is more significant
for early-stage tumors. It is likely that the effects of β6 expression
on late-stage tumors are masked by additional genetic and epigen-
etic alterations. However, based on our results, it is tempting to
speculate that the EMT process can contribute directly to tumor
progression and development of metastatic disease, and that αvβ6
represents a novel marker for this event. Moreover, the hazard
ratio analysis confirmed that this receptor is a stand-alone vari-
able, similar to those of stage and age at diagnosis, for colorectal
tumors, and indicated that high expression of this integrin is an
independent and important risk factor for patient survival. Over-
all, these findings establish αvβ6 as a novel prognostic indicator
for human colon cancer and validate our EMT model as a valuable
tool for the identification of clinically relevant markers.
Finally, a potentially important consequence of our work is that
αvβ6 may be an attractive therapeutic candidate for colon can-
cer and, in particular, that it may be possible to selectively target
invading and metastasizing cells. Furthermore, because elevated
β6 expression is predictive of outcome in early-stage disease, it may
also be a feasible target for earlier intervention and treatment. Tar-
geting integrin function has emerged as a legitimate therapeutic
strategy for several disease types. In this regard, the current devel-
opment and characterization of function-blocking anti-β6 anti-
bodies (29) offer the promise of new reagents and novel treatments
not only for diseases such as fibrosis but also, in light of the find-
ings presented here, for aggressive colorectal cancer.
Cell culture and EMT induction. LIM 1863 cells (11, 30) were grown as
organoids in RPMI 1640 (GIBCO; Invitrogen Corp.) supplemented with
5% FCS. To induce the EMT, LIM 1863 cells were seeded in 24-well plates
with a combination of TNF-α (10 ng/ml) and TGF-β1 (2 ng/ml) for 24
hours (10). Morphological changes in the organoids were assessed by
phase-contrast microscopy. The cell line HEK293 (human embryonic
kidney) was grown in DMEM supplemented with 10% FCS and antibiot-
ics (penicillin and streptomycin) (GIBCO; Invitrogen Corp.). Mink lung
epithelial cells stably transfected with a plasmid containing the luciferase
cDNA downstream of a TGF-β–sensitive portion of the plasminogen acti-
vator inhibitor 1 promoter were grown in DMEM supplemented with
4.5 g/l glucose and 10% FCS.
Antibodies. The murine mAb directed against αv integrin, 13C2, was
a kind gift from M. Horton (Imperial Cancer Research Fund, London,
United Kingdom). The anti-β6 mAbs 3G9 and 2G2 were obtained from
S. Violette (Biogen Idec, Cambridge, Massachusetts, USA), and the speci-
ficity of these antibodies has recently been reported (29). The anti-β6
antibody CSβ6 and function-blocking monoclonal 10D5 were purified
in our laboratories, and control mouse Igs were purchased from Sigma-
Aldrich. The polyclonal E-cadherin antibody (H-108) was purchased
from Santa Cruz Biotechnology Inc.
Real-time quantitative PCR. Real-time quantitative PCR was performed
on RNA samples that were first reverse-transcribed and then treated with
DNase. The primers and probes (listed below) were designed using Prim-
er Express software version 1.0 (Applied Biosystems), based on mRNA
sequences obtained from the National Center for Biotechnology Infor-
mation database (31). All reactions were performed in an ABI PRISM
7700 Sequence Detection System (PerkinElmer Applied Biosystems).
Reactions were carried out in triplicate in a 50-μl reaction volume con-
taining 25 μl of 2× TaqMan PCR Master Mix, a 50-nM concentration
of each forward and reverse primer, a 100-nM concentration of dual-
labeled probe, and 1 μg of total cDNA. Conditions for all PCR reactions
were 2 minutes at 50°C and 10 minutes at 95°C, followed by 40 cycles
of 95°C for 10 seconds and 60°C for 1 minute. Normalization to actin
In vivo αvβ6 protein expression in human colorectal metastases. Rep-
resentative H&E (A and C) or β6 immunohistochemistry (B and D)
in lymph node (A and B) or liver tissue (C and D) containing human
colorectal metastases. β6 Immunoreactivity in both samples is restrict-
ed to the metastasized tumor cells. Scale bars: 100 μm (A and B) and
50 μm (C and D).
346 The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 2 February 2005
(housekeeping gene) was performed for each sample. Cycle threshold
values were exported into an Excel (Microsoft Corp.) worksheet for cal-
culation of fold changes according to the ΔΔCT method. The primers
and dual-labeled probe (TET/TAMRA) were as follows: β6 integrin sense,
5′-CTACCTGTGGTGACCCCTGTAAC-3′; β6 integrin antisense, 5′-
GCTTGGCCAGCTGCTGAC-3′; β6 integrin probe, 5′/5TET/CTAAAC-
Expression vector and luciferase reporter gene constructs. An 1,134-bp frag-
ment corresponding to nucleotides –926 to +208 of the human ITGβ6
promoter was cloned from human genomic DNA (BD Biosciences —
Clontech) by PCR and subcloned into the pGL2 luciferase reporter vec-
tor (Promega Corp.). The fragment was inserted into the NheI-XhoI site
upstream of the luciferase gene of the pGL2 vector.
DNA transfection assays. Cotransfections of 2.5 × 105 HEK293 cells
were carried out with varying amounts of reporter gene construct DNA
and expression vector DNA using 4 μl LipofectAMINE (Invitrogen
Corp.) as previously described (32). The cells were harvested 16 hours
after transfection and assayed for luciferase activity. Transfections
for every construct were performed independently and in duplicate.
Cotransfection of a second plasmid for determination of transfection
efficiency was omitted, because potential artifacts with this technique
have been reported and because many commonly used viral promoters
contain potential binding sites for Ets factors (33).
Site-directed mutagenesis. Site-directed mutagenesis of the ITGβ6 was per-
formed using the QuikChange Site-Directed Mutagenesis Kit (Stratagene)
according to the manufacturer’s recommendations. In brief, PCR primers
encoding the ITGβ6 promoter Ets site, –66 to –63, and flanking sequences
were used, with TTAA substituted for GGAA (mutation underlined): 5′-
PCR was performed with PfuTurbo polymerase (Stratagene) using the
wild-type ITGβ6 promoter luciferase reporter construct as a template. The
PCR reaction was digested with DpnI, and the undigested plasmids were
transformed into DH5α bacteria. Individual plasmid preparations were
sequenced to verify incorporation of the Ets site mutation.
Electrophoretic mobility shift assay. Electrophoretic mobility shift assays
were performed as described previously (34). In brief, 2 μl of in vitro trans-
lation product and 0.1–0.2 ng [32P]dATP-labeled double-stranded oligo-
nucleotide probes (5,000–20,000 cpm) were run on 4% polyacrylamide
gels containing 0.5× Tris-Borate-EDTA buffer. Oligonucleotides used as
probes were as follows (Ets consensus binding site underlined): human
ITGβ6 promoter site –73 to –51, 5′-AGAACAAGGAAGTAAATCATGTT-3′
Immunoprecipitation and immunoblot analysis. To assess β6 integrin protein
levels, cells were extracted in a Triton X lysis buffer (1% Triton X-100, 50
mM Tris, 150 mM NaCl) containing protease inhibitors (pepstatin, PMSF,
aprotinin, and leupeptin) for 1 hour. Extracts were clarified by centrifuga-
tion. For immunoprecipitations, the lysates were precleared twice with
anti-mouse IgG-agarose (Sigma-Aldrich). Immunoprecipitations were
carried out indirectly using the monoclonal 13C2 or isotype-matched
control IgG and captured with anti-mouse IgG-agarose. Extracts and pre-
cipitates were analyzed by SDS-PAGE, and proteins were transferred to
nitrocellulose by electrophoresis. Residual protein sites were blocked in
Tween/Tris-buffered saline (TBST) containing 5% skim milk. The filters
were incubated with an mAb directed against β6 (2G2) in TBST plus 2.5%
skim milk for 1 hour and developed using ECL.
Xenograft studies. LIM 1863 organoids (approximately 3 × 106 cells) were
inoculated s.c. into the flanks of female nude mice. The animals were
6–7 weeks of age at the time of tumor implantation. Tumor xenografts
were harvested 16 weeks later, and formalin-fixed, paraffin-embedded
sections were used for immunohistochemistry. All animal studies were
approved by the Institutional Animal Care and Use Committee (IACUC)
of Beth Israel Deaconess Medical Center.
Flow cytometry. Cells were incubated with either the anti-β6 mAb 3G9 or
an isotype-matched control IgG (10 μg/ml in PBS) for 30 minutes. After
washing, cells were resuspended in FITC-conjugated secondary antibody
for 30 minutes in the dark, prior to FACScan analysis (BD Biosciences).
Immunohistochemistry. Immunohistochemistry was performed using
the avidin-biotin complex protocol on paraffin-embedded tissue. Tis-
sue microarray slides were generated as previously described (35). Briefly,
formalin-fixed, paraffin-embedded tissue blocks containing colon car-
cinoma specimens were retrieved from the archives of the Yale Univer-
sity Department of Pathology (New Haven, Connecticut, USA) under
Human Investigative Committee (HIC) protocol no. 8219. To generate the
microarray, areas of invasive carcinoma were selected for coring and place-
ment into a recipient master block using a Tissue Microarrayer (Beecher
Instruments Inc.). Cores measured 0.6 mm in greatest dimension and were
spaced 0.8 mm apart. The tissue microarray was sliced into 5-μm sections
and adhered to slides by an adhesive tape-transfer system (Instrumed-
ics Inc.) and UV cross-linking. For the metastasis staining, 4 randomly
selected cases of liver metastasis and 2 samples of lymph node containing
colorectal metastases were selected for study from the Beth Israel Deacon-
ess Medical Center, Department of Pathology (tissue procurement was
approved by the Institutional Review Board). For immunohistochemistry,
the slides were deparaffinized, and endogenous peroxidase activity was
blocked by incubation in a hydrogen peroxide/methanol buffer. Antigen
retrieval was then performed by incubation of the slides in a pepsin solu-
tion (Zymed Laboratories Inc.) at 37°C. Overnight incubation at 4°C with
primary antibody (2G2, 2 μg/ml; E-cadherin, 0.04 μg/ml) in 0.1% BSA was
preceded by blocking with 1× casein solution (Vector Laboratories Inc.).
Negative controls were performed by substitution of the primary anti-
body with PBS. The following day, biotinylated anti-mouse IgG (1:200;
DakoCytomation) was applied to the slides, which were subsequently
treated using a Vectastain ABC kit (Vector Laboratories Inc.). Slides were
incubated for 10 minutes with a peroxidase solution (DAKO EnVision+
System, Peroxidase [DAB]; DakoCytomation), stained with hematoxylin,
dehydrated, cleared, and mounted.
Migration assay. Migration assays were performed by assessment of the
ability of cells to migrate toward 3T3-conditioned medium using 6.5-mm
Costar Transwell chambers (8 μm pore size; Corning Inc.). Where indicated,
Transwell membranes were coated on both surfaces with 20 μg/ml laminin-1
(Roche Diagnostics Corp.) or 50 μg/ml human cellular fibronectin (Sigma-
Aldrich) by immersion overnight at 4°C, before aspiration and 3 washes in
PBS. LIM 1863 organoids were treated with TNF-α and TGF-β to induce
an EMT and added to the upper chambers, and 3T3-conditioned medium
was added to the lower wells of the chambers to induce chemotaxis. For
antibody-inhibition studies, cells were seeded in medium containing either
no antibody, anti-β6 (10D5), or an isotype-matched control IgG antibody
(100 μg/ml). After 3 days, cells were removed from the upper face of the
filters using cotton swabs, and the cells that had migrated to the lower sur-
face were fixed in methanol. Filters were mounted onto microscope slides
using VECTASHIELD mounting medium with DAPI (Vector Laboratories
Inc.), and migration was quantified by visual counting using fluorescence
microscopy. Assays were carried out in triplicate, and the results presented
are the means of 8 random fields from each well.
TGF-β bioassay. In 96-well plates, LIM 1863 cells were cultured in the
presence and absence of TNF-α and TGF-β for 24 hours (in 100 μl of
RPMI plus 5% FCS). After 3 gentle washes with DMEM containing 10%
FCS, 2 × 104 reporter cells (17) were seeded onto the LIM 1863 cells in
100 μl per well in the absence or presence of the β6-blocking antibody
The Journal of Clinical Investigation http://www.jci.org Volume 115 Number 2 February 2005
3G9 (30 μg/ml) or isotype-matched control (CSβ6). Cells were cultured
for an additional 16–20 hours, and lysates were assayed for luciferase
activity as previously described (16).
Statistical methods. Kaplan-Meier method was used to estimate median sur-
vival, 5-year survival, and 10-year survival along with their 95% confidence
intervals. Log-rank test was used to examine the differences in the overall
survival between groups. The proportional hazard ratio model was used
to obtain more precise estimates of the effects of protein expression on
survival by adjusting for the effects of age, gender, tumor type, and cancer
stage. In all analyses, only disease-related deaths were treated as events.
This work was supported by NIH grants CA 80789 and CA107548 (to
A.M. Mercurio) and by the Harvard Digestive Diseases Center (NIH
grant DK34854; to R.C. Bates). We wish to thank Jeffrey Goldsmith
(Beth Israel Deaconess Medical Center), David Rimm (Yale Universi-
ty, New Haven, Connecticut, USA), and Shelia Violette (Biogen Idec,
Cambridge, Massachusetts, USA) for invaluable contributions to this
manuscript by way of reagents, techniques, and expertise.
Received for publication August 26, 2004, and accepted in revised
form November 30, 2004.
Address correspondence to: Richard C. Bates, Department of Pathol-
ogy, Division of Cancer Biology and Angiogenesis, Beth Israel Dea-
coness Medical Center, Research North, Room 220, 99 Brookline
Avenue, Boston, Massachusetts 02215, USA. Phone: (617) 667-2816;
Fax: (617) 975-5531; E-mail: firstname.lastname@example.org.
1. Hay, E.D. 1995. An overview of the epithelio-
mesenchymal transformation. Acta Anat. 154:8–20.
2. Savagner, P. 2001. Leaving the neighborhood:
molecular mechanisms involved during epithelial-
mesenchymal transition. Bioessays. 23:912–923.
3. Theiry, J.P. 2003. Epithelial-mesenchymal transi-
tions in development and pathologies. Curr. Opin.
Cell Biol. 15:740–746.
4. Oft, M., et al. 1996. TGF-beta1 and Ha-Ras collabo-
rate in modulating the phenotypic plasticity and
invasiveness of epithelial tumor cells. Genes Dev.
5. Oft, M., Heider, K.H., and Berg, H. 1998. TGFβ sig-
naling is necessary for carcinoma cell invasiveness
and metastasis. Curr. Biol. 8:1243–1252.
6. Bhowmick, N.A., et al. 2001. Transforming growth
factor-β1 mediates epithelial to mesenchymal
transdifferentiation through a Rho-A-dependent
mechanism. Mol. Biol. Cell. 12:27–36.
7. Thiery, J.P. 2002. Epithelial-mesenchymal tran-
sitions in tumour progression. Nat. Rev. Cancer.
8. Hynes, R.O. 1992. Integrins: versatility, modula-
tion, and signaling in cell adhesion. Cell. 69:11–25.
9. Ruoslahti, E. 1991. Integrins. J. Clin. Invest. 87:1–5.
10. Bates, R.C., and Mercurio, A.M. 2003. Tumor
necrosis factor-α stimulates the epithelial to
mesenchymal transition of human colonic organ-
oids. Mol. Biol. Cell. 14:1790–1800.
11. Bates, R.C., Buret, A., van Helden, D.F., Horton,
M.A., and Burns, G.F. 1994. Apoptosis induced
by inhibition of intercellular contact. J. Cell Biol.
12. Bates, R.C., et al. 2003. Flt-1-dependent survival
characterizes the epithelial-mesenchymal transition
of colonic organoids. Curr. Biol. 13:1721–1727.
13. Oikawa, T., and Yamada, T. 2003. Molecular biol-
ogy of the Ets family of transcription factors. Gene.
14. Busk, M., Pytela, R., and Sheppard, D. 1992. Char-
acterization of the integrin αvβ6 as a fibronectin-
binding protein. J. Biol. Chem. 267:5790–5796.
15. Lehmann, K., et al. 2000. Raf induces TGFβ produc-
tion while blocking its apoptotic but not invasive
responses: a mechanism leading to increased malig-
nancy in epithelial cells. Genes Dev. 14:2610–2622.
16. Munger, J.S., et al. 1999. The integrin αvβ6 binds
and activates latent TGFβ1: a mechanism for regu-
lating pulmonary inflammation and fibrosis. Cell.
17. Abe, M., et al. 1994. An assay for transforming
growth factor-beta using cells transfected with
a plasminogen activator inhibitor-1 promoter-
luciferase construct. Anal. Biochem. 216:276–284.
18. Breuss, J.M., Gillett, N., Lu, L., Sheppard, D., and
Pytela, R. 1993. Restricted distribution of integrin
beta 6 mRNA in primate epithelial tissues.
J. Histochem. Cytochem. 41:1521–1527.
19. Breuss, J.M., et al. 1995. Expression of the β6
integrin subunit in development, neoplasia and
tissue repair suggests a role in epithelial remodel-
ing. J. Cell Sci. 108:2241–2251.
20. Zambruno, G., et al. 1995. Transforming growth
factor-beta 1 modulates beta 1 and beta 5 integrin
receptors and induces the de novo expression of
the alpha v beta 6 heterodimer in normal human
keratinocytes: implications for wound healing.
J. Cell Biol. 129:853–865.
21. Pilewski, J.M., Latoche, J.D., Arcasoy, S.M., and
Albelda, S.M. 1997. Expression of integrin cell
adhesion receptors during human airway epithelial
repair in vivo. Am. J. Physiol. 273:L256–L263.
22. Hamidi, S., et al. 2000. Expression of
alpha(v)beta6 integrin in oral leukoplakia. Br. J.
23. Regezi, J.A., Ramos, D.M., Pytela, R., Dekker, N.P.,
and Jordan, R.C.K. 2002. Tenascin and β6 integrin
are overexpressed in floor of mouth in situ carcino-
mas and invasive squamous cell carcinomas. Oral
24. Pittet, J.F., et al. 2001. TGF-β is a critical mediator
of acute lung injury. J. Clin. Invest. 107:1537–1544.
25. Morris, D.G., et al. 2003. Loss of integrin
alpha(v)beta6-mediated TGF-beta activation
causes Mmp-12-dependent emphysema. Nature.
26. Bates, R.C. 2003. Gastrointestinal matrix, organi-
zation and significance. In Encyclopedia of gastroen-
terology. L.R. Johnson, editor. Elsevier Science. San
Diego, California, USA. 219–223.
27. Prieto, A.L., Edelman, G.M., and Crossin, K.L.
1993. Multiple integrins mediate cell attachment
to cytotactin/tenascin. Proc. Natl. Acad. Sci. U. S. A.
28. Kemperman, H., Driessens, M.H., La Riviere, G.,
Meijne, A.M., and Roos, E. 1995. Adhesion mecha-
nisms in liver metastasis formation. Cancer Surv.
29. Weinreb, P.H., et al. 2004. Function-blocking
integrin αvβ6 monoclonal antibodies: distinct
ligand-mimetic and non-ligand mimetic classes.
J. Biol. Chem. 279:17875–17887.
30. Whitehead, R.H., Jones, J.K., Gabriel, A., and Lukies,
R.E. 1987. A new colon carcinoma cell line (LIM
1863) that grows as organoids with spontaneous
differentiation into crypt-like structures in vitro.
Cancer Res. 47:2683–2689.
31. National Center for Biotechnology Information.
32. Oettgen, P., et al. 1996. Characterization of NERF,
a novel transcription factor related to the Ets factor
Elf-1. Mol. Cell. Biol. 16:5091–5106.
33. Farr, A., and Roman, A. 1992. A pitfall of using a
second plasmid to determine transfection efficien-
cy. Nucleic Acids Res. 20:920.
34. Dube, A., Akbarali, Y., Sato, T.N., Libermann, T.A.,
and Oettgen, P. 1999. Role of the Ets transcription
factors in the regulation of the vascular-specific
Tie2 gene. Circ. Res. 84:1177–1185.
35. Rimm, D.L., Camp, R.L., Charette, LA., Olsen, D.A.,
and Provost, E. 2001. Amplification of tissue by
construction of tissue microarrays. Exp. Mol. Pathol.