Molecular Biology of the Cell
Vol. 20, 4140–4152, October 1, 2009
Caveolae Mediate Growth Factor-induced Disassembly of
Adherens Junctions to Support Tumor Cell Dissociation
Lidiya Orlichenko,*†Shaun G. Weller,*†Hong Cao,* Eugene W. Krueger,*
Muyiwa Awoniyi,* Galina Beznoussenko,‡Roberto Buccione,‡
and Mark A. McNiven*
*Mayo Clinic, Department of Biochemistry and Molecular Biology and the Miles and Shirley Fiterman Center for
Digestive Diseases, Rochester, MN 55905; and‡Consorzio Mario Negri Sud, Santa Maria Imbaro (Chieti) 66030, Italy
Submitted October 17, 2008; Revised July 6, 2009; Accepted July 22, 2009
Monitoring Editor: Keith E. Mostov
Remodeling of cell–cell contacts through the internalization of adherens junction proteins is an important event during
both normal development and the process of tumor cell metastasis. Here we show that the integrity of tumor cell–cell
contacts is disrupted after epidermal growth factor (EGF) stimulation through caveolae-mediated endocytosis of the
adherens junction protein E-cadherin. Caveolin-1 and E-cadherin closely associated at cell borders and in internalized
structures upon stimulation with EGF. Furthermore, preventing caveolae assembly through reduction of caveolin-1
protein or expression of a caveolin-1 tyrosine phospho-mutant resulted in the accumulation of E-cadherin at cell borders
and the formation of tightly adherent cells. Most striking was the fact that exogenous expression of caveolin-1 in tumor
cells that contain tight, well-defined, borders resulted in a dramatic dispersal of these cells. Together, these findings
provide new insights into how cells might disassemble cell–cell contacts to help mediate the remodeling of adherens
junctions, and tumor cell metastasis and invasion.
Polarized epithelial cells such as those in ductular organs,
including the pancreas, form and maintain their tubular
tissue architecture through regulated associations with
adjacent cells (Hogan and Kolodziej, 2002; Lubarsky and
Krasnow, 2003; Zegers et al., 2003). The integrity of these
lateral interactions is mediated, in part, by adherens junc-
tions (AJs), of which the transmembrane protein E-cadherin
(E-cad) is a major component. On the extracellular side,
homophilic antiparallel interactions between E-cad mole-
cules present on adjacent cells mediate the assembly and
maintenance of AJs, whereas on the intracellular side the
cytoplasmic tail of E-cad is associated with an array of actin
cytoskeletal proteins as well as signaling molecules such as
catenins, small GTPases, and nonreceptor tyrosine kinases
(Perez-Moreno et al., 2003; Gumbiner, 2005). Although main-
tenance of stable junctions is important for tissue integrity
and the functional properties of polarized epithelia, AJs are
also dynamic structures undergoing cycles of assembly and
disassembly. Indeed, reorganization of AJs is a key aspect of
tissue morphogenesis both during normal development as
well as tumor cell metastasis, when the structural integrity of
AJs is compromised as tumor cells lose polarity and subse-
quently dissociate before migration (Thiery, 2002; Cavallaro
and Christofori, 2004).
Pancreatic cancer is a particularly deadly disease and is
listed as one of the top five most lethal cancers in the United
States (Jemal et al., 2006). Although less prevalent than other
cancers, its mortality rate is well over 90% within 6 mo of
diagnosis. This exceptionally high lethality is due to a lack
of early diagnostic tools, the dispersed organization of the
pancreas within the abdomen, and a significant propen-
sity of neoplastic cells to disseminate and migrate from
the pancreas to nearby organs (Shi et al., 2001; Freelove
and Walling, 2006).
A reduction in E-cad protein has been implicated as a pre-
requisite for migratory activity and the development of an
invasive metastatic phenotype in pancreatic cancers (Imamichi
et al., 2007). It has also been described as an independent
prognostic factor for patient survival (Karayiannakis et al.,
2001; Garcea et al., 2005). Indeed, the loss of E-cad expression
has been shown to facilitate peritoneal dissemination of
pancreatic cancer cells (Furuyama et al., 2000) and has been
found to be associated with high-grade and advanced-stage
pancreatic tumors (Pignatelli et al., 1994). In contrast, several
recent studies have demonstrated that pancreatic carcino-
mas maintain normal levels of E-cad (Menke et al., 2001;
Alldinger et al., 2005; Toyoda et al., 2005), suggesting that
additional mechanisms of tumor cell dissemination also ex-
ist (see Cavallaro and Christofori, 2004).
It is known that stimulation of cells with growth factors
can affect the stability of AJs by altering the internalization
and vesicle trafficking dynamics of E-cad. However, the
precise endocytic mechanisms used remain unclear, as both
clathrin- and caveolae-mediated endocytosis have been im-
plicated in growth factor–stimulated internalization of AJ
components (Bryant and Stow, 2004; D’Souza-Schorey, 2005;
Ivanov et al., 2005). As examples, in hepatocyte growth
This article was published online ahead of print in MBC in Press
on July 29, 2009.
†These authors contributed equally to this work.
Address correspondence to: Mark A. McNiven (mcniven.mark@
Abbreviations used: AJ, adherens junction; Cav1, caveolin-1; E-cad,
E-cadherin; PM, plasma membrane; siRNA, small interfering RNA;
TER, transepithelial electrical resistance.
4140© 2009 by The American Society for Cell Biology
factor–treated Madin-Darby canine kidney (MDCK) cells, a
clathrin-dependent pathway was suggested to mediate E-
cad internalization from the basolateral domain (Palacios et
al., 2002) before degradation in a lysosomal compartment
(Palacios et al., 2005). Similarly, clathrin-mediated endocyto-
sis and subsequent lysosomal degradation of E-cad has also
been implicated in mouse mammary epithelial cells treated
with TGF? in combination with sustained activation of Raf;
however, a caveolin-mediated pathway was not ruled out
(Janda et al., 2006). Finally, epidermal growth factor (EGF)
signaling in A431 epidermoid carcinoma cells was shown to
trigger E-cad endocytosis by a clathrin-independent path-
way that was sensitive to cholesterol depletion, possibly via
caveolae (Lu et al., 2003). Thus, the endocytic mechanism
used to internalize E-cad might vary depending on cell type
as well as specific types of growth factor stimulation.
We have recently reported that normal epithelial cells as
well as pancreatic tumor cells form exceptionally large num-
bers of caveolae at cell borders in response to EGF treatment
(Orlichenko et al., 2006). Caveolae are small flask-shaped
endocytic structures of 50–90 nm in diameter rich in choles-
terol and sphingolipids while also containing a protein coat
of caveolin oligomers. Caveolae not only mediate the inter-
nalization of a variety of cargo molecules, rather caveolae
and caveolin membrane domains may also represent compart-
mentalized signaling platforms (Liu et al., 2002; Pelkmans and
Helenius, 2002; Parton et al., 2006; Parton and Simons, 2007).
Indeed, the dramatic assembly of caveolae structures we
observed in EGF-stimulated cells appears to be dependent
upon Src-mediated phosphorylation of a specific tyrosine
residue (Y14) at the N-terminus of caveolin-1 (Cav1; Li et al.,
1996; Lee et al., 2000; Orlichenko et al., 2006). Caveolae for-
mation occurs rapidly in stimulated cells at the onset of AJ
disassembly; thus, we hypothesized that caveolae could rep-
resent the endocytic pathway used by activated cells to
internalize E-cad and other AJ components.
In this present study we provide direct evidence suggest-
ing an active role for caveolae formation in the internaliza-
tion of E-cad from the borders of EGF-stimulated MDCK
cells and pancreatic tumor cells (BxPC-3, PANC-1, and
HPAF-II). Cav1 and E-cad colocalized at cell borders under
resting conditions, but were internalized and transported
together after EGF treatment, accumulating at large cyto-
plasmic vesicles coated with Cav1. Concomitant with this
morphological cointernalization was an enhanced physical
interaction between Cav1 and E-cad, as indicated by coim-
munoprecipitation of these proteins. When testing for a
direct role of caveolae formation in AJ internalization, we
found that expression of a RFP red fluorescent protein
(RFP)-tagged Cav1 tyrosine phospho-mutant (Cav1Y14F-
mRFP) or reduction of Cav1 protein levels by small interfer-
ing RNA (siRNA) treatment significantly increased the lev-
els of E-cad at AJs. Consistent with this retention of E-cad at
cell borders, MDCK cells stably expressing Cav1Y14F-mRFP
formed tightly packed cell colonies with extended cell–cell
contacts and exhibited an increase in transepithelial electri-
cal resistance (TER). Finally, we observed that different pan-
creatic tumor cell types exhibit an intrinsic inverse correla-
tion in Cav1 and E-cad protein levels. Moreover, when Cav1
levels are increased by exogenous expression of Cav1 in cells
that normally express low levels of Cav1, the ability of these
cells to disseminate is dramatically altered. Together, these
findings provide strong support for the concept that caveo-
lae play a role in the internalization of AJ proteins such as
E-cad and subsequent disassembly of cell–cell contacts. The
implications of Cav1 expression and caveolae formation in
the metastasis of pancreatic tumor cells are discussed.
MATERIALS AND METHODS
Human neoplastic pancreatic ductular epithelial cells (PANC-1, BxPC-3, and
HPAF-II) and normal MDCK epithelial cells were from American Type Cul-
ture Collection (Rockville, MD). PANC-1 cells were maintained in DMEM
(Mediatech, Herndon, VA) supplemented with 10% FBS (Invitrogen, Carls-
bad, CA) and 5% FCS, respectively. MDCK cells were maintained in EMEM
(Mediatech) containing 10% FBS. BxPC-3 cells were grown in RPMI 1640
(Mediatech) containing 10% FBS and supplemented with glutamine, 10 mM
HEPES, 4.5 mM glucose, and 1.5 mM sodium bicarbonate. HPAF-II cells were
maintained in EMEM containing 10% FBS and supplemented with 0.1 mM
nonessential amino acids and 1 mM sodium pyruvate. Media for all cell lines
contained 50 ?g/ml penicillin and 50 ?g/ml streptomycin (Invitrogen, Carls-
bad, CA). All cells were grown at 37°C in 5% CO2. Cells were grown in plastic
tissue culture dishes for biochemical analyses, on acid-washed coverslips for
fluorescence microscopy, and on carbon-coated and glow-discharged gridded
coverslips (Bellco Glass, Vineland, NJ) for electron microscopy (EM).
Antibodies and Constructs
The anti-caveolin-1 (Cav1) polyclonal antibodies were generated in rabbits
injected with a peptide comprising a sequence present in the N-terminus of
rat ?-Cav1 (amino acids 5–33: KYVDSEGHLYTVPIREQGNIYKPNNKAMA)
and affinity-purified as previously described (Henley and McNiven, 1996).
Monoclonal anti-Cav1 and anti-Cav1 PY14 antibodies were purchased from
Transduction Laboratories (Lexington, KY); the monoclonal anti-E-cadherin
antibody was from BD Biosciences (San Jose, CA); the anti-ZO-1 antibody was
from Santa Cruz Biotechnology, (Santa Cruz, CA); the Src, Erk1/2, active-Src
and Erk antibodies were from Cell Signaling Technology (Danvers, MA). Sec-
peroxidase were from Invitrogen. Fluorescently-conjugated secondary goat anti-
rabbit and goat anti-mouse antibodies were from Invitrogen. Cav1 was PCR-
amplified from rat liver cDNA using the following primers containing HindIII
(forward) and EcoRI (reverse) restriction enzyme sites: forward: 5?-AAGCTTAG-
The PCR product was digested with the appropriate restriction enzymes and
cloned into the pEGFP-N1vector (Clontech, Palo Alto, CA) to generate the
Cav1-GFP construct. This construct was then used as template along with the
Stratagene QuickChange Site-Directed Mutagenesis kit (Stratagene, La Jolla,
CA) to generate the Cav1Y14F-GFP construct. The Cav1Y14F insert was
subsequently subcloned into the pDsRed-N1vector (mRFP, Clontech) to
generate the Cav1Y14F-mRFP construct. The green fluorescent protein (GFP)-
tagged version of E-cadherin was a kind gift of W. James Nelson. Constructs
were purified using the Qiagen Plasmid Purification kit (Qiagen, Valencia,
CA) and sequenced by the Mayo Molecular Biology Core Facility (Mayo
Clinic, Rochester, MN). Constructs were verified using the DNASTAR Seq-
Man program (DNASTAR, Madison, WI).
Generation of Stable Cell Lines
MDCK, BxPC-3, and HPAF-II cells stably expressing fluorescently tagged
(mRFP or GFP) wild-type (wt) Cav1 or Cav1Y14F were generated according
to a protocol from Bio-Rad (Hercules, CA). Briefly, MDCK, BxPC-3, and
HPAF-II cells plated in two 75-ml flasks were allowed to grow until they
reached 70% confluency. Cells were then collected into 15-ml tubes and
washed three times with ice-cold PBS without disturbing the pellet by spin-
ning cells down in a low-speed centrifuge for 5 min. After that, cells were
preincubated with DNA (20 ?g/?l) on ice for 10 min, followed by electropo-
ration using the Bio-Rad Gene Pulser II system (Bio-Rad). The transfected
cells were then incubated on ice for 10 min, transferred to complete media,
and incubated overnight to allow for expression of Cav1. MDCK, BxPC-3, and
HPAF-II cells stably expressing wt Cav1-GFP, Cav1Y14F-GFP, or Cav1Y14F-
mRFP and resistant to 0.1, 0.15, and 0.4 mg/ml G-418 (Invitrogen), respec-
tively, were selected and used for the described experiments.
Transfection, Fluorescence Microscopy, and EM
The GeneJammer transfection reagent (Stratagene) was used for transfection
of pancreatic tumor and normal cells according to the manufacturer’s proto-
col. EGF was from Invitrogen, and EGF treatment of serum-starved cells
before processing for fluorescence microscopy was performed as previously
described (Orlichenko et al., 2006). Cells were fixed with 3% formaldehyde,
washed three times in Dulbecco’s PBS (D-PBS), permeabilized in 0.1% Triton
for 2 min, blocked for 1 h in blocking buffer (5% goat serum, 5% glycerol, and
0.04% sodium azide in D-PBS), and incubated with primary antibodies, when
indicated, for 1 h. Cells were then washed three times in D-PBS for 10 min
each followed by a 1-h incubation with secondary antibodies. After that, cells
were washed three times in D-PBS for 10 min and mounted in ProLong
antifade reagent (Invitrogen). Fluorescence micrographs were acquired using
either a Zeiss Axiovert 35 epifluorescence microscope (Carl Zeiss, Thorn-
wood, NJ) equipped with a Hamamatsu Orca II camera (Hamamatsu Pho-
Caveolae Mediate Disassembly of Cell Junctions
Vol. 20, October 1, 20094141
tonics, Hamamatsu City, Japan) or Zeiss LSM510 confocal microscope (Carl
Zeiss, Jena, Germany).
Control untransfected MDCK cells or MDCK cells stably expressing
Cav1Y14F-mRFP were processed for EM as previously described (Henley et
al., 1998) and imaged using a JEOL 1200 electron microscope (JEOL Ltd.,
Tokyo, Japan) to analyze cell junctions.
Live time-lapse imaging of HPAF-II cells stably expressing wt Cav1-GFP or
Cav1Y14F-GFP was performed on a Zeiss Axiovert 35 epifluorescence micro-
scope equipped with a heated stage using a Hamamatsu Orca II camera and
IPLab imaging software (Scanalytics, Fairfax, VA). Cells were plated in
35-mm culture dishes containing a glass bottom and viewed with a 63 ? 1.4
NA lens. MEM medium (Mediatech) containing 10% FBS and buffered with 15
mM HEPES (pH 7.2) was used for imaging. Five minutes before imaging, cells
were stimulated with 30 ng/ml EGF. Images were acquired every 15 s using
a 500-ms exposure time over a 25-min time period. Alternatively, time-lapse
studies were performed on PANC1 cells using a Zeiss LSM 510 confocal
3D Reconstruction of Images
For images rendered in 3D, EGF-stimulated HPAF-II cells grown on glass
coverslips were fixed, permeabilized in 0.2% Triton, costained with antibodies
against Cav1 and E-cadherin (E-cad), and imaged using a Zeiss LSM510
confocal microscope. At least 23 slices were acquired per image at 3-?m
intervals, and 3D reconstruction was performed using LSM510 software.
Measurement of TER
To determine TER, untransfected MDCK cells or MDCK cells stably express-
ing GFP, wt Cav1-GFP, or Cav1Y14F-mRFP were seeded on BD Falcon cell
culture inserts with an 8-?m pore size (Franklin Lakes, NJ) and allowed to
reach confluency. Measurements were subsequently taken 24 and 48 h after
confluency using a Millipore Millicell-ERS resistance system (Billerica, MA).
After subtracting the background resistance value of a blank filter in medium,
all values were normalized to filter size.
Caveolin-1 Knockdown via siRNA
SiRNA duplexes targeting human Cav1 mRNA as well as a control scrambled
siRNA were purchased from Dharmacon (Lafayette, CO), and cells were
transfected according to the manufacturer’s instructions. Briefly, cells seeded
on coverslips (for fluorescence microscopy) or on 10-cm tissue culture dishes
(for biochemical assays) were grown under normal conditions until they reached
30–50% confluency. Cells were then washed in low serum OPTI-MEM medium
(Invitrogen) and transfected with 75 ?M Cav1 siRNA using Oligofectamine
reagent (Invitrogen). Five hours after transfection, an equal volume of medium
containing 30% FBS was added to the OPTI-MEM, and cells were allowed to
recover for the time specified in the corresponding figure legends (from 24 to
72 h). The siRNA sequence targeting human Cav1 was AACCAGAAGGGACA-
CACAGUU, a sequence previously shown to effectively knockdown Cav1 ex-
pression (Cho et al., 2003; Orth et al., 2006).
Immunoprecipitation, Biotinylation Studies, SDS-PAGE,
and Western Blotting
Immunoprecipitation of Cav1, SDS-PAGE, and Western blotting were per-
formed as previously described (Orlichenko et al., 2006). A stripping protocol
was used for assessment of protein expression levels of E-cad, Cav1, and PY14
Cav1 on the same membrane. For stripping, 20 ml of stripping buffer (14.6 ml
H2O, 4 ml 10% SDS, 139 ?l MeOH, and 1.125 ml 1 M Tris-HCl, pH 7.0) was
applied to the membrane and incubated at 50°C for 30 min. The membrane
was then washed two times for 10 min each with D-PBS containing 0.05%
Tween 20 (D-PBST) and reblocked with 5% fat free milk in D-PBST. Horse-
radish peroxidase (HRP)-conjugated secondary antibodies were detected us-
ing ECL reagents (Amersham, Arlington Heights, IL) according to the man-
To determine the level of internalized E-cad protein in the MDCKts-VSrc
cell line after v-Src activation, a cell surface biotinylation assay was performed
essentially as described previously (Cao et al., 2005). After the biotin labeling
of surface protein, cells were lysed and E-cad protein was immunoprecipi-
tated using a commercially available mAb (BD Biosciences). The ratio of
biotinylated surface Ecadherin levels to that of total immunoprecipitated
protein was compared with or without a 7-h shift to the permissive temper-
ature to activate v-Src. Sulfo-NHS-LC-biotin and HRP-conjugated streptavi-
din were from Pierce (Rockford, IL).
Quantitative analysis of randomly selected fields of cells was performed using
IPLab imaging software on images acquired using a Zeiss Axiovert 35 epi-
fluorescence microscope. The level of E-cad staining at individual cell junc-
tions was measured by plotting the signal intensity along a linear cross-
sectional image trace that incorporated multiple cell junctions. The average
intensity of E-cad signal for individual cell borders was calculated by aver-
aging the intensity for registered pixels within regions of cell junctions.
Analysis of E-cad localized at cell junctions was performed on at least 15 cell
borders for each experimental condition.
Quantitative analysis of Western blots was performed based on the densi-
tometry of scanned bands using the Bio-Rad Image Analysis System with
Molecular Analyst Software and a Bio-Rad Model GS-700 Imaging Densitom-
eter (Bio-Rad). Bands were normalized to a background of the film of equal
area and to a loading control.
EGF-induced Caveolae Formation Leads to an Increased
Association of Cav1 and E-cad during the Disassembly of
We and others have observed that Cav1 is highly enriched
along the borders of a variety of normal and neoplastic
cultured epithelial cells (Palacios et al., 2002; Lu et al., 2003;
Janda et al., 2006; Orlichenko et al., 2006; Figure 1, a–c).
Interestingly, when resting cells were examined by EM, very
few caveolae structures were observed at the peripheral sites
where Cav1 protein is concentrated. As shown in the elec-
tron micrographs of cultured human pancreatic tumor cells
(PANC-1), the borders between adjacent cells are generally
intact and display few invaginations (Figure 1d). In dramatic
contrast to resting cells, stimulation of cells with EGF (30
ng/ml) for 10–30 min induced a marked assembly of caveo-
lar structures at the plasma membrane (PM), resulting in a
loss of cell–cell contacts and separation of adjacent cells
(Figure 1e). Assembly of caveolae under these conditions
was rapid and extensive, as indicated by the long “caveolar
towers” seen extending off the cell borders into the cyto-
plasm (Figure 1e). In addition, this dramatic assembly of
caveolae was observed in polarized epithelial cells (MDCK
and normal rat kidney [NRK]) as well as human pancreatic
ductular tumor cells (PANC-1 and BxPC-3); therefore, we
assume that this phenomenon is exhibited by many epithe-
lial cell types.
Although Cav1 and E-cad showed a marked colocaliza-
tion at cell borders in resting cells (Figure 1, a–c and a?–c?),
whether this localization represented a participation of Cav1
in growth factor-stimulated internalization of AJ proteins or
rather was simply coincidental remained to be determined.
Therefore, we stimulated MDCK cells, as a control epithelial
cell type, and human pancreatic tumor cells (BxPC-3) with
EGF for 5–30 min before fixation and immunostaining for
Cav1 and E-cad (Figure 2, a–b?). By 20 min after EGF
treatment, a punctate distribution of Cav1 could be noted
along the borders between adjacent cells, consistent with the
formation of caveolar structures. However, most striking
was the formation of large (0.5–2 ?m), spherical, intracellu-
lar compartments that were coated along the periphery with
Cav1 and filled with E-cad (Figure 2, a–b?). Furthermore,
although EGF treatment induced the formation of these
structures in MDCK cells and increased their number in
BxPC-3 cells, caveolar clusters also formed spontaneously
without EGF exposure in the various pancreatic tumor cell
types tested (BxPC-3, PANC-1, and HPAF-II; Supplemental
Figure S1), consistent with observations by others that pan-
creatic tumor cells possess amplified signaling pathways
(Korc et al., 1992; Yamanaka et al., 1993; Oikawa et al., 1995;
Liu et al., 1998; Alldinger et al., 2005).
In support of the observations described above gathered
from static images, dual-color time-lapse confocal micros-
copy was performed on cultured PANC-1 cells. After serum
starvation and stimulation with 100 ng/ml EGF for 1 h, cells
expressing fluorescent protein–tagged versions of E-cad and
Cav1 formed invaginations of the cell borders that contained
both proteins. These Ecad/Cav1-containing invaginations
were reminiscent of the “caveolar towers” seen at the ultra-
L. Orlichenko et al.
Molecular Biology of the Cell4142
structural level After EGF treatment in these cells (Figure 1e,
Orlichenko et al., 2006). Short time-lapse observations re-
vealed small vesicles carrying both E-cad and Cav1 proteins
in the act of budding from these border invaginations to-
ward the cell interior (Supplemental Figure S2 and Supple-
mental Movies 1 and 2). Interestingly, the Panc1 cell model
utilized here facilitated this live time imaging as they possess
tenuous, irregular cell borders with only modest peripheral
E-cad that appears predominantly internalized within cyto-
plasmic vesicles. Thus, these cells appear to constitutively ve-
siculate and transport E-cad from the cell borders.
The dramatic formation of caveolae along with Cav1-
coated, E-cad–containing vesicles in EGF-treated cells sug-
gested that these conditions might induce an increased as-
sociation between Cav1 and E-cad. First, to help discern
whether the E-cad–containing endocytic structures actually
represented endosomal compartments or rather clustered
patches at the cell surface, 3D reconstruction of HPAF-II
cells stimulated with EGF and stained for E-cad and Cav1
was performed. Indeed, as depicted by the representative
3D reconstruction in Supplemental Figure S3, most of the
spherical Cav1- and E-cad–containing structures were
within the central cytoplasm of each cell. Next, as EGF-
induced phosphorylation of Cav1 might alter its interactions
with proteins, we monitored Cav1 tyrosine 14 phosphoryla-
tion in MDCK and BxPC-3 cells treated with EGF by West-
ern blot using an anti-phospho-Cav1 antibody (PY14). Both
cell types showed a marked increase in Cav1 phosphoryla-
tion after EGF treatment, although significant levels of Cav1
were phosphorylated even under resting conditions in
BxPC-3 cells (Figure 2c). This observation is consistent with
the fact that these cells have amplified EGF receptor signal-
ing (L.O., S.G.W., and M.A.M. unpublished observations;
Arnoletti et al., 2004; Ali et al., 2005), modest cell–cell bor-
ders, and spontaneously form Cav1-coated endosomes in
the absence of EGF treatment (Supplemental Figure S1, b
and b?). Finally, to test for an association between Cav1 and
E-cad and whether this putative interaction might be in-
creased upon EGF stimulation, Cav1 was immunoprecipi-
tated from MDCK and BxPC-3 cells at various time points
after EGF stimulation. Subsequently, the immunoprecipi-
tates were Western blotted for E-cad and phospho-Cav1,
and the signal from the respective bands was quantitated
using densitometry. Modest levels of E-cad were coimmu-
noprecipitated with Cav1 from both cell types in the absence
of EGF; however, upon stimulation with EGF, a significant
increase (twofold) in the amount of coimmunoprecipitated
proteins was observed (Figure 2, d and e). Interestingly, this
increase paralleled the observed increase in Cav1 phosphor-
ylation in both cell types, with MDCK cells exhibiting a
slower response (30–60 min) compared with BxPC-3 cells
(5–20 min). Coimmunoprecipitation of E-cad and Cav1 may
not represent a direct binding of these proteins, and we
predict that these proteins are more likely to interact indi-
rectly as part of a complex within an endocytic compart-
ment. The fluorescence images (Figure 2, a–b?, and Supple-
mental Figures S1 and S2) also suggest an indirect
interaction, as the internalized E-cad cargo was not always
observed to overlap with Cav1, but instead appeared to be
surrounded by a Cav1-coated container.
Disrupting Caveolae Formation Results in a Stabilization
of Cell Borders
We have recently shown that EGF-stimulation of cultured
cells induces a four- to eightfold increase in caveolae-like
structures at the cell surface as assessed by EM (Orlichenko
et al., 2006). These invaginations were confirmed as caveolae
based on their size and shape, strong labeling with a Cav1
antibody by immuno-EM, sequestering of HRP-conjugated
cholera toxin, and by immunofluorescent internalization of
toxin. Importantly, expression of a Cav1 tyrosine phospho-
mutant Cav1Y14F, unable to be phosphorylated by Src at
this residue, greatly reduces the formation of these invagi-
nations along with a marked correlating decrease in toxin
uptake. This mutant protein acts to inhibit caveolin assem-
bly and provides a useful tool to test the role of caveolae in
the internalization of cell borders. In addition, EGF-induced
phosphorylation of Cav1 appears to strongly correlate with
E-cad internalization and a marked increase in the associa-
tion between Cav1 and E-cad (Figure 2 and Supplemental
Figure S1). To expand on these findings, we tested whether
expression of the Cav1Y14F mutant would prevent the in-
ternalization of E-cad from cell borders and thereby alter cell
normal and pancreatic tumor epithelial cells before EGF stimula-
tion. (a–c?) Fluorescence micrographs of normal MDCK cells (a and
a?), the human pancreatic cancer cells BxPC-3 (b and b?), and
PANC-1 (c and c?). All of the cell types display marked colocaliza-
tion of endogenous Cav1 (a, b, and c) and E-cad (a?, b?, and c?) at the
PM (arrows) under serum-starved conditions. (d) Electron micro-
graph of a region of cell–cell contact between two tightly apposed
PANC-1 cells (arrow), demonstrating that relatively few caveolae
(arrowhead) are present under serum-starved conditions. (e) Elec-
tron micrograph of PANC-1 cells stimulated with 30 ng/ml EGF for
20 min, showing a proliferation of individual caveolae and the
formation of clustered “caveolar towers” (arrowheads) along an
internalizing cell border (CB, arrow) of two adjacent cells. Scale
bars, (a–c?) 10 ?m; (d and e) 1 ?m.
E-cad and Cav1 colocalize at sites of cell–cell contact in
Caveolae Mediate Disassembly of Cell Junctions
Vol. 20, October 1, 20094143
morphology. Interestingly, MDCK cells stably expressing
mRFP-tagged Cav1Y14F showed marked differences in cell–
cell contacts, both at the light microscopy and EM levels. As
shown in Figure 3, colonies of transfected MDCK cells could
be identified using phase microscopy by appearance alone.
Cells expressing Cav1Y14F-mRFP formed more tightly as-
sociated colonies than the surrounding untransfected cells,
which appeared less organized. At the EM level, comparison
of transverse sections of untransfected MDCK cells (Figure
3c) to MDCK cells stably expressing Cav1Y14F-mRFP (Fig-
ure 3, d and e) revealed a dramatic difference in the border
morphology between the two cell populations.
Control cells displayed delicate, well-defined cell borders
in close physical association with each other. In comparison,
mutant cells possessed large, convoluted cell–cell borders
that were three to four times the length of control cells.
Observation of MDCK and BxPC-3 cells expressing mutant
Cav1 using fluorescence microscopy (Figure 4, a–b? and
d–e?) also suggested that the cell–cell contacts were excep-
tionally large in these cells. Indeed, quantitation of E-cad
fluorescence intensity indicated more than a twofold in-
crease in E-cad at the cell borders of Cav1Y14F-mRFP–ex-
pressing cells compared with adjacent untransfected cells
(Figure 4, c and f). These extensive E-cad–positive mem-
brane domains observed by light microscopy apparently
represent the excessively long borders protruding into the
cell center that were so prominent in electron micrographs
(Figure 3, d and e). To provide a biochemical method to
measure E-cad clearance from the cell surface while further
correlating caveolae assembly with cell junction internaliza-
tion, we performed surface biotinylation experiments of cul-
tured cells in the presence or absence of active src kinase.
Because HPAF and BxPC-3 cells are very tightly associated
making E-cad surface labeling difficult, we utilized MDCK
cells. These particular cells have been engineered to stably
express v-Src (activated at 35°C; Behrens et al., 1993) and
have been utilized by many others to study E-cad internal-
ization (Fujita et al., 2002) and cell dissemination during
migration (Behrens et al., 1993). We have previously utilized
this cell model to demonstrate Src-induced caveolin assem-
bly at cell borders (Orlichenko et al., 2006). In this current
study, cells stably expressing either Wt Cav1 or mutant
CavY14F were shifted to the permissive temperature for 7 h
before surface biotinylation to label all surface proteins.
E-cad protein was then immunoprecipitated from cells, run
on SDS-PAGE and probed with streptavidin-conjugated
HRP. Quantitation of three separate experiments showed a
significant reduction in E-cad internalization in the CavY14F
mutant expressing cells compared with that of the Wt Cav1-
expressing cells (Figure 4, g and h), consistent with the
morphological measurements (Figure 4c).
Caveolae have been linked to regulation of a variety of
different cell signaling cascades including Src and Erk kinase
as well as rac (Engelman et al., 1998; Grande-Garcia et al.,
2007; Patel et al., 2008). Altered activity of these pathways by
manipulation of caveolin could alter cell–cell contacts indi-
rectly rather than by caveolae formation. To address this, we
assessed active levels of phosphorylated forms of Src and
ERK in cells expressing either Wt-Cav1-GFP or CavY14F-
GFP proteins. We viewed this approach as preferable to a
Cav1 siRNA knockdown approach that would decrease the
levels of Cav1 protein. Instead, all cells examined would
express equal levels of Cav1, with the only difference being
a change in a single tyrosine substitution that alters caveolae
assembly. HPAF cells expressing either GFP-tagged wt or
mutant Cav1 were stimulated with EGF for 15 min and then
lysed and blotted for phospho-active forms of C-Src or Erk
kinases. As shown in Supplemental Figure S4, we observed
no appreciable change in the levels of these activated ki-
nases, suggesting that changes in E-cad internalization were
not due to altered signaling cascades. In addition, as shown
later, we have conducted cell separation/dissemination
studies with these caveolin mutant proteins.
motes the internalization of E-cad into Cav1-
coated endosomes. (a–b?) Fluorescence micro-
graphs of MDCK (a and a?) and BxPC-3 cells (b
and b?) treated with EGF for 20 min, showing the
redistribution of Cav1 (a and b) and E-cad (a? and
b?) from the PM to intracellular locations. (a?,
a?, b?, and b?) Higher magnification of the
boxed regions in a and a? and in b, b?, respec-
tively, emphasizing the EGF-induced localization
of Cav1 (a? and b?) and E-cad (a? and b?) to large,
Cav1-coated endocytic structures (arrows). Merged
images are shown in color. (c) Lysates from
MDCK and BxPC-3 cells treated with EGF (100
ng/ml) for the indicated times were analyzed by
Western blot using anti-E-cad, anti-Cav1, and an-
ti-phospho-Cav1Y14 (PY14) antibodies. An in-
crease in Cav1 phosphorylation at tyrosine 14
upon EGF treatment is observed in both MDCK
and BxPC-3 cells; however, this phosphorylation
is more prominent in the tumor cells. (d) Lysates
from MDCK and BxPC-3 cells that were stimu-
lated with EGF (100 ng/ml) for the indicated
times were subjected to immunoprecipitation us-
ing an anti-Cav1 antibody, and the samples were
subsequently analyzed by Western blot using an-
tibodies against E-cad, Cav1, and phospho-
Cav1Y14 (PY14). A marked increase in the asso-
ciation between E-cad and Cav1 is observed in
both cell types after EGF stimulation, although
EGF stimulation of cultured cells pro-
this increased association occurs more rapidly in BxPC-3 cells. (e) Graph depicting results from densitometric quantitation of immunopre-
cipitation experiments similar to those shown in panel d (n ? 3; results represent the average ? SEM). Scale bars, (a–b?) 5 ?m; (a?–b?) 2 ?m.
L. Orlichenko et al.
Molecular Biology of the Cell 4144
Because the morphology of cell borders in Cav1Y14F-
expressing cells was markedly different at the light and
ultrastructural levels, we next tested whether these differ-
ences translated into functional alterations at cell–cell con-
tacts. This was done in three ways: first, by comparing the
membrane dynamics at cell borders of living HPAF-II cells
stably expressing wt Cav1-GFP versus the tyrosine phos-
pho-mutant Cav1Y14F-GFP using time-lapse fluorescence
microscopy (Figure 5, a–c); second, by measuring the TER
exhibited by confluent monolayers of untransfected MDCK
cells or MDCK cells stably expressing GFP alone, wt Cav1-
GFP or Cav1Y14F-mRFP (Figure 5d); and third, by measur-
ing cell dissemination. With live cell imaging, cells express-
ing Cav1-GFP were observed to display very dynamic
borders, undergoing extensive membrane protrusions and
retractions and at times appearing to separate (Figure 5, a
and b). Most remarkable was the frequent formation of
Cav1-GFP–coated membrane tubules and vesicles that ex-
tended rapidly from the cell borders, reflecting the internal-
ization process viewed in fixed cells by EM (Figure 1e) and
fluorescence microscopy (Figure 2, a–b?). In marked con-
trast, HPAF-II cells expressing Cav1Y14F-GFP possessed
static, well-defined borders that were lined with substantial
amounts of Cav1Y14F-GFP, which remained relatively sta-
ble even after EGF stimulation (Figure 5c). A colocalization
between Cav1 and endogenous E-cad at surface invagina-
tions is more evident in fixed cells expressing the Cav1Y14F
mutant protein (Figure 5, e–e?) consistent with the premise
that caveolae formation is required for this internalization
process. In addition, TER was increased two- to threefold as
early as 24 h after confluency in MDCK cells stably express-
ing Cav1Y14F-mRFP compared with untransfected cells or
cells expressing GFP or Cav1-GFP (Figure 5d).
As an alternative to overexpression of mutant Cav1, we
next tested the effects of altering Cav1 function in MDCK
and BxPC-3 cells using an approach that would not require
exogenous protein expression. Namely, siRNA technology
was used to reduce Cav1 expression levels, and the effects
on E-cad localization at cell borders was assessed, we pre-
dicted that interfering with caveolae formation through re-
duction of Cav1 protein levels would also prevent caveolae-
based internalization of AJ proteins, thus leaving more
E-cad at the cell surface than in control cells, which was
similar to cells expressing mutant Cav1. We and others have
previously confirmed that treatment of cultured cells with
the siRNA sequence used to target human Cav1 substan-
tially reduces Cav1 protein levels, as assessed by Western
blot analysis (Cho et al., 2003; Orth et al., 2006). MDCK and
BxPC-3 cells were either mock-treated, as a control, or
treated with Cav1 siRNA for 48 (MDCK) and 72 (BxPC-3) h
and then fixed and stained with antibodies against Cav1 and
E-cad. Images from control and Cav1 knockdown cells were
acquired and adjusted in an identical way. As shown in
Figure 6, in comparison to mock-treated cells (Figure 6, a, a?
and d, d?), cells treated with Cav1 siRNA (Figure 6, b, b? and
e, e?) exhibited markedly more (up to 2.5-fold) E-cad at their
expression of a phospho-mutant Cav1 (Cav1Y14F-
of cell colonies and cell–cell contacts. (a–b?) Corre-
sponding phase (a and b) and fluorescence (a? and
b?) micrographs of mixed populations of parental
untransfected MDCK cells and MDCK cells sta-
bly expressing mRFP-tagged Cav1Y14F. The two
populations of cells exhibit noticeable differences
in cell–cell attachment, spreading, and general
morphology, allowing them to be distinguished
from one another even by phase microscopy at low
levels of magnification. Dashed lines provide fidu-
cells. (c–e) Electron micrographs of parental un-
transfected MDCK cells (c) and MDCK cells stably
expressing Cav1Y14F-mRFP (d and e). When sec-
tioned in the transverse orientation, the junctions
between the parental untransfected cells appear
thin and easy to resolve (c, arrows). In contrast, the
Cav1Y14F-mRFP-expressing cells display convo-
luted cell borders of exceptional length (d and e;
arrows) that extend inward toward the cell center.
Further, these borders are remarkably thick, consis-
tent with fluorescence micrographs of the E-cad
staining in cells expressing Cav1Y14F-mRFP (Fig-
ure 4). Scale bars, (c–e) 1 ?m.
Preventing caveolae assembly through
Caveolae Mediate Disassembly of Cell Junctions
Vol. 20, October 1, 20094145
borders. In some instances, the borders of siRNA-treated
cells were so enlarged that they appeared to form large
sheets of E-cad that extended into the cytoplasm. (Figure 6,
b?, c, e?, and f). As the siRNA probe was made to human
Cav1, we observed a more efficacious reduction of Cav1 in
the human cancer cells than in the canine MDCK cells, and
this difference was also reflected by the substantial increase
in E-cad at the cell borders of siRNA-treated BxPC-3 cells
compared with the significant but more modest increase
observed in MDCK cells (Figure 6, graphs). Regardless of
the cell type used, these findings support the premise that a
decrease in Cav1 protein, and thus caveolae, translates into
greater levels of E-cad at cell borders.
Pancreatic Tumor Cells with Abnormal Cell–Cell
Adhesions Exhibit Spontaneously Altered Cav1 Levels
The pancreatic tumor cell lines examined in this study ex-
hibit markedly distinct degrees of cell–cell adhesion; there-
fore, we next tested whether these differences in morpho-
logical phenotypes might correlate with the endogenous
protein levels of Cav1 and E-cad. Western blot analysis
indicated that BxPC-3 cells express equal levels of Cav1 and
E-cad protein (data not shown), similar to that observed in
MDCK cells, while PANC-1 cells express modest E-cad pro-
tein levels, but normal levels of Cav1 protein (Figure 7a; see
also Lin et al., 2005), corresponding with the weak cell–cell
interactions displayed by these cells. Similarly, staining of
these cells using immunofluorescence showed that many of
these loosely associated cells express robust levels of Cav1,
but little E-cad protein (Figure 7, b and b?). Interestingly,
small populations of PANC-1 cells do interact to form semi-
tight monolayers and stain positive for E-cad (see also Fu-
ruyama et al., 2000), but express very little Cav1 protein.
Thus, this cell line displays an inverse correlation between
Cav1 and E-cad protein expression, consistent with the pre-
diction that cells without Cav1 protein cannot form caveolae
and therefore E-cad levels at the cell borders remain high.
Examination of HPAF-II pancreatic tumor cells both by
Western blot and immunofluorescence provided further
support to the concept that an inverse correlation in Cav1
and E-cad protein levels translates to effects on AJ integrity.
Compared with the other normal and neoplastic cells exam-
ined in this study, HPAF-II cells express exceptionally high
levels of E-cad protein and very little (10%) Cav1 (Figure 7a;
cad is reduced in both normal and neoplastic
cells expressing the Cav1 tyrosine phospho-mu-
tant Cav1Y14F. (a–b? and d–e?) Fluorescence mi-
crographs of mixed populations of parental un-
transfected cells and cells stably expressing
mRFP-tagged Cav1Y14F (a, b and d, e) that were
fixed and stained for E-cad (a?, b? and d?, e?) after
EGF stimulation (30 ng/ml) for the indicated
times. Markedly increased levels of E-cad are
present at the cell borders of Cav1Y14F-mRFP–
expressing cells (arrows), compared with the
modestly stained untransfected adjacent cells.
Fluorescence micrographs of MDCK cells are
shown in (a–b?) and of BxPC-3 cells in (d–e?).
Dashed white lines provide fiduciary marks in-
dicating the interface between transfected and
untransfected cells. (c and f) Graphs depicting
results of fluorescence quantitation of E-cad lev-
els at cell borders. E-cad levels at the borders of
Cav1Y14F-mRFP–expressing cells are increased
two- to threefold compared with untransfected
cells. Results represent the average ? SEM for ?15
cells for each condition. (g) Western blot analysis of
MDCK cells stably expressing a temperature-sensi-
tive version of v-Src that were also transiently ex-
pressing either WTCav1 or Cav1Y14F. E-cadherin
protein was immunoprecipitated from cells after
surface biotinylation after a permissive tempera-
ture shift to activate v-Src at 35°C for 7 h. Surface
E-cadherin was detected using streptavidin-conju-
gated HRP. (h) Graph showing the relative level of
surface E-cadherin after a 7-h permissive tempera-
ture shift in cells expressing either WTCav1 or the
Cav1Y14F mutant. Cells expressing the Cav1Y14F
protein do not internalize surface E-cad, whereas
the wt-expressing cells had a 40% reduction in sur-
face protein after v-Src activation. Data represent
the average level of surface E-cadherin normalized
to total immunoprecipitated protein over three
separate experiments. Error bars, SEM. Scale
bars, (a–b? and d–e?) 10 ?m.
EGF-stimulated internalization of E-
L. Orlichenko et al.
Molecular Biology of the Cell4146
see also Rajasekaran et al., 2004; Lin et al., 2005). These cells
form remarkably strong AJs, because they assemble into
tight colonies that cannot be dissociated during passage
unless exceptionally high concentrations of trypsin are com-
bined with prolonged incubations. Additionally, these cells
do not disseminate upon treatment with high serum or EGF
stimulation. Immunofluorescence analysis confirmed that
very high levels of E-cad are present at the borders of
HPAF-II cells, and further, these cells form unusually devel-
oped AJs that are so extensive as to extend into the central
cytoplasm (Figure 7c; see also Rajasekaran et al., 2004). Im-
portantly, this unique staining pattern was not observed in
any other cell type we have examined, except those treated
with siRNA to reduce cellular Cav1 levels. Thus, as for the
PANC-1 cells, there is also a strong inverse correlation be-
tween Cav1 and E-cad protein levels in HPAF-II cells.
The spontaneously low levels of Cav1 protein exhibited
by the HPAF-II tumor cells could be a bona fide cause for the
very high E-cad levels and unusually strong cell–cell con-
tacts demonstrated by these cells. Therefore, we tested
whether artificially elevating levels of Cav1 protein through
exogenous expression of wt Cav1-GFP might compromise
the markedly adherent phenotype exhibited by these cells as
well as increase their propensity to disseminate after EGF
treatment by facilitating E-cad internalization. This was
tested using a variety of different approaches to look for a
confirming trend. First, cells were microinjected (Figure 7,
d–e?) or transfected with constructs encoding Cav1-GFP or
GFP, as a control, and allowed to recover (12 h). Subse-
quently, colonies of HPAF-II cells were serum-starved over-
night and then stimulated with 30 ng/ml EGF for 16–24 h
before fixation and staining for E-cad. Cells were counted as
disseminated if three of four borders were no longer in
contact with adjacent cells. Although ?26% of the Cav1-
GFP–expressing HPAF-II cells acquired a motile phenotype
and disseminated outward from the tightly associated colo-
nies (Figure 7, d? and e?), only 12% of the cells expressing the
Cav1Y14F mutant exhibited a dissemination response to
EGF. Further, HPAF-II cells expressing Cav1-GFP could be
easily identified not only by the Cav1-GFP (Figure 7, d? and
e?), but also by the dramatic reduction of E-cad staining in
the cytoplasm and at the cell borders. Indeed, compared
with the surrounding untransfected cells, Cav1-GFP–ex-
pressing HPAF-II cells had little, if any, E-cad at the cell
borders. In support of the experiments using a CavY-14F
mutant protein to prevent caveolae assembly, a phospho-
mimetic form (Cav1Y14D), was utilized to accentuate assem-
bly to test if this might increase cell separation. This mutant
has been utilized by others and demonstrated to increase the
internalization of cell adhesions (Joshi et al., 2008) and also
increase cholesterol transport from the cell surface (Caldieri
et al., 2008).
We have provided additional characterization of this mu-
tant protein through expression in NRK cells and measuring
the effects on caveolae formation by ultrastructural exami-
nation. As displayed in Supplemental Figure S5, EM of
nontransfected control NRK cells show normal numbers of
caveolae (0.8/10 ?m of PM), whereas those expressing the
Cav1Y14F mutant assemble markedly less (0.58/10 ?m of
PM). These findings are consistent with our original findings
(Orlichenko et al., 2006). In contrast, cells expressing the wt
Cav1 or the Cav1Y14D mutant assemble many more caveo-
phospho-mutant Cav1Y14F exhibit reduced bor-
der dynamics as well as an increase in transepi-
thelial electrical resistance. (a–c) Select still fluo-
rescence images from movies of HPAF-II cells
stably expressing wt Cav1-GFP (a and b) or the
Cav1 tyrosine phospho-mutant Cav1Y14F-GFP (c).
Imaging was initiated approximately 5 min after
the addition of EGF (30 ng/ml); time frames (min-
utes and seconds) are indicated at the top of each
image. Cells expressing wt Cav1-GFP exhibit very
dynamic borders from which many Cav1-coated
vesicles and tubules are liberated (a and b; arrows).
In contrast, cells expressing Cav1Y14F-GFP display
static, linear borders (c; arrows) that exhibit little
membrane dynamics. (d) Graph depicting the
transepithelial electrical resistance (TER) of normal,
untransfected (NML) MDCK cells, control MDCK
cells stably expressing GFP, MDCK cells stably ex-
pressing GFP-tagged wt Cav1 (WTCav1-FP), and
MDCK cells stablyexpressing
Cav1Y14F (Cav1Y14F-FP). Measurements were
taken 24 and 48 h after cells had reached conflu-
ency. Cells expressing Cav1Y14F-mRFP showed a
two- to threefold increase in TER, compared with
untransfected MDCK cells or cells stably express-
ing GFP alone or GFP-tagged wt Cav1 (n ? 3;
results represent the average ? SD). (e–e?) To
look for colocalization of E-cad in Cav1-coated
membrane invaginations, cells expressing wt or
mutant Cav1Y14F were fixed and stained for
both proteins. Cav1-positive punctate protru-
sions of the cell membrane were observed to be
markedly positive for E-cad, particularly in cells
expressing the mutant Cav1Y14F protein. Scale
bars, 10 ?m.
Cells expressing the Cav1 tyrosine
Caveolae Mediate Disassembly of Cell Junctions
Vol. 20, October 1, 20094147
lae (1.3/10 ?m of PM and 1.06/10 ?m of PM). Importantly,
the caveolae in the Y14D-expressing cells appear more clus-
tered and distal to the PM, as if internalized, compared with
the control or wt–expressing cells. Although we cannot
make firm conclusions on caveolae association with the PM
without ruthenium red EM approaches, these findings do
support the concept that increased phospho-Cav1 protein
sustain the internalization of E-cad vesicles into the cell
interior. Indeed, we observed a considerable number (35%)
of the large caveosome-like, E-cad–associated structures in
the CavY14D expressing cells compared with controls (Sup-
plemental Figure S5). Thus, cells expressing the phosphomi-
metic form of Cav1 form markedly more caveolae than
control cells or cells expressing Cav1Y14F, are more effective
in the internalization of cell adhesions (Joshi et al., 2008) and
surface cholesterol (Caldieri et al., 2008) and form large
E-cad containing caveosomes.
After the characterization of the Cav1 phosphomimetic–
expressing cells we tested if this might support or accentuate
tumor cell dissemination. Cells expressing this Cav1Y14D
protein also internalized borders and moved away from
adjacent untransfected cells (Figure 7f). Interestingly, these
migratory mutant Cav1-expressing cells often left a distal
appendage, revealing its original attachment site within the
colony before transfection. As a third approach, HPAF cells
expressing either wt or mutant Cav1 were cotransfected
with an active form of Src kinase (c-Src Y530F) to further
potentiate any observed effects of Cav1 expression on cell–
cell contacts. A stable cell line of Cav1-expressing HPAF
cells were utilized, as were transiently transfected cells (Fig-
ure 7g). We observed a near twofold increase in dissemina-
tion rates of the Cav1 wt stable-expressing cells compared
with the Cav1Y14F mutant, whereas the effect observed in
the transiently transfected cells was even more dramatic
(six- to sevenfold increase of Cav1 wt, Cav1Y14D vs.
Cav1Y14F). These “rescue” experiments provide graphic
support indicating an important role for caveolae in the
localization and internalization of E-cad in these pancreatic
tumor cells. Finally, to confirm the effect of the Cav1Y14D-
GFP on cell dissemination and morphology, MDCK cells
were transiently transfected to express this protein then
were fixed and stained for E-cad after a 48-h incubation. As
the MDCK cells do not express high levels of Src or EGFR
like the PANC1 cells, we expected that the effect of the
phospho-mimetic Cav1 form to be accentuated in this cell
type. Indeed, transfected cells exhibited a dispersed pheno-
type leaving long vestigial appendages as the HPAF cells
(Figure 7h, arrow) but exhibited a six- to sevenfold increase
in dissemination compared with wt-expressing cells.
In this study we have made several important observations
implicating both caveolae and Cav1 in the EGF-stimulated
internalization of the AJ protein E-cad in both normal epi-
thelial cells and human pancreatic ductular tumor cells. We
find that Cav1 and E-cad colocalize at cell junctions in rest-
ing cells, and furthermore, that as the epithelial cells begin to
separate after treatment with EGF, E-cad is internalized into
numerous endocytic structures coated with Cav1 (Figure 2,
a–b? and Supplemental Figures S1 and S2). Concomitant
with this vesiculation is an increased association between
Cav1 and E-cad (Figure 2, d and e). Src-mediated phosphor-
ylation of Cav1 occurs at tyrosine 14 (Li et al., 1996; Lee et al.,
2000) and is induced upon treatment of MDCK and BxPC-3
cells with EGF (Figure 2c). Interestingly, we observed that
expression of the tyrosine phospho-mutant Cav1Y14F pre-
vents caveolae formation in these cells as well as induces a
marked change in cell morphology, a remarkable increase in
the complexity of the adhesion sites between adjacent cells,
and an accumulation of E-cad at cell–cell contacts (Figures 3
and 4). In support of these findings, living cells stably ex-
increases the levels of E-cad at cell borders. (a, a?
and d, d?) Fluorescence micrographs of MDCK (a
and a?) and BxPC-3 cells (d and d?) stained for
Cav1 (a and d) and E-cad (a? and d?) under
normal growth conditions, indicating the basal
levels of E-cad present at cell borders (arrows).
(b, b? and e, e?) MDCK cells (b, b?) and BxPC-3
cells (e, e?) treated with Cav1 siRNA for 48 and
72 h, respectively, showing a decreased level of
Cav1 (b and e, arrows) and corresponding in-
crease in E-cad levels at cell borders (b? and e?,
arrows). (c and f) Higher magnification images
focused on E-cad distribution at the cell borders
of siRNA-treated MDCK (c) and BxPC-3 (f) cells.
Abnormally high levels of E-cad accumulate at
cell borders in these cells, reflecting a lack of
border internalization and resulting in the forma-
tion of robust junctions that appear to extend
large laminar protrusions into the central cyto-
plasm (arrowheads). Graphs depict results from
fluorescence quantitation of E-cad levels at cell
borders in mock-treated and Cav1 siRNA-treated
cells. Reduction of Cav1 levels results in a 2- to
2.5-fold increase in E-cad levels at cell borders.
Results represent the average ? SEM for ?15
cells for each condition. Scale bars, (a–b? and
d–e?) 10 ?m.
SiRNA-mediated knockdown of Cav1
L. Orlichenko et al.
Molecular Biology of the Cell4148
pressing this mutant Cav1 showed markedly less membrane
dynamics at cell borders when imaged using time-lapse
microscopy and over a twofold increase in TER (Figure 5).
Similarly, reduction of Cav1 protein levels by siRNA treat-
ment also led to an increase in E-cad at the borders of
EGF-stimulated cells (Figure 6). Finally, we found that an
inverse correlation between Cav1 levels and cell–cell adhe-
sion exists in several pancreatic tumor cell types. Most re-
markably, HPAF-II cells, which normally form very adher-
ent colonies of cells and express exceedingly low levels of
Cav1, could be induced to disseminate from the cell colony
if transfected with a construct allowing for exogenous ex-
pression of Cav1. To our knowledge these findings provide
the first demonstration that directly preventing caveolae
assembly, through either expression of a mutant Cav1 or
siRNA-mediated knockdown of Cav1, not only decreases
the internalization of E-cad, but also leads to a marked
proliferation of cell–cell contacts.
EGF-induced Internalization of Adherens Junction
Proteins into Caveolin-coated Endosomes
Although there has been substantial interest in the potential
role of caveolae as a mechanism for internalizing compo-
nents of cell junctions, there have been few, if any, studies
that have actually conducted a comparative analysis of this
process using both EM and light microscopy. The conspic-
uous formation of E-cad–containing, Cav1-coated endo-
somes correlated temporally with an increased association
between these proteins observed by biochemical methods
(Figure 2). Currently it is not known whether this interaction
endogenous levels of Cav1 protein that inversely corre-
late with the integrity of E-cad–based cell–cell contacts.
(a) Lysates from PANC-1 and HPAF-II cells, two pan-
creatic tumor cell types, and MDCK cells were analyzed
by Western blot using antibodies against E-cad, Cav1,
and the junctional protein ZO-1 (as a loading control).
Note the inverse correlation between endogenous levels
of E-cad and Cav1 in both tumor cell types, compared
with the MDCK cells. (b–c?) Fluorescence micrographs
of PANC-1 (b and b?) and HPAF-II (c and c?) cells
stained for E-cad (b and c) and Cav1 (b? and c?) also
indicate this inverse correlation while additionally
showing the effects on cell–cell adhesion. PANC-1 cells
exist as a heterogeneous population, and indeed, dis-
persed cells appear to have little E-cad at the borders (b, *)
with normal Cav1 (b?) levels, whereas cells associated in
a cluster have E-cad at their borders, but show very little
Cav1 staining (arrows). HPAF-II cells are more homo-
geneous and possess remarkably strong interactions,
with substantial levels of E-cad (c, arrowhead) present at
the borders, whereas Cav1 (c?) staining is modest. (d–e?)
Fluorescence micrographs of HPAF-II cells microinjected
to recover from microinjection then serum-starved, treated
for 24 h with 30 ng/ml EGF, fixed, and stained for E-cad.
Microinjected cells (*) expressing Cav1-GFP (d? and
e?) display little E-cad (d and e) and have a higher
propensity to break away from the tightly associated
cell colonies and move outward. (f) An HPAF-II cell (*)
expressing a phospho-mimetic Cav1Y14D-GFP after
conventional transfection that can be seen migrating
away from adjacent, tightly clustered, nontransfected
cells leaving a remnant appendage of its original lo-
cation (arrow). Note the high levels of E-cad in the
untransfected cells (red), whereas the disseminated
cell has no E-cad. (g) HPAF-II cells stably or tran-
siently expressing either WTCav1-GFP or Cav1Y14F-
GFP have different migratory behaviors when subjected
to active Src coexpression. Quantitation of dissemina-
tion by HPAF cells expressing different forms of exog-
enous Cav1 protein. The majority of cells (?60%) ex-
pressing either GFP, WTCav1GFP, or Cav1Y14D were
separated or detached from neighboring cells, whereas
cells expressing Cav1Y14F-GFP were less migratory.
Only 39% of the cells stably expressing CavY14F and 5%
of the HPAF cells transiently expressing this mutant
CavY14F protein displayed a disseminated phenotype.
Data are the average of three experiments for the stable
lines and two experiments for the transient transfec-
tions. Error bars, SEM. (h) A single MDCK cell trans-
fected (*) to express Cav1Y14F-GFP has migrated away
from its original attachment site (arrow) to the periphery of the colony of untransfected cells. (i) Quantitation of transfected MDCK cells that
have dispersed away from adjacent cells. Importantly, cells expressing the Cav1Y14D protein display a six- to sevenfold increase compared
with wt-expressing cells.
Pancreatic tumor cells display differences in
Caveolae Mediate Disassembly of Cell Junctions
Vol. 20, October 1, 2009 4149
represents a direct binding of E-cad to Cav1 or rather an
association of these two proteins as part of a multiprotein
complex. Although we observed the formation of Cav1-
coated endosomes filled with E-cad in all four types of cells
tested in this study, it is important to note that only the
non-neoplastic MDCK cells required stimulation with EGF
to form these organelles. In contrast, the pancreatic tumor
cells appeared to cointernalize E-cad and Cav1 spontane-
ously, although this could be increased by EGF stimulation.
This is consistent with observations that MDCK cells require
activation of Src, such as through expression of v-Src or after
growth factor stimulation, to disassemble cell borders before
separation (Palacios et al., 2001, 2005; Orlichenko et al., 2006),
whereas PANC-1 and BxPC-3 cells exhibit weak junctions,
form modest monolayers, and have elevated signaling cas-
cades as part of the neoplastic condition (L.O., S.G.W., and
M.A.M. unpublished observations; Korc et al., 1986; Arno-
letti et al., 2004; Ali et al., 2005). Thus, these cells may exhibit
a more dynamic turnover of E-cad, contributing to a more
constitutive formation of Cav1-coated, E-cad–containing en-
dosomes as well as the malignant phenotype.
Cell–Cell Contacts Are Robust in Cells with Compromised
Our findings are consistent with the premise that epithelial
cells use caveolae to internalize E-cad and disassemble cell
junctions in response to EGF stimulation. Of equal impor-
tance is the inverse finding that inhibiting the assembly of
caveolae through siRNA-mediated reduction of Cav1 levels
or expression of the Cav1Y14F mutant results in an accumu-
lation of E-cad at cell borders, altered cell morphology, and
an extensive elaboration of cell–cell contacts. These findings
are intriguing as they both support and contrast with a
study where E-cad internalization after treatment with EGF
was analyzed in A431 cells, an epidermoid carcinoma cell
line. Similar to our studies, acute treatment of A431 cells
with EGF induced E-cad internalization and disruption of
cell–cell junctions. Furthermore, the EGF-stimulated inter-
nalization of E-cad could be attenuated by cholesterol-de-
pleting drugs that interfere with caveolae assembly (Lu et al.,
2003). However, prolonged EGF treatment of A431 cells led
to a decrease in both Cav1 and E-cad expression at the
transcriptional level and a subsequent enhancement in cell
invasion. In addition, reduction of Cav1 levels through an
expression of antisense Cav1 RNA also resulted in decreased
expression of E-cad and an enhancement in cell invasion.
cell types studied. We did not analyze Cav1 and E-cad local-
ization or protein levels after treatment of cells for prolonged
time periods with EGF. Thus, our current study does not
implicate transcriptional regulation, possibly a more down-
stream effect, but instead provides evidence that under con-
ditions of acute EGF treatment, caveolae serve as an endo-
cytic mechanism for the internalization of AJ proteins. The
fact that we observed similar cellular responses—an increase
in the size of cell–cell contacts, an accumulation of border-
localized E-cad, and an increase in TER—upon expression of
a mutant Cav1 (Cav1Y14F) or reduction in endogenous
Cav1 protein via siRNA treatment seems relevant and im-
portant. Moreover, these findings are consistent with the
observations made in the pancreatic tumor cells regarding
expression levels of Cav1 and E-cad and AJ integrity.
In the cell models used for our study, we observed an
inverse correlation between Cav1 and E-cad protein levels
under a variety of experimental conditions. Using fluores-
cence microscopy, we observed that siRNA-mediated reduc-
tion of Cav1 protein levels leads to a marked (two- to three-
fold) increase in E-cad at the cell borders both in MDCK and
BxPC-3 cells, compared with untreated cells. In this case,
quantitative analysis was performed using imaging software
to measure the fluorescence intensity of E-cad at the cell
borders only, not accounting for total cellular E-cad. Perhaps
most compelling is the spontaneous inverse correlation of
E-cad and Cav1 levels in two of the pancreatic tumor cell
types studied here. Western blot analysis of total cell lysates
from PANC-1 cells indicated that these cells express rela-
tively normal levels of Cav1 (see also Lin et al., 2005), com-
pared with MDCK cells, yet more modest E-cad levels.
Interestingly, immunofluorescence analysis revealed that
these cells appear to grow as two populations with distinct
morphologies: one population forms modest junctions that
stain for E-cad but contain little Cav1, whereas the second
population of cells are generally completely dissociated
from each other and express higher levels of Cav1, but
reduced levels of E-cad (see also Furuyama et al., 2000). In
contrast to the distinct populations of PANC-1 cells,
HPAF-II cells form exceptionally tight monolayers. Further-
more, supportive of the inverse correlation observed in
PANC-1 cells but even more graphic in nature, HPAF-II cells
express little Cav1 (see also Lin et al., 2005), as assessed by
Western blot analysis, but relatively high amounts of E-cad
(see also Rajasekaran et al., 2004). Thus, the phenotypes
observed in the specific cell types studied here with reduced
expression levels of Cav1, either through experimental ma-
nipulation (siRNA treatment) or simply as a result of endog-
enous regulation, are consistent with the concept that a
reduction in Cav1 levels and/or caveolae function potenti-
ates E-cad–based cell–cell adhesion.
Caveolae Function in Pancreatic Tumor Cell
Dissemination and Invasion
The multiple experimental approaches used in this study
suggest that Cav1/caveolae play a central role in facilitating
the dissociation of pancreatic tumor cells, and potentially
other epithelial-based tumors, with activated signaling cas-
cades. Currently the literature implicating caveolin levels in
neoplasia is extensive but also complex (for recent reviews
see Goetz et al., 2008; Tanase, 2008). For example, several
studies have shown a correlation between a decrease in
Cav1 expression leading to an acceleration in the develop-
ment of mammary lesions, tumorigenesis, and invasiveness
(Williams et al., 2003, 2004). In addition, normal Cav1 levels
have been implicated in down-regulating RhoC GTPase and
thereby attenuating migration and invasion of pancreatic
tumor cells (Lin et al., 2005). In contrast, increased Cav1
levels have been implicated in promoting prostate tumor
progression (Williams and Lisanti, 2005), whereas increased
Cav1 levels have also been implicated in breast tumors (Van
den Eynden et al., 2006) and pancreatic adenocarcinoma
(Suzuoki et al., 2002).
A contribution made by our current study is a mechanistic
connection between Cav1 protein and the formation/assem-
bly of caveolae vesicles that leads to tumor cell dissemina-
tion. Certainly there are other cellular processes that could
contribute to this dispersal process in addition to caveolae
formation. However, the studies indicating an increase in the
ability of HPAF-II cells to disseminate when exogenously ex-
pressing wt Cav1 protein (Figure 7) are particularly compel-
ling, because these cells express little endogenous Cav1 and
form remarkably adherent colonies. Furthermore, they are
consistent with recent publications implicating Cav1 and
caveolae formation in cell migration (Navarro et al., 2004;
Grande-Garcia et al., 2007). Thus, although caveolae have
been implicated in both facilitating and attenuating cell mi-
L. Orlichenko et al.
Molecular Biology of the Cell4150
gration (Navarro et al., 2004; Lin et al., 2005), our findings
suggest a potential positive role for caveolae in mediating
the dissemination process in some types of pancreatic tumor
cells through their effects on the disassembly of AJs and
subsequent cell separation. As such, interfering with caveo-
lae or Cav1 function could represent a therapeutic target in
an effort to help prevent or control the metastasis of some
tumors (van Golen, 2006).
The authors thank Dr. H. M. Thompson (Mayo Clinic) for help in preparing
the manuscript. This work was supported by National Institutes of Health/
National Cancer Institute Grant CA 104125 (M.A.M.) and the Optical Mor-
phology Core of NIHDK 84567.
Ali, S., El-Rayes, B. F., Sarkar, F. H., and Philip, P. A. (2005). Simultaneous
targeting of the epidermal growth factor receptor and cyclooxygenase-2 path-
ways for pancreatic cancer therapy. Mol. Cancer Ther. 4, 1943–1951.
Alldinger, I., et al. (2005). Gene expression analysis of pancreatic cell lines
reveals genes overexpressed in pancreatic cancer. Pancreatology 5, 370–379.
Arnoletti, J. P., Buchsbaum, D. J., Huang, Z. Q., Hawkins, A. E., Khazaeli,
M. B., Kraus, M. H., and Vickers, S. M. (2004). Mechanisms of resistance to
Erbitux (anti-epidermal growth factor receptor) combination therapy in pan-
creatic adenocarcinoma cells. J. Gastrointest. Surg. 8, 960–969.
Behrens, J., Vakaet, L., Friis, R., Winterhager, E., Van Roy, F., Mareel, M. M.,
and Birchmeier, W. (1993). Loss of epithelial differentiation and gain of
invasiveness correlates with tyrosine phosphorylation of the E-cadherin/
beta-catenin complex in cells transformed with a temperature-sensitive v-SRC
gene. J. Cell Biol. 120, 757–766.
Bryant, D. M., and Stow, J. L. (2004). The ins and outs of E-cadherin traffick-
ing. Trends Cell Biol. 14, 427–434.
Caldieri, G., Giacchetti, G., Beznoussenko, G., Attanasio, F., Ayala, I., and
Buccione, R. (2008). Invadopodia biogenesis is regulated by caveolin-medi-
ated modulation of membrane cholesterol levels. J. Cell Mol. Med. (in press).
Cao, H., Weller, S., Orth, J. D., Chen, J., Huang, B., Chen, J. L., Stamnes, M.,
and McNiven, M. A. (2005). Actin and Arf1-dependent recruitment of a
cortactin-dynamin complex to the Golgi regulates post-Golgi transport. Nat.
Cell Biol. 7, 483–492.
Cavallaro, U., and Christofori, G. (2004). Cell adhesion and signalling by
cadherins and Ig-CAMs in cancer. Nat. Rev. Cancer 4, 118–132.
Cho, K. A., Ryu, S. J., Park, J. S., Jang, I. S., Ahn, J. S., Kim, K. T., and Park, S. C.
(2003). Senescent phenotype can be reversed by reduction of caveolin status.
J. Biol. Chem. 278, 27789–27795.
D’Souza-Schorey, C. (2005). Disassembling adherens junctions: breaking up is
hard to do. Trends Cell Biol. 15, 19–26.
Engelman, J. A., Chu, C., Lin, A., Jo, H., Ikezu, T., Okamoto, T., Kohtz, D. S.,
and Lisanti, M. P. (1998). Caveolin-mediated regulation of signaling along the
p42/44 MAP kinase cascade in vivo. A role for the caveolin-scaffolding
domain. FEBS Lett. 428, 205–211.
Freelove, R., and Walling, A. D. (2006). Pancreatic cancer: diagnosis and
management. Am. Fam. Physician 73, 485–492.
Fujita, Y., Krause, G., Scheffner, M., Zechner, D., Leddy, H. E., Behrens, J.,
Sommer, T., and Birchmeier, W. (2002). Hakai, a c-Cbl-like protein, ubiquiti-
nates and induces endocytosis of the E-cadherin complex. Nat. Cell Biol. 4,
Furuyama, H., Arii, S., Mori, A., and Imamura, M. (2000). Role of E-cadherin
in peritoneal dissemination of the pancreatic cancer cell line, panc-1, through
regulation of cell to cell contact. Cancer Lett. 157, 201–209.
Garcea, G., Neal, C. P., Pattenden, C. J., Steward, W. P., and Berry, D. P. (2005).
Molecular prognostic markers in pancreatic cancer: a systematic review. Eur.
J. Cancer 41, 2213–2236.
Goetz, J. G., Lajoie, P., Wiseman, S. M., and Nabi, I. R. (2008). Caveolin-1 in tumor
progression: the good, the bad and the ugly. Cancer Metastasis Rev. 27, 715–735.
Grande-Garcia, A., Echarri, A., de Rooij, J., Alderson, N. B., Waterman-Storer,
C. M., Valdivielso, J. M., and del Pozo, M. A. (2007). Caveolin-1 regulates cell
polarization and directional migration through Src kinase and Rho GTPases.
J. Cell Biol. 177, 683–694.
Gumbiner, B. M. (2005). Regulation of cadherin-mediated adhesion in mor-
phogenesis. Nat. Rev. Mol. Cell. Biol. 6, 622–634.
Henley, J. R., Krueger, E. W., Oswald, B. J., and McNiven, M. A. (1998).
Dynamin-mediated internalization of caveolae. J. Cell Biol. 141, 85–99.
Henley, J. R., and McNiven, M. A. (1996). Association of a dynamin-like
protein with the Golgi apparatus in mammalian cells. J. Cell Biol. 133, 761–
Hogan, B. L., and Kolodziej, P. A. (2002). Organogenesis: molecular mecha-
nisms of tubulogenesis. Nat. Rev. Genet. 3, 513–523.
Imamichi, Y., Konig, A., Gress, T., and Menke, A. (2007). Collagen type
I-induced Smad-interacting protein 1 expression downregulates E-cadherin in
pancreatic cancer. Oncogene 26, 2381–2385.
Ivanov, A. I., Nusrat, A., and Parkos, C. A. (2005). Endocytosis of the apical
junctional complex: mechanisms and possible roles in regulation of epithelial
barriers. Bioessays 27, 356–365.
Janda, E., Nevolo, M., Lehmann, K., Downward, J., Beug, H., and Grieco, M.
(2006). Raf plus TGFbeta-dependent EMT is initiated by endocytosis and
lysosomal degradation of E-cadherin. Oncogene 25, 7117–7130.
Jemal, A., Siegel, R., Ward, E., Murray, T., Xu, J., Smigal, C., and Thun, M. J.
(2006). Cancer statistics, 2006. CA Cancer J. Clin. 56, 106–130.
Joshi, B., et al. (2008). Phosphorylated caveolin-1 regulates Rho/ROCK-de-
pendent focal adhesion dynamics and tumor cell migration and invasion.
Cancer Res. 68, 8210–8220.
Karayiannakis, A. J., Syrigos, K. N., Polychronidis, A., and Simopoulos, C.
(2001). Expression patterns of alpha-, beta- and gamma-catenin in pancreatic
cancer: correlation with E-cadherin expression, pathological features and
prognosis. Anticancer Res. 21, 4127–4134.
Korc, M., Chandrasekar, B., Yamanaka, Y., Friess, H., Buchier, M., and Beger,
H. G. (1992). Overexpression of the epidermal growth factor receptor in
human pancreatic cancer is associated with concomitant increases in the
levels of epidermal growth factor and transforming growth factor alpha.
J. Clin. Invest. 90, 1352–1360.
Korc, M., Meltzer, P., and Trent, J. (1986). Enhanced expression of epidermal
growth factor receptor correlates with alterations of chromosome 7 in human
pancreatic cancer. Proc. Natl. Acad. Sci. USA 83, 5141–5144.
Lee, H., et al. (2000). Constitutive and growth factor-regulated phosphoryla-
tion of caveolin-1 occurs at the same site (Tyr-14) in vivo: identification of a
c-Src/Cav-1/Grb7 signaling cassette. Mol. Endocrinol. 14, 1750–1775.
Li, S., Seitz, R., and Lisanti, M. P. (1996). Phosphorylation of caveolin by src
tyrosine kinases. The alpha-isoform of caveolin is selectively phosphorylated
by v-Src in vivo. J. Biol. Chem. 271, 3863–3868.
Lin, M., DiVito, M. M., Merajver, S. D., Boyanapalli, M., and van Golen, K. L.
(2005). Regulation of pancreatic cancer cell migration and invasion by RhoC
GTPase and caveolin-1. Mol. Cancer 4, 21–34.
Liu, N., Furukawa, T., Kobari, M., and Tsao, M. S. (1998). Comparative
phenotypic studies of duct epithelial cell lines derived from normal human
pancreas and pancreatic carcinoma. Am. J. Pathol. 153, 263–269.
Liu, P., Rudick, M., and Anderson, R. G. (2002). Multiple functions of caveo-
lin-1. J. Biol. Chem. 277, 41295–41298.
Lu, Z., Ghosh, S., Wang, Z., and Hunter, T. (2003). Downregulation of caveo-
lin-1 function by EGF leads to the loss of E-cadherin, increased transcriptional
activity of beta-catenin, and enhanced tumor cell invasion. Cancer Cell 4,
Lubarsky, B., and Krasnow, M. A. (2003). Tube morphogenesis: making and
shaping biological tubes. Cell 112, 19–28.
Menke, A., Philippi, C., Vogelmann, R., Seidel, B., Lutz, M. P., Adler, G., and
Wedlich, D. (2001). Down-regulation of E-cadherin gene expression by colla-
gen type I and type III in pancreatic cancer cell lines. Cancer Res. 61, 3508–
Navarro, A., Anand-Apte, B., and Parat, M. O. (2004). A role for caveolae in
cell migration. FASEB J. 18, 1801–1811.
Oikawa, T., Hitomi, J., Kono, A., Kaneko, E., and Yamaguchi, K. (1995).
Frequent expression of genes for receptor tyrosine kinases and their ligands in
human pancreatic cancer cells. Int. J. Pancreatol. 18, 15–23.
Orlichenko, L., Huang, B., Krueger, E., and McNiven, M. A. (2006). Epithelial
growth factor-induced phosphorylation of caveolin 1 at tyrosine 14 stimulates
caveolae formation in epithelial cells. J. Biol. Chem. 281, 4570–4579.
Orth, J. D., Krueger, E. W., Weller, S. G., and McNiven, M. A. (2006). A novel
endocytic mechanism of epidermal growth factor receptor sequestration and
internalization. Cancer Res. 66, 3603–3610.
Caveolae Mediate Disassembly of Cell Junctions
Vol. 20, October 1, 2009 4151
Palacios, F., Price, L., Schweitzer, J., Collard, J. G., and D’Souza-Schorey, C. Download full-text
(2001). An essential role for ARF6-regulated membrane traffic in adherens
junction turnover and epithelial cell migration. EMBO J. 20, 4973–4986.
Palacios, F., Schweitzer, J. K., Boshans, R. L., and D’Souza-Schorey, C. (2002).
ARF6-GTP recruits Nm23–H1 to facilitate dynamin-mediated endocytosis
during adherens junctions disassembly. Nat. Cell Biol. 4, 929–936.
Palacios, F., Tushir, J. S., Fujita, Y., and D’Souza-Schorey, C. (2005). Lysosomal
targeting of E-cadherin: a unique mechanism for the down-regulation of
cell-cell adhesion during epithelial to mesenchymal transitions. Mol. Cell.
Biol. 25, 389–402.
Parton, R. G., Hanzal-Bayer, M., and Hancock, J. F. (2006). Biogenesis of
caveolae: a structural model for caveolin-induced domain formation. J. Cell
Sci. 119, 787–796.
Parton, R. G., and Simons, K. (2007). The multiple faces of caveolae. Nat. Rev.
Mol. Cell Biol. 8, 185–194.
Patel, H. H., Murray, F., and Insel, P. A. (2008). Caveolae as organizers of
pharmacologically relevant signal transduction molecules. Annu. Rev. Phar-
macol. Toxicol. 48, 359–391.
Pelkmans, L., and Helenius, A. (2002). Endocytosis via caveolae. Traffic 3,
Perez-Moreno, M., Jamora, C., and Fuchs, E. (2003). Sticky business: orches-
trating cellular signals at adherens junctions. Cell 112, 535–548.
Pignatelli, M., Ansari, T. W., Gunter, P., Liu, D., Hirano, S., Takeichi, M.,
Kloppel, G., and Lemoine, N. R. (1994). Loss of membranous E-cadherin
expression in pancreatic cancer: correlation with lymph node metastasis, high
grade, and advanced stage. J. Pathol. 174, 243–248.
Rajasekaran, S. A., Gopal, J., Espineda, C., Ryazantsev, S., Schneeberger, E. E.,
and Rajasekaran, A. K. (2004). HPAF-II, a cell culture model to study pancre-
atic epithelial cell structure and function. Pancreas 29, e77–83.
Shi, X., Friess, H., Kleeff, J., Ozawa, F., and Buchler, M. W. (2001). Pancreatic
cancer: factors regulating tumor development, maintenance and metastasis.
Pancreatology 1, 517–524.
Suzuoki, M., et al. (2002). Impact of caveolin-1 expression on prognosis of
pancreatic ductal adenocarcinoma. Br. J. Cancer 87, 1140–1144.
Tanase, C. P. (2008). Caveolin-1, a marker for pancreatic cancer diagnosis.
Expert Rev. Mol. Diagn. 8, 395–404.
Thiery, J. P. (2002). Epithelial-mesenchymal transitions in tumour progres-
sion. Nat. Rev. Cancer 2, 442–454.
Toyoda, E., Doi, R., Koizumi, M., Kami, K., Ito, D., Mori, T., Fujimoto, K.,
Nakajima, S., Wada, M., and Imamura, M. (2005). Analysis of E-, N-cadherin,
alpha-, beta-, and gamma-catenin expression in human pancreatic carcinoma
cell lines. Pancreas 30, 168–173.
Van den Eynden, G. G., Van Laere, S. J., Van der Auwera, I., Merajver, S. D.,
Van Marck, E. A., van Dam, P., Vermeulen, P. B., Dirix, L. Y., and van Golen,
K. L. (2006). Overexpression of caveolin-1 and -2 in cell lines and in human
samples of inflammatory breast cancer. Breast Cancer Res. Treat. 95, 219–228.
van Golen, K. L. (2006). Is caveolin-1 a viable therapeutic target to reduce
cancer metastasis? Expert. Opin. Ther. Targets 10, 709–721.
Williams, T. M., Cheung, M. W., Park, D. S., Razani, B., Cohen, A. W., Muller,
W. J., Di Vizio, D., Chopra, N. G., Pestell, R. G., and Lisanti, M. P. (2003). Loss
of caveolin-1 gene expression accelerates the development of dysplastic mam-
mary lesions in tumor-prone transgenic mice. Mol. Biol. Cell 14, 1027–1042.
Williams, T. M., and Lisanti, M. P. (2005). Caveolin-1 in oncogenic transfor-
mation, cancer, and metastasis. Am. J. Physiol. Cell Physiol. 288, C494–C506.
Williams, T. M., Medina, F., Badano, I., Hazan, R. B., Hutchinson, J., Muller,
W. J., Chopra, N. G., Scherer, P. E., Pestell, R. G., and Lisanti, M. P. (2004).
Caveolin-1 gene disruption promotes mammary tumorigenesis and dramat-
ically enhances lung metastasis in vivo. Role of Cav-1 in cell invasiveness and
matrix metalloproteinase (MMP-2/9) secretion. J. Biol. Chem. 279, 51630–
Yamanaka, Y., Friess, H., Kobrin, M. S., Buchler, M., Beger, H. G., and Korc,
M. (1993). Coexpression of epidermal growth factor receptor and ligands in
human pancreatic cancer is associated with enhanced tumor aggressiveness.
Anticancer Res. 13, 565–569.
Zegers, M. M., O’Brien, L. E., Yu, W., Datta, A., and Mostov, K. E. (2003).
Epithelial polarity and tubulogenesis in vitro. Trends Cell Biol. 13, 169–176.
L. Orlichenko et al.
Molecular Biology of the Cell4152