uvomorulin expression, such as MCF-7, T47D, and MDA-. MB-i. 75-7. (Fig. 1 ,. a-c) showed ... be detected in cell lines MDA-MB- .... MDA-MB-468. -. +. -. Fused.
Vol. 2, 365-3 72, August 1991
cell Growth
Cell Adhesion Molecule Uvomorulin Breast Cancer Cell Lines: Relationship and Invasive Capacities
Connie L Sommers, Erik W. Thompson, Rolf Kemlen, Edward P. Gelmann, and Stephen W. Byers’
Jeffrey
A. Torn,
Vincent T. Lombardi
Cancer Research Center [C. L. S., E. W. 1., J. A. T., E. P. G., S. W. B.] and Department ofAnatomy and Cell Biology [C. L. S., E. W. T., E. P. G., S. W. B.], Georgetown University, Washington, DC 20007, and Max-Planck-Institut f#{252}r Immunobiologie, Stubeweg 51, D-7800 Freiberg, Federal Republic of Germany [R. K.]
Abstract adhesion in carcinoma cells may step in the acquisition of an invasive, phenotype. We have examined the
be an
Expression in Human to Morphology
carcinoma cells. Metastasis represents a critical event in the course of tumor progression, usually leading to tumor lethality in cancer patients. Invasion and metastasis of tumor cells into their surrounding connective tissue and to distant sites is a multistep process. Carcinoma cells must first detach from neighboring tumor cells, invade through the basement membrane, and migrate into surrounding tissue. The invading cells may then enter the vascular or lymphatic circulation. Finally, the cells may then attach to the wall of a capillary at a second site and pass
Loss of cell-cell
& Differentiation
through
the
basement
membrane
into
a target
or-
fibroblastoid, highly invasive, vimentin-expressing breast cancer cell lines do not express uvomorulin. Of the more epithelial-appearing, less invasive, keratinexpressing breast cancer cell lines, some express uvomorulin, and some do not. We examined the morphologies of the cell lines in the reconstituted basement membrane matrix Matnigel and measured the ability of the cells to traverse a Matnigel-coated filter as in vitro models for detachment of carcinoma cells
gan, where growth resumes. CAMs may play a role both in the detachment of tumor cells from a primary tumor and in reimplantation at a distant site. Other factors that are likely to be involved in invasion and metastasis are collagenases and other proteases, motility factors, and substrate adhesion molecules (1). In this study, we examine the expression of the CAM uvomorulin in breast cancer cell lines and propose that loss of uvomorulinmediated cell-cell adhesion may be one event required for initial local invasion potentially leading to metastasis in these cells. Uvomorulin (E-cadherin, cell-CAM 120/80, L-CAM) is a 120 kilodalton cell surface glycoprotein involved in calcium-dependent cell-cell adhesion (see Refs. 2-5 for reviews). It is expressed early in mouse development and
from
is involved
important metastatic expression of the epithelial-specific molecule uvomorulin (E-cadhenin, L-CAM)
in human
neighboring
breast
cells
cancer
and
cell adhesion cell-CAM 120/80, cell
invasion
lines.
We
through
find
that
basement
morphology
in Matnigel.
responsible Matnigel
We
show
that
uvomorulin
for the fused colony morphology
since
treatment
is
in
of uvomorulin-positive
MCF-7
cells with an antibody to uvomorulin caused the cells to detach from one another but did not induce invasiveness
in these
cells,
as measured
by their
ability
to cross a Matrigel-coated polycarbonate filter in a modified Boyden chamber assay. Two uvomorulinnegative, vimentin-negative cell lines are also not highly invasive as measured by this assay. We suggest that loss of uvomorulin-mediated cell-cell adhesion may be one progression
phenotype. Introduction may
Received
12/26/90. requests for reprints should be addressed, and Cell Biology, Georgetown University,
To whom of Anatomy I
20007.
play a role
in the invasion
and metastasis
heart
Compaction
of
preimplantation
em-
of
(6).
The
cell
lines
which
appeared
fibroblast-like
ex-
pressed vimentin, whereas the epithelioid cell lines expressed only keratins. There is increasing evidence that vimentin may be a marker for poor prognosis in breast cancer(8-10), and vimentin-expressing breast cancer cell lines are more invasive as measured by in vitro assays.3 Here, we find that fibroblastoid, invasive, vimentin-expressing breast cancer cell lines do not express uvomo-
2
at the Department Washington, DC
tissue
To investigate a possible correlation between CAM expression and invasiveness in human breast cancer, we analyzed 1 7 human breast cancer cell lines for the expression of uvomorulin. We have previously described the morphology of these cell lines, which correlated with expression of the intermediate filament protein vimentin (7).
of many changes involved in the of a carcinoma cell to an invasive
CAMs2
in the
bryos. In the adult, it is expressed specifically in epithelial cells. Disruption of uvomorulin function in nontumorigenic MDCK epithelial cells by specific antibody treatment resulted in a change from a cobblestone, epithelioid appearance to a more elongated, fibroblastic morphology when the cells were plated on plastic. A small percentage of the antibody-treated cells also acquired the ability to invade collagen gels and embryonal
membrane into surrounding tissue. Colonies of uvomorulin-positive cells have a characteristic fused appearance in Matrigel, whereas uvomorulin-negative cells appear detached. Cells which are uvomorulin negative and vimentin positive have a stellate
The abbreviations
used are: CAM,
Madin-Darby canine kidney cells; Iscove’s minimal essential medium; 3
E. W.
Thompson
et a!.,
submitted
cell adhesion
molecule;
cDNA, complementary SSC, standard saline for
publication.
MDCK DNA; citrate.
cells, IMEM,
365
366
Uvomorulin
Expression
in Breast Cancer
Cell Lines
rulin. Another subset of keratin-expressing (vimentinnegative), less invasive cell lines also does not express uvomorulin. In general, these cell lines are not more invasive than other vimentin-negative, uvomorulin-positive
assay
breast cancer cell lines, as measured by an in vitro in which cells must penetrate a basement mem-
brane matrix-coated, porous, polycarbonate filter. These results indicate that although loss of uvomorulin-mediated cell-cell adhesion may facilitate invasion and metastasis, absence of uvomorulin expression does not, by itself, in vitro
lead to an invasive phenotype assay described here.
as detected
by the
embryos (13, 14). As illustrated in Fig. 4, b and of MCF-7 cells with either the gp84 or the DECMA-i antibody caused the cells to detach from each other in the Matrigel outgrowth assay, giving a morphology similar to that seen in uvomorulin-negative, vimentinnegative cells. This effect was seen whether the cells were treated with antibody at the time of plating or whether antibody was added the day after plating. Cells began to detach within 4 h after antibody treatment. In addition, this effect was reversible since, after removal ofgp84 and DECMA-1 antibodies by changing the media, the cells exhibited the fused, uvomorulin-positive pheplantation
e, treatment
notype Results
Seventeen uvomorulin
breast
cancer
expression
cell by
lines
were
analyzed
immunofluorescence
for micro-
and Northern blot analysis. Cell lines positive for uvomorulin expression, such as MCF-7, T47D, and MDAMB-i 75-7 (Fig. 1 a-c) showed bright staining at cell-cell borders by immunofluorescenCe microscopy similar to that previously described for MCF-7 (1 1) and other epithelial cells such as MDCK cells (6). Negative cell lines showed no specific cell-cell border staining (Fig. 1, d-f). A summary of uvomorulin expression in the 1 7 cell lines examined is presented in Table 1. scopy
,
Northern
blot
analysis
was
performed
to
investigate
whether uvomorulin mRNA was present in the breast cancer cell lines that were negative by immunofluorescence microscopy. As shown in Fig. 2a, a band of approximately 5 kilobases was seen in three breast cancer cell lines that were positive by immunofluorescence microscopy (MCF-7, BT474, and MDA-MB-361 No uvomorulin mRNA could be detected in cell lines MDA-MB231 and HS578T, even when MDA-MB-231 polyadenylated RNA was used (Fig. 2b). To show the presence of intact RNA in the MDA-MB-231 and HS578T lanes in Fig. 2a, the blot was subsequently hybridized to a vimentin cDNA probe. Both MDA-MB-231 and HS578T express the intermediate filament protein vimentin as previously described (7), whereas MCF-7, BT474, and MDA-MB361 do not express vimentin. It is interesting to note that the MCF-7ADR line, selected for resistance to adriamycin (12), has lost expression of uvomorulin and acquired vimentin expression. We investigated the morphology of the breast cancer cell lines plated in the reconstituted basement membrane matrix Matrigel, since potentially invasive carcinoma cells are in contact with basement membrane in vivo. The uvomorulin-positive, vimentin-negative cells form large, well-circumscribed colonies of cells in Matrigel in which individual cell borders are not distinguishable, and cells appear fused (Fig. 3, a and b). Uvomorulin-negative, vimentin-negative cells form clusters of spherical cells in Matrigel in which cell borders are easily distinguishable (Fig. 3, c and d). The uvomorulin-negative, vimentinpositive cells have a stellate morphology in Matrigel (Fig. 3, e and 1). To investigate whether the phenotype of fused cobnies versus colonies composed of separated cells was indeed dependent on uvomorulin expression, uvomo).
rulin-positive
MCF-7
polyclonal antiserum monoclonal antibody been shown to block
cells
were
treated
with
gp84
rabbit
to uvomorulin and DECMA-1 rat to uvomorubin, both of which have cell-cell interaction in mouse preim-
seen
in
untreated
cells
(Fig.
4c).
As
negative
cells were treated with similar concentrations of normal rabbit serum (Fig. 4d) or a rat monoclonal antibody to CD44/hyaluronate receptor (15, 16) (Fig. 41). CD44/hyaluronate receptor is an actin-assoCiated cell adhesion molecule involved in lymphocyte homing and in interactions with extracellular hyaluronate (16). After treatment with normal rabbit serum or CD44/ hyaluronate receptor antibody, MCF-7 colonies in Matrigel remained fused, demonstrating the specific involvement of uvomorulin in this phenotype. As an in vitro measure of invasive capacities, we tested the abilities of some of the breast cancer cell lines to cross a Matrigel-coated, porous, polycarbonate filter in a modified Boyden chamber assay. Three uvomorulin-positive, vimentin-negative cell lines were tested, as were two uvomorulin-negative, vimentin-negative and two uvomorulin-negative, vimentin-positive lines (Table 2). The vimentin-negative lines were less invasive than the vimentin-positive lines, as previously reported.3 Of the vi mentin-negative cell lines, the uvomorulin-negative lines CAMA1 and SK-BR-3 were not more invasive than uvomorulin-positive cells as measured by this assay. Since these cell lines were not more invasive, we also tested the effect on invasiveness of disrupting uvomorubin function in uvomorulin-positive MCF-7 cells. In a separate experiment, MCF-7 cells were treated with the same concentration of DECMA-1 antibody to uvomorulin that produced an effect in the Matrigel outgrowth assay and were assayed in the modified Boyden chamber chemoinvasion assay. The antibody did not increase invasiveness in MCF-7 cells; in fact, invasiveness was somewhat decreased. A similar effect was observed in the uvomorulin-positive cell lines T47D and BT474 (data not shown). As can be observed by Comparing Tables 2 and 3, some interassay variation was observed; therefore, the cells listed in Table 2 were assayed at one time, as were controls,
MCF-7
the cells
listed
in Table
3. Treatment
of MCF-7
cells
with
the anti-CD44/hyaluronate receptor antibody had no effect on invasion (Table 3). Therefore, in this assay system, loss of uvomorulin function did not, by itself, lead to increased invasiveness. Discussion
A significant suggested by ogy to neural gene deleted may represent
role for CAMs in malignant progression is several recent studies. A gene with homolCAM, DCC, is a putative tumor suppressor in 70% of coborectal cancers (1 7). This gene a novel CAM whose loss contributes to
malignant
progression
tion of resulted
B-lymphoblastoid in a specific
in coborectal
cancer.
cells by the down-regulation
Transforma-
myc oncogene of the integrin
Cell
Growth
& Differentiation
367
d
e
fig.
1.
“Materials 231;t,
Uvomorulin expression and Methods” usingthe BT549.
as detected by rabbit polyclonal
immunofluorescence antibody against
microscopy. uvomorulin,
Immunofluorescence gp84. a, MCF-7; h, T47D;
microscopy c,
was
MDA-MB-175-7;
performed
as described e, MDA-MB-
d, HS578T;
in
368
Uvomorulin
Tab!e
Expression
1
in Breast
Uvomorulin
Cancer
expression
Cell
in human
Lines
breast
can cer cell
Uvomorulin”
MDA-MB-361 MDA-MB-175-7 BT483
mRNA
IF
+++ +++ +++
+ + +
Fused
+ + + +
Fused Fused Fused Fused
-
++ ++ ++ ++ +
-
-
-
-
-
-
-
-
-
-
-
-
Spherical Spherical Spherical Spherical
+ 4+ + +
-
-
Stellate
-
-
Stellate
-
-
-
-
-
T47D ZR-75-B BT474 MCF-7 MDA-MB-468
-
SK-BR-3 MDA-MB-134 CAMA1 MDA-MB-453 MDA-MB-231 MDA-MB-436
BT549
,
HS578T MDF-7ADR Vimentin mRNA
b
Uvomorulin
Morphology . in Matrigel
Vimentin
Cell line
expression
as detected
10 days,
then
Fused
Stellate Stellate SteIIate
-
by Northern
microscopy
Fused Fused
-
-
as detected
or immunofluorescence Methods. ( Spherical for about
lines
blot
by
analysis.
Northern
(IF),
blot
as described
analysis
(mRNA)
in ‘Materials
and
stellate.
mentin negative, (2) uvomorulin negative/vimentin negative, and (3) uvomorulin negative/vimentin positive. Although loss of uvomorulin-mediated cell-cell adhesion may allow cells to detach from one another, an event which is likely to be required for local invasion, some breast cancer cell lines that did not express uvomorulin were not highly invasive, as measured by the in vitro assay described here. Clearly, additional changes other than uvomorulin loss are involved in the progression to a more invasive phenotype. One important event in the acquisition of invasive properties seems to involve a change in morphology accompanied by the onset of vimentin expression as represented by the uvomorulinpositive/vimentin-positive breast cancer cell lines, which were highly invasive. Other changes that are necessarily involved in invasion through basement membrane are increases in collagenase and other protease activities involved in breaking down basement membrane molecubes (19). This progression may be analogous to epithelial-mesenchymal transitions in embryonic development, in which keratin-expressing, uvomorulin-positive epithehal cells lose their surface polarity and give rise to migrating, uvomorulin-negative, vimentin-expressing mesenchymal cells. These cells can later reorganize into
epithelia CAM lymphocyte decrease in cell-cell
function-associated antigen 1 and a adhesion (18). We believe that these
studies and others point to the involvement of decreased homotypic cell-cell adhesion in tumor progression. Loss of cell-cell adhesion may occur by loss of expression, as
observed
in our
study,
or by reduction
unstable expression, or mutation. We have characterized breast cancer have varying degrees of invasive potential expression.
the
We
following
have
observed
phenotypes:
three
or differentiate
tives (20). transition,
into
connective
If primary carcinoma cells their ability to metastasize
enhanced. A correlation
between
can
uvomorulin
tissue undergo could be
derivasuch a greatly
expression
and
of expression,
metastatic ability has been observed in the murine ovarian carcinoma cell line OV2944 (2 1 ). A subline of OV2944 that was highly metastatic in nude mice had variable but low levels of uvomorulin compared to two weakly met-
cell lines that for uvomorulin
astatic
groups
of cells
(1) uvomorulin
with
positive/vi-
a
sublines,
which
consistently
expressed
high levels
of uvomorulin. These data also suggest that a loss or instability of uvomorulin expression correlates with increased metastatic ability. A more direct test of the
b H
H
ci,
(0
(Y)
c’_)
N
th
co
I-