Download Cytoskeleton and thyroglobulin expression change during

605
Development 102, 605-622 (1988)
Printed in Great Britain © The Company of Biologists Limited 1988
Cytoskeleton and thyroglobulin expression change during transformation
of thyroid epithelium to mesenchyme-like cells*
GARY GREENBURGt and ELIZABETH D. HAY
Department of Anatomy and Cellular Biology, Harvard Medical School, 220 Longwood Avenue, Boston, MA 02115, USA
"This paper is dedicated to the memory of Stephen P. Meier. During his postdoctoral fellowship in our laboratory (1972-1975), Dr
Meier expressed considerable interest in epithelial-mesenchymal transformation, and later (Anderson & Meier, 1981) he analysed the
role of somitomeres in mesenchymal cell migrations
t Present address: The Rockefeller University, New York, NY 10021, USA
Summary
In considering the mechanism of transformation of
epithelium to mesenchyme in the embryo, it is generally assumed that the ability to give rise to fibroblastlike cells is lost as epithelia mature. We reported
previously that a definitive embryonic epithelium, that
of the anterior lens, gives rise to freely migrating
mesenchyme-like cells when suspended in type I collagen matrices. Here, we show that a highly differentiated epithelium that expresses cytokeratin changes to
a vimentin cytoskeleton and loses thyroglobulin during epithelial-mesenchymal transformation induced
by suspension in collagen gel. Using dispase and
collagenase, we isolated adult thyroid follicles devoid
of basal lamina and mesenchyme, and we suspended
the follicles in 3D collagen gels. Cells bordering the
follicle lumen retain epithelial polarity and thyroid
phenotype, but basal cell surface organization is soon
modified as a result of tissue multilayering and
elongation of basal cells into the collagenous matrix.
Cytodifferentiation, determined by thyroglobulin immunoreactivity, is lost as the basal epithelial cells
move into the matrix after 3-4 days in collagen. By
TEM, it can be seen that the elongating cells acquire
pseudopodia, filopodia and mesenchyme-like nuclei
and RER. Immunofluorescence examination of intermediatefilamentsshowed that freshly isolated follicles
and follicles cultured on planar substrata react only
with anticytokeratin. However, all of the mesenchyme-like cells express vimentin and they gradually
lose cytokeratin. These results suggest that vimentin
may be necessary for cell functions associated with
migration within a 3D matrix. The mesenchymal cells
do not revert to epithelium when grown on planar
substrata and the transformation of epithelium to
mesenchyme-like cells does not occur within basement
membrane gels. The results are relevant to our understanding of the initiation of epithelial—mesenchymal
transformation in the embryo and the genetic mechanisms controlling cell shape, polarity and cytoskeletal
phenotype.
Introduction
type of ourchordate ancestors (Hay, 1964). Epithelial
cells show apical-basal polarity, maintain contiguity
via lateral cell junctions and usually sit on a basal
lamina; they do not invade collagenous matrices
when placed on top of them (Overton, 1977; Hay,
1984). In the higher vertebrate embryo, mesenchymal
cells first arise from the primitive streak and then
A fundamental question in cell and developmental
biology concerns the mechanism(s) by which tissue
phenotypes are created. The first tissue to emerge in
the early vertebrate embryo is epithelial in nature
and, indeed, epithelium is the principal body tissue
Abbreviations: BM, basement membrane (basal lamina);
3D, three dimensional; RER, rough endoplasmic
reticulum; TEM, transmission electron microscopy; TSH,
thyroid-stimulating hormone; ECM, extracellular matrix;
PBS, phosphate-buffered saline.
Key words: cytoskeleton, thyroglobulin, thyroid
epithelium, mesenchyme, collagen gel, epithelialmesenchymal transformation, cell shape, polarity.
606
G. Greenburg and E. D. Hay
from the mesodermal epithelia (e.g. somites) that the
primitive streak mesenchymal cells form. Mesenchymal cells are inwandering, elongate or stellate-shaped
cells that typically give rise to fibroblasts, chondroblasts and the like. As they emerge from an embryonic epithelium, they lose epithelial cell polarity and
acquire the ability to invade and move through ECM.
In the embryo, transformation from epithelium to
mesenchyme is carefully controlled and only occurs in
appropriate areas at predictable times. The genetic
program responsible for this change in tissue phenotype seemingly is either not present or is inactivated
in the definitive epithelia of the embryo and adult.
Here and there in the literature of pathology, however, it is said that epithelia, such as are found in the
lens (Henkind & Prose, 1967; Font & Brownstein,
1974) and certain neoplasms (Battifora, 1976; Dulbecco et al. 1981) give rise to fibroblast-like cells
and/or collagen fibrils, suggesting that at least some
epithelia retain activatable programs for forming
mesenchyme. In a preliminary note (Greenburg &
Hay, 1982), we published light microscopic observations that notochord, embryonic limb epidermis,
corneal endothelium and epithelium, and anterior
lens epithelium give rise to elongated bipolar cells
when the tissue is placed within a drop of collagen
that is then allowed to gel. These individual cells
invade, and migrate through, the gel. We studied
embryonic lens epithelium in detail because in isolating it, there is no possibility of a preexisting
mesenchymal contaminant. We showed (Greenburg
& Hay, 1982,1986) that the epithelial cells transforming to mesenchyme lose cytodifferentiation (as
judged by crystallin expression and ultrastructure)
and the epithelial phenotype (they switch from type
IV to type I collagen synthesis and develop mesenchymal polarity, cell shape and RER).
We publish here for the first time data showing that
thyroid follicular epithelium can also transform into
mesenchyme-like cells when cultured in type I collagen gels and that this change involves the intermediate filament cytoskeleton. Unlike the lens, which
expresses vimentin intermediate filaments, thyroid
epithelial cells have well-developed cytokeratin and
desmosomes. A major difference between epithelium
and mesenchyme is that epithelial intermediate filaments are typically composed of keratin and mesenchymal intermediate filaments of vimentin (Franke
et al. 1979a,b, 1982). In this paper, we examine the
changes in cytoskeletal expression that take place
during the transformation of thyroid epithelium to
mesenchyme and we speculate in the Discussion as to
the role these filaments play in epithelialmesenchymal transformation and cell migration. We
also evaluate the fate of differentiated organelles and
products (thyroglobulin) during this transformation
and we describe its inhibition by basal lamina. The
data have bearing on the question of genetic control
of cell shape, cell polarity and cytoskeletal phenotype, and the general problem of the mode of activation of the program for epithelial-mesenchymal
transformation in the embryo.
Materials and methods
Thyroid follicle isolation
Adult male rats were sacrificed by CO2 asphyxiation and
the skin sterilized by iodine and alcohol. The neck was
opened and the thyroid glands removed using sterile
techniques. Calf thyroid glands were obtained from Trelagans Meat Co. (Cambridge, MA). The glands were cleaned
of connective tissue and minced. Tissue pieces were then
incubated in 1 % collagenase (Nitsch & Wollman, 1980a,b)
and 5 % dispase in Hank's Balanced Salt Solution (HBBS)
for 4h at 37°C in a stirring bottle. During incubation,
follicle epithelium is separated from mesenchymal cells,
basal lamina and other ECM. This solution was passed over
a 150 fim pore-size stainless steel filter to remove aggregates
(Fig. 1). The passthrough was gravity sedimented (left to sit
in a conical centrifuge) for 3 min and the supernatant passed
through a 60[im filter. In some cases, it was then passed
through a 30^m filter (rat isolation only). Thyroid follicles
remain on the filter, while dissociated cells pass through.
The follicles were collected and placed in 10 % FCS (fetal
calf serum, Flow Laboratories, Rockville, MD) in Dulbecco's Modified Eagle's medium H-16 (DME, Gibco
Laboratories, Grand Island, NY). Initially, isolated follicles were fixed and embedded in Spurr (see below) for
transmission electron microscopy to assess their purity.
Later, only visual evaluation of follicles by phase-contrast
microscopy was used to rule out mesenchymal contaminants at the time of isolation.
Primary cell culture
Primary cultures of thyroid epithelium were prepared by
allowing isolated follicles to attach to tissue culture plastic.
After 1-2 days in culture, epithelial cells have spread out
from the explant as a monolayer. Cells were cultured in
10% FCS-DME with 10mM-glutamine, 2-5Jugmr1 fungizone and SOjUgml"1 gentamycin (Schering, Kenilworth,
NJ). The culture medium was changed every 3 days.
Collagen gel culture
Type I collagen was extracted from adult rat tail tendon
with 0-5 % acetic acid and used to prepare native collagen
gels (Ehrmann & Gey, 1956) by a modification of the
method of Elsdale & Bard (1972). Collagen was first
extensively dialysed against acidified 1/10 Ham's F12 medium and then diluted for l-5mgml~'. Next, 1-4 ml was
neutralized at 4°C with 0-2ml F12, 0-2 ml sodium bicarbonate (ll-76mgml~1), and 0-2ml FCS, to form a cold
collagen solution (final concentration lmgml" 1 ). A pellet
of thyroid follicles (centrifuged at lOOOg) was mixed into
this cold collagen solution. Then, 0-3-0-5 ml of the folliclerich collagen solution was pipetted onto a 35 mm plastic
Epithelial-mesenchymal transformation in collagen
tissue culture dish to form a drop lcm in diameter. The
drops were incubated for 15min at 37 °C to firmly gel the
collagen before addition of DME medium supplemented
with 10% FCS, lOmM-glutamine, l-S/igmP 1 fungizone,
and 50 ^ig ml"1 gentamycin. Cultures were maintained up to
22 days in vitro.
Thyroid epithelium was also cultured on the upper
surface of collagen gels prepared by pipetting 0-5 ml of
gelling collagen solution onto the tissue culture dish and
allowing it to polymerize at 37°C for 30min. The follicles
were added to the collagen covered dish in 2 ml of culture
media.
Basement membrane (BM) gel culture
Extracts of basal lamina were obtained from Dr Hynda K.
Kleinman (National Institute for Dental Research, Bethesda, MD). This material was prepared from the EHS
tumour with 2M-urea in 0-15M-NaCl, 0-5M-Tris-HCl,
pH7-4 (Kleinman et al. 1986) to a final concentration of
4mgml~1. Sterile extracts were kept frozen at —20cC.
Thyroid follicles were placed within the gel at room
temperature by first centrifuging isolated follicles at 1000 g
for 5min, aspirating off the supernatant and then mixing
the follicles with the gel solutions. 0-1 ml samples of this cell
mixture were placed as small drops on tissue culture dishes.
The cultures were allowed to sit at room temperature for
5min, incubated for 15min at 37 °C, and then medium
added. They were fixed at daily intervals up to 16 days in
vitro.
Light and electron microscopy
Live thyroid follicles suspended within collagen gels were
observed and photographed with a Zeiss inverted phasecontrast microscope. Some cultures were observed with
Nomarsky interference microscopy. For TEM, tissues or
cultures were fixed for 30min in 2% paraformaldehyde,
0-025 % picric acid, and 2-5 % glutaraldehyde in cacodylate
buffer (0-1 M) and postfixed in 1 % osmium tetroxide in
0-lM-cacodylate buffer, pH7-4, at 4°C for 3Omin. They
were stained en bloc in 1 % uranyl acetate in H2O,
dehydrated and embedded in Spurr (DER 736 embedding
kit, Tousimis Research Co., Rockville, MD). Thin sections
were cut on a Sorvall MT2-B ultramicrotome (DuPont
Instruments-Sorvall Biomedical Division, Newtown, CT)
and stained with 0-2 % lead citrate.
Indirect immunofluorescence microscopy
For immunofluorescence studies of whole thyroid, tissue
was either fixed for 30 min in 3 % formalin in PBS followed
by rinsing in PBS or was not fixed and was infiltrated with
Optimal Cold Temperature compound, OCT (Lab-Tek
Products, Naperville, IL), for 15 min. The tissues were
embedded in OCT and 5/im-thick sections were cut on an
A/O cryostat. Sections were mounted on gelatin-coated
slides and stored at —20°C. Prior to addition of antibody,
sections were warmed to room temperature and rinsed for
15 min in PBS. Sections were overlaid with 25/xl of
antisera at the appropriate dilution and incubated for
45 min at 37°C. Slides were then rinsed three times for
15 min in PBS, treated with secondary antibody (goat antirabbit IgG-FITC), incubated and washed as described
607
above. For immunofluorescence of whole collagen gels,
gels were eitherfixedin methanol (for anti-keratin and antivimentin staining) at room temperature for 5 min or in 3 %
buffered formaldehyde (for anti-thyroglobulin staining)
and permeabilized with acetone at —20°C for 5 min. For all
staining for thyroglobulin, PBS was used as the rinse buffer,
while for all staining for keratin and vimentin, 0-05 MTris-HCl buffer (pH7-4) (Greenburg, 1985) was used.
Following 10 min rinse in buffer, the whole mounts were
incubated with 30^1 antiserum diluted in the appropriate
buffer for 15 min, rinsed in buffer for 10min, incubated in
secondary antibody for 10 min and again rinsed in buffer for
10min. Short incubation times were necessary to obtain
consistently acceptable levels of nonspecific binding of
secondary antibody to the gel. Coverslips were mounted
over the sections with 10 % glycerol in PBS buffer.
Fluorescence was observed with a Zeiss photomicroscope
III equipped with x25 and x63 oil immersion objectives
under epi-illumination. Photographs were recorded on
Kodak Tri-X film for exposure times ranging from 5 s to
lmin and were developed in Microdol diluted 1:3. Identical exposure times were used for experimentals and
controls, and these were printed at the identical exposure
settings. Rabbit antibodies against rat thyroglobulin were
obtained from Dr P. Reed Larsen, Harvard Medical
School, Boston, MA (Izumi & Larsen, 19786); rabbit
antibodies against BHK vimentin (Starger et al. 1978) and
bovine hoof cytokeratin (Jones et al. 1982) were obtained
from Dr Robert D. Goldman, Northwestern University,
Chicago, IL; and anti-cytokeratin antibodies were obtained
from Miles Scientific (Naperville, IL). The specificity of the
antiserum was verified by the lack of staining in tissues
known not to contain the specific antigen and the absence of
staining with nonimmune rabbit serum and with secondary
antibody alone.
Results
Isolation and culture of thyroid follicles
Rat and bovine thyroid follicles were isolated with
collagenase-dispase as described in Materials and
methods. This procedure gives both intact and
broken follicles. By 24 h in culture, only intact follicles are seen, suggesting that broken follicles reform
the original tissue architecture. The purity of the
follicles isolated with collagenase has been described
(Nitsch & Wollman, 1980a,fr). The use of dispase
greatly increases the yield of contamination-free follicles, the purity of which was confirmed here by
electron microscopy and phase-contrast microscopy.
Basal lamina, collagenous ECM, mesenchymal cells
including c-cells and all blood vessels, except an
occasional capillary, are removed by collagenasedispase and subsequent sieving of the preparation
(Fig- 1).
Isolated follicles were suspended in the 3D environment of native type I collagen where they
present flat basal surfaces at first (Fig. 2A). During
608
G. Greenburg and E. D. Hay
Thyroid follicle isolation
n
2. Mince tissue
3. Incubate for
4h in 1%
collagenase
5 % dispase
4. Sieve through
150 jum steel mesh
5. Gravity sediment
three times
Collect supernatant
1. Isolate
thyroid gland
Suspend in gelling
rat tail collagen
6. Sieve through 60 //m,
.then 30/im
nitex screens
Fig. 1. Diagram of the method of thyroid follicle
isolation, following a modification of the procedure of
Nitsch & Wollman (l980a,b). Samples were checked by
light and electron microscopy for absence of
contaminating mesenchyme.
the first 36 h, the basal surface of the follicle becomes
irregular and cells begin to elongate into the gel.
Seemingly, the cell attached to the lumen divides and
its offspring begins to move out into the matrix
(arrow, Fig. 3). The lumen of the follicle is maintained (Fig. 3) and the cells abutting the lumen
remain epithelial.
By 48 h, some cells derived from the new basal
layer of the follicle are migrating freely in the gel and
by 5 days in culture (Fig. 2B), many elongate, bipolar
cells can be seen some distance from the explant. The
increased number of cells observed around each
follicle must result from proliferation of epithelial
cells that detach to move into the gel. The morphology of these cells is virtually indistinguishable
from that of mesenchymal cells migrating within a 3D
gel (Elsdale & Bard, 1972; Bard & Hay, 1975) and the
mesenchyme-like cells derived from other epithelia
placed within collagen gels (Greenburg & Hay, 1982,
1986).
These conclusions are based on examination of
hundreds of explants. In one experiment, the frequency of formation of mesenchyme-like cells from
10
Fig. 2. Thyroid follicles (rat) suspended in type I collagen gels are round with a smooth basal surface initially (A).
(B) By 5 days in culture, mesenchyme-like cells (arrows) have elongated from the entire basal surface to migrate within
the gel. (C) In contrast, when follicles are cultured on plastic (5 days) they remain epithelial. (D) When viewed in cross
section, cells growing on a planar surface coated with collagen consist of tightly apposed, cuboidal cells forming a
confluent monolayer that does not invade into the underlying collagen gel (10 days). B and C are the same
magnification shown in A.
Epithelial-mesenchymal transformation in collagen
609
The morphology of epithelia derived from follicles
plated onto polymerized type I collagen gels is similar
to that on tissue culture plastic. In thin sections of
plastic-embedded material, the epithelial phenotype
of the cells is readily apparent (Fig. 2D). The thyroid
epithelial cells spread uniformly over the collagen gel
as a confluent monolayer of tightly adherent, low
cuboidal cells. The cells do not invade into the gel and
thus obey the rules of a typical epithelium (Hay,
1984).
Ultrastructure
Fig. 3. Thyroid epithelial cell elongating into the collagen
gel. Thyroid follicle (bovine) were cultured for 3 days
within collagen gels and then fixed and embedded in
plastic. This 1 fim plastic section shows a thyroid follicle
containing a lumen filled with a dense colloid, probably
containing thyroglobulin. A single elongating cell, still
attached to neighbouring epithelial cells at the cell apex,
moves into the gel by extending a pseudopodium (arrow)
from its basal surface.
follicles suspended within type I collagen gels was
calculated by counting follicles in gel cultures. Counts
of representative cultures showed that of 66 follicles
cultured for 5 days, 45 (68%) have at least two cells
elongating from the surface and most had given rise,
as well, to individual mesenchyme-like cells migrating
away from the explant. 9 (14 %) showed surface
changes but did not have elongating cells and 12
(18%) showed no changes. In 7-day-old cultures,
80% of the follicles counted contained elongating
cells and mesenchyme-like cells, 8 % had only surface
changes and 12 % showed no changes. In all cases,
the follicles that showed no changes were less than
one-quarter the size of normal follicles and were
composed of small groups of cells without a lumen.
In contrast to follicles suspended within collagen
gels, follicles that are plated onto tissue culture dishes
in culture medium adhere to the plastic surface and
give rise to cells that spread as epithelia (Fig. 2C).
The small polygonal cells of the monolayers are
closely apposed with a characteristic epithelial shape.
The basal epithelial surface of freshly isolated follicles
suspended in collagen gels is initially flat as observed
in the electron microscope. Within 36 h, it becomes
disrupted. Large cell processes, or pseudopodia,
tipped withfinerfilopodiaextend into the gel (Fig. 4).
The processes are in close contact with the striated
collagen fibrils of the gel, possibly using them as
anchorage points. Despite the change in the basal cell
surface, desmosomal and tight junctions remain intact in cells attached to the lumen. By 3 days, several
cell layers have formed at the basal cell surface
presumably as a result of proliferation and the basal
cells with surfaces in contact with the collagen gel
have elongated into the gel (Fig. 5A,B).
A dramatic transformation in nuclear and cytoplasmic fine structure occurs concomitant with this cell
elongation. The extending portion of the elongating
cell becomes rich infinefilamentsand free ribosomes.
A striking change in the appearance of the RER takes
place. In rat thyroid cells, RER is dilated and
branching (RER1, Fig. 5B), whereas, in mesenchymal cells, it is thin and elongate (RER2, Fig. 5C).
Occasionally, an elongating cell attached to a follicle
shows cytoplasm that is different in opposing regions
of the same elongating cell (Fig. 5B). The proximal
end of this cell contains dilated thyroid-type RER
(RER1, Fig. 5B,D), whereas the distal end contains
long, thin RER (RER2, Fig. 5B,E). The nucleus of
this cell is becoming vesicular (N2, Fig. 5B) as compared with the dense nucleus characteristic of thyroid
epithelium (Nl, Fig. 5B). The change that takes place
in the nucleus can also be appreciated by comparing
the nucleus at the arrow in Fig. 5A with those in
nearby follicular cells.
In freely migrating mesenchyme-like cells (Fig.
5C,F) the cytoplasm has become entirely of the
mesenchymal type. The cells, which contact collagen
on all surfaces, have lost the epithelial specializations
that characterize the original tissue (basal lamina,
microvilli, tight junctions, desmosomes). The cytoplasm is highly filamentous and filled with welldeveloped secretory organelles. Long arrays of RER
610
G. Greenburg and E. D. Hay
run along the cell axis (RER2, Fig. 5C) and the Golgi
complex is well developed. The numerous long filopodia at the tips of the new pseudopodium are in
close contact with striated collagen fibrils. At the
nuclear level, the nucleolus (nuc, Fig. 5C) is now very
well developed in the large, vesicular nucleus and this
change is accompanied by a large increase in free
polysomes in the cytoplasm.
Fig. 4. Electron micrograph of the basal surface of bovine thyroid epithelium suspended within a collagen gel. The
basal surface is devoid of a basal lamina (removed during isolation) and extends cell processes into the gel. The former
flat surface has acquired pseudopodia and filopodia which are in close contact with striated type I collagen fibrils.
Bovine thyroid cells contain thinner profiles of RER than do rat cells. The cytoplasm contains tonofilaments and the
desmosomes are still present on the lateral surface.
Epithelial-mesenchymal transformation in collagen
Expression of thyroglobulin
Thyroglobulin, a major thyroid protein, was used as a
marker for cytodifferentiation. To determine
whether a change or loss in cytodifferentiation occurred in thyroid epithelial cells undergoing
epithelial-mesenchymal transformation, cultures
were examined by immunofluorescence microscopy
for the presence of thyroglobulin. Thyroid epithelial
cells cultured on plastic or polymerized collagen gels
were compared with whole mounts of thyroid follicles
suspended within collagen gels. Because the turnover
of thyroglobulin in the cell is rapid (Izumi & Larsen,
1978a,b), any cell not synthesizing thyroglobulin
would be devoid of staining. Primary cultures of
thyroid epithelial cells growing on collagen or on
plastic react positively for intracellular thyroglobulin
(Fig. 6A,B, respectively). The staining is predominantly seen as punctate spots concentrated near the
nucleus. These results suggest that the cells are
actively synthesizing thyroglobulin. Thyroid follicles
suspended within gels stain intensely for thyroglobulin (Fig. 6C,D). It was not possible to determine
whether staining is intralumenal in whole-mount
cultures, but it is likely that both cytoplasm and
colloid (seen in Fig. 3) are staining. The production
of thyroglobulin by follicular cells suspended in
collagen gels confirms the electron microscopic data
that the follicles that persist in the gels contain
differentiated epithelial cells around the lumens.
In contrast to the follicular epithelial cells, the
mesenchyme-like cells derived from the epithelium
do not react with anti-thyroglobulin antibodies
(Fig. 6E,F). Thyroid cytodifferentiation is lost before
cells elongate to acquire mesenchymal cell shape and
ultrastructure.
Expression of cytokeratin and vimentin intermediatetype filaments
The expression of intermediate filament proteins was
assessed by indirect immunofluorescence using
monospecific antibodies against either cytokeratin or
vimentin. The pattern of cytokeratin in thyroid follicular epithelial cells in situ was examined by overlaying frozen (5 ,um thick) sections of unfixed rat thyroid
gland with anti-keratin antibodies (data not shown).
The mesenchymal cells (fibroblasts, c-cells, vascular
endothelium) are negative for cytokeratin in sections
of glands fixed in situ. Primary cultures of thyroid
epithelium growing on the surface of collagen gels
retain cytokeratin staining and do not express vimentin even when subconfluent, indicating that the epithelial cytoskeletal phenotype is retained in vitro, at
least during culture periods up to 22 days.
Whole mounts of collagen gel cultures were examined by indirect immunofluorescence to observe the
611
3D organization of cultures transforming to mesenchyme-like cells. 3-day cultures were stained with
either anti-vimentin (Fig. 7A-C) or anti-keratin antibodies (Fig. 7D-F). The intact thyroid follicles suspended within the gels do not stain for vimentin as
demonstrated by the absence of staining in collagen
gels. All of the elongating cells and migrating mesenchyme-like cells, however, stain intensely for vimentin (Fig. 7A-C). The staining pattern observed is
intracellular and filamentous and runs the length of
the cell. The mesenchyme-like cells derived from the
follicles have therefore turned on vimentin expression.
Discrete staining for vimentin appears in the epithelial cells prior to overt cell elongation into the gel
(arrow, Fig. 7A). Vimentin is localized to only a few
epithelial cells of the thyroid follicles. This indicates
that initiation of transformation may occur at different times following suspension of the follicle in the
gel. It is clear from these results that vimentin
expression is an early event in the transformation of
thyroid epithelium to mesenchyme.
Do the mesenchyme-like cells still contain cytokeratin? When 3-day cultures were stained with antikeratin antibodies, the thyroid epithelial cells of
follicles are seen to stain intensely. In addition,
vimentin-rich cells elongating into or migrating within
the gel (Fig. 7D) continue to react intensely for
cytokeratin. Although the newly formed mesenchyme-like cells have acquired the mesenchymal
phenotype and express vimentin, they continue to
stain for cytokeratin. It is not possible using immunofluorescence to determine if the persistence of cytokeratin in the cytoplasm is due to the continued
synthesis of keratin or slow turnover after synthesis
has ceased.
To determine if expression of both vimentin and
cytokeratin continue in cells that have been cultured
for longer times, 22-day-old cultures were stained for
both vimentin and keratin. All mesenchyme-like cells
migrating freely within the gel stain for vimentin. In
contrast to 3-day-old cultures, the majority of cells no
longer stain for cytokeratin. Many cells can now be
found that are negative for keratin (Fig. 7E,F). The
few cells that do continue to stain for keratin are
indistinguishable morphologically from those that do
not. These results do not provide an unequivocal
answer to the question of cytoskeletal stability, but
suggest, by the consistent staining of vimentin, that
the predominant intermediate filaments in the mesenchyme-like cells are of the vimentin type and that
there is a slow turnover and loss of cytokeratin
intermediate filaments from the cytoplasm. In contrast, the epithelial cells of the follicles in contact with
the follicle lumen contain only cytokeratin, even after
long culture times.
612
G. Greenburg and E. D. Hay
m
Epithelial-mesenchymal transformation in collagen
Stability of the mesenchymal phenotype
How stable is the newly acquired mesenchymal
phenotype? To examine this question, cultures were
set up with thyroid follicles either suspended within
the collagen gel or placed on the surface of polymerized gels. The cells were cultured for several weeks
and then examined by phase-contrast microscopy. As
described, follicles cultured within the collagen gels
give rise to mesenchyme-like cells that migrate
radially away from the follicles. Eventually these cells
migrate to the periphery of the gel. At that point,
they may migrate out of the gel onto the plastic
culture dish (Fig. 8A). The cells within the gel at the
periphery are elongate and, when the cells migrate
out onto the plastic, they retain the mesenchymal
morphology and become somewhat flatter due to the
adhesiveness of the plastic substratum. By contrast,
cells on top of the collagen gel that grow over the
surface upon reaching the gel edge, migrate onto the
plastic and retain the epithelial phenotype (Fig. 8B).
They are small, closely apposed, polygonal cells,
which grow as colonies. These results indicate that
once the follicular epithelial cells transform to mesenchyme-like cells within the gel, this phenotype is
stable even when the cells are migrating on a twoFig. 5. (A) Electron micrograph of a rat thyroid cell
elongating into the gel. Thyroid follicles remain intact
after 3 days in culture and contain areas of multilayering
from which single cells elongate into the gel (arrow).
(B) The cytoplasm of the elongating cell changes from
that of thyroid epithelium to one which is highly
filamentous and contains increased numbers of
polysomes. Mitochondria (w), which are not well
preserved in this preparation, are very numerous. The
short convoluted RER of the rat thyroid epithelium
contain distended cisterna (RER1). The mesenchyme-like
cell shown here still contains thyroid type RER1 in its
proximal cytoplasm. During transformation to
mesenchymal phenotype, the RER in the distal part of
this cell has become narrow and extended (RER2, B),
characteristic of mesenchymal cells (RER2, C). The
nucleus of the thyroid cell is dense (Nl), whereas that of
the transforming cell is becoming vesicular (N2).
(C) Electron micrograph of an individual, bipolar
mesenchyme-like cell derived from rat thyroid epithelium
migrating within the same collagen gel shown in Fig. 5A.
The tapering cytoplasm contains abundant arrays of the
mesenchymal type RER (RER2). The centrally located
nucleus contains a prominent nucleolus (nuc) and wispy
chromatin. (D) Enlargement of the RER1 area of the
proximal part of the elongating cell in Fig. 5B.
(E) Enlargement of the RER2 area of the same cell. D
and E are both the magnification shown in D. (F) Light
micrograph of a mesenchyme-like cell derived from rat
thyroid epithelium. The leading edge contains the
pseudopodia and filopodia characteristics of mesenchymal
cells, m, mitochondrion; nuc, nucleolus.
613
dimensional surface. This agrees with our previous
results showing that lens-derived mesenchyme-like
cells invade collagen gels when placed on the gel
surface; they do not reform epithelia (Greenburg &
Hay, 1986).
Inhibition of epithelial-mesenchymal transformation
by BM gel
Different results are obtained when thyroid follicles
are suspended in hydrated gels composed exclusively
of basal lamina components. After 36-48 h, when
follicles suspended in type I collagen gels have begun
to extend cell processes into the gel, no sign of change
in the basal surface of the follicles is observed in BM
gels (Fig. 9A). By 5 days, no cells have elongated or
formed mesenchyme within the basal lamina gel.
Even after several weeks in culture, no mesenchymelike cells have formed from the follicles embedded in
basal lamina. Of the several hundred follicles suspended within BM gels, less than 1 follicle in 100
shows any surface changes or cell elongation. Mesenchyme-like cells are never found.
There are several possible explanations for the
results obtained in these experiments. The BM gel
may not suppport mesenchymal cell migration. This is
not the case, however, because corneal fibroblasts
readily elongate and migrate when placed within
gelling solutions of basal lamina gels (Fig. 9B). A
second possible explanation is that basal lamina
components stabilize the epithelial basal surface,
preventing extension of cell processes and cell dissociation. Basal lamina gel cultures were fixed at
intervals similar to those for collagen gels, embedded
and viewed by transmission electron microscopy.
Unlike the basal surface of follicles suspended in type
I collagen gels, the basal surface in contact with the
basal lamina gel is flat (Fig. 10). In contrast, thyroid
cell surfaces in contact with type I collagen gels
extend pseudopodia and filopodia into the gel (see
Fig. 4 for comparison).
In BM gels, the thyroid follicle remains morphologically differentiated. Electron-dense material,
probably representing thyroglobulin, is present in
secretory granules (Fig. 10) and in the lumen (not
shown). The lateral cell membranes remain closely
apposed and the fine structure of the cytoplasm is
unchanged. Apical tight junctions and desmosomes
are present on the lateral membrane. The cell monolayer in contact with basal lamina components does
not multilayer. Although no true basal lamina forms,
the basal cell surface isflatand regular. An organellefree region underlies the basal surface where an
organized actin mat is located.
614
G. Greenburg and E. D. Hay
Fig. 6. Immunofluorescent staining of thyroglobulin in rat epithelial and mesenchyme-like cells. Thyroid epithelial cells
were cultured on the surface of collagen (A) or plastic (B) for 13 days, fixed, permeabilized and stained with antithyroglobulin antiserum. The closely apposed epithelial cells reacted positively and staining was observed throughout
the cytoplasm, but excluded from the nucleus. (C,D) A single thyroid follicle examined early after suspension
fluoresced intensely for thyroglobulin. (E,F) In contrast, mesenchyme-like cells derived from a follicle (5 days in
culture) are negative when stained with anti-thyroglobulin (arrows). All micrographs are at the magnification shown
in C.
Discussion
The suspension of highly differentiated, adult glandular epithelium in 3D type I collagen matrices results in
increased motile activity of the epithelial basal surface. This surface acquires pseudopodia and filopodia, structures characteristic of mesenchymal surfaces. Cells elongate from the basal surface, detach
from the follicle and invade the surrounding gel as
single, bipolar cells (Fig. 11). These cells are morphologically indistinguishable from embryonic mesenchymal cells, and the coordinated series of events leading
to their formation strikingly resembles that observed
in epithelial-mesenchymal transformation in the embryo. Loss of surface polarity is accompanied by
changes in cytoplasmic organization and ultrastructure. Organelles of the cytoplasm no longer have the
thyroid cell morphology; the cytoplasm becomes
Epithelial-mesenchymal transformation in collagen
filamentous and contains mesenchyme-like RER.
The mesenchyme-like cells begin to express vimentin
as they leave the follicle and they eventually lose all
detectable keratin. Morphological changes in tissue
phenotype that occur during the transformation of
thyroid cells to mesenchyme-like cells are also accompanied by a loss of cytodifferentiation. Thyroglobulin, localized in thyroid epithelial cells either in
follicles or in monolayer culture, is not expressed in
the migrating mesenchyme-like cells. The change to
the mesenchymal phenotype seems to be stable, in
that the cells do not revert to epithelium when grown
on planar substrata.
Vimentin expression is an early event in the transformation of epithelium to mesenchyme-like cells and
occurs in elongating cells still attached to follicles.
Similarly, expression of vimentin by primitive streak
cells located within the embryonic ectoderm has been
reported to occur during the formation of primary
mesenchyme (Jackson etal. 1981; Franke etal. 1982).
Expression of vimentin by epithelial cells in long-term
culture (Franke et al. \919b) and by parietal endoderm (Lane et al. 1983) may be due to the cells
separating from each other (Lane et al. 1983). However, none of the thyroid epithelia cultured on a
planar collagen substratum under the conditions described here express vimentin, even when subconfluent. The change in cytoskeletal phenotype of
thyroid cells moving into collagen gels seems to be
related to acquisition of the ability to express mesenchyme-specific proteins.
Although further study is required, it is tempting to
conclude that the persistence of cytokeratin for a
week or two in the mesenchyme-like cells derived
from thyroid epithelium may result from the slow
turnover of this protein. During the proliferation of
the mesenchyme-like cells, cytokeratin could be
diluted out gradually during cytokinesis, leaving only
filaments of vimentin type. It is likely that cytokeratin
synthesis is turned off as the cells lose desmosomes,
express vimentin and extend cell processes into the
gel. A vimentin cytoskeleton may be required for cell
invasion and migration within the collagen matrix.
The primitive streak epithelium that gives rise to
vimentin-rich primary mesenchyme in the embryo
loses cytokeratin (Jackson et al. 1981). Coexpression
of vimentin and cytokeratin is not observed in mesenchymal cells (Franke et al. 1982), although it occurs
in certain epithelia (Lane etal. 1983). It is remarkable
that mesenchyme-like cells derived from thyroid
follicles are able to migrate while carrying keratin
remnants a week or more after leaving the thyroid
follicle.
The persistence of cytokeratin in the cytoplasm of
the newly forming mesenchyme-like cells is strong
proof of their origin from the follicular epithelium,
615
since no keratin is present in mesenchymal cells
surrounding follicles in situ. Indeed, the keratin stain
might be a useful marker for these cells in future
experiments to determine their developmental potential in the embryo. The conclusion that these mesenchyme-like cells originate from the follicular epithelium itself is also firmly supported by the
observation that the follicles studied here contain no
mesenchyme at the beginning of the experiment, that
no mesenchymal cells appear in BM gels and that
stages in the transition to mesenchyme can be observed in collagen gels. Indeed, we illustrated an
elongating cell that had acquired mesenchyme-type
RER in its distal elongating cytoplasm, while still
containing thyroid-type RER in its proximal cytoplasm. It seems clear that the thyroid cells dedifferentiate (lose thyroglobulin) as they move out into the
gel.
The fact that others have not described this
phenomenon in collagen gels may be due to its
misinterpretation as mesenchymal contamination.
Mauchamp and coworkers reported that thyroid epithelial cysts (inside-out follicles) reverse polarity to
form follicles when suspended within gelatin or 3D
collagen gels (Mauchamp etal. 1979; Chambord etal.
1981, 1984). During the polarity reversal, new apical
surfaces appear inside the aggregate without loss of
epithelial integrity (Kitajima et al. 1985; Barriere et
al. 1986). In suspension culture, low serum promotes
normal epithelial polarity (Nitsch & Wollman,
1980a,b). Wollman and coworkers also reported that
gels of type I collagen maintain normal polarity of
follicles, but they found no effect with gelatin (Garbi
et al. 1984a,b). We agree that the layer of thyroid
epithelial cells bordering the colloid-filled follicular
lumen are apically polarized in collagen gels. However, in 80 % of the follicles we studied in collagen
gels, basal cells lose thyroglobulin and transform into
mesenchyme-like cells. These freely wandering cells
might have been considered to be fibroblast contaminants by other authors and thus ignored (see also
Garbi et al. 1984b). Mammary gland epithelium has
been reported to form closed, epithelial structures in
collagen gels (Yang et al. 1979; Hall et al. 1982;
Haeuptle et al. 1983). The possibility of mesenchymal
cell formation in this system needs to be reexamined.
We have previously reported that embryonic epidermis, corneal epithelium, notochord and lens, and
adult corneal endothelium and lens, epithelia that
normally do not give rise to mesenchyme, do so when
suspended within collagen gels (Greenburg & Hay,
1982). In the example we previously studied in detail,
the anterior lens epithelium, cells proliferate from the
former apical surface to migrate out into the gel,
while cells in contact with lens capsule remain lenslike (Greenburg & Hay, 1982, 1986). Thus, basal
616
G. Greenburg and E. D. Hay
lamina stabilizes the epithelial phenotype of the basal
cells. On the other hand, we saw here that thyroid
luminal surfaces protected from exposure to the
collagen gel are maintained as apical surfaces. What
is remarkable is that surface polarity is lost as cells
detach into collagen from either the apical or the
basal surface, if these cells are in intimate contact
with type I collagen fibrils in a 3D configuration. It is
tempting to conclude that as new cell surfaces are
produced in the progeny of the thyroid and lens
cells, the cell surfaces contacting collagen gel adapt
to this new environment; cytoskeletal organization
and receptors for ECM must change to permit the
elongating cells to move out into the gel.
Additional evidence that the presence of a basal
lamina stabilizes the basal epithelial cell phenotype
Epithelial-mesenchymal transformation in collagen
617
>8A
Fig. 8. The stability of the mesenchyme-like phenotype is illustrated in these phase-contrast micrographs. Live cultures
of rat thyroid cells were examined after several weeks in culture. The plastic-gel interface was photographed in cultures
of follicles suspended within collagen gels (A) or placed on the surface of the collagen gels (B). When migrating cells
reach the gel edge, they leave the gel and migrate onto the plastic culture dish. The morphology of thyroid-derived
mesenchyme-like cells that have migrated out of the gel onto the plastic surface remains mesenchymal (A). In contrast,
the epithelial morphology is maintained in cultures after cell sheets that were growing on the surface of the gel migrate
over the edge onto the plastic dish (B). Both micrographs are the same magnification.
was obtained in this study. Thyroid follicles suspended in basement membrane gels do not give rise
to mesenchymal cells. The basal surface in contact
with the basal lamina is flat; no pseudopodia are
Fig. 7. Immunofluorescence localization of vimentin and
keratin in whole mounts of rat thyroid follicle and
mesenchyme-like cells transforming from the basal
surface. Cultures were stained after 3 days (A-D) or 22
days (E,F). (A) When thyroid follicles are reacted with
anti-vimentin, the epithelial cells of the follicle are
negative. Elongating cells, including pseudopodia, react
positively for vimentin. In the follicle shown here, a few
cells that have not yet extended cell processes into the gel
stain positively for vimentin (arrow). The majority of
follicle cells, however, remain negative. (B) Another
example showing that the follicle (/) does not react with
anti-vimentin, but cells elongating from the surface and
mesenchyme-like cells within the gel (arrow) fluoresce
brightly indicating the presence of vimentin intermediate
filaments. (C) A fully formed mesenchymal cell derived
from a follicle reacts well with anti-vimentin. (D) A
mesenchyme-like cell migrating freely within the gel
stains positively for keratin in a 3-day culture. As
expected, thyroid follicles fluoresce intensely when
stained with anti-keratin as well. (E,F) Phase contrast
and immunofluorescence showing that, by 22 days in
culture, the mesenchymal cell has lost reactivity with antikeratin. All micrographs are at the magnification shown
in C.
projected into the BM gel. However, the gel (Kleinman et al. 1986) does support migration of mesenchymal cells (Fig. 9; Greenburg, 1985; Bilozur & Hay,
1988). One might speculate that the presence of basal
lamina helps to maintain the normal basal polarity
(Sugrue & Hay, 1986) of epithelial ECM receptors.
Neither BM nor rat tail collagen gels contain fibronectin (Kleinman et al. 1986; Bilozur & Hay, 1988).
However, laminin and type IV collagen in BM gels
would be expected to interact with and stabilize the
epithelial basal surface ECM receptors. While this
may be the explanation for the effect of BM on
maintenance of the differentiated follicle suspended
within a gel, factors other than molecular composition of ECM must also be involved in stabilizing the
epithelial phenotype.
A case in point is the behaviour of thyroid epithelium cultured on top of a type I collagen gel. The
cells that migrate onto the planar collagen surface,
while somewhat flatter than those composing the
original follicle, are absolutely typical epithelial cells
and they express thyroglobulin, in contrast to the
mesenchyme-like cells that arise within type I collagen gel. For lack of better terminology, this effect
can be said to involve physical as well as molecular
factors. Presumably, the basal surface on top of a
collagen gel sees the same type I collagen fibrils as it
618
G. Greenburg and E. D. Hay
Fig. 9. Bovine thyroid follicles (A) and avian corneal fibroblasts (B) cultured in basal lamina for 5 days. The follicles
remain epithelial (A), but these gels do support migration of definitive fibroblasts (B). Both micrographs are at the
same magnification.
would when the epithelium is suspended in collagen.
However, from a physical point of view, the substratum has a planar form in the former case and the
cell must sense this. Epithelial cells supported on
Millipore filters can flatten the basal surface, maintain
epithelial polarity, and respond to soluble ECM
molecules (Sugrue & Hay, 1981). Soluble ECM
molecules do not promote mesenchymal differentiation or epithelial-mesenchymal transformation.
Finally, it should be noted that the apical surface of
an epithelium responds quite differently if an overlying collagen gel is added after the epithelium has
attached to the planar surface of an initial collagen
gel. The cut ends of the epithelium seem to migrate
on the second gel until a closed epithelial structure is
formed (Yang et al. 1979; Hall et al. 1982).
In the embryo, without exception, epithelialmesenchymal transformation occurs from the basal
surface of the epithelium. The basal lamina disappears before the mesenchymal cells emerge, but
changes occur in the epithelium prior to this point in
time: intercellular spaces increase and the staining
pattern of the presumptive mesenchyme changes
(Solursh et al. 1979; Nichols, 1981; Franke et al.
1982). Thus, it seems likely that a genetic program is
switched on which changes the cell surface (including
ECM receptors), cytoskeleton and polarity prior to
BM breakdown. Indeed, the emerging mesenchymal
cells may now express proteases (Valinsky & Le
Douarin, 1985) that aid in removal of basal lamina.
Some mesenchymal cells in the embryo, mainly those
derived from primitive streak, can reexpress the
epithelial phenotype. They produce basal lamina
prior to or at the time of reaggregation (Ekblom,
1984). The majority of embryonic mesenchymal cells
never reexpress the epithelial phenotype; they comprise the so-called secondary or late-formed mesenchyme (Hay, 1968) and they form connective tissue or
muscle. Thus, in the embryo it seems that genetic
programs precisely control the timing of epithelialmesenchymal transformation. While the composition
and deposition of adjacent ECM may play a major
role, it is the cells that initiate the process.
The discovery that confrontation of definitive epithelia with an unexpected 3D collagen gel triggers
epithelial-mesenchymal transformation in an otherwise stable epithelium, should provide a basis for
experiments to explain the manner in which such
genetic programs are turned on. Cells confronted on
either basal or apical surfaces with this physically
aberrant form of ECM lose polarity and migrate out
into the ECM. The epithelium unequivocally still
possesses the information to form mesenchyme-like
cells. During the transformation, nuclear as well as
cytoplasmic morphology is markedly altered and
upregulation of the mesenchymal genetic program is
closely tied to changes in cell shape. We have turned
on this genetic program by a presumably simple
process, and the explanation for the phenomenon
may shed light on the mode of activation of this
program in embryos (Hay, 1984), neoplasms (Battifora, 1976; Dulbecco et al. 1981) and lens pathologies
Epithelial-mesenchymal transformation in collagen
(Henkind & Prose, 1967; Font & Brownstein, 1974).
The cell shape change is accompanied by changes in
gene expression for specific proteins (thyroglobulin,
lens crystallins),ECM (Greenburg & Hay, 1986), and
10
619
cytoskeleton (keratin, vimentin). Moreover, we show
here and elsewhere (Greenburg & Hay, 1986) that,
like secondary mesenchyme in the embryo, the mesenchyme-like cells cannot be induced to reform
1.0 urn
Fig. 10. Electron micrograph of the basal surface of a bovine thyroid follicle cultured within basal lamina gel. The basal
surface remains flat and no mesenchyme-like cells or filopodia form, even after 10 days in culture.
620
G. Greenburg and E. D. Hay
Fig. 11. Summary diagram
illustrating the events occurring
during epithelial-mesenchymal
transformation of thyroid
follicles suspended in type I
collagen gels.
epithelium on a planar substratum. It will be interesting in the future to determine if they have acquired
the connective tissue developmental potentials of
true secondary mesenchyme.
We thank Dr Hynda K. Kleinman for her generous gift of
BM gel and Drs Robert D. Goldman and P. Reed Larsen
for antisera. We also thank Bev Pugrabi for editorial and
secretarial assistance. This work was supported by a grant
ROI-HD00143 from the National Institutes of Health. Gary
Greenburg was a recipient of an NRSA Predoctoral Fellowship GMO7226-08.
cells and from preformed monolayers. J. Cell Biol. 91,
157-166.
CHAMBARD, M., VERRIER, B., GABRION, J. & MAUCHAMP,
J. (1984). Polarity reversal on inside-out thyroid follicle
cultured within collagen gel: reexpression of specific
functions. Biol. Cell 51, 315-326.
DULBECCO, R., HENAHAN, M., BOWMAN, M., OKADA, S.,
BATTIFORA, H. & UNGER, M. (1981). Generation of
C. B. & MEIER, S. (1981). The influence of
the metameric pattern in the mesoderm on migration
of cranial neural crest cells in the chick embryo. Devi
Biol. 85, 385-402.
BARD, J. B. L. & HAY, E. D. (1975). The behavior of
fibroblasts from the developing avian cornea:
Morphology and movement in situ and in vitro. J. Cell
Biol. 67, 400-418.
fibroblast-like cells from cloned epithelial mammary
cells in vitro: a possible new cell type. Proc. natn.
Acad. Sci. U.S.A. 78, 2345-2349.
EHRMANN, R. L. & GEY, G. O. (1956). The growth of
cells on a transparent gel of reconstituted rat tail
collagen. J. natn. Cancer Inst. 16, 1375-1401.
EKBLOM, P. (1984). Basement membrane proteins and
growth factors in kidney differentiation. In The Role of
Extracellular Matrix in Development (ed. R. L.
Trelstad), pp. 173-206. New York: Alan R. Liss, Inc.
ELSDALE, T. & BARD, J. (1972). Collagen substrate for
studies on cell behavior. J. Cell Biol. 54, 626-637.
FONT, R. L. & BROWNSTEIN, S. (1974). A light and
electron microscopic study of anterior subcapsular
cataracts. Am. J. Ophthal. 78, 972-984.
BARRIERE, H., CHAMBARD, M., MAUCHAMP, J. &
GABRION, J. (1986). Polarity reversal of inside-out
FRANKE, W. W., APPELHANS, E., SCHMID, E.,
FREUDENSTEIN, C , OSBORN, M. & WEBER, K. (1979a).
References
ANDERSON,
thyroid follicles cultured within collagen gel: an
ultrastructural study. Biol. Cell 57, 39-51.
BATTIFORA, H. (1976). Spindle cell carcinoma:
Ultrastructural evidence of squamous origin and
collagen production by the tumor cells. Cancer 37,
2275-2282.
BILOZUR, M. E. & HAY, E. D. (1988). Neural crest
migration in 3D extracellular matrix utilizes laminin,
fibronectin, and collagen. Devi Biol. 125, 19-33.
CHAMBARD, M., GABRION, J. & MAUCHAMP, J. (1981).
Influence of collagen gel on the orientation of epithelial
cell polarity: Follicle formation from isolated thyroid
Identification and characterization of epithelial cells in
mammalian tissues by immunofluorescence microscopy
using antibodies to prekeratin. Differentiation 15, 7-25.
FRANKE, W. W., GRUND, C , KUHN, C , JACKSON, B. W.
& ILLMENSEE, K. (1982). Formation of cytoskeletal
elements during mouse embryogenesis. III. Primary
mesenchymal cells and the first appearance of vimentin
filaments. Differentiation 23, 43-59.
FRANKE, W. W., SCHMID, E., BREITKREUTZ, D., LUDER,
M., BOUKAMP, P., FUSENIG, N. E., OSBORN, M. &
WEBER, K. (1979b). Simultaneous expression of two
different types of intermediate filaments in mouse
Epithelial-mesenchymal transformation in collagen 621
keratinocytes proliferating in vitro. Differentiation 14,
35-50.
GARBI, C , MAJUMDER, A. L., TACCHETTI, C. &
WOLLMAN, S. H. (1984b). Migration of epithelial
cells
from separated thyroid follicles embedded in collagen
gel and fate of the follicles. J. Cell Biol. 97, 315a.
GARBI, C , NITSCH, L. & WOLLMAN, S. H. (1984a).
Embedding in a collagen gel stabilizes the polarity of
epithelial cells in thyroid follicles in suspension culture.
Expl Cell Res. 151, 458-465.
GREENBURG, G. (1985). Epithelial-mesenchymal
transformation in vitro. Ph.D. thesis. Harvard
University, Boston.
GREENBURG, G. & HAY, E. D. (1982). Epithelia
suspended in collagen gels can lose polarity and express
characteristics of migrating mesenchymal cells. /. Cell
Biol. 95, 333-339.
GREENBURG, G. & HAY, E. D. (1984). The formation of
mesenchyme-like cells from thyroid follicular
epithelium cultured in collagen gels. J. Cell Biochem.
Suppl. 8B, 181.
GREENBURG, G. & HAY, E. D. (1986). Cytodifferentiation
and tissue phenotype change during transformation of
embryonic lens epithelium to mesenchyme-like cells in
vitro. Devi Biol. 115, 363-379.
HAEUPTLE, M. T., SUARD, Y. L. M., BOGENMANN, E.,
REGGIO, H., RACINE, L. & KRAEHENBUHL, J. P. (1983).
Effect of cell shape change on the function and
differentiation of rabbit mammary cells in culture.
J. Cell Biol. 96, 1425-1434.
HALL, H. G., FARSON, D. A. & BISSELL, M. J. (1982).
Lumen formation by epithelial cell lines in response to
collagen overlay: a morphogenetic model in culture.
Proc. natn. Acad. Sci. U.S.A. 79, 4672-4676.
HAY, E. D. (1964). Secretion of a connective tissue
protein by developing epidermis. In The Epidermis (ed.
W. Montagna & W. C. Lobitz), pp. 97-119. New York:
Academic Press.
HAY, E. D. (1968). Organization and fine structure of
epithelium and mesenchyme in the developing chick
embryo. In Epithelial-Mesenchymal Interactions (ed.
R. Fleischmajer & R. E. Billingham), pp. 31-50.
Baltimore: W&W Co.
HAY, E. D. (1984). Cell-matrix interaction in the
embryo: cell shape, cell surface, cell skeletons, and
their role in differentiation. In The Role of
Extracellular Matrix Development (ed. R. L. Trelstad),
pp. 1-31. New York: Alan R. Liss, Inc.
HENKIND, P. & PROSE, P. (1967). Anterior polar cataractelectron microscopic evidence of collagen. Am. J.
Opthal. 63, 4003-4007.
IZUMI, M. & LARSEN, P. R. (1978a). Correlation of
sequence changes in serum thyroglobulin,
triiodothyronine, and thyroxine in patients with
Grave's disease and subacute thyroiditis. Metabolism
27, 449-460.
IZUMI, M. & LARSEN, P. R. (19786). Metabolic clearance
of endogenous and radioiodinated thyroglobulin in rats.
Endocrinology 103, 96-100.
JACKSON, B. W., GRUND, C , WINTER, S., FRANKE, W. W.
& ILLMENSEE, K. (1981). Formation of cytoskeletal
elements during mouse embryogenesis. II. Epithelialdifferentiation and intermediate-sized filaments in early
postimplantation embryos. Differentiation 20, 203-216.
JONES, J. C , GOLDMAN, A. E., STEINERT, P. M., YUSPA,
A. & GOLDMAN, R. D. (1982). Dynamic aspects of the
supramolecular organization of intermediate filament
networks in cultured epidermal cells. Cell. Motil. 2,
197-213.
KITAJIMA, K., YAMASHITA, K. & FUJITA, H. (1985). Fine
structural aspects of the shift of zonula occludens and
cytoorganelles during the inversion of cell polarity in
cultured porcine thyroid follicles. Cell Tissue Res. 242,
221-224.
KLEINMAN, H. K., MCGARVEY, M. L., HASSELL, J. R.,
STAR, V. L., CANNON, F. B., LAURIE, G. W. &
MARTIN, G. R. (1986). Basement membrane complexes
with biological activity. Biochem. J. 25, 312-318.
LANE, E. B., HOGAN, B. L. M., KURKINEN, M. &
GARRELS, J. I. (1983). Coexpression of vimentin
and
cytokeratins in parietal endoderm cells of early mouse
embryo. Nature, Lond. 303, 701-704.
MAUCHAMP, J., MARGOTAT, A., CHAMBARD, M.,
CHARRIER, B., REMY, L. & MICHEL-BECHET, M.
(1979).
Polarity of three-dimensional structures derived from
isolated hog thyroid cells in primary culture. Cell
Tissue Res. 204, 417-430.
NICHOLS, D. H. (1981). Neural crest formation in the
head of the mouse embryo using a new histological
technique. J. Embryol. exp. Morph. 64, 105-120.
NITSCH, L. & WOLLMAN, S. H. (1980a). Suspension
culture of separated follicles consisting of differentiated
thyroid epithelial cells. Proc. natn. Acad. Sci. U.S.A.
77, 472-476.
NITSCH, L. & WOLLMAN, S. H. (1980/?). Ultrastructure of
intermediate stages in polarity reversal of thyroid
epithelium in follicles in suspension culture. J. Cell
Biol. 86, 875-880.
OVERTON, J. (1977). Response of epithelial and
mesenchymal cells to culture on basement lamella
observed by scanning microscopy. Expl Cell Res. 105,
313-323.
SOLURSH, M., FISHER, M., MEIER, S. & SINGLEY, C. T.
(1979). The role of the extracellular matrix in the
formation of the sclerotome. J. Embryol. exp. Morph.
54, 75-98.
STARGER, J. M., BROWN, W. E., GOLDMAN, A. E. &
GOLDMAN, R. D. (1978). Biochemical and
immunological analysis of rapidly purified 10 nm
filaments from baby hamster kidney (BHK-21) cells. J.
Cell Biol. 78, 93-109.
SUGRUE, S. P. & HAY, E. D. (1981). Response of basal
epithelial cell surface and cytoskeleton to solubilized
extracellular matrix molecules. J. Cell Biol. 91, 45-54.
622
G. Greenburg and E. D. Hay
S. P. & HAY, E. D. (1986). The identification of
extracellular matrix (ECM) binding sites on the basal
surface of embryonic corneal epithelium and the effect
of ECM binding on collagen production. J. Cell Biol.
102, 1907-1916.
VALINSKY, J. E. & LE DOUARIN, N. M. (1985).
Production of plasminogen activator by migrating
cephalic neural crest cells. EMBO J. 4, 1403-1406.
SUGRUE,
YANG, J., RICHARDS, J., BOWMAN, P., GUZMAN, R.,
ENAMI, J., MCCORMICK, K., HAMAMOTO, S., PITELKA,
D. & NANDI, S. (1979). Sustained growth and three-
dimensional organization of primary mammary tumor
epithelial cells embedded in collagen gels. Proc. natn.
Acad. Sci. U.S.A. 76, 3401-3405.
{Accepted 12 November 1987)