Content uploaded by C. Schmidhauser
Author content
All content in this area was uploaded by C. Schmidhauser
Content may be subject to copyright.
Available via license: CC BY-NC-SA 4.0
Content may be subject to copyright.
Available via license: CC BY-NC-SA 4.0
Content may be subject to copyright.
Extracellular Matrix Regulates Expression of the TGF-/31 Gene
Charles H. Streuli,* Christian Schmidhauser,* Michael Kobrin,* Mina J. Bissell,* and Rik Derynckw
* Cell and Molecular Biology Division, Lawrence Berkeley Laboratory, Berkeley, California 94720; ~ Department of
Developmental Biology, Genentech Inc., South San Francisco, California 94080; and w Departments of Growth and Development, and
Anatomy, Programs of Cell Biology and Developmental Biology, University of California, San Francisco, California 94143-0640
Abstract.
Transforming growth factor-# (I'GF-#) is a
potent regulator of cell proliferation and modulates the
interactions of cells with their extracellular matrix
(ECM), in part by inducing the synthesis of various
ECM proteins. Three different isoforms of TGF-/$ are
synthesized in a defined pattern in specific cell popula-
tions in vivo. In the specific case of TGF-fll, this well-
defined and limited expression stands in sharp contrast
to its synthesis by virtually all cells in culture.
Using mammary epithelial cells as a model system,
we evaluated the substratum dependence of the expres-
sion of TGF-/31. The level of TGF-fll expression is
high in cells on plastic, but is strongly downregulated
when cells are cultured on a reconstituted basement
membrane matrix. In contrast, TGF-fl2 mRNA levels
in cells on either substratum remain unchanged. Using
the chloramphenicol acetyl transferase gene as reporter
gene under the control of the TGF-/31 promoter, we
show that transcription from this promoter is suppressed
when the cells are in contact with either endogenously
synthesized or exogenously administered basement
membrane. TGF-/$1 promoter activity is strongly in-
duced by the absence of basement membrane, i.e., by
direct contact of the cells with plastic. This modulation
of transcription from the TGF-fll promoter occurs in
the absence of lactogenic hormones which allow full
differentiation. Our results thus indicate that basement
membrane is an important regulator of TGF-B 1 syn-
thesis, and explain why most cells in culture on plastic
express TGF-/~I in contrast with the more restricted
TGF-/~I synthesis in vivo. We propose that there is a
feedback loop whereby "I'GF-/~l-induced synthesis of
basement membrane components is repressed once a
functional basement membrane is present. Finally, these
results together with our current knowledge of regula-
tion of TGF-/~I and TGF-fl2 synthesis, suggest that, in
vivo, TGF-/31 may play a major role in regulating the
ECM synthesis and the ceI1-ECM interactions, whereas
TGF-/Y2 may be more important in morphogenetic
processes.
M
UCH attention has been focused in recent years
on the biology of transforming growth factor-fl
(TGF-fl),* a potent regulator of cell proliferation.
TGF-/$ can exert a variety of effects, depending on the nature
of the target cell and the physiological conditions. It stimu-
lates the proliferation of many cell types ofmesenchymal ori-
gin, yet is growth inhibitory for many other cell types. A ma-
jor activity of TGF-fl is its ability to stimulate the synthesis
and deposition of various extracellular matrix (ECM) pro-
teins and to increase the expression of integrins, receptors
that mediate cellular interactions with ECM proteins. This
is frequently accompanied by an increased synthesis of pro-
tease inhibitors and a repression of the synthesis of ECM-
Charles H. Streuli's present address is Department of Cell and Structural
Biology, University of Manchester, Stopferd Building, Oxford Road, Man-
chester M13 9PT, England.
Mike Kobrin's present address is Department of Medicine, University of
California, Medical Sciences 1, C240, Irvine, CA 92717.
1. Abbreviations used in this paper:
CAT, chloramphenicol acetyltransfer-
ase; ECM, extracellular matrix; EHS, Engelbreth-Holm-Swarm; TGF-fl,
transforming growth factor-ft.
degrading proteases (for review see Roberts and Sporn,
1990). Taken together, these activities can result in increased
celI-ECM interactions and adhesiveness. In comparison
with other secreted factors, TGF-/~ can be considered as the
most potent regulator known to date of ECM formation and
celI-ECM interactions.
Our current knowledge of the localization and biological
activities of TGF-fl suggests that it exerts its activities in the
context of organ and tissue remodeling (Heine et al., 1987;
Thompson et al., 1989; Fitzpatrick et al., 1990; Gatherer et
al., 1990; Millan et al., 1990; Pelton et al., 1990a,b, 1991;
Schmid et al,, 1990). This implies that TGF-/5 may function
as a physiological regulator of woundhealing, developmental
tissue differentiation and morphogenesis, and tumor devel-
opment. Since most cells have the capacity to express and
secrete TGF-fl and have cell surface TGF-fl receptors
(Roberts and Sporn, 1990), it is generally believed that
TGF-fl exerts its role by autocrine and paracrine mecha-
nisms. In addition, since TGF-fl binds to several ECM pro-
teins and since ECM contains TGF-/3, it can be assumed that
the ECM and especially the basement membranes represent
a reservoir from which TGF-fl can be retrieved so that it can
exert its various activities on the surrounding cells.
9 The Rockefeller University Press, 0021-9525/93/01/253/8 $2.00
The Journal of CeU Biology, Volume 120, Number 1, January 1993 253-260 253
Several species of TGF-/5 have been identified and charac-
terized by cDNA cloning. Three different TGF-B isoforms,
each encoded by separate genes, are expressed in mammalian
call systems. These TGF-B species have a high degree of
structural identity in their mature polypeptide sequences and
have similar biological activities in culture. Their cellular
synthesis is subject to differential regulation and they are
differentially localized in vivo. In situ hybridization and
isoform-specific immunohistochemistry of the TGF-~I, -/52,
and -/~3 mRNAs and proteins have documented that the three
TGF-B species have a well-defined and characteristic expres-
sion pattern during murine development (Heine et al., 1987;
Thompson et al., 1989; Fitzpatrick et al., 1990; Gatherer et
al., 1990; Millan et al., 1990; Pelton et al., 1990a,b, 1991;
Schmid et al., 1990). Their cell and tissue distribution is
partially overlapping, frequently resulting in the expression
of more than a single TGF-/~ form in specific cells and tis-
sues. This localized expression pattern of the individual
TGF-/3 forms is also reflected to some degree in cells in cul-
ture. TGF-/T2 and -B3 expression by cells in culture is appar-
ent only in select cell lines and primary cultures, which are,
as far as can be assessed, in accordance with cell types nor-
mally expressing these TGF-/3 species in vivo. In contrast,
the expression pattern of TGF-B1 in cultured cells clearly
contradicts the localized TGF-B1 expression in vivo. For ex-
ample, during mouse development TGF-B1 expression oc-
curs mainly in cell types of hematopoietic origin, endothelial
cells and epithelia, especially in skin. Yet, cell culture ex-
periments indicate that virtually all cells synthesize TGF-/~I
in culture, in disagreement with the findings in vivo.
In this study, we sought to understand the basis of the dis-
crepancy between the virtually universal expression of TGF-
ffl in cultured cells and its localized expression in vivo. In
particular, we have paid attention to the influence of the sub-
stratum on the expression of TGF-B1 by mammary epithelial
cells. It is indeed conceivable that removal of the cells from
their natural environment and especially the absence of
ECM might contribute to the discrepancy between their be-
havior in culture and in vivo. Thus, we hypothesized that
ECM itself might be able to regulate TGF-B1 expression, and
that the absence of a physiological substratum when cells are
cultured on plastic dishes, which resembles a "wounded" en-
vironment, might induce the otherwise suppressed produc-
tion of TGF-B1. Therefore, it is likely that increased TGF-B1
synthesis would subsequently induce the secretion and depo-
sition of several ECM proteins in an autocrine manner, con-
tributing to an establishment of more physiological cell-
ECM interactions.
We have evaluated our hypothesis using epithelial cells de-
rived from the mouse mammary gland. These cells have
been used previously to explore the influence of ECM on the
regulation of gene expression and differentiation (Lee et al.,
1984, 1985; Li et al., 1987; Chen and Bissell, 1989;
Barcellos-Hoff et al., 1989; Aggeler et al., 1991; Schmid-
hauser et al., 1990, 1992; Streuli and Bissell, 1990; Streuli
et al., 1991). Our findings document that the absence of a
basement membrane has a profound and specific effect on the
expression of TGF-B1, resulting in an upregulation of TGF-
ffl expression when cells are maintained on plastic as sub-
stratum. Using the TGF-B1 promoter linked to a reporter
gene, we show that this effect is due to an enhanced transcrip-
tion from the TGF-B1 promoter. The induction of TGF-B1 ex-
pression in the absence of ECM and its suppression by the
presence of a basement membrane may provide a feedback
loop resulting in a balance between TGF-BI expression and
basement membrane formation.
Materials and Methods
Substrata and Cell Culture
EHS matrix was prepared from EHS tumors passaged in C57BL mice
(Kleinman et al., 1986) and used as described (Barcellos-Hoff et al., 1989;
Streaili et al., 1991). Collagen I was prepared from rat tails (Lee et al., 1984)
and stored at a concentration of 2-3 mg/ml. Thick collagen gels were made
with 150 #1 collagen I per cm 2 and either left attached to the dish or
released into the medium 24 h after plating (Streuli and Bissell, 1990).
Primary mammary epithelial cells were prepared from 14.5-d-pregnant
CD-1 mice and cultured on physiological substrata in the presence of insulin
(5 #g/ml) and lactogenic hormones (1 #g/ml hydrocortisone, 3 t~g/ml pro-
lactin) as described elsewhere (Lee et aL, 1985; Emerman et al., 1988;
Streuli et al., 1991). The mouse IDamrnary epithelial cell line CID-9 and
its transfected derivatives were cultured routinely in DME/FI2 medium
containing 5% FCS and 5 #g/ml insulin (Schmidhauser et al., 1990). For
CAT assays and differentiation assays, these cells were plated onto relevant
substrata and cultured with 2 % serum for 24 h. The medium was then
changed to serum-free DME/FI2 containing insulin and lactogenic hor-
mones as above, and the cells were fed every 2 d.
mRNA Analysis
RNA was extracted from tissues and cultured cells by the gnanidinium thio-
cyanate method as described (Sambrook et al., 1989). Folyadenylated RNA
was isolated by adsorption to oligo d(T)-ce]lniose (Sambrook et al., 1989).
Northern analysis for the presence of specific mRNAs was done as follows.
Total RNA was electrophoresed in 1% agarose, 5 % formaldehyde gels, fol-
lowed by etectrotransfer to nylon membranes (Oenescreen or Hybond-N)
and UV cross-linking. Equal loading was confu'med by comparing intensi-
fies of ethidium bromide-stained ribosomal RNAs. Membranes were
probed with TGF-~I (Derynck et al., 1986) or -f12 (Miller et al., 1989)
specific cDNA fragments 32p-radiolabeled by the random priming method
(Boehringer Mannbeim Corp., Indianapolis, IN) according to the manufac-
turer's directions. All hybridizations were performed under high stringency
conditions (Derynck et al., t988). Unhybridized probe was removed by
washing membranes with O.1x SSC, 0.2% SDS at 60~
Quantitation of TGF-[3 Secretion
The conditioned medium of cultured ceils was collected and assayed for
TGF-/~-induced inhibition of proliferation of CCL64 mink lung cells with
or without acid activation, as previously described (Arrick et al., 1990;
Meager, 1991).
TGF-[31 Promoter-CAT Expression Plasmid
The 1,373-bp BamHI-SacII restriction fragment containing the TGF-B1
promoter and the first 11 bp of the long 5' untranslated region (Kim et al.,
1989) was isolated from the recombinam k-phage BMA1, which carries the
first exon of the human TGF-BI precursor gene (Derynck et al., 1987). This
restriction fragment was placed upstream of the sequence coding for the
CAT coding sequence in such a way that the ATG initiator codon followed
the SaclI recognition site by only four basepalrs: CCGCGG CAAA ATG.
The sequence immediately preceding the ATG was designed to achieve an
optimal translation efficiency (Kozak, 1989). This sequence context was in-
troduced by site-directed mutagenesis on the template incorporated into
M13 mpl8 as described (Zoller and Smith, 1984). All sequences were deter-
mined using the dideoxy sequencing method on single-stranded templates
(Sambrook et al., 1989). The final plasmid pBIP-CAT incorporated the tran-
scription unit linking the TGF-B1 promoter to the CAT coding sequence,
in the same plasmid background as the plasmid S/A-CAT (Jakobovits et al.,
1988).
Transformations and Selection
Rapidly growing CID-9 cells were transfected with pBIP-CAT by the cal-
cium phosphate procedure (Sambrook et al., 1989). 40 #g plasmid DNA
The Journal of Cell Biology, Volume 120, 1993 254
(and 4 #g pSV2neo DNA) were applied per 85-ram culture dish and stable
transfected cells were selected using 400 ~g G418/ml medium. After 4 wk,
surviving clones were pooled and expanded. Cells were plated at equal den-
sity (6.2 x 104 per cm 2) on plastic culture dishes, on Engelbreth-Holm-
Swarm (EHS) matrix, or on collagen I gels for various assays as described
previously (Schmidhauser et al., 1990).
CAT Assays
To measure the CAT enzyme levels, as an indication of promoter activity,
expressed from the TGF-/~I promoter in transfected cells cultured under
different conditions, cells were harvested with dispase (40-60 win, 37~
washed to remove dispase, lysed in 0.5% NP-40 and further solubilized by
freeze thawing. To normalize the extracts, cell protein was quantified in a
micro Bradford assay; excess EHS matrix which would interfere with the
protein assay was removed efficiently by the dispase treatment. Aliquots
representing equal amounts of total cell protein (10 #g) were incubated with
14C-chioramphenicol (0.1 #Ci) and acetyl coenzyme A (70 #g) for 6 h
at
37~ extracted with ethyl acetate and separated by TLC. The levels of
acetylated chioramphenicol were quantitated by scraping the radioactive
spots from the TLC plate and counting in a scintillation counter (Sambmok
et ai., 1989). All measurements were done in duplicate and the difference
between the duplicate values was <5%.
Protein Assays
To confirm that the cell lysates had been normalized correctly, 3 #g of the
cell protein from CAT assay lysates were separated on reducing SDS-poly-
acrylawide gels and then stained with silver. Western blotting was per-
formed by separating 1 #g of cell protein on reducing SDS-polyacrylamide
gel, transferring the protein onto Immobilon-P transfer paper, and then
probing the blot with an antibody that recognizes a spectrum of mouse
caseins (Lee et al., 1985).
Results
ECM-dependent Regulation of TGF-fll mRNA Levels
To assess whether TGF-/?I was expressed by mammary gland
and whether its expression was developmentally regulated,
the steady-state levels of mRNA from glands of CD-1 mice
at
different stages of the pregnancy cycle were compared by
Northern analysis (Fig. 1). The presence of TGF-/31 mRNA
was apparent in the mammary tissue of nonpregnant mice
(Fig. 1, lane/). During two periods of active tissue remodel-
ing, i.e., during pregnancy (Fig. 1, lane 2) and gland involu-
tion after lactation (Fig. 1, lane 4), the level of TGF-fll
mRNA in the nmmnmry gland was more abundant. In con-
trast, TGF-fll mRNA levels were downregulated during
lactation when milk is being synthesized;
at
this stage no
TGF-/Yl transcripts were detected in Northern blots of un-
Figure 2.
TGF-/~I mRNA is
upregulated in cells cultured
on plastic dishes. RNA was
prepared from midpregnant
mammary gland (lane 1), and
from uncultured epithelial cells
isolated from the gland (lane
2). These epithelial cells were
also plated onto plastic dishes
(lane 3) or EHS matrix (lane
4) and cultured for 6 d before
extracting total RNA. A North-
era blot of these RNAs (10 #g)
was then hybridized with a
eDNA for human TGF-fll.
fractionated mRNA (Fig. 1, lane 3). Such temporal varia-
tions during development of the mammary gland also occur
with the expression of TGF-/~2 and -f13 mRNA (Robinson et
al., 1991).
The presence of TGF-/~I transcripts in mammary gland
prompted us to ask whether any of the TGF-fl species were
expressed by primary epithelial cells in culture. The epithe-
lied cell component of the mammary glands of mid-pregnant
mice were separated from adipose and stromed tissue by col-
lagenase digestion, followed by differential centrifugation
(Lee et al., 1985; Emerman and Bissell, 1988; Streuli et al.,
1991). TGF-fll mRNA was clearly present in these uncul-
tured
epithelial
cells (Fig. 2, lane 2). The higher level of
"IGF-fll mRNA from undigested gland tissue (Fig. 2, lane
2) presumably reflects a significant contribution from the
stromed tissue, which was removed during the epithelial cell
isolation procedure. Experiments with epithelium-free
mammary fat pads have confirmed that, indeed, the mam-
mary stroma does express significant quantities of TGF-fll
during pregnancy (Robinson et al., 1991).
Polyadenylated mRNA was then prepared from cultured
cells and was shown to contain TGF-ffl and TGF-B2 tran-
scripts (Fig. 3). TGF-fl3 was barely detectable (not shown).
Because considerable phenotypic changes occur in mam-
mary epithelial cells cultured on different substrata (Aggeler
Figure I. The
levels of TGF-/J 1
mRNA are modulated during
the pregnancy cycle. Total
RNA was extracted from the
mammary glands of 8-wk-old
virgin mice (lane 1), and from
the glands of 14.5-d-pregnant
(lane 2), 2-d-lactating (lane
3), and 5-d-involuting mice
(lane 4). 10 #g RNA, sep-
arated in 1% agarose-formal-
dehyde gels, was transferred
to nylon membranes and the
resulting Northern blot was
hybridized with a eDNA for
human TGF-31.
Figure 3.
Expression of TGF-31, but not TGF-fl2, is substratum de-
pendent. Northern blots of polyadenylated mRNAs (10/~g) isolated
from primary mouse mammary epithelial-cells that had been cul-
tured for 6 d either on plastic dishes (lanes 1) or on EHS matrix
(lanes 2), were probed in parallel with radiolabeled eDNAs coding
for human TGF-B1 and human TGF-fl2. The four previously identi-
fied TGF-/?2 transcripts were evident in mammary epithelial cells.
A mouse glyceraldehyde phosphate dehydrogenase eDNA (mGAP)
provided a suitable control for confirming equal loading of the two
different mRNAs.
Streuli et al.
Extracellular Matrix Regulates Expression of the TGF-~I Gene
255
et al., 1991; Streuli and Bissell, 1990), we compared the
steady-state levels of TGF-fl mRNAs in cells plated on plas-
tic dishes (Fig. 3, lanes/) and on a basement membrane-
rich matrix derived from EHS tumour (EHS matrix, or
"Matrigel") (Fig. 3, lanes 2). When these cells were cultured
on this matrix, TGF-~ mRNA was drastically reduced in
comparison with cells on plastic (Fig. 3). The level of TGF-
/32 mRNA, on the other hand, was unchanged.
Measurement of TGF-/~ protein levels by a sensitive bioas-
say was frequently impossible due to the presence of an in-
hibitor of the assay in the conditioned medium (data not
shown). In addition, antibody-based TGF-fl assays did not
provide sufficient sensitivity to measure TGF-/$ levels under
our conditions. However, when only low levels of this inhibi-
tor were apparent, it was clear that the levels of total TGF-fl
(thus including both TGF-/31 and -/32) was significantly
higher when the cells were cultured on plastic, compared
with cells cultured on ECM. As an example, after 3 d the
concentration of secreted TGF-/$ in the conditioned medium
was 2.1 ng/ml, when grown on ECM, versus 3.1 ng/ml from
cells on plastic. Between d 3 and 5, the cells on ECM
secreted 1.3 ng/ml TGF-/~, whereas cells on plastic produced
5.7 ng/ml TGF-/3. In yet another experiment, the level of
TGF-/$ expressed by the cells on plastic was 2.1-fold higher
than the TGF-/~ secreted by the cells on EHS matrix. Thus,
although it was not possible to discriminate between both
TGF-/3 isoform protein levels, our mRNA and protein results
do demonstrate that these different substrata have dramatic
and specific effects on the expression of TGF-/Sq, but not
TGF-I$2 mRNAs. We know from previous studies (Li et ai.,
1987; Barcellos-Hoff et al., 1989), which are also confirmed
here (see below), that milk protein expression is induced on
the basement membrane matrix and inhibited on plastic.
Functional differentiation into polarized milk secreting cells
and TOF-/$1 expression are therefore regulated in opposite
directions.
To ask whether culturing the cells on plastic induced TGF-
/31 expression or whether ECM inhibited expression, we
compared TGF-ffl mRNA levels in isolated mammary epi-
thelial cells before cell culture with those cultured on either
plastic or on EHS matrix (Fig. 2, lanes 3 and 4, respec-
tively). The TGF-ffl mRNA level in the cells cultured on
EHS matrix was similar to that in uncultured cells and was
not inhibited. However, the much higher TGF-ffl mRNA
level in cells cultured on plastic stands in sharp contrast with
that in uncultured cells. Thus the absence of a suitable ECM
as substratum resulted in an increased TGF-~ expression,
whereas culture of the epithelial cells on the basement
membrane-rich EHS matrix maintained the TGF-/31 expres-
sion at similar levels to those seen in vivo.
ECM Regulates TGF-[31 Expression at the
Transcriptional Level
To evaluate whether this substratum dependence was a result
of transcriptional regulation, we measured the expression of
an easily assayable reporter gene under the control of the
TGF-/~I promoter. We isolated a restriction fragment con-
taining the TGF-/31 promoter (Kim et al., 1989) from a
recombinant phage containing the first exon of the human
TGF-/~I gene (Derynck et al., 1987). Previous studies have
shown a high degree of sequence conservation between the
human and mouse TGF-/31 promoters and especially their
putative regulatory sequences (Geiser et al., 1991). The
1,373-bp BamI-l/-SaclI fragment containing the promoter
region and the first 11 bp of the 5' untranslated region was
linked upstream from the sequence encoding chlorarnpheni-
col acetyltransferase (CAT). The resulting expression plas-
mid TGF-fllP-CAT containing the TGF-/31 promoter and 5'
untranslated sequence fused to the CAT coding sequence was
then transfected into the mammary epithelial cell line CID-9,
which has been shown to behave and differentiate very simi-
laxly to freshly isolated primary mammary epithelial cells
(Schmidhauser et al., 1990). In addition, these cells also dis-
played a downregulated TGF-fll mRNA expression when
cultured on basement membrane (data not shown). Cotrans-
fection of the TGF-/31 promoter plasmid with a neomycin-
resistance encoding plasmid pSV2-Neo allowed the selection
and generation of stable transfected cell clones, that con-
tained integrated pTGF-fllP-CAT sequences.
The stable transfected cell lines were then cultured on
plastic culture dishes for 6 d and compared with those on
EHS matrix. The levels of accumulated CAT enzyme, ex-
pressed from the TGF-/31 promoter, were quantitated using
the standard assay for ~4C-chloramphenicol conversion
(Fig. 4, lanes I and 2). There was a considerable difference
in the CAT enzyme levels (10-13 times more CAT enzyme
in cells cultured on plastic), and thus in the TGF-fll promoter
activity, in the ceils cultured on the two types of substrata.
The lack of extracellulax matrix therefore results in a dra-
matic increase of transcription from the TGF-fll promoter,
and the difference in TGF-fll mRNA levels between cells cul-
tured on plastic and ECM is due largely to differences in
transcriptional activity.
Basement Membrane Itself Provides Signals to
Regulate TGF-(31 Expression
The pTGF-/31P-CAT plasmid transfected CID-9 cells used
above had been cultured so far in the presence of lactogenic
hormones, prolactin and hydrocortisone. However, the com-
bined presence of both these components and EHS matrix
results in the acquisition of a differentiated phenotype char-
acteristic of mammary epithelial cells in vivo. A marker of
this differentiated phenotype is the expression of caseins
which constitute the major proteins in milk. To examine
whether the low expression from the TGF-/31 promoter is
related to functional differentiation or is solely the result of
cultivation on ECM, the transfected cells were cultured on
plastic or on ECM in the absence of prolactin and hydrocorti-
sone (Fig. 4, lanes 3 and 4). There was no functional
differentiation under these conditions, as assessed by the ab-
sence of caseins. However, CAT enzyme accumulated to
similar levels in either the presence or absence of the lacto-
genie hormones. Thus, the difference in transcription from
the TGF-ffl promoter is not linked to the acquisition of
differentiation-specific functions per se, but is solely a re-
sponse to the nature of the substratum.
An explanation for the effect of the substrata might be that
the mammary epithelial cells on ECM or plastic secrete a
factor that influences the level of TGF-fll promoter activity.
It is for example known that conditioned medium from mam-
mary cells cultured on plastic suppresses the synthesis of one
of the milk proteins, whey acidic protein (Chen and Bissell,
1989). A simple experiment was therefore performed to see
whether secreted factors were responsible for the observed
The Journal of Cell Biology, Volume 120, 1993 256
Figure 5. TGF-/~ promoter ac-
tivity is downregulated on
floating collagen I gels. p#lP-
CAT-transfected cells were
cultured on thick collagen I
gels for 6 d. The gels either
were left attached to the cul-
ture dish (lanes 1) or were
floated into the medium after
2 d (lanes 2). Total cell protein
was then separated for (A)
CAT assays of TGF-/~ pro-
moter activity, or (B) Western blotting with a casein-specific anti-
body. In the experiment inA, the levels of acetylated chlorampherti-
col expressed as cpm per/~g cell protein per min enzymatic reaction
time (total reaction time was 6 h) were as follows: (lane 1) 7.76;
(lane 2) 0.28.
Figure 4. ECM controls the activity of the TGF-/~I promoter.
Pooled clones of CID-9 ceils, stably transfected with p/~IP-CAT,
were plated at equal density (6.2 x 104 per cm z) on plastic culture
dishes (lanes 2, 4, 6, and 8) or on EHS matrix (lanes 1, 3, 5, and
7). The cells were then cultured either with (lanes 1, 2, and
5-8)
or without (lanes 3 and 4) prolactin. In one experiment (lanes 5-8),
medium conditioned from cells cultured on plastic dishes or on
EHS matrix was added for 4 d; this medium was either added back
to the homologous dishes (lanes 5 and 6) or to the converse culture
conditions (lanes 7and 8). (A) CAT assays of TGF-/~I promoter ac-
tivity in the cells cultured under these different conditions. In this
experiment the levels of acetylated chloramphenicol expressed as
cpm per #g cell protein per min enzymatic reaction time (total re-
action time was 6 h) were as follows: (lane 1) 0.21; (lane 2) 2.09;
(lane 3) 0.26; (lane 4) 3.27; (lane 5) 0.24; (lane 6) 2.20; (lane 7)
0.16; and (lane 8) 1.54. (B) Cell proteins separated by polyacryla-
mide gel electrophoresis and stained with silver confirmed that the
ceil lysates indeed had been normalized correctly, and showed that
only on the EHS matrix and only in the presence of prolactin (lanes
1, 5, and 7) were substantial amounts of/~-casein synthesized. (C)
That these bands were actually casein was shown by Western blot-
ting 1/~g of ceil protein onto Immobilon-P transfer paper, and then
probing the blot with an antibody that recognizes a spectrum of
mouse caseins.
modulation of TGF-~ expression. The transfected CID-9
cells were cultured in the presence of conditioned medium
from cells on either plastic dishes or the basement mem-
brane EHS matrix in the presence of lactogenic hormones
(Fig. 4, lanes 5-8). CAT assays and parallel casein assays
showed that the TGF-/31 promoter activity was not affected
by these different conditioned media, and, thus, did not de-
pend on the presence of secreted soluble factors.
Formation of an Endogenous Basement Membrane
Downregulates TGF-[31 Promoter Activity
We have shown previously that the differentiation of mam-
mary ceils that occurs as a result of cultivation on a floating
type I collagen substratum, correlates with the formation of
an endogenous basement membrane (Streuli and Bissell,
1990). We therefore evaluated the influence of collagen type
I matrix on TGF-/5'I promoter activity in the presence of lac-
togenic hormones. If downregulation of TGF-/31 expression
on EHS matrix is due to the presence of basement membrane
components, transcription from the TGF-~I promoter
should be reduced after flotation of the collagen I gel. On
attached collagen I gels, CID-9 cells behave similarly to
those on plastic; in addition, they do not differentiate and do
not synthesize a basement membrane (Streuli and Bissell,
1990). This is in contrast to their behavior on floating colla-
gen I gels, where the cells contract the gel and synthesize a
basement membrane, and, as a result of these changes,
differentiate and secrete milk proteins. Under these condi-
tions, the formation of a basement membrane can be demon-
strated by immunostaining. The transfected CID-9 cells
were therefore cultured on attached or floating collagen I
gels and the CAT activities directed from the TGF-/~I pro-
moter were measured. The cells on attached collagen gels
displayed a high level of promoter activity, analogous to cells
on plastic, whereas on the floating gels, the cells exhibited
a greatly suppressed promoter activity as seen with cells on
EHS matrix (Fig. 5). Thus, as with an exogenous basement
membrane, the presence of an endogenous basement mem-
brane suppresses the transcription from the TGF-/31 pro-
moter. Since we have already shown that such suppression
of the promoter is not dependent on the differentiated pheno-
type, we can conclude that the basement membrane itself
provides signals to regulate TGF-/31 transcription.
Discussion
In this study, we have established that expression of the TGF-
/~1 gene is regulated negatively by extraceUular matrix. Tran-
scription from the TGF-/31 promoter is high in the absence
Streuli et al.
Extracellular Matrix Regulates Expression of the TGF-~I Gene
257
of ECM and is considerably lower in the presence of EHS
matrix or an endogenously synthesized basement mem-
brane. The much lower TGF-BI promoter activity in the
presence of basement membrane is not a result of differentia-
tion per se, since inhibition of functional differentiation (by
removing lactogenlc hormones) does not alter TGF-/3 pro-
moter activity. This type of regulation of TGF-/3 expression
is specific for TGF-/31 since the level of TGF-/32 mRNA in
the same epithelial cells is not affected by the absence of
ECM.
An evaluation of the available data on TGF-/31 expression
by different cell types reveals a major discrepancy between
the TGF-/31 synthesis in vivo and in cell culture. In vivo, only
a defined subset of cell populations expresses TGF-/31 (Wilcox
and Derynck, 1988; Thompson et al., 1989; Fitzpatrick et
al., 1990; Gatherer et al., 1990; MiUan et al., 1990; Pelton
et al., 1990a,b, 1991; Schmid et al., 1990), whereas in cul-
ture its expression is virtually ubiquitous. On the basis of our
current data, we believe that this discrepancy can be ex-
plained by the presence or absence of basement membrane,
since the absence of ECM results in a significant increase in
TGF-BI expression. There is considerable evidence that
TGF-/3 is able to induce the synthesis of various ECM pro-
teins (Massagur 1990; Roberts and Sporn, 1990; Kahari et
al., 1991). Also manunm~ cells cultured on plastic substra-
tum produce high levels of ECM proteins (Streuli and Bis-
sell, 1990), which may be a result of the high level of TGF-BI
expression. Thus, the absence of ECM induces TGF-/31 ex-
pression which in turn stimulates the synthesis of ECM pro-
teins in an autocrine fashion. We have now documented that,
once the ECM is deposited, TGF-B1 gene expression is
strongly downregulated. This may explain the base-line ex-
pression of ECM proteins we observe when cells synthesize
an endogenous basement membrane (Streuli and Bissell,
1990). It is thus possible that some type of ECM-induced
negative feedback loop regulates TGF-/31 expression.
Our results also indicate that caution has to be exercised
when evaluating the physiological relevance of TGF-/S'I syn-
thesis using cells cultured in the absence of basement mem-
brane, i.e., the commonly used cell culture conditions. The
expression of various ECM proteins and proteins involved in
ceU-matrix interactions, such as proteases, protease inhibi-
tors and integrins, is strongly modulated by TGF-/3. It is thus
conceivable that their expression levels may be dependent
upon the autocrine control of endogenously synthesized
TGF-/3, as we have recently demonstrated for a TGF-/31 over-
expressing tumor cell line (Arrick et al., 1992). Thus, the
enhanced TGF-/31 synthesis in the absence of ECM may also
affect the levels of synthesis of these proteins under autocrine
control of TGF-B, resulting in expression levels which are
not representative of the physiological conditions in vivo.
We have shown that the presence of EHS matrix, a base-
ment membrane preparation, and an endogenously synthe-
sized basement membrane keep the TGF-/31 in an uninduced
state, while a type I collagen matrix by itself does not down-
regulate TGF-/31 expression. Which component of the base-
ment membrane affects TGF-/31 expression is unknown at
present. It is, however, likely that the responsible factor must
be deposited in an insoluble form in the basement mem-
brane, since the conditioned medium from cells on plastic,
which contains a variety of soluble ECM proteins, did not
affect the TGF-B1 promoter activity. However, if a simple
negative feedback mechanism exists, it is likely that an ECM
component whose expression is under control of TGF-B is
responsible for the downregulation of TGF-~I expression.
The mechanism by which the cell recognizes the absence or
the presence of ECM is presumably based on signaling
through cell surface receptors for the appropriate matrix
components. Occupation of the receptors would then result
in a repressed transcriptional activity, whereas the TGF-/31
transcription would be induced in the absence of the ligand.
It has been established already that ECM components can
influence gene expression. Our previous work using the
same mammary epithelial cells has shown that cell-ECM in-
teractions in the absence of cell-cell interaction and polarity
is sufficient to induce expression of B-casein and that this in-
duction of gene expression requires a signal transfer medi-
ated through integrins (Streuli et al., 1991). In addition,
interaction between the cells and the ECM, and more ex-
plicitly integrin-mediated interaction with fibronectin frag-
ments, induces collagenase and stromelysin gene expression
(Werb et al., 1989). Our current study indicates that base-
ment membrane has the ability to modulate the transcription
of TGF-/~I, a factor that determines the cell-matrix interac-
tion. On the basis of this finding, it would be no surprise if
the expression of other regulators of cell-matrix interac-
tions, such as the bone morphogenetic proteins or other
members of the TGF-B superfamily, also would be modu-
lated by cell-ECM interactions.
The transcription and mRNA levels of TGF-/31, -/32, and
-B3 are differentially regulated and this is presumably related
to their different promoter structures. Since the expression
of TGF-~, and not TGF-/~2, is strongly affected by ECM,
it is likely that this specific modulation is related to the pres-
ence of a TGF-~I specific promoter element that is absent in
the TGF-/$2 promoter, e.g., an AP-1 site (Malipiero et al.,
1990). We have recently demonstrated the presence of an
ECM-responsive enhancer in the 5' region of the B casein
promoter (Schmidhauser et al., 1992). The best inducer of
TGF-/32 and -/33 currently known is retinoic acid (Glick et
al., 1989). Whereas TGF-/32 mRNA levels can be enhanced
up to 10-30-fold by retinoic acid, the level of TGF-B1 mRNA
is increased only 1.5-3-fold. These studies, however, were
performed using cells cultured on plastic substratum, result-
ing in an already high level of TGF-/~I transcription in un-
treated cultures; it is therefore possible that a greater differ-
ence could be observed if cells were cultured on ECM. The
only dramatic induction of TGF-/~I mRNA expression
reported so far has been the auto-induction of TGF-/31 by
TGF-/$ itself (Van Obberghen-Schilling et al., 1988; Bascom
et al., 1989). We now show that a dramatic difference in
TGF-/$ expression can be achieved also by the absence or
presence of ECM. These findings suggest that both the levels
of active TGF-/~I, as weft as the cell-ECM interactions may
be major determinants of the rates of TGF-~I synthesis by
epithelial and other cells in vivo.
The regulation of TGF-B1 and not TGF-/~2 expression by
ECM suggests that these two TGF-/3 forms have different
normal physiological roles in vivo. The regulation of TGF-
B2 by retinoids (and steroids) suggests that TGF-B2 may play
a role in various differentiation and morphogenetic processes
during development. On the other hand, TGF-/31 expression
in vivo may be important in the deposition and establishment
of basement membrane and extracellular matrix. In addi-
The Journal of Cell Biology, Volume 120, 1993 258
tion, auto-induction of TGF-~I expression suggests a role in
localized amplification of the TGF-/~I effects. Thus, expres-
sion of TGF-/31 would be upregulated whenever the interac-
tion of the cells with the extracellular matrix or basement
membrane is disturbed, e.g., following wounding (Sieweke
et al., 1990).
In conclusion, our data show that the expression of the
TGF-/31 gene, but not the TGF-/T2 gene, is strongly affected
by the absence or presence of ECM and suggests a possible
feedback loop in vivo, whereby the loss of of the proper cel-
lular interactions with the ECM induces "I'GF-g/1 expression.
TGF-/31 then in turn induces the synthesis of several ECM
proteins. When conditions are right for the ECM compo-
nents to come together to form a basement membrane, TGF-
B1 expression is suppressed. Further studies will be needed
to further define this regulation at the molecular level and to
understand its relevance in vivo.
We thank Drs. Caroline Damsky and Zena Werb (University of California,
San Francisco) for their critical review of the manuscript and Dr. Scan
Lawler for his help in some experiments.
This work was supported in part by Genentech Inc. and in part by the
Health Effects Research Division, Office of Health and Environmental re-
search, U.S. Department of Energy (contract DE-AC-03-76 SF0{X)98) and
a gift for research from Monsanto Co. to M. J. Bissen. C. H. Streuli was
a Fogarty International Research Fellow and C. Schmidhauser is a fellow
of the Schweizerische Nationalfonds.
Received for publication 22 January 1992 and in revised form 24 July 1992.
References
Arrick, B. A., M. Korc, and R. Derynck. 1990. Differential regulation of ex-
pression of three transforming growth factor t species in hnman breast can-
cer cell lines by estrndiol. Cancer Res. 50:299-303.
Arrick, B. A., A. R. Lopez, F. Elfman, R. Zoner, C. H. Damsky, and R.
Derynck. 1992. Altered metabolic and adhesive properties and enhanced
Uimorigenesis associated with increased expression of transforming growth
factor-tl. J. Cell Biol. 118:715-726.
Aggeler, J., J. Ward, L. M. Blackie, M. H. Bercellos-Hotf, C. H. Streuli, and
M. J. Bissell. 1991. Cytoditferentiation of mouse mammary epithelial cells
cultured on a reconstituted basement membrane reveals striking similarities
to development in vivo. J. Cell Sci. 99:407-417.
Barcellos-Hoff, M. H., J. Aggeler, T. G. Ram, and M. J. Bissell. 1989. Func-
tional differentiation and alveolar morpbogenesis of primary mammary cul-
tures on reconstituted basement membrane. Development (Camb.). 105:
223-235.
Bascom, C. C., J. R. Wolfshohl, R. J. Coffey, L. Madisen, N. R. Webb, A. R.
Purehio, R. Derynck, and H. L. Moses. 1989. Complex regulation of trans-
forming growth factor tl, j~2, and t3 mRNA expression in mouse fibroblasts
and keratinocytes by transforming growth factor tl and ~2. Mol. Cell. Biol.
9:5508-5515.
Chert, L. H., and M. J. Bissell. 1989. A novel regulatory mechanism for whey
acidic protein geue expression. Cell Regul. 1:45-54.
Derynck, R., J. A. Jarrett, E. Y. Chert, and D. V. Gueddel. 1986. The murine
transforming growth factor-t precursor. J. Biol. Chem. 261:4377-4379.
Derynck, R., L. Rhee, E. Y. Cben, and A. Van Tilburg. 1987. Intron-exon
structure of the human transforming growth factor-t precursor gene. Nucleic
Acids Res. 15:3188-3189.
Derynck, R., P. B. Lindquist, A. Lee, D. Wen, J. Tamm~ J. L. Graycar, L.
Rhee, A. J. Mason, D. A. Miller, R. J. Coffey, H. L. Moses, and E. Y.
Cheo. 1988. A new type of transforming growth factor-B, TGF-03. EMBO
(Eur. Mol. Biol. Organ.) J. 7:3737-3743.
Emerman, J. T., and M. J. Bissell. 1988. Cultures
of mammary
epithelial cells:
extracellular matrix and functional differentiation. Adv. Cell Cul. 6:137-
159.
Fitzpatrick D. R., F. Denhez, P. Kondaiah, R. J. Akhurst. 1990. Differential
expression of TGF beta isoforms in muriue palatogenesis. Development
(Camb.). 109:585-595.
Gatherer, D., P. ten Dijke, D. T. Baird, and R. J. Akhurst. 1990. Expression
of TGF-t isoforms during first trimester human embryogenesis. Develop-
ment. 110:445--460.
Glick, A. B., K. C. Flanders, D. Danielpour, S. H. Yuspa, and M. B. Sporn.
1989. Retinoic acid induces TGF-ff2 in cultured keratinocytes and mouse
epidermis. Cell Regul. 1:87-97.
Geiser, A. G., S. J. Kim, A. B. Roberts, and M. B. Sporn. 1991. Characteriza-
tion of the mouse TGF-tl promoter and activation by the Ha-ras oncogene.
Mol. Cell. Biol. 11:84-92.
Heine, U. I., E. F. Munoz, K. C. Flanders, L. R. Ellingsworth, H. Y. Lam,
N. L. Thompson, A. B. Roberts, and M. B. Sporn. 1987. Role of transform-
ing growth factor-beta in the development of the mouse embryo. J. Cell Biol.
105:2861-2876.
Heine, U. I., E. F. Munoz, K. C. Flanders, A. B. Roberts, and M. B. Sporn.
1990. Colocalization of TGF-beta 1 and collagen I and HI, fibronectin and
glycosaminoglycans during lung branching morphogenesis. Development
(Camb.). 109:29-36.
Jakobovits, E. B., U. Schlokat, J. L. Vannice, R. Derynck, and A. D. Levin-
son. 1988. The human TGF-c~ promoter directs transcription initiation from
a single site in the absence of a TATA sequence. Mol. Cell. BioL 8:5549-
5554.
Kahari, V. M., H. Larjava, and J. Uitto. 1991. Differential regulation of ex-
tracellular matrix proteoglycan (PG) gene expression. Transforming growth
factor-beta 1 up-regulates biglycan (PGI) and versican (large fibroblast PG)
but down-regulates decorin (PGII) mRNA levels in human fibroblasts in cul-
tore. J. BioL Chem. 266:10608-10615.
Kleinman, H. K., M. L. MeGarvey, J. R. Hassell, V. L. Star, F. B. Cannon,
G. W. Laurie, and G. R. Martin. 1986. Basement membrane complexes with
biological activity. Biochemistry. 25:312-318.
Kim, S.-J., A. Glick, M. B. Sporn, and A. B. Roberts. 1989. Characterization
of the promoter region of the human TGF-tl gene. J. Biol. Chem. 264:
402-408.
Kozak, M. 1989. The scanning model for translation: an update. J. Cell Biol.
108:229-241.
Lee, E.Y.-H., G. Parry, and M. J. Bissell. 1984. Modulation of secreted pro-
teins of mouse mammary epithelial cells by the collagenous substrata. J. Cell
Biol. 98:146-155.
Lee, E.Y.-H., W.-H. Lee, C. S. Kaetzel, G. Parry, and M. J. Bissell. 1985.
Interaction of mouse mammary epithelial cells with collagen substrata:
Regulation of casein gene expression and secretion. Proc. Natl. Acad. Sci.
USA. 82:1419-1493.
Li, M. L., J. Aggeler, D. A. Farson, C. Hatier, J. Hassell, and M. J. Bissell.
1987. Influence of a reconstituted basement membrane and its components
on casein gene expression and secretion in mouse mammary epithelial cells.
Proc. Natl. Acad. Sci. USA. 84:136-140.
Malipiero, U., M. Holler, U. Weruer, and A. Fontana. 1990. Sequence analy-
sis of the promoter of the glioblastoma-derived T-cell suppressor factor/
transforming growth factor (TGF')-/~2 gene reveals striking differences to the
TGF-tl and -t3 genes. Biochera. Biophys. Res. Comm. 171:1145-1151.
Meager, A. 1991. Assays for transforming growth factor-t. J. Immunol.
Methods. 141:1-14.
Massagu6, J. 1990. The transforming growth factor-t family. Annu. Rev. Cell
Biol. 6:597-641.
Millan, F. A., F. Denhez, P. Kondaiah, and R. J. Akhurst. 1991. Embryonic
gene expression patterns of TGF beta 1, beta 2 and beta 3 suggest different
developmental functions in vivo. Development (Camb. ). 111:131 - 143.
Miller, D. A., A. Lee, R. W. Pelion, E. Y. Chert, H. L. Moses, and R.
Derynck. 1989. Murine transforming growth factor-~2 cDNA sequence and
expression in adult tissues and embryos, biol. Endocrinol. 3:1108-1114.
Pelton, R. W., M. E. Dickinson, H. L. Moses, and B. L. M. Hogan. 1990a.
In situ hybridization analysis of TGF-t3 RNA expression during mouse
development: comparative studies with TGF-tl and -if2. Development
(Camb.). 110:609-620.
Pelton, R. W., B. L. M. Hogan, D. A. Miller, and H. L. Moses. 1990b.
Differential expression of TGF-t 1, -/~2 and -t3 during murine palate forma-
tion. Dev. Biol. 141:456-460.33.
Pelton, R. W., B. Saxena, M. Jones, H. L. Moses, and L. I. Gold. 1991. Im-
manohistochemical localization of TGF-tl, TGF-/Y2, and TGF-/~3 in the
mouse embryo: expression patterns suggest multiple roles during embryonic
development. J. Cell Biol. 115:1091-1105.
Roberts, A. B., and M. B. Sporn. 1990. The transforming growth factor-betas.
In Peptide Growth Factors and Their Receptors. M. B. Sporn and A. B.
Roberts, editors. Springer-Verlag, Berlin. 419-472.
Robinson, S. D., G. B. Sllberstein, A. B. Roberts, K. C. Flanders, and C. W.
Daniel. 1991. Regulated expression and growth inhibitory effects of trans-
forming growth factor-t isoforms in mouse mammary gland development.
Development (Camb.). 113:867-878.
Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning: a
laboratory manual. Cold Spring Harbor Laboratory Press, Cold Spring
Harbor, N'Y.
Schmid, P., D. Cox, G. Bilbe, R. Maier, and G. K. McMaster. 1991. Differen-
tial expression of TGF beta 1, beta 2 and beta 3 genes during mouse embryo-
genesis. Development (Camb.). 111:117-130.
Schmidhanser, C., M. J. Bissell, C. A. Myers, and G. F. Casperson. 1990.
Extrnceilular matrix and hormones transcriptionally regulate bovine B-casein
5' sequences in stably transfected mouse mammary cells. Proc. Natl. Acad.
Sci. USA. 87:9118-9122.
Schmidhanser, C., G. F. Casperson, C. A. Myers, K. T. Sanzo, S. Bolten, and
M. J. Bissell. 1992. A novel transcriptional enhancer is involved in pro-
la~tin- a~l r matrix-dependent regulation of B-casein. Mol. Biol.
Cell. 3:699-709.
Sieweke, M. H., N. L. Thompson, M. B. Sporn, and M. J. Bissell. 1990. Medi-
Streuli et al. Extracellular Matrix Regulates Expression of the TGF-Ol Gene 259
ation of wound related Rous sarcoma virus tumorigenesis by TGF-/5. Science
(Wash. DC). 248:1656-1660.
Streuli, C. H., and M. J. Bissell. 1990. Expression ofextracellular matrix com-
ponents is regulated by substratum. J. Cell Biol. 110:1405-1415.
Streuli, C. H., N. Bailey, and M. J. Bissell. 1991. Control of mammary epithe-
lial differentiation: basement membrane induces tissue-specific expression in
the absence of cell-cell interaction and morphological polarity. J. Cell Biol.
115:1383-1395.
Thompson, N. L., K. C. Flanders, J. M. Smith, L. R. Ellingsworth, A. B.
Roberts, M. B. Sporn. 1989. Expression of transforming growth factor-
beta 1 in specific cells and tissues of adult and neonatal mice. J. Cell Biol.
108:661-669.
Van Obberghen-Schilling, E., N. S. Roche, K. C. Flanders, M. B. Sporn, and
A. B. Roberts. 1988. Transforming growth factor-/31 positively regulates its
own expression in normal and transformed cells. J. Biol. Chem. 263:7741-
7746.
Werb, Z., P. M. Tremble, O. Behrendtsen, E. Crowley, and C. H. Damsky.
1989. Signal transduction through the fibronectin receptor induces collagen-
ase and stromelysin gene expression. J. Cell Biol. 109:877-889.
Wilcox, J. N., and R. Derynck. 1988. Developmental expression of transform-
ing growth factors-c~ and -/~ in mouse fetus, biol. Cell. Biol. 8:3415-3422.
Zoiler, M. J., and M. Smith. 1984. Oligonucleotide-directed mutagenesis: a
simple method using two oligonucleotide primers and a single-stranded DNA
template. DNA (NF). 3:479-488.
The Journal of Cell Biology, Volume 120, 1993 260
