(Received for publication, December 24, 1996, and in revised form, January 21, 1997)
From the Department of Surgery, Division of Urology,
and the University of Michigan Comprehensive Cancer Center, University
of Michigan, Ann Arbor, Michigan 48109, the
Department of Medicine and Cell Biology,
Washington University School of Medicine, St. Louis, Missouri 63110, and the ** Department of Pathology, Washington University School of
Medicine, St. Louis, Missouri 63110
Epithelial cells are dependent upon
adhesion to extracellular matrix for survival. We show that loss of
1 integrin receptor contact with extracellular matrix signals the
inhibition of G1 cyclin-dependent kinase
activity. This loss of cyclin-dependent kinase activity
leads to accumulation of the hypophosphorylated (active) form of
the retinoblastoma tumor suppressor protein (Rb). We present
evidence that in epithelial cells deprived of matrix contact, the
growth suppression signal elicited by hypophosphorylated Rb opposes
stimulatory signals from serum growth factors, leading to a cell cycle
conflict that triggers apoptosis. This apoptotic pathway is modulated
by Bcl-2 through a novel mechanism that regulates Rb phosphorylation.
We present evidence that the Rb-dependent apoptotic pathway
functions in vivo in the apoptosis of the prostate glandular epithelium following castration.
Adhesion of epithelial and endothelial cells to an extracellular matrix (ECM)1 transmits signals that suppress apoptosis (for review, see Refs. 1 and 2). In response to decreases in hormone levels in the prostate following castration or in the mammary gland following weaning, proteases are released that disrupt epithelial cell-ECM contacts (3, 4). This loss of ECM contact is thought to trigger apoptosis in the epithelial cells.
Integrins are heterodimeric cell surface receptors that mediate cell
adhesion to the ECM (2). Ligation of 1 integrins to ECM not only
provides an adhesive interaction, the cytoplasmic domain transmits
signals that regulate growth and prevent apoptosis (5-7). In the
absence of ECM contacts, adhesion-dependent cells arrest in
the G1 phase of the cell cycle due, in part, to the loss of
activity of G1-specific cyclin-dependent
kinases (cdks), cyclin E-cdk2 and cyclin D-cdk4/6 (8, 9). Growth arrest also correlates with a dephosphorylation and activation of the retinoblastoma tumor suppressor protein (Rb), a transcriptional repressor that controls transition of cells from G1 to S
phase by regulating expression of cell cycle genes (10). The activity of Rb is modulated by G1-specific cdks that phosphorylate
and inactivate Rb in late G1.
Here, we report that in epithelial cells, the disruption of 1
integrin contact with ECM triggers a loss of G1 kinase
activity, resulting in the dephosphorylation (activation) of Rb. We
demonstrate that this activation of Rb and its interaction with E2F are
required for induction of apoptosis. Additionally, we present evidence that the apoptotic regulator, Bcl-2, can inhibit this apoptosis by
regulating the activity of Rb. Finally, we provide evidence that
apoptosis of prostate epithelium following castration occurs through
this Rb-E2F-dependent pathway.
LNCaP cells, Bcl-2-expressing LNCaP cells, and the DU145
cell line were cultured as described (11, 12). For suspension assays,
cells were grown to 75% confluence and detached by trypsinization or
by treatment with a blocking anti-1 integrin antibody (LIA 1/2)
followed by culture on polyHEMA (15 µg/ml) (5). Western blots were
done as described (13) using anti-Bcl-2 (Dako M887), anti-Rb (Oncogene
Science OP28-A; Pharmingen 14001A), anti-p16 (Calbiochem Ab-1),
anti-p27 (Calbiochem Ab-2), anti-Bax (Pharmingen 13666E and 13686E),
anti-E1a (M. Green, St. Louis University, St. Louis, MO), anti-p21
(Pharmingen 15091A), anti-E2F-1 (KH-20 from K. Helin, Danish Cancer
Society, Copenhagen, Denmark), anti-Cyclin D1 (Pharmingen 14726E), and
anti-Cyclin E (Santa Cruz C-19). For Rb overexpression assays, 2 µg
each of the CD20 expression vector (pCMVCD20) (14) and the Rb or Rb-706
expression vector (15) was transfected into LNCaP cells. Cells were
immunostained with rhodamine-conjugated anti-CD20 (Becton-Dickinson),
and TUNEL (TdT-mediated dUTP-biotin nick end labeling) activity was
detected using fluorescein isothiocyanate-conjugated avidin and
biotinylated dUTP. At least 200 CD20(+) cells were examined for TUNEL
activity in each assay in three independent experiments. Indicated
amounts of pDN-E2F-1 were cotransfected with 1 µg of the pTA-E2F-CAT
reporter (16), and CAT activity was determined as described (13). For
stable clones, pDN-E2F-1, pCMV-E1a-12S, or pCMV-E1a-12S-928 was
transfected into LNCaP cells, and clones were isolated in 500 µg/ml
of G418.
250-g male Sprague-Dawley rats were castrated, and at the indicated times the ventral prostate was excised, fixed in 10% formalin and embedded in paraffin. Immunohistochemical studies were performed on 5-micron sections overlaid with rabbit affinity-purified polyclonal Rb antibody (SC-50, Santa Cruz). The avidin-biotin-peroxidase method was performed (VectaStain ABC kit; Vector Laboratories), and slides were developed by immersion in 3,3-diaminobenzidine tetrachloride (0.5 mg/ml), containing 0.003% hydrogen peroxide followed by a hematoxyin counterstain. Nonimmune rabbit serum was used as a negative control. Northern blotting for Rb mRNA was done as described (17).
We found that primary cultures of human breast and
tracheal epithelial cells, human umbilical vein endothelial cells, mink lung epithelial cells, and an Rb(+) human prostate epithelial cell
line, LNCaP (11), all undergo apoptosis after they are detached from
matrix, either by trypsinization or by the addition of a blocking
anti-1 integrin adhesion receptor antibody (Fig. 1A and results not shown). Apoptosis was
confirmed by trypan blue exclusion assay, a positive TUNEL assay,
laddering of chromosomal DNA, and nuclear condensation as determined by
electron microscopy. These results are consistent with previous studies
demonstrating that epithelial and endothelial cells undergo apoptosis
when they lose contact with extracellular matrix (1, 5, 6).
Binding of 1 integrin receptors to extracellular matrix ligands not
only provides an adhesive interaction, the
1 cytoplasmic tail
interacts with cytoskeletal components and protein kinases and
transmits signals that regulate cell cycle progression (7, 9). The
transition from G1 to S phase is a key regulatory
checkpoint in the cell cycle, which is controlled by G1
cdks. Therefore, we examined expression of proteins that control
G1 cdk activity. A rapid decrease in the level of cyclin D1
and cyclin E was observed (within 4 h) in LNCaP cells deprived of
matrix contact (Fig. 1B). These cyclin regulatory subunits
form active complexes with cdk4/6 and cdk2, respectively, that can
hyperphosphorylate and inactivate Rb (10). We also found that the level
of the cdk inhibitors p21 (SDI, WAF1, and CIP1) and p27 (KIP1), which
block the activity of G1 cdks (10), increased within
12 h following detachment from matrix (Fig. 1B). In
contrast, there was little change in the level of the cdk inhibitor p16
(INK4B and MTS1) (Fig. 1B). The decrease in cyclins coupled
with the increase in p21 and p27 suggested a loss of G1 cdk
activity in cells deprived of matrix contact. Accordingly, we found
that hypophosphorylated (activated) Rb accumulated when LNCaP cells (as
well as the other epithelial and endothelial cells) were deprived of
matrix contact (Fig. 1B and results not shown). This
accumulation of hypophosphorylated Rb followed the rapid decrease in
G1 cyclins and the increase in cdk inhibitors. As
hypophosphorylated Rb accumulated, apoptotic cells were detected. When
LNCaP cells were allowed to reattach, the changes in cyclin and cdk
inhibitor expression and the accumulation of hypophosphorylated Rb were
reversed (Fig. 1B and results not shown).
The importance of Rb in this process was suggested by the finding that epithelial cells with a mutated or inactive form of Rb, such as the DU145 prostate epithelial cell line, failed to undergo apoptosis when deprived of matrix contact (Fig. 1A and results not shown). As with Rb(+) LNCaP cells, DU145 cells appeared to lose G1 cdk activity when detached from matrix (results not shown), suggesting that the signal to decrease cdk activity is sent but these cells do not undergo apoptosis, perhaps because they lack Rb.
Our finding that accumulation of hypophosphorylated Rb may be important
for apoptosis in epithelial cells deprived of matrix contact was
surprising given that hypophosphorylated Rb has been implicated in the
inhibition of apoptosis (18, 19). To test whether Rb as a downstream
target of integrin signaling (8, 9) promotes apoptosis, LNCaP cells
were transfected with an expression vector for Rb (overexpression of Rb
leads to accumulation of the hypophosphorylated protein) or a control
mutant Rb (Rb-706) (15). Cells were cotransfected with an expression
vector for the B cell marker CD20, which was used to identify
transfected cells as described previously (14). Overexpression of wild
type Rb, but not mutant Rb, resulted in apoptosis of LNCaP cells (Fig. 2A). The fact that Rb induced apoptosis in
adherent epithelial cells supports the idea that Rb is a downstream
target of 1 integrin signaling such that increasing its level of
expression and activity bypasses signals arising from cell adhesion or
detachment from matrix.
We then reasoned that blocking endogenous Rb function may prevent apoptosis when epithelial cells are detached from matrix. Therefore, we derived LNCaP clones that stably expressed the adenovirus E1a protein, which binds to Rb and inhibits its function (20). E1a prevented apoptosis when cells were detached from matrix (Fig. 2B). In fact, the E1a-expressing cells continued to proliferate in suspension (results not shown). Because E1a has cellular targets in addition to Rb, we derived LNCaP cell lines stably expressing an E1a construct containing a point mutation at nucleotide 928, which selectively blocks interaction with Rb without disrupting other E1a interactions or functions (20). E1a-928 did not block apoptosis, suggesting that the anti-apoptotic activity of E1a in epithelial cells is solely the result of inhibiting Rb function. These results are then a second way of demonstrating a role for Rb in inducing apoptosis of epithelial cells that lose matrix contact.
Rb is a transcriptional repressor that is targeted to cell cycle genes through interaction with the E2F family of transcription factors (13, 16). Once Rb is concentrated at the promoter through interaction with E2F, it binds surrounding transcription factors, preventing their interaction with the basal transcription complex, thereby blocking transcription. Previously, it has been demonstrated that deletion of the E2F-1 gene, which prevents the recruitment of Rb to cell cycle genes, leads to a loss of cell cycle control and ultimately to tumor formation (21). LNCaP cell lines were created that stably express a dominant-negative form of E2F-1 (DN-E2F) that can bind DNA but is unable to recruit Rb to the promoter (22). DN-E2F blocked E2F function in transfection assays, and stable expression of DN-E2F prevented apoptosis in epithelial cells detached from matrix (Fig. 2B and results not shown). These experiments then provide a third method of demonstrating that Rb activity is required for inducing apoptosis in epithelial cells that are deprived of matrix contact. Further, the results indicate that a specific interaction between Rb and E2F is required to induce this apoptotic pathway.
Apoptotic Regulators of the Bcl-2 Family Can Determine Sensitivity to Apoptosis Induced by Matrix DeprivationOur observations along
with those of other laboratories suggest that accumulation of
hypophosphorylated Rb is a general response of
adhesion-dependent cells to the loss of matrix contact. Why do some cells respond to this growth suppression signal by undergoing apoptosis, whereas others do not? It has been shown that fibroblasts deprived of matrix contact accumulate hypophosphorylated Rb but in
contrast to epithelial cells do not undergo apoptosis (8, 9). Apoptotic
pathways can be regulated by a family of proteins that act as
pro-apoptotic (e.g. Bax) and anti-apoptotic (e.g. Bcl-2) modulators (18). We observed an inverse correlation between Bcl-2 expression and sensitivity to apoptosis (Fig. 3,
A-C). Little or no Bcl-2 was detected by Western blot in
cultured LNCaP cells that undergo apoptosis when deprived of matrix
contact (Fig. 3B). Likewise, Bcl-2 is not evident in the
glandular prostate epithelium, which undergoes apoptosis following
castration (23). However, Bcl-2 levels are high in the
castration-resistant basal epithelium and stromal fibroblasts of the
prostate gland (stromal fibroblasts from other tissues also express
Bcl-2) (23). Additionally, primary cultures of prostate stromal
fibroblasts also expressed high levels of Bcl-2 (Fig.
3C).
The ratio of pro- and anti-death Bcl-2 family members is thought to determine whether cells are susceptible to apoptosis (18). In contrast to Bcl-2, the pro-apoptotic protein Bax was expressed in LNCaP cells, and its level did not change when cells were detached from matrix (Fig. 3B). Therefore, the high Bax to Bcl-2 ratio that we observe in the epithelial cells is consistent with susceptibility to apoptosis. We decreased this ratio by using LNCaP clones that stably express Bcl-2 (Fig. 3C) and found that apoptosis was inhibited when Bcl-2-expressing cells were deprived of matrix contact (Fig. 3A). The ratio of pro- and anti-apoptotic proteins may then be one mechanism that determines whether cells will respond to the loss of matrix contact by growth suppression or apoptosis.
How might Bcl-2 inhibit apoptosis in epithelial cells that lose matrix contact? Surprisingly, we found that expression of Bcl-2 in LNCaP cells prevented the accumulation of hypophosphorylated Rb that occurs when epithelial cells are deprived of matrix contact (Fig. 3D). Although Bcl-2 did not affect expression of cdk inhibitors in adherent cells, the levels of p21 and p27 failed to increase when the Bcl-2-expressing cells were detached from matrix (Fig. 3E and results not shown), providing an explanation for why Rb continued to be phosphorylated. We also examined the ability of Bcl-2 to induce hyperphosphorylation of Rb in another set of experiments. When the concentration of serum growth factors is decreased, hypophosphorylated Rb normally accumulates, and the cell growth arrest (10). We found that when parental LNCaP cells were cultured in reduced growth factor medium for 24 h, hypophosphorylated Rb did indeed accumulate and cell growth was arrested; however, under the same conditions, Rb continued to be hyperphosphorylated in the Bcl-2-expressing LNCaP cells, and the cells proliferated (results not shown). These experiments are then a second method of demonstrating that Bcl-2 promotes hyperphosphorylation of Rb in the epithelial cells. In support of our findings, it has been reported that Bcl-2 can also promote accumulation of hyperphosphorylated Rb in B cells (24).
In contrast to the Bcl-2-expressing LNCaP cells, expression of Bcl-2 in
fibroblasts does not appear to prevent accumulation of
hypophosphorylated Rb when cells are detached from matrix (Refs. 8 and
9 and results not shown). Bcl-2 can act at multiple points in the
apoptotic pathway; in addition to inducing phosphorylation of Rb, Bcl-2
can regulate the activity of transcription factors (i.e.
NF-B and NFAT), regulate an antioxidant pathway, and inhibit interleukin 1
-converting enzyme-like proteases (caspases) that ultimately execute the apoptotic process in many cells (18). Therefore,
Bcl-2 may block apoptosis of fibroblasts that are deprived of matrix
contact through one of these other mechanisms. Alternatively, the
intracellular environment in fibroblasts may be such that growth
suppression in cells deprived of matrix contact does not initiate a
conflicting signal that triggers apoptosis.
To
determine whether an Rb-dependent apoptotic pathway may be
responsible for apoptosis of epithelial cells in vivo, we
examined expression and activation of Rb in the rat ventral prostate
following castration. The glandular epithelium constitutes
approximately 85% of the cells in the ventral prostate, and nearly
80% of these cells undergo apoptosis within 7 days following
castration (25). In response to decreasing levels of androgen,
proteases that digest the epithelial cell extracellular matrix are
released (4). These proteases are thought to trigger apoptosis at least
in part by digesting the extracellular matrix substrate for glandular epithelial cells (1). We found an increase in apoptotic cells 2 days
after castration, and a peak of apoptotic cells was seen by day 4 (Fig.
4 and results not shown). The apoptotic process was
essentially complete by day 7. No evidence of apoptosis was seen in
stromal fibroblasts or basal epithelial cells. Expression of Rb in the
glandular epithelium preceded this apoptosis. Approximately 2% of the
prostate glandular epithelium showed expression of Rb in the nucleus
before castration. The number of Rb(+) cells increased to 10% at day 1 following castration, 60% at day 2, and 78% at day 4. This increase
in Rb was paralleled by an increase in Rb mRNA (results not shown).
By day 7 following castration when apoptosis had diminished, Rb was
only evident in approximately 2% of the remaining glandular epithelial
cells (similar to the pattern before castration), and the expression of
Rb mRNA had returned to base line (results not shown). Similar to
our results in the LNCaP prostate epithelial cell line that had been
deprived of matrix contact (Fig. 1B), it has been reported
that the levels of cyclin D1 and cyclin E decrease in the prostate
following castration (26), and we found that the level of p27 increases
(results not shown). Together, these results suggest that
G1 cdk activity diminishes following castration and that
the Rb that accumulates in the glandular epithelium following
castration is likely to be in the hypophosphorylated form. A Western
blot of prostate tissue 2 days after castration revealed that Rb was
indeed exclusively in the hypophosphorylated form (results not shown).
The correlation between expression of hypophosphorylated Rb and
apoptosis following castration together with our results in cultured
cells suggests that accumulation of hypophosphorylated Rb may be
important for signaling a common apoptotic pathway in epithelial cells
in vitro and in vivo.
Apoptosis is thought to be a default pathway in cells where opposing or conflicting cell proliferation signals arise. For example in c-myc-induced apoptosis, overexpression of c-myc can initiate such a conflict by promoting entry of cells into the cell cycle when other essential cell cycle growth factors are not present (19), and Rb-mediated growth arrest would be expected to alleviate such a conflict. In contrast, Rb-mediated growth suppression appears to precipitate a conflict in epithelial and endothelial cells that lose matrix contact. This conflict could arise when the Rb growth suppression signal opposes mitogenic signals from serum growth factors that promote progression through the cell cycle. In support of this possibility we found that epithelial cells that are no longer growth factor-responsive due to either serum starvation or growth arrest by contact inhibition are unable to undergo Rb-dependent apoptosis (Fig. 1A and results not shown).
Although Rb has been shown to inhibit multiple
p53-dependent apoptotic pathways (18, 19), accumulation of
hypophosphorylated Rb and growth arrest in G1 occurs in
multiple p53-independent pathways (24, 28). One such p53-independent
pathway is induced by release of the lipid second messenger ceramide,
which causes accumulation of hypophosphorylated Rb, G1
arrest, and apoptosis (29, 30). Apoptosis occurs normally in the
prostate luminal epithelium of p53(/
) mice following castration,
indicating that this process is p53-independent (27). Therefore,
accumulation of hypophosphorylated Rb and G1 arrest appear
to have opposing roles in signaling different apoptotic pathways
(which appears to be determined at least in part by whether there is a
role of p53 in the process). It is likely that the role of Rb is to
simply arrest the growth of cells in G1, and it is the
intracellular and extracellular environment that dictates how a cell
responds to this growth suppressive signal.
We thank V. Dixit, S. Korsmeyer, S. Weintraub, and S. Weiss for critical comments and R. Buttyan, S. Ethier, S. Frisch, M. Green, M. Holtzman, K. Helin, W. Kaelin, G. Nunez, K. Pienta, and F. Sanchez for antibodies, plasmids, and cell lines.