(Received for publication, July 16, 1996, and in revised form, December 20, 1996)
From the Cancer Research Program, Garvan Institute of Medical Research, St. Vincent's Hospital, Sydney, New South Wales 2010, Australia
Estrogens induce cell proliferation
in target tissues by stimulating progression through
G1 phase of the cell cycle, but the underlying
molecular targets remain undefined. To determine the role of the
cyclin/cyclin-dependent kinase (CDK)/retinoblastoma protein
(pRB) pathway in this response we treated MCF-7 breast cancer cells
with the pure estrogen antagonist ICI 182780 to inhibit estrogen-induced gene expression and induce G1 phase
arrest. Subsequent treatment with 17-estradiol resulted in the
synchronous entry of cells into S phase commencing at 12 h. The
proportion of cells in S phase reached a maximum of 60% at 21-24 h.
Cells subsequently completed mitosis and entered a second
semisynchronous round of replication. Entry into S phase was preceded
by increased activity of both Cdk4 and cyclin E-Cdk2 and
hyperphosphorylation of pRB, all within the first 3-6 h of estradiol
treatment. The increase in Cdk4 activity was accompanied by increases
in cyclin D1 mRNA and protein, indicating that an initiating event
in the activation of Cdk4 was increased cyclin D1 gene expression. In
contrast, the levels of Cdk2 and the CDK inhibitors p21
(WAF1/CIP1/SDI1) and p27 (KIP1) in total cell lysates and in cyclin E
immunoprecipitates were unaltered at these early time points. However,
an inhibitory activity was present in antiestrogen-pretreated cell
lysates toward recombinant cyclin E-Cdk2 and was relieved by estradiol
treatment. This activity was attributable predominantly to p21. These
apparently conflicting data were resolved by performing gel filtration
chromatography, which revealed that only a minority of cyclin E-Cdk2
complexes were active following estradiol treatment. Active complexes
eluted at a higher molecular weight than inactive complexes, were
relatively deficient in both p21 and p27, and contained Cdk2 with
increased threonine 160 phosphorylation, consistent with a mechanism of activation of cyclin E-Cdk2 involving both reduced CDK inhibitor association and CDK-activating kinase-mediated phosphorylation of Cdk2.
These results provide an explanation for the early activation of both
cyclin D1-Cdk4 and cyclin E-Cdk2 complexes that accompany G1-S phase progression in response to estradiol.
Estrogenic steroids have several major roles in mammalian physiology. These include control of development of the reproductive tract and secondary sex organs, e.g. the mammary gland, regulation of the estrus cycle, control of lactation, and effects on the bone, liver, and cardiovascular systems (1). Estrogens are also intimately linked with the development and progression of a number of human cancers, notably breast cancer. Observations made a century ago showed that ovariectomy in cases of premenopausal breast cancer could lead to tumor regression (2). Subsequent research in experimental models of carcinogen-induced mammary carcinoma revealed that estrogen was essential for both the initiation and progression of the disease (3). More recently, studies with human breast cancer cell lines xenografted to athymic nude mice have demonstrated an absolute requirement for estrogen in tumor formation and growth (4). Such observations provided the rationale for the introduction of antiestrogen therapy, the current treatment of choice for hormone-responsive breast cancer. The subsequent demonstration of the efficacy of the antiestrogen tamoxifen in decreasing both disease progression and the development of contralateral breast cancer (5) provide further compelling evidence for the pivotal role of estrogen-regulated cell proliferation in breast cancer. Despite this, little is known of the molecular basis of cell proliferation control by estrogen.
Early studies on cell proliferation in the rodent uterus and mammary gland in vivo demonstrated that estrogen increases the proportion of cells synthesizing DNA by recruiting noncycling cells into the cell cycle and reducing the duration of G1 phase in already cycling cells (reviewed in Ref. 6). Using breast cancer cells synchronized at the G1/S boundary or at G2/M to test the effect of estrogen added at different stages of the cell cycle, Leung et al. (7) concluded that the sensitive cells were in early G1 phase, immediately following mitosis. These data supported observations that both nonsteroidal (8) and steroidal antiestrogens (9, 10) arrest ER-positive breast cancer cells in G1 phase. More precise mapping of the point of antiestrogen action within G1 phase using cells synchronized by mitotic selection identified a window of sensitivity in early to mid-G1 phase (11). Together these data are compatible with a model whereby estrogens and antiestrogens, through their interactions with the ER, regulate the transcription of genes that control key points in G1 progression.
Candidate genes that might fulfill this role include "immediate early" and "delayed early" genes with established roles in signal transduction and cell cycle control, particularly c-myc, which encodes a transcription factor (c-Myc) with an established role in control of proliferation and apoptosis, genes encoding components of the AP-1 transcription complex, c-fos and c-jun, and the more recently described cyclin-dependent kinase (CDK)1 complexes involved in G1 progression. Previously published studies have focused predominantly on c-myc and c-fos. Regulation of these genes is among the earliest detectable responses to estrogens and antiestrogens, being apparent within 30 min. Expression of c-myc and c-fos is apparently directly transcriptionally regulated by estrogen in immature rat uteri (12, 13) and in breast cancer cell lines (14) with kinetics that mimic those following treatment of growth arrested cells with peptide mitogens in other cell types (15). Furthermore, treatment with antisense c-myc oligonucleotides specifically inhibits the increase in cell numbers observed after estrogen treatment of MCF-7 breast cancer cells, providing strong evidence that c-myc expression is necessary for estrogen-induced proliferation (16).
The induction of c-fos and c-myc by estrogen in some respects parallels the response to growth factor mitogens. Nevertheless, there are differences in the signaling pathways utilized by estrogen and growth factors, illustrated by the different mechanisms for regulation of c-fos and c-myc, i.e. direct transcription regulation by the ER versus activation of a cascade of signaling molecules. Treatment of ZR-75-1 breast cancer cells with a series of growth factors elicited ribosomal S6 kinase activation and S6 phosphorylation, but activation of the kinase by estrogen could not be detected (17). Although transient activation of mitogen-activated protein kinases following estrogen treatment of MCF-7 cells has been demonstrated (18), another recent study using MCF-7 cells rescued with estrogen from cell cycle arrest induced by hydroxymethylglutaryl-CoA reductase inhibitors showed that estrogen-induced reinitiation of cell cycle progression was independent of activation of mitogen-activated protein kinases (19). Thus, cell cycle control by estrogen will not necessarily parallel that by growth factors.
Progress through G1 phase requires inactivation of the pRB protein by phosphorylation and the consequent release of a number of factors including the E2F family of transcription factors (20, 21). These transcription factors then activate transcription of genes whose products are required for S phase progression (20, 21). Phosphorylation of pRB is mediated by the action of the G1 phase CDKs (Cdk4, Cdk6, and Cdk2), which are activated by cyclin binding (Cdk4 and Cdk6 by D-type cyclin binding, and Cdk2 by cyclin E binding). Control of G1 CDK activity is achieved by several mechanisms including transcriptional activation of D-type cyclins and cyclin E, the rate-limiting regulatory subunits of the G1 cyclin-CDK complexes; activation and inactivation of the enzyme complexes by phosphorylation/dephosphorylation events; and the abundance and action of two families of CDK inhibitors (CDKIs) (22, 23). Modulation at any of these levels of regulation could regulate pRB phosphorylation and hence G1 phase progression.
D-type cyclins are induced as delayed early response genes by a variety of mitogens in many cell types, and removal of growth factors in G1 phase leads to their rapid down-regulation (24), consistent with the notion that these cyclins act as mitogenic sensors linking extracellular signals with cell cycle progression (25). An essential role for cyclin D1 in mammary gland development is demonstrated by the absence of lobular-alveolar structures in transgenic mice with disruption of the cyclin D1 gene (26, 27). There is accumulating evidence that D-type cyclins may also play a role in mediating the effects of growth factors and steroid hormones on breast cancer cell cycle progression by binding and activating Cdk4 and Cdk6. The abundance of cyclin D1 increases following growth factor and progestin stimulation of breast cancer cell proliferation (28) and declines rapidly following exposure to growth-inhibitory antiestrogens (10, 28). Similarly, cyclin E expression also increases following growth factor stimulation of breast cancer cells (28) but at times later than for cyclin D1 and compatible with its established role in the control of the G1 to S phase transition. Finally, ectopic expression of cyclin D1 in T-47D breast cancer cells is sufficient for Cdk4 and Cdk2 activation, pRB phosphorylation and G1-S phase progression (29). Thus, G1 cyclins and their associated kinases are potential downstream targets of estrogen-induced mitogenesis. Our recent demonstration that a decreased rate of G1 progression in antiestrogen-treated breast cancer cells is preceded by decreased cyclin D1 gene expression, cyclin D1-Cdk4 activity, and reduced pRB phosphorylation supports such a view (10).
To further test this hypothesis, we exploited the unique properties of steroidal antiestrogens that, in marked contrast to their nonsteroidal predecessors, are pure antagonists devoid of estrogen agonist activity (9, 30). Thus, exposure of estrogen-dependent cells to these compounds will block estrogen-induced gene expression with consequent inhibition of cell proliferation. Using a model in which subsequent "rescue" of these growth-arrested cells with estradiol leads to synchronous progression of cells through the cell cycle, we identified activation of both cyclin D1-Cdk4 and cyclin E-Cdk2 complexes as early events in estradiol action. Subsequent experiments were directed at identifying potential mechanisms of estrogen-induced CDK activation.
Stock solutions of
7-[9-(4,4,5,5,5-pentafluoropentylsulfinyl)nonyl]estra-1,3,5,(10)-triene-3,17
-diol
(ICI 182780) and 17
-estradiol (estradiol) were prepared as follows.
ICI 182780 (a kind gift from Dr. Alan Wakeling, Zeneca Pharmaceuticals,
Alderley Park, Cheshire, United Kingdom) was dissolved in ethanol to
10
2 M, and a working dilution of
10
5 M in RPMI 1640 medium was prepared from
this stock just prior to each experiment. Estradiol (Sigma, Castle
Hill, New South Wales, Australia) was dissolved in ethanol at 2 × 10
4 M. Stock cultures of MCF-7 cells
(Michigan Cancer Foundation, Detroit, MI) were cryopreserved and
maintained as described previously (8). Cells were cultured in RPMI
1640 medium supplemented with 5% fetal calf serum, insulin (10 µg/ml), and gentamicin (10 µg/ml). For experiments investigating
the effects of ICI 182780 and estradiol on cell cycle phase
distribution, 25-cm2 flasks were seeded with 0.8 × 105 cells. Experiments investigating the effects of
estradiol on mRNA and protein levels, cyclin-CDK complex formation,
and kinase activity employed 150-cm2 tissue culture flasks
seeded with 1.5 × 106 cells. The growth kinetics,
including changes in cell cycle phase distribution, were identical
under both experimental conditions. Cells were allowed to proliferate
for 2 days. ICI 182780 or vehicle was then added to the medium.
Following 48 h of ICI 182780 pretreatment and without a change of
medium, estradiol (100 nM), or vehicle was added directly
to the medium. The final concentration of ethanol in the tissue culture
medium was less than 0.06% and had no effect on the rate of cell
proliferation. At the completion of experiments, cells were harvested
by brief incubation with 0.05% (w/v) trypsin, 0.02% (w/v) EDTA for
flow cytometry as described previously (8) or as described below. Cell
cycle phase distribution was determined by analytical DNA flow
cytometry as described previously (31).
Total cellular RNA was extracted
from duplicate 150-cm2 flasks by a guanidinium
isothiocyanate-cesium chloride procedure, and Northern blot analysis
employing 10-20 µg of total RNA/lane was performed as described
previously (32). Filters were hybridized with cDNA labeled with
[-32P]dCTP to a specific activity of approximately
109 dpm/µg by random primer extension (Multiprime,
Amersham Australia, North Ryde, Australia). Messenger RNA abundance was
quantitated by analysis of autoradiographs using PhosphorImager
analysis (Molecular Dynamics, Sunnyvale, CA). Plasmids containing the
human cDNAs used in this study were provided by the following
investigators: cyclins D1 (pHsCYCD1-H124) and D3 (pD3-H347), Drs. Yue
Xiong and David Beach (Cold Spring Harbor Laboratory, Cold Spring
Harbor, NY); cyclin E (in pBSKS), Dr. Steven Reed (Scripps Research
Institute, La Jolla, CA); c-myc (p9C-myc), Dr. Geoff Symonds
(R. W. Johnson Pharmaceutical Research Institute, Sydney, Australia).
Restriction enzyme digestion was used to obtain fragments for labeling
as follows: for cyclin D1, a 1.3-kb EcoRI fragment
comprising the entire open reading frame; for cyclin D3, a 0.5-kb
PstI fragment corresponding to 60% of the open reading
frame; for cyclin E, a 2.5-kb EcoRI fragment comprising the
entire open reading frame; for c-myc, a 0.45-kb
PstI fragment corresponding to the entire second exon.
Equivalent loading of RNA samples was confirmed by hybridization with
an oligonucleotide probe complementary to a region of 18 S ribosomal
RNA.
Cells were lysed as follows. Cell monolayers were washed twice in ice-cold phosphate-buffered saline and then scraped into ice-cold lysis buffer A (50 mM HEPES, pH 7.5, 150 mM NaCl, 10% (v/v) glycerol, 1% Triton X-100, 1.5 mM MgCl2, 1 mM EGTA, 10 µg/ml aprotinin, 10 µg/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, 200 µM sodium orthovanadate, 10 mM sodium pyrophosphate, 100 mM NaF, and 1 mM DTT). The lysates were incubated for 5 min on ice, and the cellular debris was cleared by centrifugation (15,000 × g, 5 min, 4 °C). Equal amounts of total protein (20-40 µg) were separated by SDS-PAGE and then transferred to nitrocellulose filters. Proteins were visualized using the ECL detection system (Amersham, Australia) after incubation (2 h at room temperature or overnight at 4 °C) with the following primary antibodies: cyclin D1 (PRAD1), cyclin D2 (C-17), cyclin D3 (C-16), cyclin E (HE12), cyclin B1 (GNS1), p15 (K-18), p27 (C-19), Cdk2 (M2), Cdk4 (C-22), and Cdk6 (C-21) from Santa Cruz Biotechnology Inc. (Santa Cruz, CA); cyclin D1 (DCS-6) from Novacastra Laboratories Ltd. (Newcastle-on-Tyne, UK); p21 (a kind gift from Dr. David Beach, Cold Spring Harbor Laboratory, NY) or p21 (catalog number C24420; Transduction Laboratories, Lexington, KY); c-Myc (9E10; ATCC, Rockville, MD); or pRB (G3-245) and cyclin A (BF683) from PharMingen (San Diego, CA). Protein abundance was quantitated by analysis of Western blots using densitometry (Molecular Dynamics, Sunnyvale, CA). Quantitation of protein levels by this method was linear over the protein concentrations and exposure times tested.
Generation of Recombinant ProteinsA GST-pRB fusion protein
was employed as the substrate for the Cdk4 activity assay.
Escherichia coli transformed with a pGEX expression vector
containing a GST-pRB construct encoding amino acids 773-923 (kindly
supplied by Dr. Ed Harlow, Massachusetts General Hospital Cancer
Center, Charlestown, MA) were induced by the addition of 0.4 mM isopropyl--D-thiogalactopyranoside and
incubated for 3 h at room temperature. The bacterial pellets were
then lysed by sonication, and the fusion protein was purified by
affinity chromatography on glutathione-agarose beads and then eluted
with 15 mM reduced glutathione. GST-p16 was prepared as described previously (33).
cDNAs for Cdk2 (from Dr. Tony Hunter, Salk Institute, La Jolla, CA) and cyclin E were amplified by polymerase chain reaction and cloned separately into the baculoviral transfer vector pVL1392 (PharMingen). Cyclin E was cloned with an N-terminal GST cassette and Cdk2 with an N-terminal 6-histidine tag. The identity of both vectors was confirmed by sequencing. Viruses were generated according to the manufacturer's instructions, and Sf9 insect cells were co-infected with high titer pVL1392 GST-cyclin E and pVL1392 6His-Cdk2 viral stocks. Active cyclin E-Cdk2 complexes were purified using glutathione affinity chromatography.
Cyclin-dependent Kinase AssaysFor Cdk4
activity assays, MCF-7 cell monolayers were washed twice with
phosphate-buffered saline and then scraped into ice-cold phosphate-buffered saline and pelleted by centrifugation (15,000 × g, 5 min), and the pellets were frozen in liquid nitrogen
and then resuspended in 1 ml of ice-cold lysis buffer B (50 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM
EDTA, 2.5 mM EGTA, 0.1% Tween 20, 10% glycerol, 10 mM -glycerophosphate, 1 mM NaF, 0.1 mM orthovanadate, 10 µg/ml leupeptin, 10 µg/ml
aprotinin, 1 mM DTT, and 0.1 mM
phenylmethylsulfonyl fluoride). The lysate was placed on ice and
vortexed vigorously at intervals for 60 min and then centrifuged at
15,000 × g for 5 min at 4 °C, and the supernatant
was stored at
80 °C. Equivalent amounts of protein were precleared
by incubation with preimmune rabbit serum conjugated to protein
A-Sepharose (Zymed, San Francisco, CA) for 45 min at 4 °C. Cdk4
complexes were immunoprecipitated with rabbit polyclonal anti-human
Cdk4 antiserum (catalog number 14936E, PharMingen) conjugated to
protein A-Sepharose for 3 h at 4 °C. Immunoprecipitates were
washed four times with ice-cold lysis buffer B and three times with
ice-cold 50 mM HEPES pH 7.5, 1 mM DTT and then
either incubated with 30 µl of 50 mM HEPES, pH 7.5, 1 mM DTT or 30 µl of the same buffer containing
approximately 10
5 M bacterially expressed
GST-p16 or GST for 1 h at 30 °C. The supernatant was aspirated,
and the immunoprecipitates were used for the kinase assay.
For both the cyclin E-Cdk2 kinase activity and Cdk2 activity assays, lysates were prepared with lysis buffer A as described above. Equivalent amounts of lysate were precleared with protein A-Sepharose (1 h, 4 °C) and then immunoprecipitated with protein A-Sepharose conjugated to either an anti-cyclin E polyclonal antibody (C-19; Santa Cruz Biotechnology) or an anti-Cdk2 polyclonal antibody (M2; Santa Cruz Biotechnology) for 3 h at 4 °C. The immunoprecipitates were then washed once with ice-cold lysis buffer A, twice with ice-cold lysis buffer A containing 1 M NaCl, once again with ice-cold lysis buffer A, and then three times with ice-cold 50 mM HEPES, pH 7.5, 1 mM DTT.
The kinase reactions were initiated by resuspending the beads in 30 µl of kinase buffer (50 mM HEPES, pH 7.5, 1 mM DTT, 2.5 mM EGTA, 10 mM
MgCl2, 20 µM ATP, 10 µCi of
[-32P]ATP, 0.1 mM orthovanadate, 1 mM NaF, 10 mM
-glycerophosphate) containing
either 10 µg of GST-pRB773-923 (Cdk4
immunoprecipitates), 10 µg of histone H1 (from Sigma; cyclin E
immunoprecipitates), or 3 µg of histone H1 (Cdk2 immunoprecipitates) as a substrate. After incubation for either 30 min (Cdk4), or 15 min
(cyclin E and Cdk2) at 30 °C, the reactions were terminated by the
addition of 15 µl of 3 × SDS sample buffer (187 mM
Tris-HCl, pH 6.8, 30% (v/v) glycerol, 6% SDS, 15% (v/v)
-mercaptoethanol). Unlabeled ATP (1 µl of a 10 mM
stock solution) was added to the Cdk4 activity assays following
termination to reduce background. The samples were then heated at
95 °C for 2 min and separated using 12% SDS-PAGE, and the dried gel
was exposed to x-ray film. Relative band intensities were quantitated
by densitometric analysis as described above.
Immunoprecipitation of cyclin D1, cyclin E, and Cdk2 was performed using the method described above (for immunoprecipitating cyclin E and Cdk2 for kinase activity assays), except that the cyclin E antibodies were chemically cross-linked with dimethyl pimelimidate (Sigma) to protein A-Sepharose to reduce background (34). Antibodies used were rabbit polyclonal antisera to human cyclin D1 (29), human cyclin E (C-19; Santa Cruz Biotechnology), and human Cdk2 (M2; Santa Cruz Biotechnology). Supernatants were Western blotted to determine the degree of immunodepletion achieved with each antibody. The immunoprecipitated proteins were resuspended in 1 × SDS sample buffer, separated by SDS-PAGE, and transferred to a nitrocellulose membrane, and then proteins were detected using the antibodies described for Western blotting above. A rabbit polyclonal anti-p27 antiserum (C-19; Santa Cruz Biotechnology) and a mouse monoclonal anti-p27 antibody (catalog number K25020; Transduction Laboratories) were both used on cyclin E immunoprecipitates. Protein A-horseradish peroxidase (Zymed Laboratories, San Francisco, CA) was employed rather than a goat anti-rabbit secondary antibody to reduce background following the use of primary rabbit polyclonal antibodies. Equivalent lane loading was demonstrated by the rabbit polyclonal IgG used for immunoprecipitation that cross-reacted with secondary antibodies used for Western blotting.
Assay for Inhibitory Activity toward Cyclin E-Cdk2Cell lysates were prepared with lysis buffer A as described above. Heat treatment was performed by boiling cell lysates for 3 min followed by centrifugation (15,000 × g, 5 min, 4 °C) to remove insoluble material. CDKIs were immunodepleted by incubating cell lysates with either rabbit anti-p27 polyclonal antibody (catalog number sc-528-G; Santa Cruz Biotechnology), mouse anti-p21 monoclonal antibody (catalog number C24420; Transduction Laboratories), both antibodies, or no antibody for 3-16 h at 4 °C followed by precipitation with protein A-Sepharose beads. Complete immunodepletion of p21 could only be achieved in heat-treated lysates, whereas only 30-40% immunodepletion was possible in unheated lysates. Active cyclin E-Cdk2 complexes purified from Sf9 cells were incubated with 50 µg of cell lysate protein (either heat-treated or untreated, CDKI-depleted or nondepleted) for 1 h at room temperature. Recombinant cyclin E-Cdk2 complexes were bound to glutathione-agarose beads for 1 h at 4 °C and washed once with lysis buffer A and twice with 50 mM HEPES, pH 7.5, 1 mM DTT, and histone kinase activity was assayed. The sample was electrophoresed on duplicate 12% SDS-PAGE gels. One gel was dried, and autoradiography was performed for assessment of histone phosphorylation, while the other was transferred to a nitrocellulose filter and Western blotted for assessment of cyclin E-Cdk2-associated p21 and p27.
Gel FiltrationCell lysates prepared in lysis buffer A were
passed through a 0.22-µm MILLEX-GV4 filter (Millipore, Lane Cove, New
South Wales, Australia) and then fractionated on a Superose 12 column
using a fast protein liquid chromatography system (both from Pharmacia Biotech Inc., Uppsala, Sweden). The column was developed at a flow rate
of 0.5 ml/min with 20 mM HEPES, pH 7.5, 250 mM
NaCl, 1 mM EDTA, 0.1% (v/v) -mercaptoethanol, and 40 500-µl fractions were collected. Column calibration was performed
under the same conditions using a high molecular weight gel filtration
calibration kit (Pharmacia) containing ferritin (440 kDa), catalase
(232 kDa), and aldolase (158 kDa). Protein was concentrated prior to
Western blotting. 200 µl of each fraction was placed at
70 °C
for 1 h with 1 ml of acetone and 10 µg of carrier bovine serum
albumin protein, and protein pellets were collected by centrifugation (15,000 × g, 5 min, 4 °C) and then resuspended by
boiling for 4 min in 30 µl of SDS sample buffer. For cyclin E-Cdk2
kinase assays, 300 µl of each fraction was used. For Western blotting of cyclin E immunoprecipitates, 500 µl of each fraction was used.
To establish a sensitive experimental system for the study of specific estradiol-regulated events associated with cell cycle progression, we first characterized a model in which MCF-7 breast cancer cells were growth-arrested with the pure estrogen antagonist ICI 182780 and then "rescued" by the addition of estradiol. Initial experiments performed to optimize conditions demonstrated that maximal cell cycle inhibitory effects were achieved with 10 nM ICI 182780, reducing S phase from 40% to a minimum of 10% over 48 h, and maintaining this level of inhibition for a further 48 h (data not shown). Following 48 h of ICI 182780 pretreatment, cells were rescued with estradiol. A concentration of 100 nM estradiol was chosen, since this concentration reproducibly induced maximal stimulation of cell cycle progression in this experimental setting (data not shown).
A representative time course for changes in cell cycle phase
distribution following this treatment strategy is shown in Fig. 1. In exponentially growing cultures, the percentages of
cells in G1, S, and G2 + M phases were 50, 40, and 10, respectively. With antiestrogen pretreatment for 48 h,
cells accumulated in G1 phase such that the phase
distribution changed to 85, 10, and 5% respectively. Following
estradiol treatment, there was little change in cell cycle phase
distribution over the first 10 h, but thereafter the proportion of
G1 phase cells decreased from 86 to a minimum of 28% at
24 h (Fig. 1A). This decrease was mirrored by a 6-fold
increase in the proportion of cells in S phase from 10 to 60% (Fig.
1B), indicating a synchronized population of cells progressing from G1 to S phase. The subsequent decline in
the proportion of S phase cells after 24 h was accompanied by an
increase in the proportion of cells in G2 + M phases from
5% prior to 21 h to a maximum of 25% at 27 h (Fig.
1C). Estrogen treatment, therefore, was sufficient for cells
to progress from G1 to S phase and then to G2 + M phases as indicated by the 5-6-fold increase of cells in S phase and
then cells in G2 + M phases. The subsequent decline of the
proportion of cells in S and G2 + M phases was accompanied by semisynchronous reentry into G1 phase after 24 h,
indicating that cell division had occurred. Thus, estradiol treatment
induced synchronous progression through at least one cell cycle. The
changes in G1 (28-60%) and S phase cells (60-33%)
between 24 and 36 h reflect cells entering a second cycle but with
a markedly decreased degree of synchrony.
Estradiol Induces c-myc Expression
The model system was
further validated by examining the effects of estradiol rescue on
c-myc, an estradiol-regulated gene with an established role
in cell proliferation in this cell line (14, 16). In this model, c-Myc
protein was markedly increased at 2 h (Fig.
2A), the first time point studied, reached a
maximum ~ 8-fold increase at 4 h, and then subsequently
declined but remained elevated about 4-fold above controls until
24 h. The control lanes illustrated in Fig. 2A
demonstrated that c-Myc protein levels were at the limit of detection
in cells treated with ICI 182780 for 48-72 h (lane b) and
significantly lower than in exponentially growing cells treated with
vehicle (lane c) or estradiol alone (lane d).
Changes in the expression of c-Myc protein were preceded by similar
changes in the expression of c-myc mRNA, which reached a
maximal level at 2 h and thereafter declined but was still
elevated above controls at 24 h (Fig. 2B). Thus, the
temporal changes in c-myc gene expression following
estradiol treatment in this model are consistent with previously
published data for this and other target tissues (12, 14).
pRB Phosphorylation following Estradiol Treatment
Cells that
possess wild-type pRB require its inactivation by phosphorylation for
progress through G1 phase (20, 21). We therefore determined
whether estradiol altered pRB phosphorylation at times compatible with
a role in G1-S phase progression in breast cancer cells.
After 48 h of ICI 182780 pretreatment, almost all pRB was
hypophosphorylated (Fig. 3, time 0), in
agreement with our previous observations (10). Following estradiol
treatment, an increase in more slowly migrating, phosphorylated forms
of pRB was first apparent at 6 h. The proportion of phosphorylated pRB increased at subsequent time points with a corresponding decrease in hypophosphorylated pRB, such that after 12 h, when cells
commenced their synchronous entry into S phase (Fig. 1B),
little or no hypophosphorylated pRB remained. These later changes in
pRB phosphorylation were accompanied by an increase in the total amount
of pRB protein; at 24 h, total pRB levels were increased
approximately 4-fold relative to cells treated with ICI 182780 alone.
Controls for these experiments included cells that were in exponential growth and had not received either ICI 182780 or estradiol (lanes c) and cells that received estradiol but no antiestrogen pretreatment (lanes d). In both of these cases, cells proliferated at equal rates with similar cell cycle phase distributions associated with exponential growth (8) and had similar patterns of pRB phosphorylation, i.e. both hypo- and hyperphosphorylated pRB were present, reflecting an asynchronous cell population and hence cells in all phases of the cell cycle. Parallel control lanes a and b show cultures pretreated with antiestrogen for 48 h and then treated with estradiol (lane a) or ethanol vehicle (lane b). The continuous presence of ICI 182780 maintained pRB in the hypophosphorylated state for a further 24 h (lane b), while estradiol rescue resulted in the time-dependent increase in phosphorylation referred to above.
Effects of Estradiol on Cyclin, CDK, and CDKI ExpressionSince estradiol induced significant changes in pRB
phosphorylation, the effects of estradiol on components of the Cdk2,
Cdk4, and Cdk6 complexes, which are the major kinases responsible for pRB phosphorylation, were investigated. We first studied the expression of the regulatory subunits for these kinases, the G1
cyclins D1, D3, and E. The earliest and most pronounced changes in
protein expression were for cyclin D1. Cyclin D1 levels increased
2.6-fold at 4 h (Figs. 4A and
5B), reached maximum levels at 8 h
(5-7-fold), prior to any change in the percentage of S phase cells,
and then decreased to 2-fold above control at 24 h. In contrast,
cyclin D3 levels increased as the proportion of cells in S phase
increased i.e. after 12 h (Fig. 4A) to reach
a maximum of 3.3-fold above control at 24 h. Cyclin D2 is
expressed at very low levels in these cells (35) with no detectable
change in expression following estradiol treatment (data not shown).
Surprisingly, antiestrogen-induced growth inhibition was associated
with an approximate 4-fold increase in cyclin E levels above those
observed in exponentially growing cells (lane b versus lanes
c and d in cyclin E controls). Following estradiol
treatment, there was a less than 25% increase in the expression of
cyclin E from 2-16 h relative to controls (Fig. 4A), but
levels declined slightly at 24 h as cells progressed through S
phase.
Cyclins A and B1 remained at control levels until 16 h after estradiol rescue (Fig. 4B), at which time the proportion of cells in S phase was increased significantly (Fig. 1B) and were further increased at 24 and 30 h, consistent with the established role for these cyclins in S and G2 + M phases. Increased expression of cyclin B1 protein following estrogen treatment of MCF-7 cells has been reported previously (36).
Since the earliest and largest effect on G1 cyclin protein expression following estradiol rescue was that on cyclin D1, further experiments were undertaken to examine the effects of estradiol on cyclin D1 mRNA levels. Cyclin D1 mRNA levels increased 1.8-fold at 2 h (Fig. 5A), prior to the increase in cyclin D1 protein (Fig. 5B), continued to increase to a maximal 3.0-fold increase at 6 h, and declined slightly thereafter but were still greater than 2-fold above control at 24 h. Cyclin D1 protein levels were unchanged in the first 2 h of treatment but increased rapidly between 2 and 6 h to reach a 5-7-fold maximal increase at 8 h. The disparity between the maximum -fold induction of cyclin D1 mRNA and protein was seen in several independent experiments, and although it was not studied further it suggests that estradiol may also influence cyclin D1 protein turnover. The increase in cyclin D1 mRNA expression was inhibited by both actinomycin D (5 µg/ml) and cycloheximide (20 µg/ml), indicating a role for both transcription and protein synthesis in mediating this effect (data not shown). Elevated cyclin D3 mRNA levels were detected at 10 h, prior to the increase in protein levels (Fig. 5, A and B), reaching a maximum ~ 3-fold increase at 24 h. Cyclin E mRNA levels were not markedly changed by estradiol treatment, i.e. less than a 30% increase (data not shown).
The effects of estradiol rescue on expression of Cdk2, Cdk4, and Cdk6
were also investigated by Western blotting (Fig. 6). There was an approximately 2-fold increase in the levels of both Cdk2
and Cdk4 after 12 h similar to those observed during stimulation of a round of replication in growth-arrested breast cancer cells with
mitogenic growth factors (28). Surprisingly, estradiol treatment
resulted in a 70% reduction in the level of expression of Cdk6 over
the 24 h of treatment.
Following 48 h of antiestrogen pretreatment, the levels of the
CDKIs p21 and p27 were elevated 3-4-fold above those of exponentially growing cells (data not shown) in agreement with our previous observations made after 24 h of antiestrogen treatment (10). Subsequent estradiol rescue did not alter the expression of p21 until
8 h, after which time levels progressively declined to 40% of
control at 24 h (Fig. 7). The expression of p27 was
unaltered until 12 h following estradiol but thereafter declined
to 50% of control at 24 h (Fig. 7). However, at this time both
inhibitors were still present at levels well above those seen in
untreated, exponentially growing control cells. The Cdk4/6 inhibitor
p15INK4B was only detected at very low levels in cells that
were growing exponentially or treated with antiestrogen or estradiol
(data not shown). The p16INK4A gene is homozygously deleted in
this cell line (37, 38), thus precluding a role in the estradiol
effect.
Activation of G1 CDKs
The effects of estradiol on
Cdk4, cyclin E-Cdk2, and total Cdk2 activities were investigated using
appropriate immunoprecipitates and substrates. Cdk4 activity
(GST-pRB773-923 substrate) was elevated 4.6-fold at 3 h after estradiol treatment (Fig. 8, A and
B), maximally elevated at 6 h (6.6-fold), and
thereafter declined. The initial changes in Cdk4 activity were
temporally similar to those of changes in expression for cyclin D1
protein (Fig. 4), suggesting that an activating mechanism for Cdk4 was increased cyclin D1 expression. Somewhat surprisingly, Cdk4 activity decreased prior to the decline in cyclin D1 protein levels, a similar
result to that reported for cyclin D1-associated kinase activity
following estrogen rescue of MCF-7 cells from simvastatin arrest (39).
Inhibition of immunoprecipitated kinases with a GST-p16 fusion protein
reduced activity to control levels (Fig. 8A), demonstrating
the specificity of this assay for Cdk4 activity. Incubation with GST
alone had no significant effect on activity (data not shown).
After antiestrogen pretreatment, cyclin E-Cdk2 kinase activity (histone H1 substrate) was approximately 60% of that in exponentially growing cells (Fig. 8A). Estradiol treatment restored this activity by 1.5-fold at 4 h and 3.0-fold at 6 h (Fig. 8, A and B), when both the increase in Cdk4 activity (Fig. 8A) and the shift in pRB phosphorylation (Fig. 3) were first noted. Cyclin E-Cdk2 kinase activity continued to increase, reaching a maximum at 16 h approximately 7-fold relative to control levels and equivalent to about 4-fold the activity in exponentially growing cells. Similarly, total Cdk2 activity (histone H1 substrate) was markedly decreased by antiestrogen treatment, and subsequent estradiol treatment increased activity 2.2-fold at 6 h (Fig. 8, A and B), with continued increases paralleling changes in cyclin E-Cdk2 activity until 10 h. Further pronounced increases in total Cdk2 activity occurred after 12 h, approaching levels in exponentially growing cells (Fig. 8A) and coincident with the marked increases in S phase (Fig. 1B) and cyclin A gene expression (Fig. 4B).
The substantial and early changes in both Cdk4 activity and cyclin E-Cdk2 activity between 4 and 6 h indicated that both kinases were likely to contribute to the initial changes in pRB phosphorylation following estradiol treatment. Subsequent large increases in both cyclin E-Cdk2 activity and total Cdk2 activity were likely to contribute to the pronounced pRB phosphorylation observed at later time points (Fig. 3).
Estradiol Increases Cyclin D1-Cdk4 Complex FormationTo
confirm that the early and large changes in Cdk4 activity were due to
increased cyclin D1 protein expression as appeared likely, we
investigated changes in cyclin D1-Cdk4 association following estradiol
treatment. Whole cell lysates were immunoprecipitated with anti-cyclin
D1 antiserum and Western blotted for cyclin D1, Cdk2, Cdk4, and Cdk6
and the CDKIs p21 and p27. Under these conditions at least 95% of the
cyclin D1 was immunoprecipitated, and cyclin D1, Cdk4, p21, and p27
were not detected in preimmune serum immunoprecipitates. Metabolic
labeling of MCF-7 cells followed by cyclin D1 immunoprecipitation demonstrated that proteins corresponding to the electrophoretic mobility of Cdk4 and Cdk2 were substantially more abundant than proteins migrating at the expected position of Cdk6. These labeled proteins were all absent in preimmune serum immunoprecipitates (data
not shown). Antiestrogen pretreatment decreased the levels of both
cyclin D1 and Cdk4 in cyclin D1 immunoprecipitates, consistent with the
decrease in total cyclin D1 protein. Following estradiol rescue, the
levels of immunoprecipitated cyclin D1 increased ~ 3-fold at
4 h and remained elevated until 16 h (Fig.
9B), paralleling the temporal changes in
total cyclin D1 protein, although the relative changes are of lower
magnitude (see Fig. 4A). The relative levels of Cdk4 in the
complexes at early time points paralleled the levels of both cyclin D1
protein (Fig. 9B) and Cdk4 activity (Fig. 8) and were
maximal at 8 h (3.4-fold). Subsequently, cyclin D1-associated Cdk4
declined, reaching control levels at 24 h, although the coincident
reduction in Cdk4 activity was more pronounced (Fig. 8C).
Levels of cyclin D1-associated Cdk2 also increased following estradiol
rescue, but immunoprecipitates of cyclin D1 were unable to
phosphorylate histone H1 (data not shown), indicating that cyclin
D1-Cdk2 complexes were inactive, as described by others (40, 41).
After antiestrogen pretreatment, p21 and p27 associated with cyclin D1 increased approximately 2-fold relative to levels in exponentially growing cells (data not shown). In response to estradiol, cyclin D1-associated p21 and p27 were both further increased at 4 and 8 h (Fig. 9B), despite relatively constant total cellular levels of these proteins between 0 and 8 h (Fig. 7). Cyclin D1-associated p21 reached a maximal 4.5-fold increase between 4 and 8 h and then declined to control levels at 16 h. These results are compatible with the recent suggestion that p21 may act as an assembly factor for cyclin D1-Cdk4 complexes (42). A second, more slowly migrating form of cyclin D1-associated p21 reported previously (42) was evident in the original Western blot from 4 to 24 h but was absent in the 0 h and control lanes. Cyclin D1-associated p27 was maximal at 8 h (2.0-fold), declining to control levels between 16 and 24 h.
These data demonstrate a major increase in cyclin D1-Cdk4 complex formation as early as 4 h after estradiol treatment, indicating that the increased Cdk4 activity at this time was likely to represent activation by cyclin D1.
Effect of Estradiol on Cyclin E-associated ProteinsThe early increase in cyclin E-Cdk2 activity was not accompanied by significant increases in the expression of cyclin E or Cdk2 or by decreases in the expression of the CDKIs p21 and p27 (Figs. 4, 6, and 7). Experiments were performed to investigate whether these proteins had a role in activating cyclin E-Cdk2 by examining their association with cyclin E. Cell lysates were harvested after estradiol treatment, and cyclin E immunoprecipitates were Western blotted for cyclin E, Cdk2, p21, and p27 (Fig. 9C). The levels of these proteins remained unchanged in cyclin E complexes until 10 h after estradiol treatment. At 16 h, the level of cyclin E-associated p27 and p21 both declined, particularly that of p27. There was an increase in the relative abundance of the rapidly migrating, Thr-160 phosphorylated form of Cdk2 at 16 h (43). CAK, suggested to be cyclin H-Cdk7 in mammalian cells (44), catalyzes the phosphorylation of Thr-160 on Cdk2. However, Cdk7 activity measured using a GST-Cdk2 substrate was unaltered following estradiol treatment (data not shown). Therefore, while activation of cyclin E-Cdk2 at 16 h was likely to involve the loss of CDKI proteins and phosphorylation of Cdk2 on Thr-160, these experiments did not identify a mechanism for the activation of cyclin E-Cdk2 prior to 16 h.
Using a similar experimental design, the immunoprecipitation of Cdk2 demonstrated that Cdk2-associated cyclin A was greatly elevated at 24 h (data not shown), corresponding to increased total cellular cyclin A levels (Fig. 4B) and probably represents the mechanism by which estradiol further increased total Cdk2 activity between 16 and 24 h (Fig. 8).
Inhibitory Activity toward Cyclin E-Cdk2 Decreases following Estradiol TreatmentTo further investigate the activation of
cyclin E-Cdk2, the potential role of inhibitory activity toward cyclin
E-Cdk2 was examined. Cell lysates from antiestrogen- and
estradiol-treated cells were incubated in an in vitro assay
with active recombinant cyclin E-Cdk2 complexes prepared from
baculovirus-infected Sf9 cells. Recombinant cyclin E-Cdk2 that was
incubated with lysates from cells pretreated with ICI 182780 for
48 h was less active and contained more p21 than when incubated
with lysates from cells harvested 8 h after estradiol treatment
(Fig. 10A). Boiled lysates from both
estradiol- and ICI 182780-treated cells contained similar levels of
inhibitory activity toward cyclin E-Cdk2 (Fig. 10B) and similar levels of cyclin E-Cdk2 association with p21 and p27. Immunodepletion of p21 was sufficient to remove most of this inhibitory activity in lysates that were not boiled, while immunodepletion of both
p21 and p27 was required to remove most of the inhibitory activity in
boiled cell lysates. Immunodepletion of p21 resulted in increased p27
binding to recombinant cyclin E-Cdk2, presumably because these CDKIs
compete for identical binding sites. Similar results were obtained in
parallel experiments with recombinant cyclin A-Cdk2 (data not shown).
These data suggest that the level of recombinant cyclin E-Cdk2 activity
was determined by the level of association with CDKIs from the cell
lysates. This association was reduced following incubation with lysates
from estradiol-treated cells but could be increased by heat-treating
the cell lysates. This suggests that estradiol may prevent CDKI
association with recombinant cyclin E-Cdk2 by a mechanism that is
heat-labile. The decrease in inhibitory activity toward recombinant
cyclin E-Cdk2 following estradiol treatment was temporally similar to but less pronounced than the increase in endogenous cyclin E-Cdk2 activity (compare Fig. 10C with Fig. 8).
Activation of Cyclin E-Cdk2 Is Accompanied by Decreased CDK Inhibitor Association and Thr-160 Phosphorylation of Cdk2
Experiments by others (45) have described the activation of
cyclin E-Cdk2 following release from high molecular weight complexes; therefore, we investigated this possibility using gel filtration chromatography. Lysates of antiestrogen-pretreated MCF-7 cells prepared
with and without 8 h of estradiol treatment contained one major
peak of cyclin E protein eluting at approximately 160 kDa (fraction 24, Fig. 11A). Following estrogen treatment,
there was a small but consistent increase in cyclin E migrating in
higher molecular weight complexes (fractions 18-22), and these
complexes contained the majority of cyclin E-Cdk2 kinase activity.
Consequently, the specific activity of these higher molecular weight
complexes was 10-fold greater than the bulk of the cyclin E eluting at
the lower molecular weights. Comparison of the composition of cyclin E
immunoprecipitates eluting at these different molecular weights revealed that the larger complexes were markedly depleted of both p21
and p27 (Fig. 11B), in contrast to previous results for
cyclin E immunoprecipitates from whole cell lysates (Fig. 9).
Furthermore, in the high specific activity complexes, most of the Cdk2
was of the rapidly migrating, Thr-160 phosphorylated form (Fig.
11B) (43), again in contrast to previous results for total
cyclin E-associated Cdk2 (Fig. 9). In a separate experiment,
phosphorylation of cyclin E-associated Cdk2 following 8 h of
estradiol treatment was also observed by 32P labeling, and
phosphoamino acid analysis demonstrated that this was predominantly
threonine phosphorylation (data not shown). Thus, estradiol treatment
results in the formation of high molecular weight cyclin E-Cdk2
complexes that represent only a minority of the total cyclin E present
and are of high specific activity presumably as a consequence of the
relative lack of p21 and p27 and increased phosphorylation of Cdk2 on
Thr-160.
Despite the biological importance of estrogens and their involvement in the control of cell division in breast cancer and in target organs including the uterus and mammary gland, little is understood of the underlying programs of gene expression that mediate estrogen action. Estrogens act primarily through binding to the nuclear ER. This directly regulates the transcription of estrogen-responsive genes, which in turn influence numerous additional gene products. Recent advances in the definition of mechanisms controlling the cell cycle have allowed examination of the role of estrogens in the regulation of the expression and activity of genes known to be involved in this process. This study focuses on early estrogen responses in human breast cancer cells to identify target genes that are potential regulators of estrogen-stimulated cell cycle progression through G1 phase.
ER-positive breast cancer cell lines provide the best studied in vitro models of estrogen action, but large estrogenic proliferative responses have generally been difficult to obtain, even with removal of endogenous steroids from the culture medium, placing limitations on the study of changes in gene expression underlying these subtle effects on cell cycle progression. The use of synchronized cell populations is one strategy for increasing sensitivity, and several approaches have been adopted. These include the use of hydroxymethylglutaryl-CoA reductase inhibitors (i.e. lovostatin or simvastatin (46)), isoleucine deprivation (36), or nocodazole arrest (7). The system with perhaps the most specificity is one in which cells are first growth-arrested by estrogen antagonists and then rescued by estrogen, since some selective magnification of ER-regulated gene responses compared with more generalized responses to cell cycle progression might be expected. Rescue from growth arrest induced by nonsteroidal antiestrogens, i.e. tamoxifen and hydroxytamoxifen, is described in several earlier studies (14, 47, 48), but the partial agonist activity of nonsteroidal antiestrogens presents limitations in defining ER-mediated events. This limitation is compounded for tamoxifen, since it is not particularly potent and may act through non-ER-mediated mechanisms at effective concentrations (1). In the present study, we have refined the estrogen rescue model through the use of the estrogen antagonist ICI 182780, which as a consequence of its pure antagonist activity and its high affinity for the ER (9), is up to 2 orders of magnitude more potent than tamoxifen in inhibiting breast cancer cell growth. Estrogen rescue of ICI 182780-arrested MCF-7 cells resulted in a high degree of synchrony, with changes in cell cycle phase distribution approaching those achieved by mitotic selection (11, 31). However, the average duration of both G1 and S phases was greater following estrogen rescue than in mitotically selected cells, i.e. ~15 h compared with 9 h for G1 phase and ~12 h compared with 9 h for S phase.
Using this model, this study has defined a series of molecular events preceding changes in cell cycle phase distribution after estrogen rescue. Increased expression of c-myc and cyclin D1 mRNA and protein were the earliest observed responses to estrogen, with consequent cyclin D1-Cdk4 complex formation and activation of Cdk4. This was closely followed by activation of cyclin E-associated Cdk2 and increased pRB phosphorylation. Further increases in pRB phosphorylation were coincident with dramatic increases in Cdk2 activity as cells entered S phase. This is probably due to the marked increase in cyclin A abundance, but decreases in expression of the CDKIs p21 and p27 at these times may also contribute to these effects.
With the eventual aim of defining the ER target genes responsible for initiation of cell cycle progression, we have focused on the early events implicated in pRB phosphorylation, activation of cyclin D1-Cdk4 and cyclin E-Cdk2. In current models of G1 control, Cdk4 is thought to be critical for the earliest effects on pRB phosphorylation (20). The relative importance of Cdk2 in addition to Cdk4 as an early pRB kinase in the estrogen response is not yet clear; although Cdk2 activity is more slowly induced than Cdk4, pRB phosphorylation is first observed when the activities of both kinases are substantially increased. While ectopic expression of either cyclin D or E shortens G1 phase in various cell types (49-52), only overexpression of cyclin D1 leads to rapid pRB phosphorylation. In contrast, overexpression of cyclin E has a delayed effect on pRB phosphorylation despite rapid induction of cyclin E-Cdk2 kinase activity (53).
A likely explanation for Cdk4 activation following estrogen treatment is that it is the direct consequence of increased cyclin D1-Cdk4 complex formation resulting from estrogen-induced expression of cyclin D1 protein, a property shared with a number of other mitogens. In breast cancer cells, progestins, insulin-like growth factor-1, insulin, serum, and basic fibroblast growth factor induce cyclin D1 mRNA, protein, and cyclin D1-CDK complex formation (28) as do many other mitogens in a variety of other cell types (24, 54). Two studies published during the course of this work have also addressed the mechanisms of cell cycle control by estrogen using rescue from hydroxymethylglutaryl-CoA reductase inhibition (39) or from serum starvation and methionine/glutamine deprivation (55). Although these methods appear to produce a less dramatic effect on cell cycle progression than in the antiestrogen rescue model, both studies support the conclusion that estradiol activates Cdk4 by a mechanism involving increased cyclin D1 expression.
Several studies provide evidence for an important role of cyclin D1 in G1 progression in breast cancer cells. Ectopic expression of cyclin D1 is sufficient and rate-limiting for G1-S phase progression in pRB-positive breast cancer cells and results in increases in cyclin D1-Cdk4 and Cdk2 activities (29, 51).2 Cell cycle progression in MCF-7 cells can be arrested by inhibition of cyclin D1 function by microinjection or electroporation of anti-cyclin D1 antibodies (56). These observations suggest that activation of cyclin D1-Cdk4 might be a critical component of the estrogen response, a hypothesis supported by recent experiments investigating the effects of antiestrogens, which suggested that these compounds probably achieve their growth-inhibitory effects by reducing cyclin D1 expression (10). In these studies, ICI 182780 decreased the expression of cyclin D1 by 50%, which preceded reductions in cyclin D1-Cdk4 kinase activity, pRB phosphorylation, and S phase in MCF-7 cells.
Estrogen regulation of cyclin D1 protein expression appears to be at least partially explained by corresponding changes in mRNA levels, although the relatively greater increase in protein that was evident may indicate alterations in protein turnover. The mRNA cap-binding protein eIF-4E is reported to increase cyclin D1 translation rates (57) and may be involved in estrogen action, particularly given its transcriptional regulation by c-Myc (58). Although experiments with actinomycin D suggest that the effects of estrogen on cyclin D1 mRNA levels are transcriptionally mediated, the ability of cycloheximide to abolish mRNA induction shows that this is not necessarily a direct effect of ER on the cyclin D1 gene promoter and implies a requirement for de novo synthesis of intermediary proteins, which mediate either cyclin D1 gene transcription or mRNA stability. Studies on the cyclin D1 gene promoter have identified several regulatory regions including an AP-1 site (59), providing a link between estrogen-induced AP-1 activity (60) and cyclin D1 induction. A recent study confirms that this AP-1 site is within the promoter region responsible for estrogen regulation of this gene (39). It is also possible that c-Myc, itself directly estrogen-regulated via an atypical ER response element (61) positively regulates cyclin D1 expression (62). However, other studies do not show increases in cyclin D1 following c-Myc induction (63-65). Moreover, constitutive c-myc expression represses cyclin D1 mRNA expression in fibroblasts (66, 67). Nevertheless, cyclin D1 and c-Myc expression are dependent upon each other for colony-stimulating factor-1-dependent cell cycle rescue of fibroblasts expressing a mutant colony-stimulating factor-1 receptor (68). An additional complexity in the relationship between c-Myc and cyclin D1 is provided by a recent study which describes the ability of c-Myc to activate cyclin D1-Cdk4 independent of alterations in the levels of the known components of the complex (45).
In our model, cyclin E-Cdk2 was activated shortly after Cdk4 and within the first 4-6 h following estradiol treatment. A similar observation was recently reported in a model in which estrogen was used to rescue MCF-7 cells from serum starvation and methionine/glutamine deprivation (55). In contrast to the action of other mitogens, where activation of cyclin E-Cdk2 occurs through increases in total cyclin E and Cdk2-associated cyclin E (69, 70), we detected only small changes in either cyclin E mRNA or protein or Cdk2-associated cyclin E at the time of complex activation. This was unexpected in view of a recent model proposing that cyclin E levels are regulated transcriptionally by E2F following cyclin D1-Cdk4 phosphorylation of pRB (71, 72), which is supported by our previous studies in breast cancer cells in which ectopic expression of cyclin D1 led to pRB phosphorylation and induction of cyclin E mRNA (29). It is possible that the observed increase in the cyclin E protein level with estradiol treatment was small because of the already high level of cyclin E protein following antiestrogen pretreatment. This may be related to the decrease in cyclin E-Cdk2 activity following antiestrogen pretreatment, since cyclin E degradation, which is ubiquitin-dependent, requires activation of cyclin E-Cdk2 (73, 74). Elevated levels of cyclin E following antiestrogen pretreatment may also contribute to the difficulty in detecting alterations in cyclin E-associated CDKIs following estradiol treatment, since only a small fraction of the total cyclin E-containing complexes were active at early time points.
Our results suggest that estrogen activates cyclin E-Cdk2 by promoting the formation of cyclin E-Cdk2 complexes that lack the CDKIs p21 and p27, both of which directly inhibit cyclin E-Cdk2 activity (75-78). Although there was no change in the level of CDKIs in either whole cell lysates or cyclin E immunoprecipitates, there was a reduction in inhibitory activity toward recombinant cyclin E-Cdk2 following estradiol treatment, which was attributable to p21. An inhibitory role for p27 in vivo cannot be excluded by these experiments and has been described following a combination of estradiol treatment and release from glutamine/methionine deprivation (55), although it was unclear whether there was an actual decrease in cyclin E-Cdk2-p27 association in this system. Further activation of cyclin E-Cdk2 by phosphorylation of Cdk2 on Thr-160, a target of CAK, is also likely to result from a lack of association of CDKIs with cyclin E-Cdk2. CAK activity (measured as Cdk7 activity) did not change, as noted in other models of cell cycle progression (reviewed in Ref. 79). However, both p21 and p27 prevent CAK-mediated phosphorylation of Thr-160 on Cdk2 (77, 80); therefore, their absence from cyclin E-Cdk2 complexes would permit CAK-mediated phosphorylation and activation of Cdk2.
Removal of inhibitory phosphorylation sites on Cdk2 (Thr-14 and Tyr-15) by the Cdc25A phosphatase may also have a role in the activation of cyclin E-Cdk2. This has been described in fibroblasts following activation of conditional alleles of myc (45), and Cdc25A has recently been identified as a transcriptional target of c-Myc (81). However, Cdk2 tyrosine phosphorylation was not detected in our model by phosphoamino acid analysis or by anti-phosphotyrosine antibodies.2 Furthermore, following activation of conditional myc alleles, Cdc25A activates a 120-kDa cyclin E-Cdk2 complex (45), in contrast to our model where the active cyclin E-Cdk2 complexes are of significantly greater molecular mass. This presumably represents major differences in the composition of the active complexes in these two systems, which may reflect different activating mechanisms.
Activation of cyclin E-Cdk2 by decreased cyclin E-Cdk2-CDKI association
in the absence of changes in the levels of these proteins in cell
lysates appears to be a mechanism so far uniquely confined to the
mitogenic action of estrogen. The mechanism is similar to that
described for the activation of cyclin E-Cdk2 in fibroblasts by
platelet-derived growth factor and platelet-poor plasma, which involves
sequestration of free p27 by cyclin D-Cdk4, but these changes occur in
combination with decreased total p27 and increased cyclin E (82). Cdk4
performs a conceptually similar role in the mechanism of cell cycle
arrest by transforming growth factor-, releasing p27 (following
elevation of p15INK4B), which then binds to and inhibits Cdk2
(83). Evidence for a similar mechanism following estradiol treatment
includes increased levels of cyclin D1-associated p21 and p27 at a time
when cyclin E-Cdk2 becomes active and the increased binding of
endogenous p21 and p27 to recombinant cyclin E-Cdk2 following
heat-treatment of cell lysates. Alternatively, it is possible that
estradiol may induce other proteins that prevent cyclin E-Cdk2-CDKI
association, such as a CDKI inhibitor (64). In this regard, it is of
note that active cyclin E-Cdk2 can interact with p107, and inactive cyclin E-Cdk2 can interact with p21, but these interactions are mutually exclusive (84), presumably because p107 and p21 contain similar motifs that are required for binding to cyclin-CDK complexes (85). Therefore, it is possible that interactions between cyclin E-Cdk2
and p107 may contribute to the increase in molecular weight of cyclin
E-Cdk2 complexes with high specific activity following estradiol
treatment.
Several studies have suggested that cyclin E-Cdk2 has critical substrates in addition to pRB. In cells that possess functional pRB, inhibition of cyclin D1 (56, 86), Cdk4/6 (87, 88), or cyclin E (89) prevents G1-S phase progression. However, in cells that lack functional pRB, only inhibition of cyclin E (89, 90) prevents G1-S phase progression. Furthermore, expression of both cyclin D1 and cyclin E has an additive effect on G1 phase progression in fibroblasts (53). Therefore, the early activations of both cyclin D1-Cdk4 and cyclin E-Cdk2 by estrogen are likely to represent two critical and potentially interrelated G1 regulatory events. Further studies are under way to address the relative contribution of each of these cyclin-CDK complexes to G1 progression in this experimental paradigm.
We thank Jenny Hamilton, Andrea Brady, Christine Lee, Douglas Campbell, Alex Swarbrick, Dr. Carsten Schmitz-Peiffer, and Carol Browne for their contribution to some experimental procedures. We are grateful to Dr. Richard Lilischkis for generating recombinant cyclin E and Cdk2 baculoviruses.