Insulin/insulin-like growth factor-I and estrogen cooperate to stimulate cyclin E-Cdk2 activation and cell Cycle progression in MCF-7 breast cancer cells through differential regulation of cyclin E and p21(WAF1/Cip1).

Estrogens and insulin/insulin-like growth factor-I (IGF-I) are potent mitogens for breast epithelial cells and, when co-administered, induce synergistic stimulation of cell proliferation. To investigate the molecular basis of this effect, a MCF-7 breast cancer cell model was established where serum deprivation and concurrent treatment with the pure estrogen antagonist, ICI 182780, inhibited growth factor and estrogen action and arrested cells in G(0)/G(1) phase. Subsequent stimulation with insulin or IGF-I alone failed to induce significant S-phase entry. However, these treatments increased cyclin D1, cyclin E, and p21 gene expression and induced the formation of active Cdk4 complexes but resulted in only minor increases in cyclin E-Cdk2 activity, likely due to recruitment of the cyclin-dependent kinase (CDK) inhibitor p21(WAF1/Cip1) into these complexes. Treatment with estradiol alone resulted in a greater increase in cyclin D1 gene expression but markedly decreased p21 expression, with a concurrent increase in Cdk4 and Cdk2 activity and subsequent synchronous entry of cells into S phase. Co-administration of insulin/IGF-I and estrogen induced synergistic stimulation of S-phase entry coincident with synergistic activation of high molecular mass (approximately 350 kDa) cyclin E-Cdk2 complexes lacking p21. To determine if the ability of estrogen to deplete p21 was central to these effects, cells stimulated with insulin and estradiol were infected with an adenovirus expressing p21. Induction of p21 to levels equivalent to those following treatment with insulin alone markedly inhibited the synergism between estradiol and insulin on S-phase entry. Thus the ability of estradiol to antagonize the insulin-induced increase in p21 gene expression, with consequent activation of cyclin E-Cdk2, is a central component of the synergistic stimulation of breast epithelial cell proliferation induced by simultaneous activation of the estrogen and insulin/IGF-I signaling pathways.

Estrogens are critical for the development and normal physiological function of female reproductive tissues including the mammary gland and uterus (1). Estrogens also play a pivotal role in the initiation and progression of breast cancer, where their mitogenic properties are thought to play a causative role in the disease process (2). Although the mechanisms through which estrogens stimulate cell proliferation are becoming increasingly well understood (3)(4)(5)(6)(7), there is compelling evidence that several mitogenic growth factors, particularly those of the epidermal growth factor and IGF 1 families, interact with ERmediated signaling to regulate cell proliferation in target tissues (8,9). The molecular basis of these interactions is currently of major research interest (10 -12), but the mechanisms remain to be fully defined.
In breast epithelial cells, insulin (13) and IGF-I (14) are potent mitogens and synergize with estrogen to stimulate cell proliferation (15,16). Several potential mechanisms have been proposed to account for these effects. Estrogens alter the expression of IGF ligands, receptors, and binding proteins, suggesting potential autocrine and paracrine mechanisms for estrogen-stimulated mitogenesis (17). By up-regulating IGF-IR, estrogen can sensitize MCF-7 cells to the mitogenic effects of high concentrations of insulin (acting through the IGF-IR) and IGF-I (16,18). Other studies have shown that estrogens may impinge directly on IGF-I signal transduction downstream of the IGF-IR.
Ligand stimulation of the IGF-IR activates its tyrosine kinase activity, leading to phosphorylation of key intracellular substrates including insulin receptor substrate proteins 1-4 (19). Phosphorylated insulin receptor substrate-1 acts as a docking site for several SH2-containing proteins including the p85 subunit of phosphatidylinositol 3-kinase and Grb2 (20). These molecules in turn link IGF-IR to specific pathways controlling cellular proliferation, i.e. the phosphatidylinositol 3-kinase-Akt/protein kinase B and the Ras-Raf-mitogen-activated protein kinase pathways. In MCF-7 cells, proliferation in response to IGF-I appears to be mediated predominantly via the phosphatidylinositol 3-kinase pathway (21). Estrogen can activate insulin receptor substrate-1 by increasing gene expression and protein phosphorylation (12,22) and stimulates the formation of IGF-IR-insulin receptor substrate-1-phosphatidylinositol 3-kinase complexes (23). Ligand-bound ER␣, but not ER␤, can activate IGF-I signal transduction via a rapid nongenomic effect mediated by direct interaction of ER␣ with the IGF-IR. This interaction appeared dependent on ER␣ phosphorylation by ERK1/2 and resulted in IGF-IR phosphorylation and activation of the Ras-Raf-mitogen-activated protein kinase cascade (11). IGF-I and estradiol can also act synergistically to increase Akt protein expression and enzymatic activity (24). Finally, direct activation of p21 ras -mitogen-activated protein kinasesignaling pathways by estrogen has also been proposed (25).
In addition to estrogen regulation of components of insulin/ IGF-I-signaling pathways there is also evidence that polypeptide growth factors acting via type I receptor tyrosine kinases modulate estrogen action. Ligands for both the epidermal growth factor receptor and IGF-IR can regulate ER gene expression (26,27) and ER-mediated transcription (28) in an estrogen-independent manner. Studies in which functional domains of ER were deleted revealed a requirement for the activation function-1 domain in ligand-independent activation of ER by epidermal growth factor (29). Furthermore, ER function can be modulated by mitogen-activated protein kinase phosphorylation of Ser-118 within the activation function-1 domain after epidermal growth factor (30) or insulin/IGF-I (31) stimulation. This specific phosphorylation event provides one potential mechanism for growth factor activation of ER signaling.
The data outlined above provide evidence for mechanisms by which estrogen and insulin/IGF-I-signaling pathways are cross-modulated. However, their mitogenic effects are ultimately mediated at the level of regulation of kinases that govern transition from G 0 /G 1 to S phase of the cell cycle, i.e. the G 1 cyclin-dependent kinases (CDKs) cyclin D-Cdk4/6 and cyclin E-Cdk2 (32,33). Several of these cell cycle regulatory molecules have been identified previously as targets of insulin/IGF-I (34) or estrogen-induced mitogenesis in breast cancer cells (3)(4)(5)7). These mitogenic signals are thought to converge on cyclin D-Cdk4/6 since activity of this kinase is required for either serum or estrogen-induced mitogenesis (35). Insulin or IGF-I stimulate cyclin D1, D3, and E gene expression (34), in common with other mitogens in a spectrum of cell types. In MCF-7 cells, increased cyclin D1 gene expression and stabilized cyclin D1 mRNA in response to IGF-I is mediated via the phosphatidylinositol 3-kinase pathway (36). Estrogen also stimulates cyclin D1 gene expression and subsequent activation of cyclin D1-Cdk4 (3,5) as well as inducing early activation of cyclin E-Cdk2. However, unlike the situation observed after growth factor stimulation, this is not preceded by a major increase in cyclin E gene expression (5). Rather, estrogen-induced Cdk2 activation involves a novel mechanism wherein a small fraction of the total cyclin E-Cdk2 pool is present in a high molecular mass complex depleted of the CDK inhibitor, p21 WAF1/Cip1 (5,7). The known synergistic interactions between insulin/IGF-I and estrogen in stimulating breast cancer cell proliferation may, thus, converge at both cyclin D-Cdk4/6 and cyclin E-Cdk2, although this has not been investigated to date.
In this study we investigated the cell cycle regulatory events accompanying synergistic interactions between estrogen and insulin/IGF-I in mediating cell cycle progression in MCF-7 cells. Cells were initially arrested in a quiescent state by serum withdrawal and treatment with the pure estrogen antagonist, ICI 182780 (37). Subsequent treatment with insulin/IGF-I and estradiol resulted in synergistic stimulation of S-phase entry. Analysis of the expression and function of cyclin D1-Cdk4 and cyclin E-Cdk2 during the G 0 /G 1 -S-phase transition revealed cooperative regulation by insulin/IGF-I and estrogen of several molecules regulating these kinases. Of primary significance was the observation of synergy at the level of cyclin E-Cdk2 activation as a likely key mediator of the synergistic cell cycle progression.
To investigate the effects of insulin and estradiol on cell cycle progression, cells were allowed to proliferate for 2 days, after which the medium was removed, and cells were washed twice with phosphatebuffered saline. Medium was replaced with a defined serum-free phenol red-free RPMI 1640 medium supplemented with transferrin (24 g/ml) and gentamicin (10 g/ml). This medium was changed once daily for 2 days. After 48 h of serum deprivation in this defined medium, ICI 182780 was added directly to the medium to a final concentration of 10 nM. After 24 h of ICI 182780 pretreatment, vehicle, estradiol (100 nM), insulin (10 g/ml), or IGF-I (10 nM) was added directly to the medium, and after incubation for the desired time, cells were harvested by brief incubation with trypsin (0.05% w/v), EDTA (0.02% w/v). Cells were prepared for analytical DNA flow cytometry as previously described (39). DNA flow cytometry was performed on a FACSCalibur laser-based flow cytometer (Becton Dickinson, Immunocytometry Systems, Mountview, CA), and cell cycle-phase distribution was analyzed with a software program Modfit (Becton Dickinson, Immunocytometry Systems, Mountview, CA).
Cyclin-dependent Kinase Assays-MCF-7 cell monolayers were washed twice with phosphate-buffered saline then scraped into ice-cold phosphate-buffered saline and pelleted by centrifugation (15,000 ϫ g, 5 min). For Cdk4 activity assays, the pellets were frozen in liquid nitrogen for storage at Ϫ70°C. Immediately before assaying, the pellets were thawed and resuspended in 1 ml of ice-cold lysis buffer B (50 mM HEPES, pH 7.5, 1 mM DTT, 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 sodium orthovanadate, 10 g/ml leupeptin, 10 g/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 then centrifuged at 15,000 ϫ g for 5 min at 4°C. For cyclin E-Cdk2 activity assays, the pellet was resuspended in 1 ml of ice cold lysis buffer A and frozen in liquid nitrogen for storage at Ϫ70°C. The lysate was placed on ice, vortexed vigorously at intervals for 10 min, and centrifuged at 15,000 ϫ g for 5 min at 4°C, and the supernatant was collected.
Equivalent amounts of protein were precleared by incubation with protein A-Sepharose (Zymed Laboratories Inc., San Francisco, CA) for 45 min at 4°C. Cdk4 or cyclin E complexes were immunoprecipitated with either rabbit polyclonal human Cdk4 (H-22) or anti-cyclin E polyclonal antibody (C-19), respectively (Santa Cruz Biotechnology Inc), for 2 h at 4°C. Similarly, control samples were immunoprecipitated with antibodies preincubated with the respective Cdk4 or cyclin E peptides (Santa Cruz Biotechnology Inc.), also for 2 h at 4°C. This was followed by a 30-min incubation at 4°C with protein A-Sepharose for conjugation to the antibodies. Cdk4 immunoprecipitates were washed twice with ice-cold lysis buffer B, twice with ice-cold 50 mM HEPES pH 7.5, and 1 mM DTT, and the supernatant was aspirated. The immunoprecipitates were used for kinase assays with glutathione S-transferase-pRb 773-928 (generated from a plasmid kindly supplied by Dr. Ed Harlow, Massachusetts General Hospital Cancer Center, Charlestown, MA) as substrate. Cyclin E immunoprecipitates were washed twice with icecold lysis buffer A containing 1 M NaCl, once with ice-cold lysis buffer A, and then twice with ice-cold 50 mM HEPES, pH 7.5, 1 mM DTT. Histone H1 (Sigma) was employed as the substrate for the kinase activity assays on cyclin E immunoprecipitates.
Detection of Cdk4-, cyclin E-, and p130-associated Proteins-Immunoprecipitation of Cdk4, cyclin E, and p130 was performed using the method described above (for immunoprecipitating cyclin E and Cdk4 for kinase activity assays), except that the cyclin E and p130 antibodies were chemically cross-linked to protein A-Sepharose to reduce background (40). The antibodies used were rabbit polyclonal antisera to human Cdk4 (H-22), human cyclin E (C- 19), and human p130 (C-20) from Santa Cruz Biotechnology Inc. The supernatants obtained after immunoprecipitation were Western-blotted to determine immunoprecipitation efficiency of each antibody, with a 90 -95% efficiency consistently achieved. The immunoprecipitated proteins were resuspended in 1ϫ SDS sample buffer, separated by SDS-PAGE, and transferred to nitrocellulose membrane, and the proteins were detected using the antibodies described for Western-blotting above.
Gel Filtration Chromatography-Cell lysates prepared in lysis buffer A were passed through a 0.22-mm MILLEX-GV4 filter (Millipore, Lane Cove, NSW, Australia) and fractionated on a HiLoad 16/60 Superdex 200 column (Amersham Pharmacia Biotech) using a fast protein liquid chromatography system (Amersham Pharmacia Biotech). Proteins were eluted at 1.2 ml/min at 4°C in the following buffer: 20 mM HEPES, pH 7.5, 250 mM NaCl, 1 mM EDTA, 0.1% (v/v) ␤-mercaptoethanol, and 0.01% (v/v) Tween 20. The column void volume was ϳ45 ml and 16 3-ml fractions were collected between 45 and 93 ml (termed fractions 16 -31). Column calibrations were performed under the same conditions using a high molecular mass gel filtration calibration kit (Amersham Pharmacia Biotech) containing ferritin (440 kDa), catalase (232 kDa), and aldolase (158 kDa). The eluted proteins was concentrated before Western blotting, whereby 200 l of each fraction was placed at Ϫ70°C overnight with 1 ml of acetone and 10 g of carrier bovine serum albumin. Protein pellets were collected by centrifugation (15,000 ϫ g, 20 min, 4°C) and then resuspended by boiling for 4 min in 30 l of SDS sample buffer. For cyclin E-Cdk2 kinase assays, 300 l of each fraction was used. For Western blotting of cyclin E immunoprecipitates, 700 l of each fraction was used.
Adenoviral p21 Infection-The human p21 adenovirus (Ad-p21⌬C) was obtained from Dr. Joseph Nevins (Duke University Medical Center, Durham, NC) and amplified in HEK293 cells. Virus was purified by CsCl density gradient centrifugation, and titers were determined by a focus-forming assay using standard techniques (41). MCF-7 cells were rescued from growth arrest with insulin, estradiol, or both as described above and infected with Ad-p21 or control adenovirus at matched titers from 4 -24 h. Cells were harvested 24 h after mitogenic rescue for protein analysis or analytical DNA flow cytometry.

Synergistic Effects of Estradiol and Insulin/IGF-I on G 1 -Sphase Progression-Previous
studies have shown that treatment of exponentially growing MCF-7 cells with ICI 182780, a pure estrogen antagonist that inhibits estrogen-mediated gene transcription (42), arrests cells in the G 0 /G 1 phase of the cell cycle (37,43). Cell cycle progression is re-initiated after the addition of estradiol to ICI 182780-treated cells in serumsupplemented medium (5). In the present study this model was modified to examine the interaction of insulin/IGF-I and estrogen in mediating cell cycle progression in a serum-free environment. Exponentially growing MCF-7 cells were growtharrested by serum deprivation for 72 h to inhibit growth factorsignaling pathways with concomitant treatment for the final 24 h with ICI 182780 to inhibit estrogen action. This pretreatment reduced S phase from 40 to 3-5% and led to accumulation of quiescent cells (37). As is evident in Fig. 1A, stimulation of arrested cells with insulin (10 g/ml) or IGF-I (1 nM) alone resulted in a minor increase in S-phase cells to ϳ10 -12% at 26 h. Insulin dose-response experiments in serum-free medium without ICI 182780 indicated a maximal response between 100 ng/ml and 10 g/ml insulin (data not shown), and a similar maximal response was achieved with 1 nM IGF-I. A supraphysiological concentration (10 g/ml) of insulin, known to activate both insulin and IGF-I receptor signaling in breast epithelial cells (44,45), was used in all subsequent experiments, in agreement with the design of earlier studies demonstrating synergistic effects on breast epithelial cell proliferation (15,16,46). In marked contrast, estradiol (100 nM) alone stimulated a significantly greater S-phase entry of 43%. However, when administered together, insulin/IGF-I and estradiol stimulated the greatest G 1 -S progression, with a maximum 70% S phase after 26 h. Also evident in Fig. 1A was the more highly synchronous entry of cells into S phase after co-stimulation with insulin/ IGF-I and estradiol when compared with estradiol alone.
A time course of S-phase entry is presented in Fig. 1B. The first significant increase in the proportion of cells in S phase was apparent 14 h after estradiol treatment alone or in combination with insulin. The S-phase fraction reached a maximum between 24 and 26 h and declined thereafter, coincident with a rise in G 2 /M phase (data not shown). Interestingly, S-phase entry stimulated by insulin and estradiol was significantly greater than the sum of the two individual S-phase responses of either insulin or estradiol alone at all time points between 14 and 28 h. Together these data demonstrate a synergistic effect of insulin/IGF-I and estradiol on MCF-7 cell cycle progression and characterize an experimental paradigm for further elucidating potential molecular mechanisms responsible for these effects.

Effects of Estradiol and Insulin on Pocket Protein
Phosphorylation and Expression-Changes in the abundance and phosphorylation state of the pocket proteins, pRb, p130, and p107, accompany different stages of G 0 -to S-phase transition (32). In diverse cell types, pRb and p130 are predominantly hypophos-  Fig. 1 were harvested at the time points indicated, and whole cell lysates were prepared, subjected to electrophoresis on 7% SDS-PAGE, and Western-blotted for pRb, p107 and p130 as described under "Experimental Procedures." B, graphical presentation of total protein levels after mitogenic stimulation. C, the ratio of hyperphosphorylated pRb and p130 to total protein is represented graphically. ‚, vehicle control; OE, insulin; Ⅺ, estradiol; and f, insulin plus estradiol. This experiment is representative of data derived from three independent experiments. phorylated in arrested cells and hyperphosphorylated during G 1 -S transition (32). In agreement with these findings, pRb and p130 ( Fig. 2A) existed predominantly in their hypophosphorylated states in growth-arrested MCF-7 cells (Fig. 2C). Insulin stimulation resulted in only a minor increase in the phosphorylation of pRb and p130, whereas estradiol induced a significant increase in the phosphorylation of both proteins, as evidenced by the predominance of hyperphosphorylated forms after 15-20 h. Stimulation of cells with both insulin and estradiol had the most profound effect on the degree of pRb and p130 phosphorylation (Figs. 2, A and C). The hyperphosphorylated form was evident as early as 10 -15 h after combined treatment, and subsequently, the vast majority of total pRb and p130 was in the hyperphosphorylated state. Although differentially phosphorylated species of p107 were not discernable in these experiments, the abundance of this protein was low in arrested cells, increased slightly by insulin treatment, and induced 4-fold by estrogen treatment (Fig. 2B). There appeared to be no further increase after insulin and estrogen treatment. A similar pRb protein expression profile was also observed. In contrast, p130 protein levels were increased at 5-10 h by all treatments and then decreased to control levels (Fig. 2B). Together these data demonstrate that insulin and estradiol differentially regulate pocket protein abundance and phosphorylation (pRb, p130), in accordance with their differential effects on S-phase entry.
Effects of Estradiol and Insulin on Activation of Cdk4 and Cdk2-Since changes in pocket protein phosphorylation are mediated predominantly by the G 1 CDKs (33), we examined effects on cyclin D1-Cdk4 and cyclin E-Cdk2 activities. After serum deprivation and anti-estrogen pretreatment, Cdk4 activity in vehicle-treated control cells was low and continued to decline over 25 h to levels indistinguishable from those seen in the peptide-blocked controls (Fig. 3A). After insulin stimulation, Cdk4 activity was increased significantly between 5 and 15 h and then declined to control levels by 25 h (Figs. 3, A and  C). Estradiol stimulated Cdk4 activity between 5 and 25 h. Co-stimulation with insulin and estradiol resulted in a gradual increase in Cdk4 activity to reach a maximum ϳ2-fold increase by 10 h, thereafter remaining elevated. These data suggest that both insulin and estrogen stimulate a modest increase in Cdk4 activity, although there was no evidence of synergy.
In arrested cells cyclin E-Cdk2 activity was essentially undetectable (Fig. 3, B and C). After insulin stimulation, cyclin E-Cdk2 activity increased at 10 h, reaching a maximum 2-fold at 15-20 h, after which it returned to control levels by 25 h. An initial 2-fold increase was evident at 5 h after estradiol stimulation, with a further increase to a 5-fold maximum at 25 h. Stimulation of cells with estradiol and insulin together resulted in the most dramatic effect on cyclin E-Cdk2 activation; a 3-fold activation was observed as early as 5 h, and cyclin E-Cdk2 activity continued to increase to an 8-fold maximum at 20 h, thereafter declining to 4-fold at 25 h as cells began to exit S phase. Thus, insulin and estradiol together result in synergistic activation of cyclin E-Cdk2, and this precedes S-phase entry by at least 5 h (compare Figs. 3C and 1B).
Effects of Estradiol and Insulin on Cyclin D1-Cdk4 Complexes-CDK activity is determined predominantly by the com -FIG. 3. Activation of cyclin D1-Cdk4 and cyclin E-Cdk2 after insulin and/or estradiol stimulation. MCF-7 cells treated according to the protocol outlined in Fig. 1 were harvested at the time points indicated and whole cell lysates were prepared and immunoprecipitated with anti-Cdk4 (A) or anti-cyclin E antibodies (B). The immunoprecipitates were assayed for Cdk4 activity and cyclin E-associated kinase activity as described under "Experimental Procedures." Autoradiography of cyclin D1-Cdk4 (A) and cyclin E-Cdk2 kinase assays (B). C, graphical presentations of changes in cyclin D1-Cdk4 and cyclin E-Cdk2 kinase activities after mitogenic stimulation. OE, insulin; Ⅺ, estradiol; and f, insulin plus estradiol. These experiments are representative at least three independent experiments. position of the cyclin-Cdk complexes and in turn by the levels of constituent cyclins, CDKs and CDK inhibitors (33). We studied both total protein expression and cyclin-Cdk complex composition by Western blot analysis. In growth-arrested cells, protein levels of cyclin D1 and p21 were low in both whole cell lysates (Fig. 4A) and Cdk4 complexes (Fig. 4B), consistent with the lack of Cdk4 activity (Fig. 3A). Insulin stimulation resulted in a 3-fold induction of cyclin D1 protein expression, apparent as early as 5 h and coincident with a 4-fold induction of p21 levels (Fig. 4A). This resulted in an initial maximum 3-fold increase in Cdk4-associated cyclin D1 and p21 at 5-10 h, which declined by 50% after 20 h (Fig. 4B). This increased complex formation corresponds with the insulin-induced peak of Cdk4 activity noted between 5 and 15 h (Figs. 3, A and C). Estradiol stimulated a 5-fold increase in total cyclin D1 but decreased total p21 levels dramatically such that they were essentially undetectable after 15 h (Fig. 4A). Despite an increase in total cyclin D1, estradiol induced only a minor increase in cyclin D1 association with Cdk4, presumably due to the lack of availability of p21 as an assembly factor (47). A pronounced 8-fold induction of cyclin D1 by co-treatment with insulin and estradiol was accompanied by a biphasic effect upon p21. Levels of p21 were initially elevated 4-fold at 5 h and thereafter gradually declined (Fig.  4A). Cyclin D1 and p21 association with Cdk4 was also biphasic, an initial increase coincident with an increase in Cdk4 activity, declining to control levels by 25 h despite elevated Cdk4 activity (Fig. 3C and 4B). The initial event is likely due to insulin induction of p21, and the latter, to estradiol inhibition of p21 gene expression. These results demonstrate that insulin and estradiol have differential effects on cyclin D1-Cdk4 assembly, and this is likely mediated in large part through differential effects on total cellular p21 protein levels.

Effects of Estradiol and Insulin on Cyclin E-Cdk2
Complexes-In addition to cyclin D1-Cdk4, cyclin E-Cdk2 also plays a pivotal role in mediating G 1 -to S-phase progression (33). Total protein levels of both p27 (Fig. 4A) and Cdk2 (data not shown) were not significantly regulated by insulin or estradiol stimulation. However, cyclin E and p21 were differentially regulated. In control cells, low cyclin E, Cdk2 (data not shown), and p21 protein levels were evident in both whole cell lysates (Fig. 4A) and cyclin E complexes (Fig. 5A). The amount of p27 and p21 associated with cyclin E remained constant over 25 h. Cyclin E and p21 were consistently up-regulated by insulin to ϳ3and 4-fold above control, respectively (Fig. 4A). Insulin also caused a profound increase in p21 association with cyclin E-Cdk2, first apparent at 10 h and increasing through to 25 h (Fig. 5A). Estradiol had only minor effects on cyclin E protein expression but significantly decreased p21 expression. Downregulation of p21 resulted in decreased association with cyclin E to almost undetectable levels by 15 h, consistent with increasing cyclin E-Cdk2 activity (Figs. 3, B and C). Co-stimulation with insulin and estradiol revealed a 2-fold up-regulation of cyclin E, which declined after 20 h (Fig. 4A). This was associated with a 2-fold increase in the amount of p21 associated with cyclin E-Cdk2 at 5 and 10 h (Fig. 5A). Thereafter, both total p21 and its association with cyclin E-Cdk2 gradually declined to undetectable levels by 25 h. These changes are consistent with increasing cyclin E-Cdk2 activity (Fig. 3, B and  C). Furthermore, examination of the ratio of p21 to cyclin E revealed a time-dependent decline in the amount of p21 bound to cyclin E-Cdk2 complexes after estradiol stimulation alone or in combination with insulin, but there was little difference between these two treatments (Fig. 5B).
In marked contrast to the differential effects on p21 gene expression and CDK complex formation, neither insulin nor estradiol had major effects on the total cellular levels of p27 or its accumulation into cyclin D1-Cdk4 or cyclin E-Cdk2 complexes (Figs. 4 and 5). The only exception was at late time points, significantly after S-phase entry, where estradiol alone or in combination with insulin decreased p27 recruitment into cyclin E-Cdk2 complexes. These data imply that p27 plays a relatively minor role in regulating the early cell cycle events responsible for synergism between insulin/IGF-I and estradiol.
Insulin and Estradiol Act Synergistically to Induce and Activate High Molecular Mass Cyclin E-Cdk2 Complexes-Since our previous work has shown that estradiol treatment results in the formation of high specific activity, high molecular mass cyclin E-Cdk2 complexes that represent only a minority of the Fig. 1 were harvested at the time points indicated, and whole cell lysates were prepared for Western blot (A) and immunoprecipitation with an anti-Cdk4 antibody as described under "Experimental Procedures" (B). The samples were subjected to electrophoresis on 12% SDS-PAGE and Western-blotted for cyclins D1 and E and CDK inhibitors p27 and p21 (A) and cyclin D1, p27, and p21, as indicated (B). Graphical presentations of total protein levels (A) and Cdk4-associated proteins (B) are shown below the Western blots. ‚, vehicle control; OE, insulin; Ⅺ, estradiol; and f, insulin plus estradiol. This experiment is representative of three independent experiments. total cellular cyclin E pool (5), we investigated the effects of insulin/IGF-I on the formation of such complexes 10 h after mitogenic stimulation. The major peak of cyclin E (Fig. 6A, fractions 23-24) eluted from the gel filtration column at ϳ160 kDa and lacked significant cyclin E-Cdk2 activity (Fig. 6B). The majority of activity was associated with the higher molecular mass 350-kDa cyclin E-Cdk2 complexes (Fig. 6B, fraction  20). Insulin stimulated a ϳ2-fold increase in cyclin E-Cdk2 activity despite very low levels of cyclin E in the high molecular mass fractions (Fig. 6A, fractions 17-22). In agreement with earlier findings (5), there was a small but consistent increase in cyclin E migrating as higher molecular mass complexes after estradiol stimulation (Fig. 6A, fractions 18 -22), coincident with a maximum 4-fold increase in cyclin E-Cdk2 activity (Fig.  6B). Furthermore, co-stimulation with estradiol and insulin induced a 5-fold increase in this high molecular mass cyclin E complex (Fig. 6A, fractions 19 -22), and this was associated with a maximal 8-fold increase in cyclin E-Cdk2 activity. Thus the addition of insulin/IGF-I to estradiol resulted in the synergistic formation and activation of high molecular mass cyclin E-Cdk2 complexes.

MCF-7 cells treated according to the protocol outlined in
Active 350-kDa Cyclin E-Cdk2 Complexes Are Accompanied by Increased p130 Association-Since the high molecular mass cyclin E-Cdk2 complexes formed after estradiol treatment lacked p21 (5), the elution profiles of both p21 and p27 were investigated in the same experiments. Both CDK inhibitors eluted predominantly in a complex of ϳ160 kDa, coincident with the major peak of cyclin E protein (Fig. 6, A, C, and D). Not unexpectedly, these complexes showed little or no cyclin E-Cdk2 activity. Additionally, in control-and insulin-treated cells, some p21 and p27 molecules (in the case of p27, constituting a significant proportion of the total protein) eluted at low molecular mass, i.e. ϳ70 kDa (Fig. 6C, fractions 28 -32 and Fig.  6D, fractions 26 -30). The presence of these species presumably indicates cyclin E-Cdk2 complex saturation by p21 and/or p27. These low molecular mass forms of p21 and p27 were not detected after estradiol stimulation or with estradiol and insulin, indicating that after these treatments, all the p21 and p27 was bound in higher molecular mass complexes.
The higher molecular mass of the active cyclin E complexes is due in part to p130 association (6). To investigate whether the increased complex formation induced by estradiol and insulin treatment was due to increased association with p130, we immunoprecipitated p130 from the eluted fractions and Western-blotted for cyclin E. Cyclin E was associated with p130 in the active high molecular mass 350-kDa complexes (Fig. 7A) but not in the inactive 160-kDa complexes (Fig. 7B). In addition, a greater degree of cyclin E association with p130 was induced by insulin and estradiol stimulation compared with estradiol alone. This is in marked contrast to control and insulin-stimulated cells, where there was little or no detectable cyclin E association with p130. These data demonstrate that an increased association of cyclin E with p130 is correlated with the synergistic activation of cyclin E-Cdk2 complexes.
Adenoviral Expression of p21 Inhibits Synergism between Insulin and Estrogen on Cell Cycle Progression-Differential induction of p21 by insulin and estrogen (Figs. 4A and 6C) resulted in differential regulation of the cyclin D1-Cdk4 and cyclin E-Cdk2 complexes (Figs. 4B and 5A). The synergistic activation of cyclin E-Cdk2 (Figs. 3B and 6B) by insulin and estradiol appears to depend on the estrogen-induced downregulation of p21 (Fig. 5A) with consequent loss of p21 and accumulation of p130 in high molecular mass active cyclin E-Cdk2 complexes (Figs. 6 and 7). To investigate more directly the possible role of p21 in the synergism between insulin and estradiol, we first attempted to mimic the effects of estradiol by decreasing p21 levels using antisense oligonucleotides to the p21 mRNA. Unfortunately these experiments were unsuccessful, as MCF-7 cells that had undergone serum deprivation and ICI 182780 treatment could not tolerate the further manipulations required to deliver the p21 antisense oligonucleotides to the cell.
In an alternative approach, cells treated with insulin and estradiol were infected with an adenovirus expressing p21 to restore p21 levels to those observed after treatment with insulin alone. In agreement with earlier experiments, insulin induced p21 levels ϳ4-fold at 24 h (compare Figs. 4A and 8A) and a hypophosphorylated form of pRb (compare Figs. 2A and 8B). This was consistent with a low (8%) S-phase fraction (compare Figs. 1A and 8C) in uninfected cells or those infected with a control adenovirus lacking p21. Similarly, low levels of p21 were seen in estradiol plus insulin-stimulated cells regardless of infection with control adenovirus (Fig. 8A), and this was accompanied by a synergistic 51% S-phase entry after 24 h (Fig. 8C). This synergistic S-phase fraction was coincident with the appearance of pRb in a predominantly hyperphosphorylated state (Fig. 8B). Infection of insulin plus estradiol-stimulated cells with the p21 adenovirus resulted in the expression of a faster migrating p21 species (ϳ19 kDa), distinct from the endogenous p21. This p21 adenovirus contains a truncated human p21 cDNA that lacks the C-terminal 21 amino acids but preserves the CDK and cyclin binding domains sufficient for full CDK inhibitory activity (48,49). Infection, which led to increases in p21 protein to a maximum level almost equivalent FIG. 5. Differential association of CDK inhibitors p27 and p21 with cyclin E-Cdk2 complexes after stimulation with insulin and/or estrogen. MCF-7 cells treated according to the protocol outlined in Fig. 1 were harvested at the time points indicated, and whole cell lysates were prepared and immunoprecipitated (IP) with an anti-cyclin E antibody. The immunoprecipitates were subjected to electrophoresis on 12% SDS-PAGE and Western-blotted (A) for cyclin E, p27, and p21. B, graphical representation of the ratio of p21 to cyclin E after quantitation of A by densitometry. ‚, vehicle control; OE, insulin; Ⅺ, estradiol; and f, insulin plus estradiol. This experiment is representative of 3 independent experiments.
to that induced by insulin alone, resulted in inhibition of insulin plus estradiol-stimulated cell cycle progression as reflected in the decreased S-phase fraction (Fig. 8C). Infection with 12 ϫ 10 8 plaque-forming units/ml p21 adenovirus induced p21 levels 4-fold and reduced the S-phase fraction from 51 to 27%. This reduction in the S-phase fraction was coincident with a decrease in the hyperphosphorylated form of pRb (Fig. 8B). These data demonstrate that modulation of p21 levels inhibits the synergistic effects of insulin and estradiol on cell cycle progression, identifying a pivotal role for p21 in the cooperative interaction underlying the synergy between insulin/IGF-I and estradiol.

DISCUSSION
The importance of estrogen action in the regulation of breast cancer cell proliferation has been demonstrated both in vitro and in vivo (50). Epidemiological studies demonstrate a causative role for estrogen exposure in the etiology of breast cancer (2), whereas the clinical efficacy of oophorectomy and pharmacological agents that inhibit the synthesis or action of estrogen further emphasize the fundamental importance of estrogeninduced mitogenesis in breast cancer (51). It is now evident, however, that estrogens do not exert their mitogenic actions alone but act in concert with other hormones, growth factors, and cytokines to regulate breast epithelial cell proliferation.
One such family of mitogenic growth factors is the IGFs. These molecules play a pivotal role in the normal regulation of cell proliferation, differentiation, and apoptosis, and it is, thus, not unexpected that abnormalities in the regulation and function of IGF ligands, binding proteins, and receptors have been demonstrated in breast cancer (52). Although it is evident that insulin/IGF-I and estrogen act synergistically to stimulate breast cancer cell proliferation in vitro (15,16), the underlying molecular mechanisms remain to be fully defined. Several studies have demonstrated estrogen regulation at different levels of the IGF signal transduction cascade, as outlined in the Introduction, and these provide possible mechanisms for synergism. Since stimulation of these signaling pathways ultimately leads to cell proliferation in breast cancer cells, this study focused on the potential synergy between insulin/IGF-I and estrogen on the regulation of downstream targets governing G 0 /G 1 -S-phase progression.
The data presented here identify differential effects of estrogen and insulin/IGF-I on cell cycle targets that control S-phase entry, defining new molecular mechanisms of synergy. Using a combination of serum starvation and treatment with an estrogen antagonist, we arrested MCF-7 cells in an apparent quiescent, G 0 state (37). In marked contrast to cells that had been growth-arrested by serum deprivation alone, where addition of insulin/IGF-I re-initiated cell cycle progression (34), only a minority of MCF-7 cells entered the S phase after insulin/IGF-I treatment in the presence of anti-estrogen (Fig. 1). It has previously been suggested that anti-estrogens can inhibit IGF-IR signaling (53), and although in the absence of data comparing responses in our model with cells growth-arrested by serum deprivation alone the possibility that anti-estrogen pretreatment partially attenuates IGF-IR signaling cannot be excluded, this is unlikely to be the explanation. In our model  Fig. 1 were harvested at 10 h, and whole cell lysates prepared and subjected to separation by gel filtration chromatography as described under "Experimental Procedures." Active (19,20) and inactive (23,24) fractions (determined from Fig. 6B) were immunoprecipitated with an anti-p130 antibody. The immunoprecipitates were subjected to electrophoresis on 12% SDS-PAGE and Western-blotted for cyclin E. Control sample was whole cell lysates of MCF-7 that had been stimulated with insulin and estradiol for 10 h. This experiment is representative of three independent experiments. treatment with insulin/IGF-I alone increased cyclin D1, cyclin E, and p21 expression (Fig. 4), a pattern of expression typical of the proliferative response to diverse mitogenic growth factors. Furthermore, active p21-associated cyclin D1-Cdk4 complexes were formed (Figs. 3 and 4). These data provide convincing evidence of an intact IGF-IR mitogenic-signaling pathway following anti-estrogen pretreatment and imply that the ability of anti-estrogens to block insulin/IGF-I-induced mitogenesis in breast cancer cells occurs predominantly downstream of Cdk4 activation.
The most likely explanation for the inability of insulin/IGF-I to induce cell cycle progression is the inability to activate cyclin E-Cdk2 complexes, an essential requirement for cell cycle progression (33). Some insight into reasons for the failure of insulin to activate cyclin E-Cdk2 in anti-estrogen-pretreated cells can be gained from examination of anti-estrogen effects on cyclin-CDK complex composition and activity. An integral component of the growth inhibitory response of ICI 182780 is decreased cyclin D1 gene expression (43) and the consequent rapid redistribution of p21 from cyclin D1-Cdk4 to cyclin E-Cdk2 complexes, inhibiting the activity of the latter kinase (5,37). This is followed at later time points by a 4 -5-fold increase in total cellular p21, a modest increase in p27 (37,43), and the accumulation of inactive cyclin E-Cdk2 complexes loaded with p21 and p27 inhibitors (5,37). Thus, although the induction of cyclin D1 and cyclin E by insulin/IGF-I might be expected to change the dynamics of CDK complex formation through the sequestration of p21 and p27 into newly formed cyclin D1-Cdk4 complexes and hence facilitate the formation of active cyclin E-Cdk2 complexes, the concurrent increased p21 gene expres-  Fig. 1 were subsequently infected with the indicated amount of either p21 expressing or control adenovirus 4 h after insulin (I) or insulin and estradiol (I ϩ E) treatment. These cells were harvested at 24 h, and whole cell lysates were prepared, subjected to electrophoresis on 12% SDS-PAGE, and Western-blotted for p21 (A) and pRb (B) as described under "Experimental Procedures." C, cells treated according to the above protocol were harvested after a 24-h stimulation for DNA analysis by flow cytometry. This experiment is representative of at least three independent experiments. pfu, plaque-forming units.

FIG. 9. Proposed model of synergy between insulin and estrogen in stimulating G 1 -S-phase progression.
Graphical representation of the proposed molecular mechanisms described in the text underlying the regulation of cyclin D1-Cdk4, cyclin E-Cdk2 activity by insulin and estrogen. See "Discussion" for details. LMW, low molecular mass. sion favors maintenance of inactive cyclin E-Cdk2. Indeed it is evident from the gel filtration data presented in Fig. 6 that after stimulation with insulin alone, a significant proportion of both p21 and p27 remains in a state unbound to cyclin E-Cdk2, apparently in excess of that required to inhibit the majority of the cyclin E-Cdk2 complexes.
We and others have previously shown that anti-estrogenarrested cells can be stimulated by estrogens to reenter the cell cycle in a semi-synchronous manner (5). Numerous lines of evidence indicate that cyclin E-Cdk2 activation is critical to G 1 -S-phase progression (33). Cyclin E-Cdk2 is activated by estradiol in MCF-7 cells several hours before S-phase entry (4, 5, 7), providing a potential mechanism for estrogen-mediated early cell cycle progression. In the current model, estrogen stimulated cyclin E-Cdk2 activity in the absence of significant increases in cyclin E or Cdk2 protein levels, a result in agreement with previous observations in a serum-containing medium (5). This is in marked contrast to previous studies in fibroblasts where mitogenic activation of cyclin E-Cdk2 resulted predominantly from increases in cyclin E gene expression (54,55).
The mechanism of activation of cyclin E-Cdk2 by estrogen has been investigated in some detail (5,7) and involves at least two distinct mechanisms. In the acute phase of the response no detectable changes in the total cellular levels of the major components of the cyclin E-Cdk2 complex, i.e. cyclin E, Cdk2, p21, and p27, were apparent, but significant changes occurred in association of the CDK inhibitors with CDK complexes. Thus, as a result of estrogen-induced cyclin D1 gene expression, p21 and to a lesser extent p27 are sequestered into cyclin D1-Cdk4 complexes (5,7). In addition, estrogen inhibits p21 gene transcription, rapidly depleting the pool of newly synthesized p21 and facilitating the formation of active cyclin E-Cdk2 complexes lacking p21. 2 When both potential pathways were inhibited by the simultaneous inhibition of estrogen-induced cyclin D1 gene expression with antisense oligonucleotides and elimination of the decrease in p21 by adenoviral p21 infection, cyclin E-Cdk2 activation by estrogen was completely eliminated. 2 A consequence of these events is that a minority of the total cyclin E-Cdk2 complexes are active and, despite being depleted of p21 and p27, elute at a higher molecular mass, i.e. ϳ350 kDa on gel filtration chromatography. Although all the components of this active oligomeric complex have yet to be defined, the pRb-related protein, p130, is associated with active cyclin E-Cdk2 after estrogen activation. This interaction is likely facilitated by the high abundance of p130 after anti-estrogen arrest (37) and through competition between p21 and p130 for a common binding site on cyclin E-Cdk2, recognizing the RXL motif common to both proteins (48,56).
Delineation of the mechanisms through which insulin/IGF-I and estrogen regulate cyclin D1-Cdk4 and cyclin E-Cdk2 activity provides insight into the synergistic activation of the latter complex after combined treatment, illustrated in the model presented in Fig. 9. As noted above, insulin/IGF-I failed to initiate substantial S-phase entry despite the induction of cyclin D1 and formation of active p21-associated cyclin D1-Cdk4 complexes. Although cyclin E expression increased, little activation of cyclin E-Cdk2 occurred, presumably due to the presence of high levels of p21 and consequent formation of inactive p21-associated cyclin E-Cdk2 complexes. Treatment with estrogen alone resulted in synchronous entry of the majority of cells into S phase. Like insulin/IGF-I, estrogen induced cyclin D1 expression and led to the formation of active p21-associated cyclin D1-Cdk4 complexes. However, in contrast with insulin/ IGF-I, estrogen activated cyclin E-Cdk2 via a combination of p21 sequestration into newly formed cyclin D1-Cdk4 complexes and inhibition of p21 gene expression, favoring the formation of high molecular mass, active cyclin E-Cdk2-p130 complexes (Fig. 9).
Data presented in this manuscript indicate that the synergy between insulin/IGF-I and estrogen in initiating cell cycle progression is not due to cooperative activation of cyclin D1-Cdk4 despite evidence that cyclin D1-Cdk4 activity is essential to the effects of both mitogens (35). Rather, the synergy is associated with enhanced activation of cyclin E-Cdk2 via a significant increase in high molecular mass cyclin E complexes containing p130. Two distinct mechanisms contribute to this effect; one mechanism is the induction of cyclin E after insulin/IGF-I stimulation, and the other mechanism is the ability of estrogen to attenuate the insulin-induced increase in total cellular p21 (Fig. 4) and the concurrent recruitment of p21 into cyclin E-Cdk2 complexes (Fig. 5). Since cyclin E and its associated Cdk2 activity are rate-limiting for G 1 -to S-phase progression (33), the markedly increased formation of active cyclin E-Cdk2 complexes after combined treatment (Figs. 3 and 6) would be expected to increase pRb phosphorylation and S-phase entry, as demonstrated in Figs. 1 and 2. Our ability to inhibit the synergistic interaction between insulin/IGF-I and estrogen after infection with a p21-expressing adenovirus adds further support for a pivotal role for estrogen depletion of p21 in this synergistic mitogenic response.