Transcriptional Repression by Rb-E2F and Regulation of Anchorage-Independent Survival

Abstract

Mutations that lead to anchorage-independent survival are a hallmark of tumor cells. Adhesion of integrin receptors to extracellular matrix activates a survival signaling pathway in epithelial cells where Akt phosphorylates and blocks the activity of proapoptotic proteins such as the BCL2 family member Bad, the forkhead transcription factor FKHRL-1, and caspase 9. Insulin-like growth factor 1 (IGF-1) is a well-established epithelial cell survival factor that also triggers activation of Akt and can maintain Akt activity after cells lose matrix contact. It is not until IGF-1 expression diminishes (∼16 h after loss of matrix contact) that epithelial cells deprived of matrix contact undergo apoptosis. This suggests that IGF-1 expression is linked to cell adhesion and that it is the loss of IGF-1 which dictates the onset of apoptosis after cells lose matrix contact. Here, we examine the linkage between cell adhesion and IGF-1 expression. While IGF-1 is able to maintain Akt activity and phosphorylation of proapoptotic proteins in cells that have lost matrix contact, Akt is not able to phosphorylate and inactivate another of its substrates, glycogen synthase kinase 3β (GSK-3β), under these conditions. The reason for this appears to be a rapid translocation of active Akt away from GSK-3β when cells lose matrix contact. One target of GSK-3β is cyclin D, which is turned over in response to this phosphorylation. Therefore, cyclin D is rapidly lost when cells are deprived of matrix contact, leading to a loss of cyclin-dependent kinase 4 activity and accumulation of hypophosphorylated, active Rb. This facilitates assembly of a repressor complex containing histone deacetylase (HDAC), Rb, and E2F that blocks transcription of the gene for IGF-1, leading to loss of Akt activity, accumulation of active proapoptotic proteins, and apoptosis. This feedback loop containing GSK-3β, cyclin D, HDAC-Rb-E2F, and IGF-1 then determines how long Akt will remain active after cells lose matrix contact, and thus it serves to regulate the onset of apoptosis in such cells.

Adhesion of epithelial cells to the surrounding extracellular matrix is required for cell survival. Apoptosis of epithelial cells that are deprived of matrix contact is important for biologic processes such as involution of the mammary gland following weaning and of the prostate following androgen ablation therapy for cancer treatment. Loss of steroid hormones under these conditions in the mammary gland and the prostate triggers release of proteases that degrade the surrounding matrix, resulting in a loss of cell anchorage and epithelial apoptosis (1, 20, 68, 69, 78). Mutations that allow anchorage-independent survival are a hallmark of neoplastic transformation and are critical for tumor progression as cells lose traditional matrix contacts when tumors expand and metastasize (47).

Interaction of cells with the extracellular matrix is mediated by integrin receptors on the cell surface (30). In epithelial and endothelial cells, disruption of integrin contacts leads to apoptosis (7, 29, 60). Ligation of integrins to the extracellular matrix can activate phosphatidylinositol 3-kinase (PI-3K) and its downstream target kinase, Akt (41). This PI-3K/Akt pathway is required for cell survival, and expression of a constitutively active form of PI-3K or Akt prevents apoptosis of epithelial cells deprived of matrix contact (39, 40). A role for constitutive activation of this survival pathway in tumors is illustrated by the finding that the genes for Akt and for the regulatory subunit of PI-3K are amplified in tumors, and versions of both of these genes have been found as transforming oncogenes in retroviruses (6, 38). Additionally, the PTEN phosphatase, which negatively regulates PI-3K, is a tumor suppressor whose mutation can lead to activation of PI-3K/Akt (66, 67). PI-3K is also activated by insulin-like growth factor 1 (IGF-1), and addition of IGF-1 to cells deprived of matrix contact is sufficient to maintain activation of the PI-3K/Akt pathway and prevent apoptosis (71). Accordingly, IGF-1 has been shown to be a potent survival factor in a number of tumors.

Several proapoptotic proteins have been identified as downstream targets of Akt. One of these is the Bcl-2 family member Bad (18, 21). Phosphorylation of Bad triggers association with 14-3-3 proteins and loss of apoptotic activity. Akt also phosphorylates the forkhead transcription factor FKHRL-1, and as with Bad, this phosphorylation leads to association with 14-3-3 proteins and loss of FKHRL-1 function (10). Unphosphorylated FKHRL-1 activates genes with insulin response elements such as Fas ligand (10) and IGF binding protein 1 (33). Akt also phosphorylates procaspase 9, and this phosphorylation prevents cleavage, which is required for activation (12). When epithelial cells lose matrix contact and Akt activity diminishes, these proapoptotic proteins become activated and apoptosis ensues.

Glycogen synthase kinase 3β (GSK-3β) is also phosphorylated and inhibited by Akt (17, 62, 72). Like the proapoptotic regulators discussed above (Bad, FKHRL-1, and procaspase 9), GSK-3β also is important in regulating cell survival—overexpression of GSK-3β triggers apoptosis, and expression of a dominant-negative form of the protein prevents apoptosis when the PI-3K/Akt signaling pathway is blocked (56). These results indicate that inhibition of GSK-3β is also critical for Akt to promote cell survival. Since Akt-dependent inhibition of GSK-3β appears essential for Akt to block apoptosis, it seems that GSK-3β must somehow be linked to the other proapoptotic targets of Akt. However, this linkage is still unclear. One target of GSK-3β is the cell cycle regulatory protein cyclin D1 (23). Its phosphorylation of cyclin D1 leads to ubiquitin-mediated degradation of the protein (23). Accordingly, when cells are deprived of matrix contact and Akt activity is lost (and thus, active GSK-3β accumulates), the level of cyclin D1 declines (7, 19, 84). Interestingly, the cyclin D1 gene is frequently amplified and the protein is overexpressed in breast, head, and neck carcinomas (16, 24, 53), suggesting that maintaining expression of the protein is important as tumors expand and cells lose their traditional matrix contacts. Cyclin D1 acts as a regulatory subunit for G₁ cyclin-dependent kinase 4 (cdk4) and cdk6 (64, 65, 76). A primary target for cyclin D-cdk4-cdk6 is the cell cycle regulatory protein Rb, which arrests cells in the G₁ phase of the cell cycle by inhibiting transcription of genes required for S phase (76). Phosphorylation of Rb by cyclin D-cdk4-cdk6 inhibits Rb activity, and thus, hypophosphorylated (active) Rb accumulates in cells deprived of matrix contact (where cyclin D1 levels are diminished) (19, 31, 84). While this linkage between Akt and the cell cycle control proteins (via GSK-3β) is interesting, it remains unclear how this might relate to activity of the proapoptotic targets of Akt and thus whether regulation of cell cycle control proteins via GSK-3β is involved in the PI-3K/Akt survival pathway.

Here, we provide evidence that a major role for GSK-3β in the PI-3K/Akt survival pathway is its regulation of cyclin D expression. We show that maintenance of cyclin D1 expression in epithelial cells deprived of matrix contact is sufficient to prevent apoptosis, thus providing a possible explanation for why the gene is frequently amplified in carcinomas. Additionally, we provide evidence that the hypophosphorylated Rb that accumulates when epithelial cells lose matrix contact and cyclin D1 levels diminish forms a repressor complex that blocks IGF-1 expression. This downregulation of IGF-1 leads to loss of Akt activity and accumulation of active proapoptotic proteins. We conclude that a role of GSK-3β in the PI-3K/Akt pathway is to control expression of IGF-1 (via regulation of cyclin D1 and Rb) and thereby regulate the timing of apoptosis after epithelial cells lose matrix contact.

MATERIALS AND METHODS

Cell culture and transfection assays.

Primary human tracheal epithelial cells were maintained previously as described (45). Human umbilical vein endothelial cells (HUVEC) were maintained as recommended by the supplier (Clonetics). The human prostate epithelial cell line LNCaP was maintained in RPMI medium with 10% fetal bovine serum. The human mammary epithelial cell line MCF10A was maintained in Dulbecco modified Eagle medium with 5% horse serum, 10 μg of insulin per ml, 50 μg of hydrocortisone per ml, 20 μg of epidermal growth factor per ml, 2 mM glutamine, and 1 mM sodium pyruvate. For culture in suspension, subconfluent cell monolayers were trypsinized and placed in culture dishes with constant agitation. Due to the protective effect of insulin, MCF10A cells were insulin starved for 2 days prior to trypsinization. Where indicated, we included LY294002 (BIOMOL) at 50 μM, PD98059 (BIOMOL) at 100 μM, HNMPA(AM)3 (BIOMOL) at 35 μM, IGF-1 or insulin (Sigma) at 10⁻⁷ M, trichostatin A (TSA; Wako Bioproducts) at 10 nM for HUVEC, and tumor necrosis factor alpha (TNFα; Sigma) at 20 ng/ml. At the indicated times after culture in suspension, apoptosis was scored by both trypan blue exclusion and terminal deoxynucleotidyl transferase-mediated dUTP-biotin nick and labeling (TUNEL) reaction as previously described (19). Apoptosis was confirmed by electron microscopy and DNA laddering.

The human osteosarcoma cell line U2OS was maintained in Dulbecco modified Eagle medium with 10% fetal bovine serum. U2OS cells were transfected using Effectene (Qiagen), and chloramphenicol acetyltransferase (CAT) activity was determined as previously described (77). Where appropriate, 200 nM TSA (Wako Bioproducts) was added 18 h after transfection. The IGF-1 CAT constructs were kindly provided by R. Baserga (44). The mouse pro-B-cell line FL5.12 was maintained in IMD medium with 10% fetal bovine serum and 10% WEHI-3B conditional medium as a source of interleukin-3 (IL-3). FL5.12 cells were stably transfected using electroporation as previously described (82). At the indicated times after culture without IL-3, apoptosis was analyzed by trypan blue exclusion.

Western blot analysis and kinase assays.

Western blot assays were done as previously described (19) using antibodies for cyclin D1, cyclin A, cyclin E, cdk4, cdk2, Rb (PharMingen), caspase 3 and Bad (Transduction Labs), phospho-Bad (Ser 112 and Ser 136), Akt, phospho-Akt (Thr 308 or Ser 473), GSK-3β and phospho-GSK-3β (New England Biolabs), PTEN and p27 (Santa Cruz Biotechnology, Inc.), and α-tubulin (Sigma). For the phospho-Bad experiments, cells were initially deprived of matrix contacts for 12 h. Kinase assays for immunoprecipitated cdk4 and cdk2 were done as previously described (83) from cells cultured in suspension for 6 h or maintained as adherent monolayers. Lysates for the cdk4 kinase assay were first precleared with cdk2 antiserum. The C-terminal region of Rb (amino acids 792 to 928; Santa Cruz) was used as a substrate for the cdk4 assay, and histone H1 (Boehringer Mannheim) was used as a substrate for the cdk2 assay. Kinase assays for immunoprecipitated Akt were done using an Akt kinase assay kit (New England Biolabs).

Survival assays using transient transfections.

LNCaP cells were transfected by the calcium phosphate method. Cells newly plated on 60-mm-diameter plates were transfected with 2 μg of SV₂luc and 4 μg of the indicated expression vector. Twenty-four hours after transfection, cells were trypsinized and half of the culture was placed in petri dishes (matrix deprivation); the other half was allowed to readhere to tissue culture plastic. After another 24 h, cells were collected and luciferase activity was determined. MCF10A cells were transfected using Effectene (Qiagen). Cells newly plated on 60-mm plates were transfected with 1 μg of cytomegalovirus luc and 5 μg of the indicated expression vector. Cells were detached similarly as LNCaP and collected after 48 h. Expression vectors for CD2-p110 and CD2-p110KD were kindly provided by D. A. Cantrell. DN MEK1 (S218A/S222A) was provided by K. L. Guan, MyrAkt and A2 MyrAkt were provided by R. A. Roth, V12 Ras was provided by Alan Hall, and E2F-DB was provided by K. Helin.

Stable expression of cyclin D1.

LNCaP cells were transfected with the cyclin D1 expression vector RcCMVcyclin D1 (kindly provided by A. Arnold, Massachusetts General Hospital, Harvard University) or the empty vector using GeneFector (Venn Nova Inc.). G418-resistant colonies were selected in the presence of G418 at 500 μg/ml. Individual colonies were expanded and evaluated by Western blot analysis for cyclin D1.

RT-PCR.

Cellular RNA was isolated using RNA STAT (Tel-Test, Inc., Friendswood, Tex.). Reverse transcription (RT) reactions with 3 μg of RNA were primed using random hexamers (Boehringer Mannehein). The PCR conditions used were previously described (48). Primers were IGF-1 (5′-GTACTTCAGAAGCAATGGGAAAAATCAGCAGTCTTCC-3′ and 5′-TGCGCAATACATCTCCAGCCTCCTTAGATCACA-3′ [315 bp]), cyclin D1 (5′-TCGCTGGAGCCCGTGAAAAAGAGC-3′ and 5′-CAAAGGAAAAAACAACCAACAACAAGGAGAATG-3′ [700 bp]), and GADPH (5′-AACATCATCCCTGCCTCTACTG-3′ and 5′-TTGACAAAGTGGTCGTTGAGG-3′ [314 bp]).

RESULTS

The PI-3K/Akt pathway and survival of epithelial cells.

Previous studies have shown that depriving epithelial and endothelial cells of contacts with the extracellular matrix leads to apoptosis (30, 60, 75). In the following studies, we examined this apoptosis in primary cultures of epithelial and endothelial cells and breast and prostate epithelial cell lines with similar results. Depriving these adhesion-dependent cells of matrix contact resulted in apoptosis in approximately 24 h (Fig. 1A). This coincides with the cleavage and activation of caspase 3 (Fig. 1B). It has been demonstrated that there is a loss of PI-3K activity and the activity of the downstream kinase Akt under these conditions (41). Consistent with previous findings, overexpression of a constitutively active p110 subunit of PI-3K or expression of a constitutively active myristylated form of Akt (Myr-Akt) prevented apoptosis when cells were deprived of matrix contact (Fig. 1C and D; 40). Additionally, directly blocking PI-3K activity with the inhibitor LY294002 led to apoptosis of epithelial cells that were matrix adherent (Fig. 1A), suggesting that this pathway is essential even when cells are matrix attached. In contrast to inhibition of PI-3K activity, neither the MEK inhibitor PD98059 nor a dominant-negative MEK which blocks activation of ERK1 and -2 decreased cell viability (Fig. 1A), suggesting that activation of the mitogen-activated protein kinase pathway is not essential for epithelial cell survival.

FIG. 1.

Regulation of apoptosis in cells deprived of matrix contact. (A) LNCaP prostate epithelial cells either adherent to or detached from the matrix (Det.) were transfected with the indicated expression vectors and/or treated with the indicated inhibitors (PD is the MEK inhibitor PD98059; LY is the PI-3K inhibitor LY294002). Cells were detached for 24 h unless otherwise indicated. Apoptosis was monitored by a TUNEL assay and by trypan blue exclusion. Apoptosis was further confirmed in cells by DNA laddering and electron microscopy. Similar results were seen in primary cultures of HUVEC, primary human tracheal epithelial cells, and MCF10A mammary epithelial cells. (B) Caspase 3 is cleaved to an active form when cells are deprived of matrix contact. A Western blot is shown for caspase 3 at different time points after LNCaP cells were detached from the matrix. The arrow indicates the cleaved, active form of the protein. (C) LNCaP prostate epithelial cells were cotransfected with the indicated expression vectors along with the pSV-luciferase reporter. Retention of luciferase activity was used as a measure of cell viability in these assays as described previously (74). CD2-p110, Myr-Akt, and V12 Ras are constitutively active constructs; CD2-p110KD is a kinase-dead mutant control. The results shown are all representative of at least three independent experiments, each done in duplicate. (D) MCF10A cells were cotransfected with the indicated expression vectors along with the pCMV-luciferase reporter. E2F-DB is a dominant-negative E2F. Results are standardized with respect to an empty-vector control for V12 Ras and CD2-p110, a myristoylation mutant form of MyrAkt, and E2F-DB, an E132 binding mutant form of E2F-DB.

Expression of constitutively active mutant Ras proteins, which are found in a variety of tumors, was sufficient for anchorage-independent survival of epithelial cells (Fig. 1A, C, and D) (29). Ras not only activates Raf and, ultimately, ERK1 and -2, but it also activates PI-3K (79). The ability of activated Ras to prevent apoptosis in cells deprived of matrix contact was prevented by treatment with LY29004, but neither the MEK inhibitor PD98059 nor a dominant-negative MEK, which block activation of ERK1 and -2, had any affect (Fig. 1A). Thus, it appears that it is activation of PI-3K by Ras which leads to survival of epithelial cells deprived of matrix signaling. Indeed, previous studies with mutant forms of Ras that discriminate between its activation of PI-3K and Raf have also suggested that it is the activation of PI-3K by Ras which is essential for anchorage-independent survival of epithelial cells (40).

When cells were deprived of matrix contact, Akt activity, as assessed by phosphorylation of the protein on Thr 308 and Ser 473 and by kinase activity of immunoprecipitated protein, decreased (Fig. 2A and results not shown). This decrease in Akt activity did not occur until between 12 and 24 h after matrix detachment (Fig. 2C and D) in MCF10A cells, which are PTEN⁺, or in LNCaP cells, which are PTEN⁻ (Fig. 2E). This loss of Akt phosphorylation appears to just precede the first evidence of the onset of apoptosis at approximately 24 h following detachment of cells from the matrix. Depriving cells of matrix contact also led to loss of phosphorylation of Akt target proapoptotic proteins such as Bad (on Ser 136; phosphorylation of Ser 112 was not affected) (Fig. 2B and results not shown). Accordingly, when cells were allowed to reattach to the matrix, Akt was activated and Bad was phosphorylated on Ser 136 (Fig. 2A and B).

FIG. 2.

Phosphorylation of Akt and Bad is regulated by cell adhesion and by IGF-1. (A) Western blot for phosphorylated, active Akt (pAkt) (Thr 308; similar results were seen with an antibody specific for Ser 473) in extracts from LNCaP cells either adherent, deprived of matrix contact (Det.) for 24 h with or without the addition of IGF-1 for the indicated times, or allowed to reattach for 6 h. The blot was reprobed with a pan-Akt antibody to detect total Akt. (B) Western blot analysis using antibodies specific for Bad phosphorylation on Ser 136 or Ser 112. LNCaP cells were detached for 12 h and then pretreated with LY294002 (LY) at 50 μM (where indicated) for 10 min prior to addition of IGF-1 (10⁻⁷ M). Bad (total) indicates that a pan-Bad antibody was used to detect total Bad protein. The same blot was reprobed with the three different antibodies. (C) Western blot for phosphorylated active Akt (pAkt) (Thr 308) in extracts from LNCaP cells deprived of matrix contact for the indicated times. The blot was reprobed with a pan-Akt antibody to detect total Akt. (D) Western blot for phosphorylated Akt (pAkt) in extracts from MCF10A cells deprived of matrix contact for the indicated times. The blot was reprobed with a pan-Akt antibody to detect total Akt. (E) Western blot for PTEN in adherent LNCaP cell: HUVEC, and MCF10A cell extracts.

IGF-1 and epithelial cell survival.

IGF-1 is a well-known survival factor for epithelial cells that is required for a number of epithelial cells to propagate in culture (4, 51) and is overexpressed in epithelial tumors (13, 28, 34, 58). As with cell adhesion, IGF-1 activates the PI-3K/Akt pathway, and addition of exogenous IGF-1 prevented apoptosis of cells that lose matrix contact (Fig. 2A and 3A). Accordingly, addition of LY29004 prevents this protective effect of IGF-1. We found that treatment of cells deprived of matrix contact with IGF-1 led to rapid (within 10 min) phosphorylation of Akt on Thr 308 and Ser 473 and phosphorylation of Bad on Ser 136 (Fig. 2A and B and results not shown).

FIG. 3.

IGF-1 regulates the onset of apoptosis in cells deprived of matrix contact. (A) MCF10A mammary epithelial cells were detached from the matrix for 48 h; where indicated, the cells were detached in the presence of 10⁻⁷ M IGF-1 or insulin with or without 50 μM LY294002 (LY). Apoptosis was analyzed by trypan blue exclusion. The results shown are representative of at least three independent experiments. (B) RT-PCR analysis for IGF-1 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) control mRNA. Wild-type (WT) LNCaP cells or a clone stably expressing cyclin D1 (clone 3 in Fig. 8A; similar results were seen with clone 5) were detached from the matrix for the indicated periods of time, and cellular RNA was analyzed by RT-PCR. (C) HUVEC were treated with the insulin receptor inhibitor HNMPA(AM)3, and the effect of the inhibitor on apoptosis of cells deprived of matrix contact was analyzed by trypan blue exclusion. Similar results were seen with LNCaP cells. (D) Western blot of phosphorylated active Akt (pAkt) (Thr 308) in LNCaP cells treated with the insulin receptor inhibitor. The blot was reprobed with a pan-Akt antibody to detect total Akt.

Epithelial cells express IGF-1, and it has been shown that IGF-1 can be overexpressed in tumors, where it acts as a survival factor (4, 51). Therefore, we wondered whether expression of endogenous IGF-1 might have a role in regulating apoptosis of epithelial cells deprived of matrix contact. We found that the cells do express IGF-1 mRNA and that the IGF-1 message level diminished by 16 h after cells were deprived of matrix contact (Fig. 3B)—this preceded the first evidence of apoptosis in the cells (at about 24 h; Fig. 1A). These results demonstrated that expression of IGF-1 is dependent upon cell adhesion, and they raised the possibility that PI-3K/Akt remains active (via IGF-1) when cells are deprived of matrix contact until IGF-1 diminishes. Indeed, we found that the loss of Akt activity approximately coincides with the onset of apoptosis. These results suggested that the loss of IGF-1 may dictate the onset of apoptosis. If this were the case, we reasoned that blocking of IGF-1 signaling should lead to a more rapid onset of apoptosis when epithelial cells are deprived of matrix contact. Indeed, when cells were treated with the insulin receptor inhibitor HNMPA(AM)3, which blocks signaling by insulin or IGF-1 (61), the loss of Akt activity and the onset of apoptosis were accelerated in cells deprived of matrix contact (Fig. 3D). Taken together, the above results indicate that IGF-1 can maintain PI-3K/Akt activity in cells that have lost matrix contact and that it may be the downregulation of IGF-1 expression after epithelial cells lose matrix contact which actually dictates the onset of apoptosis. This then raised the question of how IGF-1 expression is linked to cell adhesion.

Cell adhesion is required for Akt phosphorylation of GSK-3β.

The kinase GSK-3β is another substrate for Akt, and as with Bad, FHKRL-1, and caspase 9, phosphorylation by Akt blocks GSK-3β activity (17, 62, 72). This inhibition of GSK-3β activity may also be required to prevent apoptosis, as evidenced by the finding that overexpression of GSK-3β triggers apoptosis (56). Conversely, expression of a dominant-negative form of GSK-3β can prevent apoptosis in cells where PI-3K/Akt signaling is blocked (56). While treatment of cells deprived of matrix contact with IGF-1 led to activation of Akt and phosphorylation of Bad (Fig. 2A and B), GSK-3β was not phosphorylated (Fig. 4A). However, when cells were allowed to reattach to the matrix, GSK-3β was again phosphorylated and LY29004 blocked this phosphorylation. Therefore, the ability of Akt to phosphorylate and inactivate GSK-3β is dependent upon cell adhesion to the matrix.

FIG. 4.

Cell adhesion regulates the subcellular localization of Akt and its ability to phosphorylate GSK-3β. (A) LNCaP cells were treated as described in the legend to Fig. 2A and Western blot assayed for phosphorylated GSK-3β (GSK-3α is also recognized by this phosphorylation-specific antibody against GSK-3β). The blot was reprobed with a pan-GSK-3β antibody to show total GSK-3β. (B) Immunoprecipitation kinase assay showing phosphorylation of GSK-3β. Total Akt was immunoprecipitated from LNCaP cells treated as indicated and used to phosphorylate GSK-3β purified from bacteria. (C to E) Immunofluorescent staining for active Akt (Thr 308) (red) and GSK-3β (green). Nuclear staining with 4′,6′-diamidino-2-phenylindole (DAPI) is shown in blue-purple. The merge is shown at the top of each panel. Panel C shows adherent cells, and panel D shows cells detached (Det.) from the matrix for 16 h and then treated for 15 min with IGF-1. Panel E is the same as panel D, but the cells were not stimulated with IGF-1 (the cells are also shown at a higher magnification). Similar results were seen when using an antibody specific for Akt phosphorylated on Ser 473. LY, LY294002.

Cell adhesion regulates the subcellular localization of Akt.

There are several reasons why Akt may not be able to phosphorylate GSK-3β when cells lose matrix contact. First, Akt may not be fully activated by IGF-1 in the absence of matrix contact—this could be due to altered phosphorylation of Akt itself or perhaps loss or inactivation of a third protein required for efficient interaction of Akt and GSK-3β. Alternatively, Akt may be fully active, but when cells lose matrix contact, the localization pattern of Akt and/or GSK-3β changes such that the proteins become segregated. To address the first possibility, we asked whether Akt (immunoprecipitated from cells deprived of matrix contact but treated with IGF-1) could phosphorylate GSK-3β in vitro. Indeed, we found that immunoprecipitated Akt efficiently phosphorylated GSK-3β (Fig. 4B). These results suggested that Akt was fully active and raised the possibility that loss of cell contact with the matrix changes the localization of Akt and/or GSK-3β in the cell such that GSK-3β is no longer accessible to Akt.

Using antibodies that specifically recognize Akt phosphorylated on Thr 308 or Ser 473 (phosphorylation of both sites is required for Akt activation), we found that in adherent cells activated Akt was cytoplasmic and appeared to be concentrated at the plasma membrane (Fig. 4C to E), as previously suggested for activated Akt (2, 32), and GSK-3β appeared to colocalize with active Akt. However, when cells were deprived of matrix contact, active Akt was no longer localized at the plasma membrane and, importantly, it no longer colocalized with GSK-3β (Fig. 4C to E). Active Akt was evident for at least 16 h following detachment of cells from the matrix, and although treatment of such cells with IGF-1 did increase the immunostaining for active Akt somewhat, its localization did not change (neither the expression nor the localization of GSK-3β was affected by treatment of cells in suspension with IGF-1) (Fig. 4C to E and results not shown). We conclude that Akt may no longer be able to phosphorylate GSK-3β when cells are detached from the matrix because Akt no longer colocalizes with GSK-3β.

Downregulation of G₁ cyclins triggers loss of cdk4 and cdk2 activity in epithelial cells deprived of matrix contact.

How might GSK-3β be involved in regulation of epithelial cell apoptosis? Recently, cyclin D1 was identified as a target of GSK-3β—this phosphorylation of cyclin D1 targets the protein for ubiquitination and rapid turnover (23). Cyclin D1 acts as a regulatory subunit for cdk4, which phosphorylates and inactivates Rb (64, 65, 76). This active Rb, in turn, forms a complex with histone deacetylase (HDAC) that interacts with E2F bound to the promoter of genes, repressing transcription (9, 48, 49). Interestingly, the IGF-1 gene contains E2F sites that serve as silencer elements (44, 57); thus, we hypothesized that this HDAC-Rb-E2F may repress IGF-1 expression.

To investigate the pathway between GSK-3β and IGF-1, we began by examining the effect of cell adhesion on the activity of cdks that regulate Rb function. Expression of D, E, and A cyclins was rapidly downregulated when epithelial cells were deprived of matrix contact; in contrast, the cdk2 inhibitor p27 was upregulated—likewise, direct blockage of PI-3K activity with LY294002 led to downregulation of the cyclins and upregulation of p27 (Fig. 5A to C and results not shown) (19, 25, 27, 31, 84). Loss of cyclin expression was evident between 1 and 5 h after cells were deprived of matrix contact; however, apoptosis was not evident until approximately 24 h (Fig. 1A). Therefore, the loss of cyclin precedes apoptosis. This downregulation of cyclin D1 appears to be a posttranscriptional process—while the protein declines by 5 h after cells are deprived of matrix contact, the mRNA level does not decline until between 24 and 48 h (Fig. 5D). Furthermore, transcription of the cyclin D1 gene is dependent upon mitogen-activated protein kinase signaling (42); however, addition of the MEK inhibitor PD98059 did not lead to a decline in cyclin D1 until 72 h of treatment (Fig. 5C). These results provide further evidence that this rapid loss of cyclin D in cells deprived of matrix contact is not the result of a block in transcription but is likely the result of GSK-3β-mediated turnover of the protein. Accordingly, IGF-1 treatment was not able to maintain cyclin D1 levels in cells deprived of matrix contact (Fig. 5E). In addition to loss of cyclin expression, there was an increase in cyclin-dependent inhibitors p21 and p27, but not in p16, when cells were deprived of matrix contact (Fig. 5B) (19, 27, 84). Accordingly, this loss of cyclin expression and increase in cdk inhibitor expression led to a loss of cdk4 activity and cdk2 activity when cells were deprived of matrix contact for 6 h (Fig. 5F and G).

FIG. 5.

Loss of G₁ cyclin expression, induction of the cdk2 inhibitor p27, and loss of cdk4 and cdk2 activity in cells deprived of matrix contact. (A and B) Western analysis for cyclin D1 and p27 in MCF10A cells deprived of matrix contact (det.) for increasing lengths of time or allowed to reattach. (C) Western analysis for cyclin D, cyclin E, and p27 in LNCaP cells deprived of matrix contact or treated with LY294002 (LY) or PD98059 (PD). (D) Cyclin D1 mRNA levels decay more slowly than cyclin D1 protein when cells are deprived of matrix contact or treated with LY294002. Cyclin D1 and control glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA from MCF10A cells were analyzed by RT-PCR. (E) Anchorage is required for IGF-1 to maintain cyclin D1 expression. A Western blot for cyclin D1 is shown examining MCF10A cells detached for 48 h or detached in the presence of 10⁻⁷ M IGF-1, similar to Fig. 3A. (F and G) cdk4 and cdk2 activities are lost in cells deprived of matrix contact, but stable expression of cyclin D1 overcomes the loss of cdk4 activity. cdk4 and cdk2 were immunoprecipitated from wild-type (WT) or cyclin D1-expressing (clone 3 in Fig. 8A; similar results were seen with clone 5) adherent LNCaP cells or cells placed in suspension for 6 h. Phosphorylation of the Rb C-terminal region (Rb-C; amino acids 792 to 928) or histone H1 was used to assess kinase activity. Western blots for total cdk4 or cdk2 are shown.

An HDAC-Rb-E2F repressor complex regulates the onset of apoptosis in epithelial cells deprived of matrix contact.

A major target of G1 cdk activity is Rb, and accordingly, hypophosphorylated Rb accumulated by 4 h after cells were deprived of matrix contact (where both cdk4 and cdk2 activities are lost) (Fig. 6A). Likewise, LY294002 treatment to directly block PI-3K activity also resulted in rapid accumulation of hypophosphorylated Rb (Fig. 6B and C and results not shown). We hypothesized that formation of an HDAC-Rb-E2F repressor complex at E2F sites on the IGF-1 gene promoter may be responsible for repression of the gene when cells lose matrix contact. To test this possibility, we used a dominant-negative form of E2F to displace such repressor complexes from the E2F sites. This mutant form of E2F retains a DNA binding domain but lacks the Rb binding domain, so it is unable to recruit HDAC-Rb. We have demonstrated previously that overexpression of this dominant-negative E2F displaces wild-type E2F from promoters and overcomes growth suppression by Rb (83). Expression of the dominant-negative E2F prevented apoptosis of epithelial cells deprived of matrix contact (Fig. 7A), suggesting that an Rb-E2F repressor complex is indeed involved in the apoptotic process. As a control, we examined the effect of this dominant-negative E2F on another apoptotic model, the FL5.12 pro-B-cell line, in which withdrawal of IL-3 leads to apoptosis within 12 to 24 h. In contrast to inhibition of adhesion-dependent apoptosis in epithelial cells, dominant-negative E2F had no effect on the extent or the kinetics of this IL-3-dependent apoptosis (Fig. 7C).

FIG. 6.

Hypophosphorylated (active) Rb accumulates in cells deprived of matrix contact. (A to C) Western blot showing that hypophosphorylated Rb (pRb indicates hypophosphorylated Rb; ppRb indicates hyperphosphorylated Rb) accumulates in MCF10A cells (A) and LNCaP cells (B and C) deprived of matrix contact (det.) or treated with LY294002 (LY; similar results were seen in HUVEC and human tracheal epithelial cells) and that stable expression of cyclin D1 prevents this accumulation. WT, wild-type cells. Cyclin D1, clone 3 in Fig. 8A (similar results were also seen with clone 5).

FIG. 7.

Transcriptional repression by HDAC-Rb-E2F. (A) Maintenance of cyclin D1 or expression of E2F-DB prevents apoptosis in cells deprived of matrix contact. Percent apoptosis was determined by trypan blue exclusion and confirmed by TUNEL reaction. WT, wild-type cells. Cyclin D1, clone 3 in Fig. 8A (similar results were seen with clone 5). The data shown are representative of at least three independent experiments. (B) Western blot of FL5.12 clone stably expressing E2F-DB. (C) Withdrawal of IL-3 results in apoptosis at the indicated times in both wild-type (open bars) and E2F-DB-expressing (closed bars) FL5.12 cells. Percent apoptosis was determined by trypan blue exclusion. The data shown are representative of at least three independent experiments. (D) Adherent and detached HUVEC were treated with the HDAC inhibitor TSA (48), and apoptosis was evaluated by trypan blue exclusion. (E) Adherent HUVEC were treated with TNF-α with or without TSA for 48 h. Apoptosis was evaluated by trypan blue exclusion. (F) The indicated reporter constructs (2 μg) were transfected into 60-mm plates of U2OS cells. Where indicated, TSA was added 18 h after transfection. (G) IGF-1 CAT (2 μg) was cotransfected along with the indicated expression vector (2 μg for Rb, 0.5 μg for E2F-1) into 60-mm plates of U2OS cells. An empty vector was transfected into plates that did not receive Rb or the E2F-1 expression vector. Det., detached; LY, LY294002.

If association of HDAC with Rb-E2F is also important for the apoptotic process, we reasoned that treatment of cells deprived of matrix contact with the HDAC inhibitor TSA should reverse the apoptosis. Indeed, TSA significantly inhibited this apoptosis (Fig. 7D), suggesting that HDAC activity is important. In contrast, treatment with TSA did not prevent apoptosis triggered by TNF-α treatment of HUVEC (Fig. 7E). In transfection assays, TSA activated expression of the IGF-1 gene promoter (Fig. 7F). This activation was similar to that observed with the control adenovirus major late promoter, which is classically under repression by HDAC (48). In contrast, TSA had no effect on the simian virus 40 promoter-enhancer, which is not repressible by HDAC. Furthermore, expression of Rb caused repression of the IGF-1 reporter and expression of E2F-1 activated it (Fig. 7G). Coexpression of both E2F-1 and Rb causes decreased activity from the reporter. Thus, it appears that an HDAC-Rb-E2F repressor complex (regulated by GSK-3β and cyclin D-cdk4) is important for repression of IGF-1 expression when cells lose matrix contact. Futhermore, the onset of apoptosis is delayed until IGF-1 expression diminishes.

Maintenance of cyclin D1 expression prevents apoptosis of epithelial cells deprived of matrix contact.

Cyclin D-cdk4 phosphorylation of Rb specifically blocks its interaction with HDAC (35), and the evidence we provide above shows that HDAC activity is important for Rb-E2F function in apoptosis of epithelial cells that lose matrix contact. Therefore, we reasoned that the loss of cyclin D expression may be key to the accumulation of a functional HDAC-Rb-E2F complex when cells lose matrix contact. To determine whether this is indeed the case, we isolated clones stably expressing cyclin D1 (Fig. 8A). Normally, cyclin D1 diminishes by 5 h after cells are deprived of matrix contact (8, 19, 84); however, cyclin D1 levels were maintained for more than 24 h in clones stably expressing the protein (Fig. 8B). Accordingly, cdk4 activity was preserved (Fig. 5F), Rb was maintained in its hyperphosphorylated form (Fig. 6C), and IGF-1 expression was maintained (Fig. 3C). More importantly, maintenance of cyclin D1 expression in this fashion prevented the apoptosis of cells deprived of matrix contact (Fig. 7A). Cyclin D1 levels did eventually diminish after 48 h in suspension (Fig. 8B), and this loss was accompanied by an increase in apoptosis (results not shown).

FIG. 8.

Stable expression of cyclin D1. (A) Western blot of LNCaP clones stably expressing cyclin D1. Clones 3 and 5 were used in the studies here with similar results. (B) Western blot analysis of cyclin D1, cyclin A, and cdk4 in wild-type (WT) cells or clone 3 (Cyclin D1). det., detached; LY, LY294002.

Taken together, our results point to the downregulation of cyclin D1 (and loss of cdk4 activity) as a key event in the regulation of the timing of apoptosis when cells lose matrix contact. While the cells stably expressing cyclin D1 were able to survive in the absence of matrix contact, they did not proliferate, presumably due to the continued inhibition of cdk2 activity. Thus, while overexpression of cyclin D1, which occurs frequently in head, neck, and breast carcinomas (16, 24, 53), can afford tumor cells protection from apoptosis as tumors expand and lose traditional matrix contacts, other mutations or genetic changes are required to allow the cells to proliferate under such conditions.

DISCUSSION

Apoptosis in adhesion-dependent cells deprived of traditional matrix signaling, such as that which occurs in rapidly proliferating and metastasizing tumor cells, appears to be regulated by Akt and its ability to phosphorylate and inactivate proapoptotic proteins such as procaspase 9, Bad, and FKHRL-1 in an adhesion-dependent fashion. Akt also phosphorylates another kinase, GSK-3β, blocking its activity. As with the proapoptotic proteins, the Akt-dependent blockage of GSK-3β activity is also crucial to the prevention of apoptosis. However, it has been unclear how GSK-3β might be related to these aforementioned targets of Akt, whose roles in promoting apoptosis are well established. Here, we provide evidence that GSK-3β is part of a feedback loop that determines how long Akt remains active (and thus able to maintain phosphorylation and inactivation of proapoptotic proteins) after epithelial cells lose matrix contact. This pathway, via its regulation of IGF-1 expression, then dictates when apoptosis will ensue after cells lose matrix contact (Fig. 9). Why might cells need such a regulatory loop to, at least temporarily, buffer them from an apoptotic response when they lose matrix contact? One possible explanation is that cells must be able to resist a relatively transient loss of matrix signaling, for example, when undergoing mitosis or perhaps when migrating.

FIG. 9.

HDAC-Rb-E2F and the regulation of apoptosis in cells deprived of matrix contact. Adherent cells have active PI-3K and Akt, which inhibit proapoptotic effectors of Akt and prevent the formation of an active Rb repressor complex. When epithelial cells are deprived of matrix contact, Akt activity is lost; however, apoptosis does not ensue immediately. The onset of apoptosis does not occur until the Rb complex represses expression of IGF-1. This then dictates the timing of apoptosis in anchorage-deprived cells. See the text for details.

Our results suggest that this feedback loop is triggered in cells that lose matrix contact when colocalization of active Akt and GSK-3β is lost. Akt activity is maintained in cells in suspension by the cellular expression of IGF-1 and under these conditions is still able to phosphorylate proapoptotic proteins and thus maintain inhibition of apoptosis. However, as active unphosphorylated GSK-3β accumulates in the cells in suspension (due presumably to its lack of localization with active Akt), it leads to turnover of cyclin D1. The resulting loss of cyclin D-cdk4 kinase activity leads to accumulation of hypophosphorylated Rb and assembly of an HDAC-Rb-E2F repressor complex which, in turn, represses the IGF-1 gene. With the loss of IGF-1, PI-3K activity, and thus Akt activity, diminishes, leading to accumulation of active proapoptotic proteins and apoptosis.

Short-circuiting of this cell adhesion signaling pathway allows anchorage-independent survival and is important for tumor progression. A number of mutations that constitutively activate this survival pathway have been identified. One such mutation is activation of Ras, which constitutively activates the PI-3K/Akt pathway and is present in approximately 30% of carcinomas. Amplification of the gene for Akt itself has also been observed in tumors and likewise leads to anchorage-independent survival (5, 14, 50, 59). The phosphatase PTEN inhibits PI-3K action by dephosphorylation of the D3 position of PIP3, a product of PI-3K (22, 66). PTEN is a tumor suppressor whose mutation leads to constitutive activation of PI-3K and anchorage-independent survival—it is mutated in breast, prostate, and endometrial cancers (3, 43, 73, 81). Elevated expression of IGF-1 and the IGF-1 receptor is also seen in tumors (11, 28, 34, 58). Amplification of the gene for cyclin D1 is common in carcinomas (16, 36, 46, 53), and the gene for Rb is also frequently mutated in a subset of tumors.

Mutation of the INK4a locus encoding p16^Ink4a is one of the most common mutations in tumors. p16^Ink4a binding to cdk4 blocks kinase activity, thereby leading to accumulation of hypophosphorylated Rb and formation of the HDAC-Rb-E2F repressor complex. However, the cdk2 inhibitors p21 and p27 are required to facilitate the formation of an active cyclin D-cdk4 complex (15, 37, 65). Binding of p16^Ink4a to cdk4 or cdk6 displaces p21 and p27, freeing them to inhibit cdk2 activity. Thus, p16^Ink4a can trigger growth arrest at two levels (inhibition of cdk4 and cdk2). Because of the relationships among p16^Ink4a, cyclin D1, and Rb, mutations in p16^Ink4a and Rb, as well as amplification-overexpression of the gene for cyclin D1 and mutation of Rb, appear to be mutually exclusive in tumors (55). However, amplification or overexpression of the gene for cyclin D1 occurs frequently in tumor cells where p16^Ink4a is mutated (16, 36, 46, 52–54). Indeed, in two-thirds of the cancer cell lines examined where the gene for cyclin D1 is amplified and overexpressed, there is concomitant mutation of p16^Ink4a (46). Amplification of the gene for cyclin D1 and mutation of p16^Ink4a have also been observed concomitantly in primary squamous carcinomas of the head and neck (24% of the tumors examined had amplification of the gene for cyclin D1, and 63% of these also had inactivating mutations in p16^Ink4a) (53). Moreover, these studies may have underestimated the number of tumors that are actually p16^Ink4a negative because they did not address inactivation of the Ink4a locus by methylation. The results imply a level of cooperation between overexpression of cyclin D1 and loss of p16^Ink4a in tumor formation.

Clearly, loss of p16^Ink4a is not sufficient to activate cdk4 unless expression of the D cyclin regulatory subunits is maintained in tumor cells, and overexpression of cyclin D1 cannot fully activate cdk4 in p16^Ink4a-positive cells because p16^Ink4a binds at least a portion of the kinases blocking their activity. We suggest that amplification of the gene for cyclin D1 in tumors serves primarily to prevent loss of cdk4 activity and apoptosis in epithelial cells that lose matrix signaling, whereas a subsequent mutation of p16^Ink4a is required to constitutively activate cdk4 (and thereby sequester p21 and p27), thus promoting cell cycle progression by activating both cdk4 and cdk2.

We were surprised initially that accumulation of hypophosphorylated Rb would be associated with apoptosis of epithelial cells deprived of matrix contact because hypophosphorylated Rb has been shown to inhibit p53-dependent apoptosis (26, 80). Overexpression of E2F-1 triggers apoptosis, and furthermore, a significant portion of the apoptosis observed in Rb^−/− mice is eliminated when the mice are crossed into an E2F-1^−/− background (70). In the absence of functional Rb, the resulting free E2F-1 that accumulates seems to trigger apoptosis at least in part by activating the alternate reading frame gene at the INK4a locus (63), which in turn blocks MDM2-mediated turnover of p53, leading to accumulation of p53 and apoptosis. However, less is known about p53-independent apoptosis, such as that which occurs when epithelial cells lose matrix signaling. Our results suggest that HDAC-Rb-E2F, through its repression of the gene for IGF-1, has an important role in regulating the onset of apoptosis in epithelial cells deprived of matrix contact.