Caveolin-1 Regulates the Antagonistic Pleiotropic Properties of Cellular Senescence through a Novel Mdm2/p53-Mediated Pathway

Antagonistic pleiotropy is an evolutionary theory of aging introduced by G.C. Williams in the 1950s based on the concept that some genes or processes increase the odds of successful reproduction early in life but have harmful effects later in life (1). Because these deleterious effects appear mostly after the organism's reproduction cycle is over, natural selection cannot eliminate them. Cellular senescence represents an example of antagonistic pleiotropy (reviewed in ref. 2). By inducing a permanent withdrawal from the cell cycle, cellular senescence acts as a tumor suppressor mechanism and prevents potential cancer cells from proliferating in young individuals (3). However, accumulation of senescent cells over time may contribute to aging and age-related diseases, including cancer. In fact, senescent fibroblasts secreting metalloproteinases, growth factors, and inflammatory cytokines (4–6) promote the growth and progression of preneoplastic breast cancer cells toward malignancy and potentiates the transformed phenotype of breast cancer cells (7–10).

Oxidative stress can be considered an example of antagonistic pleiotropy. Reactive oxygen species have been shown to mediate normal biological functions, such as smooth muscle cell contraction, regulation of protein tyrosine phosphatase activity, generation of thyroid hormone, cell differentiation, and mitogenic signaling (reviewed in ref. 11). However, oxidative stress has been shown to induce premature senescence in culture not dependent upon telomere shortening (12, 13). Because oxidative stress induces cellular senescence and free radical levels within cells increase over time, it is possible that oxidative stress may promote the deleterious effects of cellular senescence, such as aging and chronic diseases, including cancer.

Caveolae are invaginations of the plasma membrane enriched in cholesterol. Caveolin-1 is the structural protein component of caveolar membranes. Caveolin-1 acts as a scaffolding protein to concentrate and functionally regulate signaling molecules (14, 15). Several independent lines of evidence indicate that caveolin-1 may act as an antiproliferative protein (16–19). Consistent with this idea, we have previously shown that overexpression of caveolin-1 is sufficient to arrest mouse embryonic fibroblasts (MEF) in the G₀-G₁ phase of the cell cycle, reduce their proliferative life span, and promote premature cellular senescence (20, 21). In addition, we have shown that oxidative stress induces premature senescence by stimulating caveolin-1 gene transcription through p38 mitogen-activated protein kinase/Sp1-mediated activation of two GC-rich promoter elements (22). However, the signaling pathway that links caveolin-1 to stress-induced premature senescence (SIPS) remains unknown. In addition, the possible role of caveolin-1 as an antagonistic pleiotropic protein is unexplored.

Here, we propose caveolin-1 as a novel antagonistic pleiotropic molecule that contributes to explain, at the molecular and cellular levels, the antagonistic pleiotropic properties of oxidative stress, cellular senescence, and p53.

Glutathione S-transferase fusion proteins pull-down assay. The glutathione S-transferase (GST)–caveolin-1 (GST-Cav-1) fusion protein constructs were transformed into Escherichia coli (BL21 strain; Novagen, Inc.). After induction of expression through addition of 5 mmol/L isopropyl-β-d-galactoside (Sigma), GST-Cav-1 constructs were affinity purified on glutathione-agarose beads using the detergent sarcosyl for initial solubilization. Equal amounts of GST-Cav-1 and GST alone were incubated overnight at 4°C with cell lysates. After binding, the beads were extensively washed and resuspended in 3× sample buffer and subjected to SDS-PAGE.

Coculture studies. Three independent clones of either wild-type (WT) or caveolin-1 null MEFs were mixed and cultured as one population. Cells were plated into 100-mm dishes and subjected to oxidative stress when ∼50% confluent. Oxidative stress was induced by treating MEFs with 150 μmol/L hydrogen peroxide for 2 h. After hydrogen peroxide treatment, cells were washed with PBS and cultured in complete medium for 4 d. MEFs were then serum starved for 3 d. Serum-starved Ras^G12V-transformed NIH 3T3 (37,500 cells) or MDA-MB-231 (37,500 cells) cells were layered on top of serum-starved MEFs and cultured for 7 d. Ras cocultures were quantified by counting the number of (a) nuclei after 4′,6-diamidino-2-phenylindole (DAPI) staining and (b) Ki67-positive cells in 30 random fields per experimental point. MDA-MB-231 cocultures were quantified by counting the number of colonies after crystal violet staining. Crystal violet staining was performed by incubating the cells with 10% crystal violet in 70% ethanol for 2 min followed by extensive washes with PBS. Quantification of crystal violet staining was performed as follows: the image was preprocessed by cropping the central area of each plate, converting to the HSV colorspace and finding connected regions of pixels with saturation of >0.2 on a [0, 1] scale. To reduce noise, only colonies with an area of >32 pixels were counted (≥0.3 mm²).

Extracellular matrix and soluble factors. WT and caveolin-1 null MEFs were cultured and treated as listed above for coculture studies. Conditioned medium was then collected from respective cultures and saved. Plates were washed twice with PBS, and cells were removed by incubating in Cell Dissociation Buffer (enzyme free, PBS based; Life Technologies). Dishes were washed with PBS twice to remove any residual debris, cells, or dissociation buffer. Conditioned medium was replaced onto the respective dishes. Ras^G12V-transformed NIH 3T3 cells (37,500) were then added on the dish containing extracellular matrix and soluble factors. Cells were grown for 7 d. Results were quantified by counting the number of either nuclei after DAPI staining or Ki67-positive cells in 30 random fields per experimental point.

Tumorigenesis assays. The animal protocol described in this article were reviewed and approved by the Institutional Animal Care and Use Committee at the University of Pittsburgh. Nude (nu/nu) mice (5–6 wk from Charles River) were injected s.c. into the dorsal flap with either 3.25 × 10⁵ Ras^G12V-transformed NIH 3T3 or MDA-MB-231 cells alone or with 1 × 10⁶ MEFs (volume, 100 μL). 1 × 10⁶ MEFs alone were injected as a negative control. Four weeks after the injection, tumors were excised and the x and y axes were measured to determine the tumor volume (size) according to the following formula: volume = 0.5 × width² × length.

Oxidative stress promotes interaction between caveolin-1 and Mdm2. Oxidative stress is a well-established inducer of cellular senescence in culture (12, 13). To begin to investigate the oxidant-promoted and caveolin-1–dependent signaling events that are required for SIPS, we treated WI-38 human diploid fibroblasts with subcytotoxic levels of hydrogen peroxide for 2 hours. Cells were then washed and cultured in normal medium for different periods of time. We found that oxidative stress promoted premature senescence of WI-38 cells, as shown by activation of the p53/p21^Waf1/Cip1 pathway (Fig. 1A) and senescence-associated β-galactosidase (SA-β-gal) activity staining (Fig. 1B). To investigate the molecular mechanism underlying activation of p53 after oxidative stress, we focused on the modulation of Mdm2, a well-known negative regulator of p53, by caveolin-1. We found that oxidative stress promoted the interaction between endogenous caveolin-1 and Mdm2, as shown by coimmunoprecipitation studies in WI-38 human diploid fibroblasts (Fig. 1C). Consistent with this result, whereas caveolin-1 was mainly expressed at the plasma membrane and endogenous Mdm2 in the nucleus under resting conditions, Mdm2 was found in caveolin-1–enriched domains at the plasma membrane and in the cytoplasm after oxidative stress in WI-38 cells (Fig. 1D and Supplementary Fig. S1A–C). In support of these data, we show in Supplementary Fig. S1D, E, and F that nuclear p53 levels were low in cells wherein Mdm2 was expressed in the nucleus before oxidative stress and elevated in cells wherein Mdm2 left the nucleus upon oxidant stimulation. Thus, by sequestering Mdm2 away from p53, caveolin-1 seems to stabilize p53 after oxidative stress.

The scaffolding domain of caveolin-1 interacts with the caveolin binding domain of Mdm2 in vitro. To determine what domain of caveolin-1 is required for the interaction with Mdm2, we performed pull-down assays using a series of caveolin-1 deletion mutants fused to GST. Figure 2A and B and Supplementary Fig. S2A show that the caveolin scaffolding domain of caveolin-1 (residues 82–101) is sufficient for binding to Mdm2. Caveolin-1 directly binds to signaling molecules through a direct interaction between the caveolin scaffolding domain of caveolin-1 and caveolin binding motifs of signaling molecules. These caveolin binding motifs include crucial aromatic amino acid residues that are separated by a variable number of X residues (23). A list of peptide ligands previously identified by screening phage display libraries using a GST fusion protein containing the caveolin-1 scaffolding domain is shown in Supplementary Fig. S2B. Interestingly, both mouse and human Mdm2 have a potential caveolin binding motif between amino acids 48 and 60 (Supplementary Fig. S2B). To directly assess the importance of residues 48 to 60 of Mdm2 in binding to caveolin-1, we generated and expressed an Mdm2 mutant (Φ→A Mdm2), in which the four aromatic residues of the putative caveolin binding motif were mutated to alanines (Fig. 2C). GST pull-down assays show that WT Mdm2, but not Φ→A Mdm2, bound to the caveolin-1 scaffolding domain (Fig. 2D and Supplementary Fig. S2C), indicating that residues 48 to 60 of Mdm2 indeed represent a caveolin binding motif.

Caveolin-1 expression activates p53 by preventing binding of Mdm2 to p53. Interestingly, the caveolin binding motif within the Mdm2 sequence overlaps with the p53 binding domain (residues 23–108; Supplementary Fig. S3). Thus, we postulated that, by binding to Mdm2, caveolin-1 may prevent binding between Mdm2 and p53. To test this hypothesis, we exogenously expressed Mdm2 in fibroblasts with increasing amount of a caveolin-1–expressing vector. The interaction between Mdm2 and p53 was then assessed by coimmunoprecipitation analysis. Figure 3A illustrates that Mdm2 immunoprecipitated p53 in the absence of overexpression of caveolin-1. In contrast, Mdm2 immunoprecipitated caveolin-1, but not p53, when caveolin-1 was overexpressed (Fig. 3A). As a result of less interaction between Mdm2 and p53 after caveolin-1 overexpression, total endogenous p53 expression was increased (Fig. 3A).

Expression of the caveolin-1 scaffolding domain is sufficient to activate the p53/p21^Waf1/Cip1 pathway and induce premature senescence. Because the interaction between caveolin-1 and Mdm2 is mediated by the scaffolding domain of caveolin-1 (Fig. 2B) and overexpression of caveolin-1 is sufficient to prevent the Mdm2/p53 interaction and stabilize p53 (Fig. 3A), we then asked whether the expression of the caveolin-1 scaffolding domain is sufficient to activate the p53/p21^Waf1/Cip1 pathway. To this end, we took advantage of the cell permeable peptide cavtratin, consisting of the homeodomain of the Drosophila transcription factor antennapedia (penetratin), coupled to the caveolin-1 scaffolding domain (24). Expression of cavtratin, but not the control peptide antennapedia lacking the caveolin scaffolding domain, up-regulated the protein expression of p53 and p21^Waf1/Cip1, a downstream target of p53, in WI-38 cells (Fig. 3B). Interestingly, expression of cavtratin was sufficient to induce premature senescence in WI-38 cells (Fig. 3C).

Lack of caveolin-1 prevents the oxidant-induced activation of the p53/p21^Waf1/Cip1 pathway and induction of premature senescence. Based on the above data, one would not expect a full activation of the p53/p21^Waf1/Cip1 pathway after oxidative stress in cells that do not express caveolin-1. To address this point, MEFs were derived from WT and caveolin-1 null mice that do not express caveolin-1 and were subjected to oxidative stress. Endogenous Mdm2 was immunoprecipitated with a monoclonal specific antibody probe and immunoprecipitates subjected to immunoblotting analysis using polyclonal antibodies against either caveolin-1 or p53. Figure 4A shows that oxidative stress promoted the interaction between caveolin-1 and Mdm2 in WT but not caveolin-1 null MEFs. Interestingly, the interaction between Mdm2 and p53, as assessed by coimmunoprecipitation studies, was significantly higher in caveolin-1 null MEFs compared with WT MEFs (Fig. 4A). Consistent with the notion that Mdm2 is a negative regulator of p53, we show in Fig. 4B that the oxidant-induced stabilization of p53 is dramatically compromised in the absence of caveolin-1. Thus, more interaction between Mdm2 and p53 in caveolin-1 null cells results in reduced total p53 expression. Reduced activation of p53 in caveolin-1 null cells was confirmed by reduced up-regulation of p21^Waf1/Cip1 after oxidative stress (Fig. 4B). Because activation of the p53/p21^Waf1/Cip1 pathway plays a key role in SIPS, we subjected WT and caveolin-1 null MEFs to SA-β-gal staining at different time points after oxidative stress. Figure 4C and D shows that the number of SA-β-gal–positive cells was significantly reduced in caveolin-1 null MEFs compared with WT MEFs.

We conclude that caveolin-1 expression is required for activation of the p53 pathway and induction of premature senescence after oxidative stress (Supplementary Fig. S4).

Caveolin-1 mediates the antagonistic pleiotropic properties of cellular senescence. To investigate whether caveolin-1 represents a key mediator of the antagonistic pleiotropic properties of cellular senescence, we asked whether (a) caveolin-1 expression in cancer cells induces senescence and, thus, inhibits cancer cell growth and (b) caveolin-1–dependent senescence of fibroblasts promotes proliferation of cancer cells and tumorigenesis.

To this end, caveolin-1 was overexpressed in oncogenic Ras (Ras^G12V)–transformed NIH 3T3 cells, a well-established model system used to study cell transformation, wherein caveolin-1 mRNA and protein levels are dramatically reduced (17). We show in Supplementary Fig. S5A and Fig. 5A that overexpression of caveolin-1 in these cells was sufficient to induce cellular senescence compared with Ras^G12V-transformed NIH 3T3 cells expressing vector alone.

We then subjected WT and caveolin-1 null MEFs to oxidative stress to induce cellular senescence (150 μmol/L H₂O₂ for 2 hours, followed by culturing of cells for 7 days in normal medium) and examined the ability of these cells to promote cell proliferation of Ras^G12V-transformed NIH 3T3 cells in coculture studies. Ras^G12V-transformed NIH 3T3 cells were seeded onto an equal number of hydrogen peroxide–treated WT and caveolin-1 null cells (50–70% confluent lawns). Both Ras^G12V-transformed NIH 3T3 cells and MEFs were serum-starved for 72 hours before coculture experiments to (a) avoid any influence of growth factors contained in culture medium; (b) arrest caveolin-1 null MEFs, which otherwise would have proliferated given the fact that oxidative stress failed to induce senescence in the absence of caveolin-1 (Fig. 4); and (c) arrest the proliferation of Ras^G12V-transformed NIH 3T3 cells.

We first examined cell proliferation by staining the cocultures with DAPI and quantified the number of smaller and more intensely stained Ras^G12V-transformed NIH 3T3 nuclei (Supplementary Fig. S5B and C). Alternatively, actively proliferating Ras^G12V-transformed NIH 3T3 cells were quantified by markers of cellular proliferation: Ki67 staining and BrdUrd incorporation (Supplementary Fig. S5B, Fig. 5B, and Supplementary Fig. S5D, respectively). Growth-arrested MEFs resulted negative for Ki67 staining (data not shown). We show that cellular proliferation of Ras^G12V-transformed NIH 3T3 was dramatically enhanced when cocultured with WT MEFs subjected to oxidative stress compared with Ras^G12V-transformed NIH 3T3 cells alone (Supplementary Figs. S5B–D and Fig. 5B). Importantly, the ability of hydrogen peroxide–treated caveolin-1 null MEFs to potentiate the growth of Ras^G12V-transformed NIH 3T3 cells was dramatically inhibited (Supplementary Figs. S5B–D and Fig. 5B).

The ability of senescent MEFs to stimulate proliferation of cancer cells in a caveolin-1–dependent manner was not limited to Ras^G12V-transformed fibroblasts. We show in Supplementary Fig. S5E that H₂O₂-treated/serum-starved WT MEFs promoted a 3.5-fold increase of cell growth of serum-starved MDA-MB-231, a breast cancer epithelial cell line, as assessed by counting the number of colonies formed in a Petri dish after DAPI staining (colonies were defined as groups of cells containing at least 20 cells). In contrast, the capacity of caveolin-1–lacking MEFs subjected to oxidative stress to enhance colony formation of MDA-MB-231 cells was reduced by ∼30% compared with H₂O₂-treated caveolin-1–expressing MEFs (Supplementary Fig. S5E). A more striking difference was noted when only larger colonies (≥0.3mm²) were detected by crystal violet staining. In fact, no colonies were formed by serum-starved MDA-MB-231 cells, WT MEFs, and caveolin-1 null MEFs alone (Fig. 5C and Supplementary Fig. S5F). Colonies formed when MDA-MB-231 cells were cocultured with senescent WT MEFs (Fig. 5C and Supplementary Fig. S5F). The number of colonies was dramatically reduced (250% reduction) when MDA-MB-231 cells were cocultured with caveolin-1 null MEFs subjected to oxidative stress (Fig. 5C and Supplementary Fig. S5F) compared with H₂O₂-treated WT MEFs.

Secreted factors mediate the ability of caveolin-1–expressing senescent MEFs to stimulate the proliferation of RasG12V-transformed NIH 3T3 cells. How do caveolin-1–expressing senescent MEFs stimulate the proliferation of Ras^G12V-transformed NIH 3T3 cells? Stimulation of cell proliferation may occur by either direct cell-cell interaction and/or secreted insoluble extracellular matrix factor/secreted soluble factor contained in the medium. To directly address these possibilities, we evaluated the proliferation properties of serum-starved Ras^G12V-transformed NIH 3T3 cells after culturing the cells with conditioned medium from hydrogen peroxide–treated WT and caveolin-1 null MEFs (cells were treated as described above for coculture studies) onto the same dishes used to grow MEFs after the cells were removed by calcium chelation without altering the extracellular matrix layer. We show in Supplementary Fig. S5G and Fig. 5D that the combination of conditioned medium/extracellular matrix factors from H₂O₂-treated WT but not caveolin-1 null MEFs stimulated the proliferation of Ras^G12V-transformed NIH 3T3 cells. These results indicate that secreted factors released by senescent MEFs in a caveolin-1–dependent fashion are sufficient to stimulate Ras^G12V-transformed NIH 3T3 cell proliferation.

Caveolin-1 expression is required for senescent fibroblast-induced tumorigenesis in vivo. To assess whether caveolin-1 plays an important role in tumorigenesis promoted by senescent fibroblasts in vivo, we injected either Ras^G12V-transformed NIH 3T3 or MDA-MB-231 cells, alone or with hydrogen peroxide-treated WT and caveolin-1 null MEFs, into immunocompromised (nu/nu) mice. Mice were sacrificed 4 weeks later, and tumor size was determined. Injection of either WT or caveolin-1 null MEFs alone was done as controls and did not form tumors (data not shown). As expected, both Ras^G12V-transformed NIH 3T3 (Supplementary Fig. S6A and B) and MDA-MB-231 (Fig. 6A and B) cells alone formed tumors 4 weeks after injection. Interestingly, WT but not caveolin-1 null MEFs that were subjected to oxidative stress potentiated tumorigenesis of both Ras^G12V-transformed NIH 3T3 (Supplementary Fig. S6A and B) and MDA-MB-231 (Fig. 6A and B) cells. The small increase of tumor size that caveolin-1 null MEFs induced when cocultured with MDA-MB-231, but not Ras^G12V-transformed NIH 3T3 cells, may either reflect a variation due to the different cancer cell line used in the experiment or indicate that, in the case of MDA-MB-231 cells, there may be a caveolin-1–independent pathway in fibroblasts that contributes, although minimally, to potentiation of tumorigenesis. Together, these data indicate that senescent fibroblasts create a microenvironment that stimulates the growth of neoplastic cells in vivo in a caveolin-1–dependent manner.

The p53 pathway responds to a variety of intrinsic and extrinsic stress signals that can disrupt the fidelity of DNA replication (25). Activation of p53 initiates a program that may induce cellular senescence (26). Thus, stresses that would compromise genome stability can be accommodated by preventing cell cycle progression through p53 activation. This is a tumor suppressor action of p53 that is beneficial to young organisms, as shown by the fact that the p53 gene is mutated in ∼50% of all human cancers (27). However, p53 action has been linked to the aging process (28). Mouse models of gain of p53 function show increased tumor suppression but decreased longevity and early aging phenotypes (29, 30). Thus, the p53 tumor suppressor gene is believed to act in an antagonistically pleiotropic manner (31). Interestingly, Scrable and colleagues have analyzed data from mouse models of accelerated aging and determined that the harmful effects of p53 on aging are not attributable only to the p53 gene but involve also the NAD⁺-dependent deacetylase SIRT1 (28), suggesting that genes that have modifying effects on p53 activity but evolve independently may contribute to the deleterious actions of p53 late in life.

What are the signals that activate p53? These include γ or UV irradiation, alkylation of bases, depurination of DNA, and reaction with oxidative free radicals (32). Activation of p53 occurs through different mechanisms, including dissociation of Mdm2 from p53. In fact, p53 is a target of the E3 ubiquitin ligase activity of Mdm2, resulting in proteosomal degradation of p53 (33). Mdm2 also suppresses p53 transcriptional activity and shuttles p53 out of the nucleus (34).

Here, we showed that caveolin-1 is a novel Mdm2 binding protein. We found that oxidative stress promotes the interaction between caveolin-1 and Mdm2, which is sequestered in caveolar membranes away from p53. We identified amino acids 48 to 60 of Mdm2 as the caveolin-1 binding motif. Interestingly, the caveolin-1 binding motif of Mdm2 overlaps with the p53 binding domain of Mdm2 (amino acids 23–108). We showed that, by binding to Mdm2, caveolin-1 prevents the interaction between p53 and Mdm2. As a result, p53 gets stabilized and p21 is up-regulated. Activation of this pathway is compromised in caveolin-1 null MEFs, which do not express caveolin-1 and do not undergo cellular senescence after oxidative stress. These data are consistent with several independent lines of evidence, suggesting that caveolin-1 may function as a “tumor suppressor protein” in mammalian cells (35, 36).

H-ras G12V is a mutation that accumulates in normal adult tissue and represents one of the most common somatic mutations that lead to cancer development (37). We show here that overexpression of caveolin-1 in caveolin-1–negative Ras^G12V-transformed NIH 3T3 cells is sufficient to induce cellular senescence. These data are consistent with previously published data, wherein we show that an adenoviral-mediated overexpression of caveolin-1 in caveolin-1–negative MCF7 breast cancer epithelial cells was sufficient to restore stress-induced cellular senescence (22). Because cellular senescence is a natural tumor suppressor mechanism (38–42), it may represent one of the molecular mechanisms through which caveolin-1 acts as a tumor suppressor protein. Because most malignant tumors eventually acquire cells that can proliferate indefinitely and, in many cases, are resistant to senescence-inducing signals (43), our data indicate that caveolin-1 replacement therapy in caveolin-1–negative cancer cells may represent a way to revert these cells to a “senescence-prone” phenotype, making them good targets of therapeutic drugs.

We also found that caveolin-1 expression in senescent fibroblasts is required for their ability to secrete insoluble extracellular matrix and/or soluble growth factors and accelerate the proliferation and tumorigenesis of neoplastic cells in vitro and in vivo, respectively. These data are consistent with the known secretory properties of senescent fibroblasts and support the theory that the senescent microenvironment may contribute, together with multiple factors including somatic mutations, to cancer development in older individuals. Based on these data, we propose caveolin-1 as a major regulator of the antagonistic pleiotropic properties of cellular senescence (Fig. 6C). We can speculate that interventions aimed at targeting caveolin-1 expression in senescent fibroblasts may represent a novel therapeutic approach, in combination with chemotherapy and radiation, for the treatment of late life cancer.

No potential conflicts of interest were disclosed.

Grant support: NIH grant R01AG022548 (F. Galbiati) and NIH Predoctoral Training Grant in Pharmacological Sciences grant T32GM008424 and NRSA grant F31 AG032182 (J.N. Bartholomew).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dave Bradley for the quantification of crystal violet staining.