Combined loss of INK4a and caveolin-1 synergistically enhances cell proliferation and oncogene-induced tumorigenesis: role of INK4a/CAV-1 in mammary epithelial cell hyperplasia.

Tumorigenesis is a multistep process that involves a series of genetic changes or "multiple hits," leading to alterations in signaling, proliferation, immortalization, and transformation. Many of the molecular factors that govern tumor initiation and progression remain unknown. Here, we evaluate the transformation suppressor potential of caveolin-1 (Cav-1) and its ability to cooperate with a well established tumor suppressor, the INK4a locus. To study the effects of loss of caveolin-1 on cellular transformation, we established immortalized primary mouse embryonic fibroblasts (MEFs) expressing and lacking caveolin-1 by interbreeding Cav-1 (+/+) and Cav-1 (-/-) mice with INK4a (-/-) mice. Analysis of these cells reveals that loss of caveolin-1 confers a significant growth advantage, as measured via cellular proliferation and cell cycle analysis. Loss of caveolin-1 in the INK4a (-/-) genetic background results in constitutive hyperactivation of the p42/44 MAP kinase cascade, decreased expression of p21(Cip1), as well as cyclin D1 and PCNA overexpression, consistent with their hyperproliferative phenotype. Importantly, in cells lacking Cav-1 expression, transformation by activated oncogenes (H-Ras(G12V) or v-Src) results in increased tumor growth in vivo (up to >40-fold). Finally, INK4a (-/-)/Cav-1 (-/-) mice demonstrate disturbed mammary epithelial ductal morphology, with hyperplasia, increased side-branching, and fibrosis. Our results provide important new evidence for the transformation suppressor properties of Cav-1 and the first molecular genetic evidence that Cav-1 cooperates with a tumor suppressor, namely the INK4a genetic locus.

Caveolin-1 (Cav-1) 1 was originally identified as a phosphorylated target of v-Src in Rous sarcoma virus-transformed avian fibroblasts (1). Since then, Cav-1 has been shown to be an important structural marker protein for caveolae, flask-shaped membrane invaginations or microdomains that are enriched in cholesterol and sphingolipids (2,3). While the functions of caveolae are currently being elucidated, there is clear evidence that they are involved in vesicular trafficking and signal transduction (4). The signaling properties of caveolae were revealed when it was observed that a number of signaling molecules are localized to caveolae membranes, including Src family tyrosine kinases, H-Ras, epidermal growth factor receptor, endothelial nitric-oxide synthase, and G-proteins (5,6). In addition, caveolin-1 possesses a specific modular protein domain, termed the "caveolin scaffolding domain" (CSD) that serves to bind and inactivate many of these signal transducing molecules, thereby negatively influencing their downstream signaling capacity (3,7). Disruption of caveolae or down-regulation of caveolin-1 is sufficient to activate mitogenic signaling via many of these pathways (6,8).
The hypothesis that Cav-1 possesses a transformation suppressor function has emerged due to an accumulation of findings from genetic, molecular, and clinical approaches (9). The human CAV-1 gene is localized to the D7S522 locus on human chromosome 7q31.1, a region that is commonly deleted in a number of human malignancies, and sporadic CAV-1 mutations (P132L and C133R) have been identified in up to 16% of patients with primary breast cancer (10, 11). Caveolin-1 mRNA and protein levels have also been found to be down-regulated in a number of human primary tumors and tumor-derived cell lines (9). Likewise, Cav-1 mRNA and protein levels are dramatically down-regulated in NIH 3T3 cells transformed with a variety of activated oncogenes such as v-Abl, Bcr-abl, H-Ras G12V , and polyomavirus middle T antigen, suggesting that caveolin-1 expression may be important in suppressing cell transformation (12).
The INK4 family of proteins (p16 INK4a , p15 INK4b , p18 INK4c , and p19 INK4d ) is comprised of cyclin-dependent kinase inhibitors (CKIs) that negatively regulate progression through the G 1 phase of the cell cycle by binding to and inhibiting cyclin D/cdk4 -6 complexes (13). Cyclin D/cdk4 -6 normally inactivates Rb and other related family members by phosphorylation during the mid-G 1 phase, allowing for transcription of genes required for entry into S-phase (13). The INK4a locus is a well characterized tumor suppressor locus that, when mutated, confers increased susceptibility to a number of human malignancies, including melanomas and pancreatic carcinomas (14 -18). In terms of human cancers, INK4a and p53 account for the majority of inactivating mutations present in tumors regardless of type, age, or location (19,20). The importance of INK4a as a tumor suppressor locus was dramatically confirmed in mice when its disruption by homologous recombination resulted in mice with increased susceptibility toward carcinogenic treatments; they also developed spontaneous tumors at an earlier age (21). Surprisingly, ablation of this single locus, namely INK4a, was sufficient to result in the immortalization of mouse embryonic fibroblasts (MEFs) in culture, thereby providing one of the most important properties first acquired by transformed cells; that is, the ability to proliferate continuously and to escape senescence (21). Interestingly, the INK4a locus encodes two completely distinct and non-overlapping proteins, p16 INK4a and p19 ARF , because of alternative exon splicing and overlapping reading frames (22). Both proteins are involved in cell cycle control through different, yet slightly overlapping, regulatory pathways (23). Additionally, antisensemediated inactivation of p16 INK4a or p19 ARF is sufficient to immortalize MEFs indefinitely (23). Each protein also possesses tumor suppressor function in vivo as targeted deletion of either protein in mice, with retention of the other, resulted in increased susceptibility toward spontaneous and/or carcinogen-induced tumors (24 -26).
To elucidate the role of caveolin-1 during early cell immortalization and transformation, we now evaluate whether loss of caveolin-1 gene expression shows cooperativity with another well established tumor suppressor gene, the INK4a locus. Here, we have designed a simple experimental strategy based on INK4a (Ϫ/Ϫ)-null mice to generate immortalized primary cell lines lacking or expressing caveolin-1.
An overview of this molecular genetic strategy is presented in Fig. 1. First, we generated mice that are deficient in both INK4a and Cav-1 (INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) double knockout (dKO) mice). For this purpose, we used INK4a (Ϫ/Ϫ) mice that lack both the p16 INK4a and p19 ARF protein products, as inactivation of the entire INK4a locus occurs in patients, is more clinically relevant, and produces far better stable immortalization experimentally. Then, we used the resulting INK4a (Ϫ/Ϫ)/ Cav-1 (Ϫ/Ϫ) mouse embryos to derive specific fibroblast cell lines (MEFs) that are clearly immortalized, because of the loss of the INK4a locus. In doing so, we have established immortalized MEFs both expressing and lacking Cav-1. These novel MEF cell lines have defined genetic changes, as opposed to generating immortalized MEFs through the 3T3 protocol. Unfortunately, the 3T3 protocol selects for cells with multiple unknown genetic changes that ultimately lead to cellular immortalization and even cell transformation. Immortalized MEFs generated using the 3T3 protocol have been found to have various genetic abnormalities in p53, Rb, INK4 family proteins, Ras, etc.
We now show that analysis of these INK4a-immortalized MEFs reveals that the combined loss of INK4a and Cav-1 results in a significant proliferative advantage, compared with the individual loss of either Cav-1 or INK4a alone. Interestingly, INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) MEFs show constitutive hyperactivation of the p42/44 MAP kinase cascade, with decreased p21 Cip1 expression and cyclin D1 overexpression, providing several potential molecular mechanisms to account for their hyperproliferative phenotype. However, INK4a (Ϫ/Ϫ)/ Cav-1 (Ϫ/Ϫ) MEFs are immortalized, but not transformed, as they fail to form tumors when injected subcutaneously into nude mice.
Cav-1 shows cooperativity with tumor suppressor genes, during both immortalization and cell transformation.
Animal Studies-All animals used for these experiments were in the C57Bl/6 background. Mice were housed and maintained in a barrier facility at the Institute for Animal Studies, Albert Einstein College of Medicine (AECOM). Animal protocols used for this study were approved by the AECOM Institute for Animal Studies. Mice were kept on a 12-hour light/dark cycle with ad libitum access to chow (Picolab 20, PMI Nutrition International) and water. Cav-1 KO mice were generated, as we previously described (28). INK4a KO mice were generated by Dr. Ron DePinho (Harvard Medical School), and p53 KO mice were obtained commercially from JAX. Genotyping of Cav-1 and p53 mice was performed by PCR, as previously described (28,29). INK4a locus genotyping for mice and cell lines was performed using a three primer PCR strategy. The sequences (5Ј to 3Ј) of the three primers were as follows: 1) INK4a MTS-1 forward: TCCCTCTACTTTTTCTTCTGAC, 2) INK4a MTS-1 reverse: CGGAACGCAAATATCGCAC, 3) INK4a Null: CTAGT-GAGACGTGCTACTTC. General thermocycling conditions were used with an annealing temperature of 55°C and a total of 33 cycles. The wild-type INK4a band (278 bp) and the knockout INK4a band (313 bp) were visualized using 2.0% agarose gels.
MEF Isolation-Primary MEFs were isolated on day E13.5 by using sterile techniques to remove the mouse embryos in utero. Following decapitation and liver removal, the cells were dispersed by thoroughly mincing the remaining tissue in a 100-mm dish, using a sterile razor blade and pipetting with a 1000-l tip or 18-gauge needle. Cells were then trypsinized in 1 ml of 0.05% trypsin, 0.53 mM EDTA (Invitrogen, Life Technologies, Inc.) for 20 min at 37°C. 10 ml of complete medium (DMEM, 10% fetal bovine serum (FBS), glutamine, penicillin/streptomycin) were added to the cells to inactivate trypsin and to resuspend the dissociated cells. Cells were cultured at 37°C in a 5% CO 2 incubator (passage 1). After several days, the MEF cells were passaged and a portion of the cells were frozen in liquid nitrogen (passage 2).
Cell Culture and Proliferation Assays-All cells were cultured in DMEM, supplemented with 10% fetal bovine serum, 2 mM glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. For measuring the growth rate of the MEFs, 10 5 cells from passage 4 were seeded in multiple 60-mm sterile culture dishes. One dish was counted each day at the same time by trypsinizing the cells, resuspending them in an appropriate amount of medium, and counting a 10-l aliquot using a hemocytometer. The medium was changed each day for the plates that were not counted. At least three independent MEF cell isolates were assayed per genotype. Statistical significance (p value) was determined by the Student's t test.
Flow Cytometric Analysis-For cell cycle analysis, 10 6 cells were plated on a 10-cm dish and cultured under normal growth conditions for 24 h. Asynchronously growing cells were then trypsinized, washed twice with medium, and resuspended in 1 ml of PBS. In order to fix the cells, 9 ml of ice-cold 80% ethanol was added slowly to each sample, and then the cells were incubated for 30 min on ice. Fixed cells were then washed twice with PBS and resuspended in 1 ml of PBS containing 10 g/ml propidium iodide (Sigma) and 0.25 mg/ml RNase A (Sigma) and incubated for 30 min at room temperature. Cells were then subjected to univariate cell cycle analysis using a BD Biosciences FACScan flow cytometer. Analysis of the G 0 /G 1 , S, and G 2 /M cell cycle phase fractions was performed using CELLQUEST software. For each MEF cell isolate, three independent experiments were performed that yielded virtually identical results.
Immunoblot Analysis-Tissues were harvested, snap frozen in liquid nitrogen, and then 150 -200 mg of tissue was homogenized in 2 ml of lysis buffer (10 mM Tris, pH 7.5, 150 mM NaCl, 1% Triton X-100, 60 mM octylglucoside), containing protease inhibitors (Roche Applied Science). For cell lysates, MEFs were plated at a density of ϳ1-2 ϫ 10 6 cells in complete medium and cultured for 18 -24 h. Subconfluent cells were then collected into an appropriate volume of lysis buffer. For phosphospecific immunoblotting, cells were scraped into boiling lysis buffer to denature endogenous phosphatases. Tissue or cell lysates were centrifuged at 12,000 ϫ g for 10 min to remove insoluble debris. Protein concentrations were analyzed using the BCA reagent (Pierce), and the volume required for 20 g of protein was determined. Samples were then separated by SDS-PAGE (10 or 12% acrylamide) and transferred to 0.2 m of nitrocellulose. The nitrocellulose membranes were stained with Ponceau S (to visualize protein bands), followed by immunoblot analysis. All subsequent wash buffers contained 10 mM Tris, pH 8.0, 150 mM NaCl, 0.05% Tween-20, which was supplemented with 1% bovine serum albumin (BSA) and 4% nonfat dry milk (Carnation) for the blocking solution and 1% BSA for the antibody diluent. Primary antibodies were used at a 1:500 -1,000 dilution. Horseradish peroxidase-conjugated secondary antibodies (anti-mouse 1:10,000 dilution (Pierce) or anti-rabbit 1:5,000 dilution (BD Pharmingen)) were used to visualize bound primary antibodies with the Supersignal chemiluminescence substrate (Pierce).
In Vivo Reporter Assays-MEFs were plated at a density of 2 ϫ 10 5 cells per well in 6-well plates in complete growth medium. The following day, the cells were transiently transfected with 0.5 g of the cyclin D1 promoter or the c-Fos responsive element reporter construct per well using the Effectene Transfection Reagent (Qiagen, Valencia, CA). A pSV-␤-gal plasmid (Promega), an SV40-driven vector expressing ␤-galactosidase, was co-transfected to serve as an internal control for transfection efficiency. Thirty-six hours post-transfection, the cells were lysed in 200 l of extraction buffer, 100 l of which was used to measure luciferase activity, as described (30). Another 50 l of the cell lysate was used to conduct a ␤-galactosidase assay, as previously described (31). Each experimental value, expressed as relative light units from a 30-s count, has been normalized using its respective ␤-galactosidase activity and represents the mean of three separate transfections performed in parallel. All experiments were independently performed three times and yielded virtually identical results.
Retroviral Transduction-Retroviral constructs, pBABE-H-Ras G12V or pBABE-v-Src, were transfected into Phoenix cells using calcium phosphate precipitation. Approximately 6 -8 h post-transfection, cells were washed once with PBS, and the media was changed to complete growth medium (DMEM, 10% FBS, glutamine, penicillin/streptomycin). Forty-eight hours post-transfection, the supernatant from transfected Phoenix cells was harvested, passed through a 0.45-m filter to remove cellular debris, and mixed with polybrene (5 g/ml). 2 ϫ 10 5 cells of each cell line were then infected with a 1:1 mixture of viralcontaining supernatant and complete growth medium. Aspiration and infection was repeated every 12 h for a total of 3 cycles. To generate stable H-Ras G12V -and v-Src-transfected cell lines, cells were then placed in complete growth medium supplemented with puromycin (2.5 g/ml) for 1 week. A stable pool of transfected MEFs was generated in this fashion. Transfections were also performed in parallel using the empty vector control plasmid (pBABE-puro).
Soft Agar Assays-Anchorage-independent growth was assessed, essentially as we previously described (12). Briefly, 10 4 cells of each cell line were suspended in 2 ml of DMEM containing 10% FBS and 0.33% SeaPlaque low melting temperature agarose. These cells were then plated over a 3-ml layer of solidified DMEM containing 10% FBS and 0.5% agarose, and the cells were allowed to settle to the interface between these layers at 37°C for 30 min. The plates (at least ten for each MEF cell line) were then allowed to harden at room temperature for at least 30 min before returning to the 37°C incubator. The plates were fed every 2-3 days by overlaying with 2 ml of complete medium containing 0.33% agarose. After 10 -14 days, colonies were photographed under low magnification (ϫ10) using a stereomicroscope.
Tumorigenicity in Nude Mice-Female athymic immunodeficient (nude) mice were obtained at 6 -8 weeks of age from the NCI, National Institutes of Health. Briefly, MEFs were suspended in sterile PBS at a concentration of 10 7 cells/ml, and 100 l were injected subcutaneously into each flank of the nude mice. Mice were sacrificed at 2-3 weeks, depending on the cell line (2 weeks for v-Src-transfected versus 3 weeks for H-Ras G12V -transfected) at which point tumors were excised to determine their weight and size using calipers. The formula (X 2 Y)/2, where X is the length and Y is the width, was used in order to estimate tumor volume. At least 10 injections were performed for each cell line.
Whole Mount Analysis of Mammary Glands-Fourth (inguinal) mammary glands from 4-month-old mice were excised, spread onto glass slides, fixed, and stained essentially as we previously described (32). Briefly, mammary glands were fixed in Carnoy's fixative (6 parts of 100% ethanol: 3 parts of CHCl3: 1 part of glacial acetic acid) for 2-4 h at room temperature. The samples were then washed in 70% ethanol for 15 min and changed gradually to distilled water. Once hydrated, the mammary squashes were stained overnight in carmine alum (1 g of carmine (Sigma C1022) and 2.5 g of aluminum potassium sulfate (Sigma A7167) in 500 ml of distilled water). The samples were then dehydrated using stepwise ethanol concentrations and left in xylene to clear the fat. Mammary gland whole mounts were stored in methyl salicylate. Whole mounts were digitally photographed with a ruler using a stereomicroscope, employing the same magnification and lighting conditions for each genotype.
Histological Analysis-Portions of tumors or inguinal mammary glands were excised, formalin-fixed for 24 h, and embedded in paraffin. Sections were cut at 5 microns and stained with hematoxylin and eosin. As demonstrated in Fig. 2A, caveolin-1 levels remain unaffected by targeted disruption of the INK4a locus, which produces a null mutation. We also performed a broader tissue analysis of caveolin-1 expression in the INK4a (Ϫ/Ϫ) mice, including adipose, heart, intestine, skeletal muscle, and the mammary gland, with virtually identical results; caveolin-1 expression is unaffected in the absence of INK4a (data not shown). In contrast, caveolin-1 levels are dramatically downregulated in p53 (Ϫ/Ϫ)-null lungs ( Fig. 2A). Thus, we directed our efforts toward establishing immortal cell lines both expressing and lacking Cav-1 by using INK4a-null mice, rather than p53-null mice, because Cav-1 levels are unperturbed by genetic ablation of the INK4a locus. This finding also suggests that Cav-1 is more likely to cooperate with the INK4a tumor suppressor locus since the regulatory pathways controlling Cav-1 expression appear to not be linked to the p16 INK4a / p19 ARF protein products.
Cell cycle analysis of early passage asynchronously growing INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) MEFs revealed a small increase in the S-phase fraction and a large decrease in the G 0 /G 1 fraction, compared with INK4a (Ϫ/Ϫ)/Cav-1 (ϩ/ϩ) MEFs (Fig. 3C). This is suggestive of a role for caveolin-1 in arresting the cell cycle in the G 0 /G 1 phase (34). Similar results were obtained with later passage INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) MEFs (Ͼ40 passages) which demonstrated a profile of decreased G 0 /G 1 fractions, albeit with much larger increases in S-phase fractions that are characteristic of immortalized cells that have been cultured for prolonged periods of time. These findings also indicate that Cav-1 and INK4a show strong cooperativity with regard to their proliferative potential, because their combined absence enhances the pro-proliferative effects observed by deletion of either one alone.
Loss of Caveolin-1 Expression Induces Pro-proliferative Changes in the Levels of Several Important Cell Cycle Regulatory Proteins, including p21 Cip1 , Cyclin D1, and PCNA-In order to characterize the observed alterations in proliferation rates and cell cycle parameters at the molecular level, we next performed immunoblotting with antibodies directed against a panel of important cell cycle regulatory proteins, including the cyclins (A, E, D1), Cdks (2, 4, 6), p21 Cip1 , p27 Kip1 , and PCNA. Fig. 5A shows that in INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) dKO MEFs, the levels of the cell cycle inhibitor p21 Cip1 are dramatically decreased, compared with INK4a (Ϫ/Ϫ)/Cav-1 (ϩ/ϩ) MEFs. Conversely, the pro-proliferative proteins, cyclin D1 and PCNA, were both overexpressed in INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) MEFs (Fig. 5A). Additional molecular profiling demonstrated that expression of other important cell cycle regulators remained unchanged, including another member of the Cip/Kip family of cell cycle inhibitors, p27 Kip1 , as well as other cyclins, and cyclin-dependent kinases (Fig. 5, A and B). We next examined transcriptional regulation of the cyclin D1 gene, using a cyclin D1 promoter luciferase reporter construct to determine if the increased levels of cyclin D1 are due to increased transcription/translation or increased stabilization of cyclin D1. Table I shows that an absence of caveolin-1 results in a ϳ4-fold increase in reporter activity, indicating that cyclin D1 overexpression is due, at least in part, to its transcriptional up-regulation in INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) MEFs.
Furthermore, we also employed a reporter construct containing a c-Fos responsive element. In response to serum or growth factor stimuli, the ERK subgroup of MAP kinases becomes activated, translocates to the nucleus, and potentiates the transcriptional activity of AP-1 dimers, especially those responsive to fos promoters (39). Analysis using a c-Fos responsive element reporter construct reveals that c-Fos activity is ϳ2-fold greater in INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) MEFs, compared with INK4a (Ϫ/Ϫ)/Cav-1 (ϩ/ϩ) MEFs (Table I). Increased c-Fos activity is consistent with both the observed increases in cyclin D1 expression and the hyperactivation of ERK 1/2 and provides further insight into the molecular events underlying the hyperproliferative phenotype of these Cav-1 (Ϫ/Ϫ)-deficient cells.

Loss of Caveolin-1 Dramatically Enhances the Ability of Oncogene-transformed INK4a (Ϫ/Ϫ) MEFs to Generate Tumors in
Nude Mice-Because INK4a (Ϫ/Ϫ) MEFs are immortalized but not transformed, they offer a very sensitive means to measure the capacity of a single gene to affect cell transformation. Therefore, in order to study how loss of Cav-1 affects cell transformation, we next transformed our INK4a (Ϫ/Ϫ) MEF cell lines with two potent activated oncogenes, H-Ras G12V and v-Src.
H-Ras G12V -mediated Transformation-We recombinantly expressed an activated oncogene, H-Ras G12V , in these primary MEFs using a well established retroviral expression system (pBABE-H-Ras G12V -puro) and selection in puromycin. Using this viral transduction system, stable cell lines or "transduced pools" can be created without the selection of individual clones, avoiding the problems associated with clonal variability. As a critical negative control, populations of MEFs were also generated using the empty vector alone.
First, we examined the stable MEF cell lines for H-Ras G12V and caveolin-1 expression levels. Both INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) and INK4a (Ϫ/Ϫ)/Cav-1 (ϩ/ϩ) MEFs demonstrated equivalent levels of H-Ras G12V expression by immunoblotting with an antibody that recognizes both the wild-type and mutant-transfected forms of H-Ras (Fig. 6A). Note that caveolin-1 expression levels remain unchanged in H-Ras G12V -transfected versus non-transfected INK4a (Ϫ/Ϫ)/Cav-1 (ϩ/ϩ) MEFs. This result is in contrast to previous findings with non-transfected and H-Ras G12V -transfected NIH 3T3 cells, where Cav-1 levels are down-regulated in the presence of the activated H-Ras oncogene (12,36). A possible explanation for this discrepancy is that functional INK4a products are required for Cav-1 downregulation. Conversely, this result may not be INK4a-dependent, but rather may highlight the importance of using primary cells with a defined genetic background, unlike NIH 3T3 cells which were immortalized using the 3T3 protocol and have acquired a multitude of genetic changes through genomic in- stability over prolonged periods in culture.
In order to elucidate differences in the transformed phenotype among these INK4a (Ϫ/Ϫ) stable cell lines, we next used well established tumorigenicity assays. First, we performed soft agar assays on H-Ras G12V -transfected MEFs to determine if anchorage-independent growth is affected by the absence of caveolin-1. Soft agar growth is a well established in vitro assay that provides an index of cellular transformation; cells that are generally more transformed are able to grow more independently of cell-substrate or cell-cell adhesive contacts, which provide anti-proliferative signals.
Serrano et al.   's t test).  (Fig. 6B). Thus, in the context of H-Ras G12V -mediated transformation, loss of Cav-1 resulted in an ϳ8-fold increase in tumor weight. Similarly, caliper measurements indicated that loss of Cav-1 resulted in an ϳ11-fold increase in tumor volume (cm 3 ). Therefore, in the INK4a (Ϫ/Ϫ)/H-Ras G12V background, loss of Cav-1 expression allowed these primary MEFs to manifest a full blown transformed phenotype, as evidenced by their capacity to form large, aggressive tumors in nude mice.

FIG. 6. Loss of caveolin-1 dramatically enhances the ability of H-Ras G12V -transformed INK4a (؊/؊) MEFs to form tumors in vivo (nude mouse injections). A, Western blot analysis of H-
Gross examination of the tumors derived from H-Ras G12Vtransfected INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) MEFs revealed much larger tumors that demonstrated areas of necrosis and hemorrhage, suggesting that the tumor cells were proliferating at a pace greater than the capability of the tumor blood vasculature to supply oxygen and nutrients (Fig. 6B, see insets a and b).  (Fig. 7B). This increase was much more dramatic than that observed from the H-Ras G12V -transfected tumor cells, with the differences being greater than ϳ40-fold, compared with ϳ8-fold for the H-Ras G12V tumor cell lines.
In light of previous findings indicating that caveolin-1 can inhibit the autoactivation of c-Src (40), we next performed immunoblotting with a phosphospecific antibody that only recognizes the activated tyrosine-phosphorylated form of Src in order to understand the large differences we observed in tumorigenesis. As predicted, we found a dramatic increase in the tyrosine-phosphorylated activated species of v-Src in the v-Srctransformed INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) MEFs, suggesting that the increased activity of v-Src in the absence of Cav-1 may account for their enhanced tumorigenicity (Fig. 7C).
The similarities between the results obtained from H-Ras G12V and v-Src-mediated neoplastic transformation of INK4a (Ϫ/Ϫ) MEFs provide new data to support the hypothesis that caveolin-1 functions as an in vivo transformation suppressor or tumor susceptibility gene, and that its absence confers a greater capacity for tumorigenesis. Importantly, our results also indicate that Cav-1 and INK4a are able to cooperate during oncogene-induced tumorigenesis.

INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) Mammary Glands Show Aberrant Epithelial Compartment
Morphology-Aside from the above cellular data demonstrating that loss of Cav-1 and INK4a cooperate to enhance proliferative and tumorigenic capabilities within the cell, we attempted to find an in vivo example of cooperativity between Cav-1 and INK4a by performing histopathological analyses. We have previously reported that Cav-1-null mice demonstrate mild mammary epithelial hyperplasia in Cav-1 (Ϫ/Ϫ) virgin female mice as early as 6 weeks of age (41). In order to determine whether the loss of INK4a potentiates the mammary epithelial hyperplasia observed in Cav-1 (Ϫ/Ϫ) mice, inguinal mammary fat pads from 4-month old virgin female mice were excised, fixed in ethanol/acetic acid, and stained with carmine dye to visualize the epithelial architecture of the glands. In support of our previous findings, INK4a (ϩ/ϩ)/Cav-1 (Ϫ/Ϫ) mammary epithelial ducts are thickened and enlarged, consistent with hyperplasia, compared with normal INK4a (ϩ/ϩ)/Cav-1 (ϩ/ϩ) (wild-type) ducts (Fig. 8, top  panels). Moreover, mammary epithelial ducts in INK4a (Ϫ/Ϫ)/ Cav-1 (ϩ/ϩ) mammary glands do not demonstrate hyperplasia, and more closely resemble those of wild-type ducts in thickness and appearance (Fig. 8, bottom left panel).
In summary, this readily apparent increase in the degree of mammary epithelial ductal thickness, branching, and fibrosis provides an in vivo tissue example of cooperativity between the combined absence of Cav-1 and INK4a in imparting a more dysregulated proliferative and/or morphological tissue phenotype in INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) dKO mice. DISCUSSION Neoplastic transformation is the process by which a normal cell acquires malignant properties, which enable it to subvert growth inhibitory or apoptotic signals, such as contact inhibition, a lack of growth factors, loss of anchorage to the matrix surface, or telomere shortening (42). The number of genetic changes that must occur within a single cell to render it transformed depend on a variety of factors including the cell type, the genetic mechanism (e.g. deletion, mutation, or silencing), and the particular genes involved. However, it is widely accepted that the process of transformation requires multiple cellular events and that there are few instances during which a single aberration can confer the transformed property (42). This characteristic is likely to have evolved in mammalian cells to provide a system of checks and balances to tightly regulate growth and survival of the organism.
Generally speaking, if a gene exerts an effect on the potential for a cell to become transformed, that gene can be classified as either a tumor suppressor or oncogene. There are a whole host of tumor suppressors and oncogenes with relatively different potencies, with regard to their ability to suppress or promote transformation. For example, p53 and Rb are two well characterized and extremely potent tumor suppressors, while Myc and Ras, by analogy, are two dominant proto-oncogenes. The multi-step nature of transformation requires that the relevant oncogenes or tumor suppressors must cooperate with each other. That is, the cellular pathways intrinsic to each oncogene or tumor suppressor do not antagonize the functions or signaling effects of one another, or transformation will be less likely to occur. In this report, we have demonstrated that the combined loss of both caveolin-1 and INK4a gene products shows cooperativity, i.e. enhancing the proliferative capability and transformation susceptibility of a cell more than loss of either alone could achieve, using mouse embryonic fibroblasts. The ability to continuously proliferate irrespective of growth inhibitory, pro-senescent, or apoptotic signals is an important feature of transformed cells. While it is known that loss of INK4a is sufficient for cells to become immortalized, we show that the absence of caveolin-1 imparts an even greater proliferative capacity within these cells.
Our current findings provide novel support for the hypothesis that caveolin-1 functions as a "transformation suppressor" or "negative tumor modulator." These key findings are briefly outlined below : 1) To our knowledge, this report is the first demonstration of the utilization of INK4a (Ϫ/Ϫ) mice to generate immortal primary cell lines expressing or completely lacking a particular protein (in this case, caveolin-1). This technique has implications for investigators seeking to establish immortal, but not transformed, primary cell lines from knockout and transgenic mice. In fact, investigators attempt to accomplish the immortalization process through many means, including the re-insertion of genes (telomerase, SV-40 T antigens) or passaging primary cells through a long 3T3 protocol, which results in cells with any number of undefined genetic mutations. Once INK4a (Ϫ/Ϫ)-immortalized cell lines are established, they enable one to easily compare the effects of the presence and absence of a particular gene on the processes of transformation and tumorigenesis after transduction with an oncogene, such as H-Ras G12V , v-Src, c-Myc, c-Neu, etc. In essence, we have outlined a novel general experimental strategy that others may use to study the effect of a given gene on cellular transformation and tumorigenesis.
2) We establish that the observed down-regulation of caveolin-1 levels in H-Ras G12V -transduced cells from previous reports appears to be dependent on the INK4a locus, as we show that caveolin-1 levels do not change in these INK4a (Ϫ/Ϫ) cells after oncogenic transduction. This result is in contrast to previous findings with normal and H-Ras G12V -transfected NIH 3T3 cells, where Cav-1 levels are down-regulated in the presence of the activated H-Ras oncogene (12,36). One possible explanation for this discrepancy is that functional INK4a gene products are required for Ras-mediated Cav-1 down-regulation. Alternatively, this result may simply highlight the importance of using primary cells with a defined genetic background, unlike NIH 3T3 cells, which were immortalized using the 3T3 protocol and have acquired a multitude of genetic changes over prolonged periods.
3) Our results show that the effects of caveolin-1 on cell proliferation are p16 INK4a -and p19 ARF -independent, since the loss of the INK4a locus does not abolish the increases in cell proliferation observed after loss of caveolin-1. In fact, it appears that the proliferative phenotype is even accentuated, implying that the effects resulting from the loss of caveolin-1 and p16 INK4a /p19 ARF can cooperate to increase cellular proliferation. This is an important observation as typically a large number of tumors appear to arise or are associated with INK4a mutations, and the loss of caveolin-1, such as occurring as a result of a germline or somatic mutation, would serve as an "additional promoting event" to develop larger, more invasive tumors. 4) We demonstrate that caveolin-1 ablation in the absence of INK4a results in constitutive hyperactivation of ERK1/2. In a previous report, using primary MEFs, we have shown that there is no difference in the activation state of ERK1/2 in the presence or absence of caveolin-1 (28). Therefore, it appears that deletion of the INK4a locus unmasks the negative regulatory effect of Cav-1 on the activation state of ERK1/2 in primary MEFs. 5) We provide the first molecular profile of the effect of loss of caveolin-1 on well established cell cycle components, including cyclins, cyclin-dependent kinases, as well as cell cycle inhibitors. We demonstrate that complete loss of caveolin-1 (i.e. mimicking the effects of genetic mutation) results in PCNA and cyclin D1 overexpression, p21 Cip1 down-regulation, and basal hyperactivation of ERK 1/2. This is interesting in light of the findings that other cyclins (including A-and E-type cyclins) and inhibitors (p27 Kip1 ) are unaffected. In addition, this is the first report showing associations between caveolin-1 and PCNA expression, as well as c-Fos activation.
6) Our data are the first to describe how genetic loss of caveolin-1 affects tumorigenesis, achieved through H-Ras G12V and v-Src oncogenic manipulation. Previous tumor studies were limited to overexpression of Cav-1 or antisense strategies, which do not completely abolish Cav-1 expression as a bi-allelic null mutation would accomplish, therefore making the current study more relevant in terms of human physiological mutations. Given this, we observe overwhelming differences in tumor size achieved by genetic ablation of Cav-1. In addition, loss of caveolin-1 is shown to cooperate with the effects of loss of INK4a, suggesting that loss of both in vivo would result in larger, more aggressive tumors. 7) The findings in the mammary gland provide an in vivo example of how the combined loss of caveolin-1 and INK4a cooperate to result in the dysregulation of tissue architecture, which might make that particular tissue more susceptible to tumorigenesis. This is the first demonstration of the effect of loss of a particular gene (i.e. INK4a) on exacerbating the mammary epithelial cell hyperplasia phenotype observed in the Cav-1 (Ϫ/Ϫ) mice.
In order to directly assess whether loss of Cav-1 affects susceptibility to cellular transformation, we have separately transformed our immortalized MEF cell lines with two different constitutively activated oncogenes, namely H-Ras G12V and v-Src. We and other groups have previously shown that H-Ras localizes to caveolae microdomains and that caveolin-1 directly binds to H-Ras through the caveolin-scaffolding domain (residues 82-101) (8,(43)(44)(45). Interestingly, while caveolin-1 clearly binds wild-type H-Ras, it does not bind the mutationally activated H-Ras G12V with nearly the same affinity (43). This suggests that caveolin-1 preferentially associates with the inactive form of H-Ras and that its association not only serves to compartmentalize this signal-transducing molecule, but also to potentially negatively regulate its activation.
Aside from H-Ras, the Src family of tyrosine kinases represents another group of lipid-modified signaling molecules that have been shown to be enriched in caveolae (40, 43, 46 -48). Historically, Cav-1 was first identified as a tyrosine-phosphorylated target of v-Src in Rous sarcoma virus-transformed avian fibroblasts, implicating a role for Cav-1 in mediating cell transformation (1). In direct support of this finding, we have identified tyrosine 14 in the N-terminal portion of caveolin-1 as the principal substrate for v-Src (46). In further support of a functional relationship between Cav-1 and Src family members, caveolin-1 directly binds c-Src through the caveolin scaffolding domain (residues 82-101) and both co-localize by immunofluorescence microscopy in COS-7 cells (40). Interestingly, as with H-Ras, caveolin-1 appears to have a higher specificity for the inactive form of Src, as caveolin-1 forms a stable complex with c-Src, but not with the constitutively activated v-Src (40). Additionally, co-expression of caveolin-1 and c-Src or Fyn (a related Src family tyrosine kinase), inhibits their autophosphorylation in vivo, a process that normally renders them active. Taken together, these observations suggest that caveolin-1 both sequesters and negatively regulates the activation state of Src family tyrosine kinases.
In direct support of these in vitro results, we show here that an absence of caveolin-1 significantly potentiates the ability of v-Src to transform INK4a-deficient MEFs, resulting in striking increases in tumor mass. Also, the increased susceptibility of Cav-1-deficient cells to transformation appears to be due to a relative increase in the phosphorylated (Tyr-416) activated species of v-Src.
While loss of Cav-1 is not sufficient to transform a cell, it is now clear that this protein modulates the ability of a cell to become transformed and we have therefore referred to caveolin-1 as a transformation or tumor susceptibility gene. It is currently believed that the number of genetic or molecular alterations that must occur within a single cell to render it transformed within the context of a tissue is most likely greater than the number required for cells in culture. This may be due to the organizational complexity of a tissue, such as the role of the supporting stroma in inhibiting or confining cells that become transformed. Nevertheless, we have found evidence of dysregulated proliferation and aberrant morphology within the mammary glands of INK4a (Ϫ/Ϫ)/Cav-1 (Ϫ/Ϫ) dKO mice. This dysregulated proliferation is manifested as mammary epithelial hyperplasia, with increased wall thickness and moderate fibrosis. The fibrosis may be another primary disturbance due to the combined loss of INK4a and Cav-1 or may be a secondary effect as a result of injury to the surrounding stroma and, therefore, serving as a compensatory response to confine the epithelial hyperproliferation. Taken together, these findings suggest that the combined loss of INK4a and Cav-1 can cooperate to cause disturbances in normal tissue architecture.
In summary, we have definitively demonstrated a role for caveolin-1 in cellular transformation and oncogene-induced tumorigenesis and that the loss of caveolin-1 shows cooperativity with loss of INK4a in this process. Deficiency of caveolin-1 in the INK4a-null genetic background results in dysregulated cell proliferation, both within cultured fibroblasts and in the mammary gland in vivo. Most importantly, the complete absence of caveolin-1 clearly increases the susceptibility of cells to transformation by activated oncogenes (H-Ras G12V and v-Src). These findings provide the strongest evidence to date that caveolin-1 possesses transformation suppressor properties and clearly functions as a tumor susceptibility gene.