Genetic Screening Reveals an Essential Role of p27kip1 in Restriction of Breast Cancer Progression

In estrogen-dependent breast cancer cells, estrogen, through estrogen receptor α (ERα), enhances c-Myc and cyclin D1 expression, down-regulates p27, and activates cyclin E/cyclin-dependent kinase (CDK)-2 to promote G₁-S transition (1, 2). Because most primary breast cancers express ERα and require estrogen to grow, estrogen antagonists and aromatase inhibitors such as tamoxifen and letrozole are used to treat estrogen-dependent breast cancers (3–5). Although these treatments are initially effective, acquired resistances are a major problem. In most cases, development of drug resistance is not due to a loss or mutation of ERα (4, 6).

Overexpression or activation of receptor tyrosine kinases (RTK) and Ras oncoproteins is common in cancers. Human epidermal growth factor receptor 2 (HER2)/neu overexpression happens in 20% to 30% breast cancers and correlates with more aggressive cancer phenotypes and tamoxifen resistance (7, 8). Ras and RTKs including HER2, insulin-like growth factor I receptor, and epidermal growth factor receptor activate the phosphatidylinositol 3-kinase (PI3K)/Akt pathway (9–12). This pathway plays a pivotal role in cell survival, proliferation, motility, tumorigenesis, and metastasis through phosphorylation and subsequent relocalization of key regulatory molecules such as p27 (13–15).

In cell nucleus, p27 associates with cyclin E-CDK2 and inhibits Rb hyperphosphorylation to keep cells in G₁ phase (16). Mitogenic stimuli cause p27 phosphorylation, ubiquitination, degradation, and translocation to the cytoplasm and increase cyclin E-CDK2 activity, leading to G₁-S transition through Rb phosphorylation and E2F activation (16). In breast cancers overexpressing HER2, Akt phosphorylates p27 and keeps p27 in the cytoplasm, which precludes p27-induced G₁ arrest (13–15). Therefore, both total p27 reduction and p27 exclusion from the nucleus of breast cancer cells are associated with poor prognosis and estrogen independence (16, 17). Down-regulation of p27 also enhances MCF-7 breast cancer cell growth in the presence of antiestrogens (18, 19). However, despite many studies correlating p27 levels and locations with tumorigenesis, the lack of p27 null mutation in human tumors has made it difficult to understand the exact role of p27 in breast cancer progression (20). Furthermore, the fact that oncogene-induced mammary tumorigenesis is accelerated in p27^+/− mice but suppressed in p27^−/− mice suggests a complex role of p27 in breast cancer (21).

E2F is the key transcriptional factor for cell cycle progression (16). E2F1 interacts with amplified in breast cancer 1 (AIB1), and this interaction potentiates E2F1 target gene expression (22, 23). AIB1 is a transcriptional coactivator for nuclear receptors and other transcription factors including E2F1 (22, 24). AIB1 is overexpressed in ∼60% of human breast tumors (25). Overexpression of AIB1 in mouse mammary epithelium causes mammary carcinomas (26), whereas inactivation of AIB1 in mice suppresses oncogene and carcinogen-induced mammary tumorigenesis (27, 28). These findings indicate that AIB1 is an oncogene. AIB1 activity is not only determined by its concentration but also regulated by phosphorylation (29). In addition, E2F1, through up-regulation of the adaptor protein growth factor receptor binding protein 2–associated binder 2 (Gab2), strongly activates Akt (30), which promotes cell motility, invasion, and cancer metastasis (31, 32). Thus, we hypothesize that the serial events initiated by p27 inactivation may promote breast cancer cell migration, invasion, and metastasis.

In this study, we have done a genetic screening by using the enhanced retroviral mutagen (ERM) system (33) and identified the p27 loss-of-function clone. We show that p27 deficiency in T47D cells causes estrogen-independent and antiestrogen-resistant growth; stimulates AIB1 phosphorylation and its coactivator activity for E2F1; promotes Akt activity, cell motility, and invasion; and results in lung metastasis in ovariectomized nude mice injected i.v. with these cells.

Retroviral infection. T47D cells (10⁴) were infected with MSCV-Eco-zeo and MSCV-SV-tTA viruses (multiplicity of infection, 10:1) and selected in medium with 200 μg/mL G418 (33). The surviving T47D^tTA cells were expanded to 10⁷, transferred into six-well plates (5 × 10⁵ per well), and spin-infected with ERM retroviruses. Each of the three ERM retroviral vectors carried an ERM promoter, an expression tag in one of the three reading frames, and a splice donor (Fig. 1B; ref. 33). Two days after ERM infection, cells were transferred to 100-mm dishes for a 30-day selection in RPMI 1640 containing 10 μg/mL insulin, 10% charcoal dextran–stripped FCS (CSFCS), and 1 μg/mL puromycin. For p27 expression, the COOH-terminally poly-histidine–tagged p27 cDNA was taken from the pCMV-p27 plasmid (a kind gift from Dr. Mong-Hong Lee, Department of Molecular and Cellular Oncology, University of Texas MD Anderson Cancer Center, Houston, TX) and subcloned into the pQCXIH retroviral plasmid (BD Biosciences). The pQCXIH-p27 retroviruses were produced in PT67 packaging cells.

Reverse transcription-PCR for identification of ERM-targeted genes. Colonies were isolated and cultured in RPMI 1640 with 10 μg/mL insulin and 10% CSFCS. RNA was prepared and reverse transcribed with primer 5′-GCAAATACGACTCACTATAGGGATCCNNNN(G/C)ACG. PCR was done using ERM-tag primer (5′-ACCATGGGGAGCAGCAAGAGCAAACCAAAAGACCCCAGCCAACGC) and T7 primer. PCR products were sequenced and GenBank was searched to identify ERM-targeted genes.

Cell cycle analysis, colony formation assay, immunochemical analyses, kinase assay, transfection assay, and migration and invasion assays. These assays were carried out by following standard methods as previously described (28, 29, 34). For details, see Supplementary Materials and Methods.

Tumor growth and experimental metastasis assays. Mouse protocols were approved by Baylor College of Medicine Animal Care and Use and Committee. For tumor growth in mammary glands, 8 × 10⁶ cells were mixed with Matrigel and injected into both fourth mammary fat pads of ovariectomized athymic nude mice (35). Each type of cells was injected into six mice without pellets and six mice with 60-day estradiol-releasing pellets (1.7 mg/pellet from Innovative Research). Four weeks after injection, tumors were excised for histology and immunohistochemistry. Other organs including lung, liver, and mesenteric lymph nodes were examined for metastasis. For experimental metastasis, 10⁶ cells were injected into the tail lateral vein of ovariectomized nude mice. Three weeks after injection, their lungs were examined for metastasis and processed for histology and immunohistochemistry as described (28).

Identification of estrogen-independent growth mutants. T47D cells express ERα and depend on estrogen to grow and survive. A genome-wide screening using ERM was used to identify genes with gain-of-function or loss-of-function mutations that could make T47D cells survive in estrogen-free medium. To prepare cells for ERM infection, T47D cells were first infected with MSCV-Eco-zeo viruses to express the mouse ecotropic receptor and with MSCV-SV-tTA to express tTA tetracycline-off regulator. The obtained stable T47D^tTA cells were subsequently infected with ERM and growth selected in estrogen-free medium. As expected, no colonies were found from uninfected T47D^tTA cells after culturing in the estrogen-free medium. In contrast, ∼150 colonies formed from a total of 5 million ERM-infected T47D^tTA cells in the selection medium. Surviving colonies were isolated and the ERM-targeted genes were identified in each clone with the use of reverse transcription-PCR (RT-PCR) that amplified ERM expression tag and its trapped 3′ exon sequences (33). Of 15 clones analyzed, we obtained 12 RNA sequences containing ERM tag. Each of the 12 clones only produced one ERM tag-fused RNA sequence, indicating that each clone contains a single ERM integration. Among these estrogen-independent growth mutants (EIGM), EIGM-1 was a loss-of-function mutation for p27, which was characterized in this study. The other 11 clones included 1 small GTPase (EIGM-2–EIGM-4), 2 components of the Wnt signaling pathway (EIGM-5 and EIGM-6), 1 transcription factor (EIGM-7), 1 Golgi protein (EIGM-8), 1 bone marrow stromal cell antigen (EIGM-9), 1 enolase (EIGM-10), and 2 hypothetical proteins (EIGM-11 and EIGM-12).

Disruption of the single p27kip1 allele in T47D cells results in estrogen-independent growth. The sequence of EIGM-1 matched the second exon of the p27 gene; thus, it was designated as T47D^tTA/p27m. Bromodeoxyuridine (BrdUrd) labeling assay confirmed the estrogen-independent growth feature of T47D^tTA/p27m cells. About 93% T47D^tTA parent cells cultured in the estrogen-free medium were arrested at G₁ and <0.3% cells were labeled with BrdUrd. In contrast, only 66% T47D^tTA/p27m cells cultured in the estrogen-free medium were in G₁ and as many as 8.6% T47D^tTA/p27m cells were labeled with BrdUrd, which were similar to T47D^tTA and T47D^tTA/p27m cells cultured in the growth medium with complete serum (Fig. 1A).

The p27 gene contains two exons separated by a 511-bp intron. Exons 1 and 2 encode for the NH₂-terminal 158 and the COOH-terminal 40 (159–198) amino acid residues, respectively (Fig. 1B). PCR and sequence analyses revealed that the ERM integration site was between bp 174 and 175 of intron 1 (Fig. 1B and C). In the RT-PCR product from T47D^tTA/p27m cells, the ERM tag was fused in-frame with the second exon of p27, indicating that the 337-bp intron sequence between the splice donor site of the ERM tag and the second exon of p27 was excised during RNA processing (Fig. 1B). Accordingly, the inserted ERM promoter generated a 10-kDa protein that was recognized by antibodies against p27 COOH terminus and ERM tag, indicating that this fusion protein contains ERM tag (62 amino acids) and the COOH-terminal 40 amino acids of p27 (p27-C40; Fig. 1D and data not shown). The p27 NH₂-terminal antibody recognized a smaller and less abundant protein in T47D^tTA/p27m cells than the full-length p27 in T47D^tTA cells (Fig. 1D). Because this smaller p27 protein was not detectable by the p27 COOH-terminal antibody and its apparent molecular mass (∼10 kDa) matched the predicted size translated from exon 1 and its extending intron sequence, we concluded that ERM insertion in the p27 intron interfered the splicing between exons 1 and 2. This resulted in a truncated p27 protein with the NH₂-terminal 158 amino acids and the COOH-terminal 21 amino acids encoded by the extended intron sequence before meeting the TAA stop codon at 63 to 65 bp of intron 1 (Fig. 1B and D). This NH₂-terminal protein was designated as p27-N158. Interestingly, no wild-type p27 allele and full-length p27 protein were detected in T47D^tTA/p27m cells by PCR and immunoblotting (Fig. 1C and D), indicating that T47D cells only contain a single p27 wild-type allele and T47D^tTA/p27m cells only contain a disrupted p27 allele.

T47D^tTA/p27m cells are resistant to antiestrogens although they still express functional ERα. To test whether disruption of p27 reduces the antiproliferative effects of antiestrogens, we compared cell proliferation rates of T47D^tTA/p27m cells and T47D^tTA cells treated with 4-OH-tamoxifen and ICI 182780 for 4 days. The vehicle-treated T47D^tTA and T47D^tTA/p27m cells exhibited similar S-phase fractions (∼10%), whereas 4-OH-tamoxifen treatment reduced the S-phase fraction of T47D^tTA cells to ∼3%. However, 4-OH-tamoxifen treatment only slightly decreased the S-phase fraction of T47D^tTA/p27m cells to 7%. ICI 182780 treatment diminished the S-phase fraction of T47D^tTA cells to ∼3%, but only partially reduced the S-phase fraction of T47D^tTA/p27m cells to ∼5.5%, indicating that T47D^tTA/p27m cells were also less sensitive than T47D^tTA cells to ICI 182780 (Fig. 2A). Consistent results were also obtained from cell growth of T47D^tTA and T47D^tTA/p27m cells cultured in growth medium containing vehicle, 4-OH-tamoxifen, or ICI 182780 and in estrogen-free medium with CSFCS (Fig. 2B). To validate the specific role of p27 in these experiments, we restored stable p27 expression in T47D^tTA/p27m cells by retroviral infection. The restored p27 was a poly-histidine–tagged bigger protein compared with the endogenous p27 in T47D^tTA cells (Fig. 2B). Restoration of p27 in T47D^{tTA/p27m/+p27} cells rescued the growth-inhibitory sensitivities of these cells to 4-OH-tamoxifen, ICI 182780, and CSFCS (Fig. 2B). These results indicate that p27 function is required for T47D cells to respond to the antiproliferative effects of antiestrogens.

To examine the long-term effects of antiestrogens on T47D^tTA/p27m cell survival and growth, we compared the colony formation capability of T47D^tTA/p27m cells with that of T47D^tTA cells over a 3-week culture period. In growth medium, T47D^tTA/p27m and T47D^tTA cells formed colonies equally well, giving ∼45 colonies/cm². 4-OH-Tamoxifen treatment significantly reduced the density of colonies formed from T47D^tTA cells to 15 colonies/cm². In contrast, 4-OH-tamoxifen treatment did not significantly change the colony formation capability of T47D^tTA/p27m cells (Fig. 2C). ICI 182780 treatment profoundly reduced both the number and size of the colonies formed from T47D^tTA cells. However, ICI 182780 reduced only the colony size of T47D^tTA/p27m cells but not their total number of colonies (Fig. 2C). These results show that T47D^tTA/p27m cells are fully resistant to 4-OH-tamoxifen and partially resistant to ICI 182780.

It is a possibility that loss of ERα causes breast cancer cells to become insensitive to antiestrogens. However, immunoblotting analysis revealed that ERα levels in T47D^tTA and T47D^tTA/p27m cells were comparable. The estradiol-induced or ICI 182780–inhibited transcriptions of PR_A and PR_B from the estrogen-responsive PR gene also were comparable in these cells (Fig. 2D). Thus, the antiestrogen resistance observed in T47D^tTA/p27m cells is not due to a loss of ERα expression or function.

The p27-N158 and p27-C40 proteins are loss-of-function mutants. Most p27 immunoreactivity was detected in the nucleus of T47D^tTA cells by p27 NH₂-terminal antibody, whereas the p27-N158 immunoreactivity was detected in both cytoplasm and nucleus of T47D^tTA/p27m cells with the same antibody (Fig. 3A). Immunoblotting analyses further identified p27 in both cytoplasmic and nuclear fractions of T47D^tTA cells; on the contrary, the mutant p27-N158 was mainly found in the cytoplasmic fraction of T47D^tTA/p27m cells (Fig. 3B,, top). The latter observation is consistent with the absence of p27 nuclear localization sequence (amino acids 159–169) in p27-N158 (Fig. 1B). These results indicate that although p27-N158 contains the NH₂-terminal domains for binding cyclins and CDKs, it mainly stays in the cytoplasm and does not function as a nuclear CDK inhibitor.

The nuclear p27 in T47D^tTA cells was also detected by the p27 COOH-terminal antibody. Interestingly, the p27-C40, detected by the same antibody, was mainly found in the cytoplasm of T47D^tTA/p27m cells, although weaker immunoreactivity was also detected in their nuclei (Fig. 3A). Immunoblotting analyses confirmed that p27-C40 was distributed in both cytoplasmic and nuclear fractions of T47D^tTA/p27m cells (Fig. 3B,, bottom). Because p27-C40 expression is controlled by the tetracycline-off ERM promoter, we examined the effect of p27-C40 on cell proliferation by treating cells with tetracycline. The p27-C40 level was significantly reduced in T47D^tTA/p27m cells on tetracycline treatment (Fig. 3C). However, the distribution of cell fractions in all proliferation phases, determined by flow cytometry, was identical between vehicle- and tetracycline-treated T47D^tTA/p27m cells (data not shown). These results indicate that change of p27-C40 concentration does not affect cell proliferation.

p27 deficiency reduces the inhibitory effects of antiestrogens on CDK2 activity. In estrogen-dependent breast cancer cells, antiestrogens induce p27, inhibit cyclin E-CDK2 activity, reduce Rb and Rb-related p130 hyperphosphorylation, suppress E2F, and arrest cell cycle at G₁ (16). In T47D^tTA cells, 4-OH-tamoxifen induced a slight, and ICI 182780 and estrogen-free condition induced a significant, p27 increase. In T47D^tTA/p27m cells, these treatments failed to induce p27-N158 levels, suggesting that antiestrogens or estrogen-free conditions do not induce p27 promoter activity in the estrogen-independent T47D^tTA/p27m cells. The level of p27-C40 was decreased when T47D^tTA/p27m cells were cultured in estrogen-free medium, due presumably to lower ERM promoter activity under this culture condition (Fig. 3D and data now shown). Cyclin E and CDK2 levels were identical between T47D^tTA and T47D^tTA/p27m cells under all examined conditions (Fig. 3D).

To examine the effect of p27 mutation on the kinase activity of cyclin E-CDK2 complexes, cyclin E was immunoprecipitated from cell extracts with equal protein amount and the kinase activity of precipitates was assayed using histone H1 as substrate. The cyclin E–associated CDK2 kinase activity was reduced by 33% and 66%, respectively, when T47D^tTA cells were treated with 4-OH-tamoxifen and ICI 182780. When T47D^tTA cells were cultured in estrogen-free medium, the cyclin E–associated CDK2 kinase activity was reduced by 87%. In contrast, the cyclin E-CDK2 kinase activity in T47D^tTA/p27m cells was not significantly reduced with 4-OH-tamoxifen and only reduced by 37% and 57% with ICI 182780 and estrogen-free medium (Fig. 3D). In agreement with the increase of CDK2 activity, the ratios of hyperphosphorylated Rb to hypophosphorylated Rb were increased in T47D^tTA/p27m cells treated with 4-OH-tamoxifen, ICI 182780, and estrogen-free medium compared with the ratios of hyperphosphorylated Rb to hypophosphorylated Rb in T47D^tTA cells receiving the same treatments (Fig. 3D). Similarly, the ratios of hyperphosphorylated p130 to hypophosphorylated p130 in T47D^tTA/p27m cells treated with 4-OH-tamoxifen, ICI 182780, and estrogen-free medium were higher than the ratios of hyperphosphorylated p130 to hypophosphorylated p130 in T47D^tTA cells (Fig. 3D). The increase in these ratios was mainly due to the decrease in the hypophosphorylated inhibitory forms of Rb and p130. These results indicate that p27 deficiency desensitizes the inhibition of CDK2 activity by antiestrogens and estrogen depletion, causing sustained hyperphosphorylation and inactivation of tumor suppressors including hypophosphorylated Rb and p130, even in the presence of antiestrogens or in the absence of estrogen.

Loss of p27 stimulates AIB1 phosphorylation and E2F1 transcriptional activity. The elevated CDK2 activity is known to activate E2F transcriptional activity, which plays an essential role in G₁-S transition of the cell cycle. To measure E2F1 activity, we transfected T47D^tTA and T47D^tTA/p27m cells with an E2F1-responsive luciferase reporter and measured the luciferase activity. The reporter activity was ∼2-fold higher in T47D^tTA/p27m cells compared with T47D^tTA cells when these cells were treated with vehicle, 4-OH-tamoxifen, ICI 182780, or estrogen-free medium (Fig. 4A). These results indicate that p27 deficiency significantly stimulates E2F1 activity in T47D^tTA/p27m cells.

AIB1 is a strong coactivator for E2F1 and plays an important role in breast cancer proliferation and antiestrogen resistance (22, 23, 36, 37). Because AIB1 cellular concentration and phosphorylation levels determine its coactivator activity, we assessed if p27 deficiency would alter AIB1 protein level and/or phosphorylation status using specific antibodies detecting total AIB1 and its individual phosphorylated sites known to be important for its coactivator activity (29). Whereas total AIB1 remained unchanged in T47D^tTA/p27m cells compared with T47D^tTA cells (Fig. 4B and C), its phosphorylation levels on five of the six known sites, with the exception of Ser⁵⁴³, were significantly enhanced in T47D^tTA/p27m cells, including phosphorylated Thr²⁴ (pT24) and phosphorylated Ser⁵⁰⁵ (pS505), pS857, pS860, and pS867 in the absence of estrogen treatment (Fig. 4C). Interestingly, 17β-estradiol treatment stimulated the phosphorylation of these sites in T47D^tTA cells but not in T47D^tTA/p27m cells (Fig. 4C), suggesting that the phosphorylation levels of these sites induced by p27 deficiency are similar to the levels induced by estrogen in T47D^tTA cells.

Next, we analyzed the effect of each AIB1 phosphorylation site on E2F1 reporter by comparing the coactivator activity of AIB1 with its mutants containing a threonine (T) or serine (S) to alanine (A) mutation. These mutant and wild-type AIB1 vectors showed similar expression levels in MCF-7 breast cancer cells. MCF-7 cells were used because they gave a better efficiency than T47D cells when multiple plasmids were cotransfected. Our experiments showed that expression of E2F1 or AIB1 alone had limited stimulation on E2F1 reporter activity, but coexpression of E2F1 and AIB1 robustly potentiated E2F reporter activity (Fig. 4D). Expression of the T24A, S505A, S543A, and S860A AIB1 mutants also coactivated E2F1 activity as wild-type AIB1 did. Nevertheless, coactivator activities were partially impaired for the S867A mutant and completely diminished for the S857A mutant (Fig. 4D). These results suggest that phosphorylation on two of the six sites of AIB1 has a positive effect on E2F1-mediated transcription. Because disruption of p27 increased the phosphorylation on five of the six known sites including pS857 and pS867, the hyperphosphorylated AIB1 in T47D^tTA/p27m cells should serve as a stronger coactivator for E2F1 compared with the hypophosphorylated AIB1 in T47D^tTA cells.

Disruption of p27 increases Akt activity. The adaptor protein Gab2 is a direct E2F target (30). Indeed, Gab2 was significantly increased in T47D^tTA/p27m cells compared with T47D^tTA cells; restoration of p27 caused a reversible decrease in Gab2 in T47D^tTA/p27m/+27 cells (Fig. 5A). The E2F-induced Gab2 plays an essential role in mediating E2F-dependent activation of the PI3K/Akt signaling pathway in U-2OS osteosarcoma cells (30). To address whether p27 deficiency–induced hyperactive E2F and overexpressed Gab2 would also potentiate Akt activation in breast cancer cells, we assayed total Akt and active Akt, the phospho-Ser⁴⁷³-Akt (pS473-Akt; ref. 38), by immunoblotting. The total Akt was comparable in T47D^tTA/p27m and T47D^tTA cells (Fig. 5A). However, the pS473-Akt was ∼3-fold higher in T47D^tTA/p27m cells than in T47D^tTA cells. Restoration of p27 in T47D^tTA/p27/+27 cells did not change total Akt but decreased active Akt (Fig. 5A). Accordingly, Akt protein immunopurified from T47D^tTA/p27m cells showed higher kinase activity than Akt from T47D^tTA cells as determined by in vitro kinase assays using glycogen synthase kinase (GSK)-3β as substrate (ref. 39; Fig. 5A). These results show that inactivation of p27 in T47D cells causes superactivation of Akt, probably by enhancing the transcriptional capability of E2F.

Inactivation of p27 promotes cell mobility and invasion. The PI3K/Akt signaling pathway plays a central role in cancer cell motility, invasion, and metastasis (31, 32). To assess the effects of p27 deficiency–triggered Akt activation on cell motility, we measured the extent of individual cell migration on extracellular matrix proteins by using blue fluorescent beads to determine the track area cells had moved through in 18 h. On culture plates coated with each of the four extracellular matrix proteins, the track area of T47D^tTA/p27m cells was significantly larger than the track area of T47D^tTA cells (Fig. 5B). Quantitative analysis of the track areas revealed that T47D^tTA/p27m cells migrated 2.1-, 4.6-, 4.5-, and 3-fold faster than T47D^tTA cells on fibronectin-, laminin-, collagen I–, and collagen IV–coated culture plates, respectively (Fig. 5C). These results indicate that inactivation of p27 in T47D cells remarkably promotes cell motility.

We also evaluated the capability of T47D^tTA, T47D^tTA/p27m, and T47D^{tTA/p27m/+p27} cells to invade through the extracellular matrix by using Matrigel Invasion Chambers. The average number of T47D^tTA/p27m cells that invaded through the Matrigel layer increased >3-fold compared with T47D^tTA and T47D^{tTA/p27m/+p27} cells (Fig. 5D). These results indicate that p27 deficiency significantly enhances the invasive capability of breast cancer cells.

T47D^tTA/p27m and T47D^tTA cells showed similar cell adhesion capability on fibronectin-, laminin-, collagen I–, or collagen IV–coated culture plates (data not shown). Immunoblotting and immunocytochemistry revealed that the levels and distribution patterns of E-cadherin, an epithelial marker for mammary epithelial association (40), were similar between T47D^tTA and T47D^tTA/p27m cells (data not shown). These results indicate that p27 deficiency does not alter E-cadherin expression and location as well as mammary epithelial cell-cell interaction.

I.v. injection of p27-deficient T47D^tTA/p27m cells induces lung metastasis in ovariectomized nude mice. T47D cells are estrogen-dependent human breast cancer cells, which can only form nonmetastatic local tumors in ovariectomized nude mice treated with estrogen (35). To test whether p27 deficiency in estrogen-dependent breast cancer cells could promote tumor formation in an ovarian hormone–independent manner, we compared tumor development and progression of T47D^tTA/p27m cells with T47D^tTA cells in ovariectomized nude mice. Both T47D^tTA and T47D^tTA/p27m cells developed local tumors in the mammary fat pats of all nude mice (n = 6) treated with estradiol pellets, but failed to develop tumors in the mammary fat pads of nude mice treated with placebo (Fig. 6A and data not shown). Histologic examination revealed that T47D^tTA tumors had a smooth surface and a relatively tight cell-cell association (Fig. 6A). In contrast, the edge of T47D^tTA/p27m tumors frequently protruded into the surrounding stromal tissues (Fig. 6A). These results suggest that loss of p27 function may make T47D^tTA/p27m tumors more invasive.

To address if the local tumors formed from T47D^tTA/p27m cells might develop systemic metastasis, we examined the lung and other internal organs of mice bearing T47D^tTA/p27m tumors in their mammary fat pats. However, we did not find any metastasis that developed in any secondary organs, suggesting that T47D^tTA/p27m tumors formed in the mammary gland are unable to accomplish the entire process of distant metastasis.

To determine whether p27 deficiency–induced cellular alterations could potentiate metastasis of breast cancer cells after intravasation, we compared the capability of T47D^tTA/p27m cells with that of T47D^tTA cells to form metastatic tumors in the lung using a well-established experimental metastasis method (41). T47D^tTA cells did not develop any tumors in the lung of ovariectomized nude mice (n = 5) 3 weeks after injection i.v. (Fig. 6C). In contrast, all of the seven ovariectomized nude mice developed lung tumors 3 weeks after receiving T47D^tTA/p27m cells i.v. In these mice lacking ovarian steroids, many tumors were visible on the lung surfaces, and big invasive tumors were observed on lung sections (Fig. 6B and C). In these tumor cells, strong immunoreactivity of p27-C40 was detected in the cytoplasm, which was similar to that seen in T47D^tTA/p27m cells in Fig. 3A. In addition, the mammary epithelial marker K8 was also detected in the lung tumor cells of nude mice injected with T47D^tTA/p27m cells, but not in the lung of nude mice injected with T47D^tTA cells (Fig. 6D), indicating that these metastatic tumors originated from T47D^tTA/p27m cells. These results show that loss of p27 in T47D estrogen-dependent breast cancer cells is sufficient to permit metastatic tumor formation in the lung in an ovarian hormone–independent manner, once these tumor cells enter the circulation.

Development of estrogen independence and tamoxifen resistance is a major problem in breast cancer endocrine therapy. Characterization of the genetic changes responsible for the transition from estrogen dependence to estrogen independence and from tamoxifen responsiveness to tamoxifen resistance is essential for understanding the mechanisms for this key step in progression. In this study, we found that the estrogen-dependent T47D cells contain only one p27 allele, suggesting that p27 is already decreased in the original T47D tumor. Our genome-wide screening using ERM identified the p27-deficient T47D^tTA/p27m cells that still express functional ERα but grow in an estrogen-independent manner and are fully resistant to tamoxifen. These observations recapitulate the similar features of ERα-positive breast cancers that progress from tamoxifen sensitive to tamoxifen resistant and are consistent with an important role of p27 reduction and cytoplasmic retention in breast cancer initiation and progression (13–15, 17–19). Because our results show that disruption of p27 is sufficient to permit estrogen-dependent and tamoxifen-sensitive breast cancer cells to grow in the absence of estrogen and in the presence of tamoxifen and that these growth phenotypes can be rescued by p27 restoration, we conclude that functional p27 is required for tamoxifen-mediated inhibition of ERα-positive breast cancer growth.

T47D^tTA/p27m cells also exhibited a partial resistance to ICI 182780 compared with T47D^tTA cells. ICI 182780 only reduced the size of T47D^tTA/p27m cell colonies and inhibited 37% of CDK2 activity. Previous studies have shown that both p27 and p21cip1 can contribute to antiestrogen-mediated cell cycle arrest in breast cancer cells (18). Our analyses also reveal that p21 is more obviously induced by ICI 182780 in T47D^tTA/p27m cells than in T47D^tTA cells and more cyclin D1-CDK4 and cyclin E-CDK2 were associated with p21cip1 in T47D^tTA/p27m cells (data not shown). Therefore, the partial ICI 182780 resistance of T47D^tTA/p27m cells may be accredited to the compensatory role of p21cip1 when p27 function is lost.

E2F1 transcriptional activity is significantly elevated in T47D^tTA/p27m cells because p27 deficiency activates CDK2, leading to Rb phosphorylation and E2F1 activation. Recently, AIB1 has been identified as a major coactivator for E2F1; E2F1 interacts with AIB1 NH₂ terminus and recruits AIB1 to the promoter of E2F1 target gene (22). Other studies also have shown that growth factors, cytokines, and estrogen can stimulate AIB1 phosphorylation at specific sites and AIB1 phosphorylation usually alters AIB1 coactivator activity for steroid receptors (29, 42). Nevertheless, the mechanisms leading to AIB1 phosphorylation are complex and many kinases including p38, IκB kinases, extracellular signal–regulated kinase, c-jun NH₂-terminal kinase, and GSKs have been shown to phosphorylate AIB1 (29). Because CDK2 and Akt are stimulated in p27-deficient T47D^tTA/p27 cells, these two kinases might also play a role in AIB1 phosphorylation. Our data show that loss of p27 in T47D^tTA/p27m cells enhances AIB1 phosphorylation on five of the six mapped sites, and the phosphorylation on four of these five sites was also enhanced by estrogen in T47D^tTA cells (Fig. 5C), suggesting that p27 deficiency causes an AIB1 phosphorylation pattern similar to that caused by estrogen. Importantly, mutation of two of these sites (S857 and S867) either attenuated or diminished AIB1 coactivator activity for E2F1. These results indicate that the phosphorylated AIB1 is a better coactivator for E2F1. Thus, p27 deficiency–induced AIB1 phosphorylation further potentiates E2F1 target gene transcription following CDK2 activation and Rb phosphorylation. This double enhancement of E2F1 activity may play a crucial role in breast cancer progression into estrogen independence and tamoxifen resistance.

Intriguingly, T47D^tTA/p27m cells can grow in estrogen-free medium but are unable to develop tumors in the mammary fat pads of ovariectomized nude mice with or without tamoxifen treatment, suggesting that tumor formation in the mammary fat pads in the absence of ovarian hormones is more difficult than cell growth in estrogen-free medium for T47D^tTA/p27m cells. With estrogen replacement, both T47D^tTA and T47D^tTA/p27m cells developed solid tumors in the mammary fat pads. Although T47D^tTA/p27m tumors are morphologically more invasive, both T47D^tTA and T47D^tTA/p27m tumors formed in the mammary fat pads are not metastatic, suggesting that p27 deficiency alone is not sufficient for these tumor cells to progress into fully metastatic cancer cells.

The next question is why T47D^tTA/p27m cells lacking p27, but not T47D^tTA cells with p27, develop lung metastasis when injected into the circulation of ovariectomized mice. To develop lung metastasis, tumor cells, after intravasation or injection into the circulation, must attach to the endothelium of small lung vessels, migrate or invade through the vessel wall, and form tumors in the lung (41). Our analyses show that T47D^tTA and T47D^tTA/p27m cells have similar adhesion capability to extracellular matrix proteins, suggesting that p27 deficiency does not change cell adhesion properties and the increase in T47D^tTA/p27m cell metastatic potential is unlikely due to the change of endothelial attachment. On the other hand, cell migration and invasion abilities are correlated with metastasis. Our data show that p27 deficiency in T47D^tTA/p27m cells dramatically promotes their motility and invasiveness compared with T47D^tTA cells. Thus, p27 deficiency–induced increase in cell motility and invasion may be responsible for the ability of T47D^tTA/p27m cells to develop lung metastasis.

The mechanisms for p27 deficiency–induced cell motility, invasiveness, and lung metastasis may be very complex. We showed that T47D^tTA/p27m cells have an enhanced AIB1 phosphorylation, which makes AIB1 a stronger coactivator for E2F1. Previous studies have established that E2F1 strongly activates Akt through a direct up-regulation of Gab2 expression (30). In this study, p27 deficiency in T47D^tTA/p27m cells indeed increases E2F1 transcriptional activity, Gab2 protein, and Akt activation following AIB1 phosphorylation. Akt plays a central role in cancer cell motility, invasion, and metastasis (31, 32). Therefore, p27 deficiency–enhanced AIB1 phosphorylation may significantly contribute to cell motility, invasion, and metastasis by activating Akt. This interpretation is consistent with the lower Akt activity and less lung metastasis of the mammary gland tumors induced by oncogenes and carcinogens in AIB1-null mice compared with wild-type mice (27, 28).

Furthermore, it has been shown that overexpression of a p27 cytosolic mutant increases total Akt by inhibiting its turnover (43). Nevertheless, it is unlikely that the disrupted p27 fragments in T47D^tTA/p27m cells can play the same role that the full-length cytosolic p27 mutant does because there is no change in total Akt in T47D^tTA/p27m cells. In addition, p27 deficiency may release its direct inhibition on cell motility. Although p27 has been reported as both an inhibitor and a stimulator of cell migration (44–47), a recent article has suggested that the controversial observations about the role of p27 in cell motility may be due to the type of cell motility assayed (48). This study showed that p27 binds and impairs the function of the microtubule-destabilizing protein stathmin and thereby plays an inhibitory role in amoeboid cell migration, a type of cell movement used by tumor cells in tissues (48). Disruption of p27 in T47D^tTA/p27m cells may impair the direct inhibitory role of p27 in cell motility and tumor cell metastasis. Finally, down-regulation of p27 causes an up-regulation of G protein–coupled receptor 48 (GPR48), and GPR48 can enhance cancer cell invasiveness and metastasis (49). Therefore, multiple pathways may be involved in the p27 deficiency–induced increase in breast cancer cell motility, invasiveness, and metastasis.

Although p27 null mutations are rare, ∼50% human tumors exhibit decreased p27. The p27 reduction and cytoplasmic mislocalization correlate with more aggressive phenotypes, including HER2/neu overexpression, high proliferating indices, active invasive behavior, and high mortality (13–15, 50). These observations suggest that p27 abundance inversely correlates with breast cancer progression. However, studies using transgenic mice suggest that a threshold of p27 seems to be required for normal mammary epithelial proliferation and mammary tumorigenesis induced by oncoproteins. For example, p27^+/− mice exhibit a faster mammary epithelial growth and mouse mammary tumor virus (MMTV)-neu–induced tumorigenesis, but p27^−/− mice exhibit a delayed mammary gland morphogenesis and MMTV-neu–induced tumorigenesis (21). Molecular analysis suggests that certain p27 is required for stabilization of cyclin D1 and assembly and nuclear translocation of cyclin D1-CDK4 complexes. Thus, CDK4 in p27^−/− mammary epithelial cells cannot be activated, which in turn suppresses the neu-induced, cyclin D1-dependent mammary epithelial transformation (21). Similarly, cyclin D1 is decreased and CDK4 activity is reduced manyfold in our T47D^tTA/p27m cells (data not shown). However, T47D^tTA/p27m cells exhibit comparable proliferation as T47D^tTA cells when cultured in growth medium. These results indicate that T47D^tTA/p27m cells are no longer dependent on cyclin D1-CDK4 for survival and growth. Although it is unknown whether the T47D parent tumor has acquired cyclin D1-CDK4–independent growth, the increase in cyclin E-CDK2 activity after loss of p27 in T47D^tTA/p27m cells is cyclin D1-CDK4 independent and should be responsible for the antiestrogen insensitivity of these cells.

Based on this and previous studies discussed above, the effect of p27 gene dosage on breast cancer is likely dependent on the specific stages of tumor initiation and progression. It seems reasonable to propose that p27 reduction in the mammary epithelium and early stages of tumorigenesis may promote cell proliferation and accelerate hormone-, oncoprotein-, and carcinogen-induced transformation while still maintaining cyclin D1 and CDK4 activity. Once these transformed cells acquire a cyclin D1-CDK4–independent growth feature, a further reduction or loss or mislocalization of p27 may facilitate breast tumor cells to develop more aggressive cancer phenotypes including estrogen-independent growth, antiestrogen insensitivity, and metastasis. During this process, p27 insufficiency– or p27 deficiency–induced increase in AIB1, E2F1, Gab2, and Akt activities and decrease in p27 inhibitory role in cell motility may have a major contribution to breast cancer progression.

Grant support: NIH grants DK58242, CA119689, CA112403 (J. Xu), and CA84208 (Z. Songyang) and American Cancer Society Research Scholar Award RSG-05-082-01 (J. Xu).

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 Lan Liao for experimental assistance.