We have studied the role of the cyclin-dependent kinase (Cdk) inhibitor p27Kip1 in postnatal mammary gland morphogenesis. Based on its ability to negatively regulate cyclin/Cdk function, loss of p27 may result in unrestrained cellular proliferation. However, recent evidence about the stabilizing effect of p27 on cyclin D1–Cdk4 complexes suggests that p27 deficiency might recapitulate the hypoplastic mammary phenotype of cyclin D1–deficient animals. These hypotheses were investigated in postnatal p27-deficient (p27−/−), hemizygous (p27+/−), or wild-type (p27+/+) mammary glands. Mammary glands from p27+/− mice displayed increased ductal branching and proliferation with delayed postlactational involution. In contrast, p27−/− mammary glands or wild-type mammary fat pads reconstituted with p27−/− epithelium produced the opposite phenotype: hypoplasia, low proliferation, decreased ductal branching, impaired lobuloalveolar differentiation, and inability to lactate. The association of cyclin D1 with Cdk4, the kinase activity of Cdk4 against pRb in vitro, the nuclear localization of cyclin D1, and the stability of cyclin D1 were all severely impaired in p27−/− mammary epithelial cells compared with p27+/+ and p27+/− mammary epithelial cells. Therefore, p27 is required for mammary gland development in a dose-dependent fashion and positively regulates cyclin D–Cdk4 function in the mammary gland.
Initiation of cell division involves the integration of several growth signals, both stimulatory and inhibitory, that converge on the cell cycle machinery. The cell cycle is promoted by a family of cyclin-dependent kinases (Cdks), and Cdk activity is negatively regulated by Cdk inhibitors (CKIs). Thus, CKIs lie at a critical point in this integrated network, serving as checkpoint controls for cell cycle progression (Sherr and Roberts 1999). The CKIs can be subdivided into two categories, the Ink4 proteins, which are inhibitors of cyclin D–Cdk2/4 complexes, and the Cip/Kip proteins, which inhibit all cyclin–Cdk complexes. The Cip/Kip family includes p27Kip1, p21Cip1/Waf1, and p57Kip2. p27Kip1, originally identified as a Cdk inhibitory activity induced by antimitogenic signals (Polyak et al. 1994; Toyoshima and Hunter 1994), is known to bind to cyclin D–Cdk4, cyclin E–Cdk2, and cyclin A–Cdk2 complexes (Soos et al. 1996).
Stimulation of cyclin D–cdk4 activity in early G1 and cyclin E–cdk2 activity in mid-G1 results in the sequential phosphorylation of their mutual target, the retinoblastoma gene product, pRb. Cdk-mediated hyperphosphorylation of pRb causes release of pRb-mediated inhibition of E2Fs and other factors that drive the cell cycle forward by transactivating genes required for S phase entry (Nevins 1992; Kato et al. 1993; Johnson et al. 1993; Duronio and O'Farrell 1995). Overexpression of p27 reversibly arrests cells in G1, suggesting that p27 can limit the G1 to S transition. Inhibition of G1 progression by p27 requires the inhibition of Cdk activity, including cyclin D–Cdk4 and cyclin E–cdk2 complexes (Sherr 1996). A direct physical association of p27 with cyclin–Cdk complexes can restrain Cdk activity and maintain Rb in a hypophosphorylated state that sequesters E2F (Soos et al. 1996). By this mechanism, p27 inhibits progression into S phase (Weintraub et al. 1992, Weintraub et al. 1995). Consistent with these observations, cell culture studies have shown that expression of p27 is elevated in quiescent cells but declines as cells actively undergo DNA synthesis (Nourse et al. 1994).
Genetic evidence also suggests that p27 is a critical regulator of cellular proliferation in vivo. Mice with homozygous disruption of the p27 gene (p27−/−) are 15–30% larger than their wild-type (p27+/+) or hemizygous (p27+/−) littermates due to hyperplasia in several organs, including spleen, thymus, pituitary, ovaries, and testes (Fero et al. 1996; Kiyokawa et al. 1996; Nakayama et al. 1996). Tumors of the intermediate lobe of the pituitary were observed in p27−/−mice by 6 mo of age, and both p27+/+ and p27−/− mice were more susceptible to radiation- or chemically induced lung and colon tumors (Fero et al. 1998). Therefore, p27 acts as a tumor suppressor in many tissues, perhaps through its ability to inhibit cell cycle progression. Although homozygous gene inactivation has not yet been reported in any human tumors, hemizygous deletions have been detected (Pietenpol et al. 1995; Spirin et al. 1996; Takeuchi et al. 1998). Multiple human tumors exhibit abnormally low levels of p27 protein, including tumors of the breast (Clurman and Porter 1998; Loda et al. 1997). Decreased p27 levels in breast tumors correlate with a poor patient prognosis, implying a regulatory role of p27 in cell cycle control in transformed mammary epithelium (Porter et al. 1997; Tan et al. 1997).
High cyclin D1 protein levels have been observed in a large percentage of breast tumors (Gillett et al. 1996; Fredersdorf et al. 1997; Robker and Richards 1998; Medina 1996). In addition, transgenic mice overexpressing cyclin D1 in the mammary glands develop ductal hyperplasias and tumors (Wang et al. 1994), suggesting that increased cyclin D1–Cdk4 activity may lead to uncontrolled growth of the mammary epithelium. Conversely, cyclin D1–deficient (cyl−/−) mouse mammary glands display developmental abnormalities associated with a defect in cell cycle progression, including decreased proliferation, reduced lateral branching, and impaired differentiation (Fantl et al. 1995; Sicinski et al. 1995). Normally, the rudimentary ductal network of quiescent epithelial cells in the immature mouse mammary gland rapidly proliferates in response to increased production of pubertal hormones, such as estrogen, while undergoing extensive lateral epithelial branching. Pregnancy induces a second round of proliferation followed by differentiation to form the secretory epithelium responsible for milk production. The cessation of nursing induces involution, a process that eliminates the majority of the mammary epithelium through apoptosis within the time span of several days (for review see Medina 1996). When stressed with the enormous proliferative pressure associated with pregnancy, cyl−/− mammary glands display delayed proliferation, decreased lateral branching, and incomplete differentiation; thus, cyl−/− mammary glands cannot support lactation (Fantl et al., 1994; Sicinski et al. 1995). These observations suggest that cyclin D1 is critical for mammary gland morphogenesis and function. Given the opposing nature of p27 to cyclin D1 activity, it might be expected that mammary glands from p27-deficient mice would display a hyperproliferative phenotype, in direct contrast with the decreased growth of the cyclin D1–deficient mammary gland.
Although several studies have shown that high levels of p27 inhibit both Cdk2 and Cdk4 activities, the activity of cyclin E–Cdk2 complexes appears to be more potently antagonized by p27 than any other Cdk-containing complex. Furthermore, there is increasing evidence to suggest that p27 may, in fact, promote the activity of Cdk4-containing complexes in primary mouse embryonic fibroblasts. Assembly, stability, and function of cyclin D1–Cdk4 complexes are greatly impaired in the absence of p27 in this cell culture system (Cheng et al. 1999). However, p27-deficient mouse fibroblasts did not exhibit overall cell cycle defects. This observation might be explained by the potential presence of other signaling pathways that allow for cell cycle progression in the absence of pRb phosphorylation, and these cells therefore may not be completely dependent on cyclin D1–cdk4 activity for cell cycle control. On the other hand, a cell type that displays growth defects in the absence of cyclin D1, such as the mammary epithelium, would provide an elegant model system in which to test this hypothesis. If p27 is required for the assembly and function of cyclin D1–cdk4 complexes, then loss of p27 in the mammary epithelium would result in a phenotype that mimics the hypoplastic phenotype of the cyclin D1–deficient mammary epithelium.
We investigated these hypotheses in p27+/−and p27−/− mouse mammary glands at several stages of postnatal mammary gland morphogenesis. Transplantation of p27−/− epithelium into precleared mammary fat pads of wild-type mice allowed us to analyze the mammary glands at unique physiological stages, including pregnancy, lactation, and involution, thus circumventing the inability to observe p27−/− mammary gland differentiation due to the infertility of p27−/− females. We report that p27+/− mammary glands displayed increased proliferation and delayed involution. Intriguingly, p27−/− mammary glands displayed a decrease in epithelial proliferation with a marked delay in differentiation, similar to the phenotype of the cyl−/− glands. Cyclin D1–cdk4 assembly and activity, nuclear localization of cyclin D1, and the stability of cyclin D1 were all markedly reduced in p27−/− mammary glands. These results suggest that the absence of p27 results in a block in cell cycle progression in mammary epithelial cells, probably a result of the loss of cyclin D1–cdk4 activity.
Materials and Methods
Mouse Strains and Mammary Gland Transplantations
All mice were derived from p27+/− P1 on a mixed C57Bl/6-129/SvJ background strain, a gift from Dr. Andrew Koff (Memorial Sloan-Kettering Cancer Center, New York, NY) (Kiyokawa et al. 1996). All mice were housed in the Animal Care Facilities at Vanderbilt University School of Medicine in accordance with Association of Assessment and Accreditation of Laboratory Animal Care guidelines. Mice were genotyped by PCR analysis of genomic DNA using the following primers from the endogenous p27 allele to generate a 206-bp product: primer 1, 5′-tca aac gtg aga gtg tct aac gg-3′; and primer 2, 5′-agg ggc tta tga ttc tga aag tcg-3′. The following primer from the neo cassette disrupting the p27 gene (5′-ata ttg ctg aag agc ttg gcg g-3′) was used to generate a 298-bp product with primer 1.
Mammary glands from virgin mice were collected in the diestrus phase of the estrous cycle as determined by cytological smear of vaginal cells. Before 6 wk of age, mice were not analyzed for phase of estrous. Mammary gland transplantations were performed as described (Edwards et al. 1996). In brief, the right and left no. 4 inguinal mammary glands of 21-d-old p27+/+ recipient females were cleared of endogenous epithelium by surgically removing the mammary fat pad from the nipple to the lymph node. The no. 4 mammary glands of 6-wk-old p27+/+or p27−/− female donors were harvested, and a 2-mm3 segment was used for transplantation into the center of the remaining p27+/+ recipient fat pad. Mammary glands were analyzed 6–12 wk after transplantation. Female mice were mated after 8 wk of age or 8 wk after transplantation. Mammary glands from pregnant mice were harvested at 16.5 days postcoitum (dpc). After 10 d of lactation, pups were withdrawn from p27+/+ or p27+/− mothers, and glands were harvested at 1, 3, 5, and 21 days postforced wean [dpfw]). For reconstituted glands, pups were immediately withdrawn within 2–4 h of birth to analyze the “lactational” phenotype of the reconstituted mammary glands. (This was done to prevent precocious involution of the reconstituted mammary gland; during the transplant procedure, the primary epithelial duct was severed from the nipple. At lactation, the sealed primary duct becomes distended with milk, a condition which induces involution within 24 h.) As indicated, some mice of each genotype were given a 90-d release estrogen (0.1 mg)/progesterone (10 mg) pellets (Innovative Research of America), implanted subcutaneously between the scapulae. Mammary glands were harvested after 60 d of estrogen/progesterone exposure.
Western Blot Analyses
Mammary glands were harvested and homogenized immediately as described previously (Lenferink et al. 2000). Total protein (20 μg) was separated by SDS-PAGE and transferred to nitrocellulose membranes. Western blot analyses were performed as described previously (Brantley et al. 2000) using the following antibodies: p27 (Transduction Laboratories); pRb and cyclin D1 (BD PharMingen); cyclin E, Cdk2, and Cdk4 (Santa Cruz Biotechnology, Inc.); α-lactalbumin and β-casein (Neomarkers); keratin-14 (Amersham Pharmacia Biotech); and signal transducer and activator of transcription (Stat)5a and phospho-Stat5 (Santa Cruz Biotechnology, Inc.).
Mammary glands were harvested and immediately fixed in 10% formalin (VWR Scientific). Hematoxylin-stained whole mount preparations of no. 4 mammary glands were prepared as described previously and photodocumented (Lenferink et al. 2000). Paraffin-embedded mammary glands were sectioned (5 μm), rehydrated, and stained with Mayer's hematoxylin and eosin B–phloxine (Sigma-Aldrich). For immunohistochemistry, sections were treated as described previously (Brantley et al. 2000) using rabbit polyclonal p27 antibody (Santa Cruz Biotechnology, Inc.) or cyclin D1 antibody (Upstate Biotechnology).
BrdU and TUNEL Analyses
A sterile solution of BrdU (10 mg/ml; Sigma-Aldrich) in PBS (pH 7.4) was administered to mice by intraperitoneal injection (0.1 mg/kg). Mammary glands were harvested after 3 h, paraffin-embedded, and sectioned. Immunohistochemical detection of BrdU incorporation was performed using a mouse monoclonal anti-BrdU antibody (Zymed Laboratories) according to the manufacturer's instructions. Detection of apoptosis in paraffin sections by terminal deoxynucleatidyl transferase–mediated dUTP–biotin nick end labeling (TUNEL) analysis was performed using terminal deoxynucleotidyl transferase (Intergen Co.) to label the ends of fragmented DNA with digoxigenin, which was immunohistochemically localized using a mouse monoclonal antidigoxigenin antibody (Intergen Co.) according to the manufacturer's instructions.
Isolation and Culture of Primary Mammary Epithelial Cells
10 inguinal mammary glands per genotype were harvested from 21-d-old siblings and digested at 37°C for 4 h in 100 U/ml hyaluronidase and 3 mg/ml collagenase A (Sigma-Aldrich) in PBS (pH 7.4). The resulting cell suspension was washed five times in PBS plus 10% FCS, plated on dishes coated with growth factor–reduced Matrigel (Becton Dickinson) in primary mammary epithelial cell (PMEC) medium (serum-free DME:F12 [50:50; GIBCO BRL], 5 ng/ml EGF [Clonetics], 5 ng/ml 17-β estradiol [Sigma-Aldrich], 5 ng/ml progesterone [Sigma-Aldrich], and 50 ng/ml insulin [Clonetics]), and cultured at 37°C, 5% CO2. In some cases, monolayers were treated with cycloheximide (1 μg/ml; Sigma-Aldrich) for 0–90 min, washed with PBS, and lysed in 1% Nonidet P-40 (vol/vol). Extracts were collected as described above. Cyclin D1 protein was detected by Western blot analysis of equal amounts of protein (20 μg). For three-dimensional cultures, PMECs were plated within a thick layer of growth factor–reduced Matrigel. PMEC medium was layered on top of the polymerized Matrigel. Cultures were photodocumented using the Olympus DP-10 digital camera adapted to an Olympus CK-2 phase–contrast microscope.
Proliferating unsynchronized PMECs were harvested by trypsinization, fixed in ice-cold methanol, and labeled with 50 μg/ml of propidium iodide (Sigma-Aldrich) as described previously (Lenferink et al. 2000). A total of 10,000 stained nuclei per sample was analyzed in a FACSCalibur™ flow cytometer (Becton Dickinson). DNA histograms were modeled using Modfit-LT Software (Verity).
Immunoprecipitation and Kinase Assays
Mammary gland extracts were prepared as described above. 500 μg of total protein was used for immunoprecipitation (IP) as described previously (Lenferink et al. 2000) using polyclonal antibodies against cyclin D1, Cdk2, and Cdk4 (Santa Cruz Biotechnology, Inc.). The precipitates were either used for in vitro kinase assays or resolved by SDS-PAGE followed by Western blot analysis. For kinase assays, IP products were washed in ice-cold PBS, then equilibrated to ice-cold kinase buffer (50 mM Hepes, pH 7.5, 10 mM MgCl2, 2.5 mM EGTA, 1 mM DTT, 0.1 mM NaF, 0.1 mM Na3VO4, 1 mM ATP) for 20 min and resuspended in 20 μl of kinase buffer. Cdk2 IPs were assayed for kinase activity against 0.1 μg histone H1 (HH1; Roche Molecular Biochemicals), whereas Cdk4 IPs were assayed against 0.1 μg of a 46-kD fragment of pRb (Santa Cruz Biotechnology, Inc.). Kinase reactions were performed in a final volume of 30 μl in the presence of 5 μCi [γ-32P]ATP (specific activity 3,000 Ci/mmol; Amersham Pharmacia Biotech) for 45 min at 30°C. The entire reaction volume was electrophoresed on an 8% polyacrylamide gel at 60 mA for 3 h at 4°C. Gels were dried at 80°C and exposed to autoradiographic film (16 h at −80°C).
Whole Organ Culture of Mammary Glands
Inguinal (no. 4) mammary glands were harvested from 5-wk-old mice under sterile conditions. Only the portion of the mammary gland between the nipple and the lymph node was used for culture. The mammary glands were cultured at the air–liquid interface on Costar 8-μ nitrocellulose filters in a chemically defined medium consisting of DME/F12 supplemented with 5 ng/ml 17-β estradiol, 5 ng/ml progesterone, 50 ng/ml recombinant human insulin, 5 μg/ml prolactin (Sigma-Aldrich), and 5 ng/ml EGF. Whole mammary glands were cultured in this manner for 0–10 d at which point whole mount hematoxylin staining was used to visualize the epithelium.
p27 Expression Is Detected at All Stages of Postnatal Mammary Gland Morphogenesis
The steady state protein levels of p27 were measured at various stages of mammary gland development in wild-type C57Bl/6X129Sv/J female mice (Fig. 1 A). Similar levels of p27 protein expression were detected at all stages examined by Western blot analysis, including virgin, midpregnancy, lactation, and involution. Immunohistochemical detection of p27 in wild-type virgin mammary gland sections revealed intense staining for p27 within the epithelium with lower staining throughout the stroma (Fig. 1B and Fig. E). Stromal staining was not due to background trapping of secondary antibody, since sections of wild-type mammary glands that were incubated in the presence of secondary antibody alone did not produce any stromal staining (data not shown). Staining for epithelial p27 protein was seen in the cytoplasm and nuclei of wild-type mammary glands, but was primarily localized to the nuclei of both intact p27+/− glands (Fig. 1 C) and wild-type fat pads reconstituted with p27+/− epithelium (Fig. 1 F). Although intermediate p27 staining was detected in p27+/− glands compared with wild-type glands, detection of p27 expression was absent in intact p27−/− mammary glands (Fig. 1 D) and in the epithelium of wild-type mammary fat pads transplanted with p27−/− cells (Fig. 1 G). Blood vessels from wild-type fat pads transplanted with p27−/− tissue stained positive for p27 (Fig. 1 G, arrow), since the vasculature was derived from the wild-type host (for review see Young 2000). These results confirm the loss of p27 gene expression in p27−/− mice and demonstrate epithelial expression of p27 at various stages of postnatal mammary gland morphogenesis.
Loss of One p27 Allele Causes Increased Mammary Epithelial Growth, Whereas Loss of Both p27 Alleles Results in Decreased Growth
To determine the consequences of a reduction or loss in p27 expression on mammary gland morphogenesis, whole mount preparations and histological sections of virgin p27+/+, p27+/−, and p27−/− mammary glands were examined (Fig. 2). At 21 d of age, mammary glands from p27+/− mice demonstrated a consistent increase in the size of terminal end buds (TEBs) compared with age-matched wild-type controls (Fig. 2 A). TEBs are thought to represent the most highly proliferative region of the progressing mammary epithelium (for review see Medina 1996). Furthermore, p27+/− mammary epithelium consistently penetrated a greater area of the mammary fat pad (compare the relative location of the TEBs to the centrally located lymph node [see asterisk] of the no. 4 mammary gland), suggesting an increased rate of ductal growth (Fig. 2 A). Mammary epithelium in p27−/− mice displayed the opposite phenotype, characterized by a decrease in the size and number of TEBs and a reduced penetrance through the fat pad (Fig. 2 A). These differences persisted through 35 (Fig. 2B and Fig. C) and 70 d (Fig. 2D and Fig. G). At 35 d, TEBs were present in wild-type glands and penetrated to the distal edge of the lymph node. The strikingly prominent TEBs of the p27+/− mammary glands progressed well past the lymph node, whereas the leading edge of the p27−/− mammary epithelium lacked obvious TEBs. It was also apparent that the primary ducts of the p27−/− mammary glands had a substantial decrease in secondary branches compared with wild-type or p27+/− mammary glands.
At 70 d, TEBs regressed from wild-type glands, but remained apparent in age-matched p27+/− mice at identical phases of estrous (Fig. 2D and Fig. E). Mammary glands from p27−/− mice did not display TEBs. The decreased epithelial stromal ratio in the p27−/− mammary glands was apparent in histologic sections, and the diameter of the ductal lumena appeared to be decreased in size at this stage when the mammary glands were fully matured (Fig. 2F and Fig. G). The number of TEBs per no. 4 mammary gland at 35 d (Fig. 2 B) and the number of epithelial structures visible per 100× field in sections taken from mammary glands at 70 d (Fig. 2 F) were counted for all three genotypes and are shown in Fig. 2 I.
Although p27−/− mice are known to have altered ovarian function, it is thought that basal levels of the ovarian hormones that regulate virgin mammary gland morphogenesis are present in relatively equal concentrations in wild-type versus p27-deficient mice (Tong et al. 1998). However, ovarian production of nidatory estrogen required for embryonic implantation is reduced in p27-deficient animals. Based on this, we wanted to determine if increased concentrations of ovarian hormones could rescue the hypoplastic phenotype of p27-deficient mammary glands. Slow release (90-d) estrogen/progesterone pellets were implanted under the dorsal skin of 180-d-old virgin mice. After 60 d, the mammary glands were harvested for analysis. Mammary epithelium from all three genotypes displayed lobular bud formation (Fig. 2 H). The epithelium of p27+/− mice was more abundant than that seen in wild-type glands and displayed areas which may represent focal hyperplasias (Fig. 2 H, arrowhead). Mammary epithelium in p27−/− glands still exhibited reduced secondary branching compared with controls. These results demonstrate that in the absence of p27, mammary epithelium has fewer TEBs during early development and persistently reduced secondary branching. In contrast, p27+/− mammary glands display an increased rate of ductal development with increased numbers of TEBs.
Altered Cell Growth of the Mammary Epithelium Due to Loss of p27 Is an Epithelial Cell Autonomous Phenotype
To determine whether the p27−/− phenotype was intrinsic to the mammary epithelium and not secondary to stromal or hormonal influences, p27−/− mammary tissue was transplanted into wild-type mammary fat pads cleared of endogenous breast epithelium. This technique has been used to reconstitute mammary fat pads with donor mammary epithelium, which will permeate the fat pad and differentiate in a manner consistent with the native gland. p27 protein expression was detected in mammary glands reconstituted with p27+/+ or p27+/− epithelium (see Fig. 1E and Fig. F) but not in mammary glands reconstituted with p27−/− cells (Fig. 1 G). After 6 wk, mammary glands reconstituted with p27+/+, and p27+/− epithelium exhibited extensive ductal branching with prominent TEBs (Fig. 3A and Fig. B, arrows). In contrast, p27−/− epithelium displayed decreased lateral branching. There was a marked reduction of TEBs in the mammary glands reconstituted with p27−/− tissue, similar to what is seen in intact mammary glands from p27−/− animals (see Fig. 2C, Fig. E, and Fig. G). These characteristics persisted for >12 wk after transplantation (data not shown). These results suggest that the hypoplastic phenotype of p27−/− mammary glands is independent from hormonal or stromal factors.
PMECs harvested from p27+/+, p27+/−, and p27−/− mice were cultured in a three-dimensional extracellular matrix in the presence of mitogenic hormones to further investigate the epithelial autonomy of the p27−/− and p27+/− phenotypes (Fig. 3C and Fig. D). Tubulogenesis in p27+/+ and p27+/− PMECs occurred within 5 d of culture (21.4 ± 7.3 and 32.1 ± 5.3 branches per 400× field in p27+/+ and p27+/− PMECs, respectively). p27−/− PMECs failed to induce tubular structures within this time frame (0 branches per 400× field). After 15 d in culture, p27−/− PMECs began to organize into rudimentary tubules (data not shown), suggesting that p27−/− mammary epithelial cells are capable of growth and tubulogenesis but in a delayed fashion compared with cells containing p27. In contrast, complex branch formations occurred at a more rapid rate in p27+/− PMECs compared with wild-type cells.
Loss of One p27 Allele Causes Increased Proliferation of the Mammary Epithelium; Loss of Both p27 Alleles Causes Decreased Proliferation
We next measured mammary epithelial proliferation by labeling virgin 5-wk-old mice with BrdU and assessing BrdU incorporation into chromosomal DNA (Fig. 4, A–C). When TEBs were examined, a 2.6-fold increase in BrdU incorporation into p27+/− compared with p27+/+ mammary epithelial cells was observed (Fig. 4A and Fig. E; P < 0.001, n = 6). Although TEBs are rare in p27−/− mammary glands, the structures which most closely resembled TEBs were analyzed for BrdU incorporation and displayed a 2.6-fold decrease compared with wild-type TEBs (Fig. 4A and Fig. E; P < 0.006, n = 6). BrdU incorporation into ductal epithelium was also examined (Fig. 4 B). As expected, the overall rate of BrdU labeling was decreased in ductal epithelium compared with that in TEBs for each genotype analyzed. However, quantification of BrdU-labeled cells demonstrated that p27+/− ducts contained 1.8-fold more BrdU-positive nuclei than wild-type ducts, whereas p27−/− glands displayed a 2-fold decrease in BrdU incorporation compared with wild-type ducts (3.1 ± 0.36%, 5.58 ± 0.49% [P < 0.04], and 1.38 ± 0.12% [P < 0.02] for p27+/+, p27+/−, and p27−/−, respectively). The altered epithelial proliferation correlated with increased epithelial content in p27+/− glands and decreased epithelial content in p27−/− glands (in 10 randomly chosen 400× fields, p27+/+ equals 7,832 cells, p27+/− equals 10,490 epithelial cells, and p27−/− equals 5,502 epithelial cells). Proliferation in wild-type mouse fat pads that had been transplanted with p27+/+, p27+/−, or p27−/− mammary epithelium was examined at 5 wk after transplantation (Fig. 4 C). Again, BrdU incorporation in p27+/− mammary epithelium (31.2 ± 4.6%) was elevated compared with wild-type mammary epithelium (17.1 ± 4.4%), whereas BrdU labeling was decreased in p27−/− epithelium (6.7 ± 3.0%). Labeling of purified PMECs in culture demonstrated that a monolayer of p27+/− PMECs had a higher rate of BrdU incorporation compared with wild-type PMECs (Fig. 4D and Fig. E), whereas p27−/− PMECs had a decreased rate of BrdU labeling. Flow cytometric analysis of PMEC nuclei stained with propidium iodide demonstrated a larger G1 with a smaller S phase fraction in p27−/− cells compared with wild-type cells (Fig. 4 F, P < 0.045, S phase comparison, Student's t test). In contrast, p27+/− cells displayed an increase in the percentage of cells in S phase (P < 0.02, Student's t test) with a lower proportion of cells in G1 compared with wild-type cells, consistent with an accelerated G1 to S transition. These results suggest that p27 function is critical for the growth of the mammary gland. Whereas loss of one p27 allele results in markedly increased epithelial cell proliferation, loss of both p27 alleles reduces breast epithelial proliferation.
Total Loss of p27 Affects the Formation and Function of Cyclin D1–Cdk4 Complexes in Mammary Epithelial Cells
To elucidate the mechanism by which proliferation is increased in p27+/− and decreased in p27−/− mammary glands, the abundance of cyclin D1, cyclin E, Cdk2, and Cdk4 proteins was examined. Western blot analyses of virgin mammary extracts revealed that levels of Cdk2, Cdk4, and cyclin E were not affected significantly by the relative amount of p27 protein, whereas cyclin D1 levels were significantly lower in p27−/− glands (Fig. 5 A). Phosphorylated pRb was detected in lysates from all three genotypes, as shown by the slowly migrating immunoreactive species. However, hyperphosphorylation of ppRb was greater in p27+/+and p27+/− lysates compared with p27−/− lysates. Cyclin D1 from p27+/+ and p27+/− mammary glands coprecipitated similar levels of p27 and Cdk4, even though there was significantly less total p27 available in p27+/− glands compared with p27+/+ glands (Fig. 5 B). In contrast, experiments performed with Cdk2 antibodies showed a relative decrease in the amount of Cdk2-associated p27 in p27+/− mammary extracts but increased levels of Cdk2-associated cyclin E. As expected, cyclin D1 antibodies were unable to precipitate p27 in p27−/− glands. However, most strikingly coprecipitation of Cdk4 with cyclin D1 was eliminated in the absence of p27, even though both proteins were present in the p27−/− extracts. Antibodies directed against Cdk2 were able to coprecipitate cyclin E in the absence of p27. These data suggest that the gene dosage of p27 influences the association between cyclin D1 and Cdk4 but not of cyclin E with Cdk2.
The kinase activities of immunoprecipitated Cdk4 complexes were measured in p27+/+, p27+/−, and p27−/− mammary gland extracts using a 46-kD COOH-terminal fragment of Rb as a substrate (Fig. 5 C). Cdk4 immunoprecipitates from p27+/− extracts were able to phosphorylate p46-Rb at a level similar to that observed with p27+/+ extracts. Cdk4 activity against p46-Rb was barely detectable in lysates from p27−/− glands. Phosphorylation of HH1 by Cdk2 immunoprecipitates was comparable in extracts from each genotype tested. These results suggest that the loss of p27 does not affect the kinase activity of Cdk2, whereas Cdk4 relies on the presence of p27 in mammary epithelium for its interaction with cyclin D1 and maximal kinase activity against Rb.
To examine cyclin D1 stability in the absence of p27, PMECs were treated with cycloheximide, and their content of cyclin D1 protein was followed over time (Fig. 5 D). The level of cyclin D1 in p27+/+ and p27−/− PMECs decreased as a function of exposure to cycloheximide. However, cyclin D1 levels decreased at a faster rate in p27−/− PMECs (t1/2 < 15 min) compared with wild-type PMECs (t1/2 ∼30–45 min). This is consistent with previous cell culture studies in which cyclin D1 displayed a reduced stability in the absence of p27 (Cheng et al. 1999).
Nuclear Localization of Cyclin D1 Is Impaired in p27−/− Mammary Glands
The subcellular localization of cyclin D1 in the mammary epithelium was examined by immunohistochemistry (Fig. 6). Cyclin D1 staining was observed in glands from each genotype, but it was markedly reduced in p27−/− mammary glands, consistent with its rapid turnover (see Fig. 5 D). Cyclin D1 staining was observed in both the cytoplasmic compartment and nuclei of virgin p27+/+, with the majority of nuclei being completely devoid of cyclin D1 (Fig. 6). Cyclin D1 was evident in 17.8% of p27+/− nuclei, a 1.8-fold increase in the frequency of cyclin D1 nuclear localization over wild-type mammary epithelial cells (P < 0.045). In contrast, p27−/− epithelium displayed almost no nuclear cyclin D1 (0.0625%; P < 0.001). These results suggest that in the absence of p27, nuclear localization of cyclin D1 is severely impaired.
Complete Absence of p27 Impairs Growth and Differentiation of the Mammary Epithelium during Pregnancy and Lactation
Comparison of mammary glands from pregnant (16.5 dpc) p27+/+ and p27+/− mice revealed no morphological differences (Fig. 7A and Fig. B). Lobuloalveolar development appeared to occur normally in p27+/− glands (Fig. 7E and Fig. F). Because p27−/− females are infertile, the effect of p27 deficiency on the mammary epithelium during pregnancy was investigated in wild-type mammary glands reconstituted with either p27−/− or wild-type tissue (Fig. 7C and Fig. D). Contralateral control mammary glands reconstituted with wild-type epithelium appeared histologically indistinguishable from intact wild-type glands at 16.5 dpc (Fig. 7C and Fig. G compared with A and E). However, severely impaired growth of the p27-deficient epithelium was apparent (Fig. 7D and Fig. H), whereas the native mammary gland from the same animal displayed the morphological changes associated with pregnancy (data not shown). This suggests that p27 is required in the mammary epithelium for proper morphogenesis during pregnancy.
No morphological differences were observed between p27+/+ and p27+/− lactating mammary glands (Fig. 8A, Fig. B, Fig. E, and Fig. F). Indeed, p27+/− lactating alveoli were filled with lipid-containing secretions, and p27+/− mothers were able to successfully nurse litters from sequential pregnancies, suggesting that p27+/− mammary glands are able to differentiate (Fig. 8B and Fig. F). Examination of mammary glands reconstituted with p27+/+ epithelium showed histological evidence of lactation, including distended alveoli filled with lipid-containing secretions and a marked increase in the ratio of epithelial to stromal cells (Fig. 8C and Fig. G). In contrast, mammary glands reconstituted with p27−/− tissue (n = 12) showed minimal increase in epithelial tissue with an abundance of visible stroma throughout the gland (Fig. 8D and Fig. H). Although p27−/− ducts became distended with eosinophilic secretions, distention of the alveoli was never observed, suggesting an inability to form lobuloalveolar structures and achieve functional differentiation.
To visualize the development and differentiation of the p27−/− epithelium in the context of its native mesenchyme, whole organ culture of mammary glands was used (Fig. 8I and Fig. J). Glands were cultured for 1–10 d in a chemically defined medium containing estrogen, progesterone, prolactin, EGF, and insulin, a combination of factors that have been used to mimic pregnancy and induce differentiation of whole mammary glands in culture. Glands harvested from 4-wk-old wild-type or p27-deficient mice and cultured for 1 d both displayed the characteristic phenotypes of their respective genotypes (Fig. 8 I). Note the thin ducts of the p27−/− glands and the apparent lack of TEBs. After 10 d in culture, wild-type mammary glands responded to the lactogenic hormones by producing abundant lobuloalveolar clusters with an increase in the total epithelial content of the cultured gland (Fig. 8 J). In contrast, p27−/− glands did not demonstrate lobuloalveolar differentiation, nor did they display the increase in epithelia observed in the wild-type organ cultures. These results confirm the in vivo data obtained with mammary gland transplants and extend this observation to include the development of p27−/− epithelium in the context of its native mesenchyme.
To further determine the extent to which p27−/− epithelium was able to differentiate on a molecular level, the expression of milk proteins was examined. A low level of β-casein was detected in lactating glands reconstituted with p27−/− epithelium at 16.5 dpc, whereas it was strongly expressed in lactating wild-type glands and glands reconstituted with wild-type mammary tissue (Fig. 8 K, left). The expression of α-lactalbumin was not detected in glands reconstituted with p27−/− epithelium. This decreased expression of lactational proteins was not due to a decrease in p27−/− epithelial content relative to wild-type glands, since data were normalized to the content of keratin-14, a marker of epithelial cells. The expression and phosphorylation status of Stat5a were also examined, since this transcription factor is known to be required to initiate the transcription of several milk protein genes (Henninghausen et al. 1997). Although the levels of Stat5a were equal regardless of the genotype examined, it was observed that phosphorylation of Stat5a was decreased at 16.5 dpc in wild-type mammary fat pads reconstituted with p27−/− epithelium but not in fat pads reconstituted with wild-type epithelium (Fig. 8 K, right). These results further support that p27−/− mammary epithelial cells are unable to differentiate in response to pregnancy and lactation.
Loss of One p27 Allele Decreases Apoptosis and Delays Involution of the Mammary Gland
Involution of p27+/− and p27+/+ mammary epithelium was examined at 1, 3, 5, and 21 dpfw (Fig. 9, A–H). By 3 dpfw, delayed involution was apparent in p27+/− glands with an excess of epithelium and a reduced level of visible adipocytes compared with p27+/+ glands (Fig. 9B compared with F). At 21 dpfw, the wild-type mammary gland resembled the relatively quiescent ductal network of a virgin gland, whereas abundant epithelium remained in p27+/− mammary glands, including persistent lobuloalveolar structures and ducts distended with lipid-containing secretions (Fig. 9D compared with H). These changes persisted as long as 40 dpfw (data not shown).
TUNEL analysis was performed to detect cells undergoing apoptosis during involution (Fig. 9, I–L). Approximately 64% of the total number of p27+/+ epithelial nuclei stained TUNEL positive at 3 dpfw (Fig. 9I and Fig. M), consistent with previous reports (Strange et al. 1992). However, only 29.9% of p27+/− nuclei were TUNEL positive at this stage (Fig. 9K and Fig. M; P < 0.007). The peak of apoptosis in wild-type involuting mammary glands occurred at 3 dpfw. In contrast, apoptosis in p27+/− mammary glands did not peak until 5 dpfw (43.6%), after which the rate of apoptosis declined. These results suggest that p27+/− glands undergo reduced and delayed apoptosis, resulting in a persistence of breast epithelium after cessation of lactation. Involution in p27−/− mammary epithelium is not reported herein, since differentiation and growth did not occur in a manner that would make results interpretable.
We report the first in vivo evidence that the absence of p27 within the mammary gland impairs cell cycle progression, whereas loss of only a single p27 allele has an opposite effect, an acceleration of G1 to S progression. In this manner, stoichiometric levels of p27, ranging from no p27 (by homozygous-targeted disruption of the p27 gene) to abnormally high levels of p27 (by gene overexpression or increased protein stability) may determine the growth potential of the mammary epithelium. We demonstrate that haploid gene inactivation of p27 causes a decrease in p27 levels within the mammary epithelium, correlating with increased proliferation, increased lateral branching, elevated lobuloalveolar development, and delayed involution. These observations are consistent with previous studies demonstrating that antisense-mediated inactivation of p27 results in increased cell proliferation (Coates et al. 1996) and the low levels of p27 protein reported in several human tumors (Clurman and Porter 1998), further supporting its role as a tumor suppressor.
However, intriguingly p27−/− mammary glands showed decreased proliferation. This defect appears to be intrinsic to the epithelium, since p27−/− epithelium transplanted into wild-type females also exhibited decreased proliferation. This confirms that the hormonal environment of p27−/− females is not the ultimate determinant of this hypoproliferative phenotype. Epithelial–stromal interactions are known to influence mammary gland morphogenesis. Since a small amount of p27−/− stromal tissue is transplanted along with the p27−/− epithelium, one concern was that the p27−/− stroma may have caused the retarded proliferation of the mammary epithelium. This possibility was addressed by purifying primary mammary epithelial cells from p27−/− mice (Fig. 3). Under these ex vivo conditions, p27−/− cells recapitulated the decreased branching and proliferation observed in vivo. Our results confirm that p27 is required in the mammary epithelium for proliferation and morphogenesis to proceed at a normal rate.
Because proliferation occurs in p27−/− glands, albeit delayed, it is likely that there are other mechanisms in place that incompletely circumvent the loss of p27 with the subsequent loss of cyclin D1–Cdk4 activity. Many signals are responsible for regulating proliferation within the mammary epithelium, including estrogen, growth hormone, prolactin, progesterone, adrenal steroids, and various growth factors (Medina 1996). These pathways may be able to partially compensate for the loss of p27 and the loss of cyclin D1–Cdk4 activity. Interestingly, previous reports suggest a possible role for cyclin D1 in directly binding ER and potentiating ER-mediated transactivation of ER-responsive genes (Zwijsen et al. 1997). The ability of cyclin D1 to interact with ER is independent of its interaction with Cdks and therefore may not require the presence of p27. Other studies have shown that cyclin E–cdk2 complexes have additional functions, downstream of the Rb pathway, that drive the cell cycle into S phase (Lukas et al. 1997). It is possible that this or another signaling pathway may allow a low level of proliferation to occur, independent of the level of Rb phosphorylation. Indeed, the level of Rb phosphorylation is reduced in p27−/− mammary glands, consistent with the delay in G1 progression, but Rb phosphorylation is not completely eliminated by the loss of p27 (Fig. 5). It is possible that low levels of Rb phosphorylation by cyclin E–Cdk2 are sufficient to initiate a reduced rate of G1 progression. Alternatively, other cyclin D isoforms (such as cyclin D2 and cyclin D3) may be able to interact with Cdks to initiate Rb phosphorylation in p27−/− mammary glands. Many cell types express combinations of D-type cyclins, such as D1–D2 or D2–D3 (Tam et al. 1994). However, we did not detect cyclin D2 in virgin glands from all three mouse genotypes (data not shown). This is consistent with previous studies demonstrating that cyclin D1 is the predominant D-type cyclin within the mammary gland. During periods of increased proliferation such as pregnancy, cyclin D2 and cyclin D3 mRNAs were not detected by in situ hybridization in the mouse mammary gland, whereas levels of cyclin D1 mRNA were dramatically elevated (Sicinski and Wienberg, 1997). This observation highlights the dependence of the mammary epithelium on cyclin D1 to undergo rapid proliferation and suggests that Rb phosphorylation by cyclin D2– or D3–containing complexes in the mammary gland is unlikely. However, these possibilities would require further investigation.
The effects of p27 deficiency became most pronounced during pregnancy. Cyclin D1–deficient mammary glands also suffer from proliferative defects that are most pronounced during pregnancy (Sicinski and Weinberg 1997). Like cyclin D1 deficiency, loss of p27 results in impaired lobuloalveolar development and production of lactational proteins, suggesting that p27−/− mammary glands would be unable to support lactation. Strikingly similar defects were observed in mice that are deficient for the genes encoding the prolactin receptor or STAT5a (Henninghausen et al. 1997; Ormandy et al. 1997). It is interesting to speculate that the prolactin receptor, which operates through the JAK/STAT signaling pathway, may ultimately increase expression of cyclin D1 to initiate rapid proliferation and differentiation of the mammary epithelium during pregnancy. If cyclin D1, prolactin receptor, and Stat5a are involved in this linear pathway that controls growth and differentiation, then dependence of cyclin D1 on p27 would explain the similarity of the p27−/− mammary gland to those that are devoid of either cyclin D1, prolactin receptor, or STAT5a.
Although we present evidence that the mammary glands of p27-deficient mice are severely hypoplastic, many other organs of p27-deficient mice are hyperplastic, containing an increased number of cells. These include liver, retina, gonads, and pituitary (Fero et al. 1996; Kiyokawa et al. 1996; Nakayama et al. 1996). This observation suggests that different organs respond to the loss of p27 in different manners. This speculation is supported by a recent report of p27+/− and p27−/− mice exhibiting increased susceptibility to chemical- or radiation-induced tumor formation in lungs, stomach, and colon, whereas mammary tumors were not reported in p27−/− mice (Fero et al. 1998). The observations presented herein that p27 deficiency in the mammary epithelium results in the same phenotype as the cyclin D1–deficient mammary epithelium is in direct contrast to a recent report in which p27 deficiency can actually correct the retinal hypoplasia of cyclin D1–deficient animals (Tong and Pollard 2001). The correction of the cyl−/− retinal phenotype by loss of p27 suggests that p27 may function antagonistically to cyclin D1–cdk activity within the retina. These data taken with our observations presented herein demonstrate the tissue specificity of p27 function and confirm that p27 can indeed perform dual roles. Mammary hypoplasia due to p27 deficiency may be related to the unique dependence of the mammary epithelium on cyclin D1 to initiate rapid proliferation, such as during puberty and pregnancy. Since p27 is required for cyclin D1–Cdk4 complex formation and function, the mammary gland may also have a specific requirement for p27, whereas other epithelia may be able to compensate for the loss of p27 and/or the loss of cyclin D1–Cdk4 function.
The inability of p27−/− mammary epithelial cells to form lobuloalveolar structures and lactate (Fig. 8 and Fig. 9) suggests that p27 plays a direct role in differentiation of the mammary gland. There is precedent evidence regarding the role of p27 in differentiation; for example, the complete loss of p27 prevents terminal differentiation of the ovarian corpus luteum, oligodendrocytes, and skin keratinocytes (Robker and Richards 1998; Tong et al. 1998). On the other hand, this impaired differentiation may be secondary to the proliferative defect due to loss of p27. For example, a critical level of mammary epithelial cells may be required to allow terminal differentiation to proceed, or perhaps any given mammary cell must divide a minimum number of times before the onset of differentiation. These speculative hypotheses require further investigation.
The data presented herein support the notion that p27 levels regulate proliferation within the mammary epithelium to the extent that a reduction of p27 produces the opposite effect on proliferation compared with the complete loss of p27. This apparently paradoxical observation is perhaps expected given the recently identified role of p27 in cyclin D1–Cdk4 stabilization and function (Cheng et al. 1999). These mechanisms by which p27 regulates proliferation, differentiation, and function of the mammary epithelium are proposed as follows. p27 protein encoded by two functional p27 alleles (p27+/+) allows for stabilization of cyclin D1–Cdk4 complexes with abundant p27 remaining for association with and inhibition of cyclin E–Cdk2 complexes for the normal control of G1 progression. Heterozygous inactivation of a single p27 allele (p27+/−) reduces the level of p27. Upon mitogenic signaling, cyclin D1–Cdk4 complexes sequester p27, resulting in a relatively greater reduction of free p27 in p27+/− than in p27+/+ cells. Under these conditions, cyclin E–Cdk2 complexes bind in stoichiometric excess with the reduced levels of free p27, a situation in which Cdk2 can phosphorylate p27 and target it for ubiquitin-mediated degradation (Pagano et al. 1995; Sheaff et al. 1997). In this manner, a reduction of p27 (as it occurs in p27+/− cells) is permissive for cyclin E–Cdk2 function, Rb phosphorylation, and an accelerated G1 to S transition. The delayed postlactational involution noticed in p27+/− glands (and its potential implications for epithelial transformation) remain unexplained. However, the excessive proliferation may indirectly abrogate this process if in fact exit from the cell cycle is required for the epithelium to undergo postlactational apoptosis. In the absence of p27 (p27−/−), stable association of cyclin D1 and Cdk4 does not occur, resulting in reduced Rb phosphorylation and cell cycle arrest in early G1. Whether this delayed proliferation indirectly accounts for the impaired lactational differentiation of these glands will require further research. Nonetheless, a direct requirement of both p27 alleles for the formation of TEBs and lactation cannot be ruled out at this time. The data presented herein support a model for the role of p27 in both activation and inhibition of cell cycle progression and demonstrate a critical role for p27 in regulation of mammary epithelial proliferation, morphogenesis, and function.
The authors would like to thank Drs. Jennifer Pietenpol for critical review of this manuscript, Andrew Koff for providing a p27+/− breeding pair of mice, and Robert Whitehead for valuable suggestions, and Ms. Sandy Olsen for technical support.
This work was supported in part by National Institutes of Health training grant T32 CA09592 (to R.S. Muraoka), National Institutes of Health grant R01 CA80195 (to C.L. Arteaga), a Department of Veteran Affairs Clinical Investigator Award (to C.L. Arteaga), Vanderbilt-Ingram Cancer Center support grant CA68485, and a postdoctoral research fellowship award from the Susan G. Komen Foundation (to A.E.G. Lenferink).
A.E.G. Lenferink's present address is Biotechnology Research Institute, National Research Council, Montreal, PQ, Canada H4P 2R2.
Abbreviations used in this paper: Cdk, cyclin-dependent kinase; CKI, Cdk inhibitor; dpc, days postcoitum; dpfw, days postforced wean; HH1, histone H1; IP, immunoprecipitation; PMEC, primary mammary epithelial cell; Stat, signal transducer and activator of transcription; TEB, terminal end bud; TUNEL, terminal deoxynucleotidyl transferase–mediated dUTP–biotin nick end labeling.