Overexpression of the Low Molecular Weight Cyclin E in Transgenic Mice Induces Metastatic Mammary Carcinomas through the Disruption of the Arf-p53 Pathway

In tumor cells, cyclin E deregulation results in the appearance of five low molecular weight (LMW) isoforms. When overex-pressed in breast cancer cells, these forms of cyclin E induce genomic instability, resistance to inhibition by p21 and p27, and resistance to antiestrogen therapy. Additionally, the LMW forms of cyclin E strongly correlate with decreased survival in patients with breast cancer. However, the oncologic role of the LMW forms of cyclin E in breast cancer tumorigenesis is yet to be determined. To this end, we generated transgenic mice expressing full-length cyclin E alone (M46A), full-length and the EL4 isoforms (EL1/EL4), or the EL2/3 isoforms of cyclin E (T1) under the control of the mouse mammary tumor virus promoter. Compared with full-length cyclin E, LMW cyclin E overexpression induces delayed mammary growth during the pubertal phase and abnormal cell morphology during lactation. Both primary mammary tumor formation and metastasis were markedly enhanced in LMW cyclin E transgenic mice. LMW cyclin E overexpression in mammary epithelial cells of mice is sufficient by itself to induce mam-mary adenocarcinomas in 34 of 124 (27%) animals compared with 7 of 67 (10.4%) mice expressing only the full-length cyclin E (P < 0.05). In addition, metastasis was seen in 25% of LMW cyclin E tumor–bearing animals compared with only 8.3% of tumors in the full-length cyclin E background (P < 0.05). Moreover, LMW cyclin E overexpression selects for inactiva-tion of p53 by loss of heterozygosity and spontaneous and frequent inactivation of ARF. Therefore, LMW cyclin E over-expression strongly selects for spontaneous inactivation of the ARF-p53 pathway in vivo, canceling its protective checkpoint function and accelerating progression to malignancy.


Introduction
Cyclin E is a G 1 cyclin necessary for the transition from G 1 to S phase of the normal cell cycle (1).Deregulation of cyclin E accelerates the entry of the cells into S phase but causes inefficient progression through S phase.The untimely expression of cyclin E has been shown to interfere with the replication complex assembly as cells exit mitosis (2).An oncogenic role for cyclin E has been suggested by studies of cyclin E-deficient cells which are resistant to transformation by myc alone or myc in combination with ras, a dominant negative p53, or E1A, suggesting that cyclin E is a key component in oncogenic signaling (3).Constitutive overexpression of cyclin E protein at all phases of the cell cycle is one of the features observed in breast cancer cell cycle and thought to result in premature DNA replication, genomic instability (4,5), and carcinogenesis (6).
The cyclin E gene has been found to be amplified and the cyclin E protein constitutively expressed in several breast cancer cell lines (7,8).In addition to overexpressing the full-length 50-kDa cyclin E protein, some of these cell lines overexpress up to five low molecular weight (LMW) isoforms (8,9).Our laboratory has shown that tumor-specific processing of full-length cyclin E by an elastaselike protease generates two sets of doublets [EL2/EL3 (Trunk 1) and EL5/EL6 (Trunk 2)] that represent hyperactive LMW cyclin E isoforms (10).In addition, the EL4 band at 40 kDa represents an alternative translation site at Met 46 .These LMW isoforms are unique to tumor cells, suggesting that they may play an important role in tumorigenesis.
Clinical studies have indicated that cyclin E overexpression occurs in 25% of breast cancer tumors and is linked to poor prognosis (11).To determine the clinical significance of the LMW isoforms, we measured cyclin E expression in 395 women with breast cancer.Full-length and LMW cyclin E levels were evaluated by Western blot analysis.Full-length cyclin E, LMW cyclin E, and total ( full-length + LMW) cyclin E levels were then compared on univariate analysis with standard clinical factors to include patient age, tumor size, nodal status, and stage of disease, as well as biological markers including estrogen and progesterone receptor status, HER-2/neu status, ploidy, proliferation index, cyclin D1, and cyclin D3.Although several clinical, histologic, and molecular markers were significantly associated with outcome on univariate analysis, the majority lost significance when multivariate analysis was done.High total cyclin E levels or high levels of the LMW isoforms, however, retained significance.In fact, high levels of total or LMW cyclin E were found to be the most powerful discriminants of disease-free and overall survival, outperforming criteria currently used clinically, including positive nodal status, late stage (stage III-IV) disease status, and negative estrogen receptor status.Importantly, in this study, cyclin E was a better prognostic marker than nodal status, a finding that held even for nodenegative, stage I patients (11).These findings suggest that there may be utility for the determination of level of cyclin E expression in breast tumor specimens to better counsel patients about their prognosis.
In a breast tumor cell model, we have already shown that overexpression of the LMW forms of cyclin E, but not the full-length form, results in their hyperactivity due to increased affinity for cyclin-dependent kinase (Cdk)-2 and resistance to inhibition by the Cdk inhibitors p21 and p27 (5,12).LMW forms of cyclin E are resistant to the growth-inhibiting effects of antiestrogens, and they induce chromosomal instability (5).Collectively, the biochemical and functional differences between the full-length and the LMW isoforms of cyclin E provide a molecular mechanism for the poor clinical outcome observed in breast cancer patients harboring tumors expressing high levels of the LMW forms of cyclin E. Nonetheless, it remains uncertain whether generation of LMW cyclin E represents a necessary event for breast tumor initiation or if constitutive overexpression of full-length cyclin E protein is sufficient for oncogenic transformation of the mammary epithelium as has been suggested (13).
We hypothesize that it is the unique properties of LMW cyclin E rather than constitutive expression of the cyclin E protein that is responsible for cyclin E driven mammary oncogenesis.To address this hypothesis, we generated several different transgenic mice expressing either the full-length cyclin E alone (cyclin E-M46A), full-length (EL1) and the EL4 isoform (cyclin EL1/EL4), or the EL2 and EL3 isoforms (cyclin E-Trunk 1) under the control of the mouse mammary tumor virus (MMTV) promoter.

Materials and Methods
Generation and analysis of transgenic mice.Three constructs were generated for the development of transgenic mice (Fig. 1A).These include (a) the MMTV-cyclin EL1/EL4, which, when expressed in mammary glands of transgenic mice, generates both the full-length form, EL1, and the EL4 isoform; (b) MMTV-cyclin E-M46A, which harbors a mutation at codon 46 replacing methionine with alanine thereby generating a construct that codes only for the full-length form; and (c) MMTV-cyclin E-T1 (Trunk 1), which codes for the LMW isoforms of cyclin E, EL2, and EL3.The EcoRI fragments containing each of the cyclin E sequences (EL1/EL4, M46A, and Trunk 1) were cloned into the EcoRI site of the pBS-MMTV-pA plasmid, which consists of the Bluescript backbone (Stratagene), the MMTV long terminal repeat upstream of the Hras1 leader sequence, and a multiple cloning site upstream of the SV40 splice site and polyadenylation signal (14).Transgenic mice were generated by pronuclear microinjection of each of the three 6.0-to 6.2-kb NotI-XhoI doubly digested MMTV-cyclin E-pA fragments into fertilized oocytes collected from superovulated FVB mice as described (15).The Southern blot of tail genomic DNA digested with EcoRI was used to identify founder animals.
p53-deficient mice generated by gene targeting (16) were obtained from Dr P. Leder (Harvard Medical School, Boston, MA) who backcrossed them on a FVB background for more than seven generations.Mice were screened for their cyclin E and p53 status by PCR.Briefly, a small piece of tail was cut from each animal at the time of weaning and used to isolate genomic DNA by standard procedures.Primers used for the detection of the 422-bp PCR product from the MMTV-cyclin E transgene were 5 ¶-CCACAGAGCGGTAAGAAGCA-3 ¶ (5 ¶ sense primer) and 5 ¶-CCCATTCAT-CAGTTCCATAG-3 ¶ (3 ¶ antisense primer).A pair of primers (sense primer, 5 ¶-GATGTGCTCCAGGCTAAAGTT-3 ¶; antisense primer, 5 ¶-AGAAACG-GAATGTTGTGGAGT-3 ¶) amplifying a 525-bp PCR product from the mouse b-casein gene was used as an internal control in each reaction.For evaluating the p53 status of offsprings, PCR using three primers allowed for detection of both the normal (330 bp) and mutant p53 alleles (250 bp) in a single reaction.These primers consisted of a sense 5 ¶ primer specific for exon 1 (5 ¶-AGTTCCCCACCTTGACACCTGA-3 ¶) and two antisense primers specific for exon 2 of p53 (5 ¶-AGAGCAAGAATAAGTCAGAAGC-3 ¶) and the neo cassette (5 ¶-GGTATCGCCGCTCCCGATTCGCAG-3 ¶).All mice were provided routine care in accordance with institutional guidelines and monitored for the development of mammary tumors.Mice were sacrificed on tumor development or aged for up to 24 months-ill and distressed mice were euthanized by asphyxiation using carbon dioxide following the guidelines of the Institutional Animal Care and Use Committee.Necropsies were done and soft tissue organs were fixed in 10% neutral buffered formalin.Fixed tissues were embedded in paraffin and sectioned at 4 Am, routinely stained with H&E, and microscopically examined by a pathologist.
Histology of the mammary gland, 5-bromo-2 ¶-deoxyuridine incorporation, and terminal deoxyribonucleotidyl transferase-mediated dUTP nick end labeling assays.For histologic analysis, 6-Am sections were cut and stained with H&E.To assess cell proliferation, mice were injected i.p. with 0.25 mg 5-bromo-2 ¶-deoxyuridine (BrdUrd)/g of body weight 2 h before sacrifice.BrdUrd incorporation was detected on sections by immunohistochemistry using a cell proliferation kit (Amersham) following the manufacturer's instructions.The numbers of BrdUrdpositive cells in wild-type and MMTV-cyclin E mammary glands were counted in 10 fields under a 40Â objective lens.To detect apoptotic nuclei, formalin-fixed paraffin-embedded sections were analyzed by TdT digoxygenin nick-end labeling with ApopTag Plus Fluorescein in situ apoptosis detection kit (Chemicon International, Inc.) following the manufacturer's instructions.The numbers of terminal deoxyribonucleotidyl transferasemediated dUTP nick end labeling (TUNEL)-positive cells in wild-type and MMTV-cyclin E mammary glands were counted in 10 fields under a 40Â objective lens.
Immunohistochemistry. For immunohistochemistry, the sections were incubated in 1% H 2 O 2 to block endogenous peroxidase activity.To retrieve nuclear antigens on paraffin-embedded sections, slides were incubated for 20 min in 10 mmol/L sodium citrate buffer (pH 6.0) at 90jC.The sections were then incubated for 60 min in 5% FCS overnight with primary antibodies, followed by 2-h incubation at room temperature with appropriate secondary antibodies.Nuclei were counterstained with hematoxylin.Rabbit polyclonal anti-cyclin E (Santa Cruz Biotechnology) was used for the transgenic cyclin E. For detection, the Vectastain ABC Elite kit (Vector Labs) was used.The numbers of positively stained cells in wildtype and MMTV-cyclin E mammary glands were counted in 10 fields per slide under a 40Â objective lens (three slides per mouse and three mice per genotype).
Western blot analysis.Cell lysates were prepared and subjected to Western blot analysis as previously described (17).Briefly, 50 Ag of protein were subjected to SDS-PAGE and transferred to Immobilon P overnight at 4jC at a constant voltage of 35 mV.The blots were blocked overnight at 4jC in BLOTTO (5% nonfat dried milk in 20 mmol/L Tris, 137 mmol/L NaCl, 0.05% Tween, pH 7.6).After being washed, the blots were incubated in primary antibodies for 3 h.Primary antibodies used were cyclin E (HE-12, Santa Cruz Biotechnology), Cdk4 (C-22, Santa Cruz Biotechnology), proliferating cell nuclear antigen (PCNA; PC10, Santa Cruz Biotechnology), cyclin D1 (A-12; Santa Cruz Biotechnology), p27 (K25020, BD Biosciences-Transduction Laboratories), Cdk2 (Transduction Laboratories), p19 ARF (Abcam, Inc.), and actin (Chemicon International).For rabbit primary antibodies, blots were incubated with goat anti-rabbit immunoglobulinhorseradish peroxidase conjugate at a dilution of 1:5,000 in BLOTTO for 1 h and finally washed and developed by using the Renaissance chemiluminescence system as directed by the manufacturer (Perkin-Elmer Life Sciences, Inc.).For mouse primary antibodies, we used a secondary antibody from eBioscience (called Mouse TrueBlot) that detects only immunoglobulin G (IgG) in its native confirmation and not the denatured form.DTT was added to a final concentration of 50 mmol/L in the sample buffer immediately before use.Blots were incubated with Mouse IgG TrueBlot at a 1:1,000 dilution in BLOTTO for 1 h at room temperature, washed, and developed as described above.
Examining p53 loss in MMTV-cyclin E/p53 +/À tumors.Genomic DNA was isolated from tumor tissue segments for Southern blot analysis.The tumors segments were homogenized and digested with proteinase K (0.5 mg/mL; Roche Aplied Science) in 700 AL of digestion buffer at 55jC overnight.After RNase treatment and phenol-chloroform extraction, the nucleic acids were precipitated with ethanol.Fifteen micrograms of DNA were digested with BamHI, electrophoresed in 1% agarose, and transferred onto a Nytran filter.A 605-bp 32 P-labeled KpnI fragment of plasmid LR10 was used to identify the 6.5-kb mutant and 5.0-kb normal p53 alleles.Semiquantitative analysis of band intensity ratios was done with IpQuant Image analysis software.
Mutational analysis of the p53 gene in mammary tumors.PCR analysis followed by direct sequencing was carried out to screen for mutations of exons 2 to 11 of the p53 gene.Primers were located in the

Results
Generation of mice expressing full-length and LMW cyclin E in mammary cells.To assess the oncogenic role of cyclin E-LMW as compared with full-length cyclin E, we examined the consequences of overexpressing these isoforms in the mammary glands of transgenic mice using the MMTV promoter.Three constructs were generated for the development of transgenic mice (Fig. 1A).Six founder mice for cyclin EL1/EL4, two founder mice for cyclin E-M46A, and two founder mice for cyclin E-T1 were generated and confirmed to express the transgene.Because several founders expressing the same level of cyclin EL1/EL4 protein were found, we chose three founders (lines 668, 552, and 678) as representative lines for further analysis.Southern blotting, by comparison with known amounts of the MMTV-cyclin E vector, estimated copy number to be 2 to 8 at a single integration site in the seven established lines (Fig. 1A).Expression of the transgene in the mammary glands in all the seven lines was confirmed by Western blot assay (Fig. 1B).Relatively similar expression of the total cyclin E protein was seen in each independent line of cyclin E transgenic mice and these levels were substantially higher than wild-type, nontransgenic control animals.In the EL1/EL4 strains, EL4 was expressed at a 5-to 6-fold higher levels than the EL1 isoform, suggesting that translation at Met 46 is much more efficient than translation at the first methionine (Fig. 1B).Cyclin E Figure 1.Generation of cyclin E transgenic mice.A, scheme of the MMTV constructs used to generate the transgenic mice.The vectors used code for the cyclin E isoforms found in tumor cells.The three cyclin E constructs were engineered with a 3 ¶ FLAG sequence at the COOH-terminal end.The copy number for each of the seven lines is indicated in the table below the scheme.B, Western blot analysis of mammary glands at lactation day 10 from cyclin E-overexpressing transgenic mice.The wild-type mammary gland, shown here as negative control, was taken from a nontransgenic (Non-TG) 3-mo-old primiparous female on lactation day 10 to allow for accurate comparison.Protein extracts (50 Ag/lane) were boiled in the sample buffer in the presence of 0.1 mol/L DTT.Filters were probed with mouse monoclonal anti-cyclin E (HE-12, Santa Cruz Biotechnology), anti-Cdk2 (Transduction Laboratory), anti-p27 (Transduction Laboratory), anti-PCNA (Santa Cruz Biotechnology), and rabbit polyclonal antibody anti-Cdk4 (Santa Cruz Biotechnology).For the monoclonal antibodies, we used Mouse TrueBlot as secondary antibody (eBioscience), a horseradish peroxidase-conjugated antimouse IgG that preferentially detects the nonreduced form of mouse IgG, eliminating interference by the heavy and light chains.The filters were reprobed with anti-actin to control for loading.The levels of proteins were measured by densitometric scanning of the corresponding bands and normalized using actin values.The normalized values are indicated at the bottom.C, mammary-specific LMW cyclin E overexpression induces early morphologic alterations.Histologic comparisons of nontransgenic (a and b) and transgenic MMTV-cyclin EL1/EL4 (c and d), MMTV-cyclin E-M46A [line 1194 (e and f ) and line 1202 (g and h )], and MMTV-cyclin E-T1 [line 571 (I and j) and line 573 (k and l)] lactating mammary glands reveal that cyclin EL1/EL4 expression and, more dramatically, cyclin E-T1 expression result in anisocytoses and apoptosis.H&E staining shows that acini of the normal (a) and the transgenic cyclin E-M46A (e-g ) lactating glands are composed of a single layer of secretory epithelial cells with milk-filled lumen.Acini of cyclin EL1/EL4 and cyclin E-T1 are atypical.Many cells have large pleomorphic nuclei (d, j, and l; black arrows ) whereas others have some nuclei which are markedly smaller (triangles ) indicating anisonucleosis.Some cells are binucleated (d, j, and l ; open arrows ) and the increased presence of apoptotic bodies in the lumen and lining of the alveoli suggests increased apoptosis.All samples are from primiparous females on lactation day 10.Representative sections of nontransgenic (b), cyclin EL1/EL4 (d), cyclin E-M46A (f and h), and cyclin E-T1 (j and l) stained immunohistochemically for cyclin E using an antibody specific for human cyclin E (Santa Cruz Biotechnology).Magnification, Â400.Bar, 100 Am.Insets, arrows, cyclin E-positive alveolar cells (alv ) and cyclin E-negative myoepithelial cells (myo ).D, expression of LMW cyclin E in mammary cells leads to increased percentage of cells in S phase and increased percentage of apoptotic cells during pregnancy.a, BrdUrd incorporation into the nuclei of mammary epithelial cells at day 16.5 of pregnancy in nontransgenic, M46A, cyclin EL1/EL4, and cyclin E-T1 transgenic mice.b, quantification of BrdUrd incorporation in the nuclei of mammary epithelial cells at day 16.5 of pregnancy in nontransgenic, M46A, cyclin EL1/EL4, and cyclin E-T1 transgenic mice.For each mouse, 1,000 cells per section were enumerated and three mice per line were examined.The S-phase fraction was calculated by dividing the number of BrdUrdlabeled cells by the total number of cells.c, TUNEL analysis of nontransgenic, M46A, cyclin EL1/EL4, and MMTV-cyclin E-T1 transgenic at day 16.5 of pregnancy.d, quantification of TUNEL-positive cells at day 16.5 of pregnancy in nontransgenic, M46A, cyclin EL1/EL4, and cyclin E-T1 transgenic mice.Twenty Â200 magnification fields of view were randomly counted and the apoptotic index was calculated by dividing the number of TUNEL-positive (FITC-positive) cells by the total number of nuclei (4 ¶,6-diamidino-2-phenylindole stained).
expression and nuclear localization were also confirmed by immunohistochemical analysis of paraffin-embedded sections of the mammary glands obtained from the same glands as those used for Western blot analysis (Fig. 1C-d, f, h, j, l), whereas it was not detected in the mammary glands from age-matched nontransgenic animals (Fig. 1C-b).Additionally, overexpression of cyclin E was seen both in ductal and lobular cells of the mammary gland but not in the myoepithelial cells (Fig. 1C, insets), showing that the cyclin E transgene is being expressed in the same types of cell in the mammary glands of all the lines generated.
The expression levels of endogenous Cdk2, Cdk4, p27, and PCNA were also examined in each of the strains to examine if overexpression of cyclin E had an effect on the endogenous key cell cycle regulators.Although Cdk2, Cdk4, and PCNA showed no correlation with the expression of the cyclin E transgene, p27, a negative regulator of the cell cycle, accumulated in all lines expressing the LMW forms of cyclin E (3-to 5.5-fold over wildtype).The lack of correlation between cyclin E overexpression and PCNA levels indicates that cyclin E expression and proliferation rates are independent of each other.These results are in agreement with our breast cancer prognostic data (11) indicating that the correlation of high cyclin E to decreased survival is much more striking than that of rapid proliferation.
Overexpression of LMW cyclin E results in abnormal development of the mammary epithelium.Microscopic examination of H&E-stained (Fig. 1C-a, c, e, g, l, k) and immunohistochemically anti-cyclin E-stained sections (Fig. 1C-b, c, f, h, j, l) of mammary glands on lactation day 10 from 2-month-old primiparous wild-type (Fig. 1C-a and b) and cyclin E transgenic mice (Fig. 1C-c, d, e, f, g, h, l, j, k, l) shows an alteration in cellular morphology, which is most pronounced in those lines that overexpress the LMW forms of cyclin E. The transgenic lines have enlarged alveolar cells containing enlarged, variable-sized nuclei (anisonucleosis) and multiple nuclei in the same cell (binucleate and trinucleate cells).Binucleate alveolar cells were also observed in nontransgenic mice.However, the numbers of binucleate and trinucleate cells were markedly increased in the T1-573 and EL1-678 lines compared with nontransgenic and M46A-1194.This increase in multinucleate cells ranged from 3 to 5 per alveolus compared with 0 to 2. The nuclear size ranges from 6 to 13 Am (10.1 F 0.8 Am) for EL1-678 (Fig. 1C-c and d), 5 to 17 Am (11.5 F 1.5 Am) for T1-571 (Fig. 1C-i and j), 3 to 12 Am (8.6 F 1.1 Am) for T1-573 (Fig. 1C-k and l), and 4 to 6 Am (5.0 F 0.3 Am) for both the nontransgenic wild-type (Fig. 1C-a and b) and the full-length M46A lines (Fig. 1C-e and f ).Collectively, this figure shows by immunohistochemistry (with a human specific antibody that does not cross-react with mouse cyclin E) that the human cyclin E protein is expressed in >90% of the epithelial cells (ductal and lobular) but not in myoepithelial cells in all the cyclin E transgenic lines but not in nontransgenic mice.Second, it shows an alteration in cellular morphology (enlarged alveolar cells containing enlarged, variable-sized nuclei), which is lowest in full-length cyclin E and highest in the EL1/EL4 and T1 cyclin E transgenic mice (P < 0.05, versus M46A and nontransgenic).
We also investigated the effect of cyclin E overexpression on the development of mammary glands by whole-mount preparations and found that the mammary glands of cyclin EL1/EL4 and T1 transgenic mice underwent delayed ductal growth whereas M46A showed a less pronounced phenotype alteration as compared with nontransgenic mice (Supplementary Fig. S1).To more directly assess the contribution of proliferation versus apoptosis in the transgenic phenotype, we examined mammary glands from the T1-573 line, EL1/ EL4 line, M46A, and nontransgenic mice for S-phase fraction by BrdUrd staining (Fig. 1D-a and b) and for apoptosis by the TUNEL assay (Fig. 1D-c and d) at pregnancy day 16.Nontransgenic, M46A, EL1/EL4, and T1 transgenic female mice were pulsed labeled with BrdUrd before they were sacrificed, and the S-phase fractions were calculated as a percentage of BrdUrd-positive cells in the ductal and alveolar regions per the total number of epithelial cells.The apoptotic indices were calculated as a percentage of the TUNELpositive cells in the ductal and alveolar regions per the total number of epithelial cells.The S-phase fraction of the alveolar epithelium of T1 and EL1/EL4 mice at pregnancy day 16.5 was markedly increased (5-fold) relative to the nontransgenic mice (24.9 F 4.1% for T1 and 20.4 F 2.0% for EL1/EL4 versus 4.5 F 1.9%; P < 0.01) whereas it was moderately increased (2.2-fold) in M46A mice (9.8 F 1.9%; P < 0.01).A concomitant increase in the apoptotic index of the alveolar epithelium at pregnancy day 16.5 was also observed in T1 and EL1/EL4 mice (15-fold; 2.9 F 0.2% for T1 and 2.7 F 0.2% for EL1/EL4 versus 0.19 F 0.07%; P < 0.01) and M46A mice (3-fold; 0.57 F 0.08%; P < 0.01) relative to the nontransgenic mice.We also observed an increased percentage of p53 positively stained cells at pregnancy 16.5 in the T1 and EL1/EL4 transgenic lines (3.10 F 0.46% for T1 and 2.70 F 0.35% for EL1/EL4 versus 0.16 F 0.04%) and to a lesser extend in M46A mice (0.41 F 0.06%) when compared with the control nontransgenic mice (data not shown).Based on these results, the delayed ductal growth could be explained by impaired progression through S phase (accumulation of cells in S phase) and increased apoptotic rate due to replicative stress.Consistent with this, we observed increased percentage of cells with p53 activation in LMW cyclin E mammary glands when compared with M46A and control nontransgenic mammary glands.
LMW cyclin E transgenic mice develop metastatic mammary tumors.MMTV-cyclin E lines and control littermates were maintained as breeding colonies and monitored for transformation phenotypes up to 27 months.For this analysis, data from each of the individual transgenic lines derived using the same construct were pooled to give total tumor formation for full-length cyclin E (M46A), cyclin EL1/EL4, and cyclin E-T1.
Malignant mammary gland tumors were observed in the wildtype and all transgenic lines as listed in Table 1A.Only 10.4% (7 of 67) of the mice overexpressing the full-length cyclin E (i.e., M46A) developed mammary tumors at 24 months.This was significantly different (P = 0.016) than the rate of spontaneous mammary tumors at 24 months in wild-type inbred FVB mice (3 of 57, 5.3%).However, transgenic mice expressing LMW cyclin E showed an increase in the rate of mammary tumor formation (25.4% in EL1/EL4 and 30.2% in cyclin E-T1; P < 0.05, compared with M46A) although the average latency to tumor onset was similar among the three cyclin E transgenes (17-18.6 months).Both nulliparous and multiparous transgenic females developed these adenocarcinomas, indicating that pregnancy was not required for neoplastic transformation.
In addition to the increased frequency, tumors in transgenic mice expressing the LMW forms of cyclin E showed an increased metastatic potential.No metastatic tumors were seen in any of the nontransgenic mice, and in line M46A, only 1 of 7 (14.3%)mammary tumor-bearing mice developed lung metastasis (Table 1A).However, in transgenic mice expressing the LMW forms of cyclin E, metastasis were seen in 16.7% and 25% of EL1/EL4 and cyclin E-T1 transgenics, respectively.The predominant site of metastasis was the lung although nodal and cardiac metastases were also observed.
Whereas the cyclin E tumors have a range of tumor morphologies expressed, the predominant type of tumor observed for the LMW forms was glandular adenocarcinomas.Examples of the various morphologies are included in Fig. 2B.The mammary gland tumors were classified according to previously reported consensus reports (18,19) and ranged across a spectrum of mammary gland tumors that have been associated with activation of the Wnt pathway (20).
Analysis of expression of cyclin E in 11 representative primary tumors revealed that 73% (8 of 11) of the mammary tumors retained the transgenic protein expression (Fig. 2C).Immunohistochemistry for cyclin E correlated with the Western blot analysis: in all Western blot positive tumors, 45% to 75% of the tumor epithelial cells were stained with cyclin E by immunohistochemistry (Fig. 2D-b, for example).All the Western blot negative tumors were also negative for cyclin E by immunohistochemistry although >90% of the mammary epithelial cells in the contralateral gland of these transgenic mice expressed cyclin E at time of development of the index tumor (Fig. 2D, compare c-d and e-f ).The observation that not all primary tumors from transgenic mice expressed cyclin E suggests that whereas cyclin E is required for tumor initiation, it may not be required for tumor maintenance.Immunohistochemistry was also negative for estrogen receptor a in 8 of 9 (89%) LMW cyclin E-overexpressing tumors examined (data not shown).
Premalignant mammary lesions in cyclin E transgenic model.In humans, breast cancer is characterized by a multistage development of mammary invasive carcinoma, which is believed to progress from atypical ductal hyperplasia to in situ carcinoma and subsequently to invasive carcinoma.To determine whether the cyclin E transgene drives oncogenesis through similar steps, we examined the contralateral mammary gland of mammary tumorbearing mice for the presence of premalignant lesions.Mammary gland atypical ductal hyperplasia-characterized by one of several alterations that included variation in cell size, cell nuclei, open chromatin, and multiple layers of cells with disorganized architecture-and mammary intraepithelial neoplasia were identified in cyclin E transgenic animals (Fig. 3A and B).The frequencies of premalignant lesions were comparable between the cyclin E transgenic lines (Fig. 3B, table).
Cyclin E-overexpressing mammary tumors do not exhibit genomic instability.We have previously shown that overexpression of LMW cyclin E in MCF-7 cells resulted in genomic instability, as evidenced by the presence of multiple chromosomal aberrations and an aneuploid genome (21).However, using Spectral Karyotyping to examine metaphase spreads derived from three LMW cyclin E-T1 mammary tumors, we found that the acquisition of a very unstable genome is not required to promote mammary tumor progression in the setting of LMW cyclin E-T1 overexpression in vivo (Supplementary Fig. S2; Table 1).
Mammary adenocarcinomas from cyclin E-overexpressing mice sustained ARF loss of function.Because the mechanism of LMW cyclin E-induced mammary tumorigenesis seemed to be independent of enhanced proliferation (Fig. 1) and genomic instability (Supplementary Fig. S2), we next investigated whether cyclin E deregulation could cooperate with the ARF-mdm2-p53 pathway in promoting tumorigenesis.When Minella et al. (22) studied the consequences of deregulated cyclin E expression in primary cells, they found that cyclin E initiates a p53-dependent response that prevents excess Cdk2 activity by inducing expression of the p21Cip1 Cdk inhibitor.Loss of this response may thus be required before deregulated cyclin E can become fully oncogenic in cancer cells.
One way for cancer cells to bypass this barrier would be functional inactivation of p53.We first evaluated the p53 status in LMW cyclin E-induced mammary tumors.We examined 23 mammary tumors for both deletions (Southern blot analysis; Fig. 4) and point mutations (direct sequencing; data not shown) in the entire p53 gene.Southern blot analysis did not detect any gross genomic rearrangements and did not show evidence of loss of one allele of p53 as measured by the band intensity ratios between the wild-type allele and the pseudogene (Fig. 4).Direct sequencing of exons 2 to 11 detected no mutations in any of the 11 mammary tumors analyzed (seven EL1/EL4, four T1; data not shown).These results suggest that the p53 gene is not directly targeted for instability by LMW cyclin E. We next sought to determine whether regulation of p53 function could be affected in these tumors.We examined if Overexpression of the LMW cyclin E in transgenic mice induces metastatic mammary adenocarcinomas/carcinomas.A, Kaplan-Meier curve of mammary tumor-free survival in nontransgenic and each of the MMTV-cyclin E lines.In line MMTV-cyclin E-M46A, 7 of 67 (10.4%) mice developed mammary tumors with an average latency of 17 mo.In line MMTV-cyclin E-EL1-EL4, 18 of 71 (25.4%) mice developed mammary tumors with an average latency of 18.6 mo.In line MMTV-cyclin E-T1, 16 of 53 (30.2%) mice developed mammary tumors with an average latency of 17.2 mo.In nontransgenic mice, 3 of 57 (5.3%) developed mammary tumors with an average latency of 20 mo (P < 0.05, compared with each of the MMTV-cyclin E lines).B, morphologies of mammary tumors overexpressing cyclin E. H&E staining of MMTV-T1 (a and b), MMTV-EL1/EL4 (c, d, and f ), and MMTV-M46A (e) tumor sections shows that terminal-stage adenocarcinomas have varied morphologies: acinar (a; Â400), glandular (b; Â400), papillary (c ; Â400), solid (d ; Â400), adenosquamous (e; Â200), and myoepithelial (f ; Â200).C, protein extracts (50 Ag/lane) from mammary tumors were loaded on a 10% SDS gel and subjected to Western blot analysis and the expression of the human cyclin E transgene or the endogenous cyclin D1, p27, and actin was detected with specific antibodies.D, immunohistochemistry for human cyclin E protein in mammary tumors (b, d , and f ) and the contralateral mammary gland of the same animal (a, c , and e).This staining showed loss of cyclin E expression in mammary tumors from animal #1180 in d (line EL1/EL4-668) and from animal #1067 in f (line T1-571), whereas expression is present in the contralateral glands (c and e, respectively).Arrows, cyclin E expression in luminal cells.The positive tumor is from animal #369 in b (line EL1/EL4-678) and is used as a positive control.
ARF, which controls p53 function by binding to mdm2, is modulated in cyclin E transgenic mice.We examined ARF expression level by Western blot analysis using tumor lysates from cyclin E transgenic mice and saw that loss of ARF protein expression occurs frequently in mammary tumors arising in cyclin Eoverexpressing mice.One hundred percent (3 of 3) of M46A, 89% (8 of 9) of cyclin EL1/EL4, and 73% (8 of 11) of cyclin E-T1overexpressing mammary tumors examined had sustained ARF loss (Fig. 4).Under unstressed conditions (no cyclin E overexpression), ARF protein expression is not detectable as assessed by Western blot analysis on nontransgenic mice mammary gland (Supplementary Fig. S3).
Loss of one allele of p53 not only decreased latency but also increased the incidence of mammary tumor formation in M46A, EL1/EL4, and T1 lines (P < 0.01, M46A versus M46A/p53 +/À ; P < 0.01, EL1/EL4 and EL1/EL4/p53 +/À ; P < 0.01, T1 versus T1/ p53 +/À ).Eleven of the 17 cyclin E-T1/p53 +/À had more than one mammary tumor (10 mice with 2 tumors and 1 mouse with 3 tumors for a total of 28 tumors), and most of the time, the tumors in the same animal were of different morphologic types.The p53 +/À mice did not develop any mammary tumors during the observation period of 16 months.
The increased incidence of tumor formation in p53 heterozygote cyclin E transgenic mice confirmed that LMW cyclin E is cooperating with p53 to mediate breast tumorigenesis.To directly examine the status of p53 in the tumors of cyclin E/p53 bitransgenic mice, we assessed the cyclin E/p53 +/À mammary tumors for loss of the wild-type p53 allele, scored as loss of heterozygosity by Southern blot analysis (Fig. 5B).Semiquantitative analysis of p53 band intensity ratios was done with an image analysis software.Of the 12 cyclin E-T1/p53 +/À mammary tumors examined, all showed significant loss of the wild-type p53 allele from 60% to 88%.Of the three cyclin EL1/EL4/p53 +/À mammary tumors, one showed a nearly complete loss (86%) and the other two showed a 42% loss.These results suggest that LMW cyclin E-overexpressing cells have a selective pressure for loss of the wild-type p53 allele and these cells become highly tumorigenic and transforming in the mammary gland.
The LMW cyclin E/p53 +/À mice developed lung metastases: 20% of EL1/EL4/p53 +/À mice and 35.2% of T1/p53 +/À mice.In M46A/p53 +/À mice, none of the five malignant tumors developed metastases (Table 1B).When we compared the incidence of metastasis between cyclin E lines and their respective cyclin E/p53 +/À lines, we did not find any statistical differences.However, when we combined and compared the incidence of metastasis in all tumors formed in M46A animals, independent of p53 background, to incidence of metastasis in all tumors formed in EL1/EL4 and T1, again independent of p53 background, we found a statistically significant increase in the metastatic potential of the tumors formed in LMW cyclin E transgenic animals (Fig. 5C; P < 0.001).Thus, although loss of one allele of p53 decreases latency and increases incidence of primary mammary tumor formation in full-length or LMW cyclin Eoverexpressing mice, the loss of p53 does not further enhance the metastatic potential of LMW cyclin E. Therefore, the metastatic phenotype caused by LMW cyclin E seems to be independent of p53.

Discussion
Multiple studies have shown that overexpression of cyclin E is a relevant prognostic marker in breast cancer patients.However, the mechanism through which cyclin E deregulation contributes to oncogenesis in vivo has yet to be established.In particular, an important area of debate is whether overexpression of the full-length protein is sufficient to initiate mammary tumor formation or whether the LMW forms of cyclin E have unique properties, independent of constitutive expression, that drive oncogenesis.Additionally, because breast cancer metastasis is the disease that has high mortality, it is important to determine if full-length and LMW cyclin E have different abilities in mediating metastasis.
In this study, we have addressed all these questions and are able to confirm that LMW cyclin E is a stronger oncogene in breast cancer tumorigenesis than full-length cyclin E.Although full-length cyclin E-overexpressing animals showed a slight increase in formation of mammary tumors with low potential for metastasis, the incidence of both primary mammary tumor formation and metastasis was markedly enhanced in LMW cyclin E transgenic mice.The observation of higher oncogenicity in LMW cyclin E-overexpressing transgenic mice suggests that these isoforms have a gain-of-function when compared with fulllength cyclin E. Because M46A and LMW cyclin E mice expressed equivalent amount of transgenic protein, the differences in the observed phenotype most likely represent the intrinsic properties of LMW cyclin E isoforms.One obvious property of the LMW cyclin E is their resistance to inhibition by p21 and p27 (5,12), which increases the cyclin E-associated kinase activity and interferes more dramatically with DNA replication, DNA repair, apoptosis, and mitosis.Deregulation of these processes has been shown to increase genomic instability, which accelerates tumor progression by selecting for events that inhibit apoptosis and favor growth or by generating additional mutations that may be required to develop a significant metastatic burden (23).However, when we analyzed the tumors derived from cyclin E transgenic mice, we did not see gross changes in the genome, suggesting that alternate mechanisms were involved in tumor initiation.
One possible alternate mechanism suggested by our data is the cooperation between cyclin E and the ARF-mdm2-p53 pathway.We found that cyclin E-derived mammary tumors frequently showed loss of ARF and that tumor formation was enhanced with loss of wild-type p53.We also found that these mutations in ARF and p53 were mutually exclusive.None of the cyclin E-overexpressing tumors in the wild-type p53 background showed allelic loss of p53 by Southern blot or inactivation by point mutations.Instead, we found that ARF protein expression was lost in the majority of such tumors (3 of 3 in M46A, 8 of 9 in EL1/EL4, and 8 of 11 in T1 tumors).In contrast, all cyclin E/p53 +/À mammary tumors showed significant loss of the wild-type p53 allele, with few tumors showing loss of ARF protein.
Such alterations of the ARF-mdm2-p53 pathway in cyclin E-overexpressing mammary gland tumors are reminiscent of human cyclin E-overexpressing tumors, which frequently contain TP53 mutations (24).It may be that alteration of the p53 pathway represents one of the genetic changes required for LMW cyclin E-induced tumorigenesis.These in vivo findings may seem paradoxical compared with studies in cultured cells in which ARF did not seem to be implicated in cyclin E-induced p53 activation (22) but are consistent with other recent studies (25,26).These studies identified cyclin E overexpression in early precursor lesions of human breast, urinary bladder, lung, and colon as an ''oncogenic stress'' that activates a DNA damage response network that delays or prevents cancer.Thus, induction of this stress response by cyclin E overexpression is a potent tumor suppression mechanism and explains the selective pressure for alterations of the ARF-mdm2-p53 pathway as shown in our transgenic mouse model.Of particular relevance to our model system is the appearance of premalignant lesions in cyclin E transgenic animals.Because women with a history of atypical hyperplasias and in situ carcinomas have f5and 10-fold increased relative risks, respectively, of eventually developing invasive breast cancer, we were interested if our transgenic model would show some of these early lesions.We found atypical ductal hyperplasia and mammary intraepithelial neoplasia present in the contralateral gland of a tumor-bearing mouse or in the mammary tissue adjacent to the tumor.Thus, as in human breast cancer, our transgenic model follows sequential steps of atypia, mammary intraepithelial neoplasia, and invasive/metastatic tumors.Therefore, the LMW cyclin E transgenic model system can be used to investigate the multistep progression of mammary tumorigenesis and metastasis and to test chemopreventive treatment.
In conclusion, the two most striking findings of our study are (a) the role of LMW in the initiation of primary tumor formation and (b) the increased metastatic potential of LMW cyclin E expression when compared with full-length cyclin E. The inability of LMW cyclin E cells to arrest in G 1 and the consequent replication of A, Kaplan-Meier curve of mammary tumor-free survival in p53 +/À and each of the MMTV-cyclin E/p53 +/À lines.In line MMTV-cyclin E-M46A/p53 +/À , 5 of 17 (29.4%)mice developed mammary tumors with an average latency of 12.6 mo.In line MMTV-cyclin EL1-EL4/p53 +/À , 5 of 13 (38.4%)mice developed mammary tumors with an average latency of 11.6 mo.In line MMTV-cyclin E-T1/p53 +/À , 17 of 26 (65.4%) mice developed mammary tumors with an average latency of 10.7 mo.In p53 +/À mice, 0 of 17 (0%) developed mammary tumors (P < 0.01, compared with each of the MMTV-cyclin E lines; P = 0.56, M46A/p53 +/À versus EL1-EL4/ p53 +/À ; P = 0.02, M46A/p53 +/À versus T1/ p53 +/À ).B, levels of ARF and cyclin E in extracts of tumors from MMTV-cyclin E/p53 +/À transgenic mice were assessed by immunoblotting with antibodies specific for each protein.Tumor DNAs were also analyzed by Southern blot after Bam HI restriction digestion.The locations of the pseudogene, the mutant p53 allele, and the wild-type allele are indicated.The wild-type/mutant band intensity ratio is shown below each lane.Controls show near-equal hybridization at wild-type and Trp53 mutant bands in nontumor genomic DNA of p53 +/À mice (tail) but reduced hybridization for the wild-type band in genomic DNA from mammary adenocarcinomas of LMW cyclin E/p53 +/À mice.C, Kaplan-Meier curve of metastasis-free survival in M46A and LMW cyclin E lines independent of p53.A statistically significant increase in the metastatic potential of the tumors formed in LMW cyclin E transgenic animals was found (P < 0.001, log-rank test).
damaged DNA may result in the aberrant expression of gene products that induce tumors that lead to metastasis.Alternatively, dysregulated Cdk activity and cell cycle progression may directly facilitate the growth of new tumors at site of metastasis.Determination of how LMW cyclin E induces metastasis represents an exciting challenge and may facilitate the design of efficacious anticancer strategies.

Figure 2 .
Figure 2. Overexpression of the LMW cyclin E in transgenic mice induces metastatic mammary adenocarcinomas/carcinomas.A, Kaplan-Meier curve of mammary tumor-free survival in nontransgenic and each of the MMTV-cyclin E lines.In line MMTV-cyclin E-M46A, 7 of 67 (10.4%) mice developed mammary tumors with an average latency of 17 mo.In line MMTV-cyclin E-EL1-EL4, 18 of 71 (25.4%) mice developed mammary tumors with an average latency of 18.6 mo.In line MMTV-cyclin E-T1, 16 of 53 (30.2%) mice developed mammary tumors with an average latency of 17.2 mo.In nontransgenic mice, 3 of 57 (5.3%) developed mammary tumors with an average latency of 20 mo (P < 0.05, compared with each of the MMTV-cyclin E lines).B, morphologies of mammary tumors overexpressing cyclin E. H&E staining of MMTV-T1 (a and b), MMTV-EL1/EL4 (c, d, and f ), and MMTV-M46A (e) tumor sections shows that terminal-stage adenocarcinomas have varied morphologies: acinar (a; Â400), glandular (b; Â400), papillary (c ; Â400), solid (d ; Â400), adenosquamous (e; Â200), and myoepithelial (f ; Â200).C, protein extracts (50 Ag/lane) from mammary tumors were loaded on a 10% SDS gel and subjected to Western blot analysis and the expression of the human cyclin E transgene or the endogenous cyclin D1, p27, and actin was detected with specific antibodies.D, immunohistochemistry for human cyclin E protein in mammary tumors (b, d , and f ) and the contralateral mammary gland of the same animal (a, c , and e).This staining showed loss of cyclin E expression in mammary tumors from animal #1180 in d (line EL1/EL4-668) and from animal #1067 in f (line T1-571), whereas expression is present in the contralateral glands (c and e, respectively).Arrows, cyclin E expression in luminal cells.The positive tumor is from animal #369 in b (line EL1/EL4-678) and is used as a positive control.

Figure 3 .
Figure 3. Premalignant mammary lesions in cyclin E transgenic model.A, H&E staining of histologic sections of atypical ductal hyperplasia (ADH ; a), mammary intraepithelial neoplasm [MIN ; a putative ductal carcinoma in situ (DCIS ); b], a glandular carcinoma (c ), and lung metastasis (d ) from a 16-mo-old MMTV-cyclin E-T1 transgenic mouse.B, the table shows the number of atypical ductal hyperplasia, mammary intraepithelial neoplasm, or atypical ductal hyperplasia + mammary intraepithelial neoplasm present in the contralateral gland of mammary tumor-bearing mice.

Figure 4 .
Figure 4. Loss of ARF expression and intact p53 locus in cyclin E-overexpressing mammary tumors.Levels of ARF and cyclin E in extracts of tumors from MMTV-cyclin E transgenic mice were assessed by immunoblotting with antibodies specific for each protein.Tumor DNAs were also analyzed by Southern blot after Bam HI restriction digestion.The locations of the pseudogene and of the wild-type allele are indicated.The wild-type/pseudogene band intensity ratio was quantified for each lane.Near-equal hybridization for the wild-type and pseudogene band in genomic DNA from mammary adenocarcinomas of LMW cyclin E mice shows the integrity of the p53 locus.

Figure 5 .
Figure 5. Cooperation of cyclin E transgene with heterozygote p53 in mediating mammary oncogenesis.A, Kaplan-Meier curve of mammary tumor-free survival in p53 +/À and each of the MMTV-cyclin E/p53 +/À lines.In line MMTV-cyclin E-M46A/p53 +/À , 5 of 17 (29.4%)mice developed mammary tumors with an average latency of 12.6 mo.In line MMTV-cyclin EL1-EL4/p53 +/À , 5 of 13 (38.4%)mice developed mammary tumors with an average latency of 11.6 mo.In line MMTV-cyclin E-T1/p53 +/À , 17 of 26 (65.4%) mice developed mammary tumors with an average latency of 10.7 mo.In p53 +/À mice, 0 of 17 (0%) developed mammary tumors (P < 0.01, compared with each of the MMTV-cyclin E lines; P = 0.56, M46A/p53 +/À versus EL1-EL4/ p53 +/À ; P = 0.02, M46A/p53 +/À versus T1/ p53 +/À ).B, levels of ARF and cyclin E in extracts of tumors from MMTV-cyclin E/p53 +/À transgenic mice were assessed by immunoblotting with antibodies specific for each protein.Tumor DNAs were also analyzed by Southern blot after Bam HI restriction digestion.The locations of the pseudogene, the mutant p53 allele, and the wild-type allele are indicated.The wild-type/mutant band intensity ratio is shown below each lane.Controls show near-equal hybridization at wild-type and Trp53 mutant bands in nontumor genomic DNA of p53 +/À mice (tail) but reduced hybridization for the wild-type band in genomic DNA from mammary adenocarcinomas of LMW cyclin E/p53 +/À mice.C, Kaplan-Meier curve of metastasis-free survival in M46A and LMW cyclin E lines independent of p53.A statistically significant increase in the metastatic potential of the tumors formed in LMW cyclin E transgenic animals was found (P < 0.001, log-rank test).