|Home | About | Journals | Submit | Contact Us | Français|
Low molecular weight (LMW) isoforms of cyclin E are post-translationally generated in breast cancer cells and are associated with aggressive disease and poor prognosis. In this study, the specificity of LMW cyclin E to cancer cells was determined by measuring cyclin E expression in tumor and non-tumor tissue from 340 breast cancer patients. Our results reveal the LMW isoforms were detected significantly more frequently in breast tumor tissue than in adjacent non-tumor breast tissues (p < 0.0001). The biologic consequences of the LMW isoforms were studied using a non-tumorigenic mammary epithelial cell line transfected with the cyclin E isoforms and resulted in increased clonogenicity, the inability to enter quiescence in response to growth factor deprivation and genomic instability compared to the full-length cyclin E. Biochemical differences between the full-length and the LMW isoforms were also evident. Biacore analyses show that the LMW isoforms have more efficient binding to CDK2 compared to full-length cyclin E, which could account for the unique biologic consequences observed with the expression of LMW cyclin E. The LMW isoforms of cyclin E are tumor specific, and are biochemically and biologically distinct from the full-length cyclin E which could provide a novel role in breast cancer progression.
Cyclin E, complexed to its kinase partner, cyclin dependent kinase (CDK) 2, regulates G1 to S phase progression of the cell cycle.1,2 The cyclin E/CDK2 complex is normally negatively regulated by the cyclin dependent kinase inhibitors (CKIs) p21 and p27. The balance between G1 to S phase progression and the ability to halt the cell cycle is delicately controlled by cyclin/CDKs and CKIs in normal cells, but is often aberrant in cancer cells.3 In addition, normal cells, but not cancer cells, can exit the cell cycle before the restriction point if an environmental stressor is detected, such as a lack of nutrients. However, after the restriction point, cells are committed to replicate their DNA. Due to the crucial role that regulated expression and cyclin E activity plays in maintaining proliferative homeostasis, any defects in its expression could have a critical effect in tumorigenesis. Therefore, it is not surprising that cyclin E expression is aberrant in many types of cancers including colorectal, gastric, ovarian, melanoma as well as breast cancer.4–7
The form of cyclin E predominantly expressed in normal and tumor cells is the full-length 50 KDa isoform, considered wild-type and referred to as EL. However, tumor cells are uniquely capable of post-translationally cleaving the full-length cyclin E in to low molecular weight (LMW) isoforms. Elastase cleaves the full-length cyclin E in to 44 kDa and 33 kDa isoforms, which are then phosphorylated to generate doublets that have been designated as Trunk 1 (T1, 44–45 kDa) and Trunk 2 (T2, 33–35 kDa). Clinical studies have indicated that cyclin E overexpression occurs in 25% of breast cancer tumors and is associated with poor prognosis.8 We previously determined the clinical significance of LMW isoforms of cyclin E in 395 women with breast cancer by measuring cyclin E expression with western blot analysis. We showed that high levels of total or LMW cyclin E were the most powerful discriminants of disease-free and overall survival, outperforming currently used clinical criteria including nodal status, stage and estrogen-receptor status.8 For example, in multivariate analysis of factors predicted of disease-specific and overall survival, a high level of the lowmolecular-weight isoforms of cyclin E was strongly associated with poor outcome with a hazard ratio of 2.1.8 While, these findings suggest that there may be utility for determining the level of cyclin E expression in breast tumor specimens to better define prognosis in breast cancer patients, the specificity of the LMW isoforms of cyclin E to tumor cells has been questioned recently.9 Therefore, in the current study, we prospectively assessed the expression of LMW cyclin E versus the full-length form in breast tumor tissue and adjacent normal tissue from 340 breast cancer patients to validate our hypothesis that the LMW isoforms of cyclin E are in fact tumor specific.
LMW isoforms of cyclin E, compared to full-length cyclin E, de-regulate the cell cycle and have unique biochemical properties. For example, the overexpression of LMW cyclin E in MCF-7 breast cancer cells results in resistance to the growth inhibiting effects of anti-estrogens,10 due to their resistance to inhibition by p21 and p27.11 Compared with transgenic mice overexpressing full-length cyclin E, those expressing LMW cyclin E developed metastatic tumors and had significantly decreased survival.12 In this study, to directly link the LMW isoforms of cyclin E to the process of breast tumorigenesis, we assessed the phenotype of non-tumorigenic mammary epithelial cells, 76NE6, engineered to express the LMW cyclin E to determine whether the cells exhibited a phenotype characteristic of neoplastic transformation. This study shows that the LMW isoforms of cyclin E are indeed tumor-specific, and are both biologically and clinically relevant.
The LMW isoforms of cyclin E have been associated with an aggressive phenotype in several types of cancer including breast, ovarian, gastric and colorectal cancers as well as melanoma.8,15–18 To determine whether the LMW isoforms are specific to tumor tissue, breast cancer tissue along with adjacent non-tumor tissue samples were prospectively collected from 340 women with stage I or II breast cancer at the M.D. Anderson Cancer Center. Figure 1A shows a representative western blot of cyclin E expression in the tumor samples and normal adjacent tissue samples from six patients who are representative of those whose tumors express high levels of LMW-E. The full-length protein was seen in both the paired normal and tumor tissue samples. However, the LMW cyclin E isoforms were seen predominantly in the tumor samples, irrespective of the length of exposure time of the western blots (middle). Densitometric analyses were performed on each of the samples to quantitate both the full-length and LMW cyclin E expression. On average, the expression of full-length cyclin E expression in the tumor samples was equivalent to that of the normal samples (p =0.058, sign test). However, the frequency of LMW cyclin E expression in the tumor samples was significantly greater than that of the normal tissue samples (p < 0.0001, Fig. 1B). These data suggest that the LMW cyclin E forms are processed and occur predominantly in tumor tissue and not normal tissue. With this evidence that LMW cyclin E may be involved in breast cancer development, we next wanted to examine the biologic consequences of LMW cyclin E expression in vitro and in vivo to see if it is directly linked to the tumorigenic process.
We have previously shown that the LMW isoforms of cyclin E are hyperactive as a result of increased binding to CDK2.11 To determine whether objective and quantitative differences exist between the cyclin E isoforms when binding to cyclin E’s kinase partner CDK2, which could give rise to the unique biologic functions observed after expression of LMW cyclin E, surface plasmon resonance (SPR) technology was applied using a Biacore 2000 analyzer. Figure 2A shows the results of 4 different concentrations of the full-length (EL) or LMW (T1 and T2) isoforms of cyclin E binding to purified CDK2 at a low surface density (~1,370 RUs of CDK2, left graph), and at higher density CDK2 coating (~8,400 RUs, right graph). At all concentrations examined, we detected the highest level of binding to the CDK2 when the LMW-T2 isoform was injected, followed by LMW-T1; EL was the lowest. Furthermore, the increased binding by the LMW forms was independent of the amount of CDK2 bound to the chip (compare Fig. 2A right to left). The differences in binding to CDK2 became more prominent with increased concentrations of the cyclin E isoforms, with LMW-T1 binding to CDK2 41% more efficiently, and LMW-T2 binding 60% more efficiently than EL (full-length) did (Fig. 2B).
To ensure that the change in refractive index detected and reported by the analyzer was actually due to cyclin E binding to the CDK2, a separate experiment was performed in which each of the analytes was injected over the chip and the bound protein was immediately eluted and subjected to western blot analysis for cyclin E content (Fig. 2C). Cyclin E was present in each of the recovered EL, LMW-T1 and LMW-T2 samples but not in the uninfected lysate control, confirming that the change in refractive index detected was indeed the result of cyclin E binding to the ligand. These data quantitatively show that the hyperactivity associated with the LMW isoforms of cyclin E is due to increased binding efficiency to CDK2. Therefore, the LMW isoforms of cyclin E are biochemically distinct from full-length cyclin E.
To determine the biological effects of LMW cyclin E expression in a non-tumorigenic mammary epithelial cell line, stable clones of 76NE6 cells overexpressing full-length (EL) or LMW (T1) isoforms of cyclin E at physiological levels were generated.19 We examined whether the overexpression of the hyperactive, LMW isoform T1, has functional consequences that could play a role in neoplastic transformation. Overexpression of the T1 LMW isoform resulted in a growth advantage to 76NE6 cells in a clonogenic assay compared to overexpression of the EL isoform or an empty vector control (Fig. 3A). The LMW-T1 cells form colonies more readily than the EL clones did and both of these clones formed colonies more readily than the empty vector clones did. Therefore, overexpression of the hyperactive LMW isoforms of cyclin E provided a growth advantage to mammary epithelial cells.
Deregulated cell proliferation is a hallmark of cancer cells and can occur when the cancer cell acquires self-sufficiency in growth signals or a resistance to anti-growth signals.20 To determine whether 76NE6 cells expressing LMW cyclin E had a decreased need for supplemental growth factors, we examined whether LMW-T1 cells, EL cells and empty vector cells responded differently in response to growth factor removal. Growth was arrested in the G0/G1 phase in 100% of the 76NE6 cells with the empty vector and in 64% of the 76NE6-EL cells. In contrast, the 76NE6-T1 cells did not respond to a lack of nutrients by arresting their cell cycle (Fig. 3B). When growth factors were added back to the empty vector and EL cells, they synchronously reentered the cell cycle within the first 10 hours (Fig. 3B), whereas the LMW-T1 cells remained distributed throughout the cell cycle.
To determine whether the 76NE6-empty vector or EL cells were exiting the cell cycle into a G0 phase or arresting in G1 phase while lacking growth factors, Ki67 was measured as a marker of cell cycle activity. Ki67 is expressed in all phases of the cell cycle (G1, S, G2 and M), but not G0. The individual cells were examined for both DAPI (to label DNA of all cells) and Ki67 staining at 0, 24, 72 and 96 hours after growth factor deprivation (Fig. 3C). We next examined the percent Ki67 positive cells as a function of DAPI staining over time of growth factor deprivation in the empty vector, EL and T1 overexpressing cells. The staining was quantitated and graphically shown in Figure 3D. For each time point, the Kruskal-Wallis test was used to compare the fractions of proliferating cells for the 3 conditions (vector alone, EL, T1). These analyses revealed that there are no significant differences for the 24-hr (p = 0.688) and 48-hr (p = 0.191) time points, but there are significant differences at the 72-hr (p = 0.020) and 96-hr (p = 0.013) time points. Next pair wise comparisons, using Mann-Whitney test was used at the 72-hr and 96-hr time points to see where the differences lie. For the 96-hr time point, EL is significantly different from T1 (p = 0.021). Specifically, in cells transfected with empty vector or EL, only 4.1% and 27% of the cells respectively remained in the cell cycle (as detected by Ki-67 positivity) following 96 hrs of growth factor deprivation. However, the majority, 63%, of T1 transfected cells did not succumb to G0 arrest and remained in active cell cycle (Fig. 3D). These results suggest that the LMW-T1 overexpressing cells circumvented the regulatory mechanism in place in normal cells when challenged with a lack of nutrients and resisted a quiescent state.
Deregulation of the cell cycle can result in the unfaithful transmission of genetic information manifested as gross chromosomal aberrations. To determine whether deregulation of the cell cycle by cyclin E affects the genomic fidelity of the cell, gross chromosomal aberrations were assessed by karyotype analysis on metaphase arrested 76NE6-cyclin E clones (Fig. 4A). The empty vector karyotype was considered normal with all chromosomes being intact. However, chromosomal aberrations were identified in the karyotypes of both the EL and LMW-T1 cells (Fig. 4A), including: dicentric chromosomes, ring chromosomes, chromatid breaks, chromosome fragments and telomere fusion. Figure 4B shows the percentage of metaphases that had at least one chromosome with any type of aberration. LMW-T1 overexpressing cells exhibited significantly more overall chromosomal abnormalities than the EL overexpressing cells. The empty vector expressing cells exhibited significantly less (or no) chromosomal aberrations. We conclude that 76NE6 cells became genetically unstable subsequent to LMW cyclin E expression.
The LMW isoforms of cyclin E are specifically overexpressed in tumor tissue but not in normal tissue. The frequent appearance of the LMW forms of cyclin E and their correlation with poor prognosis in breast cancer patients suggests that they play specific roles in the development or progression of this malignancy. In this study, the overexpression of LMW cyclin E led to distinct biochemical and biological characteristics, including genomic instability, in mammary epithelial cells. Our results illustrate a potential role for the LMW forms of cyclin E in breast cancer tumorigenesis.
Using 340 breast cancer samples we have provided quantitative evidence that the LMW cyclin E are specific to tumor cells and tissues and not just a function of the levels of full length cyclin E. The analysis is Figure 1 also shows that following over-exposure of western blots, that no LMW was detected in the normal adjacent tissue as compared to tumor tissue. Hence, when using quantitative conditions (as we have done here), LMW derivatives of cyclin E1 are undetectable in normal cells, are tumor-specific and biologically important.
The ability of LMW-T1 cells to more readily colonize than the EL or empty vector clones in clonogenic assays, suggests that the LMW-T1 clone does acquire a growth advantage compared to the EL (full-length cyclin E) or empty vector clones. The observed growth advantage could be the result of the LMW-T1 cells being insensitive to anti-growth signals as opposed to increased proliferation. We showed that LMW cyclin E prevents cells from exiting the cell cycle and entering quiescence upon growth factor deprivation (Fig. 3B–D). The cells overexpressing LMW cyclin E remain active in the cell cycle whereas a normal cell senses that the environment is not optimal for cell proliferation. This resistance to anti-growth signals is likely coupled to the self-sufficiency in growth signals and uncontrolled progression through the cell cycle. While there are several mechanisms that LMW cyclin E may used to disrupt the cell cycle, the lack of response by cells overexpressing the LMW isoforms of cyclin E to antigrowth signals could be the result of their resistance to CKIs. Indeed, the accumulation of p27 has been reported to be necessary for cells to enter quiescence.21
Because cells overexpressing the LMW cyclin E isoforms do not enter quiescence under growth factor deprived conditions, it is possible that other cues that typically stop proliferation, such as DNA damage, would be ignored by these cells. Replicating aberrant DNA leads to an unstable genome. Genomic instability has been proposed to be a way in which cells undergo neoplastic transformation because the unstable genome helps the cell acquire the traits of a tumor cell phenotype. Cells expressing the LMW-T1 isoform had significantly more genomic aberrations than did those expressing full-length cyclin E or empty vector cells.
Functional assays confirmed that the cell cycle deregulation observed with expression of LMW cyclin E is secondary to the increased kinase activity associated with these complexes. Aberrant G1 to S phase transitions are a means for tumor cells to facilitate their growth, thus raising the question of how the LMW isoforms achieve their increased kinase activity. The resistance to CKIs by the LMW isoforms of cyclin E is one mechanism by which the LMW isoforms achieve a hyperactive phenotype compared with full-length cyclin E. Another mechanism is the ability of LMW cyclin E to bind to CDK2 more efficiently. We showed that CDK2 binds to 40% to 60% more of the LMW isoforms than to the fulllength cyclin E.
Understanding the mechanisms of LMW cyclin E in tumor formation is critical because patients with breast cancers that overexpress LMW cyclin E have an extremely poor outcome. Therefore, LMW cyclin E may serve as a novel target for therapeutic intervention in this subset of breast cancers. To this end, it is important that the LMW isoforms of cyclin E are tumor-specific if they are to serve as a therapeutic target (Fig. 1). Therefore, future directions should include the design of novel drugs, specifically targeting the formation of the LMW isoforms of cyclin E or their increased kinase activity.
Tumor and adjacent normal tissue were obtained at the time of surgical intervention from 340 patients with stage I or II breast cancer. Patients signed a study-specific informed consent form for participation in this protocol, which was approved by the institutional review board at The University of Texas M.D. Anderson Cancer Center. Following surgical resection of the primary tumor and regional lymph nodes, specimens were examined in pathology, and fresh tumor and normal tissue were collected by the pathologist and were processed in our laboratory to obtain protein lysates. Protein lysates were subjected to western blotting followed by densitometric analysis with use of ImageQuant Total Lab software (Amersham Biosciences, Piscataway, NJ). Each band was quantified and the LMW cyclin E bands were added together. Levels of LMW cyclin E and full-length cyclin E in tumor specimens were normalized against the levels of full-length cyclin E in adjacent normal tissue.
Cyclin E binding to CDK2 was monitored in real time with a Biacore 2000 instrument (Biacore AB, Uppsala, Sweden). CDK2 was purified from sf-9 cells using a protocol provided by Dr. Laurent Meijer (CNRS, Roscoff, France). The resultant CDK2 was tested for purity on a silver stained sodium dodecyl sulfate-polyacrylamide gel and tested for activity using a kinase assay. The freshly prepared CDK2 was immobilized on a CM5 chip (Biacore AB) using amine coupling. Surface preparation: The CM5 chips were normalized with 40% BIAnormalizing solution (Biacore AB). Two of the channels on each chip were then coated with the ligand, CDK2. This was achieved by injecting CDK2 diluted 1:1 with 0.1M sodium acetate, pH = 3.5, at a flow rate of 5 µL/minute for 50 minutes in HBS-P buffer (0.01 M HEPES pH 7.4, 0.15 M NaCl, 0.005% v/v tween20) over the activated surface. Bovine Serum Albumin (BSA) was coupled to one of the flow channels of each chip, to be used as a non-specific binding control: This was accomplished by injecting 50 nM BSA over the flow channel using the same conditions as for the CDK2. One of the flow channels was left blank to serve as a background control. Finally, the uncoupled groups on the surface were deactivated by injecting 1 M ethanolamine-HCl, pH 8.5 over all of the channels.
Cyclin E samples were prepared from sf-9 cell lysates overexpressing each of the LMW cyclin E isoforms (EL, T1 or T2). Samples were diluted to 1 µg/µL, 500 ng/µL, 250 ng/µL and 125 ng/µL concentrations using sterile water. 70 µL of analyte was injected over all four channels of the chips at a flow rate of 30 µL/minute in 50 mM Tris pH 7.5, 10 mM MgCl2, 0.005% v/v Tween20. Each analyte was injected twice in increasing concentrations, with water injections between each concentration for normalization.
After each injection, the surface was regenerated with 5 µL of HBS-EP buffer (0.01 M HEPES pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005% v/v surfactant P20).
In the experiments in which bound protein was collected for further tests, the flow rate was set to 30 µL/ minute. One hundred microliters of each cyclin E sample was coinjected with 60 µL of HBS-EP buffer and 100 µL of the eluted sample was recovered and subjected to western blotting.
Scrubber2 software (Center for Biomolecular Interaction Analysis, University of Utah, (http://www.cores.utah.edu/interaction/scrubber.html) was used for all binding analyses, which compares molar equivalents of the analyte to account for the differences in size between the isoforms of cyclin E.
The 76NE6 cell line was maintained in DFCI-1 media as described.13 pcDNA 4.0 plasmids (Invitrogen, Carlsbad, Ca) containing either the cyclin EL-Flag, T1-Flag or no insert and the gene for Zeocin resistance were linearized for transfection. Stable pools were generated by selecting for the cells containing the vector using 20 µg/mL of Zeocin for 16 days and then maintaining the cells in 10 µg/mL of Zeocin.
A total of 100, 500 or 1,000 cells were plated in 100 mm tissue culture dishes. The plates were incubated for 14 days and then stained with 0.05% crystal violet.
1 × 106 cells of each 76NE6 clonal cell line: empty vector (4.0), full-length cyclin E (EL) and LMW cyclin E (T1) were plated in ten 100 mm plates for cell cycle analysis and 3 × 105 cells of each clone were plated in duplicate wells of ten 6- well plates for Ki-67 staining experiments in complete DFCI-1 medium.13 24 hours after plating the cells, the media was changed to DFCI-3,13 to deplete the cells of growth factors. To characterize the cells during synchronization, they were harvested at the 0 hour time point. A 100 mm plate and the duplicate wells of the 6-well plates were harvested for flow cytometric analysis or Ki-67 staining every 12 hours thereafter for 120 hours, as previously described.14 To characterize the synchronous cells, the DFCI-3 media was changed to DFCI-1 media after 72 hrs and this was the 0 hr time point. Cells were collected every 3 hours for 33 hours thereafter for flow cytometric analysis and western blotting.
3 × 105 cells of each clone (empty vector, EL and T1) were plated in duplicate wells of 6-well plates on top of cover-slips that were sterilely placed in each well. The cells were synchronized as described above. At each time point, cover slips were washed with cold phosphate-buffered saline (PBS) and fixed for 1 minute by adding 500 µL of cold fixative (1:1 v/v mix of methanol and acetone). After 3 more washes, cells were incubated with 50 µL of fluorescein isothyocyanate (FITC) conjugated Ki-67 antibody (Abcam, Cambridge, MA), diluted 1:10 with PBS for 2 hours, then washed again. ProLong Gold anti-fade reagent with 4',6-Diamidino-2-phenylindole (DAPI, Invitrogen) was dropped onto microscope slides, which were then sealed with clear nail polish. Slides were analyzed within 24 hours using the 20X objective of a Nikon Optiphot microscope with an attached digital camera. The percentage of Ki-67 positive cells was determined by taking the ratio of Ki-67 positive cells to the number of DAPI positive cells.
1.5 × 106 cells of each 76NE6 clone (empty vector, EL and T1) were plated on 100 mm plates. Fresh medium was added to the cells at 6 and 18 hours until they reached ~80% confluency, at which point 20 ng/mL Colcemid was added to each plate for 4 hours. Cells were microscopically observed for the presence of metaphase cells (enlarged, translucent cells that can come off the plate with gentle tapping). The metaphase spreads were performed and analyzed for genetic instability by the Molecular Cytogenetics Facility at M.D. Anderson Cancer Center. Statistical differences were determined using the student t test with a 95% confidence interval. p values less than 0.05 were considered statistically significant.
All values for LMW cyclin E in the tumor and normal samples, and all values for full-length cyclin E in the tumor samples were normalized to the full-length cyclin E values in the normal tissue samples. For statistical analyses to determine correlation, we used the sign test, comparing the ratio of paired samples’ deviation from 1. For determination of p values for Ki67 data presented in Figure 3, for each time point, the Kruskal-Wallis test was used to compare the fractions of proliferating cells for the 3 conditions (4, EL, T1). To compare EL to T1 Mann-Whitney test was then used. All other p values were calculated using a 2-sided student’s t test. For bar graphs, error bars indicate 95% confidence intervals.
We would like to thank Flavio Palalon and the The T.C. Hsu Molecular Cytogenetics Core facility for technical assistance and Tamara Locke for editing the manuscript. This study was funded in part by the following grants: National Institutes of Health CA87458 (to K. Keyomarsi); National Cancer Institute P50CA116199 (to K. Keyomarsi); Clayton Foundation (to K. Keyomarsi) Susan G. Komen Breast Cancer Foundation BCTR0504200 (to K.K. Hunt); Breast Cancer Prevention and Treatment Research Fund, The University of Texas M.D. Anderson Cancer Center (to K.K. Hunt).