Cyclooxygenase-2 overexpression in MCF-10F human breast epithelial cells inhibits proliferation, apoptosis and differentiation, and causes partial transformation
Article first published online: 26 APR 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 116, Issue 6, pages 847–852, 10 October 2005
How to Cite
Lu, S., Yu, G., Zhu, Y. and Archer, M. C. (2005), Cyclooxygenase-2 overexpression in MCF-10F human breast epithelial cells inhibits proliferation, apoptosis and differentiation, and causes partial transformation. Int. J. Cancer, 116: 847–852. doi: 10.1002/ijc.21142
- Issue published online: 28 JUL 2005
- Article first published online: 26 APR 2005
- Manuscript Accepted: 2 MAR 2005
- Manuscript Received: 7 JAN 2005
- The Canadian Breast Cancer Research Alliance
- Canadian Breast Cancer Research Alliance
- Program in Food Safety, Nutrition and Regulatory Affairs (University of Toronto)
- human breast epithelial cells;
- cell proliferation;
To investigate the effects of cyclooxygenase-2 (COX-2) overexpression on breast cancer development, we stably transfected MCF-10F human breast epithelial cells with an expression vector containing human COX-2 cDNA oriented in the sense (10F-S) or antisense (10F-AS) direction. As expected, 10F-S cells expressed elevated levels of COX-2 protein, whereas this protein was undetectable in the 10F-AS cells. Prostaglandin E2 production in these cells reflected COX-2 levels. The 10F-S cells had a significantly decreased rate of proliferation compared to 10F-AS or parental cells, and a delay in progression through the G1 phase of the cell cycle. COX-2 overexpression also caused resistance to detachment-induced apoptosis (anoikis) as well as an inhibition of differentiation in cells cultured in Matrigel. Furthermore, after ∼ 20 passages in culture, 10F-S cells developed fibroblast-like features, expressed vimentin, and formed foci of dense growth when cultured at confluence, suggesting that the cells were undergoing epithelial to mesenchymal transition (EMT). The 10F-S cells, however, were unable to grow in soft agar or form tumors in nude mice, suggesting that they were only partially transformed. Our observations suggest that COX-2 overexpression in human breast epithelial cells will predispose the mammary gland to carcinogenesis. © 2005 Wiley-Liss, Inc.
Cyclooxygenase (COX) is the rate-limiting enzyme in the biosynthesis of eicosanoids from arachidonic acid. There are 2 COX isoforms: the constitutive form, COX-1, is involved in processes such as parturition and platelet aggregation;1 the inducible form, COX-2, is involved in inflammatory reactions1 as well as ovulation, implantation, perinatal renal development and remodeling of the ductus arteriosus.2COX-2 expression is induced by a variety of proinflammatory agents, growth factors, tumor promoters and mitogens.1, 3 Nonsteroidal anti-inflammatory drugs (NSAIDs) inhibit either COX-1, COX-2 or both isoforms, depending on the structure of the drug, and are widely used for the treatment of rheumatoid- and osteo-arthritis.
Recently, overexpression of COX-2 has been found to be a general feature of neoplasms, particularly those of epithelial origin, in both experimental animals and humans. In humans, upregulation of COX-2 has been reported for colon cancers associated with familial adenomatous polyposis, as well as sporadic colorectal cancer, and for cancers of the stomach, lung, esophagus, liver, bile duct, pancreas and skin.4 NSAIDs have been shown to inhibit the development and/or growth of a variety of carcinomas in experimental animals including colon,5 skin6 and lung.7
A number of lines of evidence have linked COX-2 to the development of breast cancer. Approximately 50% of human breast tumors express COX-28, 9 and recent epidemiological studies have suggested an inverse association between regular use of NSAIDs and the risk of breast cancer.10, 11 We and others have shown that NSAIDs are potent inhibitors of rat mammary carcinogenesis.12, 13COX-2 overexpression has been shown to increase proliferation, inhibit apoptosis, and enhance the invasiveness of breast cancer cells and to induce angiogenesis.14 A recent study has shown that transgenic mice overexpressing COX-2 in mammary epithelial cells develop mammary tumors.15 In that study, however, mammary tumors only developed in mice that had been pregnant several times. Precocious development of the mammary glands, not mammary tumors, occurred in virgin animals.15 As suggested by the authors, the relatively low level of COX-2 expression in the virgin mice, compared to the much higher level of expression in the multiparous mice during pregnancy and lactation, may explain these differences. It is not clear, however, whether COX-2 overexpression causes any phenotypic changes in normal mammary epithelial cells or whether COX-2 overexpression itself is sufficient to induce malignant transformation. Indeed, Liu et al.15 speculate that a second mutation may be necessary for complete transformation of the mammary epithelium.
In our study, we have stably overexpressed COX-2 in MCF-10F cells. These cells were established from normal human breast tissue and do not grow in soft agar and are not tumorigenic in nude mice. They provide a valuable tool with which to investigate differentiation, immortalization and carcinogenesis in a nonneoplastic breast epithelial system.16 The susceptibility of the mammary gland to tumorigenesis is influenced by its normal development, particularly puberty and pregnancy that are associated with marked alterations in cell proliferation, apoptosis and differentiation.17 Our objective here was to determine whether COX-2 overexpression causes any changes in these parameters in MCF-10F cells that may predispose them to malignant transformation. Furthermore, we sought to determine whether overexpression of COX-2 is itself sufficient to cause malignant transformation.
Material and methods
MCF-10F cells from the American Type Culture Collection (ATCC, Rockville) were grown in a 1:1 mixture of Dulbecco's modified Eagle's medium (DMEM) and Ham's F12 medium with 20 ng/ml epidermal growth factor, 100 ng/ml cholera toxin, 0.01 mg/ml insulin, 500 ng/ml hydrocortisone and 5% chelexed horse serum.18 MCF-7 human breast cancer cells were obtained from ATTC (cultured in DMEM/F12 1:1 with 10% FBS) and HCA-7 human colon cancer cells (cultured in DMEM with 5% FBS) were generously provided by Dr. S. Kirkland (Imperial College School of Medicine, London, UK).19
Human COX-2 cDNA, generously provided by Dr. S.M. Prescott (University of Utah, Salt Lake City, UT),20 was cloned into the eukaryotic expression vector pIRES2-EGFP (CloneTech, Palo Alto, CA) in both sense and antisense orientations and the vectors were transfected into MCF-10F cells with selection in medium containing 1.5 mg/ml of G418. Cells transfected with the sense vector were designated 10F-S, those transfected with the antisense vector 10F-AS, and the parental cells 10F-P.
COX-2 protein expression and prostaglandin E2 (PGE2) measurements
The 10F-P, 10F-S and 10F-AS cells were cultured to 70–80% confluence. The cells were harvested and lysed in RIPA buffer (PBS containing 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 100 ng/ml PMSF and 66 ng/ml aprotinin). Samples containing 50 μg protein were separated on 12% SDS-polyacrylamide gels. COX-2 was detected by using a rabbit polyclonal antibody (Cedarlane Laboratories Limited, Hoenby, ON, Canada). Culture medium was analyzed for PGE2 by an ELISA kit (Cedarlane Laboratories Limited) and levels normalized to cell protein.
The 10F-P, 10F-S and 10F-AS cells were seeded at an initial density of 5 × 104 cells per well in 24-well plates and cultured for 4 days. After removal of the medium, cells were washed once in PBS, fixed in 100% methanol for 10 min, stained with 0.5% crystal violet in 20% methanol for 10 min and washed with tap water.21 After air-drying, the stained cells were solubilized in 1% SDS. Absorbance of the retained crystal violet was determined at 595 nm. The cell density is reported as the ratio of cells at different time points to cells at day 0 expressed as a percentage.21
Cell cycle and cyclin A, D1 and E measurements
The 10F-P, 10F-S and 10F-AS cells were grown to 70% confluence on 10 cm plates. Some plates were used for protein isolation, while others were used for FACS analysis. Cells were washed with PBS, trypsinized, fixed with 70% ethanol, and stored at −20°C prior to staining for 30 min at room temperature with propidium iodide (20 μg/ml) in PBS containing 0.1% Triton X-100 and 0.2 mg/ml DNase-free RNase A and analyzed on a FACSCalibur flow cytometer (Becton and Dickenson, Franklin Lakes, NJ). For cyclin analysis, proteins from whole cell lysates were separated on 10% SDS-polyacrylamide gels. Cyclin A, D1 and E levels were determined using rabbit anti-human cyclin A, D1 and E (catalog numbers 06-138, 06-137 and 06-459, Upstate Biotechnology, Lake Placid, NY).
Apoptosis and Bcl-2, Bcl-XL, Bax and Bak measurements
To assess detachment-induced apoptosis (anoikis), cells were seeded at a density of 105 cells/ml in 6-well poly(2-hydroxyethyl methacrylate) (HEMA)-coated dishes22 and then harvested after 24 and 48 hr. Apoptosis was quantified using a TUNEL-based Apoptosis Detection kit (catalog number TA5354, R&D Systems, Inc., Minneapolis, MN) according to the manufacturer's protocol. For Bcl-2, Bcl-XL, Bax and Bak measurements, proteins from whole cell lysates were separated on 13% SDS-polyacrylamide gels and probed with rabbit polyclonal anti-Bcl-2 or rabbit polyclonal anti-Bcl-XS/L (catalog numbers sc-492 and sc-1041, Santa Cruz Biotechnology, Santa Cruz, CA), rabbit polyclonal anti-Bax or rabbit polyclonal anti-Bak (catalog numbers 06-499 and 06-536, Upstate Biotechnology).
Cell differentiation in Matrigel
Six-well culture plates were coated with 1 ml/well BD Matrigel Matrix (catalog numbers BD 354234, BD Biosciences, Mississauga, ON, Canada) as supplied by the manufacturer and incubated overnight at 37°C. Cells were suspended in complete medium and seeded at a density of 1 × 106 per well and incubated at 37°C for 24–72 hr, after which time morphology was assessed by light microscopy.23, 24
Cell transformation assays
Cells were passaged until there were obvious morphological changes. At this time, proteins from whole cell lysates were separated on 10% SDS-polyacrylamide gels and probed with mouse monoclonal anti-E-cadherin (catalog number 13-5700, Zymed Laboratories, Inc., San Francisco, CA), mouse monoclonal anti-keratin (catalog number MS-343-P, Lab Vision Corporation, Fremont, CA) or mouse monoclonal anti-vimentin (catalog number MS-129-P, Lab Vision Corporation). At this stage, some cells were maintained at confluence to observe focus formation. Anchorage-independent growth was assessed by measuring growth in soft agar. Briefly, ∼ 5 × 103 cells were suspended in 2 ml of 0.3% (w/v) agarose dissolved in culture medium and poured onto a bed of 0.5% agarose in 6-well plates. Colony formation was followed for 4 weeks.25 MCF-7 cells were used as a positive control. Tumorigenicity was assessed in nude mice. Under anesthesia, a 1.7 mg, 60-day release 17β-estradiol pellet (catalog number SE-121, Innovative Research of America, Sarasota, FL) was implanted into the interscapular region of 8-week-old nude mice (CD1-Nu/Nu, Charles River Laboratories, St. Constant, Quebec, Canada) (5 mice/group). At the same time, ∼106 (100 μl) cells were injected into the mammary fat pad.26 After 8 weeks, mice were sacrificed and the tumor growth was assessed. MCF-7 human breast cancer cells were used as a positive control.
All data were expressed as means ± SEM. Differences between groups were analyzed by Student's t-test.
Preparation of human breast epithelial cells (MCF-10F) that constitutively overexpress COX-2
COX-2 is expressed at a low level in MCF-10F human breast epithelial cells.27 To examine the effect of COX-2 expression in these cells, they were transfected with the eukaryotic expression vector pIRES2-EGFP containing human COX-2 cDNA in the sense (10F-S) or antisense (10F-AS) orientations. This vector leads to stable gene expression. A number of clones of 10F-S and 10F-AS cells were prepared that had similar properties and the results from 1 representative clone are presented here. To evaluate relative COX-2 expression levels in the clones, we performed Western blot analysis as well as COX assays. HCA-7 colon cancer cells that are known to overexpress COX-228 were used as a positive control. As shown in Figure 1a, the 69 kDa COX-2 protein was expressed at a high level in both the HCA-7 and 10F-S cells, to a small extent in the parental cells (10F-P), but was absent in the 10F-AS cells. Similar results were obtained in early- as well as late- (∼ 20) passage cells. PGE2 production in these cells reflected the COX-2 levels (Fig. 1b) whereas the PGE2 in 10F-P and 10F-AS cells was likely due to COX-1.
COX-2 overexpression causes decreased cell proliferation and cell cycle arrest
Since COX-2 is induced by growth factors and tumor promoters, we initially hypothesized that the stable expression of COX-2 might promote cell growth. It is clear from Figure 2a, however, that COX-2 overexpression decreased the proliferative rate of 10F-S cells compared to 10F-P or 10F-AS cells that had similar growth rates. As shown in Figure 2b, 10F-S cells had a significantly higher percentage of cells in G1 phase and a significantly lower percentage of cells in S phase compared to parental line or 10F-AS cells. Furthermore, 10F-S cells expressed a lower level of cyclin E than 10F-P and 10F-AS cells, but there were no differences in cyclins A or D1 (Fig. 2c).
COX-2 overexpression alters susceptibility to apoptosis
To determine whether expression of COX-2 in MCF-10F cells promotes resistance to detachment-induced apoptosis (anoikis), we cultured the clones in polyHEMA-coated plates, a procedure that is known to induce anoikis in normal epithelial cells.29 The 10F-S cells were less susceptible to anoikis than 10F-AS or 10F-P cells, although this difference was not significant until 48 hr (Fig. 3a). We showed, furthermore, that 10F-S cells expressed somewhat higher levels of Bcl-2 and Bcl-XL at both 24 and 48 hr. At 24 hr, 10F-AS and 10F-S cells expressed higher levels of Bax than 10F-P cells, whereas at 48 hr, 10F-S cells clearly expressed higher levels of this protein than the other 2 cell lines. Levels of Bak were similar among the 3 cell lines at both time points (Fig. 3b).
COX-2 overexpression inhibits differentiation
It is well established that the extracellular matrix is required for normal differentiation of mammary epithelial cells.30 Therefore, to investigate the effect of COX-2 expression on differentiation, we cultured the cells on Matrigel-coated plates, a method that has been used by other investigators to study epithelial differentiation.23, 24 Under these conditions, 10F-P and 10F-AS cells aggregated to form smooth spheroids (Fig. 4a), structures previously observed to form from well-differentiated mammary epithelial cells.23, 24 The 10F-S cells, however, formed irregular 3-dimensional structures with many fibroblast-like cells showing invasion into the Matrigel (Fig. 4b), suggesting differentiation of these cells was inhibited.
COX-2 overexpression causes partial transformation
To determine whether overexpression of COX-2 would eventually lead to neoplastic transformation, we passaged cells many times. After about 20 passages, some of the 10F-S cells developed long projections and they began to show the loss of contact inhibition (Fig. 5a). In contrast, throughout the culture period, 10F-P (Fig. 5a) and 10F-AS cells (not shown) consisted of uniformly well flattened, polyhedral cells that grew in monolayers, though individual cells varied somewhat in size and shape. Furthermore, unlike 10F-P and 10F-AS cells that only expressed the epithelial markers E-cadherin and keratin, after 20 passages 10F-S cells also expressed the fibroblastoid marker vimentin (Fig. 5b). When cells that had been passaged 20 times were maintained at confluence for 5 weeks, 10F-S cells formed 12–17 foci of dense growth per plate (Fig. 5c,d), whereas 10F-P cells formed only 0–3 foci per plate and 10F-AS cells formed no foci (Fig. 5d). Cells from the foci, however, neither grew in soft agar nor formed tumors in nude mice. MCF-7 human breast cancer cells were used as a positive control for these experiments.
Like most other tissues, mammary gland homeostasis is maintained by a dynamic balance of cell proliferation, apoptosis and differentiation, but particularly large changes in these parameters occur during gland development at puberty and during pregnancy, lactation and involution.31 Factors that disturb this balance may influence the susceptibility of the mammary gland to tumorigenesis. In our study, we have shown that a number of phenotypic changes occur in human breast epithelial cells overexpressing COX-2 that may be associated with an increased susceptibility of these cells to transformation.
We have demonstrated that overexpression of COX-2 in MCF-10F human breast epithelial cells leads to decreased cell proliferation and a delay in progression through the G1 phase of the cell cycle. Similar phenotypes induced by COX-2 overexpression have also been reported in rat intestinal epithelial cells,32 immortalized endothelial cells, NIH3T3, COS-7, bovine microvascular endothelial cells and human embryonic kidney cells (239 cells).33 The COX-2-induced G1 arrest in MCF-10F cells was associated with a downregulation of cyclin E but no changes in the expression levels of cyclins A or D1. In rat intestinal epithelial cells, however, G1 delay was associated with down-regulation of cyclin D1.32 Since cyclins D1, E and A are involved in the passage of cells through the restriction point and entry into S phase,34 our results confirm other reports that regulation of cyclins is cell-type specific.35, 36 G1 delay in rat intestinal epithelial cells was shown to be associated with an increased resistance to butyrate-induced apoptosis and thus to a survival advantage that may play a role in carcinogenesis.32, 37 Therefore, we next examined the effect of COX-2 overexpression on the susceptibility of MCF-10F cells to apoptosis.
Apoptosis that is induced by inadequate or inappropriate cell-matrix interactions (anoikis) is involved in a wide diversity of tissue-homeostatic, developmental and oncogenic processes.38 Recent reports have shown that resistance to anoikis contributes significantly to the malignancy of mammary cancers.39 We found that MCF-10F cells overexpressing COX-2 were more resistant to anoikis than 10F-P or 10F-AS cells. This resistance was associated with higher levels of expression of the anti-apoptotic proteins Bcl-2 and Bcl-XL. Upregulation of Bcl-2 has been found to be associated with resistance to apoptosis in mammary epithelial cells during lactation and involution and in mammary tumors in MMTV-COX-2 transgenic mice,15 as well as in rat intestinal cells that overexpress COX-2.37 We found no differences, however, in the expression of the proapoptotic protein Bak among the 10F-S, 10F-AS and 10F-P cells and, surprisingly, higher levels of proapoptotic Bax were present in 10F-S cells compared to the 10F-AS and 10F-P cells. These results suggest that neither Bak nor Bax play a direct role in conferring resistance to anoikis in the breast epithelial cells overexpressing COX-2. Our observations suggest that resistance to apoptosis in human breast epithelial cells overexpressing COX-2 may be conferred by anti-apoptotic Bcl-2 and Bcl-XL.
Differentiation, together with proliferation and apoptosis, tightly regulates mammary gland development and remodeling.40 In our study, we have demonstrated that overexpression of COX-2 in MCF-10F cells leads to the inhibition of differentiation when cells are cultured in matrigel. Similar inhibition of differentiation was seen in rat intestinal epithelial cells that overexpressed COX-2.37 It is well known in rodent models that pregnancy and age-related mammary gland differentiation are associated with resistance to mammary carcinogenesis.41 Indeed, susceptibility to mammary carcinogenesis has recently been related to the extent of morphological differentiation of the gland and the expression of genes involved in differentiation.42 Our observations, therefore, suggest that COX-2-induced inhibition of the differentiation of human breast epithelial cells will increase the susceptibility of the breast to carcinogenesis. Our observations also help to explain why mammary tumors only occur in MMTV-COX-2 transgenic mice after multiple pregnancies.15 Liu et al.15 suggested that mammary tumorigenesis in these mice is mainly related to the resistance of their mammary epithelial cells to apoptosis. Our results suggest that an inhibition of differentiation during pregnancy and lactation in these cells may also contribute to mammary tumorigenesis in these animals.
We reasoned that the alterations in proliferation, apoptosis and differentiation induced by COX-2 may predispose the breast epithelial cells to malignant transformation. After ∼20 passages in culture, the 10F-AS and 10F-P cells showed no morphological changes. After the same number of passages, however, the MCF-10F-S cells showed fibroblast-like features and formed foci of dense growth when cultured at confluence. Furthermore, at this time, all 3 cell lines expressed the epithelial markers E-cadherin and keratin, but only 10F-S cells also expressed vimentin, a fibroblastoid marker. These findings suggest that 10F-S cells were undergoing epithelial to mesenchymal transition (EMT), a process that characterizes progression towards a malignant phenotype.43, 44 The 10F-S cells, however, were unable to grow in soft agar and did not form tumors in nude mice showing that they are not fully transformed. We have now cultured the 10F-S cells for an additional 10 passages and observed no further changes. Our results suggest that COX-2 overexpression in breast epithelial cells induces partial transformation. Phenotypic changes in mammary epithelial cells, expressing the int-1 oncogene similar to those we observed, were also interpreted as evidence of partial transformation.45 It is possible that COX-2-induced suppression of growth associated with an increased resistance to apoptosis confers a survival advantage such that repeated passaging of the cells selects additional genetic/epigenetic changes that lead to EMT. Thus, EMT may be an indirect consequence of COX-2 expression and, indeed, this may be how COX-2 promotes mammary tumorigenesis.
In summary, we have shown that overexpression of COX-2 in human breast epithelial cells inhibits proliferation, apoptosis and differentiation. Since these parameters are normally tightly regulated, COX-2 expression may disrupt homeostasis, thereby predisposing the gland to carcinogenesis. Furthermore, after many passages in culture, the cells overexpressing COX-2 displayed evidence of partial transformation. During pregnancy, lactation and involution, when mammary epithelial cells undergo cycles of proliferation, apoptosis and differentiation, expression of COX-2 may have more profound effects than at other times. This may explain why mammary tumors occur only after multiple pregnancies in MMTV-COX-2 transgenic mice.15 Although a full-term pregnancy is normally protective for breast cancer development,46 our observations suggest that, as in mice, overexpression of COX-2 in the breast epithelium in women during pregnancy may be a risk factor. It is clear that control of COX-2 levels in the mammary gland at all stages may be important for breast cancer prevention.
MCA is the recipient of a Natural Sciences and Engineering Research Council of Canada Industrial Research Chair and acknowledges support from the member companies of the Program in Food Safety, Nutrition and Regulatory Affairs (University of Toronto).
- 42Susceptibility of rats to mammary gland carcinogenesis by the food-derived carcinogen 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) varies with age and is associated with the induction of differential gene expression. Am J Pathol 2004; 165: 191–202., , , .