Background
The processes of invasion and metastasis that cause mortality in patients are extraordinarily distinctive features of breast cancer progression [
1]. Although lymph-node metastasis, large tumor size, and poorly-differentiated histopathological grade are commonly considered to be established prognostic markers related to metastasis [
2], distant metastasis still occurs in 20-30% of the patients with negative lymph-node involvement [
3]. Thus far, Human Epidermal Growth Factor Receptor 2 (HER-2/neu) [
4], c-myc [
5] and HOXB9 [
6] have emerged as predictors of the risk of metastasis in breast cancer. The aberrant expression of these factors may induce the expression of growth and angiogenic factors in tumors, leading to increased local concentrations of these factors within the tumor microenvironment and thus favoring tumor progression [
6]. Recently, a new genomic test (gene-expression profiling) has been suggested to predict the clinical outcome more accurately than the traditional clinical and pathological standards [
7,
8]. However, it is an open question as to whether this method will enter into the clinical routine for staging and grading [
9]. Although these new markers and methods have been implicated, the molecular mechanism of breast cancer metastasis remains far from being fully understood due to the heterogeneity of this cancer and represents a new prerequisite for developing better treatment strategies.
The polycomb (PcG) proteins constitute a global system with important roles in multi-cellular development, stem cell biology and cancer [
10]. B-lymphoma Moloney murine leukemia virus insertion region-1 (Bmi-1), a member of the PcG family of transcription repressors, has emerged as a Myc-cooperating oncogene in murine lymphomas [
11,
12]. Bmi-1 can not only lead human mammary epithelial cells (HMECs) to bypass senescence and immortalize, but it also can play a key role in human breast cancer [
13,
14]. Moreover, a significant correlation has been observed between Bmi-1 expression and axillary lymph node metastasis in invasive ductal breast cancer [
15]. These findings suggest that Bmi-1 could be involved in the carcinogenesis and metastasis of breast cancer. Although increasing evidence has shown that Bmi-1 expression is associated with unfavorable prognosis [
16,
17], other studies have not confirmed these findings [
18,
19]. Bmi-1 protein is detected in only 25% of African breast cancer patients and is associated with a low histological grade [
18]. Additionally, higher Bmi-1 mRNA expression has been observed in early-stage patients without lymph node metastasis [
20]. In contrast, up-regulation of Bmi-1 was shown to be associated with the invasion of nasopharyngeal carcinomas and to predict poor survival [
21]. In colon cancer and gastric cancer, Bmi-1 expression is significantly correlated with nodal involvement, distant metastasis and clinical stage [
22‐
24]. Furthermore, metastatic melanoma tissues and cell lines show much higher expression of Bmi-1 than primary melanoma tissues and cell lines [
25]. Furthermore, knockdown of Bmi-1 contributes to decreased invasiveness of cervical cancer cells and gastric cells [
4,
26]. These findings indicate that Bmi-1 contributes to increased aggressive behavior of cancer cells. Bmi-1 overexpression can promote epithelial-mesenchymal transition (EMT) in NPECs, whereas Bmi-1 knockdown can reverse EMT and reduce the metastasis of nasopharyngeal carcinoma cells (NPC) [
27]. Although Bmi-1 overexpression alone did not result in oncogenic transformation of MCF-10A cells, overexpression of Bmi-1 together with H-Ras did induce an aggressive and metastatic phenotype, with the unusual occurrence of breast cancer brain metastasis [
28]. In spite of the aforementioned link between Bmi-1 and cancer, very few studies have focused on the molecular mechanism and clinical outcome of Bmi-1 in breast cancer metastasis.
The metastasis of cancer is a complex and multi-step process, including a series of successive and dynamic events along with alterations to cell morphology and biological function [
29]. After acquiring the ability to undergo EMT, cancers are prone to metastasize and establish secondary tumors at distant sites [
30,
31]. During EMT, epithelial cells acquire mesenchymal-like properties, which increase cell motility, and lose epithelial-like properties, which decrease intercellular adhesion [
32,
33]. Loss of E-cadherin is a hallmark of the invasive phase of cancer, and E-cadherin can be repressed by certain dominant transcriptional factors, such as Snail, ZEB, Twist, and basic Helix-Loop-Helix family proteins (bHLH) [
34‐
36]. Snail-induced EMT is an important breakthrough in the study of metastasis, providing new insights into the molecular mechanisms of tumor invasion [
37,
38]. Moreover, Snail expression is associated with E-cadherin repression and metastasis in breast cancer cells, as well as in other cancer cell types [
39‐
43]. In addition to Snail, numerous agents are involved in breast cancer EMT, such as components of the Six1, YB-1 and miRNA-200 families [
44‐
46]. Therefore, it is important to understand whether Bmi-1 can regulate EMT during breast cancer progression and metastasis.
The present study focuses on the expression patterns and roles of Bmi-1 in breast cancer tissues and cells to investigate the involvement of Bmi-1 in breast cancer metastasis. We demonstrate that Bmi-1 not only is increased in breast cancer tissues compared with adjacent non-cancerous tissues but also is associated with clinical features, such as tumor size, lymph node involvement, distant metastasis and clinical stage. High Bmi-1 expression predicts an unfavorable patient prognosis and serves as a high risk indicator in breast cancer. Furthermore, we also shed light on the biological impact of Bmi-1 on the invasive and metastatic properties of breast cancer cells. The overexpression of Bmi-1 enhances the motility and invasiveness of immortalized HMECs, facilitates concurrent EMT-like molecular changes, and promotes the stabilization of Snail and the activation of the Akt/GSK3β pathway. Consistent with these observations, repression of Bmi-1 in highly metastatic breast cancer cells markedly reduces cell motility and invasion, as well as tumorigenesis and lung metastases in nude mice. In addition, repression of Bmi-1 reverses the expression of EMT markers and inhibits the Akt/GSK3β pathway. Taken together, these results provide evidence that breast cancers expressing Bmi-1 exhibit aggressive and metastatic properties.
Discussion
Breast cancer, a common malignant disease in women, is prone to invade into adjacent regions and to metastasize to lymph nodes and distant organs. To develop novel treatments and cures, it is imperative to address the factors underlying tumorigenesis, invasion and metastasis. In this study, we identified and functionally characterized Bmi-1 as an important player in breast cancer progression. The current study first illustrated the expression of Bmi-1 in primary breast cancer tissues, followed by demonstrating the association between the Bmi-1 expression and clinicopathologic parameters and finally addressed the role of Bmi-1 in breast cancer prognosis in a large series of 252 samples. In this study, differential expression of Bmi-1 was detected between primary cancer tissues and the matched adjacent non-cancerous tissues. Bmi-1 expression was significantly up-regulated in breast cancer tissues compared with the adjacent non-cancerous tissues, which was echoed by our previous reports [
15,
50]. Only 35.9% of matched adjacent non-cancerous tissues displayed the high Bmi-1 expression, whereas as many as 72.2% of the cancer tissues were defined by high Bmi-1 expression. The incidence of high Bmi-1 expression shown in this study was much higher than previously demonstrated (53.2%) [
19]. The difference may reflect differences in Bmi-1 status in the samples used in different studies, which obtained tissue samples from patients with different disease stages, or samples from different populations. As shown in Additional File
2, table S2, 46.4% of the Chinese samples recruited in our study were from the early stage (I,II), while 75.8% of Korean samples used in Choi's study were from the early stage (I,II)[
19]. In addition, no samples of stage □ were recruited in Choi's study [
19]. Furthermore, the difference may come from variations in antigen retrieval, antibody dilution, development time and the positive criteria adjusted, especially the score of the positive number (Additional File
3,
4,
5, Table S3,S4,S5). For example, we used EDTA buffer to retrieve antigen in our research, as opposed to citrate buffer used in Choi's study [
19]. In addition, the development time in our study was 10 min compared with 5 min in Choi's study [
19]. The criteria used in the immunohistochemistry varied in different studies. Choi
et al. tested Bmi-1 expression by tissue microarray, which might not be a good representative of the whole paraffin-embedded tissue. Furthermore, cells were considered positive for Bmi-1 only when nuclear staining was observed [
19]. However, both nuclear and cytoplasm staining were observed in our samples. However, to further confirm Bmi-1 expression in breast cancers, multi-center studies are required.
We also compared the Bmi-1 mRNA expression in 34 paired tissues, including breast cancer and matched adjacent non-cancerous tissues. There was a significant differential between the breast cancer tissues and the adjacent non-cancer tissues, which corresponded to the protein levels in the tissues. In our analysis, high Bmi-1 expression showed an obvious correlation with larger tumor size, lymph node involvement, organ metastasis and advanced clinical stage. Over 90% of the late-stage (stages III/IV) samples showed high expression, whereas less than 50% of the early-stage (stagesI/II) samples showed high expression. These results revealed that higher Bmi-1 expression was related to more aggressive behavior, which was further supported by its expression in distant metastases. Of patients without distant metastasis, 67.1% showed high Bmi-1 expression, which is in contrast to 95.6% of patients with distant metastasis. These studies indicated that a high level of Bmi-1 protein might contribute to the invasion and progression of breast cancer. Although only 35.9% of the adjacent non-cancerous tissues displayed high Bmi-1 expression, 69.2% stained positive for Bmi-1. Because hyperplasia is known to occur often in adjacent non-cancerous tissues, we speculated that the Bmi-1 staining had originated from hyperplasia.
In this study, the Bmi-1 protein seemed to localize in the nucleus of the breast cancer cells and in the cytoplasm of the non-cancer cells. It has been reported that phosphorylation can explain differential subcellular localization of some of the polycomb family genes, such as Nervous System Polycomb 1(NSPc1) and M33 [
52,
53]. Previously, it has been reported that there is a rich proline/serine region at the carboxyl terminus of the Bmi-1 protein, where phosphorylation often occurs [
54]. However, whether the phosphorylation of Bmi-1 is a direct cause or merely associated with the nuclear-cytoplasm shuttling events remains to be determined. It is also important to note that Bmi-1 predicted poor prognosis in breast cancer, in accordance with other reports [
21,
24,
55‐
58]. In addition, there were significantly different outcomes between the patients expressing high and low levels of Bmi-1 by subset analysis, which suggested that Bmi-1 may be used to predict the clinical outcome. In addition, it provides a potential therapeutic target for the future treatment of breast cancer.
In our study, Bmi-1 was not significantly correlated with ER and PR expression, which is consistent with a previous report indicating that Bmi-1 mRNA expression had no significant correlation with ER or PR expression [
20], but it is inconsistent with other previously-published data [
15,
18,
59]. Nevertheless, a statistically significant association was observed between Bmi-1 expression and survival when ER or PR was included in our analysis. ER
+ cancer cells depend on estrogen for their growth, so they can be treated with drugs that block the effect of estrogen. Patients with ER present were offered adjuvant hormone therapy in our study. High Bmi-1 expression was associated with unfavorable survival, irrelevant to ER or PR presence, indicating that hormonal therapy did not affect the prognostic role of Bmi-1. Patients with metastatic breast cancer may take tamoxifen for varying lengths of time, depending on the cancer's response to this treatment and other factors [
60]. When used as adjuvant therapy for early-stage breast cancer, tamoxifen is generally prescribed for five years. However, the ideal length of treatment with tamoxifen is not known. Thus, different lengths of treatments with tamoxifen depending on the individual responses to it may have resulted in a different outcome in our study. In addition, the combination of endocrine therapy with other therapy remains a research issue [
61]. Generally, patients who are positive for HER-2 had a worse prognosis [
62]. However, these 252 patients did not receive anti-HER-2 therapy because Herceptin (a humanized monoclonal antibody directed at the HER2 ectodomain) was not in use at that time in China. Although there was no significant correlation between Bmi-1 expression and HER-2 status, patients with high Bmi-1 expression showed poor survival stratified by HER-2, just as with ER and PR. Having demonstrated the significance of Bmi-1 in the overall survival prognosis of breast cancer patients, it will be our next focus to investigate the prognostic value of Bmi-1 in terms of disease-free survival and cancer-specific survival. Additionally, a possible correlation between Bmi-1 expression and outcome after hormonal therapy and chemotherapy warrants investigation and would require a large number of samples.
Metastatic relapse remains a major challenge in breast cancer management. Many factors are involved in tumor progression, including changes in cell adhesion, cell communication, increased migration or motility and invasiveness [
63]. In this study, Bmi-1 was shown to contribute to each of these events in clinical samples and in cell lines. To address the role of Bmi-1 in tumor progression, Bmi-1 was overexpressed in two immortalized HMEC lines, 76N-TERT and MCF-10A. Conversely, RNA interference was used to decrease the expression of Bmi-1 in MDA-MB-435S, an estrogen-independent breast cancer cell line derived from a mammary ductal carcinoma [
64]. MDA-MB-435S cells can form progressively growing tumors in the lungs and regional lymph node metastases following injection into the mammary fat pad of 3-4 week old athymic nude mice [
65]. In this study, the overexpression of Bmi-1 alone could not fully transform 76N-TERT or MCF-10A cells. Furthermore, Bmi-1 expression did not alter the morphology of these cells in 3-D Matrigel culture. In addition, the spindle-shaped phenotype and non-contact inhibited, disorganized proliferation of MDA-MB-435S cells [
66] was not altered by the inhibition of Bmi-1. This result was in accordance with a previous observation that Bmi-1 alone did not result in an EMT phenotype in MCF-10A cells, but that co-overexpression of Bmi-1 and Ras readily did [
28]. Additionally, we examined the potential oncogenic role of Bmi-1 by the injection of Bmi-1-expressing MCF-10A cells into mice. Even injection of 1 × 10
7 MCF-10A/Bmi-1 cells did not result in tumor formation after two months in nude or SCID mice. Unlike in immortalized NPECs, Bmi-1 alone was not sufficient to induce the typical EMT morphological changes in immortalized HMECs. The induction of morphological alterations associated with EMT by Bmi-1 might depend on the cell type. To our knowledge, the immortalized NPECs were derived from squamous epithelium, whereas the immortalized HMECs originated from glandular epithelium. In addition, the morphologic changes of EMT might be directed by differential oncogene activation. Ras [
67] and ILEI [
68] can lead to EMT, tumor formation and metastasis. These results suggest that additional oncogenic events, such as H-Ras expression or loss of expression of tumor suppressor genes could be involved in the EMT of immortalized HMECs induced by Bmi-1. Thus, we suggest that Bmi-1-induced EMT is cell-type specific.
One thing worth mentioning is that although E-cadherin, a useful molecule to protect breast cancer from metastasis [
69], was not detected in MDA-MB-435S cells, the MDA-MB-435S/shBmi-1 cells still manifested reduced motility. To our knowledge, several highly metastatic cancer cells, including MDA-MB-435S cells, lack E-cadherin expression [
70]. Low E-cadherin expression can be caused by gene mutations or promoter methylation [
71,
72], as well as by regulation by inhibitors such as Twist [
73]. After EMT, mesenchymal FosER cells completely lacked E-cadherin but formed neither tumors nor metastases [
74], indicating that loss of E-cadherin expression might be necessary but not sufficient for tumor progression. Similarly, although E-cadherin expression was decreased by Bmi-1 overexpression, the HMECs did not form tumors in the current study. As we know, besides E-cadherin, many other genes are involved in breast cancer metastasis, such as β-Catenin [
75] and N-cadherin [
76]. Numerous studies have linked aberrant E-cadherin with the development of metastasis in cancer [
77], whereas other studies have presented different results indicating that cells from distant metastases and nodal involvement consistently expressed E-cadherin, often at higher levels than in the primary tumor [
78,
79]. It appears that translational regulation and post-translational events are probable mechanisms for E-cadherin re-expression [
80]. It is possible that loss of E-cadherin is a transient phenomenon that allows malignant cells to invade vascular channels and tissues. Disseminated mesenchymal cancer cells seem to undergo the reverse transition, mesenchymal-epithelial transition (MET), at the metastatic site to allow micrometastases to give rise to a secondary neoplasm. In this regard, cancer cells from the secondary site re-express markers of epithelial cells such as E-cadherin. However, whether re-expression of E-cadherin occurs in Bmi-1 overexpressing cancer cells in metastatic site, and if so, what is the underlying mechanism requires further investigation. In addition, our data suggest that Bmi-1 has a critical effect on breast cancer tumorigenesis and lung metastasis. We believe that this is an extremely important observation in terms of studying breast cancer lung metastasis because the lung is the most common location of breast cancer metastasis. We suggest that Bmi-1 contributes to the metastasis of breast cancer. Crosstalk between different pathways, recognized as a mechanism for expanding the cellular communication signaling network, is currently receiving increased attention. The activated PI3K/Akt pathway has been well documented in various human malignancies and sometimes correlates with an aggressive phenotype [
81]. Our previous data also indicated that down-regulation of Bmi-1 by an RNA interference (RNAi) approach was accompanied by down-regulation of Akt/protein Kinase B (PKB) activity [
50]. In our current study, we demonstrated that Bmi-1 induced invasion, which might be associated with activation of the Akt pathway in breast cancer cells. As metastasis can occur in early stages of tumor development, some genes may constantly regulate tumor development. They may not only facilitate primary tumor initiation but also promote tumor transformation and metastasis [
82]. The expression pattern of Bmi-1, together with functional studies, indicate that Bmi-1 plays a prominent role in breast cancer progression and metastasis and opens the door for future studies addressing Bmi-1-targeted therapy in breast cancer.
Materials and methods
Tissue Samples
Paraffin-embedded breast cancer samples were obtained from 252 female Chinese patients (median age: 47 years, range: 26-78 years) diagnosed with breast cancer in 1999-2001 at Cancer Center, Sun Yat-sen University, Guangzhou, China. Of the 252 breast cancer samples, 39 matched adjacent non-cancerous tissues were obtained from the above-mentioned patients. All of the samples were treated by surgical excision. Among them, 239 cases had follow-up records and the median follow-up time was 59 months (range: 4-78 months). Clinical and pathologic factors were evaluated, including age, TNM classification, clinical stage, presence of steroid receptors and HER-2 expression. HER-2 expression was only analyzed in 159 cases, while the other samples were not analyzed. A total of 147 cases were analyzed for ER, PR and HER-2 expression. Thirteen cases were missing records of patient age, survival time and survival status, but included the TNM classification, clinical stage, presence of steroid receptors and HER-2 expression. To use these clinical materials for research purposes, prior patients' consent and approval from the Institute Research Ethics Committee were obtained. The observation period was from 1999 to 2006. The clinical stages of all the patients were classified according to the 2002 TNM staging of UICC (International Union against Cancer).
Immunohistochemistry in Clinical Samples
The 4 μm paraffin-embedded sections of breast cancer were deparaffinized with xylene, rehydrated and treated with 3% hydrogen peroxide in methanol to quench the endogenous peroxidase activity. Subsequently, antigen retrieval was performed by heating in a microwave oven with EDTA (pH 8.0). One percent bovine serum albumin (BSA) was used to block non-specific binding, followed by incubation of the sections with a mouse monoclonal anti-Bmi-1 antibody (1:100, Upstate Biotechnology, Lake Placid, USA) overnight at 4°C. After washing with phosphate buffered saline, sections were incubated with biotinylated secondary antibody, followed by a further incubation with the streptavidin-horseradish-peroxidase complex. The sections were then immersed in 3, 3'-Diaminobenzidine (DAB) for 10 min, counterstained with 10% Mayer's hematoxylin, dehydrated, and mounted in crystal mount. The primary antibody was replaced by non-immune mouse IgG of the same isotype to serve as negative controls. To minimize variations in the immunopositive cells, all sections were stained in DAB for the same amount of time. Two pathologists, blinded to the clinical outcome, scored the results of the staining independently. Measurements of ER, PR and HER-2 were routinely performed as previously described [
83].
Cell lines, Vectors and Plasmids
Immortalized HMECs (76N-TERT and MCF-10A) and radiation-transformed cells (76R-30) were cultured in Keratinocyte-SFM medium (Invitrogen, Grand Island, NY) supplemented with bovine pituitary extract. MDA-MB-435S cells were maintained in DMEM/F12 (Gibco) supplemented with 10% fetal calf serum (HyClone). SK-BR-3, ZR-75-30 and BCAP-37 cells were grown in RPMI 1640 with 10% fetal calf serum. The pMSCV-Bmi-1 and Bmi-1 short hairpin RNA (shRNAi) constructs were generated as described previously [
21,
51]. Retrovirus expressing Bmi-1 was produced and transfected into 76N-TERT and MCF-10A cells, as described previously [
13]. The plasmid with shBmi-1 was introduced into MDA-MB-435S cells, which showed strong ability to metastasize [
48]. The sequences of shRNA were as follows: shBmi-1 1# GUUCACAAGACCAGACCAC and shBmi-1 2# GACCAGACCACUACUGAAU [
51]. pMSCV and PRS plasmids were used as controls. All retrovirally infected cells were maintained under Puromycin selection and used as stable cells.
RT-PCR, Real-time PCR and Western Blot Analysis
Total RNA from fresh tissues and cell lines was isolated using Trizol reagent (Invitrogen, Grand Island, NY), according to the manufacturer's instructions, and 1.0 μg of total RNA treated with DNAase was used for cDNA synthesis by random hexamers. Genes were amplified by PCR from cDNA. The forward primer for Bmi-1 was 5-'CTGGTTGCCCATTGACAGC'-3, the reverse primer was 5-'CAGAAAATGAATGCGAGCCA'-3. The forward primer for GAPDH was 5-'AGCCGTTCGGAGGATTATTCG'-3, the reverse primer was CTTCTCCTCAGCAGCCAGAG. The products were analyzed by agarose gel electrophoresis and confirmed by appropriate size.
Real-time PCR was carried out using an ABI PRISM 7500 Sequence Detection System (Applied Biosystems, Foster City, CA). Reactions were performed in triplicate repeats in two independent experiments. The geometric mean of the GAPDH (glyceraldehyde-3-phosphate dehydrogenase) housekeeping gene was used as an internal control to normalize the variability in expression levels. The forward primer for Bmi-1 was 5-'CTGGTTGCCCATTGACAGC'-3, the reverse primer was 5-'CAGAAAATGAATGCGAGCCA'-3 and the probe was FAM-CAGCTCGCTTCAAGATGGCCGC-TAMRA. The forward primer for E-cadherin was 5-'GAACAGCACGTACACAGCCCT'-3, the reverse primer was 5-'GCAGAAGTGTCCCTGTTCCAG'-3 and the probe was FAM-ATCATAGCTACAGACAATGGTTCTCCAGTTGCT-TAMRA. The forward primer for GAPDH was 5-'GACTCATGACCACAGTCCATGC'-3, the reverse primer was 5-'AGAGGCAGGGATGATGTTCTG'-3 and the probe was FAM-CATCACTGCCACCCAGAAGACTGTG-TAMRA.
Immunoblotting was carried out as described [
84]. The blots were probed with mouse anti-Bmi-1, anti-E-cadherin, anti-β-catenin, anti-fibronectin and anti-vimentin antibodies (BD, Transduction Laboratories, Lexington, UK) as well as with rabbit anti-p-GSK, anti-t-GSK (Cell Signaling Technology, Inc. USA), anti-snail (Abcam, Cambridge Science Park, Cambridge, UK), anti-p-Akt (Santa Cruz Biotechnology, CA, USA) and goat anti-t-Akt (Santa Cruz Biotechnology, CA, USA) antibodies. The membranes were stripped and re-probed with mouse anti-α-tubulin (Sigma Aldrich, Inc. St Louis, Missouri, USA) to confirm equal loading of the samples.
Wound Healing Assay
Cells were seeded in six-well plates and cultured under permissive conditions until 90% confluence. After starving the cells for 24 h in medium without EGF or FBS, the confluent cell monolayer was lightly and quickly scratched with a pipette tip to produce a straight line. The debris was removed and the edge of the scratch was smoothed with PBS washing. The wound healing assays were done in growth factor-free medium, further excluding any effect due to a potential proliferation difference. The open gap was then inspected and photographed microscopically at indicated times, and is shown in the Figures at a 200X magnification. The migration activity was calculated as the number of cells entering into the rectangle. Experiments were repeated a minimum of three times.
Proliferation Assay
1 × 105 cells were plated on a P60 plate. Every 24 h, cells were trypsinized and counted under a light microscope at least three times until the sixth day. Experiments were repeated a minimum of three times.
Boyden Chamber Assay
This assay measures the ability of cells to invade a Matrigel matrix overlying a membrane containing 8-μm pores. Cells were seeded in medium deprived of EGF or FBS in the top chamber (BD), whereas medium containing EGF or FBS was added to the bottom chamber. After an appropriate cultivation time, the chambers were fixed with 1% paraformaldehyde and stained with hematoxylin. The number of cells in ten random fields of view was enumerated at 200X or 400X magnification for each filter. Three independent experiments were performed and the data are presented as the mean ± SD.
Three Dimensional Matrigel Culture
Matrigel (1.2 mg/ml, BD) was coated on the bottom of a 24-well plate. After Matrigel polymerization, cells were seeded into the well with growth medium containing 2% Matrigel. The cells were cultivated at 37°C incubation and alterations to the morphologic phenotype were monitored at 200X magnification every other day. Experiments were repeated a minimum of three times.
Anchorage-Independent Growth in Soft Agar
The soft agar assay was used to determine the propensity for anchorage-independent growth. Cells were plated in a 60-mm dish using 2 ml of growth medium, including 0.33% agar on the top of a bottom layer containing 0.66% agar. The cells were fed every two days with 1 ml medium. Colonies were photographed and counted in ten random fields of view at 200X magnification using light microscopy. Each experiment was done in triplicate.
Confocal Immunofluorescence Microscopy
Cells were seeded onto glass slides for 24 h, washed with PBS, fixed in 4% paraformaldehyde and permeabilized with 0.5% Triton X-100 for five minutes. After blocking with BSA, cells were stained with anti-snail primary antibody followed by FITC-conjugated anti-rabbit IgG. To visualize the nucleus, 4' 6-Diamidino-2-phenylindole (DAPI) staining was also performed, as previously described [
85]. Immunofluorescence was detected by fluorescence microscopy (Olympus).
Mouse Injections, Necropsy, Histopathology
The ability to form tumors and metastasize was analyzed by injecting cells with repressed Bmi-1 into nude mice. Mice were bred and maintained under SPF conditions in the Department of Animal Center, Cancer Center, Sun Yat-Sen University, as approved by the China Care Committee Institute. Ten healthy female nude mice, which were four- to six- weeks old, were randomly assigned to each group. Each mouse was injected in the fat pad with 2 × 10
6 cells in PBS solution. Tumor growth was measured by caliper, and tumor volume was calculated according to the formula: length × width
2 × 0.52, as described previously [
86]. All mice were sacrificed on the sixth week after injection. The primary tumor and lung tissues of each mouse were removed, weighed and embedded in 10% paraffin. Each tissue was chopped into small pieces. Total protein was extracted to detect Bmi-1 expression from the primary xenografts. Each section from the primary xenografts and lung tissues was subjected to H&E staining, according to standard protocols, for histological examination and metastasis evaluation. The nodes of lung metastasis were quantified by counting metastatic lesions in ten sections (10 μm per section in a series). Data were collected by counting the total numbers of metastatic lesions from ten sections. Sections of primary tumors and lung lesions were used to detect the expression of the markers (Bmi-1, β-catenin, fibronectin) by IHC, as described previously.
Statistical analysis
The Chi-Square test was employed to evaluate the differences in Bmi-1 expression between the two categories of tissues. For assessment of the correlation between clinical features and Bmi-1 expression in breast cancer, P values were calculated by using the Chi-Square test or the Fisher's exact test. Relative risks (RRs) of death associated with Bmi-1 expression and other predictive variables were estimated by using the univariate and multivariate Cox proportional hazards model. The overall survival curve was plotted using Kaplan-Meier survival analysis and compared by the log-rank test. Result variations for the chamber invasion assays, wound healing assay, soft agar assay, tumor volume, tumor weight and lung metastasis lesions in mice, described as mean ± SD, were assessed using the two-tailed Student's t test. A value of P < 0.05 was considered significant (two tailed) by using SPSS 16.0.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
BHG and YF were involved in the design, performed the experiments, and drafted the manuscript. RZ, LHX, MZL participated in sample collection. HFK and LBS reviewed the manuscript. MSZ conceived the idea for the study, contributed to the overall experiment design and revised the manuscript. All authors read and approved the final manuscript.