Background
Gastric cancer, with highly aggressivity, is one of common diagnosed and the leading causes of cancer death, so novel biomarker effective therapeutic strategies are being urgently sought [
1,
2]. As an oncogene, overexpression of B cell-specific Moloney murine leukemia virus integration site 1 (Bmi-1) is present in a wide variety of tumors and associated with poor prognosis [
3,
4]. Bmi-1 may also be involved in cancer metastasis and treatment resistance in some kinds of cancer [
5‐
7]. Our previous studies have found that Bmi-1 expression in gastric cancer is associated with lymph node metastasis and clinical stage and is an independent prognostic factor in patients with gastric cancer [
8,
9]. Importantly, Bmi-1 plays an essential role in maintaining self-renewal of hematopoietic stem cells [
10,
11]. It is regarded as one of the stem cell markers and a crucial regulator of prostate stem cell self-renewal and malignant transformation [
12]. In addition, it is also involved in self-renewal of breast cancer and glioma cancer stem cells (CSCs) [
13,
14]. However, the underlying mechanisms of Bmi-1 regulating stemness remain unclear.
CSCs are a small proportion of cells with stem cell features in tumor tissues and considered as a source of tumor formation [
15,
16]. Although CSCs only account for a very small fraction of cancer cells, they demonstrate important features including self-renewal, resistance to treatment, and high metastasis potential, all of which make them the source of cancer relapse and treatment failure [
17,
18]. Researchers have isolated subgroups of stem cells from many types of tumors. The investigation on gastric CSCs was initiated later than that on other cancers. Scholars have preliminarily isolated gastric CSCs or stem cell-like cells from gastric cancer cell lines or tissues [
19‐
22]. Takaishi et al. conducted serum-free culture of human gastric cancer cell lines to obtain microspheres and found them possessing CSCs properties [
19]. CD44
+CD24
+ and EpCAM
+/CD44
+ cell subgroups have also been confirmed to have CSCs properties [
21]. Recently, Chen et al. isolated CD44
+CD54
+ cells subgroup from gastric cancer tissue and peripheral blood of patients with gastric cancer and found that this cell subgroup had self-renewal ability and formed transplanted tumor which had features very similar to those of the original tumor [
22]. All these studies have shown the existence of gastric CSCs. Side population (SP) cells isolated from gastric cancer cells and gastric cancer tissues also have CSCs features [
23,
24]. Bmi-1 was found to be overexpressed in SP cells [
24]. However, there are also studies showing no increase of Bmi-1 expression in gastric CSCs [
19]. Therefore, the results of Bmi-1 expression in gastric CSCs are still conflicting, and further research is required to clarify whether Bmi-1 plays a role in regulation of stemness in gastric cancer.
In this study, we intended to explore the role and mechanisms of Bmi-1 in regulating stem cell-like features of gastric cancer. Here, we show that Bmi-1 positively regulates stem cell-like properties via upregulating miR-21, and miR-34a negatively regulates stem cell-like characteristics by negative feedback regulation of Bmi-1 in gastric cancer.
Methods
Details regarding cell strains and cell culture methods, vectors construction and virus infection, spheroid colony formation assay, chemo-sensitivity experiment, cell migration assay, Western blot, immunohistochemical analyses(IHC), immuno-fluorescence staining, Quantitative real time RT-PCR (QRT-PCR), chromatin immunoprecipitation(ChIP), dual fluorescence report assay, and in vivo tumorigenesis are provided in the supplementary materials and methods (see Additional file
1).
Clinical samples and analyses
The expression of Bmi-1 in 101 paraffin-embedded primary site specimens of gastric cancer and 72 ovarian metastatic specimens originated from gastric cancer, and the expression of Oct4, Sox2, Gli1, CD44, and CD133 in 101 primary site specimens of gastric cancer was tested using IHC.
Another 74 fresh gastric cancer tissues and paired normal mucosal tissues were used to detect the expression of Bmi-1 messenger RNA (mRNA) and miRNAs by QRT-PCR. Details are provided in supplementary methods.
Statistics
All statistical analyses were done by using the SPSS 19.0 software package, and two-tailed P values of less than 0.05 were considered significant. In IHC assays of gastric cancer samples, Pearson χ
2 test was used to determine the correlation between Bmi-1 expression and clinicopathologic characteristics, and Spearman's Rank correlation assay was used to determine the correlation between Bmi-1 and stem cell markers expression. Among 21 pairs of samples, the matching McNemar test was used to detect the difference of Bmi-1 expression between primary and metastatic lesions. In QRT-PCR analysis of fresh tissues, the expression of Bmi-1, miR-21, or miR-34a was not normally distributed. Hence, the distribution was established by using Log10 and geometric averages. The correlation between Bmi-1 and miR-21/miR-34a expression levels was analyzed by the Pearson coefficient test. The correlation between Bmi-1, miR-21, or miR-34a expression and clinicopathologic characteristics was analyzed by ANOVA.
Discussion
Over the past two decades, research on CSCs has attracted great interest and also made great progress. As an oncogene and stem cell marker, Bmi-1 can maintain the self-renewal of CSCs in some tumors, including breast cancer and glioma, but whether Bmi-1 is implicated in the regulation of gastric CSCs is still unclear. In this study, we isolated stem cell-like cells by serum-free microsphere culture and found that Bmi-1 was overexpressed in microsphere cells. Furthermore, Bmi-1 overexpression increased microsphere formation rate, anti-cancer drug resistance, cell migration, and stem cell markers expression, while Bmi-1 knockdown decreased these parameters in GC cells Clinical samples analysis showed that Bmi-1 expression in GC tissues was associated with regional lymph node metastasis and distant ovary metastasis and positively correlated with the expression of stem cell markers. These results suggested that Bmi-1 positively regulates the stem cell-like characteristics of GC cells.
Bmi-1 is a transcriptional inhibitor and the earliest research found that Bmi-1 downregulated the expression of p16 and p19 by inhibiting INK4a/ARF and thus regulates cell proliferation and senescence [
29]. Douglas et al. found that Bmi-1 may promote the development of Ewing’s sarcoma through p16 independent mechanisms, and Bmi-1 knockdown induced expression changes of hundreds of downstream genes [
30]. We previously found that Bmi-1 can influence breast cancer cells proliferation and tumorigenicity through regulating pAKT [
31]. Song et al. found that Bmi-1 may inhibit PTEN expression, then activate PI3K/AKT/Snail pathway and therefore regulate EMT and metastasis of nasopharyngeal cancer cells [
41]. These results indicated that Bmi-1 may play its role by regulating many downstream target genes. However, the mechanisms of Bmi-1 have not been fully elucidated, especially in regulating stemness.
miRNAs plays important roles in the process of cell proliferation, apoptosis, and carcinogenesis by interacting with mRNAs, IncRNAs, and other endogenous RNAs [
43,
44]. Recently, some studies suggested that miRNAs were involved in regulation of CSCs. We speculated that Bmi-1, as an exogenous gene-silencing factor, might also regulate the expression of a variety of genes via miRNAs and found that miR-21, miR-34a, and miR-125a-5p were regulated by Bmi-1. Of these three miRNAs, miR-21 was most significantly affected by Bmi-1. In gastric cancer tissues, miR-21 had a significantly positive correlation with Bmi-1 expression. It was reported that miR-21 is highly expressed in many kinds of malignant tumor tissues and might act as an oncogene [
45]. In breast cancer and colorectal cancer, miR-21 is involved in regulation of EMT and stemness [
46,
47]. In gastric cancer, overexpression of miR-21 promotes cell growth, invasion, drug resistancem and EMT by inhibiting PTEN and P53 [
48,
49]. It has also been found that miR-21 expression in gastric CSCs is higher than in parental cells [
50]. In our study, in vitro experiments found that miR-21 was highly expressed in microsphere which enrich stem cell-like cells, and it may enhance the self-renewal, drug resistance, migration, and stem cell markers expression in GC cells, and clinical sample investigation found that miR-21 was highly expressed in GC tissues and positively correlated with lymph node metastasis and nerve invasion. These results suggested miR-21 positively regulates stem cell-like properties of GC cells. Further, we conducted co-transfection to simultaneously change the expression of Bmi-1 and miR-21 and found that upregulation of miR-21 restored the stem cell-like characteristics of GC cells inhibited by Bmi-1 knockdown; and downregulation of miR-21 inhibited the enhancement of stem cell-like characteristics induced by Bmi-1 overexpression. Therefore, it confirmed that miR-21 may mediate the function of Bmi-1 in regulating the stem cell-like characteristics of GC cells. Since miR-21 may regulate the expression of a variety of downstream target genes, Bmi-1 may form a complex regulatory network via miR-21. We tested two important downstream target genes of miR-21, p53 and PTEN, and did find Bmi-1 may regulate the downstream p53 and PTEN-AKT pathways via miR-21. Previous studies have shown that Bmi-1 may directly bind to PTEN promoter in nasopharyngeal carcinoma cells [
41]. In this study, we also found that Bmi-1 binds to PTEN promoter in GC cells, suggesting Bmi-1 may directly inhibit the expression of PTEN. Taken together, these results suggested that in addition to the direct effect, Bmi-1 may indirectly regulate PTEN-AKT via miR-21 therefore forming bypass pathways to enhance its activating effect on AKT.
MiR-34a acts as a tumor suppressor and is downregulated in some kinds of cancer [
51,
52]. It negatively regulates stem cell-like characteristics of glioma cells through downregulating c-Met and NOTCH [
53] and inhibits the growth, invasion and metastasis of GC by targeting PDGFR and MET [
51]. In present study, we found that miR-34a is downregulated in GC tissues and negatively correlates with aggressive tumor phenotype. Furthermore, in vitro studies showed that miR-34a inhibited cells proliferation and microsphere formation, decreased drug resistance, and migration potential. These results suggested that miR-34a acts as a tumor suppressor, and negatively regulates stem cell-like characteristics in GC cells. However, why Bmi-1 upregulates miR-34a which has opposite functions? We conceived there might be a negative feedback pathway between Bmi-1 and miR-34a and Bmi-1 inhibit its own overexpression and functions by inducing miR-34a. It was reported that c-Myc is a target gene of miR-34a [
37], and c-Myc binds to t Bmi-1 promoter and upregulates its expression. Therefore, miR-34a may regulate Bmi-1 via c-myc. Our experiments showed that after miR-34a upregulation, Bmi-1 and c-Myc expression was inhibited, while c-Myc overexpression increased the inhibited expression of Bmi-1, confirming miR-34a downregulates Bmi-1 by targeting c-Myc. Further, we found that Bmi-1 overexpression restored stem cell-like characteristics inhibited by miR-34a upregulation. These results confirm our assumption that miR-34a negatively regulates stem cell-like characteristics of GC cells by negative feedback regulation of Bmi-1. The complicated negative feedback loop can also explain why we did not found the correlation between the expression of Bmi-1 and miR-34a in GC tissues. In addition, our study found a new mechanism and pathway by which miR-34 negatively regulates stemness through downregulating Bmi-1.
What is the mechanism of Bmi-1 to regulate miR-21 and miR-34a? Studies have shown that Bmi-1 regulates NF-kB activity by influencing its nuclear-plasma distribution [
39,
54], and NF-kB can binds to the promoter region of miR-21 and miR-34a [
40,
55] . Thus, we hypothesized that Bmi-1 may regulate miR-21 and miR-34a through activating NF-kB. In our study, reporter assay and western analysis confirmed that Bmi-1 may increase NF-kB aggregation in cell nucleus and activate its transcriptional activity. ChIP test confirmed the binding of NF-kB to miR-21 and miR-34a promoter, and Bmi-1 increases the binding. Furthermore, NF-kB inhibitors can inhibit the upregulation of miR-21 and miR-34a induced by Bmi-1 overexpression. These data confirmed Bmi-1regulates miR-21 and miR-34a expression via activating NF-kB. Further, as it was found that AKT can activate NF-kB, we suspected Bmi-1 may regulate NF-kB and miR-21/miR-34a via activating AKT. We found that AKT overexpression can really activate NF-kB and increase miR-21/miR-34a expression, while AKT inhibitor treatment can inhibit the increased NF-kB activity and upregulation of miR-21 and miR-34a induced by Bmi-1 overexpression, indicating that Bmi-1 upregulates NF-kB activity and miR-21/miR-34a expression via activating AKT.
Acknowledgements
The authors deeply thank Professor Bing-ya Liu for providing the GC cells, Professor Dimri GP for providing the Bmi-1, Bmi-1 shRNA, and pSRa-mAkt plasmid and Professor Li Jun for providing the pNF-kB-luc constructs.