Background
Hepatocellular carcinoma (HCC), which includes over 90 % of liver cancers, is the leading cause of cancer mortality worldwide [
1]. Because HCC is resistant to chemoradiotherapy and has a high recurrence and metastatic rate after surgery, HCC patients have a very low five-year survival rate [
2,
3]. In recent years, accumulating evidence has suggested that in HCC, a population of cells with stem cell-like features, known as cancer stem cells (CSCs) or tumor-initiating cells (TICs), is essential for the recurrence, metastasis and resistance to chemoradiotherapy seen in HCC [
4]. Previous studies have shown that CD90, EpCAM, CD133, CD24, OV-6 and CD44 can be used as CSC markers in HCC, and cells expressing these markers possessed CSC characteristics such as self-renewal, tumor generation and aggressive growth [
5‐
10]. However, the optimal marker for identifying CSC populations in HCC remains disputed.
MicroRNAs (miRNAs) are a large group of small, noncoding RNAs that are important regulators of post-transcriptional gene expression. miRNAs bind to the 3'-untranslated region (UTR) of the target mRNA by base pairing and suppress protein expression through translational repression or mRNA degradation [
11‐
13]. miRNAs play key roles in a variety of physiological and pathological processes, including embryo development, cellular differentiation, stem cell self-renewal and tumor progression [
14]. In human solid tumors, many miRNAs have been found to participate in CSC maintenance [
15‐
17]. To date, in HCC, miR-130b has been shown to promote CD133
+ CSC tumorigenicity and self-renewal [
18], whereas miR-181 inhibition reduces the number of EpCAM
+ CSCs and tumor-initiating ability [
19]. However, it remains unclear which miRNA regulates the stemness of CD90
+ HCC CSCs.
Mitogen-activated protein kinase kinase kinase 8 (MAP3K8), also known as tumor progression locus 2 (TPL2) or cancer Osaka thyroid (COT), is a member of the serine/threonine protein kinase family. MAP3K8 activation is critically involved in inflammation and has variable effects on tumors [
20,
21]. In lung and intestinal cancer, MAP3K8 is a tumor suppressor gene [
22‐
24]. However, MAP3K8 is predominantly considered a tumor-promoting oncogene in several tumor types, such as breast cancer, colon cancer, renal cell carcinoma, endometrial cancer, gastric cancer, nasopharyngeal carcinoma, thymoma and lymphoma [
20]. MAP3K8 is up-regulated in multiple tumor types and is closely related to tumorigenesis and/or cancer progression [
25‐
27]. However, the role of MAP3K8 in HCC initiation and progression remains unknown. In this study, we found that miR-589-5p was down-regulated in CD90
+ CSCs and examined the effects of miR-589-5p expression and its target protein MAP3K8 on CD90
+ CSCs and the clinical outcomes of HCC.
Methods
Patients and tissue specimens
Tumor specimens were obtained from 2006 to 2008. All of the patients underwent surgical resection of primary, pathologically confirmed HCC at the Institute of Hepatobiliary Surgery, Southwest Hospital, Third Military Medical University. The tumor stage was based on the Edmondson grade [
28]. Sixty-six formalin-fixed, paraffin-embedded tumor specimens were used for IHC, and forty frozen tumor specimens were used for RNA extraction. All of the patients were followed for 5 years.
Cell lines and culture
For routine culture, MHCC97H, MHCC97L and HepG2 HCC cell lines purchased from the Shanghai Cell Collection (Shanghai, China) were maintained in high-glucose DMEM (Gibco, 8113035) containing 10 % FBS in a 5 % CO2 incubator at 37 °C, whereas the SMMC-7721 cell line was maintained in RPMI medium 1640 (Gibco, 8112340). Both media contained 10 % fetal bovine serum (FBS, Gibco, 10099141), penicillin (500 U/ml) and streptomycin (500 μg/ml).
Flow cytometry
Cells isolated from HCC cell lines, spheres and tumor xenografts were labeled with anti-human antibodies at 4 °C for 20 minutes according to the manufacturer’s instructions (antibodies are described in Additional file
1: Table S1). For unconjugated primary antibodies, the cells were subsequently incubated with FITC- or PE-conjugated secondary antibodies at 4 °C for 20 minutes. The labeled cells were detected using a FACSAria II system (BD Biosciences). The FcR Blocking Reagent (Miltenyi, 130-059-901) was used to increase antibody specificity.
Cell sorting
The MHCC97H and MHCC97L HCC cells were dissociated into single cells and labeled with anti-CD90 antibodies at 4 °C for 20 minutes. Subsequently, the cells were magnetically labeled with anti-mouse IgG1 MicroBeads (Miltenyi) according to the manufacturer’s instructions. In brief, the cell suspension was loaded onto a MACS column, which was placed in the magnetic field of a MACS Separator. The unlabeled cells (CD90- cells) passed through the column, and the magnetically labeled cells (CD90+ cells) were retained. After removing the column from the magnetic field, the magnetically retained CD90+ cells were eluted.
miRNA microarray
The miRNA expression profiles of CD90+ and CD90- cells isolated from the HCC cell lines MHCC97H and MHCC97L were compared using a 5th generation miRCURY LNA™ microRNA Array (v.14.0) (Exiqon). RNA extraction, RNA quality control, RNA labeling, array hybridization and data analysis were performed at the Kangchen Biotechnology Corporation (Shanghai, China). Then, the scanned images were imported using GenePix Pro 6.0 software (Axon).
Quantitative real–time PCR
For qRT-PCR of mRNA targets, total RNA was extracted from cancer cells using RNAiso Plus (TaKaRa, 09108B). cDNA synthesis was performed according to the manufacturer’s instructions (TaKaRa, DRR047A), and qRT-PCR was performed with SYBR Premix Ex Taq II (TaKaRa, DRR081A) using a LightCycler system (Roche). The PCR reaction conditions for all of the assays were 94 °C for 30 seconds, followed by 40 cycles of amplification (94 °C for 5 seconds, 60 °C for 30 seconds and 72 °C for 30 seconds). GAPDH mRNA was used to normalize RNA inputs. The qRT-PCR primers are listed in Additional file
1: Table S2.
For qRT-PCR of miRNAs, small RNAs were extracted from cancer cells or tumor tissues using RNAiso for small RNAs (TaKaRa, D340A). miRNAs were converted to cDNA using a cDNA synthesis kit (TaKaRa, DRR047A), and qRT-PCR was performed with SYBR Premix Ex Taq II (TaKaRa, DRR081A) using a LightCycler system (Roche). The PCR reaction conditions for all of the assays were 95 °C for 20 seconds, followed by 40 cycles of amplification (95 °C for 10 seconds, 60 °C for 20 seconds and 70 °C for 5 seconds). U6 was used to normalize the RNA inputs. All of the primers were from the Bulge-Loop™ miRNA qRT-PCR primer set (RiboBio, MQP-0102, China).
Luciferase gene reporter
The
in silico predictions of the potential binding regions in MAP3K8 mRNA for miRNAs were performed using TargetScan (
http://www.targetscan.org). Wild-type and mutant MAP3K8 3′-UTR luciferase plasmids were generated using the pmiR-RB-REPORT™ vector (RiboBio, China). The full-length wild-type MAP3K8 3′-UTR is 1463 bp. The wild type sense sequence was 5′-GATATGCACC GGTCTCAAGG TTCTCATTTC-3′, and the mutant sense sequence was 5′- GATATGCACC GGTCTCAAGG AAGACATTTC-3′. Exponentially growing 293 T cells were transfected with wild-type or mutant vectors using Lipofectamine® 2000 reagent (Invitrogen, 11668027, USA) according to the manufacturer's instructions. The miR-589-5p mimics or non-target control (RiboBio, NC#22, China) were co-transfected with the vectors for 48 hours, and then luciferase activity was measured.
For the clone formation assay, 500 cells were sorted by MACS and seeded per well in 6-well plates. After 10 days of culture, the clones were fixed using methanol and dyed with hematoxylin, and the number of clones (>50 cells) was assessed microscopically.
For the sphere formation assay, 1000 cells were sorted by MACS and seeded per well in ultra-low attachment 6-well plates (Costar, 3741). The cells were cultured in DMEM/F12 media (Sigma) containing B27 supplement (Gibco, 17504-044), antibiotics, 20 ng/ml EGF (Peprotech, AF-100-15) and 20 ng/ml bFGF (Peprotech, 100-18B). Fresh medium was added every 3-5 days. After 2 weeks of culture, spheres with a diameter >75 μm were counted. For FACS analysis, the spheres were collected and dissociated into single cells using trypsin.
Cell invasion and migration assays
The invasion and migration assays were performed in 24-well Millicell hanging inserts (Millipore) with or without a Matrigel layer (BD Biosciences) according to the manufacturer's instructions. Briefly, 1 × 105 cells were seeded into the top chamber, and DMEM with 10 % FBS was added to the bottom chamber as a chemoattractant. After a 48 hour incubation at 37 °C, the numbers of cells that invaded the Matrigel (invasion) or membrane (migration) were counted in 10 fields using a 40× objective lens.
To assess tumor formation in nude mice, CD90+ and CD90- cells were sorted and injected (amounts ranging from 1 × 103 to 5 × 105) subcutaneously into different sides of 6-week-old male nude mice for controlled visualization and comparison. The mice were maintained under standard conditions and were examined for tumor formation for 12 weeks. After the tumors formed, the mice were sacrificed, and xenografts were harvested for IHC and primary culture. The fresh tumor xenografts from the nude mice were cut into small pieces and plated in a cell culture flask, and tumor cells migrated out from these pieces. DMEM containing 15 % FBS was used to initially establish the primary cultures, and DMEM containing 10 % FBS was used for subsequent maintenance.
To assess the effect of miR-589-5p on HCC tumorigenesis, 3 days after 1 × 105 CD90+ MHCC97H cells were subcutaneously injected into nude mice, micrON™ agomir-589-5p (25 nmol, 50 μl) or control RNAs (RiboBio, China) were injected into the same site every 3 days within the next 2 weeks. The mice were maintained under standard conditions and were examined for tumor formation for 12 weeks.
miR-589-5p mimic/antagomir transfection
The miR-Ribo™ miR-589-5p mimic/antagomir and negative control miRs are commercially available (RiboBio, China), and the experiments were performed according to the manufacturer's instructions. In brief, 5 × 105 cells were seeded per well in 6-well plates. The miR-589-5p mimics/antagomir (or control miRs) and Lipofectamine® 2000 were diluted in Opti-MEM® (Gibco, 31985-062, USA) separately, were mixed gently and were added to the culture plates. The final concentration of mimic was 50 nM, and the final concentration of antagomir was 100 nM. After a 24 hour incubation at 37 °C, the cells were used for additional experiments.
siRNA transfection
The siRNAs and negative control RNAs were synthesized and purified by Sangon Biotech (Shanghai, China). Synthesized siRNAs were transfected into sorted CD90+ MHCC97H and MHCC97L cells with Lipofectamine® 2000 according to the manufacturer’s protocol. The siRNAs for MAP3K8 were sense: 5′-GCGCCTTTGGAAAGGTATATT-3′ and antisense: 5′-TATACCTTTCCAAAGGCGCTT-3′. The negative control siRNAs were sense: 5′-TTCTCCGAACGTGTCACGTTT-3′ and antisense: 5′-ACGTGACACGTTCGGAGATT-3′. The final concentration of siRNAs was 25 nM. After a 24 hour incubation at 37 °C, the cells were used for further experiments.
Western blot analysis
Prepared cells were lysed in radioimmunoprecipitation assay (RIPA) buffer supplemented with a protease inhibitor (Roche, Branford, CT). Total proteins (30 μg/well) were separated by electrophoresis on 12 % sodium dodecyl sulfate-polyacrylamide gels. Subsequently, protein samples were transferred onto nitrocellulose membranes (Pierce, Thermo Fisher Scientific, Waltham, MA) and incubated with corresponding primary antibodies (antibodies are described in Additional file
1: Table S1). The membranes were incubated with horseradish peroxidase-conjugated secondary antibodies and developed using SuperSignal™ chemiluminescence reagent (Pierce, Thermo Fisher Scientific) according to the manufacturer’s instruction. Protein expression levels were normalized against GAPDH.
Immunohistochemistry
HCC, paired non-tumor tissues and tumor xenografts from nude mice were fixed with formalin and embedded in paraffin. Then, samples were sectioned (5 μm) and attached to poly-L-lysine coated slides. The slides were deparaffinized, treated with 3 % H
2O
2 at 37 °C for 1 hour to block endogenous peroxidase activity, and heated in 10 mM citrate buffer at 120 °C for 2 min and 10 sec for antigen retrieval. After incubation with the primary antibody (antibodies are described in Additional file
1: Table S1) at 4 °C overnight, the samples were incubated with a secondary peroxidase-conjugated antibody for 60 min at 37 °C and then developed with DAB (Dako, 00080066). Hematoxylin was used as a counterstain. The staining intensity of CD90 was described as low or high according to the percentage of positively stained cells. Because CD90 is expressed at a basal level in most liver cells, only cells expressing a significantly higher level of CD90 were considered positive [
29].
Statistics
The data were analyzed using SPSS 18 software (SPSS Corp., Chicago, IL, USA) and are presented as the mean values ± standard error of mean (SEM) or the median with the range. When two groups were compared, Student’s t test or the Mann-Whitney U test was used. Chi-square or Fisher’s exact methods were used for clinical statistical analyses. The Kaplan–Meier and Cox regression methods were used for survival analyses. p < 0.05 was considered statistically significant and is indicated by *. p < 0.01 was considered highly statistically significant and is indicated by **. Experiments were performed in three independent repeats in triplicate.
Discussion
According to the cancer stem cell theory, CSCs are only a small subset of cells within a tumor, and this population tends to be stable in various environment. An ideal CSC marker should distinguish this subset and be expressed in all primary tumors and cell lines. To date, CD90, CD133 and EpCAM have been used as distinguishing phenotypic markers for enriching HCC CSCs from both primary tumors and cell lines [
7‐
9]. In this study, these potential CSC markers were examined by flow cytometry, and the size of the CD133
+ and EpCAM
+ populations varied greatly among the different HCC cell lines. In contrast, CD90 was much more consistently expressed in all of the tested HCC cell lines, ranging from 0.9 % to 3.1 %. In addition, we found that in the cell spheres, the proportion of CD90
+ cells increased in all cell lines but only up to 11.8 %. Moreover, in every HCC tumor sample examined, only a fraction of the tumor cells showed significant positive staining for CD90, ranging from 1.5-15.1 %. Therefore, CD90 is an ideal CSC marker that is stably expressed in a small cell population.
In HCC, several miRNAs have been shown to regulate CSCs and to play cancer promoting or suppressing roles. It has been reported that exogenous miR-181 increased EpCAM
+ HCC cell quantity and tumor-initiating ability [
19]. In CD133
+ HCC cells, miR-130b was overexpressed and enhanced chemoresistance, tumorigenicity and self-renewal [
18], whereas miR-150 was down-regulated and significantly inhibited tumor sphere formation and cell growth [
30]. In this study, we found that miR-589-5p expression was down-regulated in CD90
+ HCC cells by comparing the miRNA expression profiles of CD90
+ and CD90
- cells, and this result was confirmed by qRT-PCR. Overexpression of miR-589-5p suppressed the CSC characteristics of CD90
+ HCC cells such as stem cell-associated gene expression (Oct4, Sox2 and Nanog), cell sphere formation, invasiveness and tumorigenicity both
in vitro and
in vivo. However, overexpression of miR-589-5p had no impact on the regulation of stemness in CD90
- HCC cells, because CD90
- HCC cells do not possess CSC characteristics [
7]. Moreover, transfection of miR-589-5p antagomir in the whole cell population suppressed the expression of miR-589-5p, but failed to increase CD90
+ population (Additional file
2: Figure S4). This might due to the low abundance of miR-589-5p in HCC cell lines, antagonizing miR-589-5p did not significantly inhibit miR-589-5p functions. Hence, these data suggest that miR-589-5p is down-regulated in CD90
+ HCC cells and suppresses stem cell characteristics.
MAP3K8 has been reported to be overexpressed in various human tumors and to promote cell transformation, proliferation, migration, and invasion by activating extracellular signal–regulated kinase (ERK), Rac1, and focal adhesion kinase (FAK) [
31,
32]. However, few studies have focused on the role of MAP3K8 in HCC development. One recent study determined that MAP3K8 knockout mice exhibited a significantly lower incidence of liver tumors compared with wild-type mice in diethylnitrosamine-induced tumor formation model [
33]. In this study, the
in silico analysis predicted that MAP3K8 was a potential downstream target of miR-589-5p. Luciferase reporter assays showed that miR-589-5p directly bound to the 3ˈ-UTR of MAP3K8 mRNA, and exogenous miR-589-5p decreased MAP3K8 expression at both the mRNA and protein levels. Moreover, inhibition of MAP3K8 by siRNA significantly reduced the expression of Oct4, Sox2 and Nanog and suppressed self-renewal, migration and invasion. The above findings indicate the importance of MAP3K8 in human HCC tumorigenesis and progression by promoting CD90
+ CSC stemness characteristics. Overall, miR-589-5p appears to decrease the population of CD90
+ cells and impair stem cell characteristics partly by silencing MAP3K8.
The status of CSCs might be a key determinant of cancer behavior [
34‐
37]. Our clinical study indicated that the expression levels of CD90 and miR-589-5p were significantly inversely correlated in the HCC clinical specimens, and CD90
+ HCC samples or samples with decreased miR-589-5p expression showed more vascular invasion and reduced disease-free and overall survival. Moreover, the combination of CD90
High and miR-589-5p
Low predicted even poorer prognosis. These results might be explained by the high invasive and metastatic capacities of CD90
+ HCC and the alteration of stemness by miRNAs. Additionally, our
in vivo study demonstrated that CD90
+ HCC cells initiate tumor xenografts in immunodeficient mice, whereas CD90
- cells and miR-589-5p-transfected CD90
+ cells do not. One mouse injected with 1 × 10
5 CD90
- cells grew a small tumor by the 11
th week, but this tumor xenograft contained CD90
+ cells (Additional file
2: Figure S5), suggesting that CD90
+ cells are required to re-establish the cellular hierarchy and to generate tumors in HCC. Thus, CD90 overexpression and miR-589-5p down-regulation indicate more aggressive HCC and poor clinical outcomes.
In summary, the binding of miR-589-5p to the MAP3K8 3ˈ-UTR inhibits MAP3K8 expression and suppresses CD90+ CSC characteristics, and the expression status of CD90 and miR-589-5p determines the behavior of HCC. Thus, CD90 and miR-589-5p are useful predictors of HCC progression, and miR-589-5p and MAP3K8 might be novel molecular targets for HCC treatment.
Acknowledgements
Authors thank Dr. Jun Li (Chongqing Cancer Institute & Hospital & Cancer Center) for his contribution to professional writing revision.