Abstract
Cancer stem cells (CSCs) play significant roles in tumor initiation. MicroRNA-135a (miR-135a) induced the formation of a CD133+ subpopulation from a human papillomavirus-immortalized cervical epithelial cell line. Compared with the CD133− cells, the CD133+ cells expressed higher levels of miR-135a and OCT4, exhibited significantly higher tumorsphere forming capacity and the time required for tumorsphere formation was shortened in the second generation. Serum induction suppressed the expression of CD133, OCT4 and miR-135a, but increased expression of involucrin in the miR-135a-induced CD133+ cells. The miR-135a-induced CD133+ cells were tumorigenic in a limiting dilution approach in vivo. The cells expressed significantly higher level of active β-catenin and OCT4 than the CD133− counterpart. Wnt3a enhanced the expression of OCT4 and CD133 in cervical cancer cells but failed to enhance CD133 transcription in normal cervical cells. Wnt3a stimulation also increased tumorsphere size and self-renewal of miR-135a-induced CD133+ subpopulation. Wnt/β-catenin inhibition suppressed tumorsphere formation while Wnt3a partially nullified the inhibitory effect. Taken together, miR-135a induced the formation of a subpopulation of cells with CSC properties both in vitro and in vivo and the Wnt/β-catenin signaling pathway is essential to maintain its tumorigenicity.
Introduction
Cervical cancer is the third most prevalent cancer in women worldwide (1). Although several factors were found closely related with cervical carcinogenesis, the biological mechanisms causing the disease are still poorly understood. Besides, there are only a few therapeutic options for this cancer and its recurrence is not infrequent (2). Evidences showed that a rare population of cells termed cancer stem cells (CSCs) is necessary in giving rise to and sustaining tumor growth. These cells possess enhanced ability to self-renew, undergo multilineage differentiation and survive in adverse tumorigenic environment (3–6).
It was known that normal adult stem cells acquire mutations and eventually leading to a malignant phenotype. CSC was first identified in human acute myeloid leukemia (7) and subsequently in liver, endometrial, colorectal and cervical cancer (8–11). Studies suggest that CSCs in cervical carcinoma are capable of initiating tumor development; and express markers like NANOG and nucleostemin (11–14). The survival of CSCs after chemotherapies and their self-renewal ability are the major causes of cancer relapse (12).
Most CSCs investigations focus on isolation and characterization of CSCs. Some cell surface markers, such as CD133, CD26 and CD44, are effective in isolation of CSCs in solid tumors (10,12,15,16), but subpopulations without these markers also display stem cell related features (17,18). CD133+ has recently been identified as a surface marker of CSCs for cervical cancer (19). CSC-like cells can be formed by forced expression of reprogramming factors into natural cancer cell lines (20–22). The relevance of genesis of these CSC-like cells to in vivo tumorigenesis remains unknown.
Aberrant microRNAs (miRNAs) expressions are commonly observed in pathological conditions including cervical cancer (23–25). Recently, we demonstrated that microRNA-135a (miR-135a), a miRNA dysregulated in early cervical cancer development, transformed human papillomavirus (HPV)-immortalized cervical epithelial cells into cancer cells by activating Wnt/β-catenin signaling through suppression of the pathway negative regulator seven in absentia homolog 1 (SIAH1) (26). In addition, we found that the expression of miR-135a was higher in squamous cell carcinoma (SCC) samples than in CINI, II and III samples. The expression of the miRNA was also higher in cervical cancer cell lines than in an immortalized cervical epithelial cell line (26). We hypothesized that the miRNA induced the formation of CSCs during transformation of the non-tumorigenic cells to the tumor cells. Here, we examined the formation of CSC-like subpopulation after introducing miR-135a into a non-tumorigenic HPV-infected normal cervical epithelial cell line NC104-E6/E7. The results showed that miR-135a induced the formation of CD133+ cells both in vitro and in vivo and the Wnt/β-catenin signaling pathway is essential to maintain the tumorigenecity of the cells.
Materials and methods
Cell cultures and transfections
HPV-16+ CaSki and SiHa, HPV-18+ HeLa and HPV-16/18− C33A cells were obtained from American Type Culture Collection (Manassas, VA). Transfections were done using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). OCT4-EGFP reporter construct (phOct4-EGFP) (27) with the human OCT4 promoter cloned upstream of EGFP was a grateful gift from Dr. W. Cui (Department of Gene Expression and Development, Roslin, Midlothian, UK). HeLa, Caski, SiHa and C33A were authenticated by Department of Pathology, The University of HK in 2012. The other cell lines, NC104-E6/E7 and NC104-E6/E7-135a were authenticated by Genetica Cell Line Testing (Burlington, NC) in 2019.
Tumor specimen collection and processing
The Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster, Hong Kong approved this study (approval no. UW 11-062). Tumor specimens were obtained from patients, staged according to the criteria of the International Federation of Gynecology and Obstetrics (FIGO). The tumor specimens were minced and subjected to enzymatic digestion using collagenase IA and deoxyribonuclease (Worthington Biochemical, Lakewood, NJ). After successive filtration through 100 and 40 µm cell strainers, the single cell suspension was collected for further manipulation.
Isolation of CD133+ cells from cell lines and tumor tissues
Single cells from cell lines and tumor tissues were positively selected using the CD133 MicroBead Kit (clone AC133) from MiltenyiBiotec (Bergisch-Gladbach, Germany). The purity of cells collected was confirmed by immunofluorescent staining for tumorsphere formation assay.
Tumorsphere formation assay and differentiation
CD133+ cells seeded in low density (500 cells/cm2) were cultured in serum-free medium supplemented with 10 ng/ml of bFGF, B27 (1:50), N2 (1:100), 5 µg/ml insulin (Gibco) and 20 ng/ml epidermal growth factor (Sigma–Aldrich, St Louis, MO) in low attachment plate (Corning, Lowell, MA). The number of primary tumorspheres formed (spheroids with ≥50 cells) was counted. Single cells dissociated from the primary tumorspheres were propagated at a density of 1000 cells/cm2 for secondary tumorspheres generation. All images of tumorspheres were captured under a Nikon Eclipse TE300 inverted microscope (Nikon, Japan). For β-catenin modulation, 270 pM (100 ng/ml) of recombinant human Wnt3a (R&D Systems, Minneapolis, MN) in phosphate-buffered saline and 5 µM XAV939 in dimethyl sulfoxide (DMSO) were added as Wnt/β-catenin pathway activator and inhibitor, respectively.
To determine the tumorsphere-initiating cell frequency, isolated CD133+ and CD133− cells were diluted to different concentrations and cultured in 12-well low attachment plates as described above. For in vitro differentiation, the dissociated cells from tumorspheres were spun onto glass slides using a Cytospin 2 (Shandon, Cheshire, UK) and cultured in 10% fetal bovine serum media for 14 days.
Total RNA isolation, mRNA and miRNA quantitative real-time PCR
Total RNA was extracted using the mirVana™ PARIS™ miRNA isolation Kit (Ambion, Austin, TX). Reverse transcription was performed using the Taqman®Reverse Transcription Reagents (mRNA) and miRNA specific primers with the Taqman®MicroRNA Reverse Transcription Reagents (miRNA) (Applied Biosystems, Carlsbad, CA). Gene expressions were analyzed using the Taqman® Gene Expression Assay on an Applied Biosystems 7500 Real-Time PCR system. All mRNA data were normalized to the 18S rRNA. The Taqman®miRNA Assays were used for miRNA quantification. MiRNA expressions were normalized to miR-16 or to RNU6B small nucleolar RNA (clinical biopsies). Data were analyzed using the 2−ΔΔCt method (28).
Inhibition of Wnt/β-catenin signaling during miRNA-135a-induced transformation
MiR-135a was force-expressed as described (26). To inhibit Wnt/β-catenin signaling, 5 µM of XAV939 in DMSO was added during the time of miR-135a induction. Same volume of DMSO was added in the control group. Expression of CD133 was measured by Taqman® Gene Expression Assay after three successive miR-135a induction.
Immunohistochemical validation
Paraffin sections of 5 µm thickness were deparaffinized and rehydrated. The slides were heated in the target antigen retrieval solution (Dako, Glostrup, Denmark) and endogenous peroxidase was quenched with 3% H2O2 and blocked with 5% serum. The sections were incubated with the anti-CD133 antibody (1:50, Abcam, Cambridge, UK) at 4°C overnight, followed by successive incubation with biotinylated polyclonal rabbit anti-mouse IgG (Dako), StrepAB Complex/HRP (Vector Laboratories, Burlingame, CA) and 3,3′-diaminobenzidine (Dako). The sections were counterstained with hematoxylin.
Staining intensity of each slide was scored semiquantitatively as described (29) by two independent observers (CON Leung and TM Ye) without knowledge of the clinical status of the patients. The intensity of signals and percentage of stained tumor cells were examined in five random fields with ~100 cells/field. The signal intensity was recorded as negative (0), weak (1), moderate (2) and strong (3), while the tumor cells were scored as no positive tumor cells (0), 1–25% (1), 26–50% (2), 51–75% (3) and 76–100%. A composite ‘histoscore’ was calculated as a product of the mean staining intensity (0–3) and the mean percentage of positive cells (0–4), with a maximum value of 12. The final histoscore was categorized as negative (0–1), weak (1–3), moderate (4–7) and strong (8–12).
Preparation of protein extracts and western blotting
Cellular extracts were prepared using Radioimmunoprecipitation assay buffer supplemented with a Protease Inhibitor Cocktail (Calbiochem, San Diego, CA). Anti-OCT4, anti-involucrin, anti-Green fluorescent protein (GFP) (1:1000, Abcam) and anti-β-actin (1:5000, Sigma–Aldrich) antibodies were used. Horseradish peroxidase-conjugated anti-goat, anti-rabbit or anti-mouse IgG (1:5000, GE HealthCare, Piscataway, NJ) were used where appropriate. Specific signals were visualized using enhanced chemiluminescence method (AbFrontier, Korea). Cropped gel images were used in the manuscript and full-length gel images will be provided upon request.
Immunofluorescent staining and confocal microscopy
Cells of interest were seeded and grown on cover slips or spun on glass slides. They were fixed with 4% paraformaldehyde. Following incubation with 10% serum, the samples were incubated with anti-CD133 antibody. The cells or slides were washed and incubated successively with Alexa-Fluor®-565 anti-mouse IgG (Invitrogen) and 4′,6-diamidino-2-phenylindole. They were observed after washing and mounting with aqueous fluorescence mounting medium (Dako). For cells transfected with phOct4-EGFP, they were fixed at 48 h post-transfection and incubated with 4′,6-diamidino-2-phenylindole or stained with the anti-CD133 antibody. All images were captured using a Zeiss LSM700 inverted confocal microscope (Gottingen, Germany).
In vivo tumorigenicity assay
The study protocol was approved by the Committee of the Use of Live Animals in Teaching and Research at the University of Hong Kong (CULATR no. 1828-09). Tumorigenicity was determined by mixing 500 CD133+ or CD133− cells with 1000 mouse fibroblast 3T3 cells in 100 µl of culture medium/Matrigel (BD Biosciences, San Jose, CA) mixture (1:1, v/v), which were inoculated subcutaneously into both sides of the posterior flanks of 6–8-week-old Severe combined immunodeficient mice. Animals with no sign of tumor burden were terminated at 3 months post-inoculation, and the injection sites were dissected to ensure absence of tumor. Negative control mice were inoculated with 3T3 fibroblasts and Matrigel without cervical cancer cells. For secondary tumor analysis, 100 CD133+ cells isolated from primary tumor were inoculated into Severe combined immunodeficient mice.
Statistical analysis
The Mann–Whitney rank sum test and Kruskal–Wallis one-way analysis of variance (ANOVA) were used for statistical evaluation. The results were presented using the SigmaPlot 11.0 (Jandel Scientific, San Rafael, CA). All statistical analyses were shown as mean ± standard error. A difference with P < 0.05 was considered significant. Tumorsphere-initiating cell frequency was calculated using Extreme limiting dilution analysis (30).
Results
Identification of CD133+ subpopulation in cervical cancer biopsies
No CD133+ cells were observed in normal tissues and weak cytoplasmic CD133 expression was found in cervical intraepithelial neoplasia biopsies (CINI-III, Supplementary Figure 1, available at Carcinogenesis Online). The CD133 expression was dramatically elevated by more than 3-fold in SCC biopsies when compared with the CIN groups. The histoscores of CD133 expression are summarized in Supplementary Table 1, available at Carcinogenesis Online. There was a significant elevation of CD133 expression when the precancerous cells became malignant. Interestingly, the percentages of CD133+ tumor cells in SCC specimens were highly variable. Most samples possessed >50% CD133+ cells (9/12, 75%) while others had 70–90% CD133+ cells (3/12, 25%, Supplementary Table 2, available at Carcinogenesis Online).
CD133+ subpopulation has high tumorsphere formation ability
CD133+ and CD133− cells were isolated from three fresh clinical samples by anti-CD133 antibody column and their tumorsphere formation ability was examined in non-adherent serum-free culture. The tumorsphere formation rates of CD133+ cells were 9-fold higher than that of the CD133− cells (2.25 versus 0.25%). The CD133+ subpopulation formed tumorspheres with larger diameter when compared with its negative counterparts (Supplementary Figure 2A and B, available at Carcinogenesis Online). Morphologically, the CD133+ tumorspheres were composed of viable cells while the CD133− tumorspheres consisted of clumps of non-healthy cells (Supplementary Figure 2C, available at Carcinogenesis Online). Based on the size of the tumorspheres, the CD133− tumorspheres were estimated to possess less than 50 cells while the CD133+ tumorspheres were composed of more than 200 cells. In addition, majority of the cells in the CD133+ tumorspheres retained the CD133 expression after 14 days in culture (Supplementary Figure 2D, available at Carcinogenesis Online). These observations revealed that the CD133+ cells possessed higher tumorsphere forming capacity than the CD133− cells.
Coexpression of CD133 with OCT4 in cervical cancer cell lines
The CD133+ cells in four cervical cancer cell lines, HeLa, C33A, Caski and SiHa were studied (Supplementary Figure 3, available at Carcinogenesis Online). Some side populations of cancer cells express stem-cell markers such as OCT4 (31). Therefore, we determined the expression of OCT4 in the CD133+ cells using an OCT4-EGFP reporter with the EGFP expression driven by the OCT4 promoter (27). CD133 immunofluorescence staining coupled with OCT4-EGFP reporter showed the presence of OCT4+ and CD133+ cells. Cells coexpressing the two molecules are presence in all the cervical cancer cell lines examined (Supplementary Figure 3, available at Carcinogenesis Online).
CD133+ cervical cancer cells possess characteristics of CSCs in vitro
CD133+ cells isolated from SiHa (HPV-16+) and C33A (HPV-16/18neg) were cultured at low density (260 cells/cm3) in non-adherent serum-free condition. These cells displayed significant higher tumorsphere forming ability in terms of tumorsphere number and diameter on day 35 of culture when compared with their negative counterparts (Figure 1A, left panel).
Tumorsphere forming ability of CD133+/− cells isolated from cervical cancer cell lines. (A) Representative figures of tumorspheres formed, left panel shows primary tumorspheres while the right panel shows secondary tumorspheres. (B) Bar charts showing the number and diameter of the tumorspheres formed (*P < 0.05 and **P < 0.01). Percentages of CD133+ or CD133− cells formed tumorspheres were indicated on the bars. (C) mRNA expression of miR-135a, CD133 and OCT4 in the CD133+/− tumorspheres (*P < 0.05). (D) Serum induction reduced expression of CD133 in CD133+ cells.
To determine the self-renewal capacity of the CD133+ subpopulation, the primary tumorspheres were dissociated into single cells and cultured in the same condition. Interestingly, these cells took only 18 days to form secondary tumorspheres of size comparable to that of the primary tumorspheres on day 35 of culture (Figure 1A, right panel), consistent with enhanced self-renewal ability of the secondary tumorspheres. The CD133+ secondary tumorspheres were significantly larger than their CD133− counterpart on day 18 in culture (Figure 1B). The expressions of CD133, OCT4 and miR-135a in the primary tumorspheres from the CD133+ subpopulation were significantly higher than that from the CD133− subpopulation (Figure 1C).
The CD133+ cells from day 14 primary tumorspheres of SiHa were isolated and cultured in either 10% serum or serum-free medium to determine their differentiation capacity. In a 2-week culture, the cells grew to a monolayer as their parental cells. Upon serum-induced differentiation, the CD133+ cells had reduced CD133 expression while the expression of CD133 in the CD133− cells remained undetectable (Figure 1D).
MiR-135a-transformed CD133+ cells express stemness markers
MiR-135a was transiently force-expressed in an immortalized cervical epithelial cell line NC104-E6/E7 that do not express CD133 as determined by qPCR (data not shown). After transfection, OCT4 was expressed in the treated cells (Figure 2A–E). MiR-135a forced expression elevated the expression of OCT4 mRNA for around 2.5-fold when compared with the scramble control (Figure 2F).
Expression of phOCT4-EGFP reporter construct in miR-135a-transformed and untreated NC104-E6/E7 cells and tumorsphere forming ability of CD133+/− subpopulation of miR-135a-transformed NC104-E6/E7 in vitro. (A–D) EGFP reporter construct was activated in miR-135a-transformed NC104-E6/E7 cells but not in untreated NC104-E6/E7 cells at 48 h post-transfection. (E) Western blot analyses of GFP and OCT4 expressions in miR-135a and miR-Ctrl transfected cells. (F) OCT4 mRNA expression in miR-135a and miR-Ctrl transfected cells (*P < 0.05). (G) Representative images of primary and secondary tumorspheres formed by either CD133+ or CD133− subpopulations isolated from miR-135a-transformed NC104-E6/E7 cells. Quantitative analyses of (H) tumorspheres number, (I) diameter and (J) miR-135a expression of primary and secondary tumorspheres formed by CD133+ or CD133− subpopulations isolated from miR-135a-transformed NC104-E6/E7 cells (N = 6, **P < 0.01, ***P < 0.001). (K) Western blot analyses of OCT4 and β-catenin expression of CD133+/− tumorspheres formed from miR-135a-transformed NC104-E6/E7 cells. (L) The mRNA level of multidrug resistance genes and stem cell-associated genes was enhanced in tumorspheres from CD133+ subpopulations isolated from miR-135a-transformed NC104-E6/E7 cells (N = 2, at least 150 tumorpsheres of each subpopulation were used for analyses). (M) Western blot analysis of β-catenin and SIAH1 expression in cells formed by miR-135a induction with or without XAV939 inhibition. Relative expression of β-catenin and SIAH1 to β-actin (right panel, N = 3, *P < 0.05). (N) QPCR quantification of CD133 in control and XAV939-treated cells after successive miR-135a forced expression.
The miR-135a-induced CD133+ and the CD133− subpopulations were isolated from the transformed cells and examined for their self-renewal ability. The number of CD133+ tumorspheres was 3.3-fold higher and the diameter was 3.8-fold larger than the CD133− group (Figure 2G–I). Besides, the diameter and number of 2nd passage CD133+ tumorspheres were also larger than that of the CD133− cells. The CD133+ population expressed around 23.5-fold more miR-135a than the negative counterpart (Figure 2J). The tumorspheres formed from the miR-135a-transformed CD133+ cells had higher expression of OCT4 and β-catenin protein relative to the CD133− cells (Figure 2K). The upregulation of β-catenin is consistent with our previous findings on induction of β-catenin expression by miR-135a forced expression (26). The CD133+ cell-derived tumorspheres also had upregulated expression of two multidrug resistance genes (ABCB1 and ABCG2) and the stem cell-associated genes (ALDH, OCT4 and CD44) (Figure 2L). The CD133+ subpopulation exhibited a higher in vitro tumorigenic cell frequency ranging from 1/258 to 1/599 with a mean of 1:393. For the CD133− subpopulation, the tumorigenic cell frequency was significantly lower with a mean of 1:1968 (1/1016–1/3812) cells (Table 1).
Tumorsphere-initiating cell frequency of miR-135a-induced CD133+ and CD133− subpopulations
| No. of cells | No. of tumorspheres/no. of assays | Tumorsphere-initiating frequency (TIF, 95% CI) | Differences in TIF | ||||
|---|---|---|---|---|---|---|---|
| 1000 | 400 | 200 | 100 | 25 | |||
| CD133+ | 11/12 | 8/12 | 4/12 | 3/12 | 1/12 | 1:393 (1/258–1/599) | P < 0.05 |
| CD133− | 4/12 | 3/12 | 2/12 | 0/12 | 0/12 | 1:1968 (1/1016–1/3812) |
| No. of cells | No. of tumorspheres/no. of assays | Tumorsphere-initiating frequency (TIF, 95% CI) | Differences in TIF | ||||
|---|---|---|---|---|---|---|---|
| 1000 | 400 | 200 | 100 | 25 | |||
| CD133+ | 11/12 | 8/12 | 4/12 | 3/12 | 1/12 | 1:393 (1/258–1/599) | P < 0.05 |
| CD133− | 4/12 | 3/12 | 2/12 | 0/12 | 0/12 | 1:1968 (1/1016–1/3812) |
Tumorsphere-initiating cell frequency of miR-135a-induced CD133+ and CD133− subpopulations
| No. of cells | No. of tumorspheres/no. of assays | Tumorsphere-initiating frequency (TIF, 95% CI) | Differences in TIF | ||||
|---|---|---|---|---|---|---|---|
| 1000 | 400 | 200 | 100 | 25 | |||
| CD133+ | 11/12 | 8/12 | 4/12 | 3/12 | 1/12 | 1:393 (1/258–1/599) | P < 0.05 |
| CD133− | 4/12 | 3/12 | 2/12 | 0/12 | 0/12 | 1:1968 (1/1016–1/3812) |
| No. of cells | No. of tumorspheres/no. of assays | Tumorsphere-initiating frequency (TIF, 95% CI) | Differences in TIF | ||||
|---|---|---|---|---|---|---|---|
| 1000 | 400 | 200 | 100 | 25 | |||
| CD133+ | 11/12 | 8/12 | 4/12 | 3/12 | 1/12 | 1:393 (1/258–1/599) | P < 0.05 |
| CD133− | 4/12 | 3/12 | 2/12 | 0/12 | 0/12 | 1:1968 (1/1016–1/3812) |
Wnt/β-catenin activation mediates miR-135a-induced CD133+ cell formation
MiR-135a was force-expressed in NC104-E6/E7 cells with or without XAV939 treatment for suppression of Wnt/β-catenin signaling during the induction process. Upon XAV939 inhibition, the expression of β-catenin was significantly lower while that of SIAH1 was comparable with the control group receiving miR-135a forced expression only (Figure 2M). QPCR revealed that CD133 was only expressed in the miR-135a force-expressed NC104-E6/E7 cells without XAV939 treatment (Figure 2N), demonstrating the importance of Wnt/β-catenin during miR-135a induction.
Serum challenge suppressed expression of miR-135a, OCT4 and CD133 in the CD133+ subpopulation
Differentiation of the CD133+-miR-135a-transformed cancer cells were induced in serum-containing medium. The expression of miR-135a, CD133 and OCT4 decreased progressively with time of serum culture (Figure 3A–C). Western blotting and immunohistochemical staining demonstrated that serum culture reduced OCT4 and CD133 expression and increased that of involucrin, a marker of keratinocyte terminal differentiation (Figure 3D and E).
Expression of miR-135a, OCT4 and CD133 of CD133+ subpopulation from miR-135a-transformed NC104-E6/E7 cells upon 10% serum differentiation for 18 days and in vivo tumorigenicity of miR-135a-transformed CSC-like cells. Real-time PCR analyses of (A) miR-135a, (B) CD133 and (C) OCT4 expression at days 0, 6, 13 and 18 upon 10% serum differentiation (n = 5, a,b denotes P < 0.05 when compared with day 0). (D) Western blot analyses of involucrin (epithelial cell differentiation marker), CD133 and OCT4 expression of freshly sorted CD133+ subpopulation subjected to 10% serum differentiation and normal culture medium, respectively. (E) Immunofluorescent staining of involucrin marked the success of 10% serum differentiation on the CD133+ subpopulation. (F) Tumors formed by inoculation of miR-135a-induced CD133+ cells. (G) Immunofluorescent staining of GPF indicated the tumor was formed by miR-135a-induced CD133+ cells. Inserts showed views with higher magnification. (H) Cells of different NC ratios were found in the CD133+ cells derived tumor. Insert d showed cells with higher NC ratios than cells in insert c. (I) Secondary tumors formed by inoculating 100 miR-135a-induced CD133+ cells from primary tumor.
Small number of CD133+ cells formed tumor in vivo
Five hundred of either miR-135a-induced CD133+ subpopulation or CD133− cells were inoculated into immunodeficiency mice for in vivo tumor development. Out of six mice inoculated with the CD133+ cells, five of them formed tumors after 50 days. The mean tumor weight was 1.97 ± 1.05 g (Figure 3F). No tumor was formed when CD133− cells were inoculated.
To confirm that the CD133+ cells but not the supporting 3T3 mouse fibroblasts were responsible for the growth of tumor, a GFP-expressing NC104-E6/E7-135a cell line, NC104-E6/E7-135a-GFP was produced. The in vivo tumor formation assay was repeated with the GFP-expressing cell line and confirmed GFP expression in most of the cells in the tumors formed by the CD133+ cells of NC104-E6/E7-135a-GFP (Figure 3G).
Different parts of the tumors formed by the CD133+ cells in mice had different morphology. Figure 5C shows the histology of one of the tumors. Region (a) of the tumors possesses features similar to those in biopsies of SCC; it contained more cells with eosinophilic cytoplasm, irregular nuclei and distinct cell borders when compared with Region (b) of the same tumor (Figure 3H). Hematoxylin and eosin staining showed that the cells in Region (a) had a higher nuclear-to-cytoplasm ratio (NC ratio) than those in Region (b) (Figure 3H, inserts a–d).
One hundred CD133+ cells isolated from the primary tumors were inoculated into Severe combined immunodeficient mice. Three out of four of the inoculated mice formed tumor after 35 days, indicating that the miR-135a-transformed CD133+ cells were able to form secondary tumors. The mean tumor weight was 1.93 ± 0.67 g (Figure 3I).
CD133+ cells expressed high level of active β-catenin, stemness and multidrug resistant genes
The CD133+ and CD133− subpopulations in the NC104-E6/E7-135a cell line were subjected to immunofluorescence microscopy. The active β-catenin signal was mainly localized to the nuclei of the cells. H-scoring indicated that the level of active β-catenin was higher in the CD133+ subpopulation by 60% (P < 0.05, Figure 4A and B). QPCR demonstrated that expression of CD133, Oct4, ABCG2, ALDH1A and ALDH1A3 was significantly higher in CD133+ cells subpopulation than the CD133− counterpart (Figure 4C, P < 0.05).
Expression of CSC-associated genes in the miR-135a-induced CD133+ and CD133− subpopulations. (A) Immunofluorescence staining of active β-catenin in the CD133+ and CD133− subpopulations isolated from the NC104-E6/E7-135a cells. Arrows indicated nuclear expression of active β-catenin (Red). (B) H-scoring indicated active β-catenin is significantly higher in the CD133+ subpopulation. (C) Quantification of various CSC-associated genes in the subpopulations (*P < 0.05).
Wnt3a increased transcription of CD133 and Oct4 but not proliferation
Recombinant Wnt3a (rWnt3a) significantly increased active β-catenin by 2.5-fold at a concentration of 270 pM (100 ng/ml) in the NC104-E6/E7-135a cells (Supplementary Figure 4A, available at Carcinogenesis Online, P < 0.05). The mRNA expression of CD133 and OCT4 in the NC104-E6/E7-135a cells was significant increased by 3.8 ± 1.42-fold (P < 0.05) and 7.0 ± 1.27-fold (P < 0.0001, Supplementary Figure 4A, available at Carcinogenesis Online, lower panel), respectively. The treatment did not affect proliferation of the cells (Supplementary Figure 5, available at Carcinogenesis Online). Treatment with 270 pM rWnt3a significantly upregulated active β-catenin by 3.5-fold in the C33A cells (Supplementary Figure 4B and C, available at Carcinogenesis Online, P < 0.05). The treatment significantly increased CD133 and OCT4 expression in the C33A cells (Supplementary Figure 4B, available at Carcinogenesis Online, middle and lower panels). CD133 was not detected in the NC104-tert cells; the same epithelial keratinocytes for generating NC104-E6/E7 cells by immortalization with hTERT. rWnt3a did not trigger CD133 expression but significantly increased OCT4 expression (2.9 ± 0.88-fold) (Supplementary Figure 6, available at Carcinogenesis Online).
β-Catenin enhanced tumorsphere formation and stemness genes expression in CD133+ cells
Treatment with 270 pM of rWnt3a increased the number and size of the CD133+ cell tumorspheres by 21 and 45% (P < 0.001 and P < 0.01), respectively. In contrast, inhibition of β-catenin with XAV939 reduced the number and size of the tumorspheres by 23% (P < 0.001) and 30% (P < 0.05), respectively. Simultaneous application of rWnt3a and XAV939 partly nullified the effect of each other and significantly reduced the number of tumorsphere formed when compared with the control (19%, P < 0.05). rWnt3a treatment in CD133− cells did not significantly increase in the number of tumorspheres formed. Due to the small size of tumorspheres formed, their size was not determined. CD133− cells treated with XAV939 failed to form any tumorspheres (Figure 5A–C).
Wnt stimulation positively regulated CD133+ cells tumorsphere formation ability and CSC-associated genes expression. Effect of Wnt stimulation/inhibition on tumorsphere formation ability of the miR-135a-induced CD133+/− subpopulations. (A) Representative figures of tumorspheres formed upon 240 h of treatment. Number (B) and size (C) of the tumorspheres formed. Percentages of CD133+ or CD133− cells formed tumorspheres were indicated at the top of the bars. (D) Expression of active β-catenin detected by western blotting. Expression of (E) CCND1 and c-Myc upon various treatments. (F) Relative expression of CSC-associated genes in the CD133+ subpopulation upon treatment with rWnt3a or XAV939 (*P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant).
The expression level of active β-catenin immunoreactivities and mRNA of its two direct targets, cyclin D1 (CCND1) and c-Myc, were examined (Figure 5D and E). The protein level of active β-catenin in tumorspheres increased after rWnt3a treatment and decreased upon XAV939 treatment when compared with the control (Figure 5D, left panel). Besides, the treated tumorspheres expressed 2.29 ± 0.53-fold and 2.5 ± 0.48-fold higher CCND1 and c-MYC, respectively, than the control (P < 0.05). In contrast, XAV939-treated tumorspheres expressed 0.49 ± 0.24-fold and 0.6 ± 0.28-fold less CCND1 and c-MYC, respectively. There were cell clumps formed by the CD133− cells, but most of them were too small to be categorized as tumorspheres (<50 cells).
Tumorspheres formed after rWnt3a stimulation expressed significantly higher level (1.7–6.0-fold) of OCT4 (P < 0.001), CD133 (P < 0.001), ALDH1A1 (P < 0.001), ALDH1A3 (P < 0.05) and ABCG2 (P < 0.01). In contrast, tumorspheres formed after β-catenin suppression expressed significantly lower levels of OCT4 (0.27 ± 0.05-fold, P < 0.001), CD133 (0.27 ± 0.02-fold, P < 0.05) and ABCG2 (0.6 ± 0.06-fold, P < 0.05). The expression of ALDH1A1 (0.5 ± 0.44-fold, ns) and ALDH1A3 (0.6 ± 0.30-fold, ns) was also suppressed but did not reach statistically significance level (Figure 5F).
Continuous Wnt/β-catenin activation is necessary for high tumorigenecity of the miR-135a-induced CD133+ cells
Cells of primary tumorspheres formed after treatment with Wnt modulators were dispersed and subjected to secondary tumorsphere formation assay in the presence of the same or different Wnt modulators. The cells were treated with (i) rWnt3a, (ii) XAV939 or (iii) Control (DMSO) during the primary and secondary tumorsphere formation (Figure 6A). Secondary tumorspheres were allowed to form for 20 days.
Images of secondary tumorspheres formed by miR-135a-induced CD133+ subpopulation under various treatments. (A) Secondary tumorspheres formed under treatment of rWnt3a and XAV939. (B) Number of secondary tumorspheres f ormed under different treatment (a,b; d,f and e,f denoted P < 0.05; b,c: P < 0.01; ANOVA). (C) Size of secondary tumorspheres formed under different treatment (a,b; b,c; d,f and e,f denoted P < 0.05).
The number and size of the secondary tumorspheres formed upon continuous rWnt3a treatment were significantly increased by 1.98 ± 0.11-fold (P < 0.001) and 1.44 ± 0.02-fold (P < 0.001) when comparing to the control group (Figure 6B and C). Continuous XAV939 treatment significantly decreased number and size of tumorspheres by 0.68 ± 0.10-fold and 0.74 ± 0.12-fold, respectively. Notably, there were no significant differences in both parameters of the secondary tumorspheres between the (i) untreated CD133+ cells (control), (ii) secondary tumorspheres of DMSO-treated CD133+ cells (Control) with rWnt3a treatment during primary tumorsphere formation, (iii) secondary tumorspheres formed under XAV939 from cells formerly treated with rWnt3a, (iv) secondary tumorspheres treated with rWnt3a which were formerly treated with XAV939 and (v) secondary tumorspheres of DMSO-treated CD133+ cells (Control) with XAV939 treatment during primary tumorsphere formation.
SCC CD133 immunopositive cells exhibited higher β-catenin level
CD133+ and CD133− cells were isolated from SCC biopsies (n = 3) and were inoculated into immunosuppressive mice. CD133+ cells formed tumors while CD133− cells failed to form tumor in vivo. Immunohistochemistry showed enriched β-catenin immunoreactivities in the cells with high NC ratio (>0.6) and some, but not all of the cells with enriched β-catenin cells exhibited CD133 immunoreactivities (Supplementary Figure 7, available at Carcinogenesis Online). β-Catenin signals were predominantly expressed in the cytoplasm but there were signals in some of the nucleus. Overall, CD133 was expressed in 6.38 ± 2.01% of the cells with high NC ratio. Importantly, all CD133 cells expressed higher β-catenin with H-score of >2. There was no CD133 expression in cells with low NC ratio (<0.6).
Discussion
Transformation of normal cells to cancer cells or CSC-like cells is important for studying the mechanisms of cancer cells transformation and disease progression. We identified a subpopulation of the transformed cells expressing CD133 and possessing CSC-like properties. A recent study identified CD133+ cells in tumorspheres from four low-passage HPV+ cervical cancer derived cell lines that expressed stemness and epithelial mesenchymal transition markers (32). Our study provides a comprehensive characterization of the CD133+ cells derived from the miR-135a-treated NC104-E6/E7 cell line. Apart from analyzing the expression of Oct4 and CD133 as markers of CSC-like cells, we characterized the tumorigenic potential of the CD133+ cells from fresh cervical cancer biopsies and after in vitro miR-135a-induced transformation by both in vitro and in vivo models. The results were correlated with expression of CD133 and have higher tumorigenic potential as in the archived cancer samples. The impact of Wnt signaling on the CSC-like properties of the cells was also investigated.
Majority of the SCC tumor specimens displayed a low to moderate percentage of CD133+ cells, in line with the CSC theory that tumors are generated and maintained by a small subset of cells capable of differentiation into tumor bulks (6). By using flow cytometry, it was reported that CD133 was expressed in 1.3–50.6% (median 10.4%) of the cervical cancer tissues (19) which is similar to our data obtained by immunostaining (6.11 ± 1.7%). The percentages were much lower in precancerous tissues, suggesting an increase in the number of CD133+ cells is a factor for disease progression. Epithelial tumors compose of a heterogeneous population of cells at multiple differentiation stage with variable and complex surface protein expression (16). CD133+ cells are one of the commonest studied CSC-like subpopulations of epithelial cancers (33) and other CSC-like subpopulations such as CD44+/CK17+, CD49f+ and aldehyde dehydrogenase (ALDHhigh) have been reported in cervical cancer (11,12,34).
Investigations on formation of CSCs are limited. Most CSC studies focused on isolation and characterization of CSCs from a tumor bulk. Induced pluripotent stem cell technology was used to reprogram non-tumor or tumor cells to CSCs (35). It was proposed that patient-derived induced pluripotent stem cells could be used as a model to study association of genotype and drug response (31). This is the first study showing that a single miRNA can transform HPV-infected cells into CSC-like cells. The findings indicate the possibility of reducing complexity in generating CSC-like cells, thus allowing easier generation of CSCs for precision oncology.
In our previous study, we demonstrated that high miR-135a expression suppressed SIAH1 and led to increased β-catenin expression (26). SIAH1 degrades β-catenin by the ubiquitin proteasomal degradation pathway and hence suppressing SIAH1 will increase the total amount of intracellular β-catenin level. Two observations showed that Wnt/β-catenin signaling is downstream of the action of miR-135a on induction of CSC-like cells. First, inhibition of Wnt abolished the ability of miR-135a in transforming CSC-like cells. Second, Wnt3a stimulated the transcription of CD133, OCT4 and stemness gene expression. It is likely that the effect of miR-135a on induction of CD133+ cells is not independent of Wnt ligand, and miR-135a would enhance the action of Wnt ligand by prolongation of the Wnt/β-catenin signals via reduced degradation of β-catenin.
MiR-135a can be a tumor suppressor or oncogene depending on the cell context. MiR-135a inhibits CSC-driven medulloblastoma development by repressing Arhgef6 (36), and modulates senescence of tendon stem/progenitor cell through regulating expression of Rock1 (37). The actions of miR-135a on CSC-like cells formation in other models need further investigation. One interesting point to note is that the CD133+ cells from cervical cancer cell lines have higher miR-135a expression when compared with the CD133− cells. This highlighted the role of miR-135a in cervical cancer carcinogenesis.
Activation of Wnt/β-catenin signaling is related to cervical cancer tumorigenesis (26,38,39). The expression of β-catenin is higher in 73% of cervical cancer specimens and mutation of the β-catenin gene is only observed in 20% of the studied cases (40), indicating that the β-catenin increase in cervical cancer is not associated with mutation of the β-catenin gene in most cases. The alternative possibilities are overexpression of positive regulators or downregulation of negative regulators of the β-catenin signaling pathway. In fact, the expression of positive regulators, Wnt10b, -14, DVL1 and FZD10, is high in some cervical cancer cell lines (41–43). On the other hand, some of the negative regulators are inactivated in cervical cancer samples, including hypermethylation of CpG islands in the promoters of axin, APC genes and DICKKORF (44–47), and phosphorylation of Ser9 of GSK3β (48).
Wnt/β-catenin signaling is associated with regulation of CSC-like cells in a variety of cancers, including prostate (49), colorectal (50) and breast cancer (51). However, little is known on how Wnt/β-catenin signaling transforms cervical CSC and thereby enhances the expression of CD133 and OCT4. Immunoprecipitation shows that CD133 forms a complex with E‐cadherin and β‐catenin, indicating that the three molecules are closely related (52). We found that Wnt/β-catenin activation is a prerequisite for CD133+ cell formation. Suppression of Wnt/β-catenin signaling inhibited formation of the CD133+ cells and reduced tumorigenicity of the transformed CD133+ cells.
Increasing evidence supports the use of CD133 as a biomarker for cancer diagnosis (53,54) and nuclear localization of β-catenin are associated with poor survival in cervical cancer (55). There was no evidence indicating a direct relationship between Wnt/β-catenin signaling and CD133. Recently, it was found that CD133 may act as a permissive factor for Wnt/β-catenin signaling in kidney cells. Immunoprecipitation showed that CD133 formed a complex with β-catenin and CD133 knockout cells have lower basal and activated β-catenin level (56). Whether cancer cells have similar mechanisms awaits further investigations.
The miR-135a-induced CD133+ subpopulation possesses higher tumorigenecity than its CD133− subpopulation and their tumorsphere forming ability can be enhanced or suppressed by Wnt/β-catenin stimulation or inhibition, respectively. In addition, a continuous Wnt/β-catenin stimulation is required to maintain and enhance its tumorsphere forming ability, consistent with the coexpression of β-catenin and CD133 in cervical cancer biopsies. Thus, the Wnt/β-catenin signaling pathway can be the potential target for treatment of cervical cancer. Based on our findings and previous investigation, suppression of Wnt/β-catenin signaling not only suppresses cervical cancer transformation but also suppresses the tumorigenecity of the CD133+ cells, thus reducing overall disease progression. Similarly, in breast cancer, knockdown/loss of Wnt1 decreases the number of CSC-like Sca-1+ and aldefluor-positive cells and their tumorigenecity (51). In lung cancer A549 cell line, knockdown of β-catenin reduces expression of OCT4, colony formation ability, migration and drug resistance abilities (57). Several drugs, including the Wnt/β-catenin inhibitor XAV939, are undergoing clinical trial for treatment of various cancers (58).
Notably, the CD133+ subpopulation was only found in cervical cancer cell lines, but not in a non-cancerous cell model or immortalized cervical epithelial cell line with normal cytology. High dose of rWnt3a did not induce expression of CD133 in these cells. Thus, Wnt activation alone is not able to induce the formation of CD133+ cells. The observation implies that either long-term and/or multiple Wnt/β-catenin stimulation is needed for formation of cervical cancer or additional factor is necessary for the formation of CD133+ subpopulation.
In summary, we demonstrated that the miR-135-induced CD133+ cells possess CSC-like properties and Wnt/β-catenin signaling is essential for their tumorigenecity. In pancreatic and liver cancers, there is a correlation between the proportion of CD133+ cells and poor differentiation grade (59,60). In colorectal cancer, CD133 can be used for prediction of non-response to chemotherapy (61). The active β-catenin expression level was higher in the CD133+ subpopulation of the NC104-E6/E7-135a cells. Wnt/β-catenin activation elevated the β-catenin expression level as well as CD133 expression. Our findings provide grounds to better understand the pathogenesis of cervical cancer through understanding the formation of CSC-like cells. The possible use of CD133 as a prognostic indicator for cervical cancer, and the correlation of CD133+ cells with the differentiation grading and chemoresistance of cervical cancer remain to be investigated.
Abbreviations
- CSC
cancer stem cell
- DMSO
dimethyl sulfoxide
- GFP
green fluorescent protein
- HPV
human papillomavirus
- SCC
squamous cell carcinoma
Funding
The work was supported by the General Research Fund, Hong Kong Special Administrative Region to William SB Yeung (GRF784413).
Acknowledgements
The authors acknowledge the technical assistance of confocal microscopy and flow cytometry analysis at Faculty Core Facility, Li Ka Shing Faculty of Medicine, the University of Hong Kong.
Conflict of Interest Statement: None declared.
Authors’ contributions
CON Leung and RTK Pang: Conception and design, experimental work and manuscript writing. WSB Yeung: Conception and design, financial support and manuscript writing. WD and SW Tsao: Provision of cell lines. TM Ye, DCK Yuen and NZ: Experimental work. HYS Ngan and ANY Cheung: Provision of clinical samples.
References
