Connexin 43 Reverses Malignant Phenotypes of Glioma Stem Cells by Modulating E-Cadherin§


  • Author contributions: S.-C.Y.: conception and design, performance of experiments, data analysis and interpretation, and manuscript writing; H.L.X., X.F.J., J.H.C., W.H.C., Q.L.W., and Y.H.X.: provision of study material or patients; Y.F.P., B.W., and X.Z.Y.: data analysis and interpretation and manuscript writing; Y.L., J.J.D., J.Y.J., X.H.Y., S.L.X., X.F.J., and W.H.S.: data analysis and interpretation; J.M.W. and X.Z.: data interpretation and manuscript revision; X.W.B.: conception and design, financial support, administrative support, data analysis and interpretation, manuscript writing, and final approval of manuscript.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS November 30, 2011.


Malfunctioned gap junctional intercellular communication (GJIC) has been thought associated with malignant transformation of normal cells. However, the role of GJIC-related proteins such as connexins in sustaining the malignant behavior of cancer stem cells remains unclear. In this study, we obtained tumorspheres formed by glioma stem cells (GSCs) and adherent GSCs and then examined their GJIC. All GSCs showed reduced GJIC, and differentiated glioma cells had more gap junction-like structures than GSCs. GSCs expressed very low level of connexins, Cx43 in particular, which are key components of gap junction. We observed hypermethylation in the promoter of gap junction protein α1, which encodes Cx43 in GSCs. Reconstitution of Cx43 in GSCs inhibited their capacity of self-renewal, invasiveness, and tumorigenicity via influencing E-cadherin and its coding protein, which leads to changes in the expression of Wnt/β-catenin targeting genes. Our results suggest that GSCs require the low expression of Cx43 for maintaining their malignant phenotype, and upregulation of Cx43 might be a potential strategy for treatment of malignant glioma. STEM CELLS 2012; 30:108-120.


Biological processes in multicellular organisms are regulated accurately by the substantial exchange of feedback information transferring in complicated networks [1]. Gap junctional intercellular communication (GJIC) through gap junction, which is composed of a family of more than 20 connexins (Cxs), is the most important format of direct communication between adjacent cells. Gap junction allows direct intercellular exchange of ions (e.g., Ca2+), second messengers, small metabolites, and peptides for normal function of cells [2–6]. Signals of contact inhibition, apoptosis, differentiation, and localization are transferred from adjacent cells through gap junction to maintain cellular homeostasis [2, 3].

Uncontrollable proliferation, poor differentiation, and high invasiveness are biological features of cancers. It has been reported that dysfunction of GJIC may affect cell proliferation, differentiation, and localization [7]. For instance, malfunction and disruption of GJIC are correlated with tumorigenesis in pulmonary and hepatic cells [8–10]. Several carcinogens are shown to decrease Cxs expression in normal epithelial cells [11], while enhancement of GJIC function has a profound effect for growth inhibition of cancers such as glioma with unknown mechanism [12–19]. Overall results from the studies of relationship between Cxs and glioma invasion have been controversial [20, 21].

Cancer stem cells (CSCs), although as a small fraction of cancer cell population, have been believed to play crucial roles in tumor malignancy. The isolation and characterization of CSCs have been assisting our better understanding of tumor initiation, progression, and recurrence [22–24]. Accumulating evidence from our and other laboratories has showed that CSCs in malignant glioma, that is, glioma stem cells (GSCs), possess the capacities of indefinite proliferation, differentiation multipotency, and higher tumorigenicity [25–40]. GSCs are more aggressive than the committed glioma cells. Therefore, GSCs might be the resources of invasiveness and recurrence of malignant glioma. However, the state of GJIC and related protein expression in GSCs remain unknown, and GJIC's contribution to such capabilities as invasion and tumorigenicity needs to be revealed.

In this study, we first evaluated the structural basis and function of GJIC in GSCs isolated from either human glioblastoma multiforme (GBM) cell line or primary glioma specimens and found the reduced expression of Cx proteins, Cx43 in particular, due to hypermethylated promoter region in gap junction protein α1 (GJA1) gene (which encodes Cx43). Reconstitution of Cx43 expression resulted in the inhibition of self-renewal, invasiveness, and tumorigenicity of GSCs. We found that influencing the expression of E-Cadherin and its coding proteins might be one of the undermined mechanisms. This study indicates the requirement of the low expression of Cx43 for maintaining GSCs in their malignant phenotype and a potential strategy for treatment of malignant glioma by upregulation of Cx43.


Cell Line, Primary Tumors, Tumorspheres, and Adherent GSCs

Human GBM cell line U87 was purchased from American Type Culture Collection (ATCC). Specimens from six cases of primary human malignant glioma (Supporting Information Fig. S1) were obtained clinically with the informed consent of the glioma patients as approved by the Institutional Research Ethics Board at Southwest Hospital, Chongqing, China. Tumorspheres from U87 cells and primary specimens were isolated and characterized as previously described [32, 33]. In addition, GSCs were cultured in the adherent condition using the method described by Pollard et al. [40].

Electron Microscopy

For observation under scanning electron microscope (SEM), the cells on coverslips after culture were fixed with 2.5% glutaraldehyde and postfixed in 1% osmium tetroxide. After washing with phosphate-buffered saline (PBS), the samples were dehydrated by gradient ethanol. The samples were then transferred to a critical point dryer in 100% ethanol and dried in CO2. Coverslips were mounted on aluminum sample stubs, gold coated by sputtering, and then observed under SEM (KYKY-EM3200 Digital SEM, Beijing, China). For examination by transmission electron microscope (TEM), the tumor cells were fixed with 2.5% glutaraldehyde and postfixed in 1% osmium tetroxide. After dehydration with gradient ethanol and soaking in acetone, the samples were then embedded in Epon812 resin and acetone followed by 100% Epon812 resin for 1 hour. Finally, Epon812 resin was solidified. Ultrathin sections were made with an LKB ultramicrotome (NOVA, Sweden) and stained with uranyl acetate and lead citrate for examination by TEM (JEM-2000Ex, JEOL, Japan).

Fluorescence Recovery After Photobleaching

Different groups of tumor cells were incubated with 5-carboxyfluorescein diacetate (5(6)-CFDA, 10 μg/ml, Molecular Probes). Fine-shaped cells neighboring on three to four other cells were selected under a laser confocal microscopy (LCM, Leica SP-5, Germany) and photobleached to 50%–90% of their original fluorescence intensity. The cells were examined for recovery of fluorescence 4 minutes later. After the transduction of Cx43, the recovery of GJIC was evaluated as follows: tumorspheres were plated onto glass coverslips. Alexa Fluor 568 hydrazide (500 μg/ml, Molecular Probes) was loaded into tumorspheres using Influx Pinocytic Cell-Loading Reagent (Molecular Probes). After recovered for 10 minutes, the samples were placed under a LCM for fluorescence recovery after photobleaching (FRAP) analysis.


The cells/human glioma specimens were fixed with 4% paraformaldehyde or cold acetone and then incubated with antibodies against CD133 (mouse monoclonal IgG1, Abcam, U.K.), nestin (mouse monoclonal IgG1, Chemicon), CD44s (mouse monoclonal IgG1, Neo Markers), β-tubulin III (mouse monoclonal IgG1, Chemicon), glial fibrillary acidic protein (GFAP, rabbit polyclonal, Dako, Denmark), myelin basic protein (MBP, rabbit polyclonal, Chemicon), E-cadherin (rabbit polyclonal, Santa Cruz), sex determining region Y-box 2 (Sox2, rabbit polyclonal, Novus Biologicals), Oligo2 (rabbit polyclonal, Millipore), B-cell specific moloney murine leukemia virus insertion site 1 (Bmi-1, rabbit polyclonal, Santa Cruz), and Cx43 (rabbit polyclonal, Zymed). Appropriate secondary antibodies (fluorescein isothiocyanate (FITC) goat anti-rabbit, tetramethyl rhodamin isothiocyanate (TRITC) goat anti-rabbit; Cy3 red goat anti-mouse; Cy5 goat anti-mouse; Molecular Probes) were used. The cells were counterstained with 4,6-diamidino-2-phenylindole (DAPI, Sigma) for cell nuclei and then examined under a LCM. Immunohistochemistry staining of tumor-bearing tissues from nude mice for nestin (Chemicon), Cx43 (mouse monoclonal IgG1, Abcam), and Ki-67 (rabbit polyclonal, Zhongshan, Beijing, China) was performed by Super Vision immunohistochemistry kit (Boster, Wuhan, China). Double-labeling immunostaining of human glioma specimens for stem cell/progenitor markers (octamer-binding transcription factor 4, Oct4, rabbit polyclonal, Santa Cruz; Oligo2; Sox2) and Cx43 (Abcam) was performed by DouSPTM Detection System (Maixin Biotech, Fuzhou, China) according to manufacturer's instructions.

Reverse-Transcription Polymerase Chain Reaction (RT-PCR) and Real-Time RT-PCR

Total RNA was isolated from different cell groups by Tripure (Roche Co., Swiss), following the instructions of manufacturer. RT-PCR was carried out using RNA PCR kit 3.0 (TaKaRa, Japan). The sequence of primers, the product size, and the annealing temperature were shown in Supporting Information Table S1 (Nos. 1–6). The results of RT-PCR were further verified by real-time RT-PCR using SYBR PrimeScript PCR kit II (TaKaRa, Japan). The sequence of primers, product size, and annealing temperature were also shown in Supporting Information Table S1 (Nos. 7–16).

Western Blotting

The equivalent amounts of protein were subject to Western blotting analyses. Primary antibodies, rabbit polyclonal anti-GAPDH (Beyotime, Haimen, China), Cx43 (Abcam), and E-cadherin (Santa Cruz) were used according to the manufacturer's instructions. After application of horseradish peroxidase-labeled secondary antibodies, chemiluminescence (SuperSignal West Pico, Pierce) was quantified using ImageQuant 5.0.


Cx43 or E-cadherin pulldown assays were performed with anti-Cx43 antibody (Abcam or Santa Cruz) or anti-E-cadherin antibody (Santa Cruz or BD, mouse monoclonal IgG1) using the Dynabeads Coimmunoprecipitation Kit (Invitrogen) following the manual instructions.

Flow Cytometry

Cell suspension was pelleted and stained with anti-Cx43 for 1 hour at 4°C, and then appropriate secondary antibodies (FITC Green goat anti-rabbit; Molecular Probes) were added for 1 hour at 4°C. Analysis of fluorescence intensity was performed by flow cytometry on Coulter Epics XL (Beckman). Flow cytometry was also used to estimate the transduction efficacy of adenovirus vector. We isolated CD133-positive cells, which were termed GSCs, from human primary glioma specimens using fluorescence-activated cell sorting (FACS). Briefly, cell suspension was pelleted and stained with anti–CD133–phycoerythrin (Miltenyi Biotec, Germany) for 0.5 hour at 4°C, then sorted CD133-positive by flow cytometry on the FACSAria II (BD).

Promoter Methylation

Chromatin in approximately 1 × 106 cells were isolated and sonicated by a sonicator (Diagenode, Belgium). To perform methylated DNA immunoprecipitation (MeDIP), sonicated chromatin was incubated with goat anti-human 5-methlcytidine antibody overnight and rabbit anti-goat IgG-coupled magnetic beads for 2 hours at 4°C. After elution, methyl DNA was purified. Two pairs of primers (Supporting Information Table S1, Nos. 17 and 18) were designed to amplify two potential methyl regions in GJA1 gene (NM_000165) promoter. Enrichment of methylated DNA fragments obtained from the MeDIP assay was determined by real-time PCR using SYBR PrimeScript PCR kit (TaKaRa, Japan). GAPDH (primers, Supporting Information Table S1, No. 19) served as an internal control.

Overexpression and Knocking Down

Total RNA was isolated from U87 cell line and full length of GJA1 cDNA (AF151980) was amplified. Sequence of PCR primers, products size, and annealing temperature were shown in Supporting Information Table S1 (No. 20). Then, the human GJA1 cDNA was cloned into HindIII/BamHI sites of pAdTrack-CMV to obtain the pAdTrack-CMV-Cx43, and the prospective recombinant adenovirus (Ad-Cx43) was produced. Control virus AdEasy-green fluorescent protein (GFP; Ad-GFP, Mock) was identical to Ad-Cx43 except that it did not contain Cx43 expression cassette. Viruses were amplified in HEK-293 cells. After the determination of virus concentration by an adenovirus titer immunoassay kit (QuickTiter, Cell Biolabs), the suspension containing Ad-Cx43 and Mock (5.0 μl/ml stem cell culture medium, 5–20 multiplicity of infection) were used to infect tumor cells for at least 72 hours. After the confirmation of transduction efficacy by flow cytometry and real-time RT-PCR, cells were collected for further investigation. The expression of Cx43 and E-cadherin in different groups was knocked down by shRNA Lentiviral particles (SHGLY-NM_000165, Sigma) or siRNA (CDH1-2558, 2198, 1246, Ibsbio, Co., Ltd., Shanghai, China). Sequence of these shRNAs and siRNAs were shown in Supporting Information Table S2.

Proliferation Assay

Cells dissociated from different groups were prepared into single cell suspension and seeded in 96-well plates at approximately 1,000 cells per well in 0.1 ml of Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS. On days 0, 1, 2, 3, 4, 5, 6, 7, and 8, culture medium was replaced with 0.1 ml of fresh DMEM containing 10% FBS and 0.01 ml Cell Counting Kit-8 (CCK-8, Dojindo, Japan) solution was added. After incubation at 37°C for 1.5 hours, the plates were agitated for 15 minutes, and the optical density of the solution was measured at 490 nm in a photometer (μQuant, Bio-TEK).

Matrigel Invasion Assay

Cells dissociated from different groups were cultured in the serum-free stem cell medium. On the day of experiment, the 24-well micropore polycarbonate membrane filter with 12.0 μm pores (Becton Dickinson) was coated with 10–12 μl Matrigel (Sigma, approximately 8–12 mg/ml). Then 200-μl cell suspensions (2 × 105 per ml) were seeded onto the top chamber in serum-free DMEM/F12 culture medium. The bottom chamber was filled with DMEM/F12 containing 10% FBS as chemoattractants. After 24 hours of incubation in a 5% CO2 humidified incubator at 37°C, the membranes were fixed and stained by crystal violet, and the cells on the upper surface were carefully removed with a cotton swab. The invasive cells on the lower surface were observed and counted under ×200 of microscopy.

Xenografts in Nude Mice

Nude mice were purchased from the Center of Animals (Third Military Medical University, Chongqing, China). The cells dissociated from different groups were cultured in the serum-free stem cell medium and then injected subcutaneously (5 × 104 cells) and orthotopically (5 × 103 cells) into nude mice. Tumor-bearing tissues were fixed and cut with a vibratome horizontally into 8-μm coral sections. The sections were mounted on slides and stained with immunohistochemistry, Harris hematoxylin and alcoholic eosin (H&E). Mice were hosted according to the guidelines of the Third Military Medical University Animal Committee.

Gene Expression Array and Gene Set Enrichment Analysis

RNA was isolated from different groups of U87 tumorsphere (Mock vs. Ad-Cx43) and labeled following manufacturer's instructions. Universal Reference RNA (Stratagene, LA Jolla, CA) was labeled with Cy3-dUTP, and experimental samples were labeled with Cy5-dUTP. Each labeled sample was hybridized to an Agilent whole genome gene expression array following manufacturer's instruction, and the hybridized arrays were then scanned using a GenePix 4000B scanner (Molecular Devices, Sunnyvale, CA). To analyze sets of related genes that might be systematically altered in the transduction of Cx43, we used a gene set enrichment analysis (GSEA), a method which includes searching for consistent but subtle changes in gene expression by incorporating pathway annotations [41]. Details on the methodology for this analysis and the software as well as the information on the biological data sets are available on the websites of and∼rnusse/pathways/targets.html. Briefly, the genes from experimental microarray were first ranked according to their expression differences (signal to noise ratio) between Ad-Cx43 transduced and Mock groups. The standard GSEA null hypothesis is that the rank ordering of the genes in a given comparison is random and not associated with the order of the genes and/or treatment. The extent of association was then measured by a nonparametric running sum statistic termed the enrichment score (ES), and the maximum ES (MES) was recorded over each gene set. Totally, 83 target genes of Wnt/β-catenin signaling pathway were observed and the enrichment profile was drawn according to the ES score and positions of gene set members on the rank ordered list.

Statistical Analysis

Results were presented in the form of mean (X) ± SD. The data from FRAP and real-time PCR were analyzed by ANOVA and Student's t-test. The data from primary tumorsphere formation and Matrigel invasion assay were analyzed by ANOVA. The SPSS 10.0 statistical software was used.


Loss of Gap Junction-Like Structure and Dysfunction of GJIC in GSCs

Tumorspheres, CD133-positive glioma cells, and GSCs expanded in adherent culture were isolated and identified from U87 and human glioma specimens as previously described [32, 33, 40] (Supporting Information Figs. S2, S3).

The cellular surface in U87 tumorsphere was smooth with rare prominences. All cells were connected to one another tightly. Most cells in U87 tumorspheres were in an immature or primitive state as shown by rare organelles such as mitochondria and endocytoplasmic reticulum. Few gap junction-like structures were found between these cells. However, gap junction-like structures were detected among descendants differentiated from U87 tumorspheres, which contained abundant organelles (Fig. 1A).

Figure 1.

Dysfunctional gap junction between GSCs. (A): Gap junction deficiency in GSCs. Although the cells in U87 tumorsphere were tightly connected (photograph of scanning electron microscopy, scale bar = 10 μm), no gap junction between these cells was observed under transmission electron microscope (TEM, scale bar = 100 nm). Gap junction-like structures (denoted by red arrow) were detected under TEM between descendant cells differentiated from U87 tumorsphere (scale bar = 200 nm). Conjunction region was denoted by yellow box. (B): Fluorescence recovery after photobleaching (FRAP) analysis of GJIC before reconstitution of connexins. One cell of interest (denoted by red arrow) was exposed to a short pulse laser that caused bleaching of the fluorescence. Recovered fluorescence, resulting from diffusion of unbleached dye from adjacent cells, was measured and plotted as a function of time when recovery was achieved. The fluorescence values were normalized against a reference cell (denoted by white arrow). Scale bar = 50 μm. (C): Recovery rates of different cell groups. Data from FRAP were analyzed by ANOVA and Student's t tests. Abbreviations: ANOVA, analysis of variance; GSCs, glioma stem cells.

FRAP is a noninvasive approach for measuring cell gap junction coupling. In a tumorsphere, adherent non-sphere-forming cells, adherent GSCs, and descendants differentiated from tumorspheres were loaded with the fluorescent dye 5(6)-CFDA, and one pulse of intense argon laser bleached the dye in a cell of interest. Subsequently, the unbleached 5(6)-CFDA in adjacent cells diffused back into the cell via gap junctions (Fig. 1B). The recovery rate in spheres (U87 tumorsphere, 5.02 ± 1.33%; glioma specimen-2 tumorsphere, 5.58 ± 4.98%; glioma specimen-3 tumorsphere, 4.05 ± 4.77%) was significantly lower than that of the controls (U87 adherent non-sphere-forming cells, 28.32 ± 7.07%, descendents differentiated from U87 tumorsphere, 31.43 ± 7.40%; glioma specimen-2 adherent non-sphere-forming cells, 30.96 ± 17.71%; glioma specimen-3 adherent non-sphere-forming cells, 20.14 ± 3.87%) (Fig. 1C). The recovery rate in glioma specimen-1-derived tumorsphere was also significantly lower (6.35%).

The size of cells and the condition of cell–cell contact, in other words cell density around targeting cells, varied differently between sphere and adherent cells. Thus, we examined the functional status of gap junction between those GSCs expanded in adherent condition. As shown in Figure 1B and 1C, the recovery rate in adherent GSCs was also low (4.00 ± 2.79%). These results suggest that the significant reductions in gap junction coupling in most GSCs occur not only in spheres but also in GSCs cultured under adherent conditions.

Expression of Connexins Were Lower in GSCs

We then examined the levels of Cxs, which are major components of gap junction. As shown in Figure 2, except for GJB6 (Coding for Cx30, NM_006783) and GJB1 (Coding for Cx32, BC039198), the relative mRNA levels of those gap junction protein genes (GJB2-Cx26, NM_004004; GJA3-Cx46, NM_021954; GJA1-Cx43, AF151980; and GJA8-Cx50, NM_005267) were lower in U87 tumorspheres than in U87 adherent non-sphere-forming cells (Fig. 2A, 2B), with GJA1 mRNA at the lowest level. Cx43 protein level was high in adherent non-sphere-forming or CD133 glioma specimen cells, with considerably lower level in U87 tumorspheres or CD133+ glioma specimen cells (Fig. 2C). Flow cytometry analysis confirmed the low expression of Cx43 in U87 tumorspheres (Fig. 2D). Immunofluorescent staining revealed low expression of Cx43 in tumorspheres and adherent GSCs (derived from U87 and glioma specimens 1–3), while all corresponding adherent non-sphere-forming cells with detectable Cx43, displayed as lamellar/spot at the borders between adjacent cells (Fig. 2E).

Figure 2.

Low expression of connexins (Cxs) in GSCs. (A): Expression of Cx-coding genes detected by RT-PCR. (B): Expression of Cx-coding genes detected by real-time PCR. Data were analyzed by Student's t-test. (C): Expression of Cx43 protein detected by Western blotting. (D): Expression of Cx43 detected by flow cytometry. U87 tumorsphere group was represented by yellow curve, and negative control was represented by blue curve. (E): Immunofluorescence staining of Cx43 (green/red) with the nuclei counterstained by 4,6-diamidino-2-phenylindole (blue). Scale bar = 50 μm. Abbreviations: FITC, fluorescein isothiocyanate; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GJA, gap junction protein α; GJB, gap junction protein β; GSCs, glioma stem cells; RT-PCR, reverse-transcription polymerase chain reaction.

Hypermethylation of GJA1 Gene Promoter Reduced Expression of Cx43 in GSCs

Immunostaining for Cx43 and markers for stem cell and progenitors such as Oligo2, Sox2, and Oct4 in human glioma specimens showed that Cx43 was detected in most differentiated tumor cells with no expression of stem cell markers (Fig. 3A, 3B); in contrast, Cx43 was not detected in the stem cell marker-positive tumor cells. The results were consistent with the in vitro findings.

Figure 3.

Double staining of stem cell markers/Cx43 in human primary glioma specimens and measurement of hypermethylation of GJA1 promoter in glioma stem cells (GSCs). (A): Fluorescent immunostaining for Cx43 (red, yellow arrows), stem cell markers Oct4, Oligo2, and Sox2 (green, white arrows), with the nuclei counterstained by DAPI (blue). Scale bar = 50 μm. (B): Double immunohistochemistry for Cx43 (red) and stem cell markers Oct4, Oligo2, and Sox2 (dark brown), under light microscopy ×200. (C–E): Methylated DNA by immunoprecipitation for analysis of promoter methylation. (C): Amplification regions and primers of real-time PCR. (D): Real-time PCR products from region 1. (E): Real-time PCR products from region 2. Data were analyzed by Student's t-test. Abbreviations: DAPI, 4,6-diamidino-2-phenylindole; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GJA1, gap junction protein α1; PCR, polymerase chain reaction.

For understanding the possible mechanism of the reduced expression of Cx43, we examined methylation status in the promoter of GJA1 gene. With methyl-DNA immunoprecipitation and real-time PCR (Fig. 3C–3E), we found that methylation levels of two regions in GJA1 promoter were much higher in U87 tumorspheres (Fig. 3D, 3E, region 1, 6.84 × 10−2; region 2, 1.03 × 10−1) than in adherent non-sphere-forming cells and differentiated descendant cells (region 1, 1.96 × 10−3 and 4.49 × 10−3; region 2, 1.3 × 10−3 and 1.38 × 10−4, respectively).

Reconstitution of Cx43 Inhibited In Vitro Self-Renewal, Proliferation, and Invasiveness of GSCs

An adenovirus expression vector containing GJA1 gene was constructed and transduced into different cell groups (Supporting Information Fig. S4A, S4B) with an efficacy of up to 70% (Supporting Information Fig. S4C). Enhanced expression of GJA1 in U87 tumorspheres at mRNA and protein levels was detected by real-time RT-PCR (Supporting Information Fig. S4D) and immunostaining (Supporting Information Fig. S4E). After the transduction of shRNA Lentiviral particles, reduced expression of GJA1 in adherent non-sphere-forming U87 cells at mRNA level was detected by real-time RT-PCR (Supporting Information Fig. S4F). Alexa Fluor 568 hydrazide, a red fluorescent dye, was applied to evaluate the functional status of GJIC after the reconstitution of Cx43 (Supporting Information Fig. S5A–S5C). There was no significant difference of the recovery rate between Ad-Cx43-reconstituted (Ad-Cx43) U87 tumorspheres and the control groups (Supporting Information Fig. S5C). U87 cells reconstituted with Cx43 formed fewer primary tumorspheres in serum-free stem cell medium 7 days postseeding (1.30 ± 1.42 tumorspheres/low-power field), when compared with a large quantity of tumorspheres formed by Wt U87 (16.10 ± 3.57) and Mock-transduced cells (17.10 ± 4.82) (Fig. 4A, 4B). The mRNA level of GFAP, an astrocytic differentiation marker of glioma, was increased in Ad-Cx43-transduced U87 cell line (Supporting Information Fig. S6A). Conversely, the expression of CD133, a stem cell marker, was decreased in U87 sphere after the Ad-Cx43 transduction (Supporting Information Fig. S6B). Therefore, reconstitution of Cx43 in GSCs promoted cell differentiation.

Figure 4.

The in vitro self-renewal, proliferation, and invasion of glioma stem cells (GSCs)/non-GSCs. (A): Formation of U87 primary tumorsphere after the reconstitution of Cx43. (B): The number of primary tumorspheres derived from Ad-Cx43 transduced U87 cell line cells was counted 7 days after seeding (ANOVA test). Light microscopy ×100. (C): Morphology of U87 tumorsphere cells after the reconstitution of Cx43 in Dulbecco's modified Eagle's medium + 10% FBS for 10 days; light microscopy ×200. (D): Growth curve of GSCs. (E, F): Invasion assay using 10% FBS as stimuli. (E): Invasion of GSCs after the reconstitution of Cx43. (F): Invasion of non-GSCs after the knocking down of Cx43. Statistics analysis was done by ANOVA test. Abbreviations: ANOVA, analysis of variance; FBS, fetal bovine serum; OD, optical density; shRNA, short-hairpin RNA; Wt, wild type.

The potential to proliferate by differentiation is another important property of GSCs. In DMEM + 10% FBS, the proliferation of Mock U87 tumorsphere increased by 27.16-fold on day 7. However, Ad-Cx43 U87 tumorsphere showed 55.50% reduction in proliferation when compared with Mock U87 cells, demonstrating that Cx43 reconstitution inhibited the proliferation of differentiating precursor cells (Fig. 4C, 4D). The histological features of the growth state of different sphere groups on day 10 were shown in Figure 4C. Mock cells grew in multiple layers with round shape in an extremely compact appearance. However, Ad-Cx43 cells grew in a loose monolayer with elongated shapes.

Spontaneous invasion of tumor cells was reduced after reconstitution of Cx43 (10% FBS: Wt U87 tumorsphere, 199.40 ± 33.92 cells/Hf; Mock-Ad U87 tumorsphere, 191.00 ± 46.52 cells/Hf; Ad-Cx43 U87 tumorsphere, 45.80 ± 12.91 cells/Hf) (Fig. 4E). On the contrary, invasion of those adherent non-sphere-forming cells was enhanced after the knocking down of Cx43 (10% FBS: Wt adherent non-sphere-forming U87 cells, 29.40 ± 11.35 cells/Hf; Mock-Lentivirus adherent non-sphere-forming U87 cells, 36.80 ± 16.82 cells/Hf; Lentivirus-shRNA-Cx43 adherent non-sphere-forming U87 cells, 95.10 ± 19.18 cells/Hf) (Fig. 4F).

Reconstitution of Cx43 Inhibited Tumorigenicity of GSCs

The subcutaneous and orthotopical tumorigenicity of GSCs isolated from U87 cell line and human glioma specimens 4–6 were evaluated after reconstitution of Cx43. After 8 week of tumor cell implantation, visible tumors were found in mice implanted with Wt and Mock U87 tumorspheres. However, the size of tumors in mice receiving Ad-Cx43 U87 tumorspheres was much smaller (Fig. 5A, 5B; Supporting Information Table S3). In the H&E-stained sections, the tissue from Ad-Cx43 group showed better differentiated tumor cells with thinner shape, less dense nuclei, and more abundant cytoplasm than those from mock and Wt groups (Fig. 5C). We performed immunostaining and found less proliferation, better differentiation, and increased expression of Cx43 in Ad-Cx43 group compared with the controls (Fig. 5D). The orthotopical tumorigenicity of GSCs was also inhibited after reconstitution of Cx43. Neural symptoms and dyscrasia were observed in mice receiving Mock U87 and glioma specime 4–6 tumorspheres, and xenografts were formed in the brains 4–8 weeks postimplantation. Under light microscopy, tumor tissues were observed in the brain of each mouse in this group (Fig. 5E, Supporting Information Table S3). In contrast, tumor was difficultly found in mouse brains implanted with 5 × 103 Ad-Cx43 transduced tumorspheres (U87 tumorspheres, glioma specimen 4–6 tumorspheres). There were significant differences in survival time of the mice between the groups (Fig. 5F). These results indicated that reconstitution of Cx43 profoundly inhibited the tumorigenicity of GSCs.

Figure 5.

Xenografts of glioma stem cells (GSCs) after Cx43 reconstitution. (A–C): Subcutaneous implantation of 5 × 104 U87 tumorsphere cells. (A): Gross appearance of xenografts formed 9 weeks after implantation (white arrow); upper-left inset represents a tumor formed 12 weeks after implantation. (B): The tumor sizes in different mouse groups. (C): Hematoxylin and eosin (H&E) staining, light microscopy ×200. (D): Immunostaining of subcutaneous tumor bearing tissues. (E): Orthotopical implantation of 5 × 103 GSCs in the mouse brains. H&E staining, ×200 under light microscopy. The tumor was indicated by blue arrow. (F): Survival curves of nude mice showing significant difference between Ad-Cx43 and Mock groups. Abbreviation: Wt, wild type.

Reconstitution of Cx43 Upregulated E-Cadherin Protein Expression

E-Cadherin is an important factor involved in the activation of Wnt/catenin signaling pathway and epithelial–mesenchymal transition (EMT) or mesenchymal–epithelial transition, which is crucial for invasion and stemness of cancer cells. Therefore, we examined the change of E-cadherin expression after reconstitution of Cx43 in GSCs (spheres and adherent GSCs) by Western blotting and immunostaining. As shown in Figure 6, E-Cadherin expression was upregulated in Cx43-reconstituted groups compared with the almost absence of this protein in Wt and Mock groups (Fig. 6A–6C). The spatial distribution of these two proteins was observed in the adherent GSCs. As shown in Figure 6B, we could see the localization of Cx43 and E-cadherin overlapped in the cell membrane and cytoplasm. Furthermore, the colocalization of Cx43 and E-cadherin in human glioma specimens was also revealed by fluorescent immunostaining (Fig. 6D, Supporting Information Fig. S7).

Figure 6.

Measurement of E-cadherin protein after transduction of Cx43 in GSCs. (A): Immunostaining of U87 sphere for GFP (green), E-Cadherin (red) under laser confocal microscope. Scale bar = 10 μm. (B): Immunostaining of adherent GSCs for Cx43 (red) and E-cadherin (purple). GFP fluorescence (green) and the nuclei were counterstained by 4,6-diamidino-2-phenylindole (DAPI) (blue). Scale bar = 50 μm. (C): Western blotting for Cx43 and E-cadherin. (D): Immunostaining of glioma specimen for Cx43 (green) and E-cadherin (red). The nuclei were counterstained by DAPI (blue). Scale bar = 5 μm. Abbreviations: GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GFP, green fluorescent protein; GSCs, glioma stem cells; Wt, wild type.

Interaction of Cx43 with E-Cadherin Resulted in the Changes of GSCs Invasiveness Via Wnt/β-Catenin Pathway

Downregulation of E-cadherin by siRNA (Supporting Information Fig. S8) could not alter the invasion of those GSCs without Cx43 restoration (202.17 ± 42.91 cells/Hf vs. 236.00 ± 48.63 cells/Hf) (Fig. 7A). However, for those Cx43-reconstituting GSCs with significantly reduced invasiveness (26.00 ± 12.66 cells/Hf), knocking down the expression of E-cadherin with siRNA can partially restore their invasion (101.83 ± 19.43 cells/Hf). These results suggest that GSC invasiveness is affected by a possible correlation between E-cadherin and Cx43.

Figure 7.

Interaction of Cx43 with E-cadherin resulting in the decreased invasiveness via the Wnt/β-catenin pathway. (A): Invasion assay after knocking down of E-cadherin (10% FBS as attractants, statistics by ANOVA test). (B): Expression of E-cadherin mRNA after Cx43 reconstitution. The data were analyzed by Student's t-test. (C): Coimmunoprecipitation. (D): Gene set enrichment analysis analysis of gene-expression profile for Wnt/β-catenin targets in GSCs transduced with Mock or Ad-Cx43. The top part of this plot shows the progression of the running enrichment score and the maximum peak therein. The middle part shows the genes in Wnt/β-catenin targeting gene set as “hits” against the ranked list of genes. The lower part shows the histogram for the ranked list of all genes in the expression data set. (E): The corresponding heat maps show the expression values for the top subset of genes of Wnt/β-catenin targeting gene set, which contributes most to the enrichment score in Mock group. Results are transformed into colors, where red indicates a high and blue indicates a low expression value. (F): Proposed mechanism of Cx43 restoration inhibiting the malignant phenotypes of GSCs. Abbreviations: ANOVA, analysis of variance; FBS, fetal bovine serum; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GSC, glioma stem cells; MET, mesenchymal-epithelial transition; siRNA, small interfering RNA; TCF, T cell–specific transcription factor; IP, immunoprecipitation.

To gain insight into the mechanisms by which Cx43-regulated E-cadherin protein inhibit invasiveness/tumorigenicity of GSCs, we performed real-time PCR and Cx43/E-cadherin pulldown assays to examine an interaction between these two genes/proteins. We found that reconstitution of Cx43 in GSCs not only changed the mRNA levels of E-cadherin (Fig. 7B) but also formed a protein complex of Cx43 with E-cadherin. As demonstrated by coimmunoprecipitation analysis (Fig. 7C), E-cadherin was detected in Cx43 immunoprecipitates and vice versa, Cx43 detected in E-cadherin immunoprecipitates, whereas no signals were detected in immunoprecipitates of 5% inputs group.

Because E-cadherin can recruit β-catenin to the cell membrane by forming a complex with this protein and inhibit the activation of Wnt/β-catenin signaling pathway [42, 43], we performed microarray analysis to compare the expression profile of Wnt/β-catenin target genes between Mock and Ad-Cx43 transduction groups. GSEA-P software package [41] and a list of target genes of Wnt/β-catenin signaling (∼rnusse/pathways/targets.html) were applied for interpreting gene expression data following the instruction from inventors. Genes were ranked according to their correlation to Ad-Cx43 transduction, and then the position of each gene-set member was identified, and a MES for each gene set was calculated. Gene set enrichment profile and their corresponding heat maps are shown in Figure 7D, 7E. Wnt/β-catenin downstream target genes were significantly lower after Ad-Cx43 transduction such as Wnt3a, Atoh1, POSIN, etc. Of note, stemness-related genes such as Sox2, Nanog, Oct4, and Sox2, which are known Wnt/β-catenin downstream target genes, were among the genes that were reduced after Ad-Cx43 transduction. Thus, it is likely that Cx43 can influence the expression of E-cadherin and form a complex with E-cadherin to inhibit the expression of Wnt/β-catenin targeting genes, then to decrease the invasiveness/tumorigenicity of GSCs.


Cancer stem cell theory is challenging the knowledge of cancer development and therapies. Cancer cells have been recognized as a heterogeneous population, even though they share such malignant features as uncontrollable proliferation, poor differentiation, and potential invasiveness. However, the resource of heterogeneity during cancer progression has not been clarified. Recent studies have revealed that CSCs give rise to progenitor, committed, or differentiated cancer cells, which contribute to the heterogeneity by their self-renewal and multilineage differentiation [22–24]. Therefore, CSCs are now proposed as a crucial cell subpopulation for cancer development and malignant phenotypes including invasion, metastasis, and recurrence, and thus of importance for establishment of novel treatment strategies. Malignant glioma is the most common lethal cancer in the brain. Despite the development of neurosurgery, chemotherapy, and radiotherapy during the past decades, the mean survival time of patients with malignant glioma has been limited within 2 years. Recent findings of tumor-initiating cells in glioma, that is, GSCs, provide a promising future of more effective treatment by targeting these seeding stem cells. Therefore, investigation for potential target of GSCs to inhibit their invasion and tumorigenicity is of significant importance in translational cancer research.

In this study, we found the loss of gap junction structures and dysfunction of GJICs in GSCs. We further found the reduced expression of gap junction proteins, especially Cx43, due to hypermethylation of its gene promoter. Most interestingly, we for the first time revealed that exogenous overexpression of Cx43 significantly inhibited the self-renewal, tumorigenicity, and invasiveness of GSCs, indicating that this molecule functions as a regulator for malignant phenotypes of GSCs. For further understanding the possible mechanism, we examined the invasion and stemness-related molecules such as E-cadherin, and found that restoration of Cx43 could upregulate and interact with E-cadherin. Furthermore, we found that the expression of Wnt/β-catenin targeting genes was changed, which might be due to an interaction between Cx43 and E-cadherin. The results might interpret the previous findings from others that decreased Cx43 was associated with the rate of glioma cell proliferation [14] and restoration of Cxs decreased the proliferation, tumorigenicity, and drug resistance of glioma cells [16–19].

Cx43 plays a critical role in maintaining the integrity of gap junctions [7, 44]. Our in vitro study showed that GJIC was dysfunctional and Cx43 was lost in CD133-positive, tumorsphere-forming GSCs. The result was confirmed by double-immunostaining of human primary glioma specimens. In contrast, differentiated glioma cells expressed relatively higher levels of Cx43, forming obvious gap junctions that maintained efficient GJIC. When GSCs were seeded into serum-containing medium, the intracellular Cx43 level and intercellular GJIC were partially restored in the descendant cells along with the expression of differentiation markers. Similar results were obtained in other types of CSCs from human colon carcinoma cell line HCT116 and breast cancer cell line MCF-7 (our unpublished data) as well as CD44+/CD24−/low breast CSCs [38, 45]. Therefore, Cx43−/low might be an additional helpful feature indicative of the stemness of cancer cells.

The expression and function of Cx43 are regulated by complicated mechanisms [46–48]. The expression of GJA1 was regulated by promoter methylation and hypermethylation of the GJA1 promoter was detected in tumor cells [46, 47]. In this study, we found hypermethylation of GJA1 promoter in GSCs, which might be one of the mechanisms to mediate suppression of Cx43 expression in the stem cells. DNA methyltransferases such as Dnmt3a and Dnmt3b are directly regulated by the core pluripotency transcription factors Oct4, Sox2, and Nanog [49]. This might be responsible for the phasic change of GJA1 mRNA levels accompanied by the downregulation of Sox2 and Oct4 during the differentiation of GSCs. This phenomenon is also consistent with the repression of CD133, a well-known GSC marker, caused by the promoter hypermethylation [50–52].

There is a close relationship between Cx43 and malignancy of gliomas [13–21]; however, the underlying mechanisms are still poorly understood. Furthermore, the relationship between the expression of Cx43 and the self-renewal, tumorigenicity, and invasiveness of CSCs remains unknown. Some signaling molecules transferred through GJIC have been hypothesized as the possible reason for the inhibition of proliferation and invasion of tumor cells after the restoration of Cx43. However, as shown in this study and others [53, 54], the functional status of GJIC was not significantly altered between GSCs after transduction of Ad-Cx43, although the self-renewal, tumorigenicity, and invasion capacities of these cells were obviously inhibited. Cx43 overexpression could affect EMT, a critical process involved in the generation of CSCs and the dissemination of cancer cells [55, 56]. Cx43 via its C-terminal or cytoplasmic loop can also influence the phosphorylation of some signaling molecules that are critical to tumor invasion and angiogenesis [57, 58]. Furthermore, Cx43 could be directed to the nucleus of cancer cells given its putative nuclear targeting sequence encoded within the carboxyterminal domain to directly affect expression of genes that are critical for self-renewal, tumorigenicity, and invasion capacities [59]. Thus, inhibition of tumor growth by Cx43 overexpression could be mediated by a mechanism that is independent of significant GJIC.

Thus, we have investigated the underlying mechanism for Cx43 to influence the malignant phenotypes of GSCs. E-Cadherin, a glycoprotein anchoring β-catenin to cell membrane to regulate the activity of Wnt/β-catenin signaling pathway, is implicated as a key factor in different cellular processes including invasion, stemness, and EMT [60]. Upregulation of this protein can inhibit the formation of β-catenin/TCF complex [43], which can initiate the transcription of a set of invasiveness and stemness-related genes. In tumor, E-cadherin is frequently downregulated or extinguished and strongly enhances the activity of Wnt/β-catenin signaling pathway [61–65]. A series of studies have revealed a possible correlation between Cx43 and E-cadherin expression. For example, immunoelectron microscopic analyses showed a colocalization of Cx43 and E-cadherin at cell–cell contact sites during gap junction formation [66]. As detected by immunohistochemical staining, a significant relationship between these two proteins was found in various tumors, such as lung cancer [47] and gastric cancer [67]. Hernandez-Blazquez and Nambara et al. [68, 69] have reported that E-cadherin controls intracellular trafficking and function of Cx43. In addition, antibodies against E-cadherin inhibited the assembly of gap junctions [70]. However, it remains unclear how these two proteins interact with each other. In our study, reconstitution of Cx43 in GSCs could upregulated E-cadherin expression. We further found that Cx43 was clearly colocalized with E-cadherin, which is in agreement with a previous report that extensive colocalization of Cx43 and β-catenin was detected by immunocytochemistry [71]. Of note, knocking down the expression of E-cadherin enhanced the invasiveness of Cx43-reconstituting GSCs.

Wnt/β-catenin pathway regulates cell fate decisions during development and plays a central role in modulating the delicate balance between stemness and differentiation in several adult stem cell niches [72, 73]. In many tissues, activation of this signaling has also been associated with cancer [74–76], including glioma [76]. Heterogeneous intracellular distributions of those key Wnt/β-catenin signaling molecules are observed within CSCs and their differentiated descendants. In particular, tumor cells located at the invasive front and those migrating into the adjacent stromal tissues show nuclear β-catenin staining. Hence, different levels of Wnt/β-catenin signaling activity reflect stemness heterogeneity of tumor cell and are likely to account for distinct cellular activities such as proliferation and invasiveness, which prompt tumor growth and malignant behavior, respectively [75].

E-Cadherin can recruit β-catenin and form a complex to inhibit Wnt/β-catenin signaling pathway [42, 43]. We performed microarray analysis to compare the expression changes of Wnt/β-catenin target genes between Mock and Ad-Cx43 transduction groups. Based on our results obtained from gene expression array and GSEA (Fig. 7D, 7E) and some recent findings by others [56, 71], we hypothesize that forming a complex with E-cadherin and influencing the expression of Wnt/β-catenin targeting genes, such as Sox2, Nanog, Oct4, fibroblast growth factors (FGFs), ID2, etc., might be the candidate mechanism for Cx43 to influence the dissemination and tumorigenicity of GSCs (Fig. 7F). In Cx43-reduced rat epidermal keratinocytes, E-cadherin protein translocated from plasma membrane to intracellular compartments. At the same time, these cells exhibited EMT features as defined by significantly more cells expressing vimentin compared with controls [56]. In addition to Wnt/β-catenin pathway, both Hedgehog and Notch signaling pathways are also important in the regulation of stemness and invasiveness of cancer cells [77]. We found that in the Cx43-reconstituting GSCs, the expression of both Shh and Notch2NL was downregulated by twofold but nearly twofold expression of Notch2 was increased (data not shown). However, further analysis indicated that the activation of Hedgehog and Notch signaling pathways was hardly correlated with Cx43 overexpression (data not shown). Thus, Cx43 overexpression may mainly target stemness genes that are directly regulated by Wnt/β-catenin pathway.


In this study, we provide evidence for the first time that dysfunction of GJIC is an important feature of GSCs. Reconstitution of Cx43, the key component to maintain the functional GJIC, does not alter functional status of GJIC in GSCs but ablates their self-renewal, invasiveness, and tumorigenicity although the interaction with E-cadherin and its coding protein to downregulate the express of those Wnt/β-catenin targeting genes. Our results identify a potential role of Cx43 in maintaining the malignant phenotype of GSCs.


This study was supported by grants from National Basic Research Program of China (973 Program, No. 2010CB529403) and the National Natural Science Foundation of China (NSFC, Nos. 30725035 and 30700863).


The authors indicate no potential conflict of interest.