Gap junctional intercellular communication (GJIC) has been suggested to be necessary for cellular proliferation and differentiation. We wanted to investigate the function of GJIC in mouse embryonic stem (ES) cells using pharmacological inhibitors or a genetic approach to inhibit the expression of connexins, that is, the subunit proteins of gap junction channels. For this purpose, we have analyzed all known connexin genes in mouse ES cells but found only three of them, Cx31, Cx43, and Cx45, to be expressed as proteins. We have demonstrated by coimmunoprecipitation that Cx31 and Cx43, as well as Cx43 and Cx45, probably form heteromeric gap junction channels, whereas Cx31 and Cx45 do not. The pharmacological inhibitors reduced GJIC between ES cells to approximately 3% and initiated apoptosis, suggesting an antiapoptotic effect of GJIC. In contrast to these results, reduction of GJIC to approximately 5% by decreased expression of Cx31 or Cx45 via RNA interference in homozygous Cx43-deficient ES cells did not lead to apoptosis. Additional studies suggested that apoptotic death of ES cells and adult stem cells reported in the literature is likely due to a cytotoxic side effect of the inhibitors and not due to a decrease of GJIC. Using the connexin expression pattern in mouse ES cells, as determined in this study, multiple connexin-deficient ES cells can now be genetically engineered in which the level of GJIC is further decreased, to clarify whether the differentiation of ES cells is qualitatively or quantitatively compromised.
Disclosure of potential conflicts of interest is found at the end of this article.
Functional gap junction channels are formed by docking and opening of two hemichannels in contacting membranes of adjacent cells. Each hemichannel is composed of six protein subunits termed connexins. To date, 20 connexin isoforms have been found in the mouse genome, 21 in the human genome . Expression of connexin proteins and functional gap junctional communication between stem cells has been controversially discussed. One hypothesis, on which the “stem cell theory of cancer” was based, claimed that the absence of gap junctional coupling is characteristic of adult stem cells and that gap junctional coupling is a hallmark of differentiated cells. These noncoupling cells were thought to be target cells for the onset of carcinogenetic events, because they are constitutively “immortal” until induced to express connexin genes and to differentiate . This hypothesis was based on findings that many cancer cells and several adult stem cell lines, for example, keratinocyte stem cells  and corneal epithelial stem cells , apparently do not express connexins and lack gap junctional intercellular communication (GJIC). In contrast to these findings with adult stem cells, connexin protein expression had been already verified in early embryos during all stages of development, beginning at the two-cell stage. Trafficking of connexin proteins to the plasma membrane and functional GJIC was first observed at the eight-cell stage, when adhesion of the blastomeres occurred, and was demonstrated at the blastocyst stage from which embryonic stem cells are derived . Furthermore, Oyamada et al.  described expression of Cx43 and Cx45 mRNA as well as protein in mouse embryonic stem cells. More recently, Cx43, Cx45 [7, 8], and Cx40  protein expression was reported in human embryonic stem cells.
In this study, we wanted to identify all connexins that are expressed in mouse embryonic stem (ES) cells and participate in gap junctional intercellular communication. Furthermore, we wanted to investigate the consequences of experimental inhibition of GJIC in HM1 cells , which have often been used for generating transgenic mice with altered connexin genes .
HM1 ES cells were grown on gelatin-coated culture plates. Cells from confluent culture dishes were trypsinized, washed in phosphate-buffered saline (PBS), and collected by centrifugation. RNA extraction was performed using Trizol (Invitrogen, Carlsbad, CA, http://www.invitrogen.com). Reverse transcription-polymerase chain reaction (RT-PCR) analyses of ES cell RNA were performed as previously described by von Maltzahn et al. . In the case of the Cx23 PCR analysis, the upstream primer Cx23KXhoIUSP (5′-ccgctcgagcggccaccatgtctctaaattacatcaagaac) and the downstream primer mCx23EcoDSP (5′-ggaattcctcatcgtttaaagagaaaacatgatc) were used. PCR mixtures (50 μl) contained 20 mM Tris-HCl (pH 8.4), 250 μM dNTPs, 1.25 mM MgCl2, 50 mM KCl, 2 μM of each primer, and 1 unit of Taq DNA polymerase (Promega, Madison, WI, http://www.promega.com). PCR was carried out for 40 cycles using a PTC-200 Thermal Cycler (BioRad, Hercules, CA, http://www.bio-rad.com) with the following PCR program: first denaturing step at 94°C (3 minutes), denaturing at 94°C (1 minute), annealing at 55°C (1 minute), elongation at 72°C (2 minutes), and final elongation (7 minutes). In the case of the Panx1 PCR analysis, the upstream primer mPanx1-USP (5′-ttgaccatggccatcgcccacttgg) and the downstream primer mPanx1-DSP2 (5′-ctacagatccaacaggggtgcgtc) were used. For the Panx2 PCR analysis, we used the upstream primer mPanx2-USP (5′-agtcggcggacatggcgacc) and the downstream primer mPanx2-DSP2 (5′-ccgctcggtggtagcagttgtcg). In the case of Panx3 PCR analysis, the upstream primer mPanx3-USP (5′-attctcagcagcatcatgtcgc) and the downstream primer mPanx3-DSP1 (5′-ttcaaagtaccgttccttcc) were used. PCR mixtures (50 μl) contained 20 mM Tris-HCl (pH 8.4), 200 μM dNTPs, 1 mM MgCl2, 50 mM KCl, 2 μM of each primer, and 1 unit of Taq DNA polymerase (Promega). PCR for Panx1 and Panx2 was carried out for 40 cycles with the following PCR program: first denaturing step at 94°C (3 minutes), denaturing at 94°C (1 minute), annealing at 65°C (1 minute), elongation at 72°C (2 minutes), and final elongation (7 minutes). In the case of the Panx3 PCR, the annealing step was performed at 55°C.
X-Gal Staining of Connexin-Deficient ES Cells Expressing a lacZ Reporter Gene
HM1 ES cells in which Cx26-coding  or Cx30.3-coding  DNAs were heterozygously replaced by lacZ reporter-coding DNA were cultured on gelatin-coated glass coverslips and stained with X-gal (X-gal: 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) solution. HM1 ES cells in which Cx43-coding DNA was flanked by loxP sites  were treated with Cre protein delivered into the cells by protein transduction and stained with X-gal solution.
Immunofluorescence and Immunoblot Analysis
For immunofluorescence analyses, HM1 ES cells were cultured on gelatin-coated glass coverslips in a 24-well culture plate. The cells were fixed in absolute ethanol and incubated for 1 hour in blocking solution (4% bovine serum albumin, 0.1% Triton X-100 in PBS). Afterward, the cells were treated with antibodies to Cx31 (1:100; Biotrend, Cologne, Germany, http://www.biotrend.com) and Cx45 (1:250; Chemicon, Temecula, CA, http://www.chemicon.com) or with a monoclonal mouse antibody directed against the stem cell marker protein Oct4 (1:400; Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com). To detect Cx43, we used a monoclonal Cx43 antibody (1:200; Zymed, San Francisco, http://www.invitrogen.com) that was directly coupled to DyLight 547 (Pierce, Rockford, IL, http://www.piercenet.com). Secondary Alexa 488- or Cy5-labeled antibodies (1:2,000; Mobitec, Göttingen, Germany) were used to detect and visualize the primary Cx31 and Cx45 antibodies. All antibodies were diluted in blocking solution. Micrographs were taken with a laser scanning microscope (LSM 510meta; Carl Zeiss, Jena, Germany, http://www.zeiss.com).
For immunoblot analyses, HM1 ES cells were lysed with 1 × Laemmli buffer (250 mM Tris, 12% [wt/vol] SDS, 40% [vol/vol] glycerin). Proteins were separated by SDS-polyacrylamide gel electrophoresis (PAGE) and transferred on nitrocellulose membrane (Amersham Biosciences, Little Chalfont, U.K., http://www.amersham.com). The membrane was blocked using 5% milk powder in wash buffer (8.5 mM Tris-HCl, 1.7 mM Tris-Base, 50 mM NaCl, 0.1% Tween 20 in double-distilled water) for 1 hour and incubated with primary antibodies against Cx26 (monoclonal mouse; 1:750; Zymed), Cx31 (polyclonal rabbit; 1:500; Biotrend), Cx32 (polyclonal mouse; 1:500; Zymed), Cx37 (polyclonal rabbit, 1:2500 ), or Cx45 (monoclonal mouse; 1:500; Chemicon) for 2 hours at room temperature. To detect Cx43, we used rabbit polyclonal antibodies raised in our laboratory against amino acid residues 359–381 of the Cx43 C-terminal region (1:400 diluted) . For the detection of Panx1, we used rabbit polyclonal Panx1 antibodies (no. 48-8000; 1:500; Invitrogen). Secondary horseradish peroxidase-conjugated antibodies (1:20,000; Dianova, Hamburg, Germany, http://www.dianova.de) were used to detect the primary antibodies. Antibodies were diluted in blocking solution. To visualize bound antibodies, enhanced chemiluminescence (ECL) reagent (SuperSignal West Pico Chemiluminescent Substrate; Pierce) was used as recommended by the manufacturer. ECL blots were exposed to x-ray film (SuperRX; Fujifilm, Tokyo, http://www.fujifilm.com).
HM1 Cell Culture and Tracer-Coupling Analysis
HM1 ES cells  were cultured on gelatin-coated culture dishes. The following tracers were applied by iontophoresis as previously described by Elfgang et al. : Lucifer yellow, ethidium bromide, and neurobiotin. Lucifer yellow CH (Li+ salt; 4% wt/vol solution in 1 M LiCl; Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) was injected for 10 seconds using negative current of 20 nA. Ethidium bromide (Merck, Darmstadt, Germany, http://www.merck.com) was injected for 5 seconds as 5 mg/ml in 1 M LiCl with positive current of 20 nA. Five minutes after injection, dye transfer was recorded by counting the number of fluorescent neighboring cells. Neurobiotin (N-2 [aminoethyl]-biotinamide hydrochloride; Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com) was iontophoretically injected for 3 seconds in 0.1 M Tris buffer, pH 7.6, using positive current of 20 nA. Afterward, cells were washed twice with PBS and fixed for 5 minutes in 1% glutaraldehyde solution. Then, cells were washed with PBS, incubated overnight at 4°C in 1% Triton X-100, washed with PBS, and incubated with horseradish peroxidase-avidin D (1% in PBS; Vector Laboratories) for 90 minutes. To visualize neurobiotin indirectly by avidin D-conjugated horseradish peroxidase, the HistoGreen POD Substrate Kit (Linaris, Wertheim-Bettingen, Germany, http://www.linaris.de/linaris_eng/index.shtml) was used. Neurobiotin tracer transfer was quantified by counting the labeled neighboring cells. As negative control, to exclude plasma bridges, 0.4% rhodamin-3-isothiocyanate dextran (Sigma-Aldrich) was injected into single HM1 ES cells.
ES cells were cultivated as described above, washed in PBS-containing 0.02% NaN3 (Sigma-Aldrich), and harvested in PBS supplemented with 1× Complete (Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com). Membrane fractions were enriched by cell fractionation, and subsequently, lysis buffer (20 mM TEA, 20 mM EDTA, 10 mM dithiothreitol, 2% [wt/vol] n-lauryl maltoside, pH 9.2) was added. Connexons were isolated in a sucrose gradient (10%–40%) by ultracentrifugation (120,000g, 20 hours, 4°C), and afterward, sucrose and detergents were removed by dialysis. Connexons were found in the range of 14%–20% sucrose.
Dimethylpimelimidate (DMP) was used to covalently couple Cx31, Cx43, or Cx45 antibodies to protein G-Sepharose beads (13 mg of DMP per 60 μl of Sepharose). Beads were washed with immunoprecipitate-wash buffer (0.2 M TEA in Tris-HCl). Connexon-containing fractions were incubated for 30 minutes (4°C) with Sepharose G gel to avoid unspecific binding of proteins to the Sepharose matrix. Afterward, the fractions were incubated with antibody-coupled beads overnight (4°C) for immunoprecipitation. Then, beads were washed with PBS and transferred to 1 × Laemmli buffer to release connexons. Proteins were separated by SDS-PAGE and processed for immunoblotting as described above.
Transmission Electron Microscopy of Isolated Connexons
Isolated connexin fractions were allowed to adhere to carbon-coated formvar-covered nickel grids (150 mesh) for 3 minutes and were fixed with 2% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.3) for 20 minutes at room temperature. After being washed in double-distilled water, connexons were negatively stained with 1% uranyl acetate for 3 minutes. Samples were examined at an acceleration voltage of 80 kV using a CM 120 transmission electron microscope (Philips Electron Optics, Eindhoven, The Netherlands, http://www.fei.com) equipped with a LaB6 filament.
Pharmacological Inhibition of Gap Junctional Communication and Apoptosis Assay
Gap junctional coupling in HM1 ES cells was disrupted using the pharmacological gap junction inhibitors 18-α-glycyrrhetinic acid (18-AGA) (0–120 μM; Sigma-Aldrich) and carbenoxolone (0–120 μM; Sigma-Aldrich). 18-AGA and carbenoxolone were diluted in a dimethyl sulfoxide (DMSO; Merck)/ethanol (3:2 vol/vol) solvent to obtain a 10 mM stock solution that was further diluted in ES cell medium to the final concentration. The P2 receptor antagonist pyridoxal-5′-phosphate-6-azophenyl 2′,4′-disulfonate (PPADS) (Sigma-Aldrich) was diluted in double-distilled water and used at concentrations of 20–200 μM. Apoptosis was measured 5 hours after adding the gap junction inhibitors using the ApoONE Caspase3/7 assay (Promega) or the AnnexinV fluorescein isothiocyanate apoptosis detection kit (Calbiochem, San Diego, http://www.emdbiosciences.com) or by poly(ADP-ribose) polymerase (PARP) cleavage, detected by immunoblot analyses with polyclonal rabbit antibodies (1:1,000; New England Biolabs, Ipswich, MA, http://www.neb.com).
Generation of Cx43-Deficient ES Cells
HM1 ES cells in which one allele of Cx43 was replaced by CreERT-coding DNA and the second allele was flanked by loxP sites followed by lacZ reporter-coding DNA  were treated with Cre protein delivered into the cells by protein transduction . ES cell clones in which Cx43 was deleted were isolated.
Short Interfering RNA-Mediated Downregulation of Cx31 and Cx45
Cx31 and Cx45 in RNA were downregulated by lipofection of predesigned smartPool short interfering RNA (siRNA) constructs (Dharmacon, Chicago, http://www.dharmacon.com) targeted to Cx31 and Cx45. Experiments were performed in 96-well plates according to the recommendation of the manufacturer. DharmaFECT1 (Dharmacon) was used as a transfection reagent. Protein lysates were harvested 96 hours after transfection. Microinjection experiments were performed at the same time.
Connexin Transcript and Protein Expression
As an initial screen, we performed RT-PCRs of HM1 cell cultures using connexin-specific primer combinations. Total RNA of HM1 cells was harvested and processed for RT-PCR individually. Using genomic DNA (or, in the case of Cx36 and Cx39, plasmid DNA) as positive controls, we found signals for amplicons of Cx26, Cx30.3, Cx31, Cx32, Cx37, Cx43, and Cx45 mRNAs (Fig. 1A–1C). Thus, 7 of 20 Cx genes are expressed as transcripts in mouse ES cells. Furthermore, we performed RT-PCR analyses for Panx1, Panx2, and Panx3 and found Panx1 mRNA in HM1 ES cells (Fig. 1D). Subsequently, connexin-specific antibodies were used to determine the presence and localization of gap junctions in HM1 cells. By immunoblot analyses, only Cx31, Cx43, and Cx45 proteins were found, whereas Cx26, Cx32, and Cx37, as well as Panx1, were not detected (Fig. 1D, 1E). Cx31, Cx43, and Cx45 could be clearly demonstrated as characteristic punctate immunosignals in the plasma membranes of adjacent cells (Fig. 2A–2E). The results of X-gal staining in HM1 cells, in which Cx26 (12) or Cx30.3 coding DNA (13) had been replaced heterozygously by LacZ coding DNA, demonstrated that beta galactosidase was not expressed in these cells (Fig. 2E, 2F). HM1 cells in which Cx43-coding DNA  had been replaced heterozygously by lacZ-coding DNA stained positively for β-galactosidase (Fig. 2H). In the case of Cx45, we studied the expression of the enhanced green fluorescent protein reporter gene in HM1 cells that had been used for the generation of Cx45-deficient mice  (Fig. 2I). Cx31 knock-in lacZ mice had previously been generated in our laboratory using R1 embryonic stem cells. R1 Cx31 knock-in ES cells were shown to be positive for X-gal staining, as well as the inner cell mass of Cx31 knock-in lacZ blastocysts . Thus, using three different approaches to demonstrate the presence of distinct connexin isoforms, only Cx31, Cx43, and Cx45 were found to be expressed as proteins in HM1 ES cells.
Gap Junctional Intercellular Communication
To assess the extent of gap junctional intercellular communication in ES cells tracer injections were performed (compare .). The most widespread coupling pattern was observed after injection of neurobiotin, the tracer with the lowest molecular mass (287 Da; charge, +1). On the average, 148 cells (n = 60) were positive for this marker (Fig. 3A). Injection of the fluorescent dye ethidium bromide (394 Da; charge, +1) labeled, on the average, 6 neighboring cells, and Lucifer yellow (443 Da; charge, −2) labeled 11 (Fig. 3B, 3C). To exclude the possibility of plasma bridges between adjacent cells, we injected the fluorescent dye rhodamin dextran (10,000 Da) and observed labeling only in the injected cell (Fig. 3D). Taken together, GJIC among HM1 cells was very abundant using tracers of low molecular mass, whereas charge selectivity appeared to have no major effect.
Heteromeric and Heterotypic Channel Formation
Gap junction hemichannels (connexons) were enriched by ultracentrifugation, and connexon-containing fractions were identified by detection of Cx43 via immunoblot analysis (Fig. 4A). Connexin hemichannels could be visualized by transmission electron microscopy of the connexon-containing fractions (Fig. 4B). Coimmunoprecipitation analyses using the hemichannel-containing fractions led to the conclusion that Cx31 and Cx43, as well as Cx43 and Cx45, form heteromeric connexin hemichannels. Cx31 and Cx45 proteins were not coimmunoprecipitated and thus are presumably not located in heteromeric channels (Fig. 4C–4F). Furthermore, immunosignals for all three ES cell connexins were found to be partially colocated, presumably in the same gap junctional plaque (Fig. 2D).
Inhibition of GJIC by Pharmacological Inhibitors
We wanted to investigate the physiological role of gap junctional coupling between mouse ES cells. Thus, we cultured HM1 cells in presence of the previously described pharmacological gap junction inhibitors 18-α-glycyrrhetinic acid or carbenoxolone or the solvent DMSO/ethanol. This reduced GJIC to approximately 3% of the original coupling observed in control HM1 cells treated with the DMSO/ethanol solvent. In the first series of experiments, death of ES cells was detected 5 hours after adding the inhibitors. This led to further analyses of whether this cellular death was due to apoptosis or necrosis. Initially, dose-response curves of the inhibitors were deduced by measuring the caspase3/7 activity and the number of living cells by relative fluorescence units (Fig. 5A–5E). The results indicated that the inhibitors themselves (Fig. 5A, 5B) and not the DMSO/ethanol solvent (Fig. 5C) evoked the response and that apoptosis rather than necrosis caused cell death. To prove that apoptosis was induced by the loss of gap GJIC and not by an unspecific side effect, coupling-deficient HeLa wild-type cells and HeLa Cx43 transfectants were cultured under the same conditions and the same concentration of gap junction inhibitors. Neither HeLa wild-type nor Cx43-transfected HeLa cells were perturbed in their growth ability or underwent apoptosis in the presence of the inhibitors (Fig. 5D, 5E). These findings were confirmed by an AnnexinV assay (Fig. 5F) or by monitoring PARP cleavage (Fig. 5F, 5G).
Recently, 18-AGA and the glycyrrhetinic acid derivative carbenoxolone were reported to be antagonists of purinergic receptors (P2 receptors; ). Therefore, we treated HM1 cells with the P2 receptor antagonist PPADS (50–200 μM) to demonstrate that apoptosis was not caused by unspecific side effects of the gap junction inhibitors on P2 receptors, especially P2X7 receptors, which were described to be involved in apoptosis . Our results showed that even high concentrations of PPADS did not lead to apoptosis in ES cells measured by AnnexinV assay (Fig. 5F). These results indicate that pharmacological inhibition of GJIC is associated with apoptotic cell death in ES cells.
Inhibition of GJIC by RNA Interference
To check whether gap junctional coupling in ES cells is required for cell survival, we used a genetic approach to verify the results obtained with the pharmacological gap junction inhibitors. Therefore, we generated homozygous Cx43-deficient ES cells (Fig. 6A), in which we further downregulated Cx31 and Cx45 by RNA interference (Fig. 6B). The deletion of Cx43-coding DNA reduced GJIC, measured by neurobiotin microinjection, to 10% of the coupling observed in HM1 cells (Fig. 6C–6E), demonstrating the high contribution of Cx43 to GJIC in ES cells.
On the average, GJIC could be further reduced to 5% of the original coupling by downregulating the expression of Cx31 and Cx45 using RNA interference (Fig. 6B, 6C, 6E, 6F), but it could not be completely abolished, so that in the end, an average of eight cells remained functionally coupled (Fig. 6F). In approximately 20% of the injected cells, a complete loss of GJIC was found (Fig. 6G). Although GJIC in these cells was strongly reduced compared with HM1 ES cells and in some cases was even completely abolished, no increase in apoptosis could be detected (Fig. 6H, 6I). These results indicate that either apoptosis after treatment with gap junction inhibitors was caused by an unspecific side effect beyond inhibition of GJIC or that the 5% remaining coupling was sufficient to protect HM1 ES cells from apoptosis.
Efficacy of Pharmacological Inhibitors of Gap Junctional Coupling
Whereas blocking of GJIC with the pharmacological gap junction inhibitors 18-AGA and carbenoxolone led to apoptosis in ES cells, we could not find evidence for increased apoptosis in Cx43-deficient ES cells treated with siRNA targeted against Cx31 and Cx45 mRNAs. For this reason, we decided to analyze the efficacy of 18-AGA with ES cells in more detail. We applied different concentrations of 18-AGA and measured the intercellular transfer of microinjected neurobiotin.
Up to a concentration of 40 μM 18-AGA, no significant reduction in GJIC could be observed, but we detected a strong reduction to 3% of the initial GJIC when increasing concentrations up to 80 μM 18-AGA were used. Further increase of the inhibitor concentration led to no further reduction in neurobiotin transfer (Fig. 7A).
Subsequently, we performed an AnnexinV assay to measure apoptosis at different inhibitor concentrations, from 20 to 120 μM. We observed increased apoptosis (in approximately 40% of the cells) at a concentration of 80 μM, which marks the point of maximum blockage. Increasing the concentration of 18-AGA to 120 μM increased apoptosis to approximately 80% in these cells, although there was no further reduction in GJIC (Fig. 7B).
In addition, we analyzed the effect of 18-AGA on Cx43-deficient ES cells (Fig. 7A). Surprisingly, even inhibitor concentrations of 100 μM did not affect GJIC, whereas the apoptosis test showed a dose-response curve similar to that measured with HM1 ES cells (Fig. 7B).
While this study was under way, several reports on the expression of gap junction proteins in human ES cells were published [7, 8, 22]. The first study, by Wong et al. , demonstrated that Cx43 and Cx45 are both expressed in human ES cells using immunofluorescence and immunoblotting analyses. Furthermore, functional GJIC was demonstrated in these cells by scrape loading. In a more recent study, Huettner et al.  showed that electrical coupling occurred in human ES cells and that transcripts for almost all human connexin isoforms were present except Cx40.1 and Cx50. Furthermore, they found Cx40 as an additional connexin protein in human ES cells. In our study, we focused on HM1 mouse ES cells, which are derived from the inner cell mass of E12TG2a blastocysts  and were used in our laboratory to generate several connexin-deficient mice [10, 13, 14]. Thus, these cells displayed the main characteristics of ES cells, that is, germ line transmission and expression of stem cell marker proteins such as Oct-4.
We combined RT-PCR, immunofluorescence, and immunoblotting analyses, as well as analyses of reporter gene expression, to identify connexin isoforms that may contribute to GJIC in mouse ES cells. Seven connexin genes were expressed as transcripts. This is less than Huettner et al.  found in human ES cells. In HM1 mouse ES cells, we found Cx31, Cx43, and Cx45 proteins located in the plasma membrane. It remains elusive why such a large number of transcripts in mouse ES cells is present but not translated and why in human ES cells nearly all connexin isoforms are transcribed . These findings may give rise to the speculation that there could be a regulatory system involving microRNAs that regulates the translation of Cx26, Cx30.3, Cx32, and Cx37 independently of their transcription in ES cells.
By tracer injections, we demonstrated functional GJIC in ES cells. In our study, we injected various tracers that differ in charge as well as molecular mass and could show that small tracers, such as neurobiotin (287 Da), permeate rather easily through gap junction channels in ES cells compared with tracers of greater molecular mass, such as Lucifer yellow (433 Da) or ethidium bromide (394 Da). Their charge (neurobiotin, +1; ethidium bromide, +1; Lucifer yellow, −2) apparently did not have a major influence. The relatively strong expression of Cx43, as demonstrated by immunofluorescence, in combination with the analysis of the coupling pattern in Cx43-deficient (Cx43del) ES cells compared with wild-type ES cells suggests that Cx43 has a major impact on cell-cell coupling in ES cells.
The three connexin isoforms detected in ES cells were further investigated regarding their ability to form heteromeric hemichannels. We showed that Cx31 and Cx43 proteins, as well as Cx43 and Cx45 isoforms, interact with each other, probably within the same hemichannel, whereas Cx31 and Cx45 do not. The ability to form heteromeric channels could increase the complexity and the possibilities for regulation of GJIC in ES cells.
Inhibition of gap junctional coupling by pharmacological compounds revealed that inhibition of GJIC coincided with the onset of cell death mediated by apoptosis. Induction of apoptosis by 18-AGA and carbenoxolone could not be observed in HeLa cells that lacked gap junctional coupling or in strongly coupled HeLa cells expressing Cx43.
Suadiciani et al.  reported that different pharmacological gap junction inhibitors, such as heptanol, octanol, carbenoxolone, flufenamic acid, and mefloquine, were potent antagonists of P2X7 receptors, which could contribute to the initiation of apoptosis  and pseudoapoptosis  in some cell types. To verify that apoptosis in ES cells was triggered by inhibition of GJIC and not by unspecific side effects of the pharmacological inhibitors on P2 receptors, we treated ES cells with the P2 receptor antagonist PPADS  and found that even high concentrations of PPADS did not trigger apoptosis, which we had observed after treatment of ES cells with gap junction inhibitors.
In parallel to the pharmacological approach to abolish GJIC, we chose a genetic approach with the intention to verify the results obtained by 18-AGA and carbenoxolone. Therefore, we adopted a strategy combining the use of Cx43-deficient ES cells with the knockdown of Cx31 and Cx45 expression via RNA interference. With this approach, we could inhibit GJIC, on the average, to 5% of the original coupling observed in HM1 cells. This extent of downregulation of GJIC is similar to what we had achieved with the use of pharmacological inhibitors (approximately 3%). Under these conditions, approximately 20% of the injected cells did not show any tracer transfer. Surprisingly, this strong reduction of GJIC did not increase apoptosis in ES cells. Although we could reduce protein expression of Cx31 and Cx45 to approximately 15% of the original expression level by RNA interference, GJIC was only decreased to approximately half of the coupling observed in Cx43-deficient ES cells. This could be due to a strong overexpression of these connexins, so that approximately 15% of the expressed proteins were able to mediate 50% of the original coupling. Another explanation for this observation could be the involvement of additional, nonconnexin gap junctional channels in ES cells, which may cause the 5% residual coupling after deletion of Cx43 and downregulation of Cx31 and Cx45. During our investigations, we found pannexin1 (Panx1) mRNA but not pannexin2 or pannexin3 mRNA in HM1 ES cells. Panx1 has been reported to form functional gap junction channels in some cell types  that are sensitive to gap junction inhibitors, such as, for example, carbenoxolone . Therefore, besides the three connexin proteins expressed in mouse ES cells, Panx1 could be another possible candidate protein for mediating cell survival. However, we could not detect any Panx1 protein by immunoblot analysis of HM1 ES cells.
To finally decide whether GJIC in ES cells protects against apoptosis or is due to a cytotoxic side effect of the gap junction inhibitors used, we analyzed the efficacy of the inhibitor on HM1 cells and Cx43del ES cells in more detail. We noticed that the number of apoptotic cells was strongly increased when the inhibitor was applied at concentrations of 80–120 μM, although GJIC did not further decrease. Furthermore, 18-AGA had little effect on GJIC between Cx43-deficient ES cells but induced apoptosis, as in HM1 ES cells. Therefore, we conclude that the gap junction inhibitor acts on Cx43 but not on Cx45 or Cx31. Similar findings were published by He et al.  with regard to Cx31.
18-AGA reduced GJIC in HM1 ES cells to approximately 3% of the coupling observed in control cells treated with the DMSO/ethanol solvent. This is less coupling than observed in Cx43-deficient ES cells treated with the inhibitor (approximately 10%). This could be due to heteromeric channels composed of Cx43 and Cx31 or Cx43 and Cx45 in HM1 cells. Our coimmunoprecipitation studies on purified connexons indicate that these heteromeric connexons probably occur in ES cells. Therefore, the effect of 18-AGA on Cx43 could also influence heteromeric channels containing Cx43 and Cx31 or Cx43 and Cx45. The 3% residual coupling in HM1 ES cells treated with 18-AGA could be due to homomeric Cx31 and Cx45 channels.
In recent years, it has been discussed that GJIC may play a role in cell death and survival and that this effect could be mediated by the exchange of cell fate modulators, such as, for example, ATP, Ca2+, or cAMP . Several studies were published that demonstrated that blocking gap junctions with pharmacological inhibitors led to apoptosis in different types of cells, for example, GFSHR-17 granulosa cells , rat neural crest cells , and cortical neural progenitor cells . Our initial experiments with HM1 ES cells using the pharmacological inhibitors 18-AGA and carbenoxolone seemed to point in the same direction. However, a more detailed study of the efficacy of 18-AGA demonstrated that this compound induced apoptosis independently from its effect on GJIC. Similar data were published by Satomi et al. , who described that 18-AGA induced apoptosis and G1 cell cycle arrest in human hepatocellular carcinoma cells (HepG2) and speculated that this effect could be independent of the inhibitory effect of 18-AGA on GJIC. Therefore, we conclude that results obtained from experiments with these inhibitors should be carefully rechecked, especially with regard to induction of apoptosis.
Since the different connexins expressed in mouse ES cells have now been unequivocally identified, further genetic means can be used to ablate them as completely as possible. The three connexin proteins (Cx43, Cx45, and Cx31) expressed in ES cells were detected throughout early embryonic development at the morula stage, as well as the blastocyst stage, in mouse  and rat . Furthermore, we found expression of these three connexins during differentiation of HM1-mouse ES cells within the embryoid body (data not shown). Gap junctions in ES cells may be required for coordinated proliferation and differentiation. In addition, the loss of GJIC in ES cells could make them more susceptible to apoptosis when exposed to stress conditions, such as, for example, stress factors, hypoxia, and so forth. These and other hypotheses can now be addressed using multiple connexin-deficient ES cell lines of low passage number.
Mouse ES cells express seven different connexin mRNAs, of which three (Cx31, Cx43, and Cx45) are translated into proteins and participate in functional GJIC. In contrast to human ES cells, Cx40 does not seem to be expressed in mouse ES cells. The identified connexin proteins are likely to form heteromeric connexin hemichannels in ES cells, consisting of Cx31 and Cx43 or Cx43 and Cx45 proteins. Interaction of Cx31 and Cx45 within the same hemichannel could not be shown.
Furthermore, we addressed the question of which biological function connexins might fulfill in embryonic stem cells. Results with pharmacological gap junction inhibitors suggested that gap junction channels could mediate the exchange of survival factors between ES cells and thus prevent apoptosis. In addition, we used homozygous Cx43-deficient ES cells and downregulated Cx31 and Cx45 via RNA interference. With this strategy, we could reduce GJIC to an extent similar to that observed after treatment of ES cells with pharmacological inhibitors. However, this decrease in coupling did not lead to increased apoptosis. In additional studies, we analyzed the efficacy of 18-AGA on ES cells in more detail. Although 80 μM 18-AGA induced maximum reduction of GJIC in ES cells, the number of apoptotic cells further increased with higher concentrations of inhibitor. Although 18-AGA had no effect on GJIC in Cx43-deficient ES cells, apoptosis was increased when the inhibitor was applied at the same concentration as in HM1 ES cells. These results and the fact that reduction of GJIC by siRNA in Cx43-deficient ES cells did not affect cell survival suggest that apoptosis observed after treatment of ES cells with pharmacological gap junction inhibitors is very likely due to an unspecific side effect of the inhibitor rather than to the decrease of GJIC.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
We thank the following colleagues: Dr. Frank Edenhofer (Life and Brain, Bonn University, Bonn, Germany) for providing Oct-4 antibodies and the Cre protein for protein transduction, Dr. Ute Preuss for providing PARP antibodies, Dr. Dominik Eckardt for Cx43-deficient ES cells, Dr. Alex Simon (University of Arizona, Tucson, AZ) for the Cx37 antibodies, and Dr. Goran Söhl for the Cx23 primer. This study was supported by grants of the German Research Association (SFB 645, B2) and the Bonn Forum in Biomedicine (to K.W.).