Connexin 32-mediated cell-cell communication is essential for hepatic differentiation from human embryonic stem cells

Gap junction-mediated cell-cell interactions are highly conserved and play essential roles in cell survival, proliferation, differentiation and patterning. We report that Connexin 32 (Cx32)-mediated gap junctional intercellular communication (GJIC) is necessary for human embryonic stem cell-derived hepatocytes (hESC-Heps) during step-wise hepatic lineage restriction and maturation. Vitamin K2, previously shown to promote Cx32 expression in mature hepatocytes, up-regulated Cx32 expression and GJIC activation during hepatic differentiation and maturation, resulting in significant increases of hepatic markers expression and hepatocyte functions. In contrast, negative Cx32 regulator 2-aminoethoxydiphenyl borate blocked hESC-to-hepatocyte maturation and muted hepatocyte functions through disruption of GJIC activities. Dynamic gap junction organization and internalization are phosphorylation-dependent and the p38 mitogen-activated protein kinases pathway (MAPK) can negatively regulate Cxs through phosphorylation-dependent degradation of Cxs. We found that p38 MAPK inhibitor SB203580 improved maturation of hESC-Heps correlating with up-regulation of Cx32; by contrast, the p38 MAPK activator, anisomycin, blocked hESC-Heps maturation correlating with down-regulation of Cx32. These results suggested that Cx32 is essential for cell-cell interactions that facilitate driving hESCs through hepatic-lineage maturation. Regulators of both Cx32 and other members of its pathways maybe used as a promising approach on regulating hepatic lineage restriction of pluripotent stem cells and optimizing their functional maturation.


Results
Lineage restriction and maturation of hESCs to hepatocytes. Based on previous studies, a stepwise 3-stage protocol was used to induce hESCs differentiation to hepatocytes (hESC-Heps) (Fig. 1a). More than 95% of hESCs expressed pluripotency markers, including SRY (sex determining region Y)-box 2 (SOX2) and octamer-binding transcription factor 4 (OCT4) (Fig. 1b,c and Supplementary Fig. S1). SOX17-positive definitive endoderm (DE) was efficiently induced in the first three days during stage 1, in which forkhead box protein A2 (FOXA2) and GATA-binding protein 4 (GATA4) were dramatically increased. In the next hepatic specification stage, expression of liver transcriptional factors, HNF4A and HNF1B, were initiated, while DE marker SOX17 was reduced and pluripotent markers SOX2, NANOG and OCT4 disappeared by day 8. In the final stage of differentiation (hepatocyte maturation), high levels of alpha-fetoprotein (AFP) were expressed in the early stage of formation of hepatocyte-like cells (Pre-Hep) at about day 15; cells were found to express high levels of mature hepatocyte markers, such as albumin (ALB) and cytokeratin 18 (CK18) at about day 25 (Fig. 1b,c and Supplementary Fig. S1). Meanwhile, many functional genes of hepatocytes, including transferrin (TF), alpha-1anti-trypsin (AAT), drug metabolism phase I and phase II enzymes, phase III transporters, and detoxification-related nuclear receptors, were significantly increased in hepatocytes derived from hESCs (Fig. 1c,d). Importantly, expression of Cx32 correlated with acquisition of mature functions, and the localization of Cx32 changed from cytoplasm to membrane during the differentiation process (Fig. 1d,e). This protocol produced similar results in the H1 hESC cell line ( Supplementary Fig. S3a).

Cx32 regulators affected the expression of hepatic markers in hESC-Heps.
To explore the effects of Cx32 on hepatocyte differentiation, VK2 and 2-aminoethoxydiphenyl borate (2-APB), which have been shown to directly increase 33 or inhibit 36 Cx32 gap junctions in vitro, respectively and effectively improve 33 or block 37 hepatic gap junction communication in vivo, were used to test Cx32 relevance. When added to the last stage of differentiation, VK2 caused a dose-dependent induction of Cx32, ALB and AAT expression. Cx32 was induced about 3-fold by VK2 at 50 μ M ( Supplementary Fig. S2a). In contrast, addition of 2-APB to the last stage of differentiation caused reduction of these genes, and down-regulated Cx32 by 3-fold at 50 μ M ( Supplementary Fig. S2b). Therefore, subsequent differentiation was carried out at 50 μ M of VK2 and 2-APB.
By day 20 of differentiation, cells induced with the treatment of VK2 were large and homogeneously polygonal shaped with bright junctions. A small fraction became binucleated (arrows), and these displayed more typical hepatocyte morphology than cells in DMSO-treated control group (Fig. 2a). To compare the gene expression  GATA4, FOXA2, HNF4A, and HNF1B), key liver proteins (ALB, TF, AAT), and gap junction proteins (Cx32 and Cx43). Pluripotency factors disappear by the DE stage, followed by a transient expression of early endodermal genes and the gradual appearance of hepatocytespecific proteins. hESC, human embryonic stem cell; DE, definitive endoderm; HPC, hepatic progenitor cell; Pre-Hep, the early stage of hepatocyte-like cells at about day 15; Hep, hepatocytes derived from hESCs at about day 25. PHH, primary human hepatocytes. (d) mRNA expression of drug metabolism phase I and phase II enzymes, phase III transporters, and nuclear receptors. (e) Immunostaining of Cx32 and stage-specific makers of cells at different stages of differentiation. Scale bars, 50 μ m. Data represent mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001. of the hepatocytes induced under these conditions, a repertoire of hepatic markers were analyzed by qRT-PCR. These included plasma proteins (ALB and AAT) and genes related to cell-cell communication (Cx32 and CLDN1), the urea cycle (CPS1 and OTC1), drug metabolism phase I enzymes (CYP3A4), phase II enzymes (UGT1A1, UGT1A3, UGT1A4 and UGT2B7), phase III transporters (OATP1B1 and MDR1), and hepatic nuclear receptors (SHIP, FXR, LXRA and PPARα). The observed relative levels of expression were similarly higher in hESC-Heps treated with VK2 than untreated cells (Fig. 2b). In addition, immunostaining and flow cytometry data showed that cells treated with VK2 demonstrated more homogeneous and enhanced expression of ALB, Cx32, CK18, CPS1 and E-cadherin (ECAD) than untreated cells (Fig. 2c,d and Supplementary Fig. S4). In our previous study, we found that overexpression of Cx32 in hepatic progenitor-like cells WB-F344 resulted in improved expression of hepatocyte markers when induced to hepatic differentiation (data not shown). These results indicated that enhanced expression of Cx32 by genomic or chemical induction could promote hepatocyte differentiation.
On the contrary, cells differentiated in the presence of 2-APB showed poorer morphology and looser contacts between cells with a higher nuclei/plasma ratio (Fig. 2a). Moreover, the gene expression of hepatic markers were lower in cells treated with 2-APB, and only several scattered cells were positively stained with these mature hepatic markers when induced with 2-APB ( Fig. 2b-d and Supplementary Fig. S4). In addition, Cx32 was knocked-down in hESCs-derived hepatic progenitor cells using siRNA, and the expression levels of hepatocyte markers were dramatically decreased when the cells were further induced to differentiation ( Supplementary Fig. S5). These results suggested that inhibition of Cx32 resulted in impaired hepatic differentiation, which further proved the vital role of Cx32 in hepatocyte differentiation.
Inhibition of p38 MAPK pathway promoted hepatic differentiation of hESCs. p38 MAPK pathway was shown to regulate negatively the function of Cx32-mediated gap junctions 35 . We used a p38 MAPK inhibitor, SB, in the last step of hepatocyte differentiation, and observed more binucleated polygonal hepatocytes and an increased expression of hepatic markers than untreated controls (Fig. 3a). Cx32, ALB and AAT were induced about 3-fold by SB at 10 μ M ( Supplementary Fig. S2c). The expression of hepatic markers, including Cx32, ALB, AAT, OTC1, UGT1A4, MDR1, etc. were increased in hESC-Heps treated with SB than untreated cells (Fig. 3b). Additionally, immunostaining and flow cytometry data showed that cells treated with SB demonstrated more homogeneous and enhanced expression of ALB, Cx32, CK18, CPS1 and ECAD than untreated cells (Fig. 3c,d and Supplementary Fig. S4). On the contrary, anisomycin, an activator of p38 MAPK, disrupted hepatocyte differentiation and decreased expression of hepatic markers dramatically ( Fig. 3b-d and Supplementary Fig. S4). This consistent outcome of treatment by VK2 which up-regulated Cx32 and SB which inhibited p38 MAPK were also observed in the hepatic differentiation of the H1 hESC cell line ( Supplementary Fig. S3b).
To confirm these findings, fetal human hepatocytes (FHHs) isolated from fetuses at 16-20 weeks of gestation were differentiated with SB for 4 days and compared to DMSO-treated controls. Surprisingly, SB treatment drove expression of many mature hepatocyte-specific markers, including Cx32, ALB, ASGR1, TAT and TF, and suppressed expression of fetal hepatocyte genes, AFP and Cx43. In the meanwhile, functional genes involved in drug metabolism were improved by SB treatment, such as phase I and phase II enzymes, phase III transporters and detoxification-related nuclear receptors ( Supplementary Fig. S6). These results confirmed that SB influences hepatic maturation.
Effect of SB, VK2 and 2-APB on GJIC activity. The establishment of an elaborate GJIC network between cells is important for hepatocyte differentiation. As we observed up-regulated expression of Cx32 which represented the major gap junction protein in hepatocytes, FRAP was used to evaluate the effect of SB, VK2 and 2-APB on GJIC activity of the hESC-Heps. The fluorescence recovery of representative cells before, just after and 1, 3, or 5 min after bleaching were recorded. Fluorescence recovery was significantly enhanced by both SB and VK2 treatment, while 2-APB blocked the recovery of fluorescence, as expected ( Fig. 4a). Quantitative data demonstrated that hESC-Heps in the control group exhibited approximately 18% recovery at 5 min after bleaching. In contrast, the recovery rate was almost 2-and 3-fold faster with SB or VK2 treatment, respectively (Fig. 4b). Thus, SB and VK2 were able to improve the GJIC activity, and this was mediated primarily by Cx32 in hepatocytes differentiated from hESCs.
Hepatocyte functions were susceptible to Cx32 regulators. The secreted levels of ALB detected in the hESC-Heps culture supernatant were much higher in SB or VK2 groups than that in the control group, and were lower, as expected, in the 2-APB group (Fig. 5a). By day 25, the hESC-Heps induced with SB or VK2 produced urea at about 3 to 6 times higher than those in the control group, whereas 2-APB reduced urea production by one fold (Fig. 5b). PHHs were used as a positive control in the above quantitative assays. Additional functions included low density lipoprotein (LDL) uptake, cytoplasmic accumulation of neutral triglycerides and lipids, and glycogen storage (Fig. 5c). We observed consistently better results in SB and VK2 groups than those in the control group or 2-APB group.
Furthermore, bile canaliculi functions were evaluated by carboxydichlorofluroscein diacetate assays. The 5 (and 6)-Carboxy-2′ ,7′ -dichlorofluorescein diacetate (CDFDA) readily diffused into cells, where it was metabolized to fluorescent CDF and excreted into the bile canaliculi by the multidrug resistance-associated protein 2 (MRP2). Of note, hESC-Heps supplied with SB or VK2 demonstrated denser CDF accumulation and more functional bile canaliculi (arrows) formation in comparison with the controls (no treatment group). However, 2-APB treatment resulted in even lower CDF accumulation and much less bile canaliculi formation (Fig. 5c).
To evaluate CYP450 activity in hESC-Heps, we monitored the metabolism of testosterone, dextromethorhan and omeprazole, the substrates metabolized by CYP3A4, CYP2D6 and CYP2C19, respectively. Their respective major metabolites, 6-β -OH-testosterone, dextrorphan and 5-OH-meprazole were determined using high pressure liquid chromatography (HPLC). Normalized data showed that treatment with SB caused a 2-and 1.5-fold increase in testosterone and dextromethorphan metabolism respectively. VK2 induced about 2-fold increase in metabolism of both the two substrates, compared to untreated cells. In contrast, 2-APB severely reduced the activity to metabolize all three substrates (Fig. 6a). Furthermore, a sensitive and selective bioluminescent assay also confirmed CYP3A4 activity was improved by SB and VK2, and was inhibited by 2-APB (Fig. 6b). Importantly, the CYP3A4 activity in SB-or VK2-treated hESC-Heps were inducible by rifampicin (Fig. 6b). Furthermore, rifampicin treatment induced mRNA expression levels of CYP3A4, CYP2A6, CYP2B6, CYP2C8 and CYP2C9 in both SB and VK2 induced hepatocytes (Fig. 6c). SB and VK2 drive hepatic maturation. To explore the extent of expression of characteristic fetal and mature hepatic markers, we carried out RNA sequencing (RNA-Seq) analysis on SB-and VK2-treated hESC-Heps, comparing them to untreated controls, PHHs and FHHs. The gene expression profile of SB-and VK2-induced hESC-Heps clustered closer to adult than that to fetal hepatocytes (Fig. 7a,b). Genes involved in regulation of liver development and maturation and in typical liver functions, including glucose metabolism, fatty acid metabolism, drug metabolism, and complement and coagulation, were up-regulated in SB-and VK2-treated hESC-Heps, compared to untreated controls as well as to FHHs (Fig. 7c). Furthermore, we analyzed 452 genes up-regulated in SB-treated hESC-Heps and 1427 genes up-regulated in VK2-treated hESC-Heps compared to controls by KEGG. We found that the enriched pathways included PPAR signaling pathway, metabolism of xenobiotics by CYP450, pentose and glucuronate inter-conversions, glycine, serine and threonine metabolism, complement and coagulation cascades and others (Fig. 7d,e). Taken together, these data suggested a global effect of SB and VK2 stimulation. Stabilization of Cx32-mediated GJ by inhibition of Cx32 phosphorylation. Pathway analysis of RNA-seq data showed that with SB-treatment, MAP kinase phosphatases (DUSP4 and DUSP9) which inhibit the p38 pathway were up-regulated ( Supplementary Fig. S7a). Similarly, under the treatment of VK2, MAP kinase phosphatases (DUSP8, DUSP9 and DUSP14), as well as a protein tyrosine phosphatase (PTPN7), which is also an inhibitor for the p38 pathway, were up-regulated ( Supplementary Fig. S7b). The bioinformatics data suggested that both SB and VK2 might affect the hepatic differentiation process in similar ways through negative regulation of the p38 signaling pathway.
Cxs are dynamic polytopic membrane proteins that exhibit unprecedented short half-lives of only a few hours. MAP kinases, including p38, have been reported to phosphorylate Cxs and contribute to gap junction internalization and degradation 34,38,39 . Western blotting showed that phosph-p38 (p-p38) was dramatically down-regulated in hESC-Heps treated with SB, indicating efficient inhibition of p38 activity by SB (Fig. 8a). To investigate the effects of SB on Cx32 phosphorylation, we immune-precipitated Cx32 from untreated and SB-treated hESC-Heps  The metabolic products of testosterone (6-β -OH-testosterone), dextromethorphan (dextrorphan), and omeprazole (5-OH-meprazole), assays for CYP3A4, 2D6 and 2C19 activities, respectively, were determined by liquid chromatography-tandem mass spectrometry. Data were calculated using area under the curve and normalized to hESC-Heps without treatment (controls). (b) Induction of CYP3A4 activity in response to 72 hours of stimulation with PXR agonist rifampicin. (c) The mRNA levels of the induced CYP enzymes in ESC-Heps treated with SB or VK2 were measured by qRT-PCR. Fold inductions were normalized to the levels in cells without inducer treatment, respectively. Data represent mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001. and then probed the precipitates with anti-phosphoserine/threonine by Western blotting. SB-treatment reduced the levels of serine/threonine-phosphorylated Cx32 (Fig. 8b). Similarly, VK2-induced hepatocytes also showed much lower activity of p38 and less phosphorylated Cx32 than untreated cells (Fig. 8a,b). These results indicated that both SB and VK2 might stabilize Cx32 by inhibiting its phosphorylation, thus maintaining Cx32-mediated gap junctions (Fig. 8c).

Discussion
Gap junctions form a web of cellular communication for the interconnected cells, which facilitate the spread of signals and cell-to-cell synchronization. Especially in liver, they play an essential role in amplifying injury, shown with the gap junction-dependent progression of liver injury induced by some drugs 37 . In cultured hepatocytes, sharing intracellular components via gap junction-mediated cell-to-cell communication resulted in synchronized death or alternatively led to the protection against toxic stimuli to death by an equal distribution of death versus survival molecules 40 . In these scenarios, Cx32 acted as one of the major members to compose gap junction structure. Until now, it has not been reported whether Cx32 and Cx32-mediated GJIC activities play any roles during the in vitro process of hepatic lineage restriction when pluripotent stem cells are guided to reach to a mature hepatocyte fate. Therefore, study on Cx32 expression as well as Cx-mediated GJIC activation on both hepatic differentiation from hESCs and maturation of hESC-Heps will be helpful to enlarge our knowledge on understanding Cx32 and the cell-cell interactions during liver organogenesis. Furthermore, it could be useful to improve the present in vitro induction strategies of hepatic differentiation and maturation from hESCs to more productive level. Here we demonstrated the first evidence during process of hepatic differentiation from hESCs, which is that modulation of Cx32 as a key hepatic gap junction protein, to concisely regulate the hepatic lineage restriction and maturation of hESCs. Up-regulation of Cx32 expression and GJIC activity by the treatment of VK2 was shown to enhance the expression of hepatic markers and improve hepatic functions of hESC-Heps. On the other hand, inhibition of Cx32 by 2-APB blocked GJIC activity between cells, shown with the impeded hepatocyte maturation. These findings indicate that the establishment of natural and routine GJIC activity is critical for functional maturation of hESC-Heps. GJIC activity is regulated by Cx biosynthesis, transport and assembly as well as the formation and removal of gap junctions from the cell surface. Indeed, modulation of Cx32 expression levels from our findings supported the positive function of Cx32 for hESC hepatic differentiation and maturation. On the other hand, a stably maintained up-regulation as well as homeostasis for GJIC is also important for this process. Phosphorylation of Cxs has been implicated in the regulation of gap junction turnover. All members of MAP kinase family, including p38, ERKs, and JNK, have been reported to phosphorylate Cxs, leading to gap junction internalization and eventually degradation 34,38,39 . Previous studies showed that p38 inhibitor, SB, blocked the down-regulation of Cx32 and well maintained gap junction structures in vivo after partial hepatectomy 35 . In this study, we demonstrated that the mechanism of p38-mediated phosphorylation for Cx32 degradation exists in the dynamically process from hESCs to differentiation into hepatocytes. When SB was applied to the stem cells during differentiation, p38 phosphorylation-dependent degradation of Cx32 can be successfully blocked and resulted in promoting hepatic differentiation of hESCs. In contrast, activation of p38 by the activator anisomycin blocked differentiation and maturation of hESC-Heps. Taken together, these observations supply the evidence that the mechanism of MAP kinase down-regulation on Cx also existed in the in vitro dynamic process from hESCs to hepatic cells. Significantly, that the lineage maturation of hESC-Heps could be modulated by restraint effect of p38 inhibitor (e.g. SB) on gap junction internalization, besides the direct modulation on up-regulation of Cx32 expression levels by VK2.
Bile acid secretion and drug metabolites modification via apical bile canaliculi are based on the critical function of hepatocytes' epithelial polarization 41,42 . Gap junction plaques are frequently associated with tight junction strands in hepatocytes. Previous studies indicated that Cx32-mediated GJIC induced expression and function of tight junctions, which may affect cell polarity [43][44][45] . Our data indicated that the increased levels of expression and functions of Cx32 by SB or VK2 resulted in coordinated, functional polarization and bile canaliculi formation, whereas inhibition of Cx32 by 2-APB resulted in poorly polarized cells.
Moreover, for the first time, our results demonstrated that hESC-Heps carried the capacities to express the significantly increased levels of phase I and II enzymes, as well as phase III transporters, after the treatment with SB or VK2. In addition, they carried the significantly improved functional biotransforming systems, shown by HPLC analysis of drug metabolism. More importantly, the metabolic functions of hESC-Heps derived after SB-or VK2-treatments were enhanced by a common CYP inducer Rifampicin, with a significant increasing of some nuclear receptors, key mediators regulating drug-metabolizing enzymes and transporters 46 , which indicated those hESC-Heps developed a full biotransformation system. Remarkably, the gene expression profiles of hESC-Heps after SB-or VK2-treatments clustered closer to adult than to fetal hepatocytes. Overall, the hESC-Heps induced with SB-or VK2 showed improved maturation for hepatic characters, and these findings suggested that up-regulation of Cx32 may be very helpful for maturation of hepatic functions.
Our findings first proved the existence of regulation mechanism on GJIC during the in vitro dynamic process from hESCs to hepatocytes. Up-regulation of Cx32 as well as GJIC improved the hepatic differentiation from hESCs and following up hepatic maturation, which also raise the possibility that other members of Cxs and gap junction channels could be regulated for improving hepatic differentiation and maturation. Our modification on induction strategy could greatly benefit in guiding hepatic lineage restriction from pluripotent stem cells to hepatocytes, which presently faces the challenges to overcome that derived hepatic cells mostly have a fetal hepatocyte phenotype. In addition, our findings also supply new information to realize the developmental mechanism of liver organogenesis, which is communication via Cx32-mediated gap junctions may play essential for hepatic lineage restriction and maturation of fetal hepatocytes in embryonic livers.

Materials and Methods
Human Subjects. Fetal liver tissues were obtained from abortion in the Chinese PLA General Hospital Cell Culture. The human embryonic stem cell (hESCs) lines, H9 and H1, were grown in feeder-free conditions in six-well Nunclon surface plates (Nunc) coated with Matrigel (BD Biosciences) and maintained in mTESR1 media (Stem Cell Technologies). Cells were passaged at a 1:3~4 ratio using dispase (Invitrogen). All Matrigel plates were coated with a 1:80 dilution in Advanced DMEM-F12 (Life Technologies) and incubated at room temperature for at least 1 hr before use.

Isolation and Culture of Human Fetal Hepatocytes.
Fetal liver tissues at 21 gestational weeks were obtained from abortion with informed patient consent. Fetal liver cells were obtained as previously described 47 . The fetal liver tissue was cut into 1-3 mm 3 fragments for digestion in RPMI 1640 containing type IV collagenase and deoxyribonuclease (Sigma) at 37 °C for 15-20 min, and the cells were plated on Matrigel, Laminin, and collagen IV (3:1:1) mixed gel coated-plate in HCM (Lonza) supplemented with 10% fetal bovine serum. Twelve hours later, cells were washed with RPMI medium and subsequently cultured in HCM with 25 ng/ml HGF, 10 ng/ml OSM and 1 μ M Dex in the presence or absence of SB203580 for 4 days before messenger RNA (mRNA) analysis.
Fluorescence Recovery after Photobleaching (FRAP). FRAP measurements were performed by LSM 510 META using modifications of methods previously described 48 . After the cells incubated with 1 μ M Calcein-AM (Invitrogen) for 25 min in the confocal plates, a single cell was bleached for 25 s using a 488-nm laser beam. The selected cells were recorded for the pre-and post-bleach scans at low laser power, and recovery of fluorescence was recorded over 5 min at 15-s intervals. Two unbleached cells in the same visual field were selected as reference to subtract the loss of photobleaching during the acquisition process. Subsequent analysis of recorded frame sequences was done using LSM Image Examiner software (ZEISS). All FRAP measurements were performed on 15-30 cells per experiment and repeated at least three times. Carboxydichlorofluroscein Diacetate Assay. To evaluate bile canaliculi function, hESCs-derived hepatocytes were incubated with 2 μ M of 5 (and 6)-carboxy-2′ ,7′ -dichlorofluorescein diacetate (CDFDA, Sigma) for 30 min. Cells were subsequently washed with ice-cold PBS and imaging was performed on a confocal microscope (PerkinElmer). 5(6)-carboxy-2′ ,7′ -dichlorofluorescein (CDF)-positive canaliculi were normalized to cell number.
Quantitative real-time PCR analyses. Total RNA was isolated using an RNeasy extraction kit. RNA was reverse transcribed using Superscript II reverse transcriptase (Invitrogen) according to the manufacturer's instructions. Quantitative real-time PCR (qRT-PCR) was performed with SYBR Green real-time PCR master mix (TOYOBO) on a Bio-Rad iQ5 Real-Time PCR detection system (Bio-Rad). The data were analyzed using the delta-delta Ct method. The primers are listed in Supplementary Table S1.
Small interfering RNA transfection. The hESCs-derived hepatic progenitor cells were transfected with Cx32 siRNA (siCx32) or the non-specific siRNA (NSsiRNA) (Sigma) using Lipofectamine 2000 (Invitrogen). Following transfection, cells were incubated at 37 °C in a CO 2 incubator for 48 h before being harvested for the detecting the knockdown efficiency. And the cells were further differentiated to hepatocytes for evaluating the effect of Cx32 knockdown on hepatocyte differentiation. RNA sequencing. The mRNA was extracted using RNeasy Micro Kit (Qiagen) and followed by reversed transcription and amplification (REPLI-g WTA Single Cell Kit, Qiagen). RNA sequencing was performed on HiSeq X ten (Illumina). The low quality parts of raw reads were filtered. Reads were mapped to UCSC GRCh38 reference by Tophat version 2.0.11. Read counts of each gene were summarized using HTSeq version 0.6.1p1. DESeq2 package was employed to detect differentially expressed genes (fold change above 1 with adjusted p-values below 0.01). Original data were uploaded to the Gene Expression Omnibus database (accession number GSE76098). Immunofluorescent Staining. Cells were fixed with 4% paraformaldehyde for 20 min at room temperature and blocked with 10% goat or donkey serum for 1 h, followed by incubation with primary antibodies at 4 °C overnight. Labeled isotype-specific secondary antibodies were added and incubated 1 h at room temperature. Cells were counterstained with 4′ ,6-diamidino-2-phenylindole (DAPI) for visualization of cell nuclei and observed using a confocal microscopy (PerkinElmer) and Volocity Software (PerkinElmer). Antibodies used in this study were summarized in Supplementary Table S2.

Flow cytometry
Single cell suspensions were obtained by dissociation with Accutase for 3-5 min. Cell surface antigen staining was performed in PBS at 4 °C. Intracellular staining was performed with the BD Cytofix/Cytoperm ™ Kit (BD Biosciences) according to the manufacturer's instructions. Briefly, cells were fixed and permeabilized with BD Cytofix/Cytoperm solution for 20 min at 4 °C. Intracellular antigen staining was performed in BD Perm/Wash solution.
The stained cells were analyzed or sorted with BD FACSCalibur (BD Biosciences), and the data was analyzed using the Flowjo software (TreeStar). The sources and concentrations of primary, secondary antibodies and isotype controls are listed in Supplementary Table S2.
Immunoprecipitation and Western blotting. Cells were harvested in lysis buffer (50 mM Tris-HCl, pH 7.4, 0.25 mM sodium-deoxycholate, 150 mM NaCl, 2 mM EDTA, 0.1% sodium dodecyl sulfate, 1% Triton X-100) containing protease and phosphatase inhibitors (Roche). Lysates were sonicated for 30 s, maintained on ice for 30 min, and then spun at 15,000 rpm for 15 min at 4 °C. The supernatant was precleared with Protein G-Sepharose (GE Healthcare) and 100 μ g total protein was incubated with 2 μ g Cx32 antibody (Sigma) at 4 °C for 2 hrs with rotation. Thereafter, 100 μ l of Sepharose beads was added, and the sample was rotated at 4 °C for 4 hrs. The beads were collected and washed thoroughly in lysis buffer, and the precipitated proteins eluted with loading buffer (2 × Laemmli buffer plus dithiothreitol). Proteins were separated by 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis, transferred to polyvinylidene difluoride membranes, and probed with anti-Cx32, antiphosphoserine/threonine (Abcam), and horseradish-peroxidase-conjugated anti-rabbit or anti-mouse IgG antibodies. Proteins were detected by enhanced chemiluminescence HRP substrate (Millipore).
Statistics. Data are shown as means and standard deviations. For most statistic evaluation, 2-tailed Student's