Connexin-based channels contribute to metabolic pathways in the oligodendroglial lineage

Oligodendrocyte precursor cells (OPCs) undergo proliferation, migration and dynamic interactions with axons before myelination (Kang et al., 2010; Richardson et al., 2011; Rivers et al., 2008). In addition to myelination, oligodendrocytes also act as an energy source for axons by actively participating in monocarboxylate transporter 1 (MCT1, also known as SLC16A1)-mediated lactate delivery (Funfschilling et al., 2012; Lee et al., 2012). Recently, it has been found that OPCs directly promote angiogenesis to support the highly energy-consuming myelination process (Yuen et al., 2014), and oligodendroglial cells take up more energy substrates than neurons do at least in culture (Sanchez-Abarca et al., 2001). Thus, oligodendroglial cells require an extraordinary metabolic demand to support their development and function (Harris and Attwell, 2012; Nave, 2010). However, the detailed energy metabolism pathways of the oligodendroglial lineage are currently unclear.

In the present study, we considered the first step of energy consumption: energy substrate uptake through selective transporters (Hirrlinger and Nave, 2014). It is well known that astrocytes are the major energy source for neurons because they take up glucose through Gluts and provide the metabolite lactate for neurons through the ‘MCT-mediated astrocyte neuron lactate shuttle’ (Pellerin and Magistretti, 1994,, 2012). Similarly, oligodendrocytes can absorb extracellular glucose and/or lactate through Glut1 (also known as SLC2A1) and MCT1, respectively (Hirrlinger and Nave, 2014; Morrison et al., 2013; Rinholm et al., 2011; Saab et al., 2013). By contrast, OPCs do not express MCT1 (Lee et al., 2012), and there is a lack of evidence showing the expression of other Gluts in OPCs. Therefore, our main question is whether OPCs can obtain energy supply through other pathways such as non-selective energy uptake channels.

A typical feature of glial cells is their high expression of connexins, which can form gap junctions and/or hemichannels in different glial cell types. For instance, connexin 43 (Cx43, also known as GJA1) and connexin 30 (Cx30, also known as GJB6) are mainly expressed in astrocytes (Ransom and Giaume, 2013), whereas Cx47, Cx32 and Cx29 (also known as GJC2, GJB1 and GJC3, respectively) are present in oligodendroglial cells (Parenti et al., 2010; Theis et al., 2005). Recently, pathological changes of oligodendroglia or demyelination found in transgenic mice with different subsets of connexins (i.e. a Cx43 and Cx30 double knockout) suggest that glial connexins might participate in the regulation of the myelination or remyelination processes (Li et al., 2014; Markoullis et al., 2012a,,b,, 2014). In astrocytes, Cx43 gap junction channels and hemichannels are permeable to glucose, lactate and other metabolic substrates (Giaume et al., 2013; Ransom and Giaume, 2013). At a more integrated level, the astroglial networking mediated by gap junctions sustains neuronal activity through the intercellular trafficking of metabolites (Rouach et al., 2008). Based on these findings, it has been hypothesized that gap junction communication between astrocytes and oligodendrocytes might allow oligodendrocytes to obtain energy supplies from astrocytes (Hirrlinger and Nave, 2014; Morrison et al., 2013). However, it is not clear whether this gap-junction-mediated energy substrate pathway also exists between OPCs and astrocytes. Given that connexin-based channel functions are related to the development of astrocytes and oligodendrocytes (Tress et al., 2012; Venance et al., 1995; Von Blankenfeld et al., 1993), it is worthwhile exploring whether connexin-based channels also contribute to energy uptake in OPCs.

To address these questions, we took advantage of the ability of the fluorescent glucose analog, 2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-deoxyglucose (2-NBDG), which can permeate connexin-mediated gap junction channels and hemichannels (Retamal et al., 2007b; Rouach et al., 2008), to help analyze the glucose trafficking in oligodendroglial lineage cells. Here, we demonstrate that 2-NBDG can be taken up by OPCs through connexin hemichannels in a Ca²⁺-dependent manner, whereas it is transferred between astrocytes and oligodendrocytes through gap junction channels. We also show that connexin-hemichannel-mediated glucose uptake supports OPC proliferation. Taken together, our findings indicate that connexin-based channels contribute to energy uptake pathway in oligodendroglial lineage cells.

Oligodendroglial cells pass through distinguished development stages with specific biomarkers expression. Briefly, platelet-derived growth factor a (PDGFRa) is used as an early developmental marker to identify OPCs. Immature oligodendrocytes are identified by O4 (marker specific for the oligodendroglial lineage; Sommer and Schachner, 1981; Bansal et al., 1989) or CNPase (also known as CNP, 2′,3′-cyclic nucleotide 3′-phosphodiesterase), and MBP and CC1 (also known as APC, adenomatous polyposis coli) are both mature oligondendrocyte markers (Emery, 2010). In our study, most OPCs (PDGFRa positive) differentiate into immature oligodendrocytes (O4 positive) on the third day, and into mature oligodendrocytes (MBP positive) on the sixth day after being induced to differentiate in vitro (Fig. 1). Double immunostaining results showed that only Cx29 and Cx47, but not Glut1, were detected in OPCs and immature oligodendrocytes (Fig. 1A,B). At more advanced stages of differentiation, namely after 6 days differentiation in culture, all three connexins (i.e. Cx29, Cx32 and Cx47) and Glut1 expression were observed in mature oligodendrocytes (Fig. 1C). The percentages of oligodendroglial cells positive for connexins at different time in culture showed that PDGFRa⁺ OPCs dominantly express Cx47 and Cx29 (Fig. 1D), and the same expression pattern was also observed in the corpus callosum of postnatal developing mice (Fig. S1). The expression levels of connexins and Glut1 were further determined by western blot analysis as well as quantitative PCR (qPCR) upon differentiation of OPCs in cultures (Fig. 1E,F). Moreover, Glut2 and Glut3 (neuronal specific) were not found in OPCs (Fig. S1D). Based on the different expression patterns of connexins and Gluts between OPCs and oligodendrocytes, we hypothesized that oligodendroglial connexins might differently contribute to metabolic pathways (i.e. glucose uptake) in oligodendroglial lineage cells.

To examine the possibility of metabolic coupling between astrocytes and oligodendroglial lineage cells, we studied glucose trafficking in an astrocyte-oligodendroglia co-culture system by dye coupling (Giaume et al., 2012). Briefly, the cells were patched with a whole-cell recording pattern, and the intercellular diffusion of the dye was monitored after 20 min of recording. At the beginning of the whole-cell configuration (1 min), 2-NBDG or sulforhodamine B (SRB) filled up CC1⁺ mature oligodendrocytes. After 20 min, these probes diffused from oligodendrocytes to the astrocyte layer underneath (Fig. 2A). Specifically, 2-NBDG was time dependently transferred to astrocytes (Fig. 2B), and this intercellular diffusion was blocked by carbenoxolone (CBX) (Fig. 2A). However, when the same experiment was performed on OPCs (PDGFRa⁺), 2-NBDG and SRB only filled up the recorded OPCs but were not detected in the underlying astrocytes after 20 min recording (Fig. 2C), indicating that glucose can be transferred between oligodendrocytes and astrocytes but not between OPCs and astrocytes. This result raised the question of how glucose enters OPCs.

As functional gap junctions between astrocytes and OPCs were not detected in the early developmental stages of oligodendroglial cells (Fig. 2) whereas Cx29 and Cx47 are already expressed, we examined their hemichannel function by performing an ethidium bromide (EtBr) uptake assay in cultured OPCs (Giaume et al., 2012). As illustrated in Fig. 3, OPCs (PDGFRa⁺) exhibited hemichannel activity in normal culture conditions (solution with 1 mM Ca²⁺), which could be blocked by CBX and La³⁺ but not by the Glut1 inhibitor STF31. This hemichannel activity was increased in Ca²⁺-free condition known to trigger hemichannel opening (Fig. 3A). The uptake assay was also performed on mature oligodendrocytes (CC1⁺) under the same condition, which showed that EtBr uptake activity in oligodendrocytes was less than that in OPCs (Fig. 3A,a1,B,b1). To test whether hemichannels in OPCs and oligodendrocytes were permeable to glucose, we performed a dye uptake assay with 2-NBDG (Retamal et al., 2007a). The uptake ratio for this fluorescent glucose analog was similar to that monitored by the uptake of EtBr, indicating that both OPCs and oligodendrocytes uptake glucose from the extracellular medium through connexin hemichannels; however, hemichannel-dependent glucose analog uptake was more pronounced in OPCs than that in oligodendrocytes (Fig. 3A,a2,B,b2). In addition, 2-NBDG uptake in OPCs could be blocked by hemichannel inhibitors but not by the Glut1 inhibitor STF31 or Cytochalasin B, which has also been shown to inhibit Gluts (Griffin et al., 1982) (Fig. 3A,B; Fig. S2). Thus, hemichannels might be the main contributor for glucose uptake in oligodendroglial cells, specifically at the OPC stage.

Because it has been reported that intracellular Ca²⁺ ([Ca²⁺]_i) elevation triggers the opening of Cx43 and Cx32 hemichannels in other cell types (De Vuyst et al., 2006; Wang et al., 2013), we wondered whether hemichannel activity in the oligodendroglial lineage was also dependent on Ca²⁺. Thus, we monitored cytoplasmic [Ca²⁺]_i in OPCs and oligodendrocytes in normal culture conditions following Rhod-2 loading. Real-time recordings showed that most OPCs exhibited spontaneous ‘oscillatory’-like Ca²⁺ signaling with peak and plateau transients, whereas oligodendrocytes showed ‘flat’ Ca²⁺ signaling (Fig. 4A,B). As OPCs were characterized by higher [Ca²⁺]_i signal and hemichannel activity compared to oligodendrocytes, we focused our investigation on OPCs. The cytoplasmic [Ca²⁺]_i and the hemichannel activity were monitored under different conditions, including chelating cytoplasmic Ca²⁺ with BAPTA-AM or increasing [Ca²⁺]_i with ionomycin. We found that treatment of BAPTA-AM inhibited the spontaneous oscillatory-like Ca²⁺ signaling and reduced the 2-NBDG or EtBr uptake in OPCs (Fig. 4C–E). However, the relative 2-NBDG uptake rate was significantly increased in the ionomycin-treated group, and this effect was inhibited by CBX treatment (Fig. 4D). Taken together, these results indicate that hemichannel activity depends on intracellular Ca²⁺ elevation in OPCs.

Glucose is considered as the most important energy source in the brain and OPC development might rely on it. To determine the role of connexin-hemichannel-mediated glucose uptake in supporting OPC development, we firstly tested the effect of glucose on OPC proliferation. Purely cultured OPCs were fed with OPC-proliferation media containing three different concentrations of glucose (0, 0.75 and 1.5 mg/ml). Usually 1.5 mg/ml is considered as the normal extracellular glucose concentration. Olig2 is an oligodendroglial lineage marker which is expressed through the whole development process, and we use Ki67 and Olig2 double-positive cells to mark the proliferating OPCs as described previously (Niu et al., 2012b). When exposed to media with reduced concentrations of glucose (0 mg/ml and 0.75 mg/ml), the number of viable cells and Ki67 and Olig2 double-positive proliferating OPCs were significantly decreased in a concentration-dependent manner (Fig. 5A,B). Moreover, in the presence of a normal glucose concentration, the blockade of connexin hemichannels in OPCs with CBX or La³⁺ resulted in a significant decrease of OPC proliferation, indicated by decreased viable cell numbers or Ki67 and Olig2 double-positive OPC numbers (Fig. 5C–E). These results indicate that glucose uptake through hemichannels contributes to OPC proliferation.

Cx43 is highly expressed by astrocytes, and works as hemichannels and/or homomeric gap junction channels (Cx43–Cx43) between astrocytes (Evans et al., 2013; Mitterauer, 2015; Ransom and Giaume, 2013; Ye et al., 2009), and also forms heteromeric gap junctions (Cx43–Cx47) between astrocytes and oligodendrocytes (Orthmann-Murphy et al., 2007; Theis et al., 2005). It has been reported that in astrocytes, Cx43 deletion increases glucose uptake by a compensatory upregulation of Glut transporters and/or glucose metabolism enzymes (Gangoso et al., 2012). To determine the influence of astroglial glucose over-consumption on the oligodendroglial lineage, we used the astrocyte Cx43 conditional knockout mouse (hGFAP^Cre/+:Cx43^fl/fl, Cx43-KO). Interestingly, it was found that the glucose concentration in the medium collected from Cx43-KO astrocytes was significantly lower than that from the wild-type astrocytes (Fig. 6A). Similar results were found in wild-type astrocytes treated with the connexin-based channel blocker CBX (Fig. 6B).

In addition, we performed an EtBr uptake assay in corpus callosum of acute brain slices from postnatal day (P)14 mice or astrocyte–OPC co-cultures to detect hemichannel function in OPCs under different conditions. We found that the hemichannel function in OPCs was blocked by CBX and increased in Ca²⁺-free solution, but was not affected by Cx43 deletion in astrocytes (Fig. 6D,E). However, decreased OPC proliferation was found in the corpus callosum of postnatal Cx43-KO mice as well as OPCs co-cultured with astrocytes from Cx43-KO mice (Fig. 6F,G; Fig. S3), whereas no significant difference in PDGF, bFGF, CNTF and lactate levels were found after Cx43 deletion in astrocytes (Fig. S4). Moreover, the expression of oligodendroglial connexins was not significantly altered by Cx43 knockout in astrocytes (Fig. 6C).

Finally, we performed rescue experiments and found that the decrease in OPC proliferation observed after Cx43 deletion in astrocytes could be compensated by an external glucose supply in astrocyte–OPC co-cultures (Fig. 7A). To further confirm that astrocytes affect OPC proliferation in a non-cell-autonomous manner, we designed sandwich co-cultures in which OPCs were seeded on coverslips in contact with astrocyte medium but not directly on astrocyte monolayers, as shown in the diagram (Fig. 7B). In these sandwich co-cultures, similar results were obtained as that in the astrocyte–OPC co-cultures (Fig. 7C,D). Taken together, these results indicate that Cx43 deletion in astrocytes affects glucose level in the extracellular medium but does not impact on the connexin hemichannel function of OPCs.

OPCs and oligodendrocytes need to go through a series of energy-consuming developmental events, including proliferating to enrich the population in central nervous system (CNS) tissues, migrating and/or distributing into their destinations and wrapping axons to form myelin (Kang et al., 2010; Richardson et al., 2011; Rivers et al., 2008). However, the mechanisms underlying their energy metabolism, in particular the pathways involved in energy metabolite uptake and trafficking remain unidentified. Functional tests have shown that astrocytes and oligodendrocytes are connected by gap junctions to form panglial networks (Griemsmann et al., 2014; Ransom and Giaume, 2013; Tress et al., 2012). Moreover, astroglial connexin channels (gap junction channels and hemichannels) are permeable to glucose derivatives (2-NBDG) in both in vitro and ex vivo models (Blomstrand and Giaume, 2006; Retamal et al., 2007a), thus they provide the basis to form metabolic intercellular networks (Rouach et al., 2008). Interestingly, gap junction channels are ∼65% less permeable to the phosphorylated 2-NBDG-6P than to 2-NBDG (Rouach et al., 2008), indicating that the phosphorylated form of the fluorescent glucose analog is restricted in the glial cells. As glial hemichannels exhibit similar permeability properties with gap junction channels (Giaume et al., 2013), it is expected that once absorbed in OPCs by hemichannels, 2-NBDG is phosphorylated and then stays trapped within the cells for further metabolic requirement, such as glycolysis. Based on these findings, and to further understand the metabolic role of glial connexins in metabolic supply, we investigated whether connexin-based channels contribute to glucose trafficking pathways in oligodendroglial cells. Here, we provide evidence demonstrating that a glucose analog cannot exchange between OPCs and astrocytes due to the absence of gap junction, whereas hemichannel activity supports glucose uptake in OPCs, which is essential for their proliferation. In addition, for the first time, we show that a glucose analog can be transferred through oligodendrocyte–astrocyte gap junctions, which provides the basis for a panglial metabolic route that previously had only been suggested but not demonstrated (Morrison et al., 2013).

During brain development, oligodendroglial-specific Cx47 and Cx32 are detected at embryonic stages, whereas Cx29 appears at P0. Cx47 expression increases successively in regions populated with developing oligodendroglia and declines from early postnatal stage to adulthood (Parenti et al., 2010). Here, we used OPC cultures to further demonstrate the dynamic expression pattern of connexins during oligodendroglial development in vitro, suggesting a role for connexin-based channels in this process (Li et al., 2014). Although the combination of different connexins between oligodendrocytes and astrocytes (Cx47–Cx43, Cx47–Cx30, Cx32–Cx30 and Cx32–Cx26) has been identified, and a heterotypic gap junction activity has been confirmed by dye coupling and/or electrical measurements (Magnotti et al., 2011), the types of metabolic substrates that are exchanged between these two glial cell types have not been investigated yet. Current data have shown that this ‘panglial’ networking of oligodendrocytes and astrocytes serves to spatially buffer K⁺, allow water transport and serve as bi-directional channels for the spreading of Ca²⁺ waves (Kamasawa et al., 2005; Menichella et al., 2006; Parys et al., 2010; Wallraff et al., 2006). Here, we provide direct evidence that oligodendrocyte–astrocyte gap junctions are also permeable to the fluorescent glucose analog 2-NBDG. Considering the capacity of astrocytes to uptake, store and supply energy substrates (Rouach et al., 2008), it is likely that glucose and/or its metabolites might be transferred from astrocytes to oligodendrocytes through oligodendrocyte–astrocyte gap junctions as recently hypothesized (Morrison et al., 2013). In this regard, our findings might provide a new understanding of a connexin-channel-mediated metabolic supply alternative pathway besides Glut- or MCT-mediated pathways in the oligodendroglial lineage.

However, using 2-NBDG, we could not observe such gap-junction-mediated glucose transport between OPCs and adjacent astrocytes. Instead, we showed for the first time that connexin-based hemichannels in OPCs are responsible for glucose analog uptake from the extracellular environment, though there is still a lack of evidence to show which connexin protein plays the dominant role in this process. These in vitro results imply that a functional gap junction channel might not exist at early OPC developmental stages until they differentiate into GalC-positive oligodendrocytes (Venance et al., 1995). Consistently, it has been demonstrated that cells positive for the cell surface ganglioside A2B5, and NG2 glial cells do not electrically or dye couple with astrocytes whereas mature oligodendrocytes do (Bergles et al., 2000; Lin and Bergles, 2004; Xu et al., 2014). Importantly, given that neither Gluts (i.e. Glut1, Glut2 and Glut3) nor MCT1 was detectable in OPCs in our studies and in studies by other (Lee et al., 2012), whereas 2-NBDG uptake in OPCs can be blocked by CBX but not by Glut inhibitors such as STF31 and Cytochalasin B, it is likely that connexin-mediated hemichannels serve as a major metabolic supply pathway in OPCs. Thus, the crucial developmental step of myelination might depend on a connexin-channel-mediated glucose supply (Rinholm et al., 2011; Yan and Rivkees, 2006).

Normally it is thought that glial hemichannels are either kept closed, to maintain cellular integrity (Spray et al., 2006), or active, as in certain brain areas such as the hippocampus (Chever et al., 2014) and the olfactory bulb (Roux et al., 2015). Moreover, they are triggered to open by some stimulation, such as inflammation, mechanical stress or ischemia (Batra et al., 2012; Johansen et al., 2011; Karpuk et al., 2011). In our study, however, we demonstrated for the first time that connexin-mediated hemichannels are responsible for glucose uptake in OPCs, implying a role for oligodendroglial connexin hemichannels under physiological conditions. We further revealed that the activation of hemichannels in the oligodendroglial lineage depends on intracellular Ca²⁺ signaling, which can trigger connexin hemichannel opening in other cell types (De Vuyst et al., 2006). Interestingly, we observed that OPCs exhibit stronger spontaneous Ca²⁺ oscillations than oligodendrocytes do (Fig. 4) as recently reported (Cheli et al., 2015), which might explain why more glucose uptake through connexin hemichannels occurred in OPCs compared to oligodendrocytes (Fig. 3). In this regard, our data shed light on different energy substrate uptake mechanisms and pathways between OPCs and oligodendrocytes. In OPCs, connexin hemichannels provide a main route for glucose entry, whereas in oligodendrocytes, glucose and lactate transporters, supplemented by the gap junctions between astrocytes and oligodendrocytes, as well as their hemichannels contribute to the energy substrate uptake and traffic.

As elegantly stated by A. Harris (2007), “Although connexin-based channels are considered as poorly selective channels leading to an idea that they are ‘non-specific large conductance channels', numbers of studies demonstrate that there are dramatic, unanticipated and connexin-specific differences in the channel permeability to cytoplasmic molecules, particularly among those thought to directly mediate intercellular signaling (i.e. cAMP, IP₃,…). The data strongly indicate that certain biological molecules have highly specific interactions within connexin pores that enable surprising degrees of selective permeability that cannot be predicted from simple considerations of pore width or charge selectivity” (Harris, 2007). Based on these statements, it seems reasonable to consider that although we show here that astrocyte–oligodendrocyte gap junction channels and OPCs hemichannels are permeable to 2-NBDG, it does not automatically imply that all cytoplasmic or extracellular signaling molecules with small molecular mass can permeate through these channels.

Finally, we used a conditional astrocytic Cx43 knockout (Cx43-KO) mouse as a glucose-deprivation model to confirm the importance of connexin-channel-based energy support on the OPC proliferation in both in vivo and co-culture systems. In these mice, glucose uptake in Cx43 lacking astrocytes is increased by compensatory upregulation of Glut1, Glut3 and type I/II hexokinase expression (Gangoso et al., 2012; Tabernero et al., 2006). As predicted, we observed that OPC proliferation was inhibited by Cx43 deletion in astrocytes without impacting on OPC hemichannel properties. Moreover, this effect can be rescued by a compensatory extracellular glucose supply, indicating the importance of astrocytes in maintaining the glucose levels of the CNS extracellular matrix. Another point that needs caution is that lactate is also an important metabolic substrate for the brain during the early postnatal period (Barros, 2013; Rinholm et al., 2011). Given the extensive permeability of hemichannels for various metabolites (Giaume et al., 2013), there is no evidence to exclude the uptake of lactate or other energy substrates through hemichannels in OPCs. In fact, it has already been reported that gap junction channels in astrocytes are permeable to lactate in vitro and ex vivo (Tabernero et al., 1996; Rouach et al., 2008). Given that various connexin channels exhibit different functional properties, including size selectivity, charge selectivity and voltage or chemical gating, we still have no evidence to show that connexin channels in OPCs are permeable or impermeable to other substrates, such as Ca²⁺, K⁺ and metabolites.

Because OPCs need to go through a series of highly energy-consuming developmental events (Barateiro and Fernandes, 2014) and likely cannot get enough energy support from Glut1, MCT1 or gap junctions as oligodendrocytes do, they might rely more on the hemichannel-mediated energy supply pathway (Fig. 8). This dependence might partly explain why OPCs are more vulnerable to oxygen-glucose deprivation than oligodendrocytes (Fern and Moller, 2000; Ziabreva et al., 2010). Moreover, likely due to deficient Cx47-based hemichannel properties, Cx47-null mice exhibit decreased oligodendroglial lineage cell numbers at P14, but gene knockdown does not influence myelination in adult mice (Tress et al., 2012). Finally, abnormal connexin expression is involved in limiting OPC recruitment and consequent remyelination failure in multiple sclerosis patients and in autoimmune encephalomyelitis mice (a model of multiple sclerosis) (Markoullis et al., 2012a,,b,, 2014).

In this study, we focused on connexin-mediated energy supply pathways in oligodendroglial lineage cells, providing direct evidence that glucose can be delivered between astrocytes and oligodendrocytes through gap junction channels, whereas spontaneous intracellular Ca²⁺ signaling triggers connexin hemichannel activity that enables glucose uptake in OPCs that supports their proliferation. Identification of the dual functions of connexin-based channels in the oligodendroglial lineage, along with their metabolic roles, might provide valuable insights into the energy supply pathways of oligodendroglial cells and might result in new therapeutic strategies for energy-related diseases.

Cx43^flox/flox and hGFAP^Cre/+ mice were purchased from the Jackson Laboratory. These two mouse strains were paired to produce offspring, either hGFAP^Cre/+:Cx43^fl/fl (Cx43-KO) or hGFAP^+/+:Cx43^fl/fl (wild type). The offspring were genotyped by PCR for Cre and floxed Cx43 alleles. More than four animals were tested in each group. All animal studies were conducted with the approval of the Laboratory Animal Welfare and Ethics Committee of the Third Military Medical University (TMMU) Administrative Panel on Laboratory Animal Care.

Acute brain slices were prepared from brains of P14 Cx43-KO and wild-type littermate mice using a vibroslicer (Thermo Scientific, Microm HM650V) as previously described (Geiger et al., 2002). Slices were incubated with the hemichannel-permeable fluorescent tracer ethidium bromide (EtBr) (394.3 g/mol, 4 µM final concentration, ThermoFisher, Grand Island, NY, USA) in oxygenated artificial cerebro-spinal fluid (ACSF) for 10 min at room temperature. After fixation with 4% paraformaldehyde (PFA), the slices were incubated with anti-platelet-derived growth factor a receptor (PDGFRa) primary antibody (1:200, cat. no. sc-338; Santa Cruz Biotechnology, Dallas, TX, USA) overnight at 4°C after pre-incubation in the blocking buffer for 1 h (containing 0.2% gelatin and 1% Triton X-100), followed by incubation in fluorescent-conjugated secondary antibody at room temperature for 1 h. Images were taken using a confocal laser-scanning microscope (Olympus, IV1000) at the selected specific wavelength (561 nm). More than eight images were captured in each group for statistical analysis.

Primary astrocyte cultures were prepared from brain hemispheres of Cx43-KO and wild-type littermate mice pups (P1–P3) as previously described (Giaume et al., 2012). The purified astrocytes were plated into six-well plates (10⁵ cells per well). The OPC cultures were prepared as previously described (Niu et al., 2012b). Briefly, the mixed glial cells were isolated from cortex of P1–P3 animals and enriched in OPC growth medium followed by a two-passage purifying processes. The purified OPCs were induced to differentiate by using OPC differentiation medium [DMEM/F12 + 1% N2 supplement + 5 g/ml NAC + 1% FBS; Dulbecco's modified Eagle's media/F12 (DMEM/F12, cat. no. SH30023, Hyclone, Logan, Utah, USA), N2 supplement (cat. no. 17502048, Life Technologies), fetal bovine serum (FBS, cat. no. SV30087, Hyclone), N-acetyl-l-cysteine (NAC, cat. no. 0LA0011, AMERSCO, Solon, OH, USA)]. For co-cultures, OPCs were seeded either on an astrocyte monolayer to set up astrocyte–OPC co-cultures or on poly-D-lysine-coated glass coverslips placed above the astrocytes to set up a sandwich co-culture system. We used 4.5×10⁴ cells per well in 24-well plates and 10⁶ cells per well in 60-mm dishes.

OPCs or oligodendrocytes were incubated with 4 µM EtBr or 500 µM 2-NBDG (342.3 g/mol, neutral, Life Technologies, Carlsbad, CA) for 30 min at room temperture in different conditions (Giaume et al., 2012), including normal culture medium (OPCs are in OPC proliferation medium and oligodendrocytes are in OPC differentiation medium) or culture medium pre-incubated with carbenoxolone (CBX, Sigma, St Louis, MO, USA), La³⁺ ions (lanthanum, Sigma), Glut1 inhibitor (STF31, Tocris Bioscience, Bristol, UK), Cytochalasin B (Sigma), Cytochalasin D (Enzo Life Sciences, East Farmingdale, NY) or Ca²⁺-free solution. Then, the cells were fixed with 4% PFA for 15 min at room temperature. Immunofluorescence pictures were captured using a confocal laser-scanning microscope (Olympus, IV1000) at an appropriate wavelength (EtBr, 561 nm; 2-NBDG, 488 nm). The EtBr fluorescent signal in the nuclei and the 2-NBDG fluorescent signal in cell bodies were analyzed using Mac Imaris 7.4.0 software. The mean fluorescence intensity per µm³ was quantified in arbitrary units (AUs) (Giaume et al., 2012).

Mouse pups of different ages (P7, P14, P21) were anesthetized with 1% pentobarbital and transcardially perfused with 4% PFA. Brains were dissected and cryoprotected in 30% sucrose at 4°C. Serial coronal sections (20 μm) were obtained using a cryostat microtome (MS 1900, Leica, Wetzlar, Germany). OPCs on the cover slips were fixed as previously described (Niu et al., 2012a). Brain sections or cell cultures were blocked with 0.5% bovine serum albumin (BSA) and 0.2% Triton X-100 for 1 h and then incubated with primary antibodies overnight at 4°C followed by the fluorescence-conjugated secondary antibodies at room temperature for 1 h. Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI, 0.1 μg/ml, ThermoFisher). Rabbit anti-PDGFRa (1:200, cat. no. sc-338; Santa Cruz, Dallas, TX, USA) or rat anti-PDGFRa (1:100, cat. no. ab61219; Abcam, Cambridge, UK) antibodies were used to label OPCs. Mouse anti-O4 (1:100, cat. no. O7139; Sigma) antibody was used to label immature oligodendroglial cells, and goat anti-MBP (1:300, cat. no. sc-13914; Santa Cruz Biotechnology) or mouse anti-CC1 (1:1000, cat. no. OP80; Millipore, Billerica, MA) antibodies were used to label mature oligodendrocytes. Mouse anti-Olig2 (1:200, cat. no. MABN50; Millipore) antibody was used to label all oligodendroglial lineage cells. Rabbit anti-Ki67 (1:1000, cat. no. RM-9106; Thermo Scientific) antibody was used to detect proliferative cells. Rabbit anti-Cx29 (1:100, cat. no. sc-68377; Santa Cruz Biotechnology), rabbit anti-Cx32 (1:100, cat. no. C3595; Sigma), rabbit anti-Cx47 (1:200, cat. no. 364700; Life Technologies; or 1:100, sc-30335-R; Santa Cruz Biotechnology), mouse anti-Glut1 (1:500, cat. no. ab40084; Abcam), goat anti-Glut2 (1:100, cat. no. sc-7580; Santa Cruz Biotechnology) and mouse anti-Glut3 (1:100, cat. no. ab41525; Abcam) antibodies were used to detect connexin proteins and Glut in oligodendroglial cells. Mouse anti-Cx43 (1:100, cat. no. 610062; BD, Erembodegem, Belgium) antibody was used to label Cx43 in astrocytes. Appropriate Alexa-Fluor-conjugated secondary antibodies included donkey anti-mouse, rabbit and goat IgG (1:1000, Life Technologies). The immunofluorescent signal was determined using a fluorescence microscope (Olympus BX-60) or a confocal laser-scanning microscope (Olympus, IV 1000) with excitation wavelengths appropriate for Alexa Fluor 488 (488 nm, Life Technologies), Alexa Fluor 568 (568 nm) or Alexa Fluor 647 (647 nm). Cell counting and fluorescence intensity analyses were conducted on nine randomly chosen fields under a 20× objective lens for each sample using an Image Pro Plus image analysis system.

SDS-PAGE western blotting was performed as previously described (Niu et al., 2010). Proteins were transferred to polyvinylidenedifluoride membranes and visualized by chemiluminescence (ECL Plus, GE Healthcare, Marlborough, MA, USA) after incubation with antibodies. β-actin (1:1000, cat. no. sc- 47778; Santa Cruz Biotechnology) was used as the loading control. Quantification of band intensity was analyzed using the Image Pro Plus software. The primary antibodies included: rabbit anti-PDGFRa (1:500, cat. no. sc-338; Santa Cruz Biotechnology), goat anti-MBP (1:1000, cat. no. sc-13914; Santa Cruz Biotechnology), mouse anti-Olig2 (1:1000, cat. no. MANB50; Millipore), rabbit anti-Cx29 (1:500, cat. no. sc-68377; Santa Cruz Biotechnology), rabbit anti-Cx32 (1:500, cat. no. C3595; Sigma), rabbit anti-Cx47 (1:1000, cat. no. 364700; Life Technologies; 1:500, cat. no. sc-30335-R; Santa Cruz Biotechnology) and mouse anti-Glut1 (1:2000, cat. no. ab40084; Abcam).

The number of viable cells was estimated using the CCK-8 assay, which provides an effective and reproducible measurement of OPC proliferation (Xiao et al., 2008). Briefly, OPCs were cultured at a density of 1.5×10⁴ cells per well in poly-D-lysine-coated 96-well plates containing 100 µl OPC proliferation medium [DMEM/F12 + 1% N2 supplement + 10 ng/ml bFGF + 10 ng/ml PDGF-AA; PDGF-AA (cat. no. 100-13A, Peprotech, Rocky Hill, NJ, USA), bFGF (cat. no. 100-18B, Peprotech)]. After 12 h, the cells were subjected to different conditions, and at the indicated time points, 10 µl of CCK-8 solution reagent (Dojindo, Cell Counting Kit-8) was added to each well according to the manufacturer's instructions, followed by a 4 h incubation. The relative viable cell numbers were determined by measuring the absorbance at 450 nm using a microplate reader (Bio-Rad, Model 680). Three independent experiments were performed in triplicate.

Glucose concentrations were determined using a glucose assay kit (Abcam) according to the manufacturer's instructions. Supernatants from 48 h cultured astrocytes in different conditions were collected for the glucose assay. Treatment with CBX (50 µM, 48 h) was used to mimic Cx43 deletion in astrocytes. The optical density (OD) value was measured according to the absorbance at 570 nm using a microplate reader.

Astrocyte–OPC co-cultures were used for dye-trafficking experiments. Recorded cells were patched in current-clamp mode using an Axon Instruments system (Axopatch 200B). The oligodendroglial cells were patched with a whole-cell recording technique, using 5–10 MΩ glass electrodes filled with an internal solution as previously described (Giaume et al., 2012; Rouach et al., 2008). To monitor intercellular diffusion, 2-NBDG (2 mg/ml) and sulforhodamine B (SRB) (558.7 g/mol, 2 negative charges, 1 mg/ml, ThermoFisher) were injected into either oligodendrocytes or OPCs during whole-cell recordings and images were captured after 1 and 20 min to confirm the diffusion of dye among cells. The connexin channel blocker CBX was applied in astrocyte-oligodendrocyte co-cultures to verify connexin-channel-dependent glucose transport. Images of intercellular diffusion of these fluorescent dyes were captured after 1 or 20 min, and the coupling of these cells was analyzed with an Olympus CCD and Clampex 10.0 software. Dye-trafficking experiments were carried out for more than six cells in each group.

OPCs were cultured on glass-bottomed dishes in OPC proliferation medium and then loaded with the fluorescent Ca²⁺-sensitive dye Rhod-2 (5 μM, Life Technologies) for 30 min at 37°C. The loaded cells were rinsed three times with modified Ca²⁺ imaging buffer solution before confocal imaging. OPCs were incubated in fresh modified Ca²⁺ imaging buffer (NaCl 125 mM; KCl 5 mM; CaCl₂ 1 mM; MgSO₄ 1.2 mM; glucose 5 mM and HEPES 25 mM) with 2-NBDG (500 µM) and were real-time recorded using a confocal laser-scanning microscope (Olympus). During recording, cytoplasmic Ca²⁺ concentration and 2-NBDG uptake were measured by exciting Rhod-2 at 581 nm and exciting 2-NBDG at 488 nm, respectively. To chelate intracellular Ca²⁺, OPCs were pretreated with medium containing BAPTA-AM (25 µM, ThermoFisher) for 20 min before recording. Ionomycin (1 µM, ThermoFisher), a Ca²⁺ ionophore that allows Ca²⁺ ions to enter the cells through pores made in the plasma membrane, was applied to increase the intracellular Ca²⁺. The real-time Ca²⁺ and 2-NBDG levels were recorded using a confocal laser-scanning microscope (Olympus, IV1000). Recordings were analyzed with Fluoview Image Processing Software (V2.1) and GraphPad Prism 5 Software. Ca²⁺ fluorescence signals and 2-NBDG fluorescence signals were analyzed in regions of interest (ROIs) covering the somata of oligodendroglial cells. Normalized changes of Rhod-2 fluorescence intensities were calculated as ΔF/F=(F−F0)/F0 (F, fluorescence intensity; F0, baseline intensity) (Otsu et al., 2015). And the relative changes of 2-NBDG fluorescence intensities were calculated as the (F−F0)/S [F, 2-NBDG fluorescence intensity at end time point; F0, baseline 2-NBDG fluorescence intensity; S, time (seconds) from start point of treatment to the end time point of experiment].

Total ribonucleic acid (RNA) was isolated from brains or cultured astrocytes of wild-type and Cx43-KO mice using TRIzol (Life Technologies) and the RNeasy Plus Mini Kit (Qiagen, Hilden, Germany). Real-time quantitative PCR (qPCR) was performed with the C1000 Touch™ Real-time PCR Detection System (Bio-Rad) and GoTaq^® qPCR Master Mix (Promega, Sunnyvale, CA, USA). The oligonucleotide primers, amplification procedure, and melt curve analysis were performed. For each sample, three independent repeats were performed. Each sample was tested in triplicate.

Statistical significance between groups was determined with GraphPad Prism 5 software. Statistical analyses were performed by one-way analysis of variance (ANOVA) followed by Tukey's post-hoc test. Comparisons between two experimental groups were made using an unpaired t-test. A probability of P<0.05 was considered statistically significant.

The authors wish to thank Jia Lou for her assistance in preparing the figures.