Connexin Signaling Is Involved in the Reactivation of a Latent Stem Cell Niche after Spinal Cord Injury

The ependyma of the adult spinal cord is a latent stem cell niche that is reactivated by spinal cord injury contributing new cells to the glial scar. The cellular events taking place in the early stages of the reaction of the ependyma to injury remain little understood. Ependymal cells are functionally heterogeneous with a mitotically active subpopulation lining the lateral domains of the central canal (CC) that are coupled via gap junctions. The ependyma of the adult spinal cord is a latent stem cell niche that is reactivated by spinal cord injury contributing new cells to the glial scar. The cellular events taking place in the early stages of the reaction of the ependyma to injury remain little understood. Ependymal cells are functionally heterogeneous with a mitotically active subpopulation lining the lateral domains of the central canal (CC) that are coupled via gap junctions. Gap junctions and connexin hemichannels are key regulators of the biology of neural progenitors during development and in adult neurogenic niches. Thus, we hypothesized that communication via connexins in the CC is developmentally regulated and may play a part in the reactivation of this latent stem cell niche after injury. To test these possibilities, we combined patch-clamp recordings of ependymal cells with immunohistochemistry for various connexins in the neonatal and the adult (P > 90) normal and injured spinal cord of male and female mice. We find that coupling among ependymal cells is downregulated as postnatal development proceeds but increases after injury, resembling the immature CC. The increase in gap junction coupling in the adult CC was paralleled by upregulation of connexin 26, which correlated with the resumption of proliferation and a reduction of connexin hemichannel activity. Connexin blockade reduced the injury-induced proliferation of ependymal cells. Our findings suggest that connexins are involved in the early reaction of ependymal cells to injury, representing a potential target to improve the contribution of the CC stem cell niche to repair. SIGNIFICANCE STATEMENT Ependymal cells in the adult spinal cord are latent progenitors that react to injury to support some degree of endogenous repair. Understanding the mechanisms by which these progenitor-like cells are regulated in the aftermath of spinal cord injury is critical to design future manipulations aimed at improving healing and functional recovery. Gap junctions and connexin hemichannels are key regulators of the biology of neural progenitors during development and in adult neurogenic niches. We find here that connexin signaling in the ependyma changes after injury of the adult spinal cord, functionally resembling the immature active-stem cell niche of neonatal animals. Our findings suggest that connexins in ependymal cells are potential targets to improve self-repair of the spinal cord.


Introduction
The central canal (CC) is a latent stem cell niche whose cells may have retained part of the developmental programs needed to re-pair the spinal cord after traumatic injury (Göritz and Frisén, 2012;Sabelström et al., 2014). After spinal cord injury (SCI), ependymal cells resume their mitotic activity (Mothe and Tator, 2005;Cizkova et al., 2009) to play an important part in limiting the spread of the lesion by the contribution of cells to the glial scar and producing neurotrophic factors that promote the survival of neurons around the injury site (Sabelström et al., 2013). The manipulation of the ependymal stem cell niche to optimize the cellular events needed for endogenous repair requires understanding the mechanisms by which proliferation, migration, and differentiation of CC-contacting progenitors are regulated in the normal and injured spinal cord. In particular, it is important to explore the early functional changes induced by injury in CCcontacting progenitors.
Both in reptiles and neonatal rats, ependymal cells have heterogeneous membrane properties (Russo et al., 2008;Marichal et al., 2012), with voltage-gated K ϩ currents that may regulate their behavior as described for some neural progenitors (Chittajallu et al., 2002;Schaarschmidt et al., 2009). In addition, some cells lining the CC are electrically and metabolically coupled via gap junctions (Russo et al., 2008;Marichal et al., 2012). Gap junction communication and connexin hemichannels are multifaceted regulators during the development and function of neural networks (Elias and Kriegstein, 2008). Indeed, connexins play an important role in the regulation of neural stem cells (Bruzzone and Dermietzel, 2006), with both connexin 43 (Cx43) and connexin 26 (Cx26) expressed differentially during development (Nadarajah et al., 1997;Bittman and LoTurco, 1999). The importance of connexins as regulators of proliferation is maintained in adult neurogenic niches as KO of Cx30 and Cx43 in progenitors of the dentate gyrus decrease their proliferation (Kunze et al., 2009). As early as 4 h after complete transection of the rat spinal cord, Cx43 mRNA increases and remains elevated at least 4 weeks after injury (Lee et al., 2005). The interference with Cx43 function after SCI reduces the reactive astrogliosis and neuronal death (O'Carroll et al., 2008) and improves functional recovery (Cronin et al., 2008). However, it is unknown whether connexins and gap junction coupling in the CC stem cell niche are regulated by injury. As postnatal development progresses, the ability of the spinal cord for self-repair declines and is paralleled by a reduction in the stem cell-like properties of ependymal cells, which become less competent to generate new oligodendrocytes (Li et al., 2016). The fact that progenitors contacting the lateral aspects of the neonatal CC have a higher proliferative rate (Marichal et al., 2012) suggests that coupling via connexins may be important to maintain an active CC stem cell niche. We hypothesized that membrane properties and/or communication via connexins in the CC are developmentally regulated and play a part in the reactivation of this latent stem cell niche after injury. To test these ideas, we combined patch-clamp recordings of ependymal cells in neonatal (P9 -P11) and adult (P90 -P150) mice with immunohistochemistry for various connexins in the normal and injured spinal cord. We found that coupling among ependymal cells was reduced in adulthood but increased after injury, resembling the immature CC. These changes were paralleled by an increase in the immunoreactivity of Cx26. In addition, we found that injury reduced the activity of connexin hemichannels in ependymal cells of the adult spinal cord. Finally, the pharmacological blockade of connexins reduced the injury-induced reactivation of ependymal cell proliferation. Collectively, our findings show that connexin signaling is involved in the early reaction of ependymal cells to injury and represents a potential target to improve the contribution of the CC stem cell niche to repair.

Materials and Methods
Animals. Neonatal (P9 -P11) and adult (P90 -P150) C57BL/6J mice were used. In some experiments, FoxJ1CreER-R26RtdTomato transgenic mice (gift from Prof. Jonas Frisén, Karolinska Institutet) were also used to facilitate the identification of ependymal cells. This transgenic mouse expresses CreER under the control of the FoxJ1 promoter, which is active in cells with motile cilia resulting in a selective and robust expression of tdTomato in ependymal cells (Meletis et al., 2008). To induce the expression of tdTomato in adult mice, we injected tamoxifen (Sigma Millipore; 2 mg, 20 mg/ml in corn oil, i.p.) for 5 d and allowed 5 d between the last injection and surgery to ensure clearance (Meletis et al., 2008). To induce recombination in neonatal animals, we applied 3 daily subcutaneous injections of tamoxifen (P4 -P6) at a concentration of 75 g/g of body weight (Cai et al., 2013). Pups were kept with their mother until use. All experimental procedures were approved by our local Committee for Animal Care (protocol #006-5-2017).
SCI. Animals were anesthetized with ketamine (100 mg/kg, i.p.), xylacine (10 mg/kg, i.p.), and diazepam (5 mg/kg, i.p.). Injury of the dorsal aspect of the spinal cord was performed as described by Frisén et al. (1993). Briefly, after laminectomy, the dorsal funiculus at low thoracic level (T13) was cut transversely with microsurgical scissors (depth ϳ0.8 mm), and the lesion was extended rostrally to comprise about one spinal cord segment. Recovery from anesthesia was promoted with flumazenil (0.5 mg/kg, i.p.), yohimbine (2 mg/kg, i.p.), and tramadol (3 mg/kg, i.p.) for pain relief. A second dose of tramadol was applied 24 h after surgery. Sham-injured animals were used as controls by performing all the procedures described above but without injuring the cord.
Morphological identification of patch-clamp recorded cells. During whole-cell patch-clamp recordings, cells were filled with a fluorophore and biocytin. Cells were first imaged in living slices by epifluorescence; and after completion of the recording period, slices were fixed by immersion in 4% PFA in 0.1 M PB, pH 7.4, overnight. Following PB rinsing, the slices were incubated in PB containing 0.3% Triton X-100 with streptavidin- Invitrogen,Thermo Fisher Scientific), mounted in glycerol, and imaged in a confocal microscope (Olympus, VF300; or Carl Zeiss, LSM800 Airyscan).
5-Ethynyl-2-deoxyuridine (EdU) proliferation assay. EdU (40 mg/kg, i.p.) was injected twice daily (4 h interval) from day 3 to day 5 after SCI. The animals were anesthetized as already described at day 5 after injury and perfused with 4% PFA in 0.1 M PB, pH 7.4. EdU was revealed with the Click-iT AlexaFluor-647 imaging kit (Invitrogen, Thermo Fisher Scientific).
Connexin hemichannel permeability assay. The activity of connexin hemichannels in ependymal cells was evaluated by the uptake of the hemichannel permeable dye ethidium bromide (EtBr; 4 M, 15 min). Briefly, spinal cord slices were cut as described above and incubated in Ringer's solution devoid of calcium 15 min before and during the application of EtBr with or without the application of the broad connexin blocker carbenoxolone (Cbx; 200 M), the cell-permeant Cx43 hemichannel blocking peptide TAT-Gap19 (YGRKKRRQRRR-KQIEIKKFK, 100 M), or the control peptide TAT-Gap19scrambled (YGRKKRRQRRR-IEKFKIKQK, 100 M). After incubation in these solutions, slices were rinsed in normal Ringer's solution for 15 min, then fixed for 24 h in 4% PFA in PB, pH 7.4, rinsed in PB (3 times, 10 min), and finally mounted in glycerol for imaging. Ependymal cells were imaged in an LSM800 Airyscan confocal microscope (Carl Zeiss). Stacks of consecutive confocal images with high-bit depth color (16 bit) were taken at 1 m intervals with the 561 nm laser for EtBr at 2.5% of laser intensity and detector settled at 743 V. Dye uptake analysis was performed in an ROI defined by the outline of the ependyma. As described by Chever et al. (2014), fluorescence intensity was digitized in arbitrary units of 65,000 shades of gray with ImageJ (National Institutes of Health) (Schneider et al., 2012). EtBr uptake (as fluorescence units) was expressed as the difference between the fluorescence measured in ependymal cells (ϳ75 cells per slice) and the background fluorescence from a field with no labeled cells.
In vivo blockade of connexins in the injured spinal cord. To locally block signaling via connexins around the lesion epicenter, 50 l of Pluronic F127 (20% w/v in saline at 5°C-10°C) containing the connexin blocking peptide Gap26 (10 mM, VCYDKSFPISHVR, GenScript) was injected within the lesion. The liquid Pluronic F127, kept at low temperatures, easily filled the lesion site and became a hydrogel as warmed up in contact with the tissue. Pluronic F127 dissolved in saline or containing the scrambled version of Gap26 (Gap26Scr, PSFDSRHCIVKYV, GenScript) was injected within the injured spinal cord as controls.
Experimental design and statistical analyses. Mice of either sex were randomly assigned to experimental groups. Sham-injured mice were used as controls. When using injured mice, we checked that loss of weight at death did not exceed 15%. The following criteria were used for including the data from patch-clamp experiments: a stable holding current, an RMP more hyperpolarized than Ϫ40 mV, and an access resistance Յ40 M⍀. At least a 10 min control period with stable electrophysiological properties was allowed before drug application. Electrophysiological data were analyzed with Clampfit 10.0 (Molecular Devices). All images obtained for quantitative analysis were taken using the same preset parameters and analyzed with ImageJ. For quantitative analysis of immunofluorescence data, a minimum of five sections were analyzed and averaged per biological replicate (5 animals for each experimental condition, unless otherwise stated). To analyze the expression of connexins in the apical and basal portions of ependymal cells, the immunoreactive puncta were converted into particles using a macro developed in IJ1macro language (ImageJ, National Institutes of Health) and quantified in two regions: an inner ring (0 -5 m from the CC lumen) and an outer ring (5-20 m from the CC lumen). The confocal images were acquired with an LSM800 Airyscan confocal microscope (Carl Zeiss). Statistical analyses were performed with Prism 7.0 (GraphPad). Statistical significance was set at p Ͻ 0.05. The statistical test chosen for different experiments is noted next to the corresponding result. Numerical values are reported as mean Ϯ SEM.

Results
Neural progenitors have membrane properties that regulate their behavior. For example, voltage-gated K ϩ channels (Chittajallu et al., 2002;Pardo, 2004;Schaarschmidt et al., 2009) and gap junction coupling via connexins (Bruzzone and Dermietzel, 2006) have been shown to affect the ability of neural stem cells to proliferate. Ependymal cells in the spinal cord of neonatal rodents are functionally heterogeneous with some displaying complex electrophysiological phenotypes (Marichal et al., 2012). Thus, we asked whether the membrane properties of ependymal cells vary as the CC matures to become quiescent. To address this question, we made patch-clamp recordings of ependymal cells in acute spinal cord slices of neonatal (P9 -P11) and adult (P90 -P150) mice. The membrane properties of ependymal cells in neonatal animals were dominated by prominent leak currents ( Fig. 1 A, B) with input resistances (IRs) of 116.8 Ϯ 11.3 M⍀ (n ϭ 48) and RMP of Ϫ79.0 Ϯ 1.3 mV (n ϭ 48). Leak subtraction allowed the discrimination of two cell types: Type 1 cells with a passive electrophysiological phenotype ( Fig. 1A; n ϭ 25) and Type 2 cells with sizable outward currents ( Fig. 1B; n ϭ 11). In a few recordings, a small inward current was observed (n ϭ 4, data not shown).
The electrophysiological phenotypes of ependymal cells in the lateral domains of the CC of adult mice were similar from those in neonatal animals ( Our previous study in neonatal rats showed that most cells in the lateral domains of the CC are coupled via gap junctions (Marichal et al., 2012). Thus, we assessed gap junction coupling among ependymal cells by injection of biocytin, which permeates through functional gap junctions revealing dye coupling. As expected, most recordings of ependymal cells (47 of 68) in the lateral domains of the neonatal CC revealed clusters of dye-coupled cells ( Fig. 2A-C). The morphology of clusters of dye-coupled cells was heterogeneous, with some covering a substantial portion of the CC ( Fig. 2A, right, insets, B 2 ,C 1 ,C 2 ) and some comprising most of the lateral aspect of the CC (Fig. 2B 3 ), whereas others were limited to a narrower extension (Fig. 2B 1 ). Both the counterstaining with DAPI ( Fig. 2A, insets, B 3 ) and recombination of the reporter gene tdTomato in ependymal cells of FoxJ1CreER mice (Fig. 2B 1 ,B 2 ; ϳ60% of the whole ependymal cell population) (Meletis et al., 2008) confirmed that the cell bodies of dye-coupled cells were confined to the ependymal cell layer. Some cells within the clusters had processes that gathered to form a bundle projecting to the dorsal ( Fig. 2A, arrowheads, C 1 , arrowheads) or the ventral (Fig. 2C 2 , arrowheads) raphe. As expected for ependymal cells, a careful inspection of the apical surface of dye-coupled cells revealed a variable number of cilia projecting to the CC lumen ( Fig. 2A, bottom left, inset; Movie 1). The gap junction blocker Cbx (100 M) increased the IR (Fig. 2D 1 ; p Ͻ 0.05; Wilcoxon matched pairs signed rank test, n ϭ 6), indicating that dye coupling resulted from gap junction coupling. Because Cbx has been reported to interact with targets other than connex-ins (e.g., GABA A receptors and calcium channels among others) (Juszczak and Swiergiel, 2009;Connors, 2012), we also tried the more specific peptidic antagonist Gap26 (Evans et al., 2012). In line with the results obtained with Cbx, Gap26 also increased the IR of cells that were dye-coupled with neighboring cells (Fig. 2D 2 ; p Ͻ 0.05, Wilcoxon matched pairs signed rank test, n ϭ 7). In a minority of cases, patch-clamp recordings in the CC of neonatal animals resulted in single labeled cells (Fig. 2E). However, the situation in the adult CC was different. Figure 2F shows the morphology of a patch-clamped single ependymal cell with two conspicuous cilia protruding into the CC lumen. Although clusters of dye-coupled cells in the CC of the normal adult mouse could be observed (Fig. 2G,H ), the relative frequency of clusters of gap junction-coupled cells was significantly lower than in the immature spinal cord (15 of 56; Fig. 2I; p Ͻ 0.0001, Fisher's exact test). However, the mean volume of clusters of gap junction-coupled cells was not significantly different between neonatal and adult animals ( Fig. 2J; p ϭ 0.40, Mann-Whitney U test). Both in neonatal and adult mice, clusters of gap junction-coupled cells often surrounded other ependymal cells that were not dye-coupled (Fig. 2B 2 ,G, inset, arrowheads; Movie 2). Regardless the age, the electrophysiological phenotypes of single cells were not statistically different from clustered cells ( p ϭ 0.91, Fisher's exact test).
To explore the molecular basis of gap junction coupling in the CC and whether there is some developmental plasticity, we made show that the response to voltage steps was completely passive (Type 1 cell). B, A subpopulation (Type 2) of ependymal cells in the neonatal spinal cord displayed voltage-gated outward currents (B 1 ). The I-V plot in Type 2 cells is nonlinear (B 2 ). The current plots at the beginning (red dots) and the end (black dots) suggest that voltage-gated outward currents have a component with time-and voltage-dependent inactivation. C, D, The electrophysiological phenotypes of ependymal cells in the adult spinal cord (P Ͼ 90) were similar to their neonatal counterpart with Type 1 passive cells (C) and Type 2 cells (D). E, Current responses to voltage steps applied from a holding potential of Ϫ90 mV (E 1 ) and Ϫ30 mV (E 2 ) in an ependymal cell of the adult spinal cord. Notice the slow kinetics and lack or inactivation of the outward currents from a holding potential of Ϫ30 mV typical of a delayed rectifier (I KD ). Subtraction of currents at Ϫ90 and Ϫ30 mV reveals a fast inactivating component characteristic of A-type (I A ) currents. F, No statistically significant differences were found for the IRs (F 1 , p ϭ 0.29) and RMPs (F 2 , p ϭ 0.3) between neonate and adult ependymal cells. G, Ratio of Type 1 and Type 2 cells in neonate and adult ependyma. ns, non significant.

Figure 2.
Gap junction coupling between ependymal cells is developmentally downregulated. A, Cluster of dye-coupled cells contacting the CC in the neonatal spinal cord (stack of 58 confocal optical sections). Some cells in the cluster had processes that project to the dorsal raphe, reaching the pia matter (arrowheads). Several cilia (bottom left, inset, arrows) are seen on the apical pole of clustered cells (stack of sections 12-21). Right, Insets, Two confocal optical sections with DAPI counterstaining to illustrate that dye-coupled cells are within the ependymal cell layer. B, Clusters of dye-coupled cells cover variable portions of the CC. B 1 , A small cluster of dye-coupled cells comprising a discrete extension of the dorsolateral aspect of the CC. B 2 , A cluster covering the dorsolateral half of the CC (stack of 64 optical sections). Some dye-coupled cells covered the whole lateral aspect of the CC (B 3 , stack of 86 optical sections). C, Clustered cells often showed processes (arrowheads) that projected toward the dorsal (C 1 ) or ventral (C 2 ) raphe (stacks of 43 and 60 optical sections, respectively). Numerous varicosities (C 1 ,C 2 , asterisks) could be observed at some distance of the CC. D 1 , Cbx (100 M) and Gap26 (D 2 , 200 M) increased the IR in cells that were dye-coupled to their neighbors. Top, Representative currents induced by a voltage (Figure legend continues.) immunohistochemistry for various connexins in neonatal and adult mice. Similar to the developing cortex (Dermietzel et al., 1989;Nadarajah et al., 1997) and the CC in turtles (Russo et al., 2008), we found that Cx43 and Cx26 were expressed in ependymal cells of the CC of neonatal mice (Fig. 3 A, B). Because Cx30 is expressed in astrocytes (Rash et al., 2001) and ependymal cells in the brain (Kunzelmann et al., 1999), we asked whether ependymal cells express Cx30 in addition to Cx43 and Cx26. In neonatal mice, we found a very weak immunostaining for Cx30 with few scattered puncta in the gray and white matter, whereas ependymal cells lining the CC were devoid of Cx30 (Fig. 3C). Cx43 was also strongly expressed in the CC of the adult spinal cord with a similar spatial profile than in neonatal animals (Fig. 3D). However, unlike in the neonatal CC, the immunoreactivity for Cx26 was very low in ependymal cells of the adult spinal cord (Fig. 3E). Whereas Cx30 in the neural parenchyma increased substantially in the spinal cord of adult mice, the ependymal cell layer remained devoid of Cx30 (Fig. 3F ). Despite that Cx50 has been reported in ependymal cells isolated from the spinal cord of rats (Rodríguez-Jimenez et al., 2016), we found no immunoreactivity within the ependymal cell layer of neonatal or adult mice (data not shown).
In turtles, Cx43 is spatially segregated to the apical compartment of ependymal cells (Russo et al., 2008). To explore the relative distribution of Cx43 and Cx26 in the apical and basal compartments of ependymal cells, we digitally converted connexin puncta into particles and quantified their density in two concentric rings (0 -5 m and 5-20 m from the CC lumen; Fig.  3G). Although Cx43 puncta were expressed at different distances from the CC lumen, a statistically significant higher density of puncta was observed in the apical poles of ependymal cells close to CC lumen in neonatal but not in adult mice ( Fig. 3H; p Ͻ 0.05 and p ϭ 0.079, respectively, two-way ANOVA with Holm-Sidak post hoc, n ϭ 5 animals per group). The overall density (apical ϩ basal compartments) of Cx43 puncta in neonatal and adult mice did not change significantly ( Fig. 3H; p ϭ 0.072, two-way ANOVA). In contrast to Cx43, the densities of Cx26 in the apical and basal compartments of ependymal cells of the neonatal and adult spinal cord were not significantly different ( Fig. 3I; p ϭ 0.16 and p ϭ 0.93, respectively, two-way ANOVA with Holm-Sidak post hoc, n ϭ 5 animals per group). As suggested by the representative images shown in Figure 3B, E, the density of Cx26 puncta in the adult ependyma was greatly reduced compared with their neonatal counterpart (neonatal: 7.2 Ϯ 1.4 puncta/100 m 2 ; adult: 1.00 Ϯ 0.5 puncta/100 m 2 , Fig. 3I; p Ͻ 0.0001, two-way ANOVA).
Gap junction coupling has been shown to be important for the maintenance of the phenotype of active progenitor cells (Cheng et al., 2004;Bruzzone and Dermietzel, 2006). The low rate of gap junction coupling among ependymal cells may be related to the very low proliferative activity of this stem cell niche in adulthood (Alexovic Matiasová et al., 2017). Thus, we speculated that reactivation of "dormant" progenitor-like cells in the CC by injury  0.40). A, B 3 , C, G, Data obtained from C57BL/6J mice. In this and subsequent figures, dorsal is facing upward. *p Ͻ 0.05, ****p Ͻ 0.0001, ns ϭ non significant. formed a dorsal hemisection of the spinal cord and made patch-clamp recordings of ependymal cells 5 days post injury (DPI) in slices cut from the rostral and caudal stumps (1.2 mm at both sides of the lesion epicenter; Fig. 4A). We found no statistically significant differences between the RMP of ependymal cells from injured and sham-injured mice (Fig. 4B 1 ; RMP-injured: Ϫ78.6 Ϯ 3.48 mV, n ϭ 15; RMP sham: Ϫ76.0 Ϯ 2.77, n ϭ 11; p ϭ 0.54, Mann-Whitney U test). However, the IR was lower in cells recorded in the injured spinal cord (Fig. 4B 2 ; IR-injured: 57.35 Ϯ 9.4 M⍀, n ϭ 17; IR sham: 155.4 Ϯ 40.4 M⍀, n ϭ 11; p Ͻ 0.05, Mann-Whitney U test). We found no differences on the relative frequency of cells with passive membrane properties (Type 1) and those with voltage-gated outward currents (Type 2) (Fig. 4B 3 ; p ϭ 0.66, Mann-Whitney U test). Whereas intracellular biocytin staining still revealed some single ependymal cells (Fig. 4C), the frequency of clusters of dye-coupled cells over single cells in injured animals compared with sham controls was significantly higher (Fig. 4D-F; p Ͻ 0.01, Fisher's exact test, 38 of 56 and 15 of 37, respectively). The relative frequency of clusters of gap junctioncoupled cells in adult injured animals was comparable with that of normal neonatal animals (Figs. 2I, 4F; p ϭ 0.83. Fisher's exact test), and there was no statistically significant difference between the incidence of clusters between normal and sham-injured adult mice ( p ϭ 0.35, Fisher's exact test). We then asked whether the increase in the relative frequency of gap junction coupling among ependymal cells in the lateral domains of the CC may be associated with a larger size of clusters of gap junction-coupled cells. To answer this question, we made 3D reconstructions of the clusters of gap junction-coupled cells and calculated the total volume (in m 3 ) using ImageJ (Fig. 4G). We found no significant differences between the average size of clusters between injured and shaminjured adult animals ( Fig. 4G; p ϭ 0.48, Mann-Whitney U test).
Next, we explored the molecular basis for the upregulation of gap junction coupling induced by SCI. The overall density and location of Cx43 puncta within the ependymal cell layer were not different in the spinal cord stumps (0 -2000 m from the lesion epicenter) of the injured compared with sham control animals ( Fig. 5A-C; p ϭ 0.26, two-way ANOVA). In contrast to Cx43, the expression of Cx26, which was low in the intact adult CC (Fig. 3E) and in sham-injured controls (Fig. 5D), increased significantly within the ependymal cell layer of injured animals (Fig. 5 E, F; p Ͻ 0.01, Kruskal-Wallis test with Dunn's post hoc), resembling the immunostaining for Cx26 in neonatal mice (Fig. 3B). As in the normal spinal cord, there was no immunoreactivity for Cx30 and Cx50 in the ependymal cell layer of the rostral and caudal stumps of the injured cord at 5 DPI (data not shown).
A hallmark of the reactive CC is that ependymal cells resume mitotic activity (Mothe and Tator, 2005). Thus, we compared the spatial profile of the proliferative reaction of ependymal cells in the CC with that of the increased expression of Cx26 at 5 DPI. The thymidine analog EdU was injected twice daily (40 mg/kg, i.p.) from day 3 to day 5 after injury (Fig. 5G 1 ,G 2 ). EdU ϩ nuclei in the CC peaked around the lesion epicenter (Fig. 5H 2 ,I ) to reach control levels at ϳ2000 m away from the lesion site (Fig. 5H 1 -H 3 ,I ). The spatial profile of the increased expression of Cx26 was slightly wider than for EdU uptake (Fig. 5I ). There was a signifi- cant difference in the density of Cx26 puncta within the epicenter (0 -2 mm around the epicenter) compared with rostral ( p Ͻ 0.001, Kruskal-Wallis test) and caudal ( p Ͻ 0.0001) regions (2-4 mm from the epicenter), which had a density similar to shaminjured animals ( Fig. 5J; p ϭ 0.99 and p ϭ 0.1, respectively, Kruskal-Wallis test, with Dunn's post hoc). Similarly, the difference in EdU uptake between the epicenter and the rostral and caudal regions was significantly different (Fig. 5K; p Ͻ 0.0001 and p Ͻ 0.001, respectively, Kruskal-Wallis test, with Dunn's post hoc). This shows that the increase of Cx26 occurs locally around the injury site and is spatially correlated with the increased proliferation of ependymal cells.
In addition to forming gap junctions, unopposed connexin hemichannels play an important role in the nervous system, both in normal and pathological conditions (Bennett et al., 2003(Bennett et al., , 2012. Thus, we explored the presence of functional connexin hemichannels in ependymal cells of the adult intact spinal cord and the eventual plasticity during the acute phase of SCI. We found that incubation of adult spinal cord slices with EtBr resulted in robust EtBr uptake in many ependymal cells (Fig. 6A 1 ), which was strongly reduced when slices were preincubated in 100 M Cbx (Fig. 6A 2 ,D 1 ; p Ͻ 0.0001, Mann-Whitney U test). Because Cbx is a nonselective connexin antagonist, we used the TAT-Gap19 peptide, which binds to the intracellular loop of Cx43 blocking hemichannel activity without altering Cx43 gap junctions (or other connexins) in the time span of our experiments (Abudara et al., 2014). In the presence of the inactive TAT-Gap19 scrambled peptide (TAT-Gap19Scr), the uptake of EtBr was similar to that observed in control conditions (Fig. 6B 1 ; p ϭ  0.40, Mann-Whitney U test). However, uptake of EtBr in the presence of 100 M TAT-Gap19 was significantly reduced (Fig.  6B 2 ,D 1 ; p Ͻ 0.0001 Mann-Whitney U test). Next, we asked , and ratio of Type 1 and Type 2 cells (B 3 ) in ependymal cells from sham-injured (sham) and injured (5 DPI) adult spinal cord. The IR between the two groups was significantly different ( p Ͻ 0.05), whereas both the RMP and the proportion of Type 1 and Type 2 cells were not. C, Single ependymal cell recorded at 5 DPI (stack of 7 confocal optical sections). Recombination of tdTomato confirms the ependymal identity of the recorded cell. D, Cluster of dye-coupled cells at 5 DPI covering the dorsolateral (D 1 ), ventrolateral (D 2 ), and ventral (D 3 ) aspects of the CC (stacks of 47, 21, and 25 optical sections, respectively). D 3 , The cluster has a bundle of processes projecting to the ventral sulcus (arrowhead). D 3 , Inset, The expression of tdTomato in ependymal cells shows the cluster was within the ependymal cell layer. E, Large cluster of dye-coupled cells spanning the lateral domain of the CC at 5 DPI (E 1 , stack of 22 optical sections). Most of dye-coupled cells matched the expression of tdTomato. The cluster corresponded to a Type 1 cell phenotype (E 2 ). Cbx (100 M) increased the apparent IR of the recorded cell, suggesting the gap junction coupling with neighboring cells (E 3 ). F, The proportion of clusters of dye-coupled ependymal cells over single cells was significantly higher ( p Ͻ 0.01) at 5 DPI than in sham-injured animals. G, The volume of clusters of gap junction-coupled cells was similar in the CC of sham-injured and injured spinal cord ( p ϭ 0.48). *p Ͻ 0.05, **p Ͻ 0.01, ns ϭ non significant.
whether hemichannel activity was also present in the reactive ependyma. Interestingly, at 5 DPI, we found no significant difference in EtBr uptake between the control and in medium containing Cbx (Fig. 6C 1 ,D 1 ; p ϭ 0.68, Mann-Whitney U test). Additionally, EtBr uptake at 5 DPI was significantly lower compared with the uninjured ependyma ( p Ͻ 0.0001) or TAT-Gap19Scr ( p Ͻ 0.01) controls (Fig. 6D 2 ; Kruskal-Wallis test with Dunn's post hoc). This suggests that functional Cx43 hemichannels are downregulated during the early stages of the reaction of the ependyma to injury.
Our findings suggest that connexin signaling may be involved in the injuryinduced reactivation of the dormant CC stem cell niche. To test this idea, we blocked connexin signaling in vivo immediately after SCI. We locally applied Gap26 and Gap26Scr in Pluronic F127 hydrogel or Pluronic F127 with saline (vehicle) at the time of injury and injected EdU (two daily injections, from day 3 to day 5 after SCI; Fig. 7A; n ϭ 4 animals per group). The proliferation response of ependymal cells to injury was strong when vehicle was applied and not significantly different from the proliferation rate observed when using Gap26Scr (Fig.  7B 1 ,B 2 ,C; p ϭ 0.9, Kruskal-Wallis test with Dunn's post hoc). However, the blockade of connexins by Gap26 significantly reduced the proliferative reaction of ependymal cells to injury (Fig. 7B 3 ,C; p Ͻ 0.05, Kruskal-Wallis test with Dunn's post hoc), suggesting that connexin signaling is involved in the "awakening" of ependymal cells. The average proliferation when using Pluronic F127containing vehicle or Gap26Scr was higher than that of the injured cord without the gel (Fig. 5 I, K vs Fig. 7C). To test whether Pluronic F127 may affect proliferation, we applied the gel to the surface of sham-injured cords (Fig. 8A). We found that Pluronic F127 did not change significantly the number of EdU ϩ nuclei in the ependyma or the dorsal surface of the cord (Fig. 8 B, C; p ϭ 0.27 and p ϭ 0.50, respectively, Nested t test; n ϭ 3 animals per group). Furthermore, the gel seemed quite inert as did not generate a substantial inflammation judged by the number of Iba1 ϩ cells ( Fig. 8D; p ϭ 0.59, Nested t test; n ϭ 3 animals per group). It may be possible that Pluronic F127 gel within the injury site exerts some mechanical strain that could increase background proliferation as described in some stem cells (Vining and Mooney, 2017).

Discussion
Communication via gap junctions is a major regulator of the behavior of neural stem cells (Elias and Kriegstein, 2008). Here, we report that gap junction coupling within the CC of mice decreases from early postnatal life into adulthood, when the CC stem cell niche becomes quiescent (Sabourin et al., 2009;Alexovic Matiasová et al., 2017). In the early stages of the reaction to in-jury, gap junction coupling among adult ependymal cells increased, resembling their immature counterparts. Similar to stem cell niches in the embryo, Cx43 and Cx26 were expressed in the CC with a developmental downregulation of Cx26 that was reverted by injury. Our findings suggest that connexin signaling is involved in the early reaction of ependymal cells to injury, representing a potential target to modulate the contribution of the CC to self-repair.

Developmental plasticity of gap junction coupling in the CC
Our data showed that the membrane properties of ependymal cells remained similar during postnatal development and after injury of the adult spinal cord. However, gap junction coupling between ependymal cells in the lateral domains of the CC was a common feature in neonatal mice but less frequent in adults. Proliferating radial glia in the developing cortex are dynamically coupled via gap junctions (LoTurco and Kriegstein, 1991) and pharmacological uncoupling prevents these progenitors entering the S-phase of the cell cycle (Bittman et al., 1997). Thus, the reduction of coupling between ependymal cells as the spinal cord matures is in line with the fact that proliferation in the CC peaks D, E, Expression profile of Cx26 in sham-injured and 5 DPI mice, respectively. F, The density of Cx26 puncta increased dramatically at 5 DPI compared with control animals ( p Ͻ 0.01). G 1 , Experimental design to study the correlation between proliferation and Cx26 expression. G 2 , An EdU ϩ nucleus from a cell expressing Cx26. H, Representative images of EdU uptake and Cx26 immunohistochemistry at rostral (H 1 ), epicenter (H 2 ), and caudal (H 3 ) levels. I, Spatial profile of Cx26 density (green) and EdU ϩ nuclei (magenta) at 5 DPI. The lesion epicenter is located at 0 in the abscissa. J, Bar graph of the density of Cx26 puncta in a region around the lesion epicenter (0 -2 mm around the epicenter) and in neighboring rostral and caudal regions (2-4 mm away from the epicenter). The density of Cx26 puncta in the sham-injured spinal cord is also shown (gray bar). K, Number of EdU ϩ nuclei in the same regions as in the plot in J. All images are single confocal optical planes. G-K, Data obtained in C57BL/6J mice. **p Ͻ 0.01, ***p Ͻ 0.001, ****p Ͻ 0.0001, ns ϭ non significant.
Gap junction coupling in the CC of neonatal mice was correlated with the expression of Cx26 and Cx43, resembling the expression profile of connexins in neural progenitors of the developing brain (Dermietzel et al., 1989;Nadarajah et al., 1997). The fact that Cx26 puncta in the neonatal CC were evenly distributed compared with Cx43 suggests that connexin signaling in ependymal cells shares key features with bona fide stem cell niches of the developing mammalian brain (Nadarajah et al., 1997;Elias and Kriegstein, 2008). Progenitor-like cells in the CC of turtles (a vertebrate with self-repair capabilities) are coupled by Cx43 located close to the CC lumen, whereas Cx26 is expressed more distally (Russo et al., 2008), suggesting that the spatial profile of Cx43 and Cx26 expression is a phylogenetically conserved feature of active stem cell niches.
The low expression of Cx26 in the adult ependyma is in line with early studies showing that Cx26 in the brain disappears 3-6 weeks after birth (Dermietzel et al., 1989). The expression of Cx26 in BrdU ϩ cells of the developing cortex is significantly higher than Cx43 (Bittman and LoTurco, 1999), suggesting that the low Cx26 in adult ependymal cells may be linked to a nonproliferative phenotype. In line with this, the spatial profile of Cx26 upregulation matched the increase in EdU uptake at 5 DPI.
Cx30 appears late in astrocytes during postnatal development and was also reported in ependymal cells of the brain (Kunzelmann et al., 1999). Although we found a similar temporal profile of Cx30 expression in the spinal cord parenchyma, ependymal cells lacked Cx30. This difference may be related to the fact ependymal cells in the spinal cord are different from those in the brain (Mirzadeh et al., 2008;Alfaro-Cervello et al., 2012).

Injury-induced plasticity of connexin signaling in the CC
Resumption of proliferation is an early event of the reaction of the CC to SCI (Mothe and Tator, 2005;Meletis et al., 2008;Sabelström et al., 2013). The increased incidence of coupling among ependymal cells at 5 DPI suggests that gap junction coupling is a component of the cellular mechanisms implied in the There is a significant difference in connexin hemichannel activity between control and 5 DPI ( p Ͻ 0.0001 control vs 5 DPI; p Ͻ 0.01 between TAT-Gap19 and 5 DPI; p ϭ 0.99 between TAT-Gap19 and control). All images represent single optical sections. All data were obtained in C57BL/6J mice. **p Ͻ 0.01, ****p Ͻ 0.0001, ns ϭ non significant. Figure 7. The proliferation of ependymal cells in response to SCI is reduced by connexin blockade. A, Experimental design to evaluate the role of connexin signaling in the reaction of ependymal cells to SCI. B, Numerous nuclei within the ependyma are labeled with EdU at 5 DPI when vehicle (B 1 ) or the scrambled version of Gap26 (B 2 , Gap26Scr 10 mM) was applied. However, the number of EdU ϩ nuclei was lower when Pluronic F127 hydrogel containing the connexin blocker Gap26 (B 3 , 10 mM) filled the lesion site. C, Bar graph represents that the number of EdU ϩ nuclei at 5 DPI was not significantly different between Pluronic F127containing saline (vehicle) or scrambled Gap26 (Gap26S, 10 mM) but decreased significantly when the hydrogel contained 10 mM Gap26 ( p Ͻ 0.05). The images represent stacks of 51 optical sections. All data were obtained in adult C57BL/6J mice. *p Ͻ 0.05, ns ϭ non significant. early reaction to injury and resembles a return to a state of connexin signaling similar to the active CC stem cell niche of neonatal mice (Sabourin et al., 2009), which has higher intrinsic regenerative potential (Li et al., 2016). Injury overruled the developmental downregulation of Cx26 in ependymal cells, whereas the expression of Cx43 did not change, in contrast with the upregulation of Cx43 reported in progenitors of the dentate gyrus after traumatic brain injury (Greer et al., 2017). Collectively, these results suggest that the injury-induced upregulation of Cx26 may be the molecular mechanism underlying the recoupling of ependymal cells in the reactive CC. However, we cannot rule out that recoupling is mediated by a change in Cx43 gating or assembly into gap junctions (Pogoda et al., 2016). Indeed, Cx43 has been proposed to underlie coupling of progenitor-like cells in the CC of turtles because its spatial profile of expression matches the location of clusters of dyecoupled cells (Russo et al., 2008).
Our EtBr uptake experiments indicate that, in the intact CC of adult mice, there is a population of functional Cx43 hemichannels that decrease after injury. A parsimonious interpretation of our data is that unopposed hemichannels may be recruited for gap junction assembly in ependymal cells reactive to injury. Alternatively, a negative interaction with upregulated Cx26 after injury may reduce the population of permeable Cx43 hemichannels (Gemel et al., 2004). In glial cells, Cx43 seems neuroprotective when assembled into gap junctions (Kozoriz and Naus, 2013) or neurotoxic in the form of hemichannels (Orellana et al., 2009). We speculate that the reduction of hemichannel activity may enhance the viability of CC progenitor-like cells within injured tissue.

Connexins and SCI
The role played by connexins in the response of the nervous system to injury is controversial, with reports suggesting that increased expression of Cx43 is detrimental (Frantseva et al., 2002a,b), whereas others show that Cx43 downregulation results in an expansion of the lesion size (Siushansian et al., 2001;Nakase et al., 2004). In the rat spinal cord, Cx43 mRNA increases as early as 4 h after complete transection and remains elevated at least 4 weeks after injury (Lee et al., 2005). The interference with Cx43 function after SCI reduces the reactive astrogliosis and neuronal death (O'Carroll et al., 2008) and improves functional recovery (Cronin et al., 2008).
The fact that the size of clusters of gap junction-coupled cells was similar in neonatal, intact, and injured adult CC suggests that ependymal cells form small-world networks to regulate proliferation as in embryonic progenitors (Malmersjö et al., 2013). In line with this idea, our in vivo experiments show that connexin blockade around the lesion largely prevented the injury-induced resumption of ependymal cell proliferation. The molecular mechanisms by which connexins may regulate the behavior of progenitor cells are not clear. Strong metabolic coupling of spinal precursors at their apical poles, a critical cell compartment in neural progenitors (Taverna et al., 2014), may allow an efficient communication of molecules that regulate cell division or differentiation (Bruzzone and Dermietzel, 2006). Gap junctions may provide genetic communication among neighboring progenitors as they allow the passage of microRNA, with Cx43 being more efficient than Cx26 (Zong et al., 2016). Cx43 and Cx26 may also affect the proliferation, migration, and differentiation of spinal precursors by nonconventional mechanisms via molecular interactions with a variety of proteins (Kardami et al., 2007). Interestingly, the role of connexins in neuronal migration during development in the cortical plate is mediated by the facilitation of cell-cell adhesion and not by their functions as channels (Elias et Figure 8. Pluronic F127 gel applied on the dorsal aspect of the spinal cord does not affect proliferation. A, Experimental design to test whether Pluronic F127 affects proliferation. Sham-injured animals served as controls (A 1 ); whereas in the test group, Pluronic F127 was applied on the dorsal surface of the spinal cord (A 2 ). EdU was injected to evaluate proliferation, and immunohistochemistry for Iba1 was performed to evaluate inflammation (A 3 ). B, Representative images of EdU uptake in the CC of control (B 1 ) and Pluronic F127 (B 2 ) mice. No significant differences were found between the two groups (B 3 ; p ϭ 0.27). C, EdU uptake in the tissue underlying the dorsal surface of the spinal cord (A 1 ,A 2 , dotted boxes) in control (C 1 ) and Pluronic F127 (C 2 ). The quantification of the number of EdU ϩ nuclei was not significantly different (C 3 , p ϭ 0.50). D, Representative images of Iba1 immunohistochemistry in the dorsal surface of the cord (A 1 , A 2 , boxed areas) in control (D 1 ) and Pluronic F127 mice (D 2 ). The number of Iba1 ϩ cells in both groups showed no significant differences (D 3 ; p ϭ 0.59). All data were obtained in adult C57BL/6J mice. ns ϭ non significant.
al., 2007). Although Cx43 and Cx26 are both expressed in the CC, developmental and injury-induced plasticity seemed related mostly to Cx26, a connexin associated to mitotically active progenitor cells (Bittman and LoTurco, 1999). Interestingly, the duo Cx43/Cx26 is expressed in the skin and has a major role in wound healing (Wong et al., 2016). Cx26 is expressed in proliferating keratinocytes and downregulated as differentiation proceeds to be weakly detected in the normal epidermis (Kretz et al., 2004). Persistent expression of Cx26 maintains the epidermis in a hyperproliferative state (Djalilian et al., 2006), suggesting that Cx26 is a key regulator of proliferation in epithelia. Remarkably, Cx26 is upregulated in keratinocytes within the edge of a wound (Pedersen et al., 2003). The fact that Cx26 in the adult CC is upregulated around the lesion epicenter resembles the events in the wounded skin and suggests that Cx26 may be part of a general mechanism to regulate cell proliferation during healing. In addition, Cx26 increases cell migration in a scrape wound model by lowering cell adhesion (Polusani et al., 2016) and interacts with actin to drive migration in the developing cortex (Elias et al., 2007), raising the possibility that Cx26 may be also involved in the migration of the ependyma-derived cells toward the scar. Cx43 could also affect ependymal cell proliferation as Cx43 is necessary for the mitogenic effect of FGF in neural progenitors (Cheng et al., 2004). Future studies should address the specific functional role of Cx26 and Cx43 in the biology of ependymal cells by selective genetic deletion using loxP transgenic mice. A better understanding of the role played by gap junction coupling and connexin hemichannels in the normal and injury-activated CC will provide valuable clues to improve the contribution of this stem cell niche to self-repair.