Wnt produced by stretched roof-plate cells is required for the promotion of cell proliferation around the central canal of the spinal cord

During embryogenesis and homeostasis, secreted signaling proteins such as Wnts and BMPs regulate cell proliferation and differentiation in a spatially coordinated manner. How physiological changes regulate the activity ranges of these signaling proteins remains to be determined, however. In general, secreted signaling proteins diffuse into the extracellular space, creating a concentration gradient around the producing cells. During morphogenesis, cells dynamically change their shape. Marked changes in the morphology of signal-producing cells during tissue morphogenesis presumably result in the activation of signaling in neighboring cells. However, it is unclear whether changes in cell morphology during tissue morphogenesis actually create new signaling targets.

The developing spinal cord has been thoroughly characterized as a model system for research that is aimed at understanding the spatial regulation of secreted signaling proteins. During early development of the vertebrate spinal cord, the most dorsal region, known as the roof plate, functions as an organizing center by secreting Wnt and BMP (Chizhikov and Millen, 2004; Lee and Jessell, 1999). Several researchers have hypothesized that concentration gradients of these signaling proteins generate patterns of interneuron subtypes in the dorsal spinal cord (Chizhikov and Millen, 2005; Lee et al., 2000; Muroyama et al., 2002). Interestingly, as this patterning event nears completion, the morphology of the spinal cord changes dynamically; that is, the neural tube lumen gradually shrinks and a median septum forms along the dorsoventral axis (Böhme, 1988). A recent study in zebrafish embryos revealed that the roof plate stretches during this morphological change (Kondrychyn et al., 2013). This finding prompted us to ask whether the processes of these stretched roof-plate cells are involved in intercellular communication and whether stretching of roof-plate cells induces changes in those cells that respond to signals from the roof plate.

In adult mice, neural progenitor/stem cells localize around the shrunken lumen of the spinal cord (Hamilton et al., 2009; Johansson et al., 1999; Meletis et al., 2008; Weiss et al., 1996), and this area is referred to as the central canal. The central canal is connected to the ventricle, in which neural progenitor/stem cells localize in the forebrain (Lois and Alvarez-buylla, 1993; Morshead et al., 1994; Weiss et al., 1996). As the proliferation of neural progenitor/stem cells in the brain is regulated by secreted signaling molecules, including Wnt (Adachi et al., 2007; Chenn and Walsh, 2002; Hirabayashi et al., 2004; Kuwabara et al., 2009; Lie et al., 2005; Machon et al., 2007; Munji et al., 2011; Toledo et al., 2008; Varela-Nallar and Inestrosa, 2013; Wrobel et al., 2007), it is plausible that spinal cord progenitor/stem cells are regulated in a similar manner. For better understanding of this regulation, the sources of these signals should be clarified. One intriguing possibility is that stretched roof-plate cells are a source of signaling molecules that target the neural progenitor/stem cells that surround the central canal.

Wnt1 and Wnt3a are specifically expressed in the roof plate (Parr et al., 1993) and they play numerous roles in the developing spinal cord, including regulating the proliferation and specification of dorsal interneurons (Muroyama et al., 2002) before the mid-gestation stage. However, owing to poor resolution of current digoxigenin-based in situ hybridization methods, whether Wnt expression is maintained in stretched roof-plate cells remains unclear. Furthermore, genetic studies of Wnt1- and Wnt3a-deficient mutant mouse embryos yielded no information regarding the roles of the respective proteins in later stages, as neither Wnt1- nor Wnt3a-deficient single mutants exhibit a roof plate-related phenotype, and most of the double mutants die before embryonic day (E)12.5 (Ikeya et al., 1997).

In this study, we found that expression of Wnt1 and Wnt3a is maintained in stretched roof-plate cells. Analyses of mouse embryos with a specific defect in Wnt secretion in roof-plate cells provided evidence that supported our hypothesis that Wnt secretion by roof-plate cells is required for the proper activation of Wnt signaling in, as well as the proper proliferation of, the neural stem/progenitor cells surrounding the central canal.

To examine changes in the morphology of roof-plate cells during development of the spinal cord in mice, we first monitored expression of the ezrin gene, which encodes a FERM domain-containing actin-binding protein that localizes on the apical side of epithelial cells and is expressed in roof-plate cells on E10.5 (Fig. 1A; Saotome et al., 2004). With shrinking of the spinal cord lumen after E13.5, we found that ezrin-positive roof-plate cells stretched along their dorsoventral axis (Fig. 1B-D′), consistent with observations in other species (Böhme, 1988; Kondrychyn et al., 2013). Of note, the stretched cells re-oriented so that the nuclei of these cells were aligned along the midline (Fig. 1B′).

Ezrin expression was also observed along the apical surface of the lumen from the mid-gestation stage (in the area indicated by yellow dashed brackets in Fig. 1B-D) and this complex expression pattern made it difficult to precisely follow changes in the morphology of the stretched roof-plate cells. To overcome this problem, we minimally labeled stretched roof-plate cells using a Wnt1-CreERT allele in conjunction with the R26R-Confetti reporter (Fig. 1E-K; Snippert et al., 2010; Zervas et al., 2004). Pregnant mice carrying E11.5 or E13.5 embryos were treated with a low dose of 4-hydroxytamoxifen (4OH-TM) and embryos were recovered on E13.5 or E15.5, respectively (Fig. 1E). These labeling experiments demonstrated that the nucleus of each roof-plate cell (marked by open arrowheads in Fig. 1F-H) moved in the ventral direction beginning on E13.5 and that the cell process stretched along the dorsoventral axis so that it continued to face the lumen (closed arrowheads in Fig. 1D′,F-K). Most of the processes that were labeled with membrane-bound CFP maintained contact with the lumen on both E13.5 (83.3%, n=5/6) and E15.5 (100%, n=8/8) (Fig. 1F,I), which suggests that the processes had stretched to keep their tips facing the shrinking lumen. Electron microscopic (EM) observation showed that the tips of the stretched cells were in contact with the lumen on both E13.5 and E18.5 (Fig. 2A,C). EM observation also showed that the narrow tip of the process of each roof-plate cell was aligned side by side with that of its neighboring cells (Fig. 2A,C). Furthermore, the processes of the roof-plate cells were enriched in intermediate filaments, which could be clearly identified based on their structure (Fig. 2E,G). In the area close to the lumen, a basal body structure was specifically observed, which suggested the formation of cilia (Fig. S1J). In addition, numerous vesicles were seen at the tip of the processes, suggesting active secretion of proteins via secretory vesicles (Fig. S1J). In contrast, ezrin-immunostaining of roof-plate cells also showed that the dorsal tip of the cell process extended to the spinal cord pia after E15.5 (braces in Fig. 1C,D). These results indicate that each roof-plate cell stretched along its dorsoventral axis to maintain physical contact with both the pia and shrinking lumen.

Previous research has shown that the cell processes that are stretched along the midline in the dorsal spinal cord express nestin (Ševc et al., 2009). To examine the extent of the contribution of progeny roof-plate cells to the formation of midline processes, we followed roof-plate progeny among nestin-positive midline cells. In this experiment, roof-plate progeny cells were labeled using a Wnt1-CreERT allele in conjunction with the R26R-tdTomato reporter (Madisen et al., 2009; Zervas et al., 2004). Pregnant mice carrying E13.5 embryos were injected with a dose of 4OH-TM for sufficient CreERT induction (2 mg of 4OH-TM per pregnant mouse), and the embryos were recovered on E18.5 (Fig. S1A-C). As shown in Fig. S1C, most of the roof-plate progeny cells labeled on E13.5 were localized along the midline of the dorsal spinal cord by E18.5. Co-staining of these tdTomato-positive cells with anti-nestin antibody revealed that the roof-plate progeny cells constituted 92.5% of the nestin-positive midline cells (197 of 213 nestin-positive cells; Fig. S1D-F). Given that the efficiency of tamoxifen-induced Cre activation is <100%, this result indicates that almost all of the nestin-positive midline cells were derived from roof-plate cells.

Similar to stretched roof-plate cells, neural progenitor cells also have a characteristic long process. The nuclei of neural progenitor cells demonstrate cell cycle-dependent movement known as interkinetic nuclear migration, during which mitosis occurs on the apical side. To determine whether elongated roof-plate cells also exhibit characteristics similar to those of neural progenitor cells, we examined the position of mitotic roof-plate cells by monitoring the phosphorylation of histone H3 (pHH3) on E12.5 and E13.5 (Fig. S1G-I). Although most of the pHH3-positive cells were detected on the apical side along the spinal cord lumen, except at the roof plate, which is marked by ezrin expression, pHH3 signals were not localized on the apical side of roof-plate cells on E12.5 (Fig. S1G,I). Furthermore, a pHH3 signal was hardly detected at the roof plate on E13.5, although the pHH3-positive cells along the spinal cord lumen were still detected (Fig. S1H,I). Therefore, the elongation of roof-plate cells did not appear to be a process that is common to all neuroepithelial cells.

To determine whether Wnt1 and Wnt3a expression is maintained throughout embryogenesis in stretched roof-plate cells, we performed in situ hybridization experiments (Fig. S2A-D). However, only faint staining of Wnt1 and Wnt3a mRNA was detected in stretched roof-plate cells (Fig. S2C,D). We therefore sought to evaluate Wnt1 and Wnt3a expression through direct detection of the endogenous distribution of Wnt1 and Wnt3a proteins; we used immunostaining with specific antibodies and also generated a line of knock-in mice in which endogenous Wnt3a was replaced with egfp-Wnt3a, which encodes Wnt3a fused in-frame with EGFP at the N-terminus [Wnt3a^{tm(gfp-Wnt3a)}; Fig. S3A,B]. In knock-in embryos, egfp-Wnt3a recapitulated the expression of endogenous Wnt3a, as evidenced by similarity between the GFP expression pattern and that of Wnt3a protein (Fig. S3F-H,I-L). Furthermore, in immunohistochemistry experiments using an anti-GFP antibody, we detected GFP-Wnt3a with higher sensitivity than in experiments using an anti-Wnt3a antibody. Additional confirmation that egfp-Wnt3a recapitulated the activity of endogenous Wnt3a was obtained by the absence of any change in Axin2 expression in homozygotes, as Axin2 is a well-known target of Wnt/β-catenin signaling (Fig. S3M,N). In addition, no obvious abnormalities were observed, which would have been expected if Wnt3a activity had been compromised (Fig. S3C-E,O; Takada et al., 1994).

In the immunostaining of wild-type embryos with anti-Wnt1 and anti-Wnt3a antibodies and of knock-in embryos with anti-GFP antibody, specific signals were observed in the dorsal-most region of the spinal cord on E10.5, as was previously shown by in situ hybridization (Fig. 3A; Fig. S2E-I; Fig. S3I-L; Parr et al., 1993). Signals were also detected along the apicobasal axis of roof-plate cells from E13.5 to E18.5, although there was some decrease in signal intensity on E18.5 (Fig. 3B-D). Thus, embryonic Wnt1 and Wnt3a expression continues during and shortly after the morphological change in roof-plate cells. In addition to these midline signals, Wnt1, Wnt3a and GFP-Wnt3a proteins were detected in the margin of the dorsal spinal cord on E10.5 and E13.5. Importantly, GFP-Wnt3a was detected up to the tip of the process of stretched roof-plate cells (Fig. 3C-D′), which indicates that endogenous Wnt was delivered to the tip of the process that faces the lumen of the central canal.

We next assessed Wnt signaling activity by monitoring Axin2 expression, which is activated by Wnt/β-catenin signaling (Fig. 3E-H′; Jho et al., 2002), or using a transgenic embryo harboring the ins-TOPGAL (TOPGAL) reporter gene (Fig. 3I-L′; Moriyama et al., 2007). Before roof-plate cell stretching, Axin2 is expressed in the dorsal spinal cord (Fig. 3E; Jho et al., 2002), consistent with the distribution of Wnt1 and Wnt3a (Fig. 3A; Fig. S2E,H). From E13.5 to E15.5, Axin2 expression was evident in stretched roof-plate cells (Fig. 3F,G), whereas its expression was reduced on E18.5 (Fig. 3H). In addition, a web-like pattern of weak Axin2 expression was observed in the middle of a transverse section of the spinal cord (Fig. 3G,H), which most likely corresponded to the previously reported activation of Wnt/β-catenin signaling in developing vessels (Stenman et al., 2008). On the other hand, as has been previously shown by a number of TOPGAL reporter analyses, Wnt signaling in the roof-plate cells was hardly detectable by this method; although this signaling in neural crest cells was detectable on E10.5 (Fig. 3I). From E15.5 to E18.5, TOPGAL expression was also detected in ependymal cells that surround the central canal, towards which the tips of the stretched roof-plate cells faced (Fig. 3K-L′). Therefore, consistent with the expression of Wnt1 and Wnt3a in stretched roof-plate cells, Wnt/β-catenin signaling that was detectable by Axin2 expression was activated in these cells. In addition, Wnt/β-catenin signaling that was detected by the TOPGAL reporter was also activated in cells in close proximity to the tip of the roof-plate cell processes.

Next, we examined the role of Wnt expressed in the roof-plate cells by inhibiting its secretion by these cells. We generated embryos with a specific conditional knockout (cKO) of Wls, which is required for Wnt secretion (Goodman et al., 2006; Bartscherer et al., 2006; Bänziger et al., 2006). The knockout mutant embryos were constructed using a floxed Wls allele and Wnt1-Cre transgene (Carpenter et al., 2010; Danielian et al., 1998). Although abundant expression of Wls was detected in the dorsal midline of the spinal cord of littermate controls on E13.5 (Fig. 2I), almost no immunoreactivity to Wls was detected in this region in the Wls cKO embryos (Fig. 2J). In contrast to littermate controls, Wls cKO embryos exhibited abundant expression of Wnt1 and Wnt3a proteins across an area wider than the roof plate (Fig. 2K-N). Considering that progeny cells of roof-plate cells were distributed outside the roof plate at this stage (Fig. S4), defective secretion of Wnt proteins is likely to be the reason for the prolonged persistence of Wnt1 and Wnt3a proteins within the progeny cells in Wls cKO embryos.

As in the case of Wnt1 and Wnt3a compound mutant embryos, patterning of the dorsal neural tube was impaired in the Wls cKO embryos on E10.5 (Fig. S5; Muroyama et al., 2002). On E13.5, Wls cKO embryos exhibited a malformed head, a predictable defect that is caused by a lack of Wnt signaling in the roof plate (Fig. S6A-E). An examination of transverse sections of the spinal cord revealed a shortened dorsal (but not ventral) spinal cord (Fig. S6F-H), probably resulting from a defect in cell proliferation caused by the loss of Wnt signaling during the early stages of roof-plate formation. EM observations showed that the alignment of the tips of the stretched roof-plate cells was disturbed in the Wls cKO embryos (Fig. 2A-D). In addition, EM observations and immunostaining for nestin showed that the bundle of the processes of roof-plate cells appeared to be thinner and was frequently branched in the Wls cKO embryos (Fig. 2E,F,O,P; Fig. S6I,I′), although the cytoskeletal structures appeared to have formed properly (Fig. 2G,H). In contrast, no obvious change in the distribution of acetylated tubulin was detected in the Wls cKO embryos (Fig. S6J,K). These results indicate that Wnt secreted by roof-plate cells is required for the coordinated changes in roof-plate cell morphology that occur during normal development.

To further explore this morphological defect in Wls cKO mutant embryos, we examined the development of roof-plate cells in earlier stages of embryogenesis (Fig. 4). On E13.5, when stretching is first observed in normal embryos, roof-plate cells appeared to be elongated along the midline based on ezrin expression, forming a ‘train’ of cells, as described above (Fig. 4B,E,H). By contrast, in Wls cKO mutant embryos the ezrin-positive cells did not rearrange in a train-like manner (Fig. 4C,F,I). Thus, defective coordination of roof-plate cells on E13.5 in Wls cKO mutant embryos (Fig. 4J) appears to have induced the morphological abnormality that is evident in these cells in later stages. Because Wnt signaling was active in roof-plate cells during stretching, it appears to be plausible that Wnt activity was directly involved in this rearrangement. However, as the number of ezrin-positive cells was slightly increased in Wls cKO mutant embryos (Fig. S7), it also appears to be possible that the defective rearrangement of ezrin-positive roof-plate cells was inhibited by crowding of roof-plate cells in these embryos.

As described in the section ‘Wnt expression during the stretching of roof-plate cells’, expression of the TOPGAL reporter was detected in ependymal cells on E18.5 (Fig. 3L′). As the ventral tip of the stretched roof-plate cells faced the dorsal surface of the central canal, we hypothesized that Wnt proteins secreted by roof-plate cells were required for this Wnt/β-catenin signaling detected by the TOPGAL reporter. To support this idea, we observed the delivery of GFP-Wnt3a proteins up to the tip (Fig. 3C′,D′), and numerous vesicles at the tip, which suggested active secretion of proteins via secretory vesicles (Fig. S1J). We further tested this hypothesis from several different aspects.

As Wls is essential for the secretion of Wnt proteins, we examined the expression of TOPGAL reporter in ependymal cells in Wls cKO embryos. In contrast to littermate controls, Wls cKO embryos exhibited a significant reduction in the frequency of TOPGAL-positive cells among ependymal cells on E15.5 and E18.5 (Fig. 5A-F). Furthermore, this reduction was more severe in the ependymal cells that lined the dorsal half of the central canal (Fig. 5D-F). It therefore appears to be plausible that Wnt proteins secreted by roof-plate cells are specifically required for activation of Wnt/β-catenin signaling in some of the dorsal ependymal cells. On the other hand, our findings also suggested that some other Wnt proteins that had not been secreted by roof-plate cells activated Wnt/β-catenin signaling in other ependymal cells.

However, as it still appeared to be possible that the defects in Wnt signaling in dorsal ependymal cells in Wls-cKO embryos could have been a secondary consequence of impaired Wls function in earlier developmental stages, we next tried to suppress the function of Wls after the onset of roof-plate stretching by using a floxed Wls allele and Wnt1-CreERT transgene (Carpenter et al., 2010; Zervas et al., 2004). To distinguish Wls cKO embryos generated using Wnt1-CreERT from those generated by Wnt1-Cre, henceforward, we refer to the former as Wnt1-CreERT; Wls cKO and the latter as Wnt1-Cre; Wls cKO embryos. Pregnant mice were injected with 4OH-TM every other day from day 13.5 post coitum (Fig. 6A-C). The resulting Wnt1-CreERT; Wls cKO embryos exhibited no obvious abnormalities in terms of the morphology of stretched roof-plate cells, as assessed by immunohistochemistry using an anti-nestin antibody and by EM observations (Fig. 6D-G; Fig. S8A-C). In contrast, Wnt signaling was specifically reduced in the ependymal cells in these embryos (Fig. 6H-J; Fig. S8D-G). This result indicates that the reduced Wnt activation that was observed in ependymal cells in Wnt1-Cre; Wls cKO embryos, as described above, was not a secondary result of the morphological defect in the stretched roof-plate cells or any other defect that occurred before E13.5.

On the other hand, it also remained possible that roof-plate progenies had differentiated into ependymal cells during or after the stretching and that Wnt signaling remained activated in these cells. To examine the contribution of progeny roof-plate cells to the formation of ependymal cells, we labeled roof-plate progenies using a Wnt1-CreERT allele in conjunction with the R26R-tdTomato reporter (Fig. 6N-P). On E18.5, we hardly detected roof-plate progenies among the ependymal cells; only in a few sections did we observe roof-plate progenies in the most dorsal portion of the area surrounding the central canal, and the average number of these progenies in this area was slightly increased on postnatal day 4 (Fig. 6O,P; Fig. S9). As the number of roof-plate progenies among the ependymal cells was significantly low compared with that of the roof plate-dependent increase in the number of Wnt-active cells (Fig. 6P; Fig. S8G,H), the results obtained from these two types of Wls-cKO embryos strongly suggest that Wnt secretion by stretched roof-plate cells is required for activation of Wnt signaling in the ependymal cells that surround the dorsal central canal on E18.5. As most ependymal cells are derived from the p2 and pMN domains in the ventral spinal cord, in which Wnt/β-catenin signaling is not activated before stretching, rather than from roof-plate cells (Fig. 6O; Fig. S1C; Yu et al., 2013), our results appear to indicate that stretching of the roof-plate cells altered the target cells of the Wnt signals that are secreted by roof-plate cells.

Neural stem cells have been identified in the ependymal zone that surrounds the central canal in the adult spinal cord (Johansson et al., 1999; Martens et al., 2002; Weiss et al., 1996). Therefore, we next investigated whether Wnt proteins secreted by roof-plate cells could be required for the proliferation of ependymal cells. On E18.5, Wnt1-Cre; Wls cKO mutant embryos exhibited a significant decrease in the frequency of Ki67 (also known as Mki67)-immunoreactive proliferating cells among the ependymal cells (Fig. 5G-I). Furthermore, ependymal cells surrounding the dorsal half of the central canal exhibited more severe defects on E18.5 than the cells surrounding the ventral half (Fig. 5I). A similar decrease in the frequency of Ki67-immunoreactive proliferating cells was observed in Wnt1-CreERT; Wls cKO embryos recovered from pregnant mice that had been injected with 4OH-TM every other day from day 13.5 post coitum (Fig. 6K-M). Furthermore, compared with the contribution of roof-plate progenies to the formation of ependymal cells (Fig. 6N-P), the effect of Wls-deficiency on proliferation of ependymal cells was evident. Based on these results, we concluded that Wnt secretion by stretched roof-plate cells was required for normal proliferation of the ependymal cells surrounding the central canal, at least on E18.5.

In normal adult mice, neurogenesis rarely occurs in the spinal cord; rather, after injury to the spinal cord, ependymal cells generate progeny that undergo multiple fates, which suggests that ependymal cells exhibit latent neural stem cell properties (Johansson et al., 1999; Meletis et al., 2008; Shechter et al., 2007; Habib et al., 2016). We therefore examined whether Wnt secretion by roof-plate cells is required to induce the proliferation of ependymal cells after spinal cord injury in adult mice. As Wnt1-Cre; Wls cKO mice die immediately after birth, we compared the effect of spinal cord injury between control and Wnt1-CreERT; Wls cKO mice from a mother that had been injected with 4OH-TM every other day from E12.5 to E16.5 (Fig. 7A). In control embryos, proliferation of ependymal cells, as assessed by the presence of cells that are positive for both Ki67 and vimentin, was generally activated 2 days after spinal cord injury (Fig. 7B; Fig. S10A-D). In contrast, injury-induced proliferation of ependymal cells was suppressed in the Wnt1-CreERT; Wls cKO mice (Fig. 7C,D). Furthermore, cells in the dorsal half of the central canal exhibited a more severe defect than the cells in the ventral half (Fig. 7D). These data demonstrate that Wnt secretion by roof-plate cells is required for the normal proliferation of dorsal ependymal cells after injury to the spinal cord. In contrast, proliferation of ventral ependymal cells was not decreased after spinal cord injury, suggesting that Wnt ligands were reactivated in the ventral spinal cord independently of roof-plate cells.

Interestingly, expression of Wnt1 (but not Wnt3a) was reactivated on the injured side of the spinal cord and along the stretched roof-plate cells extending between the pia and central canal 2 days after spinal cord injury (Fig. 7E-G; Fig. S10E,F). Furthermore, Wls expression was also reactivated in the stretched roof-plate cells (Fig. S10G-H′), which suggests that Wnt1 secretion was reactivated in stretched roof-plate cells after spinal cord injury. In contrast, apparent reduction of Wnt1 reactivation was observed in stretched roof-plate cells of Wnt1-CreERT; Wls cKO mice, although reactivation in other areas of the spinal cord was only slightly affected by spinal cord injury (Fig. 7F-J). Therefore, reactivation of Wnt1 secretion by roof-plate cells appears to play a role in promoting ependymal cell proliferation and Wnt signaling.

The roof plate functions as an organizing center during the development of the dorsal spinal cord. Considerable evidence has shown that Wnt produced by roof-plate cells plays multiple roles before the stretching of the roof plate (Ikeya et al., 1997; Ikeya and Takada, 1998; Muroyama et al., 2002; Riccomagno et al., 2005). For example, Wnt1 and Wnt3a expressed by roof-plate cells play roles in regulating the proliferation and specification of neural progenitor cells in the dorsal neural tube (Lee et al., 2000; Muroyama et al., 2002). After stretching, the apical tips of the processes of roof-plate cells become close to the ependymal cells that surround the central canal. Notably, our experiments using GFP-Wnt3a knock-in embryos showed that endogenous Wnt proteins are delivered to the tip of the elongated processes. Use of the TOPGAL reporter showed that Wnt signaling is activated in the ependymal cells that surround the central canal, but that this signaling is impaired in Wnt1-CreERT; Wls cKO embryos, in which Wls expression is impaired almost exclusively in elongated roof-plate cell progeny after E13.5. Therefore, Wnt proteins secreted from the roof-plate cells appear to have another role, in addition to their earlier roles in the organizing center. Considering that most ependymal cells originate from cells in the ventral region, p2 and pMN domains (Yu et al., 2013) of the spinal cord, our results strongly suggest that the stretching morphogenesis of signal-producing cells results in a change in the signal-receiving cells during the development of the spinal cord.

Interestingly, our data also indicate that Wnt proteins play a role in the morphogenesis of signal-producing cells. In Wnt1-Cre; Wls cKO embryos, the development of roof-plate cells, which produce Wnt1 and Wnt3a, is impaired. That is, coordinated rearrangement of the stretched roof-plate cells does not occur on E13.5, and this defect likely causes morphological abnormalities in these cells during later stages of development. These findings suggest that Wnt proteins actively regulate the change in cell morphology, thus revealing a novel function. Similarly, signaling by other molecules is required for the formation and maintenance of cell processes that are involved in the intercellular transfer of signaling proteins (Hsiung et al., 2005; Roy et al., 2014; Inaba et al., 2015). However, we cannot exclude another possible explanation for the abnormal morphology of roof-plate cells in Wnt1-Cre; Wls cKO embryos. The reduced proliferation of dorsal spinal cord cells in response to the reduction in Wnt secretion by roof-plate cells before their stretching could, in turn, affect the stretched morphogenesis of the roof-plate cells. In addition, as roof-plate cell progeny are distributed across a wide area of the spinal cord, we must consider the distribution of these cells in Wnt1-Cre; Wls cKO embryos. More-extensive analyses could perhaps clarify the mechanism by which roof-plate cells stretch during normal development of the spinal cord.

It is thought that intercellular signals regulate neural progenitors in both embryos and adults. For instance, Wnt signaling regulates the population of neural stem/progenitor cells in the subventricular zone (SVZ) of the cerebrum and the subgranular zone of the hippocampus in adult mammals (Adachi et al., 2007; Bowman et al., 2013; Lie et al., 2005; Maretto et al., 2003). Brain injury leads to an increase in Wnt signaling in the SVZ (Piccin and Morshead, 2011). However, despite the importance of these intercellular signals, which cells provide signaling proteins to the neural cell precursors remains a matter of debate. In the spinal cord, neural stem/progenitor cells are among the population of ependymal cells that surround the central canal, which is connected to the brain ventricle (Johansson et al., 1999; Meletis et al., 2008; Weiss et al., 1996). In this study, we showed that the proliferation of ependymal cells, including neural stem/progenitor cells, is also regulated by intercellular signals. The proliferation of ependymal cells is promoted via the close proximity of roof-plate cells resulting from their stretching. As Wnt signaling by ependymal cells is enhanced as a result of roof-plate cell stretching, it is plausible to conclude that Wnt is a ligand that is secreted by stretched roof-plate cells to activate neural progenitor/stem cells in the spinal cord. An interesting area of future study is whether brain neural progenitors are also regulated by the long-range transport of Wnt proteins that is mediated by stretched roof-plate cells, as in the case of the spinal cord.

Finally, we recognize that Xing et al. (2018) also reported that roof-plate cells are responsive to Wnt signaling by utilizing Axin2-CreERT2 mice during the revision period of this paper. They showed that the Wnt-responsive cells contribute to the ependymal cells of the spinal cord in the adult mouse. In contrast, our study identified Wnt ligands that are expressed in stretched roof-plate cells and showed evidence that secretion of these Wnt ligands increased Wnt signaling in ependymal cells and was specifically required for proliferation. Furthermore, we also showed the requirement for Wnt in roof-plate cells for response to spinal cord injury. Thus, Wnt signals appear to regulate ependymal cells in several distinct ways.

This study was performed in accordance with the Guidelines for Animal Experimentation of the National Institutes of Natural Sciences, with approval of the Institutional Animal Care and Use Committee of the National Institutes of Natural Sciences (#15A183, #16A071, #17A062). Every effort was made to minimize animal suffering during experimental procedures.

EIIa-Cre (Lakso et al., 1996), ins-TOPGAL (Moriyama et al., 2007), R26R-Confetti (Snippert et al., 2010), R26R-tdTomato (Madisen et al., 2009), Wls flox (Carpenter et al., 2010), Wnt1 KO (McMahon and Bradley, 1990), Wnt1-Cre (Danielian et al., 1998), Wnt1-CreERT (Zervas et al., 2004) and Wnt3a KO (Takada et al., 1994) mice have been previously described. To generate egfp-Wnt3a knock-in mice, FLAG-EGFP was integrated into exon 1 of the gene encoding Wnt3a, immediately behind the signal peptide, and the gene was inserted into the DT-A-pA/loxP/PGK-Neo-pA/loxP vector (www.cdb.riken.go.jp/arg/cassette.html). The neomycin cassette was removed by crossing with EIIa-Cre mice. Genotyping was carried out by PCR using the following primers: Wnt3a-E1F, 5′-ATGGCTCCTCTCGGATACCT-3′; Wnt3a-I1R, 5′-ACTTTCACTGCCCTCCCTTC-3′; and GFP-R1, 5′-TCACGAACTCCAGCAGGACCATG-3′. The wild-type, homozygous and heterozygous genotypes were indicated by a 290-bp fragment, a 756-bp fragment and both fragments, respectively.

Before E13.5, embryos were fixed in 3.5% paraformaldehyde (PFA) for 30 min or overnight at 4°C. Embryos older than E15.5 were perfused with 4% PFA via the heart. Eight-week-old mice were overdosed with sodium pentobarbital (70 mg/kg) and perfused via the heart with PBS followed by 4% PFA. The spinal cord was dissected and post-fixed overnight at 4°C.

Immunohistochemistry was performed on cryosections of tissue as described below. Sections were incubated overnight at 4°C with the following primary antibodies : anti-β-galactosidase (PM049, MBL International; 1:400), anti-ezrin (ab4069, Abcam; 1:400), anti-GFP (598, MBL International; 1:500), anti-Gpr177 (ab176376, Abcam; 1:400), anti-Ki67 (ab16667, Abcam; 1:1000), anti-mash1 (556604, BD Pharmingen; 1:400), anti-nestin (60051A, BD Pharmingen; 1:500), anti-phospho-histone H3(Ser10) (06-570, Upstate; 1:400), anti-RFP (PM005, MBL International; 1:500), anti-tubulin, acetylated (T6793, Sigma-Aldrich; 1:1000), anti-vimentin (V2258, Sigma-Aldrich; 1:500), anti-Wnt1 (ab15251, Abcam; 1:200) and anti-Wnt3a (1:10 dilution; Takada et al., 2006). The cryosections were then incubated with secondary antibodies at a 1:500 dilution for 2 h at room temperature. In this study, the following secondary antibodies were used: goat anti-mouse IgG Alexa Fluor 488 (A-11029, Invitrogen), goat anti-mouse IgM Alexa Fluor 488 (A-21042, Invitrogen), goat anti-mouse IgG Alexa Fluor 647 (A-21235, Invitrogen) and goat anti-rabbit IgG Alexa Fluor 555 (A-21429, Invitrogen). The tissue sections were counterstained with DAPI (Dojindo). Wnt1 signals were amplified using a TSA Plus kit (PerkinElmer). To obtain Wnt3a signals, embryos were fixed in 3.5% PFA for 30 min at 4°C, and then the antigen was retrieved by incubation in Antigen Unmasking Solution (Vector Laboratories) for 15 min at 95°C. Fluorescent images were acquired using an inverted confocal microscope (Nikon A1Rsi).

Cryosections were subjected to in situ hybridization as follows. Briefly, sections were re-fixed with 4% PFA for 15 min at room temperature, washed with PBS, treated with 6 µg/ml of proteinase K for 20 min, incubated in 0.2 M HCl for 10 min and then acetylated using 0.1 M triethanolamine and 0.25% acetic anhydride. The sections were then incubated in hybridization buffer (50% formamide, 5× SSC, 1% SDS, 50 µg/ml tRNA and 50 µg/ml heparin) overnight at 55°C. Probes were fragmented to ∼300 bases by alkaline hydrolysis. The next day, the sections were washed with 2× SSC/50% formamide, treated with 10 µg/ml RNaseA in TNE for 30 min at 37°C, and then washed consecutively with 2× SSC, 0.2× SSC, and Tris-buffered saline with Tween 20 (TBST). The sections were incubated with 1% blocking reagent (Roche) in TBST for 1 h and then treated with a 1:500 dilution of anti-digoxigenin-AP Fab fragments (Roche) overnight at 4°C. The following day, the sections were washed with TBST and alkaline phosphatase buffer (NTMT), and signals were developed using BM Purple (Roche). The following probes were used: Axin2 (Jho et al., 2002), Wnt1 (Parr et al., 1993) and Wnt3a (Roelink and Nusse, 1991).

R26R-Confetti mice were crossed with Wnt1-CreERT mice. To induce Cre-mediated recombination, pregnant mice were injected with 1 mg of 4OH-TM dissolved in ethanol, dimethyl sulfoxide and sesame oil on E11.5 or E13.5. The spinal cord was harvested on E13.5 or E15.5 after perfusion fixation. Sections were cut using a vibrating microtome at a thickness of 400 µm. Images were captured using a multiphoton microscope (Leica TCS SP8 MP).

Eight-week-old mice were anesthetized via intraperitoneal injection of a mixture of medetomidine (0.3 mg/kg), midazolam (4 mg/kg) and butorphanol (5 mg/kg). A laminectomy was performed at the level of T13. A right lateral hemisection injury was created with the tip of a scalpel. After the procedure, the mice were awakened via intraperitoneal injection of atipamezole (0.3 mg/kg). Mice were sacrificed by perfusion with PFA 1, 2 or 6 days after spinal cord injury.

Embryos were perfused via the heart with a mixture of 2% PFA, 2.5% glutaraldehyde and 100 mM sodium cacodylate buffer. The spinal cord was dissected and post-fixed for 2 h at room temperature in the same fixative.

Statistical analyses were performed using Excel and R software (version 3.5.1). Differences were assessed for statistical significance using the Student's t-test. A P value of <0.05 was considered indicative of statistical significance. Error bars in figures represent the standard error of the mean for each group.

We thank Drs Abe and Kiyonari for generating the GFP-Wnt3a knock-in allele, Drs Kadomatsu and Morozumi for their support in producing spinal cord injury, Ms Misaki and Takahashi for their technical support in EM observation, Drs Ikenaka, Ishino and Kitadate for their technical advice regarding experiments and the Functional Genomics Facility and Spectrography and Bioimaging Facility of the NIBB Core Research Facilities for their technical support. All members of S.T.’s laboratory are gratefully acknowledged for their helpful discussions.