Microtubules (MTs) regulate numerous cellular processes, but their roles in brain morphogenesis are not well known. Here, we show that CAMSAP3, a non-centrosomal microtubule regulator, is important for shaping the lateral ventricles. In differentiating ependymal cells, CAMSAP3 became concentrated at the apical domains, serving to generate MT networks at these sites. Camsap3-mutated mice showed abnormally narrow lateral ventricles, in which excessive stenosis or fusion was induced, leading to a decrease of neural stem cells at the ventricular and subventricular zones. This defect was ascribed at least in part to a failure of neocortical ependymal cells to broaden their apical domain, a process necessary for expanding the ventricular cavities. mTORC1 was required for ependymal cell growth but its activity was downregulated in mutant cells. Lysosomes, which mediate mTORC1 activation, tended to be reduced at the apical regions of the mutant cells, along with disorganized apical MT networks at the corresponding sites. These findings suggest that CAMSAP3 supports mTORC1 signaling required for ependymal cell growth via MT network regulation, and, in turn, shaping of the lateral ventricles.
Microtubules (MTs) are fundamental for a wide range of cellular processes such as cell division, polarization, migration, differentiation and growth. MT disruption dysregulates these processes, leading to severe abnormalities in cell and/or body functions, and in turn developmental disorders (Kapitein and Hoogenraad, 2015; Lasser et al., 2018; Matamoros and Baas, 2017). MTs are classified into two major categories, centrosomal and non-centrosomal. Non-centrosomal MTs emanate from sites other than the centrosome, such as the Golgi apparatus, nuclear envelope and apical cortex (Muroyama and Lechler, 2017; Sanchez and Feldman, 2017; Toya and Takeichi, 2016; Wu and Akhmanova, 2017). Although general functions of MTs in cellular and developmental processes have been well studied, specific roles of the non-centrosomal MT subsets are less known.
The CAMSAP3 (Nezha)/Patronin family proteins are responsible for assembly of non-centrosomal MTs (Baines et al., 2009; Goodwin and Vale, 2010; Jiang et al., 2014; Meng et al., 2008; Tanaka et al., 2012). These proteins specifically bind to the MT minus ends, and are localized at various subcellular sites to which they tether MTs, resulting in formation of MT networks distinct from those organized by the centrosomes. For example, CAMSAP3 is localized at the apical cortex of intestinal epithelial cells, producing apicobasal arrays of MTs in these cells (Muroyama et al., 2018; Toya et al., 2016). CAMSAP2 and CAMSAP3 are differentially distributed in neuronal dendrites and axons, controlling their extension and polarity (Pongrakhananon et al., 2018; Yau et al., 2014). CAMSAP2 is also required for endothelial cell polarization in zebrafish (Martin et al., 2018), and Patronin is a key determinant of the anterior-posterior axis in Drosophila oocytes (Nashchekin et al., 2016). Thus, evidence has accumulated that this family of proteins plays many important roles in regulation of cell structure and function, but it remains unknown how much their functions contribute to the generation of higher-order structures such as tissues and organs.
The brain ventricles are cavities that are filled with cerebrospinal fluid (CSF). CSF circulates through multiple compartments of the ventricles, supporting the homeostasis of neurons and glia, and also maintaining quiescent adult neural stem cells (NSCs) (Delgado et al., 2014; Kokovay et al., 2012; Lowery and Sive, 2009). A major pool of adult NSCs is located in the zone adjacent to the surface of lateral ventricles, called the ventricular-subventricular zone (V-SVZ), in which NSCs directly contact the CSF with their apical processes (Lim and Alvarez-Buylla, 2016; Mirzadeh et al., 2008). Once the ventricle is stenosed or closed, CSF circulation is occluded, leading to hydrocephalus (Kahle et al., 2016; Kousi and Katsanis, 2016; Lowery and Sive, 2009). At the areas of closure, neurogenesis is no longer maintained in the adult V-SVZ (Shook et al., 2012). The ventricles are lined with ependymal cells that develop multiple cilia on their apical surfaces, and these cilia play a crucial role in CSF circulation (Banizs et al., 2005; Ibañez-Tallon et al., 2004; Sawamoto et al., 2006).
In the present study, we show that CAMSAP3 is important for lateral ventricle shaping. Through phenotypic analysis of a mouse line in which Camsap3 is mutated so as to abolish its function, we found that CAMSAP3 dysfunction causes various defects at the lateral ventricles, including enhancement of their closure and depletion of adult NSCs. Exploring mechanisms underlying these defects showed that CAMSAP3 supports mTORC1 signaling, which is important for the growth of ependymal cells, possibly through control of MT-dependent lysosomal positioning. Loss of this mechanism inhibited the normal expansion of the lateral ventricles, leading to their increased closure and associated events. These observations suggest that the CAMSAP3–non-centrosomal MT system is indispensable for constructing robust lateral ventricles, which support brain homeostasis.
Increased ventricle stenosis and fusion in Camsap3 mutant brains
We used a Camsap3 mutant mouse, Camsap3dc/dc, in which the coding sequence for the C-terminal CKK domain of CAMSAP3, required for its binding to MTs, had been deleted from both alleles, resulting in expression of a C-terminal domain-truncated, non-functional CAMSAP3 protein (Toya et al., 2016). Brains from adult mutant mice were slightly smaller than those from wild-type (WT) mice, having distinctly smaller olfactory bulbs (Fig. 1A-C). Histological analysis of these brains at postnatal day (P) 28.5 showed no obvious gross defects in their cytoarchitecture, except for ventricular abnormalities (Fig. 1D). In WT brains, the lateral ventricle was widely open at the dorsal regions, with narrowing at more ventral regions where stenosis or fusion occurred to variable extents among the individuals. The lateral ventricles in mutant brains exhibited much smaller cavities at the dorsal regions and more extensive stenosis or fusion at the ventral regions than in WT brains. Such ventricular defects were not clearly detectable in fetal brains (Fig. S1), suggesting that the observed ventricular defects were elicited at postnatal stages. Immunostaining WT brains at P28.5 for CAMSAP3 showed that this protein is concentrated along the ependymal cell zones that line the lateral ventricles, and its mutated counterpart showed a similar distribution (Fig. 1G), which implied that the Camsap3 mutant phenotypes observed above might be related to ependymal cell structure or function.
Therefore, we closely examined the layers of ependymal cells by immunostaining for an ependymal cell marker S100β (a calcium binding protein). In WT ventricular surfaces (VSs), S100β was detected through the VSs. In mutant brains, however, the regions lacking S100β+ ependymal cells were increased (Fig. 1H-J). We also used another marker, parvalbumin (PV), which is specifically expressed in ependymal cells at the stenosed region (Filice et al., 2017), and found that the number of PV+ cells was increased in mutant brains (Fig. S2A,B), confirming the enhanced stenosis in mutant brains. These results indicate that Camsap3 mutation promotes ventricular stenosis or fusion at the boundaries between the striatum and septum.
We also analyzed the anterior lateral ventricles using en face views of the whole-mount striatum, identifying three distinct regions on the VS of both genotypes: (1) the S100β-negative region, (2) the region with brighter S100β signals, and (3) the region with darker and dotted S100β signals (Fig. 1K,L). Comparing these observations with those obtained using coronal sections, we can assume that region 1 corresponds to the fused region, region 2 to the stenosed region, owing to its PV expression (Fig. S2C-E), and 3 to the ‘conventional’ ependymal cell layer that faces to the open ventricle. Analysis of these specimens showed that the fused and stenosed regions covered larger parts of the VSs in the mutant striatum than in WT, which resulted in a decrease of the conventional area in the mutants (Fig. 1M,N).
Ventricle stenosis and fusion frequently cause hydrocephalus (Kahle et al., 2016; Kousi and Katsanis, 2016), which is anatomically characterized by abnormal enlargement of the ventricle. However, Camsap3 mutant mice did not show any sign of ventricle enlargement, at least in the telencephalon, rather they showed ventricle narrowing.
Decrease of adult NSCs in the V-SVZ of the Camsap3 mutant striatum
Ependymal cells are part of the NSC niche at the V-SVZ (Fig. 2A) (Kokovay et al., 2008; Lim and Alvarez-Buylla, 2016; Paez-Gonzalez et al., 2011). Therefore, loss of ependymal cells could affect maintenance of adult NSCs. To determine whether adult NSCs are normally present at the striatal V-SVZ of Camsap3 mutant mice, we immunostained for VCAM1, a cell surface sialoglycoprotein that had been proposed as a marker for adult NSCs (Kokovay et al., 2012). En face views (top views) of the striatal VSs in a whole-mount specimen confirmed that VCAM1 was exclusively localized on the surface of clustered cells having small apical domains (Fig. 2B), which can be anatomically identified as NSCs. Immunostained coronal sections showed that VCAM1 signals were detected as puncta distributed along the VS in both WT and Camsap3 mutant brains at P28.5 (Fig. 2C). These puncta closely associated with a tip of GFAP+ processes, which ensures that VCAM1 immunosignals represent parts of NSCs (Fig. S3A). Importantly, such VCAM1 signals were greatly reduced at the regions where stenosis or fusion was morphologically recognized, regardless of genotype, suggesting that NSCs were eliminated as a result of ventricle closure. Whole-mount samples confirmed that VCAM1+ clusters were mostly restricted to the conventional layer in both genotypes (Fig. 2D,E). In addition, even at the conventional area, the number of VCAM1+ clusters/area was decreased in about a half of the mutant mice examined (Fig. 2F). Overall, the number of VCAM1+ clusters was significantly decreased in the mutants (Fig. 2G).
We next examined adult NSCs using another approach, a BrdU-label retention assay. This assay exploits the slowly dividing nature of adult NSCs (Fig. 2H) (Doetsch et al., 1999; Furutachi et al., 2015); once labeled, they retain the labels for a long period after the labeling protocol has ended. By contrast, rapidly dividing cells, such as transit amplifying cells, dilute the BrdU labels via multiple rounds of cell division. We administered P28.5 mice with BrdU for one week, followed by a non-administrating period of two weeks, until sacrifice (Fig. 2H). The results showed that fewer BrdU-retaining cells were detected in the V-SVZ of Camsap3 mutant striatum at P49.5 (Fig. 2I-K), consistent with the idea that NSCs were decreased. On the other hand, we did not find any sign of cell proliferation defects in embryonic striata, as assessed by immunostaining for Ki67 or PH3, as markers for dividing cells (Fig. S1D-F).
Adult NSCs in the V-SVZ produce interneurons that migrate to the olfactory bulb (OB) (Lim and Alvarez-Buylla, 2016). Therefore, we expected less production of OB interneurons in Camsap3 mutant mice, and this might explain why they have smaller OBs. To test this idea, we estimated the number of BrdU-retaining neurons in the entire granule cell layer (GCL) of the OB and found that the number was decreased in mutant OBs (Fig. S4). Although the ratio of BrdU-retaining neurons to GCL neurons and the density of cells expressing NeuN, a neuronal marker, in the GCL were comparable between WT and mutant OBs (Fig. S4A-C,F), the total number of BrdU-retaining neurons was decreased in mutant OBs (Fig. S4E), owing to the reduction in their overall size (Fig. S4D; Fig. 1A-C). Furthermore, we found that the decrease of BrdU-retaining neurons was restricted to the calretinin (CalR)+ subpopulation of GCL neurons (Fig. S4H,I), leaving other subpopulations statistically unaffected (Fig. S4J-N). This result is consistent with a previous observation (Merkle et al., 2007) that CalR+ neurons are generated from NSCs located at the surface of the anterior portion of the lateral ventricle, which was abnormally closed in Camsap3 mutant mice. Together, these results consistently suggest that the morphological defects in the lateral ventricles resulted in a decrease of adult NSCs at the V-SVZ of Camsap3 mutant striatum.
In addition, we noted that VCAM1+ cells had smaller apical domains in the mutants, although they formed clusters as WT cells did, and the number of cells per cluster did not change (Fig. S3B-E), suggesting the possibility that CAMSAP3 dysfunction may also have affected morphology of NSCs. On the other hand, we did not find morphological differences in GFAP+ processes of VCAM1+ cells between WT and mutant striata (Fig. S3A). Thus, the architectural changes in NSCs that remain in the Camsap3-mutanted striatum appear to be limited to apical domain size, as far as we analyzed.
Failure of ependymal cells to grow normally at Camsap3 mutant neocortices
Then, we began to explore mechanisms by which loss of functional CAMSAP3 led to the narrowing of the lateral ventricle. This morphological change included shortening of the neocortical VSs along the mediolateral axis (Fig. 1D-F). We paid attention to this abnormality, because we can presume that the reduced growth of the neocortical VSs would lead to narrower opening of the ventricular cavity. To seek the cellular basis of the shortening of neocortical VSs in the mutants, we first observed morphology of individual ependymal cells in P28.5 brains, by immunostaining for N-cadherin as a cell junction marker, as well as for γ-tubulin, a component of centrosomes or basal bodies (BBs), which are multiplied to produce multiple cilia in differentiated ependymal cells. We found that differentiation of ependymal cells normally took place at mutant neocortices, as assessed by appearance of a multiple BB cluster in each cell. Remarkably, however, their apical domain size was significantly reduced in adult mutant brains (Fig. 3A,B). Moreover, the number of ependymal cells was also decreased along the mediolateral length of the neocortex in the mutant brains (Fig. 3C,D). Overall, the mediolateral length of VSs was reduced by 60% in the mutant neocortex at P28.5.
The ependymal cell apical domain is known to enlarge during the postnatal period at the striatum in a planar cell polarity (PCP) fashion (Redmond et al., 2019). We therefore examined whether this would also be the case at the neocortex by observing developmental processes of ependymal cells. At embryonic day (E) 18.5, a stage before ependymal cell differentiation, the neocortical VSs were solely covered by mono-ciliated radial glial cells, defined as ependymal precursor cells (Spassky et al., 2005) (Fig. 3A). The apical domain size in these cells did not differ between WT and mutant brains (Fig. 3B). At P7.5, centriole amplification had begun in some cells to produce multiple cilia, which was accompanied by apical domain growth in both genotypes (Fig. 3A). Usually, the apical domain size was the smallest in cells with a monomeric BB, and it increased along with centriole amplification, reaching the maximum in cells that have acquired multiple BBs. The ratio of such differentiating cells (with multiple BBs) to undifferentiated cells (with a monomeric BB) did not greatly differ between the two genotypes (Fig. S5A), nor did their apical domain sizes differ at this stage (Fig. 3B). At P14.5, however, a difference in the apical domain size between WT and mutant cells became evident. Although the VSs at this age were covered primarily by multiciliated cells in both genotypes, those in mutant neocortices displayed narrower apical domain areas than WT cells (Fig. 3A,B). By P28.5, apical domain growth was largely completed, and the difference in apical domain size between WT and mutant cells was thus fixed (Fig. 3A,B). To summarize, apical domain growth in Camsap3-mutated ependymal cells was impaired during postnatal periods. Closer cytological observations further indicated that the number of BBs and the areas occupied by BBs at the apical domain of ependymal cells did not particularly differ between WT and mutant neocortices; that is, only the non-ciliated areas failed to grow normally in the mutants.
We also analyzed apical domain size in ependymal cells lining the striatum of P28.5 brains, at the dorsal portions which are not closed, and found that the size was only slightly reduced in the mutants (Fig. S3F,G). We did not perform this analysis for the closed portions, as normal ependymal cells did not always persist there. Concerning the reduction in ependymal cell number in mutant neocortices, we could not determine how this occurred. Ependymal cell differentiation starts early in the postnatal period and, once differentiated, they become postmitotic (Spassky et al., 2005). We did not find any decrease in cell number at E18.5, suggesting that cell proliferation proceeded normally in the mutant neocortices. We also examined potential involvement of cell death. However, our immunostaining for cleaved caspase 3 did not detect any clear apoptotic figures at neocortical ventricular zones in either WT or mutants at P7.5 and P28.5. We suspected that some delamination of ependymal cells might have been induced at the mutant neocortices, but we did not capture any images that support this idea.
Downregulation of mTORC1 signaling in Camsap3 mutant ependymal cells
Previous work has suggested that mTORC1 signaling regulates the apical domain size of radial glial cells (Foerster et al., 2017). We therefore examined whether an mTORC1-dependent mechanism is also involved in ependymal cell growth. We immunostained neocortical VSs of WT brains at various stages for phosphorylated S6 ribosomal protein (p-S6RP), an mTORC1 signaling activity readout (Fig. 4A) (Ruvinsky and Meyuhas, 2006), observing its distribution. At E18.5, p-S6RP was only faintly detected in radial glial or ependymal precursor cells (Fig. 4B). At P7.5, cytosolic p-S6RP became detectable in some of the ventricular cells (Fig. 4B,C), and expression of active mTORC1 at this stage was also confirmed by western blotting (Fig. 4D). Triple immunostaining for p-S6RP, γ-tubulin and N-cadherin showed that p-S6RP was detected in enlarged cells, particularly in those having amplifying centrioles (Fig. 4E-G) that are identified by the appearance of deuterosomes (Al Jord et al., 2014; Shahid and Singh, 2018), suggesting that activation of mTORC1 signals coincides with ependymal cell differentiation. In cells that had acquired multiple BBs, however, p-S6RP signals dropped, although some remained along the plasma membranes (Fig. 4B,E-G). These observations suggest that mTORC1 is transiently activated in developing ependymal cells at the WT neocortex. To test whether this activation of mTORC1 is important for apical domain growth of ependymal cells, we injected postnatal WT mice with the mTORC1 inhibitor rapamycin. Daily injections of the inhibitor from P6.5 for eight consecutive days resulted in a marked decrease of the apical domain size (Fig. S5B), confirming that mTORC1 signaling drives apical domain growth of ependymal cells.
We then asked what happened to mTORC1 signals in Camsap3 mutant ependymal cells. Total phosphorylation level of mTOR was not much changed in the mutant neocortex (Fig. 4D). Immunostaining of neocortical VSs for p-S6RP in mutant cells showed its distribution pattern was similar to that in WT cells in all the stages examined (Fig. 4B). Nevertheless, in P7.5 samples, the overall intensity of cytosolic p-S6RP immunosignals significantly decreased in mutant cells (Fig. 4B,H), and also the number of p-S6RP+ cells tended to decrease in them (Fig. 4E,F). Collectively, these results suggest that mTORC1 activity is downregulated in Camsap3-mutated ependymal cells at postnatal periods, although its overall activity in neocortical lysates did not differ between WT and mutants (Fig. 4D).
As an alternative mechanism that may reduce the apical domain size in ependymal cells, we considered the possibility that the apical domain had actively constricted, a phenomenon that occurs widely during epithelial morphogenesis (Takeichi, 2014). Such constriction is generally executed by activation of myosin II, which is associated with the adherens junctions (AJs)–F-actin complex (Takeichi, 2014). However, immunostaining for an activated (phosphorylated) form of myosin regulatory light chain 2 did not show any difference in its distribution at the apical domains or junctions between WT and mutant cells at least in P28.5 brains (Fig. S5C), suggesting that this possibility is unlikely.
Reduction of lysosomes at apical regions of Camsap3 mutant ependymal cells
mTORC1 is activated on lysosomes (Betz and Hall, 2013; Sancak et al., 2010) (Fig. 5A), and we confirmed that mTORC1 is closely associated with lysosomes in ependymal cells (Fig. 5B). Previous work showed that lysosomal positioning is crucial for mTORC1 activation. For example, forced displacement of lysosomes from the plasma membranes was sufficient to downregulate mTORC1 signaling (Korolchuk et al., 2011), as signaling modules upstream of mTORC1, including Akt, accumulate beneath the plasma membranes (Finlay and Cantrell, 2011; Gao et al., 2014). In ependymal cells at P7.5, an activated form of Akt (phospho-Akt) was detected as punctate immunosignals not only at the centriole amplification stage but also at the multiple BB stage in both WT and Camsap3 mutant neocortices (Fig. 5C,D). By shifting focus planes, we confirmed that phospho-Akt was most abundant around the level of AJs, which are located at subapical regions of the cell (Fig. 5E,F). We also examined other signaling pathways upstream of mTORC1, such as ERK and AMPK, and found that their active forms were rarely or only faintly detected at the corresponding regions of these cells in both genotypes at P7.5.
As we did not find any difference in phospho-Akt distribution between WT and Camsap3 mutant neocortices, we considered the possibility that Akt might not be able to effectively activate mTORC1 in the mutants. To test this idea, we examined the distribution of lysosomes, by which mTORC1 is carried, using the lysosomal marker Lamp2 in ependymal cells at P7.5. The results showed that Lamp2+ structures tended to decrease in the apical portions of mutant cells (Fig. 5G-I). These observations suggest that mTORC1 became less accessible to Akt in the mutant ependymal cells, which may explain why mTORC1 activity was reduced in them.
Depletion of apical MT networks in Camsap3 mutant ependymal cells
Lysosomes are transported along MTs (Bonifacino and Neefjes, 2017; Pu et al., 2016). Therefore, we investigated whether MT networks showed any changes due to CAMSAP3 dysfunction in neocortical ependymal cells. We first observed CAMSAP3 distribution at various developmental stages by immunostaining, finding that its staining signals were relatively weak at E18.5, but increased by P7.5, persisting to P28.5 (Fig. S6A). Interestingly, the upregulation of CAMSAP3 selectively occurred in cells with either amplifying centrosomes or multiple BBs (Fig. 6A,B), suggesting some functional relations between this upregulation and mTORC1 activation that also occurred during centriole amplification, although the upregulation of CAMSAP3 persisted to the multiple BB stage unlike mTORC1. CAMSAP3 was concentrated at an apical portion of each cell, more precisely, at a level more apical than N-cadherin+ AJs that are present at a subapical region of lateral cell membranes (Fig. 6A,B). Comparison of CAMSAP3 and γ-tubulin localizations showed that, although γ-tubulin+ centrosomes or BBs were clustered at a particular region of the cytoplasm, CAMSAP3 was scattered throughout the apical cytoplasm, being located more apical than the cluster of BBs (Fig. 6A; Fig. S6B). Mutated CAMSAP3 proteins were also concentrated at the apicalmost domain of cells with amplifying centrosomes or multiple BBs at P7.5 and later (Fig. 6A; Fig. S6A), suggesting that the ability of CAMSAP3 to bind MTs is not involved in its localization observed here.
We then observed MT networks at the levels where N-cadherin-bearing AJs and BBs can be focused on, using anti-tyrosinated tubulin antibodies that react with cytoplasmic MTs, but less so with ciliary MTs that are generally stable. This allowed us to detect the former MTs unambiguously. In WT ependymal cells with a monomeric centrosome, MTs radiated from it (Fig. S6C). However, in cells undergoing centriole amplification, MTs not directly associated with BBs increased. In those with multiple BBs, MTs became organized in a meshwork over the BB and non-BB zones (Fig. S6C). This MT meshwork was oriented in parallel to the apical cortex, unlike the perpendicular arrangement of MTs observed in intestinal epithelial cells (Toya et al., 2016). In Camsap3 mutant cells, MTs radiating from a centrosome or a small cluster of BBs were also observed. However, in those with multiple BBs, the MT meshwork tended to be less developed at the non-BB zone (Fig. S6C).
Next, we closely analyzed MT organization in P7.5 WT ependymal cells along the apicobasal axis by serial optical sectioning of the images. At the apicalmost levels of WT ependymal cells, where CAMSAP3 was concentrated, MT filaments were detected, showing their association with CAMSAP3 puncta (Fig. 6C-E), as also confirmed by coronal sections of the cortex (Fig. 6F). As expected, these MTs were located above the MT population associated with γ-tubulin+ BBs (Fig. S6D). This apicalmost population of MTs was, however, depleted in Camsp3 mutants (Fig. 6C-F; Fig. S6D), although MTs were normally detectable below the CAMSAP3-localizing zones, where BBs are present. These observations suggest that CAMSAP3 produces apicalmost MT networks in ependymal cells and the lack of functional CAMSAP3 causes their depletion (Fig. 6G).
In addition, we observed actin networks that are known to contribute to special and functional organization of BBs in multiciliated cells (Mahuzier et al., 2018; Werner et al., 2011). In multiciliated ependymal cells of P7.5 WT brains, we detected actin filaments interspacing BBs, and similar actin networks were detectable in mutant cells (Fig. S7), suggesting that actin assembly takes place without notable defects in Camsap3 mutants.
We also examined whether the PCP of ependymal cells, which is characterized by the biased localization of a BB cluster within individual cells, was affected by Camsap3 mutation. Each of the multiciliated ependymal cells has two aspects of PCP, ‘translational’ and ‘rotational’ polarity (Mirzadeh et al., 2010). We analyzed the translational polarity in individual cells creating two parameters (Fig. S6E,F), and found that BB clusters were anteriorly positioned in both genotypes, rather more strictly in mutant cells (Fig. S6E). The extent of BB displacement was also indistinguishable between them (Fig. S6F). These observations suggest that translational polarity was normally established in Camsap3-mutated ependymal cells. Finally, we examined the motility of cilia, the coordinated motion of which in a multiciliated cell is generally disturbed by defects in rotational polarity, by live imaging. Our results showed no detectable difference in their beating pattern between WT (Movie 1) and mutant (Movie 2) ependymal cells.
Our present findings demonstrate that the CAMSAP3–non-centrosomal MT system is important for morphogenesis of the lateral ventricles. Loss of it brought about narrowing as well as increased stenosis or fusion of the ventricles. We sought primary causes of this ventricular deformation, finding that ependymal cells at the neocortex failed to undergo normal expansion of their apical domain, a final process of their differentiation. This failure apparently interfered with normal extension of the neocortical VSs along the mediolateral axis, and, in turn, widening of the ventricular cavity. These findings are consistent with previous observations that narrower ventricles are more prone to be stenosed under normal and pathological conditions (Ibañez-Tallon et al., 2004; Lowery and Sive, 2009; Shook et al., 2012).
CAMSAP3 proteins were concentrated at the apicalmost domain of ependymal cells, as observed in intestinal epithelial cells (Toya et al., 2016), suggesting that these different cell types share a common mechanism to place CAMSAP3 at particular subcellular sites. However, unlike in intestinal epithelial cells, the C-terminal domain-truncated CAMSAP3, expressed in the mutant mice used here, also accumulated at similar sites in the case of ependymal cells, which suggests that non-functional CAMSAP3 behaves differently between these cells. At the subcellular regions in which CAMSAP3 accumulated, MTs organized into networks that are oriented in parallel to the apical plasma membranes. In the absence of functional CAMSAP3, such MT networks were diminished. These observations suggest that CAMSAP3 serves to organize MT networks that are located at the apicalmost regions of the cells and arranged in a horizontal orientation, although it remains to be determined whether it also regulates apicobasally oriented MTs as found in the intestines (Toya et al., 2016). On the other hand, BBs, which are another set of subcellular structures that associate with MTs, were located below the CAMSAP3-disributuing zone, and MTs present at the BB zones remained unchanged in Camsap3 mutant brains.
A recent study showed that CAMSAP3 associates with BBs in a one-to-one fashion in multiciliated nasal epithelial cells (Robinson et al., 2020), but this kind of spatial organization of CAMSAP3 and BBs was not observed in ependymal cells. We also did not find any abnormality of ciliary movement nor cytological PCP in Camsap3-mutated ependymal cells. These findings suggest that CAMSAP3 differently functions in multiciliated cells derived from different organs.
We showed that mTORC1 signals became upregulated during ependymal differentiation and this process was important for their further growth. Our results suggest that the CAMSAP3-mediated MT networks likely support this process. It is known that mTORC1 activation occurs on lysosomes (Betz and Hall, 2013; Sancak et al., 2010), and, in Camsap3 mutants, lysosomes were less accumulated at the apical regions of ependymal cells, coinciding with the reduction of mTORC1 activity. Given that phospho-Akt, a kinase working upstream of mTORC1, was enriched in the apical portions of ependymal cells, these observations suggest that the lysosomal mispositioning observed in mutant cells ought to have reduced mTORC1 that is accessible to Akt, resulting in the lesser activation of mTORC1, as was shown in a previous study (Korolchuk et al., 2011). Thus, we propose a scenario that CAMSAP3 maintains a population of MTs which serves for redistribution of components required for mTORC1 signaling, such as lysosomes, and loss of this mechanism results in reduced mTORC1 activation, and in turn a failure of ependymal cell growth and the resultant narrowing of the lateral ventricle. One important question remains to be answered: why was VS growth less affected along the dorso-ventral axis of the brain? Alternative or additional mechanisms might be involved in VS extension at this side of ventricles, but cytological changes due to stenosis and fusion in the mutant ventricles prevented us from pursuing this question further.
Ventricle closure and resultant loss of ependymal cells were accompanied by depletion of adult NSCs at the striatal V-SVZ, confirming that the normal configuration of ependymal layers is important for maintaining NSCs. The depletion of NSCs was possibly due to loss of direct access to the CSF or of ependymal supports for their maintenance.
The ventricle occlusion and resultant prevention of CSF circulation are thought to cause hydrocephalus (Kahle et al., 2016; Kousi and Katsanis, 2016). Our mutant mice, however, did not show any signs of hydrocephalus despite the abnormal closure of the lateral ventricle. It is likely that the ventricular defects observed in Camsap3 mutant brains still allowed CSF circulation, saving them from hydrocephalus. We also should emphasize that ciliary beating, required for CSF circulation (Banizs et al., 2005; Ibañez-Tallon et al., 2004), appeared to be normal in the mutant ependymal cells.
We unexpectedly discovered that apical accumulation of CAMSAP3 coincided with the onset of ependymal growth, which requires activation of mTORC1. Ependymal cells likely have a developmental program for coordinating these independent cellular systems to work together, in order to attain their maturation. Elucidating the molecular basis of such a program would provide deeper insights into the mechanisms of how cells orchestrate multiple cellular machineries for their polarization, differentiation and growth.
MATERIALS AND METHODS
A Camsap3 (Nezha) mutant mouse line, Camsap3dc/dc, has been previously reported (Toya et al., 2016). Briefly, the genomic sequence that encompasses the 14th and 17th exons was flanked by loxP sites, and deleted by crossing the floxed mice with the β-actin Cre transgenic ones. The resultant mice were simply designated as Camsap3 mutant mice in this work. The mutant mice were analyzed after heterozygous mice had been backcrossed for at least four generations to C57BL/6N. Noon of the day on which the vaginal plug was detected was designated as E0.5, and E19.5 was defined as P0.5. For all experiments, male or female mice were used. The experiments using mice were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee of RIKEN Kobe Branch.
5-Bromo-2′-deoxyuridine (BrdU, Sigma-Aldrich) was dissolved in drinking water at a final concentration of 0.8 mg/ml and given to P28.5 mice for seven consecutive days.
Rapamycin (Merck) was dissolved in ethanol and diluted in water containing 5% (v/v) PEG400 (Nacalai Tesque) and 5% (v/v) Tween80 (Nacalai Tesque) at a final concentration of 40 µg/ml. P6 mice were intra-peritoneally injected with either rapamycin at 1 mg/kg body weight or vehicle only once a day for eight consecutive days.
Fixation and sectioning
Embryos were dissected and perfused with 4% (w/v) paraformaldehyde (PFA) in phosphate-buffered saline (PBS; 0.1 M, pH 7.4) prewarmed at 37°C. Postnatal mice were anesthetized with sodium pentobarbital (100-200 mg/kg body weight; Abbott) and transcardially perfused with 1% (w/v) PFA in 37°C PBS, followed by 4% (w/v) PFA in 37°C PBS (Miller, 1981). For CAMSAP3, α-tubulin or tyrosinated tubulin immunostaining, PHM buffer (60 mM Pipes, 25 mM Hepes, 2 mM MgCl2) (Schliwa and Van Blerkom, 1981; Tanaka et al., 2012) was used instead of PBS. Brains were removed, postfixed in the same fixative for 2 h at room temperature (RT) and subsequently overnight at 4°C, and cryoprotected by immersion in 10-30% (w/v) sucrose series in PBS. Fixed brains were embedded in OCT compound (Sakura Finetek), quickly frozen on isopentane cooled with liquid nitrogen and cut coronally at 20 µm using a cryostat (HM500M, MICROM). For en face imaging of the VS of the neocortex or the striatum, a single coronal slice was cut at the level of the anterior edge of the callosal commissure at 600 µm for E18.5 brains, 800 µm for P7.5 brains and 1 mm for P14.5, P28.5 and P49.5 brains using a microslicer (DTK-3000W, Dosaka EM). Usually, the posterior cut surface of the slice contained the midline crossing of the anterior commissure. Then, a piece containing the V-SVZ of the neocortex or of the striatum was dissected from the slice, embedded in OCT compound, frozen on isopentane cooled with liquid nitrogen and cut in parallel to the VS at 10 µm using the cryostat. Alternatively, the dissected V-SVZ was used as a wholemount.
Immunofluorescence staining and microscopical imaging
Antigen retrieval was performed by incubating sections in a citrate buffer (10 mM citrate, 0.05% Tween-20, pH 6.0) for 10 min at 98°C heated with a microwave (MI-77, Azumaya) or in HistoVT one (Nacalai Tesque) for 20 min at 70°C. For BrdU detection, sections were incubated with 2 M HCl for 30 min at RT. After two washes with PBS, sections were blocked in PBS with 0.3% (v/v) Triton X-100 (PBSX) containing 10% (v/v) normal horse serum (NHS) for 1 h at RT followed by incubation with primary antibodies diluted in PBSX containing 5% NHS overnight at 4°C. Sections were washed twice with PBS and incubated for 2 h at RT with corresponding secondary antibodies and subsequently incubated for 30 min at RT with DAPI (Sigma-Aldrich). Stained sections were washed twice with PBS and mounted with either Fluorsave (Chemicon) or 2,2'-thiodiethanol (Harvard Center for Biological Imaging, hcbi.fas.harvard.edu/tips-and-tricks). For staining F-actin, phalloidin-CF488 (Biotium) was mixed with the secondary antibodies and sections were mounted with 3,3′-thiodipropanol (Tejedo et al., 2019). Whole-mount immunostaining was performed essentially in the same way as described above, except that no antigen retrieval was performed. Stacks of images were taken along the z-axis at optimal intervals using a confocal laser scanning microscope (LSM780, Zeiss) with 20×/0.80 NA and 40×/1.30 NA objective lenses or using Airyscan (LSM880, Zeiss) with 63×/1.40 NA and 100×/1.46 NA objective lenses. Acquired images were processed using Zen (Zeiss) and Photoshop CS5 (Adobe Systems).
Antibodies for immunofluorescence
The rabbit antibody against CAMSAP3 was generated as previously described (Tanaka et al., 2012) (1:500) and the rat antibody against N-cadherin, MNCD2, has been previously generated (Matsunami and Takeichi, 1995) (1:100). The following primary antibodies were purchased: mouse anti-α-tubulin (Sigma-Aldrich, T9026, 1:1000), rabbit anti-β-catenin (Sigma-Aldrich, C2206, 1:400), mouse anti-β-catenin (BD Transduction Laboratories, 610153, 1:500), rat anti-BrdU (Bio-Rad, OBT0030, 1:200), rabbit anti-calbindin (Millipore, AB1778, 1:250), rabbit anti-calretinin (Millipore, AB5054, 1:500), rabbit anti-cleaved caspase 3 (Cell Signaling Technology, 9661, 1:400), mouse anti-γ-tubulin (Sigma-Aldrich, T6557, 1:1000), rabbit anti-γ-tubulin (Sigma-Aldrich, T3559, 1:500), rat anti-Ki67 (Thermo Fisher Scientific, 14-5698-82, 1:250), rat anti-Lamp2 (Abcam, ab13524, 1:200), rabbit anti-mTOR (Cell Signaling Technology, 2983, 1:100), mouse anti-NeuN (Merck, MAB377, 1:200), mouse anti-parvalbumin (Sigma-Aldrich, P3088, 1:500), rabbit anti-phospho-Akt (Ser473) (Cell Signaling Technology, 4060, 1:100), rabbit phospho-histone H3 (Ser10) (Upstate Biotechnology, 06-570, 1:500), rabbit anti-phospho-MLC2 (Thr18/Ser19) (Cell Signaling Technology, 3674, 1:100), rabbit anti-phospho-S6RP (Ser235/Ser236) (Cell Signaling Technology, 4858, 1:100), rabbit anti-S100β (Dako, IS504, 1:2), mouse anti-tyrosinated tubulin (Sigma-Aldrich, T9028, 1:400), rabbit anti-tyrosine hydroxylase (Merck, AB152, 1:250), rat anti-VCAM1 (BioLegend, 105701, 1:100) and mouse anti-ZO-1 (Invitrogen, 339100, 1:200). The following secondary antibodies were used: CF-488 conjugated anti-rat (Biotium, 20027, 1:500), CF-568 conjugated anti-mouse (Biotium, 20105, 1:500), CF-647 conjugated anti-rabbit (Biotium, 20047, 1:500), Alexa Fluor-488 conjugated anti-rat (Invitrogen, A-11006, 1:500), Alexa Fluor-568 conjugated anti-mouse (Invitrogen, A-11031, 1:500), and Alexa Fluor-647 conjugated anti-rabbit (Invitrogen, A-21245, 1:500).
Recording of cilia motility
P28 mice were deeply anesthetized with sodium pentobarbital and brains were dissected. Two consecutive coronal slices were cut at 450 µm at the level of the anterior edge of the callosal commissure using a vibratome (VT1000S; Leica) and placed on a slide. The slices were soaked with DMEM (Gibco) and a cover glass was placed over the sample. Cilia beating was recorded using the Olympus microscope BX53, to which a CCD camera (HAS-U2, Ditect) with a 40×/0.75 NA objective lens was attached, at 400 frames per second. The movies were made using ImageJ.
To quantify the extent of ventricle fusion and stenosis in coronal sections, sections that contained three structures – the callosal commissure, the striatum and the septum – were selected at 200 µm intervals. Typically, three sections were selected for each mouse. The combined length of the stenosed and fused regions were measured along the VS using Metamorph (version 7.7, Molecular Devices) and divided by the length of the striatal VS in each section. The calculated values were averaged for each mouse.
The length of the neocortical VS was measured essentially in the same way as described above. To normalize the difference in the hemisphere size among brains, we calculated the proportion of the neocortical VS to the mediolateral length of the hemisphere, which was defined by the line that started at the midline of the callosal commissure, extended in parallel to the mediolateral axis of the hemisphere and ended at the pia matter of the neocortex.
To quantify the extent of ventricle fusion and stenosis on the striatal wholemount, the striatal VS was classified into three regions based on the appearance of S100β signals, the area lacking S100β signals being regarded as the fused region, the area with brighter S100β signals as the stenosed region, and the remaining S100β+ area as the conventional region that faces to the open ventricle (see also Results section). The surface area of each region was measured using Metamorph and its proportion to the entire VS was calculated.
To quantify the PV signal intensities in ependymal cells on the striatal wholemount, boxed regions were put along a line that was drawn so that it crossed the boundaries between the stenosed and conventional areas. The signal intensity in each boxed region was measured using Metamorph.
To count the number of VCAM1+ clusters on the striatal VS wholemount, we excluded the one-third anterior part as well as one-third dorsal part of the striatal VS where weak VCAM1 signals could be detected in S100β+ ependymal cells. VCAM1 signals in the remaining part were binarized and the number of VCAM1+ profiles larger than 7 µm2 were automatically counted using Metamorph. The number of clusters per 1000 µm2 of the conventional region or of the conventional and stenosed regions was calculated.
To quantify the number of BrdU+ cells in the striatal V-SVZ, coronal sections were selected as described above. Typically, five or six sections were selected for each mouse. BrdU signals were binarized using Metamorph and BrdU+ profiles larger than 17 µm2, which was about half the size of the nucleus, were recognized as BrdU+ cells. These cells were manually counted in the V-SVZ, except for BrdU- and S100β double-positive cells, which are unlikely NSCs. The V-SVZ was defined as a high-cell density region along the striatal VS. The number of BrdU+ cells per 1 mm-length V-SVZ was calculated in each section. The calculated values were averaged for each mouse.
To quantify the number of proliferating cells in the striatal VZ of E18.5 mice, coronal sections were selected as described above. Ki67+ cells were automatically counted using ITCN plugin (imagej.nih.gov/ij/plugins/itcn.html) of ImageJ. Phospho-histone H3+ cells were manually counted. The VZ was defined as the parenchymal area within 20 µm from the striatal VS. The number of cells per 100 µm-length VZ was calculated in each section and averaged for each mouse.
To calculate the density of BrdU+ cells in the medial part of the GCL of the OB, coronal sections that were anterior to the accessory OB and contained the V-SVZ were selected at 200 µm intervals. Typically, three or four sections were selected for each mouse. The 400 µm-wide boxed region spanning the entire depth of the GCL was placed at the center of the line connecting the dorsal and ventral edges of the medial GCL using Metamorph. BrdU+ cells, which were defined as described above, were manually counted and NeuN+ cells were automatically counted using ITCN plugin of ImageJ. The number of BrdU+ cells was divided by the number of NeuN+ cells in each section. The same calculation was carried out for all selected sections and averaged for each mouse.
To estimate the total number of BrdU+ cells in the entire GCL in a single section, the area of the entire GCL was measured using Metamorph and divided by the area of the GCL in the boxed region (see above). The calculated value was multiplied by the number of BrdU+ cells in the boxed region. The estimation was carried out for all selected sections and summed for each mouse.
We manually counted BrdU+ cells expressing one of the following markers, calretinin, calbindin and tyrosine hydroxylase, in the medial part of the GCL or of the glomerular layer (GL). The medial part of the GL was defined the same way as that applied for the GCL. The count was summed across all selected sections in each mouse.
To measure the size of the telencephalon and the OB, the dorsal view of the brain was captured using a stereoscopic microscope (M420, Leica) attached to a CCD camera (DC500, Leica). The telencephalon and the OB were outlined and their areas were measured using Metamorph.
Apical domain size of a cell was measured using Metamorph by tracing cell boundaries visualized by N-cadherin or β-catenin immunofluorescence. To count the number of cells on the neocortical VS along the mediolateral axis, the 100 µm-wide boxed region that spans the entire mediolateral length of the VS was placed at the center of the apicobasal length of the neocortical VS. The number of cells in the boxed region was manually counted.
Phospho-S6RP or phospho-Akt signal intensities were measured in cells undergoing centriole amplification, which were recognized based on the appearance of γ-tubulin signals. Ten consecutive optical planes taken along the z-axis (namely the apicobasal axis) at 0.16 µm-intervals were selected for each cell so that the middle plane was placed at the plane where N-cadherin+ junctions were the best in focus. Cells on the VS sharply steep to the optical plane were excluded from the analysis. The signal intensity was measured in each optical plane using Metamorph and the highest value was recorded for each cell. The background fluorescence was separately measured and subtracted.
To quantify the proportion of phospho-S6RP+ cells in each cell type (cells with monomeric BB, amplifying centrioles or multiple BBs) and the proportion of each cell type in the phospho-S6RP+ cells, cells showing phospho-S6RP immunofluorescence signals of an intensity higher than 10 luminance (arbitrary unit in Metamorph) were recognized as phospho-S6RP+ cells. To quantify the phospho-Akt distribution, six optical planes taken along the z-axis at 0.48 µm-intervals were selected for each cell so that the middle plane was placed at the plane where N-cadherin+ junctions were the best in focus. The signal intensity was measured in each optical plane using Metamorph, which was normalized to the intensity at the plane of N-cadherin+ junctions. The distribution was calculated in each cell and averaged for each genotype.
To quantify Lamp2 signal intensities in the apical region of the cytoplasm, they were measured using Metamorph in three optical planes at 0.48 µm-intervals taken along the z-axis from the apicalmost of the cell. In the case of cells located on the VS steep to the optical plane, the inclination was normalized. In such cases, the apicalmost domain of the cell as well as the apical edge of β-catenin+ junctions were captured in the consecutive optical planes at 0.16 µm-intervals, each of which contained part of them (see Fig. 5G). To measure the Lamp2 intensity in the apicalmost domain of such cases, the region surrounded by visible segments of the apical edge of β-catenin+ junctions and by a line drawn to connect the ends of the segments was regarded as a part of the apicalmost domain in each plane. Lamp2 intensity at the parts of the apicalmost domain was separately measured and summed. The line(s) was kept unmoved in the following basal-side planes; the region surrounded by the line(s) and the lateral edge of β-catenin+ junctions and being 0.48 µm-beneath the apicalmost region was defined as the second-apicalmost region in each plane. The same method was applied to the remaining planes. The background fluorescence was eliminated before measurement using Zen. The measured intensities were divided by the total areas of the apical domain.
The PCP of ependymal cells was analyzed using ImageJ. A line was drawn from the centroid of an ependymal cell to the centroid of the BB cluster, and the angle between the line and the anterior axis was measured: 0° was set at the angle parallel to the anterior axis and the angle increased clockwise. The distance between the centroid of the ependymal cell and the centroid of the BB cluster was measured and divided by the distance of the line extended to the cell edge.
P7.5 neocortical tissues were suspended in ice-cold lysis buffer [20 mM Tris-HCl (pH 7.5), 1 mM EGTA, 1 mM EDTA, 50 mM NaCl, 1% (v/v) glycerol, 1% (v/v) Triton X-100] containing 0.25 mM NaF, 1 mM PMSF, 1 mM NaVO3 and a proteinase inhibitor cocktail (Roche) and lysed with a sonicator (UR-20P, TOMY). After centrifugation at 20,000 g at 4°C for 10 min, the lysate was added with 2-fold concentrated sample buffer [125 mM Tris-HCl (pH 6.8), 20% (v/v) glycerol, 2% SDS, 0.006% Pyronin Y, 1% 2-mercaptoethanol], separated in a 5-20% (w/v) gradient SDS/PAGE gel (SuperSep Ace, FUJIFILM) and transferred to a PVDF membrane (Millipore). Blots were blocked with Tris-buffered saline (TBS) with 0.05% (v/v) Tween-20 (TBST) containing 5% (w/v) skim-milk for 1 h at RT followed by incubation with primary antibodies diluted in TBST containing 3% bovine albumin overnight at 4°C. Blots were washed three times with TBST and incubated for 1 h at RT with peroxidase-tagged secondary antibodies (Amersham, NA934 and NA931, 1:1000). Chemiluminescence was obtained using Novex ECL Chemiluminescent Substrate Reagent Kit (Invitrogen) and detected with a CCD imager (LAS-4000, GE Healthcare). The following primary antibodies were used: rabbit anti-mTOR (Cell Signaling Technology, 2983, 1:1000), rabbit anti-phospho-mTOR (Ser2448) (Cell Signaling Technology, 5536, 1:1000) and mouse anti-GAPDH (Merck, G8795, 1:1000). Scanned images were quantified using Metamorph after monochrome inversion.
Data are presented as scatter plot with mean±s.e.m. except for phospho-Akt distribution, which was presented as mean±s.d. Statistical significance was determined by the unpaired two-tailed Student's t-test or Watson's U2 test.
We thank Yoko Inoue for technical support and the RIKEN Kobe light microscopy facility for imaging experiments. We are also grateful to Hiroshi Hamada for his support.
Conceptualization: T.K., M.T.; Formal analysis: T.K.; Investigation: T.K.; Resources: H.S., M.K.; Writing - original draft: T.K.; Writing - review & editing: T.K., M.T.; Supervision: M.T.; Funding acquisition: M.T.
This work was supported by the program Grant-in-Aid for Scientific Research (S) (grant number 25221104) from the Japan Society for the Promotion of Science to M.T.
Peer review history
The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.195073.reviewer-comments.pdf
The authors declare no competing or financial interests.