Basolateral rather than apical primary cilia on neuroepithelial cells committed to delamination

Delamination of epithelial cells is a fundamental process in tissue morphogenesis during development. A striking example of delamination is the development of the mammalian cerebral cortex, which originates from the dorsolateral telencephalic neuroepithelium. The primary stem and progenitor cells of this tissue, the neuroepithelial cells (NECs), exhibit apical-basal polarity, a central parameter in the delamination process (Götz and Huttner, 2005). Notably, these cells remain in contact with the basal lamina throughout the cell-cycle via their basal plasma membrane, are anchored to each other by an adherens junction (AJ) belt at the apical-most end of their lateral plasma membrane and possess an apical plasma membrane that constitutes the ventricular surface (Farkas and Huttner, 2008). As in other epithelial cells with apical-basal polarity (Weisz and Rodriguez-Boulan, 2009), a primary cilium (from here onwards referred to simply as ‘cilium’) emanates from the apical plasma membrane of NECs (Cohen and Meininger, 1987; Dubreuil et al., 2007; Farkas and Huttner, 2008; Hinds and Ruffett, 1971; Sotelo and Trujillo-Cenoz, 1958).

In the course of centrosome duplication during the cell cycle, the cilium’s basal-body and its associated second centriole each become the template centriole of the two centrosomes that during M-phase form the poles of the mitotic spindle (Bettencourt-Dias and Glover, 2007; Nigg and Raff, 2009). Consequently, given that the cilium persists during the cell cycle until the onset of M-phase, its apical location in NECs is considered to be the main reason why their nuclei undergo mitosis at the ventricular surface (Taverna and Huttner, 2010). This in turn is why NECs and the highly related radial glial cells into which they transform with the onset of neurogenesis (Kriegstein and Alvarez-Buylla, 2009) have been referred to collectively as apical progenitors (APs) (Farkas and Huttner, 2008; Taverna and Huttner, 2010). APs span the entire cortical wall, with their nuclei undergoing interkinetic nuclear migration (Taverna and Huttner, 2010) in the ventricular zone (VZ).

During cerebral cortex development, AP divisions are symmetric initially, with each division generating two daughter APs, both of which exhibit apical-basal polarity and re-establish a cilium at their apical plasma membrane after mitosis, utilizing the centrosome inherited from the mother AP (Farkas and Huttner, 2008; Kriegstein and Alvarez-Buylla, 2009). With the onset of cortical neurogenesis, an increasing proportion of APs switch to asymmetric divisions. These result in AP self-renewal and the generation of either a neuron destined for the preplate/cortical plate or, more frequently, a distinct type of progenitor, referred to as basal progenitor (BP) or intermediate progenitor cell, destined for the subventricular zone (SVZ) (Götz and Huttner, 2005; Kriegstein and Alvarez-Buylla, 2009). At birth, depending on the orientation of the cleavage plane relative to the apical plasma membrane and AJ belt, these neurons and BPs can be delaminated already, i.e. are no longer integrated into this belt and lack apical plasma membrane (Fietz and Huttner, 2010; Konno et al., 2008). Alternatively, the newborn neurons and BPs can initially be integrated into the AJ belt and possess apical plasma membrane, and subsequently delaminate from this belt and become devoid of apical plasma membrane (Konno et al., 2008).

Mature CNS neurons are known to carry a cilium on their perikaryal surface (Lee and Gleeson, 2010; Louvi and Grove, 2011). It is unknown, however, at which stage in the neuronal differentiation process the centrosome inherited from the neuronal progenitor cell is utilized for ciliogenesis (Han and Alvarez-Buylla, 2010). Similarly, assuming that BPs (like most cell types) carry a cilium, it is unknown when during the BP life cycle the centrosome inherited from the AP is utilized for ciliogenesis. In this context, an important question arising is whether ciliogenesis is in any way related to BP/neuron delamination.

The process of ciliogenesis includes two major steps, (1) membrane association of the centrosome’s older centriole, which becomes the cilium’s basal body, and (2) growth of the microtubule-based axonemes in concert with the evagination of the ciliary membrane (Marshall, 2008; Pazour and Bloodgood, 2008; Pedersen et al., 2008; Rohatgi and Snell, 2010; Rosenbaum and Witman, 2002; Santos and Reiter, 2008; Satir and Christensen, 2007; Sorokin, 1968). These steps can occur either at the plasma membrane or at intracellular membranes, such as Golgi-derived or endosomal membrane vesicles that subsequently fuse with the plasma membrane.

Given that the cilium has been shown to be a key organelle in cell signalling (Berbari et al., 2009; Goetz and Anderson, 2010; Lancaster and Gleeson, 2009), the cilium protruding from the apical plasma membrane of APs is in a strategic location to detect signals in the ventricular fluid (Han and Alvarez-Buylla, 2010). This option no longer exists for neurons and BPs after they have delaminated from the AJ belt and become devoid of apical plasma membrane. Key questions arising from these considerations are when those newborn neurons and BPs that are initially integrated into the AJ belt and possess apical plasma membrane re-establish a cilium, and on which membrane. Here, we show that such newborn BPs re-establish the cilium on their basolateral, rather than apical, plasma membrane. Re-establishing a basolateral (bl)-cilium therefore emerges as the first known cell biological feature that precedes BP delamination, a key aspect of cerebral cortex evolution.

Unless otherwise indicated, C57BL/6 mice were used. The transgenic mouse line expressing Tbr2-GFP was the Tg(Eomes::GFP) BAC transgenic strain described previously (Gong et al., 2003; Kwon and Hadjantonakis, 2007). All animal studies were conducted in accordance with German animal welfare legislation.

Fixed embryos at embryonic day (E) 9.5-17.5 were embedded in 4% low-melting agarose (ScienceServices) in 0.1 M phosphate buffer pH 7.4, followed by the preparation of transverse 200-μm-thick vibratome sections (VT1200S, Leica). Sections were fixed in 1% glutaraldehyde in 0.1 M phosphate buffer for at least 1 hour, post-fixed in 1% osmium tetroxide, contrasted with 1% uranyl acetate, dehydrated and flat-embedded in Epon replacement (Carl Roth).

Ultrathin (70 nm) plastic sections of dorsal telencephalon (dTel) and hindbrain were cut on a microtome (Leica, UCT) and post-stained with uranyl acetate and lead citrate according to standard protocols. Images were taken on a Morgagni EM at 80 kV (FEI) with a Morada or Veleta camera (Olympus) and ITEM software (Olympus).

Vibratome sections of E12.5-13.5 mouse embryos prepared as above were embedded in 12% gelatin in 0.1 M phosphate buffer. Gelatin-embedded tissue pieces containing dTel were infiltrated with 2.3 M sucrose for Tokuyasu cryosectioning (140 nm for immunofluorescence, 70 nm for EM). For both immunofluorescence and immuno-electron microscopy (immuno-EM), sections were immunolabelled as previously described for ultrathin cryosections (Dubreuil et al., 2007) using mouse mAbs against actin (Linaris, 1:10), pan-cadherin (Sigma, 1:50), β-catenin (BD, 1:50), polyglutamylated tubulin (Enzo, 1:100), rat mAb against integrin-α6 (Chemicon, 1:5); rabbit antibodies against Arl13b (ProteinTech, 1:100), β-catenin (Sigma, 1:200), Cep164 (ProSci, 1:20), GFP (Abcam, 1:150), Odf2 (Abcam, 1:10), Tbr2 (Abcam, 1:50) and γ-tubulin (Sigma, 1:200); and goat antibody against GFP (MPI-CBG, 1:150), followed by Cy3- or Cy2-conjugated goat anti-mouse, goat anti-rabbit, donkey anti-rabbit or donkey anti-goat secondary antibodies for immunofluorescence (Dianova, 1:400) and goat anti-mouse IgG/5-nm or 10-nm-gold (Dianova, 1:30) or proteinA/10-nm-gold (University Utrecht) for EM. The lipid GM1 was detected using Alexa488-coupled cholera toxin-subunit B (Sigma, 1:50) followed by a rabbit anti-Alexa488 antibody (Invitrogen, 1:500) and proteinA/10-nm-gold. Immunofluorescence sections were embedded in Mowiol 4-88 and viewed using a conventional fluorescence microscope [Olympus BX61 or Zeiss Axiovert 200M, both with Diagnostic Instrument cameras with IPLab (BD) or MetaMorph (Molecular Devices) software]. EM micrographs were taken as described above.

E11.5-13.5 Tbr2-GFP embryos were dissected and fixed in 4% paraformaldehyde in 0.1 M phosphate buffer, and 200-μm-thick vibratome sections were cut, permeabilized with 0.1% Triton X-100 and blocked with 0.5% bovine serum albumin (Sigma), 0.2% gelatin in PBS. Sections were incubated with rabbit anti-GFP antibodies (Abcam ab290, 1:1500) followed by goat anti-rabbit-IgG antibody coupled to ultrasmall gold (Aurion, The Netherlands, 1:100). Sections were fixed in 1% glutaraldehyde, processed for silver enhancement (R-Gent SE-EM, Aurion) and post-fixed in 1% osmium tetroxide followed by embedding as described above for conventional transmission EM.

Pre-embedding labelling for prominin-1 was performed as described (Dubreuil et al., 2007), either without modification or by replacing the protein A/10-nm-gold by ultrasmall gold and silver enhancement as described above. As the latter labelling was, in principle, the same as that used to detect Tbr2-GFP, which resulted in labelling throughout the cytoplasm, the lack of prominin-1 labelling on the bl-plasma membrane was not due to lack of penetration of the antibodies or detection reagents used.

Apical versus bl-cilia were quantified on ultrathin plastic sections that were oriented largely parallel to the apical-basal axis of ventricular cells. To avoid scoring the same cilium in different sections, only every ≥4th consecutive section was analysed. Quantification was confined to the apical-most ≈5 μm of the VZ. Only cells that, in the section analysed, either showed apical plasma membrane or at least formed apical AJs with neighbouring cells were used for quantification. We considered only cilia that showed a basal body in the section analysed and that protruded from the membrane for at least 400 nm, using the width of the basal body (≈200 nm) as the measuring unit, provided that they did not show the ‘chubby’ appearance of nascent cilia (Sorokin, 1962; Sotelo and Trujillo-Cenoz, 1958). Cilia were scored as apical if they, in the section analysed, either protruded into the ventricular lumen or were present in a membrane profile that was within 500 nm or less of the ventricular surface, which, hence, could be assumed to be obliquely sectioned apical plasma membrane. Cilia were scored as bl-type if they either protruded into the intercellular space or were present in a membrane profile that was basal to the AJ belt, which, hence, could be assumed to be obliquely sectioned bl-plasma membrane. Statistical significance was determined using an unpaired t-test (with Welch’s correction) in R (http://www.R-project.org/).

In utero electroporation of dTel of E12.5 embryos with pCAGGS-GFP (0.3 μg/μl) in combination with, or without, pCMV-SPORT6-Insm1 (1.5 μg/μl) was performed as described (Farkas et al., 2008).

The effect of Insm1 overexpression on the occurrence of bl-cilia was determined for GFP-positive cells 12 and 18 hours after electroporation. Using ultrathin, post-embedding GFP-immunogold-labelled frozen sections, the apical-most region of the neuroepithelium was analysed. As some of the sections analysed were consecutive and, therefore, the same single cilium could be scored in multiple sections, the cilia numbers of every third section were pooled and summed up, yielding three values, as follows: section numbers 1+4+7, etc.; section numbers 2+5+8, etc.; section numbers 3+6+9, etc. Section and embryo numbers were: control 12 hours: 20 sections from one embryo; Insm1 12 hours: 45 sections from two embryos; control 18 hours: 31 sections from three embryos; Insm1 18 hours: 55 sections from two embryos. Statistical significance was determined using an unpaired, one-tailed t-test. Sections of the same series were subjected to Tbr2 and GFP double immunofluorescence; each count was the sum of two to three sections.

The numbers of bl-cilia and spot-like AJs basal to the apical AJ belt (i.e. excluding the apical-most ≈2 μm region of the VZ) were counted by scanning β-catenin- and polyglutamylated tubulin-immunogold-labelled ultrathin frozen sections that contained at least one bl-cilium under the electron microscope. A bl-cilium and a spot-like AJ were scored as colocalized if the gap between these two structures (in the case of the bl-cilium, typically the basal body) was ≤500 nm; when including the size of the basal body and that of a spot-like AJ, such a bl-cilium/spot-like AJ assembly was found to be ≤1.2 μm.

For calculating the probability of colocalization of bl-cilia and spot-like AJs, the radial thickness of the VZ was measured on micrographs of the ultrathin frozen sections analysed. The probability of colocalization was calculated assuming a binominal distribution according to P=1–(1–size of assembly/length of lateral membrane)^{number of junctions per lateral membrane}.

A ‘ventricular cell’ (VC) is defined as a cell that contains apical plasma membrane and is integrated in the AJ belt. This term thus comprises neuroepithelial/radial glial cells carrying apical cilia and cells with ventricular contact that carry bl-cilia.

‘Basolateral-type cilia’ is the collective term for bl-cilia of VCs and cilia of delaminated cells; that is, cells that show neither apical plasma membrane nor contact with the apical AJ belt. The latter cilia comprise both cilia protruding into the intercellular space and cilia surrounded by membrane in the section analysed.

EM of the VZ of mouse embryonic day (E) 12.5-13.5 dTel revealed numerous primary cilia protruding from the apical plasma membrane of neuroepithelial cells (NECs) into the ventricular lumen (Fig. 1A), as described previously (Cohen and Meininger, 1987; Dubreuil et al., 2007; Hinds and Ruffett, 1971; Sotelo and Trujillo-Cenoz, 1958). Remarkably, however, some cilia also emerged from the basolateral (bl) plasma membrane of NECs (Fig. 1B-F). Importantly, these cells appeared to be unaltered with regard to other features of apical cell polarity, such as, exhibiting an apical plasma membrane and apical AJs (Fig. 1C,E). In light of the findings described below that such cells with bl-cilia lose apical polarity and, hence, neuroepithelial character, we shall from here onwards refer to these cells as ventricular (rather than neuroepithelial) cells (see note on terminology in Materials and methods).

EM analysis also provided evidence that the bl-plasma membrane was the final destination of bl-cilia. The two centrioles at the base of some bl-cilia had undergone duplication (Fig. 1E,F), indicating that the respective ventricular cell (VC) had progressed beyond the G1 phase of its cell cycle (Bettencourt-Dias and Glover, 2007; Nigg and Raff, 2009; Santos and Reiter, 2008; Seeley and Nachury, 2010). Ciliogenesis in cycling cells is known to take place early in G1. Hence, bl-cilia did not constitute an intermediate in apical ciliogenesis, being in transit to the ventricular surface, but were mature cilia.

Bl-cilia either, in the minority of cases, fully protruded from the bl-membrane into the intercellular space, occasionally being largely surrounded by bl-membrane of a neighbouring NEC (Fig. 1C,D), or, in the majority of cases, were embedded in a ciliary pocket with a small opening to the intercellular space (Fig. 1E; supplementary material Fig. S1A-F). The majority of bl-cilia pointed basally. The fine structure of bl-cilia resembled that of canonical cilia (including apical cilia of NECs), with a 9+0 microtubule doublet axoneme (Fig. 1G,H) extending from a basal body with nearby centriolar satellites (Fig. 1E). In addition, transitional fibres (also called alar sheets or distal appendages) and subdistal appendages, two structures that are thought to tether the cilium to the plasma membrane, and a necklace structure of the plasma membrane at the ciliary base were observed (Fig. 1D,F) (Cohen and Meininger, 1987; Dawe et al., 2007; Satir and Christensen, 2007). Similar to what has been described for apical cilia of NECs (Cohen and Meininger, 1987), a rootlet structure was rarely observed for bl-cilia. The length of the bl-cilia was relatively short (up to 3 μm) and in the same range as that of apical cilia of NECs (Cohen and Meininger, 1987). Together, these observations indicate that bl-cilia share major structural features with apical cilia.

We searched for intermediates in apical and bl-ciliogenesis. Nascent cilia were observed in intracellular vesicles located near, and probably destined for, the ventricular surface (supplementary material Fig. S1G-N), as well as on the apical plasma membrane (Fig. 2A), sometimes apparently just after fusion of an intracellular vesicle (supplementary material Fig. S1O-R). Nascent cilia were also observed on the bl-plasma membrane (Fig. 2B). Likely precursor structures to these were detected in newborn VCs (as indicated by the midbody still being present; Fig. 2C,D); namely, (1) basal bodies tethered directly to the plasma membrane (Fig. 2C) and, albeit less frequently, (2) small chubby cilia in intracellular vesicles that pointed basally and were found basal to AJs (Fig. 2D). Two distinct pathways of ciliogenesis have been described: (1) formation at an intracellular membrane vesicle followed by exocytosis and (2) formation directly at the plasma membrane (Pedersen et al., 2008; Rohatgi and Snell, 2010; Sorokin, 1962; Sorokin, 1968; Sotelo and Trujillo-Cenoz, 1958). Our data are consistent with apical ciliogenesis being initiated predominantly at intracellular vesicles and bl-ciliogenesis predominantly directly at the lateral cell surface.

We investigated whether the occurrence of bl-cilia reflected the loss of cell polarity and, perhaps, the redistribution of apical membrane constituents to the lateral plasma membrane. Immunogold EM showed that prominin 1 (also known as CD133), which in epithelial cells is a marker of an apical membrane-microdomain (Röper et al., 2000; Weigmann et al., 1997) and found on apical cilia (Dubreuil et al., 2007), was still localized at the apical surface of VCs carrying a bl-cilium (Fig. 2E-H). This apical prominin 1 was observed at the midbody of bl-cilia-forming cells that had just arisen from division (Fig. 2E,F), as is known to be the case for APs forming an apical cilium (Dubreuil et al., 2007), and on the ventricular surface proper of cells bearing a mature bl-cilium (Fig. 2G,H). Similarly, using cholera toxin binding, GM1, a lipid known to be enriched on the apical plasma membrane of epithelial cells (Janich and Corbeil, 2007), was largely detected on the ventricular surface and intracellular vesicles beneath it both in VCs with apical cilia (Fig. 2I) and in VCs with bl-cilia (Fig. 2J). Conversely, the bl-membrane protein α6-integrin (Haubst et al., 2006; Wodarz and Huttner, 2003) was still basolateral in both types of VCs (Fig. 2K,L). We conclude that VCs forming bl-cilia still exhibit apical-basal membrane polarity.

The switch from apical to bl-ciliogenesis implies a reduced interaction of the post-mitosis centrosome with apical membrane components, an increased interaction with bl-membrane components, or both. There has been significant progress in identifying the molecular players involved in the biogenesis of cilia (Ishikawa and Marshall, 2011; Pedersen et al., 2008; Santos and Reiter, 2008). These include proteins involved in trafficking of nascent cilia to the plasma membrane and in delivery of additional membrane to the growing cilium, as well as proteins associated with transitional fibres and subdistal appendages and, hence, implicated in the interaction of the centrosome with the membrane, such as Cep164 and Odf2 (Dawe et al., 2007; Emmer et al., 2010; Francis et al., 2011; Graser et al., 2007; Ishikawa et al., 2005; Li and Hu, 2011; Pazour and Bloodgood, 2008; Sfakianos et al., 2007). We investigated whether Odf2 and Cep164 are associated with the membrane-tethered basal body of neuroepithelial cilia, and of bl-cilia in particular. Immunogold EM revealed that this was indeed the case for both apical and bl-cilia, with immunoreactivity being observed on transitional fibres and, for Odf2 even more so, on the subdistal appendages (supplementary material Fig. S2). These observations suggest that at least some components of the machinery mediating the interaction of the basal body with the membrane are the same for apical and bl-cilia of VCs.

To obtain clues relating to the significance of bl-cilia, we investigated the frequency of their occurrence in the mouse dTel at various stages of neurogenesis (E9.5 to E17.5). Specifically, we determined by EM, in the apical-most ∼5-μm region of the VZ, whether mature cilia protruded from the apical surface of NECs (apical cilia) or belonged to the bl-type (see note on terminology in Materials and methods). The latter comprised, in the majority of cases, mature bl-cilia on VCs (although their localization at the cell surface was not corroborated by analysis of serial sections in every single case), with the remainder being cilia on cells that had already delaminated from the AJ belt. Bl-cilia were not observed at E9.5 (Fig. 3), i.e. before the onset of neurogenesis in the dTel (Götz and Huttner, 2005; Kriegstein and Alvarez-Buylla, 2009). Interestingly, bl-cilia appeared at the onset of neurogenesis (E10.5) and their proportion relative to apical cilia increased with the progression of neurogenesis, reaching a value of nearly 50% at E17.5 (Fig. 3). These observations suggested that the appearance of bl-cilia is somehow linked to neurogenesis.

During neurogenesis, APs, in addition to self-renewal, give rise to two further cell types: neurons and BPs (Fietz and Huttner, 2010; Götz and Huttner, 2005; Kriegstein and Alvarez-Buylla, 2009). In the dTel, the progeny of neurogenic APs are mostly BPs, whereas in the hindbrain almost all are neurons (Haubensak et al., 2004). Neurogenesis in the hindbrain starts earlier than in the dTel, and the increase in neuron-generating APs over time is less steep than in the cortex (Götz and Huttner, 2005; Haubensak et al., 2004). Consistent with a link between the appearance of bl-cilia and neurogenesis, we observed, in the apical-most ∼5-μm region of the hindbrain VZ, bl-type cilia already at E9.5, and the increase in their proportion from E10.5 to E13.5 was less steep than that in the dTel (Fig. 3). These data provided further support for the notion that the appearance of bl-cilia is linked to the generation of BPs and neurons from APs.

Newborn neurons and BPs initially might still be in contact with the ventricle and be integrated into the AJ belt, from which they subsequently delaminate (Fietz and Huttner, 2010; Konno et al., 2008; Kriegstein and Alvarez-Buylla, 2009). We therefore explored whether the cells with apical plasma membrane, apical AJs and a bl-cilium observed in the E12.5-13.5 dTel (Fig. 1) were nascent BPs. For this purpose, we made use of a mouse line expressing cytosolic GFP under the control of the promoter of Tbr2 (Gong et al., 2003; Kwon and Hadjantonakis, 2007), a transcription factor specifically expressed in BPs of dTel (Englund et al., 2005). Importantly, nascent BPs can initiate Tbr2 expression when they are still located at, or near, the ventricular surface (Kowalczyk et al., 2009).

Pre-embedding immunogold labelling for GFP followed by EM revealed immunoreactivity indicative of Tbr2 expression in the cytoplasm of mitotic cells located at the basal boundary of the VZ (mitotic BPs), but not of apical mitotic cells (mitotic APs), as expected (Fig. 4A-B′). Few non-mitotic cells with ventricular contact showed Tbr2-GFP immunoreactivity (Fig. 4C-E). Remarkably, in >95% of the cases in which a cilium was contained in the ultrathin section of such an immunolabelled VC, it was localized basolaterally (Fig. 4I). This indicated that nascent BPs, which still exhibit ventricular contact and integration into the AJ belt and thus have not yet undergone delamination, place their cilium on the basolateral rather than apical plasma membrane. This constitutes a major difference to the APs from which BPs arise, which carry apical cilia (Fig. 4D, inset). Furthermore, ∼90% of the VCs with bl-cilia were Tbr2-GFP-positive, although in some cases the level of immunoreactivity was low (Fig. 4C,J). This implies that a transcriptional programme characteristic of BPs is initiated by the time that a bl-cilium is formed in AP progeny.

In many (Fig. 4D,E; supplementary material Fig. S3), but not all (Fig. 4C), instances, the width between the apical AJs (as observed in ultrathin sections) was less in the Tbr2-GFP-positive, bl-cilia-carrying VCs than in the Tbr2-GFP-negative, apical cilia-carrying APs. This indicates that the formation of a bl-cilium in nascent, still ventricular, BPs precedes apical constriction, which in turn precedes their delamination. Accordingly, Tbr2-GFP-positive cells lacking ventricular contact exhibited cilia emerging from what corresponded to the previous bl-membrane (Fig. 4F,F′). These cells were BPs that had delaminated from the AJ belt but the nuclei of which were still located in the VZ. The cilia of these delaminated BPs were typically found on the retracting apical process rather than in the vicinity of the cell body (Fig. 4F). Similarly, cilia were also observed on Tbr2-GFP-positive BPs that had further delaminated in that their nuclei were located in the SVZ (Fig. 4G-G′). Here, the cilium was typically observed closer to (Fig. 4G′), or even on (data not shown), the cell body. In line with previous observations (Lee and Gleeson, 2010; Louvi and Grove, 2011), the newborn neurons that had inherited Tbr2-GFP from their BP mothers (Kowalczyk et al., 2009; Kwon and Hadjantonakis, 2007) and had migrated from the SVZ basally to form the developing cortical plate, showed cilia that, in most cases, emerged from a plasma membrane pocket close to their nucleus (Fig. 4H,H′).

Given that in the VZ of the dTel, cells with bl-cilia are predominately nascent BPs, we investigated whether conditions leading to increased delamination of BPs also increased bl-cilia. We addressed this issue by overexpression of the transcription factor insulinoma-associated 1 (Insm1), which promotes the generation of BPs from APs (Farkas et al., 2008) (Fig. 5A). Specifically, we determined the occurrence of apical and bl-cilia in control- (GFP only) and Insm1- (Insm1+GFP) transfected cells in the apical-most region of the VZ (Fig. 5B,C) 12 and 18 hours after in utero electroporation of the dTel of E12.5 embryos, using immunogold labelling of ultrathin cryosections for GFP to identify transfected VCs (Fig. 5D-G). Insm1 overexpression caused a significant increase in the proportion of VCs with bl-cilia (Fig. 5A). As most of these cells constituted daughter cells of transfected APs that re-established a cilium after mitosis, we conclude that establishing a cilium on the bl-plasma membrane precedes the increased delamination of VCs to become BPs in response to Insm1 overexpression.

What are the cues that direct bl-cilia and cilia of delaminated BPs and neurons to specific sites of the plasma membrane? We noticed that bl-cilia were not randomly distributed along the bl-plasma membrane, but that the majority was found just basal to the AJ belt (Fig. 1C,E,G). Furthermore, we observed that apical cilia (as judged from their basal bodies/centrioles) were largely centred with respect to the AJ belt (supplementary material Fig. S4G). These data raised the possibility that junctional components such as cadherins might be involved in cilia positioning. Of relevance for bl-cilia positioning, cadherins are not confined to apical AJs but are known to occur throughout the cortical wall (Kadowaki et al., 2007; Marthiens and ffrench-Constant, 2009). This reflects puncta adhaerentia (Hinds and Ruffett, 1971), which are spot-like AJs along the bl-plasma membrane of VCs and on the plasma membrane of delaminated BPs and neurons.

We therefore investigated a possible spatial relationship between bl-type cilia and spot-like AJs. Indeed, EM revealed bl-cilia of VCs, and cilia of delaminated BPs and neurons, in the vicinity of electron-dense plaques that resembled AJs (Fig. 6A-C′; supplementary material Fig. S4A-F). The identity of these junctions as spot-like AJs was corroborated by double immunofluorescence on ultrathin frozen sections for cadherin and the centrosomal marker γ-tubulin (Fig. 6D-D′), or for β-catenin and the ciliary membrane marker Arl13b (Fig. 6E-E′), by double-immunogold EM for β-catenin and polyglutamylated tubulin, another marker of cilia (Hammond et al., 2008; Sloboda, 2009) (Fig. 6H-H′), and by immunogold EM for cadherin (Fig. 6I-I′) and actin (Fig. 6J,J′). In accordance with these observations on embryonic mouse cerebral cortex, in the SVZ of foetal human cerebral cortex, centrosomes were found to be located close to spot-like AJs (supplementary material Fig. S4H-L).

We quantified and statistically tested the spatial relationship between AJs and bl-cilia (for an example, see supplementary material Fig. S5). Indeed, EM analysis of the VZ of E13.5 dTel revealed that >80% of the bl-cilia of cells that had probably delaminated from the apical AJ belt were found in the vicinity (within 500 nm) of spot-like AJs, which on average were six per cell (Fig. 7). This frequency of colocalization of bl-cilia and spot-like AJs was much greater than would be expected if bl-cilia were randomly distributed with respect to spot-like AJs, the probability in the latter case being 1 in 22 (Table 1). We conclude that the occurrence of bl-cilia in the vicinity of spot-like AJs is highly significant and suggests a role of spot-like AJs in the positioning of bl-cilia.

The present results uncover a novel cell biological feature of newborn NECs destined to change fate, that is, to delaminate from the apical AJ belt: the re-establishing of their primary cilium at the basolateral, rather than apical, plasma membrane (Fig. 8). Our finding that virtually all Tbr2-GFP-positive VCs (i.e. nascent BPs) exhibit a bl-cilium indicates that re-establishing the cilium at the bl-plasma membrane constitutes an obligatory step in their delamination process. This conclusion is not contradicted by the existence of a few Tbr2-GFP-negative VCs with bl-cilia, as these were probably either nascent BPs that had not yet started to express Tbr2-GFP to a detectable level, or nascent neurons (which do not express Tbr2) (Englund et al., 2005). The appearance of bl-cilia with the onset, and their increase in abundance with the progression, of neurogenesis are consistent with the notion that formation of a bl-cilium is a hallmark of delamination.

Interestingly, re-establishing of the cilium at the bl-plasma membrane preceded the actual delamination step, i.e. it occurred in cells that were still integrated into the AJ belt and exhibited apical-basal polarity, reflected, in particular, by the presence of apical plasma membrane (Fig. 8). Importantly, the increased delamination of BPs upon overexpression of Insm1 (Farkas et al., 2008) was preceded by increased formation of bl-cilia, providing strong support for a link between bl-ciliogenesis and BP delamination. A bl-cilium thus emerges as the earliest cell biological feature described to date that is observed specifically in those VCs that delaminate from the apical AJ belt to become BPs or neurons. Although the latter cells do not transgress the basal lamina but remain within the tissue corresponding to the original neuroepithelium, it will be interesting to investigate whether re-establishing the cilium at the basolateral rather than apical plasma membrane also occurs when cells delaminate from an epithelium, including transgression of the basal lamina; for example, in the formation of the neural crest from the neural plate (Acloque et al., 2009; Hay, 2005) or of the endocrine pancreas from the endoderm (Cole et al., 2009; Rukstalis and Habener, 2007). On a more general note, the question arises whether bl-ciliogenesis is a common feature of epithelial-mesenchymal transition and whether it could be involved in cancer metastasis.

Although the mechanism underlying bl-ciliogenesis remains to be investigated further, the statistically highly significant occurrence of bl-cilia in the vicinity of spot-like AJs suggests a role of the latter in the positioning of bl-cilia (Fig. 8; supplementary material Fig. S6). Such a role would be consistent with previous data obtained in other systems showing a spatial relationship between centrosomes and AJs (Bornens, 2008; Dupin et al., 2009). It might also be interesting to explore whether Lkb1 (Stk11 – Mouse Genome Informatics) might be involved in ciliary positioning in NECs as this protein kinase has been found to influence the location of the cilium with regard to plasma membrane domains, albeit in pancreatic β-cells, which do not exhibit proper apical-basal polarity (Granot et al., 2009).

The present observation of a bl-cilium in nascent BPs and neurons has intriguing implications for centrosome inheritance. In the course of centrosome duplication during the cell cycle, the cilium’s basal body and its associated second centriole, i.e. the older (‘first generation’) and younger (‘second generation’) centriole, respectively, dissociate from one another, with each ending up, together with the newly synthesized (‘third generation’) centrioles, in one of the two centrosomes (Bettencourt-Dias and Glover, 2007; Nigg and Raff, 2009) (Fig. 8). It has recently been shown that upon AP asymmetric division, the centrosome containing the first-generation centriole is preferentially found in the daughter cell remaining an AP and the centrosome containing the second-generation centriole in the progeny that delaminates and eventually becomes a neuron (Wang et al., 2009). This implies (at least in mouse) that the first-generation centriole becomes the basal body of the apical cilium of the daughter AP, and the second-generation centriole the basal body of the bl-cilium of the nascent daughter BP/neuron. Extrapolating from observations with 3T3 cells in culture (Anderson and Stearns, 2009), this in turn may imply that the daughter AP grows its cilium earlier than the daughter BP/neuron.

Given that cilia are key sensory organelles (Berbari et al., 2009; Goetz and Anderson, 2010; Lancaster and Gleeson, 2009), our finding of bl-cilia on subsequently delaminating BPs/neurons has intriguing implications with regard to signalling. Apical cilia are in a strategic location to sense signals specifically present in the ventricular fluid, whereas bl-cilia are not. Re-establishing the cilium at the bl-plasma membrane may thus deprive VCs of sensing a ventricular signal. Instead, bl-cilia might allow sensing signals originating from the lateral plasma membrane of neighbouring NECs or those present in the intercellular space. Such alteration in signalling could promote the cell fate changes associated with the transition from APs to BPs and neurons, including delamination. In addition, with regard to spatial constraints and cytoskeletal organization, re-establishing the cilium at the basolateral rather than apical plasma membrane might be an advantage, if not a prerequisite, for constriction of the apical-most region of a VC for subsequent delamination.

Importantly, the appearance of neural progenitors that delaminate from the apical AJ belt to form the SVZ, notably the outer SVZ, is thought to be a crucial step in the expansion of the cerebral cortex during mammalian, and in particular primate, evolution (Fietz et al., 2010; Hansen et al., 2010; Reillo et al., 2011). Thus, re-establishing the cilium at the basolateral rather than apical plasma membrane of a VC that retains mitotic activity might turn out to be a cellular biological step that is at the core of cortical expansion.

We thank Dr Robert Hevner for kindly providing the Tbr2-GFP mouse line; Drs Johannes Vogt (Charite, Berlin) and Simone Fietz for human foetal tissue; Katja Langenfeld for excellent technical assistance; Drs Yannis Kalaidzidis and Andreas Ettinger for help with statistics; Jussi Helppi and other members of the animal facility, and the electron microscopy facility, of the Max Planck Institute of Molecular Cell Biology and Genetics for excellent support; and Dr Caren Norden for helpful comments on the manuscript.