Mitotic spindle orientation distinguishes stem cell and terminal modes of neuron production in the early spinal cord

Spinal cord cells include motor neurons, sensory neurons and multiple classes of interneuron that together mediate movement and sensation of the body. The molecular mechanisms that specify these different kinds of neuron in the spinal cord have been well-characterized(Jessell, 2000; Price and Briscoe, 2004). There is also evidence for crosstalk between patterning factors that regulate cell type and genes that drive cell cycle exit and neuronal differentiation(Scardigli et al., 2003; Scardigli et al., 2001)(reviewed by Bertrand et al.,2002). Neuron production is itself regulated through the Notch signalling pathway; differentiating neurons deliver lateral inhibitory signals, such as Delta1, which ensure that neighbouring cells do not differentiate simultaneously (reviewed by Lewis, 1996). However, despite this wealth of understanding at the level of gene regulation, the cell behaviour underlying neuron production in the spinal cord remains poorly characterized.

The spinal cord is so well-studied because it is a relatively simple neuroepithelium. At early stages it comprises a bilaterally symmetrical tube each side of which is just one cell wide, with each cell contacting both the outer basal and inner apical (ventricular/luminal) surfaces. Within this tissue nuclei migrate between these surfaces and their position reflects progression through the cell cycle; mitosis occurs apically and nuclei move towards the basal surface during G1. After the nuclei arrive at the basal surface, cells then either re-enter S phase or exit the cell cycle and become neurons. This process, known as interkinetic nuclear migration(Frade, 2002), and subsequent neurogenesis have not been monitored in the spinal cord in vivo.

Neurons can be produced by asymmetric, stem cell mode divisions that produce a neuron and a progenitor or by symmetric terminal divisions that produce two neurons. Retroviral labelling studies indicate that both asymmetric and symmetric modes of neuron production underlie sensory neuron generation in the latestage spinal cord(Wildner et al., 2006). However, the first neurons to be born in the chick spinal cord are reticular spinal interneurons located in dorsal and intermediate regions, followed by ventral motor neurons (Sechrist and Bronner Fraser, 1991), and the modes of division that produce these crucial components of the spinal cord have not been determined.

Work in the mammalian cortex suggests that whether a division generates asymmetric or symmetric cell fates depends on the orientation of the mitotic spindle and hence cleavage plane position with respect to the apical surface of the neuroepithelium (Chenn and McConnell, 1995). A parallel cleavage plane places one daughter cell at the apical surface and the other more basally, whereas a perpendicular cleavage plane generates daughter cells that inherit equal portions of apical and basal membrane; mitotic spindle orientation may thus mediate cell-fate choice, as a parallel cleavage plane could lead to asymmetric segregation of determinants localized at apical or basal surfaces (reviewed by Gotz and Huttner, 2005). Chenn and McConnell propose that in the early cortex perpendicular cleavage plane divisions give rise to two progenitors, whereas parallel divisions generate a neuron and a progenitor (Chenn and McConnell, 1995). This and other studies(Kosodo et al., 2004) thus suggest a link between division orientation and acquisition of a neurogenic(neuron generating) ability.

Experimental evidence for the importance of mitotic spindle orientation for cell-fate choice in the mammalian cortex and retina has come from recent work in which its position has been manipulated by interfering with components of an apical protein complex, including Inscuteable and AGS3(Betschinger and Knoblich,2004; Sanada and Tsai,2005; Zigman et al.,2005). Disruption of this apical complex leads to an increase in perpendicular cleavage plane divisions and in symmetric cell fates. The relationship between cleavage plane orientation and cell-fate choice has not been addressed in the spinal cord, nor is it known whether this tissue, like the cortex, displays a transition from asymmetric to symmetric modes of neuron production as development proceeds. Neuron-generating divisions in the cortex are further distinguished by longer cell cycle times, because of extension of G1 phase (Calegari and Huttner,2003; Takahashi et al.,1995) (reviewed by Gotz and Huttner, 2005). It has been suggested that a longer G1 allows time for the asymmetric localization and accumulation of cell fate determinants. Although overall cell cycle time gets longer as development proceeds in many neuroepithelia, including the spinal cord(Kauffman, 1968), it has yet to be established whether this is a specific characteristic of all neuron-generating divisions.

Here we have developed a novel long-term time-lapse imaging assay to monitor single cells in the early chick spinal cord as they transit mitosis and then either divide again or undergo neuronal differentiation. Our study defines key parameters of cell behaviour in this tissue and surprisingly reveals that neurons are generated concurrently by symmetric and asymmetric division. Furthermore, we find that mitotic spindle orientation does not correlate with acquisition of neurogenic ability, but serves within neurogenic cells to distinguish modes of neuron production.

Embryos at Hamburger and Hamilton (HH)(Hamburger and Hamilton, 1951)stage 9-10 were electroporated using standard techniques(Itasaki et al., 1999), with a concentration of 0.125 μg/μl enhanced yellow fluorescent protein(eYFP)-α-tubulin (BD Sciences/Clontech) and/or Gap43-mRFP1(Poggi et al., 2005). Embryos were incubated at 38°C for 4 or 16 hours (giving slices for imaging at HH10 and 12). Slices of ∼150 μm were taken with a microknife, embedded in rat tail collagen type I in coverslip-based petri dishes (WillCo-dish glass-bottom dish, Intracel, Royston, UK; GWst-3522, coated with poly-L-lysine, Sigma) and cultured in Neurobasal medium, without phenol red(Gibco), supplemented with B-27 to a final 1× concentration with L-glutamine and gentomycin [after Placzek and Dale(Placzek and Dale, 1999)]maintained at 37°C with 5% CO₂/air for ∼4 hours before imaging. For fibroblast growth factor (FGF) treatment, media containing 200 ng/ml Human FGF4 (R&D Systems) replaced normal medium at the start of the experiment and was changed once (at 18 hours) approximately halfway through the imaging period.

Slices were imaged on a DeltaVision Spectris microscope workstation(Applied Precision, LLC, Issaquah, WA) in a Solent environmental chamber kept at 37°C. The chamber stage was buffered with 5% CO₂/95% air mix and maintained in a humid environment. Images were captured using a 40×1.35 NA objective lens with the Hg-arc lamp. Thirty optical sections (exposure time=50 milliseconds, 512×512 pixels, bin=2) spaced by 1 μm to 2.5μm were imaged at either 7-minute or 1.5-minute intervals, for up to 38 hours. The point-visiting function in the software allowed up to eight slices to be imaged during each experiment. Data was deconvolved and analyzed using constrained iterative deconvolution(McNally et al., 1999;Swedlow, 1997). At the end of time-lapse experiments, slice culture dishes were immediately washed in 37°C PBS and fixed in fresh 37°C 4%formaldehyde (FA) for 40 minutes to preserve microtubules/eYFP-α-tubulin, washed in PBS and 37°C PBS 0.5% Triton X-100, and subsequently labelled with antibodies against GFP and neuronal markers (see below).

Embryos at day 1.5 (HH10), day 2 (HH13), day 2.5 (HH16) and day 3 (HH18-19)were fixed in fresh 4% FA and subjected to in situ hybridization for NeuroM (Roztocil et al.,1997), a bHLH transcription factor that marks newly born neurons(Diez del Corral et al., 2002; Roztocil et al., 1997) (see below), using standard techniques. Sections through neural tube were then processed for immunocytochemistry for HuC, an RNA binding protein characteristic of differentiating neurons(Marusich et al., 1994) (see below), followed by staining with the chromatin marker DAPI.

Embryos at HH10 were exposed to 100 μl of 0.1 mM BrdU for 1 hour or 4 hours, fixed in 4% FA and processed and cryosectioned using standard techniques (three embryos per treatment, four sections from each embryo). Sections were treated using standard immunocytochemistry techniques,post-fixed in 4% FA for 10 minutes and subjected to a standard BrdU detection protocol (Gunhaga et al.,2000). Total progenitor cell population was determined by subtracting calculated proportion of differentiated neurons (3%, see Fig. S1 in the supplementary material) from total cell number, and simultaneous equations were used to calculate cell cycle time(Storey, 1989). Mitotic cells were determined by distinct morphology of DAPI-stained nuclei undergoing mitosis at the apical surface, and M-phase duration was calculated using cell cycle time and mitotic index (see Fig. S2 in the supplementary material).

Embryos at 1.5 and 3 days were fixed in FA, processed and cryosectioned using standard techniques. Sections through neural tube flanked by somites 1-3 were then processed for immunocytochemistry for cleaved Caspase-3 (see below)and stained with DAPI. Apoptotic cells were scored in four sections in three embryos for each timepoint, day 1.5 (1±0.6%) and day 3(0.2±0.05%) (s.d. is the variance between embryos). Images of cells in slice culture were analyzed for apoptosis (characteristic blebbing and eruption of cells) at 16 hours in time-lapse movies (15 slices, four experiments) (1.5±0.5%) (s.d. is the variance between experiments). The number of cells undergoing apoptosis and the total number of labelled cells visible at this time were recorded. This was used to calculate the proportion of cells dying at a fixed timepoint, which could then be compared with that observed in fixed tissue sections.

Embryos at day 1.5 and 3 were fixed in fresh 4% FA and stained with DAPI. The angle of the cleavage plane of cells in anaphase and telophase was measured relative to the apical surface of the neural tube in sections through neural tube flanked by somites 1-3 (in 10-12 sections from three to four separate embryos). Images were captured on a Leica DM IRB, 20× 0.3 NA lens with a Hamamatsu Snapper DIG16 camera. Data were analyzed onscreen using OpenLab (Improvision) software. In slice cultures the apical surface was identified by the position of the flattened end feet of cell processes, which form the lumen wall. When the lumen is not absolutely vertical, stereoscopic viewing of the images was used to determine the position of a dividing cell in relation to the apical surface. Visualization of the lumen was also assisted by brightfield images. To measure the angle of the cleavage plane relative to the apical surface, data was rendered into a three-dimensional view (using the volume viewing tool in Softworx software) and assigning the lumen (apical surface) to the horizontal plane. The image was then rotated around 360°in the y-axis to enable accurate measurement of the angle bisecting the two daughter cells at cytokinesis [similar to the rendering and angle measurements made in Haydar et al. (Haydar et al., 2003)].

Standard whole-mount in situ hybridization techniques were performed to detect NeuroM. Immunocytochemistry was also performed using standard procedures. Differentiated neurons were detected using mouse anti-HuC/HuD(1:200) (Molecular Probes, A-21271) or mouse anti-neurofilament-associated antigen 3A10 (1:50) (Developmental Studies Hybridoma Bank). Cleaved Caspase-3(Asp175) rabbit monoclonal antibody (Cell Signalling, 9664) was used to detect apoptotic cells, following the manufacturer's instructions. YFP signals in slice cultures were enhanced with a rabbit anti-GFP (1:100) (MBL, 598) after the imaging period. BrdU was detected with sheep anti-BrdU (1:100) (AbCam,ab1893). Secondary antibodies used were anti-mouse Alexa Fluor-488 (1:500)(Molecular Probes, A21121), anti-mouse Alexa Fluor-568 (1:200) (Molecular Probes, A-11031), or anti-mouse Rhodamine Red (1:200) (Jackson ImmunoResearch,715-295-150), anti-sheep FITC (1:200) (Jackson ImmunoResearch, 713-095-147),anti-rabbit biotin (1:1000) (Jackson ImmunoResearch, 111-065-144) with Cy3-conjugated streptavidin (1:4000) (Jackson ImmunoResearch, 016-160-084),and anti-rabbit Alexa Fluor-488 (1:200) (Molecular Probes, A-21206).

Statistics were performed using either Minitab or Sigma Stats. Statistical tests are justified in the Figure legends. In all descriptive statistics where sample means are used, variance is indicated by 1 s.d.

To elucidate cell behaviour underlying neurogenesis in the early spinal cord we used low-level electroporation to introduce a plasmid driving expression of eYFP-α-tubulin and/or GAP43-mRFP1 into the neural tube of day 1.5 chick embryos (stage HH10). Whole embryo slices were taken between the level of somites 1-6 and each slice embedded in collagen on a coverslip-based petri dish and maintained in serumfree medium(Fig. 1A). Using a low concentration of plasmid we labelled ∼5% of the cells located in dorsal and intermediate regions of the neural tube, allowing easy visualization of the microtubule cytoskeleton (eYFP-α-tubulin) and cell membrane(GAP43-mRFP1) of individual cells. Time-lapse, three-dimensional fluorescence images were recorded on a wide-field DeltaVision imaging system and subjected to image deconvolution (McNally et al.,1999; Swedlow, 1997). In all cases, we recorded three-dimensional data stacks spanning at least 30 μm at 7-minute intervals for up to 38 hours, unless otherwise stated (see Materials and methods). This imaging period corresponds to development from HH10/11 to ∼HH16, during which newly born neurons rise from 3 to 15% of cells in the neural tube (see Fig. S1 in the supplementary material).

To validate this approach we compared cell cycle time, length of mitosis and incidence of cell death in this region of the neural tube at day 1.5 in fixed embryos with that observed in neural tube in slices subjected to our imaging protocol. In fixed tissue, the proportion of cells incorporating BrdU at two separate timepoints indicated a cell cycle time of ∼16 hours (see Materials and methods and Fig. S2 in the supplementary material). In living slices, cell cycle times were directly determined from the length of time between cell divisions. These show striking heterogeneity, ranging from ∼9 to 28 hours (87 cells in 20 slices; see Fig. S2 in the supplementary material)but with an average (16 hours 10 minutes±4 hours) similar to that calculated in our fixed tissue and in the zebrafish hindbrain(Lyons et al., 2003). The duration of mitosis in slices (28±9 minutes) was similar to that calculated in the embryo (32±8 minutes), indicating that imaged cells are not delayed at cell cycle checkpoints (see Materials and methods and Fig. S2 in the supplementary material). Immunocytochemical detection of the cleaved form of Caspase-3, characteristic of apoptotic cells, also indicated that imaging does not increase the incidence of cell death: similar proportions of apoptotic cells were observed in sections of fixed neural tube at day 1.5(1±0.6%) and day 3 (0.2±0.05%) and in imaged slices at 16 hours(1.5±0.5%) (see Materials and methods). These findings indicate that cell behaviour in these tissue slices is comparable to that observed in the embryo and that this tissue slice assay can be used for detailed analysis of normal cell behaviour in the chick neural tube.

To monitor cellular dynamics during neurogenesis in the spinal cord, cells were labelled with GAP43-mRFP1 alone and/or YFP-α-tubulin. Labelling the cell membrane with GAP43-mRFP1 revealed that as cells round up on division at the apical surface they retain a thin membranous process that is inherited by one daughter cell in all divisions in which we were able to observe this structure (n=12 cells) (Fig. 1B). Following initiation of the cleavage furrow revealed with GAP43-mRFP1, a midbody identified with eYFP-α-tubulin forms between the two halves of the cell and then becomes displaced towards the apical surface(Fig. 1C, Fig. 2A; see Movie 1 in the supplementary material). This structure, a remnant of the mitotic central spindle (Glotzer, 2005; Nagele and Lee, 1979)consisting of a dense concentration of microtubules and many microtubule-associated proteins (Skop et al., 2004), relocated in all 25 cases examined to form the new apical process in each daughter cell (Fig. 1C, Fig. 2A; see Movie 1 in the supplementary material). The resumption of interkinetic movement following division is then accompanied by the movement of eYFP-α-tubulin into the basal process(Fig. 2B; see Movie 2 in the supplementary material). Once nuclei of both sibling cells have migrated to the basal side of the neuroepithelium, cells may then remain in this configuration in contact with both apical and basal surfaces for up to 15 hours (n=5 cells) (Fig. 2A; see Movie 2 in the supplementary material). If a cell is going to differentiate into a neuron, the endfoot of its apical process then releases from the apical surface and the process appears to lose tension. The apical process, as visualized with eYFP-α-tubulin, then gradually withdraws across the surface of neighboring cells until it disappears into the basally located cell body (Fig. 2A,B). A striking reorientation of the polarity of the cell is then manifest, such that it is shifted by 90° and the cell now extends processes along the dorsoventral, rather than mediolateral, axis of the neural tube (Fig. 2A,B; see Movie 2 in the supplementary material). Subsequent fixation and labelling demonstrated that such cells express neuronal markers (neurofilament or HuC; Fig. 2C) in 19/19 cases (six slices, three separate experiments), confirming that these novel and dramatic changes in cell morphology are indicative of neuron birth.

To determine the division modes underlying neuron production in the early neural tube we monitored the generation of individual lineages. Each lineage comprises at least one complete cell cycle, i.e. the generation of two daughter cells by mitosis at the apical surface and the subsequent re-entry into mitosis of at least one of these daughter cells and/or neuronal differentiation. Progenitor cells were identified by their completion of a second successful cell division, whereas cells that became neurons exhibited the striking stereotyped series of morphological changes described above. Of the 51 lineages monitored, 34 involved only progenitor-generating divisions(referred to hereafter as PP divisions)(Fig. 3; see Movie 3 in the supplementary material), whereas the remaining 17 lineages produced neurons(Table 1). Of these, seven exhibited an asymmetric mode of division, producing a neuron and a further progenitor (referred to hereafter as PN divisions)(Fig. 2B; see Movie 2 in the supplementary material), and another six gave rise to two neurons each(referred to hereafter as NN divisions)(Fig. 4; see Movie 4 in the supplementary material) (four more lineages produced a neuron and a sibling of undetermined fate). This indicates that asymmetric and symmetric modes of neuron production take place concurrently in the early spinal cord.

To examine the relationship between mitotic spindle/cleavage plane orientation and cell-fate choice we first wished to establish that similar proportions of cells divide in particular orientations with respect to the apical surface in tissues fixed immediately or monitored in vivo. Cleavage plane orientations were measured relative to mitotic spindle position in fixed/DAPI-stained sections of the neural tube at day 1.5 and 3 and in living tissue slices spanning this developmental interval (see Materials and methods). Histograms of these measurements after binning the data into 30°ranges (Fig. 5A,B) show that cells dividing with cleavage planes in all orientations were observed. We did not detect a preferential orientation at day 1.5 in fixed samples, whereas by day 3 we observed a slight preference for either parallel (28%) or perpendicular (58%) orientations (Fig. 5A). In living tissue, imaged from day 1.5 to day 3, cleavage planes were still more likely to be perpendicular to the apical surface and least likely to be parallel (Fig. 5B). These live data broadly correspond with the spread of division orientations observed in fixed tissue over this time period and we find no statistically significant difference between these data sets(Fig. 5A,B). We also made measurements of 20 cleavage plane angles in movies with data taken at 1.5-minute intervals of cells double-labelled with eYFP-α-tubulin and the cell membrane marker Gap43-mRFP1 to help visualize the cleavage furrow. These movies allowed us to confirm that our measurement of final cleavage plane position is accurate in vivo using data collected at 7-minute intervals(see Fig. 5C; see Movie 1 in the supplementary material).

Having established that the range of cleavage plane orientations observed in vivo is representative of normal cell behaviour, we investigated the relationship between cleavage plane orientation and the subsequent fates of daughter cells. This was analyzed in 47 lineages and in all cases we were able to identify all cells at all time points and determine the position of basal and apical surfaces (Table 1). Of the 34 cells undergoing PP divisions, the majority (21/34, 62%) divided with a perpendicular cleavage plane (Fig. 3; see Movie 3 in the supplementary material; Table 1). By contrast, all six cells that divided to give rise to two neurons exhibited a perpendicular cleavage plane (see Table 1)(Fig. 4; see Movie 4 in the supplementary material). We conclude that there is a good correlation between generation of symmetric neuronal fates and a perpendicular cleavage plane and that symmetric progenitor-generating divisions are not exclusively perpendicular, although they show a tendency to divide with this orientation.

Strikingly, in contrast to the NN divisions the seven PN-generating divisions had cleavage planes in any orientation. For example, Fig. 2B (see Movie 3 in the supplementary material) shows a lineage arising from a division that generated a neuron and a progenitor with a cleavage plane orientated 34° relative to the apical surface, followed by division of the daughter progenitor cell with a cleavage plane at 14° to the apical surface. However, another PN division had a cleavage plane at 75°(Fig. 6; see Movie 5 in the supplementary material). One explanation for this last division comes from observations in the mammalian cortex where only small differences about the perpendicular axis appear sufficient to asymmetrically apportion apical membrane between daughter cells (Kosodo et al., 2004). Indeed, we find that all but one of the NN divisions are oriented greater than 80° and are thus closer to perpendicular than the PN divisions (see dot plot of all cleavage plane orientations, Fig. 6B). Furthermore,statistical comparison of PN and NN division orientations indicates that these are distinct groups (see Table 1).

Importantly, knowing the cell fate outcome of divisions with specific cleavage plane orientations provides further insight into the role of mitotic spindle orientation in cell-fate choice. Our data indicate that many divisions with a parallel cleavage plane (6/9) give rise not to a neuron and a progenitor, but to two progenitors (Table 1). Simply positioning the mitotic spindle along the apico-basal axis (parallel cleavage plane) is therefore insufficient to generate asymmetric cell fates in this context. Furthermore, statistical comparison provides no evidence for a difference between PN and PP divisions that occur in all orientations, whereas NN division orientations are all perpendicular and are statistically different from PP as well as PN divisions(Table 1). Together, these data suggest that regulation of mitotic spindle position is only crucial within cells that have neuron-generating potential and, moreover, that assignment of this neurogenic ability is separable from division orientation.

Work in the developing mammalian cortex suggests that neuron-generating divisions are characterized by a longer cell cycle than progenitor-generating divisions and that this provides time for daughter cells to become different(Calegari et al., 2005; Calegari and Huttner, 2003; Takahashi et al., 1995). Here we use our direct visualization of mitosis and neuron formation to make these measurements for the first time in the spinal cord. We find that in the chick neural tube the cell cycle length of cells dividing to produce one neuron and one progenitor is longer than that for cells that generate only further progenitors (see Fig. 7A). These data support a role for cell cycle extension in the acquisition of the ability to generate neurons.

FGF is known to inhibit neuronal differentiation in many contexts,including chick neural tube explants taken from the same region as our slices(Diez del Corral et al.,2002). Here, FGF has been shown to inhibit expression of a bHLH transcription factor, NeuroM(Diez del Corral et al.,2002), which acts within the neurogenesis cascade that leads to neuronal differentiation (Roztocil et al.,1997). We therefore next assessed whether this loss of neuron-generating ability in response to FGF signalling is also accompanied by a decrease in cell cycle length. All 30 lineages analyzed in FGF-treated slices consisted of only progenitor-generating divisions. All of these lacked the apical process release and withdrawal behaviour that is characteristic of newly born neurons (Fig. 7B;see Movie 6 in the supplementary material). Exposure to FGF did not alter the range of cleavage plane orientations of divisions at the apical surface(Fig. 7C). This finding indicates that FGF does not attenuate neuron production by directing mitotic spindle position. Importantly, the cell cycle length of cells in neuroepithelium exposed to FGF was similar to that of normal progenitor-generating PP divisions (Fig. 7A), but shorter than those of PN divisions(Fig. 7A), consistent with all cells behaving as progenitor generators. In a subset of FGF-treated lineages we also observed additional novel cell behaviours, including incomplete interkinetic nuclear migration movements and increased non-apical divisions(Fig. 7D) characterized by even shorter cell-cycle times that are statistically different from PP, PN and apically dividing cells in FGF-treated tissue (see Fig. 7A,B,B′; see Movie 6 in the supplementary material). These data indicate that loss of neurogenic factors in the presence of FGF is accompanied by adoption of a more rapid cell cycle characteristic of cells that only generate further progenitors.

Our analysis of cell behaviour within the living spinal cord reveals the presence of a dynamic cell population that exhibits dramatic and sequential changes in cell shape and subcellular organization during progression through the cell cycle and neuronal differentiation. The retention of a thin basal process composed of membrane and lacking microtubules as cells undergo mitosis at the apical surface confirms that cells maintain contact with both basal and apical surfaces throughout the cell cycle (see Afonso and Henrique, 2006). A similar membranous basal process has been observed in other neuroepithelia(Das et al., 2003; Miyata et al., 2001; Saito et al., 2003) and the relocation of eYFP-tubulin into this structure following mitosis is consistent with a microtubule-based mechanism mediating interkinetic nuclear migration movements (Messier and Auclair,1974). The apical displacement and incorporation of the microtubule-rich midbody into the new apical processes on completion of cytokinesis is the first observation that the post-mitotic midbody gives rise to a specific structure.

Importantly, we demonstrate that all cells in the early spinal cord,whether prospective progenitors or neurons, possess an apical and basal process and exhibit nuclear movement towards the basal surface following mitosis. Prospective neurons with basally located nuclei retain this apical process (see Minaki et al.,2005) and may do so for many hours prior to the first morphological manifestation of neuronal differentiation; the release of this process from the apical surface. This step must thus involve loss of junctional complexes, including adherens junctions (see Afonso and Henrique, 2006),between the differentiating cell and its neighbours. A similar behaviour has also been reported prior to neuronal differentiation in the mammalian cortex,although here prospective neurons do not retain apical contact but migrate away and then form a new process that extends to the apical surface and is retracted on differentiation (Noctor et al., 2004). In the early spinal cord, this loss of apical contacts is followed by reorganization of the microtubule cytoskeleton as the apical process is withdrawn towards the cell body, which then becomes realigned along the basal surface. Together, these observations indicate that an early step in the birth of a neuron is a dramatic change in neuroepithelial cell polarity,as indicated by loss of apical contacts, followed by a reorganization of the microtubule cytoskeleton.

We find that divisions taking place at the same time at the apical surface of the early neural tube can be symmetric divisions that generate two progenitors or two neurons, or asymmetric divisions that produce a progenitor and a neuron. This appears to differ from cell behaviour in the mammalian cortex, where these modes of division are more clearly temporally or spatially separated (Cai et al., 2002; Haubensak et al., 2004; Miyata et al., 2004; Noctor et al., 2004). The time and position of neuron births from both symmetric and asymmetric divisions,which we observe in dorsal and intermediate regions of the early rostral spinal cord, suggest that these cells are reticular spinal interneurons(Sechrist and Bronner Fraser,1991). Symmetric division is also a strategy used to generate some early neurons in the zebrafish hindbrain(Lyons et al., 2003), but this will quickly deplete the progenitor cell population and it is surprising that this mode of neuron production should take place so early in neurogenesis.

This is the first study in which mitotic spindle orientation and subsequent daughter cell fates have been monitored continuously in the vertebrate neural tube. Overall, perpendicular cleavage plane divisions generate daughter cells that adopt the same fates, but these can be either two neurons or two progenitors. As this occurs in the same tissue at the same time, this suggests that individual cells within this early neuroepithelium are not equivalent and that only some cells have neuron-generating (neurogenic) potential. This conclusion is supported by the peppered expression pattern of proneural bHLH transcription factors such as Ngn2 within the early spinal cord(Ma et al., 1999; Diez del Corral et al., 2003)(see Fig. S3 in the supplementary material). If this logic is applied to cell fate outcomes following non-perpendicular cleavage planes, it suggests that only those cells with neurogenic factors have the potential to generate a neuron and a progenitor, whereas non-neurogenic cells will continue to give rise to only progenitors, regardless of spindle orientation. Indeed, this is consistent with the daughter cell fates adopted following divisions with a parallel cleavage plane; most of these divisions give rise to two further progenitors. Overall, these data suggest that the mechanism of mitotic spindle orientation operates in the context of a heterogeneous cell population within the early spinal cord and that here assignment of neurogenic fate is not linked to division orientation.

Our data also reveal that NN division orientations are exclusively perpendicular and statistically distinct from those of both PN and PP divisions, whereas PN and PP division orientations are indistinguishable from each other. This suggests that mitotic spindle orientation is only crucial in neurogenic cells, where it distinguishes between stem cell and terminal modes of neuron production. This conclusion is consistent with data in the mammalian cortex showing that a slight deviation of the cleavage plane from 90° is sufficient to generate asymmetric fates in neurogenic cells(Kosodo et al., 2004), and with the increase in neurons born of terminal divisions following imposition of a perpendicular cleavage plane in this context(Sanada and Tsai, 2005). However, this contrasts with increased proliferation of some retinal cells following imposition of perpendicular cleavage plane divisions(Zigman et al., 2005), which indicates that here mitotic spindle orientation can regulate the switch to neuron production (see Harris,2005).

Nevertheless, there is growing evidence that simply altering mitotic spindle orientation is not sufficient to elicit neuron production, as observed recently in the chick caudal neural tube prior to the normal onset of neuron production (Roszko et al.,2006) and, while our paper was under revision, in the Drosophila optic lobe (Egger et al., 2007). Furthermore, neurogenic determination prior to terminal mitosis has recently been reported in the zebrafish retina where it is mediated by ath5 expression(Kay et al., 2005; Poggi et al., 2005). Determination of a neurogenic fate prior to mitoses that generate neurons in the chick neural tube is also suggested by Ngn2-positive cells, which rapidly incorporate BrdU (Murciano et al.,2002), and by the expression pattern of Dyrk1a(Hammerle et al., 2002).

Tis21, which distinguishes neurogenic divisions in mammalian neural tissue (Iacopetti et al.,1999; Kosodo et al.,2004), can drive cell cycle exit(Tirone, 2001) and it has been suggested that it functions to extend the G1 phase of the cell cycle, a characteristic of cells in the neuron-generating phase of cortical development(Calegari and Huttner, 2003; Takahashi et al., 1995)(reviewed by Gotz and Huttner,2005). It also regulates expression of the neurogenic gene Math1, providing a link between neurogenic status and cell cycle control (Canzoniere et al.,2004). Based on studies in fixed tissue in the later hindbrain and early neural plate, cell cycle times in the early chick spinal cord have been estimated to be ∼8 hours (Hammerle et al., 2002). Our calculations from BrdU labelling and direct measurements reveal a longer average cell-cycle time and a surprising heterogeneity in cell cycle times in the living neuroepithelium. Furthermore,as in the cortex, we show that neurogenic divisions are indeed longer than progenitor-generating divisions. The cell cycle preceding a symmetrical terminal division is predicted to be even longer than asymmetric neurogenic divisions (Calegari and Huttner,2003). This may explain why to date we have not recorded the full cell cycle in a symmetric terminal division. This extension of cell cycle length in neurogenic divisions further helps to distinguish neurogenic from non-neurogenic progenitors.

Consistent with the link between neurogenic factors and cell cycle regulation (reviewed by Ohnuma and Harris,2003), we further show that FGF-mediated inhibition of neuronal differentiation, which involves repression of neurogenic genes, also entails cell cycle acceleration, and a similar observation has been made in cortical progenitor cells in vitro (Lukaszewicz et al., 2002). FGF probably interferes with neurogenesis at multiple levels (Olivera-Martinez and Storey,2007), and we also see an increase in non-apical divisions and breakdown of interkinetic nuclear migration in a subset of cells; a response difference that may reflect cell cycle phase on first exposure to FGF(Hitomi and Stacey, 1999). Finally, the lack of effect of FGF on the range of division orientations at the apical surface reinforces the observation that the mechanism of mitotic spindle position and the assignment of neurogenic cell fate are separable processes.

Development of a long-term live imaging assay in the spinal cord has allowed us to follow individual cells throughout neurogenesis. We have revealed great cell heterogeneity within this tissue, measured accurately mitotic spindle orientation, identified novel cell behaviour associated with neuronal differentiation and imaged continuously for long enough to ascertain cell fate outcomes following oriented divisions. Our observations confirm a link between cell cycle duration and neuron generation and support the important conclusion that although mitotic spindle orientation can regulate cell fate symmetry, it need not determine neurogenic cell fate. This new long-term, high-resolution imaging assay now opens the way to analysis of signalling dynamics on a cell-by-cell basis during neurogenesis in the spinal cord, and to detailed investigation of the links between the mechanics of mitosis and neuron production.

We are grateful to Ruth Diez del Corral for help establishing slice culture methods, Sam Swift for microscopy advice and support, Chris Allan for computing support, Janice Aitken for artwork of a stage 10 embryo, Jonathan Gordon and David Martin for statistical advice, and Jon Clarke, Ruth Diez del Corral, Domingos Henrique and an anonymous reviewer for critical reading of various versions of this manuscript. A.C.W. was supported by a Wellcome Trust four-year PhD program studentship, J.R.S. is a Wellcome Trust Senior Research Fellow, and K.G.S. was an MRC Senior Non-Clinical Research Fellow when this work began and is currently funded by a grant from the MRC (G0600234).