The Phox2b transcription factor coordinately regulates neuronal cell cycle exit and identity

All neurons of the vertebrate CNS are derived from the neuroepithelium, which forms the wall of the embryonic neural tube and generates a vast array of neuronal phenotypes. Early in CNS development, neuroepithelial cells generate more neuroepithelial cells by symmetrical divisions. At a given stage, which varies according to rostrocaudal and dorsoventral position, the dividing neuroepithelial progenitors start generating neuronal precursors, which leave the neuroepithelium or ventricular zone (VZ), generally after completion of their last mitosis, and begin to differentiate. Hence, exit from the cell cycle and the decision to become a neuron are tightly linked (Altman and Bayer, 1984; Rakic, 1988; McConnell, 1995; Huttner and Brand, 1997). Lateral inhibition mediated by the Delta-Notch intercellular signalling pathway is thought to negatively regulate cell cycle withdrawal, thus limiting the number of neurons produced (Chitnis et al., 1995; de la Pompa et al., 1997; Henrique et al., 1997; see Lewis, 1998, for review). In addition, some of the growth factor requirements for neuroepithelial proliferation have been identified (Cameron et al., 1998). However, the cell-intrinsic determinants that positively regulate cell cycle exit in the vertebrate neural tube remain largely unknown.

In Drosophila, differentiation along a neural pathway requires the activity of the proneural genes that code for basic-helix-loop-helix (bHLH) transcription factors (Jan and Jan, 1993). In vertebrates, a number of related genes have proneural activity in that they induce ectopic or supernumerary neurons when misexpressed in Xenopus or zebrafish embryos (for reviews, see Lee, 1997; Anderson, 1999; Brunet and Ghysen, 1999). The ectopic neurons generated express genes characteristic of post-mitotic neurons (Ma et al., 1996; Blader et al., 1997). This implies that the vertebrate proneural genes promote cell cycle withdrawal in addition to conferring neuronal properties, but positive regulation of cell cycle exit by these genes has not been demonstrated in the neural tube. Rather, lack of Mash1, a gene of the bHLH class homologous to Drosophila proneural genes, has been found to cause a decrease in progenitor proliferation in the VZ (Casarosa et al., 1999; Horton et al., 1999). Recently, the winged helix transcription factor XBF-1 has been found to promote neuronal differentiation and to inhibit cell proliferation when misexpressed at low doses in the Xenopus neural plate (Hardcastle and Papalopulu, 2000). Paradoxically, however, inactivation of its mouse orthologue caused reduced proliferation and premature differentiation of neuroepithelial cells (Xuan et al., 1995).

The period of acquisition of neuronal identity starts before and extends beyond completion of the final mitosis. This has been studied most intensively during motor neuron development (Ericson et al., 1997; Sharma et al., 1998; Tanabe et al., 1998; Briscoe et al., 1999; Briscoe et al., 2000). Three general classes of motor neurons can be distinguished: the somatic motor (sm) neurons that innervate most skeletal muscles in the body, the branchiomotor (bm) neurons that innervate the branchial arch-derived muscles of the face and jaw, and visceral motor (vm) neurons that innervate sympathetic and parasympathetic ganglia. The dividing progenitors of most sm and of the spinal vm neurones, i.e. of the motor neurones that send axons ventrally from the neural tube, express the LIM homeobox genes Lhx3 (Lim3) and Lhx4 (Varela-Echevarría et al., 1996; Ericson et al., 1997). In embryos lacking both genes, post-mitotic spinal motor neurons are still generated in proper numbers, but acquire identities reminiscent of cranial bm or vm neurons (Sharma et al., 1998; Tanabe et al., 1998). Conversely, ectopic expression of Lhx3 in the hindbrain leads to the generation of ventrally projecting neurons at the expense of a cranial vm phenotype, but does not seem to influence the number of neurons generated (Sharma et al., 1998). Likewise, the homeobox gene MNR2 is switched on by chick motor neuron progenitors just before their final division. When misexpressed in the spinal cord region of the chick neural tube, ectopic MNR2 is able to induce an sm differentiation programme (Tanabe et al., 1998), characterised by expression of the Islet1 and HB9 transcriptional regulators, which are required for post-mitotic sm differentiation (Pfaff et al., 1996; Arber et al., 1999; Thaler et al., 1999). However, timing and extent of neurogenesis appear to be controlled by an MNR2-independent process (Tanabe et al., 1998). In the mouse, an MNR2 homologue has not been found, and the closely related HB9 gene may have subsumed MNR2 function in this species. In mice lacking HB9, motor neurons transiently express interneuron markers and acquire inappropriate subtype identities, but they are still generated in correct numbers (Arber et al., 1999; Thaler et al., 1999). Upstream of these transcription factor genes, another set of homeodomain proteins subdivides the VZ into different progenitor domains and is responsible for conferring distinct progenitor identities (Briscoe et al., 2000). Nkx6.1 in particular is expressed in the motor neuron progenitors and induces a sm phenotype when misexpressed in the dorsal spinal cord. However, no effects of Nkx6.1 on neurogenesis per se have been reported. Thus, although exit from the cell cycle and acquisition of subtype-specific (as well as generic) neuronal identity appear tightly coordinated in time, how this is achieved has not been elucidated.

In the embryonic mouse (Pattyn et al., 1997) and chick (J. Gay and J. F. B., unpublished results) hindbrain, the closely related paired-type homeobox genes Phox2a and Phox2b are expressed by all bm and vm, but not by sm neurons. Phox2b, but not Phox2a, is already expressed in the dividing progenitors and expression of both genes persists in the differentiating mantle layer neurons (Pattyn et al., 1997; Pattyn et al., 2000). The analysis of the Phox2b knockout phenotype has uncovered two successive functions of this transcription factor in hindbrain motor neurons. In the progenitors, Phox2b activity is required for high level expression of the bHLH family member Mash1. In the post-mitotic precursors, it is necessary for all further differentiation, both generic and type-specific (Pattyn et al., 2000). Here, we demonstrate that in the absence of Phox2b, post-mitotic motor neurons are not generated in proper numbers in the VZ. Conversely, forced expression of Phox2b in the presumptive spinal cord of chick embryos, where it is normally not expressed, promotes expression of early post-mitotic markers and migration into the mantle layer. In addition, Phox2b induces ectopic expression of motor neuron markers in the dorsal spinal cord. Together, these results suggest that Phox2b coordinately regulates the decisions to exit the cell cycle and to become a particular type of neuron.

The generation of the Phox2bLacZ/LacZ mice used in this study has been described (Pattyn et al., 2000). In this line, targeted insertion of the LacZ gene allows for visualisation of the mutant cells that would normally express the inactivated gene.

The coding regions of mouse Phox2b (Pattyn et al., 1997) and mouse Otlx2 (Pitx2) (Mucchielli et al., 1996) were cloned between the EcoRI and XhoI sites of the pCAGGS vector, which drives expression by a CMV/actin hybrid promoter (Koshiba-Takeuchi et al., 2000). GFP was expressed from the pCAGGS-AFP vector (Momose et al., 1999).

Chick embryos 44-52 hours old (HH 12-14) were electroporated in ovo essentially as described (Funahashi et al., 1999). The mPhox2b (3 mg/ml) or Pitx2 (5 mg/ml) expression vectors plus pCAGGS-AFP (0.8 mg/ml) or the empty vector (3 mg/ml) plus pCAGGS-AFP (0.8 mg/ml) were injected into the lumen of the neural tube at spinal cord levels. A pair of electrodes flanking the neural tube delivered six pulses (25 V, 30 mseconds each) from a BTX square electroporator. Embryos were allowed to develop at 38°C before processing at 24 or 48 hours after electroporation (24 h.a.e. or 48 h.a.e., respectively). GFP expression was used to monitor efficiency and extent of electroporation.

Antisense RNA probes for ChAT (Tanabe et al., 1998), EGFP (Clontech), Dll1 (Bettenhausen et al., 1995), Delta1 (Henrique et al., 1997), E. coli LacZ (Pharmacia), Hes5 (Takebayashi et al., 1995), Islet2 (Tsuchida et al., 1994), NeuroM (Roztocil et al., 1997), cPhox2a (Ernsberger et al., 1995), mPhox2b (Pattyn et al., 1997), Pitx2 (Mucchielli et al., 1996) and class III β-tubulin (Lee et al., 1990) were labelled using a DIG-RNA labelling kit (Roche). In situ hybridization and combined in situ hybridization and immunohistochemistry on cryosections were done as described (Hirsch et al., 1998), except that for chick embryos post-hybridization washes were in 50% formamide, 2×SSC, 0.1% Tween 20. For immunohistochemistry, the following antibodies were used: monoclonal anti-BrdU (Sigma), monoclonal anti-GFP (Roche), monoclonal anti-Islet1/2 (Tsuchida et al., 1994) and rabbit anti-mouse Phox2b (Pattyn et al., 1997). TUNEL analysis was done as described (Pattyn et al., 2000). BrdU incorporation and detection in mouse (Pattyn et al., 2000) and chick (Sechrist and Marcelle, 1996) embryos were done as described. Pictures were taken with a Kappa DX30 CCD camera using Kappa software and assembled using Adobe photoshop.

The number of Dll1+ cells in the VZ from Phox2b knockout embryos was assessed on Dll1/BrdU double-stained sections through the hindbrain of embryonic day (E)10.5 embryos. The LacZ-expressing area was delimited by β-galactosidase staining of adjacent sections. A minimum of 680 Dll1+ cells were counted for each genotype on every third section, and 16 sections from four embryos were analysed for each genotype.

The effects of Phox2b misexpression were quantified at 24 h.a.e. by counting the number of BrdU+, Delta1+ and NeuroM+ cells in the electroporated region on transverse spinal cord sections from mPhox2b-transfected and control embryos. Positive cells were counted on every fifth section throughout the optimally transfected region both within the transgene-expressing area, as determined by GFP expression on adjacent sections, and in the equivalent area of the nontransfected side. To normalize for slight differences in developmental stage and rostrocaudal position, the results were expressed as the difference in positive cell numbers between the transfected and the nontransfected side (number of positive cells in the transfected minus the number of positive cells in the nontransfected area). A minimum of five sections was analyzed per embryo, and the results given as the difference in the number of positive cells per section of at least four embryos analyzed.

The distribution of GFP fluorescence along the mediolateral axis of the neural tube was analysed at 24 and 48 h.a.e. on transverse spinal cord sections. On each section, the transfected area of the neural tube was divided into three bins of equal dimensions, from the luminal to the pial surface: a luminal bin, an intermediate bin and a pial bin. The mean fluorescence intensities in each bin and in the total areas were measured using a Zeiss confocal microscope and NIH imaging software for 4-10 sections of four (control transfections) or six (mPhox2b transfections) embryos. Results were expressed in relative fluorescence intensities by dividing the value for each bin by the mean fluorescence of the total transfected area to obtain a measure of the relative distribution of the transfected cells across the mediolateral extent of the neural tube wall.

In the vertebrate neural tube, the dividing progenitors form a single layer of neuroepithelial cells, in which the nuclei migrate in a cell cycle-dependent manner. The neuronal precursors that have exited the cell cycle migrate radially out of the VZ towards the mantle layer and start their post-mitotic differentiation programme (Sauer, 1936; Guthrie et al., 1991). In the following, we operationally define the VZ as the compartment delimited internally by the lumen of the tube, and externally by the outermost nuclei in S phase as determined by a BrdU pulse and the mantle layer as the BrdU-negative area lying basal to it.

Our previous analysis of bm neuron generation has shown that in Phox2b knockout embryos, mantle layer cells do not switch on generic or type-specific differentiation genes and maintain expression of markers normally expressed only in the VZ, although the partition of the wall of the neural tube into an apical VZ and a basal mantle layer is preserved. Moreover, the mantle layer is reduced in size (Pattyn et al., 2000). As shown before, the number of cells in the lateral aspects of the mutant neural tube that do not incorporate BrdU – i.e. post-mitotic precursors – is massively reduced in the ventral region of rhombomere 4 (r4) from E10.5 Phox2b^LacZ/LacZ hindbrain, which normally gives rise to the facial bm neurons (Pattyn et al., 2000; Fig. 1A,B). A cluster of LacZ⁺ cells that had not incorporated BrdU after a 3 hour pulse was still found lateral to the BrdU⁺ cells in the homozygous mutants. This shows that after completion of their last mitosis, they have followed the usual pattern of radial migration from the VZ to the mantle layer. Although the results show that fewer post-mitotic precursors accumulate in the mutants, they leave open several possibilities as to how this could come about, such as (1) a reduced rate of progenitor proliferation, (2) the generation of fewer post-mitotic cells in the VZ, (3) a reduced rate of exit from the VZ or (4) preferential death of post-mitotic mantle layer cells. We observed increased cell death in the mutant territory, but the apoptotic cells were distributed throughout the wall of the neural tube and there was no sign of preferential death of cells in the mantle layer (Pattyn et al., 2000). We then counted the number of cells in the mutant territory that had incorporated BrdU after a 1 hour pulse. If anything, this value was slightly higher for Phox2b^−/− than for Phox2b^+/− embryos (+15%, results not shown).

To test whether fewer post-mitotic neuronal precursors are generated within the mutant VZ, we examined the expression of Dll1, a mouse homologue of Drosophila Delta, which is the earliest known marker expressed after the last S phase (Bettenhausen et al., 1995; Myat et al., 1996; Henrique et al., 1997). Dll1 expression displayed a salt-and-pepper pattern throughout the VZ of heterozygotes (Fig. 1C), presumably corresponding to cells that have just completed their last mitosis close to the lumen and are on their way out of the VZ (Myat et al., 1996; de la Pompa et al., 1997; Dunwoodie et al., 1997; Henrique et al., 1997). In the LacZ⁺ region from mutant embryos, Dll1 was still expressed (Fig. 1D). However the Dll1 expression pattern was altered in two ways. First, mutant precursors fail to downregulate Dll1 upon exit from the VZ, as they normally do (compare Fig. 1C and D). The persistence of VZ markers on the post-mitotic mantle layer cells in the mutants had also been observed for other genes (Pattyn et al., 2000). Since the mutant cells that have become post-mitotic still migrate to the mantle layer, this results in the accumulation of Dll1⁺ cells in the lateral aspects of the neural tube. Second and more relevant to the question under study, Dll1 expression in the VZ was reduced by 50% (Fig. 1C-E). To test whether this decrease was due to the selective death of Dll1⁺ cells, we performed Dll1/TUNEL double-labelling experiments. Virtually none of the TUNEL⁺ cells in the VZ were also Dll1⁺ (Fig. 1F), arguing that Dll1 is not activated at its normal rate. The failure of detecting doubly labelled TUNEL⁺/Dll1⁺ cells cannot be ascribed to mRNA degradation in dying cells, since virtually all TUNEL⁺ cells could be labelled with a LacZ probe (not shown). These data thus indicate that, at any point in time, a smaller proportion of the mutant progenitors become post-mitotic, switch on Dll1 and migrate out of the VZ.

We reasoned that the reduced numbers of Dll1-expressing cells should result in a lesser degree of Notch-mediated lateral inhibition in the VZ. A major output of Notch signalling is the induction of the Enhancer of split and related genes (Weinmaster, 1997; Wettstein et al., 1997; Ohtsuka et al., 1999). We therefore examined the Enhancer of split homologue Hes5, whose normal expression throughout the VZ of the embryonic neural tube is reduced or abrogated in mouse mutants with disrupted Notch signalling (de la Pompa et al., 1997). Hes5 has also been implicated in lateral inhibition in another neurogenic region, the olfactory placode (Cau et al., 2000). In the VZ of ventral r4 from wild-type or heterozygous embryos, Hes5 transcripts were not uniformly expressed, clusters of cells with high expression levels juxtaposing cells where Hes5 expression was barely detectable (Fig. 1G). A similar pattern has been described in the olfactory placode (Cau et al., 2000). In Phox2b-deficient embryos, Hes5 expression was reduced in the normally Phox2b⁺ area (Fig. 1H). Hes5 transcripts in the homozygous mutants were only detectable in the regions where expression was highest in the heterozygotes while the areas of weak expression appeared negative.

The foregoing data suggested to us that in Phox2b mutants, exit from the cell cycle is defective. Alternatively, Phox2b could be required for the survival of committed precursors at a stage that precedes Dll1 expression. To test whether Phox2b activity is sufficient to drive progenitors to leave the cell cycle and to trigger their post-mitotic differentiation, we turned to gain-of-function experiments in chick embryos. We used misexpression in chick embryos since in ovo electroporation offers a convenient means for temporally and spatially controlled expression in one side of the neural tube (Funahashi et al., 1999; Itasaki et al., 1999). We electroporated a mouse Phox2b (mPhox2b) expression vector into the presumptive spinal cord region at Hamburger and Hamilton (1951) stage (HH) 12-14 and examined the embryos 24 hours (HH 18-19) or 48 hours (HH 22-23) later. At these stages, the chick spinal cord normally expresses neither Phox2b nor Phox2a. As in mouse (Tiveron et al., 1996; Hirsch et al., 1998), a small population of interneurons just dorsal to the sulcus limitans expresses Phox2a, but not Phox2b, at slightly later stages (>HH24) (results not shown). A GFP expression plasmid was coelectroporated alongside to visualise the transfected cells. As a control, we expressed either GFP alone or the homeobox gene Pitx2 (formerly called Otlx2), which is also a paired-class homeodomain protein (Mucchielli et al., 1996). Virtually all GFP-expressing cells coexpressed Phox2b or Pitx2 (Fig. 2A,B), which allowed us to use either the homeobox transgenes or the GFP reporter as markers of the transfected cells.

We first tested whether mPhox2b induced expression of two early markers of post-mitotic VZ cells, the chicken Dll1 orthologue Delta1 and NeuroM, which codes for a bHLH transcription factor and, like its mouse orthologue Math3 (Takebayashi et al., 1997), is transiently expressed by post-mitotic VZ cells en route to the mantle layer (Roztocil et al., 1997). 24 h.a.e. there was a marked increase in Delta1- and NeuroM-expressing cells in the mPhox2b-transfected VZ (Fig. 3E-H). The dorsalmost Phox2b-transfected cells consistently failed to respond with Delta1 or NeuroM induction, possibly because they constitute neural crest-derived cells. An increase in Delta1 or NeuroM expression was never observed when Pitx2 (Fig. 3A-D) or GFP alone (not shown) were electroporated. Cell counts revealed that the differences in the numbers of positive cells between the mPhox2b-transfected and the nontransfected sides of the embryos were highly significant. In absolute terms, we counted approximately 16 Delta1⁺ or NeuroM⁺ cells/section on the mPhox2b-transfected compared to approximately 6 Delta⁺ or NeuroM⁺ cells on the nontransfected side, a >2.6 fold increase. Pitx2 expression produced a decrease in positive cells in the transfected area, whereas GFP expression alone was without effect (Fig. 3J). Phox2b misexpression thus produces a mirror image of the phenotype observed in Phox2b-deficient embryos, where both Dll1 (Fig. 1) and Math3 (Pattyn et al., 2000) are downregulated in the VZ. Most Delta1⁺ cells in the transfected region coexpressed mPhox2b, suggesting that Phox2b acts cell-autonomously (Fig. 3I). In line with the normally observed expression pattern of Dll1/Delta1 on VZ cells en route to the mantle layer, the mPhox2b-induced increase in Delta1 expression was transient and confined to VZ cells. At 32 h.a.e., an increase in the number of Delta1 cells was not observed any more, and at 48 h.a.e., when the great majority of the transfected cells has shifted to the mantle layer and expresses generic neuronal differentiation markers (see below), virtually none of the mPhox2b⁺ cells also express Delta1 (not shown). This finding shows that the normal temporal relationship between expression of Delta1 and markers of later stages of differentiation is preserved in ectopic Phox2b-induced neurons.

As expected from the increase in post-mitotic Delta1⁺, NeuroM⁺ neurons, there was a marked reduction in the number of BrdU-incorporating cycling progenitors in the mPhox2b-transfected area at 24 h.a.e. (Fig. 4E,F) compared to control transfections (Fig. 4A,B). We counted around half as many BrdU⁺ cells on the mPhox2b-transfected than on the nontransfected side compared to an 11% decrease after electroporation of GFP alone (Fig. 4I). We tested expression of proliferating-cell nuclear antigen (PCNA) as an additional marker of cycling cells. PCNA is a nuclear protein that is part of the elongation apparatus and essential for DNA replication (Tsurimoto, 1998). In the wall of the neural tube, DNA-bound PCNA has been found in late G₁ to mid-S phase nuclei (Takahashi and Caviness Jr, 1993). We found a clear reduction in the number of PCNA⁺ nuclei after transfection with mPhox2b (Fig. 4G,H), but not after transfection with GFP alone (Fig. 4C,D), again suggesting that Phox2b promotes cell cycle exit.

Does mPhox2b also drive the cells to migrate to the mantle layer? At 24 h.a.e., the vast majority of the transfected cells were found in the VZ and had the typical morphology of neuroepithelial cells, in both mPhox2b-transfected and control neural tubes. One day later, however, the great majority of mPhox2b-transfected cells were found in the lateral region of the neural tube while in the controls, most electroporated cells were still in the neuroepithelium and looked like neuroepithelial cells (Fig. 5A-H). Quantification of the GFP fluorescence showed that at 48 h.a.e., 55% of the mPhox2b-transfected cells were found in the lateral third of the neural tube compared to 33% in the GFP control (Fig. 5I-K). We confirmed that the Phox2b-transfected cells that had translocated laterally were indeed post-mitotic differentiating mantle layer neurons using BrdU incorporation and class III β-tubulin (Lee et al., 1990) as markers (Fig. 6). At 48 h.a.e., few, if any, of the mPhox2b⁺ cells incorporated BrdU after a 3 hour pulse. Most were located external to the BrdU⁺ VZ progenitors, showing that they had shifted to the mantle layer, and many of them expressed β-tubulin. Together, these data demonstrate that Phox2b positively regulates the decisions to become post-mitotic, to migrate to the mantle layer and to initiate generic neuronal differentiation.

In the mouse (Pattyn et al., 1997) and chick (J. Gay and J. F. B., unpublished results) hindbrain, Phox2b is expressed in the bm and vm, but not in the sm neurons, and is required for their formation as shown by targeted gene inactivation in the mouse (Pattyn et al., 2000). We therefore tested whether the ectopically mPhox2b-expressing neurons acquire a bm or vm identity. In the ventral hindbrain, where the cranial motor neurons are being born, Phox2b is already expressed in the cycling progenitors and persists on the post-mitotic neurons. The LIM homeobox gene Islet1 is expressed most prominently in all post-mitotic motor neuron precursors, but also in a small population of dorsal interneurons (Ericson et al., 1992; Liem et al., 1997), while Islet2 is specific for sm neurons (Varela-Echevarría et al., 1996; Ericson et al., 1997). Finally, all motor neurons and no other neurons in the spinal cord express choline acetyltransferase (ChAT) at the stages studied (see Fig. 7D). At 48 h.a.e., many mantle layer cells that expressed mPhox2b could be labelled by an antibody recognising both Islet1 and Islet2. There was thus a clear increase in Islet1/2⁺ cells in the area that expressed the transgene (Fig. 7E-H), an effect not seen in GFP (Fig. 7A,B) or Pitx2 (not shown) control electroporations. Islet2 was only expressed ventrally in differentiating spinal motor neurons, showing that Islet1/2 immunoreactivity is due to Islet1, but not Islet2 expression (Fig. 7I,J). While ChAT expression was confined to the ventrally differentiating motor neurons in control transfections and on the nontransfected side, it was detected dorsolaterally throughout the mPhox2b-transfected area (Fig. 7C,D,K,L). Hence, forced expression of Phox2b generates ectopic neurons that, like cranial bm or vm neurons, are Islet1⁺, ChAT⁺ and Islet2⁻.

We used chicken Phox2a (cPhox2a) as an additional marker to confirm the bm or vm phenotype of the ectopically Phox2b-expressing neurons. In both mouse (Pattyn et al., 1997) and chick (J. Gay and J. F. B., unpublished results) embryos, Phox2a, whose expression follows that of Phox2b, is also expressed by cranial bm and vm but not by sm neurons. Unlike Phox2b, Phox2a is downregulated on cranial motor neurons beyond E13 in the mouse (Tiveron et al., 1996; Jacob et al., 2000). In Phox2b-transfected embryos examined at 20 h.a.e. (HH 18), many of the transfected cells also expressed cPhox2a

(Fig. 7M), an effect not seen after electroporation of GFP alone (not shown). Ectopic cPhox2a expression was transient; its expression had become weaker by 32 h.a.e. (HH 20) (Fig. 7N) and had declined to barely detectable levels by 48 h.a.e. (not shown). At 32 h.a.e., many of the cPhox2a⁺ cells had shifted to the mantle layer and were thus most likely differentiating post-mitotic neurons. Hence, Phox2b-expression also induces Phox2a, which is never expressed by sm neurons, further suggesting that Phox2b activates at least some aspects of a bm or vm differentiation programme. The transitory character of this expression may be related to the downregulation of Phox2a that is normally observed in cranial motor neurons at later stages.

Neurogenesis in the vertebrate neural tube is marked by a rapid succession of events in precursor cells. Before the first post-mitotic neurons are generated, different homeodomain transcription factors already subdivide the VZ into distinct progenitor domains (Briscoe et al., 2000). Once neurogenesis is under way, the progenitors committed to become a neuron complete their last mitosis close to the lumen in most regions of the neural tube. They then migrate away from the VZ while transiently activating genes such as Math3/NeuroM or the Ebf family members (Roztocil et al., 1997; Garel et al., 1997; Pattyn et al., 2000), which mark the earliest steps of generic neuronal differentiation. This is accompanied or soon followed by expression of downstream transcription factor genes that specify neuronal sub-type identity (see for reviews, Goulding, 1998; Edlund and Jessell, 1999). These decisions must be regulated coordinately, but how this is achieved is poorly understood. The evidence presented in this study shows that the homeodomain transcription factor Phox2b promotes neural tube progenitors to exit the cell cycle and to start a generic neuronal differentiation programme, and that its expression is sufficient to initiate differentiation towards a motor neuronal phenotype. Thus, Phox2b appears to be a critical determinant that successively regulates commitment to become a neuron and execution of a sub-type-specific differentiation programme.

In mice lacking Phox2b, there is a massive reduction in post-mitotic mantle layer cells in the region where the bm neurons of the facial nerve normally develop (Pattyn et al., 2000). Our previous results showed that the depletion of the mantle layer is not due to an early death of mantle layer cells, as observed for example in Islet1 knockout embryos (Pfaff et al., 1996). However, they did not allow us to conclude that Phox2b is required for proper exit from the cell cycle since other possibilities remained, such as a reduced rate of proliferation of the mutant progenitors or a reduced rate of exit from the VZ. We here show that expression of Delta1/Dll1, the earliest known marker expressed by VZ cells after completion of their last S phase (Myat et al., 1996; de la Pompa et al., 1997; Dunwoodie et al., 1997; Henrique et al., 1997), is defective in the VZ of mutant embryos. Since this defect is not due to a requirement of Phox2b for the survival of Dll1⁺ cells, this suggests that Phox2b positively regulates cell cycle exit of the progenitors in which it is expressed. Our gain-of-function experiments in the chick neural tube fully confirm the role of Phox2b as a positive regulator of cell cycle exit. Forced expression of Phox2b reduces the numbers of BrdU-incorporating and PCNA-expressing cycling cells and upregulates Delta1/Dll1 and NeuroM/Math3, which are transiently expressed by early post-mitotic precursors on their way out of the VZ (Myat et al., 1996; de la Pompa et al., 1997; Henrique et al., 1997; Roztocil et al., 1997). Hence Phox2b clearly acts on progenitor cells in the VZ. Lack of Phox2b has the opposite effect: the number of Delta1/Dll1⁺ cells in the VZ is reduced by half (this study) and NeuroM/Math3 expression is extinguished (Pattyn et al., 2000). It could be argued that Phox2b, rather than driving cells to exit the cell cycle, increases the survival of progenitors committed to a neuronal fate, in particular since spontaneous cell death has been observed in the dorsal chick spinal cord at the stages studied (Homma et al., 1994). However by TUNEL analysis, we found spontaneous cell death to occur only in the dorsalmost part of the VZ and its extent not to be reduced by Phox2b transfection (results not shown). This makes it very unlikely that the effects of Phox2b misexpression can be explained by a positive effect on survival.

2 days after transfection, most cells ectopically expressing Phox2b have relocated to the mantle layer and express the generic neuronal marker class III β-tubulin, an effect not seen in control transfections. Hence, Phox2b expression is sufficient to drive cells to become post-mitotic, to start a neuronal differentiation programme and to migrate to the mantle layer. In the absence of Phox2b, neuroepithelial cells are still able to become post-mitotic, albeit at reduced rates. This argues that rather than being absolutely required for cell cycle withdrawal, Phox2b positively regulates the rate at which it occurs.

At what step in the pathway does Phox2b act? The bHLH transcription factors with proneural activity are another class of transcriptional regulators that appear to control timing and extent of neuronal differentiation (Fode et al., 1998; Ma et al., 1998). In Phox2b knockout embryos, Mash1, a homologue of Drosophila proneural genes, is expressed at strikingly reduced levels in the bm progenitors and is downregulated prematurely in the sympathetic ganglion primordia (Pattyn et al., 1999; Pattyn et al., 2000). One possibility thus is that Phox2b acts by maintaining or boosting expression of proneural genes, a possibility corroborated by our finding that Phox2b upregulates expression of Delta1/Dll1, which is thought to be a target of transcription factors with proneural activity (Chitnis et al., 1995; Casarosa et al., 1999). Alternatively, Phox2b may function downstream of, or in parallel with, proneural gene products, by promoting expression of bHLH transcription factors that lie further downstream in the regulatory cascade (Fode et al., 1998; Ma et al., 1998; Cau et al., 2000) or by cooperating with bHLH factors in the control of cell cycle regulators. Finally, Phox2b acts in the context of the Delta-Notch lateral inhibition system. As judged by the reduction in Hes5 expression (de la Pompa et al., 1997; Casarosa et al., 1999; Ohtsuka et al., 1999; Cau et al., 2000), Notch signalling is activated to a lesser degree in the Phox2b mutant territory, probably as a consequence of reduced expression of the Notch ligand Dll1. This should initiate a feedback loop (de la Pompa et al., 1997; Wettstein et al., 1997; Lewis, 1998), counteracting the negative effect of Phox2b deficiency on Dll1 expression and cell cycle exit. Conversely, increased Delta1 expression in the gain-of-function experiments should initiate a feedback loop that opposes the effect of Phox2b misexpression. It is thus easy to understand that in Phox2b mutants not all cells remain cycling progenitors, and that not all cells misexpressing Phox2b can be driven to express Delta1 or NeuroM at any one time.

Our analysis of the loss-of-function phenotype has shown that Phox2b is absolutely required for the generation of the bm and vm neurons in which it is expressed (Pattyn et al., 2000). Here we show that Phox2b is also sufficient to induce characteristics of a motor neuronal phenotype in the lateral spinal cord, where motor neurons normally never develop. The ectopic Phox2b-expressing cells switch on the motor neuronal markers Islet1/2 and ChAT, but not Islet2, which is specific for sm neurons. Forced expression of Phox2b also transiently activates Phox2a, whose expression in motor neurons is confined to bm and cranial vm neurons. Hence, the expression of Phox2b in neural progenitor cells appears sufficient to activate a differentiation programme resembling that of bm or hindbrain vm neurons, but in the absence of specific markers we cannot distinguish between these two fates.

Motor neurons are normally generated only in the ventral neural tube in response to sonic hedgehog signalling from the notochord and floor plate (Briscoe et al., 1999; Briscoe et al., 2000; for a review, see Edlund and Jessell, 1999). Generation of alar plate neurons depends on BMPs and related molecules that are expressed in the dorsalmost regions of the neural tube (Liem et al., 1995; Lee et al., 2000), and exposure of progenitors to BMPs inhibits motor neuron generation (Basler et al., 1993; Liem et al., 1995). Our results show that like MNR2 and Hb9 (Tanabe et al., 1998), Phox2b is able to override such negative control of motor neuronal differentiation in the dorsal spinal cord. In the hindbrain, however, the dorsal Phox2b-expressing precursors do not give rise to motor neurons, but to sensory relay neurons, to neurons of the reticular formation and to noradrenergic neurons (Pattyn et al., 1997, A. P. and J. F. B., unpublished). Other studies have shown that forced expression of Phox2b or Phox2a in the chick neural crest pathway (Stanke et al., 1999) or of Phox2a in the zebrafish embryo (Guo et al., 1999) generates ectopic noradrenergic neurons. One possibility thus is that Phox2b must cooperate with other factors to specify dorsal sub-type identities in the hindbrain, and that these factors lack in the spinal cord.

Taken together, our studies show that a single transcription factor is able both to increase cell cycle exit of neuronal progenitors and to initiate at least some aspects of sub-type-specific differentiation of the neurons thus generated. Transcriptional regulators, which assume this role in neuronal types other than bm and vm neurons, remain to be identified.

We thank the following for probes: M. Ballivet, R. Beddington, F. Guillemot, T. Jessell, H. Ogawa and C. W. Ragsdale. C. Moretti is thanked for help with the quantitative analyses. This work was supported by institutional grants from CNRS and Université de la Méditerranée and by specific grants from the European Community Biotech programme (BIO4-CT98-0112) and the Association pour la Recherche sur le Cancer.