Within the developing vertebrate nervous system, specific subclasses of neurons are produced in vastly different numbers at defined times and locations. This implies the concomitant activation of a program that controls pan-neuronal differentiation and of a program that specifies neuronal subtype identity, but how these programs are coordinated in time and space is not well understood. Our previous loss- and gain-of-function studies have defined Phox2b as a homeodomain transcription factor that coordinately regulates generic and type-specific neuronal properties. It is necessary and sufficient to impose differentiation towards a branchio- and viscero-motoneuronal phenotype and at the same time promotes generic neuronal differentiation. We have examined the underlying genetic interactions. We show thatPhox2b has a dual action on pan-neuronal differentiation. It upregulates the expression of proneural genes (Ngn2) when expressed alone and upregulates the expression of Mash1 when expressed in combination with Nkx2.2. By a separate pathway, Phox2brepresses expression of the inhibitors of neurogenesis Hes5 andId2. The role of Phox2b in the specification of neuronal subtype identity appears to depend in part on its capacity to act as a patterning gene in the progenitor domain. Phox2b misexpression represses the Pax6 and Olig2 genes, which should inhibit a branchiomotor fate, and induces Nkx6.1 and Nkx6.2, which are expressed in branchiomotor progenitors. We further show that Phox2b behaves like a transcriptional activator in the promotion of both, generic neuronal differentiation and expression of the motoneuronal marker Islet1. These results provide insights into the mechanisms by which a homeodomain transcription factor through interaction with other factors controls both generic and type-specific features of neuronal differentiation.
INTRODUCTION
The vertebrate CNS is derived from the neuroepithelium, a single layer of proliferating cells with stem cell characteristics that line the lumen of the neural tube. During neurogenesis, the dividing neural stem cells generate neuronal precursors that migrate away from the ventricular zone (VZ),generally after completion of their last mitosis, and begin to express generic and type-specific neuronal genes. A key question in early CNS development is how specific subclasses of neurons are produced in proper numbers at the correct times and locations. This process involves the concomitant activation of programs that commit neural progenitors to differentiation and of programs directing neuronal subtype identity. Considerable advances have been made in defining the mechanisms that govern each of these programs in several model systems (reviewed by Cepko,1999; Guillemot,1999; Jessell,2000; Briscoe and Ericson,2001; Ohnuma et al.,2001). However, the question of how they are integrated to yield the highly reproducible pattern of neurogenesis has been addressed only very recently (for a review, see Bertrand et al., 2002).
Generic neuronal differentiation in vertebrates, as in Drosophila,is regulated by basic helix-loop-helix (bHLH) transcription factors, which promote it, and the Notch signaling pathway, which inhibits it. Several vertebrate genes encoding bHLH proteins (such as Ngn1, Ngn2 andMash1) are thought to be equivalent to Drosophila proneural genes and to confer competence to become a neuron (for a review, seeKageyama and Nakanishi, 1997;Brunet and Ghysen, 1999;Bertrand et al., 2002). Pioneering studies in Xenopus and zebrafish have demonstrated that many of the neurally expressed bHLH factors promote neuronal differentiation during primary neurogenesis (Ma et al.,1996; Blader et al.,1997). Recent work in mouse and chick embryos shows that they function similarly in the neural tube of higher vertebrates(Mizuguchi et al., 2001;Novitch et al., 2001;Scardigli et al., 2001). A key property of the early expressed bHLH proteins is that they restrict their own proneural action by activating the expression of genes that inhibit neurogenesis. This `lateral inhibition' has been well documented during primary neurogenesis in Xenopus and in the chick retina, where neural bHLH proteins up-regulate the expression of Delta that activates the Notch receptor on neighboring cells (Chitnis et al., 1995; Henrique et al.,1997; Koyano-Nakagawa et al.,1999). Activated Notch in turn inhibits expression and activity of the neural bHLH genes via its effector genes, the bHLH transcriptional repressors of the Hes family, which have been shown to inhibit neurogenesis in a variety of settings (Ishibashi et al.,1995; Wettstein et al.,1997; Kageyama and Nakanishi,1997; Kageyama and Ohtsuka,1999; Ohtsuka et al.,1999; Castella et al.,2000; Nakamura et al.,2000; Kondo and Raff,2000; Cau et al.,2000; Kabos et al.,2002).
Another class of molecules that have been implicated in the control of neurogenesis are the Id HLH factors. Id proteins block differentiation and promote proliferation in diverse cell types, including neural cells, mainly by acting as dominant-negative inhibitors of positive regulatory bHLH proteins(for a review, see Norton,2000). The four vertebrate Id family members are expressed in the embryonic neural tube in partially overlapping patterns(Jen et al., 1997). In mice double mutant for Id1 and Id3, the neural tube shows signs of premature neuronal differentiation(Lyden et al., 1999). Conversely, overexpression of Id2 leads to overgrowth of the neural tube (Martinsen and Bronner-Fraser,1998), and forced expression of Id1 or Id2blocks neuronal differentiation (Cai et al., 2000; Toma et al.,2000). However, how the expression of the Id family members is regulated in the neural tube and their precise mode of action in neural cells have not been elucidated.
In general, newly born CNS neurons acquire phenotypes that reflect their site of origin in the VZ. How this comes about has been best studied in the ventral spinal cord. In response to a gradient of sonic hedgehog secreted from ventral axial structures, the VZ is partitioned along the dorsoventral axis into discrete domains that express particular combinations of homeodomain (HD)transcription factors. These HD proteins appear to specify the subtype identity of their neuronal progeny through the action of a different set of HD proteins, which are switched on around the time of the last mitosis (for a review, see Jessell, 2000;Briscoe and Ericson, 2001;Lee and Pfaff, 2001). Most of the early expressed HD proteins appear to function as transcriptional repressors and are thought to specify neuronal identity by repressing alternative fates (Muhr et al.,2001; Vallstedt et al.,2001).
How the different molecular machineries that direct generic and type-specific aspects of neuronal differentiation are coordinated in any class of neurons remains poorly understood. One way this seems to be achieved is by the fate-specifying action of proneural genes themselves. Rather than merely drive a `generic' pathway of neuronal differentiation, they also participate in the specification of neuronal types(Fode et al., 2000;Gowan et al., 2001;Scardigli et al., 2001;Parras et al., 2002). A striking recent example of the neural-fate determining properties of a bHLH factor is provided by Olig2 function in spinal motoneuron progenitors(Mizuguchi et al., 2001;Novitch et al., 2001;Lu et al., 2002).
We have previously documented such a dual action on pan-neuronal and type-specific differentiation for the fate-determining HD protein Phox2b. In the ventral hindbrain, Phox2b is expressed by the progenitors of the two main classes of cranial motor neurons, the branchimotor (bm) and visceromotor (vm) neurons (collectively termed bm/vm neurons), and by their postmitotic descendants, but not in somatic motor (sm) neurons(Pattyn et al., 1997). In the progenitors, Phox2b is necessary for cell cycle exit in proper numbers. In the postmitotic precursors, Phox2b function is required for all aspects of type-specific and generic differentiation. Conversely,forced expression of Phox2b in the spinal cord promotes pan-neuronal differentiation and emigration from the VZ and imparts a phenotype, which resembles that of bm/vm neurons (Pattyn et al., 2000; Dubreuil et al.,2000). In the Phox2b-expressing cells, the early postmitotic markers Delta1 and Math3/NeuroM are induced prematurely, and the neurons thus generated ectopically expressPhox2a, choline acetyltransferase (ChAT) andIslet1, but not Islet2, as do bm and vm neurons. However,the molecular interactions by which Phox2b accomplishes this have not been elucidated.
We have begun to examine the genetic interactions by which Phox2bpromotes both, generic and type-specific aspects of neurogenesis. Ectopic expression studies provide evidence that Phox2b drives pan-neuronal differentiation by upregulating Ngn2 in the absence and ofMash1 in the presence of Nkx2.2 and by repressing the negative regulators of neurogenesis chick Hes5b and Id2. Initiation of a bm/vm fate represents a third activity that implies downregulation of Pax6 and Olig2 and upregulation ofNkx6.1 and Nkx6.2. Our data reveal how an HD transcription factor, through interaction with other factors expressed in the progenitor domain, coordinately regulates pan-neuronal and type-specific differentiation.
MATERIALS AND METHODS
Expression vectors
The coding regions of mouse Hes5(Takebayashi et al., 1995),chicken Id2 (Martinsen and Bronner-Fraser, 1998), Mash1(Cau et al., 1997), mouseNgn2 (Cau et al.,1997), mouse Nkx2.2(Hartigan et al., 1996) andmPhox2b (Pattyn et al.,1997) were cloned into the pCAGGS vector that drives expression by a CMV/actin hybrid promoter(Koshiba-Takeuchi et al.,2000). GFP was expressed from the pCAGGS-AFPvector (Momose et al., 1999). Activator and repressor forms of Phox2a/b were constructed by fusing the PCR-amplified Phox2a homeobox (which is identical at the amino acid level to that of Phox2b) in frame 3′ to either theDrosophila Engrailed repressor domain [the construct is identical to the one described by Lo et al. (Lo et al.,1999) under the name pERPH] or the herpes simplex VP16transactivation domain (Triezenberg et al., 1988) and subcloning into the pCAGGS vector,yielding PHDEnR and PHDVP16, respectively. To express the isolated homeodomain, a Kozak sequence containing an initiator methionine was fused 5′ to the Phox2a homeobox before subcloning into pCAGGS. Siamois-based control constructs were as described previously(Lemaire et al., 1998). Correct expression of all constructs was verified by in situ hybridization with the appropriate probes. In all cases, expression of the transfected gene was co-extensive with that of GFP.
Electroporation
Chick embryos 44- to 52-hour-old (HH 12-14) were electroporated in ovo essentially as described (Dubreuil et al.,2000). The expression vectors were used at 1 mg/ml except for Nkx2.2 at low doses and the homeobox fusion constructs, which were used at 0.5 mg/ml, and pCAGGS-AFP (0.8 mg/ml). We always co-injectedpCAGGS-AFP to visualize the transfected area. Embryos were allowed to develop at 38°C for different time periods. After harvesting, the embryos were fixed in 4% paraformaldehyde, embedded in gelatin and analyzed on transverse neural tube sections at the transfected level.
Histological methods
Antisense RNA probes for Cash1(Jasoni et al., 1994),Delta1 (Henrique et al.,1997), EGFP (Clontech), cHes5b, Id2(Martinsen and Bronner-Fraser,1998), Islet2(Tsuchida et al., 1994),NeuroM (Roztocil et al.,1997), Ngn1 and Ngn2(Perez et al., 1999),Nkx2.2 (Briscoe et al.,1999), Nkx6.1 (Qiu et al., 1998), Nkx6.2(Cai et al., 1999),Olig2 (Zhou et al.,2001), Pax6 (kindly provided by T. Ogura), andPhox2b and choline acetyltransferase (ChAT) (kindly provided by T. Jessell) were labeled using a DIG-RNA labeling kit (Roche). In situ hybridization and combined in situ hybridization and immunohistochemistry on cryosections were carried out as described(Hirsch et al., 1998;Dubreuil et al., 2000). For immunohistochemistry, the following antibodies were used: monoclonal anti-BrdU(Sigma), monoclonal anti-Islet1/2(Tsuchida et al., 1994) and rabbit anti-mouse Phox2b (Pattyn et al.,1997). BrdU incorporation and detection in chick embryos were carried out as described (Sechrist and Marcelle, 1996). Pictures were taken with Kappa DX30, Nikon DXM 1200 or Leica DC300F CCD cameras using Kappa, ACT-1 or Leica software and assembled using Adobe Photoshop.
Quantitative analyses
In situ hybridization signals were quantified by measuring the signal intensity captured with a CCD camera on transverse spinal cord sections. On each section, the mean signal intensity was recorded for the optimally transfected area, as determined by GFP expression on an adjacent section, and in an equivalent area from the non-transfected side. The results were expressed as the difference in mean signal intensities between the transfected and the non-transfected sides and statistical significance determined by two-tailed t-test.
RESULTS
To explore the genetic interactions by which Phox2b triggers generic and type-specific neuronal differentiation, we used in ovo electroporation in chick embryos. By this method, high-level expression of exogenous genes is achieved in neuroepithelial progenitors and persists in the differentiating mantle layer (ML) neurons. Misexpression of Phox2b should thus mimic the situation in the ventral hindbrain, where Phox2b is expressed in the VZ progenitors and persists during terminal differentiation of the bm/vm neurons born in this region.
Phox2b upregulates expression of proneural genes
One way by which Phox2b could drive neuronal progenitors to become postmitotic is by inducing or boosting the expression of bHLH transcription factors with proneural activity, which should promote pan-neuronal differentiation. Ngn2 overexpression in neuroepithelial progenitors,for example, has recently been shown to promote their exit from the cell cycle, migration to the ML and expression of pan-neuronal markers(Mizuguchi et al., 2001;Novitch et al., 2001). We show in addition that, as does misexpression of Phox2b(Dubreuil et al., 2000),misexpression of Ngn2 also upregulates Delta1, the earliest known marker of postmitotic cells in the VZ(Myat et al., 1996)(Fig. 1A). When overexpressed in the chick neural tube, Mash1 behaved basically like Ngn2. It promoted cell cycle exit as assessed by BrdU incorporation, Delta1expression and relocation to the ML (Fig. 1C,D, and not shown). After transfecting GFP alone, most electroporated cells were still in the VZ (not shown)(Dubreuil et al., 2000). Most cells transfected with Ngn2 were positioned laterally already at 24 hours after electroporation (h.a.e.) (Fig. 1A), while after transfection of Mash1 andPhox2b, an equivalent effect was observed only at 48 h.a.e.(Fig. 1C,D)(Dubreuil et al., 2000). This suggests that at the doses used, Ngn2 is more potent thanMash1 or Phox2b.
Two observations made Cash1, the chicken ortholog of Mash1, an obvious candidate for mediating the effect of Phox2b on pan-neuronal differentiation. First, Mash1 is co-expressed with Phox2b in the cranial bm/vm progenitors (Pattyn et al., 2000) and is in fact the only bHLH gene we found expressed in this progenitor population (M. R. H., unpublished). Second, in the absence ofPhox2b function, Mash1 expression in the bm/vm progenitors is downregulated (Pattyn et al.,2000). However, misexpression of mouse Phox2b(mPhox2b) did not induce Cash1 in the presumptive spinal cord either at 6 or at 24 h.a.e. (Fig. 1E,G).
We thus reasoned that another bHLH factor with proneural activity might mediate the effect of Phox2b in spinal regions of the neural tube. Among the three genes studied (Cash1, Ngn1 and Ngn2),Ngn2 was prominently expressed in the HH18-20 spinal cord, in a broad ventrally located area, where the sm neurons are being born at this stage, and in some scattered cells further dorsally. Misexpression of Phox2bresulted in a marked dorsal expansion of the Ngn2 expression domain. The increase in Ngn2 expression was greatest at 6 h.a.e. and declined thereafter (Fig. 1F,F′,I,H,H′,J). Double-labeling with a Ngn2probe and anti-Phox2b antibodies showed that most cells expressingNgn2 ectopically also expressed Phox2b indicating thatPhox2b functions cell-autonomously(Fig. 1I). Cells in the dorsalmost region of the neural tube consistently failed to respond toPhox2b transfection with Ngn2 induction, in line with our observation that Phox2b was also unable to induce early postmitotic markers at this location (Dubreuil et al.,2000). Ngn1 expression was upregulated by Phox2bat 12 h.a.e. but not at 6 h.a.e. (not shown).
The failure to induce Cash1 could be explained by the requirement for a co-factor present in bm/vm progenitors, but absent in the dorsal spinal cord. Reasoning that Nkx2.2, the expression of which in the hindbrain bm/vm progenitors coincides precisely with that of Phox2b(Pattyn et al., 2000), may be the missing factor, we co-transfected Nkx2.2 together withPhox2b. This resulted in a marked induction of Cash1throughout the transfected area at 20 but not at 6 h.a.e.(Fig. 2A,A′,B,B′).Nkx2.2 expressed alone at the same dose did not induce Cash1(Fig. 2C,C′). At higher concentrations, Nkx2.2 transfected alone also promoted Cash1expression, but much more weakly than in combination with Phox2b(Fig. 2D,D′). Hence, at physiological expression levels, the combinatorial action of Nkx2.2and Phox2b may be required to induce or maintain expression ofMash1. These results raised the possibility that co-expression ofPhox2b and Nkx2.2 may also promote Ngn2 expression. However, Ngn2 expression was repressed rather than activated by transfecting Phox2b together with Nkx2.2(Fig. 2B″), which can be attributed to the negative regulation of Ngn2 in the spinal cord(Zhou et al., 2001).
Previously, we have shown that Phox2b misexpression in spinal regions of the neural tube, where it is normally never expressed, promotes the generation of neurons that migrate to the ML and induces aspects of a bm/vm phenotype (Dubreuil et al.,2000). This holds true also at hindbrain levels of the neuraxis,where Phox2b-expressing vm neurons are born at this stage. At 48 h.a.e in the caudal hindbrain, most ectopically Phox2b-expressing cells had relocated to the ML, an effect not seen after transfectingGFP alone (Fig. 3A-B′). Most of them expressed Islet1/2(Fig. 3C-C″) and the bm/vm marker Phox2a, as do bm/vm neurons, but neither Islet2nor Hb9, which are specific for sm neurons (not shown). Forced expression of Phox2b also very efficiently induced the endogenousPhox2b gene (Fig. 3B′), as it does at spinal levels (seeFig. 5B″). We then examined whether ectopic Phox2b would also upregulated Ngn2when expressed alone and Cash1 when expressed together withNkx2.2. Misexpression of Phox2b in the caudal hindbrain resulted in dorsal expansion of Ngn2 expression at 16 h.a.e., whileCash1 was not induced (Fig. 3D-E′). By contrast, Cash1 was induced by cotransfecting Phox2b together with Nkx2.2(Fig. 3F-F′), but not after transfecting Nkx2.2 alone at the same concentration(Fig. 3G-G′). Clearly,then, the effects of Phox2b misexpression on pan-neuronal and type-specific differentiation can also be observed at rostrocaudal levels of the neural tube, where Phox2b is expressed and known to be required for the specification of vm progenitors.
These results suggest that in its normal expression territory in the ventral hindbrain, Phox2b promotes neurogenesis by upregulatingMash1 in combination with Nkx2.2. When misexpressed in more dorsal regions of the neural tube, it appears to do so by upregulatingNgn2 in cooperation with as yet unknown factors.
Phox2b inhibits expression of negative regulators of neurogenesis
We next examined if Phox2b affects the expression of genes known to play a role as negative regulators of neuronal cell cycle exit and differentiation. Among such negative regulators, we focused on the Hes and Id genes, which function as effectors of Notch signaling and as natural inhibitors of bHLH factor activity, respectively(Kageyama and Nakanishi, 1997;Kageyama and Ohtsuka, 1999;Norton, 2000).
Two members of the Hes family have been reported to be expressed in the chick neural tube: Hairy1 and Hairy2(Jouve et al., 2000). In the early chick spinal cord, Hairy1 is expressed in a narrow dorsal stripe and the floor plate (Jouve et al.,2000) and thus unlikely to function as negative regulator of neurogenesis in the lateral neural tube. Strong expression of Hairy2is confined to cells adjacent to the floor plate. Weaker expression is found throughout the alar plate (Jouve et al.,2000), which was not affected by misexpression of Phox2b(not shown). More recently, additional chick homologs of Hes genes have been identified, which are most similar by sequence to mouse Hes5 (D. Henrique, personal communication). Among them, chick Hes5b is expressed in a broad lateral region of the neural tube (C. J., O. Pourquié and D. Henrique, unpublished; seeFig. 4L′,N′). Like mouse Hes5 (Kageyama and Nakanishi, 1997), chick Hes5b responds to activated Notch and is upregulated in the chicken neural tube by expression of a constitutively active Notch construct (D. Henrique, personal communication). Hes5 also behaved as a negative regulator of neurogenesis in the chick neural tube. Overexpression of mouse Hes5,which we used to distinguish enforced from endogenous expression,downregulated expression of the early postmitotic markers Delta1 andNeuroM (Fig. 4C-D′). We then investigated whether Phox2b affected its expression. Six hours after misexpression of Phox2b, we already detected a decrease in chick Hes5b expression in the transfected area in a sizeable fraction of the embryos (Fig. 4K,K′). At 24 h.a.e., chick Hes5b expression was virtually extinguished on the transfected side of all embryos(Fig. 4L,L′).
Among the Id family members, we focused on Id2, the overexpression of which has been found to cause overgrowth of the chick neural tube (Martinsen and Bronner-Fraser,1998). We found Id2 expression to be very dynamic over the time period studied. Strong expression at early stages and in caudal regions, where development is less advanced, was confined to the dorsal-most neural tube with weak expression more ventrally. Upon further development,expression became strong in a broad lateral domain(Fig. 4K″,L″,N″). Id2 overexpression had little effect on expression of Delta1(Fig. 4G,G′), but inhibited neuronal differentiation as judged from downregulation ofNeuroM (Fig. 4H,H′). As Delta1 precedes NeuroM in postmitotic cells, this result may be taken to mean that Id2 affects neurogenesis mainly at a step downstream of Delta1 expression. As in the case of chick Hes5b, a slight decrease in Id2 expression at 6 h.a.e. of Phox2b foreshadowed a massive repression at 24 h.a.e.(Fig. 4K,K″,L,L″).
Our previous results have shown that ectopic Phox2b expression promotes cell cycle exit (Dubreuil et al.,2000). It was thus possible that repression of chickHes5b and Id2, which are expressed mainly in VZ progenitors,was a mere consequence of cell cycle exit. Two types of evidence argue against this explanation: (1) at 6 h.a.e., when we detected the first signs of repression, equivalent numbers of BrdU-incorporating cells were present on the transfected and the non-transfected side (not shown); and (2) double-labeling for BrdU incorporation and chick Hes5b or Id2 expression at 18 h.a.e. showed that either gene was downregulated both in cells that had incorporated BrdU and were thus in the S-phase of the cell cycle and in BrdU-negative cells, which are a mixture of postmitotic precursors and of progenitors in other phases of the cycle(Fig. 4M). Hence,downregulation of chick Hes5b and Id2 appears to precede withdrawal from the cell cycle.
To provide evidence that inhibition of Hes and Id genes plays a role in mediating the neurogenesis-promoting activity of Phox2b, we co-expressedHes5 or Id2 together with Phox2b. Co-transfection of Hes5 antagonized the increase in the number ofDeltal-expressing cells observed after misexpression ofPhox2b (Fig. 4E,E′). Quantification of the results showed that in the presence of mouse Hes5, the increase in Delta1+cells seen after transfecting Phox2b alone was reduced by around 60%(Fig. 4F). Similarly,Id2, when co-transfected with Phox2b, consistently prevented the increase in NeuroM+ cells caused by Phox2b;the effect on Delta1 expression was more variable, but on most sections, Delta1 expression was not increased in response to Phox2b(Fig. 4I-J′).
Finally, we tested the idea that downregulation of chick Hes5b andId2 in response to Phox2b may be a consequence of increasedNgn2 expression. Forced expression of Ngn2, however,resulted in a marked upregulation of the chick Hes5b and Id2expression levels (Fig. 4N-N″). Although induction of chick Hes5b probably reflects activation of the Notch pathway, the mechanism by which Ngn2promotes Id gene expression remains undefined.
Together, these results show that, in addition to activating proneural genes, Phox2b inhibits expression of negative regulators of neuronal differentiation by a separate pathway. They also raise the possibility that downregulation of Hes5 and Id2 may be essential components of the response to Phox2b. As Hes5 expression can be taken as a read-out for Notch activity (de la Pompa et al., 1997; Ohtsuka et al.,1999), Phox2b may affect expression of any of the components involved in this signaling pathway.
Regulatory interactions between Phox2b and transcription factors expressed in the progenitor domain
We then explored the transcriptional regulations by which Phox2b specifies a bm/vm phenotype. There is now a large body of evidence to suggest that, in addition to driving generic neuronal differentiation, the bHLH factors with proneural activity play key roles in the specification of neuronal subtype identity (Perez et al., 1999;Fode et al., 2000;Mizuguchi et al., 2001;Novitch et al., 2001;Scardigli et al., 2001;Lo et al., 2002;Parras et al., 2002). We therefore tested whether the activation of a bm/vm phenotype in response to Phox2b could be mediated by Ngn2 or Mash1. Overexpression of Ngn2,although promoting premature neurogenesis, did not induce bm/vm markers (not shown). Misexpression of Mash1 resulted in the appearance of ectopic Islet1+, Islet2- cells in the dorsolateral spinal cord(Fig. 5A-A″). However,neither ectopic expression of the motoneuronal marker ChAT(Fig. 5A′′′)nor of Phox2a (not shown) could be detected, two genes that are induced by Phox2b (Dubreuil et al., 2000). In some embryos, we observed ectopic expression of chicken Phox2b in the transfected area, but this effect was slight and inconsistent. We thus conclude that neither Ngn2 nor Mash1 mediates the effect of Phox2b on bm/vm differentiation.
The current model of neuronal subtype specification in the spinal cord posits that the VZ is parcellated into different domains, each expressing a different set of transcription factors that cross-repress each other and ensure that only the appropriate type of neurons arises from each domain(Briscoe et al., 2000;Muhr et al., 2001;Novitch et al., 2001;Lee and Pfaff, 2001). Changing the transcriptional code of the progenitor domains may thus be a prerequisite for inducing ectopic bm/vm neurons. In the ventral hindbrain, the expression patterns of Phox2b and Nkx2.2 are co-extensive(Pattyn et al., 2000), raising the possibility that Nkx2.2 may be involved in the fate-specifying activity of Phox2b. However, Phox2b misexpression did not yield ectopicNkx2.2+ cells (Fig. 5B,B′).
We then tested whether Phox2b affects the expression ofNkx6.1, Nkx6.2, Pax6 and Olig2. Among them, Nkx6.1and Nkx6.2 are expressed by hindbrain bm/vm progenitors(Qiu et al., 1998;Briscoe et al., 1999;Cai et al., 1999) and are required for their proper development (J. Ericson and M. Sander, personal communication), albeit the Nkx6.1/6.2 expression domains also encompasses the progenitors of sm and V2 neurons. In both, mouse and chick, Nkx6.2 is co-expressed with Nkx6.1 in bm/vm progenitors, while in the spinal cord, the expression patterns of both genes are co-extensive in chick but not in mouse embryos (Cai et al.,1999; Vallstedt et al.,2001). The ventral limits of Pax6 and Olig2expression define the dorsal boundary of the bm/vm progenitor domain(Ericson et al., 1997;Osumi et al., 1997;Mizuguchi et al., 2001). In the absence of Pax6, hindbrain sm neurons appear to acquire a vm identity (Ericson et al.,1997; Osumi et al.,1997). Nkx6.1 and Nkx6.2 expression should thus favor a bm/vm fate, whereas Pax6 and Olig2 should suppress it. Forced expression of Phox2b resulted in dorsal expansion of the domains of Nkx6.1 and Nkx6.2 expression whereasPax6 and Olig2 were repressed(Fig. 5C-D″). These data show that Phox2b changes the pattern of transcription factor expression in the progenitor domains and suggest that to be able to initiate bm/vm differentiation, Phox2b needs to downregulate progenitor factors that are not permissive for this fate. They also suggest that Phox2b may cooperate with Nkx6.1 and Nkx6.2 in specifying a bm/vm identity.
An activator form of Phox2b mimics the auto-regulatory and neurogenesis-promoting activities of Phox2b and the induction of Islet1
Only activating functions have so far been ascribed to Phox2b(Yang et al., 1998;Lo et al., 1999;Yokoyama et al., 1999;Flora et al., 2001). To investigate whether Phox2b functions as an activator when promoting neurogenesis and bm/vm differentiation, we fused the Phox2a HD either to the trans-activating domain of the viral protein VP16 (PHDVP16) or to the repressor domain of Drosophila Engrailed (PHDEnR). The Phox2a HD could be used in place of the Phox2b HD, as the two HDs are identical at the amino acid level (Pattyn et al., 1997). In a previous study, the PHDEnR construct has been shown to behave as a dominant-negative form of Phox2a/b in neural crest cells (Lo et al., 1999). Similar fusions between the HD of Siamois, which is also a member of the paired homeogene family(Galliot et al., 1999), and the activator or repressor domains were used as controls (SHDVP16 andSHDEnR).
Expression of PHDVP16 mimicked the ability of Phox2b to elicit emigration to the ML and ectopic Islet1/2+ cells, albeit the effect on Islet1/2 expression was less pronounced. By contrast, expression ofPHDEnR had no such effects (Fig. 6A-B′). Similarly, PHDVP16 expression, but not that of PHDEnR or SHDVP16, increased the number ofNgn2+ cells within the transfected area at 6 h.a.e., when the effect of Phox2b on Ngn2 expression was greatest(Fig. 6C,C′,E-F′).Ngn1 expression was not yet upregulated at 6 h.a.e. (not shown), but was so at 20 h.a.e. (Fig. 6D,D′). None of these effects was observed after expression of the HD alone (not shown). Positive autoregulation of Phox2b and upregulation of Nkx6.1 also appeared to reflect an activator function, as PHDVP16 transfection resulted in induction of the endogenous Phox2b gene and in dorsal expansion of Nkx6.1expression (Fig. 6G-H′),whereas PHDEnR had no effect on Phox2b and repressedNkx6.1 (Fig. 6I-J′). Together, these results suggest that Phox2b acts as a transcriptional activator in promoting neurogenesis and a bm/vm fate.
Activation of the endogenous Phox2b gene potentially complicates the interpretation of the PHDVP16 phenotype, which could be ascribed to wild-type Phox2b. However, conversion of a transcriptional repressor to an activator, or the converse, generates a dominant-negative or antimorphic form that antagonizes the endogenous protein(Onichtchouk et al., 1998). In principle, then, if wild-type Phox2b acted as a repressor, PHDVP16 should block and not phenocopy its activity.
In contrast to the foregoing results, transfection of both PHDVP16and PHDEnR (but not of SHDVP16) reduced chick Hes5b and Id2 expression, an effect also observed after expression of the HD alone (Fig. 7). These results suggest that inhibition of Hes and Id genes by full-length Phox2b may be mediated by the HD, and that the function of the Phox2 HD is dominant over heterologous repressor and activator regions. Equivalent results have been obtained for the HD transcription factor Xdbx(Gershon et al., 2000). Also in this case, the isolated HD and activator and repressor fusion constructs mimicked the repressive activity of the full-length protein. One possibility is that the Phox2 HD binds to and sequesters essential co-activators required for chick Hes5b and Id2 expression. A similar mechanism has been shown to operate in the case of the inhibition of astrocyte differentiation by Ngn1 (Sun et al.,2001). Likewise, the Pou domain transcription factor Pit1 has been found to function as a DNA-binding dependent activator and a DNA binding-independent repressor (Scully and Rosenfeld, 2002). Clearly, further work is required to elucidate the mechanism by which Phox2b represses the expression of these negative regulators of neurogenesis.
DISCUSSION
Our previous loss-of-function (LOF) and gain-of-function (GOF) studies have provided evidence that the HD transcription factor Phox2b coordinates generic and subtype-specific aspects of cranial motoneuronal differentiation(Dubreuil et al., 2000). Here,we have used a GOF approach to explore the underlying molecular interactions. Phox2b appears to drive cell cycle exit and pan-neuronal differentiation by promoting expression of bHLH genes with proneural activity and relieving them from the negative influences exerted by Notch signaling and the Id2 repressor. In addition, Phox2b elicits changes in the expression of transcription factors in the progenitor domain that may underlie its capacity to initiate a bm/vm neuronal differentiation program. These results provide insights into the mechanisms by which a single transcription factor collaborates with other factors to coordinately regulate generic and type-specific neuronal differentiation. Fig. 8A,Bpresent a schematic view of the postulated genetic interactions that underlie these activities, in bm/vm progenitors and upon ectopic expression of Phox2b in the dorsolateral spinal cord or caudal hindbrain, respectively. Our GOF approach also reveals a substantial degree of redundancy built into the system, which could not have been detected by LOF studies alone.
Phox2b upregulates proneural genes and represses negative regulators of neurogenesis
One function of Phox2b, whether misexpressed in the spinal cord or in its normal expression domain in the ventral hindbrain, is to increase the probability of cell cycle exit and to activate pan-neuronal markers(Dubreuil et al., 2000). Our present data strongly suggest that this activity is mediated in part by proneural bHLH factors. Phox2b misexpression expanded the expression domain of Ngn2, which has been shown to promote generic neuronal differentiation in a wide variety of settings, including the chick neural tube(Mizuguchi et al., 2001;Novitch et al., 2001). Moreover, in combination with Nkx2.2 it induced Cash1, which is normally co-expressed with Phox2b in bm/vm progenitors, is downregulated in the absence of Phox2b(Pattyn et al., 2000) and(like Ngn2) capable of driving neuronal differentiation in the chick neural tube. Another result supporting the view that up-regulation of proneural genes is required for premature neurogenesis in response toPhox2b is that co-expression of Id2, a natural inhibitor of bHLH factor activity, greatly attenuates the effect of Phox2b on generic neuronal differentiation. There is now ample evidence suggesting thatMash1 and Ngn2, in addition to their proneural function,participate in neuronal subtype specification(Fode et al., 2000;Scardigli et al., 2001;Lo et al., 2002). The fact that Phox2b directs bm/vm differentiation, whether in combination with Mash1 in its endogenous expression domain or with Ngn2after forced expression in the spinal cord, suggests that all that Ngn2 and Mash1 provide in this context is their proneural function.
Phox2b not only upregulates expression of the Ngn2 orMash1 proneural genes, but also represses inhibitors of neurogenesis such as Hes or Id family members(Ishibashi et al., 1995;Lyden et al., 1999;Ohtsuka et al., 1999;Cai et al., 2000;Cau et al., 2000). As overexpressing Ngn2 increased rather than decreased chickHes5b and Id2 expression, downregulation of these genes cannot be a consequence of Ngn2 induction and probably occurs by a separate pathway. Several lines of evidence support the idea that downregulation of chick Hes5b and Id2 plays a causal role in the neurogenesis-promoting function of Phox2b. Forced expression ofPhox2b results in rapid repression of chick Hesb5 andId2, independently of cell cycle stage, suggesting that their downregulation precedes the last S phase. Furthermore, constitutive expression of either gene counteracts the premature production of neurons caused byPhox2b misexpression.
In the Phox2b knockout mice, Hes5 expression is downregulated rather than stimulated(Dubreuil et al., 2000) and there is no change in the expression of Id family members (data not shown). Downregulation of Mash1 and Delta and the ensuing reduced activity of the Notch pathway most probably accounts for reduced Hes5expression. In the light of our evidence that Ngn2 stimulatesId2 expression, reduced proneural activity may also explain whyId expression is unchanged in the absence of Phox2b. Hence, a sustained lack of Phox2b activity as occurs in knock out mice may induce feedback loops that prevent upregulation of Hes or Id genes and this may explain why some Phox2b-/- progenitors are still able to exit the cell cycle.
Finally, the question should be asked why forced expression ofNgn2 or Mash1 are able to promote neuronal differentiation in the presence of endogenous chick Hes5b and Id2. First, Hes5, which downregulates expression of proneural genes in addition to counteracting their activity (de la Pompa et al.,1997; Kondo and Raff,2000; Ohtsuka et al.,1999), cannot repress expression of the transfected genes driven by artificial promoters. Second, high-level expression of the transgenes as achieved by transfection may titrate out the repressor molecules. Hence,forced expression of the proneural bHLH factors may be sufficient to overcome the influence of these negative regulators of neurogenesis, while counteracting them by transcription factors such as Phox2b may be required in a physiological setting.
Phox2b changes the pattern of transcription factor expression in the progenitor domain
In the ventral hindbrain, Phox2b is expressed both by cycling bm/vm progenitors and by their postmitotic descendants. Our data argue that during its neuroepithelial phase of expression, Phox2b acts not only to promote cell cycle exit but also as a patterning gene, controlling the identity and fate of dividing progenitors. Pax6 and Olig2expression defines the dorsal limit of the Nkx2.2+,Phox2b+ bm/vm progenitor domain(Ericson et al., 1997;Briscoe et al., 1999;Mizuguchi et al., 2001). In the absence of Pax6, spinal interneurons and motoneurons do not properly develop, and in the caudal hindbrain, sm neurons appear to switch to a vm fate(Ericson et al., 1997). Olig2 directs a spinal motoneuronal fate and elicits expression of downstream factors appropriate for this fate(Mizuguchi et al., 2001;Novitch et al., 2001). According to the derepression model of spinal cord neurogenesis(Muhr et al., 2001;Lee and Pfaff, 2001), both factors should thus suppress a bm/vm fate. The pronounced downregulation of the two factors by Phox2b misexpression may thus be a necessary step in the chain of events that result in the ectopic induction of bm/vm markers.Pax6 and Phox2b appear to maintain cross-inhibitory interactions because, in the absence of Pax6, the Phox2b expression domain in the ventral hindbrain expands dorsally(Mizuguchi et al., 2001). One consequence of Phox2b overexpression is premature neuronal differentiation, which alone might explain the decrease in progenitor factor expression. We argue against this explanation, as Phox2b does not repress Ngn2 and synergizes with Nkx2.2 to activateCash1, two genes that in the spinal cord are restricted to progenitor domains. Because Olig2 expression depends on Pax6(Mizuguchi et al., 2001;Novitch et al., 2001), its downregulation may be a consequence of decreased Pax6 expression, but may also occur by a separate pathway.
Nkx6.1 and Nkx6.2 are co-expressed with Phox2bin the bm/vm progenitors (Qiu et al.,1998) and are required for their proper development (J. Ericson and M. Sander, personal communication). In line with this, Phox2bmisexpression results in dorsal expansion of their expression domains. These HD factors may thus cooperate with Phox2b in the implementation of a bm/vm phenotype both in the ventral hindbrain and after misexpression in the spinal cord. However, Nkx6.1 is expressed throughout the ventral third of the neural tube and, when ectopically expressed, directs sm and V2 neuronal fates (Briscoe et al., 2000). This shows that by itself, Nkx6.1 does not induce a bm/vm fate and that Phox2b activity is required to achieve this.
Phox2b acts as an activator in inducing Ngn2 and cranial motoneuronal markers
The available evidence suggests that Phox2b functions as a transcriptional activator rather than as a repressor, in line with the fact that it lacks an EH1 domain that confers repressor activity to other HD proteins(Muhr et al., 2001). First,Phox2b binds to and transactivates the promoters of the dopamineβ-hydroxylase and Phox2a genes(Yang et al., 1998;Yokoyama et al., 1999;Adachi et al., 2000;Hong et al., 2001;Flora et al., 2001). Second,the PHDEnR construct, which should act as a repressor, prevents induction of tyrosine hydroxylase and dopamine β-hydroxylase by BMP2 in neural crest cells (Lo et al.,1999). We show that PHDVP16 but not PHDEnRmimics the ectopic induction of Ngn2 and Islet1 in the chick spinal cord, suggesting that Phox2b functions as an activator.
Phox2b thus appears to provide an activator function, which is necessary and sufficient for the initiation of bm/vm differentiation. According to the derepression model of neuronal cell type specification in the spinal cord(Muhr et al., 2001;Vallstedt et al., 2001), the factors that activate expression of downstream determinants of neuronal identity are thought to be expressed in the neural tube in a topologically unrestricted manner. Our results suggest that such transcriptional activators can also be deployed in spatially restricted domains.
Functional redundancy among neural HD genes
When misexpressed in spinal regions, Phox2b affects the expression of Pax6, Olig2, Nkx6.1 and Nkx6.2. Nevertheless, expression of these four genes is not detectably affected in the hindbrain ofPhox2b mutant mice (data not shown), suggesting that other genes are redundant with Phox2b in this capacity. In the case of Pax6and Olig2, Nkx2.2, the expression of which does not depend onPhox2b and which in GOF experiments represses both genes(Muhr et al., 2001;Novitch et al., 2001), is probably sufficient to exclude Pax6 and Olig2 from bm/vm progenitors in the absence of Phox2b. Conversely, despite the fact thatNkx2.2 has the capacity to repress Pax6, lack of Nkx2.2 does not lead to ventral expansion of the Pax6 domain(Briscoe et al., 1999). In the hindbrain, this may be due to functional redundancy with Nkx2.9, but on the basis of our results this may also be due to the presence of Phox2b. Likewise, bm/vm neurons appear to develop normally in the absence of Nkx2.2(Briscoe et al., 1999), which again may be attributed to the persistent expression of either Nkx2.9or Phox2b.
It has been argued that redundant roles have been selected as backup against developmental error, whereby one of the genes fails to be expressed adequately at the appropriate site, thus helping to maintain the spatiotemporal precision of embryonic patterning(Cooke et al., 1997). Not surprisingly, the complex process of neuronal patterning appears to provide striking examples of this. The alternative possibility is that the redundant roles we observe are selected because they are non-redundant at other expression sites.
Synchronizing neuronal fate determination with timing and extent of neurogenesis
Neurogenesis involves the parallel activation of a program that controls commitment to cell cycle exit and generic neuronal differentiation and of a program that specifies the identity of the neurons to be generated. Recent studies suggest that one way these two programs are coordinated is by the type-specification properties of bHLH genes themselves. Examples are the bHLH transcription factors Math1 and Ngn1, the ectopic expression of which in the neural tube both drives the cells to move to the ML and fosters production of distinct classes of interneurons (Gowan et al., 2001), and Mash1, the misexpression of which in the dorsal forebrain and ventral spinal cord promotes the appearance of ventral cell types (Fode et al., 2000) and V2 interneurons (Parras et al.,2002), respectively. Another example is provided by Olig2expressed in spinal motoneuron progenitors(Mizuguchi et al., 2001;Novitch et al., 2001). When misexpressed in the spinal cord or hindbrain, it drives cells to exit the cell cycle and to migrate to the ML and, in addition, directs expression of determinants of sm or V2 neuronal fates. Furthermore, Olig2misexpression expands the domain of Ngn2 expression suggesting that Olig2, like Phox2b, promotes generic neurogenesis by boosting the expression of downstream proneural genes. Hence, Olig2 and Phox2b may have comparable roles in sm and bm/vm progenitors, respectively. However, one important mechanistic difference between Phox2b and Olig2 is that Phox2b acts as an activator in inducing Ngn2 and motoneuronal markers, whereas Olig2 functions as a transcriptional repressor. (Whether Olig2 also represses negative regulators of neuronal differentiation has not been examined.) It will be interesting to learn which transcription factors play this role in other neuronal lineages and how they work.
Acknowledgements
We thank the following for reagents: D. Anderson, M. Ballivet, M. Bronner-Fraser, J. Ericson, F. Guillemot, T. Jessell, R. Kageyama, P. Lemaire,T. Ogura, O. Pourquié and J. Rubenstein. This work was supported by institutional grants from CNRS, Université de la Méditerranée and Ecole Normale Supérieure; and by specific grants from the European Community (QLGT-CT-2001-01467), the Association pour la Recherche sur le Cancer, the Association Française contre les Myopathies and the Ministère de la Recherche.