Domain-specific control of neurogenesis achieved through patterned regulation of Notch ligand expression

The establishment of the central nervous system (CNS) depends on the generation of functionally distinct neurons produced in appropriate numbers and at correct positions. In the developing spinal cord, distinct classes of neurons are produced at specific locations along the dorsoventral (DV) axis (Jessell, 2000; Poh et al., 2002). These neurons include motoneurons (MNs) and various types of interneurons, which originate from progenitor cells organized in specific domains along the DV axis. Genetic gain- and loss-of-function experiments have revealed that the establishment of DV domains, and neurogenesis within these domains, is influenced by homeodomain (HD) transcription factors, proneural basic helix-loop-helix (bHLH) proteins and the Notch signalling pathway. However, the internal relationship between these sets of regulatory pathways is currently incompletely understood.

HD transcription factors are expressed in distinct patterns along the DV axis in the developing neural tube. Their expression is confined to neural progenitors and is controlled by a ventral-to-dorsal graded activity of Sonic Hedgehog signalling (Jessell, 2000). The combinatorial expression of these HD proteins subdivides the spinal cord into a number of distinct progenitor domains, which each gives rise to specific subtypes of postmitotic neurons (Jessell, 2000). Cross-repressive interactions between complementary pairs of HD proteins further refine and sharpen the boundaries of these progenitor domains (Briscoe et al., 2000; Muhr et al., 2001). For instance, cross-repression between the HD transcription factor Dbx2 and Nkx6.1 leads to the establishment of a sharp progenitor domain boundary between the p2 and p1 domains (Briscoe et al., 2000; Sander et al., 2000; Vallstedt et al., 2001). Genetic loss- and gain-of-function experiments have revealed the importance of HD proteins for the specification of neuronal identities (Briscoe et al., 2000; Jessell, 2000; Pierani et al., 2001). However, the precision by which neuronal subtypes are generated during CNS development implies that mechanisms controlling cell type and cell number must be stringently coordinated. Indeed, the transcription factors Olig2 and Lmx1a not only determine the fate of somatic MNs and midbrain dopamine neurons, respectively, but also influence the rate of neurogenesis by promoting expression of the proneural bHLH protein Ngn2 in ventral progenitor cells (Andersson et al., 2006; Dubreuil et al., 2002; Lee et al., 2005; Mizuguchi et al., 2001; Novitch et al., 2001; Scardigli et al., 2003; Scardigli et al., 2001; Sugimori et al., 2007). Thus, one regulatory means to control neuronal subtype identities and cell numbers appears to involve a regulatory interaction between cell fate-determining transcription factors and proneural bHLH proteins (Scardigli et al., 2003).

Like HD proteins, components of the Notch pathway (Dll1, Jag1 and Fng proteins) are expressed in distinct patterns along the DV axis in the developing spinal cord (Henrique et al., 1995; Johnston et al., 1997; Lindsell et al., 1996; Myat et al., 1996; Sakamoto et al., 1997). Notch signalling is a cell-cell communication system and activation of Notch signalling results from the interaction between the Notch receptor and its ligands, Delta and Jagged, expressed by adjacent cells. The ligand receptor interaction induces a proteolytic cleavage of the Notch protein and nuclear translocation of its intracellular domain (Notch ICD), which interacts with the DNA-binding protein CSL (RBP-Jk) to activate Notch downstream genes, such as the Hey and Hes genes. These proteins have been demonstrated to, at least in part, mediate their functions by repressing proneural proteins, such as Ngn2 (Louvi and Artavanis-Tsakonas, 2006). Also, Notch receptors are modulated by the Fringe glycosyltranserases, which extend O-glycosylation carried out by the O-fucosyltransferases (O-fut1 in Drosophila and Pofut1 in mammals) (Stanley, 2007). Fringe-mediated glycosylation modulates the ability of the Notch receptor to respond to the ligands. In many situations, Fng enhances Dll1-Notch interactions while reducing Jag1-Notch interactions (Hicks et al., 2000; Moloney et al., 2000; Shimizu et al., 2001; Xu et al., 2007). Despite these findings, we do not understand the roles and functional relationship of Notch ligands and Fng proteins in regulating neurogenesis and neuronal subtype specification along the DV axis of the neural tube. Moreover, the functional relevance of their patterned distribution along the DV axis of the neural tube has not been resolved.

Given that both HD proteins and Notch signalling are important for the control of CNS development, and that both exhibit distinct expression domains in the developing spinal cord, we have in this report addressed the relationship between HD proteins and Notch signalling. We provide evidence that HD proteins regulate the spatial distribution of Notch ligands and Fng proteins in distinct DV progenitor domains. Furthermore, we demonstrate that loss of Dll1 or Jag1 signalling leads to a domain-specific increase of neuronal differentiation but does not affect the establishment of progenitor domain boundaries. Our data also suggest that signalling between Dll1⁺ and Jag1⁺ domains is prohibited along the DV axis of the neural tube, which is in contrast to the situation in zebrafish rhombomeres and the Drosophila wing, where Notch signalling is enhanced at boundaries between domains (Wu and Rao, 1999). We discuss a model in which the different configurations of Fng expression with regard to ligand expression may explain the observed differences in the outcome of Notch signalling in different tissues. Moreover, we show how this expression pattern might allow for quantitative different levels of Notch signalling in distinct progenitor domains, as well as prevent inadequate activation of Notch signalling across boundaries between adjacent progenitor cell populations. Thus, this mechanism provides a means to individually control the numbers of specific neurons produced by neighbouring progenitor domains along the DV axis of the neural tube.

cDNA encoding chick Nkx6.1 and mouse Nkx2.2 were misexpressed using the retroviral vector RCASBP(B), and mouse Dll1, Jag1, Dbx1 and Mfng and chick Olig2 were misexpressed using the CMVβ/actin-based pCAGGS vector. DNA constructs were electroporated into the neural tube of Hamburger-Hamilton (HH) stage 10-11 chick embryos (Briscoe et al., 2000). Embryos were harvested and fixed 30-48 hours post-transfection (hpt) and processed for immunohistochemistry or in situ hybridization.

Immunohistochemistry was performed as previously described (Briscoe et al., 2000). The following antibodies were used: rat anti-Jag1 (Developmental Studies Hybridoma Bank), rat anti-PDGRα (BD Pharmingen); mouse anti-Pax7, anti-Nkx2.2, anti-Evx1/2 and anti-En1 (Developmental Studies Hybridoma Bank), mouse anti-GATA3 (Santa Cruz, sc9009), mouse anti-Ngn2 (kind gift from D. Anderson, California Institute of Technology, CA, USA) mouse anti-HuC/D (Molecular Probes), mouse anti-Brn3a (Chemicon); rabbit anti-cleaved caspase-3 (Cell Signaling), rabbit anti-Dbx1 and anti-Dbx2 (kind gift from A. Pierani, Institut Jacques Monod, Paris, France), rabbit anti-Pax2 (Nordic Biosite), rabbit anti-Nkx6.1 (Briscoe et al., 2000), rabbit anti-Jag1 (Santa Cruz), rabbit anti-Gsh1/2 (kind gift from M. Goulding, The Salk Institute, CA, USA), rabbit anti-Chx10 (Ericson et al., 1997), rabbit anti-Dll1 (Alexis Biochemicals), rabbit anti-Nestin (BD Biosciences Pharmingen), rabbit anti-Sox3 (kind gift from T. Edlund, Umeå University, Sweden); guinea pig anti-Isl1/2 (Tsuchida et al., 1994), guinea pig anti-Olig2 (kind gift from T. Jessell, Colombia University, NY, USA); goat anti-β-gal (Biogenesis), goat anti-Sox10 (Santa Cruz, N20). In situ hybridization was performed essentially as described (Tsuchida et al., 1994), using probes for mouse Jag1 and Dll1 (Lindsell et al., 1996), Lfng (IMAGE4000728), Mfng (IMAGE575363), Hes5 (Akazawa et al., 1992), Dbx1 (Invitrogen IRAVp968a07126d) and Olig2 (kind gift from B. Novitch, University of Michigan, MI, USA), and chick Jag1 and Delta1 (isolated from cDNA libary), Lfng (kind gift from C. Tabin, Harvard Medical School, MA, USA), Mfng (Geneservice ChEST679I3), Nkx6.1 (Briscoe et al., 2000), Dbx1 and Dbx2 (Pierani et al., 2001), and Gsh1 (Geneservice ChEST1010i7).

The generation of Dbx1 and Nkx6.1 mutant mice was previously reported (Pierani et al., 2001; Sander et al., 2000). Jagged1^Ndr/Ndr (Nodder) mice were identified in an ENU screen at Ingenium, Germany. These mice carry a single base-pair mutation in the extracellular part of jagged 1. The name Nodder reflects the nodding behaviour and balance defects in the heterozygous state. Mice carrying the Dll1^lacZ knock-in allele have been described previously (Hrabe de Angelis et al., 1997). Human embryos (5.5 weeks of gestation) were collected after elective routine abortions with consent given by the pregnant woman. The tissue collection was approved by the Regional Human Ethics Committee, Stockholm.

Student's t-test comparisons were performed for all statistical examinations. Bars indicate means±s.d. (standard deviation). Significance for the tests were assumed at the level of P<0.05, n=4 (*P<0.05, 0.01>**P>0.001, ***P<0.001).

To examine the role of Dll1 and Jag1 in the developing spinal cord, we applied in situ hybridization and immunohistochemistry to map the regional expression of Dll1, Jag1, Lunatic Fringe (Lfng) and Manic Fringe (Mfng) in relation to HD proteins defining different DV progenitor domains (Jessell, 2000; Poh et al., 2002). In HH stage 20 chick embryos, Dll1, Lfng and Mfng had similar expression patterns: the genes were broadly expressed in the ventral and dorsal thirds of the neural tube, and in a narrow intermediate stripe (Fig. 1A,B,D; data not shown) (Johnston et al., 1997; Lindsell et al., 1996). The ventral expression of Dll1, Lfng and Mfng coincided precisely with the expression domain of Nkx6.1 (Fig. 1E; data not shown), the intermediate stripe correlated to Dbx1 expression, whereas the dorsal expression mapped to the Gsh1/Gsh2 expression domain (Fig. 1G,H; data not shown). Jag1 expression was primarily confined to two narrow domains in the intermediate part of the spinal cord, where it was complementary to the expression of Dll1 (Fig. 1B,C,Q) (Lindsell et al., 1996). These Jag1⁺ cells correlated with the expression domain of Dbx2 but not Dbx1, Nkx6.1 or Gsh1/Gsh2 (Fig. 1E-H). Similar expression patterns were observed in the embryonic chick, mouse and human spinal cord (Fig. 1I-L; see also Fig. S1A-E,K,L in the supplementary material). Thus, the expression of Jag1, Dll1, Lfng and Mfng correlates precisely with defined DV progenitor domains (Fig. 1R), and this expression pattern appears to be conserved in higher vertebrates.

Different DV progenitor domains generate distinct neuronal subtypes that can be distinguished based on their expression of molecular markers (Helms and Johnson, 2003; Poh et al., 2002). Expression of Jag1 in cells with a Dbx2⁺/Dbx1⁻ identity (Fig. 1J,K) indicate that dorsal Jag1⁺ cells correspond to d6 progenitors generating dI6 neurons, whereas the ventral stripe delineates p1 progenitors giving rise to V1 neurons. In support for this, we could at embryonic day (E)10.5 in the mouse, detect dorsal Jag1⁺ cells that co-expressed Pax2 [expressed in all neurons derived from Dbx2⁺ cells (Matise and Joyner, 1997)] but not the dI5 marker Brn3a or the V0 marker Evx1/Evx2 (Fig. 1M,N; data not shown) (Pierani et al., 1999). Ventral Jag1⁺ cells co-expressed the V1 marker En1, but not markers for juxtaposed V0, V2a or V2b neurons (Evx1/Evx2, Chx10 or Gata3) (Fig. 1N-P; data not shown) (Matise and Joyner, 1997). Similar expression of Jag1 was detected also in the chick and human neural tube (see Fig. S1F-J,M-O in the supplementary material; data not shown). Thus, Jag1 is expressed in Dbx2⁺/Dbx1⁻ progenitors generating V1 and dI6 neurons, whereas expression of Dll1, Lfng and Mfng is confined to Dbx2⁺/Dbx1⁺ p0 progenitors generating V0 neurons and to ventral Nkx6.1⁺ progenitors producing V2 neurons and MNs.

The precise correlation between the expression of Dll1, Jag1 and Fng genes and specific DV progenitor domains implies that these genes might be regulated by HD proteins involved in DV pattern formation. However, because Notch signalling has been shown to influence boundary formation and the affinity properties of cells in certain tissues (Baek et al., 2006; Herranz and Milan, 2006), it is also possible that the Notch pathway has a reciprocal input on DV patterning.

To examine this, we analysed the expression of Jag1, Dll1, Lfng and Mfng in Nkx6.1 and Dbx1 loss- and gain-of-function experiments. We focused on these HD proteins because they confer functional identity to p2 and p0 progenitors expressing Dll1 and concomitantly suppress the establishment of p1 progenitors expressing Jag1 (Briscoe et al., 2000; Pierani et al., 2001; Sander et al., 2000). Also, repressive interactions between Nkx6 and Dbx class proteins are known to be essential for the establishment of p0, p1 and p2 progenitor domain boundaries (Briscoe et al., 2000; Muhr et al., 2001; Vallstedt et al., 2001). In chick in ovo electroporation experiments, forced expression of Nkx6.1 extinguished Dbx1, Dbx2 and Jag1 expression (Fig. 2A,B; data not shown) (Briscoe et al., 2000), and induced ectopic expression of Dll1, Lfng and Mfng, resulting in an uninterrupted expression of these genes in the intermediate neural tube (Fig. 2C,D; see also Fig. S2A in the supplementary material). Conversely, in Nkx6.1^−/− embryos at E10.5 (Sander et al., 2000), the Jag1⁺ domain was ventrally expanded into the p2 and pMN domains (Fig. 2I,J; Fig. S3 in the supplementary material), and the expression of Dll1, Lfng and Mfng was reduced (Fig. 2K-N; Fig. S2C,D in the supplementary material). However, Jag1 expression did not extend to the ventral midline, and expression of Dll1 and Lfng could still be detected in Nkx2.2⁺ p3 progenitors in Nkx6.1 mutants (Fig. 2I-N; see also Fig. S3 in the supplementary material), implying that factors in addition to Nkx6.1 influence expression of Notch ligands and Fng proteins in the ventral neural tube. Consistent with this, Ngn2 expressed in the pMN domain has been shown to directly regulate Dll1 expression (Castro et al., 2006), and we found that Nkx2.2 (see Fig. S3 in the supplementary material) and Olig2 (Novitch et al., 2001) could downregulate Jag1 expression when overexpressed in intermediate positions of the neural tube (data not shown).

Forced expression of Dbx1 did not affect the expression of Dbx2 (data not shown), but repressed the dorsal and ventral stripes of Jag1 expression (Fig. 2E,F). Also, similar to experiments with Nkx6.1 overexpression, the suppression of Jag1 was followed by a continuous expression of Dll1, Lfng and Mfng in the intermediate neural tube (Fig. 2G,H; Fig. S2B in the supplementary material). In Dbx1^−/− embryos (Pierani et al., 2001), by contrast, the intermediate stripe of Dll1, Mfng and Lfng expression was lost (Fig. 2Q-T; Fig. S2E,F in the supplementary material), and the expression of Jag1 encompassed the entire Dbx2⁺ domain (Fig. 2O,P). Together, these data establish that Nkx6.1 and Dbx1 control the DV expression profile of Notch ligands by suppressing the expression of Jag1 and promoting that of Dll1.

We next examined DV patterning in mice carrying loss-of-function mutations for Dll1 (Hrabe de Angelis et al., 1997) and Jag1 (Jag1^Ndr/Ndr; E.M.H. and U.L., unpublished). Apart from a slight enlargement of the floor plate in Dll1^−/− mutant embryos (Przemeck et al., 2003), the initial establishment of DV progenitor domains appeared unaffected in Dll1 and Jag1 mutants, as indicated by an unaffected expression patterning of HD proteins (Fig. 2Y) and a normal degree of intermingling of cells at progenitor domain boundaries (Fig. 2U,V,Z). Also, there was no change in the patterned expression of Jag1 in Dll1^−/− or of Dll1 in Jag1^Ndr/Ndr embryos (Fig. 2W,X; data not shown). Moreover, misexpression of Dll1 or Jag1 in the chick neural tube did not alter the regional expression of Nkx6.1, Dbx1, Dbx2, Dll1 or Jag1 (see Fig. S4A-L in the supplementary material). Together, these data show that the positional expression of Dll1, Jag1, Lfng and Mfng is regulated by Nkx6.1 and Dbx1, while neither Dll1 nor Jag1 have any obvious reciprocal effect on DV pattern formation.

The normal establishment of DV progenitor domains in Dll1^−/− and Jag1^Ndr/Ndr mutants prompted us to examine the generation of specific neuronal subtypes in these mice. We quantified the generation of neuronal subtypes derived from Dll1⁺ or Jag1⁺ progenitors, and restricted our analysis to the intermediate and ventral neural tube. In Dll1^−/− embryos, there was a significant overproduction of Isl1/Isl2⁺ MNs at E9.5 (Fig. 3E,E′,K), and an increased number of Chx10⁺ V2a, Gata3⁺ V2b and Evx1/Evx2⁺ V0 neurons at E10.5 (Fig. 3A,A′,C,C′,D,D′,K,Z). The generation of V1 neurons, however, was unaffected, as indicated by similar numbers of En1⁺ cells in Dll1 mutants and littermate controls at E10.5 (Fig. 3B,B′,K,Z). By contrast, in Jag1 mutants at E10.5, there was an approximately twofold increase of En1⁺ V1 neurons, whereas the production of MNs, V2a, V2b and V0 neurons was similar to that in controls (Fig. 3F-J′,K,Z). A similar region-specific effect on neurogenesis was observed also in the brainstem of Dll1^−/− and Jag1^Ndr/Ndr embryos (data not shown). In summary, these data reveal that Dll1 and Jag1 control neurogenesis in a progenitor domain-specific fashion along the DV axis of the neural tube.

We next examined the expression of the Notch target gene Hes5 (Kageyama and Nakanishi, 1997) and the proneural bHLH protein Ngn2 (Bertrand et al., 2002) in the p0— and p1 domains in Dll1^−/− and Jag1^Ndr/Ndr mutants between E9.0 and E10.5. The expression of Dbx1 or Jag1 was used as a positional landmark for the p0— and p1 domains in these experiments (data not shown). In line with the overproduction of Evx1/Evx2⁺ V0 neurons, there was a significant loss of Hes5 expression within the p0 domain in Dll1 mutants at E10.5, whereas Hes5 expression in adjacent Jag1⁺ p1 progenitors was similar to that in control embryos (Fig. 3O,P). The reduction of Hes5⁺ cells in the p0 domain was accompanied by a local increase in the number of Ngn2⁺ cells (Fig. 3L,M,Y). In Jag1^Ndr/Ndr mutants, the p0 domain appeared normal and there was instead reduced numbers of Hes5⁺ cells and an increase of Ngn2⁺ cells in the p1 domain at E10.5 (Fig. 3O,Q,L,N,Y). Also the number of cells expressing the neural progenitor marker Sox3 was reduced by ~40-50% in the p0 domain of Dll1^−/− mice, and a corresponding reduction was observed in the p1 domain of Jag1^Ndr/Ndr mutants (Fig. 3Y), suggesting that the regional overproduction of neurons in these mice occurs at the expense of undifferentiated progenitor cells. At early embryonic stages (E9-E9.5), however, we did not detect any significant difference in Hes5 or Ngn2 expression in Dll1^−/− or Jag1^Ndr/Ndr mutants compared with in littermate heterozygous controls (Fig. 3R-T; data not shown). Moreover, there was no obvious premature induction of neurogenesis in Dll1^−/− or Jag1^Ndr/Ndr mutants, as indicated by the absence of En1⁺ and Evx1/Evx2⁺ neurons at E9.5 (Fig. 3U-X). Together, these data suggest that the overproduction of neurons in Dll1 and Jag1 mutants reflect an increased pace, rather than a premature onset, of neurogenesis.

Notch signalling has also been implicated in the control of gliogenesis (Rowitch, 2004; Taylor et al., 2007). Dll1 and Jag1 mutant embryos die at E12, which precluded any extensive analysis of gliogenesis in these mice. Nevertheless, as indicated by the lack of Olig2, Pdgfrα and Sox10 expression at E11.5 (see Fig. S5C-H in the supplementary material), there was no premature specification of oligodendrocyte precursors (OLPs) in Dll1 mutants. Instead, the extensive loss of Olig2⁺ progenitors observed in Dll1 mutants after E10.5 (see Fig. S5A-D in the supplementary material) indicates a progressive depletion of pMN progenitors over time, and implies that Dll1 function is required to sustain a pool of presumptive MN/OLP progenitors at late developmental stages.

Glycosylation of Notch receptors by Fng proteins has been shown in some situations to facilitate Dll1 signalling and constrain the ability of Jag1 to activate the Notch receptor (Stanley, 2007), whereas, for example, in the somites Lfng appears instead to inhibit Dll1-Notch signalling (Dale et al., 2003). To examine how Fng proteins affected Dll1 and Jag1 signalling in the developing spinal cord, we ectopically expressed Mfng, Dll1 or Jag1 in chick by in ovo electroporation and analyzed the generation of neuronal subtypes derived from progenitor domains endogenously expressing Dll1 or Jag1. Overexpression of Mfng resulted in a ~10-20% reduction of Isl1/Isl2⁺ MNs and Evx1/Evx2⁺ V0 neurons 40 hours after electroporation (Fig. 4A,C,D). The number of V1 neurons, however, was instead increased by ~1.6-fold in response to Mfng expression (Fig. 4B,D). These data suggest that Fng augments Notch signalling within Dll1⁺ progenitor domains, but obstructs Notch signalling in Jag1⁺ p1 progenitors resulting in a reduced generation of MNs and an overproduction of V1 neurons.

Overexpression of Dll1 resulted in a significant reduction (~60%) of Isl1/Isl2⁺ MNs 45-48 hours after electroporation (Fig. 4E,F) (le Roux et al., 2003). The expression of the pan-neuronal marker HuC/D was reduced to a similar extent in the MN domain (Fig. 4E,G), indicating that the loss of Isl1/Isl2 expression indeed reflected an enhanced suppression of neurogenesis by Dll1. Forced expression of Dll1 also suppressed the generation of Evx1/Evx2⁺ V0 neurons (by ~60%; data not shown), but had a marginal effect on the production of En1⁺ V1 neurons (~11%; Fig. 4E,H,I). In corresponding experiments, ectopic expression of Jag1 had a limited influence on the production of MNs and V0 neurons (13% and 18%, respectively), but suppressed the generation of V1 neurons by ~60% (Fig. 4E,J-M; data not shown). These effects of Dll1 and Jag1 could not be explained by a regional effect on the rate of apoptosis (see Fig. S4M-O in the supplementary material). Thus, forced Dll1 expression suppresses MNs and V0 neurons derived from Dll1⁺/Fng⁺ progenitors but has little effect on V1 neurons generated from Jag1⁺/Fng⁻ progenitors. Conversely, Jag1 effectively suppresses neurogenesis in Jag1⁺/Fng⁻ progenitor domains but has a limited influence on neurons generated from Dll1⁺/Fng⁺ progenitors. Together, these findings indicate that the patterned distribution of Fng proteins modulates the capacity of progenitor cells to respond to Dll1 and Jag1 signalling. Nevertheless, in domains in which Dll1 or Jag1 mediate the suppression of neurogenesis, their effect appears to be mediated through canonical Notch downstream signalling. For instance, the suppression of MN differentiation by overexpression Dll1 in the pMN domain was associated with a selective increase in the number of Hes5⁺ (Fig. 4N) and Sox3⁺ progenitors (Fig. 4R,T), whereas the fraction of cells expressing high levels of Ngn2 was reduced (Fig. 4P,T). By contrast, there was no significant change in Ngn2 (Fig. 4Q,T) expression or the numbers of Hes5⁺ and Sox3⁺ cells (Fig. 4O,S,T) within the pMN domain in similar experiments with Jag1.

HD proteins, which are implicated in pattern formation, as well as several components of the Notch pathway, exhibit specific expression domains along the DV axis of the neural tube, but their internal relationship has not been determined. We show that the patterned expression of Notch ligands and Fng genes are controlled by HD transcription factors. Loss of Nkx6.1 led to a ventral expansion of Jag1 expression, accompanied by a reduction of Dll1, Lfng and Mfng expression. Conversely, forced ectopic expression of Nkx6.1 suppressed the expression of Jag1 and induced that of Dll1, Lfng and Mfng. Perturbation of Dbx1 caused ectopic Jag1 expression in the ‘p0 domain’ and a concomitant reduction of Dll1, Lfng and Mfng expression, whereas overexpression of Dbx1 had the opposite effect. By contrast, perturbation of Dll1 and Jag1 did not alter the expression patterns of the Nkx6.1 and Dbx1 proteins, and there was no obvious increase of cell intermingling at progenitor domain boundaries. In conclusion, these findings suggest a mechanism in which HD proteins act upstream of Notch ligands and Fng gene expression, resulting in the establishment of discrete progenitor domains with co-localized expression of Dll1 and Fng, whereas regions expressing Jag1 are devoid of Fng protein expression (see Fig. 5A).

Our data reveal that genetic ablation of Notch ligand expression domains has several consequences for neuronal differentiation. Loss of Jag1 leads to an increase of V1-interneurons, but not of V0- or V2-interneurons or MNs. Loss of Dll1 has the opposite effect and results in an increase in the number of V0- and V2-interneurons and MNs, while leaving the number of V1-interneurons unaltered. Forced expression produced the converse picture in which Jag1 suppressed V1-interneuron differentiation, whereas Dll1 reduced the number of MNs. Interestingly, forced expression of Mfng had a similar activity to overexpression of Dll1 but opposite to that of Jag1 misexpression, which suggests that Mfng enhances Dll1/Notch signalling and reduces the efficiency of Jag1/Notch signalling. This is consistent with several previous observations (Stanley, 2007; Xu et al., 2007), but is in contrast to somitogenesis, in which Lfng appears to inhibit Dll-Notch signalling (Dale et al., 2003), and to data suggesting that Notch2 can respond to Jag1 in the presence of Fng (Hicks et al., 2000; Kopan and Ilagan, 2009).

Spatially organized expression of Notch ligands and Fng proteins is not unique to the DV axis of the developing spinal cord, but is found in many developmental processes, such as rhombomere organization along the anteroposterior axis of the developing hindbrain and DV patterning of the Drosophila imaginal wing and eye discs (Cheng et al., 2004; Wu and Rao, 1999). In sharp contrast to the developing spinal cord, however, Notch signalling activity in zebrafish rhombomeres and in the Drosophila wing and eye is constrained to boundaries between segments, rather than to an entire regional territory (Fig. 5B). More specifically, Notch activation regulates the segregation and differentiation of rhombomere boundary cells in the hindbrain (Cheng et al., 2004), whereas in the fly wing Notch activity is confined to a small group of cells at the wing margin that is important for wing outgrowth (Wu and Rao, 1999). We propose that this difference in outcome of Notch signalling (i.e. within a domain versus at or across the domain boundary) reflects the principal difference of how Notch ligand expression is set up relative to the pattern of Fng expression. In the Drosophila wing, the expression of Fng overlaps with that of Jag1 (Serrate in flies) but not with that of Dll1 (Delta in flies) (Wu and Rao, 1999), whereas in the developing spinal cord Fng expression is instead superimposed on Dll1-expressing, but not on Jag1-expressing, progenitor domains. As depicted in Fig. 5, the selection of which Notch ligand that Fng is co-expressed with has important consequences for signalling. In the Drosophila wing, the presence of Fng in the Serrate expressing domain precludes signalling within the Serrate⁺/Fng⁺ territory, and the absence of Fng expression in the Delta-expressing domain precludes Notch signalling in the Delta⁺/Fng⁻ domain. This internal relationship of Fng and Notch ligands, however, enables functional Notch signalling to take place at the domain boundary, as Delta ligands can activate Notch receptors in the neighbouring cells of the Serrate⁺/Fng⁺ domain, and conversely Serrate ligands can activate Notch receptors in the neighbouring cells of the Delta⁺/Fng⁻ domain (Fig. 5B). The situation in the developing spinal cord, with its Jag1⁺/Fng⁻ and Dll1⁺/Fng⁺ domains, generates the opposite outcome; a domain-wide activation by either Dll or Jag1, and a suppression of signalling across progenitor domain boundaries (Fig. 5B). This is in line with the observed phenotypes in gain- and loss-of-function experiments with Dll1, Jag1 and Mfng, in which the establishment of progenitor domain boundaries was not altered, but which revealed a domain-specific regulation of neurogenesis along the DV axis.

The establishment of the CNS depends on the generation of functionally distinct neurons produced in appropriate numbers and in a spatially defined manner. Previous findings demonstrate that the use of different Notch ligands can mediate qualitatively distinct responses in cell fate determination. For instance, Dll4 and Jag1 have opposing effects on angiogenesis (Benedito et al., 2009), and Dll4, but not Dll1, can function to specify V2a and V2b interneuron subtype identities in a uniform population of ventral progenitors of the developing spinal cord (Del Barrio et al., 2007; Peng et al., 2007; Rocha et al., 2009) (see also Fig. 3). Despite the fact that Dll1 and Jag1 control the differentiation of distinct neuronal subtypes along the DV axis, our data do not support the notion that signalling downstream of Dll1 and Jag1 is qualitatively different, or that Dll1 or Jag1 would determine the actual identity of cells. Instead, at least with regard to the regulation of Hes5, Ngn2 and Sox3 expression, Dll1 and Jag1 appear to carry out similar signalling activities, albeit in distinct DV progenitor domains. The subtype identity of cells is therefore more likely to be determined through the regional activity of cell fate-determining HD proteins (Jessell, 2000), while the specific distribution of Notch ligands and Fng proteins along the DV axis, also controlled by HD proteins, endows distinct progenitor domains with an enhanced regulatory autonomy and the possibility to influence the pace of neurogenesis in a domain- and cell type-specific manner along the DV axis of the neural tube (Fig. 5). Hence, the combined functional activities and regulatory interactions between HD proteins and components of Notch signalling described here should contribute to a coordinated control of cell-type specification and regulation of cell number in the developing CNS.

We thank A. Pierani for providing the Dbx1 mutant mice, M. Sander for the Nkx6.1 mice and G. Sturm for the Jagged^Ndr/+ mice. We are also grateful to D. Anderson, A. Pierani, M. Goulding, T. Jessell, B. Novitch, T. Edlund and C. Tabin for supplying us with reagents. We are grateful to A Kjaeldgaard for providing us with abortion material. This work was supported by the Swedish Research Council (DBRM), the Swedish Foundation for Strategic Research (FFL-2, CEDB and OBOE), EU project EuroSyStem, the Göran Gustafsson Foundation, Swedish Brain Power, the Swedish Cancer Society, the Craig Hospital Foundation and the research funds of the Karolinska Institutet.