Notch1 mRNA is expressed in the cerebellar primordium as early as E9 (data not shown), and by E12.5 is restricted to neural progenitors in the ventricular zone (VZ) (Figure 1). BMP activity in the rhombic lip and caudal VZ is evident from the expression of Msx2, a homeodomain transcription factor that has been shown to be a target of BMP signaling 21. Note that the expression pattern of Msx2 overlaps with the expression of Math1 at this embryonic stage (Figure 1). Expression of the bHLH transcription factor Mash1 22 likely delineates the precursors of Purkinje cells and other GABAergic cerebellar neurons that arise from the ventricular zone 1923. Within the VZ, Hes5 expression reflects Notch signaling activity in the cerebellar progenitor pool, while the punctate expression of Delta1 likely indicates a subpopulation of neural precursors that are undergoing differentiation.

As a first test for the requirement of Notch signaling in the regulation of rhombic lip neural induction, we examined whether loss of Notch activity in cerebellar progenitors would lead to an increase in rhombic lip neurogenesis as measured by Math1 expression. To circumvent the early (E10) lethality of Notch1 null embryos 24, we crossed mice carrying a conditionally null allele of Notch1 25 in which the first exon of the Notch1 gene has been flanked by loxP sites ('floxNotch1') with mice expressing cre recombinase under control of the Engrailed-1 gene (En1cre) 26. As shown by β-galactosidase staining in a Rosa26 stop-lacZ background 27, Engrailed-1 directs the expression of cre recombinase widely across the mid-hindbrain region by E9.5 (Figure 2a). By E10.5, there is no detectable Notch1 mRNA remaining in the cerebellar primordium of the conditional mutant, in contrast to wild-type littermates (Figure 2b, c), with the exception of the most lateral territory, where recombination is incomplete (data not shown). Loss of Notch1 did not have an overt effect on isthmic or roof plate markers at this stage (Additional File 1). Both of the mRNAs for the bHLH transcription factors Mash1 and Math1 are expressed at higher levels in the mutant embryos, but in a complementary pattern: Mash1 expression is increased in the rostral cerebellar primordium (Figure 2d, e) whereas there is a robust increase in Math1 expression proximal to the rhombic lip in comparison to wild-type littermates (Figure 2f, g). This increase in Math1 expression is apparent at the protein level as well (Figure 2h, i), and appears to reflect an increase in both cell numbers as well as the levels of Math1 expression induced within differentiating cells.

By E12.5, mutant embryos exhibit a marked decrease in the size of the cerebellar primordium in comparison to wild-type littermates. Nevertheless, no obvious increase in cell death (terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling, Additional File 2, or caspase-3 immunohistochemistry), or decrease in proliferation (short term bromodeoxyuridine incorporation) in the residual cerebellar ventricular zone was observed at this stage (data not shown). Following the upregulation of Mash1 in the mutants at E10.5 (Figure 2d, e), Mash1 expression is decreased in all but the most rostral territory of the ventricular zone in comparison to wild type at E12.5 (Figure 2l, m). In contrast, an increase in Math1 expression is still evident in the E12.5 mutants, and scattered Math1⁺ cells are observed in a broader area of the ventricular zone in comparison with sections from wild type littermates at the same medial-lateral position (brackets in Figure 2n, o). This Math1 expression pattern in the mutant resembles that observed in the most medial sections from wild-type animals, where the cerebellar primordium is thinner and in closer proximity to the midline and roof plate. The increase in subpial distribution of Math1⁺ cells in the mutant could be accounted for by an increase in migration rate away from the rhombic lip, which is consistent with the observation that Math1 activity is required for subpial migration of rhombic lip neurons 628.

The increase in Math1 expression and concomitant decrease in Mash1 from E10.5 to E12.5 in the mutants suggests that early loss of Notch1 results in an increase in rhombic lip neurogenesis at the expense of maintaining the ventricular zone progenitor pool. Consistent with this, by E16.5, the mutant cerebellum is greatly reduced in size in comparison to wild type (Figure 3a–d), and there is a severe decrease in the Purkinje cell precursors (calbindin immunostaining) that are normally generated from the ventricular zone beginning around E12 (Figure 3e, f). In contrast to the results presented in a previous study on the En2-Cre; floxNotch1 mutant mouse 19, we did not observe any postnatal survival of mutant embryos, and even by E18.5 there was no apparent recovery of the mutant cerebellum in terms of size or morphology (data not shown). The greater severity of phenotype in our study most likely results from a more complete loss of Notch1 in the cerebellar primordium using the En1-cre knock-in mouse than was obtained in the En2-cre; floxNotch1 mutant.

Previously, we and others have found that early Math1 rhombic lip neural precursors give rise to specific hindbrain nuclei and the deep cerebellar nuclei (DCN) 78, with the peak production of the latter occurring around E11.5, suggesting that these neurons are generated after the majority of hindbrain neurons have been produced (E9.5 to E11.5). To obtain a short term fate map of the Math1⁺ cells that arise in the En1cre; floxNotch mutant cerebellum, we generated this conditional mutant on a Math1-LacZ knock-in (+/-) background 29 and analyzed the residual β-galactosidase staining present in rhombic lip lineages in coronal sections at E14.5. Strikingly, while we observed overall comparable numbers of rhombic lip-derived β-gal⁺ hindbrain neurons that collectively include the mesopontine tegmental (MPT), parabigeminal (PBG), lateral lemniscus (LL), and lateral parabrachial (LPB) neurons from rostral to caudal positions (Figure 4a–d, long arrows), we observed a dramatic decrease in DCN (short arrows). Neurons of the LPB that arise from a Math1⁺ precursor have been shown to express calbindin embryonically 8 and, interestingly, we observed an increase in calbindin staining in the mutant (Figure 4e, f, white arrows). This result suggests that, in the absence of Notch1 activity, specific early rhombic lip derived lineages are generated in excess at the expense of the hindbrain nuclei and DCN that are specified slightly later.

Surprisingly, we also observed that granule cell precursors (GCPs), a rhombic lip derived population specified after the DCN, appear to be generated to some extent in the mutant, although there is a pronounced decrease in caudal regions (Figure 4c, d). The persistence of granule cells in the mutant may reflect the fact that, while the En1cre driver used in these experiments recombines the vast majority of the mes/r1 primordium, the most lateral regions of the cerebellar primordium escape recombination (data not shown). Thus, it is possible that some granule cells are generated laterally and migrate medially to populate the rostral external granule layer (EGL). Alternatively, as discussed in more depth below, there may be distinct lineage-restricted pools of rhombic lip progenitors that are maintained independently of Notch activity until they begin to divide asymmetrically to produce neurons.

If the loss of Notch1 enhances the responsiveness of cerebellar progenitors to inductive signals that direct rhombic lip neurogenesis, then expression of Notch ligands (for example, Delta) should also render cells more receptive to these developmental cues by virtue of lateral inhibition 30. To test this hypothesis directly, we generated a pseudotyped bicistronic retrovirus that expresses full length Delta1, along with human placental alkaline phosphatase (PLAP), to allow histochemical detection of transfected cells. Approximately 10⁶ virions were injected into the ventricle of E9.5 to E10 embryos in utero using ultrasound backscatter microscopy 31, and the infected animals analyzed three weeks after birth. Figure 5a shows the results of injections of a control retrovirus expressing only PLAP analyzed at P21 by alkaline phosphatase histochemistry. Infected cells were present in all compartments of the mature cerebellum and did not show an obvious bias towards any particular cell type. Strikingly, similar injections with retroviruses expressing full length Delta1 (Figure 5b) resulted in the labeling of granule cells predominantly, as observed by position, morphology (note the PLAP staining of parallel fibers in the molecular layer), and immunohistochemical co-labeling with antibodies against Zic2 (green) 32 and PLAP (red; Figure 5e inset). Furthermore, Delta1 infections at this stage also appeared to contribute to the DCN (white arrow). Given that these retroviruses require approximately 24 hours to integrate and express the Delta1 protein, we suggest that these Delta infected cells were differentiating and expressing Math1 at around E11.5 to E12 and thus contributed to the DCN and early specified (anterior) GCP that are normally produced at that time.

To examine the effect of constitutive Notch activation in cerebellar progenitors, we performed injections with a retrovirus expressing the intracellular domain of the Notch receptor (Notch ICD), which is known to result in ligand independent activation of the Notch signaling pathway 33. Injections of Notch1 ICD expressing retroviruses at E9.5 resulted in infected cells developing primarily into Bergmann radial glia (Figure 5c) as shown by their radial morphology and immunohistochemical co-labeling with brain lipid-binding protein (BLBP; red) 34 and PLAP (green; Figure 5f inset). These gain-of-function experiments demonstrate that constitutive Notch activity in cerebellar progenitors prevents these cells from developing as rhombic lip derivatives. While several recent reports have described a role for Notch signaling in Math1⁺ lineages following their specification 3536, our results argue that, in those contexts, Notch signaling must be regulated in a dynamic manner such that subsequent phases of differentiation can occur.

While both Notch and BMP activities have been shown to be crucial during embryonic cerebellar development, how these signaling pathways interact in this context is unknown. To explore the mechanism by which Notch1 signaling in the cerebellar primordium antagonizes the induction of rhombic lip neurogenesis in progenitor cells by BMP signaling, we chose to use the chick in ovo electroporation system, which is well suited for short term gain-of-function with multiple expression plasmids 37. Stage HH 10–12 chick embryos 38 were electroporated at the mid-hindbrain boundary with various expression constructs along with a green fluorescent protein (GFP) reporter plasmid, and analyzed two days later by in situ hybridization on cryosections of the cerebellar primordium for changes in the expression of the chick Math1 homolog Cath1 (Figure 6). To test whether ectopic BMP activity could induce Cath1 expression, we electroporated a constitutively active form of the BMP receptor 1b (caBMPR) at a level sufficient to induce patterning changes but not cell death 3940. While a control electroporation of a GFP reporter plasmid alone did not induce Cath1 expression (Figure 6a, b), ectopic activation of the BMP signaling pathway in the cerebellar primordium (dashed circle) resulted in a robust induction of Cath1 expression along the dorsal surface of the cerebellar anlage (Figure 6c, d). Interestingly, when the Notch1 ICD was co-electroporated along with the caBMPR expression plasmid into the cerebellar ventricular zone, the ectopic induction of Cath1 was suppressed (Figure 6e, f; Additional File 3). Thus, although ventricular zone progenitors are competent to respond to BMP signaling and express Cath1, they are prevented from doing so by high levels of Notch activity.

The observation that Notch activity can block the induction of Cath1 by BMP signaling in cerebellar progenitors prompted us to try to determine at what level these pathways intersect within the cell. Electroporations in chick cerebellar primordia of caBMPR alone or caBMPR and Notch1 ICD were stained by immunohistochemistry for phosphorylated Smad1, a direct readout of BMP receptor signaling activity 41. Smad1 is a transcription factor that, upon phosphorylation by the BMP receptor serine/threonine kinase activity, forms a heterodimer with Smad4 and translocates to the nucleus to activate transcription of target genes (for example, Msx1/2; Figure 6h) 42. As shown in Figure 6, the levels of phosphorylated Smad1 were elevated in both caBMPR (Figure 6i) and caBMPR/Notch1 ICD (Figure 6k) electroporated tissue, demonstrating that expression of the Notch ICD does not interfere at this stage of the BMP signaling pathway. However, while electroporation of caBMPR into the cerebellar primordium (dashed oval) resulted in ectopic expression of Msx1/2 in the ventricular zone (Figure 6j), this induction was suppressed when Notch1 ICD was co-electroporated (Figure 6l). Counts of GFP⁺/Msx1/2⁺ cells in the ventricular zone of caBMPR and caBMPR/Notch1 ICD electroporated cerebella from three embryos each are shown in the graph and indicate that, while approximately 80% of ventricular zone cells transduced with caBMPR express Msx1/2, this percentage drops to around 10% when Notch1 ICD is co-transduced. Thus, expression of the Notch1 ICD antagonizes the BMP signaling pathway at the level of Msx1/2 expression.

Engrailed1-cre, floxed Notch1, Math1-LacZ, and Rosa26 stopLacZ mice were genotyped as previously described 25262729. To generate En1cre; floxNotch1 embryos, the En1-cre line was crossed with homozygous floxNotch1 animals, and the resultant En1-cre; floxNotch1/+ males crossed with homozygous floxNotch1 females. En1cre; floxNotch1; Math1-LacZ animals were generated by crossing En1cre; floxNotch1 (c/+) with Math1-LacZ; floxNotch1 (c/+). The morning of the observed plug was considered day 0.5. Embryos collected at E10.5 to E14.5 were fixed in ice cold 4% paraformaldehyde/phosphate-buffered saline (PBS) for 1 to 2 hours, washed in PBS, and equilibrated in 30% sucrose/PBS overnight. Older embryos and adults were perfused transcardially and the brains dissected prior to sucrose equilibration. For cryosectioning, embryos were mounted in Tissue-Tek OCT (VWR, West Chester, PA) and sectioned at 14 to 20 μM.

Section antisense RNA in situ hybridization was performed as previously described 52, with the following probes: Notch1, Math1, Mash1, Msx2, and Cath1. Immunohistochemistry with antibodies against Math1 (rabbit-antiserum; kind gift of J Johnson (UT Southwestern Medical Center), calbindin (rabbit antiserum; Swant, Bellinzona, Switzerland), human placental alkaline phosphatase (sheep α-PLAP antiserum; American Research Products, Belmont, MA, USA), Zic2 (rabbit antiserum; kind gift of J Aruga (RIKEN Brain Science Institute), BLBP (rabbit antiserum, kind gift of T Anthony and N Heintz (Rockefeller University), phosphorylated Smad1 (purified rabbit IgG; Cell Signaling Technologies, Danvers, MA, USA), and Msx1/2 (mouse monoclonal antibody 4G1, ascites, Developmental Studies Hybridoma Bank, Iowa City, IA, USA) was performed as previously described 33.

Preparation, injection, and histochemical analysis of control (CLE) and Notch1 ICD (CLEN) retroviruses have been described previously 33. Full-length cDNA for human Delta1 was subcloned into pCLE downstream of the EF1α promoter (CLED) and virus prepared as above. We analyzed six to eight P21 brains each for the CLED and CLEN experiments and found three to four for each that had substantial infections in the cerebellum.

In ovo electroporation was performed as described previously 40, with the following modifications. Specifically, cDNA for the Notch1 ICD was subcloned into the chick expression vector pMiwIII, such that its expression was directed by the chicken β-actin promoter. The constitutively active BMP receptor 1b and GFP constructs have been described previously 40. Plasmids were injected into the ventricle at the mid-hindbrain boundary (GFP, 0.2 μg/μl; Notch1 ICD, 1 μg/μl; and caBMPR, 0.33 μg/μl) and two electrodes placed on either side of the neural tube. Five rectangular electric pulses of 15 volts (50 ms each) were then delivered. Embryos were recovered after approximately two days further incubation, fixed for 1 hour in ice cold 4% paraformaldehyde/PBS, washed in PBS, and allowed to equilibrate overnight in 30% sucrose/PBS prior to mounting and cryosectioning. At least three electroporated embryos were analyzed for each experiment.