Mash1 and Math3 Are Required for Development of Branchiomotor Neurons and Maintenance of Neural Progenitors

Mash1 and Math3 are expressed in branchiomotor neuron progenitors in the hindbrain

To characterize the roles of Mash1 and Math3 in hindbrain development, we first examined their expression patterns by in situ hybridization. The hindbrain was cut along the roof plate and flat mounted to locate the expression domains. At embryonic day 9.5 (E9.5), Mash1 was expressed mainly in two longitudinal columns. Expression in the ventral column occurred at a high level from r2 to r4, where trigeminal and facial branchiomotor neurons arise (Fig. 1A, arrowheads). This expression continued but became less significant at E10.5 (Fig. 1C). Math3 expression was also observed in the ventral column from r2 to r4 at E9.5 (Fig. 1B, arrowheads) and continued at E10.5 (Fig. 1D). In r2 and r4, Mash1 expression occurred mainly in the ventricular zone, whereas it continued in subsets of cells migrating out of the ventricular zone at both E9.5 and E10.5 (Fig. 1E, H, K, O). Math3 expression also occurred in the ventricular zone, but Math3 was expressed mainly by the cells in the outer layers of r2 and r4 during this period (Fig. 1F, I, L, P). Some of these cells coexpressed Mash1 and Math3 (Fig. 1N, R). These results suggest that Mash1 and Math3 are expressed by both neural progenitors and differentiating neurons of r2 to r4.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Increased apoptosis in Mash1(-/-);Math3(-/-) hindbrain. TUNEL assay was performed with transverse sections of r2 (A-D, I, K) and r4 (E-H, J, L) at E10.5. More cells died in r2 and r4 of Mash1(-/-);Math3(-/-) (D, H-L). These apoptotic cells expressed Nkx2.2 (I, J) but not MAP2 (K, L), suggesting that double-mutant cells undergo apoptosis before becoming mature neurons (I-L). Apoptotic cells were also slightly increased in number in Mash1(-/-) r2 (B). Scale bar, 100 μm. WT, Wild type.

Because the ventral region of r2-r4 is known to give rise to branchiomotor neurons, we next compared the expression patterns of Mash1 and Math3 with those of Phox2b, which has been shown to control branchiomotor neuron generation (Pattyn et al., 1997, 2000). Phox2b expression was well overlapped with the Mash1 and Math3 expression domains in r2 and r4 (Fig. 1G, J, between dashed lines in M, Q). These results suggest that Mash1 and Math3 are expressed at distinct but overlapping stages by differentiating trigeminal and facial branchiomotor neurons and their progenitors.

Defects of trigeminal branchiomotor neuron development in Math3(-/-) and Mash1(-/-);Math3(-/-) mutants

To investigate the roles of Mash1 and Math3 in branchiomotor neuron development, we first analyzed trigeminal motor neurons in Mash1-null and/or Math3-null mice. We used Islet1 as a general motor neuron marker (Ericson et al., 1992). In wild-type embryos, trigeminal branchiomotor neurons arose from the ventral region of r2 and r3 between E9.5 and E10.5 (Taber Pierce, 1973; Marshall et al., 1992) and migrated dorsolaterally to form the trigeminal motor nucleus (Fig. 2A, arrowhead, Q) (Altman and Bayer, 1982), which extended in the dorsolateral position of r2 and r3. This migration was almost completed by E11.5, and only a few motor neurons remained in the ventral region (Fig. 2E). Although trigeminal motor neurons appeared to be generated normally in Mash1-null embryos (Fig. 2B, F, R), a few more motor neurons still remained in the ventral region adjacent to the floor plate of r2 and r3 at E11.5 (Fig. 2F, arrowhead), suggesting that the development of trigeminal branchiomotor neurons is slightly delayed compared with the wild type. Similarly, more motor neurons remained in the ventral region of r2 and r3 at E11.5 in Math3-null embryos (Fig. 2G, filled arrowhead), although trigeminal motor neurons were normally born at E10.5 (Fig. 2C,S). In addition, the trigeminal motor nucleus of Math3-null embryos was smaller than that of the wild type (Fig. 2G, open arrowhead). This is likely attributable to a migratory defect, because more neurons remained in the ventral region of Math3 mutants (Fig. 2G, filled arrowhead). Furthermore, the rostral neurons migrated dorsocaudally to converge on a small caudal nucleus in r2 (Fig. 2G, open arrowhead). These results indicate that Math3 is important for migratory behavior of trigeminal branchiomotor neurons.

Because the overlapping patterns of Mash1 and Math3 expression suggest redundant roles of these bHLH genes, we next examined Mash1;Math3 double-mutant embryos. In the double mutants, trigeminal motor neurons were severely reduced in number at E10.5 and E11.5 (Fig. 2D, H, filled arrowhead, T). Very few cells migrated dorsolaterally, and, as a result, the trigeminal motor nucleus was very small at E11.5 (Fig. 2H, open arrowhead) compared with the wild type, Mash1-null, and Math3-null embryos (Fig. 2E-G). At later stages, the trigeminal motor nucleus (nV) was also very small (Fig. 2L, P). To further investigate the cause of the defects of the double-mutant embryos, we examined apoptotic cell death by TUNEL assay. In wild-type embryos and Math3-null embryos, no overt apoptotic cell death was observed in r2 (Fig. 3A, C), whereas a few TUNEL-positive cells were detected in Mash1-null embryos (Fig. 3B). In contrast, TUNEL-positive cells were significantly increased in r2 of Mash1; Math3 double-mutant embryos (Fig. 3D). These apoptotic cells expressed Nkx2.2 (Fig. 3I) but were negative for the neuronal marker MAP2 (Fig. 3K), suggesting that cells undergo apoptosis before becoming mature neurons in the absence of Mash1 and Math3. These results indicate that Mash1 and Math3 are required for migration and survival of immature trigeminal branchiomotor neurons.

Defects of facial branchiomotor neuron development in Math3 and the double mutants

We next examined development of facial branchiomotor neurons in mutant embryos. Facial branchiomotor neurons are known to adopt unique migratory pathways (Garel et al., 2000). In wild-type embryos, facial branchiomotor neurons arose at r4 from E9.5-E11.5 (Fig. 2A, E, open arrowheads, U) (McKay et al., 1997; Pattyn et al., 2003b). After E10.5, they migrated caudally, keeping its position adjacent to the floor plate. After entering r6, they switched the direction and migrated dorsolaterally, forming the nucleus at r6 (Fig. 2I, M, nVII). This migration was almost completed at E14.5. In Mash1 mutant embryos, generation and migration of facial branchiomotor neurons seemed to be normal at E10.5-E12.5 (Fig. 2B, F, J, V). However, at E13.5, many cells still remained near the floor plate of r4 or on the way to r6 (Fig. 2N), although some of them in r4 could be aberrant inner-ear efferent neurons (Tiveron et al., 2003). These results suggest that in the absence of Mash1, facial branchiomotor neurons exhibit a delay of migration. However, at later stages, the facial motor nucleus was formed, and no overt defect was observed (data not shown). In contrast, Math3-null embryos exhibited abnormal migration of facial branchiomotor neurons. At E11.5, very few cells reached r6 (Fig. 2G, arrow), whereas the majority of the cells remained in the ventral region of r4 of Math3-null embryos (Fig. 2G). At E12.5, very few cells reached the dorsolateral position of r6, whereas the majority still remained in the ventral region of r4 (Fig. 2K). At E13.5, in Math3-null embryos, some of these neurons migrated dorsolaterally within r4 (Fig. 2O, arrowhead), forming an elongated nucleus between r4 and r6 (Fig. 2O, bracket), in contrast to the wild-type facial motor nucleus, which resided in r6 (Fig. 2M).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Ectopic gene expression in Math3(-/-) hindbrain. In situ hybridization for c-Ret (A, B), Sim1 (C, D), and Ebf-1 (E, F) was performed with flat-mounted hindbrain (A-D) and transverse (E, F) sections of r4 at the indicated time points. A, B, c-Ret expression was rarely detectable in the wild type although ectopically upregulated in the ventral r4 of Math3(-/-). C, D, Sim1 expression was not detected in the wild type but occurred at a high level in the ventral r4 of Math3(-/-). Thus, branchiomotor neurons of r4 were misspecified in Math3(-/-). E, F, Ebf-1, which is required for migration of branchiomotor neurons, was not significantly affected in Math3 mutant embryos, suggesting that Ebf-1 is not involved in the migratory defects of Math3 mutant facial branchiomotor neurons. Scale bar: (in F) E, F, 100 μm. WT, Wild type.

In the Mash1;Math3 double-mutant hindbrain, although many neurons were born in the ventral r4 at E10.5 (Fig. 2D, X), very few cells reached r6 at E11.5, whereas the majority of the cells remained in the ventral region of r4 (Fig. 2H). At E12.5 and E13.5, only subsets of cells reached the dorsolateral position of r6 (Fig. 2L, P, open arrowheads), whereas other subsets migrated within r4, forming an elongated nucleus between r4 and r6, which is similar to but smaller than the Math3-null nucleus (Fig. 2O, P). In Mash1;Math3 double-mutant r4, TUNEL-positive cells were increased (Fig. 3H) compared with the others (Fig. 3E-G). These apoptotic cells expressed Nkx2.2 (Fig. 3J) but were negative for the neuronal marker MAP2 (Fig. 3L), suggesting that cells undergo apoptosis before becoming mature neurons in the absence of Mash1 and Math3. These results indicate that Mash1 and Math3 are required for migration and survival of immature facial branchiomotor neurons. Strikingly, the left and right neurons, which are normally separated by the floor plate, were fused across the midline in r4 of the double mutants. This fusion started at E12.5 (Fig. 2L) and became more severe at E13.5 (Fig. 2P, arrowhead). Thus, in the absence of Mash1 and Math3, the cytoarchitecture of r4 was disrupted.

Misspecification of facial branchiomotor neurons in Math3 mutant embryos

To characterize the migratory defects of facial branchiomotor neurons in Math3 mutant mice, we examined several markers that are expressed during migration. Expression of c-Ret, a gene for the glial cell line-derived neurotrophic factor receptor, initiates in facial branchiomotor neurons only after they migrate into r5 (Garel et al., 2000). In the wild-type embryos, few, if any, cells expressed c-Ret at E10.5 (Fig. 4A). In contrast, in Math3 mutant embryos, c-Ret expression occurred ectopically in r4 (Fig. 4B), suggesting that this ectopic expression may affect migration of branchiomotor neurons. Sim1, a gene for a bHLH-PAS (motif of proteins PER-ARNT-SIM) transcription factor expressed by V3 interneurons of the spinal cord (Fan et al., 1996; Briscoe et al., 1999), was not expressed in the wild-type hindbrain (Fig. 4C). In contrast, Sim1 was ectopically expressed in the ventral region of r4 of Math3(-/-) at E11.5 (data not shown) and E13.5 (Fig. 4D). These results indicate that facial branchiomotor neurons are misspecified in Math3 mutant embryos, raising the possibility that this misspecification may lead to the migratory defect of these neurons.

In contrast, expression of the bHLH gene Ebf-1, which is required for correct migration of branchiomotor neurons (Garel et al., 2000), was not significantly affected in Math3 mutant embryos (Fig. 4E, F). Thus, these Math3-null cells are not totally transformed into other cell types, although they are severely misspecified.

Loss of neural progenitors in the ventral region of Mash1; Math3 double-mutant r4

In the absence of Mash1 and Math3, facial branchiomotor neurons on either side were fused across the midline in r4. To examine this fusion, we made transverse sections through r4 at E13.5. In the wild type, the inner surface was covered by the MAP2-negative ventricular zone, which contains neural progenitors (Fig. 5A). In contrast, in the ventral region of the double-mutant r4, the ventricular cells were missing and instead MAP2-positive neurons from either side of the neural tube were fused together over the floor plate (Fig. 5B, boxed region). Furthermore, these MAP2-positive neurons were exposed directly to the fourth ventricle (Fig. 5B′). Neural progenitors are known to have the tight junction and adherens junction at the apical side and thereby prevent neurons from migrating into the ventricular lumen: premature loss of neural progenitors leads to intermingling of neurons from the left and right walls of the neural tube (Hatakeyama et al., 2004). In the wild type, the adherens junction molecule N-cadherin was highly expressed at the apical surface (Fig. 5C), whereas it was missing in the double mutants (Fig. 5D, arrowheads), suggesting that neural progenitors are prematurely lost. In agreement with this observation, BrdU uptake and expression of Ki-67, a marker for mitotic cells, were reduced in the double mutants compared with the wild type as early as E10.5 (Fig. 5E-H), indicating that dividing neural progenitors are not properly maintained in the absence of Mash1 and Math3. The floor plate, which expresses Shh, was maintained in the double mutants as in the wild type at both E10.5 and E13.5 (Fig. 5I-L), indicating that the midline structure is not lost in the double mutants. Thus, inactivation of Mash1 and Math3 is likely to lead to premature loss of neural progenitors and their apical junctional complex and thereby allows facial branchiomotor neurons to intermingle each other from either side of the neural tube. This observation implies that Mash1 and Math3 are required not only for normal development of branchiomotor neurons but also for maintenance of neural progenitors and the cytoarchitecture.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Intermingling of r4 neurons across the midline in Mash1(-/-);Math3(-/-). Transverse sections of the ventral r4 were examined. A, B, B′, In the wild type, the inner surface of the hindbrain was covered by MAP2-negative ventricular cells, and neurons (MAP2⁺) were not present in the midline region at E13.5 (A). In contrast, many neurons were intermingled across the midline in Mash1(-/-);Math3(-/-) (B). The boxed region in B is enlarged in B′. Intermingled neurons were exposed to the lumen of the hindbrain. C, D, The adherens junction molecule N-cadherin was highly expressed at the apical side in the wild type (C), whereas it was missing in the double mutants (D, arrowheads). E, F, BrdU uptake was examined at E10.5. BrdU uptake was reduced in the double mutants (F) compared with the wild type (E). G, H, Ki-67 expression was examined by immunohistochemistry at E10.5. Ki-67 expression was reduced in the double mutants (H) compared with the wild type (G). I-L, In situ hybridization of Shh. Shh was expressed in both the wild-type and double-mutant floor plate at E10.5 and E13.5. Scale bars: (in A) A, B, 400 μm; B′, 100 μm; E-L, 100 μm. WT, Wild type; DKO, double knock-out.

Notch signaling is impaired in Mash1(-/-);Math3(-/-) double-mutant mice

Because Notch signaling is known to regulate maintenance of neural progenitors (Ohtsuka et al., 1999; Gaiano et al., 2000; Hitoshi et al., 2002), we next addressed whether Notch signaling is affected in the double-mutant embryos. We first examined expression of Hes5, a downstream Notch effector (Ohtsuka et al., 1999). In the ventral r4 of the wild-type, Mash1(-/-), and Math3(-/-) embryos, expression of Hes5 was observed at E10.5 (Fig. 6A-C). In contrast, in the double-mutant r4, Hes5 expression was significantly downregulated (Fig. 6D, between dashed lines). At E11.5, Hes5 expression remained at a very low level in the double mutants, whereas Hes5 expression was still maintained in the wild-type, Mash1-null, and Math3-null r4 (Fig. 6E-H), although it was also slightly downregulated in Mash1-null r4 (Fig. 6F). These results indicate that Notch signaling is affected in the double-mutant embryos, which accounts for loss of neural progenitors.

Because neuronal bHLH genes are known to activate expression of Notch ligands such as Delta, we next examined expression of Delta-like1 (Dll1) and Dll3. In the wild-type and Math3(-/-) embryos, Dll1 expression occurred at E10.5 and E11.5 in the ventral domain where facial motor neurons are generated, whereas it was slightly decreased in Mash1(-/-) embryos (Fig. 6I-K, M-O). In contrast, in the double mutants, Dll1 expression in the ventral r4 was downregulated at E10.5 (Fig. 6L) and mostly missing at E11.5 (Fig. 6P). Similarly, expression of Dll3 was missing in the double mutants at E10.5 (Fig. 6T, between dashed lines), in contrast to the wild-type, Mash1-null, and Math3-null embryos (Fig. 6Q-S). Thus, expression of the Notch ligands is also lost in the absence of Mash1 and Math3, which is likely to lead to downregulation of Hes5 in neighboring cells.

The loss of Notch ligand expression could be attributable to either downregulation of the gene expression in cells that should normally express Mash1/Math3 or apoptotic loss of these cells. To differentiate between these possibilities, we monitored Mash1/Math3-expressing cells in the double-mutant hindbrain. In Mash1 and Math3 mutant alleles, a neomycin-resistant (neo) gene cassette was inserted into each locus in the same and reverse orientations, respectively (Guillemot et al., 1993; Tomita et al., 2000). Thus, we could use the neo antisense and sense strand probes as an indicator for the Mash1 and Math3 promoter activities, respectively. In the wild-type embryos, Math3 was expressed by cells adjacent to the floor plate of r4 at E11.5 (Fig. 7A, arrow). In the double-mutant embryos, the antisense neo was expressed in the same domain (Fig. 7B, arrow). Similarly, the sense neo was expressed in the same domain (Fig. 7D). Thus, the cells that should normally express Mash1/Math3 did not die but still existed in the double-mutant embryos at E11.5, indicating that loss of Dll1 and Dll3 expression is attributable to downregulation of the gene expression rather than to cell death.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Impairment of Notch signaling in the ventral r4 of Mash1(-/-);Math3(-/-). In situ hybridization for Hes5 (A-H), Dll1 (I-P), and Dll3 (Q-T) was performed with transverse sections of the ventral r4 at E10.5 (A-D, I-L, Q-T) and E11.5 (E-H, M-P). In Mash1(-/-);Math3(-/-), expression of Hes5 (D, H), Dll1 (L, P), and Dll3 (T) was significantly downregulated in the ventral r4, which gives rise to facial branchiomotor neurons (between dashed lines). In Mash1(-/-), expression of Hes5 (B, F), Dll1 (J, N), and Dll3 (R) was slightly downregulated compared with the wild type and Math3(-/-). Scale bar, 100 μm. WT, Wild type.

We also noted that antisense neo-positive cells were present in a higher density in the double-mutant ventricular zone than in the wild type, which exhibited a salt-and-pepper pattern (Fig. 7, compare arrows in A, B). This is probably because Notch signaling, which represses neuronal bHLH gene expression of neighboring cells in a way known as “lateral inhibition,” is impaired in the double mutants.

The delay of oligodendrocyte development in Mash1(-/-);Math3(-/-) mouse embryos

The ventral region is known to give rise to oligodendrocytes after generating motor neurons. Because neural progenitors were significantly decreased in the double mutants, we next examined the development of oligodendrocytes in the ventral r4 of the double mutants. In r4, expression of the early oligodendrocyte marker PDGFRα was observed at E12.5 in the wild-type, Mash1(-/-) and Math3(-/-) embryos (Fig. 8A-C). In contrast, in Mash1(-/-);Math3(-/-) embryos, PDGFRα-positive cells were severely reduced in number (Fig. 8D). These results suggest that impairment of maintenance of neural progenitors may lead to reduction of oligodendrocyte progenitors. At E14.5, however, PDGFRα-positive cells were found to be distributed throughout the hindbrain in the double-mutant embryos (data not shown). Thus, development of oligodendrocytes is not dependent on Mash1 or Math3, and the initial decrease of oligodendrocyte progenitors can be recovered at later stages.

Mash1 and Math3 cooperatively regulate development of trigeminal and facial branchiomotor neurons

We found here that Mash1 and Math3 cooperatively regulate development of trigeminal and facial branchiomotor neurons. In the absence of Mash1 and Math3, these neurons exhibit migratory defects and apoptosis, resulting in formation of very small trigeminal and facial nuclei. Furthermore, in r4, neurons with migratory defects are intermingled over the midline from either side of the neural tube, and oligodendrocyte progenitors are reduced in number. Thus, Mash1 and Math3 not only promote branchiomotor neuron development but also regulate the subsequent oligodendrocyte development and the cytoarchitecture in the hindbrain. It has been shown that neurons derived from the ventral r4 are completely missing in Mash1;Phox2b double-null embryos, whereas they are still generated in Mash1-null and Phox2b-null embryos, indicating that Mash1 and Phox2b cooperatively promote neurogenesis in the ventral r4 (Pattyn et al., 2004). However, because Phox2b upregulates Math3 expression (Dubreuil et al., 2000, 2002; Pattyn et al., 2000), our present data raise the possibility that at least some aspects of Phox2b functions in rescuing Mash1 activities are mediated by Math3. Particularly, apoptosis and decrease of expression of Notch signaling molecules observed in Mash1-null mice are more severely affected in Mash1;Math3 double-null mice, indicating that Math3 may compensate Mash1 for cell survival and activation of Notch signaling.

Mash1 and Math3 function at a neuronal differentiation process in the hindbrain

Our present study established the roles of Mash1 and Math3 in the differentiation process of cranial branchiomotor neurons. In the double mutants, although trigeminal and facial branchiomotor neurons seemed to be born normally, they underwent cell death at E10.5. As a result, expression of the neuronal markers MAP2 and β-tubulin III exhibited marked reduction (data not shown). Furthermore, the majority of the surviving neurons did not properly migrate but remained in the ventral region, where they were born. Thus, in the absence of Mash1 and Math3, the neuronal fate seems to be determined, but the subsequent differentiation process is impaired, although Mash1 and Math3 are known to regulate the neuronal fate determination process. It is possible that other bHLH genes may compensate for the determination step. Alternatively, Phox2b itself has neuronal fate determination activities, as described previously (Dubreuil et al., 2000; Pattyn et al., 2004), although it remains to be determined whether Phox2b can directly activate the neuronal fate determination program without inducing neuronal bHLH gene expression.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

The cells that should normally express Mash1 and Math3 remained in Mash1(-/-); Math3(-/-). In situ hybridization was performed with transverse sections of the ventral r4 at E11.5. A, Math3 was expressed in a salt-and-pepper pattern in the ventral r4 of the wild type (arrow). B, Expression of the antisense neomycin mRNA under the control of the Math3 promoter was detected in the ventral r4 of Mash1(-/-); Math3(-/-) (between dashed lines). C, Mash1 was expressed in the ventral r4 of the wild type. D, Expression of the sense neomycin mRNA under the control of the Mash1 promoter was detected in the ventral r4 of Mash1(-/-); Math3(-/-) (between dashed lines). Thus, the cells that should normally express Mash1/Math3 remained in the double mutants. Cells adjacent to the floor plate were decreased in the double mutants (compare arrowheads in A, B). Scale bar, 100 μm. WT, Wild type.

Mash1 and Math3 regulate Notch ligand expression in r4

It was reported previously that Dll1 expression is completely lost in r4 of Phox2b;Mash1 double-mutant embryos. This loss of Dll1 expression in Phox2b;Mash1 double mutants is probably as a result of the complete blockade of neurogenesis (Pattyn et al., 2004). In the present study, we showed that Dll1 and Dll3 expression is also lost in the ventral r4 of the Mash1;Math3 double mutants, although many neurons are born. It is thus likely that Mash1 and Math3 cooperatively maintain Notch ligand expression in these neurons and thereby activate Notch signaling and downregulate Mash1 and Math3 expression in neighboring cells, resulting in a salt-and-pepper expression pattern. Consistent with this idea, in the absence of Mash1 and Math3, expression of the Notch effector Hes5 is lost, and the Mash1 and Math3 promoter activities are uniformly upregulated in the double-mutant ventricular cells. Our results thus suggest that Math3, rather than Phox2b, directly regulates Notch ligand expression in collaboration with Mash1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Impairment of oligodendrocyte development in the ventral r4 of Mash1(-/-); Math3(-/-). In situ hybridization for PDGFRα was performed at E12.5. PDGFRα expression was significantly downregulated in the ventral r4 of Mash1(-/-); Math3(-/-) (D) compared with the others (A-C). Thus, oligodendrocyte progenitors were significantly reduced in number in the double mutants. WT, Wild type.

Control of facial branchiomotor neuron migration

Facial branchiomotor neurons adopt a unique migratory pathway, and we found that in the absence of Math3, this migration is severely affected. Thus far, it has been shown that facial branchiomotor neuron progenitors and their progeny start expression of the three transcription factor genes Nkx6.1, Math3, and Ebf-1 in this order, and mutant analyses revealed that each gene is required for the normal migration. In the absence of the earliest onset gene Nkx6.1, facial branchiomotor neurons migrate dorsolaterally within r4, forming the nucleus mainly in r4 (Müller et al., 2003; Pattyn et al., 2003a), whereas in the absence of the latest onset gene Ebf-1, dorsolateral migration occurs in r5-r6, forming the nucleus in r5-r6 (Garel et al., 2000). In the absence of the intermediate onset gene Math3, some populations of facial branchiomotor neurons exhibit aberrant migration in r4-r5, whereas others exhibit normal migration within r6, forming an elongated nucleus extending from r4 to r6. Thus, inactivation of Nkx6.1, Math3, and Ebf-1 allows migration of branchiomotor neurons in r4, r4-r6, and r5-r6, respectively, in contrast to the wild type allowing dorsal migration only in r6. The onset of aberrant migration correlates well with the onset of gene expression, indicating that each gene prevents ectopic migration in a stage-specific manner. It is thus likely that the behavior of facial branchiomotor neurons should be controlled strictly by Nkx6.1, Math3, and Ebf-1 in this order and, in their absence, these neurons could prematurely undergo dorsolateral migration.

c-Ret is prematurely expressed by Math3-null branchiomotor neurons in r4, as observed in Ebf-1 mutants, and this ectopic c-Ret expression probably accounts for the aberrant dorsal migration of these neurons. Although Mash1 and Math3 are functionally redundant in the branchiomotor neuron development as shown above, most of the migratory defects are attributable only to Math3 mutation, indicating that Math3 has a unique role in migration of branchiomotor neurons.

Maintenance of neural progenitors by Mash1 and Math3

We showed that oligodendrocyte progenitors are reduced in number at E12.5 in Mash1;Math3 double mutants, although the number of oligodendrocytes becomes comparable with that of the wild-type embryos at E14.5. This recovery is probably because of the highly proliferative property of oligodendrocyte progenitors, as shown in the analysis of Pax6 mutant mice (Sun et al., 1998). The reduction of oligodendrocyte progenitors at E12.5 in the double mutants is likely attributable to decrease of neural progenitors because Notch signaling is impaired. Another interesting observation is intermingling of branchiomotor neurons across the midline in the double mutants. This defect is also likely attributable to loss of neural progenitors. Because neural progenitors have the tight and adherens junctions at the apical side to prevent neurons from scattering, premature loss of neural progenitors allows neurons to escape into the lumen (Hatakeyama et al., 2004). Thus, Mash1 and Math3 play an important role in maintenance of neural progenitors to ensure not only a sufficient supply of glial progenitors but also the structural integrity of the nervous system.