Pax-6 is involved in the specification of hindbrain motor neuron subtype

During development of the vertebrate nervous system, the neural tube gives rise to distinct cell types at specific positions along the anteroposterior (A-P) and dorsoventral (D-V) axes (reviewed in Lumsden and Krumlauf, 1996; Tanabe and Jessell, 1996). For example, the floor plate cells differentiate at the ventral midline, motor neurons appear ventrolaterally, and neural crest and roof plate cells are generated in the dorsal region. Moreover, many classes of interneurons have been identified, all of which are situated in the middle part of the neural tube. Recent studies have revealed that the initial D-V patterning and generation of multiple cell types are dependent on two localized sources of inductive signals, BMP-mediated signals that come dorsally from the non-neuronal ectoderm and sonic hedgehog (shh) secreted from the ventrally situated notochord (reviewed in Smith, 1993; Tanabe and Jessell, 1996). As development proceeds, more specialized neuronal subtypes are generated. The subclasses of motor neurons in the chick and the primary motor neuron subclasses in zebrafish can be distinguished by the combinational expression of transcription factors belonging to the LIM homeodomain family (Tsuchida et al., 1994; Appel et al., 1995; Tokumoto et al., 1995; also see Lumsden 1995; Tanabe and Jessell, 1996). From transplantation experiments in chick and zebrafish, it is suggested that rostrocaudally restricted signals regulate the subtype identity of motor neurons (Appel et al., 1995; Tanabe and Jessell, 1996). However, precise molecular mechanisms underlying specification of motor neuron subtypes are still largely unknown.

Along this line of study, great advances have recently been made in identifying molecules implicated in the regionalization of the nervous system. They include genes encoding transcription factors, secreted proteins, cell adhesion molecules and cell surface receptors and their ligands, many of which are expressed in discrete domains or regions within the neural tube. Among those, members of the Pax family are well-characterized transcription factors and are expressed during early CNS development (Stoykova and Gruss, 1994). The Pax genes were originally identified as vertebrate homologues of the Drosophila genes paired (prd) and gooseberry (gsb), and are characterized by the presence of a specific DNA-binding domain of about 130 amino acid residues called the paireddomain (for review see Mansouri et al., 1996). prd and gsb are segmentation genes belonging to the pair-rule and segmentpolarity classes, respectively, and are known to play crucial roles in establishing the segmental organization of the fly embryo through interactions with segment-polarity genes, wingless and engrailed (Noll, 1992).

Several members of the Pax family are differentially expressed along the D-V axis of the neural tube. The expression domain of Pax-6 constitutes the ventrolateral region, whereas Pax-3 is expressed in a dorsal region of the neural tube with its ventral border overlapping the Pax-6 domain (Walther and Gruss, 1991; Goulding et al., 1993; Matsuo et al., 1993; Stoykova and Gruss, 1994). Previous studies demonstrated that these characteristic expression patterns of Pax-6 and Pax-3 are under the control of the ventralizing signal from the notochord (Goulding et al., 1993; Macdonald et al., 1995). Furthermore, ectopic expression of Pax-3 inhibits normal floor plate differentiation and induces fusion of bilateral Pax-6 domains (Tremblay et al., 1996). Thus, it appears plausible to speculate that these two Pax genes play important roles in establishing precise subdivisions of the neural tube following the initial D-V patterning by inductive signals.

With respect to the Pax family, it is noteworthy that spontaneous mutant animals and human disorders have been found. For example, mutations in the mouse and rat Pax-6 genes result in an eyeless phenotype in the homozygous conditions (Hill et al., 1991; Matsuo et al., 1993), and those in the human PAX-6 gene are found in patients with a variety of eye disorders, including aniridia and Peter’s anomaly (Hanson and Van Heyningen, 1995). Several morphological analyses have also been performed in Small eye mice, suggesting involvement of Pax-6 in patterning the forebrain and differentiation and migration of neurons (Schmahl et al., 1993; Stoykova et al., 1996). Despite these studies, a clear role for Pax-6 in the development of the CNS has yet been established at the molecular level.

In the course of our analysis on craniofacial phenotypes in the Pax-6 mutant rat, we noticed that the homozygous skulls lack the hypoglossal foramen, through which one of the cranial motor nerves, the hypoglossal, passes (Osumi-Yamashita et al., 1997). With respect to the development of motor nerves, the hindbrain stands as an interesting subject since the hindbrain motor neurons constitute three distinct classes, somatic motor (SM), branchiomotor (BM) and visceral motor (VM) neurons (Guthrie and Pini, 1995; Tucker et al., 1996). SM neurons innervate muscles derived from the paraxial mesoderm, while BM neurons target the muscles of the pharyngeal arches. VM neurons are presynaptic to neurons innervating salivary glands or smooth muscle in the head and thorax. Interestingly, the axons of these motor neuron classes segregate along different pathways. SM axons leave the neural tube at the ventral exit point, while axons of BM and VM neurons grow dorsally within the neural tube and their axons subsequently converge with the incoming axons of the sensory neurons at the dorsal exit point. It is notable that these features in the hindbrain contrast to those in the spinal cord, where both SM and VM axons exit the neural tube ventrally and no BM neurons develop.

In this study, we examined in detail the development of the hindbrain neural tube to explore the role of Pax-6 in the nervous system. We investigated a novel Small eye rat strain with a point mutation in the Pax-6 gene, which was recently found in a colony of Sprague-Dawley rats maintained in one of our laboratories. In these homozygous mutant embryos, we found that the axonal projection patterns of two cranial somatic motor nerves, the abducens and hypoglossus, were altered to resemble those of BM/VM nerves. We also investigated expression of three members of Lim homeobox genes, Islet-1, Islet-2 and Lim-3, which may constitute the hindbrain LIM code (Varela-Echavarria et al., 1997a). We demonstrate that the expression of Islet-2 was specifically missing in the mutant, indicating that subtype specification of motor neurons was affected by the Pax-6 mutation. Furthermore, Wnt-7b, a member of the Wnt family of secreted molecules, was specifically lost in the region overlapping with the Pax-6-positive domain in the homozygous hindbrain. Based on these results, we propose that Pax-6 plays a vital role in the specification of motor neuron subtypes in the mammalian hindbrain.

A novel strain of Small eye rats (rSey²) was found in the Sprague-Dawley (SD) rat colony of the Safety Research Laboratories, Yamanouchi Pharmaceutical Co., Ltd. Embryos were obtained from intercrosses of heterozygous rats or of wild-type SD rats, with the day of the vaginal plug designated as 0 days. Homozygotes were distinguishable by their external morphology from embryonic day 11 (E11) onwards.

Procedures of PCR for cDNA and genomic DNA were as previously described (Matsuo et al., 1993). Positions of primers used for PCR are shown in Fig. 2. Oligonucleotides used to amplify the cDNAs are as follows: No. 1 to 6, previously described (Matsuo et al., 1993), No. 7, 5′-GTGTCATCAATAAACAGAGTTCTTC-3′; and No. 8 CTTGCGTGGGTTGCCCTGGTA-3′.

All the procedures were basically performed as described previously (Lee et al., 1995). Anti-Pax-6 antibody was kindly provided from Dr. Randall Reed and recognizes C-terminal domain of Pax-6 protein (Davis and Reed, 1996). The 2H3 monoclonal anti-165-kDa neurofilament and the 40.2D6 anti-Islet-1 antibodies were obtained from the Developmental Studies Hybridoma Bank. For whole mounts, HRP-conjugated second antibody (Chemicon) was used and immunoreactivity was detected with diaminobenzidine (DOJIN). For sections, antigen-enhancement by boiling in 0.01 M sodium citrate was performed before incubation with the primary antibody. Biotinylated anti-mouse antibody (Vector) was used as the secondary antibody and immunoreactivity was detected using ABC kit (Vector Laboratories) and Metal enhanced DAB kit (Pierce).

Methods for in situ hybridization on cryosections and whole mounts were described previously (Osumi-Yamashita et al., 1997). The antisense RNA probe for the rat Pax-6 gene was synthesized from the template plasmid described previously and recognized both the wild-type and mutant Pax-6 transcripts (Matsuo et al., 1993). Other rat cDNAs were obtained by RT-PCR in which cDNA templates were synthesized from the total RNA taken from E11.5 rat embryo brain as described in detail in Nakagawa et al. (1996). Oligonucleotides used to amplify the cDNAs are as follows: Wnt-1, 5′-CTCTTTGGC-CGAGAGTTCGTG-3′ and 5′-AGACGAGCTGTTGCAAGCTCG-3′; Wnt-3, 5′-GAAGGCTGGAAGTGGGGCGGC-3′ and 5′-ACG-CAATGGCATTTCTCCTTCCG-3′; Wnt-5a, 5′-GGTCTCTAGGTAT-GAATAACC-3′, and 5′-GCGTGCGCTCTCATAGGAAC-3′; Wnt-7a, 5′-GGCATAGTCTACCTCCGGATC-3′ and 5′-CATCCACGAA-GACCTTGGCG-3′; Wnt-7b, 5′-GGCATCGACTTTTCTCGTCGC-3′ and 5′-CGAGGTGCGGTTGCACAGAC-3′; Islet-1, 5′-GCAGCATAGGCTTCAGCAAG-3′ and 5′-GTGGAACGCCTG-GACGATG-3′; Islet-2, 5′-GACAACAGGTCTCACTGCGC-3′ and 5′-CACCTCTGCACTCTGTCCTG-3′; and Lim-3, 5′-CAGTTC-CAAGTCCGACAAGG-3′ and 5′-GGTACGAGTCAAGACTGGCT-3′. The identities of the PCR products were confirmed by sequencing the subcloned fragments using an automated sequencer (ABI 373A DNA Sequencer, Applied Biosystems). Since rat sequences were not available for some of the above genes during the initial course of this study, we used the reported mouse sequences for primer design and compared the similarity of the obtained PCR products with the mouse sequences. In all cases, more than 97% identity in nucleotide sequence was observed in each clone, which gave sufficient information to distinguish different members of each family. Therefore, we concluded that the obtained cDNA clones are the rat cognates of the mouse cDNAs. The cDNA fragments were cloned into pBluescript SK II (Stratagene), and antisense RNA probes were generated with T3 or T7 RNA polymerase. Corresponding sense probes were used to check the specificity of hybridization signals. Samples were photographed under a Nomarski or dissection microscope.

Embryos were fixed in 4% PFA in PBS. For retrograde labeling, DiI, 1,1-dioctadecyl-3,3,3′3′-tetramethylindocarbacyanine perchlorate (DiI C₁₈ (3), Molecular Probes, Inc.) was saturated in dimethylformamide and applied to the facial, glossopharyngeal or vagal ganglia, or to the mesodermal tissue that the abducent and hypoglossal nerves innervate. For anterograde labeling, DiI solution was injected at the ventrolateral region of the posterior hindbrain. Specimens were incubated for 12-24 hours at 37°C, frozen in OCT compound and sectioned. Labeled axons and neuronal cell bodies were visualized using fluorescent optics and photographed.

A novel Small eye rat strain (rSey²) was spontaneously identified in SD rat colony of one of our laboratories. rSey² exhibited abnormal phenotypes indistinguishable from those of rSey (Matsuo et al., 1993; Fujiwara et al., 1994, Osumi-Yamashita et al., 1997); adult heterozygotes showed small eyes, while homozygotes lacked eyes and nose and die at birth. Genetic crosses proved the phenotypic variant to be due to an autosomal semidominant mutation (data not shown). Impaired development of eyes and nose was morphologically detectable from E11 onward, and E12 homozygous embryos exhibited loss of lens and olfactory placodes and abnormal structure of the optic vesicle (Fig. 1). Immunohistochemistry using anti-Pax-6 antibody, which recognizes C-terminal 17 amino acid residues (Davis and Reed, 1996), revealed intense nucleus staining in the developing lens, retina and olfactory epithelium of wild-type embryo, while no specific staining was observed in homozygous tissues (Fig. 1).

To discover whether rSey² has a mutation in the Pax-6 gene, RT-PCR was performed using the same primer sets previously used for amplifying rat Pax-6 (Matsuo et al., 1993). All the fragments of region A, B and C (see Fig. 2) were amplified from cDNA obtained from E12 homozygous embryos and no difference was found in the size of amplified fragments (data not shown). However, sequence analysis of the cloned RT-PCR fragment of region B revealed that the RNA from rSey²/rSey² contained an additional one nucleotide (C) at 3′-downstream to the paired box, resulting in an abnormal stop codon at 33 bp downstream from the insertion site due to the frame-shift (Fig. 2). This suggests that the mutant gene product lacks the pairedtype homeodomain, which is consistent with the fact that the antibody against the C-terminal domain of Pax-6 protein could not stain tissues from rSey²/rSey² (Fig. 1). To ascertain the origin of the mRNA addition, genomic DNA from rSey² was amplified by PCR using primer sets of No. 7 + 8 and No. 3 +8.Fragments of 167 and 155 bp were generated from primer sets of No. 7 + 8 and No. 3 + 8, respectively. The genomic sequence of these fragments confirmed that C was inserted at the same site as that in cDNA. Thus, it is concluded that rSey²is a novel Pax-6 mutant rat strain.

The first indication of neuronal defects in the hindbrain of the Pax-6 mutant embryos was obtained by whole-mount immunostaining of cranial nerves using the anti-neurofilament antibody, 2H3. In the cranial region of both wild-type and homozygous embryos, the highly organized formation of cranial ganglia and nerve fibers becomes apparent from embryonic day (E) 11 (data not shown). Although the branchial nerves developed normally, abnormal development of two cranial somatic motor nerves, the abducent (nVI) and hypoglossal (nXII), was observed in homozygous embryos. In E12 wild-type embryos, nXII originates from the caudal hindbrain and rostral-most spinal cord, extending axons into the tongue primordium (Fig. 3A, see arrows). Axons from the first and second cervical nerves (C1 and C2) extend along nXII to form a part of the cervical ansa (Fig. 3A, arrowhead). In homozygous embryos, however, no obvious nerve fibers could be detected in the corresponding region. Instead, C2 was found to innervate the tongue primordium (Fig. 3B, arrowhead). C1 did not develop its ventral root in the homozygotes, but instead possessed dorsal roots that are not present in wild-type embryos.

In normal development, the abducent nerve (nVI) arises from the hindbrain at the levels of r5 and r6, and projects ventrally then anteriorly toward one of the eye-moving muscles, the lateral rectus (Varela-Echavarria et al., 1997a). The morphology of nVI was examined in detail from the ventral side of the neural tube after removing the jaws in E13 embryos. In the wild-type embryo, axon bundles of nIV were observed to extend towards the eye primordium, crossing the ramus palatinus (rp) of the facial nerve (Fig. 3C). In the homozygous embryos, however, no fibers corresponding to nVI developed (Fig. 3D). From this view, misextension of the ophthalmic nerve (V1) could be seen in the frontonasal region. This is probably caused in response to the loss of the lateral nasal prominences (Osumi-Yamashita et al., 1997), the target tissue that V1 innervates in the wild type. In contrast to the above disorders, axonal growth into maxillary and mandibular regions from the trigeminal ganglion (V2 and V3) looked similar in the wild-type and mutant embryos (Fig. 3A,B). Morphology of other cranial nerves that originate from the hindbrain region including the acousticofacial (nVII and VIII), 3′-site to the paired box, resulting in the frame shift, which generates an abnormal stop codon (!) at the downstream. The primers used for PCR are numbered. glossopharyngeal (nIX), and vagal (nX) nerves, also developed normally in the homozygous embryos. In addition, we could find no obvious defect or change in nerve fiber formation in the trunk region of the neural tube (Fig. 3A,B).

As described above, two motor nerves were found to be missing in Pax-6 mutant embryos. Therefore, we next asked whether this defect was due to the lack of corresponding motor neuron populations. The abducent (VI) and hypoglossal (XII) nerves originate mainly from r5 and r6, and from the lower part of r7 and cervical spinal cord, respectively (Tucker et al., 1996; Varela-Echavarria et al., 1997a; and our unpublished results). Only the results on the r5 and r7 levels are shown here as representative data, although immunostaining and in situ analyses were performed along the entire A-P neuraxis. We first used neurofilament (Fig. 4) and MAP-2 (data not shown) as general neuronal markers. Neurons positive for these markers were observed in the hindbrain in both wild-type and homozygous embryos from E11 onward. By E12, nVI and nXII displayed typical SM nerve features, i.e., their axons extended ventrally out of the neural tube in the wild-type embryo (Fig. 4A,D,K,N). As mentioned above, this shows striking contrast to the phenotype of the BM/VM nerves of nVI, nIX and nX, whose motor axons grow dorsally. At the same rhombomere levels in mutant embryos, neurofilament-(Fig. 4G,I,Q,S) and MAP-2-(data not shown) positive cells were also clearly detectable, indicating that emergence of motor neurons per se was intact. However, the loss of ventrally growing axons at the r5-r7 levels in the mutants (Fig. 4G,I,Q,S) were apparent at this stage.

The differentiation of motor neurons was also examined by immunostaining of Islet-1. Islet-1-positive cells emerged in the ventral region of the neural tube in the wild-type embryos at E11 (data not shown), which increased in number, clustering to make a motor column at E12 (Fig. 4B,E,L,O). Neurofilament was also expressed in these Islet-1-positive cells by comparing consecutive sections. In the homozygous embryos, the Islet-1-positive cells were similarly detected in the ventral region of the neural tube at E11 (data not shown) and E12 (Fig. 4H,J,R,T). Thus, the absence of ventrally extending nerve fibers is not a result of the lack of motor neurons themselves. However, there were differences in the location of motor neurons in the mutant hindbrain. It is notable that, at the r5 level in the E12 homozygous mutants, some Islet-1- and neurofilament-positive cell bodies were observed to stray out into the ventricular side (arrowheads in Fig. 4I,J), which is in contrast to the tight cluster of motor neurons in the wild-type embryos (Fig. 4D,E). This particular phenomenon was highly region- and stage-specific: we did not observe similar mislocation of motor neurons in other areas of the hindbrain or at later stages. An additional minor difference between wild-type and homozygous embryos is that Islet-1-positive cells in the mutant were located more laterally, which was most prominent at the r7 level (Fig. 4T). In advanced embryos, abducent and hypoglossal nuclei were morphologically undetectable in the homozygous brainstem (data not shown).

We further examined localization of the Pax-6 protein within the hindbrain with special attention to the location of motor neurons. At E12, Pax-6 protein was detected in the ventrolateral region of the hindbrain in the wild-type embryos (Fig. 4C,F,M,P). Importantly, consecutive sections (Fig. 4E versus F; O versus P) and double staining for Pax-6 mRNA and Islet-1 protein (data not shown) revealed that Islet-1-positive cells were present in the ventrolateral regions negative for Pax-6. In addition, consecutive sections stained with antibodies against Pax-6 and neurofilament (Fig. 4K versus M; N versus P) and double staining for Pax-6 mRNA and neurofilament (data not shown) clearly demonstrated that the Pax-6-positive cells are negative for neurofilament, and that the ventral and dorsal borders of the Pax-6-positive domain in the hindbrain correspond well to the regions flanking the ventral (e.g., nXII) and the dorsal (e.g., nX) exit points of the cranial nerves, respectively (arrows in Fig. 4M).

The above studies demonstrated that, in the postotic hindbrain of Pax-6 mutant rats, motor neurons were generated at correct locations, but their axons failed to leave the neural tube at the correct ventral exit points. As mentioned above, the region of the neural tube at the level of r5-r7 provides a particular situation, where BM and VM neurons coexist with SM neurons (refer to Fig. 8A). Thus, we next examined the patterns of axonal projection of cranial nerves at different rhombomere levels both by retrograde and anterograde labeling methods (Fig. 5).

When retrograde labeling of facial (nVII), glossopharyngeal and vagal (nX) nerves was performed in the wild-type embryos, dorsally projecting VM and BM axons were labeled with DiI, and cell bodies of these neurons were found to be lined up along the ventrolateral margin of the neural tube (arrows in Fig. 5A,C; also see small arrows in Fig. 4E,O). In contrast, DiI-labeled cell bodies of the abducent (nVI) and hypoglossal (XII) SM neurons were found to form a packed round-shaped cell cluster at a more ventral position (Fig. 5B,D; also see Fig. 4E,O). We confirmed that the position of these labeled cells at the r5 and r7 levels precisely matched with the location of Islet-1-positive cells (data not shown). In homozygous embryos, however, we found quite distinct patterns of axonal projection. At the level of r5, DiI injection into the facial ganglion labeled not only the VM/SM neurons of the facial nerve (nVII) but also the cell bodies aligned at the ventricular side of the neural tube (arrowheads in Fig. 5F; also see Fig. 4I,J). Since these cells were positive for both Islet-1 and neurofilament (see Fig. 4I,J) and there were no axons growing ventrally, the motor neurons that should have normally been somatic are most probably misspecified as branchiomotor with their axons extending dorsally and their cell bodies mislocated. Likewise, retrograde labeling of the vagal ganglion not only stained cell bodies that are located laterally (arrows in Fig. 5H: also see arrows in Fig. 4T) but also those situated more ventrally (arrowheads in Fig. 5H; also see arrowheads in Fig. 4T) than those in the wild-type embryos. This position corresponded to that of the hypoglossal nuclei in the wild-type embryos, which was revealed by comparison of the labeling patterns shown in Fig. 5D,H. In contrast to the wild-type, DiI injected at regions where the ventral fibers would have normally been expected did not stain any cells in the neural tube (Fig. 5G,I).

The alteration of axonal patterning in homozygous embryos was also examined by anterograde labeling. In the wild-type embryos, DiI injection at the ventral region of the neural tube in r5 (data not shown) and r7 (arrows in Fig. 5E) labeled typical SM-type axons exiting the neural tube ventrolaterally. In the homozygotes, however, labeled axons showed typical BM- or VM-type projection patterns, i.e., initially extending dorsally and then leaving the neural tube in the company of the vagal nerve (arrowheads in Fig. 5J). In summary, these results demonstrate that, in mutant embryos, Islet-1-positive motor neurons originating at the r5 and r7 levels do not project their axons ventrally like typical SM-type neurons, but resemble BM- or VM-type neurons in the dorsal projection of their axons. This abnormal axonal growth pattern is consistent with the lack of the abducent and hypoglossal nerves in the homozygotes.

Mislocation of motor neurons and abnormality of their axonal growing in the Pax-6 mutant hindbrain raised a question whether this defect is caused by impaired specification of motor neuron subtypes. In the chick hindbrain, distinct subclasses of motor neurons have been shown to express different sets of LIM homeobox genes; especially, postotic hindbrain SM neurons (i.e., abducent and hypoglossal nuclei) express Islet-1 + Islet-2 + Lim-3, while BM/VM neurons express only Islet-1 (Varela-Echavarria et al., 1997a). To examine whether specification of motor neurons was altered in the mutant at the molecular levels, we compared expression patterns of these LIM homoebox genes in the wild-type and mutant embryos.

Consistent with the results in the chick, all these LIM genes were expressed in distinct subclasses of motor neurons in the wild-type rat hindbrain (summarized in Fig. 8A; Varela-Echavarria et al., 1997a). Distribution of Islet-1 transcripts (Fig. 6A) were the same as that of the protein; it was expressed by motor neurons, both the ventrally clustered cells and laterally situated ones, of all cranial nerves (see small arrows in Fig. 4E,O). Islet-2 was expressed in SM-type nuclei (nIII, nIV, nVI and nXII) in the hindbrain. Compared to the Islet-1 domain, Islet-2 expression was rather confined in the ventral cluster within the Islet-1-positive domain (Fig. 6B). In the chick hindbrain, Lim-3 is shown to be expressed only by abducent (nVI) and hypoglossal (nXII) nuclei located at postotic hindbrain levels (Varela-Echavarria et al., 1997a). We also observed Lim-3 expression in the same hindbrain levels in the rat. These Lim-3-positive cells were located in the ventral Islet-1-positive region, which overlapped with that of the Islet-2-positive domain (Fig. 6C). We also noticed that only Islet-1 was expressed dorsal to the Islet-2/Lim-3-positive domain, suggesting the presence of a Islet-1-positive motor neuron subclass that does not express Islet-2 or Lim-3. These cells are detected along the ventrolateral margin of the neural tube and, by comparing retrograde labeling data (see arrows in Fig. 5A,C), probably constitute a BM/VM-type subclass of motor neurons, which is consistent with the observation in the chick hindbrain. Thus, at least two molecularly distinct subclasses of motor neurons can be distinguished in the postotic hindbrain of the wild-type rat: Islet-1⁺/Islet-2⁺/Lim-3⁺ neurons, and Islet-1⁺ neurons, which correspond to the SM and BM/VM type subclasses, respectively.

Next, we examined the Pax-6 mutant hindbrain. Islet-1 mRNA was detected in an almost similar pattern to that in the wild-type, but shifted rather laterally (Fig. 6D), and expression of Lim-3 was observed in the same manner with that in the wild-type (Fig. 6F). In contrast, no expression of Islet-2 was detected in the postotic hindbrain of the Pax-6 mutant (Fig. 6E). It is important to note that Islet-2 was expressed in oculomotor (nIII) and trochlear (nIV) nuclei, which originate from the midbrain and midbrain/hindbrain boundary levels (see Fig. 8A). We also found that SM neurons in the spinal cord and cells in the cranial and dorsal root ganglia expressed Islet-2 both in the wild-type and mutant embryos (data not shown). Thus, the loss of Islet-2 expression at r5-r7 levels are highly region-specific. These results are consistent with the idea that the SM-type motor neurons are specifically lost in the postotic hindbrain of the Pax-6 mutant.

Although both Pax-6 and Islet-2 were expressed almost throughout the hindbrain and spinal cord, loss of Islet-2 expression was observed only in the postotic hindbrain level of the mutant embryos. Thus, we suspected some secondary signaling cascade might be involved in the regulation of Islet-2 by Pax-6. In the chick and mouse neural tube, multiple members of Wnt genes, which encode secreted signaling molecules, are expressed in the ventrolateral region overlapping the Pax-6 domain (Parr et al., 1993; Hollyday et al., 1995). Therefore, we challenged the possibility that expression of some members of Wnt genes is specifically affected in the mutant hindbrain. To test this possibility, in situ analyses of five Wnt members, Wnt-1, Wnt-3, Wnt-5a, Wnt-7a and Wnt-7b, were performed on whole mounts and serial sections. In accordance with previous results in the mouse (Parr et al., 1993), rat Wnt genes showed distinct expression patterns within the neural tube. Wnt-1 and Wnt-3 were expressed in the dorsal region, where their ventral boundaries showed no overlap with that of Pax-6, both in the hindbrain and in the spinal cord, and there was no change of expression between the wild-type and homozygous embryos (data not shown). Expression of Wnt-5a, Wnt-7a and Wnt-7b was detected in the domain largely overlapping with the Pax-6 domain. In the hindbrain, however, each member showed distinct patterns along the A-P axis. Wnt-5a was not detected in r2-r6, and was present only in the lateral region at r7 (Fig. 7A). Wnt-7a was expressed throughout the hindbrain (Fig. 7C), whereas the Wnt-7b expression domain included the ventro-lateral-dorsal side of the hindbrain only caudal to r2 (Fig. 7E). In contrast, all these members possessed almost identical expression patterns throughout the spinal cord.

Among the Wnt family members, only Wnt-7b exhibited the different expression pattern between the wild-type and mutant embryos. In the homozygous hindbrain, the most ventral Wnt-7b domain specifically diminished while its expression was detected in the dorsal side of the neural tube (Fig. 7H,J), making the stripe narrower (compare Fig. 7E and F). By comparing expression of Pax-6 (e.g., Fig. 4C,M) and Wnt-7b (e.g., Fig. 7G,I) in the wild-type, it is concluded that, in the mutant hindbrain,Wnt-7b is lost from only the ventralmost subdomain in which it overlaps with Pax-6. A similar alteration of Wnt-7b expression was observed also in the spinal cord. Thus, the loss of Wnt-7b expression in the Pax-6 mutant embryos is highly specific among the members of the Wnt family.

In this study, we have demonstrated that two cranial motor nerves with characteristics of SM neurons, i.e., the abducent and hypoglossal, were specifically missing in the Pax-6 mutant rat. Other cranial nerves that develop within the same anteriorposterior levels in the hindbrain (the facial, glossopharyngeal and vagal nerves) grew normally. Similar defects were also observed in Small eye mouse embryos (Ericson et al., 1997; and our unpublished results), suggesting that the abnormality is indeed genetically associated with loss-of-function mutation of Pax-6. Embryonic motor neurons express the LIM-type transcriptional factor Islet-1 soon after they leave the cell cycle (Ericson et al., 1992; Yamada et al., 1993). Islet-1-positive cells were observed lateral to the floor plate throughout the hindbrain and spinal cord both in the wild-type and homozygous embryos from E11 onward. Thus, patterns of axonal growth but not generation of the motor neuron lineage were substantially affected in the mutant.

We showed that ventrally growing axons (the SM-type) were specifically lost in the postotic hindbrain of the Pax-6 mutant, while dorsally growing ones (BM/VM-type) remained intact. One possibility to explain this phenomenon would be that the impaired development of the postotic hindbrain motor nerves is caused by the lack of the muscles that they should innervate. However, this is quite unlikely since the tongue muscle, which is the major target of the affected hypoglossal nerve, was innervated by the cervical motor nerve (C2, Fig. 3B, arrowhead).

Another possibility that accounts for the defects in the Pax-6 mutant hindbrain is misspecification of motor neuron subtypes. Although Islet-1 is regarded as a general marker for motor neurons, several other classes of transcription factors are also expressed in the subsets of motor neurons. In particular, members of LIM homeobox genes, Islet-1, Islet-2, Lim-1 and Lim-3, define subclasses of motor neurons that segregate within the motor columns and these distinct subsets of motor neurons choose discrete axonal pathways toward different target muscles (Tsuchida et al., 1994). Subclasses of primary motor neurons in zebrafish can also be distinguished by the combinational expression of LIM genes (Appel et al., 1995; Tokumoto et al., 1995). LIM homoedomain proteins control cell fate decisions in both C. elegans and Drosophila (Way and Chalfie, 1988; Freyd et al., 1990; Cohen et al., 1992; Bourgouin et al., 1992), and are suggested to be involved in control of axonal pathfinding in vertebrates and invertebrates (Lundgren et al., 1995; Tanabe and Jessell, 1996). Most recently, hindbrain LIM codes are intensively characterized in the chick (Varela-Echavarria et al., 1997a). In particular, ventrally projecting SM neurons of abducent and hypoglosssal nuclei express Islet-1 + Islet-2 + Lim-3, while BM/VM neurons (dorsally projecting) express only Islet-1. Some atypical subtypes, such as accessory abducens generated at the r5 level, were identified as Islet-1 + Lim-3 motor neurons, though they express Islet-1 + Islet-2 + Lim-3 at the stage of axonal extension. Thus, Islet-2 is considered as a specific marker for the SM-type motor neurons in the hindbrain. This leads us to examine expression patterns of these three members of LIM genes in the Pax-6 mutant hindbrain. We confirmed that the general feature of the hindbrain LIM code is consistent with that in the chick. Furthermore, we found that expression of Islet-2 was specifically missing at the r5-r7 levels in the mutant rat hindbrain, while both Islet-1 and Lim-3 expression remained intact. The results strongly suggest that the primary defect of the motor neurons in the Pax-6 mutant hindbrain is the specific loss of the SM-type LIM code. This observation at the molecular level may well explain why ventrally projecting motor neurons are missing at the r5-r7 levels in the mutant (Fig. 5), which resulted in the loss of the abducent and hypoglossal nerves (Fig. 3). To be emphasized here is that loss of expression of Islet-2 in SM neurons was tightly linked to misextension of their axons; only the abducent and hypoglossal nerves were affected that were generated from the r5-r7 levels in which loss of Islet-2 occurred in the mutant. On the other contrary, primary axonal growth of other cranial SM neurons, the oculomotor and trochlear ones, as well as other spinal motor neurons, appeared intact and Islet-2 expression was detected in these neurons. In this sense, it is interesting to note previous morphological studies pointing out the possibility that the default state of the axonal growth and sprouting pattern in the postotic hindbrain is a dorsally growing one (Goodrich, 1930; Fritzsch and Northcutt, 1993), and thus Islet-2 expression can change the direction to ventrally.

Change in LIM code is likely to cause mislocation of motor neurons, finally resulting in loss of the abducent and hypoglossal nuclei. Interestingly, in the chick hindbrain, Islet-2 is downregulated in the accessory abducent neurons, which subsequently have a distinct LIM code (Islet-1 + Lim-3), migrate laterally and diverge from the main abducens nucleus (Varela-Echavarria et al., 1997a). In the Pax-6 mutant, Islet-2 expression was not induced, resulting in the same LIM code as that of the accessory abducens neurons; these cells migrated laterally within the neural tube. It is likely that such misspecified motor neurons cannot survive since Islet-1-positive cells decreased in number from E13 onward, resulting in loss of the nuclei. At r5 (nVI) level, misspecified motor neurons were observed to stray out from the neural tube from E11. This may be related to the fact that the specific rhombomere is unique because no cranial nerve root is formed (see Fig. 8A) and an atypical somatic motor nucleus, the accessory abducens, which has a different LIM code (Islet-1 + Lim-3; Varela-Echavarria et al., 1997a) develops. Whatever the case, as a result, the abducent nucleus was not formed in the mutant.

Since motor neurons themselves do not seem to express Pax-6, the defects are considered to be induced by extracellular signals emanating from the Pax-6-expressing region. It remains possible, however, that some cell-autonomous mechanisms may be involved in the specification of motor neuron subtypes. For example, many of ventral cell types in the neural tube are derived from Pax-6-positive progenitors and they are missing in Small eye mice (Ericson et al., 1997). Yet, the question remains why the defect in SM nerves is specifically seen in the postotic hindbrain, although Pax-6 is expressed throughout the hindbrain and spinal cord. Preliminary experiments demonstrated that some of the hindbrain markers, e.g., Hox-b1, Krox-20 and CRABP I, did not show significant differences from those of the wild-type (data not shown). In addition, branchial nerves whose roots are derived from the specific rhombomeres developed normally in the mutant (Fig. 3). Thus, we consider that overall regional specification of the hindbrain along the A-P axis is not affected in the Pax-6 mutant. Therefore, it is speculated that some other factors are involved in the specification of SM neurons in the hindbrain.

Since some members of the Wnt gene family that encode secreted molecules are expressed in the ventrolateral region overlapping the Pax-6 domain in the mouse and chick neural tube (Parr et al., 1993; Hollyday et al., 1995), we surveyed expression patterns of several Wnt genes in the rat hindbrain as candidates that regulate Islet-2 expression. Our results revealed that Pax-6 appears essential for proper expression of Wnt-7b in the ventrolateral region of the neural tube, where its expression overlaps with that of Pax-6. Of note, in normal development motor neurons juxtapose to the ventral boundary of Pax-6/Wnt-7b expression. It is therefore an intriguing possibility that Wnt-7b, which is secreted from the Pax-6-positive domain, plays an essential role in the induction of Islet-2 adjacent to the Pax-6-negative domain. In support, an analogous genetic cascade is reported in developing vertebrate limb buds; expression of Lmx-1, another member of LIM family, in the dorsal mesenchyme is induced by Wnt-7a localized in the dorsal ectoderm, thereby establishing the dorsoventral pattern in the limb bud (Vogel et al., 1995; Riddle et al., 1995). Further, fly paired regulates wingless gene (see Noll, 1992), and Wnt-1 expression in the midbrain/hindbrain boundary is controlled by Pax-2, 5, 8 (Krause et al., 1991; Rowitch and McMahon, 1995), genetic interactions also analogous to Pax-6 and Wnt-7a. With respect to the phenotypic changes of the SM neurons, Wnt-7b may be involved in either specification of neuronal subtypes or axon pathfindings of SM neurons.

Overlapping expression patterns of ventrally expressedWnt genes are regarded to be one possible explanation for the SM neuron defects in the postotic hindbrain. In the spinal cord, expression of Wnt-5a overlapped with Wnt-7b in their dorsoventral expression domains (Parr et al., 1993; and our unpublished results). In contrast, their A-P expression domains appeared complementary, i.e., Wnt-7b is expressed from r2 to the spinal cord, while Wnt-5a is not expressed in r2-r6 but is expressed in r1 and the spinal cord (Fig. 7). In r7, Wnt-5a was not expressed in the region adjacent to the motor column (our unpublished results). Therefore, in the spinal cord, the function of Wnt-7b in the ventrolateral region may be compensated by Wnt-5a, thus resulting in no obvious phenotypic anomaly in the spinal motor neurons. Although Wnt-7a shows the highest homology with Wnt-7b in its amino acid sequence (Gavin et al., 1990) and is expressed throughout the hindbrain (Parr et al., 1993; and our unpublished results), it appears that it is not able to rescue the loss of function of Wnt-7b. Among possible functions of the Wnt genes are cell-fate determination, proliferation and cell survival (Nusse and Varmus, 1992; McMahon, 1993; Perrimon, 1994). It would be interesting to examine whether Wnt-7b homozygous knock-out mice would show any defect in the axonogenesis of the abducent and hypoglossal nerves.

Another possibility to be considered is failure in pathfinding of the SM axons. Recently, great progress has been made in the isolation and characterization of various molecules involved in axonal guidance (for review see Tessier-Lavigne, 1994; Tessier-Lavigne and Goodman, 1996). The floor plate has been shown to chemoattract interneuron axons and chemorepulse axons of motor neurons (Kennedy et al., 1994; Colamarino and Tessier-Lavigne, 1995; Tamada et al., 1995; Guthrie and Pini, 1995). Very recently, it is shown that motor axon subpopulations respond differentially to netrin-1 and semaphorin D; dorsally projecting axons (BM/VM-type) were repelled by both, while ventrally projecting ones (SM-type) responded to semaphorin D but not to netrin-1 (Varela-Echavarria et al., 1997b). It is also possible that the dorsal region including the exit points of the BM/VM axon may produce diffusible tropic and/or trophic molecules that induce BM/VM axon outgrowth (Tucker et al., 1996). Considering the fact that the ventral and dorsal limits of Pax-6 expression corresponded to the exit points of SM and BM/VM nerves, respectively, it is likely that the Pax-6 domain defines the specific sites for different motor axons to leave the neural tube by controling some chemorepellent/attractant. This idea is supported by other developmental evidence in zebrafish showing that expression boundaries play crucial roles in growth cone guidance in the forebrain (Wilson et al., 1993).

The authors would like to thank Dr Randall Reed for kindly providing anti-Pax-6 antibody, Drs Sarah Guthrie for valuable unpublished informations on hindbrain LIM codes and Thomas Jessell on ventral patterning in Small eye mice, Drs Fujio Murakami, Hajime Fujisawa and Andrew Lumsden for fruitful discussion, and Dr Ruth Yu for critical reading of the manuscript. Sincere gratitude is extended to Kouichi Yahagi for maintaining animals, Sayuri Tomonari and Keiko Yamamoto for analyses of the mutant gene, and Masaaki Torii and Tadahiro Kaneko for isolation of rat Wnt cDNAs for in situ hybridization. The 2H3 and 40.2D6 antibodies developed by T. M. Jessell and J. Dodd were obtained from the Developmental Studies Hybridoma Bank maintained by the Department of Pharmacology and Molecular Sciences at Johns Hopkins University School of Medicine Baltimore, MD, and the Department of Biology at the University of Iowa, Iowa City, IA, under contract number NO-HP-2-3144 from the NICHD. This work was supported by Grants-in Aid for Scientific Research from the Japanese Ministry of Education, Science and Culture (No. 05557078, 06671805, 08457475).