open brain, a new mouse mutant with severe neural tube defects, shows altered gene expression patterns in the developing spinal cord

During embryonic development of the central and peripheral nervous system several developmental processes can be distinguished. First, neural ectoderm is induced by interactions with the mesoderm to define the neural plate (Spemann and Mangold, 1924; Baker, 1927; reviewed by Ruiz i Altaba and Jessell, 1993). The induced neural plate then folds inwards and rounds up to form the neural tube (Schoenwolf and Franks, 1984). In a third phase, the different neuron populations within the neural tube differentiate into specialized neurons and begin to establish functional neuronal circuits (Bancroft and Bellairs, 1975; Altman and Bayer, 1984). Before and shortly after closure of the neural tube, the neural crest cells delaminate from the dorsal neural tube and migrate to various regions of the embryo to differentiate into dorsal root ganglia and other components of the peripheral nervous system (Le Douarin et al., 1993).

Many studies have been performed to understand the specification of neuron populations within the central nervous system, especially in the developing spinal cord. During this specification phase, important signals come from the underlying mesodermal structure, the notochord and midline cells of the neural ectoderm, which later differentiate into floor plate cells (Jessell et al., 1989). Induction of the floor plate requires contact between notochord and the medial cells of the neural plate (Kitchin, 1949; Placzek et al., 1991), whereas differentiation of the first neuron populations, the motor neurons located lateral to the floor plate, involves signals from both floor plate cells and the notochord (Hatta et al., 1991; Yamada et al., 1993). Recently a TGF-β-like gene, dorsalin-1, has been described which is expressed in the dorsal region of the spinal cord and is thought to act as a dorsal signal (Basler et al., 1993). The specification of neurons along the dorsoventral axis of the spinal cord may thus be directed by the combination of both dorsal and ventral signals (Basler et al., 1993).

Many genes exhibit very specific expression patterns along the dorsoventral axis of the developing neural tube. For example, Pax-3 and Msx-2 gene expression during normal mouse development is detected at day 12.5 p.c. in the dorsal ventricular zone of the developing spinal cord (Goulding et al., 1991; Takahashi et al., 1992) and Pax-6 expression has been described in a subregion of the ventricular zone in the basal plate (Walther and Gruss, 1991). Goulding et al. (1993) demonstrated that the expression patterns of both Pax-3 and Pax-6 change if the specification of cells in the spinal cord is experimentally altered in chicken embryos. Similarly, alterations in the specification of neurons either in mutant mouse strains (Phelps and Dressler, 1993; Dietrich et al., 1993) or after ectopic expression of the HNF-3β gene (Sasaki and Hogan, 1994) are accompanied by changes in Pax gene expression patterns. Other genes show expression in specific populations of cells in the developing neural tube, like Shh (Echelard et al., 1993) and HNF-3β (Sasaki and Hogan, 1993) in the floor plate, Wnt-3a in dorsalmost cells including the roof plate (Roelink and Nusse, 1991; Parr et al., 1993), and Islet-1 in the developing motor neurons (Ericson et al., 1992). Therefore, the expression patterns of these genes can be regarded as molecular markers to monitor alterations in the dorsoventral specification of cells in the developing neural tube.

In this report we describe a new mouse mutation, open brain (opb), which results in severe malformations of the developing nervous system, the eyes, the skeleton and the developing limbs. The phenotype in the central nervous system was characterized by an exencephaly and an abnormal shape and structure of the spinal cord. Because of the defect in cranial neural tube closure, we named this mutation open brain. Defects in specification of cells along the dorsoventral axis in the spinal cord are accompanied by alterations in gene expression patterns.

All mouse strains (opb mutants, NMRI and C57Bl/6) were maintained at the Max-Planck-Institute of Immunebiology under special pathogen-free conditions.

Detection of vaginal plugs in the morning after matings was taken as day 0.5 postcoitum (p.c.). Embryos were prepared at different stages of development, fixed overnight with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS), dehydrated and embedded in paraffin. The tissue was sectioned at 6-8 µm and tissue sections were stained with hematoxylin and eosin. Skeletal preparations were done as described by Kessel et al. (1990).

Embryos were prepared as described above for histological analyses. In situ hybridizations on sections were performed as described by Kessel and Gruss (1991). Bright- and dark-field views of hybridized and exposed sections were recorded with a video camera, digitized, artificially colored (blue for bright-field and red for dark-field images) using Adobe Photoshop^®, and bright- and dark-field views were superimposed. Whole-mount in situ hybridizations were performed as described by Rosen and Beddington (1993) and Wilkinson (1992). For Pax-3 in situ hybridizations, a 520 nucleotide (Nu) riboprobe (Goulding et al., 1991), for Pax-6 a 260 Nu riboprobe (Walther and Gruss, 1991), for Msx-2 a 400 Nu riboprobe (Monoghan et al., 1991), for Wnt-3a a 750 Nu riboprobe (Roelink and Nusse, 1991), for Shh a 640 Nu riboprobe (Echelard et al., 1993), for HNF-3β a 1500 Nu riboprobe (Sasaki and Hogan, 1993) and for Pax-1 a 1500 Nu probe were generated (A. Neubüser and R. Balling, personal communication). Intensities of in situ hybridization signals from tissue sections were measured using Image 1.45^®. One representative section was chosen from an unaffected littermate and one from a mutant embryo. Both sections were hybridized with the same probe under the same conditions and the same parameters were used for calculations. Using the above semiquantitative analysis, the same background levels were observed in both sections. Staining of embryos for lacZ reporter gene expression was performed as described (Zakany et al., 1988).

Sections of 6 µm were cut on a cryostat and collected onto poly-L-lysine-pretreated slides. Sections were fixed with 4% PFA/PBS and incubated overnight at 4°C with antibodies against Islet-1 (diluted 1:50, kindly provided by Drs S. Morton and T. Jessell, New York, USA) in PBS, 3% fetal calf serum. Sections were washed with 0.01 M Tris, 0.15 M NaCl, 0.1% Tween20, pH 7.4, incubated with horse-radish peroxidase-conjugated secondary antibody (HRP-conjugated goat anti-mouse IgG, Sigma, diluted 1:750) washed with Tris saline, 0.1% Tween20, pH 7.4 and incubated in diaminobenzidine, 0.03% ammonium nickel sulfate, 0.003% H₂O₂. Sections were dehydrated and mounted in Eukitt (Riedel-de-Haen).

opb mutant embryos were originally discovered in litters produced to obtain homozygous transgenic mice carrying a Hoxb-6/lacZ reporter gene construct (Schughart et al., 1991). In this line, the reporter gene was expressed in the limbs, the lateral body wall and the dorsal root ganglia (DRG). However, upon further breeding of transgenic opb carriers, the mutant phenotype and the transgene segregated, suggesting that the opb mutation was not caused by the insertion of the transgene but instead occurred in the genetic background of the breeding colony used to maintain the transgenic line.

The original carriers were of a mixed outbred Albino background (NMRI). To generate a congenic mutant mouse line, opb carrier males were bred to C57Bl/6 females. Offspring were interbred and mutant embryos were identified between day 9.5 and 15.5 p.c. by their exencephaly and/or the pronounced protrusion of the spinal cord. Male and female carriers, which otherwise had no recognizable phenotype, could be identified in this way. The mutant line was maintained only through male carriers because females were killed to analyze their litters for opb mutant embryos. To date, the opb mutation has been bred for four subsequent generations onto a C57Bl/6 background, with no detectable alterations in the penetrance of the opb phenotype. However, the expressivity of cranial neural tube defects decreased somewhat with an increasing C57Bl/6 background.

The opb mutation exhibited a segregation pattern typical for an autosomal recessive mutation at a single locus with full penetrance. Identified opb carrier males were bred to wild-type C57Bl/6 females, offspring from this mating were intercrossed and embryos scored for the opb phenotype. Of 744 embryos examined from 89 intercrosses, mutant embryos were observed in 24% of the matings (expected: 25% for an autosomal recessive mutation) and within litters with mutant embryos, 23% of the embryos (expected: 25%) were affected.

Within the developing central nervous system, a severe exencephaly was apparent in the brain of day 12.5 p.c. opb mutant embryos (Fig. 1B,D). The affected region involved the forebrain, midbrain and hindbrain. This defect in the closure of the neural tube was already evident at day 9.5 p.c., at a time when the cephalic neural tube had completely closed in normal littermates (Fig. 2C). At day 10.5 p.c., opb embryos exhibit an open neural tube extending from the developing diencephalon into the future myelencephalon, to about the level of the otic vesicle (Figs 2B, 3A). The most rostral regions of the forebrain were closed (Fig. 2B).

In histological sections, the gross subdivision of the brain anlage was discernible at day 10.5 p.c. (Fig. 3A). Formation of the telencephalic vesicles was abortive. The mediobasal portions of the diencephalon were formed in a quasi-normal manner, i.e., the hypothalamus produced a median extension which touched Rathke’s pouch and resembled the infundibulum of the pituitary gland (Fig. 3A). In the midbrain, the mediobasal structures were again least affected and several regions resembling anlagen of cranial nerves were discernable. However, the lateral structures were profoundly malformed. In the hindbrain, segmentation in rhombomeres was detectable (Fig. 3A). The cranial nerve ganglia V and VIII appeared normal. In most embryos, no enhanced cell death was seen at embryonic day 10.5 p.c. However, in several embryos aged 12.5 p.c., cells within the malformed portions of the brain underwent extensive necrosis (Fig. 3B). Necrosis was present in a pseudolaminar pattern and affected exclusively the ependymal and immediately subependymal layers. In contrast, the external two thirds of the ventricular zone and the mantle layers were largely unaffected. Along the craniocaudal axis, the necrotic zone ended in the myelencephalon. This region coincided with the caudal limit of the exencephalic malformation, suggesting that cell death was caused by exposure of neural tissue to amniotic fluid. Also, exencephalic opb mutant embryos showed strong malformations of the frontonasal, maxillary and mandibular processes (Fig.1B).

The expressivity of the open neural tube defect in the brain varied somewhat and seemed to decrease with an increasing C57Bl/6 background. Of the embryos identified with prominent spinal cord and dorsal root ganglia defects, 17% (total of 46) showed a closed neural tube in the brain region. Whether all brain structures develop normally in these embryos is currently being investigated. Some embryos (3 of 8) with no obvious exencephaly developed pronounced general edemas, which may have been caused by incomplete neural tube closure. In addition to neural tube defects in the cranial region, severe malformations of the spinal cord were observed. The neural tube in day 12.5 p.c. embryos protruded from the back of the embryo along the thoracic and lumbosacral region (Fig. 1B). This protrusion was accompanied by a circular to oval shape of the neural tube and an enlargement of the central canal (Fig. 3D). The severity of the misshapen spinal cord varied from completely circular to almost normal (see also Figs 4, 5, 6) and in a few 12.5 day p.c. embryos (2%), a myeloschisis (open neural tube) was also apparent in the lumbosacral region. Also, the severity of the spinal cord defects varied between individual embryos. In the most severely affected embryos, a strongly malformed spinal cord was observed throughout the thoracic and lumbar region whereas in less severly affected embryos a smaller region along the anteroposterior axis showed a pronounced malformation. In mildly affected embryos, the spinal cord appeared almost normal throughout its entire length.

In all affected opb embryos, a floor plate was identified but, in most embryos, a roof plate structure was not histologically discernible (Fig. 3D). In the most severely affected spinal cord regions, the ventral horn, which corresponds to the future motor neuron horn, seemed to be reduced in size (Fig. 3D,F). Analyzing expression of Islet-1 protein as a marker for differentiating motor neurons (Ericson et al., 1992), it became evident that in those embryos only very few motor neuron precursors were generated (Fig. 3H). The degree of reduction of differentiating neurons of the future ventral horns correlated with the degree of the malformation of the neural tube itself. Differentiating neurons in the ventral horn were drastically reduced in embryos with a circular-shaped neural tube (Fig. 3D) whereas in embryos with almost normally shaped spinal cords differentiated neurons were present. In addition, only a few axonal projections originating from the developing motor neurons of the ventral horn into the ventral root were observed in severely affected spinal cord regions (data not shown). Axons of the presumptive ventral funiculi as well as axons in the future ventral commissure were present in all opb embryos (Fig. 3D). Thus controlaterally as well as ipsilaterally projecting relay neurons appeared to be differentiating.

We also observed severe abnormalities in the presumptive meninx primitiva at day 12.5 p.c. This anlagen, roughly defined as the region directly surrounding the neural tube and containing many blood vessels, represented only a relatively thin region of tissue around the neural tube of wild-type embryos (Fig. 3C) but was extremely enlarged in opb mutant embryos (Fig. 3D, arrow).

Furthermore, in opb embryos the organization of dorsal root ganglia (DRG) was drastically disturbed. In early generations (before the Hoxb-6/lacZ transgene and the opb mutation had segregated) DRG could easily be visualized by X-gal staining in opb mutant embryos. In normal embryos, the DRG were organized in a metameric fashion along the lateral sides of the spinal cord (Fig. 1E).

However, in opb mutant embryos, DRG were much smaller and completely disorganized (Fig. 1F). The ganglia in opb embryos were not well separated from each other or from the neural tube and metameric organization was lost completely (Fig. 1F). Most of the ganglia were found in much more dorsal positions than normal (Fig. 1F) and several ganglia were even found in a position dorsal to the neural tube (Figs 1F, 3D).

To describe the defects in opb mutant embryos in more detail, molecular markers were used to examine the organization of neuron precursors along the dorsoventral axis of the spinal cord. Expression of Wnt-3a and Msx-2 were studied to monitor the presence of roof plate cells, expression patterns of Shh and HNF-3β were examined to analyze the development of the floor plate. In addition, Pax-3 and Pax-6 genes were studied because alterations in dorsoventral specification of neurons in the spinal cord are accompanied by changes in their expression patterns (see Introduction).

In wild-type embryos, Wnt-3a was expressed at day 12.5 p.c. in the dorsalmost region of the developing spinal cord including the roof plate (Fig. 4A). In contrast, in severely affected spinal cord regions of opb mutant embryos, no expression of Wnt-3a could be detected (Fig. 4B). Similarly, the pattern of Msx-2 gene expression was altered. In normal embryos, we observed weak Msx-2 expression throughout the ependymal layer of the alar plate (Fig. 4C) and a strong signal was found in dorsalmost cells as described by Takahashi et al. (1992; Fig. 4C). In severely affected spinal cord regions of opb mutant embryos, weak expression of Msx-2 was detected in the alar plate but the strong signal in dorsalmost cells was absent (Fig. 4D). In addition, a dorsalmost region was evident in opb embryos where no expression of Msx-2 was detected (Fig. 4D, arrowhead). Pax-3 gene expression at day 12.5 p.c. is normally restricted to proliferating neurons in the ventricular layer of the alar plate (Fig. 4E; Goulding et al., 1991). In severely affected spinal cord regions of opb embryos, we observed a region of cells in the dorsalmost part in which no Pax-3 gene expression was evident (Fig. 4F, arrowhead).

In wild-type embryos, Pax-6 gene expression was described near the sulcus limitans in the basal plate of the ventricular layer (Walther and Gruss, 1991). In addition, we observed low level Pax-6 expression in normal embryos in the ependymal layer of the alar plate (Fig. 5A). Similarly as for Pax-3, a dorsalmost region was seen in opb embryos in which Pax-6 gene expression was not detected (Fig. 5B, arrowhead). In addition, the expression of the Pax-6 gene in other regions of the presumptive alar plate of opb embryos was significantly stronger than in wild-type embryos (Fig. 5B, rectangle; Fig. 5E, arrow). Semi-quantitative comparisons of in situ hybridization signals revealed about 2-to 3-fold increased levels of Pax-6 expression in the dorsal ependymal layer of opb embryos (Fig. 5D) as compared to normal littermates (Fig. 5C). In the ventral half of the spinal cord, expression of Pax-6 was normal except that the ventralmost region negative for Pax-6 expression appeared to be enlarged in opb embryos compared to wild-type (Fig. 5B,E). In less severely affected spinal cord regions expression of Pax-6 extended to the dorsalmost cells but still, an elevated level of Pax-6 expression could be observed in the dorsal half (Fig. 5E, arrow).

Shh gene expression could be found in the ventral spinal cord region (presumptive floor plate) and the notochord of both opb mutant and wild-type embryos (Fig. 6A,B). In opb mutant embryos, however, a much larger area of Shh (Fig. 6B, arrow) and also of HNF-3β (Fig. 6D, arrow) positive cells was observed indicating a more extended floor plate region in the malformed neural tube.

In summary, these results suggest that, although not clearly evident at the histological level, at the molecular level a pronounced alteration in the dorsoventral specification of cells in the spinal cord must have occurred in opb mutant embryos.

At day 9.5 p.c., a normally extended optic stalk was apparent in both wild-type and mutant embryos. In slightly older embryos, further development of the distal region of the optic stalk into a cup occurred, but in opb mutant embryos an abnormal flattened appearance of the optic cup was evident (data not shown). At day 12.5 p.c., defects in the eye anlagen of opb mutant embryos were clearly detectable. In normal embryos, the optic cup had induced the formation of a lens and two cell layers, the inner layer (future neural retina) and the outer layer (future pigment layer) were formed (Fig. 7A). However, in many opb mutant embryos, no induction of a lens vesicle was observed, the neural layer appeared to be completely absent in some embryos and the pigment layer, although present, was misshapen and greatly reduced in size (Fig. 7B, arrow). In many cases, the eye anlagen did not develop at all. The phenotype of the eye in opb embryos was variable even within a single embryo, with some embryos exhibiting the described severe eye phenotype on one side, whereas eye development on the other side proceeded to the induction of a lens vesicle. But even the more developed eye anlagen exhibited only very small lens anlagen and the ventral fissure did not close (data not shown). Interestingly, eye defects were also found in those opb mutant embryos in which the cranial neural tube appeared to have closed.

Malformations were also observed in the axial skeleton of opb mutants (Fig. 7D). Vertebrae and intervertebral discs in the thoracic and lumbar regions were formed, in several embryos; however, the anlagen of the vertebral bodies were irregular in shape and appeared to be split in half (data not shown). Neural arches were misshapen and abnormal fusions of the processi articulari were observed (Fig. 7D). The distal parts of many of the developing ribs were split and/or fused, and the angulus costae showed a hairpin-like structure (Fig. 7D, arrow and arrowheads). To study the organization of developing somites in opb embryos, whole-mount in situ hybridizations were performed with Pax-1 and Pax-3-specific probes (data not shown). In normal 10.5 day p.c. embryos, Pax-1 is expressed in the early ventral somite which will later differentiate into sclerotome (Brand-Saberi et al., 1993; Dietrich et al., 1993) whereas Pax-3 is expressed in the future dermomyotome; strongly along the caudal and ventrolateral edge of the somite and weaker in the rostral and dorsolateral regions (Goulding et al., 1993; Bober et al., 1994; Dietrich et al., 1993; Williams and Ordahl, 1994). Expression patterns of both genes revealed that in opb embryos the paraxial mesoderm in the thoracic region becomes properly segmented (in contrast to DRG). The development of dorsal somitic regions, however, appeared to be affected in opb mutant embryos where an abnormal, flame-like pattern was observed instead of the regular arch-shaped Pax-3 pattern seen in wild-type embryos. Further studies are underway to describe these defects in more detail.

Alterations in developing limbs were clearly detectable in skeletal preparations of day 15.5 p.c. old opb embryos. Both forelimbs and hindlimbs exhibited supernumerary digits (Fig. 7F). In the hindlimbs shortening of the fibula and greatly reduced tibia anlagen were observed (Fig. 7F), whereas radius and ulna of the forelimbs appeared to develop normally. Closer inspection of the cartilage anlagen suggested that duplications in preaxial regions had occurred in the autopod. In the developing hindlimb, the second and third distal tarsals and the cuboid were formed normally, but the first distal tarsal could not be unambiguously identified. Instead, an anteriorly located distal tarsal was visible which resembled an additional second tarsal anlagen. This tarsal exhibited a short anterior extension (Fig. 7F, arrowhead) that articulated with the supernumerary digit in the mutant hindlimb. Also, the navicular was elongated anteriorly and a distinct anlage of the tibiale was absent (Fig. 7F, arrow). These observations suggest that the additional digit present in the hindlimb most likely represented a duplication of digit I or digit II. Similarly, supernumerary digits in the forelimb were observed which also represented preaxial digit duplications (data not shown). Furthermore, the scapula showed various malformations such as a scalloped surface of the anterior and dorsal regions (margo cranialis and margo dorsalis) and a single perforation of the facies costalis (Fig. 7D).

We have described a new mouse mutant, open brain (opb), with severe malformations in the developing nervous system, the eyes, the axial skeleton and the developing limbs.

Malformations in the brain of opb mutant embryos were manifested as an exencephaly extending from the forebrain to the hindbrain regions. During normal mouse development, three sequential points of neural tube closure in the developing brain have been described (Sakai, 1989; Copp et al., 1990): (1) at the hindbrain-spinal cord (HB/SC) boundary, (2) at the very rostral point, in the anterior forebrain (aFB) and (3) at the forebrain-midbrain (FB/MB) boundary. From these points, closure spreads in both cranial and caudal directions from the FB/MB and HB/SC closure points and proceeds backwards from the aFB closure point. In opb embryos, the neural tube in the most rostral region of the forebrain as well as in the posterior hindbrain/spinal cord region was closed, suggesting that initial closure occurs at the aFB and HB/SC but not the FB/MB boundary. In addition, rostral movement of neural tube fusion in opb embryos is severely affected after initial closure at the HB/SC boundary and the initial fusion point starting in the aFB moves only into the most rostral forebrain regions, resulting in an open midbrain and hindbrain and partially open forebrain.

Several other mouse mutants with similar defects in neural tube closure have been described (reviewed by Copp et al., 1990). Failure of initiation of neural tube closure at the FB/MB boundary has been described in detail for the SELH/Bc mouse strain (MacDonald et al., 1989, Juriloff et al., 1989; Table 1). Although all SELH/Bc embryos fail to initiate closure at the FB/MB boundary, in most embryos de novo closure in the anterior forebrain occurs and closure proceeds backwards through the entire midbrain region to generate a completely normal neural tube. The primary cause of failure of neural tube closure in SELH/Bc embryos and most other mutant mouse strains is not known. Alterations in the expression or function of genes that regulate cell adhesiveness or cell shape might be responsible for the defect (reviewed by Copp et al., 1990; Schoenwolf and Smith, 1990). Alternatively, abnormal proliferation in either the neural tube or the surrounding tissue leading to mechanical tensions may cause failure of neural tube closure. This was demonstrated in the mouse mutant curly tail (ct) where abnormal bending of the neural tube in the lumbosacral region caused by hypoproliferation of notochord and hindgut tissues results in an open neural tube (van Straaten et al., 1993). Further analysis of opb and other mutations will help to elucidate the underlying mechanisms that play a role in neural tube closure.

Most of the other malformations in the brain of opb embryos appear to be consequences of the defect in neural tube closure. No specific abnormalities were seen in the lamination and radial arrangement of differentiating neuroectodermal precursor cells within the dysplastic areas, suggesting that cell differentiation in brain regions was unaffected.

The malformations in the spinal cord of opb embryos are characterized by alterations in both ventral and dorsal regions. This represents a unique feature of opb embryos since all other known mouse mutations with similar defects (Table 1) exhibit malformations in either the ventral or the dorsal half of the spinal cord, e.g. Danforth’s short tail (Dunn and Gluecksohn-Schoenheimer, 1940) and Splotch (Auerbach, 1954).

Specification of neurons along the dorsoventral axis involves both a ventral signal emanating from the notochord and floorplate, and a dorsal signal thought to be generated from dorsalmost cells originating from the lateral regions of the early neural plate (see Introduction). Both signals are thought to be interdependent. If the notochord is removed in chicken embryos or degenerates in mutant mouse embryos (i.e. Danforth’s short tail) the expression of Pax-3, normally restricted to the dorsal half of the neural tube extends ventrally (Goulding et al., 1993). In contrast, implantation of an additional notochord dorsolateral to the neural tube results in repression of Pax-3 and Pax-6 in the alar plate (Goulding et al., 1993).

In ventral spinal cord regions of opb mutant embryos, the notochord and floor plate differentiated. Also, expression of Shh in floor plate and notochord cells at day 12.5 p.c. (Echelard et al., 1993) and HNF-3β in floor plate cells (Sasaki and Hogan, 1993) was observed in opb embryos indicating that the initial ventral signal was established. However, the areas of Shh and HNF-3β gene expression in opb mutant embryos were considerably larger than in wild-type embryos. In the dorsal half of the neural tube, an elevated level of Pax-6 gene expression was observed and altered expression patterns of several other genes (Pax-3, Pax-6, Msx-2 and Wnt-3a) were evident in opb embryos. During normal development, Pax-3 and Msx-2 expression is detected in the dorsal ventricular zone of the developing spinal cord extending to the dorsalmost margin of the neural tube (Goulding et al., 1991; Takahashi et al., 1992). Pax-6 expression was detected in a subregion of the ventricular zone in the basal plate of the spinal cord (Walther and Gruss, 1991) and weakly throughout the dorsal ventricular zone. opb embryos, however, exhibited a dorsalmost region in which neither Pax-3, Pax-6 nor Msx-2 genes were expressed. Most strikingly, no expression of Wnt-3a, which is normally evident in dorsalmost cells, including the roof plate, (Roelink and Nusse, 1991; Parr et al., 1993) was detected in severely affected spinal cord regions of opb embryos. Likewise, the strong signal of Msx-2 gene expression normally found in dorsalmost cells (Takahashi et al., 1992) was not evident in opb embryos. The latter observations indicate that no roof plate cells were differentiating in opb embryos.

In summary, these results suggest that in opb embryos specification of cells along the dorsoventral axis of the spinal cord is perturbed. The observed alterations can be explained in two ways. First, the initial or subsequent dorsal signals may be missing or much weaker than in wild-type embryos resulting in loss of expression of genes in dorsalmost regions. Alternatively, the ventral signals may be much stronger than normal and thus the ventral domain expands dorsally causing changes of developmental fate of cells in the alar plate. Since in chicken embryos induction of an additional floor plate in dorsolateral regions of the neural tube by notochord transplantations did not result in complete loss of roof plate cells or complete suppression of dorsal genes in dorsalmost regions (Yamada et al., 19991; Placzek et al., 1991; Artinger and Bronner-Fraser, 1992), we currently favor the first explanation.

Also, in opb mutant embryos differentiating motor neurons were drastically reduced in numbers. At present, this defect cannot be explained by the observed defects in dorsoventral specification of the neural tube. It is therefore more likely that opb embryos exhibit an additional defect in the differentiation of motor neuron precursor cells.

In opb mutant embryos, DRG were found in very dorsal instead of lateral positions, they were much smaller and showed no metameric organization along the body axis. These malformations may be due to an intrinsic defect in neural crest cells and/or a defect in somitic tissue. Neural crest cells migrate in a ventral direction to their final location through very specific pathways (Serbedzija et al., 1990) and those neural crest cells destined to form DRG migrate through somitic tissue. Defects in segmentation of paraxial mesoderm could therefore result in abnormal organization of DRG. Analysis of Pax-1 and Pax-3 gene expression patterns in opb mutant embryos, however, revealed that segmentation of the paraxial mesoderm into somites was not disturbed. It has been demonstrated that the interaction with the rostral half of the somites seems to be crucial for proper metameric organization of DRG (Davies et al., 1990; Kalcheim and Teillet, 1989). If the rostral halves of the somites are extirpated in chicken embryos, only very small ganglia form, which are not organized in a metameric fashion (Kalcheim and Teillet, 1989), very similar to the defects seen in opb embryos. However, in embryos where rostral somite halves had been removed, intervertebral disc anlagen were also absent or much reduced in size (Goldstein and Kalcheim, 1992). This defect was not observed in opb. Thus, in opb embryos, the paraxial mesoderm becomes properly segmented and the rostral halves of the somites can differentiate to a large extent, suggesting that the basis for the DRG abnormalities most likely resides in an intrinsic defect of neural crest cells. Additionally or alternatively, the abnormalities observed in dorsal somitic regions of opb embryos may influence the proper organization of DRG.

Furthermore, DRG formed in abnormal dorsal and dorsalmost locations in opb embryos. It has been demonstrated in chicken embryos that the dorsoventral polarity of the developing neural tube influences the direction of migration of neural crest cells that form DRG (Keynes and Stern, 1984). If migrating neural crest cells must communicate with the neural tube to determine their direction of migration (Kalcheim et al., 1987), one would expect that the precursors for DRG in opb embryos are guided into abnormal positions by the abnormal neural tube.

Neural crest cell development is also affected in Splotch (Sp) mutant embryos (Table 1), which is caused by a mutation in the Pax-3 gene (Epstein et al., 1991). In these mutants, DRG are smaller, but metamerization of ganglia is still recognizable and displacement to a dorsalmost position does not occur. In addition, Sp mutants generally exhibit a rostrocaudal gradient in the severity of DRG defects (Auerbach, 1954). Instead, the severe defects in the organization of DRG in opb embryos become established within only a few segments in the upper thoracic region.

The observed defects in the neural tube and dorsal root ganglia in opb embryos were much more dramatic than the skeletal defects. Since it has been shown that the neural tube influences the maintenance and the specification of cells in the somitic mesoderm (Packard and Jacobson, 1976; Teillet and Le Douarin, 1983), it is possible that the defects in the axial skeleton of opb mutant embryos represent a secondary defect caused by the malformed neural tube.

The defect in the development of the eyes in opb embryos most likely represents an intrinsic defect in the developing eye anlagen and is not linked to the exencephalic phenotype per se, since eye defects were also observed in embryos in which the neural tube appeared closed. Pax-6 expression patterns were altered in spinal cord regions of opb embryos and the same could be true for the eye anlagen. Such alterations in Pax-6 gene expression could result in developmental defects of the eye because mutations in the Pax-6 gene are known to result in eye defects as described for the Small eye (Sey) phenotype (Matsuo et al., 1993). The induction of a lens anlage was observed in several opb embryos, indicating that the response of the ectoderm to induction by the optic cup (Henry and Grainger, 1990) was not disturbed. The eye defects in opb resemble defects described for homozygous Extra toes (Xt/Xt) embryos (Table 1; Franz and Besecke, 1991) and it will be interesting to compare eye development in opb, Xt and Sey embryos.

opb mutant embryos also exhibited several malformations in the developing limbs, i.e. preaxial digit duplications. Such duplications characterize many mouse mutants, but only the Xt mutation shows phenotypic alterations affecting both neural tube closure and limb development (Table 1; Johnson, 1967). The regional extent of the exencephaly observed in both Xt and opb mutant embryos is similar, but Xt does not show a defect in development of the DRG nor does it exhibit a misshapen spinal cord (Johnson, 1967). In Xt mutants, the neural tube and limb defects are caused by a mutation in the gli-3 gene (Schimmang et al., 1992), which is normally expressed in both structures. Possibly opb embryos exhibit a mutation in a gene expressed in both the developing nervous system and limbs.

The chromosomal location of the opb mutation has not yet been determined. Preliminary linkage studies suggest that opb segregates from the Sp and Xt loci (T. G. and K. S., unpublished results). The opb mutation therefore seems to represent a new mutation in a yet unknown gene locus.

In conclusion, the opb mutant represents a valuable mouse model to analyze the molecular and cellular basis of neural tube defects. opb might also serve as a valuable genetic model to study embryonic processes such as specification of cells along the dorsoventral axis in the neural tube, inductive processes during eye development and the interaction between neural tube and adjacent somitic mesoderm.

K. S. is grateful to Dr. R. Kemler for his support of this work in his laboratory. This work was supported by a grant from the DFG (German Research Foundation) awarded to K. S. We thank Drs R. Balling, B. Christ, T. Franz, H. Koseki, S. Mackem, J. Wallin, C. Ordahl and J. Wilting for their help and discussions in the interpretation of various aspects of the opb phenotype. Pax-3 and Pax-6 cDNA clones for in situ hybridizations were made available by Dr. P. Gruss, Msx-2 by Drs D. Davison and R. Hill, HNF-3β by Drs H. Sasaki and B. Hogan Shh by Drs B. St-Jaques and A. McMahon, Wnt-3a by Drs P. Salinas and R. Nusse, a Pax-1 probes by A. Neubüser and Dr. R. Balling. We thank Drs S. Morton and T. Jessell for providing an -bodies against Islet-1. We also thank Drs R. Balling, H. Koseki, S. Wood, E. Füchtbauer, K. Wertz and D. Solter for their discussion of the manuscript, T. Dammann for technical support and L. Lay for help with photographies.