Disabled-1 acts downstream of Reelin in a signaling pathway that controls laminar organization in the mammalian brain

The characterization of spontaneous and targeted mutations in mice has recently identified several genes that are required for correct cell positioning in the developing brain (D’Arcangelo and Curran, 1998). One of the most significant of these is reelin (Reln), the gene disrupted in reeler mice (D’Arcangelo et al., 1995). The ataxic phenotype of reeler mice was first described in 1951 (Falconer, 1951). Subsequent histopathological studies revealed that the reeler cerebellum is dramatically decreased in size and the normal laminar organization found in several brain regions is disrupted (Hamburgh, 1960). Reln is a large extracellular protein that is secreted from distinct neuronal populations (D’Arcangelo et al., 1995; Ogawa et al., 1995; Miyata et al., 1996; D’Arcangelo et al., 1997; Nakajima et al., 1997). Although cells that express Reln are positioned normally in reeler mice, the topography of neighboring regions is severely disrupted. This observation led to the hypothesis that Reln provides a local signal that controls cell-cell interactions critical for directing cell positioning in the developing brain.

Two novel mouse mutations, scrambler and yotari, were described recently that exhibit remarkable behavioral and histopathological similarities to reeler (Sweet et al., 1996; Goldowitz et al., 1997; González et al., 1997; Yoneshima et al., 1997). Molecular genetic studies revealed that scrambler and yotari arose from independent mutations in the disabled-1 (Dab1) gene (Sheldon et al., 1997; Ware et al., 1997). These mutations cause aberrant splicing of Dab1 mRNA that result in the synthesis of little or no Dab1 protein (Dab1). Furthermore, targeted disruption of Dab1 has also been shown to cause a reeler-like phenotype (Howell et al., 1997b). These findings demonstrate that Dab1, like Reln, is necessary for the formation of laminar structures in the developing brain. Dab1 contains a phosphotyrosine-interacting/phosphotyrosine-binding domain (PI/PTB) and it exhibits some degree of similarity to the Drosophila disabled gene. It is an intracellular protein that becomes phosphorylated on tyrosine residues during embryonic development and it can bind to the protein tyrosine kinases Src, Abl and Fyn (Howell et al., 1997a). Based on its biochemical properties, Dab1 is thought to function as an adapter molecule in the transduction of protein kinase signals.

The close phenotypic similarities among reeler, scrambler, yotari and disabled-1 null mice, suggest that Reln and Dab1 function in the same signaling pathway that influences neuronal positioning in the developing brain. Although initial studies reported that Dab1 is expressed during development (Howell et al., 1997a,b; Sheldon et al., 1997), no biochemical link between Reln and Dab1 was established. The simplest relationship between these two molecules would be that either Dab1 is required for the expression of Reln or, conversely, that Reln is required for Dab1 expression. However, Reln is expressed normally in Dab1-deficient mice and Dab1 is expressed in reeler (Goldowitz et al., 1997; González et al., 1997; Howell et al., 1997b; Sheldon et al., 1997; Yoneshima et al., 1997). Thus, the question remains open as to whether Dab1 functions independently from Reln in a signaling pathway required for normal neuronal positioning or whether its product functions as a downstream effector of Reln. Here we demonstrate that Dab1, like Reln, plays a role in the initial stages of lamination in the cerebral cortex, because the preplate fails to split in scrambler mice. A comparative analysis of the temporal and spatial patterns of Reln and Dab1 expression reveals that both genes are expressed very early during development of the cerebral cortex, cerebellum and hippocampus in adjacent cell populations, prior to the time when histological abnormalities appear in either reeler or scrambler. Furthermore, although the levels of Dab1 mRNA are comparable in reeler and normal mice, ectopic neurons in reeler contain approximately 10-fold more Dab1 protein than their normal counterparts. This finding suggests that Reln affects the turnover of Dab1 and it provides the first evidence for a biochemical link between these proteins. Taken together, our data strongly suggest that Dab1 functions downstream of Reln as a component of a signal transduction pathway that directs cell positioning in the developing brain.

Scrambler (Dab1^scm/Dab1^scm) and reeler (Reln^rl/Reln^rl) mice were obtained from breeding colonies at the University of Tennessee and at St. Jude Children’s Research Hospital, respectively. Scrambler arose on the dancer (DC/Le) strain of mice, and mutants were mated once to the C3HeB/FeJLe strain at The Jackson Laboratory (Bar Harbor, ME). The scrambler colony was maintained by sibling matings. The founder colony of B6C3Fe/J-reeler mice at St. Jude was purchased from The Jackson Laboratory. Mutant mice were obtained by mating heterozygous males to heterozygous or homozygous females. Both mouse colonies were maintained in a pathogen free environment with a light/dark cycle of 12 hours. In the text, control mice refers to wild-type and heterozygous reeler or scrambler mice, which are phenotypically normal. Females were examined each morning and if a plug was observed, we designated the stage of development as embryonic day 0.5 (E0.5). Embryos used for in situ hybridization and immunohistochemistry were harvested from females that were first anesthetized with Avertin (0.2 ml/10 g body weight) and killed by cervical dislocation. Embryos younger than E16.5 were immersion-fixed in 4% paraformaldeyhde in 0.1 M sodium phosphate buffer (PB; pH 7.2) overnight at 4°C. Older embryos, postnatal, and adult mice were perfused transcardially with the same fixative. Tissue was incubated overnight in a series of sucrose solutions (20-30%) and embedded in tissue freezing medium (Triangle Biomedical Sciences, Durham, NC) for cryostat sectioning. Sections, 10-16 μm thick, were mounted on Fisher Superfrost/Plus slides and stored at −20°C until use for in situ hybridization or immunohistochemistry.

The protocol for in situ hybridization was essentially as described previously by Simmons et al. (1989). Hybridization analysis was performed in situ using [³³P]UTP riboprobes that were generated by in vitro transcription of amplified DNA products containing the T7 polymerase promoter sequence flanking the desired nucleotide primer sequence. Antisense and sense riboprobes were generated that correspond to either the 3′ untranslated region (UTR) of Dab1 (nucleotides 1935-2116) or the open reading frame (nucleotides 618-821). For the localization of Reln, we used a riboprobe corresponding to nucleotides 5818-5973. Slides were incubated for 10 minutes at room temperature in Proteinase K (10 μg/ml) in a buffer containing 100 mM Tris and 50 mM EDTA (pH 8.0). Slides were then acetylated with acetic anhydride, rinsed in 2× SSC, dehydrated, and exposed to either sense or antisense denatured probe in the following hybridization buffer: 50% formamide, 10% dextran sulfate, 5× Denhardt’s solution, 620 mM NaCl, 10 mM EDTA (pH 8), 20 mM PIPES-Na (pH 6.8), 0.2% SDS, 50 mM DTT, 250 μg/ml denatured salmon sperm DNA and yeast tRNA. Hybridization occurred at 60°C overnight in a humid chamber containing 4× SSC and 50% formamide. Hybridized slides were exposed to 0.004 mg/ml RNase A for 30 minutes at 37°C in the appropriate buffer. Slides were washed in 2× SSC for 1 hour at 62°C followed by 0.2× SSC at 65°C for 2 hours, dehydrated in a graded series of ethanol in 0.1 M ammonium acetate, and exposed to Biomax MR film overnight (Kodak). The following morning, slides were dipped in autoradiography emulsion (Type NTB2; Kodak) and placed at 4°C in a light proof box for several days. Following development, slides were counterstained with 0.1% toluidine blue.

Early stages of cerebral cortical development were analyzed in scrambler mice by labeling cells in S-phase using the thymidine analogue, 5-bromo-2-deoxyuridine (BrdU). Timed-pregnant scrambler females were injected intraperitoneally with BrdU (5 mg/ml in saline solution; Sigma), at 50 μg/gm body weight. To label Cajal-Retzius cells and subplate neurons in the preplate, three females were injected twice, once at E10.5 and again 3 hours later, and two other females received a single injection of BrdU at E11.5 (Crandall et al., 1986; Wood et al., 1992; del Río et al., 1995; Ogawa et al., 1995; Sheppard and Pearlman, 1997). Because Cajal-Retzius and subplate cells are transient populations that disappear after birth, we analyzed the distribution of BrdU-labeled cells at E16.5. In addition, we examined the distribution of chondroitin sulfate proteoglycans (CSPGs), which are associated with the preplate and its derivatives in the marginal zone and subplate (Sheppard and Pearlman, 1997), in control and scrambler mice. Females were killed as described above and embryos were immersion fixed in a 3:1 solution of 95% ethanol and acetic acid. Brains were dissected, dehydrated in ethanol, cleared in xylene, and embedded in paraffin for sectioning. Sections were prepared at a thickness of 6 μm and incubated with either a mouse monoclonal anti-BrdU antibody diluted 1:100 (Becton-Dickinson, San Jose, CA) or anti-chondroitin sulfate at 1:600 (mouse monoclonal IgM, clone CS-56 from Sigma). The primary antibody was detected using the ABC peroxidase kit (Vector Laboratories, Burlingame, CA) followed by diaminobenzidine (DAB) as described previously (Hamre and Goldowitz, 1996). Sections were counterstained with Cresyl violet, dehydrated, and coverslips applied with Permount (Fisher Scientific).

Cryostat sections used in immunohistochemistry were blocked for 1 hour with 5% normal goat serum (NGS; Vector) diluted in 0.1 M sodium phosphate buffer (pH 7.2) containing 0.01% Triton X-100 (PB-X), then incubated in primary antibody diluted in 2.5% NGS in PB-X for 1 hour or overnight at 4°C. Primary antibodies and their dilutions used in this study were: anti-Reln CR-50 at 1:200 (Ogawa et al., 1995) and anti-MAP2 (2a + 2b) at 1:200 (mouse monoclonal IgG, clone AP-20 from Sigma). For the localization of Dab1 protein, we used a rabbit polyclonal antibody obtained from B. Howell and J. Cooper. This antibody (B3) was raised against residues 107-243 which overlaps the PI/PTB domain of Dab1 (Howell et al., 1997a). A dilution of 1:200 was adequate to distinguish control and scrambler brains (Fig. 4). For double staining, primary antibodies were incubated together as were secondary antibodies. Sections were washed in PB and then incubated with either fluorescein conjugated anti-rabbit or Texas red conjugated anti-mouse antibodies (Vector) diluted at 1:200 in 5% NGS in PB-X for 1 hour at room temperature. Slides were washed for several hours in PB and coverslips applied with Vectashield mounting medium (Vector). Slides were observed on an Olympus BX60 microscope. Images were acquired with a Hamamatsu C5810 video camera and directly imported into Adobe Photoshop (v.3.0). Contrast and brightness enhancements were applied equally to each figure. Figures were printed using a Fujix Pictrography 3000.

Brain tissue was removed from E16.5 reeler homozygous, heterozygous and wild-type embryos and the neocortex was dissected. Protein extracts from the neocortex and the remainder of the brain were prepared by Dounce-homogenizing snap-frozen tissue in 500 μl of cold lysis buffer containing 0.1% NP-40, 250 mM NaCl, 50 mM Tris-HCl (pH 7.4), 1 mM EDTA, 2 mM PMSF, 20 μM leupeptin, 50 mM NaF per 100 mg tissue. Extracts were clarified by microcentrifugation at 14,000 rpm for 30 minutes. A total of 100 μg of protein extract was loaded per lane onto a 4-12% polyacrylamide gradient gel (Novagen, Inc.), electrotransferred onto nitrocellulose membranes, and incubated with two different rabbit polyclonal antibodies specific for Dab1. The first antibody is the same as that used for immunohistochemistry studies (B3) and the second antibody was raised against a C-terminal peptide of Dab1 (Howell et al., 1997a). As a control, we used an anti-cdk5 antibody (Santa Cruz). Immunoblots were visualized by enhanced chemiluminescence (Boehringer Mannheim). Quantification of Dab1 protein levels was performed by densitometry. For analysis of Dab1 mRNA, total RNA was isolated from E16.5 reeler homozygous, heterozygous and wild-type brains, loaded at 10 μg per lane onto a 1.0% agarose-formaldehyde gel, and hybridized with a full length Dab1 DNA probe as described previously (Sheldon et al., 1997).

The laminar organization of the mouse cerebral cortex begins with the appearance of Cajal-Retzius and subplate neurons in the preplate between embryonic days 10 to 12 (E10-E12). The preplate is located near the pial surface, superficial to the ventricular zone (Marín-Padilla, 1978; Wood et al., 1992; Sheppard and Pearlman, 1997). The first wave of cortical plate neurons exits the cell cycle near the ventricular surface and migrates radially before invading the preplate, thus splitting this structure into the marginal zone, containing the Cajal-Retzius cells, and the subplate. Thereafter, newly generated neurons migrate past older neurons and insert into the developing cortical plate directly beneath the marginal zone. Corticogenesis proceeds in this inside-out manner ultimately generating a six-layered structure in mammals (Allendoerfer and Shatz, 1994; Marín-Padilla, 1998). This classic pattern of lamination is disrupted in adult reeler mice, which exhibit an inversion of cortical layers (Rakic and Caviness, 1995). The ectopic position of reeler cortical neurons is a consequence of the failure of migrating cells to bypass older cells, as the first cohort of neuronal precursors destined for the cortical plate fails to split the preplate (Caviness, 1982; Hoffarth et al., 1995; Ogawa et al., 1995; Sheppard and Pearlman, 1997). Superficially, the apparent inversion of cortical layers observed in adult scrambler mice (Sweet et al., 1996; González et al., 1997) resembles that of reeler, suggesting that a similar developmental process has been disrupted.

Disruptions in the cytoarchitecture of the cerebral cortex were apparent in scrambler mice by E16.5 and the normal radial organization of cell bodies was disorganized in a manner similar to that observed in reeler. The tight packing of neuronal cell bodies seen in normal mice was less obvious in scrambler and cell-free zones were present within the cortical plate. In reeler, these areas are referred to as the intermediate plexiform zone (IPZ), which has been shown to contain fibers innervating ectopic subplate and cortical plate neurons (Caviness, 1976; Pinto-Lord and Caviness, 1979; Yuasa et al., 1994; Sheppard and Pearlman, 1997). The borders of the cortical plate were less distinct in scrambler, suggesting that cell topography was already abnormal at E16.5. To address this possibility, we compared the distribution of microtubule associated protein (MAP2), a neuron-specific marker (Crandall et al., 1986), in control, scrambler and reeler mice. In control mice, the subplate was apparent as a dense array of MAP2-positive cells located beneath the MAP2-positive cortical plate cells (Fig. 1A). Although neuronal cell bodies in the reeler cortical plate expressed MAP2, the subplate layer was absent and MAP2-positive neurons, which resemble subplate cells morphologically, were located near the pial surface (Fig. 1B).

In the scrambler cerebral cortex, the distribution of MAP2-positive neurons resembling subplate cells (Fig. 1C) was remarkably similar to that in reeler, suggesting that subplate neurons were displaced like those in reeler.

The superficial position of subplate neurons in reeler arises from the failure of migrating cortical plate cells to invade the preplate. Thus, the preplate is not split into two layers, one containing Cajal-Retzius neurons and the other containing subplate neurons, during the initial stages of laminar organization. Instead, in the reeler cortex the majority of subplate cells is located above the cortical plate with Cajal-Retzius cells in a structure known as the superplate (Caviness, 1982; Ogawa et al., 1995). Although subplate cells are ectopically positioned in the reeler neocortex, chondroitin sulfate proteoglycans (CSPGs) continue to be associated with these cells as well as those in the marginal zone (Sheppard and Pearlman, 1997).

To determine whether the preplate fails to split in scrambler mice, we analyzed the distribution of CSPGs at E16.5 using a mouse monoclonal antibody. In addition, we labeled Cajal-Retzius cells and subplate neurons by injecting BrdU at E10.5 and E11.5 and we examined their distribution after formation of the cortical plate at E16.5. In control mice, the formation of the cortical plate splits the CSPG-positive preplate into two distinct layers, one associated with the marginal zone and the other with the subplate layer (Fig. 1D). In contrast, CSPG staining of the scrambler cerebral cortex was more broadly distributed near the pial surface and staining was also apparent in the IPZ (arrowheads in Fig. 1E), indicating that the preplate had failed to split properly in the mutant. In control mice, the majority of BrdU-positive cells labeled on E10.5 was located in both the marginal zone above the cortical plate and in the subplate layer between the cortical plate and the intermediate zone (Fig. 1F,H). In contrast, most of the BrdU-positive cells in the scrambler cerebral cortex were located in the superficial aspect of the cortical plate (Fig. 1G,I). The subplate layer, which was clearly discernible in control mice, was absent in the cerebral cortex of scrambler mice. When BrdU was injected approximately 24 hours later, at E11.5, the majority of labeled cells was located in the subplate, in the marginal zone, and in deep regions of the cortical plate in control mice (Fig. 1J). In contrast, the majority of labeled cells in scrambler was located near the pial surface (Fig. 1K), indicating that many cortical plate neurons had failed to intercalate between the subplate and the marginal zone. The ectopic location of these early-generated neurons, together with the immunohistochemical and morphological appearance of the embryonic cerebral cortex, demonstrate that the preplate fails to split properly in scrambler.

To compare the expression patterns of Reln and Dab1 in the cerebral cortex, in situ hybridization analysis was performed before formation of the cortical plate and throughout development of the cerebral cortex. At the preplate stage of development (E11.5), Reln was expressed in Cajal-Retzius cells located near the pial surface (Fig. 2A). In contrast, Dab1 was distributed homogeneously throughout the proliferative zone (Fig. 2B) underneath the cells expressing Reln. Approximately 24 hours later (E12.5), cortical plate formation begins in the lateral aspect of the cerebral cortex. In coronal sections of the cerebral cortex hybridized with a Dab1 probe, it was apparent that Dab1 expression reflected the lateral to medial gradient of cortical plate formation. Cells in the cortical plate, which is located at the most lateral aspect of the cerebral cortex at E12.5, appeared to express high levels of Dab1 compared to cells in the ventricular zone (Fig. 2C). One day later, cells expressing high levels of Dab1 were observed in more medial aspects of the cerebral cortex, where the cortical plate has formed (Fig. 2D). At E14.5, the cortical plate is several cell diameters thick and the intermediate zone is well developed. At this time, Reln expression was highest in the marginal zone (Fig. 2E), whereas Dab1 was expressed directly beneath this layer in cortical plate cells and in upper regions of the intermediate zone (Fig. 2F). Reduced levels of Dab1 were also present in the ventricular zone. Thus, Dab1 and Reln are expressed in adjacent cell populations of normal mice before the first morphological abnormalities become apparent in scrambler and reeler mice.

As corticogenesis proceeds, cells continue to migrate into the cortical plate to form the various layers of the cerebral cortex (Fig. 3A). At E16.5, Reln mRNA (Fig. 3B) and protein (Fig. 4F) persisted in cells occupying the marginal zone and, underlying this structure, cells in the cortical plate continued to express Dab1 mRNA (Fig. 3C) and protein (Fig. 4A). High levels of Dab1 protein were also detected in fibers in the intermediate zone, whereas lower levels of Dab1 were observed in the ventricular zone (Fig. 4A). Double staining with anti-Dab1 and anti-MAP2 antibodies demonstrated that Dab1 was present in neurons and their apical processes within the cortical plate and the marginal zone (not shown). In reeler, Dab1 expression was detected in cortical plate neurons, in cells within the ventricular zone, and in fibers of the intermediate zone (Fig. 4B). However, between the Dab1-positive cortical plate neurons and the pial membrane, there was a region of reduced Dab1 immunostaining (Fig. 4E). This area corresponds to the superplate, which contains Cajal-Retzius neurons and ectopically located subplate neurons.

In the cerebral cortex of postnatal day 5 (P5) mice, Reln and Dab1 were expressed in adjacent layers near the superficial aspect of the cerebral cortex (Fig. 3F,G). In addition to the persistence of Reln in the marginal zone, other cortical neurons in deeper layers expressed Reln (Fig. 3F). The expression of Dab1 appeared more intense near the surface of the cortex compared to that in deeper layers (Fig. 3G). In adult mice, both genes were expressed in the cerebral cortex, although the levels were lower compared to those in the developing brain (Fig. 3J,K).

Adult reeler and scrambler mice exhibit a dramatic reduction in the size of the cerebellum (Mariani et al., 1977; Goffinet, 1983; Sweet et al., 1996; Goldowitz et al., 1997). Granule cells, which arise from a proliferative population known as the external germinal layer (EGL), are reduced in number and Purkinje cells, which arise from a proliferative population near the fourth ventricle, are positioned ectopically in both reeler and scrambler. Although Purkinje cells are generated at the correct time, between E11 and E14, they fail to migrate radially in the direction of the EGL (Sidman and Rakic, 1973; Goffinet, 1983, 1984; Yuasa et al., 1991, 1993; Goldowitz et al., 1997). Thus, in embryonic reeler and scrambler mice, Purkinje cells fail to align into a layer beneath the EGL within the cortex of the cerebellum. The EGL is discernible on the dorsal surface of the cerebellum at E13.5 (Fig. 5A). At this time, Reln expression was observed both in the EGL (Fig. 5B) and in the nuclear transitory zone (NTZ). The NTZ represents a transient pathway through which neurons destined for the cerebellar nuclei migrate (Miyata et al., 1996). At E13.5, cells expressing Dab1 were located beneath the Reln-positive EGL (Fig. 5C) in an area known as the differentiating zone (Altman and Bayer, 1985). At E18.5, the Purkinje cell plate is well-formed and the EGL has expanded to cover the surface of the cerebellum (Fig. 5D). At this time, Reln was expressed in granule neurons in the EGL and in a few cells located deep in the cortex (Fig. 5E). Cells in the Purkinje cell plate continued to express Dab1, which appeared as a broad band beneath the EGL (Fig. 5F). An additional population of cells expressed Dab1 deep in the cerebellum (Fig. 5F, arrow).

The formation of the Purkinje cell layer in the cerebellum has been proposed to be regulated by Reln although it is not expressed in Purkinje cells themselves (D’Arcangelo et al., 1995; Miyata et al., 1996, 1997). If Dab1 functions in a Reln-dependent pathway in the cerebellum, it would be predicted to be expressed in Purkinje cells. Immunohistochemical analysis of the cerebellum at P3 demonstrated that Dab1-positive cells were located in the Purkinje cell layer, beneath the EGL (Fig. 5G). In the reeler cerebellum many of the Purkinje cells that failed to migrate radially formed heterotopic clusters that expressed high levels of Dab1 (Fig. 5H, asterisks). Although some Dab1-positive cells reached the cortex in the reeler cerebellum, they failed to align in a layer beneath the EGL (Fig 5H, arrowhead). In normal and reeler cerebellum, cells that contained Dab1 were also located in cerebellar nuclei (Fig. 5G,H, arrow).

The formation of the pyramidal layer of the hippocampus proper involves similar processes to those that occur during development of the cerebral cortex. For example, pyramidal neurons are generated in a proliferative layer near the ventricle and they migrate along radial fibers to form the pyramidal cell plate. Subsequently, pyramidal neurons align in an inside-out pattern reminiscent of that in the cerebral cortex. In contrast, there are two phases of cell migration during formation of the dentate gyrus that resemble processes involved in the generation of granule cells in the cerebellum. First, granule cell precursors originate from a restricted population of neuroepithelial cells near the ventricle. These cells migrate to establish a proliferative zone beneath the pyramidal layer. Second, postmitotic granule neurons that arise from these precursor cells migrate radially to form the dentate gyrus in an outside-in pattern (Angevine, 1965; Caviness, 1973; Schlessinger et al., 1978; Nowakowski and Rakic, 1979; Stanfield and Cowan, 1979a; Cowan et al., 1980).

The hippocampus in adult reeler and scrambler mice is characterized by malpositioning of both pyramidal neurons and granule cells (Caviness and Sidman, 1973; Stanfield and Cowan, 1979b; Sweet et al., 1996; Goldowitz et al., 1997; González et al., 1997). Recently, the organization of cells in the hippocampus proper has been shown to depend on the presence of Reln in the molecular layer (Nakajima et al., 1997). The first abnormality in the reeler hippocampus becomes obvious at approximately E16, when the orderly alignment of pyramidal cells, typical of the normal hippocampal plate is not discernible. Instead, pyramidal neurons remain scattered throughout the intermediate zone in reeler mice (Stanfield and Cowan, 1979a). In the scrambler hippocampus at E16.5, the hippocampal plate is also not apparent suggesting that, as in the reeler hippocampus, pyramidal neurons fail to align in a precise layer.

In the hippocampal primordium at E12.5 the patterns of Reln and Dab1 expression resembled those in the cerebral cortex at E11.5, in which Reln expression was localized to the marginal zone (not shown) and Dab1 was expressed throughout the ventricular neuroepithelium (Fig. 2C). By E14.5, three layers are apparent in the hippocampus. The densely packed ventricular zone contains the majority of cells in the hippocampal anlage, an intermediate zone, containing the first cells destined for the pyramidal cell layer, lies above the ventricular zone, and the marginal zone is located superficial to this layer (Fig. 6A). In addition, cells destined for the dentate gyrus are apparent as an extension of the ventricular epithelium near the medial extent of the hippocampus. At this time, most cells expressing Reln were located in the marginal zone (Fig. 6B), although a few cells in the intermediate zone were also positive. The Reln-expressing cells in the marginal zone correspond to Cajal-Retzius-like cells (Soriano et al., 1994). A thin layer of Dab1-positive cells first appeared around E14.5, directly beneath the Reln-positive marginal zone (Fig. 6C). This layer, which is continuous with Dab1-positive cortical plate cells in the cerebral cortex, likely represents the first cohort of pyramidal cells that form the hippocampal plate (Stanfield and Cowan, 1979a). Thus, Dab1 and Reln are expressed in the normal hippocampus in adjacent cell populations before morphological abnormalities are detected in either reeler or scrambler.

At E18.5 the hippocampal plate is well-formed and granule cells, which have a protracted period of genesis compared to pyramidal cells (Angevine, 1965), can only be seen in the suprapyramidal blade of the dentate gyrus (Fig. 6D). At this time, Cajal-Retzius-like cells in the marginal zone expressed Reln (Fig. 6E) adjacent to a region of cells that expressed Dab1 in the dentate gyrus (Fig. 6F). In the hippocampus proper, cells expressing Dab1 are located in the pyramidal cell layer and in the intermediate zone (Fig. 6F). Several days later, at P3, the full complement of pyramidal cells is present and the infrapyramidal blade of the dentate gyrus is becoming populated with granule cells. During this period, Reln expression persisted in Cajal-Retzius-like cells and, similar to the situation in the cerebral cortex (Fig. 6G), an additional population of cells outside of the marginal zone was also positive. At this time, there was robust expression of Dab1 in hippocampal pyramidal cells and in granule cells of the supra- and infrapyramidal blades of the dentate gyrus (Fig. 6H).

During the course of our immunohistochemical studies we noticed that reeler neurons consistently exhibited much stronger staining with anti-Dab1 antibodies compared to those in normal mice. For example, in Fig. 4A,B, the level of Dab1 staining in the cortical plate and intermediate zone is much higher in reeler than in the control cerebral cortex. In this figure, panel A was exposed about twice as long as panel B to obtain an equivalent image. In contrast, cells in the ventricular zone of both reeler and control cortex expressed similar levels of Dab1. Furthermore, ectopic Purkinje cells, but not cells in the cerebellar nuclei, of reeler mice exhibited more intense immunostaining with Dab1 antibodies (Fig. 5G,H). No difference was found in the level of MAP2 staining between normal and mutant mice. These results suggest that Dab1 expression is elevated in reeler mice. However, in situ hybridization experiments (Fig. 3C,D) and northern hybridization analysis indicated that reeler mice expressed the same level of Dab1 mRNA as that found in control mice (Fig. 7A). To resolve this conundrum, we compared the levels of Dab1 in reeler and normal mice by immunoblot analysis. As shown in Fig. 7B, reeler extracts prepared from E16.5 embryos contained approximately 5- to 10-fold more Dab1 protein than brain extracts from normal mice or heterozygous littermates. Similar results were obtained using antibodies that were raised against either the PI/PTB domain or the C-terminal region of Dab1. In some heterozygous brain extracts, more Dab1 was present than that found in wild-type extracts, but the difference in levels was variable. Thus Dab1 protein, but not mRNA, is elevated in ectopic neurons in reeler mice. This finding suggests that Reln exerts a post-transcriptional control on Dab1 expression, either at the level of translational efficiency or protein turnover.

The reeler mutation has attracted a significant degree of attention because of the widespread neuroanatomical defects apparent in the central nervous system (CNS) of the mutant mice. Although neurogenesis occurs normally in reeler, dramatic defects in cell positioning are apparent in several laminar structures of the CNS. Mouse chimera analysis suggested that the product of the reeler gene functions as an extrinsic signal on Purkinje cells as they migrate from their site of origin near the fourth ventricle to their final destination in the cerebellar cortex (Mullen, 1978; Terashima et al., 1986). This prediction was largely supported by the identification and characterization of Reln, the gene mutated in reeler mice (D’Arcangelo et al., 1995). The product of Reln was shown to be a large protein secreted by distinct cell types, such as Cajal-Retzius cells and cerebellar granule cells, during CNS development (Ogawa et al., 1995; Miyata et al., 1996; D’Arcangelo et al., 1997). However, the finding that Reln expression is largely confined to the marginal zone of the developing cerebral cortex, next to the cortical plate which is disrupted in reeler mice, implies that Reln acts through a molecular signaling pathway that ultimately impinges on the neurons that go astray in the mutant mice. The identification of reeler-like defects in scrambler, yotari and Dab1-null mice (Sweet et al., 1996; Goldowitz et al., 1997; González et al., 1997; Howell et al., 1997b; Sheldon et al., 1997; Yoneshima et al., 1997) provides a unique opportunity to investigate the respective roles of two distinct genes in the complex biological events that underlie formation of the mammalian cerebral cortex.

Reln has several hallmarks of proteins that function as components of the extracellular matrix. It contains a cleavable signal peptide at the N terminus and a series of eight EGF-like repeats similar to those of the tenascin and integrin families (D’Arcangelo et al., 1995). It is secreted by COS cells transfected with a Reln-expression clone and from primary cultures of cerebellar cells (D’Arcangelo et al., 1997; Goldowitz et al., 1997). Furthermore, immunohistochemical studies with the anti-Reln antibody, CR-50, suggest that Reln accumulates on Purkinje cell processes (Miyata et al., 1996), even though these cells do not express Reln mRNA. Taken together, these observations indicate that Reln acts as an extracellular signal that activates receptors present on the surface of cell populations that go astray in reeler and scrambler mice.

Dab1 is a cytoplasmic molecule that probably does not interact with Reln directly, however, it has several features that make it a likely candidate to be involved in a Reln-initiated signaling pathway. Dab1 is known to bind protein tyrosine kinases such as Src, Fyn and Abl via its PI/PTB. The observation that Dab1 becomes phosphorylated on tyrosine residues during neuro development demonstrates that it is also a substrate for protein tyrosine kinases (Howell et al., 1997a). These features suggest that Dab1 acts as an adapter protein that links components of a signal transduction pathway such as cell surface receptors and cytoplasmic protein kinases.

Although Dab1 mRNA is expressed normally during neuro development in reeler mice, Dab1 protein levels accumulate approximately 5- to 10-fold in embryonic brain tissue. This dramatic difference in Dab1 protein levels between reeler and wild-type mice is present as early as E12.5, and it is maximal at E16.5. In adult mice this difference in Dab1 levels is less pronounced (data not shown). The time of peak overexpression of Dab1 corresponds to the period in which neuronal migration is underway and when Reln is required for the normal positioning of neurons in CNS. This increase could arise if under normal circumstances Dab1 is degraded after fulfilling a signaling function evoked by Reln as part of a switch mechanism that controls cell positioning. Alternatively, it is possible that the absence of Reln causes an increase in the translation rate of Dab1 mRNA, leading to increased levels of Dab1. Regardless of the mechanism responsible for the increased levels of Dab1 protein in reeler, this finding establishes the first biochemical link between these proteins and suggests that Dab1 functions downstream of Reln.

Mammalian cortical plate formation occurs subsequent to the appearance of the preplate, a stratum of neurons and fibers that arises early in development. Postmitotic neuronal precursors migrate from the ventricular zone along the surface of radial glia (Rakic, 1972) towards the pial surface where they insert into the preplate, splitting this structure into the superficial marginal zone (layer I) and a deep layer of neurons known as the subplate (Marín-Padilla, 1978, 1998). Previous studies in reeler mice demonstrated that many cortical plate neurons accumulate beneath, instead of above, the subplate in a structure known as the superplate (Caviness, 1982; Ogawa et al., 1995; Sheppard and Pearlman, 1997). These observations suggested that Reln plays a direct role in the cell-cell interactions that govern cell positioning within the preplate.

In the scrambler cerebral cortex, most of the subplate neurons, which were identified immunohistochemically at E16.5 with anti-MAP2 antibodies and by labeling with BrdU, are located above the cortical plate (Fig. 1). Therefore, cortical plate neurons failed to bypass the subplate despite the presence of Reln in scrambler mice (Goldowitz et al., 1997; González et al., 1997; Sheldon et al., 1997). Furthermore, in the scrambler cerebral cortex the distribution of CSPGs, which are associated with the preplate and its derivatives, is remarkably similar to that in reeler (Sheppard and Pearlman, 1997). Both mutants display an intense staining of CSPGs near the pial surface and in the IPZ, which may indicate the presence of a few subplate neurons intermixed with cortical plate neurons. Thus, the preplate fails to split properly in the scrambler cerebral cortex. The fact that the same developmental process is disrupted in scrambler and reeler mice suggests that Reln and Dab1 participate in a common signaling pathway that controls cell positioning within the cortical plate.

In the reeler and scrambler cerebral cortex, the first wave of cortical plate neurons fails to insert into the preplate resulting in the accumulation of cortical plate cells beneath the superplate (Caviness, 1982; Ogawa et al., 1995; Sheppard and Pearlman, 1997; present study). Reln is produced by Cajal-Retzius cells located in the preplate before the arrival of cortical plate neurons. In contrast, at this time (E11.5) Dab1 is expressed in neuroepithelial cells in the ventricular zone at low and relatively uniform levels. Approximately 24 hours later, Dab1 expression levels are more pronounced in the lateral extent of the neocortex where the cortical plate is emerging directly beneath the Reln-positive cells. As more neurons arrive in the cortical plate, the overall level of Dab1 increases (Fig. 2). Although the importance of Dab1 expression in the ventricular zone is unclear, Dab1, like Reln, is required for the migration of cortical plate cells into the preplate.

In the cerebellum, adjacent populations of cells express Dab1 and Reln before anatomical abnormalities in scrambler and reeler are apparent. For example, Purkinje cells in the differentiating zone (dz) of the E13.5 cerebellum express Dab1 in a region underlying cells in the external germinal layer (EGL) that express Reln (Fig. 5). Therefore, it seems likely that the initial signaling events between the Reln-positive EGL and the Dab1-positive Purkinje cells occur when these cells are in close proximity. In the absence of Reln or Dab1, this signaling event does not occur and Purkinje cells remain clustered beneath the cortex of the cerebellum. Although granule cells are dramatically decreased in number in adult scrambler mice, their migration occurs normally (Goldowitz et al., 1997). The reduction in granule cell number may be a secondary event caused by a lack of trophic support that would be provided by correctly-positioned Purkinje cells. Thus, the cerebellar defects in adult scrambler and reeler mice may be a consequence of an initial failure of the interactions between granule cell precursors and Purkinje cells in which the Purkinje cells fail to migrate radially towards a Reln signal secreted by EGL cells. In the hippocampus of Reln and Dab1 deficient mice, pyramidal cells are not aligned in a monolayer and granule cells are widely distributed in the dentate gyrus (Stanfield and Cowan, 1979b; Sweet et al., 1996; Howell et al., 1997b; Yoneshima et al., 1997). Neither of these cells express Reln during development, which is secreted by the Cajal-Retzius-like cells in the marginal zone. However, Dab1 is expressed in the first cohort of pyramidal cells as they arrive in the intermediate zone underneath the marginal zone at E14.5 (Fig. 6), and later, at E18.5, granule cells express Dab1 when they migrate radially to form the dentate gyrus. Therefore, Reln may provide an instructive signal for both of these populations of migrating cells.

In the three brain regions analyzed here, the migration of neuronal precursors away from their site of origin appears to be unimpeded in the absence of Dab1. However, Dab1-deficient neurons fail to coalesce into discrete layers. This is most likely a consequence of the inability of the first cohort of cells expressing Dab1 to respond to a Reln-evoked signal, which is present at the terminus of the migratory route of these cells. In the scrambler cerebral cortex, the first wave of cortical plate neurons is most likely unable to respond to Reln in the preplate and, as a result, they fail to enter this region. This contrasts with the situation in Cdk5 and p35 null mice in which, although the laminar organization of the cortex is disrupted, the first cohort of cortical neurons is able to invade the preplate and form a cortical plate between the marginal zone and subplate (Gilmore et al., 1997; Kwon and Tsai, 1998). In these mice, lamination fails at the next stage in which later-generated cohorts of neurons fail to migrate past the first population which has taken up residence in the cortical plate. This may either be a consequence of a failure in migration per se or it is possible that Cdk5/p35 is required to alter the properties of the first cortical plate neurons, after they insert into the preplate, permitting them to be bypassed by later generated neurons. In either scenario, Cdk5/p35 would be envisioned to function downstream of Dab1 during corticogenesis. Thus, the mechanism of cortical plate formation involves at least two distinct processes, an early event dependent on Reln and Dab1, and a later event that also requires Cdk5/p35, in addition to Reln and Dab1.

The nature of the molecular interactions between Reln and cells that express Dab1 are not yet clear. One possibility is that Reln, which is produced and secreted into the extracellular environment, acts as a guidepost by binding to receptors on adjacent cells resulting in modulation of tyrosine kinases. For example, the majority of cortical neurons ascends to the preplate along radial fibers and presumably contacts Reln in the marginal zone during the formation of the mammalian cortical plate. Reln could conceivably bind to the leading process of these migrating cells instructing them to translocate to the preplate. Subsequently, cells would detach from the radial fiber and differentiate into mature neurons. Consistent with this idea, we found that Dab1 is expressed at high levels in cortical plate neurons at the time when they invade the preplate. In addition, Dab1 is present in neuronal processes in the marginal zone. These observations imply that Dab1 may mediate the functional interaction of Reln with cortical plate neurons. Alternatively, Reln could conceivably interact indirectly with cells containing Dab1 through effects on radial glia, which are known to express molecules required for neuronal migration and laminar organization in the cortex (Cameron and Rakic, 1994; Anton et al., 1996a,b). Finally, we cannot rule out the possibility that Reln binds to other molecules, for example diffusible factors, in the extracellular matrix (Sheppard and Pearlman, 1997) that affect migration. Hopefully, these alternatives will be resolved by the identification and characterization of proteins that interact with both Reln and Dab1 and mediate their functions.

The authors would like to thank Richard Cushing and Joli Williams for technical assistance and Drs Richard Smeyne and Eduardo Soriano for helpful comments on the manuscript. This work was supported in part by NIH Cancer Center Support CORE grant P30 CA21765, NIH grants T32 CA09346 (D. S. R.) and RO1 NS36558 (T. C.), the American Lebanese Syrian Associated Charities (ALSAC), the President’s Special Research Grant of RIKEN, the Ministry of Education, Science, and Culture of Japan (K. N.), and the University of Tennessee College of Medicine Bridge Fund and the Department of Anatomy and Neurobiology (D. G.).