C3G regulates cortical neuron migration, preplate splitting and radial glial cell attachment

The development of the cerebral cortex into a six-layered structure involves a coordinated series of events. Disruption of these events leads to mental retardation syndromes including lissencephaly(Ayala et al., 2007). Neural precursors proliferate in the ventricular and subventricular zones, exit the cell cycle and migrate to specific locations. At the commencement of corticogenesis, cells from the cortical hem and the cortical ventricular zone migrate to the pial surface to form the cortical preplate(Bielle et al., 2005; Takiguchi-Hayashi et al.,2004). The next cohort of cells arises from the cortical ventricular zone and forms layer VI neurons. These cells position themselves in the centre of the preplate, thus splitting the preplate into the marginal zone/layer I and the subplate (Bayer and Altman, 1991; Gupta et al.,2002). These early steps of cortical development are glia-independent and entail the translocation of the nuclei and cell bodies by contracting cell processes anchored at the pial basement membrane(Gupta et al., 2002; Nadarajah et al., 2001; Olson and Walsh, 2002). The succeeding cohorts of cells successively migrate through previously formed layers and position themselves superficially such that layers VI, V, IV and II/III develop in an inside-out fashion(Bayer and Altman, 1991; Gupta et al., 2002). These later steps involve migration of cortical neurons along radial glial cells,which extend cell processes from their ventricular position to the pial surface, where they are anchored in the basement membrane(Gupta et al., 2002; Olson and Walsh, 2002). Clearly, neuronal migration relies on the interaction of neurons with their extracellular environment, including neighbouring cells (radial glia) and the extracellular matrix (ECM) proteins in the basement membrane.

The ECM is a network of proteins that forms a scaffold providing mechanical support, but ECM proteins also are essential inducers of cell signalling. The ECM contains specialised components in defined regions specifically for cell signalling. For example, ECM within the marginal zones of the cerebral cortex and the cerebellum contains reelin, which regulates neuronal migration(D'Arcangelo et al., 1995). Other ECM proteins such as laminin, fibronectin and collagen, are produced by the neuroepithelium throughout neurogenesis(Liesi, 1985; Sheppard et al., 1991; Stewart and Pearlman, 1987; Thomas and Dziadek, 1993b). Here we provide functional evidence that C3G (Rapgef1, Grf2), a molecule that transduces signals from ECM proteins, is essential for cortical neuron migration.

C3G is a guanine nucleotide exchange factor for small GTPases of the Ras family (Knudsen et al., 1994; Tanaka et al., 1994). C3G can be activated by a number of extracellular signals including integrin binding(Arai et al., 2001; Arai et al., 1999; Uemura and Griffin, 1999) and reelin stimulation (Ballif et al.,2004). Phosphorylation of C3G and complex formation with the adaptor protein Crk are necessary for C3G activation(Ichiba et al., 1997). The Crk-C3G complex translocates to the cytoplasmic membrane(Ichiba et al., 1997), where C3G stimulates GTP exchange predominantly on Rap1(Ohba et al., 2001), thereby activating signalling through Rap1. C3G regulates migration in fibroblasts(Ohba et al., 2001; Voss et al., 2003) and is essential for filopodia formation and cytoskeletal changes in fibroblasts(Radha et al., 2007). In addition, C3G regulates neural precursor proliferation in the developing cerebral cortex (Voss et al.,2006).

We report here that during the development of the cerebral cortex, C3G is essential for normal neuronal migration.

Experiments were undertaken with the approval of the Royal Melbourne Hospital Research Foundation Animal Ethics Committee and conforming to the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. Noon of the day of vaginal plug detection was defined as embryonic day 0.5 (E0.5) in timed matings of C3G^gt/+ × C3G^gt/+ mice (Voss et al., 2003) on a CD1 or a CBA background. BrdU experiments were carried out as previously described(Thomas et al., 2000). Embryos were recovered, genotyped and processed for histology as reported previously(Voss et al., 2003).

Primary antibodies were directed against BrdU (Bio-Science Products 010198,1:10), β-tubulin type III (Promega G7121, 1:2000), calretinin (Chemicon AB5054, 1:100), Emx1 (Chan et al.,2001), laminin (Chemicon AB2034, 1:1000), Map2 (Sigma M4403,1:200), reelin (Abcam ab18570, 1:1000), RC2 (from Miyuki Yamamoto via the Developmental Studies Hybridoma Bank, 1:200), C3G (Santa Cruz sc-869 and sc-15359, 1:500) and phospho-tyrosine (BD Transduction Laboratories PY-20,1:1000). Secondary antibodies were Vector Laboratories BA-1400, Vector FI-2020, Molecular Probes A-21124, A-11004, A-11035, A-11029, A-11003,A-11001, Southern Biotechnology 710003 and Jackson ImmunoResearch 111-156-003.

Rap1 activation assays were conducted as described previously(Voss et al., 2006). Immunoblotting was performed as described(Voss et al., 2003; Voss et al., 2006). C3G immunoprecipitations were performed as described(Ballif et al., 2004). Briefly,cortical neurons were washed with ice-cold PBS, lysed in 400 μl nRIPA [0.15 M NaCl, 1% Triton X-100, 0.1% SDS, 1% sodium deoxycholate, 10 mM sodium phosphate (pH 7.4), 2 mM EDTA, 14 mM 2-mercaptoethanol, 50 mM NaF, 2 mM Na₃VO₄, 1 mM phenylarsine oxide and one complete protease inhibitor tablet per 40 ml]. Lysates were clarified by centrifugation and then precleared with a cocktail of protein A/protein G sepharose [protein A (G) sepharose 4 fast flow, GE Health Care]. Lysates were mixed with 20 μl protein G sepharose and 2 μg anti-C3G (sc-15359) antibody. Samples were incubated for 2.5 hours with agitation at 4°C. Resin was washed three times with 0.8 ml nRIPA buffer. Samples were eluted by boiling in 30 μl 2× sample buffer [4% SDS, 124 mM Tris-HCl (pH 6.8), 0.1% Bromophenol Blue, 20% glycerol, 200 mM DTT).

In situ hybridisation and BrdU detection were performed as described(Thomas et al., 2000). Immunofluorescence on sections was carried out as described(Voss et al., 2003). Immunofluorescence on cultured cells was performed as described(Merson et al., 2006; Voss et al., 2000).

Neural precursor cells were cultured in neural stem cell proliferation medium as previously described (Merson et al., 2006). Human 293T cells were transfected with 4.5 μg of either pCRL [reelin expression vector; gift from T. Curran(D'Arcangelo et al., 1997)] or pCDNA3.1 (empty parental vector; mock) using FuGENE (Roche). Medium was replaced 24 hours post-transfection with neural differentiation medium (as neural stem cell proliferation medium but without Fgf2, and without Egf). Conditioned medium was recovered 48 hours post-transfection. Cortical neurons were isolated and cultured as described(Herrick and Cooper, 2002). Briefly, E16.5 cerebral hemispheres were dissociated enzymatically and mechanically, passed through a sieve, collected by centrifugation and plated onto poly-l-lysine-coated tissue culture plates in neural differentiation medium with 1% foetal bovine serum.

Replication-incompetent enhanced GFP (eGFP)-expressing retrovirus was produced from a stably transfected packaging cell line, 293gp NIT-GFP (gift from F. Gage, Salk Institute, La Jolla, CA). Cells were transiently transfected with pVSV-G (Palmer et al.,1999), the supernatant harvested after 48 hours and concentrated by ultracentrifugation at 53,000 g for 3.5 hours at 4°C.

Coronal slices were prepared from embryonic forebrain as previously described (Noctor et al.,2001). Slices were obtained from the anterior half of the cerebral hemispheres, where medial and lateral ganglionic eminences could be separately identified, and cultured individually on slice culture inserts (Millicell,Millipore). Cultured slices were incubated for 48 hours with eGFP-expressing retrovirus before images were collected on an inverted Zeiss Axiovert 200-LSM 5 Pascal confocal microscope using 488 nm excitation and LP 505-530 nm emission filters. Images were captured using a 10× objective set at the sectional plane that contained the majority of migrating cells (optical slice<10 μm). Time-lapse imaging was performed at 37°C, 100% relative humidity, 5% CO₂ in air on a CO₂- and heat-controlled stage incubator (Zeiss) using minimum laser exposure to prevent photo-damage and bleaching. Images were taken every 15 minutes for 6 to 12 hours.

Telencephalic hemispheres were isolated and cultured on Matrigel or laminin as described previously (Thomas and Dziadek, 1993a). For time-lapse imaging, telencephalic hemisphere explants on Matrigel were cultured on an inverted imaging microscope (Zeiss)equipped with a heat-controlled stage incubator at 37°C, 100% relative humidity and 5% CO₂ in air. Images were taken every 5 minutes for 22 hours in phase contrast using a 20× objective. The imaging interval of 22 hours was taken from day 3 to day 4 after plating.

Data were analysed using StatView 5.0.1 Software (SAS Institute),performing analyses of variance followed by Fisher's post-hoc tests orχ ² test as indicated in the tables. The default alpha value of the software (5%) was used. Data are presented as mean ± s.e.m.

During development, C3G is expressed at low levels in all cell types (Voss et al., 2003) and at higher levels in the developing nervous system(Voss et al., 2006). In the cortical neuroepithelium, C3G expression was observed at low levels in proliferating ventricular zone cells at E12.5 and E15.5(Fig. 1A-F). Differentiating neurons near the pial surface at E12.5(Fig. 1A-C) and in the cortical plate at E15.5 (Fig. 1D-F)expressed slightly higher levels of C3G mRNA.

Mice homozygous for a C3G-null allele do not survive beyond E5.5(Ohba et al., 2001). We studied the role of C3G later in development, using a hypomorphic mutant C3G allele (C3G^gt) that expresses less than 1%normal C3G mRNA and less than 5% normal C3G protein(Voss et al., 2003). C3G is required for blood vessel maturation and, on a 129Sv-enriched background, the level of C3G protein produced by the C3G^gt mutant allele is insufficient for the differentiation of blood-vessel-supporting cells,resulting in haemorrhage in the majority of homozygous mice (80%) by E11.5(Voss et al., 2003). However,we found that crossing the C3G^gt allele onto an outbred CD1 background or to a CBA inbred background ameliorated the blood vessel fragility phenotype. On these vigorous genetic backgrounds, all homozygous embryos live beyond E12.5, the majority are externally indistinguishable from wild-type littermate controls at E12.5 and are alive at E14.5. This provided the opportunity to study the requirements for C3G during the early development of the cerebral cortex.

The C3G^gt/gt mutant cortical neuroepithelium was histologically indistinguishable from wild-type littermate controls at E9.5,E10.5 and E11.5 (not shown). At E12.5, the C3G^gt/gtcortical neuroepithelium lacked a continuous basement membrane, which was visible by differential interference contrast microscopy in wild-type controls(arrows, Fig. 1G, versus arrowheads, Fig. 1H). At E13.5, C3G^gt/gt neuroepithelial cells (arrowheads, Fig. 1J) could be seen in the area of the pericerebral vascular complex, a tissue structure outside of the neuroepithelium, which is demarcated in wild-type controls (stippled line, Fig. 1I). At E14.5, the developing wild-type cortical primordium was clearly organised in a ventricular zone with radially oriented neuroepithelial cells, an intermediate zone and a nascent cortical plate, the primordium of the cerebral cortex(Fig. 1K). The subplate was positioned between the intermediate zone and the cortical plate. The marginal zone overlay the cortical plate on its pial surface(Fig. 1K). By comparison, the C3G^gt/gt cerebral cortex primordium appeared disorganised(Fig. 1L). The subplate was indistinct and, consequently, the boundary between the intermediate zone and the cortical plate was not distinguishable. A cortical plate or a marginal zone could not be discerned (Fig. 1, compare K with L). C3G^gt/gt cortical cells lacked radial orientation and appeared rounded(Fig. 1, compare M with N). The defects in cerebral cortex development were consistently observed in a total of 21 C3G^gt/gt mutant embryos but not in 21 wild-type controls at E12.5 (four pairs), E13.5 (ten pairs) and E14.5 (seven pairs). Our findings suggest that C3G deficiency is incompatible with normal development of the cerebral cortex.

Microtubule-associated protein 2 (Map2; Mtap2-Mouse Genome Informatics) is a marker of post-mitotic neurons and identifies cortical neurons in the developing cerebral cortex. In the wild type, Map2 staining was stronger in neurons within the cortical plate than in migrating neurons in the intermediate zone, and marked the cortical plate clearly(Fig. 2A). Although strongly Map2-positive cortical neurons were present in the C3G^gt/gt cortex (Fig. 2B), they were disorganised and not structured into a cortical plate (Fig. 2, compare A with B; n=7 each for wild-type and C3G^gt/gt brains at E13.5 and E14.5).

We examined the basement membrane integrity by laminin immunofluorescence staining of C3G^gt/gt mutant and wild-type developing cortex (n=3 for each genotype at each developmental stage totaling nine C3G^gt/gt mutant and nine wild-type brains). As expected, laminin immunoreactivity was strongest in the area of the basement membrane and was also seen associated with blood vessels. In the wild type,laminin was continuous, encircling the entire neuroepithelium at its pial surface marking the basement membrane (Fig. 2C). In the C3G^gt/gt mutants, laminin deposition appeared initially irregular(Fig. 2D) and then areas with and without laminin alternated along the pial surface(Fig. 2E). These two stages of disintegration of the basement membrane were seen in C3G^gt/gt mutants at E12.5 and E13.5. At E14.5, the latter,more severe stage of basement membrane disintegration prevailed. Furthermore,cells were found to protrude from the neuroepithelium into the pericerebral space (arrow, Fig. 2G). Map2 staining (arrow, Fig. 2B) and Map2/laminin double staining (arrows, Fig. 2G,I) showed that the cells protruding into the pericerebral space were of neuronal identity. A possible cause of the heterotopic position of neuronal cells in the pericerebral space is the basement membrane disintegration.

At E10.5, the first neurons are born, migrate to the pial surface and form the cortical preplate, which can be seen between E12.5 and E13.5. These preplate cells can be observed in their final destination after incorporation of the thymidine analogue BrdU during the DNA-synthesis phase of their last cell division at E10.5. The next cohort of neurons leave the cell cycle at E11.5, migrate from the ventricular zone towards the pial surface and position themselves in the centre of the preplate, thereby splitting the preplate into marginal zone and subplate, which can be seen at E14.5. If the preplate cells have been labelled with BrdU previously, the products of preplate splitting,the marginal zone and the subplate, are visible as two distinct layers containing BrdU-labelled cells. Cells that leave the cell cycle after BrdU label incorporation are strongly BrdU positive. By contrast, cells remaining in the cell cycle distribute the BrdU to their progeny, thereby diluting it,resulting in weakly BrdU-positive cells.

Cells strongly labelled with BrdU at E10.5 were present in two discrete layers in the developing wild-type cerebral cortex at E14.5, the marginal zone at the pial surface (arrows, Fig. 3A) and the subplate (arrowheads, Fig. 3A). Between the marginal zone and the subplate, the nascent cortical plate consists of the newly arrived neurons of the future layer VI(Fig. 3A). In the C3G^gt/gt mutants, by contrast, the strongly BrdU-positive cells were almost exclusively found close to the pial surface (arrows, Fig. 3B, compare with Fig. 3A), demonstrating either that preplate splitting did not occur in the C3G^gt/gtmutant embryos or that the subplate cells did not survive (n=4 each for wild-type and C3G^gt/gt brains). We counted strongly BrdU-positive cells in both hemispheres in the wild-type and C3G^gt/gt mutant parietal cortex primordium in the marginal zone area near the pial surface versus the submarginal areas. This classification was necessary because the C3G^gt/gt mutant cortex primordia did not form a recognisable marginal zone, cortical plate or subplate. The distribution of BrdU-positive cells was significantly altered in the C3G^gt/gt mutants(Table 1; P<0.0001). Whereas approximately half the BrdU-positive cells populated the marginal zone and the other half the subplate in the wild type,>90% of the cells resided near the pial surface in the C3G^gt/gt mutants. The remaining <10% BrdU-positive cells were scattered in various positions below the area of the marginal zone,but not arranged into a subplate. The average total number of BrdU-positive cells labelled at E10.5 and analysed at E14.5 did not differ significantly between wild-type and mutant hemispheres (39.7±4.0 versus 32.5±3.1, respectively), supporting the conclusion that mutant subplate cells did not die and that the observed difference is due to a preplate splitting defect. By contrast, later in development at E11.5 and E12.5, C3G^gt/gt mutant neuroepithelial cells showed a higher rate of BrdU incorporation than wild-type cells, as we have shown previously(Voss et al.,2006). Similarly,marginal zone and subplate cells were distributed abnormally to the pial surface in E14.5 C3G^gt/gt mutants as seen by calretinin(calbindin 2 - Mouse Genome Informatics) staining, which marks marginal zone and subplate cells in the control brains(Fig. 3, compare C with D; n=3 each for C3G^gt/gt mutant and control brains). BrdU treatment at E12.5 predominantly, but not exclusively, labelled future layer VI cells, which can be seen between the marginal zone and the subplate forming the nascent cortical plate in the wild type(Fig. 3E). By contrast, C3G^gt/gt mutant cells labelled at E12.5 were scattered in positions below the unsplit preplate (Fig. 3F), confirming the finding that C3G^gt/gtmutant cells do not split the preplate and do not form a cortical plate.

Cells of the pseudostratified neuroepithelium and radial glial cells rely for attachment to the pial surface on interaction with the basement membrane. The attachment mechanism requires laminin, β1 integrin, focal adhesion kinase and integrin-linked kinase in vivo(Beggs et al., 2003; Graus-Porta et al., 2001; Halfter et al., 2002). Furthermore, C3G has been shown to be crucial for integrin-mediated cell adhesion in fibroblasts (Voss et al.,2003). Therefore, we examined the distribution of the basement membrane protein laminin in conjunction with radial glial cell processes by staining for the marker protein RC2 (Ifaprc2) (n=3 for each genotype at each developmental stage totaling nine C3G^gt/gt mutant and nine wild-type brains).

Laminin staining was continuous in the wild-type, but discontinuous in the C3G^gt/gt brains, indicating disruption of the basement membrane (Fig. 2; Fig. 4, compare A,C,E,G with B,D,F,H). The basement membrane defect became successively more severe from E12.5 to E14.5 (Fig. 4, compare B with D and F with H). RC2 immunofluorescence, marking radial glial cell processes, was observed along cell processes with ventricular-to-pial orientation, essentially parallel to each other, in the wild type(Fig. 4A,C, arrowheads in E). By contrast, radial glial fibres in the C3G^gt/gt mutants were disorganised, lacked ventricular-to-pial orientation and were entangled(Fig. 4B,D and long arrows in F). Numerous small areas of overlap between RC2 and laminin staining could be observed in the wild-type tissue by confocal microscopy (arrows, Fig. 4E) and epifluorescence(arrows, Fig. 4G). These areas of laminin and RC2 overlap represent anchoring of radial glial fibres to the basement membrane. Areas of overlap between laminin and RC2 were similar in number in C3G^gt/gt mutants and controls at E12.5, but significantly reduced in the C3G^gt/gt mutants as compared with wild type at E13.5 and E14.5 (Fig. 4E-H, 25.0±4.0 versus 64.3±6.4 per 220 μm of basement membrane length, respectively; P=0.0065). Although normal numbers of RC2 and laminin double-positive areas were observed at E12.5, the basement membrane was already disorganised and tangling of radial glial processes was already manifest (Fig. 4B), suggesting that both disintegration of the basement membrane and lack of cell attachment contribute to the abnormal phenotype of C3G^gt/gt mutants. Lack of cell anchoring to the basement membrane, as well as the disorganisation of the radial glial processes, are possible causes of abnormal neuronal migration.

Signalling of the ECM protein reelin to cortical neuroepithelial cells is a well-established mechanism regulating cortical neuron migration, and reelin-deficient mice exhibit a preplate splitting defect. In the developing cerebral cortex, reelin is strongly expressed by specialised cells in the marginal zone, the Cajal-Retzius neurons. Accordingly, we observed reelin immunoreactivity in evenly spaced, individual cells in the wild-type marginal zone (arrowheads, Fig. 5A). C3G^gt/gt mutant marginal zone cells also expressed reelin. However, they were unevenly spaced, i.e. they were congregated in some areas and absent in other areas of the pial-most aspects of the mutant cortex primordium (arrowheads, Fig. 5B; n=3 each for wild-type and C3G^gt/gt brains). Likewise, another marker of Cajal-Retzius cells, calretinin, showed that the C3G^gt/gtmutant cells were unevenly spaced (n=3 each for wild-type and C3G^gt/gt brains; Fig. 5, compare C with D). The unaltered expression of reelin suggests that the extracellular signalling molecule reelin is not affected by the deficiency in the intracellular signalling molecule C3G. The gaps in Cajal-Retzius cell spacing corresponded to the disruption of the basement membrane in the C3G^gt/gt mutant cortex at E12.5, E13.5 and E14.5 (Fig. 5E-J; n=3 for each genotype at each developmental stage totaling nine C3G^gt/gt mutant and nine wild-type brains). In addition to radially migrating cortical neurons, the tangentially migrating reelin-expressing Cajal-Retzius cells rely on the meningeal membranes for migration and retention in their appropriate position(Borrell and Marin, 2006; Zarbalis et al., 2007). Therefore, it is likely that the abnormal distribution of the Cajal-Retzius cells is a consequence of the disorganisation of the basement membrane.

To enable us to observe cortical neuron migration in real time, we conducted organotypic slice cultures of C3G^gt/gt mutant, C3G^gt/+ heterozygous and wild-type brains infected with eGFP-expressing virus to enable visualisation (n=3 each for C3G^gt/gt mutant, wild-type and C3G^gt/+heterozygous brains).

The E12.5 wild-type brain slices formed a robust cortical plate within the 48 hours of culture as revealed by FITC-phalloidin staining of filamentous actin (Fig. 6A). By contrast,the C3G^gt/gt mutant slices failed to form a cortical plate(arrows, Fig. 6B), mirroring the defects observed in vivo (Figs 1 and 2). The viral vector infects proliferating cells in the ventricular zone, which during the 48 hours of culture give rise to cells that continue to proliferate as well as to cells that differentiate into neurons. Consequently, both differentiating and proliferating cells are eGFP labelled during the imaging period. Forty-eight hours after viral infection, many eGFP-positive cells had migrated a considerable distance from the ventricular zone towards the pial surface in the wild-type slices (arrows, Fig. 6A,C). By contrast, eGFP-positive C3G^gt/gtmutant cells predominantly remained near the ventricular surface (arrowheads, Fig. 6B,D). Time-lapse movies of wild-type and C3G^gt/gt mutant brain slices (see Movies 1, 2 in the supplementary material) as well as still images(Fig. 6E-K) show a clear ventricular-to-pial orientation of the wild-type, but not of the C3G^gt/gt mutant, cortical neurons. The wild-type neurons were predominantly bipolar and showed cell processes in pial or ventricular direction (arrowheads, Fig. 6E,G), whereas the C3G^gt/gt mutant neurons were mostly multipolar and had three or more short processes in various directions (arrowheads, Fig. 6D,F,H-K). In the time-lapse movies it was apparent that the wild-type neurons moved along their ventricular-to-pial processes (see Movie 1 in the supplementary material), whereas the C3G^gt/gtmutant neurons merely sent out small processes continually, merely to retract them again and to send out another small process in a different direction (see Movie 2 in the supplementary material). Only a very few C3G^gt/gt mutant neurons moved from their position during the timelapse imaging, although some circular movement was observed (compare Movies 1 and 2 in the supplementary material). The speed of movement of C3G^gt/gt mutant cells was significantly slower than that of wild-type cells (Table 2). C3G^gt/+ heterozygous neurons did not exhibit a migration defect and were indistinguishable from wild type (not shown).

The eGFP-labelled cells were Emx1-positive and so of dorsal telencephalic origin in the wild-type and C3G^gt/gt mutant brains and were negative for GABA, thereby excluding interneuron identity (data not shown). Therefore, the absence of radial orientation of C3G^gt/gt mutant neurons cannot be attributed to an interneuron identity and is instead an abnormal function of dorsal telencephalic neurons, which ordinarily migrate predominantly radially(Tan and Breen, 1993; Tan et al., 1998).

Dorsal telencephalic neuroepithelium explant cultures were established to monitor cortical neuron migration on a complex ECM or recombinant laminin. In wild-type cultures, cell processes emanated radially and neurons emerged and migrated away from the explants along the cell processes after 3 days(arrowheads, Fig. 7A). After 6 days of culture, wild-type explants were surrounded by a corona of radially migrating neurons (arrowheads, Fig. 7C). In addition to migration along explant-derived cell processes, wild-type cells also migrated on complex matrix independently of cell processes. By contrast, C3G^gt/gt telencephalic explants did not give rise to processes or migrating cells throughout the culture period (Fig. 7, compare A with B and C with D) (n=14 each for wild-type, C3G^gt/+ and C3G^gt/gt brains). Similarly, on the defined laminin substrate, C3G^gt/gtexplants failed to generate processes or migrating neurons, whereas wild-type explants exhibited both processes and migrating neurons(Fig. 7, compare E with F). C3G^gt/+ heterozygous explants were indistinguishable from wild-type explants (not shown). As C3G^gt/gt mutant cells were unable to send out processes, we can only comment on their ability for cell-process-independent migration in this assay and conclude that, unlike wild-type cells, C3G^gt/gt mutant cells were incapable of migration on complex or defined matrix. In time-lapse movies, wild-type cortical neurons were observed migrating centrifugally away from the explants. Some wild-type cells first moved away, then towards the explants and finally reverted to centrifugal migration. A migration velocity of 10 μm/hour was observed (Fig. 7I-K and see Movie 3 in the supplementary material). The speed of 10 μm/hour is similar to that observed previously for cortical neurons(Noctor et al., 2004). By contrast, C3G^gt/gt mutant cortical neuroepithelial explants showed a complete absence of migration. Note, however, that although no cells migrated out of the C3G^gt/gt explants, movement within the C3G^gt/gt explants was apparent and the explants increased in size during the culture period(Fig. 7L-N and see Movie 4 in the supplementary material; n=8 each for wild-type, C3G^gt/+ and C3G^gt/gt brains).

Like wild-type, C3G^gt/gt mutant explant cells expressed the neuronal marker β-tubulin type III(Fig. 7G,H). Moreover, isolated wild-type and C3G^gt/gt mutant cortical neural precursor cells formed similar percentages of neurons when induced to differentiate on laminin in vitro (not shown). Therefore, C3G^gt/gtneuroepithelial cells differentiated into cells expressing neuronal marker proteins, but failed to send out processes and were unable to migrate on laminin or complex matrix. The C3G mutant cell attachment and migration defects observed in explant cultures are more severe than the migration defects observed in vivo or in brain slice cultures, presumably because the substrate offered for attachment in the explant cultures is less supportive than the extracellular environment in the brain. It is noteworthy,however, that the same substrate is sufficient for the attachment and migration of wild-type cells.

Since the preplate splitting defect in the C3G^gt/gtmutant embryos resembles the preplate splitting defect in reelin mutant mice,we postulated that C3G might be an effector of reelin signalling. Indeed, C3G is phosphorylated (activated) in response to reelin stimulation in E16.5 wild-type, but not in Dab1 mutant, cortical neurons; furthermore, the downstream target of C3G activation, Rap1 is loaded with GTP upon reelin stimulation (Ballif et al.,2004). As previously shown by others(Ballif et al., 2004), we observed that E16.5 wild-type cortical neurons treated with reelin(Fig. 8A) exhibited induction of C3G tyrosine phosphorylation (Fig. 8B) corresponding to C3G activation. As the major effect of C3G activation is Rap1 GTP loading [activation(Ohba et al., 2001)], we examined Rap1 activation in C3G mutant and wild-type cells. To obtain sufficient amounts of material for this experiment we used E10.5 neural precursor cells expanded in vitro as described previously(Voss et al., 2006). After 4 hours of mitogen withdrawal, E10.5 neural precursor cells were treated with or without reelin. GTP loading of Rap1 (activation) was higher in mock-treated wild-type cells than in mock-treated C3G mutant cells(Fig. 8C). Moreover,reelin-treated wild-type cells showed higher levels of Rap1 GTP loading than C3G mutant cells. Reelin-treated C3G mutant cells had 5-fold lower levels of GTP-loaded Rap1, even compared with unstimulated wild-type cells (Fig. 8C). C3Gmutant cells responded to reelin stimulation with modest levels of Rap1 GTP loading, which may be attributed to the fact that our C3G^gt allele is not a null allele and produces approximately 5% normal C3G protein (Voss et al., 2003).

In summary, C3G is activated by reelin and is required for Rap1 activation in response to reelin. C3G mutant mice phenocopy the reelin mutant preplate splitting defect. Together, our findings suggest that C3G is an essential effector of reelin signalling in cortical development.

In the present study, we have shown that C3G^gt/gthypomorphic mutant mice have severe defects in cortical development. The cortical malformations resulting from insufficient C3G fall into two distinct classes (Fig. 9).

Firstly, C3G deficiency causes a fundamental defect in one of the earliest steps of cortical neuron migration, i.e. C3G-deficient cortical neurons fail to split the cortical preplate. The first C3G-deficient post-mitotic neurons are positioned near the pial surface and form an apparently normal preplate. However, prospective layer VI neurons fail to migrate into the preplate and therefore fail to split the preplate into marginal zone and subplate. Instead,the second cohort remains below the preplate. This phenotype has previously been found in four cases only, all mutations of members of the reelin signalling pathway.

Secondly, we found areas of the developing cortex where the basement membrane was disrupted and groups of cortical neurons had invaded the marginal zone. This phenotype is similar to the undulating cortex previously observed in mice with mutations causing a disruption of the basement membrane or disruption of the interactions of cells with the basement membrane.

The defect in cortical preplate splitting in C3G^gt/gtmutant embryos is similar to defects found in mice carrying mutations in the reelin gene (D'Arcangelo et al.,1995; Sheppard and Pearlman,1997), ApoER2 (Lrp8 -Mouse Genome Informatics)and Vldlr (Trommsdorff et al.,1999), Src and Fyn(Kuo et al., 2005) and Dab1 (Howell et al.,1997; Rice et al.,1998; Sheldon et al.,1997). Mutations in these genes of the reelin signalling pathway cause preplate splitting defects and subsequently show defects in cortical layer formation.

Signalling through the reelin/ApoER2/Vldlr/Src/Fyn/Dab1 signalling pathway(D'Arcangelo et al., 1995; Howell et al., 1997; Kuo et al., 2005; Sheldon et al., 1997; Trommsdorff et al., 1999) is required for the development of normal cortical lamination. However, only the proximal aspects of this signalling pathway and some, but not all, downstream events are well understood (D'Arcangelo,2006; Tissir and Goffinet,2003). Reelin binding to ApoER2/Vldlr activates the Src family kinases Src and Fyn, which leads to Dab1 phosphorylation. How these events are connected to the downstream effectors, such as actin skeleton reorganisation and microtubule motor complex function, is unknown(D'Arcangelo, 2006; Tissir and Goffinet, 2003). Although additional components of the reelin signalling pathway have been identified in vitro in recent years, their function in neuronal migration in vivo has not been elucidated (D'Arcangelo,2006). In particular, it was recently reported that C3G phosphorylation is the first phosphorylation event in reelin-stimulated neurons, along with Dab1 phosphorylation(Ballif et al., 2004). However,that study did not address whether or not C3G is functionally relevant to neuronal migration. We provide direct evidence here that lack of C3G causes similar defects in early corticogenesis to reelin mutation. Furthermore, we show that in the C3G-deficient state, the effects of reelin mediated by C3G,namely increased Rap1 GTP loading, are largely abrogated. As C3G is required for filopodia formation in fibroblasts(Radha et al., 2007) and Rap1 for protrusion formation in a number of cell lines(Arthur et al., 2004),reelin-induced C3G-mediated Rap1 activation may be required for cell process extension.

Recent detailed studies have shown that there are multiple modes of neuron migration and that migration might involve a number of discrete stages. Initially, neurons leave the ventricular zone by a process of somal translocation, in which neurons retain connection to the pial surface but lose connection to cells lining the ventricle and retract to the pia. Subsequently,neurons move away from the ventricles by locomotion, a process by which cells migrate along radial glia, but may also leave glia and reside temporally in the subventricular zone before resuming migration. While dissociated from the radial glia, these cells have a characteristic multipolar phenotype(Noctor et al., 2004; Tabata and Nakajima, 2003). Interestingly, the majority of C3G-deficient neurons display a multipolar morphology and are arrested in migration in brain slice cultures. This phenotype is reminiscent of the dominant-negative and RNAi depletion phenotypes of filamin A and Lis1 (Pafah1β1 -Mouse Genome Informatics),which both cause arrest at the multipolar stage of cortical neuron migration(Nagano et al., 2004; Tsai et al., 2005). Failure to return to the bipolar morphology may impede further migration. It is therefore possible that C3G might provide a link between the well-known upstream aspects of the reelin signalling pathway and effector events involving filamin A and Lis1 function in the contexts of the actin cytoskeleton and the microtubule motor complex.

The disorientation of the radial glial processes in the C3G^gt/gt mutants surpasses that observed in Reeler mutants. In Reeler mutants, radial glial processes span the ventricular-to-pial extent of the developing cortex at an oblique angle(Hartfuss et al., 2003; Luque et al., 2003; Magdaleno and Curran, 2001). In C3G^gt/gt mutants, radial glial processes are completely disorganised suggesting, in addition to a defect in reelin signalling, the failure of a second cellular process, as discussed below.

In addition to the overlap between the C3G^gt/gt and the reelin mutant phenotypes, the C3G^gt/gt mutant phenotype shows similarities to the abnormalities caused by disruption of the basement membrane (Halfter et al.,2002) or failure to attach to the basement membrane(Beggs et al., 2003; Graus-Porta et al., 2001; Niewmierzycka et al., 2005). For example, mouse foetuses with nervous-system-specific deletion of theβ1 integrin gene (Itgb1) show a discontinuous basement membrane as evidenced by discontinuous laminin staining. Moreover, they exhibit invasion of the marginal zone by cortical plate neurons(Graus-Porta et al., 2001). Like the β1 integrin-deficient developing cerebral cortex, the C3G^gt/gt mutant cortex primordium exhibits discontinuity of laminin deposits in the basement membrane and invasion of the marginal zone and even of the pericerebral space by cortical neurons, suggesting a total collapse of basement membrane function. As interactions of laminins with integrin receptors are required to localise laminin at appropriate sites for basement membrane assembly (Schwarzbauer,1999), and as C3G is required for integrin signalling and function(Arai et al., 2001; Arai et al., 1999; Uemura and Griffin, 1999; Voss et al., 2003; Voss et al., 2006), the observed discontinuity of the basement membrane is likely to be the result of a lack of integrin signalling in the absence of C3G.

Notably, RC2-positive radial glia cell fibres, which are important for cortical neuron migration, largely fail to connect to the discontinuous laminin deposits present in the C3G^gt/gt mutants. Perhaps as a consequence of their failure to connect to the ECM, the C3G^gt/gt mutant radial glia fibres are disorganised,entangled and lack ventricular-to-pial orientation. Furthermore, our in vitro findings show that C3G is required for neuron-ECM interaction for process extension and migration on ECM. Previous reports have established that C3G is activated in response to the engagement of number of different substrates and different integrin heterodimers in a variety of cell types(Arai et al., 2001; Arai et al., 1999; Uemura and Griffin, 1999). Functionally, C3G is essential for embryonic fibroblast cell adhesion and regulation of migration (Ohba et al.,2001; Voss et al.,2003) and for the establishment/maintenance of β1 integrin-and paxillin-positive focal adhesions in embryonic fibroblasts(Voss et al., 2003). Furthermore, laminin-induced Rap1 activation is C3G dependent in differentiating neural precursor cells(Voss et al., 2006). Our current data show that, similar to fibroblasts, neurons have an essential requirement for C3G for interaction with ECM proteins.

In conclusion, we have shown that the Ras signalling molecule C3G is crucial for cortical neuron migration. Our results are consistent with the hypothesis that reelin signalling and integrin signalling converge on C3G activity in migrating cortical neurons. The C3G mutant phenotype combines both elements-reelin loss and integrin loss - and therefore exceeds either of these mutants in severity.

We thank S. Michajlovic, N. Ashman and A. Morcom for technical assistance,T. Kilpatrick for discussion, and F. Gage and T. Curran for reagents. This work was supported by the Walter and Eliza Hall Institute of Medical Research and by the Australian NHMRC.