The adhesion signaling molecule p190 RhoGAP is required for morphogenetic processes in neural development

Members of the Rho family of small GTPases, which includes the Rho, Rac and Cdc42 proteins, function as critical regulators of actin cytoskeleton organization. As such, these proteins mediate a variety of cellular processes, including migration, adhesion and shape change (Hall, 1998; Van Aelst and D’Souza-Schorey, 1997). These cellular functions of the Rho GTPases have recently been linked to several of the morphogenetic events associated with normal embryonic development in mammals and in other multicellular organisms (Settleman, 1999). Like the Ras GTPases, Rho proteins cycle between a GDP-bound inactive form and a GTP-bound active form in a tightly regulated manner (Nobes and Hall, 1994). Regulation of the nucleotide state of the Rho GTPases is accomplished by the action of three major classes of proteins: guanine nucleotide exchange factors, guanine nucleotide dissociation inhibitors and GTPase activating proteins (GAPs). GAPs function by stimulating the relatively weak intrinsic GTP hydrolyzing activity of their substrate GTPases, thereby inactivating them. Among the numerous RhoGAPs that have been described, p190 RhoGAP, together with a closely related protein, p190-B, accounts for a substantial fraction of the total Rho inhibitory activity in cultured cells (Burbelo et al., 1995; Settleman et al., 1992a; Vincent and Settleman, 1999), and both proteins are widely expressed in adult mammalian tissues (Burbelo et al., 1998; Settleman et al., 1992a). P190 RhoGAP was first identified as the major binding partner of p120 RasGAP in Src-transformed cells (Ellis et al., 1990). Thus, p190 RhoGAP may integrate signals transduced by the Ras and Rho family GTPases. P190 RhoGAP contains an amino-terminal GTPase domain (Foster et al., 1994) and a carboxy-terminal RhoGAP domain that preferentially interacts with the Rho GTPase (Settleman et al., 1992b). Although the precise role of the GTPase domain of p190 RhoGAP has not been established, it appears to regulate the ability of p190 RhoGAP to function as a Rho regulator in cultured cells (Tatsis et al., 1998).

Evidence for a role for p190 RhoGAP in regulating the actin cytoskeleton comes from several studies with cultured fibroblasts. For example, overexpression of p190 RhoGAP causes cells to become rounded with long, beaded extensions, and is associated with collapse of the actin cytoskeleton (Tatsis et al., 1998). In addition, lysophosphatidic acid (LPA)-induced actin stress fiber formation, which is Rho-mediated, is inhibited by microinjection of the RhoGAP catalytic domain of p190 RhoGAP (Ridley et al., 1993). P190 RhoGAP has also been observed to rapidly co-localize with polymerized actin upon growth factor stimulation (Chang et al., 1995).

Several lines of evidence indicate that p190 RhoGAP mediates a signal transduction pathway downstream of integrins, the receptors for extracellular matrix proteins. For example, expression of the isolated amino-terminal region of RasGAP, which is constitutively associated with p190 RhoGAP, results in a reduction in focal contacts and an impaired ability to adhere to fibronectin (McGlade et al., 1993). Furthermore, p190 RhoGAP is rapidly recruited to the plasma membrane upon antibody-induced integrin engagement, where it co-localizes with cortical actin (Burbelo et al., 1995). Finally, p190 RhoGAP is recruited to the cytoskeleton following adhesion of fibroblasts to fibronectin (Sharma, 1998). These observations suggest that p190 RhoGAP mediates an adhesion signaling pathway and that it acts through its ability to regulate the Rho GTPases, which have previously been shown to be required for integrin-dependent adhesion signaling (Clark et al., 1998; Hotchin and Hall, 1995).

We have examined the in vivo function of p190 RhoGAP by a targeted gene disruption strategy in mice. We first determined that p190 RhoGAP is highly and specifically expressed in the developing nervous system of the mouse, suggesting a role for the protein in neural development. Mice lacking a functional p190 RhoGAP exhibit several specific defects in neural development that appear to reflect a role for the protein in adhesion signaling. We also find evidence of a role for PKC in regulating p190 RhoGAP’s response to adhesion molecule activation. Taken together, our findings suggest that p190 RhoGAP is critical for several aspects of tissue morphogenesis in the developing nervous system by regulating Rho GTPase-mediated actin organization in response to the engagement of adhesion molecules.

Embryos were fixed in 4% formaldehyde by immersion (at E12.5) or transcardiac perfusion (at E15.5 and E18.5) and cryoprotected in Dulbecco’s PBS containing 20% sucrose. 20 μm sections were cut by cryostat. [³⁵S]UTP-labeled riboprobes were synthesized from a pBluescript SK+/− phagemid containing p190 RhoGAP cDNA (bp 1199-1792). Hybridizations were carried out under standard conditions. Slides were dried and exposed to β-Max film. Sections were immersed in Kodak NTB-2 autoradiography emulsion.

Whole-mount in situ hybridization was carried out essentially as previously described (Wilkinson, 1992). Wild-type E8.5 embryos were fixed in 4% paraformaldehyde at 4°C for 3-12 hours. Samples were rinsed in PBT (PBS + 0.1% Tween 20), dehydrated in methanol and stored at −20°C. After rehydration, the embryos were hybridized to riboprobes overnight at 70°C in hybridization buffer (50% formamide, 5× SSC pH 4.5, 50 μg/ml yeast RNA, 1% SDS, 50 μg/ml heparin) containing 1 μg/ml digoxigenin-labelled RNA probe. pBluescript SK+/− plasmid containing p190 RhoGAP coding sequence (bp 1694-3610) served as template for synthesis of probes. Following extensive washing, embryos were incubated overnight at 4°C with alkaline phosphatase-conjugated anti-digoxigenin antibody that had been preabsorbed with heat-inactivated mouse embryo powder. Alkaline phosphatase activity was visualized using nitroblue-tetrazolium chloride and 5-bromo-4-chloro-3-indolyl phosphate for 1-2 hours at room temperature.

Analysis of p190 RhoGAP protein expression, tyrosine phosphorylation, and RasGAP association was carried out as previously described (Hu and Settleman, 1997). Subcellular fractionation was performed by tissue homogenization in ice-cold hypotonic buffer (20 mM Hepes pH 7.5, 2 mM MgCl₂, 10 mM KCl) containing protease and phosphatase inhibitors, followed by centrifugation in a microfuge (10,000 g for 4 minutes at 4°C) to yield the crude extract. Ultracentrifugation (100,000 g for 20 minutes at 4°C) of the crude extract yielded the soluble fraction (supernatant), whereas lysis of the obtained high-speed pellet in hypotonic buffer supplemented with 0.5% NP-40 and 400 mM NaCl, followed by centrifugation (25,000 g for 10 minutes at 4°C), yielded the insoluble fraction.

Genomic clones containing p190 RhoGAP coding sequence were isolated from a 129/SvJ genomic library. The targeting strategy took advantage of the fact that the amino-terminal 80% of the p190 RhoGAP coding sequence is contained in a single large exon. The p190 RhoGAP targeting vector was constructed by inserting a 1.4 kb BamHI-EcoRI 5′ genomic fragment and a 6.5 kb XbaI 3′ genomic fragment in the pPNT vector (Tybulewicz et al., 1991) in opposite transcriptional orientation to the PGK-neo^R cassette. This resulted in a deletion of 1.5 kb of the first exon, thereby eliminating the translation initiation site. NotI linearized targeting vector was electroporated into D3 ES cells. Transfected ES cells were selected for growth in G418 and gancyclovir. Correctly targeted ES clones (5 of 500 screened) were detected by Southern blot analysis of HindIII digested genomic DNA, using a 5′ external probe. Two 3′ external probes were used to confirm the absence of rearrangements 3′ of the targeted region. ES cells from each of the five identified p190 RhoGAP heterozygous clones were injected into C57BL/6J blastocysts to generate chimeric mice. Germline transmission was determined by crosses between chimeras and C57BL/6J mice and confirmed by Southern blot analysis of tail DNA. Subsequent genotypic analysis was performed by PCR. ES cell culture and blastocyst injections were performed essentially as described by Hogan et al. (1994).

Mouse embryonic fibroblasts (MEFs) were prepared as previously described (Hogan et al., 1994) and maintained in DMEM/15% fetal bovine serum. Cos-7 cells, maintained in standard culture conditions, were transfected using the DEAE-Dextran method with the epitope-tagged p190 expression vector, RcHAp190 (Hu and Settleman, 1997). For immunostaining with anti-p190 RhoGAP or anti-HA-tag antibodies, cells were plated for the indicated times on glass coverslips that were coated with fibronectin (0.001% w/v in PBS), poly-L-lysine (0.01% w/v in H₂O) or were untreated. Cells were fixed in 4% paraformaldehyde in PBS (20 minutes), rinsed in PBS, permeabilized in 0.1% Triton X-100/PBS (2 minutes) and blocked with 1% normal goat serum, 0.1% BSA in PBS (1 hour). Primary antibodies were diluted in 0.1% BSA/PBS and applied to coverslips for 1 hour (anti-p190 RhoGAP, Transduction Laboratories, 1:50; anti-HA-Tag 12CA5, 1:1). Coverslips were rinsed in PBS and incubated in Cy3-conjugated secondary antibody (Jackson ImmunoResearch), diluted in PBS (1:300) in the presence of FITC-conjugated phalloidin (Sigma) (1:300) for 30 minutes. Coverslips were rinsed in PBS, mounted on glass slides and viewed with a Zeiss fluorescence microscope.

A quantitative cell adhesion assay was carried out in 96-well microtitre plates, essentially as previously described (McGlade et al., 1993). Briefly, wells were coated overnight at 4°C with fibronectin (0.1-10 μg/ml) diluted in PBS (25 μl/well). Wells were washed several times with PBS and blocked with PBS containing 2.5 mg/ml BSA for 2 hours at 37°C. Fibroblasts were trypsinized, centrifuged in DMEM-15% FCS (5 minutes, 1500 rpm at room temperature) and washed three times in serum-free DMEM. Cells were then resuspended in DMEM containing 2.5 mg/ml BSA, to a concentration of 3×10⁵ cells/ml, and plated in triplicate on fibronectin-coated wells (100 μl/well). Cells were allowed to adhere at 37°C, after which wells were aspirated and washed once with PBS. Attached cells were fixed in 3.7% paraformaldehyde for 30 minutes and stained overnight with 0.5% Toluidine Blue in 3.7% paraformaldehyde at room temperature. Wells were rinsed with PBS and cell adhesion was quantitated by measuring absorbance at OD 540 using an automated microtiter well plate reader.

To analyze reorganization of the actin cytoskeleton during adhesion and spreading of MEFs on fibronectin, glass coverslips were coated with 10 μg/ml fibronectin in PBS for 30 minutes at 37°C in 12-well plates. Wild-type and mutant MEFs were trypsinized and plated at low density (10⁴/well) on the fibronectin-coated coverslips for the times indicated, washed once with PBS, fixed with 4% formaldehyde in PBS for 15 minutes at room temperature and stained for actin as described above.

Newborn mice were sacrificed by CO₂ asphyxiation and fixed by immersion in Bouin’s fixative. Coronal paraffin sections were stained with Haematoxylin and Eosin. E15-E17 embryos were fixed by transcardiac perfusion with 4% paraformaldehyde. 1 μm plastic sections were taken through selected slabs at 50 μm intervals, so that the entire region containing the forebrain commissures, optic nerves and optic chiasm were sampled. Sections were heat dried onto microscope slides, stained with Toluidine Blue, coverslipped with Pro-Texx mounting medium and photographed. For Nissl staining, brains were fixed by immersion (4% paraformaldehyde) followed by cryoprotection in PBS/20% sucrose. Coronal cryostat sections were then stained with Cresyl Violet acetate (0.5% in 1% acetic acid) for 10 minutes.

Whole embryos (E10.5-E14.5) or heads (E16.5) were fixed in 4% paraformaldehyde at 4°C for 2-3 hours or overnight, respectively. Following fixation, whole embryos, dissected brains or whole heads were washed with PBS, cryoprotected in PBS-20% sucrose overnight, frozen in OCT and cut on a cryostat. For immunofluorescence, sections (10-14 μm) were air-dried for 30 minutes, washed with PBS, permeabilized in PBS-0.1% Triton X-100 for 10 minutes, blocked with PBS-10% normal goat serum (NGS) containing goat anti-mouse F(ab′)₂ fragments (10 μg/ml) for 30 minutes, washed briefly with PBS and then incubated with primary antibody diluted in PBS-1% NGS for 2 hours. Subsequently, sections were washed with PBS, followed by incubation with Cy3-conjugated secondary antibody for 1 hour. After PBS washes, sections were coverslipped using polyvinyl alcohol-based aqueous mounting medium. For F-actin staining, sections were treated as above, omitting the F(ab′)₂ fragments in the block solution. Sections were incubated with fluorescent phalloidin for 1 hour followed by PBS washes and mounting.

The cortactin primary antibody (4F11, Upstate Biotechnology, Inc.) was used at 1:200. Goat anti-mouse F(ab′)₂ fragments (Jackson ImmunoResearch) were used as indicated. The Rat-401 antibody (Developmental Studies Hybridoma Bank) was used at 1:1 dilution. The p190 RhoGAP monoclonal antibody (Transduction Laboratories) was used at 1:50. Cy3-conjugated goat anti-mouse secondary antibody (Jackson ImmunoResearch) was diluted 1:500. FITC-conjugated phalloidin (Sigma) was diluted 1:200, rhodamine-conjugated phalloidin (Sigma) was diluted 1:500.

NIH3T3 cells were transfected by calcium phosphate with either RCHAp190 (wild type p190 RhoGAP) or the ΔR6 truncation mutant corresponding to the amino-terminal deletion of the GTPase domain. The ΔR6 truncation mutant was made by deleting the amino-terminal region of p190 RhoGAP at an XbaI site that is 1.1 kb from the start codon. 24 hours post-transfection, cells were re-plated onto glass coverslips, and after an additonal 24 hours, cells were fixed and immunostained with either 12CA5 antibody (RCHAp190) or p190 RhoGAP monoclonal antibody A5D12. Cells that exhibited protein expression were scored, using fluorescence microscopy, as either being morphologically unaffected, partially collapsed, or completely collapsed.

Embryos were washed in PBS, then fixed in Karnovsky’s fixative (pH 7.5) at room temperature for 30 minutes. Subsequently, embryos were rinsed in 0.125 M sodium cacodylate buffer (pH 7.3), 310 mOsm, postfixed in 2% osmium tetroxide for 45 minutes, rinsed in 0.125 M sodium cacodylate buffer (pH 7.3) and dehydrated in graded alcohols to 100% ethanol. Following critical point drying and sputter coating with gold/palladium, samples were viewed on an Amray 1000 scanning electron microscope.

In vitro PKC assays were performed in a solution containing 20 mM MOPS buffer, pH 7.2, 25 mM β-glycerophosphate, 1 mM sodium orthovanadate, 1 mM DTT, 1 mM CaCl₂, 0.1 μM ATP and 10 μCi [γ-³²P]ATP. Where indicated, the reaction was supplemented with 20 ng purified brain PKC (Upstate Biotechnology, Inc.), 0.1 mg/ml phosphatidylserine and diglycerides and/or 100 ng of purified, baculovius-produced p190 RhoGAP protein. Following incubation at 30°C for 10 minutes, samples were analyzed by 7.5% SDS-PAGE, followed by autoradiography. For in vivo PKC phosphorylation assays, early passage MEFs were serum-starved (DMEM-0.1% FCS) overnight, preincubated for 2 hours in phosphate-free DMEM containing 0.1% phosphate-free, dialyzed FCS, then incubated in the same medium containing ³²P (200 μCi/ml) for 3 hours. Cells were then stimulated with TPA (100 ng/ml) for the indicated times, washed twice with ice-cold PBS, scraped in 1 ml of ice-cold PBS and centrifuged for 1 minute at 4000 rpm at room temperature. Cell pellets were lysed in 50 mM Hepes pH 7.4, 150 mM NaCl, 1.5 mM MgCl₂, 5 mM EGTA, 10% glycerol, 1% Triton X-100 with protease and phosphatase inhibitors. Debris was removed by centrifugation and lysates were precleared by incubation with protein A beads. Proteins were immunoprecipitated with anti-p190 RhoGAP and protein A-Sepharose for 2 hours at 4°C. Immune complexes were washed four times in 20 mM Hepes pH 7.5, 150 mM NaCl, 10% glycerol, 0.1% Triton X-100, boiled in gel loading buffer and separated by 7.5% SDS-PAGE, which was followed by autoradiography.

Cos-7 cells transfected with RcHAp190 RhoGAP were plated on glass coverslips in 12-well plates. 48 hours after transfection a subset of wells was treated for 30 minutes with the phorbol ester TPA (100 ng/ml) by direct addition to the medium. Mouse 3T3 fibroblasts were cultured overnight on glass coverslips and similarly treated with TPA. Cells were then washed once with PBS, fixed with 4% formaldehyde in PBS for 15 minutes at room temperature and stained for p190 RhoGAP and actin as described above.

We previously reported that the p190 RhoGAP gene is expressed ubiquitously in adult rodents (Settleman et al., 1992a). To determine whether the expression of p190 RhoGAP is developmentally regulated we performed in situ hybridization analysis on tissue sections of mouse embryos at several gestational stages. We found that p190 RhoGAP mRNA is specifically expressed at high levels in the brain, spinal cord and eyes of developing mouse embryos (Fig. 1D-F). In brain, expression appears to be widespread, with some enrichment in the cortical plate (Fig. 1A-C). The p190 RhoGAP protein product and its binding partner RasGAP are also expressed uniformly throughout neural development (E12-P5) (Fig. 1G). Thus, during embryogenesis, the expression of p190 RhoGAP is largely restricted to the nervous system. In addition, p190 RhoGAP protein is expressed in all regions of the mature nervous system of adult animals (Fig. 1H). We have also found that p190 RhoGAP is expressed in a variety of neuronal and glial cell lines, suggesting that it is not restricted to a particular cell type in the nervous system (data not shown).

To identify a potential role for p190 RhoGAP in neural development, we disrupted the gene in mice by targeted homologous recombination in embryonic stem cells (Fig. 2A-C). The targeting strategy involved deleting 1.5 kb of amino-terminal coding sequence from a single large exon (3.8 kb) that encodes most of the p190 RhoGAP protein. Mice heterozygous for the targeted allele develop normally and display no abnormalities, and intercrossing these animals yields p190 RhoGAP^−/− animals at the expected frequency at birth (25%). The majority of p190 RhoGAP^−/− mice are grossly indistinguishable from their normal littermates throughout gestation (Fig. 2F), although 95% die within the first 2 days after birth (Fig. 2G). The remaining mutant animals are runted and die within 3 weeks.

Immunoblots of embryo brain lysates using an amino-terminal p190 RhoGAP antibody indicated that the full-length protein is not expressed in homozygous mutant animals (Fig. 2C). However, with other p190 RhoGAP-specific antibodies, we observed that an altered form of the protein is expressed from the targeted allele by an internal translational initiation (Fig. 2D). This truncated form of p190 RhoGAP specifically lacks the amino-terminal GTPase domain (Fig. 2E), and is expressed at levels comparable to the wild-type protein.

Several lines of evidence indicate that the truncated form of p190 RhoGAP expressed in the mutant mice represents either a severely defective loss-of-function allele or a null allele. First, we have not observed any defects in more than 100 heterozygous mutant animals that have been examined. Thus, by genetic criteria, the p190 RhoGAP mutation is a loss-of-function allele and is not functioning in a dominant-interfering manner. Second, we have found that each of the known functional domains of p190 RhoGAP is defective in the mutant protein. Thus, the GTPase function is lost by the deletion of the entire catalytic region. In addition, unlike the wild-type protein, the mutant protein fails to undergo tyrosine phosphorylation or RasGAP association in brain (Fig. 3A). The GTPase domain has recently been shown to be required for normal RhoGAP activity in vivo (Tatsis et al., 1998) and we have similarly observed a substantial reduction in the ability of the GTPase-deficient form of the protein to disrupt the actin cytoskeleton in transfected cells, despite retention of a functional RhoGAP catalytic domain (Table 1). In addition, we have determined that the mutant protein fails to localize properly within cells (Fig. 3B,C). Approximately one-third (34%) of the wild-type protein is present in a detergent-insoluble fraction, whereas only a small fraction of the mutant protein (7%) is present in the same insoluble fraction of brain lysates from p190 RhoGAP^+/− animals (Fig. 3B). Differential subcellular localization is also indicated by the distinct immunofluorescent staining patterns seen in wild-type and p190 RhoGAP mutant mouse embryonic fibroblasts (MEFs) (Fig. 3C). Taken together, these results indicate that the truncated p190 RhoGAP is a severely defective loss-of-function protein, but do not rule out the possibility that a complete deletion of the gene might result in a more substantial phenotype.

To identify potential defects in cells lacking a functional p190 RhoGAP, we first examined MEFs. Homozygous mutant cells in culture were compared to wild-type and heterozygous mutant cells in several assays of growth and morphology. Overall, no significant differences were observed between the cell populations in comparisons of proliferation rates, morphology, actin organization, migration, spreading, or adhesion to extracellular matrix (Fig. 4 and data not shown). In addition, all cell lines were readily established as stable, immortalized ‘3T3’ lines that also exhibit similar properties (data not shown). Since we have previously observed that the closely related p190-B protein is also expressed in fibroblasts, and actually accounts for most of the RhoGAP activity in those cells (Vincent and Settleman, 1999), it is likely that a redundant function of p190-B masks the absence of a functional p190 RhoGAP in these cells.

A detailed histological analysis of developing embryos revealed that defects in p190 RhoGAP^−/− animals appear to be largely restricted to the developing nervous system. A small percentage of homozygous mutant animals (less than 2%) also exhibit omphalocele, which is a failed fusion of the abdominal cavity (not shown). No abnormalities in any other organs were observed grossly or microscopically. Moreover, none of numerous p190 RhoGAP heterozygous mutant animals (more than 100 examined) was found to exhibit any of the defects described below. Consistent with the observed enriched expression of p190 RhoGAP in embryonic eyes, we found that all of the p190 RhoGAP^−/− animals exhibit abnormalities in eye development (more than 50 examined). Eyes from p190 RhoGAP^−/− animals display an apparent hyperplasia of the retinal pigmented epithelium (RPE) (Fig. 5B,D). Upon dissection, we determined that eyes from p190 RhoGAP mutant embryos are significantly smaller than normal and exhibit a defect in closure of the optic fissure, leading to coloboma (Fig. 5A-D). Most likely, the observed RPE hyperplasia is a secondary consequence of the neural retina defect, as has been previously described in adult eye tissue (Campochiaro et al., 1994).

Closure of the optic fissure generally occurs between E11.5 and E14.5 and marks the end of the morphogenetic phase in the formation of the eye (Hero, 1990). To determine the cellular basis for the observed eye defect, we examined retinal sections from wild-type and mutant embryos at several relevant developmental stages. To identify potential abnormalities in actin cytoskeleton organization, we compared phalloidin-stained eye sections (Fig. 5E-J). This analysis revealed abnormal folding of the retina in the mutants, and although no obvious differences in the distribution of polymerized actin were seen, the persistent invagination of the actin-rich basal membrane in the mutants potentially reflects a subtle defect in actin organization in the absence of functional p190 RhoGAP. Moreover, the progression of the mutant phenotype through these stages of development clearly indicates that the observed defect occurs during the normal course of eye development, and does not arise subsequent to the developmental phase.

Despite the defect in closure of the optic fissure, the retinal layers appear to be normal with regard to cell number and cell type composition (data not shown). Examination of the localization of the cellular markers cortactin, a cortical actin-binding protein (Fig. 5K-O) and cadherin (not shown) indicated no defects in the polarization of retinal epithelial cells, but revealed an abnormal morphology of cells lining the optic fissure in the mutants. Thus, the eye defects in p190 RhoGAP mutant mice appear to be associated exclusively with defective morphogenesis of the neural retinal tissue, and are associated with aberrant epithelial cell shapes.

The majority of brains dissected from p190 RhoGAP^−/− newborn animals are normal in size, weight, and external appearance (see next section). However, they always exhibit a substantial cleft between the cerebral hemispheres. Histological analysis revealed that these mice completely lack a corpus callosum, the major bundle of midline-crossing cortical axons in the forebrain (50 out of 50 examined), whereas in wild-type and heterozygous mutant littermates this structure is readily seen (35 examined) (Fig. 6A,B). Callosal axons in mutant mice extend normally to the midline but fail to cross. Instead, they remain ipsilateral and form neuromas, called Probst bundles (Probst, 1901). In addition to agenesis of the corpus callosum, formation of other forebrain midline structures is disrupted in p190 RhoGAP mutants. The anterior and hippocampal commissures are absent at the midline or minimally formed in all mutants (Fig. 6B and data not shown). Furthermore, an abnormal morphology of the third ventricle is seen in all of the mutants (Fig. 6C,D).

Agenesis of the corpus callosum can be caused by defects in axon guidance and/or midline fusion of the cerebral hemispheres. To distinguish between these possibilities, we compared brain sections from wild-type and p190 RhoGAP^−/− embryos at several developmental stages. We found that outgrowth of callosal axons towards the midline is normal in mutant mice (not shown). However, fusion of the cerebral hemispheres, which is completed at E17 in wild-type embryos, is obviously delayed in the mutant mice, as indicated by the persistence of the cerebral longitudinal fissure (Fig. 6E,F). Furthermore, at E16.5, when hemisphere fusion is underway, p190 RhoGAP protein is specifically enriched in cells along the midline of the forebrain, together with F-actin, although no obvious differences were seen in the appearance of actin in the corresponding region of mutant embryos (data not shown).

In addition to the midline fusion defect in the forebrain of p190 RhoGAP mutant animals, these mice display a markedly disorganized layering of the cerebral cortex. Whereas in wild-type embryos, neurons in the cortical plate are densely packed (Fig. 6G,I), the neurons in the same region in the mutant animals are much more diffusely organized (Fig. 6H,J), suggesting a defect in neuronal migration. Birth-dating of cortical neurons by BrDu incorporation during embryonic development did not reveal any major defect in migration associated with particular layers of the developing cortex (data not shown). In addition, the density of radial glial fibers, which are used to guide the migration of cortical neurons from the ventricular zone, appears to be essentially normal in the mutants (Fig. 6K,L). Other areas of the brain, including the brain stem and cerebellum, do not display any obvious malformations in p190 RhoGAP^−/− animals.

Although most p190 RhoGAP^−/− newborn animals appear grossly normal externally, approximately 30% of the mutant animals exhibit exencephaly (Fig. 7A and 7B). Exencephaly typically reflects a defect in closure of the anterior neural tube, which normally occurs between E8.5 and E9.5 (Copp et al., 1990). Therefore, E10.5 embryos were examined by scanning electron microscopy for neural tube defects. Approximately 30% of p190 RhoGAP^−/− embryos (but no wild-type or p190 RhoGAP^+/− embryos) display a severe cranial neural tube closure defect. These embryos have an open neural tube extending from the developing forebrain to the presumptive hindbrain (Fig. 7C-F). Complete closure of the caudal neural tube was observed in all p190 RhoGAP mutant animals, although the morphology of the roof plate was abnormal, suggesting a mild closure defect in the spinal cord as well (data not shown).

Unlike in some mouse mutants displaying defects in neural tube closure, no enhanced apoptosis was observed in the neuroepithelium of p190 RhoGAP mutants at E9.5 (data not shown). However, phalloidin staining of E10.5 wild-type and mutant spinal cord sections revealed a substantially increased accumulation of F-actin along the basal edge of neuroepithelial cells at the presumptive floor plate in the mutants (Fig. 7G,H). In some mutant animals, a reduced apical constriction of neuroepithelial cells in this region is also seen (Fig. 7H). Thus, the neural tube defect in p190 RhoGAP^−/− mice appears to be associated with abnormal cell shape changes and actin assembly within the neuroepithelium. At E8.25, p190 RhoGAP mRNA is expressed specifically in the neural folds (Fig. 7I), consistent with a role for the protein in cells that participate directly in this midline fusion event. At slightly later stages, expression is more widespread, with a higher level of mRNA present in the dorsal neural crest at E8.5 (Fig. 7K) and in the developing eye at E9.0 (Fig. 7J).

The midline fusion defects seen in p190 RhoGAP^−/− animals are strikingly similar to those previously reported in mice lacking the major PKC substrate, MARCKS, including agenesis of the corpus callosum, a defect in fusion of the cerebral forebrain, an open anterior neural tube, eye defects, and occasional omphalocele (Stumpo et al., 1995). Since p190 RhoGAP is prominently phosphorylated on serine residues (Ellis et al., 1990), we investigated whether p190 RhoGAP, like MARCKs, is regulated by PKC-mediated phosphorylation. First, we determined that purified baculovirus-produced p190 RhoGAP is a direct substrate of PKC in vitro (Fig. 8A). In addition, we found that p190 RhoGAP undergoes a significant increase in its overall phosphorylation state following exposure of cells to the PKC activator, 12-O-tetradeconoyl-phorbol-13-acetate (TPA) (Fig. 8B). To assess the physiological relevance of this phosphorylation in vivo, we examined the distribution of p190 RhoGAP in cultured fibroblasts before and after stimulation with TPA. Interestingly, we observed that p190 RhoGAP is rapidly relocalized from the cytoplasm to regions of plasma membrane ruffling following exposure of cells to TPA (Fig. 8C,E). Moreover, p190 RhoGAP is colocalized with polymerized actin in the TPA-induced ruffles (Fig. 8F). Therefore, it appears that PKC can phosphorylate p190 RhoGAP in vivo, and that this phosphorylation, like the phosphorylation of MARCKS by PKC (Thelen et al., 1991), regulates the subcellular localization of p190 RhoGAP.

Since PKC, like p190 RhoGAP, has been implicated in integrin-mediated signaling to the actin cytoskeleton (Defilippi et al., 1997), we next examined the distribution of p190 RhoGAP in cells during spreading on fibronectin. As was seen in TPA-treated cells, in fibroblasts plated on fibronectin, p190 RhoGAP rapidly re-localizes to membrane ruffles, where it co-localizes with polymerized actin (Fig. 8I,J). By contrast, cells spreading on poly-L-lysine exhibit no redistribution of p190 RhoGAP protein (Fig. 8G,H). Thus, integrin-mediated cell spreading induces the same redistribution of p190 RhoGAP as does PKC activation. The translocation of p190 RhoGAP observed in both cases can be completely blocked by pretreatment of cells with PKC inhibitors (data not shown). The observed redistribution of p190 RhoGAP, as seen by immunostaining with a p190 RhoGAP antibody, was confirmed in TPA-treated cells following transfection with a plasmid that expresses an epitope-tagged p190 RhoGAP (Fig. 8K,L). As described earlier, fibroblasts derived from p190 RhoGAP mutant mice appear to spread normally on fibronectin, potentially reflecting redundancy provided by p190-B, which is also a substrate for PKC and relocalizes to membrane ruffles upon PKC activation (data not shown). We also found that in Cos cells expressing a recombinant form of p190 RhoGAP that mimics the truncation produced in the mutant mice, TPA can induce its redistribution to the plasma membrane, suggesting that the response to PKC is maintained in the altered protein (Fig. 8M,N). However, there is a difference in the appearance of immunostained retinal sections from wild-type and mutant p190 RhoGAP mice, suggesting that subcellular distribution may be regulated by additional factors (Fig. 8O,P).

Using a gene targeting strategy, we have investigated the biological function of p190 RhoGAP in mammalian development. Previous studies in cultured cells have revealed that the Rho GTPase plays an important role in regulating cell shape and that p190 RhoGAP, a potent Rho inhibitor, mediates a signal to the actin cytoskeleton in response to integrin activation. Our in vivo studies demonstrate an essential role for p190 RhoGAP in regulating a variety of morphogenetic events in neural development that are believed to require adhesion molecule-dependent modulation of the actin cytoskeleton. While some recent studies have implicated Rho GTPase signaling in tissue morphogenesis in invertebrate organisms, including Drosophila and C. elegans (Settleman, 1999), these are the first genetic studies that establish a critical role for Rho GTPase regulation in mammalian embryonic morphogenesis.

Although we cannot conclude that the targeted p190 RhoGAP allele that was used in these studies is a null allele, the fact that it is defective for all of the measurable in vivo functions of the p190 RhoGAP protein suggests that it is, minimally, a severely defective loss-of-function allele. However, it remains formally possible that a complete deletion of the gene might result in additional or more severe phenotypes. As is the case when interpreting the results of many gene targeting experiments in mice, the analysis of the p190 RhoGAP mutant phenotype is somewhat complicated by the potential redundancy provided by the closely related p190-B molecule, which is also expressed throughout the nervous system (Burbelo et al., 1998; M. B., S. M. and J. S., unpublished observation). This may account for the fact that the p190 RhoGAP mutant phenotype is restricted to a subset of neural tissues, despite widespread expression of the protein throughout the nervous system. In fact, the regional restriction of phenotypes in p190 RhoGAP mutant mice has made it difficult to isolate and compare cell populations from wild-type and mutant animals in order to further explore the cellular nature of the observed tissue defects. Thus far, we have not been able to identify cellular defects in such brain-derived cultures (M. B., S. M., and J. S., unpublished observations). Significantly, p190-B accounts for the majority of total RhoGAP activity in fibroblasts (Vincent and Settleman, 1999), indicating that in some cell types, p190-B may be the major regulator of Rho GTPase activity. Notably, we did not find evidence of compensatory up-regulation of p190-B expression in cells or tissues derived from the p190 RhoGAP mutant mice (M. B. and J. S., unpublished observation).

Consistent with the predominantly nervous system-restricted expression profile of the p190 RhoGAP gene during embryogenesis, we found that mice lacking a functional p190 RhoGAP protein exhibit defects that appear to be largely restricted to the developing nervous system. A detailed histological analysis of the mutant mice revealed that three temporally and spatially distinct midline fusion events in neural development require a functional p190 RhoGAP protein: namely, closure of the cranial neural tube, fusion of the hemispheres at the midline of the forebrain, and closure of the optic fissure. It is possible that there is also a requirement for p190 RhoGAP in adult tissues, where it is widely expressed (Settleman et al., 1992a), but the perinatal death of the mutant mice precludes such an analysis. Presently, the specific cause of death in the p190 RhoGAP mutant animals is unknown.

The presence of multiple midline fusion defects in p190 RhoGAP^−/− animals suggests that a similar cellular mechanism, which requires p190 RhoGAP function, underlies the morphogenetic process that normally leads to these distinct fusion events. Indeed, previous observations support a mechanistic relationship among the various midline fusion events associated with normal mammalian neural development. For example, mice lacking the transcription factor Pax2 exhibit a neural tube defect and a persistent optic fissure (Torres et al., 1996), and mice lacking the PKC substrate, MARCKS, as mentioned earlier, exhibit defects in neural tube closure, fusion of the forebrain hemispheres, and formation of the corpus callosum (Stumpo et al., 1995). In addition, many human patients that exhibit coloboma of the eye also have agenesis of the corpus callosum (Denslow and Robb, 1979), and approximately 5% of patients identified with a midline fusion defect also exhibit at least one additional midline fusion defect (Khoury et al., 1989).

The observation that a small percentage of the p190 RhoGAP^−/− animals (like the MARCKS mutants) exhibit omphalocele, a midline fusion defect of the abdominal wall, also supports the conclusion that the distinct phenotypes observed in these animals are all related to a similar problem with tissue fusion. Clinical studies have revealed an association between the presence of omphalocele and other midline defects, including an open neural tube, coloboma of the eye, and agenesis of the corpus callosum (Calzolari et al., 1997; Martinez-Frias 1995). Thus, it seems likely that many, if not all, midline tissue fusion events in mammalian development are mediated by a common morphogenetic mechanism, and that p190 RhoGAP is an important regulator of such fusion events.

A shared feature of these various morphogenetic processes of neural development is the regulated alteration of cell morphology that is associated with dynamic rearrangements of the cytoskeleton. For example, a role for actin reorganization in neural tube closure is well documented (Schoenwolf and Smith, 1990), and several mouse mutants lacking actin regulatory proteins, such as Mena, Abl, Vinculin and MARCKS, exhibit anterior neural tube closure defects (Koleske et al., 1998; Lanier et al., 1999; Stumpo et al., 1995; Xu et al., 1998). The MARCKS protein, which like p190 RhoGAP is regulated by PKC, associates directly with the actin cytoskeleton, and can regulate membrane ruffling and cell spreading (Myat et al., 1997), suggesting that its role in neural tube closure may also involve the regulation of actin organization. Thus, it is likely that Rho GTPase regulation of the actin cytoskeleton plays a significant role in neural tube closure.

In comparisons of the neural tubes of wild-type and p190 RhoGAP^−/− embryos, we observed a substantial increase in F-actin along the basal edge of neuroepithelial cells in mutant embryos, and defects in apical constrictions in these cells, suggesting abnormal regulation of actin polymerization at the median hinge point. Significantly, the neural tube defect in p190 RhoGAP mutants is largely restricted to the anterior neural tube, and it was recently reported that actin microfilament reorganization in the apical region of neuroepithelial cells is required for closure of the cranial neural tube but not the spinal neural tube (Ybot-Gonzalez and Copp, 1999). Possibly, p190 RhoGAP regulates the distribution of Rho-mediated actin polymerization, thereby affecting neuroepithelial cell shape changes. In light of the observation that PKC activation leads to a redistribution of p190 RhoGAP within cells, it is conceivable that signals from activated adhesion molecules, via PKC, direct p190 RhoGAP to subcellular sites of actin reorganization. Indeed, it has been observed that p190 RhoGAP rapidly accumulates at sites of antibody-mediated cross-linking of integrins in cultured cells (Burbelo et al., 1995). Loss of p190 RhoGAP activity is likely to result in Rho activation, and it has been observed in Drosophila that either a decrease or increase in Rho GTPase signaling is detrimental to several morphogenetic processes in embryonic development (Settleman, 1999), suggesting that disruption of Rho GTPase pathways in mammalian embryos may have similar consequences.

Abnormalities in actin-dependent cell shape changes may also account for the observed defects in ventricle morphogenesis, forebrain fusion, and optic fissure closure in the mutant mice. Indeed, we find that p190 RhoGAP protein is enriched in cells along the forebrain longitudinal fissure, where it co-localizes with F-actin. However, mechanisms less directly associated with cell shape change cannot be ruled out. For example, the optic fissure closure defect in Pax2 mutant mice is associated with a failure to degrade extracellular laminin at the fissure (Torres et al., 1996). Interestingly, p190 RhoGAP is required for local degradation of extracellular matrix in melanoma cells exposed to laminin (Nakahara et al., 1998), suggesting that p190 RhoGAP may affect integrin-mediated tissue morphogenesis by regulating the formation of cellular processes which secrete metalloproteases that remodel the extracellular matrix.

Several cell culture studies point to a role for p190 RhoGAP in integrin-mediated signaling from the extracellular matrix (Burbelo et al., 1995; McGlade et al., 1993; Nakahara et al., 1998; Sharma, 1998). In fact, the defects observed in p190 RhoGAP mutant mice are consistent with a role for p190 RhoGAP in adhesion signaling not only via cell-matrix interactions, but also through cell-cell adhesion. The cell-cell adhesion molecules, N-cadherin and N-CAM (Bronner-Fraser et al., 1992), at least one integrin subunit (Kil et al., 1996), and the laminin α5 chain (Miner et al., 1998) are required for proper neural tube closure. In addition, like p190 RhoGAP, integrin subunits α3 and αv and neural adhesion molecules, L1 and N-CAM, have each been implicated in neuronal migration in the developing mouse brain (Anton et al., 1999; Asou et al., 1992; Tomasiewicz et al., 1993). It is also worth noting that several of the neural adhesion molecules have been directly implicated in the guidance and fasciculation of axons (Walsh and Doherty, 1997), and we have begun to identify similar defects in a subset of axon tracts within the nervous system of p190 RhoGAP mutant mice (to be reported elsewhere). Activation of each of these various types of adhesion molecules is known to affect actin re-organization (Burden-Gulley and Lemmon, 1996), and it would therefore not be surprising to find that their effects on the cytoskeleton are mediated by the Rho GTPase and its regulators. Indeed, integrin-induced cell shape changes in fibroblasts require Rho GTPases (Clark et al., 1998; Hotchin and Hall, 1995).

Significantly, PKC has been implicated in the transduction of signals that affect cell shape both by integrins (Katz and Yamada, 1997) and neural adhesion molecules, including L1, N-CAM and N-cadherin (Bixby and Jhabvala, 1990; Kolkova et al., 2000). In addition, the MARCKS-related protein, MacMARCKS, which is a PKC substrate, is also required for neural tube closure (Chen et al., 1996), and has been found to be required for integrin-dependent spreading of macrophages (Li et al., 1996). In light of our observation that PKC as well as integrin activation induces a similar re-localization of p190 RhoGAP, we propose that in response to engagement of diverse neural adhesion molecules, p190 RhoGAP functions to transduce signals via PKC activation that direct the morphogenesis of neural tissues. Finally, it is worth noting that all of the developmental defects observed in the p190 RhoGAP mutant mice reflect defects that appear to be independent of cell fate determination or cell proliferation. Thus, it appears that the role of the Rho GTPase in mammalian development may be restricted to the regulation of changes in cell shape and cell movements that determine the morphogenesis of embryonic tissues.

We are grateful to Amin Fazeli, Piotr Sicinski, and Lena Du for help in producing knockout cells and mice, and to Andrea McClatchey for critical reading of the manuscript. We also thank Lawrence Cherkas for assistance with photography, Ed Seling for assistance with electron microscopy, and Pradeep Bhide, Philippe Soriano, Marc Tessier-Lavigne, and Ken Kosik for helpful discussions. We thank Sheila Thomas for providing the cortactin antibody and Tony Pawson for providing Rat-2 GAP-N fibroblasts. This work is supported by NIH and ACS awards to J. S.