Vinculin knockout results in heart and brain defects during embryonic development

Vinculin is a cytoskeletal protein of 117 kDa and is a major constituent of cell junctions, both cell-matrix (adhesion plaques) and cell-cell (zonulae adherens). Vinculin occurs in multimolecular complexes that are thought to function in adhesion and/or signaling between the extracellular milieu and the cell, via integrins and cadherins. The vinculin gene occurs as a single copy with only one splicing variant and apparently no close relative to take over functions in its absence. The amino-terminal of vinculin binds to talin (Jones et al., 1989) which, in turn, binds to β-integrins, and the carboxy-terminal binds to actin (Johnson and Craig, 1995b), phospholipids (Johnson and Craig, 1995a) and paxillin (Turner et al., 1990). This arrangement suggests that vinculin acts as a bridge between the extracellular matrix and the actin cytoskeleton.

The inactivation of the vinculin gene (double ‘knockout’) in F9 embryonal carcinoma (EC) and embryo-derived stem (ES) cells (Coll et al., 1995) gave a common mutant phenotype. The shape of the vinculin null F9 cells was more rounded and the cells were less spread, had no or unstable lamellipodia, reduced adhesion and increased locomotion compared to wild-type (WT) cells. Interestingly, vinculin null ES cells were able to differentiate in vitro into a variety of differentiated cell types, including rhythmically beating cardiomyocytes. Both mutant and WT ES cells formed teratocarcinomas in nude mice and these contained a range of cell types that appeared similar in range and frequency. In spite of its ubiquitous expression, vinculin appears to be dispensible for the differentiation of many cell types found in early embryogenesis.

We showed earlier that F9 embryonal carcinoma cells that lack vinculin still recognize and adhere to extracellular proteins that may guide developmental processes. In the absence of vinculin, integrins still form focal adhesion plaques because alternate links exist. For example, α-actinin can bind to β-integrin (Burridge et al., 1988; Otey et al., 1990) and actin can bind to talin (Goldmann et al., 1994; Horwitz et al., 1986), to link the integrins to the cytoskeleton. The resulting focal adhesions of the vinculin null cell are less effective than when vinculin is present, even though some compensation occurs by the increased level of α-actinin, talin and paxillin in the adhesion plaques of vinculin null cells (Volberg et al., 1995). The properties of F9 embryonal carcinoma cells after the abrogation of vinculin in vitro suggests that vinculin plays a regulatory or signaling role rather than merely a structural connecting link (Xu, W., J.-L. Coll and E. D. Adamson, 1997, unpublished data). The question that we sought to answer here was whether the function of vinculin was essential to developmental processes through its cell adhesion and migration activities or if specific functions would be revealed. For example, the loss of vinculin expression in C. elegans leads to the loss of muscular activity and lethality at the larval L1 stage (Barstead and Waterston, 1991). An in vitro study of vinculin function using antisense oligonucleotides to inhibit vinculin transcription and translation, caused loss of nerve growth cone stability and reduced axonal outgrowth (Varnum-Finney and Reichardt, 1994). These data might suggest that muscle and neuronal development could be affected in the total absence of vinculin. The cells in an embryo have an exacting program to fulfil, a process that involves cell migration, cell recognition and adhesive responses. Because vinculin alters the effectiveness of these activities, the developmental process might be aberrant in several respects. Our analyses of vinculin null embryos indicates that the developmental program fails at several points and results in defects that could be ascribed to altered adhesion and migration of cells. In order to explore this, we derived fibroblast cultures from sibling normal and mutant embryos on the 10th day of gestation to determine if these cells also had defective adhesion and locomotion. This was confirmed because greatly reduced adhesion to all substrata was demonstrated for vinculin null fibroblasts, and cells also migrated at elevated rates, thus providing the basis for the appearance of multiply defective embryos that fail to survive beyond the 10th day of gestation.

A phage genomic library (DASHII, kindly provided by T. Doetschman (University of Cincinnati, OH) from the 129/Sv mouse strain was screened using a mouse vinculin cDNA, and an 8 kb EcoRI fragment from a positive phage clone was identified, encompassing intron 2, exon 3, intron 3, exon 4 and intron 4. A 2.3 kb BglII-AvrII fragment derived from intron 2 followed by PGKneobpA was introduced into the vector and a diphtheria toxin-A gene was ligated to 950 bp of exon 4 and partial intron 4 sequences. The vector (Fig. 1A) used for the present study was modified from one described earlier (Coll et al., 1995), by removing putative splicing signals from the insertion site in the mouse vinculin gene to avoid possible resplicing of mRNA (Coll et al., 1997) at the boundaries of the insert. In addition, the insert was PGKneo in reverse transcriptional direction. The PGK promoter is stronger than the earlier pMC1neo and gave more frequent targeting events.

PCR was initially used to screen for homologous recombination with primers P1, P2 and P3. P1 sequence was 5’TCAGACCC-ATACTCGGTTCC3’, P2 was 5’CTGGCCCAAGATTCTTTGTG3’ and P3 was 5’AGACTCGTCAAGAAGGCGATAGAAGGC3’. Following the denaturation step 98°C 2 minutes, 35 cycles of PCR were performed (94°C 45 seconds; 58°C 45 seconds; 72°C 3 minutes) ending with 10 minutes at 72°C incubation. The positive ES clones were further identified by Southern blotting using a probe of 500 bp vinculin sequences that are located outside the targeting vector. The normal allele produces a DNA of 8.0 kb while the targeted DNA is 3.4 kb (see Fig. 1A).

Blastocysts were collected from pregnant female mice, strain C57BL/6. Approximately 8-10 single-targeted ES cells were microinjected into each embryo and the blastocysts were implanted into the uteri of pseudopregnant mothers. 10 male and 3 female chimeras with agouti coat coloring were produced and the males were mated to Black Swiss females. Tail biopsies of offspring were tested for germline transmission by the presence of a targeted vinculin gene by PCR and Southern blotting. Four males proved to be fertile heterozygotes and these were bred to their siblings to give the embryos and offspring described.

Embryos or yolk sac homogenates were incubated overnight at 55°C in a lysis buffer that contained 20 mM Tris (pH 7.5), 50 mM KCl, 0.45% Tween 20, 0.45% NP-40, 0.01% gelatin and 100 µg/ml proteinase K. The samples were heated at 98°C 10 minutes, then PCR was carried out for 30 cycles at 94°C 45 seconds, 58°C 45 seconds, 72°C 3 minutes using all three primers, P1, P2 and P3 together in a single PCR reaction. The primers used and the DNA bands obtained are described above.

E10.5 embryos were fixed in 4% paraformaldehyde-PBS for 6 hours at 4°C, before dehydration through a methanol series. Endogenous peroxidase activity was quenched by incubation in 20:80 vol:vol of 30% hydrogen peroxide:methanol for 5 hours at room temperature. The samples were rehydrated through a methanol series to PBS and incubated in blocking solution (20% normal goat serum [NGS], 0.2% Triton X100/PBS) overnight at 4°C. The embryos were then exposed to a 1:1 dilution of hybridoma supernatant 2H3 reactive to neurofilament protein (Developmental Studies Hybridoma Bank, Johns Hopkins University, Baltimore, MD) or with a 1:50 dilution of mouse mAb to MAP2 (Chemicon, Temecula CA) for 48 hours at 4°C. After extensive washing the embryos were exposed to horseradishperoxidase-conjugated goat anti-mouse IgG (Jackson ImmunoResearch, West Grove, PA) diluted 1:300 with 0.2% NGS, 0.1% NP-40 in PBS for 24 hours at 4°C. The embryos were washed and exposed to peroxidase substrate, hydrogen peroxide, with the chromogen, Vector™SG (Vector Labs. Inc. Burlingame, CA), for 2-5 minutes at room temperature. The reaction was stopped by rinsing in distilled H₂O. The embryos were taken through a glycerol series to 80% glycerol for photography.

Embryos from two litters of heterozygous crosses were used to derive mouse embryo fibroblasts. E10.5 of genotype Vin^+/+ and Vin^−/− were dissociated with sterile needles after removing head and heart, and washed twice in ES medium (DMEM, 10% FBS, 1× non-essential amino acids (Sigma), 1 mM sodium pyruvate, 0.1 mM β-mercaptoethanol). The dissociated tissues were cultured in 24-well tissue culture plates. After 4 days, cells were trypsinized and transferred to 60 mm tissue culture dishes. The genotype of cells was tested by immunoblotting and immunofluorescence with anti-vinculin antibodies (Sigma). MEF became spontaneously immortalized lines after several weeks in culture and were used to characterize cell adhesion and migration, and for other experiments.

Fibroblasts from mouse embryos were cultured for 60 minutes on 60 mm dishes coated with fibronectin (20 µg/ml), rinsed with cold 1 mM Na₃VO₄ (a phosphatase inhibitor) in PBS, and lysed in 500 µl 10 mM Tris-HCl, pH 7.5, 1% SDS and 1 mM Na₃VO₄. The samples were boiled for 5 minutes with 500 µl of 2× immunoprecipitation buffer (20 mM Tris-HCl, pH7.5, 300 mM NaCl, 2 mM EDTA, 2 mM Na₃VO₄, 2% NP-40). The lysates were precleared with 30 µl of 50% protein-A Sepharose CL 4B (Sigma) at 4°C for 1 hour and immunoprecipitated by incubation with 2 µg of anti-pp¹²⁵ FAK, antipaxillin (both from Transduction Labs.) or anti-α-actinin (Sigma), anti-CAS (Vuori and Ruoslahti, 1995) (a kind gift of Dr K. Vuori of this Institute) at 4°C overnight. 25 µl of a 50% suspension of protein-A Sepharose was added followed by incubation at 4°C for 2 hours. The immunoprecipitates were washed 3 times with RIPA buffer, analyzed by 10% SDS-PAGE and membranes were incubated with anti-phosphotyrosine (RC20H, Transduction Labs.).

Cell pellets were lysed in 2× RIPA buffer (20 mM Tris-HCl, pH 7.5, 300 mM NaCl, 2 mM EDTA, 0.2% SDS, 2% NP-40, 0.4 mM PMSF),incubated at 4°C for 2 hours and an equal amount of 2× sample buffer (Laemmli, 1970) was added. Aliquots containing 30 µg protein were analyzed by SDS-PAGE followed by transfer to Immobilon-P membrane (Millipore Corporation, Bedford, MA). Vinculin was detected using anti-human vinculin monoclonal antibody (clone hVin-1, Sigma Chemical Corpn., St. Louis, MO), followed by anti-mouse IgG coupled to an enhanced chemiluminescent probe (ECL, Amersham Corp, Aylesbury, UK).

96-well plates (Linbro/Titertek, Flow Lab. Inc, Virginia) were coated with fibronectin, superfibronectin (Morla et al., 1994), vitronectin, laminin and collagens Type I and IV (Sigma) at various concentrations as indicated, and incubated overnight at 4°C. The wells were rinsed 3 times with PBS and blocked with 100 µl 2% heat-inactivated bovine serum albumin (BSA) for 2 hours at room temperature and rinsed 3 times with PBS. Cells were harvested in trypsin-EDTA as described for the migration assay and 100 µl (5×10⁵ cells /ml) was added to each well in serum-free DMEM medium and allowed to attach for 25 minutes. Nonadherent cells were removed by gently washing 3 times with PBS and attached cells were fixed in 10% formalin at 20°C for 30 minutes and then stained with 1% Toluidine Blue for 60 minutes. The plates were washed extensively with water and air dried. The blue dye was eluted in 100 µl of 2% SDS solution and the absorbance was measured at OD⁶²⁰ on microtiter plate reader. Nonspecific cell adhesion was measured on BSA-coated wells and was subtracted from each reaction value.

Cell migration assays were performed using modified Boyden chambers (6-well cell culture inserts containing polyethylene terephthalate membrane, 8.0 µm pore size and 1.0×10⁵ pore density, Becton Dickinson, Franklin Lakes, NJ). The lower surface of the membrane was coated with fibronectin (20 µg/ml) or other matrix proteins overnight at 4°C. The lower chamber was filled with 1 ml of α-MEM medium with 10% FBS. Cells were harvested with trypsin/EDTA and washed once using serum-free DMEM medium containing 20 µg/ml trypsin inhibitor and resuspended to 2.5×10⁵ cells/ml. 2 ml was added to the upper chamber and the cells were allowed to migrate at 37°C, 5% CO₂ for 6 hours The upper surface of the membrane was wiped with a cotton tip to mechanically remove nonmigratory cells and the migrant cells attached to the lower surface were fixed in 10% formalin at RT for 30 minutes, stained for 20 minutes with 1% crystal violet in 100 mM borate buffer, pH 9.0 and containing 2% ethanol. The numbers of stained cells per field were counted and photographed with Nikon camera attached to the inverted phase-contrast microscope. Background migration was evaluated on BSA-coated membranes and subtracted from all data. Each assay was performed twice in triplicate wells for each cell line.

Cells were grown to confluency in 100 mm tissue culture dishes. Three or four sites in each dish were scraped from the confluent monolayer of cells with a yellow plastic pipet tip to create a cleared line. The medium was removed and replaced with fresh medium. After cells were incubated at 37°C for 6 hours, the progress of cells moving into the wound area was photographed.

Cells were allowed to attach for 2 hours on fibronectin-coated coverslips. Coverslips were fixed for 8 minutes in 4% paraformaldehyde in PBS and permeabilized for 2 minutes in 0.2% Triton X-100 in PBS. After washing 2 times in PBS, the coverslips were incubated at 20°C in 3% BSA and 1% donkey serum. Cells were stained with anti-human vinculin (Sigma), anti-α-actinin, anti-talin (both from Sigma), anti-paxillin (Transduction Labs.) or antiphosphotyrosine (PY20, ICN) for 1 hour. The coverslips were washed 4 times in PBS, 15 minutes each and then stained with fluorescein (DTAF)-conjugated F(ab’)₂ fragment of donkey anti-mouse IgG. The coverslips were washed and were mounted using SlowFade antifade solution (Molecular Probes. Inc., Eugene, OR). The specimens were examined in a Zeiss microscope fitted with epifluorescence, viewed and photographed with a ×100 objective.

We have described previously the targeting of the vinculin gene in ES and in F9 EC cells (Coll et al., 1995). The new targeting vector (Fig. 1A) used for the present study was modified to increase targeting frequency and prevent leaky expression but the procedures used were identical (see Materials and Methods). Southern blot analysis (Fig. 1B) was used to identify six targeted ES [clone R1 (Nagy et al., 1993)] cell lines and four were shown by karyotyping to have 40 chromosomes.

Two of the targeted ES cell lines (Fig. 1B) were used to generate chimeric mice by normal microinjection procedures. Germline transmission resulted from four fertile males out of ten male chimeras that were agouti in coat color. Upon breeding, heterozygous animals were selected by the presence of one targeted vinculin gene as indicated by Southern blots (Fig. 2A). Sibling matings of heterozygotes produced no live born homozygous pups, indicating lethality in mutant embryos. Timed pregnant mothers were dissected on embryonic days from E7.5 to E12.5 to establish the numbers and genotypes of the conceptuses in each litter (Table 1). In all, 332 embryos (including resorptions) and 50 new-born pups were included and all except the resorptions were genotyped. For E7.5 to 12.5 litters, a normal Mendelian ratio was evident for the +/+:+/− groups (23%:45%), but only 16±1.5% homozygous embryos survived until E10.5. The number of resorbed embryos was 19±10% indicating that some homozygous embryos might fall into this group at earlier stages. This possibility needs further confirmation and analysis, because normal WT resorption rates vary from 10 to 20% depending on the strain. Abnormal embryos were detected starting at E8.5, and these were analyzed as described below. The genotype of all offspring was confirmed by PCR (Fig. 2B) and by immunoblotting to demonstrate the lack of vinculin protein in the abnormal embryos (Fig. 2C). Thus, the vinculin genes were correctly targeted and produced no vinculin protein. A similar result is emerging for vinculin^−/− in the 129SvJ strain background, though with smaller numbers of offspring to this date.

Homozygous vinculin null embryos failed to develop beyond the 10th day of gestation and at best were two-thirds of the normal size range. The morphology of whole embryos on E10.5 is shown in Fig. 3. The most prominent feature was the failure of the neural folds to fuse in the head region (upper arrow in Fig. 3C) and failure of head structures to fuse in the ventral cranial midline. The shape of the head was aberrant, and forelimb buds were present but small and asymmetrically placed. The heart structures were small and surrounded by a much dilated pericardial cavity (arrows in Fig. 3B,C).

Whole-mount immunostaining of E10.5 embryos using antibodies to neurofilament protein emphasized the paucity of nerve development. In the trunk and caudal regions where the neural tube was fused, the wavy appearance of the spinal cord nerve tracks in mutant embryos showed that this was abnormal compared to normal WT embryos (Fig. 4A-C). Staining with antibody to MAP-2 showed large nerve bundles in the open cranial neural tube (Fig. 4E) whereas the normal embryo had developed arborized dendrites visible through the ectoderm (Fig. 4D).

At least two litters of each of day E8.5 and E9.5 embryos were analyzed by serial sectioning and H and E staining. Serial sagittal sections of E8.5 Vin^−/− embryos, stained with hematoxylin and eosin, shows that the anterior-posterior axis was smaller than normal (Fig. 5A), the uterine cavity was reduced in size, and the developmental process was greatly retarded. Mesoderm was present in both embryonic (head fold and somites) and extraembryonic (allantois) tissues but mesodermal cells were fewer and were more condensed compared to WT embryos perhaps because of reduced amounts of extracellular matrix material. The amnion and visceral yolk sacs appeared normal, but the foregut was vestigial and heart structures were barely discernible. Clearly gastrulation took place and mesoderm was formed but further development was retarded. The neural folds seen in transverse sections (Fig. 5B) were thin and distorted with a neural groove that was mishapen. The thickness of the neuroepithelial layer was already reduced even at this stage.

E9.5 embryos showed several anomalous features in the (^−/−) genotype (Fig. 5C). They were about two-thirds the size of (+/+) embryos (not to scale in Fig. 5), the brain was especially reduced and ventricles were absent or small, but inspection of numerous sections indicated that forebrain, midbrain and hindbrain structures were present. Somites were small and the heart structure developmentally abnormal or severely retarded. Transverse sections of the head at this stage (Fig. 5D) indicated that the failure of the cephalic neural folds to fuse resulted in an open diocoel. The myelocoel was small and/or distorted and serial sections showed that there was only one optic vesicle in this particular embryo. Mutant embryos were structurally fragile and did not survive fixation and embedding well compared to wild type. Higher magnification of histological sections of the head (Fig. 6A,B) showed that the cells of the neuroepithelium were disarrayed and the layer was thin and distorted. Mitoses were present but reduced in number in mutant E9.5 neuroepithelium, indicating that at least some cells were alive and still proliferating. There were also many degenerating cells and in some places, absence of head ectoderm in vinculin null embryos. However, the softer and more friable mutant tissues tended to exaggerate this defect.

Sections of the heart at E9.5 revealed the likely cause of death by E10, because the mutant structure was about half the size of the normal littermate’s, and was oriented in a downward slant, rather than perpendicular to the spinal cord. The walls were thin with few cardiomyocytes in what should be the dense layer, but there were signs of trabeculae formation (Fig. 6E,F). The endocardium was present, but never formed cushions or valves and the heart was never observed to contract. However, when heart structures were transferred to tissue culture, after several days in medium the cardiomyocytes from vinculin null embryos were observed to contract rhythmically. There were a few blood cells in the atrium of null embryos, indicating that hematopoiesis had started. Other tissues were reduced in vinculin^−/− embryos, especially the density of the somites (Fig. 6C,D). There were few signs of craniofacial, liver or gut development as degeneration leading to death occurred at this stage.

In order to determine if the vinculin knockout cells had altered properties that might explain the developmental problems, we isolated fibroblast cell lines from normal and mutant embryos (MEF) for analysis and comparison. The MEFs were derived from E10.5 embryos as populations of cells. We were able to prepare only one population of normal and one population of mutant MEFs, but these represented a mixture of fibroblasts derived from various regions in the decapitated bodies. MEFs from both genotypes were slow-growing and were established in 2 months and have given consistent results for 4 months in culture. The two genotypes produced cells with typical fibroblastic morphology but mutant MEFs were somewhat less spread with fewer cytoplasmic extensions and were more rounded, so that the cell margins were more visible (see Fig. 8 upper panel).

An important question was whether the MEFs would display a mutant phenotype such as we have described for the very different F9 embryonal carcinoma (EC) cell. Fibroblasts have much higher adhesivity to substrata, elaborate extensive actin cytoskeletons with prominent vinculin-containing 7 is an immunofluorescent micrograph showing that the intensities and locations of α-actinin (C,D), actin (E,F), paxillin (G,H), talin (I,J) and total phosphotyrosine proteins (K,L) are similar in both mutant and WT cells. Vinculin is adhesion plaques, much faster migration rates and much slower growth rates compared to F9 EC cells. We first examined their ability to express some focal adhesion proteins correctly and to confirm that vinculin was not expressed. Fig. absent in null MEFs (Fig. 7B) but present in WT cells (Fig. 7A).

An important property of embryo cells is that they should be able to migrate into their correct embryonic field. We therefore measured the migratory rates of the two MEF populations on fibronectin and on other substrata. In one assay, we tested the ability to close a ‘wound’ in a monolayer of cells growing on plastic (Fig. 8). The mutant MEFs migrated faster than WT cells and after 6 hours of wound healing, MEF vin^−/− cells had moved to fill 50% of the gap, while only 10% of the area was filled by normal MEFs. A second kind of assay (Boyden chamber) was used to measure the number of cells that crossed a membrane to reach the other side where they were stained and counted. These assays were conducted in the haptotactic mode, that is, substrata were coated on the membrane undersurface and cells migrated through the pores towards the matrix component. The vinculin^−/− MEFs moved at about twice the rate of WT cells (Fig. 9). except on ‘superfibronectin’ (Morla et al., 1994; Pasqualini et al., 1996) which consists of a mixture of FN polymerized with the extracellular III-1-C domain of FN, on which the difference in rates between wild type and mutant MEFs was reduced.

Another significant requirement of embryo cells during development is to be able to find and adhere to the appropriate location. MEFs were therefore tested in cell attachment assays on purified matrix materials used to coat wells at a range of concentrations (see Materials and Methods). Mutant MEFs had attachment levels 50% lower than WT cells for attachment to fibronectin (FN), superfibronectin (sFN), vitronectin (VN) and laminin (LN). A similar difference was observed between F9 mutant and WT cells (Coll et al., 1995). In Fig. 10, the WT MEFs showed consistently higher adhesion to each substratum compared to mutant MEFs. For Type I collagen, there was little or no adhesion of mutant MEFs even at high collagen concentrations. These results demonstrate that the mutant MEFs retain a phenotype similar to that which we first described in mutant F9 cells and implies that, in general, cells that lack vinculin, whether normal or transformed, may have a similar altered behavior.

Adhesion plaques contain several kinases, including FAK and c-Src, that may be involved with altered cell behavior. A marked difference in the levels of proteins that were tyrosine phosphorylated was detected in the two populations of embryonic cells 1 hour after seeding cells in fibronectin-coated dishes. Immunoblotting of MEF lysates showed that the constitutive level of tyrosine phosphorylation of many proteins of mutant cells was significantly higher than WT cells (Fig. 11A, lanes 1, 2). Some of the proteins that were overphosphorylated were identified by immunoprecipitation with specific antibodies followed by blotting with anti-phosphotyrosine. There were higher levels of phosphorylated paxillin (2.8-fold), FAK (3.5-fold), pCAS¹³⁰ (1.3-fold) and α-actinin (1.3-fold) in mutant MEFs (Fig. 11C). It is possible, therefore, that FAK (autophosphorylation accounts for its increased level of YP staining) is responsible for some of the increase in tyrosine phosphorylation, particularly of paxillin and pCAS¹³⁰, known substrates of FAK elicited by the stimulus of cell adhesion. We tested if src kinase was elevated in vinculin^−/− MEF lysates, by taking aliquots that contained equal amounts of protein and immunoprecipitating with anti-src antibodies. The immune complexes were incubated with γ[³²P]ATP and enolase, a substrate of c-Src and other TPKs. The WT and mutant MEFs had slightly reduced Src-kinase activities as revealed by SDS-PAGE and autoradiography of the phosphorylated enolase product (Fig. 11C, bottom line). Thus, vinculin null cells are specifically enriched with activated FAK, compared to WT cells. Although the levels of the phosphorylated versions of some of the focal adhesion proteins was elevated, immunoblotting showed that the total amounts of paxillin FAK, α-actinin, talin and actin were similar in normal and mutant MEFs (Fig. 11B). We conclude from this result that tyrosine phosphorylation may be important in the functional differences between normal and mutant MEFs.

We have shown that F9 vin^−/− cells have a similar mutant phenotype (Coll et al., 1995) and we have also transfected these cells with a vector that expresses vinculin constitutively and restored the mutant phenotype to normal (W. Xu, J.-L. Coll, and E. D. Adamson, unpublished data) and spreading, adhesion and migration are all restored to wild-type levels. Thus the altered cellular phenotype in MEFs is likely to be caused directly by the abrogation of vinculin expression, although this remains to be formally demonstrated.

Of the two major defects in the embryonic development of vinculin null mice, the most noticeable was the failure of the neural tube to fuse in the midline, with subsequent neural tube defects (NTD) at 100% frequency. The cranial neural tubes of day E9.5 and E10.5 embryos failed to develop beyond the bifurcated lateral outward-curving neural plate stage and the ectoderm failed to close around it. This process of bending of the neural folds towards the midline is mediated by a pair of dorsolateral hingepoints where the cell changes to a wedgeshape in a process regulated by the actin cytoskeleton, and therefore susceptible to the loss of actin cytoskeletoninteracting proteins such as vinculin. A similar NTD occurs in embryos lacking the MacMARCKs gene product, a protein that also regulates the actin cytoskeleton (Chen et al., 1996). In the absence of vinculin, epithelial cell-cell adhesion is reduced (Tozeren et al., 1997) and this possibly exacerbates the inability of the cells to curve and move together as cohesive concave plate.

In vinculin null embryos, cranial nerves failed to develop, observed using whole-mount immunostaining with antibodies to neurofilament protein and microtubule protein (Figs 3, 4). Even though the spinal cord was usually able to fuse, dorsal root ganglia and nerve tracks were thin and barely visible. The reduction in nervous tissue development in vinculin null embryos, could be explained by the failure of mutant axons to extend properly, as demonstrated by reduced axonal growth in vinculin-reduced neurons (Varnum-Finney and Reichardt, 1994). By extrapolation from MEF defects, the altered motility and adhesivity of mutant neural tube cells could explain the inability to effect closure.

The second and probably the lethal defect in vinculin null embryos is the malformation of the heart. Hypodynamic hearts becomes lethal at this stage (E10) of embryonic development, because the initiation and organization of tissues into organs now rapidly occurring, requires efficient vascular delivery of have properties that focus on cell interaction have given broadly similar results. The abrogation of the plakoglobin gene (γ-catenin) involved in desmosomal plaques also gave severe heart dysfunction with death at E10-16 due to ruptured ventricles (Bierkamp et al., 1996; Ruiz et al., 1996). Knockout of the fibronectin gene causes embryonic lethality but produces embryos that express several markers of mesoderm and somitogenesis. Meanwhile, morphogenesis becomes increasingly aberrant leading to death from the 10th day of gestation onwards (George et al., 1993; Georges-Labouesse et al., 1996). FN null animals had some features similar to the vinculin knockout animal, but they did not have open neural tubes and the severely defective brain development described here. In addition, the focal adhesion kinase, FAK(−^/−) mouse had some features in common except that lethality was earlier, on day E8.5 (Furuta et al., 1995). In these animals, mesoderm cells were produced, but gastrulation was abnormal and MEFs cells had reduced migration. Embryos had aberrant morphogenesis and essential organs and tissues such as the heart, somites and blood vessels did not develop beyond a primitive stage. The most similar knockout phenotype with nutrients and removal of waste products. There are two major sources of cells that form the cardiac field. The major source is from the lateral margins of the primitive streak (PS) on day E7, and later the more medial mesoderm contributes. In vinculin null animals, the population of cardiomyocytes is so severely reduced (Fig. 6F) that it is clearly insufficient to produce an aggregation of myocardiocytes that can beat rhythmically. In addition, a second source of cells from the cardiac neural crest was likely missing, reviewed by Olson (1997). Therefore, two sources of cells are reduced and likely contribute to lethality in vinculin KO embryos.

In addition to the structural incapacity due to reduction of the number of cells reaching the field, we hypothesize that vinculin and metavinculin play functional roles in cardiogenesis and in contraction mechanics. Cardiomyocytes must act in synchrony as a multicellular organ and reduced adhesivity at intercalated discs would reduce contractile strength. We regard cardiac function as a major specific role of vinculin and/or metavinculin, and this view is supported by a case of dilated cardiomyopathy recently described that was caused by lack of metavinculin, the isoform that is expressed only in muscle. The subject presenting with heart failure was shown, by gene sequencing, to have a genetic inability to splice the extra exon that forms metavinculin into the normal vinculin product, indicating that there is a specific role for metavinculin in heart function (Maeda et al., 1997). Our mutant embryos lacked both vinculin and its isoform metavinculin.

The inactivation of the genes whose products are concentrated in focal adhesion and zonulae adherens plaques (as in the case of vinculin) or that respect to heart development and lethality was that of N-cadherin (Radice et al., 1997). This protein plays an essential role in cell-cell adhesion in the heart but not in the neural tube. Accordingly, the null animals did not exhibit lack of neural tube closure. The broad similarities between these knockout animals suggest that these genes have activities in the same signaling program.

FN and other matrix factors outside the cell connect with vinculin via the integrins and talin. The stimulus of cell adhesion activates the YP-kinase activity of FAK in the focal adhesion plaque, which then phosphorylates paxillin, tensin and β-integrins (Schaller and Parsons, 1995). How signals are generated from FAK is not yet understood, but can affect cell shape, motility and adhesion as well as nuclear activities. One seeming disparity in these knockout mice is the specific effect of loss of each of these components as measured by MEF behavior. When FN gene activity is abrogated in MEFs, there is little effect on cell adhesion and migration (Georges-Labouesse et al., 1996). The abrogation of FAK activity reduces MEF migration and increases adhesion and the number of focal adhesion plaques in MEFs. In contrast, the loss of vinculin leads to reduced adhesion and increased migration, the opposite result. Therefore, the direction of the abnormalities in MEF adhesion or migration is irrelevant, but rather, any abnormality produces similar inadequacies in function during the developmental process. However, the vinculin null MEF also has increased FAK activity (Fig. 11), along with increased migratory activity (Figs 8, 9) and reduced adhesion (Fig. 10), a concordance that indicates a role for FAK in these cellular activities.

Vinculin^−/− MEFs expressed higher FA kinase-specific activity compared to WT MEFs (Fig. 11). Vinculin is not a direct substrate for FAK kinase (Turner, 1994), but vinculin and FAK both bind to paxillin (Brown et al., 1996) and paxillin is a substrate. We found that in vinculin null embryos, FAK activity and the total level of tyrosine phosphorylated proteins are elevated (about 2-to 3-fold higher) compared to WT MEFs, correlating with unstable focal adhesion complexes (Ilic et al., 1995). Increased FAK activity could be a response to the loss of vinculin and may be responsible for increased cell motility of vinculin null cells. This idea is supported by finding that greatly reduced motility results when FAK activity is abrogated (Ilic et al., 1995) and increased motility when FAK is overexpressed (Matsumoto et al., 1994). However, the loss of FAK in embryo knockout cells did not alter the expression or the tyrosine phosphorylation of most cytoskeletal proteins including vinculin and paxillin; therefore, interactions with other kinases can compensate for the absence of FAK. A role for FAK in the increased locomotion observed in vinculin null cells requires further study, and the cells derived from knockout embryos will be a useful tool.

In summary, our analyses indicate that abnormalities in the heart are the likely major lethal defects in development occurring in the absence of vinculin. It is still open to question whether vinculin plays a specific role in cardiac development or in function, and we are pursuing this line of enquiry.

We thank members of the laboratory, A. Ben-Ze’ev and R. G. Oshima for advice and many helpful discussions. J.-L. Coll initiated the study that is described here and established the F9 vinculin null cell lines. We thank Ms M. Andahazy and Ms J. Avis for excellent embryonal stem cell and transgenic services. We are grateful for skilled histopathology and photography services at the Burnham Institute. Financial support from the NIH is gratefully acknowledged, CA 28427 from the NCI (E. D. A.) and AR41816 from the NIAMSD (H. B.).