Absence of Basement Membranes after Targeting the LAMC1 Gene Results in Embryonic Lethality Due to Failure of Endoderm Differentiation

A Lambda FIX^®II genomic library (Stratagene) of the 129SVJ mouse line was screened using a PCR product comprising 372 bp of the 5′ untranslated region and the first 128 bases of exon one of the LAMC1 gene (Ogawa et al., 1988). Six different clones representing this area were isolated and mapped. Targeting construct 1 was formed by cloning the 6-kb HindIII fragment containing the first exon together with 2-kb upstream sequence and 3.5 kb of intron 1 into pUC 19. The IRES β-Geo cassette (Friedrich and Soriano, 1991; Mountford et al., 1994), which had a NotI linker added to its 3′ end was inserted into the unique NotI site in the first exon (see Fig. 1). The use of the cap-independent translation initiation sequence removed the need to place a Neo resistance cassette in frame to obtain expression by the LAMC1 promoter.

Construct 2 was an EcoRI/SacI fragment of LAMC1 cloned into KSII Bluescript (Stratagene). The sequence was interrupted at the NotI site by the insertion of the phosphoglycerate kinase (pgk) promoter (Soriano et al., 1991) 5′ to a hygromycin resistance cassette with poly A tail (Hygro). This divided the LAMC1 fragment into two arms with 6-kb homology in the 5′ arm and 2.5-kb homology in the 3′ arm (see Fig. 2).

R1 mouse embryonic stem (ES)¹ cells were grown in standard ES conditions with DME supplemented with 15% (vol/vol) fetal bovine serum (FBS), 0.1 mM β mercaptoethanol, and 1,000 U/ml of LIF (ESGRO; GIBCO BRL). 5 × 10⁶ cells were transfected by electroporation with 25 μg of linearized construct 1 and colonies selected for resistance to G418 at 380 μg/ml in the culture medium. Surviving clones were picked, expanded, and then DNA extracted for Southern blotting. DNA from the cells digested with SacI was probed with an external 3′ probe and an internal 5′ probe (see Fig. 1). In cases of correct integration, the wild-type 11-kb fragment was reduced to 7 kb when probed with the external probe and a 7.5-kb band was seen with the internal probe (see Fig. 1).

Attempts using increased G418 concentrations up to 1.5 mg/ml failed to produce ES cells in which both the LAMC1 alleles had been targeted (Mortensen et al., 1992). Therefore, the second targeting construct was used for disruption of the second LAMC1 allele in ES cells previously targeted with construct 1 (see Fig. 2 A). After correct targeting, Southern blot hybridization of SacI genomic DNA digests with probe 1 resulted in the wild-type band being lost, whereas a 5.5-kb band appeared (see Fig. 2 B). Clones so targeted were checked for the absence of expression of the LAMC1 gene by Northern hybridization of mRNA with a probe of LAMC1 cDNA.

Two independent ES cells lines were used to generate germ line chimeras. Blastocysts were isolated from C57Bl/6 mice 3.5 d post coitum (pc) (plug date = 0.5 d pc) and were injected with five to seven +/− ES cells. Blastocysts were then transferred into the uteri of pseudopregnant CD1 foster mothers. Chimeric male progeny were mated to C57Bl/6 females and offspring were tested for germline transmission by Southern blots of DNA extracted from tail biopsies. Heterozygous animals were mated together to obtain homozygous embryos.

The rabbit polyclonal primary antibodies used were: anti–laminin α1 raised against recombinant domain IVa (Schulze et al., 1996); anti–laminin γ1 raised against recombinant domain III LE3-5 (Mayer et al., 1993); anti-EHS laminin which recognizes all three subunits of laminin (Kücherer-Ehret et al., 1990); anti-nidogen raised against recombinant nidogen (Fox et al., 1991); and anti-perlecan raised against recombinant domain III3 (Schulze et al., 1995). Rabbit polyclonal antibodies against von Willebrand factor, the 200-kD neurofilament subunit, and skeletal myosin were obtained from Sigma.

Embryos were washed in phosphate buffered saline (PBS) before embedding and freezing in Tissue-Tek (Sakura Finetek Europe). Cryostat sections were fixed with 0.5% (wt/vol) paraformaldehyde in PBS for 10 min, washed with PBS, and then blocked with 5% (vol/vol) goat serum in PBS/0.1% (vol/vol) Tween 20. The primary antibodies (see below) and goat–anti rabbit Cy3 conjugate secondary antibodies (Jackson Immunodiagnostics) were used in the same solution before washing the sections in PBS and mounting in fluorescent mounting medium (Dako).

For visualization of fragmented nuclear DNA in situ, serial cryosections were fixed for 20 min in 4% (wt/vol) paraformaldehyde in PBS before staining by a modification of the terminal dUTP–biotin nick end labeling (TUNEL) method (Gavrieli et al., 1992). A TACS apoptosis detection kit™ (Trevigen) was used according to the manufacturer's instructions, fragmented DNA being end labeled with biotinylated nucleotides using the Klenow fragment, followed by detection with streptavidin-horseradish peroxidase conjugates. The sections were then counterstained with eosin.

Embryos were isolated from heterozygous matings by flushing the uterus with M2 medium and the blastocysts were cultured in M16 medium at 37°C in 5% CO₂ until they had fully expanded or hatched. Where present, the zona pellucida was removed from the expanded blastocysts by a short incubation in acid tyrode solution and the blastocysts washed in PBS before fixation in 1% (wt/vol) paraformaldehyde for 10 min at room temperature. The blastocysts were permeabilized in PBS/0.02% (vol/vol) Triton X-100 containing 2% (wt/vol) bovine serum albumin for 30 min before incubation with antibodies and subsequent fluorescence microscopy.

Undifferentiated ES cells were trypsinized, triturated, and then resuspended in DME 10% FBS at a dilution of 1,000 cells/μl. The cells were then placed in hanging drops of 20 μl on the lower surface of the lids of plastic Petri dishes containing PBS (Wobus et al., 1991). After 24 h of culture as hanging drops, the cell aggregates were plated into plastic Petri dishes and the embryoid bodies were fixed with 4% (wt/vol) paraformaldehyde after varying culture periods before sectioning and immunostaining as described above. Frozen sections were also stained for lacZ expression as previously described (Beddington and Lawson, 1990).

To monitor cell phenotypes of differentiating ES cells after formation in hanging drops, the embryoid bodies were allowed to attach to tissue culture plastic and cultured in the above medium for 21 d. Preliminary experiments showed that under these conditions small numbers of myotubes differentiated (Kuang et al., 1998). The cultures were then fixed in 4% (wt/vol) paraformaldehyde and immunostained as above.

Wild-type ES cells, and those heterozygous and homozygous for mutations in the LAMC1 alleles, were preplated in tissue culture dishes for 10 min to deplete them of the embryonic fibroblast feeder cells. The nonadherent ES cells were isolated and cultured on a gelatin-coated plate in ES media with LIF for two or three days until almost confluent. The cells were lysed and RNA extracted with guanidinium isothiocyanate (Chomczynski and Sacchi, 1987). 10 μg of total RNA was separated on a denaturing formaldehyde gel of 1% agarose and transferred by vacuum blotting to a nylon membrane (Hybond N; Amersham). After UV cross-linking of the RNA to the membrane, it was prehybridized with a 50% formamide containing buffer and hybridized against cDNA probes for laminin α1, β1, γ1, γ2, and glyceraldehyde-3-phosphate dehydrogenase (GADPH) mRNAs. After high stringency washing, the blots were exposed to autoradiographic film. The probe for the laminin γ1 mRNA was a BamHI-EcoRI fragment between bases 2,959 and 4,163 in the protein coding area (Sasaki and Yamada, 1987), whereas γ2 was a BamHI fragment spanning nucleotides 1,509–2,120 of the protein coding region (Sugiyama et al., 1995). The probe for the α1 chain was a PCR-generated fragment using GCGCATCAGAACACTCAACG (sense) and CAAGGGTGGTCATCATAAGG (antisense) primers amplifying between bases 708 and 1,203 of the protein coding region (Sasaki et al., 1988). The probe for the β1 chain was a PCR product using primers GATAACTGTCAGCACAACACC (sense) and GTGAAGTAGTAACCGGACTCC (antisense) giving a probe between bases 1,231 and 1,794 of the protein coding region (Sasaki et al., 1987).

Embryoid bodies were produced from 5 × 10³ ES cells incubated in hanging drops for 2 d (Wobus et al., 1991). About 10 embryoid bodies were then cultured in 500 μl DME without methionine, supplemented with 1% FBS and 200 μCi/ml [³⁵S]methionine (specific activity >1,000 Ci/mmol; Amersham). Labeling was carried out overnight. The medium was then collected, centrifuged, and then supernatants were stored at −70°C. Cells were washed with complete DME supplemented FBS and lysed in 50 mM Tris HCl, pH 7.5, containing 1% (vol/vol) Triton X-100, 10 mM EDTA, 0.10 M NaCl, and protease inhibitors. After trituration, insoluble material was removed by centrifugation and the supernatants were stored at −70°C.

Immunoprecipitations were carried out in the extraction buffer using Pansorbin (Calbiochem-Novabiochem) to precipitate the antibody–antigen complexes. Proteins bound to the Pansorbin were removed using boiling SDS gel electrophoresis sample buffer. Proteins were fractionated by SDS-PAGE on 5% gels under reducing conditions or on 3 and 5% gels under nonreducing conditions before fluorography.

For immunoblot analysis, embryoid bodies were cultured as above but using complete DME supplemented with 10% FBS. Cell extracts and media were concentrated by immunoprecipitation with antilaminin antiserum before SDS-PAGE under reducing conditions. After electroblotting onto nitrocellulose, the membrane was incubated with the primary antibodies which were detected using goat anti–rabbit antibodies conjugated with horseradish peroxidase (Dako). The enzymatic activity was visualized using 4-chloro-1-naphthol.

A diagram of the initial targeting of the LAMC1 gene is shown in Fig. 1 A. Of the 50 G418-resistant ES cell clones expanded and analyzed on Southern blots by hybridization with probe 1, 17 had undergone recombination at the LAMC1 gene, as shown by the appearance of a band of the expected size of 7 kb and of equal intensity to the remaining wild-type 1-kb band (Fig. 1 B). A single insertion was demonstrated by the internal probe 2 which hybridized to the expected 7.5-kb band (Fig. 1 C). Although expression of the neomycin resistance–lacZ fusion cassette was dependent on the laminin promoter (Ogawa et al., 1988) present in the targeting vector, the high frequency of homologous recombination (>30% of all G418-resistant clones) suggests that the regulatory elements needed for full expression of the LAMC1 gene in ES cells were lacking in the construct. This agrees with the recent demonstration of a strong enhancer element in the first intron of LAMC1, which was absent from our targeting construct (Chang et al., 1996). As expected, the frequency of the second targeting event with the hygromycin resistance cassette (Fig. 2 A) was much lower than that of the first: out of 200 clones, 12 had undergone the second homologous recombination shown by replacement of the wild-type LAMC1 band by a 5-kb fragment (Fig. 2 B).

Northern blotting showed the absence of laminin γ1 subunit mRNA in −/− ES cells, and the amount of LAMC1 mRNA was reduced in +/− cells relative to the wild type (Fig. 2 C). However, levels of the mRNAs coding for laminin α1 and β1 subunits were the same in undifferentiated +/+, +/−, and −/− cells (Fig. 2 C). The expression of LAMC2 remained below the level of detection in all cases (data not shown).

LAMC1 +/− animals were phenotypically normal and have been intercrossed for at least seven generations and have also been bred into a C57Bl/6 background. More than 200 heterozygous matings produced no progeny homozygous for the mutation that were either born or found in utero after day 8.5 pc, indicating early embryonic lethality (Fig. 1 D). Day 3.5 pc preimplantation blastocysts from these matings were immunostained with polyclonal antibodies specific for the laminin γ1 subunit (Fig. 3 A) and also with antibodies against laminin-1 which recognize the α1 and β1 subunits as well as γ1 (Kücherer-Ehret et al., 1990). The majority of blastocysts showed the reported pattern of immunostaining for laminin-1 (Thorsteinsdottir, 1992): the trophectodermal basement membrane was stained together with apparently intracellular staining of cells throughout the blastocyst (Fig. 3 D). However, about one-quarter of the expanded and hatched embryos showed no immunostaining for the laminin γ1 subunit (Fig. 3, B and C) and displayed only intracellular staining using the laminin-1 antibodies (Fig. 3 D).

The embryos found in decidua at day 4.5 pc appeared histologically normal (Fig. 4, A and F). However, eight out of the 40 embryos sectioned and stained did not display laminin γ1 subunit immunoreactivity, although as expected there was immunoreactivity in the stroma and epithelial lining of maternal decidua (Fig. 4 J), which also stained with the other antibodies used (Fig. 4, G–I). Embryos negative for γ1 immunoreactivity demonstrated the absence of any extracellular laminin when stained with laminin-1 antibodies, although there was an accumulation of cells with intense laminin immunoreactivity in the inner cell mass (Fig. 4 G, inset). Very limited patchy deposits of nidogen and collagen type IV were seen under the trophectodermal epithelium but there was no continuous sheet characteristic of the trophectodermal basement membrane (Fig. 4, H and I).

Of the 80 decidua examined at 5.5 d pc from heterozygous matings, 28 contained laminin γ1–negative accumulations of cells, none of which conformed to any recognizable embryonic structures (Fig. 5, B and C), whereas out of the 40 decidua examined from heterozygous/wild-type matings, only five contained no recognizable embryo. Detection of fragmented nuclear DNA by TUNEL staining of serial sections revealed very few stained cells in any embryos before 5.5 d pc (data not shown). However, at 5.5 d pc we detected low numbers of stained nuclei in the γ1-positive embryos (Fig. 5 D), and the laminin γ1–negative embryos displayed either increased numbers of TUNEL-positive cells (Fig. 5 E) or very intense labeling, indicative of extensive DNA fragmentation (Fig. 5 F).

After 48 h of suspension culture, the differentiating +/− and −/− ES cells at the periphery of developing embryoid bodies began to display intense lacZ staining (Fig. 6), indicating the expression of the LAMC1 gene. This reflects the differentiation of these cells into primitive endoderm-like cells with high levels of laminin expression (Doetschman et al., 1985). The deposition of a continuous basement membrane-like sheet of laminin, nidogen, perlecan, and collagen type IV immunoreactivity was observed towards the periphery of +/− embryoid bodies after 7 d of culture (Fig. 7, A, C, E, and G). Outside this basement membrane, there was a sheet some one to three cells thick displaying laminin immunoreactivity (Fig. 7 A). No differences in these were observed between +/+ and +/− embryoid bodies (data not shown). In contrast, in addition to the expected lack of γ1 immunoreactivity in the −/− embryoid bodies (data not shown), no deposition of basement membranes was detected when they were stained with any of the above antibodies: although there were patchy extracellular deposits of perlecan and collagen type IV, these molecules were not deposited in a continuous basement membrane-like sheet (Fig. 7, F and H). Although there was no obvious extracellular laminin staining, polyclonal antibodies to laminin-1 showed immunoreactivity to be accumulated mainly in the cells at the surface of the −/− embryoid bodies (Fig. 7 B). Staining with antibodies specific to α1 and β1 showed the same patterns of distribution as that detected with the polyclonal antibodies to laminin 1 (data not shown). Little if any intracellular or extracellular nidogen immunoreactivity was detectable in the −/− embryoid bodies (Fig. 7 D) although it was present as expected in the basement membranes of +/+ and +/− embryoid bodies (Fig. 7 C).

To determine if the differentiation of ES cells was affected by the absence of the γ1 subunit, embryoid bodies were allowed to attach to tissue culture plastic and cultured for up to 3 wk. At this time, the LAMC1 +/− cells displayed basement membrane-like sheets of laminin immunoreactivity (Fig. 8 A), small numbers of isolated von Willebrand–positive cells were seen (Fig. 8 C) and neurofilament-positive cell bodies and neurites were identified (Fig. 8 E). Furthermore, the cultures contained low numbers of myotubes that stained with antibodies to skeletal myosin (Fig. 8 G). As expected from Fig. 7, the LAMC1 −/− cells did not deposit basement membranes although individual permeabilized cells displayed intense laminin immunoreactivity (Fig. 8 B). Although no differences from controls were seen in the von Willebrand– and neurofilament-positive cells (Fig. 8, D and F, respectively), no normal myotubes were observed, myosin-positive cells being present in large aggregates which extended thin processes (Fig. 8 H).

After overnight [³⁵S]methionine metabolic labeling, a band of ∼800 kD on nonreducing SDS-PAGE was immunoprecipitated with laminin-1 antibodies from both culture medium and extracts of the +/+ and +/− embryoid bodies (Fig. 9 A, lanes 1–4). This corresponds to the expected size of laminin type 1. In contrast, neither the cell extracts nor the culture medium from the −/− embryoid bodies contained detectable laminin of 800 kD, although lower molecular weight bands were immunoprecipitated (Fig. 9 A, lanes 5 and 6).

To better characterize these bands, electrophoresis was performed on higher percentage SDS-polyacrylamide gels under nonreducing and reducing conditions. The pattern of reduced protein bands immunoprecipitated from the media of the −/− embryoid bodies differed from that of the −/− embryoid body cell extracts (Fig. 9 B, lanes 7 and 8), whereas the patterns were the same from the +/− embryoid body cell extracts and media (Fig. 9 B, lanes 5 and 6). These differences are consistent with the observation that most of the laminin immunoreactivity of +/+ and +/− embryoid bodies was seen in a basement membrane and hence extracellular (Fig. 6 A), whereas that seen in the −/− embryoid bodies appeared to be intracellular (Fig. 6 B). Although a nidogen band of 150 kD was present in the medium of these cells (Fig. 9 B, lanes 4 and 8; Fig. 9 C, lane 3), there was little or no nidogen detectable in the −/− embryoid body extracts under both nonreducing (Fig. 9 B, lane 3) and reducing conditions (Fig. 9 B, lane 7). Thus, the lack of nidogen immunoreactivity in −/− embryoid bodies (Fig. 9 D) is not due to lack of nidogen synthesis, but rather results from nidogen being lost from the embryoid bodies into the medium (Fig. 9 B, lanes 4 and 8).

The nonreduced extracts of +/− and −/− embryoid bodies were similar in that they contained a prominent band of ∼200 kD (Fig. 9 B, lanes 1 and 3). However, a novel strong band of ∼300 kD under both non–reducing and reducing conditions was immunoprecipitated from −/− embryoid body medium (Fig. 9 B, lanes 4 and 8), indicating that this secreted protein was not disulfide-bonded to other laminin subunits. To identify the novel band, immunoprecipitation and immunoblotting experiments were performed with laminin α1 subunit-specific antibodies directed against the IVa domain. In addition to the 400-kD α1 chain, an equally strong band corresponding to the 300-kD protein was also found in −/− cell extracts (Fig. 9 D, lane 2), and this band alone was detected in the −/− medium (Fig. 9 D, lane 3) although it was absent from the +/− medium (Fig. 9 D, lane 1). Immunoprecipitation of this band from the −/− cell medium with antibodies to the laminin α1 subunit also coprecipitated a band of 200 kD but no nidogen was detected (Fig. 9 C, lane 4). Conversely, immunoprecipitation of nidogen from −/− cell medium with antibodies against nidogen failed to precipitate any laminin subunits (Fig. 9 C, lane 3). Thus although the −/− ES cells secrete a modified laminin α1 subunit and nidogen, they are not associated as normal (Fig. 9 C, lanes 1 and 2), consistent with the absence of the laminin γ1 subunit in the −/− ES cells.

We have used homologous recombination to target one or both of the LAMC1 alleles coding for the laminin γ1 subunit in mouse embryonic stem cells. By so doing, we have disrupted the formation of all described laminin isoforms with the exception of laminin 5. Although the null mutation resulted in the absence of basement membranes and hence was an early embryonic lethal, surprisingly, preimplantation development appeared to be normal in that a pumping trophectodermal epithelium allowed expansion of the blastocysts. Basement membranes were first found to be necessary for differentiation of primitive endodermal cells, in their absence Reichert's membrane failing to form. In vitro analysis of the null mutation showed that the laminin γ1 subunit was necessary for the differentiation of stable myotubes and for assembly of other covalently bonded laminin subunits. In turn, the lack of intact laminin deposition was necessary for the assimilation of other components into a continuous basement membrane in vitro, consistent with the disruption of formation of the first basement membrane to be formed during development in vivo.

Although mice heterozygous for the LAMC1 gene are healthy and fertile, the analysis of different gestational ages showed that LAMC1-null mutant embryos did not survive later than day 5.5 pc. Immunostaining of pre- and postimplantation embryos up to 4.5 d pc showed laminin γ1–negative embryos to have an apparently normal morphology although they lacked basement membranes as defined by staining for other basement membrane components. Disruptions of cell–extracellular matrix interactions in the developing lung, kidney, and salivary gland have been shown to inhibit epithelial morphogenesis (Ekblom et al., 1994; Kadoya et al., 1997; Klein et al., 1988). However, the fact that the null mutant embryos described here could develop a functional pumping trophectodermal epithelium means that a basement membrane is not an absolute requirement for the differentiation of epithelia. Although this result points to the primary importance of cell–cell interactions in epithelial development (Watson et al., 1990), it does not rule out a role for basement membranes in the maintenance of specific epithelia or the differentiation of other epithelial cell properties. Indeed, although a decidual reaction occurred in the uterine wall adjacent to −/− embryos, it may be that in the absence of the trophectodermal basement membrane that the trophoblast was unable to successfully implant into the uterus.

Although both Reichert's membrane and underlying parietal endoderm cells were absent in the LAMC1-null embryos, cells staining strongly for intracellular laminin α1 or β1 subunits were seen in the inner cell mass, characteristic of primitive endodermal cells (Doetschman et al., 1985; Dziadek and Timpl, 1985). Thus, although these cells had evidently started to differentiate along the extra–embryonic endodermal pathway, the absence of a trophectodermal basement membrane had prevented further development involving the migration of parietal endodermal cells and/or the differentiation of primitive endodermal cells. The need for cellular interactions with laminin at this stage of development is consistent with the observation that embryos lacking the β1 integrin subunit or α-dystroglycan also die rapidly after day 4.5 pc (Fässler and Meyer, 1995; Stephens et al., 1995). It has previously been shown that the formation of the proamniotic cavity in the epiblast is due to death of the cells unable to interact with the extracellular matrix via a β1-containing integrin receptor (Coucouvanis and Martin, 1995). At 5.5 d pc we were unable to find anything other than disrupted laminin γ1–negative embryos, the cells of which displayed increased DNA fragmentation. The fact that the increased extent of DNA fragmentation in these embryos varied widely is consistent with a rapid onset of apoptotic cell death subsequent to disruption of embryo structure.

The early embryonic lethality of the LAMC1 knockout in vivo precludes an analysis of the roles of basement membranes in subsequent postimplantation development. However, by observing the differentiation of −/− ES cells in culture we were able to see that developing myotubes were affected. It has recently been shown that the stability of myotubes derived from LAM2A −/− ES cells is compromised (Kuang et al., 1998). Taken together with our observation of aggregates of myosin-positive LAMC1 −/− cells with processes characteristic of retraction, these results are consistent with the laminin γ1 subunit being required for the formation of α2/γ1-containing laminin isoforms that are necessary for the maintenance of myotubes.

In the absence of the γ1 subunit, no covalently bonded laminin subunits were produced by differentiating ES cells. Previous studies have indicated that intracellular laminin transport and secretion is limited by the assembly of the α1 subunit to form a triple coiled-coil α-helix with preassembled βγ dimers (Peters et al., 1985; Hunter et al., 1990; De Arcangelis et al., 1996; Yurchenco et al., 1997). However, despite the lack of extracellular laminin deposition in the embryoid bodies, our immunoprecipitation experiments clearly showed that some laminin subunits were secreted from the ES cells. However, the α1 chain had undergone cleavage to produce a fragment of ∼300 kD that was released into the medium. The size of this fragment is consistent with cleavage occurring at or close to the terminal globular domain of the α1 subunit. A similar but incomplete cutting of the laminin α1 subunit upon secretion from transfected cells has recently been demonstrated, and it was suggested that when unable to assemble into a coiled-coil structure, the α1 subunit adopts a conformation laying it open to cleavage (Yurchenco et al., 1997). Although the truncated α1 subunit in the −/− ES cell medium described here was noncovalently associated with a protein band of 200 kD, the fact that the α1 subunit was cleaved to a 300-kD fragment indicates that it is also unlikely to have been associated with any other laminin subunits via a coiled-coil interaction.

Experiments in vitro have shown that collagen type IV can self-assemble into a characteristic chicken wire network (Yurchenco and O'Rear, 1993). Furthermore, there are reports of basement membrane-like structures lacking either type IV collagen (Brauer and Keller, 1989) or laminin (Hahn et al., 1980). However, the present work demonstrates in both embryos and embryoid bodies that laminin is necessary for the incorporation of collagen type IV into a continuous basement membrane. Although we cannot rule out the hitherto undocumented existence of other laminin γ subunits, it is clear that the γ1 isoform is a prerequisite for the formation of that laminin variant necessary for the assembly of the first basement membranes in the preimplantation embryo. Furthermore, the available data are consistent with that variant being laminin type-1, the α1, and β1 subunits of which have also been shown to be expressed in the preimplantation embryo (Shim et al., 1996). In the absence of the laminin, collagen IV and perlecan were seen in disorganized deposits within the embryoid bodies. However, little if any nidogen was deposited with them, but instead it was released into the culture medium. Although nidogen has been shown to be able to bind to both collagen IV and perlecan in solid-phase binding assays in vitro (Battaglia et al., 1992; Dziadek et al., 1985), the present experiments indicate that this apparently does not occur in the absence of laminin in embryoid bodies. Clearly factors other than binding interactions between its individual components regulate basement membrane deposition in vivo. In this regard, it should be noted that basement membrane organization is disrupted by targeted deletions of the β1 integrin subunit (Fässler and Meyer, 1995; Stephens et al., 1995) or α-dystroglycan (Williamson et al., 1997), pointing to the involvement of cellular receptors for laminin in basement membrane deposition.

Taken together, the present results show that other basement membrane components require laminin for their assembly into an organized basement membrane structure. Furthermore, although cell–cell contacts may be sufficient for epithelium formation during preimplantation development, these do not form a sufficient basis for postimplantation embryonic development when basement membranes are first required for endoderm differentiation.

This work was supported by a European Union SCIENCE Programme Grant (SCC-CT90-0021) and Wellcome Trust grant to D. Edgar. N. Smyth, C. Frie, and M. Paulsson were supported by a grant from the Bundesministerium fuer Bildung und Forschung in the framework of the Centre for Molecular Medicine Cologne. P. Murray was supported by a postgraduate research studentship (G610/47) of the Medical Research Council.

ES

embryonic stem

pc

post coitum

TUNEL

terminal dUTP–biotin nick end labeling