Regulation of the differentiation and behaviour of extra-embryonic endodermal cells by basement membranes

Prior to implantation of the mammalian blastocyst, cells at the periphery of the inner cell mass (ICM) that line the blastocoel cavity differentiate to become primitive endoderm, depositing a basement membrane (BM) between themselves and the remaining undifferentiated cells of the ICM (Nadijcka and Hillman, 1974; Smyth et al., 1999). The primitive endoderm cells that remain attached to this BM differentiate to become visceral endoderm (VE) cells, whereas those lying adjacent to the trophectoderm differentiate to become parietal endoderm (PE) cells which migrate over the blastocoelic surface of the trophectodermal BM (Enders et al., 1978; Gardner and Papaioannou, 1975). As they migrate on this pre-existing BM, PE cells secrete large amounts of laminin and other BM components that are incorporated into Reichert’s membrane, a thick BM that acts as a barrier between maternal and embryonic environments, thus forming the parietal yolk sac (Salamat et al., 1995).

The mechanisms that regulate differentiation and behaviour of the extra-embryonic endoderm cells are only now becoming apparent. VE cell differentiation has been shown to be dependent on the presence of factor(s) derived from undifferentiated embryonic stem (ES) cells (Lake et al., 2000). It is likely that one such factor is the VE-inducing molecule, bone morphogenic protein (BMP)4, which is expressed in undifferentiated ES cells but subsequently is downregulated in differentiating columnar epiblast cells (Coucouvanis and Martin, 1999), the development of which is dependent on the BM (Coucouvanis and Martin, 1995; Murray and Edgar, 2000). Conversely, in conditions where only differentiated epiblast cells are present, more PE cells have been shown to develop (Lake et al., 2000), suggesting that PE cell differentiation is a default pathway that will be indirectly enhanced by the BM. However, apart from this indirect mechanism, previous work using F9 embryonal carcinoma cells has shown that PE cell differentiation can be directly promoted by parathyroid hormone related peptide (PTHrP) (Chan et al., 1990; van de Stolpe et al., 1993). Furthermore, PTHrP has been demonstrated to stimulate the migration of PE cells from ICM explants (Behrendtsen et al., 1995). Taken together, these observations indicate that PE cell differentiation is likely to be influenced by both direct and indirectly acting factors that, in turn, may be regulated by the BM. The relative importance of these factors and their interactions remain to be established.

In order to analyse the functions of BMs during embryogenesis, we have previously targeted the mouse Lamc1 gene, the mutation of which results in early embryonic lethality (Smyth et al., 1999). The Lamc1 gene encodes the laminin γ1 subunit, which is found in most if not all laminin isoforms expressed during early embryogenesis (see references in Colognato and Yurchenco, 2000). Consequently, in the absence of the γ1 subunit, no BMs were deposited in these embryos because mature trimeric laminin molecules fail to be assembled and secreted from the cells (Smyth et al., 1999). Analysis of the embryos at different stages showed that the blastocysts died very shortly after implantation, at which time it became apparent that the parietal yolk sac had failed to develop (Smyth et al., 1999). The failure of yolk sac development is likely to be a consequence of abnormal PE cell differentiation and/or migratory behaviour, but it remains to be established which is compromised by the lack of the BM.

The availability of embryoid bodies (EBs) derived from Lamc1^−/− mouse embryonic stem (ES) cells offers a unique system to define the reasons for the defects in endodermal cell differentiation that result from the lack of BM. EBs represent a well-established in vitro model for investigating peri-implantation development (Robertson, 1987). In a sequence of events that closely resembles development in vivo, cells positioned at the periphery of EBs differentiate into primitive endodermal cells which in turn differentiate into either VE or PE cells (Robertson, 1987). The results of the present analysis using EBs demonstrate that while the initial differentiation of primitive endodermal cells is unaffected by the lack of a BM, the organisation of VE and PE cells is disrupted in its absence. Furthermore, more VE and PE cells differentiated from primitive endodermal cells in Lamc1^−/− EBs, indicating that the BM helps maintain the population of primitive endodermal stem cells. The PE cells of Lamc1^−/− EBs migrate more readily on laminin and fibronectin substrates than those differentiating in control EBs which require PTHrP in order to initiate migration, suggesting not only that the major role of PTHrP is to permit PE cell delamination from the BM, but also that loss of contact with the BM is responsible for enhanced PE cell differentiation.

The production of R1 mouse Lamc1^+/− and Lamc1^−/− ES cells has been described in detail previously (Smyth et al., 1999). The clone of Lamc1^+/− ES cells used here as a control has a wild-type phenotype and was that used previously for the construction of healthy heterozygous germline animals (Smyth et al., 1999). The absence of clonal artefacts in the Lamc1^−/− cells was confirmed by rescue of phenotype by addition of laminin type-1 (Sigma) to developing EBs (Murray and Edgar, 2000). The cells were cultured on mitomycin-treated STO feeder cells in gelatinised 3.5 cm tissue culture dishes. The culture medium was DMEM (Gibco-BRL) supplemented with 15% (vol/vol) ES grade foetal bovine serum (FBS) (Gibco-BRL), 0.1 mM β mercaptoethanol, 1 mM L-glutamine, 0.1 mM non-essential amino acids and 1000 U/ml of LIF (ESGRO; Gibco-BRL). ES cells were subcultured on STO feeder layers every 2 days.

To make EBs, ES cells were trypsinised, triturated and split 1:10 by replating in bacterial petri dishes, under which conditions the cells remained in suspension and formed aggregates. The EB culture medium was as above, except that LIF was omitted and the FBS content was reduced to 10% (vol/vol). For immunostaining, EBs were fixed for 1 hour with 4% (wt/vol) paraformaldehyde, washed three times in PBS and soaked in 15% (wt/vol) sucrose overnight at 4°C. The samples were then incubated in 7.5% (wt/vol) gelatin:15% (wt/vol) sucrose for 1 hour at 37°C. 100 μl aliquots were pipetted onto a block of gelatin and allowed to set at room temperature, after which they were mounted onto cork discs with OCT cryofixative (Dako) and frozen in liquid nitrogen-cooled isopentane before storage at –80°C. For Toluidine Blue staining, EBs were fixed for 1 hour in 2% (wt/vol) glutaraldehyde:4% (wt/vol) paraformaldehyde and prepared as described previously (Robinson, 1982).

Endodermal cell counts were performed on single, Toluidine Blue-stained sections obtained at the largest diameter from serial section series of 15-33 Lamc1^+/− and Lamc1^−/− EBs (numbers for individual experiments are given in the figure legends). The sections were viewed with bright field illumination on a Leitz RM22 microscope. EBs that were growth-retarded or that had formed multiples were omitted from the analysis. Primitive endodermal cells were identified as thin, spindle-shaped cells that stained intensely with Toluidine Blue and formed a continuous monolayer on the surface of the EBs (Nadijcka and Hillman, 1974; Hogan and Tilly, 1978). These cells were clearly distinguishable from PE cells, which were thicker, more rounded and only loosely associated with the surface of the EBs (Casanova and Grabel, 1988; Enders et al., 1978). Cells with this appearance lacked apical vacuoles, distinguishing them from VE cells, and were shown in other experiments to display strong laminin immunoreactivity, consistent with their identity as PE cells. Results are presented ± s.e.m., the significance of differences being assessed by unpaired t-test; P<0.05 was regarded as significant.

For analysis of PE migration, EBs were cultured in suspension for 4 days to allow PE cells to differentiate, and then plated on tissue culture plastic, fibronectin or laminin substrates for various times and fixed with 4% paraformaldehyde (wt/vol) for 30 minutes. The fibronectin substrate was prepared by coating tissue culture chamber slides (GIBCO BRL) with gelatin and allowing fibronectin to adsorb from medium containing FBS (Grabel and Watts, 1987). The laminin substrate was prepared by pipetting a 200 μl aliquot of either 10 μg/ml or 100μg/ml laminin (Sigma) in phosphate buffered saline (PBS) onto a tissue culture chamber slide. The slide was left at room temperature for 30 hours to allow the laminin to adsorb to the plastic. 100 nM PTHrP 1-34 (a gift from Dr J. A. Gallagher, Department of Human Anatomy, University of Liverpool) was added to the culture medium at the same time as EBs were plated onto the tissue culture slides.

Cryostat sections (10 μm) were blocked with 10% (vol/vol) goat serum (GS) in PBS, which was applied for 1 hour and then aspirated before application of the primary antibody. The primary antibodies used were rabbit anti-EHS laminin, which recognises all three subunits of laminin type 1, at 1/5000 dilution (Kücherer-Ehret et al., 1990); rabbit anti-mouse α-feto-protein (AFP) serum at 1/200 dilution (ICN Biomedicals); and rat TROMA-3 diluted 1/2 (a gift from N. Smyth, Department of Biochemistry, University of Cologne), which recognises the mouse PE cell specific cytokeratin Endo C (Boller and Kemler, 1983). Incubations with primary antibodies were carried out overnight in 1% (vol/vol) GS in PBS in a humidified atmosphere at room temperature and the sections were then washed three times in PBS. For immunostaining with anti-EHS laminin and anti-AFP, the secondary antibody was TRITC-conjugated swine anti-rabbit IgG at 1/100 dilution (Dako); for TROMA-3, the secondary antibody was biotinylated rabbit anti-rat IgG at 1/50 dilution (Dako). Secondary antibodies were applied in 1% (vol/vol) GS in PBS at room temperature for 2 hours and the sections were then washed three times in PBS. Streptavidin conjugated with fluorescein isothiocyanate (FITC) (Amersham) at 1/50 dilution in 1% (vol/vol) GS in PBS was applied to TROMA-3 sections at room temperature for 2 hours and the sections were then washed three times in PBS. Sections were mounted in fluorescent mounting medium (Dako) and photographed using a Leitz RM22 fluorescent microscope.

Total RNA was extracted from Lamc1^+/− and Lamc1^−/− EBs using guanidium isothiocyanate (Chomczynski and Sacchi, 1987), and reverse transcribed using Superscript™ II (Gibco BRL). For day 2and day 6 EBs, whole populations were used, but for day 10 EBs, 10-15 cavitated Lamc1^+/− EBs and an equal number of Lamc1^−/− EBs were selected using phase contract microscopy. Primers for HNF4α1 (Nakhei et al., 1998), perlecan (Miller et al., 1997), and vHNF1 and GATA-4 (Duncan et al., 1997) were as described previously. GAPDH primers were 5′ggtgaaggtcggagtcaacgg3′ (forward) and 5′ggtcatgagtccttccacgat3′ (reverse); product size, 520 bp; annealing temperature, 54°C. Lamb1 primers were 5′gcagacacaacaccaaaggc3′ (forward) and 5′tgtacccatcacagatcccg3′ (reverse); product size, 344 bp; annealing temperature, 56°C. AFP primers were 5′acatcagtgtctgctggcac3′ (forward) and 5′agcgagtttccttggcaacac3′ (reverse); product size, 461 bp; annealing temperature, 54°C. Semi-quantitative RT-PCR was performed as previously described to determine mRNA levels relative to those of GAPDH (Squitti et al., 1999).

After 2 days of suspension culture, light microscopy of Toluidine Blue-stained resin-embedded sections showed that Lamc1^−/− and Lamc1^+/− control EBs were indistinguishable, in both cases spindle-shaped primitive endodermal cells that stained intensely with Toluidine Blue (Hogan and Tilly, 1978) formed a monolayer on the surface (Fig. 1A,B). This is typical of the organisation of these cells in the blastocyst (Nadijcka and Hillman, 1974). There were no obvious differences in the overall growth characteristics of Lamc1^−/− and Lamc1^+/− EBs. By day 6, the primitive endodermal cells had begun to diverge into two distinct morphological types in both Lamc1^−/− and Lamc1^+/− EBs (Fig. 1C-F). Type 1 were highly polarised with large apical vacuoles (Fig. 1C,D), characteristic of VE cells (Morini et al., 1999), whereas type 2 cells were not as tightly associated, and were smaller and lacked large apical vacuoles (Fig. 1E,F), consistent with the appearance of PE cells (Casanova and Grabel, 1988; Enders et al., 1978). It has been shown previously that VE cells express low levels of BM components (Behrendtsen et al., 1995; Hogan and Tilly, 1981), and consequently the BM underlying these cells in Lamc1^+/− EBs was not visible with light microscopy (Fig. 1C). In contrast, PE cells express very high levels of BM components (Behrendtsen et al., 1995; Hogan et al., 1984), and so the BM-like extracellular matrix underlying these cells in Lamc1^+/− EBs tended to be extremely thick (Fig. 1E), resembling Reichert’s membrane (Inoue et al., 1983). However, extracellular matrix material was present as discrete deposits in Lamc1^−/− EBs and did not form a continuous Reichert’s membrane-like structure (Fig. 1F), owing to the lack of secreted laminin (Smyth et al., 1999).

At day 6, the VE cells of Lamc1^+/− EBs were arranged in an epithelial monolayer at the periphery of the EB (Fig. 1C). The VE cells of Lamc1^−/− EBs, however, while being located at the periphery, formed disorganised groups in which the basal surfaces of the cells were orientated towards each other (Fig. 1D), rather than being aligned parallel to the core of the EB as they were in Lamc1^+/− EBs. The PE cells of Lamc1^+/− EBs were arranged one to two cells deep overlying the thick BM that separated them from the inner core cells of the EB (Fig. 1E). In contrast, the PE cells of Lamc1^−/− EBs usually formed multilayers at the periphery and were frequently associated with discrete deposits of extracellular matrix material (Fig. 1F).

To determine if the absence of compartmentalisation between endodermal cells and remaining inner core cells of the BM-deficient Lamc1^−/− EBs affected the numbers of endodermal cells, we counted these cells in Lamc1^−/− and Lamc1^+/− EBs at days 2 and 10 (Fig. 2A). At both time-points endodermal cells could be clearly distinguished from the inner core cells of the EBs because of their intense staining with Toluidine Blue (Hogan and Tilly, 1978). The results showed that there was no significant difference (P>0.05, unpaired t-test) between the numbers of extra-embryonic endodermal cells in Lamc1^−/− and Lamc1^+/− EBs at either day 2 or day 10 (Fig. 2A). To substantiate this result, semi-quantitative RT-PCR was used to estimate the relative expression levels of Gata4 and vHNF1 mRNAs. Both these genes are expressed in primitive, visceral and parietal endodermal cells (Arceci et al., 1993; Barbacci et al., 1999; Soudais et al., 1995), so they are an appropriate indicator for all extra-embryonic endodermal cell lineages. The results showed that the expression levels of these genes in Lamc1^−/− and Lamc1^+/− EBs were similar both at day 2 and day 10 (Fig. 2B). Thus, the absence of a BM in Lamc1^−/− EBs did not appear to affect the overall levels of extra-embryonic endodermal cell differentiation.

Although the BM did not affect the overall numbers of extra-embryonic endodermal cells, we wished to establish if it had any specific effect on the number of VE and/or PE cells that differentiated form primitive endodermal cells. Accordingly, the numbers of VE and PE cells were determined in age-matched Lamc1^−/− and Lamc1^+/− EBs by scoring them on the basis of their morphological characteristics as described above. At day 4, no significant difference (P>0.05, unpaired t-test) was observed in the numbers of primitive endodermal, VE or PE cells in Lamc1^−/− relative to Lamc1^+/− EBs (Fig. 3). By day 10, however, the numbers of both VE and PE cells were significantly increased (P<0.02) in Lamc1^−/− relative to Lamc1^+/− EBs (Fig. 3). Given that there was no difference between the overall number of extra-embryonic endodermal cells in Lamc1^−/− and Lamc1^+/− EBs (Fig. 2A), then the higher numbers of VE and PE cells observed in Lamc1^−/− EBs is consistent with an increased differentiation of primitive endodermal cells into both these lineages. This hypothesis is supported by the observation that the number of primitive endodermal cells was significantly greater (P<0.001) in the Lamc1^+/− EBs compared with Lamc1^−/− EBs (Fig. 3).

To establish enhanced VE cell differentiation definitively, we performed immunostaining with the VE cell-specific marker, AFP. We found that there were more AFP-positive cells in 10 day Lamc1^−/− EBs than in the Lamc1^+/− EBs (Fig. 4A,E), consistent with the increased numbers of cells with VE-like morphology. To confirm that the increase in differentiation of primitive endoderm to VE cells had not occurred at the expense of PE cell differentiation, we performed immunostaining with TROMA-3 monoclonal antibodies that recognise the PE-specific cytokeratin Endo C (Boller and Kemler, 1983). We found that there were also more TROMA-3-positive cells in the Lamc1^−/− compared with Lamc1^+/− EBs, and that the organisation of the Lamc1^−/− cells was disrupted compared with Lamc1^+/− controls (Fig. 4C,G).

Semi-quantitative RT-PCR was used to estimate the relative mRNA levels of AFP and another VE cell-specific marker HNF4α1 (Nakhei et al., 1998). As expected, these genes were not expressed in either EB type at day 2, before any VE cells had differentiated (Fig. 5A,B), but at later time-points the levels of both mRNAs were increased in Lamc1^−/− compared with Lamc1^+/− EBs, consistent with there being more VE cells in the Lamc1^−/− EBs (Fig. 5A,B). In contrast to the VE markers, there were relatively high levels of expression of Lamb1 and perlecan mRNAs at day 2 (Fig. 5C), reflecting the expression of BM components by primitive endodermal cells. However, the relative levels of Lamb1 and perlecan mRNAs were higher in Lamc1^−/− EBs than controls at later time points (Fig. 5C), consistent with greater numbers of PE cells.

Although the failure of parietal yolk sac development in Lamc1^−/− embryos (Smyth et al., 1999) could have been caused by either a defect in PE cell differentiation or PE cell migration, the above experiments show that PE cells do differentiate in the absence of a BM. Thus, the effect of the BM on PE migration was next examined. To do this, Lamc1^−/− and Lamc1^+/− EBs were cultured for 4 days in suspension, and then allowed to attach to different tissue culture substrates. This time-point was chosen because although PE cells have started to differentiate at day 4, there was no significant difference in the numbers of these cells between the Lamc1^−/− and Lamc1^+/− EBs, which may have affected their migration characteristics (Fig. 3). Furthermore, we have shown previously that formation of the columnar epiblast epithelium and proamniotic cavity normally begins in Lamc1^+/− EBs from day 5 (Murray and Edgar, 2000). Hence, by using day 4 EBs to assay PE cell migration, we were able to avoid any influence of the columnar epiblast epithelium on the migration of these cells.

After 30 hours of attachment to a fibronectin substrate, there was no evidence of cell migration from control EBs (Fig. 6A,B). However, after 3 days culture, some migration of PE cells (defined by intracellular laminin immunoreactivity and morphology) was noted (Fig. 6C,D), consistent with the timing of PE cell migration on fibronectin substrates previously reported (Behrendtsen et al., 1995). In contrast, after 30 hours of attachment to fibronectin, extensive PE cell migration from Lamc1^−/− EBs had already occurred (Fig. 6E,F; Fig. 7). It has previously been demonstrated that the addition of laminin inhibits PE cell migration over fibronectin substrates (Behrendtsen et al., 1995). Thus, the PE cell migration from Lamc1^−/− EBs noted in the present experiments could be due to the absence of laminin secreted from these cells. To test this hypothesis, EBs were plated directly onto laminin substrates. Counts of migrating PE cells confirmed that little migration from Lamc1^+/− control EBs occurred on either fibronectin or laminin substrates after 30 hours culture (Fig. 7). Surprisingly, extensive PE cell migration from Lamc1^−/− EBs was observed on both fibronectin and laminin substrates at this time point (Fig. 7; Fig. 8A,B), although the migration was lower (P<0.01, unpaired t-test) on laminin compared with that on fibronectin (Fig. 7), consistent with previous observations (Behrendtsen et al., 1995). Significantly, it was noted that PE cell migration from Lamc1^−/− EBs stopped abruptly at the limit of the laminin substrate (Fig. 8C-D), indicating that laminin is a permissive substrate for PE cell migration, whereas tissue culture plastic is not. Taken together, these observations indicate that the absence of secreted laminin is unlikely to explain the rapid migration of PE cells from Lamc1^−/− EBs and an alternative explanation is required.

Before displaying migratory behaviour, it is clear that PE cells must first detach from the EB. In order to differentiate between PE cell delamination and migration, we made use of the observation that tissue culture plastic is not a permissive substrate for PE cell migration from Lamc1^−/− EBs (Fig. 8C). Thus, when plated directly onto tissue culture plastic, it was found that although the PE cells were unable to migrate away from Lamc1^−/− EBs, nevertheless they were able to detach from the EBs and adhere to the substrate (Fig. 8E,F). In Lamc1^+/−images of E,G, respectively. Note that the lower levels of intracellular laminin in the PE cells from Lamc1^+/− EBs (C) compared with those migrating from Lamc1^−/− EBs (E,G) is due to rapid laminin secretion following trimer assembly. Scale bar: 30 μm.

EBs, on the other hand, the PE cells remained firmly attached to the EB showing an absence of both delamination and migration (Fig. 8G,H). These observations suggest that attachment to the BM present in control EBs is the major factor inhibiting PE cell migration on laminin or fibronectin substrates.

It has previously been shown that PTHrP promotes PE cell migration on a variety of substrates, including fibronectin and laminin (Behrendtsen et al., 1995). In the light of the above observations, we wished to determine if the action of PTHrP was on delamination rather than a stimulation of migration per se. Thus, Lamc1^−/− and Lamc1^+/− EBs were cultured for 30 hours on laminin substrates with or without the addition of PTHrP. The results demonstrate that while PTHrP was necessary for PE cell migration from control EBs (Fig. 9C,D), as previously reported (Behrendtsen et al., 1995), the addition of PTHrP to cultures of Lamc1^−/− EBs did not make any appreciable difference to PE cell migration (Fig. 9E-H). Determination of cell numbers confirmed that while PTHrP did not affect PE cell migration from Lamc1^−/− EBs, the number of migratory cells remained lower (P<0.01, unpaired t-test) in Lamc1^+/− in the presence of PTHrP than in Lamc1^−/− EBs (Fig. 10). Given the lack of effect of PTHrP on PE cell migration in BM-deficient Lamc1^−/− EBs where no delamination is required, these results suggest that the major effect of PTHrP in Lamc1^+/− EBs is to induce delamination from the BM.

In this study we have used EBs to analyse the factors influencing extra-embryonic endodermal cell differentiation and behaviour, in order to explain the peri-implantation lethality caused by the absence of BMs in Lamc1^−/− mouse embryos (Smyth et al., 1999). The results show that while the initial differentiation of primitive endodermal cells was unaffected by the absence of BMs in Lamc1^−/− EBs, the BM affects the subsequent development of VE and PE cells by regulating their numbers and organisation. Furthermore, analysis of EB outgrowths showed that PE cells migrated more readily from Lamc1^−/− EBs than from controls, which required PTHrP in order to permit delamination of the PE cells from the BM. Thus, the failure of parietal yolk sac development in vivo (Smyth et al., 1999) was not caused by failure of PE cell differentiation or intrinsic migratory ability, but is likely to have been due to the absence of their normal trophectodermal BM substrate.

The laminin and other BM components deposited between primitive endodermal cells and the rest of the ICM or embryoid body are produced by the primitive endodermal cells themselves (Cheng et al., 1998). Indeed, the increased expression of laminin and other BM components is one of the earliest indications of primitive endodermal cell differentiation, which occurs in EBs several days before deposition of a BM (Grover et al., 1983; Murray and Edgar, 2000). It is therefore not surprising that the present results demonstrate the initial differentiation of endodermal cells to be independent of the BM. It has recently been shown that increased steady-state levels of the transcription factor Oct3/4 result in the differentiation of ES cells into primitive endodermal cells (Niwa et al., 2000). However, the intercellular signalling mechanisms that regulate Oct3/4 expression are unknown, and the relationship between Oct3/4 and laminin expression is currently under investigation (manuscript in preparation).

The endodermal cells that formed in Lamc1^−/− EBs differed from controls in being unable to form an organised epithelium, although the presence of apical vacuoles in the VE cells clearly indicates that they were polarised. These observations agree with earlier work on BM-deficient dystroglycan-null EBs (Henry and Campbell, 1998) in demonstrating that VE cell polarisation can occur in the absence of the BM. However, the VE cells differentiating in β1-integrin-null EBs not only fail to form an epithelium but also fail to polarise (Aumailley et al., 2000; Stephens et al., 1993), indicating that the requirement of

β1-integrins for VE cell polarity may involve mechanisms independent of their interactions with BM ligands. This hypothesis is supported by the observation that cadherin:catenin complexes are unable to maintain the subcortical cytoskeleton characteristic of polarised cells in the absence of β1 integrins (Wang et al., 1999).

It is likely that the absence of the BM has indirect effects on VE cell differentiation, owing to its role in the development of the epiblast, which has been shown to be dependent on the BM (Coucouvanis and Martin, 1995; Murray and Edgar, 2000). While BMP4 is expressed in undifferentiated ES cells, it is downregulated in columnar epiblast cells (Coucouvanis and Martin, 1999). In the absence of the BM, failure of columnar epiblast development is accompanied by increased levels of BMP4 (Murray and Edgar, 2000). Because BMP4 in turn induces VE cell differentiation (Coucouvanis and Martin, 1999; Gu et al., 1999; Sirard et al., 1998; Yang et al., 1998), absence of the BM would be expected to result in the increased numbers of VE cells described here.

The present results show that not only VE cell differentiation but also that of the PE cells is enhanced in Lamc1^−/− EBs. PE differentiation from primitive endodermal cells has been suggested to occur as a default pathway in the absence of VE differentiation factor(s) produced by ES cells (Lake et al., 2000). This notion is supported by observations that PE differentiation is enhanced at the expense of VE differentiation in Actr1a^−/− and Smad4^−/− mutants, implicating the involvement of ES cell-derived BMP4 (Gu et al., 1999; Sirard et al., 1998; Yang et al., 1998). However, given the increased numbers of VE cells in Lamc1^−/− EBs, then the increased differentiation of PE cells reported here is unlikely to have been due to an indirect effect of the BM on epiblast development.

PE cell differentiation has been shown to be directly stimulated by trophectodermal cells and/or the PTHrP which they secrete (Behrendtsen et al., 1995; van de Stolpe et al., 1993). However, PE cells differentiate in EBs that lack a trophectoderm (Doetschman et al., 1985), and PTHrP receptor-null embryos progress normally through this stage of development (Lanske et al., 1996), indicating that PTHrP is not essential for PE cell differentiation and that alternative mechanisms must be responsible. A very early hypothesis has been made, suggesting that PE cell differentiation occurs owing to loss of contact with the BM, as a result of overcrowding caused by continuing proliferation of primitive endodermal cells (Gardner, 1982). The present experiments support this hypothesis by showing that more PE cells do indeed develop in EBs lacking a BM. However, we also demonstrate that early differentiating PE cells have the ability to migrate away from the remaining cells of Lamc1^−/− EBs without the need for additional factors such as PTHrP. Thus, the increased numbers of PE cells appearing at later time points in Lamc1^−/− EBs could be either a direct consequence of lack of BM contact and/or alternatively could be caused by the loss of lateral inhibition if early differentiating PE cells are not restrained by the BM.

When presented with permissive substrates, PE cells migrated much more readily from Lamc1^−/− than from Lamc1^+/− control EBs. Using ICM explants, it has previously been shown that laminin substrates reduce PE cell migration (Behrendtsen et al., 1995). However, a lack of secreted laminin from Lamc1^−/− cells was not responsible for enabling migration, as the PE cells migrated readily on exogenous laminin substrates. In order to explain these observations, it should be noted that, prior to migration, PE cells must normally delaminate from the BM deposited between extra-embryonic endoderm and the epiblast. Thus, any factors regulating delamination are likely to play a crucial role in affecting the migration of PE cells.

Previous work has suggested that PTHrP stimulates PE cell differentiation and migration (Behrendtsen et al., 1995; van de Stolpe et al., 1993), and the present results agree with this observation by showing that PTHrP increases the number of migrating PE cells from control EBs. However, PTHrP did not increase PE cell migration from Lamc1^−/− EBs, indicating that the main effect of PTHrP is to promote PE cell delamination from the primitive endodermal BM. It is known that the binding of PTHrP to the type I PTHrP receptor activates protein kinase C (PKC)-coupled signal transduction pathways (Abou-Samra et al., 1992). Furthermore, activation of PKC has recently been shown to regulate intracellular β1 integrin trafficking (Ng et al., 1999). A similar relocalisation of β1 integrins caused by PTHrP-induced PKC activation may therefore be involved in PE cell delamination from the BM, thereby allowing migration to proceed.

The fact that PE cells differentiate in BM-deficient Lamc1^−/− EBs indicates that the failure of parietal yolk sac formation in vivo (Smyth et al., 1999) is likely to be caused by defective PE cell migration. Indeed, both Lamc1^−/− and integrin β1-null mouse embryos lack BMs, and it has been reported that cells with the intense laminin immunoreactivity characteristic of PE cells accumulate on the surface of these ICMs shortly after implantation (Smyth et al., 1999; Stephens et al., 1995). Because outgrowth analysis showed that PE cells derived from Lamc1^−/− EBs have no migratory defect, then the failure of PE cells to form the parietal yolk sac in Lamc1^−/− embryos probably results from the absence of the trophectodermal BM, which is the normal substrate for migrating PE cells (Leivo et al., 1980; Salamat et al., 1995; Wartiovaara et al., 1979).

The only known function of PE cells is to synthesise Reichert’s membrane, which acts as a filtrative layer between the embryonic and maternal environments (Gardner, 1983). However, despite the formation of trophectodermal and primitive endodermal BMs in dystroglycan-null mutants, Reichert’s membrane fails to develop, although the PE cells do migrate (Williamson et al., 1997). Notably, dystroglycan-null mutants form apparently normal egg cylinders and do not die until after E7.5. The fact that embryonic lethality caused by to the lack of PE cell migration in Lamc1^−/− and β1 integrin-null embryos occurs prior to E6.5 indicates that the death of these embryos may result from events other than the failure to deposit Reichert’s membrane. Very recently it has been demonstrated that the ability to interact with extracellular matrices via β1 integrin receptors can regulate gene expression in endodermal cells (Aumailley et al., 2000). It is therefore possible that a defect in gene expression caused by the lack of basement membranes also contributes to the lethality seen in Lamc1^−/− embryos.

We thank Mrs Ann Currie and Ms Marion Pope for skilful technical assistance. We are grateful to Dr Jim Gallagher for his generous gift of PTHrP. This work was supported by a Postgraduate Research Studentship awarded by the Medical Research Council and a scholarship from the British Federation of Women Graduates to P. M.