PINCH1 regulates cell-matrix and cell-cell adhesions, cell polarity and cell survival during the peri-implantation stage

Adhesion between cells and adhesion of cells to the extracellular matrix (ECM) are fundamental for the development of multicellular organisms. Cell adhesion receptor proteins that mediate such interactions include cadherins and integrins. Integrins are essential for cell migration but are also involved in the induction of cell polarity and differentiation, and in the regulation of cell proliferation and programmed cell death. Integrins are heterodimeric transmembrane glycoproteins consisting of a non-covalently bound α and β subunit. The interaction of integrins with their ligands at the extracellular side leads to the connection of the integrin cytoplasmic tails with the actin cytoskeleton, to the clustering of integrins into focal adhesion sites (FAs), and finally to the activation of various intracellular signalling pathways (Brakebusch and Fässler, 2003). Since the integrin cytoplasmic domains lack both an actin binding domain as well as enzymatic activity, all of these effects are mediated by integrin-associated molecules.

PINCH1 and 2 (Zhang et al., 2002a; Braun et al., 2003) are proteins that localise to FAs where they interact with the integrin-linked kinase (ILK), which in turn binds the cytoplasmic tails of β1 and β3 integrins (reviewed by Wu, 1999; Grashoff et al., 2004). The PINCH proteins are highly similar, encoded by different genes and composed of five tandemly arranged LIM domains followed by putative nuclear localisation/export signals (Braun et al., 2003; Zhang et al., 2002a). LIM domains are double zinc-binding motifs that facilitate interactions with other proteins or nucleic acid (Kadrmas and Beckerle, 2004). Consistent with such binding properties, the first LIM domain of PINCH has been shown to bind ILK and the fourth LIM domain the SH2/SH3 adaptor protein Nck2 (or Grb4), which in turn can associate with several receptor tyrosine kinases and thereby links integrins to growth factor signalling (Tu et al., 1998). Furthermore, mammalian PINCH1 and 2 and C. elegans PINCH have been shown to be present in the nucleus of a variety of cell types (Hobert et al., 1999; Zhang et al., 2002a; Campana et al., 2003). Whether nuclear PINCH1 and 2 binds nuclear proteins and/or DNA, however, is not known.

Loss-of-function studies in C. elegans and D. melanogaster and cell biological studies with mammalian cells, in which either the interaction of PINCH1 with ILK was disrupted by overexpressing dominant-negative forms of PINCH1 or PINCH1 levels were depleted by RNA interference provided several compelling lines of evidence that PINCH1 is essential for integrin functions and actin organization (Zhang et al., 2002b; Fukuda et al., 2003). Loss of PINCH expression in flies and worms revealed a crucial role of PINCH in integrin-mediated cell adhesion and linkage of actin to integrin adhesion sites (Clark et al., 2003; Hobert et al., 1999). Cell biological studies showed that PINCH1 promotes cell adhesion, spreading, motility, survival and ECM assembly (Wu, 1999). These studies also demonstrated that the interaction of PINCH with ILK is necessary to target both proteins to FAs and to prevent degradation of the PINCH-ILK-parvin complex (Zhang et al., 2002c; Fukuda et al., 2003).

Several biochemical and cell biological studies uncovered a prominent role for ILK in promoting cell survival. The proposed molecular mechanism involves ILK recruitment into FAs and phosphorylation of Ser473 of PKB/Akt (Delcommenne et al., 1998). Phosphorylation of Ser473 by ILK and of Thr308 by 3-phosphoinositide-dependent kinase 1 (PDK1) leads to full activation of PKB/Akt and inhibition of programmed cell death. In agreement with the ability of PINCH1 to stabilise and to target ILK to FAs, depletion of PINCH1 in HeLa cells leads to diminished phosphorylation of Ser473 of PKB/Akt (Fukuda et al., 2003). Interestingly, however, depletion of PINCH1 also reduced Thr308 phosphorylation and induced apoptosis even in cells overexpressing a constitutive active form of PKB/Akt (Fukuda et al., 2003). This indicates that PINCH1 is able to modulate PKB/Akt activity in an ILK-dependent and an ILK-independent manner and furthermore, that PINCH1 functions in the survival pathway both upstream of PKB/Akt as well as downstream or alternatively, independent of PKB/Akt.

In order to directly test the physiological function of PINCH1 in vivo we generated mice and embryoid bodies (EBs) carrying a disrupted PINCH1 gene. We report here that mice lacking PINCH1 expression arrest their development at the peri-implantation stage. Absence of PINCH1 in EBs led to abnormal epiblast polarity, impaired cavitation and detachment of primitive endoderm from basement membranes (BMs). These defects were also observed in β1 integin- and ILK-deficient EBs, although they were more pronounced. Additional PINCH1-specific defects included diminished survival of primitive endoderm and abnormal cell-cell interactions of endoderm and epiblast.

A 650 bp fragment from a PINCH1 EST clone (GenBank accession number AI 195455) was used to screen a PAC library. Several clones were identified and used to generate the PINCH1 constructs. To delete the PINCH1 gene in mice we generated a conventional targeting construct by replacing part of exon 3, the entire exon 4 and a large part of intron 5 with a neomycin cassette (neo knockout allele; termed PINCH1^+/neo) and electroporated it into ES cells. Out of 360 ES clones isolated, 13 underwent homologous recombination and five of them were injected into host blastocysts to generate germline chimeric mice. Second, we also constructed a conditional PINCH1 null allele (floxed allele), in which exon 4 was flanked by loxP sites and a neo-tk cassette. Upon electroporation and G418 selection 360 surviving ES cell clones were isolated and eight of them identified as targeted clones. Three of them were used to remove the neo-tk cassette by transient expression of the Cre recombinase, which gave rise to ES cell clones either lacking the neo-tk cassette (termed PINCH1^+/fl) or ES cell clones lacking the neo-tk cassette as well as exon 4 (termed PINCH1^+/–). Three PINCH1^+/– ES cell clones were used to generate mice. Insertion of the neo cassette into exon 3 disrupts the PINCH1 gene and Cre-mediated deletion of exon 4 gives rise to splicing of exon 3 into exon 5, which leads to a reading frame shift and if at all, to a short, non-functional N-terminal PINCH1 peptide. The phenotype of PINCH1^neo/neo mice and PINCH1^–/– mice was identical and therefore, both mutations were designated as PINCH1-null or PINCH1-deficient.

To obtain PINCH1^–/– ES cells, two PINCH1 ES cell clones lacking exon 4 and the neo-tk cassette (PINCH1^+/–) were re-electroporated again with the floxed PINCH1 construct. ES cell clones with a recombination event on the wild-type allele were again Cre-treated to obtain ES cells with a homozygous deletion of PINCH1^–/–. They were identified by Southern blot assay and by PCR genotyping. The following primers were used for the PCR genotyping: 5′-CTAGGCTGGTAATGCAGGCC-3′ and 5′-CCTGCCAATGATGAATTCAC-3′ to determine the wild-type or floxed allele; 5′-CTAGGCTGGTAATGCAGGCC-3′ and 5′-CGCAGTTGGCACAGTTGAAG-3′ to determine the constitutive null allele.

The following antibodies were used for the analyses: mouse monoclonal antibodies against ILK (clone 3, Transduction lab; clone 65.1.9, Upstate Biotech); rat mAb against E-cadherin (Zymed); rat mAb against ZO-1 (Chemicon); mouse mAb against GM130 (BD Biosciences Pharmingen); mouse mAb against Ki67 (BD Biosciences Pharmingen); rabbit antibody against cleaved (activated) caspase-3 (Cell Signalling Technology); rabbit antibody against integrin β1 (Fässler and Meyer, 1995); rabbit antibodies against PKB/Akt and phospho-PKB/Akt (Thr308, Ser473) (Cell Signalling Technology); rat mAb against α-tubulin. Secondary Abs were purchased from Jackson ImmunoResearch Laboratories Inc. (West Grove, PA, USA), Molecular Probes (Eugene, OR, USA) and BioRad (Hercules, CA, USA).

To generate PINCH1- and PINCH2-specific antibodies, peptides derived from the C-termini of PINCH1 (CLKKLSETLGRK) or PINCH2 (CAQPKSVDVNSL), respectively, were synthesized and conjugated to keyhole limpet haemocyanin (KLH) using a commercial kit (Imject Maleimide Activated mcKLH Kit, Pierce Biotechnology Inc., Rockford, IL, USA). Rabbits were immunized with 400 μg of conjugate four times at 2-week intervals and bled 1 week after the last injection. The antiserum was affinity-purified using a commercial kit (SulfoLink Kit, Pierce Biotechnology Inc.); high affinity antibodies were eluted from the column using 100 mM glycine buffer, pH 2.7 and then dialyzed against phosphate-buffered saline (PBS).

To test the two affinity-purified polyclonal antibody preparations, lysates were prepared from mouse bladder and western blots were carried out. The PINCH1 antibodies recognized a single band of around 35 kDa and PINCH2 antibodies a single band around 36 kDa (see Fig. S1A in supplementary material). To exclude cross-reaction GST-tagged PINCH1 and 2 were expressed in bacteria, blotted onto membranes and cross-hybridized with both antibodies. PINCH1 antibodies recognized PINCH1 but not PINCH2 and vice versa (see Fig. S1B in supplementary material).

Apoptotic cells in implantation chambers or EBs were detected using either the In Situ Cell Death Detection kit (Roche Diagnostics) or the cleaved caspase-3 immunostaining.

Wild-type R1 cells and PINCH1-null ES cells were cultured as described previously (Li et al., 2002). To evaluate epiblast differentiation, ES cell aggregates were grown in suspension culture for 9 days as previously described (Li et al., 2002).

Normal, PINCH1-null and Ilk-null EBs were grown for 7 days in suspension, harvested, lysed and immunoblotted as described previously (Li et al., 2002). The experiments were carried out with three different samples, the bands were measured densitometrically and the protein loading was controlled by immunostaining for tubulin.

Total RNA isolation, gel separation, blotting and hybridization were performed as previously described (Fässler et al., 1995). A PINCH1 cDNA probe (Braun et al., 2003) was random primed using standard protocols.

Immunohistochemistry of embryos or tissues and immunofluorescence studies of EBs were performed as described previously (Sakai et al., 2003). For further analysis of EB morphology, EBs were fixed in glutaraldehyde, embedded in Epon, and prepared as thick and thin sections, the former used for light microscopy and the latter for electron microscopy (described by Li et al., 2002).

We used two independent strategies to disrupt the PINCH1 gene in mice. We made a conventional targeting construct by replacing part of exon 3, the entire exon 4 and a large part of intron 5 with a neomycin cassette (+/neo) and a conditional PINCH1 null allele, in which exon 4 was flanked by loxP sites and a neo-tk cassette (Fig. 1A). The neo-tk cassette was removed by transiently expressing Cre recombinase, giving rise to ES cell clones either lacking the neo-tk cassette (+/fl) or lacking the neo-tk cassette as well as exon 4 (+/–) (Fig. 1A). The PINCH1^+/neo and PINCH^+/– ES cells were used to generate mice. Both null mutations produced the same phenotype and are therefore, not further distinguished.

Mice with a heterozygous PINCH1-null mutation appeared normal. Among 391 viable, 3-week-old offspring from heterozygous intercrosses, 144 were of wild-type and 247 of heterozygous genotypes. No homozygous mutant mice were among the progeny (Fig. 1B) indicating that loss of PINCH1 expression results in embryonic lethality.

To determine when loss of PINCH1 expression arrests development, embryos derived from heterozygous intercrosses were dissected at different developmental stages, genotyped and analysed. During midgestation (E10.5-13.5) we analysed 45 implantation chambers; 15 of them were empty and the remaining 30 contained either wild-type or heterozygous embryos. At E7.5 we analysed 78 embryo chambers, of which seven contained cell remnants that were PCR genotyped homozygous for the PINCH1-null allele (Fig. 1C,D) and 15 contained so few cells that it was not possible to genotype them. At E6.5 we sectioned 30 embryo chambers and found 21 normally developed egg cylinders (putative wild-type or heterozygous) and nine embryos with various degrees of degeneration (putative PINCH1-null) ranging from complete absence of embryos to egg cylinders infiltrated by maternal blood cells (Fig. 2A). These data together with the expected number of normal appearing PINCH1-null E3.5 blastocysts (data not shown) indicate that loss of PINCH1 expression allows normal pre-implantation development, subsequent implantation into the uterus and initiation of the decidua reaction leading to the formation of a normal number of embryo chambers. PINCH1-null embryos, however, deteriorate rapidly after implantation. In order to test whether mutant embryos underwent apoptosis we performed TUNEL assays on sections derived from E5.5 and E6.5 embryos. While normal embryos showed no TUNEL-positive signals, sections of putative PINCH1-null embryos exhibited many TUNEL-positive signals (Fig. 2B).

To obtain PINCH1^–/– ES cells we re-electroporated two PINCH1^+/– ES cell clones with the floxed PINCH1 construct. ES cell clones with a recombination event on the wild-type allele were treated with Cre to delete the PINCH1 gene (Fig. 1A). Homozygous inactivation of the PINCH1 gene was confirmed by Southern and western blotting (data not shown).

To investigate the mechanism underlying the PINCH1 phenotype, we generated EBs and monitored formation of the endoderm, the embryonic BM, the epiblast and a proamnionic-like cavity by light microscopy of live and sectioned EBs for up to 12 days. In wild-type EBs, the endoderm started to form in 2 days as a single layer of tightly associated squamous cells (primitive endoderm), which gradually became cuboidal in morphology (visceral endoderm). The epiblast developed as a second layer of elongated, pseudostratified epithelial cells in 5-7 days and the remaining ICM cells underwent apoptosis and created a sharply defined central cavity (Fig. 3A and Fig. 7A). By contrast, the endoderm formed 1 day later in PINCH1-null EBs and often developed into a loosely associated, multilayer of cells that sometimes detached from the BM (Fig. 3B,F, Fig. 4B and Fig. 7C). The PINCH1-null epiblast was not polarized and consisted of multiple layers of polygonal cells that failed to elongate. In addition, the cavitation was some times incomplete (Fig. 4B). After 9 days in culture, most PINCH1-null EBs began to degenerate by an apoptotic mechanism, while no significant deterioration occurred in the normal EBs cultured for 12 days (data not shown).

To elucidate the spatial relationship between PINCH1, integrins and BM, EBs were immunostained with different antibodies. In wild-type EBs, PINCH1 and β1 integrin were expressed in endoderm and epiblast and concentrated at the BM zone co-localizing with laminin γ1 (Fig. 3C,E). PINCH2 was not expressed in either normal or PINCH1-null EBs (data not shown). Immunostaining of laminin γ1 and other BM components revealed a linear deposition beneath the PINCH1-null endoderm indicating that PINCH1 is not required for BM expression and assembly (Fig. 3D and data not shown). Expression and localization of β1 integrin was visible at the BM zone of normal endoderm and epiblast (Fig. 3E). By sharp contrast, β1 integrin localized to the subcortical area of PINCH1-null endoderm and epiblast (Fig. 3F) indicating that PINCH1 is required for β1 integrin recruitment to cell-matrix adhesion sites.

Rhodamine-phalloidin staining of wild-type EBs showed the actin cytoskeleton in the apical belt and along the lateral plasma membranes (Fig. 3G). In PINCH1-null EBs, the apical actin belt assembled, although it was less prominent. Interestingly, actin aggregates were visible adjacent to the basal plasma membranes (arrowheads in Fig. 3H) and parallel actin fibres were absent along the apicobasal axis.

To further characterize the epiblast polarity defects, EBs were immunostained with different antibodies. GM130 staining showed that the Golgi apparatus of normal epiblast repositioned from the perinuclear region to the apical junctional region with radial orientation (Fig. 4A). While most PINCH1-null epiblast cells retained their Golgi apparatus in the perinuclear region, some showed apical distribution in a nonlinear fashion (Fig. 4B). E-cadherin was enriched at the apices of normal epiblast (Fig. 4C). Although the apical accumulation of E-cadherin was less evident in PINCH-null EBs (Fig. 4D), electron microscopy revealed that adherens junctions formed between the cells facing the incomplete cavities (data not shown). The tight junction (TJ) marker ZO-1 was expressed exclusively by endodermal cells and redistributed from the lateral plasma membranes to the apices to form a punctate pattern that correlated to the ultrastructural TJs (Fig. 4E and Fig. 7G). In PINCH1-null endoderm, ZO-1 often remained in the lateral and sometimes even in the basal plasma membranes (Fig. 4F and Fig. 7H), suggesting a defect in TJ maturation and/or localization.

Since PINCH1 can modulate proliferation and survival of cells in vitro we investigated these processes during EB development. Ki67 staining revealed similar numbers of proliferating cells in wild-type and PINCH1-null EBs (Fig. 5A,B). Apoptotic cells were detected in the centre of normal EBs where the cavity formed (Fig. 5C), and occasionally in the epiblast but never in the endoderm. By contrast, apoptosis was prominent in the PINCH1-null endodermal cells (Fig. 5D). Immunostaining for phosphorylated Ser473 of PKB/Akt was visible throughout the cytoplasm of only a few cells of the endoderm or epiblast both in wild-type and PINCH1-null EBs (Fig. 5E,F). Also the number of PKB/Akt-positive cells was similar between normal and mutant EBs.

It has been reported that the stability of ILK and PINCH is interdependent (Fukuda et al., 2003). To test whether this is also the case in EBs, we performed western blot analysis with lysates obtained from PINCH1- and Ilk-null EBs. We found that ILK is threefold reduced but not absent in PINCH1-null EBs and vice versa, PINCH1 is threefold reduced but not absent in Ilk-null EBs (Fig. 6A). To test whether the remaining ILK is the result of the expression of an N-terminal PINCH1 peptide (from exon 1 to exon 3), which can be translated after Cre-mediated deletion of exon 4 of the PINCH1 gene, we isolated total RNA and performed a northern blot assay. Fig. 6B shows that deletion of exon 4 led to a decay of the mutant mRNA indicating that PINCH1 mutant EBs express little or no N-terminal PINCH1 peptide and that the reduced ILK expression is stabilized by a PINCH1-independent manner. In agreement with the western blot data, immunostaining confirmed reduced expression of ILK in PINCH1-null EBs. In wild-type EBs ILK was enriched at the BM zone of endoderm and epiblast (Fig. 6C). In PINCH1-null EBs ILK expression levels were reduced but not absent. Furthermore, ILK localized to the subcortical region of endodermal cells and epiblast-like cells (Fig. 6C). The specificity of the anti-ILK antibody was evaluated by staining Ilk-null EBs, which revealed no immunosignal (Fig. 6D).

During the development of normal EBs the primitive endoderm appears as a single layer of tightly associated squamous cells, followed by the differentiation of the elongated pseudostratified epiblast cells and the apoptosis of the remaining ICM cells, giving rise to a central cavity (Fig. 7A). In PINCH1-null EBs the endoderm formed but often detached from the BM (Fig. 7B,C), the epiblast failed to polarize and consisted of multiple layers of polygonal cells, and the cavitation was sometimes incomplete (Fig. 7C). Transmission electron microscopy of normal EBs showed that endodermal cells tightly adhered to the BM, had a cuboidal shape, numerous microvilli at their apices (Fig. 7D) and TJs symmetrically assembled at the apical contacts of two cells (Fig. 7G). The normal epiblast consisted of a single layer of pseudostratified columnar cells. By contrast, PINCH1-null endodermal cells were round, connected to each other by thin plasma membrane extensions, contained fewer microvilli and were often detached from the BMs (Fig. 7C,E,F). TJs were often observed in the middle or basal portion of cell-cell contacts and were occasionally distributed asymmetrically (Fig. 7H). The PINCH1-null epiblast had multiple layers of polygonal cells with different staining intensity in Methylene Blue-stained thick sections (Fig. 7B,C,E), suggesting heterogeneity within the cell population.

Cell culture and biochemical studies demonstrated that the interaction of PINCH1 with ILK is a prerequisite to locate both proteins to integrin-mediated adhesion sites and to prevent their proteasome-mediated degradation (Zhang et al., 2002c; Fukuda et al., 2003). These findings led to the hypothesis that PINCH1 and ILK are obligate partners in controlling cell behaviour such as cell shape modulation, motility and survival (Fukuda et al., 2003) and imply that ablation of either gene in vivo would lead to very similar, if not identical, phenotypes. This assumption is consistent with findings of genetic studies in worms and flies, which revealed that loss-of-function alleles of either the PINCH or the Ilk gene cause indistinguishable defects, characterized by abnormal muscle attachment and embryonic lethality (Hobert et al., 1999; Zervas et al., 2001; Mackinnon et al., 2002; Clark et al., 2003). In order to test whether this is also true for mice we disrupted the PINCH1 gene in the mouse germline and in ES cells, investigated the consequences in vivo and in ES cell-derived EBs, and compared the phenotypes with those reported for the Ilk-null mice (Sakai et al., 2003).

We generated two independent mouse strains carrying a constitutive PINCH1 null mutation and found that, similar to mice lacking the Ilk gene, they died shortly after implantation. Interestingly, however, the PINCH1-null mice proceeded further in their development than ILK-deficient or β1 integrin-deficient mice. PINCH1-null embryos died at around E6.5-7.5 and showed variable degrees of degeneration. In a few cases we were even able to isolate remains of PINCH1-deficient embryos at stages as late as E8.5 and E9.5. By contrast, Ilk-null or β1 integrin-null embryos are completely absorbed at around E5.5-6.5. Such a temporal difference in lethality suggests that the disruption of the PINCH1 gene can cause overlapping but must also different defects to the disruption of the Ilk or the β1 integrin gene in vivo. These differences are unlikely to be caused through compensation by PINCH2; the expression of PINCH2 is not detectable before E14.5-15.5 (Braun et al., 2003) and was also absent in normal as well as PINCH1-deficient EBs.

EBs have emerged as an elegant experimental system to investigate normal or pathological peri-implantation. As with the surface of the inner cell mass (ICM) of blastocysts the outermost cells of ES cell aggregates differentiate into primitive endoderm and deposit a BM, which is required to polarize the adjacent cells into the columnar epiblast and to induce apoptosis in the remaining cells, resulting in the first cavity called the amniotic cavity. Using EBs derived from mutant ES cells we were able to show in a previous report that mice lacking β1 integrin or ILK die from different causes at peri-implantation. While β1 integrin-deficient mice fail to deposit a BM between primitive endoderm and epiblast, Ilk-null mice produce a BM but fail to polarize the epiblast and to cavitate (Aumailley et al., 2000; Li et al., 2002; Sakai et al., 2003).

In agreement with our in vivo analysis, PINCH1-null EBs developed primitive endoderm, deposited a BM and formed, in the majority of cases, fully expanded cavities and hence clearly progressed much further than Ilk-null EBs. A more detailed analysis of PINCH1-null EBs using immunostaining and electron microscopy disclosed a few similarities with ILK-deficient EBs but strikingly, also additional defects. Similarly, though not as severely as in Ilk-null EBs, loss of PINCH1 expression gave rise to an abnormally polarized epiblast with aberrant arrangement of the Golgi stacks, partial inversion of F-actin deposition and diffuse distribution of E-cadherin. One possible explanation for the polarity defect could be that loss of PINCH1 is associated with loss of ILK expression. Such a phenomenon was reported for PINCH1-depleted HeLa cells and served as an explanation for the similar defects that were apparent in ILK- or PINCH1-depleted cells (Fukuda et al., 2003). Immunostaining of our PINCH1-null EBs revealed expression of ILK in the endoderm and epiblast. Quantitative measurements of ILK levels in whole EBs by western blotting demonstrated a reduction of ILK in PINCH1-null EBs and vice versa. Since PINCH2 is not expressed in EBs these data reveal that there is an interdependence of the two proteins, which is, however, limited and not complete. This notion was supported by a similar observation in PINCH1-null fibroblasts, which also displayed reduced, but not absent, ILK expression (R.B. and R.F., unpublished). The reduction of ILK levels in PINCH1-null EBs allows the drafting of several hypotheses. First, loss of ILK leads to severe defects of epiblast polarity and cavitation while the reduction of ILK in PINCH1-deficient EBs may account for the milder polarity defect, the ability to cavitate and also to develop further in vivo. Second, there may be functionally different pools of ILK and PINCH1 in cells; one pool of PINCH1 and ILK, in which the function and stability of both proteins depend on the interaction between one another, and additional pools of the two proteins, in which their stability and function is not interdependent.

Such ILK-independent, PINCH1-autonomous functions could be responsible for the additional abnormalities observed in PINCH1- but not in ILK-deficient EBs. The most prominent defects were severe detachment of the primitive endoderm from BMs, profound cell-cell adhesion defects affecting cells of the primitive endoderm and epiblast, and apoptosis of the primitive endoderm. A role of PINCH1 in promoting integrin-mediated adhesion was elegantly demonstrated in PINCH1-deficient worms and flies (Hobert et al., 1999; Clark et al., 2003) and therefore could also expected in EBs. However, it was unexpected that PINCH1 is required for cell-BM adhesion specifically by the primitive endoderm but not by the epiblast. Since neither cell type expresses PINCH2, the epiblast must have a different or an additional mechanism by which it anchors to the BM. The mechanism underlying this cell-type-specific defect is not known.

The presence and the severity of the cell-cell adhesion defects point to an important role of PINCH1 in initiating and/or maintaining cell-cell contacts of both epiblast and endoderm. The latter cells form tight junctions, which were often mislocalized or absent. It is not clear how PINCH1 promotes formation and localisation of cell adhesion junctions. One possibility is that PINCH1 is physically present in cell-cell adhesion sites. However, this has not been reported yet and our immunostaining of normal EBs revealed PINCH1 accumulation at the basal plasma membrane, diffuse staining throughout the cytoplasm but no significant enrichment at cell-cell adhesions. Another possibility is that PINCH1 is indirectly involved in the regulation of cell-cell adhesions. Such a role has been proposed for ILK, which can modulate E-cadherin expression through β-catenin-Lef/Tcf-dependent transcription and expression of the transcriptional repressor snail (Novak et al., 1998; Tan et al., 2001). It will be important to further investigate both possibilities using cells from PINCH1 mutant mice that exhibit abnormal cell-cell adhesion in vivo.

Depletion of PINCH1 in HeLa cells by RNA interference induced strong apoptosis associated with reduced activation of PKB/Akt (Fukuda et al., 2003). We also observed apoptotic cells in the endoderm of PINCH1-null EBs as well as in PINCH1-null embryos. Similar to the abnormal adhesion to BMs, this defect was cell-type-specific and was not observed in the epiblast. Although cell-ECM detachment can induce anoikis in many cell types including certain epithelial and endothelial cells (Frisch and Ruoslahti, 1997) we noticed cell death in detached and also equally frequently in attached endodermal cells. This indicates that PINCH1 prevents apoptosis in a rather direct manner and not by impairing integrin-mediated adhesion. A remaining puzzle is how PINCH1 prevents death of the primitive endoderm. The strong nuclear PINCH1 staining specifically in endodermal cells suggested a possible regulation through a nuclear function of PINCH1 (Campana et al., 2003). An alternative possibility is that deletion of PINCH1 results in aberrant regulation of PKB/Akt activity. PKB/Akt is a serine/threonine kinase, which transmits extracellular signals to downstream effectors, including caspases (Lawlor and Alessi, 2001). Phosphorylation of Thr308 by PDK1 and Ser473 by DNA protein kinase (Feng et al., 2004) and ILK (Delcommenne et al., 1998) regulates the kinase activity of PKB/Akt. Depletion of ILK in HeLa cells induced apoptosis and gave rise to reduced Ser473 and Thr308 phosphorylation, suggesting that apoptosis was at least in part induced by aberrant PKB/Akt activity (Fukuda et al., 2003). We did not observe a clear reduction of PKB/Akt Ser473 phsophorylation in PINCH1-null endoderm. It is possible that differences escaped detection by our immunostaining approach, or alternatively that PINCH1 modulates PKB/Akt-independent survival pathways in primitive endoderm. A PKB/Akt-independent survival pathway in addition to a PKB/Akt-dependent one has been proposed to operate in HeLa cells (Fukuda et al., 2003). The nature of the PKB/Akt-independent survival pathway is not known.

In conclusion we show that the PINCH1 function is similar to ILK function, in that it is rate-limiting for peri-implantation development. Although the time point of lethality supports the prevailing view that the stability and the functions of both polypeptides are interdependent we show that PINCH1 can also operate in an ILK-independent manner to ensure endoderm survival and cell-cell adhesion, and that the stability of both proteins does not solely depend on their interactions. Finally, we found that certain PINCH1 functions such as apoptosis and BM detachment occurred exclusively in mutant endoderm indicating cell-type-specificity of PINCH1 function.

We thank Werner Müller (University of Cologne, Germany) for the Cre expression plasmid, Kyle Legate and Carsten Grashoff for critically reading the manuscript. R.B. is supported by a PhD fellowship from the University of Copenhagen and O.K. is supported by an Erwin Schroedinger fellowship from the Austrian ScienceFoundation (FWF). The work was supported by grants from the NIH (R37-DK36425), the DFG (SFB 473), the Max Planck Society and the Fonds der Chemischen Industrie.