Cdc42 is crucial for the establishment of epithelial polarity during early mammalian development


  • Xunwei Wu,

    1. University of Copenhagen, Institute of Molecular Pathology, Copenhagen, Denmark
    Current affiliation:
    1. Cutaneous Biology Research Center (CBRC) Massachusetts General Hospital, Harvard Medical School, Charlestown, MA02129
    Search for more papers by this author
    • Drs. Wu and Li contributed equally to this work.

  • Shaohua Li,

    1. Department of Pathology and Laboratory Medicine, Robert Wood Johnson Medical School, Piscataway, New Jersey
    Current affiliation:
    1. Department of Surgery, Robert Wood Johnson Medical School, New Brunswick, NJ 08903
    Search for more papers by this author
    • Drs. Wu and Li contributed equally to this work.

  • Anna Chrostek-Grashoff,

    1. Max Planck Institute of Biochemistry, Department of Molecular Medicine, Martinsried, Germany
    Search for more papers by this author
  • Aleksandra Czuchra,

    1. Max Planck Institute of Biochemistry, Department of Molecular Medicine, Martinsried, Germany
    Search for more papers by this author
  • Hannelore Meyer,

    1. Max Planck Institute of Biochemistry, Department of Molecular Medicine, Martinsried, Germany
    Search for more papers by this author
  • Peter D. Yurchenco,

    1. Department of Pathology and Laboratory Medicine, Robert Wood Johnson Medical School, Piscataway, New Jersey
    Search for more papers by this author
  • Cord Brakebusch

    Corresponding author
    1. University of Copenhagen, Institute of Molecular Pathology, Copenhagen, Denmark
    • University of Copenhagen, BRIC, Ole Maaløes Vej 5, 2200 Copenhagen, Denmark
    Search for more papers by this author


To study the role of Cdc42 in the establishment of epithelial polarity during mammalian development, we generated murine Cdc42-null embryonic stem cells and analyzed peri-implantation development using embryoid bodies (EBs). Mutant EBs developed endoderm and underlying basement membrane, but exhibited defects of cell polarity, cell–cell junctions, survival, and cavitation. These defects corresponded to a decreased phosphorylation and membrane localization of aPKC, a reduced phosphorylation of GSK3β, and a diminished activity of Rac1. However, neither Rac1 nor the kinase function of GSK3β seem to contribute to cell polarization and cell–cell contacts. In contrast, EBs expressing dominant-negative (dn) PKCζ mimicked well the phenotype of Cdc42-null EBs, suggesting a major role of aPKC in mediating cell polarization downstream of Cdc42. Finally, aggregation experiments with endodermal cell lines suggested that Cdc42 might affect formation of adherens and tight junctions by PKCζ-dependent regulation of the protein levels of p120 catenin and E-cadherin. Developmental Dynamics 236:2767–2778, 2007. © 2007 Wiley-Liss, Inc.


Cell polarity is crucial for the function of virtually all cells and is involved in processes such as cell division, cell migration, and signaling (Nelson,2003). Cell polarization is initiated by intracellular or extracellular cues determining a specific site at the cell surface identified by a “landmark” protein such as cadherins at cell–cell contacts or integrins at cell–extracellular matrix contacts. These “landmarks” recruit signaling and scaffolding complexes and induce the local rearrangement of the actin and microtubule cytoskeleton. The reorganization of the cytoskeleton is controlled by small GTPases of the Rho family, which mediate their function by binding in their active, GTP-bound form to more than 60 effector molecules, including serine/threonine kinases such as PAKs and MEKK1, lipid kinases such as PI3-K, and cytoskeletal proteins such as N-WASP and IQGAP (Etienne-Manneville and Hall,2002; Jaffe and Hall,2005). Particularly the small GTPase Cdc42 was shown to be important for the establishment of different aspects of cell polarity from yeast to mammalian cells (Etienne-Manneville,2004).

Cell polarization is essential for embryonic development. During implantation, primitive endoderm forms at the surface of the inner cell mass (ICM) of the blastocyst and deposits an underlying basement membrane (BM). This BM induces the polarization of the adjacent ICM cells to form the primitive ectoderm. Cells not in contact with the BM undergo apoptosis, resulting in the establishment of the proamniotic cavity (Li et al.,2003). Previous work revealed that at embryonic day (E) 5.5 presumptive Cdc42-deficient embryos were smaller than normal, disorganized in structure, and lacked normal primary ectoderm (Chen et al.,2000). At E6.5, these embryos had largely degenerated. The detailed structure of Cdc42-null embryos, epithelial cell polarity, the presence of a normal BM, however, were not studied. Conditional knockout of Cdc42 in keratinocytes in vivo did not have any immediate effect on epithelial cell polarity. Only after several months adherens junction proteins were decreased due to an increased degradation of β-catenin (Wu et al.,2006a) and subtle defects in the maintenance of the BM were observed (Wu et al.,2006b). In neural progenitors in brain, on the other hand, loss of Cdc42 caused a quick disappearance of adherens junctions (Cappello et al.,2006). Whereas these data indicate cell-type specific roles of Cdc42 in the maintenance of adherens junctions, it is not known whether epithelial cell polarity including adherens and tight junctions and BM can be established at all if Cdc42 is absent.

To date, in vitro data could not confirm a crucial role of Cdc42 in the establishment of epithelial polarization, because inhibition of Cdc42 function in MDCK cells did not interfere with tight junction formation after transient loss of cell–cell interaction (Takaishi et al.,1997; Kodama et al.,1999; Kroschewski et al.,1999; Gao et al.,2002). Moreover, it was demonstrated that the Rac1-specific guanine nucleotide exchange factor Tiam1 triggers tight junction assembly in keratinocytes independently of Cdc42 through activation of Rac1 and PKCζ (Mertens et al.,2005).

Yet, a general problem of these studies was that Cdc42 or Rac1 functions were repressed by dominant-negative mutant forms of these proteins, which is not specific and dependent on the level of expression (Czuchra et al.,2005; Vidali et al.,2006). The specific importance of Cdc42 in the establishment of epithelial cell polarity in mammalian cells, therefore, is not clear. To clarify this issue, we tested now to which extent epithelial cell polarity can develop in undifferentiated cells that specifically and completely lack Cdc42.

Establishment of epithelial cell polarity during peri-implantation development can be modeled by aggregation of embryonic stem (ES) cells in embryoid bodies (EBs) in suspension (Li et al.,2003). In EBs, the outermost cells differentiate to endodermal cells that form an underlying BM. ICM cells attached to the BM become columnar epiblast epithelium, while cells not in contact with the extracellular matrix (ECM) undergo apoptosis, resulting in the formation of the proamniotic cavity.

Analyzing EB formation of Cdc42-null ES cells, which lack a functional Cdc42 gene, we show now that Cdc42 is essential for the establishment of apical cell polarity of the actin cytoskeleton in both endoderm and epiblast (primordial ectoderm that gives rise to the three germ layers) and facilitates formation of adherens and tight junctions. Similar defects were observed in EBs expressing dnPKCζ, but not in Rac1-null EBs, suggesting that Cdc42-dependent regulation of aPKC is mediating these effects, while Rac1 cross-activation by Cdc42 is not important. Furthermore, we demonstrate that Cdc42 is not required for BM formation, but rather is involved in mediating BM-dependent polarization and survival during early development.


Generation and Characterization of Cdc42-Deficient ES Cells

Heterozygous Cdc42 +/− ES cells generated by gene targeting (Wu et al.,2006a) were re-electroporated with the targeting vector and with a Cre expression plasmid to yield heterozygous fl/− and homozygous −/− Cdc42-null ES cells. Deletion of exon 2 was confirmed by polymerase chain reaction (PCR) and Southern blot and loss of protein expression was demonstrated by Western blot using two antibodies raised against the entire Cdc42 protein or the C-terminal region of Cdc42 (Fig. 1A–C). Heterozygous fl/− cells were still capable of colonizing the germ line, indicating that the prolonged culture did not impair ES cell properties.

Figure 1.

Characterization of Cdc42-null embryonic stem (ES) cells. A: Southern blot of genomic DNA of Cdc42 fl/− ES cells after a second electroporation with the targeting construct before and after transfection of Cre expression vector. Before Cre recombination (left panel): ES cell DNA was digested with EcoRI and hybridized with an external probe (+, 12.0 kb; − or fl neo, 5.2 kb). After Cre recombination (right panel): ES cell DNA was digested with XbaI and BamHI and hybridized with an internal probe (fl, 2.8 kb; −, 2.3 kb). B: Genomic polymerase chain reaction of Cdc42 fl/− and Cdc42−/− ES cell clones (fl, 0.6 kb; −, 0.15 kb). C: Western blot analysis of lysates of ES cell clones. Cdc42 protein was detected by a polyclonal (poly) and a monoclonal (mono) antibody in lysates of wild-type (+/+) and heterozygous (fl/−) ES cells, but not in three independent −/− clones. D: Phase contrast picture of fl/− and −/− ES cell colonies on feeder cells. E: Detection of active (act) and total (tot) Rho GTPases in normally growing ES cells by pull-down assay and Western blot. Cdc42−/− ES cells showed decreased amounts of active Rac1, but unchanged levels of active RhoA. Total expression levels of Rac1 and RhoA were unchanged. F: Western blot analysis of lysates of adherent ES cells for phosphorylated (phosph) and total (tot) levels of Akt (Ser 473), Gsk3β (Ser 9), PKCζ, Erk1/2, and p38 revealed no change between control and mutant ES cells.

Mutant Cdc42 (−/−) ES cell clones showed a distinct morphology with round cell bodies protruding from the normally smooth surface of an ES cell colony (Fig. 1D). Proliferation of mutant and control cells was not significantly different as assessed by cell cycle analysis using bromodeoxyuridine (BrdU) incorporation and propidium iodide staining of nuclear DNA (Control: G1 26.4 ± 1.2%, S 43.5 ± 3.4%, G2/M 30.1 ± 2.9%; Cdc42−/−: G1 27.7 ± 2.4%, S 39.5 ± 4.8%, G2/M 32.8 ± 2.5%; n = 4). Loss of Cdc42 resulted in a reduced amount of activated Rac1, whereas activation of RhoA was not affected (Fig. 1E). The expression levels of Rac1 and RhoA were not changed in mutant ES cells. Assessing the activity of Cdc42 regulated signaling pathways in adherent, undifferentiated ES cells under normal growth conditions, we could not detect changes in the phosphorylation or expression levels of Akt, GSK3β, PKCζ, Erk, and p38 (Fig. 1F).

Cdc42-Null EBs Have Nonpolarized Epiblast Cells and Impaired Cavitation, but Can Form a BM

In mice, presumably Cdc42-deficient embryos show endodermal differentiation, but neither formation of polarized epiblast nor cavitation (Chen et al.,2000). We wanted to study this process in more detail by differentiating Cdc42-deficient ES cells in EBs.

After 7 days in suspension, approximately 50% of the control EBs (Cdc42 fl/− or +/+) showed endoderm, columnar ectoderm, and cavitation (Fig. 2A). Mutant EBs displayed morphologically normal endoderm characterized by flattened, often vacuolated cells, but never columnar ectoderm and only impaired cavity formation (Fig. 2A′), closely mimicking the morphology of Cdc42-null embryos at E5.5 (Chen et al.,2000). Instead of a pseudostratified, columnar epithelium, mutant epiblast cells had a polygonal shape similar to the undifferentiated ICM cells.

Figure 2.

Cdc42-null embryoid bodies (EBs) display impaired formation of polarized ectoderm and cavity. A,A′: Semithin sections of 7-day Cdc42 fl/− and −/− EBs. Cdc42 fl/− EBs showed endoderm (en), a polarized epiblast (ep) and a cavity (cv). Cdc42−/− EBs had an endoderm, but did not show polarization of the inner cell mass (icm). Although an apoptotic zone (az) was discernible, no normal cavitation occurred. Basement membrane-like material (bm) was observed below the endodermal layer in control and mutant EBs. In addition, Cdc42-null EBs displayed also internal deposition of extracellular matrix. B: Semiquantitative reverse transcription-polymerase chain reaction of ES cells and 7-day EBs was carried out for the endodermal marker AFP, the early ectodermal marker bone morphogenetic protein-4 (BMP-4), the mesodermal marker ζ-globin, and the housekeeping gene GAPDH. Although AFP, BMP4, and GAPDH were detected in mutant and control EBs, ζ-globin was only amplified from control EBs. Mutant and control ES cells, both, lacked AFP and ζ-globin, had low levels of BMP-4, and high levels of GAPDH. C,C′: Immmunofluorescence of 7-day EBs for F-actin, laminin α1 (LNα1), and activated cleaved caspase 3. Normal EBs showed apical distribution of F-actin in endodermal and ectodermal cells. In contrast, unpolarized cortical F-action was found in Cdc42−/− EBs. Despite an impaired cavitation of Cdc42-deficient EBs, caspase-3 is activated in most inner cell mass cells. Also cells in contact with the BM are positive for activated caspase-3 (arrowheads).

To test the differentiation status of endoderm and epiblast cells, we performed semiquantitative reverse transcription (RT) -PCR to detect specific differentiation markers. At day 7, mutant and control EBs (wild-type and heterozygous) expressed mRNA of the endodermal marker AFP and the early, precavitation epiblast marker bone morphogenetic protein-4 (BMP-4; Fig. 2B). Expression of the mesodermal marker ζ-globin, however, could only be detected in control, but not in mutant EBs (Fig. 2B). Control or mutant ES cells had undetectable levels of AFP and ζ-globin, and very low amounts of BMP-4 were found (Fig. 2B).

Defective cavitation of mutant EBs was not caused by lack of apoptosis, because most mutant epiblast cells showed high levels of cleaved caspase 3 at day 7 (Fig. 2C′). Furthermore, 4′,6-diamidine-2-phenylidole-dihydrochloride (DAPI) staining and electron microscopy (EM) often revealed fragmented nuclei, another hallmark of apoptotic cell death (Fig. 3F; Fig. 4D′; Supplementary Figure S1A′,C′, which can be viewed at After day 7, progressive decay of the mutant EBs occurred, while control EBs displayed a normal morphology at least until day 9 (data not shown).

Figure 3.

Normal basement membrane (BM), but severely reduced adherens and tight junction formation in the absence of Cdc42. A–H: Ultrastructural analysis of 7-day Cdc42 fl/− (A–C) and −/− (D–H) embryoid bodies (EBs; en, endoderm; ep, epiblast; cv, cavity; az, apoptotic zone; sc, secretory cell). BM is formed in control (A, arrows) and mutant (D, arrows) EBs. Occasionally, Cdc42−/− EBs also showed internal BM fragments (E, arrows) surrounded by undifferentiated cells. Generally, Cdc42-null EBs lacked well-formed endodermal tight junctions (D, stars in lower inset) and inner cell mass (ICM) adherens-type junctions (F, stars), which were readily detected in control EBs (tight junctions: A, arrows in insert; adherens junctions: B, arrows). However, few adherens and tight junctions were found in mutant EBs (tight junction: D, arrow in upper inset; adherens junction: H, arrow). The adherens junctions, when seen were located adjacent to the apoptotic zone. H shows an instance in which an adherens junction is present at one cell–cell contact (arrow) but not the other (star). Sometimes, the null EBs contain secretory cells (sc) with prominent/dilated endoplasmic reticulum, which normally is only seen in the endoderm, deep inside the EB adjacent to an apoptotic zone (G), which might represent endoderm in inappropriate locations.

Figure 4.

Unpolarized distribution of adherens and tight junction molecules in Cdc42-deficient embryoid bodies (EBs). A–D′: Immunofluorescent staining of 7-day Cdc42 fl/− (A–D) and Cdc42−/− EBs (A′–D′) for E-cadherin, β-catenin, α-catenin, and ZO-1 as indicated. Basement membrane was identified by laminin α1 (LNα1) and nuclei by 4′,6-diamidine-2-phenylidole-dihydrochloride (DAPI). Cdc42 fl/− EBs show strongly polarized distribution of E-cadherin, β-catenin, and α-catenin toward the apical side of the ectoderm (arrowheads). No polarization of inner cell mass (ICM) was observed in Cdc42−/− EBs. D′: Whereas the ZO-1 was clearly detectable between endodermal cells of control EBs, it was poorly expressed and diffusively deposited in mutant EBs. Scale bars = 20 μm

Whereas in endodermal cells of control EBs the Golgi complex was often located between the nucleus and the basal side, it showed no preferred localization in mutant EBs, which could indicate an impaired polarized secretion (Supplementary Figure S1A,A′). Surprisingly, however, BM components laminin α1, collagen IV, nidogen-1, and perlecan were deposited normally below the endodermal cells in mutant EBs (Fig. 2C,C′ and Supplementary Figure S1B–D,B′–D′). The presence of a BM corresponded to a polarized distribution of α6 integrin at the BM zone both in control and mutant EBs (Supplementary Figure S1E,E′). Ultrastructural analysis confirmed the formation of a subendodermal BM both in control and mutant EBs (Fig. 3A,D). However, BM thickness varied more often in Cdc42 mutant EBs, which could indicate a subtle defect in the establishment of a normal BM. In a few mutant EBs, stretches of BM could also be found at internal sites, surrounded mostly by morphologically nonpolarized cells but often seen with a single ectopically located secretory cell with prominent endoplasmic reticulum cisternae (Figs. 2A′,C′, 3E,G). In contrast to normal epithelial cells, BM attachment of Cdc42-null cells seemed not necessarily to prevent cell death of epiblast cells, because several cleaved caspase-3–expressing cells were found in the direct vicinity of the BM (Fig. 2C′, arrowheads).

Cdc42 Facilitates the Establishment of Adherens and Tight Junctions

To test the role of Cdc42 in the establishment of cell–cell junctions in early development, we performed immunofluorescence light and EM. In both control and mutant EBs, β-catenin, α-catenin, and F-actin were found to be colocalized with E-cadherin close to the plasma membrane, suggesting a connection of E-cadherin with the actin cytoskeleton (Figs. 2C,C′; 4A–C,A′–C′). The distribution of the adherens junction proteins, however, was different. In 7-day-old control EBs, E-cadherin, α-catenin, β-catenin, and F-actin were concentrated toward the apical side of the epiblast, while mutant epiblast cells showed a uniform membranous distribution of these molecules (Figs. 2C′; 4A′–C′). In control EBs, adherens junctions formed close to the apical side of the ectoderm, which faces the cavity (Fig. 3B,C, arrows). Adherens junctions were mostly missing between epiblast cells of mutant EBs (Fig. 3F, stars). Occasionally, adherens junctions did form between epiblast cells adjacent to apoptotic foci in the absence of Cdc42 (∼10-fold less; Fig. 3H, arrow). In that case, however, no adherens junctions were detected at the neighboring cell–cell contacts (Fig. 3H, star), indicating that the adherens junction belt was not complete and instead corresponded to “spot-weld” junctional complexes.

In 7-day-old normal EBs, tight junctions were formed between endodermal cells as indicated by ultrastructural analysis and expression of ZO-1, which is crucial for the normal kinetics of tight junction assembly (Umeda et al.,2004; McNeil et al.,2006; Figs. 3A, inset, arrow, 4D). Cdc42-deficient endodermal cells expressed the ZO-1 at reduced levels that were deposited rather diffusely (Fig. 4D′). EM, however, revealed that, although tight junctions were mostly absent (Fig. 3D, lower inset, star), they occasionally formed between Cdc42-null endodermal cells (∼sixfold less; Fig. 3D, upper inset, arrow). Also here, no continuous tight junction belt was detected.

Rac1 Is Not Required for the Establishment of Cell Polarity and Cell–Cell Contacts in EBs

Because loss of Cdc42 leads to a significant reduction in Rac1 activity (Fig. 1E), we wanted to test to which extent this effect contributes to the impaired cell polarization and cell–cell contacts. We, therefore, generated Rac1-null ES cells (Supplementary Figure S2A,B,B′) and differentiated them in EBs. Rac1-null EBs formed endoderm, BM, epiblast, and showed normal cavitation (Fig. 5A,A′), while ectoderm was often thickened and several epiblast cells attached to the BM underwent cell death (Fig. 5A,A′; Supplementary Figure S2C′–E′, stars indicating dead cells in contact to BM). This phenotype is similar to the one of murine Rac1-null embryos (Sugihara et al.,1998).

Figure 5.

Normal formation of polarized cell–cell junctions in Rac1−/− embryoid bodies (EBs). A,A′: Phase contrast pictures of 7-day EBs show endoderm (en), epiblast (ep), and cavity (cv) in Rac1−/− and Rac1 fl/− EBs. B–C′: Immunofluorescent staining of E-cadherin (B,B′) of 7 day EBs and ZO-1 (C,C′) of 3-day EBs. Basement membrane was identified by laminin α1 (LNα1) and nuclei by 4′,6-diamidine-2-phenylidole-dihydrochloride (DAPI). E-cadherin and ZO-1 were polarized toward the apical side of endoderm and epiblast both in control (Rac1 fl/−) and mutant (Rac1−/−) EBs. D,D′: Ultrastructural analysis of 3-day EBs. Tight junctions were detected between endoderm cells (en) of both the Rac1 fl/− and −/− EBs (arrows; arrowhead in insert). Scale bar = 2 μm

In the epiblast, staining for E-cadherin was present at the cell membrane and increased toward the apical side (Fig. 5B,B′) and ZO-1 was readily detectable between endodermal cells lacking Rac1 (Fig. 5C,C′). Ultrastructural analysis confirmed the presence of normal, belt-like tight junctions between endodermal cells in Rac1-deficient EBs (Fig. 5D,D′). F-actin, Golgi complex (GM130), and microtubule organization center (MTOC; pericentrin) showed similar polarized distribution in control and mutant EBs (Supplementary Figure S2C–E,C′–E′). These data indicate that Rac1 is not required for cell polarization and cell–cell contacts during early embryonic development.

aPKC, but not GSK3β Contributes to the Establishment of Cell–Cell Junctions and Cell Polarity

Activity of aPKC, which has the isoforms PKCζ and PKCλ, is crucial for the maturation of tight junctions in MDCK cells (Suzuki et al.,2001,2002). Because Cdc42 regulates aPKC activation (Qiu et al.,2000; Yamanaka et al.,2001; Wu et al.,2006a), we wondered whether aPKC activity is also altered in Cdc42 null EBs. Indeed, both immunostaining and Western blotting indicated a significant reduction of phosphorylated aPKC (pPKCζ/λ) in Cdc42-null EBs, while total levels of PKCζ were not changed, indicating decreased aPKC activation (Fig. 6A,B,A′,B′,C). In contrast to controls, pPKCζ/λ and PKCζ were, furthermore, not polarized to the apical side of endoderm and epiblast cells of Cdc42-null EBs, where the junctional complexes are located (Fig. 6A,B, arrows, and A′,B′).

Figure 6.

Cdc42 regulates aPKC activity, which affects cell–cell junctions and cell polarity. A–B′: Immunofluorescent staining for total PKCζ and phosphorylated PKCζ/λ (pPKCζ/λ) in 7-day Cdc42 fl/− and Cdc42−/− embryoid bodies (EBs) indicated an apical localization of PKCζ and pPKCζ/λ in endoderm and ectoderm of control EBs (arrowheads), but not in Cdc42−/− EBs. Basement membrane was identified by staining for laminin γ1 (LNγ1). C: Western blot analysis of lysates from 7-day EBs showed a strong reduction of pPKCζ/λ in Cdc42−/− EBs, but no change in the total level of PKCζ. D–F′: Immunofluorescent staining for F-actin, E-cadherin, and Par3 in 7-day Cdc42 fl/− EBs with (D′–F′) or without (D–F) dominant PKCζ (dnPKCζ) expression. Polarized distribution of F-actin, E-cadherin, and Par3 were observed in Cdc42 fl/− EBs (arrowheads), but not in Cdc42 fl/− EBs with dnPKCζ expression. BM was identified by staining for laminin α1 (LNα1). Scale bars = 20 μm.

Could the reduced aPKC activation contribute to the observed defects in cell polarity and cell–cell junctions? To test this question, we overexpressed dnPKCζ in ES cells (Supplementary Figure S3A) and differentiated them in EBs (Supplementary Figure S3B). Due to the high similarity of PKCζ and PKCλ, dnPKCζ inhibits both isoforms of aPKC (Suzuki et al.,2001). The dnPKCζ EBs developed defects similar to Cdc42-deficient EBs: Loss of apical polarization of F-actin, and E-cadherin, of basal location of Golgi complex (GM130) in endodermal cells and of apical localization of Golgi and MTOC (pericentrin) in epiblast (Fig. 6D′,E′ and Supplementary Figure S3B′–D′). Furthermore, we observed loss of Par3 localization at the apical membrane of the epiblast cells (Fig. 6F,F′). EM revealed incomplete tight (dnPKCζ: 56.3% [13 of 23]; control: 100% [50 of 50]; n = 3/3) and adherens junction belts (dnPKCζ: 13.4% [9 of 67]; control 97.5% [77 of 79]; n = 3/3) in endoderm and epiblast, similar to Cdc42-null EBs (Fig. 7A,A′,B,B′; arrows indicating tight [A,A′] or adherens junctions [B], arrowheads marking cell–cell contacts without tight or adherens junctions [A′,B′]). These data show that the decreased activation of aPKC in the absence of Cdc42 is quite likely responsible for the impaired cell polarization and the defective cell–cell junctions.

Figure 7.

Defective formation of adherens and tight junctions in embryoid bodies (EBs) with stable expression of dnPKCζ. A–B′: Ultrastructural analysis of 7-day Cdc42 fl/− (A,B) and Cdc42 fl/− dnPKCζ (A′,B′) EBs (BM, basement membrane; en, endoderm; epi, epiblast; cv, cavity; az, apoptotic zone; sc, secretory cell). Cdc42 fl/− dnPKCζ EBs often lacked well-formed endodermal tight junctions at both sides of a cell (A′, arrows at tight junctions, arrowheads indicating absence of tight junctions). In most cases, no adherens junctions were detected at the cell–cell borders of epiblast cells in Cdc42 fl/− dnPKCζ EBs (B′, arrowheads), in contrast to control (B, arrows at adherens junctions). Scale bar = 2 μm.

Decreased activation of aPKC due to inactivation of Cdc42 can lead to decreased, inhibitory phosphorylation of GSK3β at serine 9, which affects the degradation of β-catenin in keratinocytes and the MTOC polarization in migrating astrocytes (Etienne-Manneville and Hall,2003; Wu et al.,2006a). In Cdc42-null EBs, phosphorylation of GSK3β at serine 9 is significantly reduced in immunofluorescent staining, both in endoderm and epiblast, indicating increased kinase activity of GSK3β (Supplementary Figure S4A, arrowheads, S4A′). To assess whether enhanced phosphorylation by GSK3β adversely affects cell–cell contacts or cell polarization during early development, we tried to rescue the phenotype of Cdc42-null EBs by overexpressing dnGSK3β (Supplementary Figure S4B). Despite high expression levels of dnGSK3β in ES cells, the phenotype of Cdc42 (−/−) dnGSK3β EBs was identical to that of Cdc42-null EBs with respect to morphology and unpolarized distribution of E-cadherin, F-actin, Golgi complex and MTOC (Supplementary Figure S4C–G).

To test whether GSK3β kinase activity plays a role at all in embryoid body formation, we investigated EB formation of normal ES cells in the presence of the GSK3β kinase inhibitor LiCl (20 mM). We did not observe any apparent change in endodermal or epiblast polarity, as tested by the distribution of Golgi (CM-130), F- actin, and ZO-1 (Supplementary Figure S5).

These data strongly suggest that GSK3β kinase activity is not necessary for normal polarization and junction formation in EBs and that the observed polarization defects in Cdc42-null EBs are independent of the increased GSK3β activity.

Decreased Protein Amounts of p120-Catenin and E-Cadherin in Aggregating Cdc42-Null Endodermal Cells

Because EBs consist of different amounts of endodermal, ectodermal, alive, and dead cells, it is difficult to compare mutant and control EBs biochemically and only severe differences will be detected. To detect perhaps more subtle changes associated with the loss of Cdc42, we forced ES cell to differentiate into the endodermal lineage by stably transfecting them with the transcription factor GATA-4 (Czuchra et al.,2005). Both, Cdc42-null and control cells, could aggregate in suspension with similar kinetic. However, Cdc42-deficient endodermal cell aggregates tended to be less compact than control aggregates (Fig. 8A, arrows). Already after 2 hr in suspension, PKCζ phosphorylation is strongly increased in normal endodermal cells. E-cadherin expression is at that time point lower than in adherent cells, but strongly increases during aggregation up to 72 hr. Cdc42-deficient endodermal cells displayed during aggregation decreased amounts of phosphorylated PKCζ and phosphporylated GSK3β, while the total levels of these proteins were not altered (Fig. 8B). Normal levels were observed of IQGAP and of activated and nonactivated p38, Akt, and Erk (Fig. 8B). Unexpectedly, Cdc42-null endodermal aggregates showed reduced protein amounts of E-cadherin and p120-catenin, while β- and α-catenin were unchanged (Fig. 8B). This impairment could contribute to the defective formation of cell–cell junctions in Cdc42-null EBs. Yet, in 7-day EBs, we could not detect reduced levels of E-cadherin or p120 catenin (Supplementary Figure S6), suggesting either too subtle defects to be detected the complex EB mixture (s. a.) or kinetic differences.

Figure 8.

Reduced protein amounts of E-cadherin and p120 catenin in Cdc42−/− endodermal cells. A: Phase contrast pictures show the morphology of Cdc42 fl/− and Cdc42−/− endodermal cells in adherent culture and at different time points in suspension. Cdc42−/− cells formed less-compact aggregates (arrows) compared with control (arrowheads). B: Western blot analysis of lysates from Cdc42fl/− and Cdc42−/− endodermal cells in adherent cultures and at different time points in suspension. Cdc42−/− endodermal cells in suspension showed reduced protein amounts of E-cadherin, p120 catenin, pGSK3β, and pPKCζ, but unchanged levels of IQGAP1, β-catenin, α-catenin, GSK3β, PKCζ, p-p38, p38, pAKT, AKT, pErk, and Erk. C: Western blot of lysates of endodermal cells treated with 10 μM of the PKCζ inhibitor Gö6983 in adherent cultures and at different time points in suspension. Gö6983-treated cells showed in suspension reduced protein amounts of E-cadherin and p120 catenin, similar to Cdc42-null cells.

To test whether the reduction of E-cadherin and p120 catenin in aggregating Cdc42-null endodermal cells is dependent on the reduced activation of PKCζ, we treated normal endodermal cells during aggregation with the PKCζ inhibitor Gö6983. Indeed, in the presence of Gö6983, there was a reduced increase of E-cadherin expression and a reduction of p120 catenin expression, exactly as observed in Cdc42-null endodermal cells (Fig. 8C). These data show that, in suspension, Cdc42 mediates activation of PKCζ, which is required for up-regulation of E-cadherin expression and maintenance of p120 expression during aggregation.


During embryogenesis, initially undifferentiated and unpolarized cells undergo differentiation and establish cell polarity. To understand the role of Cdc42 in these processes, we mimicked early embryonic development by differentiation of Cdc42-null ES cell in EBs. We found an important role for Cdc42 in the establishment of several aspects of cell polarity, which was independent of Rac1 and conceivably caused by a reduced activation of aPKC.

Cdc42 and Epithelial Polarization

In the absence of Cdc42, EBs showed severe defects in cell polarization. First, filamentous actin was homogenously distributed in a cortical manner, while control cells showed a clear accumulation of F-actin at the apical side. Second, whereas MTOC and Golgi complex are normally located between nucleus and basal (endoderm) or apical (epiblast) cell border, both polarized distributions were lost in the absence of Cdc42. Third, adherens junction proteins were distributed homogenously at membrane of Cdc42-null epiblast cells and not concentrated toward the apical cell border. Finally, Cdc42-null endodermal cells and epiblast cells facing the apoptotic zone established only punctuate cell–cell contacts containing adherens and tight junction markers, but no continuous junctional belt separating the apical cell membrane from the basolateral. Of interest, α6 integrin, was enriched at the basal side of Cdc42-null epiblast cells, similar to controls, indicating that not all aspects of cell polarity are regulated by Cdc42.

The reduced phosphorylation of PKCζ/λ and the loss of PKCζ at the cell membrane in the absence of Cdc42 suggested that Cdc42 is crucial for the activation of aPKC and for its translocation to the cell membrane, most likely in form of a Cdc42-GTP-Par6-Par3-aPKC complex as suggested by the colocalization of active PKCζ with Par3 at the apical membrane of the epiblast. This hypothesis is strongly supported by the similar polarization defects of EBs made from ES cells overexpressing dnPKCζ, which inhibits both aPKC isoforms PKCζ and PKCλ (Suzuki et al.,2001). There we observed impaired polarization of actin and microtubule cytoskeleton, of adherens junction molecules in epiblast cells, and interrupted cell–cell junction belts. Already earlier important functions of aPKC in the formation of cell–cell contacts were reported. Overexpression of dnPKCζ or dnPKCλ in MDCK cells resulted in disturbed junctional localization of ZO-1 during reformation of cell–cell contacts after transient low calcium treatment (Suzuki et al.,2001). In a wounding assay with MTD1-A cells, aPKC inhibition by dnPKCλ caused punctuate instead of belt-like cell–cell contacts (Suzuki et al.,2002). Furthermore, peptide inhibition of PKCζ in inner cell masses isolated from early blastocysts was shown to interfere with tight junction assembly (Eckert et al.,2005). Of interest, PKCζ activation was not altered in adherent ES cells in the absence of Cdc42, but strongly reduced in Cdc42-null EBs. Similarly, PKCζ phosphorylation was normal in Cdc42-null adherent endodermal cells, but did not increase in suspension, as it occurred in normal endothelial cells. These data suggest that Cdc42 is crucial for the increased activation of PKCζ in nonadherent ES and endodermal cells.

Although GTP-bound Rac1 is able to bind to the Par6-Par3-aPKC complex in vitro (Lin et al.,2000; Yamanaka et al.,2001), we did not observe impaired cell polarization in Rac1-null EBs, suggesting that Cdc42 is the major regulator of the Par6-Par3-aPKC complex during early embryogenesis.

In Cdc42-null keratinocytes, reduced activation of PKCζ results in decreased phosphorylation and increased kinase activity of GSK3β (Wu et al.,2006a). Also in Cdc42-null EBs, we detected decreased amounts of phosphorylated of GSK3β at the membrane. Because GSK3β can interact with the microtubule binding protein APC (Etienne-Manneville and Hall,2003), the increased kinase activity of nonphosphorylated GSK3β could lead to defects in the microtubule network. Yet, overexpression of a kinase-dead dnGSK3β did not rescue the polarization defects, confirming previous observations that phosphorylation of GSK3β is not required for normal development (McManus et al.,2005). Also because inhibition of the kinase by LiCl did not interfere with the formation of fully polarized EBs, kinase activity of GSK3β seems not at all to be involved in this process.

Finally, our studies with aggregated endodermal cells indicated that also reduced amounts of p120 catenin and E-cadherin could contribute to the impaired formation of belt-like cell–cell contacts. We could not observe changes in these proteins in lysates of 5-day-old EBs. However, it is possible that E-cadherin and p120 are reduced at earlier time points or that subtle changes are masked by the complex cellular composition of EBs. In endodermal cells, in suspension the expression of both E-cadherin and p120 catenin is dependent on PKCζ activation, which in turn is controlled by Cdc42. Regulation of E-cadherin levels during cell–cell aggregation, therefore, might be one pathway how PKCζ regulates the establishment of mature cell–cell contacts.

Cdc42 and BM

Surprisingly, BM could form in the absence of Cdc42. BM formation requires the polarized deposition and assembly of basement membrane components at the basal side of a cell layer. In MDCK cells, inhibition of Cdc42 function results in unpolarized secretion of normally basal proteins to apical and basolateral sides (Kroschewski et al.,1999; Müsch et al.,2001). Although endodermal cells of Cdc42-deficient EBs showed severe defects in polarization with respect to actin cytoskeleton, distribution of adherens (ZO-1) and signaling molecules (PKCζ), and orientation of Golgi complex, BM was still formed at the basal side of the endoderm. BM formation clearly does not require Cdc42, although subtle defects in kinetic or efficiency cannot be excluded. Interestingly, stretches of BM could form between the epiblast cells or the epiblast and an ectopically located endoderm-like cell deep inside mutant EBs, where the cells lacked any distinct polarization with respect to the location of the Golgi complex, adherens junction proteins, F-actin, and ultrastructure. We suspect that these endodermal-like cells have migrated in an inappropriate manner within the EBs to secrete laminin and collagen that then forms the internal BM. These ectopic BMs seem likely to further contribute, through its cell-inductive effects, to disruption of the radial polarity vector that normally exists between the peripheral endodermal/ICM BM and central cavity.

Despite the presence of a BM, the phenotype of Cdc42-deficient EBs is similar to that of LNγ1- or β1 integrin-null EBs, which lack a BM (Smyth et al.,1999; Aumailley et al.,2000). In all these mutants no polarized ectoderm is formed and cavitation is highly impaired. It is unlikely that Cdc42 is affecting adhesion of BM receptors, because fibroblastoid cells lacking Cdc42 showed normal adhesion to extracellular matrix components (Czuchra et al.,2005). It is more likely that Cdc42 is mediating polarization signals downstream of BM receptors. In line with this hypothesis, BM attachment of ectodermal cells did not prevent cell death in Cdc42-deficient EBs, indicating lack of BM-dependent survival signals in the absence of Cdc42. Rac1-null EBs showed no defect in epiblast polarization, but an even more pronounced death of BM attached cells, suggesting that both Rac1 and Cdc42 contribute to BM-dependent cell survival.


Generation of Cdc42-Null ES Cells

Heterozygous Cdc42+/− and Rac1+/− ES cells were generated as described (Wu et al.,2006a; Chrostek et al.,2006). Cdc42+/− and Rac1+/− clones containing a knockout allele were identified by Southern blot and PCR and re-electroporated with the targeting construct. Homologous recombinants were then electroporated with a Cre expression plasmid (kindly obtained by Werner Müller, University of Cologne, Germany) and selected in the presence of FIAU. Cdc42 (fl/−) and Cdc42 (−/−) ES cell clones were identified by Southern blotting and PCR. Experiments were carried out with two to three independent control and knockout clones.


The following antibodies were used for this study: rabbit anti-Cdc42 (sc-87); rabbit anti-RhoA (sc-179); goat anti-Par6 (sc-1440); goat anti-GATA4 (sc-1237; all Santa Cruz); mouse anti-Cdc42 (clone 44); mouse anti-Rac1 (clone 102); mouse anti–β-catenin (clone14); mouse anti-GM130 (clone 35; all Transduction Laboratories); mouse anti-GAPDH (clone 4G5; Chemicon); rat anti–E-cadherin (ECCD2; Zymed); rabbit anti–α-catenin (C2018; Sigma); rabbit anti-cleaved caspase-3; rabbit anti-Akt; rabbit anti-ERK; rabbit anti–GSK-3β; rabbit anti–phospho-Akt (Ser473); rabit anti–phospho-ERK (Thr202/Tyr204); rabbit anti–phospho-GSK-3β (Ser9); rabbit anti–phospho-PKCζ/λ (Thr410/403; all Cell Signaling); rabbit anti–ZO-1 (Zymed); mouse anti-p120 catenin, fluorescein isothiocyanate (FITC) -conjugated mouse anti-α6 integrin (GoH3; all Pharmingen); rat anti-laminin γ1 (05-206); rabbit anti-PKCζ; rabbit anti-Par3 (all Upstate); rabbit anti-JNK, mouse anti-p38 (2F11); rabbit anti–phospho-p38 pTp Y180/182 (all Biosource). Rabbit antibodies against nidogen1, perlecan, and collagen IV were kindly provided by Dr. Rupert Timpl, Max Planck Institute of Biochemistry, Martinsried, Germany. A rabbit antibody against laminin α1 was prepared by P. Yurchenco. As secondary antibodies FITC-, Cy3-, and Cy5-conjugated anti-mouse, anti-rat, and anti-rabbit IgG (all Jackson ImmunoResearch), anti-rabbit IgG and horseradish peroxidase–conjugated anti-mouse (all Amersham) were used. F-actin was detected by FITC-conjugated phalloidin (Molecular Probes).

Generation of Endodermal Cells and of Stably Transfected ES Cell Lines

ES cells were electroporated with pCAG-PI expression vectors for GATA4 (Fujikura et al.,2002; kindly obtained by Dr. Hitoshi Niwa, RIKEN Center for Developmental Biology, Kobe, Japan), dnPKCζ (K281W; Suzuki et al.,2001; kindly obtained by Dr. Shigeo Ohno, Yokohama University, Japan) or dnGSK3β (K85R; Pierce and Kimelman,1995; kindly obtained from Dr. Jürgen Behrens, University Erlangen, Germany) and selected by puromycin (1 μg/ml). The positive clones were confirmed by Western blot analysis of expression GATA4, PKCζ, and GSK3β. To inhibit PKCζ, endodermal cells were treated during aggregation with 10 μM Gö6983 (Calbiochem).

Pull-Down Assay and Western Blot

Pull-down analyses of feeder-free growing ES cells for GTP-bound Cdc42, Rac1 and RhoA were performed using GST-PAK1 and GST-Rhotekin as described (Zondag et al.,2000). Western blot analysis of feeder-free growing ES cells, endodermal cells, or EBs was carried out according to standard protocols using ECL detection (Amersham).

EB Formation

EB formation in suspension was carried out as described earlier (Li et al.,2002), but starting with aggregates of 5–10 ES cells. For inhibition of GSK3β kinase activity, aggregation was performed in the presence of 20 μM LiCl (Sigma).

Cell Cycle Analysis

Cell cycle was analyzed by BrdU incorporation and propidium iodide staining using the In Situ Cell Proliferation kit (FLUOS, Roche) according to the instructions of the manufacturer. Briefly, cells were labeled with BrdU for 45 min at 37°C in a humidified hood, washed with phosphate buffered saline (PBS), fixed with 70% ethanol on ice, rinsed with PBS, denatured with 4 mol/L HCl for 20 min, washed with PBS, incubated with FLUOS antibody for 45 min at 37°C in a humidified hood, washed again with PBS, and then stained with 50 μg/ml propidium iodide (Sigma) in PBS containing 20 μg/ml RNase A (Sigma) for 30 min at room temperature. Cell fluorescence was analyzed by fluorescence-activated cell sorting.

Semiquantitative RT-PCR

Total RNA was isolated with TRIzol reagent (Life Technologies) and reverse-transcribed to cDNA using SuperScript II reverse transcriptase (Life Technologies). Primers and PCR conditions for AFP, BMP-4, ζ-globulin, and GAPDH were described previously (Johansson and Wiles,1995; Levinson-Dushnik and Benvenisty,1997; Murray and Edgar,2000).

Immunofluorescence and EM

EBs were fixed with 4% paraformaldehyde (PFA) and incubated for 3 hr with 7.5% sucrose in PBS and overnight with 15% sucrose in PBS at 4°C. EBs were then embedded in OCT and frozen on dry ice. Four-micrometer cryosections were analyzed by immunofluorescent staining as described (Li et al.,2002). Plated EBs were directly fixed with PFA and analyzed by immunofluorescent staining. For ultrastructural analysis, EBs were fixed in glutaraldehyde, embedded in Epon, and prepared as thick and thin sections, the former for light microscopy, and the latter for EM as described (Li et al.,2002).


We thank Dr. Hitoshi Niwa, Dr. Shigeo Ohno, and Dr. Jürgen Behrens for expression vectors; Drs. John Collard, Arnoud Sonnenberg, and Martin Aepfelbacher for materials and help in establishing the pull-down assay; Dr. Werner Müller for the Cre expression plasmid; Dr. Rupert Timpl for antibodies; Mrs. Ursula Kuhn for expert technical assistance; Drs. Jolanda van Hengel, Atsushi Suzuki, and Klaus Ebnet for comments on the manuscript; and Dr. Reinhard Fässler for discussions, comments on the manuscript, and strong support. P.D.Y. and S.L. were funded by the National Institutes of Health.