The Apical Polarity Determinant Crumbs 2 Is a Novel Regulator of ESC-Derived Neural Progenitors§


  • Thorsten Boroviak,

    1. Department of Biomedical Science, University of Sheffield, Western Bank, Sheffield, S10 2TN United Kingdom
    Current affiliation:
    1. The Wellcome Trust Centre for Stem Cell Research, University of Cambridge, Tennis Court Road, Cambridge CB2 1QR, UK.
    Search for more papers by this author
  • Penny Rashbass

    Corresponding author
    1. Department of Biomedical Science, University of Sheffield, Western Bank, Sheffield, S10 2TN United Kingdom
    • University of Sheffield, Alfred Denny Building, Western Bank, Sheffield S102TN, United Kingdom
    Search for more papers by this author
    • Author contributions: T.B.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing; P.R.: conception and design, data analysis and interpretation, provision of study materials, financial support, manuscript writing, final approval of manuscript.

    • Telephone: 44-0-114-222-2382; Fax: 44-0-114-276-5413;

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS November 23, 2010.


ESCs undergoing neural differentiation in vitro display an intrinsic heterogeneity with a large subset of the cells forming polarized neural rosettes that maintain the neural progenitor microenvironment. This heterogeneity is not only necessary for normal development but also causes substantial technical challenges for practical applications. Here, we report a novel regulator of early neural progenitors, the apical polarity protein Crb2 (Crumbs homologue 2). Employing monolayer differentiation of mouse ESCs to model neurogenesis in vitro, we find that Crb2 is upregulated with Sox1 and Musashi at the onset of neuroepithelial specification and localizes to the apical side of neural rosettes. Stable Crb2-knockdown (KD) lines die at the onset of neural specification and fail to stabilize several apical polarity proteins. However, these cells are able to proliferate under self-renewing conditions and can be differentiated into mesodermal and endodermal lineages. Conversely, Crb2 overexpression during neural differentiation results in elevated levels of other apical polarity proteins and increases proliferation. Additionally, sustained overexpression of Crb2 reduces terminal differentiation into TuJ1-positive neurons. Furthermore, we demonstrate that Crb2 overexpression under self-renewing conditions increases glycogen synthase kinase (GSK)-3β inhibition, correlating with an increase in clonogenicity. To confirm the importance of GSK-3β inhibition downstream of Crb2, we show that Crb2-KD cells can be forced into neural lineages by blocking GSK-3β function and supplementing Epidermal Growth Factor (EGF) and basic Fibroblast Growth Factor (bFGF). Thus, this is the first demonstration that a member of the Crumbs family is essential for survival and differentiation of ESC-derived neural progenitors. STEM CELLS 2011;29:193–205


All cells of the neural lineage are derived from polarized neuroepithelial cells [1]. It has been proposed that self-renewal and differentiation of early neural progenitor cells is regulated by symmetric and asymmetric cell divisions [1–4]. Inheritance of the apical plasma membrane and its junctional complexes is associated with neural progenitor proliferation, whereas lack of its inheritance leads to lineage restriction and differentiation [1]. However, the detailed mechanisms by which the apical domain is set up and how it mediates its proliferative effect are poorly understood.

Cell polarity itself is governed by three evolutionarily conserved cell polarity complexes, that is, the apical Crumbs complex, Par complex and the basal Lgl complex (reviewed in [5, 6]). The three mammalian Crumbs homologues (Crb1, Crb2 and Crb3) are transmembrane proteins, which interact via their cytoplasmic domain with the other members of the complex, Pals1 (also known as Mpp5), PatJ (also known as Inadl), and Lin-7 [5]. The Par complex consists of the three homologues of Par6 (Par6a, Par6b, and Par6d), Par3 (Pard3 and Pard3b), the two atypical protein kinase C (aPKC) homologues, aPKCλ and aPKCζ, and Cdc42. The Crumbs complex and Par complex interact with each other by binding Par6 to either Crumbs [7] or Pals1 [8] (schematic summary shown in Fig. 2A). An important downstream target of the activated Cdc42-Par6-aPKC complex is glycogen synthase kinase-3β (GSK-3β). aPKC inhibits GSK-3β via phosphorylation at serine 9 (Ser9) [9]. GSK-3β is a core component of Wnt/β-catenin signaling [10] and a key regulator of neurogenesis [11–13]. Although several components of the Par complex have been reported to play a pivotal role in neurogenesis [14–18], and a recent study shows that Pals1 is important for cell survival in the cerebral cortex [19], almost nothing is known about the function of the Crumbs proteins in mammalian neurogenesis.

A powerful approach to the study of neurogenesis is to manipulate the neural differentiation of ESCs in vitro. ESCs can be efficiently differentiated into neural lineages in adherent monolayer cultures in the absence of serum [20], which closely mimics embryonic neural development [21–23]. Cell polarity is also maintained in vitro, as neural progenitors tend to organize themselves into neural rosettes [23–26]. The formation of neural rosettes is initiated by the acquisition of cellular polarity and marked by apical localization of ZO-1 [27], the neural stem cell marker Prominin1 [28], and members of the Par complex, including Par3 and aPKC [23]. The latter have been proposed to act as fate determinants in embryonic neural stem cells [29]. Neural rosettes are conserved among species [23, 26, 30, 31] and positive for early neural markers, such as Sox1, Sox2, Nestin, and Pax6 [23, 26, 31]. Recently, it has been shown in human ESCs that neural rosettes represent an early neural stem cell stage [26].

In this study, we aimed to identify new polarity proteins playing a regulative role in mammalian neurogenesis, by employing the ESC monolayer differentiation model [20]. We show that ESCs upregulate Crumbs homologue 2 (Crb2) at the onset of neural specification and that Crb2 localizes apically in neural rosettes. We found that Crb2-depleted cells fail to enter the neural lineage. However, they are able to self-renew as ESCs and differentiate into mesodermal and endodermal lineages. Our gain and loss of function studies demonstrate that Crb2 is essential for the stabilization of other polarity proteins. In addition, we identify Wnt/β-catenin and fibroblast growth factor (FGF) signaling as important downstream pathways of the apical junctional complex. Finally, we confirm our findings by showing that stimulation of these pathways, in a temporally defined manner, rescues Crb2-depleted cells and allows them to enter the neural lineage.


ESC Culture and Neural Monolayer Differentiation

For ESC maintenance see Supporting Information. Neural differentiation in monolayer culture was carried out as previously described [20]. Briefly, CGR8.8 ESCs were seeded at low density (1.0–1.5 × 104 cells per centimeter square) into gelatin-coated dishes. The next day, the ESC medium was changed to serum-free N2B27 medium (1:1 mixture of Dulbecco's modified Eagle's medium/F12 1:1 and Neurobasal, supplemented with N2, B27, L-glutamine and penicillin-streptomycin [all from Invitrogen, Carlsbad, California,]), which was counted as day 0 of differentiation. During subsequent neural differentiation, the N2B27 medium was changed every day.

Crb2-Knockdown Rescue

Neural differentiation was induced for the Crb2-knockdown (KD) and shGFP lines as described earlier and the cells were cultured for the first 4 days in the presence of the following small molecules in N2B27 medium as indicated in Figure 6: CHIR99021 (a kind gift from J. Nichols; 3 μM), bFGF (25 ng/ml), and Epidermal Growth Factor (EGF) (10 ng/ml; both from Sigma, St. Louis, MO,, 10 μM Z-VAD-FMK (Promega, Madison, WI http://www.promega. com). From day 4 onward, the cells were cultured in unsupplemented N2B27 until day 8.

For rescue using shRNA-resistant Crb2 cDNA, see Supporting Information.

Mesodermal and Endodermal Differentiation of ESCs

For monolayer differentiation into mesodermal and endodermal lineages, ESCs were seeded out at very low density (1 × 103cells per centimeter square) in gelatin-coated dishes. The next day medium was changed to 20% fetal calf serum (FCS)-supplemented ESC medium without Leukemia Inhibitory Factor (LIF). For subsequent mesodermal and endodermal differentiation, medium was changed every other day.

Clonogenicity Assay

ESCs were seeded out on gelatin-coated 10-cm dishes at a density of 10 per centimeter square. The cells were cultured for 7 days in ESC medium, fixed in methanol and subsequently stained with crystal violet (Sigma Aldrich).

Plasmids and Generation of Stable ESC Lines

See Supporting Information.

RNA Isolation, cDNA Synthesis, Reverse Transcription Polymerase Chain Reaction, and Quantitative Polymerase Chain Reaction

See Supporting Information.

Western Blotting, Immunofluorescence Staining, Flow Cytometry, and Luciferase Reporter Arrays

See Supporting Information.

Statistical Analysis

All statistical analysis was performed using Prism 5.03 software (GraphPad Software Inc, La Jolla, CA p values were determined using the following statistical methods: Figure 4B and 4C, two-way ANOVA with Bonferroni post-test; Figure 6F, 6G, and 6I, two-way ANOVA with Bonferroni post-test; Supporting Information Figure 4CBb, unpaired t test; Supporting Information Figure 5A, 5E, two-way ANOVA with Bonferroni post-test; Supporting Information Figure 5I, one-way ANOVA with Tukey's multiple comparison post-test

Image Processing

Images were processed using ImageJ (NIH, and Photoshop CS3 (Adobe, San Jose, CA, drawings were made using Omnigraffle 5 Pro (Omni Group, Seattle, WA


Dynamics of Neural Rosette Formation

We analyzed the localization of junctional complexes during neural monolayer differentiation of mouse ESCs, using the tight junctional protein ZO-1 [27] as a marker (Fig. 1A–1E). After 2 days in serum-free culture, the cells formed cellular junctions in a sporadic manner (Fig. 1A), which subsequently developed into a widespread epithelium by day 4 (Fig. 1B). At that stage, only very few, mostly separated, TuJ1-positive neurons were observed. By day 6, the junctional network had retracted leading to the formation of neural rosettes (Fig. 1C). The latter formed three-dimensional structures, which were maintained for extended periods of time. Extensive terminal differentiation into TuJ1-positive neurons started between day 8 and 10 (Fig. 1D, 1E). Consistent with previous studies, we found that rosettes consisted of early neural progenitors, positive for Sox1 (Fig. 1F, 1G), Nestin (Fig. 1H), Musashi (Fig. 1I), and Pax6 (Fig. 1J). They exhibited polarity markers, such as apically localized β-catenin (Fig. 1K) and N-cadherin (Fig. 1L), whereas pan-cadherin (Fig. 1M) showed a more extensive expression pattern. Furthermore, neural rosettes contained proliferative, Ki67-positive cells (Fig. 1N). Consistent with the in vivo situation in the early central nervous system, mitotic cells were located at the apical edges of the rosettes, as shown by phospho histone H3 staining (Fig. 1O). We conclude that ESCs differentiate into early neural progenitors via a polarized epithelial sheet-like state around day 4.

Figure 1.

Characterization of neural rosette formation. (A–E): Immunofluorescence images showing xy planes (and for [A–C]zx planes) of mouse ESCs during neural monolayer differentiation at days 2, 4, 6, 8, and 10 stained for the junctional marker ZO-1 and the postmitotic neuronal marker TuJ1. (F): Phase contrast and (G) GFP expression of Sox1-GFP knock-in reporter line (46C) neural rosettes at neural differentiation day 8. (H–O): CGR8.8 neural rosettes at day 8 of neural monolayer differentiation are positive for (H) Nestin, (I) Musashi, (J) Pax6, show apical localization for (K) β-catenin, (L) N-cadherin, but more uniform (M) pan-cadherin, and are positive for the proliferation (N) Ki67 and mitotic marker (O) phospho (Ser10) histone H3 (PH3). Scale bar = 20 μm (A–C), 50 μm (D–O). Dotted line in (F–O) indicates location of rosette. Abbreviations: DAPI, 4'6-diamidino-2-phenylindole; GFP, Green Fluorescent Protein.

Polarity Protein Expression and Localization in Neural Differentiation

To identify new potential regulators of neurogenesis among polarity proteins (Fig. 2A), we screened for changes in gene expression by reverse transcription polymerase chain reaction (RT-PCR) during neural differentiation. Our controls confirmed that ES cells differentiated into neural progenitors via a FGF5- positive primitive ectodermal stage [32] (Fig. 2B). We found that most polarity proteins were expressed throughout during neural differentiation (Fig. 2C). In contrast, the Crumbs gene family exhibited a highly dynamic pattern. Crb3 was present in undifferentiated ESCs and subsequently downregulated. However, Crb2 was upregulated at the time of neural specification at day 4 and maintained at later stages. Crb1 followed a similar pattern to Crb2, but its expression lagged behind, so that its major upregulation occurred at day 8. This result was verified by quantitative PCR (qPCR; Supporting Information Figure 2A). At the protein level, Crb2 upregulation correlated with the appearance of the early neural markers Sox1 and Musashi (Fig. 2D). Also, we noticed a slight increase in protein levels of PatJ, Pard3, and Par6 around day 2 and 4. Furthermore, at day 4 there was an increase in phosphorylation of aPKCλ/ζ and a stabilization of Cdc42 at day 4 (Fig. 2D), indicated by increase in its protein but not RNA levels (Fig. 2D, 2B and Supporting Information Figure 2B). These events are most likely to play a role in epithelial formation, as ESCs differentiate into neuroectodermal precursors at this stage (Fig. 1B and [20]).

Figure 2.

Expression of polarity proteins during neural monolayer differentiation. Samples were taken every second day until day 10, 0 represents undifferentiated ESCs grown up in LIF-supplemented ESC medium for 3 days. (A): Schematic representation of the three polarity complexes in an epithelial cell. (B, C): Reverse transcription polymerase chain reaction results during neural monolayer differentiation of CGR8.8 mouse ESCs for (B) markers and (C) polarity genes. Gene names are stated on the left, the number of polymerase chain reaction cycles on the right side. (D): Western Blots in a time course during neural monolayer differentiation of CGR8.8 ESCs. aPKC* antibody detects PKCζ and phospho aPKC** antibody detects Thr 410 phosphorylated PKCζ and Thr 403phosphorylated PKCλ. Extensive characterization of the Crb2 antibody is shown in Supporting Information Figure 1A and 1B, a full-length Western blot of Pals1 is shown in Supporting Information Figure 1C. Abbreviations: aPKC, atypical protein kinase C; Dlg1, Disc Large 1; Gapdh, glyceraldehyde 3 phosphate dehydrogenase; L.C., loading control; LGL, Lethal Giant Lar; ND, Neural Differentiation.

The observation that the Par complex was found to stabilize prior to the onset of neurogenesis, whereas Crb2 was upregulated with early neural markers, which suggested a distinct role for Crb2 during neural differentiation. Consistent with this, we observed that members of the Crumbs complex (Crb2, Pals1, and PatJ) localized more apical to ZO-1 staining (Fig. 3A), whereas Par complex protein (aPKC, Par3, and Cdc42) expression overlapped with ZO-1 in day 8 neural rosettes (Fig. 3B).

Figure 3.

Localization of polarity proteins of the Crumbs and Par complex. (A–D): Immunofluorescence of neural rosettes derived from CGR8.8 ESCs at day 8 stained for (A) Crb2, Pals1, and PatJ in relation to the junctional marker ZO-1 and in overlay with DAPI (blue), insets represents higher magnification image of boxed area; (B) aPKC, Par3, Cdc42, Cdc42/Rho-GTP (using a GST-CRIB domain fusion protein as a probe) in relation to ZO-1 and in overlay with DAPI (blue), insets represents higher magnification image of boxed area; (C) phospho-GSK-3β and ZO-1; and (D) triple staining for Crb2, Nestin and TuJ1. (E): Immunohistochemistry of the right hemisphere of a frontal section through an E12.5 mouse dorsal telencephalon stained for Crb2 and the postmitotic neuronal marker TuJ1. (F): Higher magnification of the neocortex shown in (E). Scale bar = 50 μm ([A–D], [F]), 100 μm (E); images (E) and (F) are tile scans. Abbreviations: aPKC, atypical protein kinase C; Crb2, Crumbs homologue 2; DAPI, 4'6-diamidino-2-phenylindole; n, neocortex; s.e., surface epithelium; v, ventricle.

In addition, we confirmed that apical Cdc42 was in its activated, GTP-bound state (Fig. 3B and Supporting Information Figure 3B). Previous reports suggested that Cdc42 together with aPKC inactivates GSK-3β via phosphorylation at Ser9 [9]. GSK-3β, a key player of the Wnt/β-catenin pathway is inhibited by phosphorylation at Ser9 [33]. This in turn prevents β-catenin from degradation and results in increased Wnt/β-catenin signaling [34]. Our immunocytochemistry showed that GSK-3[gr beta] is phosphorylated at Ser9 specifically at the apical surface (Fig. 3C). This may be a prerequisite for β-catenin enrichment within this region (Fig. 1K). Crb2 localization was also associated with the apical fate determinant and neural stem cell marker prominin1 [35, 36] in neural rosette cells (Supporting Information Figure 3C). However, Crb2 expression was excluded from basally located TuJ1-positive neurons (Fig. 3D). We observed a similar pattern in vivo, in the E12 mouse dorsal telencephalon, where Crb2 was localized apically at the ventricular zone, whereas differentiation into TuJ1-positive cells occurred basally (Fig. 3E, 3F).

From our expression studies, we conclude that Crb2 is upregulated at the onset of neural specification, major key polarity proteins are stabilized on a protein level at this stage, and Crb2 localizes apically in early, polarized neural progenitors in vitro and in vivo.

Crb2 Upregulation Is Specifically Required for Cell Survival During Neural Monolayer Differentiation

Our observations raised the question whether Crb2 upregulation was a cause or a consequence of neurogenesis. We addressed this by generating stable Crb2-KD ESC lines using short hairpin RNA interference. The KD efficiency of two independent hairpins against Crb2 was tested and confirmed by transient transfection in preliminary experiments. Hairpins targeting Crb2 exon 8 (shCrb2-1) and 9 (shCrb2-2) and one against GFP (shGFP), to account for any nonspecific effects from the selective agent or the short hairpin, were introduced into wild-type CGR8.8 ESCs. At least three independent clones of each construct were analyzed, with one illustrated here for simplicity. As Crb2 expression is barely detectable in ESCs, we did not expect the KD constructs to show a severe phenotype in undifferentiated cells. This was found to be the case, that is, all of the KD and control lines proliferated exponentially in ES medium supplemented with FCS and LIF (Fig. 4B). The Crb2-KD lines showed similar levels of the pluripotency marker Oct4 compared with controls and there was no increase in cleaved Caspase3, a marker for apoptosis (Supporting Information Figure 4A).

Figure 4.

Phenotype of Crb2 knockdown. (A): Phase contrast pictures of a shGFP control line and two Crb2-KD lines (shCrb2-1 and shCrb2-2) after seeding (middle left column), under self-renewing conditions in LIF-supplemented ES medium (left column), and after 2 (middle right column) and 4 (right column) days of neural differentiation in serum-free N2B27. (B): Quantification of cell numbers under self-renewing and (C) neurogenic conditions for the experiment shown in (A). Cell numbers were determined by counting trypsinized cells using a hemocytometer. Error bars represent standard deviation between three independent experiments. (D): Mesodermal and endodermal monolayer differentiation in ES medium without LIF of a shGFP control line, shCrb2-1, and shCrb2-2. Phase contrast pictures were taken at day 10 of mesodermal and endodermal differentiation. The relative RNA levels for pluripotency markers Nanog, Oct4, and Rex1, plus mesodermal and endodermal markers AFP and Brachyury were determined by quantitative polymerase chain reaction (PCR) in self renewing (ES blue bar) and mesendoderm differentiation day 10 (MesEndo red bar) conditions. (E): Summary of the Crb2-KD phenotype: Crb2-depleted ESCs can self-renew and differentiate into mesodermal and endodermal lineages but die under neurogenic conditions. (F): Crb2 regulates apical polarity protein stability. Western blots of ES-WT, shGFP control, shCrb2-1, and shCrb2-2 at neural differentiation day 4 for neural markers, polarity proteins, and Notch. (G): Quantitative PCR of polarity genes of the shGFP control, shCrb2-1, and shCrb2-2 at neural differentiation day 4. (H): Western blots of ES-WT, shGFP control, shCrb2-1, and shCrb2-2 at neural differentiation day 4 for GSK-3β and Ser9-phosphorylated GSK-3β. Numbers below pGSK-S9 represent Ser9-phosphorylated GSK-3β normalized to total GSK-3β levels, relative to ES-WT as determined by band quantification. RNA expression levels in (D) and (G) represent an average of three independent experiments and were normalized to Gapdh. Error bars represent standard deviation. Scale bar = 100 μm (A, D); ***, p < .0001. Abbreviations: aPKC, atypical protein kinase C; Crb2, Crumbs homologue 2; ES WT, wild-type ESCs; FCS, fetal calf serum; Notch1-FL, Notch1 full length; Gapdh, glyceraldehyde 3 phosphate dehydrogenase; Gsk, glycogen synthase kinase; Notch1-ICD, Notch 1 intracellular domain; KD, knockdown; LIF, leukemia Inhibitory Factor; ND, Neural Differentiation; n.s., non significant; pGsk-S9, ser9 phosphorylated GSK-3β; shCrb2, short hairpin against Crumbs homologue; shGFP, short hairpin GFP.

Neural differentiation of the Crb2-KD lines resulted in a 70%–75% reduction of Crb2 protein levels at neural differentiation day 4 (Supporting Information Fig. 4B), concomitant with a dramatic decrease in cell numbers and viability (Fig. 4A and 4C). Importantly, this was at the time point when Crb2 was found to be upregulated in wild-type ESCs and could also be rescued with shRNA-resistant Crb2-cDNAs (Supporting Information Fig. 4C). To address whether this apoptotic phenotype was due to neural differentiation, rather than a result of the change in culture conditions, we added GSK-3 inhibitor (CHIR99021), Mitogen-Activated Protein Kinase Kinase Kinase-1 (MEKK) inhibitor (PD0325901), and LIF to the serum-free N2B27 medium. This inhibitor cocktail (2i+LIF) is reported to maintain ESCs in their pluripotent state [37, 38], and indeed all lines grew robustly into round, ES-like colonies (Supporting Information Fig. 4C), which expressed pluripotency markers (Supporting Information Fig. 4D).

Next, we wanted to test whether Crb2-KD ESCs were generally unable to differentiate or whether this effect was specific to neural differentiation. To address this issue, we directed Crb2-KD cells into the mesodermal and endodermal lineages by culturing them in ESC medium supplemented with 20% FCS but without LIF. Under these conditions, the Crb2-KD lines proliferated similar to controls, downregulated pluripotency markers and upregulated Brachyury and Alpha-fetoprotein (AFP), markers of the mesodermal and endodermal lineage (Fig. 4D). Taken together, Crb2-KD cells were able to proliferate under self-renewing conditions with and without FCS and could be differentiated into mesodermal and endodermal lineages but specifically died at the onset of neural specification at day 4 in neural monolayer differentiation (Fig. 4E).

Crb2-KD Cells Fail to Undergo Neural Specification and to Stabilize Apical Polarity Determinants

To elucidate the potential cause of this phenotype, we next assessed the protein levels of neural markers and polarity proteins by Western blotting in the Crb2-KD lines at day 4 of neural differentiation. The Crb2-KD lines exhibited an increase in cleaved Caspase3 but also had reduced expression of the early neural differentiation markers Sox1 and Musashi (Fig. 4F). Furthermore, we found that Crb2 depletion resulted in reduced levels of other apical polarity proteins, such as PatJ, Par3, aPKC, and Cdc42 (Fig. 4F). In our system, reduction of Crb2 correlated with a decrease in levels of full-length Notch1 (Fig. 4F), which would be consistent with the hypothesis that Crb2 inhibits gamma secretase [39, 40].

Members of the Crumbs family are the only proteins of the apical Crumbs and Par complex with a transmembrane domain. To determine whether loss of the other polarity proteins was due to loss of complex stabilization, rather than active downregulation in the Crb2-KD cells, we carried out qPCR at the same stage and found no changes in gene expression of Cdc42, PatJ, Par3, and aPKCλ in Crb2-KD cells compared with controls (Fig. 4G). As our immunofluorescence analysis showed that wild-type neural rosettes exhibit substantial GSK-3β inhibition at the apical surface (Fig. 3C), the level of Ser9 phosphorylation in Crb2-KD cells was determined. We found a substantial decrease in GSK-3β inhibition in Crb2-depleted cells (Fig. 4H). We conclude that Crb2-KD cells failed to stabilize apical polarity proteins and showed decreased Notch1 levels at neural differentiation day 4.

Crb2 Overexpression Increases Proliferation but Reduces Terminal Differentiation Under Neurogenic Conditions

As the Crb2-KD lines died under neurogenic conditions at day 4, we could not use them as a tool to study the molecular mechanisms for this phenotype. Therefore, we decided to explore the effects of Crb2 overexpression. Four independent ESC clones stably expressing the full-length cDNA of Crb2 (Crb2-OVEX1–4) plus three empty vector controls were derived. Endogenous Crb2 displays two bands at ∼200 and 220 kDa on Western Blots during neural differentiation (Fig. 5A). Our overexpression construct (exogenous Crb2) produced a single band at 200 kDa in all the clones analyzed (Fig. 5A).

Figure 5.

Phenotype of Crb2 overexpression. (A): Western blot of cell lysates from undifferentiated ESCs (ND day 0), ND day 4 and 8 of an empty vector control ESCs (empty vector), and Crb2-OVEX1 probed for Crb2 and Gapdh. (B): Immunocytochemistry at ND day 8 of (A) an empty vector control and two Crb2-overexpressing lines (Crb2-OVEX1 and Crb2-OVEX2) stained for TuJ1 and ZO-1. (C, D): Relative RNA levels of Crb2 from the construct (Crb2-EXO) and endogenous (Crb2-ENDO) mRNA of empty vector control and Crb2-OVEX1–4 in (C) undifferentiated ESCs and at (D) ND day 8 as determined by quantitative polymerase chain reaction. Expression levels represent an average of three independent experiments and were normalized to Gapdh. Error bars indicate standard deviation. (E): Western blots of an empty vector control and Crb2-OVEX1–4 in undifferentiated ESCs and at ND day 8. Numbers below TuJ1 represent TuJ1 protein levels normalized to Gapdh relative to the empty vector control at ND day 8 as determined by band quantification. (F): Inverse correlation between exogenous Crb2 expression levels and TuJ1 protein levels for Crb2-overexpressing clones 1–4, dotted line represents linear regression. (G): Western blots of wild-type CGR8.8 ESCs (ES-WT), an empty vector control, and Crb2-OVEX1 and Crb2-OVEX2 lines under self-renewing conditions probed for GSK-3β and pGSK-S9. Numbers below pGSK-S9 represent Ser9-phosphorylated GSK-3β normalized to total GSK-3β levels, relative to ES-WT as determined by band quantification. (H): Representative images of colony-forming unit assays for wild-type CGR8.8 ESCs (ES-WT), an empty vector control, and Crb2-OVEX1 and Crb2-OVEX2 lines under self-renewing conditions. (I): Western blots for the samples as described for (G) probed for polarity proteins and Notch1. Numbers below paPKC represent phosphorylated aPKC normalized to total aPKC levels relative to ES WT as determined by band quantification. Scale bar = 50 μm (B). Abbreviations: aPKC, atypical protein kinase C; Crb2, Crumbs homologue 2; Crb2-OVEX1, Crb2-overexpressing line 1; DAPI, 4'6-diamidino-2-phenylindole; ES WT, wild-type ESCs; Notch1-FL, Notch1 full length; Gapdh, glyceraldehyde 3 phosphate dehydrogenase; Gsk, glycogen synthase kinase; Notch1-ICD, Notch 1 intracellular domain; ND, neural differentiation; pGsk-S9, ser9-phosphorylated GSK-3β.

The undifferentiated Crb2-overexpressing (Crb2-OVEX) lines grew robustly under self-renewing conditions (Supporting Information Fig. 5A), exhibited a typical ESC-like morphology (Supporting Information Fig. 5B) and showed Oct4 levels similar to wild-type ESCs and empty vector control (Supporting Information Fig. 5C). However, Crb2-overexpressing cells gave rise to approximately twice as many colonies in a clonogenicity assay compared with controls (Fig. 5H, Supporting Information Fig. 5H).

When inducing neurogenesis in the Crb2-OVEX lines, they not only survived but also showed a slight increase in proliferation (Supporting Information Fig. 5D). Exogenous expression of Crb2 did not affect the gross expression patterns of Crb1 and Crb3 (Supporting Information Fig. 5E). Crb2-OVEX lines exhibited a substantial reduction of terminal differentiation into TuJ1-positive neurons (Fig. 5B). Consistent with this, we found reduced expression of NeuroD1 and Mtap2 at day 8 (data not shown). Notably, the amount of terminal differentiation into mature neurons varied between the different Crb2-OVEX ESC lines. To address whether this could be due to variable exogenous Crb2 expression between the lines, we performed qPCR with exogenous and endogenous Crb2-specific primers in undifferentiated ESCs (Fig. 5C) and at neural differentiation day 8 (Fig. 5D). Different lines were found to maintain exogenous Crb2 (Crb2-EXO) at different levels after 8 days of neural differentiation. Interestingly, these levels (Crb2-EXO: clone 2 ⋙ 4 ⋙ 1 ⋙ 3, Fig. 5D) inversely correlated with postmitotic neural differentiation (TuJ1: clone 3 ⋙ 1 ⋙ 4 ⋙ 2, Fig. 3E). This inverse correlation (Fig. 5F) was further confirmed for Crb2-OVEX1 and 2 by fluorescence-activated cell sorter (FACS) analysis (Supporting Information Fig. 5F).

Crb2 Overexpression in ESCs Increases Wnt/β-Catenin Signaling and Stabilizes Other Apical Polarity Proteins

To determine which signaling pathways were affected by Crb2 overexpression, we performed a 10-pathway luciferase array assay under self-renewing conditions. We identified a significant increase in T-cell factor/lymphoid enhancer factor (TCF/LEF) luciferase reporter activity (Supporting Information Fig. 5G), suggesting an upregulation of the Wnt/β-catenin signaling. Consistent with the reporter array results, we found an increased level in inhibitory Ser9 phosphorylation of GSK-3β (Fig. 5G). As GSK-3β inhibition is reported to enhance cell viability in ESCs [37], this is consistent with our observation of increased clonogenicity (Fig. 5H, Supporting Information Fig. 5H).

Given that Crb2-KD reduced the stability of apical determinants, we examined their protein levels in the Crb2-OVEX lines by Western blotting. We found an increase in PatJ, Par3, Par6, and Cdc42 but not Pals1. Furthermore, there were elevated levels of phosphorylated aPKC and a reduction of the basal protein Lgl1 in the Crb2-OVEX lines (Fig. 5I). We also observed an increase in the full-length form of Notch1 (Fig. 5I).

Taken together, Crb2 overexpression in ESCs resulted in elevated levels of major apical polarity proteins, full-length Notch1, GSK-3 Ser9 phosphorylation, and Wnt/β-catenin signaling and thus enhanced clonogenicity. Additionally, in our system, forced expression of Crb2 in neural progenitors inhibited terminal differentiation into postmitotic neurons in our system.

The Crb2-KD Phenotype Can Be Rescued by GSK-3β Inhibition and Stimulation of FGF Signaling

Our results have shown that Crb2-KD cells failed to stabilize apical polarity determinants, including aPKC and Cdc42, concomitant with a decrease in GSK-3β phosphorylation. Conversely, Crb2 overexpression resulted in an increase of apical polarity protein levels as well as aPKC and GSK-3β phosphorylation. Cdc42-activated aPKC has been shown to phosphorylate GSK-3β [9], which is consistent with the elevated levels of phospho-GSK-3β and TCF/LEF reporter activity in Crb2-overexpressing cells. Moreover, a recent study suggests a fundamental function of GSK-3β for neural progenitor survival and homeostasis [12]. We, therefore, hypothesized that the apoptotic phenotype of Crb2-KD cells may be partially rescued by GSK-3β inhibition with the chemical inhibitor CHIR99021 during neural differentiation (Fig. 6A, 6B). However, previous reports have also shown that β-catenin/Wnt signaling in undifferentiated ESCs can maintain pluripotency [37, 41, 42], implying that GSK-3 inhibition alone would interfere with neurogenesis. To overcome this problem, we decided to stimulate neural specification by additionally adding EGF and bFGF [20, 43, 44].

Figure 6.

Rescue of the Crb2 knockdown (KD). (A): Schematic outline of the rescue strategy: Crb2-KD ESCs were cultured in the presence of small molecules in serum-free N2B27 medium for 4 days and then in serum-free N2B27 alone for a further 4 days. Analysis for survival and neural progenitor markers was performed at day 4 ([C, D] and Supporting Information Figure 6A and 6B) and 8 (E–I). (B): Schematic outline of the site of action of the small molecules used. (C, E): Phase contrast pictures of DMSO; Z-VAD-FMK (Z); Z-VAD-FMK, bFGF, and EGF; CHIR99021 (CHIR); bFGF and EGF; and CHIR99021, bFGF, and EGF-treated cells of the shGFP control and shCrb2-1 at ND day 4 (C) and 8 (E). (D): Quantification of cell numbers for the experiment shown in (B) plus for the Crb2-KD line shCrb2-2. (F): Quantification of cell numbers for the experiment shown in (E) plus for the Crb2-KD line shCrb2-2. (G): Relative RNA levels for pluripotency marker Nanog and neural progenitor marker Pax6 at ND day 8 in the shGFP control line cultured in N2B27 supplemented with 2i+LIF (shGFP-2i) and shGFP plus DMSO compared with shCrb2-1 cells cultured in either CHIR alone or CHIR+bFGF+EGF. (H): Immunocytochemistry of the DMSO-treated shGFP control and CHIR99021+bFGF+EGF-treated shCrb2-1 and shCrb2-2 stained for TuJ1 and Crb2 at ND day 8. (I): Relative RNA levels for neural lineage markers in the shGFP control line cultured in N2B27 plus 2i+LIF supplemented (shGFP-2i) compared with those in cell lines at ND day 8, DMSO-treated shGFP control (shGFP-NDd8),CHIR99021+bFGF+EGF-treated shGFP control (shGFP-NDd8-FEC), and Crb2-KD line 1 (shCrb2-1-NDd8- FEC). (G, I): Quantitative polymerase chain reaction was normalized to Gapdh. Error bars represent standard deviation. Scale bar = 100 μm (C), 50 μm (H). *, p < .05; **, p < .001; ***, p < .0001. Abbreviations: aPKC, atypical protein kinase C; bFGF, basic Fibroblast Growth Factor; CHIR, CHIR99021; DMSO, dimethyl sulfoxide; EGF, Epidermal Growth Factor; ERK, Extracellular Signal Regulated Kinase; GSK-3, glycogen synthase kinase-3; LIF, leukemia Inhibitory Factor; MEKK, Mitogen-Activated Protein Kinase Kinase Kinase-1; ND, neural differentiation; n.s., non significant; PAK1, serine/threonine protein kinase 1; shCrb2, short hairpin against Crb2; shGFP, short hairpin against Green Fluorescent Protein; Z-VAD-FMK, pan caspase inhibitor.

After 4 days of culture in bFGF-, EGF-, and CHIR99021-supplemented N2B27 medium (FEC treated), we assayed for cell survival and neural progenitor marker expression (Fig. 6C, 6D). We found an increase in cell survival in the FEC-treated Crb2-KD lines compared with dimethyl sulfoxide (DMSO) controls. Also, the FEC-treated Crb2-KD lines had a similar morphology to the mock-treated shGFP control lines, whereas the FEC-treated shGFP controls revealed an altered morphology (Fig. 6C, Supporting Information Fig. 6B). The FEC-rescued cells expressed Sox2, Nestin, and Pax6 (Supporting Information Fig. 6A), indicating that they underwent neural specification.

Critically, quantification of cell numbers at neural differentiation day 4 showed a reduction of almost 50% in the control shGFP lines (Fig. 6D). In contrast, the numbers of the Crb2-KD lines doubled (Fig. 6D), which suggests a specific survival effect, rather than a nonspecific, trophic effect of the CHIR99021-, bFGF-, and EGF-supplemented medium. To further distinguish between putative direct prosurvival effects of the individual compounds, we separately analyzed the effects of FGF stimulation and GSK-3β inhibition. We then compared these with the results obtained from the triple cocktail (FEC-treated cells). At neural differentiation day 8, only the FEC treatment promoted substantial cell survival and entry into the neural lineage. CHIR treatment alone resulted in a modest increase in cell survival at day 8. However, these cells exhibited substantially lower Pax6 expression compared with DMSO controls (Fig. 6G). In addition, they exhibited a very round, colony-like morphology (Fig. 6E) further suggesting that they failed to enter the neural lineage. Importantly, cells with just FGF/EGF treatment showed the poorest survival of all treatments at day 8 neural differentiation (Fig. 6F). We, therefore, conclude that in this context, FGF signaling stimulation promotes neural specification rather than cell survival.

In addition, to determine whether we have an indirect prosurvival effect from the compounds, we also treated the cells with Z-VAD-FMK, a well described pan-Caspase inhibitor [45, 46] both alone and in combination with bFGF/EGF (Fig. 6A, 6B). We used a concentration that had previously been published to reduce apoptosis during mouse ESC derivation [47] and obtained similar results with higher concentrations (data not shown). There was initially a modest increase in cell numbers at day 4 (Fig. 6C, 6D), but the cells still do not survive to day 8 (Fig. 6E, 6F). Importantly, Z-VAD-FMK combined with bFGF+EGF also did not produce neural progenitors, suggesting a specific role for GSK-3β.

Further characterization of the FEC-treated Crb2-KD cells showed that they produced TuJ1-positive neuron-like cells at neural differentiation day 8 in the absence of Crb2 (Fig. 6H). Analysis of these cells by qPCR showed that while all cells (shGFP, shGFP treated, and shCrb2-1 treated) upregulated a variety of neural markers including Olig2, Sox1, Pax6, and Nestin, Otx1 upregulation only occurred in untreated shGFP controls (Fig. 6I). Treated shGFP and shCrb2-1 cells expressed similar levels of Olig2, Sox1, and Pax6. However, we found that Sox2 was significantly reduced in the Crb2-deficient cells.

Taken together, by simultaneously restoring GSK-3β inhibition and FGF signaling, Crb2-depleted cells were able to survive and differentiate under neurogenic conditions.


Crb2 Is Essential for Neural Differentiation

In this study, we show that ESCs differentiate into neural rosettes via a sheet-like epithelial stage and establish an expression profile of key proteins known to be required to set up and maintain cell polarity (Fig. 7A). This expands on previously published microarray expression analysis of neural monolayer differentiation [22, 23] as these data sets do not include all major polarity proteins. In our study, we identified that Crb2 is specifically upregulated at the time of neural specification and is localized apically in neuroepithelial cells in vitro and in vivo. Crb2-KD did not effect ESC self-renewal or mesodermal and endodermal differentiation. However, under neurogenic conditions, Crb2 depletion resulted in massive cell death at the time of neural specification, thereby demonstrating that Crb2 is necessary for neurogenesis.

Figure 7.

A model for Crb2 in neurogenesis. (A): Diagrammatic summary of Crumbs gene expression according to previously described stages [22, 23, 26] of neurogenesis, together with an summary of cell morphological changes in neural in vitro differentiation. (B): Proposed model for cell signaling events downstream of Crb2 in neuroepithelial cells. Abbreviations: aPKC, atypical protein kinase C; Crb2, Crumbs homologue 2; EGF, Epidermal Growth Factor; ERK, Extracellular Signal Regulated Kinase; FGF, fibroblast growth factor; GSK-3, glycogen synthase kinase-3; MEKK, Mitogen-Activated Protein Kinase Kinase Kinase-1; ND, Neural Differentiation; PAK1, serine/threonine protein kinase 1.

Crb2 Is Required for Stabilization of Apical Polarity Proteins

Apical polarity proteins are fundamentally important for neural development, that is, Par3 and Par6 localize apically and promote proliferation of neural progenitors in the developing mouse cerebral cortex [16], overexpression and mislocalization of aPKCζ increases neural progenitor numbers [15], and very recently it has been shown that loss of Pals1, notably a member of the Crumbs complex, resulted in massive cell death during cortical development [19]. This is similar to our Crb2-KD phenotype. Furthermore, Crb2 overexpression increased the levels of several apical proteins, whereas these were decreased in Crb2-KD cells under neurogenic conditions. However, as RNA levels were unaffected, Crb2 must play an important role in the stabilization of the apical Crumbs and Par protein complexes.

Taken together, our experiments showed that Crb2 ensures apical polarity complex stability and thus substantially contributes to proliferation and survival in neuroepithelial cells.

A Model for Apical Polarity Protein Signaling in Neuroepithelial Cells

We, therefore, propose a model (Fig. 7B) in which survival, proliferation, and differentiation during neurogenesis depend on signals from apical polarity proteins, localized at the adherens junctions of early neural progenitors. The only known active kinases of the Par and Crumbs complex are Cdc42 and aPKC, which interact via the structural component Par6. Cdc42-GTP binds to the Par6-aPKC dimer, resulting in aPKC activation [48]. Both Cdc42 and aPKC are associated with neurogenesis [15, 18, 49, 50].

The important role of Cdc42 for neural induction is underlined by other studies, where Cdc42-deficient ESCs failed to differentiate into ectodermal lineages [51]. A well-described downstream target of activated Cdc42 is serine/threonine protein kinase 1 that stimulates the FGF signaling pathway [52–54]. Interestingly, Par6 overexpression is able to induce EGF/FGF independence of epithelial cells, in a Cdc42- and aPKC-dependent manner [55], providing further evidence for a possible link between apical polarity complex proteins and FGF signaling. We observed that wild-type ESCs stabilize Cdc42 at the onset of neural specification. Crb2-depleted cells failed to stabilize Cdc42 and to undergo neural specification. Consistent with this, we found that increased Cdc42 levels in Crb2-overexpressing cells led to an increase in phosphorylated aPKC. Thus, stabilization of Cdc42 and intrinsic activation of FGF signaling are important contributors to neural specification.

Moreover, the activated Cdc42-Par6-aPKC complex has previously been shown to inhibit GSK-3β via phosphorylation at serine 9 [9]. We found that Crb2-overexpressing ESCs exhibited a significant increase in Wnt/β-catenin reporter activity together with increased levels of GSK-3β serine 9 phosphorylation. Furthermore, activated Cdc42, aPKC, Ser9 phosphorylated GSK-3β, and Crb2 colocalized at the apical lumen in neural rosettes. Combined, our data indicate that Crb2 interacts with the same pathway. GSK-3β has been reported to induce neuronal cell death [13], blocked inhibition of GSK-3β impairs neural progenitor proliferation in vivo [11], and deletion of all GSK-3 isoforms in the developing central nervous system (CNS) enhances progenitor proliferation but strongly inhibits neurogenesis [12]. The latter provides an explanation, why we see that sustained Crb2 overexpression increased proliferation under neurogenic conditions but reduced the number of terminally differentiated neurons.

In our model, Crb2-mediated inhibition of GSK-3β is fundamentally important for neural progenitor integrity. We tested the model and the physiological relevance by partially rescuing Crb2-KD cells under neurogenic conditions. Restoration of GSK-3β inhibition, together with stimulation of the FGF signaling pathway, not only significantly increased survival of Crb2-KD cells but also allowed differentiation into the neural lineage. Although GSK-3β inhibition or stimulation of FGF signaling alone slightly increased cell survival at day 4, these cells still died before day 8. Hence, stimulation of both pathways is essential for neural specification and survival.

Thus, our experiments demonstrate that Wnt/β-catenin and FGF signaling can compensate for Crb2-dependent loss of apical polarity proteins.

Interestingly, characterization of rescued Crb2-KD cells at neural differentiation day 8 revealed that they showed a significant reduction in Sox2 expression. This is in agreement with another study, where overexpression of the apical determinant aPKCζ in chick has been reported to increase the number of Sox2-positive neural progenitors [15]. These results underline the importance of the apical domain for maintenance of early neural progenitors.

Crb2 and Notch Signaling

Although the Crb2-KD rescue with GSK-3β inhibition and stimulation of FGF signaling validates our model, the rescue is incomplete as cell survival does not reach wild-type levels. Therefore, alternative pathways must also play a role. One of these could be Notch signaling. Recent reports suggested that Crumbs in Drosophila and Crb2 in mammalian cells modulates Notch signaling via inhibition of γ-secretase, but these findings are currently disputed as another report failed to detect this interaction in HEK293 cells [56]. Our data support a role for Crb2 inhibiting γ-secretase, as Crb2-KD cells show reduced levels of Notch1 at neural differentiation day 4, whereas undifferentiated Crb2-overexpressing ESCs exhibited higher levels of Notch1 compared with control cells. Nevertheless, we cannot exclude the possibility that this result is due to indirect molecular effects. We tried to restore Crb2-mediated Notch inhibition in Crb2-KD cells with the chemical inhibitor N-[N-(3,5-difluorophenacetyl-L-alanyl]-S-phenylglycine t-butyl ester (DAPT), but, not surprisingly, these attempts failed (data not shown).

The hypothesis that Crb2 influences Notch cleavage would provide an attractive mechanism for spatial regulation of Notch signaling within the cell. However, further studies need to be carried out to confirm or refute a direct role for Crb2 in Notch signaling. Currently, there are more than 60 known γ-secretase substrates [39], including amyloid precursor protein (APP). The latter is processed by the γ-secretase complex, leading to the formation of amyloid-β peptides, which are considered a primary cause of Alzheimer's disease. If Crb2 inhibits γ-secretase activity, it is plausible that appropriate Crb2 expression levels may be crucial for the prevention of neurodegeneration.


Many future practical applications of stem cell biology will depend on pure populations of cells. Appreciating the mechanisms that produce and maintain the different cell types is of fundamental importance. We have identified Crb2 as a new player in neural differentiation, and as it has a large extracellular domain, it potentially could be an important therapeutic target for neuronal survival and to direct neural differentiation in vitro.


In this study, an ESC monolayer differentiation model was used to show that Crb2 is upregulated at the onset of neural specification. Gain- and loss-of-function studies demonstrated that Crb2 is essential for the stabilization of other polarity proteins. Furthermore, Crb2-depleted cells failed to enter the neural lineage but were able to self-renew as ESCs and differentiate into mesodermal and endodermal lineages. Cell survival and neural specification of Crb2-KD cells could be rescued by stimulating Wnt/β-catenin and FGF signaling.


We are grateful to M. Murtaza for her contribution to Figure 5E and M. Rivolta and A. Furley for their comments on the article. We thank J. Nichols for providing us with the GSK-3 inhibitor CHIR99021 and CGR8.8 cell-line, A. L. Bivic for the PatJ antibody, and S. Winder for the GST-PAK1-CRIB domain fusion protein. This work was partly supported by the European Commission (HEALTH-F2-2008-200234) and the Biotechnology and Biological Sciences Research Council (BB/D014840/1). T.B. is supported by a University of Sheffield PhD scholarship. Microscopy was performed in the Sheffield BioImaging Facility (Wellcome grant GR077544AIA).


The authors indicate no potential conflicts of interest.