Notch, Epidermal Growth Factor Receptor, and β1-Integrin Pathways Are Coordinated in Neural Stem Cells*

Notch1 and β1-integrins are cell surface receptors involved in the recognition of the niche that surrounds stem cells through cell-cell and cell-extracellular matrix interactions, respectively. Notch1 is also involved in the control of cell fate choices in the developing central nervous system (Lewis, J. (1998) Semin. Cell Dev. Biol. 9, 583-589). Here we report that Notch and β1-integrins are co-expressed and that these proteins cooperate with the epidermal growth factor receptor in neural progenitors. We describe data that suggests that β1-integrins may affect Notch signaling through 1) physical interaction (sequestration) of the Notch intracellular domain fragment by the cytoplasmic tail of the β1-integrin and 2) affecting trafficking of the Notch intracellular domain via caveolin-mediated mechanisms. Our findings suggest that caveolin 1-containing lipid rafts play a role in the coordination and coupling of β1-integrin, Notch1, and tyrosine kinase receptor signaling pathways. We speculate that this will require the presence of the adequate β1-activating extracellular matrix or growth factors in restricted regions of the central nervous system and namely in neurogenic niches.

Notch1 is a cell surface protein involved in the control of cell fate choices in the developing CNS (1). This transmembrane receptor is involved in stem cell maintenance (6) and promotes glial and neural fates in a stepwise manner, first by inhibiting neuronal fate and promoting glial fate and second by promoting astrocyte differentiation (7). Notch1 also plays a role in the control of neurite extension in mammalian cells and in axon growth in Drosophila (8 -10). Interestingly, in the immune system Notch1 serves two biologically contrasting functions; it is responsible for the apoptotic cell death of B lymphocytes (11), whereas it promotes the survival of T cells (12). The diverse effects of Notch1 activation observed in multiple cell types and at different stages of development suggest the presence of context-dependent control mechanisms. Growth factors (GF) and ECM molecules (acting through integrins) belong to the complex environment that surrounds NSC during development (13) and that affect Notch signaling. For example, FGF-1 and -2 inhibit neural differentiation by affecting (directly or indirectly) the Notch pathway (14), and EGFR activation leads to Notch signaling during pancreas tumorigenesis (15). Integrins may also be involved in the Notch response during angiogenesis, when Notch4expressing endothelial cells display ␤1-integrin in an active, high affinity conformation (16). Furthermore, in zebrafish the boundary cells between developing somites behave differently depending on the levels of Notch activation, and it has been suggested that the extracellular matrix (which differs at the rhombomere boundaries) plays a role in this process (17). Nevertheless, the coordination between ␤1-integrin, Notch1, and GF pathways is poorly understood.
Lipid rafts are special membrane regions that affect signaling by sorting proteins and lipids into specific membrane domains, where privileged interactions occur. Caveolae are specialized lipid rafts that contain cholesterol, sphingolipids, and caveolins (22-24-kDa membrane proteins, required for the formation of the caveolae) and that serve as scaffolds for signaling molecules.
In this paper, we explore how some of the receptors for ECM and GF (that are present on the surface of the NSC) act together with the Notch1 pathway to control the NSC responses to changes in the microenvironment. We discuss the possibility that lipid rafts may play important roles in directing the changes in signaling and the responses to environmental changes that occur during cortical development and may act as integrators of parallel and simultaneous signals originated from integrins, growth factors, and Notch receptors. We conclude that the GF and ECM composition of biological neural stem cell "niches" may affect NSC maintenance and differentiation by affecting Notch signaling, in a context-and time-dependent manner.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-FGF-2 was obtained from PeProtech, EGF was from Calbiochem, and B27 supplement was from Invitrogen. Antibodies for immunoprecipitation-blocking experiments and Western blots were obtained from Chemicon and Pharmingen (anti-␤1 integrins), Santa Cruz Biotechnology, Inc. (Santa Cruz, CA) (EGFR), Upstate Biotechnology, Inc. (Lake Placid, NY) (EGFR), and Cell Signaling Technology (phosphorylated MAPK and MAPK). Antibodies used for immunohistochemistry included polyclonal Notch1 and caveolin 1 (Santa Cruz Biotechnology), monoclonal anti-Nestin (Pharmingen), and monoclonal anti-␤ III tubulin (Sigma). All fluorescent secondary antibodies were obtained from Jackson Immunochemicals or Molecular Probes, Inc. (Eugene, OR). The EGF receptor inhibitor AG1478, (Calbiochem) was used at 20 M. Mixtures of protease and phosphatase inhibitors were from Calbiochem. The remaining products were from Sigma if not otherwise specified.
To prepare ES cell-derived NSC, we used a similar approach to the one developed by Bibel et al. (21). Briefly, we generated embryoid bodies that were exposed to retinoic acid (RA) for 4 days (as described by Bain et al. (22)). Taking into account that a switch in growth factor requirements (from FGF-2 to EGF) occurs in vivo during midneurogenesis (23,24), we sequentially exposed the RA-primed embryoid bodies to FGF-2 first and to a mix of FGF-2 and EGF second, in order to simulate the changes in the GF microenvironment that occur during embryonic CNS development.
Secondary Neurosphere Formation Assays-Intact primary neurospheres maintained 8 -10 days in vitro were mechanically dissociated, and the same number of cells for each condition was plated at low density (5000 cells in 1.5 ml, Ͻ5 cells/l) and grown for 10 days in 20 ng/ml EGF and FGF-2. The number of secondary neurospheres formed was counted, and statistically significant differences between groups were calculated using Student's t tests.
Secondary Neurosphere Formation Assays after Morpholino Treatment or EGFR Inhibition-Intact neurospheres (8 -10 days in vitro) were exposed to ␤1 antisense morpholinos obtained from GeneTools, following the manufacturer's protocols (available on the World Wide Web at www.gene-tools.com). Control groups were exposed to a missense morpholino with a random sequence. The spheres were then mechanically dissociated, and the same number of cells for each condition were plated at low density (3-4 cells/l) and grown for 10 days in different growth factor concentrations (20 or 2 ng/ml concentration of either EGF or FGF-2), as described above. The number of secondary neurospheres formed was then counted, and statistically significant differences between groups were calculated using Student's t tests. Secondary neurosphere formation assays were also done in the presence of an EGFR inhibitor (AG1478, 20 M), and these experiments were analyzed as described above.
Immunohistochemistry-Neurospheres, ES cell-derived NSC, and neonatal or embryonic brain tissue were fixed in 2-4% paraformaldehyde in phosphate-buffered saline. Tissue samples were cryoprotected in 25% sucrose and sectioned (14 m) prior to immunohistochemistry, except for cell monolayers. The samples were blocked in phosphatebuffered saline (0.1% Triton X-100) containing normal blocking serum and incubated overnight with the appropriate antibodies at 4°C, followed by incubation with the secondary antibodies and counterstaining with DAPI or sytox green. Pictures were acquired using a Zeiss Axioplan 2 fluorescence microscope and Smartcapture 2 software.
GST-NICD Construct Preparation-Briefly, the Notch intracellular domain (NICD) fragment was subcloned from the plasmid pbabe-NICD (kind gift from G. Weinmaster) into the pGEX expression vector to prepare a GST-NICD fusion protein. The fusion protein was produced in Escherichia coli, linked to agarose beads (Amersham Biosciences), and purified using standard protocols. Pull-down assays were performed by incubating total lysates obtained from neurospheres with the GST-NICD or GST-alone beads, for 2 h at room temperature. The beads were then collected, washed, and boiled, and the resulting supernatant was analyzed by SDS-PAGE and immunoblotting as follows.
Western Blots and Immunoprecipitations-For Western blotting, neurospheres or ES cell-derived NSC were lysed (10 mM Tris-HCl, 5 mM EDTA, 150 mM NaCl buffer, 1% Triton X-100) in the presence of protease and phosphatase inhibitors (5 g/ml leupeptin, 2 g/ml aprotinin, 2 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin, 2 mM sodium fluoride, 2 mM sodium vanadate, all from Sigma) or the equivalent Calbiochem mixtures. The supernatant was clarified by centrifugation at 14,000 rpm for 20 min at 4°C. Protein concentrations were determined with a Bio-Rad protein assay with bovine serum albumin as a standard, and equal amounts of protein were loaded in each well. Proteins were separated by SDS-PAGE and electroblotted onto nitrocellulose membranes (Hybond-C; Amersham Biosciences). Membranes were blocked in 10% nonfat dry milk in Tris-buffered saline for 1 h at room temperature. Blots were then incubated with the primary antibodies overnight at 4°C in milk/Tris-buffered saline containing 0.1% Tween 20 (TBS-T), followed by a 2-h incubation with the appropriate secondary peroxidaseconjugated antibody (Amersham Biosciences). Blots were developed using ECL reagents (Amersham Biosciences), following the manufacturer's instructions (Amersham Biosciences). For immunoprecipitations, the samples were lysed as previously described. To remove nonspecifically binding proteins, 150 -200 g of proteins were precleared by mixing with agarose beads (A/G plus; Santa Cruz) for 30 min at 4°C. The samples were then incubated with the adequate antibody in the presence of fresh agarose A/G beads, either at 4°C overnight or at room temperature for 2 h on a rotating platform. The beads were then washed and boiled for 10 min in Laemmli loading buffer. Equal amounts (as measured by protein assay) were loaded on 10% SDS-polyacrylamide gels and processed for immunoblotting as previously described.
Raft Isolation-Isolation of rafts was performed using sucrose gradients, as previously described (25). Briefly, neurospheres were placed on ice and suspended for 30 min in 0.2 ml of extraction buffer: 50 mM Tris-HCl, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, and a mixture of protease and phosphatase inhibitors (Calbiochem). Cell lysates were then adjusted to 40% OptiPrep and overlaid with solutions of 30 and 10% Optiprep in the extraction buffer. These gradients were centrifuged for 16 h at 35,000 rpm at 4°C in an SW50.i rotor (Beckman Instruments). Fractions of equal volume, including the raft (floating fraction) and nonraft (bottom fraction) fractions, were collected and analyzed by SDS-PAGE (10%), followed by immunoblotting. Protein assays were performed on all fractions before immunoblotting to ensure equal loading.
Reverse Transcription-PCR-RNA was extracted using the RNAeasy kit (Qiagen), and 0.1 g of RNA from each sample was used to generate cDNA using the transcriptor first strand cDNA synthesis kit (Roche Applied Science). Reverse transcription-PCR was done using the published primers for Hes5 and glyceraldehyde-3-phosphate dehydrogenase and following the conditions described by Zine et al. (26)

␤1-Integrins and Notch1
Co-localize in the Ventricular Zone, in Neurospheres, and in ES Cell-derived NSC-Stem cells from the skin and prostatic epithelia can be identified and isolated by their high ␤1-integrin expression levels (27,28). In the developing brain, ␤1-integrin is expressed in the ventricular zone (VZ) (4) (Fig. 1C) by NSC that are exposed to a changing ECM and, possibly, to variable growth factor levels (13,29). Likewise, Notch1 plays a role in NSC and is thought to direct radial glial cell (RGC) differentiation (30,31). Not surprisingly, Notch1 is expressed in the same VZ region (32) (Fig. 1B), and, interestingly, it is co-expressed with the ␤1-integrins (Fig. 1D). Co-expression of Notch1 and ␤1-integrins is also detectable in neurospheres (Fig. 1, (4,29,33) and in NSC/radial glial cell cultures, derived from ES cells (Fig. 1, H-J). The observation that both proteins are simultaneously expressed in neural progenitors and that their expression overlaps raises the hypothesis that they may cooperate or act in a coordinated fashion. To further test the hypothesis that ␤1-integrins and Notch pathways interact in neural progenitors/NSC, we used primary neurospheres and ES cell-derived NSC, the later providing an NSC-enriched population positive for RGC markers (Fig. 1, K and L) (21) (currently accepted to be NSC (34)). Notch1 is highly expressed in the ES cell-derived NSC cultures and colocalizes with the ␤1-integrin (Fig. 1, H-J). The ES cell-derived NSC cultures are therefore suitable to study, in vitro, the cooperation between ␤1-integrin, growth factors, and Notch pathways, all of which are known to be crucial for NSC and RGC maintenance and development (4,31,35).
␤1-Integrin and Growth Factor Receptors Are Required for Secondary Neurosphere Generation-To test the role of ␤1-integrin in NSC, we treated primary neurospheres with morpholino antisense oligonucleotides against ␤1-integrin to decrease the ␤1 subunit protein levels in EGF-or FGF-2-grown cells, prior to secondary neurosphere formation assays (see "Experimental Procedures"). The decrease in ␤1-integrin was confirmed by Western blot (Fig. 2E). Spheres treated with antisense or missense (control) morpholino oligonucleotides were dissociated and tested for their capacity to form new spheres (secondary neurosphere formation assay; see "Experimental Procedures"). These experiments show that a decrease in ␤1-integrin is associated with a moderate decrease in secondary neurosphere formation (Fig. 2, A-D). Interestingly, in EGF-grown (A and B) and FGF-2-grown (C and D) cells, the decrease in secondary neurosphere formation is more significant at low EGF levels (2 ng/ml; Fig. 2B) than at higher EGF levels (20 ng/ml; Fig.  2A), suggesting that in the presence of high EGF levels, the cells are less dependent on ␤1-integrin to maintain adequate levels of proliferation or survival. After morpholino treatment, the decrease in secondary neurosphere formation in FGF-2-grown cells (Fig. 2, C and D) is already apparent at high levels of FGF-2 (20 ng/ml; Fig. 2C) when compared with spheres grown with low levels of FGF-2 (2 ng/ml; Fig. 2D).
EGFR and ␤1-integrin interactions have been extensively demonstrated in three-dimensional breast culture systems (36), and it is conceivable that the high levels of EGFR found on the nestin-positive FGF-2-grown spheres (Fig. 2, F-H) may be responsible for the more acute response to a decrease in ␤1-integrin observed in FGF-2-grown cells. To test this hypothesis, spheres grown in both EGF and FGF-2 were used in a secondary neurosphere formation assay in the presence of FGF-2 and an EGFR inhibitor, AG1478 (20 M). The exposure to AG1478 resulted in a sharp decrease in the number of secondary neurospheres formed, indicating that even for FGF-2-grown spheres, the EGFR is the crucial pathway involved in proliferation (Fig. 2J), as indeed suggested by the high levels of EGFR found on the FGF-2-grown cells (Fig. 2, G and H). Consequently, in FGF-2-grown spheres with decreased levels of ␤1-integrin (EGFR-strongly positive/␤1-depleted), a lack of exposure to EGF FIGURE 1. ␤1-integrins and Notch1 co-localize in the developing VZ and in neurospheres. Notch1 and ␤1-integrins are detected in the VZ during mouse gestation (4,32) and both play important roles in neural stem cell control. Ligands for the Notch1 and ␤1-integrins (delta 1 and laminin, respectively) are available in the VZ, and their expression levels change in time (4,76). High levels of ␤1-integrins (red, C) and Notch1 (green, B) are present in the embryonic day 12.5 mouse VZ. Note that Notch1 (green, B) and ␤1-integrin (red, A) are co-expressed in the VZ (yellow, D). Note the lack of overlap in blood vessels (arrow, C and D). NSC can be cultured from embryonic or postnatal CNS tissue (19,77) in suspension cultures that give rise to spheroid structures (neurospheres) that contain NSC at the edge (29), where these cells express high levels of Notch1 (29) and lex/SSEA (33). Notch1 is expressed in sectioned neurospheres (E) together with ␤1-integrin (F), and they partially overlap (yellow, G), predominantly at the edge. The edge of neurospheres contains a nestin-rich cell population, which expresses EGFR (see Fig. 2) and ␤1-integrins (4). Neural progenitors derived from ES cells express radial glial markers, Notch1, and ␤1-integrin; ES cells can be driven toward a neural progenitor fate using diverse protocols (22,53,55) and also give rise to RGC (21) that express RGC markers. Taking into account that RGCs are now generally accepted to be NSC (34,78), the ES cellderived neural progenitors can be considered to be an NSC population that can be readily grown and maintained in large numbers to analyze complex pathway interactions in NSC. The ES-derived neural progenitors express high levels of Notch1 receptor (H) and ␤1-integrins (I), like the mouse embryonic SVZ and the edge of primary neurosphere cultures. Note that Notch1 and ␤1-integrins expression patterns overlap in the ES cellderived NSC (J, yellow). The ES cell-derived NSC cultures are therefore suitable to study, in vitro, the cooperation between ␤1 and Notch1 pathways, both of which are known to be crucial for NSC and RGC maintenance and development (4,31,35). RC2 (K) and GLAST (L) are also expressed by ES cellderived NSC.
will be severely felt (despite the high levels of EGFR expression) and cannot be compensated by (lacking) integrin activation. In the EGFgrown spheres (EGFR-positive/␤1-depleted), even low levels of EGFR will be enough to respond to the EGF in the medium. These results suggest that ␤1-integrin may be important for EGFR activation in neurospheres and point toward a potential cooperation between the two pathways, as already described for epithelial cells and fibroblasts (37,38). Furthermore, it was recently shown that a decrease in ␤1-integrin causes a reduction in neurosphere size during secondary neurosphere formation assays, due to reduced progenitor proliferation and increased cell death (39). The loss of ␤1-integrin in neurospheres also reduces the number of nestin-positive cells in a growth factor-dependent manner, and this phenotype can be rescued by exposing the cells to high growth factor levels (39). This result is consistent with our morpholino experiments, and both may be explained by the signaling confluence of EGFR and ␤1-integrins toward the MAPK pathway. In fact, in the absence of ␤1-integrins, the signaling through the MAPK pathway may become more dependent on the presence of EGF in the medium (4). Interestingly, we have also observed that the addition of EGF and FGF-2 to starved neurospheres leads to an increase of the detectable levels of NICD expression by Western blot (Fig. 2I). Other authors have observed that growth factors, such as ciliary neurotrophic factor, increase NICD expression levels (40) or affect Notch activation (15,41). Taking into account that ␤1-integrins modulate the response to GF in neural progenitors (39), we raise the hypothesis that the GF effects on the levels of NICD could be partially dependent or coordinated with the activation of ␤1-integrins. How the ␤1-integrins, GF receptors, and Notch signaling pathways are coordinated in neural progenitors remains to be elucidated, and we suggest that special membrane domains may be involved in the coordination of these pathways interactions, cross-talks, and sequential and/or simultaneous effects.
Caveolin 1, a Lipid Raft-resident Protein, Is Present on Neurospheres, on ES-derived NSC/RGC, and in the Embryonic VZ-Lipid rafts are membrane domains that act as privileged signaling platforms (42), where interactions and cross-talk between different signaling pathways may take place. Recently, lipid rafts were found to be present on neuroepithelial progenitors, where they play a role in signal transduction (43). We reasoned that the embryonic VZ, the ES-derived NSC and primary neurospheres could also contain lipid rafts. Using Western blots and immunofluorescence, we confirmed the existence of caveolin 1 (a component of lipid rafts) on ES-derived NSC, in the embryonic VZ, and on neurospheres (Fig. 3, A-H  . ␤1-integrins and EGFR are required for secondary neurosphere formation. A decrease in ␤1-integrin affects secondary neurosphere formation at low EGF and at high FGF-2 concentrations. Intact neurospheres were exposed to morpholino antisense to reduce ␤1-integrin expression and then dissociated and used in secondary neurosphere formation assays in the presence of 20 ng/ml EGF (A), 2 ng/ml EGF (B), 20 ng/ml FGF (C), and 2 ng/ml FGF (D). The experiments were carried out five times for each condition. A statistically significant decrease (p Ͻ 0.001) in the number of secondary neurospheres formed after antisense treatment was observed at 20 ng/ml EGF, 2 ng/ml EGF, and 20 ng/ml FGF-2. The strongest effect was seen for 20 ng/ml FGF-2 (C), followed in order by EGF 2 ng/ml (B) and EGF 20 ng/ml (A), suggesting that the morpholino-treated FGF-2-grown spheres are more dependent on ␤1-integrin than the EGF-grown ones. The decrease in ␤1-integrin was ascertained by Western blot (E). Samples grown in EGF or FGF-2 and either exposed to the morpholino (m) or to the missense (ms) were run in parallel after protein quantification. Note that the EGFR is markedly up-regulated at the nestin-rich edge of FGF-2-grown spheres, as detected by immunohistochemistry on sectioned neurospheres (F, nestin; G, EGFR) and confirmed with Western blot (H). The nuclei in F and G are labeled with DAPI (blue). To test the role of the EGFR activation in neural progenitor proliferation, intact mouse spheres (grown in a mix of EGF and FGF-2 for 8 days) were used for secondary neurosphere formation assays in the presence of FGF-2 and of an EGFR inhibitor (AG1478; 20 M), as described under "Experimental Procedures" (J). Control cells were exposed to Me 2 SO alone. Exposure to the EGFR inhibitor induces a statistically significant reduction (p Ͻ 0.001, n ϭ 3) in the number of secondary spheres formed, in the presence of FGF-2. Note that EGFR is highly expressed at the edge of neurospheres (G), in the region where nestin-positive progenitors are abundant (F). Exposure of neurospheres to growth factors affects NICD levels; the addition of EGF and FGF-2 to 24-h growth factor-starved neurospheres leads to an increase in detection of Notch intracellular domain (NICD) by Western blot (I). The experiment shown is representative of various replicates (n Ͼ 3). Note that as a result of EGF or FGF-2 addition (after growth factor starvation) MAPK is phosphorylated, whereas total levels of MAPK remain similar, indicating a strong and specific response to the growth factors. Equal amounts of proteins were loaded in each lane. WT, wild type. FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 context-dependent modulation of Notch (or other receptors) may require the simultaneous activation of (or the interaction with) other membrane receptors, like ␤1-integrins, promoted by specific signaling platforms like the lipid rafts, in a temporally and spatially controlled manner. Rafts can be isolated within detergent-insoluble glycosphingolipid-rich microdomains by density gradient centrifugation at 4°C (47). We examined the distribution of Notch1 in Optiprep density gradient fractions of Triton X-100 extracts prepared at 4°C. We reasoned that changes in the protein levels in rafts (membrane compartments of specific lipid composition that are privileged for receptor interactions and act as signaling platforms (42,48)) are more relevant than the overall changes in expression and allow for subtle changes to be detected in functional fractions. Lipid rafts are insoluble in Triton X-100 and float at the 10 -30% interface of the density gradient, and we used the presence of caveolin 1 to confirm that the raft fraction had been correctly identified. Note that we also observed that caveolin 1 (a raft-resident protein) is present on the cells located at the edge of neurospheres (Fig. 3C) in ES-derived NSC (Fig. 3F) and in the embryonic VZ (Fig. 3B). To test the effect of growth factors on neural progenitors, we starved (overnight) intact EGF-and FGF-2-grown neurospheres and then added EGF or FGF-2, respectively (20 ng/ml), for 24 h. Using intact neurospheres ensured that only the cells located at the edge (enriched in ␤1-integrins, EGFR, and Notch1 (29)) were exposed to the changing environment.

Notch and ␤1-Integrin Interactions in Neural Stem Cells
The lysates from spheres treated in this manner were centrifuged on Percoll gradients, and the resulting fractions were analyzed for the presence or absence of Notch1 protein in caveolin 1-containing membrane compartments, with or without growth factor activation (Fig. 4, A and B). When EGF is added to starved EGF-grown neurospheres (Fig. 4A, EGF panel), an increase in NICD is detected in the nonraft fraction (lane 1) at the expense of the caveolin 1-positive fractions (lanes 2-5 in Fig. 4B). Likewise, FGF-2 produces the same effect when added to starved FGF-2-grown neurospheres (Fig. 4A,  FGF-2 panel).
Taking into account that ␤1-integrins modulate the response to GF in neural progenitors (39), we raised the hypothesis that the GF effects on the levels of NICD could be partially mediated by ␤1-integrins. If so, activation of ␤1-integrins should also lead to activation of the Notch pathway.
In Neurospheres, the Activation of ␤1-Integrins with Mn 2ϩ Mobilizes NICD out of the Caveolin-containing Fractions-To further characterize the mechanisms involved and to challenge the role of ␤1-integrins, we stimulated ␤1-integrins in intact spheres and in ES-derived neural stem cells, using manganese (Mn 2ϩ ). Mn 2ϩ is a divalent cation known to activate ␤1-integrins (49) and to induce the redistribution of LFA-1, ␣4␤1 (50), or ␣6␤1 (47) into lipid rafts. Whereas no significant change in Notch1 distribution occurred after Mn 2ϩ treatment in EGF-grown spheres, Mn 2ϩ leads to the redistribution of Notch from caveolin-positive to caveolin-negative fractions in FGF-2-grown neurospheres (Fig.  4A, FGF-2 panel). The specificity of this effect on neurospheres was analyzed by exposing the suspension cultures to a ␤1-integrin blocking antibody (Ha2/5), prior to Mn 2ϩ exposure (Fig. 4C). We observed that preincubation of the starved neurospheres with the blocking ␤1-integrin antibody decreases the Notch mobilization induced by the addition of Mn 2ϩ (Fig. 4) in both EGF-and FGF-2-grown NS cultures, with a greater proportion of NICD remaining in the caveolar compartment. When combined with our ␤1-integrin loss of function (morpholino) data, these results suggest that a decrease in ␤1-integrin may affect neurosphere formation assays by altering Notch 1 processing/transfer to the nuclear fraction and thus preventing its downstream proliferative actions. Interestingly, integrin-mediated adhesion has been proposed to govern the presence of cholesterol-enriched microdomains or lipid rafts on the plasma membrane by controlling internalization via a caveolindependent pathway (51). Together with our results, these experiments raise the possibility that ␤1-integrins play a role in the growth factormodulated transfer of NICD between membrane compartments and/or into the nucleus, possibly via lipid rafts.

␤1-Integrin Activation with Mn 2ϩ Leads to Movement of NICD into the Nucleus in ES-derived NSC-
To test whether the activation of ␤1-integrins affects Notch internalization, we exposed cells to Mn 2ϩ (a divalent cation known to activate ␤1-integrins (49)) and evaluated the movement of NICD to the nucleus, by immunocytochemistry. Mn 2ϩ is known to activate ␤1-integrins and to induce the redistribution of ␤1-integrins into lipid rafts. Using an antibody specific for NICD (raised against an epitope exposed only after cleavage), we observed that ␤1-integrin activation with manganese led to a shift of NICD into the nucleus, detectable by immunohistochemistry (Fig. 5). This experiment was technically less demanding to do on monolayers of ES-derived NSC than on three-dimensional neurospheres, where the cell-cell contacts and large amounts of ECM present account for activation of integrins, even when GF levels are low.
␤1-Integrin Activation with Mn 2ϩ Is Followed by Changes in HES5 mRNA Expression-To study the effects of ␤1-integrin activation on the Notch1 downstream pathway, we used Mn 2ϩ to induce a change in the ␤1-integrin conformation and activation state, as previously described (49). Other molecules known to activate integrins were also used, namely ECM molecules (laminin 1, laminin 2, and fibronectin) and EGF. It is noteworthy that fibronectin and laminin can cause EGFR phosphorylation, through ␤1-integrin activation (37,38,52), and EGF may also activate ␤1-integrin, through cross-talk between the EGFR and ␤1-integrin. Cells grown in EGF and FGF-2 were starved overnight and then stimulated with Mn 2ϩ , GF, or ECM for 3 h. Changes in HES5 mRNA expression were evaluated by reverse transcription-PCR (Fig. 5). This experiment revealed that Mn 2ϩ , EGF, EGF ϩ FGF-2, and ECM all lead to HES transcription, whereas FGF-2 alone does not (Fig. 5). The GF response pattern suggests that ␤1-integrin activation may affect Notch activation and HES transcription through the EGFR, which can be activated through ␤1-integrin stimulation by ECM ligands or EGF but not by FGF-2 alone. In fact, recent evidence shows that fibronectin and laminin can cause EGFR phosphorylation through ␤1-integrin activation (52). Furthermore, in breast tumor cells, ␤1-integrin and EGFR pathways are known to be coupled and interdependent (36), and null ␤1-integrin NSC are more reliant on high growth factor levels (for survival and proliferation) than wild type NSC (39). Finally, EGFR activation is important for neurosphere generation, and by blocking this receptor with an inhibitor, the number of secondary spheres produced after passaging decreases dramatically (as previously shown in Fig. 2J).
␤1-Integrins Co-immunoprecipitate with NICD-To further understand how Notch1 and ␤1-integrins pathways could simultaneously affect stem cell behavior, we searched also for direct interactions between the two proteins. We used neurosphere lysates to look for physical interactions between Notch1 and ␤1-integrins, both of which are highly expressed by the cortical layer of the EGF-and FGF-2-grown neurospheres (4,29). Using lysates from 8 -10 days in vitro neurospheres incubated with an antibody against ␤1-integrin and blotted with anti-Notch1, we found that ␤1-integrin and Notch1 co-immunoprecipitate (Fig. 6A), indicating that they are present in the same protein complex. The specificity of this interaction was confirmed by a reverse co-immunoprecipitation. For this purpose, a GST-NICD protein (generated as described under "Experimental Procedures") was used to pull down the ␤1 and ␣6 subunits in EGF-and FGF-2-grown neurospheres (Fig. 6, B and C).

DISCUSSION
In this paper, we show that Notch1 and ␤1-integrins are co-expressed in ES-derived NSC, neurospheres, and in the mouse embryonic VZ. We  ) were starved overnight and exposed to culture medium, culture medium with growth factors, or culture medium containing Mn 2ϩ for 12 h. A and B show the same fractions of cell lysates, blotted with a Notch antibody (A) or a caveolin 1 antibody (B), to identify fractions that contain caveolin 1 and Notch1. This experiment shows that the NICD is mobilized from a caveolin-positive fraction (lane 2) to a caveolin 1-free fraction (lane 1) when EGF is added to EGF-starved neurospheres and when FGF-2 is added to FGF-2-starved neurospheres. This suggests that the addition of growth factor stimulates Notch1 transfer between compartments. ␤1-integrins can be activated by Mn 2ϩ . When Mn 2ϩ is added to FGF-2-grown neurospheres, NICD also moves between compartments, but when Mn 2ϩ is added to EGF-grown neurospheres, no NICD movement between compartments is detected. Note that the protein concentrations for each of the fractions (collected as described under "Experimental Procedures") were normalized prior to the Western blot analysis, thus ensuring equal loading of the gels. The presence of caveolin 1 was tested on all fractions, but only fractions 6 to 1 are shown here. Cav1, caveolin1. Blocking ␤1-integrins with an antibody decreases the NICD mobilization induced by Mn 2ϩ . Neurospheres grown in either EGF or FGF-2 (C) were starved overnight and then exposed sequentially to a ␤1-integrin blocking antibody for 1 h and to Mn 2ϩ overnight. The controls were not incubated with the blocking antibody and were exposed to Mn 2ϩ overnight. The neurospheres were then harvested, lysed, and spun on Optiprep gradients to isolate different membrane compartments, as previously described. The same amounts of proteins per fraction were analyzed by Western blots with anti-Notch1 and anti-caveolin 1 antibodies. This experiment shows that preincubation with a ␤1-integrin blocking antibody decreases the response of Notch1 to Mn 2ϩ , indicating that ␤1-integrins may play a role in the growth factor-dependent Notch1 activation. The growth factor-and integrin-dependent activation of Notch1 might explain the context-dependent effects of this protein throughout neurogenesis.
report that the Notch pathway cross-talks with the ␤1-integrin pathway, as indicated by the interaction detected between the NICD fragment and the ␤1-integrins. Furthermore, GF (EGF and FGF-2) cause an increase of NICD levels, and activation of ␤1-integrins with Mn 2ϩ induces NICD internalization, via caveolin1-enriched rafts, with the appearance of the NICD fragment in the nucleus. These results suggest that the GF may play a role by enhancing the level of NICD, whereas the ␤1-integrins could modulate how much of it reaches the nucleus by regulating the internalization and travel through the endocytic compartments. Therefore, the ␤1-integrin may act dually, as an NICD buffering/sequestering system and as an internalization control, activated by extracellular cues. The ECM/GF environment could play an important modulatory role on NSC behavior through ␤1-integrins, which may act by controlling the nuclear availability of NICD in a context-dependent manner (possibly by playing a role in Notch internalization, through caveolin 1-containing membrane domains), making the Notch pathway modulable by extrinsic factors such as GF and the ECM.

ES-derived Neural Progenitors Express ␤1-Integrins, Notch1, and Caveolin 1 Like the Progenitors Present in the VZ and in Neurospheres-
Previously, several groups have described methods to generate neural progenitors/NSC, glial cells, and differentiated neurons from ES cells (53)(54)(55)(56)(57). In this paper, we generate a population of neural progenitors/ NSC (that express radial glial markers, ␤1-integrins, and Notch1) by exposing ES cells, sequentially, to a variable environment. This culture system can be used to study the underlying mechanisms of neural development. In particular, changes in protein signaling that occur during NSC generation and differentiation can be analyzed in these cells, and the context-dependent effects of proteins such as Notch are more amenable to analysis using the ES-derived NSC cultures than using the heterogeneous neurosphere and VZ explant cultures.
␤1-Integrins, ECM, and GF Participate in a Complex Network of Interactions That Affect NSC Behavior-The proteins that constitute the neurogenic niche participate in more than one regulation loop. For example, during development, FGF signaling plays a role in neural and mesodermal cell induction, mediated by Ets and GATA transcription factors (58). FGF also promotes changes in the cell responsiveness to the environment by increasing the expression of ␤1-integrins, laminin, and EGFR on neural progenitors (4,23,24,59,60). Furthermore, FGF-2 affects Notch signaling (61) and regulates neuronal differentiation by poorly defined interactions with Notch (14). Taken together, these observations suggest that FGF-2 increases the levels of receptors and ligands and prepares the NSC to become responsive to subsequent waves of growth factors and to a changing ECM. Interestingly, two of FIGURE 5. ␤1-integrin activation in ES cell-derived NSC leads to movement of NICD into the nucleus and to changes in HES5 mRNA expression. To evaluate if the NICD mobilization leads to transport into the nucleus, ES cell-derived NSC were starved overnight and then exposed to Mn 2ϩ for 30 min to activate ␤1-integrins. The cells were then fixed and used for immunofluorescence. NICD was detected using a polyclonal antibody that is specific for the intracellular portion of Notch1, released after the second cleavage. A-D show the control cell expression of ␤1-integrin (A and I in red), Notch 1 (B and J in green), nuclear DAPI (C; blue in K), and a merged image (D and L) when the cells are not exposed to Mn 2ϩ . Note that Notch1 is detected mainly in the cytoplasm (B and D). E-H show the same stainings in cells that were starved overnight and then exposed to Mn 2ϩ : ␤1-integrin (E and M in red), Notch1 (F and N in green), and nuclear DAPI (G and O in blue) and a merged image (H and P). Note that NICD is detected in both the cytoplasm and the nucleus in F and H (compare B and F). Note also that in dividing cells, Notch1 and ␤1-integrin overlap (yellow; arrow in L). This experiment was technically less demanding to do on monolayers of ES cell-derived NSC than on three-dimensional neurospheres, where the cell-cell contacts and large amounts of ECM present may account for activation of integrins, even when GF levels in the medium are low. Furthermore, these cultures represent faithfully the radial glial cell/NSC population (79). Q, the addition of Mn 2ϩ , ECM, and GF to ES cell-derived NSC increases the expression of HES5 mRNA when compared with starved cells. Lanes 1-9, cells were starved overnight and then exposed to laminin 1 (lane 1, L1), laminin 2 (lane 2, L2), fibronectin (lane 3, FN), Mn 2ϩ (lanes 6 and 7, Mn), a mix of EGF and laminin 1 (lane 8, EL1), or a mix of EGF and laminin 2 (lane 9, EL2). Lanes 4 and 5 show cells that were starved (st) overnight and not exposed to either GF or ECM molecules. HES5 mRNAs were detected by reverse transcription-PCR. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA expression was used as a control.
the receptors up-regulated by FGF-2 (EGFR and ␤1-integrins) are known to cross-talk in epithelial cells (36,52), and manipulation of either pathway can overcome deficits in ␤1-integrin or EGFR signaling in NSC (39). Furthermore, in Drosophila, a dynamic interplay exists between Notch and EGFR signaling, and both antagonistic and synergistic/additive effects have been described (62,63). In vertebrates, cross-FIGURE 6. ␤1-integrins co-immunoprecipitate with NICD. A, Western blot with anti-Notch1 of lysate prepared from 10 days in vitro neurospheres grown in FGF-2 (lane 1) or EGF (lane 2). Co-immunoprecipitation of ␤1-integrin and Notch1 lysates obtained from EGF-or FGF-2-grown neurospheres; lysates from EGF-or FGF-2-grown neurospheres were used to immunoprecipitate (IP) ␤1-integrin with a polyclonal rabbit anti-␤1 antibody. The proteins were resolved on SDS gels, transferred to membranes, and blotted for the presence of Notch1 using a goat polyclonal antibody. This experiment indicates that Notch1 coimmunoprecipitates with ␤1-integrins. A band with the size of 110 (cleaved Notch 1) is detected in lanes 4 and 5 (FGF-2-grown neurospheres) and in lanes 7 and 8 (EGF-grown neurospheres). Note that it is the same band as the one detected in the Western blot (prepared from aliquots of the same lysates; lanes 1 and 2). A larger band is also observable on the blot, possibly a multiprotein complex. Lanes 3 and 6 are preclear controls, where the lysate is incubated with beads only, to detect nonspecific attachment/interactions. B and C, to check the specificity of the Notch-␤1 interaction, the reverse immunoprecipitation was done using a GST-NICD fusion protein (or GST protein alone, as a control) to pull down ␤1-integrins in lysates from EGF-grown (B) or FGF-2-grown (C) neurospheres. The pool down with GST-NICD was followed by blotting with a polyclonal antibody against ␣6␤1 and reveals the adequate ␣6 band (B and C, top arrowhead) and ␤1 band (B and C, lower arrowhead). EGF, EGF-grown spheres. FGF-2, FGF-2grown spheres. IP, immunoprecipitation. Role for a direct NICD-␤1 interaction in the ventricular zone. ␤1-Integrins are very abundant in dividing VZ cells. The arrows in D show cells that express high levels of integrins, and arrows in E point at the dividing cells. Note the condensation of laminin (laminin ␣2-rich) in the VZ (F, arrows), indicating that the adequate ligand for ␤1-integrins is available in the VZ. Furthermore, in the VZ, ␤1-integrins, and Notch1 expression overlap (see Fig. 1D). We propose that in the presence of the adequate ECM or GF, ␤1-integrins in the VZ restrict the movement of the cleaved NICD by tethering it to the membrane during symmetric (G) or asymmetric (H) divisions. In both cases (G and H), this ensures that only cells anchored to the ECM proceed to retain NSC characteristics (sustained selfrenewal, blockage of differentiation, and survival). Whereas in symmetric divisions (G) both cells are anchored and behave equally, in the second case (H) the anchorage to the ECM is not equal, and the retention of ␤1-integrins in the most apical cell conditions the relative availability of NICD in the two daughter cells, ultimately affecting cell fate. Our co-immunoprecipitation data indicate that this tethering occurs and that ␤1-integrins interact with NICD (A-C). The overlap between ␤1-integrins and actin markers like phalloidin (13) in the VZ reinforces the hypothesis that ␤1-integrins help to anchor some crucial molecules and signaling complexes in a polarized manner in the VZ. Laminin in G and H is depicted in orange and predominates along the ependymal ventricular surface (as seen in F), whereas fibronectin is predominantly expressed outside the VZ and toward the pial surface of the developing neural tube. NICD, Notch1 intracellular domain; ϩϩϩ NICD, high levels of NICD; ϩ NICD, lower levels of NICD; VZ, ventricular zone. FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 talk between transforming growth factor ␤ and Notch occurs (64), and EGFR activation leads to Notch signaling during pancreas tumorigenesis (15). These examples indicate that some functions of Notch may be context-dependent and could require complex interaction loops with other signaling pathways.

Notch and ␤1-Integrin Interactions in Neural Stem Cells
Direct interactions can also explain context-dependent signaling effects; our co-immunoprecipitation data (Fig. 6, A-C) suggest that ␤1-integrins (which are very highly expressed in the VZ, together with laminin 2; Fig. 6, D-F) may sequester the NICD fragment. This raises the possibility that the ␤1-integrin-associated Notch1 mobilization to the nucleus could be a context-dependent event, partially regulated by the extracellular environment. In fact, integrins can regulate Rac targeting by internalization of membrane domains, such as lipid rafts (65), and in Drosophila selective endocytic pathways are required for the Delta/ Serrate/LAG-2 family to activate Notch (66,67). Therefore, it is conceivable that the effect of integrin activation on Notch may require lipid raft internalization of Notch-containing membrane domains. Our finding that ES-derived NSC, the embryonic mouse VZ, and neurosphere primary cultures are all very rich in caveolin 1 (a component of lipid rafts) supports the hypothesis that these microdomains may play important roles in the coordination between signaling pathways in NSC. Lipid rafts may play a crucial role in controlling signaling in a spatial and temporal manner (68). Therefore, the role of lipid rafts and their biological significance in NSC could be to bring together in the same domain two proteins (Notch and ␤1-integrin) at defined developmental time points, allowing for interactions to occur between multiple signaling pathways or, alternatively, implementing direct physical interactions between the two proteins (Fig. 6, G and H). The integrative role of lipid rafts may help to explain the context dependence of receptor signaling. For example, whereas Notch1 and ␤1-integrins are both expressed in ES cells and ES-derived NSC, caveolin 1 is more abundant in the latter (data not shown), perhaps facilitating direct or indirect interactions between ␤1-integrin and Notch1 pathways, preferentially in neural progenitors.
Our results (Fig. 6, A-C) suggest that the cytoplasmic tail of the ␤1-integrin interacts (directly or as part of a protein complex) with the NICD fragment and alters the biological availability of the NICD fragment, modulating the amount that can reach the nucleus. The interaction between the cytoplasmic domains of Notch1 and ␤1-integrins suggests a speculative model, whereby direct linkage of the two proteins could be affected by the ligation state of the integrin. We propose that ␤1-integrins "mop up" excess free NICD (Fig. 6, G and H) under specific conditions (e.g. during cell division) in the presence of the adequate ECM-like laminin ␣2-containing laminins (which abounds in the ependymal surface of the VZ and may "anchor" the progenitor/NSC cells) or through cross-talk with tyrosine kinase receptors like the EGFR. This mechanism could be important to modulate the level of transcriptional regulation activity of NICD and to control effects on survival, proliferation, and differentiation. In turn, the "release" of the NICD fragment allows it to reach the nucleus to promote the maintenance of an undifferentiated fate (Fig. 6, G and H).
A model that relies on external factors to alter the equilibrium between bound and free intracellular signaling molecules has also been suggested for the regulation of ␤-catenin signaling, where a balance between bound and free levels of ␤-catenin is partially controlled by multiple interactions between cadherins and receptor tyrosine kinases (69). Likewise, proteins like Notch that depend on regulated intramembrane proteolysis for signaling (70) require effective mechanisms to control downstream signaling. We suggest that spatial control can be achieved by a "buffering system" that keeps levels of NICD balanced in the presence of the right ECM. The model we propose predicts that an adequate balance of free and bound NICD is maintained only in the cells that are in contact with the adequate matrix or in the presence of specific growth factors (Fig. 6, G and H), further highlighting the important role played by niches during neural development. The model predicts that changes in ␤1-integrin levels could lead to an imbalance in NICD levels. Interestingly, NICD overexpression in a chondrogenic cell line inhibits differentiation and decreases proliferation (71), and, likewise, a decrease in ␤1-integrin in chondrocytes causes diminished proliferation and changes in the G 1 /S transition and cytokinesis (72). Both of these observations may be due to an imbalance (increase) in free NICD, provoked by the NICD overexpression or by the decrease in ␤1-integrins, respectively. If changes in ␤1-integrin expression cause an imbalance in NICD levels, the increase in NICD availability that we observe when GF are added to the neural progenitor cultures may explain the proliferation rescue of the ␤1-integrin null cells by GF (39). Interestingly, the addition of GF to ␤1-integrin null cells also increases the number of nestin-expressing cells and decreases differentiation, an effect that could also be explained by activation of the Notch pathway.
An additional role for ␤1-integrin could be to provide survival signals (in a context-dependent manner that depends on the ECM composition) to counteract the potentially deleterious effect of excessive Notch activation. For example, it is known that in B lymphocytes, Notch1 induces cell cycle arrest and apoptosis (11), whereas in T cells Notch1 has an antiapoptotic function (12). Inhibition of Notch by Numb in the Drosophila serotonin lineage causes cells to differentiate, whereas cells that retain Notch signaling initiate apoptosis (73). Interestingly, the onset of mammary apoptosis in the mouse mammary gland coincides with a change of conformation between ligand-bound and unbound ␤1-integrin (74). We speculate that the equilibrium between apoptosis and survival in neural progenitors could be due to a balance between the levelsof␤1-integrinsandthelevelsofNICDandthereforehighlycontextdependent. During cortical development, once the cells abandon the laminin-rich "VZ niche" (4) (Fig. 6, G and H), they encounter a different matrix and/or growth factor environment, which may favor noncanonical Notch biological roles in differentiation or cell death. This proposed model could also account for the context-dependent effects of Notch1 described for different malignant cells (75).
In summary, Notch1 is a cell surface protein involved in the control of cell fate choices in the developing CNS (1). Our data suggest that the Notch1 pathway is partially dependent on the integrin/ECM/GF environment. ␤1-integrins may act on Notch signaling through 1) physical interaction (sequestration) of the NICD fragment by the cytoplasmic tail of the ␤1-integrin, 2) by affecting trafficking of the NICD. The ligation state of ␤1-integrins could therefore "fine tune" Notch activation/ processing in a changing ECM and growth factor environment, resulting in differential effects on cell fate according to the microenvironment present at a given moment in time. The different roles attributed to Notch during cortical development, such as the role in the sequential generation of radial glia and of neurons (30,31) or the later role of Notch in neurite extension (8), could be explained by a context-dependent modulation of the pathway, dependent on the ECM composition and GF availability (13). We conclude that, through a ␤1-dependent Notch1 pathway modulation, the ECM and GF in the immediate vicinity of NSC may participate in neural stem cell fate determination in neurogenic niches. Furthermore, our findings suggest that caveolin1-containing lipid rafts play a role in the coordination and coupling of ␤1-integrin, Notch 1, and TKR signaling pathways. We speculate that this will require the presence of the adequate ␤1-activating ECM or GF in restricted regions of the CNS and namely in neurogenic niches.