Phosphatidylinositol 3-Kinase Interacts with the Adaptor Protein Dab1 in Response to Reelin Signaling and Is Required for Normal Cortical Lamination*

Reelin is a large secreted signaling protein that binds to two members of the low density lipoprotein receptor family, the apolipoprotein E receptor 2 and the very low density lipoprotein receptor, and regulates neuronal positioning during brain development. Reelin signaling requires activation of Src family kinases as well as tyrosine phosphorylation of the intracellular adaptor protein Disabled-1 (Dab1). This results in activation of phosphatidylinositol 3-kinase (PI3K), the serine/threonine kinase Akt, and the inhibition of glycogen synthase kinase 3β, a protein that is implicated in the regulation of axonal transport. Here we demonstrate that PI3K activation by Reelin requires Src family kinase activity and depends on the Reelin-triggered interaction of Dab1 with the PI3K regulatory subunit p85α. Because the Dab1 phosphotyrosine binding domain can interact simultaneously with membrane lipids and with the intracellular domains of apolipoprotein E receptor 2 and very low density lipoprotein receptor, Dab1 is preferentially recruited to the neuronal plasma membrane, where it is phosphorylated. Efficient Dab1 phosphorylation and activation of the Reelin signaling cascade is impaired by cholesterol depletion of the plasma membrane. Using a neuronal migration assay, we also show that PI3K signaling is required for the formation of a normal cortical plate, a step that is dependent upon Reelin signaling.

ceptor and apoE receptor 2 (apoER2) function as Reelin receptors and interact with the cytoplasmic adaptor protein Disabled-1 (Dab1) through an Asn-Pro-X-Tyr (NPXY) tetra-amino acid motif in their cytoplasmic domains (for review, see Ref. 4). Binding of Reelin to its receptors induces tyrosine phosphorylation of Dab1 in neurons (5). This event is essential for its function (6) and requires the activity of Src family tyrosine kinases (SFKs) (7,8). Mice that lack either reelin, dab1, or apoer2 and vldlr display the typical reeler phenotype, which is characterized by inversion of cortical layers, cerebellar dysplasia, and ataxia.
We have recently reported that Reelin activates phosphatidylinositol-3-kinase (PI3K) in neurons, leading to activation of the serine/threonine kinase Akt (also known as protein kinase B) and inhibition of glycogen synthase kinase 3␤, a major kinase for the microtubule-associated protein tau. Reelin-dependent activation of this signaling pathway is apoE receptordependent and requires Dab1 (9).
Three classes of PI3K catalyze the formation of phosphoinositides (PIs). These membrane lipids function as second messengers and also mediate the recruitment of adaptor and scaffolding proteins to specific membrane compartments. PI3K signaling, thus, participates in diverse cellular and physiological processes including cell proliferation, inhibition of apoptosis, differentiation, cell motility, membrane trafficking, endocytosis, metabolic regulation, and neoplastic transformation (for reviews, see Refs. 10 -12). Only class 1 PI3 kinases are sensitive to the widely used inhibitors wortmannin and LY2944002, which inhibit the formation of phosphatidylinositol 3,4,5-trisphosphate and, thus, prevent the PI3K-dependent activation of Akt. Class 1 PI3Ks are heterodimers of the p110 catalytic subunit and a regulatory subunit. The latter are multidomain proteins that exist in several isoforms, of which p85␣ is the longest one. Binding of the regulatory subunit to activating membrane-associated proteins, such as insulin-regulated substrate 1 (IRS1), frequently occurs at phosphorylated tyrosine residues (for review, see Ref. 13). This relieves inhibition on p110 and also brings the complex in proximity to its substrate in the membrane (14 -18).
Dab1, like IRS1, is a PTB domain protein that is tyrosine-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  phosphorylated in response to an extracellular signal and binds to NPXY motifs in the cytoplasmic domains of several transmembrane proteins (for review, see Ref. 4). These similarities prompted us to examine whether Dab1 is directly involved in the Reelin-mediated PI3K activation. We also investigated the role of PIs in Dab1 membrane targeting and Reelin signaling. Phosphatidylinositol 4,5-bisphosphate has been reported to regulate cytoskeletal reorganization through accumulation in cholesterol-rich plasma membrane compartments called rafts or caveolae (for reviews, see Refs. 19 and 20). These membrane microdomains are enriched in tyrosine kinases and scaffolding proteins and play an important role as cellular signal transduction platforms (for reviews, see Refs. [21][22][23]. Additionally, apoER2 localizes to caveolae in stably transfected Chinese hamster ovary cells (24), and caveolar localization is crucial for the platelet-derived growth factor-induced tyrosine phosphorylation of low density lipoprotein receptor-related protein 1, another member of the low density lipoprotein receptor gene family (25).
Here we show that activation of PI3K in neurons is regulated by a Reelin-dependent interaction of phosphorylated Dab1 with the PI3K regulatory subunit p85␣. We show that the Dab1containing Reelin signaling complex is associated with the plasma membrane and that the signaling can be modulated by cholesterol depletion of the plasma membrane. Finally, we demonstrate that PI3K activity is required for formation of a normal cortical plate in an in vitro cortical slice migration assay.
Primary Rat Embryonic Neuron Culture-All animal experiments were conducted according to procedures approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Texas Southwestern Medical Center Dallas. Timed pregnant Sprague-Dawley rats were purchased from Charles River Laboratories (Wilmington, MA). Primary forebrain neurons were prepared from E18 -20 embryos as described (8) and cultured at ϳ 1 million cells/cm 2 on poly-D-lysinecoated cell culture dishes in a defined serum-free medium (Neurobasal, 2% B-27 supplement, 1 mM glutamine, supplemented with 100 units/ml penicillin and 100 g/ml streptomycin sulfate) at 37°C in a 5% CO 2 atmosphere. Between culture days 4 and 6 primary neurons were treated with partially purified Reelin or control-conditioned serum-free medium for the indicated times as described (8).
Treatment of Cultured Neurons with Inhibitors, Cholesterol Depletion, and Filipin Stain-Neurons were incubated with various inhibitor compounds for 60 min at a final concentration of Me 2 SO in the culture medium of Յ0.1% (v/v) without any detectable adverse effect. For cholesterol depletion experiments, neurons were treated with 5 mM M␤CD or with 5 mM M␤CD preloaded with cholesterol (final cholesterol concentration, 100 g/ml). For detection of free cholesterol, neurons were fixed in 2% paraformaldehyde in PBS supplemented with 50 g/ml filipin for 30 min, washed 3 times with PBS, mounted using the Pro-Long antifade kit from Molecular Probes (Eugene, OR), and analyzed by fluorescence microscopy (excitation at 360 nm).
Cellular Fractionation of Primary Embryonic Neurons-After stimulation, neurons were harvested in ice-cold hypotonic buffer A (20 mM Tricine, pH 7.8, 0.25 M sucrose, 1 mM EDTA, supplemented with protease and phosphatase inhibitor cocktails), subjected to a pressure of 500 p.s.i. for 15 min, and forced through a small hole by slowly releasing the pressure. Nuclei were removed by centrifugation at 900 ϫ g for 10 min, and the postnuclear supernatants were separated into membrane and soluble fractions by ultracentrifugation at 200,000 ϫ g for 60 min. Membrane fractions were resuspended in Tris buffer, pH 7.5 (50 mM, 0.15 M NaCl, with protease and phosphatase inhibitor cocktails) containing 1% Triton X-100, incubated at room temperature, and homogenized by repeatedly forcing them through a 27-gauge needle. The homogenate was clarified by centrifugation at 20,000 ϫ g for 15 min, and the protein contents of solubilized membrane and soluble fractions were determined using the Bio-Rad detergent-compatible assay. Denatured and reduced samples were separated by SDS-gel electrophoresis (15 g of protein/lane) and analyzed by immunoblotting.
Co-immunoprecipitation Experiments-Control or Reelin-treated neurons were washed with ice-cold PBS containing 10 mM NaF, 25 mM ␤-glycerophosphate, and 2 mM sodium orthovanadate (Sigma), centrifuged at 900 ϫ g for 5 min at 4°C, and resuspended in 1 ml of immunoprecipitation buffer (50 mM Tris, pH 7.5, 0.15 M NaCl, 1 mM MgCl 2 ,1% Triton X-100, supplemented with protease and phosphatase inhibitor cocktails). The neurons were homogenized by forcing through a 27-gauge needle and incubated on ice for 30 -60 min. Lysates were cleared by centrifugation at 20,000 ϫ g and 4°C for 30 min. 800 -1000 g of protein were mixed with 4 g of the indicated IgG or 10 l of antiserum in a total volume of 1 ml and incubated on ice for 1 h. Immune complexes were precipitated with 50 l of protein A-agarose (for rabbit polyclonal antibodies) or protein G-Sepharose (for mouse monoclonal antibodies) slurry. Pelleted beads were washed once with immunoprecipitation buffer, then three times with washing buffer (immunoprecipitation buffer containing 0.1% Triton X-100), resuspended in 50 l of SDS loading buffer, reduced with ␤-mercaptoethanol (Sigma), and boiled. Approximately 5% of the total immunoprecipitate was used to detect the respective immunoprecipitated protein; the rest was used to detect co-precipitated proteins by immunoblotting.
Embryonic Slice Culture Assay-Entire fetal brains at E13.5 were embedded in 4% low melting agarose (prepared in Dulbecco's modified Eagle's medium/Hanks' F-12 medium with glutamine, glucose, and HEPES, from BioWhittaker) and glued on a vibratome support using cyanoacrylate. 3000-m-thick coronal sections were cut, and a slice was processed for histology immediately after sectioning to verify the developmental status before culture. Sections were cultured on collagencoated polytetrafluoroethylene membranes (Transwell-COL, Costar cat. 3494) in 12-well plates, and 1.5 ml of medium was added to a level just covering the section. Cultures were incubated in serum-free Dulbecco's modified Eagle's medium/F-12 supplemented with B-27 (1/50), G-5 (1/ 100), penicillin/streptomycin for 2 days in the absence or presence of 50 M LY294002. Slices were fixed in Bouin's fixative for 2 h before embedding in paraplast, sectioning, and staining with hematoxylin/eosin.

Activation of PI3 Kinase by Reelin Requires SFK Activity-
Activation of SFKs and PI3K in neurons in response to Reelin requires Dab1 and the apoE receptors VLDLR and apoER2. To determine the order in which SFKs and PI3K are activated by Reelin we used kinase inhibitors on cultured primary embryonic neurons (Fig. 1). The SFK inhibitor PP2 blocked not only the Reelin-induced Dab1 phosphorylation and SFK activation but also the activation of PI3K by Reelin, as determined with an antibody directed against activated Akt (Fig. 1A, lane 2). Because PP2 can also inhibit c-Abl and c-Kit, we used the inhibitor STI571 to block these kinases without affecting SFK activity (for review, see Ref. 31). This had no effect on PI3K activation by Reelin (Fig. 1A, lane 6). By contrast, the PI3K inhibitors LY294002 and wortmannin had no effect on the activation of SFKs by Reelin, as detected with an antibody directed against a tyrosine-phosphorylated epitope in the activation loop (␣-pY418), whereas Akt phosphorylation at serine 473 was completely blocked (Fig. 1B, lanes 4 and 8). Thus, SFK activation precedes PI3K activation by Reelin.
Potential Mechanisms Leading to Reelin-induced PI3K Activation-Ligand-induced PI3K activation often involves binding of the SH2 domains of the regulatory subunit p85␣ ( Fig. 2A) to phosphotyrosine motifs in growth factor receptor tails or adaptor proteins. Using web-based algorithms (scansite.mit.edu and www.cbs.dtu.dk/services/NetPhos) for detecting sequence motifs that mediate binding to protein interaction domains, we identified conserved putative SH2 domain binding motifs in Dab1 (Fig. 2B) that include Tyr-198 and Tyr-220. These amino acids are specifically phosphorylated in response to Reelin stimulation. The NetPhos algorithm (29) correctly predicts the phosphorylated and nonphosphorylated residues (6,32) in the cluster of five tyrosine residues located immediately C-terminal of the Dab1 PTB domain (Fig. 2B). Phosphorylation of the potential site involving Tyr-300 has so far not been detected experimentally (6). An alternative mode of interaction between Dab1 and p85␣ could involve an SH3 domain interaction with proline-rich sequences within Dab1. Such a predicted site sur-rounds Pro-424. Conceivably, tyrosine phosphorylation of Tyr-198 and Tyr-220 might induce a conformational change exposing this site.
p85␣ would be unlikely to interact directly with the cytoplasmic domains of apoER2 and VLDLR. Although NetPhos identifies the NPXY motif as a likely interaction site for PTB domains, the other tyrosine residues in the cytoplasmic tails of apoER2 and VLDLR are not predicted to be in the proper sequence context where they could mediate SH2 domain interactions with p85␣ (Fig. 2C). Consistent with this prediction, we were unable to find any indication of tyrosine phosphorylation of apoER2 in immunoprecipitates from Reelin-treated neuronal lysates (Fig. 3A). However, a small amount of a phosphoprotein that comigrates with Dab1 was co-precipitated with apoER2 from Reelin-treated neuronal lysates and likely corresponds to tyrosine-phosphorylated Dab1 (Fig. 3A, lane 6).
Reelin Does Not Cause Tyrosine Phosphorylation of the PI3K Subunits-SFK-dependent phosphorylation of the PI3K subunit p85␣ can activate PI3 kinase by relieving inhibition of the p110 catalytic subunit. We, therefore, tested if Reelin, which activates SFKs in neurons, induces tyrosine phosphorylation of p85␣. Neuronal lysates were immunoprecipitated with a p85␣ antibody, and the precipitates were immunoblotted with a phosphotyrosine-specific antibody (Fig. 3B). The tyrosine-phosphorylated 85-kDa protein that is detected in the Reelin-induced lysate (lane 2) corresponds to tyrosine-phosphorylated Dab1 (p-Dab1; see also Fig. 3C). Although p85␣ is no longer detectable in the supernatant after immunoprecipitation with an ␣-p85␣ antibody, the 85-kDa phosphoprotein was not depleted by the p85␣ immunoprecipitation (Fig. 3B, lane 4), and no significant tyrosine phosphorylation of immunoprecipitated p85␣ was detected (lanes 5 and 6). The weakly tyrosine-phosphorylated precipitated protein band at ϳ85 kDa (lane 6) likely is phosphorylated Dab1 (see next paragraph). The strongly tyrosine-phosphorylated band at ϳ 110 kDa (top panel) corresponds to the catalytic PI3K subunit, which co-precipitates constitutively with p85␣. Tyrosine phosphorylation levels of p110 were also not affected by Reelin treatment. These results suggest that modulation of tyrosine phosphorylation of PI3K itself is not a major regulatory mechanism for Reelin-induced PI3K activation in neurons.
Reelin Induces Interaction of Dab1 and p85␣ in Neurons-A common mechanism of ligand-induced PI3K activation involves recruitment of p85␣ to autophosphorylated receptor kinases or their substrates, resulting in disinhibition of the p85␣-bound catalytic subunit p110. Because SFK inhibition blocks the tyrosine phosphorylation of Dab1 and PI3K activation by Reelin (Fig. 1A), we hypothesized that Reelin might trigger the recruitment of p85␣ to p-Dab1. To test this, we immunoprecipitated Dab1 protein from lysates of control and Reelintreated primary neuronal cultures and tested for the presence of p85␣ in the precipitated immune complexes. As shown in Fig. 3C, total Dab1 protein and p-Dab1 were immunodepleted from the immunoprecipitation supernatants (lanes 3 and 4), and p85␣ indeed co-immunoprecipitated with Dab1 in a manner that correlated with the degree of tyrosine phosphorylation of Dab1 (lanes 6 and 7). No p85␣ was immunoprecipitated from a Reelin-treated sample with a control antiserum (lane 8) or from lysates in which Dab1 phosphorylation by Reelin had been blocked by the SFK inhibitor PP2 (lane 11).
These data suggest that Reelin induces the recruitment of p85␣ into a complex with Dab1 in a tyrosine phosphorylationdependent manner. The converse experiment, in which Dab1 was co-precipitated with an ␣-p85␣ antibody in a Reelin-dependent manner supports this conclusion (Fig. 3D). Dab1 was detected only in p85␣ immunoprecipitates from Reelin-treated neuronal lysates (lane 4) but not from control-treated lysates (lane 3), lysates immunoprecipitated with an irrelevant immunoglobulin (lane 5), or lysates from Reelin-treated cells in which SFKs, and thus, Dab1 phosphorylation had been inhibited with PP2 (lane 11). Neither Fyn (Fig. 3C, lane 6 and 7) nor Src (not shown) could be detected in Dab1 immunoprecipitates from control (lane 6) or Reelin-treated (lane 7) neuronal lysates. This suggests that the interaction between SFKs and p-Dab1 that has been observed in vitro and in transfected cell lines (33) and which involves the SFK-SH2 domain is either transient under these conditions or mutually exclusive and weaker than the p85␣-Dab1 interaction. We found the latter to be stable even in the presence of 0.5 M NaCl or 60 mM octylglucopyranoside, a detergent that disrupts cholesterol-rich microdomains of the plasma membrane (data not shown).

Activation of the Reelin Signaling Cascade Occurs at the Plasma Membrane-The
Dab1 PTB domain can mediate simultaneous interaction of the protein with the NPXY motif of apoE receptor cytoplasmic domains and with PIs at the inner leaflet of the plasma membrane (34). The affinity of the Dab1 PTB domain for PIs might be important for increasing the apparent affinity for the NPXY motifs in the Reelin receptor tails, which on their own have a relatively low affinity for the Dab1 PTB domain in the micromolar range. The PI interaction might also serve to direct Dab1 into PI-rich specialized membrane compartments such as caveolae and rafts.
To investigate whether other known components of the Reelin signaling pathway are recruited into this membrane-associated complex, we separated postnuclear supernatants of control (Fig. 4, lanes 1 and 3) and Reelin-treated neurons (lanes 2 and 4) into membrane (lanes 1 and 2) and cytosolic (lanes 3 and  4) fractions. The distribution of several components of the Reelin signaling cascade between these fractions was then examined by immunoblotting. Dab1, its major physiological kinase Fyn, and the Reelin receptor apoER2 were found predominantly in the membrane fraction. By contrast, Akt was restricted almost completely to the cytosolic fraction, whereas p85␣ and Cdk5, a neuronal serine/threonine kinase that can also phosphorylate Dab1, were present in both fractions. The phosphorylation state of the proteins (p-Dab1, SFK-(Tyr(P)-418) and Ser(P)-473-Akt) did not affect their subcellular distribution. In particular, no significant redistribution of p85␣ to the membrane compartment in response to Reelin treatment was observed. It cannot be ruled out, however, that changes in the subcellular localization of a small but nevertheless important subfraction of a protein might escape detection at this level of sensitivity. Indeed, only ϳ 5% of the total p85␣ pool associate with the platelet-derived growth factor receptor upon ligand stimulation (35).
Cholesterol Depletion Affects Reelin Signaling-The relative enrichment of phosphatidylinositol 4,5-bisphosphate in raftlike domains, phospholipid-and cholesterol-rich plasma membrane patches, prompted us to examine the effects of raft disintegration on Reelin signaling in neurons. Phosphatidylinositol 4,5-bisphosphate serves not only as a PI3K substrate and precursor of phosphatidylinositol 3,4,5-trisphosphate, but it also mediates the targeting of proteins containing PI binding domains to the plasma membrane, where it participates in the regulation of endocytosis and cytoskeletal assembly. Depletion of membrane cholesterol with M␤CD was used to disrupt cholesterol-and PI-enriched membrane rafts in cultured neurons. The activation of downstream signaling components by Reelin was significantly reduced in cholesterol-depleted neurons (Fig. 5A, lanes 9 and 10), whereas total protein levels were not significantly affected by the treatment. The reduced intensity of the bands at ϳ130 and 180 kDa in the phosphotyrosine immunoblot, which are not regulated by Reelin, indicates a general reduction of kinase-mediated signaling in cholesterol-depleted neurons, consistent with an important role of raft-like membrane structures in neuronal signaling.  6, 7, 9, 10). The p85␣-subunit co-precipitates with Dab1 in a Reelin-dependent manner, correlating with the increase in p-Dab1 levels (lanes 7 and 10). No Dab1 or p85␣ protein is precipitated by the control serum (lane 8). Treatment with the SFK inhibitor PP2 prevents Dab1 tyrosine phosphorylation and co-precipitation of p85␣ (lane 11). D, Reelin-dependent co-immunoprecipitation of Dab1 with p85␣ from primary embryonic neurons. p85␣ is immunoprecipitated from lysates of control and Reelin-treated rat forebrain neurons. cortical plate in embryonic mouse brains. Wild type embryonic mouse brain slices, which express Reelin endogenously, were incubated in the absence or presence of the PI3K inhibitor LY294002. After 2 days in which the slices were allowed to proceed through development in vitro, the slices were fixed and stained for histological examination (Fig. 6A). At E13.5, when the embryonic brains had been harvested, the preplate had split, but the cortical plate was only just beginning to form (top panel). After 2 days in culture, a normal, well formed cortical plate had developed (middle panel). Formation of an ordered cortical plate was completely prevented by treatment with the PI3K inhibitor (bottom panel), although cell proliferation was not significantly affected, as judged by the comparable number of cells in the treated and untreated slices.
To demonstrate that we had successfully inhibited PI3K activity in the slices, we prepared protein lysates of identically treated slices and immunoblotted them with an antibody against total Akt and with the anti-p-Akt antibody (Fig. 6B). The effect of wortmannin, another PI3K inhibitor, could not be evaluated because this substance was not functionally stable over the 2 day incubation period. These data suggest that Reelin-dependent PI3K activation may be functionally important in the developing neocortex. DISCUSSION The large signaling molecule Reelin regulates the lamination of the neocortex and the cerebellum during embryonic development (for reviews, see Refs. 1-4). It is involved in the specifi-cation of the radial glia (36,37) and also modulates synaptic plasticity in the mature brain (38,39). Transmission of the signal to the neurons requires the Reelin receptors VLDLR and apoER2 and the cytosolic PTB domain adaptor protein Dab1, which interacts with NPXY motifs in the intracellular domains of the receptors (26,27,34). Translation of the signal into a cellular response requires tyrosine phosphorylation of Dab1 and activation of SFKs (7,8). PI3K signaling is also activated in response to Reelin, resulting in activation of Akt, inhibition of glycogen synthase kinase 3␤, and modulation of the phospho-  1 and 3) and Reelin-treated (lanes 2 and 4) cultured neurons. Postnuclear supernatants were prepared as described under "Experimental Procedures" and separated into membrane (lanes 1 and 2) and cytosolic soluble (lanes 3 and 4) fractions by ultracentrifugation. The fractions were separated by SDS-gel electrophoresis and analyzed by immunoblotting (20 g of protein/lane). Phosphorylated and non-phosphorylated Dab1 and Fyn, the main Dab1 kinase, were found predominantly in the membrane fraction whether the cells had been treated with Reelin or not. By contrast, the PI3K substrate Akt and its activated form, p-Akt (Ser-473), were mainly found in the soluble fraction. p85␣ and the serine/ threonine kinase Cdk5 were present in both fractions. ApoER2 serves as a marker protein for the membrane compartment.  1 and 2) before control (lanes 1, [3][4][7][8] or Reelin treatment (lanes 2, 5-6, 9 -10). Cholesterol depletion with M␤CD potently decreased Reelin-induced tyrosine phosphorylation of Dab1 and subsequent activation of SFKs and PI3K (lanes 9 and 10). Note that the total protein levels (Dab1, Cdk5, and the unspecific band seen with the SFK-(Tyr(P)-418 (␣-p-Y418)) antibody, marked with an asterisk) remain virtually unaltered in the cholesterol-depleted neurons. rylation state of the microtubule-stabilizing protein tau (9). It was unclear, however, how PI3K is activated and whether this precedes or is a consequence of prior SFK activation and whether PI3K signaling is necessary for Reelin-controlled neuronal migration. In this study we have now shown that SFKs are activated independent of PI3K and that SFK function is required for subsequent PI3K activation. The latter involves the formation of a complex of the regulatory p85␣ subunit of PI3K with tyrosine-phosphorylated Dab1. Inhibition of PI3K activity prevents the formation of a normal cortical plate in an in vitro slice migration assay. These results establish a contiguous biochemical pathway that begins at the plasma membrane and continues to regulate the dynamic architecture of the microtubular network.
Cultured primary embryonic forebrain neurons are a well characterized model system in which most of the biochemical mechanisms that are activated by Reelin binding at the neuronal plasma membrane have been worked out. Analysis of neuronal lysates and coimmunoprecipitation experiments now show that the regulatory p85␣ subunit of PI3K complexes with tyrosine-phosphorylated Dab1 to activate the enzyme (Fig. 3). The same principal mechanism is used by the IRS proteins in the insulin signaling pathway in primary embryonic neurons, where insulin-dependent tyrosine phosphorylation of IRS triggers its interaction with the SH2 domains of p85␣, which allows the associated p110 subunit to become active. Like Dab1, IRS contains an N-terminal PTB domain, which mediates its binding to an NPXY motif in the insulin receptor tail when it is tyrosine-phosphorylated in response to insulin. A PI binding pleckstrin homology domain contributes to the membrane association of IRS (40). In contrast to the IRS proteins, the Dab1 PTB domain strongly prefers unphosphorylated NPXY motifs (34).
Thus, in a possible scenario, phosphorylation of the VLDLR or apoER2 NPXY motif might function as a switch, turning off Dab1-dependent signaling and potentially activating another phosphotyrosine-dependent pathway. However, we found no evidence for tyrosine phosphorylation of the apoER2 intracellular domain in the absence or presence of Reelin, making such a mechanism unlikely. Instead, Dab1 phosphorylation, which appears to be amplified from a low basal level by receptor dimerization (41), SH2 domain interaction (33), and amplification of SFK activity (7,8), serves to recruit PI3K into a complex where it becomes activated. This model is supported by several lines of evidence. First, tyrosine phosphorylation of Dab1 is necessary for Reelin-triggered PI3K activation, and SFKs are required for this to occur (Fig. 1). Second, the Reelin-dependent interaction of p85␣ and Dab1 correlates with the levels of Dab1 tyrosine phosphorylation, and the SFK inhibitor PP2 prevents co-immunoprecipitation of p85␣ and Dab1 in response to Reelin treatment (Fig. 3). The phosphorylation level of neither PI3K subunit was altered by Reelin stimulation, indicating that direct tyrosine phosphorylation of this enzyme is not involved. Furthermore, Fyn and Src, the main SFKs that are involved in Reelin signaling, were not detected in p85␣ immunoprecipitates from Reelin-stimulated neurons (Fig. 3), which argues against a mechanism involving a scaffold that organizes SFKs and PI3K in an activation complex (42).
IRS-1 activation of PI3K can be suppressed by serine phosphorylation (for review, see Ref. 13). Intriguingly, Dab1 is also serine-phosphorylated by Cdk5 (43), although this is not modulated by Reelin. Cdk5, together with glycogen synthase kinase 3␤, mediates phosphorylation of tau and also regulates microtubule function by other mechanisms. This raises the possibility of a regulatory mechanism where Cdk5 might control glycogen synthase kinase 3␤ activation by Reelin.
How could Dab1 activate PI3K directly? One possible mechanism might involve the direct interaction of the p85␣ SH2 domain with phosphotyrosines in Dab1. The five clustered tyrosine residues that follow the PTB domain and that are phosphorylated in vitro and in vivo (6,32) are obvious candidates for this interaction. A predicted YVAM consensus binding motif for p85␣ at tyrosine 300 of Dab1 is another possible candidate site. However, phosphorylation of this residue has not been observed in vitro (32). Furthermore, no residual tyrosine phosphorylation was detectable in a neuronal Dab1, in which five potential Reelin-dependent tyrosine phosphorylation sites had been mutated to non-phosphorylatable phenylalanines (6).
However, this does not completely rule out a Reelin-induced phosphorylation of this site in wild type neurons, which could depend either on a conformational change or on another unidentified tyrosine kinase that requires phosphorylation of the mutated tyrosines for activation. Alternatively, a conformational change induced by phosphorylation of Tyr-198 and Tyr-220 might result in exposure of the proline-rich potential Srchomology 3 domain binding site around residue 424 (Fig. 2B). A functional role for the C-terminal region of Dab1 in neuronal migration is supported by findings in mice that express only one functional hypomorphic allele of the naturally occurring truncated p45 splice from of Dab1 (44).
The structural basis for simultaneous binding of Dab1 to NPXY motifs and to PIs in the plasma membrane that are both mediated by the PTB domain (34) has recently been reported (45). The PI interaction likely serves to enrich Dab1 in the correct membrane compartments, whereas the NPXY motif interaction confers specificity. This model is supported by our cell fractionation experiment (Fig. 4), which shows that Dab1 resides almost exclusively in the membrane fraction and that phosphorylation of Dab1 and SFKs occurs in the membrane compartment. Interestingly, cytosolic tyrosine phosphorylation of IRS-1 is not sufficient for normal Akt activation in response to insulin but requires plasma membrane association of the adaptor protein (46).
Cholesterol is another important component of the plasma membrane compartment where Reelin signaling takes place. Caveolae and rafts are cholesterol-rich membrane compartments in which many signaling proteins are concentrated, including apoER2 (24). M␤CD treatment, which depletes cellular membranes of cholesterol, indeed greatly reduced Dab1 phosphorylation, SFK activation, and Akt activation (Fig. 5), suggesting that Reelin signaling in embryonic neurons takes place in these cholesterol-rich microdomains.
It is difficult to gauge the extent of the role of PI3K for Reelin-mediated regulation of neuronal migration, since this enzyme participates in numerous cellular processes. Gene targeting of the different class I PI3K subunit isoforms in mice (for review, see Ref. 47) is not informative in this respect. Likewise, disruption of the catalytic subunits p110␣ and -␤ is embryonically lethal before the onset of neuronal migration. However, inhibition of PI3K activity with the inhibitor LY294002 in a cortical slice migration assay prevented the formation of a normal cortical plate (Fig. 6), suggesting that PI3K is necessary to mediate this Reelin-dependent step and plays a role in regulating cortical lamination during the development of the mammalian brain.