Research Article

Endothelial Dab1 signaling orchestrates neuro-glia-vessel communication in the central nervous system

See allHide authors and affiliations

Science  24 Aug 2018:
Vol. 361, Issue 6404, eaao2861
DOI: 10.1126/science.aao2861

Developing the bloodbrain barrier

During development, signals need to be dynamically integrated by endothelial cells, neurons, and glia to achieve functional neuro-glia-vascular units in the central nervous system. During cortical development, neuronal Dab1 and ApoER2 receptors respond to a guidance cue called reelin. Studying mice, Segarra et al. found that Dab1 and ApoER2 are also expressed in endothelial cells (see the Perspective by Thomas). The integration of reelin signaling in endothelial cells and neurons facilitates the communication between vessels, glia, and neurons that is necessary for the correct positioning of neurons during cortical development. This integration is also important for correct communication at the neurovascular unit required for blood-brain barrier integrity in the mature brain.

Science, this issue p. eaao2861; see also p. 754

Structured Abstract


The function of the brain relies on communication among the complex network of cells that constitute this organ. Vascularization of the central nervous system (CNS) ensures adequate delivery of oxygen and nutrients to build up and maintain homeostasis of neuronal networks. Thus, it is not surprising that blood vessels and neuronal cells share multiple parallelisms orchestrating their development in synchrony and in a mutually dependent manner in the CNS. Despite the essential role of the endothelium in brain function, the means by which signaling at the interface of endothelial cells, glial cells, and neurons is integrated temporally and spatially for proper brain development has remained largely unexplored.


Integration of signaling pathways and cellular responses among endothelial cells, glial cells, and neurons is needed to ensure proper architecture of the brain. Reelin (Reln), a large secreted glycoprotein, induces Disabled 1 (Dab1)–dependent responses in neurons to guide their migration in all layered brain structures. Secretion of reelin by Cajal-Retzius cells in the marginal zone of the cortex timely coincides with active sprouting of pial vessels ingrowing perpendicularly into the marginal zone and forming a complex vascular network needed to support brain development and function. Therefore, reelin might be in the perfect position to perform a bivalent function to timely and spatially orchestrate both neuronal migration and CNS vascularization. We reasoned that blood vessels might instruct the process of neuronal migration by a cell-autonomous function of Dab1 on endothelial cells. To investigate this, we deleted the expression of vascular Dab1 in mice and investigated the effects on CNS vascularization, neuroglial organization, and neurovascular unit function.


We found that reelin/Dab1 signaling is conserved in endothelial cells and exerts potent proangiogenic effects in the developing vasculature of the CNS by controlling endothelial cell proliferation and active filopodia extension of the vascular network. The interaction of the reelin receptor ApoER2 (apolipoprotein E receptor 2) and VEGFR2 (vascular endothelial growth factor receptor 2) mediated the proangiogenic roles of Dab1 in endothelial cells. Surprisingly, deletion of Dab1 exclusively in the vascular system induced changes in the position of postmitotic pyramidal neurons in the cortical layers of the cerebral cortex. At the cellular level, depletion of vascular Dab1 reduced the docking of the radial glia processes to the pial surface at embryonic and postnatal stages and altered the differentiation of glial cells to astrocytes. The defects in neuronal migration persisted in adult mutant animals, where stereotypical attachment of the astrocytes to penetrating vessels in the glia limitans superficialis was also found to be aberrant. The functionality of the neurovascular unit [blood-brain barrier (BBB) integrity] was also affected in reelin knockout animals, and we could attribute those defects to the lack of Dab1 signaling exclusively in endothelial cells. The increased BBB permeability was again associated with an insufficient coverage of the brain vasculature by astrocytic endfeet. Mechanistically, we determined that the astroglial attachment to the vasculature is mediated by reelin-induced deposition of laminin-α4 by endothelial cells to the extracellular compartment, which in turn enables the binding of the glial processes to the CNS vasculature via the activation of integrin-β1 in glial cells.


Our results shed new light on the function of the vasculature in CNS development and homeostasis—in particular, how signals from the endothelium orchestrate the communication among vessels, glial cells, and neurons, and how specific changes in the molecular signature of the endothelium affect a plethora of processes such as CNS vascularization, extracellular matrix composition, neuroglial cytoarchitecture, and BBB development.

Instructive functions of vascular Dab1 in the neurovascular interface.

(A) ApoER2 and Dab1 control endothelial cell proliferation and tip cell filopodia extension during CNS vascularization by cross-talking to the VEGFR2 pathway. (B and C) Vascular Dab1 also instructs radial glia organization and neuronal migration in the developing cerebral cortex (B) and the development of the blood-brain barrier (C). In both cases, Dab1 signaling in the vasculature regulates the deposition of laminin-α4 and, in turn, the activation of integrin-β1 in glial cells.


The architecture of the neurovascular unit (NVU) is controlled by the communication of neurons, glia, and vascular cells. We found that the neuronal guidance cue reelin possesses proangiogenic activities that ensure the communication of endothelial cells (ECs) with the glia to control neuronal migration and the establishment of the blood-brain barrier in the mouse brain. Apolipoprotein E receptor 2 (ApoER2) and Disabled1 (Dab1) expressed in ECs are required for vascularization of the retina and the cerebral cortex. Deletion of Dab1 in ECs leads to a reduced secretion of laminin-α4 and decreased activation of integrin-β1 in glial cells, which in turn control neuronal migration and barrier properties of the NVU. Thus, reelin signaling in the endothelium is an instructive and integrative cue essential for neuro-glia-vascular communication.

Vascularization of the central nervous system (CNS) ensures adequate delivery of oxygen and nutrients to build up and maintain homeostasis of neuronal networks (1). Apart from these metabolic functions, vessels have also been suggested to serve as niches and scaffolds for neuronal migration both during development and during adult neurogenesis (2). The orchestration of a perfect architecture of the neurovascular unit (NVU) is fundamental for brain function. Previous studies have examined the relation of the vasculature to the neuroglial components. It has been shown that the radial glia scaffold is necessary for angiogenesis because ablation of radial glia cells induces regression of cortical vessels (3), and that the expression of integrin-β8 by glial cells regulates developmental vascularization of the brain (4). Additionally, besides the well-established support of neuroblast migration by the radial glia during corticogenesis (5), it has been reported that the vasculature also contributes to neuronal navigation and positioning; for example, periventricular endothelial cells (ECs), which release γ-aminobutyric acid (GABA), guide the tangential migration of GABAergic neurons during embryogenesis (6, 7). However, despite these emerging studies, the means by which signaling at the interface of neurons, ECs, and glial cells is integrated for proper brain development remains largely unexplored.

Reelin, a well-known guidance cue for migrating neurons, regulates lamination of brain regions during development and synaptic plasticity in the adult brain (8). Binding of reelin to the apolipoprotein E receptor 2 (ApoER2) and to the very low density lipoprotein receptor (VLDLR) triggers a signaling cascade required for neuronal migration that involves the adaptor protein Disabled1 (Dab1) (913). Reelin knockout or Dab1 knockout mice exhibit aberrant cortical, hippocampal, and cerebellar architecture (14, 15). Notably, reelin signaling also has an impact on the blood and lymphatic vasculature (1618). In humans, reelin mutations cause severe developmental defects (19) and are associated with several neurological diseases (2022). While neurons are migrating and colonizing the layers in the mouse cortex guided by the expression of reelin in the marginal zone (8), pial vessels start to sprout perpendicular to the cortical marginal zone and grow into the cortex to form the complex vascular network needed to support brain development and function (23).

Reelin induces angiogenic responses

Our ex vivo culture technique allows quantification of acute EC responses and the guidance of tip cells during angiogenic sprouting in the mouse retina (24). Stimulation of explanted retinas with exogenous reelin (Reln) resulted in an increase of filopodia extensions per vessel length (Fig. 1, A and B). In vitro, stimulation of ECs with exogenous reelin promoted EC tube formation (fig. S1, A to C) and Dab1 phosphorylation (Fig. 1, C to E, and fig. S1, D and E). In the mouse retina vascularization model (25), Reln–/– and ApoER2–/– mice both showed defective extension of the superficial vascular network at postnatal day 2 (P2) (Fig. 1, F and G, and fig. S1, F and G) and a strong decrease in filopodia extensions at the vascular front at P7 (Fig. 1, H and I, and fig. S1, H and I). Reelin’s proangiogenic effects are mediated by its receptor ApoER2, because retinas explanted from ApoER2–/– mice failed to increase filopodia extensions after reelin stimulation (Fig. 1, J and K).

Fig. 1 Reelin pathway is proangiogenic in endothelial cells.

(A) Isolectin B4 (IB4) whole-mount staining of wild-type retinal explants after 4 hours of reelin (Reln) stimulation. (B) Quantification of the number of filopodia (green dots) at the vascular front (areas outlined in red) depicted in (A) (n = 9 to 15 images, 3 explants per condition). (C) Immunofluorescence of phospho-Dab1 (pDab1) in human umbilical vein endothelial cells (HUVECs) after 30 min of Reln stimulation. (D) Quantification of (C) (n = 56 to 66 images, 6 experiments). (E) Western blot analysis of Dab1 phosphorylation in HUVECs after Reln stimulation. Actin was used as a loading control. (F) IB4 whole-mount staining of wild-type (WT) and ApoER2–/– mutant vascular networks at P2. (G) Quantification of the vessel radial length in (F) (n = 20 or 21 retinas, 10 or 11 animals per genotype). (H) ApoER2–/– retinas at P7. (I) Quantification of filopodia extensions in (H) (n = 5 retinas, 3 to 5 animals per genotype). (J) Reln stimulation of retinal explant cultures for 4 hours increases filopodia sprouting at the vascular front in WT explants but not in ApoER2–/– explants. (K) Quantification of the number of filopodia related to (J) (n = 10 to 18 images, 3 explants per condition). Data were normalized to corresponding controls. Scale bars, 30 μm [(A), (H), and (J)], 10 μm (C), 200 μm (F). Data are means ± SEM. *P < 0.05, ***P < 0.001; ns, not significant.

Early postnatal retinal vascularization and endothelial tip cell formation are regulated by astrocytes within the ganglion cell layer, which produce the vascular chemotactic factor VEGF (vascular endothelial growth factor) (26, 27). Moreover, in cortical neurons in culture, VEGF and reelin pathways cross-talk to regulate the phosphorylation of the NR2 subunit of the N-methyl-d-aspartate (NMDA) receptor (28). We found that ApoER2 coimmunoprecipitated and coclustered with VEGF receptor 2 (VEGFR2) in ECs, and that such interaction was induced by VEGF-A and by reelin (Fig. 2, A to C, and fig. S2, A and B). VEGF-A activated VEGFR2 and induced Dab1 phosphorylation in ECs (Fig. 2, D to F). VEGF-C, another ligand for VEGFR2 (29), also induced Dab1 phosphorylation (fig. S2C), supporting the idea that VEGFR2 activation occurs upstream of Dab1 signaling. Reciprocally, reelin also activated VEGFR2 (fig. S2D). Costimulation with VEGF and reelin synergistically increased Dab1 phosphorylation (fig. S2, E and F) and filopodia extension in the ex vivo retina assay (Fig. 2G). Moreover, ApoER2–/– vessels were unable to respond to VEGF-mediated tip cell sprouting (Fig. 2, H and I), indicating the presence of functional cross-talk between the two receptors.

Fig. 2 Reelin pathway cross-talks with VEGF signaling in endothelial cells.

(A) Immunoprecipitation (IP) of VEGFR2 from HUVEC lysates and Western blot for VEGFR2 and ApoER2. TL, total lysates. (B) ApoER2/VEGFR2 clusters (white punctae) upon 10 min of stimulation with VEGF-A in HUVECs. PLA, proximity ligation assay. (C) Quantification of ApoER2/VEGFR2 clusters in (B) (n = 108 to 110 cells, 3 experiments). (D) Western blot of phosphorylation of Dab1 (pDab1) and VEGFR2 (pVEGFR2) after VEGF-A stimulation of HUVECs. (E) Immunofluorescence staining with a phospho-specific Dab1 antibody in HUVECs showing increased pDab1 after VEGF-A stimulation. (F) Quantification of pDab1 in (E) (n = 26 to 30 images, 3 experiments). (G) WT retinal explants exposed to a combination of Reln and VEGF-A for 4 hours show increased number of filopodia when compared to single stimulation conditions and to the control (n = 13 to 21 images, 4 or 5 explants per condition). (H) VEGF-A stimulation of WT and ApoER2–/– retinal explants. (I) Quantification of (H) (n = 9 or 10 images, 3 explants per condition). Stimulation is normalized to each control condition. Scale bars, 30 μm [(B) and (H)], 10 μm (E). Data are means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

Dab1 regulates vascular morphogenesis cell-autonomously

The retinal ganglion cell layer, underneath the astrocytes and the vessel bed, appeared as the major source of reelin (fig. S3, A and B). Both ApoER2 and Dab1 are expressed in retinal vessels (fig. S3, C and D). To uncouple the function of Dab1 in the vasculature from its function in neuronal cells, we generated inducible endothelial-specific Dab1 loss-of-function mice (Dab1iΔEC) by crossing Dab1flox/flox mice to the tamoxifen-inducible vascular-specific Cre deleter line Cdh5(PAC)creERT2. Tamoxifen injection for three consecutive days efficiently induced the vascular-specific genetic recombination and loss of Dab1 expression in ECs (fig. S4, A to C). Dab1 deletion from P1 to P3 impaired vascular radial growth in Dab1iΔEC retinas at P3 to P4 (Fig. 3, A and B). This reduction in vessel growth was persistent at P7 (fig. S4, D and E) and recovered at adult stages (fig. S4, F and G). Moreover, EC proliferation and vascular network complexity were also decreased in the Dab1iΔEC postnatal retinas (Fig. 3, C and D, and fig. S4, H and I). A reduced number of filopodia per vessel length and a reduced tip/stalk cell ratio were also observed in Dab1iΔEC mice (Fig. 3, E to H). Expression of VEGF or reelin was not changed in Dab1iΔEC or ApoER2 mutants (fig. S4J).

Fig. 3 Endothelial Dab1 is essential for retina and cortex vascularization.

(A) IB4 staining of Dab1iΔEC retinas at P3–P4 after tamoxifen (TMX) administration from P1 to P3. (B) Quantification of vessel radial length shown in (A) (n = 6 to 12 retinas, 3 to 7 animals per genotype). (C) Vascular proliferation (phospho–histone H3, pHH3) in Dab1iΔEC animals at P3–P4 after TMX administration from P1 to P3. (D) Quantification of the number of pHH3+ ECs in (C) (n = 9 to 15 retinas, 5 to 9 animals per genotype). (E) Filopodia extensions at the vascular front of Dab1iΔEC retinas at P7. (F) Quantification of (E) (n = 5 retinas, 3 animals per genotype). (G) ETS-related gene (ERG) and IB4 staining in Dab1iΔEC retinas at P7 after TMX administration from P1 to P3. (H) Quantification of the relative number of tip cells versus stalk cells at the vascular front shown in (G) (n = 6 to 10 retinas, 5 or 6 animals per genotype). (I) Development of the vasculature of control and Reln–/– embryos stained with IB4. Arrows point to defects in ingrowing sprouts from the pial vessel and branching defects in the neocortex below the marginal zone. Dashed line delineates the marginal zone. (J) Quantification of vessel density in E17.5 cortices in (I) (n = 3 animals per genotype). (K and M) Dab1iΔEC cortices stained with IB4 at E17.5 (K) or P7–P8 (M) after TMX administration at the indicated time points. The vascularization of the upper cortex is reduced in the mutant animals, with decreased area covered by vessels and fewer vascular intersections. (L) Quantification of vessel density in (K) (n = 5 or 6 animals per genotype). (N to Q) Quantification of vessel density (N) (n = 8 or 9 animals per genotype), vessel length (O) (n = 5 animals per genotype), branch points (P) (n = 5 animals per genotype), and vessel orientation in relation to the pial surface (Q) (n = 5 or 6 animals per genotype) shown in (M). Scale bars, 200 μm (A), 100 μm [(C), (I), (K), (M)], 20 μm (E), 75 μm (G). Data are means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

In the cerebral cortex, reelin is secreted by Cajal-Retzius cells during developmental stages (30) and regulates the migration of neurons during their correct positioning in the different cortical layers (8, 31). Cortical vessels expressed both ApoER2 and Dab1, and brain ECs also responded to reelin by phosphorylating Dab1 (fig. S5, A to C). The embryonic brain is vascularized by the pial and periventricular vessels (32), with sprouts from the pial vessels penetrating into the neural tube from the perineural vascular plexus around embryonic day 10.5 (E10.5) by sprouting angiogenesis and forming columnar vascular structures perpendicular to the meninges (23, 33). In the absence of reelin, vascularization of the embryonic neocortex was aberrant (Fig. 3, I and J). From E11.5 on, we found fewer sprouts penetrating from the pial vessels into the neocortex and connecting to the periventricular vessels (Fig. 3I, arrows). In control animals, penetrating vessels from the pia at E14.5 started branching directly below the marginal zone, but this pattern was lost in the Reln–/– mutants (Fig. 3I, dashed line). Vessels were also less branched at E13.5 and E14.5, which later resulted in a reduction of vessel density as quantified at E17.5 (Fig. 3J). The same aberrant phenotype was observed when deleting Dab1 exclusively from the vessels at E10.5 to E12.5 and analyzing the cortical vasculature at E17.5 (Fig. 3, K and L); this finding supported the cell-autonomous function of Dab1 observed in the retina. Dab1 deletion at early postnatal stages from P1 to P3 and analysis at P7 to P8 also unraveled a defective vessel architecture (Fig. 3M) reflected by a reduction in vessel density (Fig. 3N), vessel length (Fig. 3O), and number of branch points (Fig. 3P) as well as a preferential orientation at an 80° to 90° angle with respect to the pial surface (Fig. 3Q). These morphological defects did not impair vessel perfusion (fig. S5, D and E). As in the retina, the morphological vascular defects in the cortex were compensated at adult stages (fig. S5, F to J).

Dab1 in the endothelium is required for proper migration of neurons during cortical layering

The absence of reelin expression by Cajal-Retzius cells in the marginal zone (MZ) as well as the lack of Dab1 signaling in neurons leads to many defects in cortical lamination (8, 15, 34, 35), such as projection neurons invading the MZ or layer I. We hypothesized that the cell-autonomous function of reelin signaling in the vasculature could also contribute to the proper migration of neurons during cortical lamination. Examination of the general cortical cytoarchitecture of the early embryonic deleted (E10.5 to E12.5) Dab1iΔEC mutants analyzed at E17.5 revealed a poorly defined separation of cortical layers as well as a remarkable invasion of cells in the marginal zone (Fig. 4A) recapitulating some of the defects observed in Reln–/– mice (35). Markers for deeper and upper layers—Tbr1+ and Cux1+, respectively—showed aberrant positioning of early- and late-born neurons at E17.5 (Fig. 4, B to E). Positioning and numbers of Pax6+ apical progenitors and Tbr2+ intermediate (basal) neuronal progenitors were not affected by embryonic Dab1 vascular deletion (fig. S6, A to D), indicating that the defects in neuronal positioning in the Dab1iΔEC mutants are not a consequence of a deficient neurogenic pool. We then focused on analyzing the neuronal migration defects in more detail. At perinatal days, later-born neurons, which are destined to upper cortical layers, are still navigating along the radial glia processes before translocating into layers II/III (36). Neuronal distribution analysis in the Dab1iΔEC mice at P7–P8 revealed invasion of postmitotic neurons in layer I after perinatal Dab1 deletion (Fig. 4, F and G). Immunostaining for Cux1 confirmed that later-born neurons entered into layer I (Fig. 4, H and I). Additionally, we found an increased number of Cux1+ cells below layer IV (Fig. 4, H and J). Birth-dating neurons with bromodeoxyuridine (BrdU) injected at E15.5 and analyzed at P8 after perinatal deletion of Dab1 confirmed the presence of later-born neurons in the lower cortical layers as well as a delayed migration of E15-born neurons into the upper layers (fig. S6, E to G). General defects in brain size were not observed in the Dab1iΔEC mice (fig. S7, A to F).

Fig. 4 Vascular reelin signaling mediates neuronal positioning in the neocortex.

(A to C) DAPI (A), Tbr1 (B), and Cux1 (C) staining of Dab1iΔEC cortices at E17.5 after TMX administration from E10.5 to E12.5. (D) Quantification of the number of Tbr1+ cells in upper layers in (B) (n = 10 to 13 images, 2 or 3 animals per genotype). (E) Quantification of Cux1+ cells in the marginal zone in (C) (n = 30 to 37 images, 5 animals per genotype). (F) NeuN staining of Dab1iΔEC cortices at P7–P8 after TMX administration from P1 to P3. (G) Quantification of NeuN+ cells in layer I in (F) (n = 29 to 34 images, 5 or 6 animals per genotype). (H) Cux1 staining of Dab1iΔEC cortices at P7–P8 after TMX administration from P1 to P3. (I and J) Quantification of Cux1+ neurons in layer I (n = 19 images, 4 animals per genotype) (I) and below layer IV (n = 40 to 57 images, 7 to 10 animals per genotype) (J). Scale bars, 100 μm [(A) and (F)], 50 μm [(B), (C), and (H)]. MZ, marginal zone; CP, cortical plate; VI, layer VI; SP, subplate; L, layer. Data are means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

Endothelial Dab1 is required for vessel-glia communication

Apical radial glial cells (RGCs) extend a single basal process (radial fiber) that reaches the meningeal basement membrane (BM) and anchors via bulb endfeet structures making frequent contact with the pial vessels (5, 37). Such endfeet-BM interactions are thought to contribute to the final placement of migrating neurons in the cortex (3840). Analysis of E17.5 Dab1iΔEC mice (tamoxifen administration E10.5 to E12.5) and E17.5 Reln–/– mice revealed a reduction of docking endfeet of RGCs to the BM (Fig. 5, A and B, and fig. S8, A and B). The same results were observed after perinatal removal of vascular Dab1 (Fig. 5, C and D). Deletion of VEGFR2 in the vessels reduced vascularization but did not recapitulate the same defects in glia anchoring (fig. S8, C to F), which suggests that the effects of Dab1 in glia-EC communication are independent of VEGF/VEGFR2 signaling and distinct from the effects of reelin/Dab1 signaling on vessel growth (Figs. 1 to 3).

Fig. 5 Vascular reelin signaling mediates radial glial attachment to the pial surface.

(A) Nestin staining of Dab1iΔEC cortices at E17.5 after TMX administration from E10.5 to E12.5. White dots indicate radial glial cell (RGC) contacts to the pia. (B) Quantification of RGC contact points in (A) determined by counting the RGC endfoot docking sites along the pial surface, as detailed in the high-magnification inset in (A) (n = 4 animals per genotype). (C) Nestin staining of Dab1iΔEC cortices at P7–P8 after TMX administration from P1 to P3. (D) Quantification of RGC contact points in (C) (n = 3 or 4 animals per genotype). (E) Representative confocal images of Nestin, DAPI, GFAP (astrocytes), and IB4 staining in P4 cortices. RGCs show an aberrant morphology and a premature astrocytic differentiation in Dab1iΔEC cortices relative to control littermates. (F) NeuN staining of Dab1iΔEC cortices 4 to 7 weeks after TMX administration from P1 to P3. (G) Quantification of NeuN+ cells in layer I in (F) (n = 12 or 13 images, 3 animals per genotype). (H) Podocalyxin (Pdx), GFAP, and NeuN staining of the interhemispheric fissure in 4- to 7-week-old Dab1iΔEC mice after TMX administration from P1 to P3. Scale bars, 50 μm [(A), (C), (E), zoom (F), and (H)], 100 μm (F), 25 μm [zoom (H)]. Data are means ± SEM. **P < 0.01, ***P < 0.001.

Later in postnatal stages, RGCs differentiate into mature astrocytes after retracting their contacts from the pial and ventricular surfaces (41). We also observed a distortion of the columnar morphology of RGCs in vascular Dab1 mutants at P4 (Fig. 5E, left panels). This was accompanied by a premature morphological change of the bipolar radial glial cells toward stellate-shaped cells characteristic of mature astrocytes (GFAP+) (Fig. 5E, right panels). Moreover, analysis of Dab1iΔEC mice at 4 to 7 weeks of age after perinatal deletion revealed that the invasion of neurons in layer I persisted in adult stages (Fig. 5, F and G). The astrocytic organization at the glia limitans superficialis, with astrocytes closely sealing the meningeal surface, is characterized by GFAP+ fibrillary processes radially distributed along the interhemispheric vasculature and the penetrating vessels. This organization was disturbed in the Dab1iΔEC mice denoted by GFAP+ astrocytic processes detaching from the pial surface and its penetrating vessels (Fig. 5H).

Vascular Dab1 is necessary to form a functional blood-brain barrier

In the CNS, blood vessels are central components of the NVU that regulate homeostatic functions of the brain: the blood-brain barrier (BBB) and the neurovascular coupling (42, 43). Extravasation of the BBB-impermeable fluorescent tracer Alexa Fluor 555 cadaverine was observed in postnatal and adult mice after perinatal deletion of vascular Dab1 (Fig. 6, A to F) and in the Reln–/– mutants (Fig. 6, G to I), indicating compromised BBB integrity. Electron microscopy analysis revealed anatomically defective tight junctions as well as a functionally augmented rate of transcytosis (Fig. 6, J to L). Reln–/– vessels exposed unsealed tight junctions with an altered alignment in relation to the vessel lumen (Fig. 6, J and K). Moreover, a significantly increased number of cytoplasmic vesicles filled with the tracer horseradish peroxidase was observed in the Reln–/– endothelium (Fig. 6, J and L). The blood-retinal barrier was also impaired at P7 in Dab1iΔEC mice (Fig. 6, M to O), and defects were compensated after 4 weeks of age (fig. S9, A to C). We performed a late acute removal of Dab1 by using the Cdh5(PAC)creERT2 deleter mice and injecting tamoxifen for three consecutive days starting at P25 and evaluating the BBB permeability at P30. In this setting, removal of Dab1 in mature vessels did not seem to significantly affect the BBB maintenance (fig. S9, D to H). We also found a down-regulation of Dab1 expression at P30 relative to postnatal stages (fig. S9I). Overall, this suggests that Dab1 function is important for BBB etiology but not for physiological barrier maintenance.

Fig. 6 Endothelial Dab1 regulates NVU integrity.

(A) Fluorescent whole-brain images of Dab1iΔEC and control littermates injected with Alexa Fluor 555 cadaverine (Cad-A555) at P7. (B) Quantification of fluorescence intensity in (A) (n = 6 animals per genotype). (C) Cad-A555 and podocalyxin (Pdx) staining of P7 Dab1iΔEC cortices. (D) Fluorescent whole-brain images of adult Dab1iΔEC and control littermates injected with Cad-A555. (E) Quantification of fluorescence intensity in (D) (n = 5 to 7 animals per genotype). (F) Cad-A555 and Pdx staining of 4- to 7-week-old Dab1iΔEC cortices. (G) Fluorescent whole-brain images of 3- to 4-week-old Reln–/– and control littermates injected with Cad-A555. (H) Quantification of fluorescence intensity in (G) (n = 3 animals per genotype). (I) Cad-A555 and Pdx staining of Reln–/– cortices. (J) Transmission electron microscopy images of control and Reln–/– cortical vessels. (K) Quantification of the distribution of the tight junction (TJ) angle in (J) (n = 3 animals per genotype). (L) Quantification of the number of horseradish peroxidase (HRP)–filled vesicles relative to the lumen length in (J) (n = 3 animals per genotype). (M) Fluorescent whole-retina images of P7 Dab1iΔEC and control littermates after intravenous injection of Cad-A555. (N) Quantification of fluorescence intensity corresponding to (M) n = 6 to 8 retinas, 5 animals per genotype). (O) Representative confocal images showing an elevated cadaverine extravasation in Dab1iΔEC retinas from the blood vessels (IB4). Scale bars, 5 mm [(A), (D), and (G)], 50 μm [(C), (F), and (I)], 100 nm (J), 2 mm (M), 75 μm (O). Data are means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

Vascular Dab1 is necessary for laminin-α4 secretion and astrocytic integrin-β1 activation

Analysis of the NVU cellular components in Dab1iΔEC and Reln–/– mice revealed a prominent reduction of the coverage of cerebral vessels by the endfeet water channel aquaporin4 (Aqp4) (Fig. 7, A and B, and fig. S10, A and B). Total levels of Aqp4 were not changed, which suggested an uncoupling of astrocytic endfeet to the NVU (Fig. 7C and fig. S10C). We confirmed that pericyte ensheathment, also important for BBB integrity (44, 45), was not significantly affected in our mutants (fig. S11, A to D).

Fig. 7 Astrocytic endfeet attachment is reduced in reelin signaling mutants.

(A) Aquaporin4 (Aqp4), Pdx, and Cad-A555 staining of 4- to 7-week-old Dab1iΔEC cortices. Defects in Aqp4 coverage are detected in larger vessels (arrow) as well as smaller capillaries (arrowhead). (B) Aqp4 vessel coverage quantification as percentage of vessel area covered by Aqp4 staining in (A). Vessels of wide-ranging diameter were included in the quantifications (n = 3 to 6 animals per genotype). (C) Western blot for Aqp4 in brain lysates from Dab1iΔEC mutants. Pan-cadherin (pan-Cadh) was used as a loading control. (D) Laminin-α4 (Lama-4), laminin-α2 (Lama-2), and albumin staining of 3- to 4-week-old Reln–/– cortices. (E) Lama-4 and Aqp4 staining of Reln–/– cortices. Arrows indicate sites of colocalization of Lama-4 with Aqp4, which are decreased in the mutant (arrowheads). Boxes indicate magnification areas in (F). (F) Magnified pictures from (E) and line intensity profiles of staining intensity for Lama-4 (black lines) and Aqp4 (purple lines). Note the coincident distribution of staining in the control vessels, which is lost in Reln–/– vessels. Scale bars, 50 μm [(A), (D), (E), (F)]. Data are means ± SEM. ***P < 0.001.

The extracellular matrix (ECM) network between perivascular astrocytic endfeet and ECs regulates barrier properties of the cerebral vasculature (46). Laminins are major components of the gliovascular lamina (47). Albumin extravasation from vessels in the Reln–/– mice coincided with an impaired deposition of laminin-α4 (Lama-4, produced by endothelium) and a decrease in laminin-α2 [Lama-2, expressed at the endfeet (48)] (Fig. 7D). Moreover, deposition of Lama-4 coincided with sites of accumulation of Aqp4 staining in control animals (Fig. 7E, upper panels, arrows), which suggests that Aqp4-enriched endfeet dock on sites of endothelial Lama-4 accumulation. In agreement with this, Reln–/– mutants showed decreased Lama-4 deposition in correlation with the absence of Aqp4 staining (Fig. 7E, lower panels, arrowheads). Correlative accumulation of Aqp4 and Lama-4 can be appreciated in higher-magnification pictures and in the corresponding line intensity profile of the staining (Fig. 7F).

Lama-4 in the ECM binds to its cognate receptors integrin-α3β1 and integrin-α6β1, which are expressed by astrocytes and whose activation is required for endfeet anchorage (49, 50). Disrupted deposition of Lama-4 from ECs in vivo in Reln–/– and vascular Dab1 mutants also had an effect on the functional activation of integrin-β1. Analysis of the ratio of staining at the vessel wall versus the intracellular compartment of the endothelium revealed that Lama-4 accumulated at the vessel wall, and that such accumulation correlated with sites of integrin-β1 activation in control but not in Reln–/– or Dab1iΔEC mice (Fig. 8, A and B). Integrin-β1 is expressed by multiple cell types of the CNS, including vessels and astrocytes (51). Integrin-β1 activation also accumulated at the astrocytic endfeet, as evidenced by colocalization with the endfeet marker Aqp4 in control animals, and its coexpression was dimmer in the mouse mutants (Fig. 8C). Vascular signaling downstream of endothelial integrin-β1 has been shown to be important for the proper localization of VE-cadherin, and inactivation of endothelial integrin-β1 signaling leads to aberrant VE-cadherin distribution and extensive hemorrhaging in retinas (52). Reln–/– and Dab1iΔEC did not show any alteration in VE-cadherin arrangements in the vessels (fig. S11E), which suggests that the lack of Lama-4 deposition in the reelin signaling mutants preferentially affected the endfeet’s integrin-β1 activation.

Fig. 8 Dab1 instructive roles on astrocytic endfeet attachment are mediated by vascular laminin-α4 secretion and astrocytic integrin-β1 activation.

(A) Lama-4 and activated integrin-β1 (β1-Int) staining of Reln–/– and Dab1iΔEC cortices. Arrows indicate areas of Lama-4 and β1-Int enhanced colocalization. (B) Quantification of Lama-4 and β1-Int relative signal at the vessel wall in (A) [n = 3 (Reln–/–), n = 3 or 4 (Dab1iΔEC) animals per genotype]. (C) Representative pictures of β1-Int and Apq4 costaining, showing the activation of β1-Int in the astrocytic endfeet. (D) Lama-4 staining in reelin-stimulated brain ECs (bEND.3). (E) Representative quantification of Lama-4 fluorescent signal in (D) (n = 7 or 8 images per condition, 3 experiments). (F) Representative experiment of attachment of primary astrocytes to reelin-stimulated bEND.3 cells. The adhesion of astrocytes to the EC monolayer is impaired after blocking astrocytic β1-Int (n = 5 measurements per condition, 3 experiments). (G) Permeability assay based on an in vitro BBB model of cocultured ECs and astrocytes (n = 3 experiments). bEND.3 cells in the luminal compartment were stimulated with reelin 24 hours before astrocytic seeding in the abluminal compartment of the insert. (H) Lama-4 and brain lipid binding protein (BLBP) staining of the pial region of Dab1iΔEC embryos at E17.5 after TMX administration from E10.5 to E12.5. (I) Quantification of Lama-4 relative signal at the vessel wall in (H) (n = 3 animals per genotype). (J) Quantification of BLBP+ cells contacting pial vessels in (H) (n = 3 animals per genotype). Scale bars, 50 μm (A), 25 μm [(C) and (D)], 10 μm (H). Data are means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

Direct stimulation of brain ECs with exogenous reelin led to a specific increase in Lama-4 secretion (Fig. 8, D and E) without changes in total mRNA transcript and protein levels (fig. S12, A and B). Conversely, Lama-5, which is also expressed by brain ECs, was not affected by reelin stimulation (fig. S12, C to F). Also, total protein levels for endothelial laminins were not affected in vivo (fig. S12, G to J). Reelin-induced secretion of Lama-4 had a direct impact on the attachment of primary astrocytes to monolayers of brain ECs. Stimulation of brain ECs with reelin led to a significant increase in the attachment of seeded astrocytes; however, pretreating primary astrocytes with a blocking antibody against integrin-β1 (53) impaired the adhesion of the astrocytes onto the EC monolayer in a dose-dependent manner (Fig. 8F and fig. S13). Furthermore, an in vitro BBB model using a coculture of ECs apposed to astrocytes showed that reelin stimulation of the endothelial compartment significantly increased the tightness of the barrier and such effect was blocked when astrocytic integrin-β1 was inhibited (Fig. 8G), mimicking the leakiness we observed in the reelin signaling mutants in vivo. These results indicate that Lama-4 is required for integrin-β1–dependent astrocytic adhesion and barriergenesis.

The same defects in Lama-4 secretion were recapitulated at the missing contacts of the radial glia endfeet with the pial vessels during embryonic development (Fig. 8, H to J) that resulted in impaired neuronal migration (Figs. 4 and 5). These findings suggest that the same mechanism applies to the instructive role of vascular Dab1 in neuronal migration.


Our study provides precise mechanistic insights into an important function of the vasculature that goes beyond the supply of oxygen and nutrients and extends into an instructive structural role in building up functional architecture in the brain. Dab1 expressed in ECs is necessary to instruct the communication of vessels with the glia for the proper positioning of neurons during cortical development as well as for the correct communication at the NVU. We described how different neurovascular processes are synchronically regulated by the convergence on vascular Dab1 signaling. On one hand, reelin/ApoER2/Dab1 signaling regulates angiogenesis in the CNS by cross-talking to the VEGF/VEGFR2 pathway; on the other hand, vascular Dab1 regulates the deposition of Lama-4 and the docking of the radial glia/astrocytes on the vasculature, consequently affecting the neuronal migration at developmental stages and the etiology of BBB integrity.

Reelin in the vasculature exerts proangiogenic functions during development via the interaction with VEGF/VEGFR2 pathway. This function of Dab1 in ECs regulates proliferation and tip/stalk cell fate. The vascular growth defects observed during embryonic and postnatal stages are recovered during adulthood both in the retinal and in the cortical vessels. This suggests that other compensatory mechanisms are in place to correct for such defects in the vasculature. Dab2, a relative protein to Dab1, has been shown to be involved in developmental angiogenesis by controlling VEGFR2 endocytosis (54) and therefore could be a good candidate to compensate for the lack of Dab1 in the vessels. In agreement with this, we observed a down-regulation of Dab1 in the adult vessels, whereas Dab2 continues to be expressed during adulthood (55).

However, the neuronal defects provoked by the loss of function of Dab1 in the embryonic and early postnatal vasculature persist to adulthood and have a severe impact on neuronal positioning and BBB function. Dab1 in neurons is essential to guide correct neuronal positioning in the cortex during development (36, 56). Dab1 cell-autonomous function in neurons seems to be involved specifically in the somal translocation process during radial neuronal migration but not in glia-mediated locomotion (36). We have shown that Dab1 in the vessels is essential for both the maintenance of the RGC scaffold supporting migration of later-born neurons and their proper navigation toward the correct cortical layer during somal translocation. Dab1 deficiency in the vessels leads to detachment of the glial processes observed during both embryonic and early postnatal developmental stages, when the endfeet of the RGCs are anchored to the pial basement membrane and later when the mature cortical astrocytes extend their endfeet to surround the vessels and form a functional BBB. In agreement with our results, several studies have shown that RGC detachment of the pial surface induces deficiencies in neuronal positioning (39, 40, 57). Notably, neuroglial integrin-β1 conditional mutant mice showed important layering defects, including neuronal invasion to the marginal zone, originated by the deficient anchorage of radial glia processes to the meningeal surface (39), which we have shown to be recapitulated by the Dab1 endothelial-specific mutant mice. Furthermore, it has also been shown that proper neuronal migration in the cortex requires the selective expression of integrin-β1 by RGCs but not by neurons (58), hence the activation of integrin-β1 in glial cells is crucial for proper corticogenesis.

The basement membrane components Lama-2 and Lama-4 are involved in the attachment of the radial glia processes to the meningeal surface (59). In line with these studies, we found that vascular reelin signaling regulates the secretion of Lama-4 by ECs, which in turn has an impact on radial glia attachment to the pial surface and regulates neuronal migration. Although the regulatory mechanisms of Lama-4 deposition are still largely unknown, it has been reported that Lama-4 is accumulated in focal adhesions (60), which might also be influenced by reelin signaling (61).

Defects in Lama-4 deposition and integrin-β1 activation also altered the assembly of the astrocytic endfeet ensheathment on the vasculature and consequently resulted in increased BBB permeability. Astrocytes exert a crucial function in maintaining the BBB in different physiological and pathological conditions (62) through their endfeet anchorage to the basal lamina proteins, which are relevant elements regulating BBB function in health and disease (46). Interestingly, Lama-4 null mice suffer from multiple hemorrhages starting at embryonic stages (63). Although the severe defects derived from the complete abrogation of Lama-4 expression are not comparable to the local defects in Lama-4 deposition observed in our mutants, it is remarkable that both animal models show a loss of BBB integrity.

Changes in Lama-4 secretion could regulate both endothelial and glial integrin-β1 activation. However, our vascular Dab1 loss-of-function mutants phenocopy the defects shown by several reports after neuroglial deletion of integrin-β1 (39, 58, 59). Conversely, vascular loss of integrin-β1 has rather divergent phenotypes relative to those seen with vascular Dab1 deletion, such as hyperproliferation, sprouting, or altered VE-cadherin patterning (52, 64). In addition, we have shown that astrocytic adhesion to endothelial cells and astrocytic-induced barrier properties are dependent on reelin signaling on ECs and that astroglial integrin-β1 activation is indispensable for such processes. Hence, on the basis of previous literature and our observations, we deduce that astroglial integrin-β1 activation plays a relevant role in endothelial-astroglial interaction.

Together, our findings constitute a fundamental example of the integration of the neurovascular signaling that converges in the organization of the neuronal-(astro)glial-vascular elements that regulate the development and homeostasis of the CNS. Such integrative signaling might explain, at the molecular level, the comorbidity of neuropathological conditions and vascular dysfunction.

Materials and methods

Genetically modified mice and treatments

Dab1flox/flox mice carrying loxP sites flanking exon 2 for the Dab1 gene (36) (kindly provided by U. Mueller) were crossed with Cdh5(PAC)-CreERT2 (65) (kindly provided by R. Adams) to generate endothelial cell–specific Dab1 knockout mice (Dab1iΔEC). VEGFR2flox/flox mice (66) (kindly provided by E. Wagner) were crossed with Cdh5(PAC)-CreERT2 (65) to generate endothelial cell–specific VEGFR2 knockout mice (VEGFR2iΔEC). Cre activity was induced by intraperitoneal injection of 0.1 ml of tamoxifen (1 mg/ml) or 4-hydroxytamoxifen (4-OHT) (1 mg/ml) each day for 3 days (P1 to P3) for postnatal analysis at P3–P4, P7–P8, and postnatal weeks 4 to 7. For adult induction, 0.1 ml of 4-OHT (5 mg/ml) was intraperitoneally injected for 3 consecutive days (P25 to P27) for analysis at P30. For embryonic induction, pregnant females were injected intraperitoneally with 0.2 ml of 4-OHT (10 mg/ml) each day for 3 days (E10.5 to E12.5). Tamoxifen injectable solution was prepared in ethanol and peanut oil as described (67). Cre-negative animals were used as controls. Additionally, the Cdh5(PAC)-CreERT2 line was crossed with the ROSA26R(EYFP) reporter strain and genetic recombination was induced as described for the Dab1iΔEC mice. Reelin knockout (Reln–/–) and ROSA26R(EYFP) mice were obtained from Jackson Laboratories; ApoER2 knockout (ApoER2–/–) mice were kindly provided by J. Herz (68). For Reln–/– mice, both wild-type and heterozygous littermates were used as controls, except as mentioned. Wild-type C57BL/6J animals were used for fluorescent in situ hybridization (FISH) analysis, isolation of primary cell cultures, and retinal explant cultures. Both males and females were used for all experiments indistinctively.

All animals were genotyped by PCR. Protocols and primer sequences were used as described by the distributor or donating investigator.

For intravenous Alexa Fluor 555 cadaverine tracer injection and detection, mice were deeply anesthetized by injection of ketamine and xylazine at 180 mg/kg and 10 mg/kg of body weight, respectively; 50 μl of Alexa Fluor 555 cadaverine (1 mg/ml) was injected into the retro-orbital venous sinus in 4- to 7-week-old (Dab1iΔEC) or 3- to 4-week-old (Reln–/–) mice as described (69). For postnatal mice (P7), 40 μl of Alexa Fluor 555 cadaverine (1 mg/ml) was injected intraperitoneally. Cadaverine was allowed to circulate for 20 min in adult mice and for 2 hours in postnatal mice, respectively. Mice were then perfused intracardially with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS). Eyes from P7 animals were fixed in 4% PFA for 2 hours at room temperature (RT) before retina isolation in PBS. Eyes from 4- to 7-week-old adult mice were fixed in 4% PFA for 10 min prior to retina isolation in 4% PFA and post-fixed overnight (4% PFA, 4°C). Brains were post-fixed at RT in 4% PFA for 3 hours. For cadaverine detection, whole brains and isolated retinas (mounted in PBS) were imaged using a dissecting microscope with an attached fluorescent lamp and a Texas Red filter and subsequently processed for immunostaining. Cadaverine leakage was quantified by measuring mean fluorescence intensity of whole-brain pictures.

For BrdU experiments, pregnant females were injected intraperitoneally with 0.2 ml of BrdU solution (10 mg/ml) dissolved in sterile PBS at E15.5. Brains from the offspring were collected and analyzed at P8.

For intravenous peroxidase injection, horseradish peroxidase Type II (HRP) was dissolved at 100 mg/ml in sterile PBS, injected at 10 mg per 20 g body weight in deeply anaesthetized mice, and let to circulate for 30 min. Mice were perfused with 0.3 M HEPES, 1.5% PFA, and 1.5% glutaraldehyde (GDA) and processed for histochemistry.

For intracardiac isolectin B4 (IB4), mice were processed the same way as for regular perfusion. Prior to the PBS/4% PFA perfusion, 1 ml of IB4 (20 ng/μl) was intracardially injected.

All animal experiments were approved by the Regierungspräsidium of Darmstadt and the Veterinäramt of Frankfurt am Main.

Cell culture and treatment

For isolation of primary mouse brain endothelial cells, mouse brain microvascular fragments were processed as described (70). Briefly, capillary fragments were seeded on collagen I–coated wells and cultured in DMEM (Dulbecco’s modified Eagle’s medium) with 20% fetal bovine serum (FBS) supplemented with heparin (100 μg/ml) and EC growth supplement (ECGS; 5 μg/ml). After 2 days of puromycin selection (4 μg/ml), cells were cultured for two more days without puromycin before their experimental use.

Mouse primary lung endothelial cells (MLECs) were isolated as described (27). Lungs were isolated into dissection buffer (HBSS supplemented with 10% FBS) and treated with collagenase type II. Filtered tissue pellets were resuspended in dissection buffer and incubated with anti-rat IgG–coated magnetic beads, pre-coupled with rat anti-mouse CD31. Beads were resuspended in endothelial cell medium consisting of high-glucose DMEM GlutaMAX-I, penicillin (100 U/ml), streptomycin (100 μg/ml), 20% FBS, 0.4% endothelial cell growth supplement with heparin and plated onto gelatin-coated plates.

For isolation of mouse primary astrocytes, brains of P2–P5 mice were isolated into dissection buffer (DMEM) and subsequently incubated in 0.25% trypsin/EDTA for 20 to 30 min at 37°C. Digested tissue was homogenized with a fire-polished Pasteur pipette and centrifuged. Pelleted cells were resuspended in astrocyte medium consisting of high-glucose DMEM GlutaMAX-I, penicillin (100 U/ml), streptomycin (100 μg/ml), 10% FBS, 1% MITO+ Serum Extender and plated onto flasks. After the cultures were confluent, the astrocytes were purified by shaking the flasks on a rotator at 250 rpm at 37°C for 2 days to detach all other cell types.

Pooled human umbilical vein endothelial cells (HUVECs) were cultured in endothelial basal medium (EGM) supplemented with hydrocortisone (1 μg/ml), bovine brain extract (3 μg/ml), gentamicin sulfate (30 μg/ml), amphotericin B (50 μg/ml), EGF (10 μg/ml), and 10% FBS on gelatin-coated culture dishes at 37°C, 5% CO2.

For the detection of phospho-Dab1 (pDab1) by immunostaining upon reelin stimulation, HUVECs were starved in DMEM containing 2% FBS overnight. Stimulation was performed by adding 150 μl of reelin supernatant or green fluorescent protein (GFP) supernatant for 45 min. For the detection of pDab1 by immunostaining upon VEGF-A stimulation, HUVECs were starved in serum-free medium (EGM) for 4 hours prior to stimulation with recombinant VEGF-A (50 ng/ml) for 30 min. For costimulation with VEGF-A and reelin, HUVECs were stimulated with 40-fold concentrated reelin supernatant and recombinant VEGF-A (50 ng/ml) for 30 min. Mouse primary brain endothelial cell (MBEC) cultures were starved with Opti-MEM serum reduced medium at 37°C, 2 hours prior to stimulation. For stimulation, MBECs were treated with 40-fold concentrated reelin supernatant or GFP supernatant. For the detection of pDab1, cultures were stimulated for 15 min at 37°C. Dab1 phosphorylation was assessed by measuring pDab1 fluorescence intensity (HUVECs and MLECs) or by measuring the area covered by pDab1 fluorescence (MBECs).

For Western blot analysis, cells were starved in EGM for 1 hour prior to stimulation. Stimulation was performed by adding recombinant VEGF-A (50 ng/ml), recombinant VEGF-C (100 ng/ml), recombinant reelin fragment (100 ng/ml unless mentioned otherwise) for 15 min or reelin supernatant for 30 min.

The immortalized mouse brain microvascular endothelial cell line bEND.3 was grown in DMEM supplemented with 10% FBS and antibiotics. For the detection of Lama-4 and Lama-5 by immunostaining upon reelin stimulation, bEND.3 cells were starved in minimum essential medium (MEM) containing penicillin/streptomycin (100 μg/ml) for 2 hours. Stimulation was performed by adding recombinant reelin fragment (100 ng/ml) overnight. Cells were fixed with 4% TCA for 10 min at RT and processed for immunostaining. For quantifications, % of area covered by positive pixels of Lama-4 staining was measured using ImageJ and normalized to the number of cells (number of DAPI+ nuclei) per field. Fisher’s compared t-values test was used to estimate the significance among the biological replicates.

Preparation of reelin-containing and control supernatants

To obtain reelin-enriched supernatants and GFP control supernatants, incubation medium [DMEM, penicillin (100 U/ml), streptomycin (100 μg/ml), G418 (0.360 g/liter), 10% FBS] from reelin-transfected 293-HEK cells or GFP-transfected control 293-HEK cells (a gift from M. Goetz) was replaced by serum-free medium containing penicillin/streptomycin (100 μg/ml) and cells were incubated for 2 days at 37°C, 5% CO2. The conditioned medium was collected and concentrated 40-fold by centrifugation using filter units. Reelin content, as well as its absence in control cell supernatants, was confirmed by Western blotting using mouse anti-reelin antibody 1:1000 (Millipore; MAB5364).

Tube formation assay

Tube formation assays were performed using μ-slides angiogenesis. Experimental settings were set up as suggested by the manufacturer with minor modifications. Briefly, 2 × 104 HUVECs were cultured on growth factor–reduced Matrigel at 37°C in a humidified incubator supplied with 5% CO2 for 6 to 8 hours in the presence of concentrated reelin supernatant or control GFP supernatant. The tubular network was quantified by counting the number of branch points and by measuring total tube length.

Mouse retinal organotypic explants

Retinal explant experiments were performed as described (24). In short, eyes were enucleated from newborn pups (P3–P5) and transferred to FBS-free DMEM medium. Retinas were dissected from eyecups and vitreous bodies were removed. Retinas were flat-mounted onto the hydrophilic polytetrafluoroethylene (PTFE) membrane of culture plate inserts with the nerve fiber layer facing the membrane. DMEM with 10% FBS was layered underneath the membrane and dropped on the retinas to prevent dryness. Retina explants were incubated at 35°C in a humidified incubator with 5% CO2 for 2 to 4 hours before stimulation. For stimulation of retina endothelial tip cells, VEGF-A was diluted in DMEM or Opti-MEM I and 3% FBS to a final concentration of 1 μg/ml and layered underneath the insert membrane and dropped on the explants. 40-fold concentrated reelin and control GFP supernatants were used for reelin stimulation experiments. For the combined VEGF-A/reelin stimulation experiments, VEGF-A (100 ng/ml) was diluted in the concentrated reelin or GFP supernatants. Stimulation was carried out at 35°C in a humidified incubator with 5% CO2 for 4 hours. Explants were fixed with 4% PFA at RT for 30 min and stained with IB4 (1:200). For quantifications, numbers of filopodia at the vascular front were counted and normalized against 100 μm vessel length.

Proximity ligation assay

Endothelial cells were fixed with 4% PFA in PBS for 10 min at RT and permeabilized using 0.1% Triton X‐100 in PBS for 4 min on ice. Subsequently, the cells were washed with PBS, and blocking solution (2% BSA, 4% NDS in PBS) was applied for 30 min at 37°C in a humidified chamber. Proximity ligation assay was performed as suggested by the manufacturer. All incubation periods were performed at 37°C. In brief, primary antibodies against ApoER2 1:80 (Abcam; ab86548) and VEGFR2 1:80 (R&D Systems; AF644) in blocking solution were added for 30 min. The cells were incubated with the corresponding PLA probes for 60 min. Phalloidin-FITC 1:500 was added to visualize actin filaments. Ligation and amplification of the probes was performed for 30 min and 100 min, respectively. Cells were incubated with DAPI, washed, and mounted. Incubation times with antibodies occurred at 4°C and overnight for the primary antibodies and 1 hour at 37°C for the PLA probe antibodies. For quantification, the number of PLA probe punctae per cell was counted.

Immunohistochemistry, immunocytochemistry, and FISH

All samples from mutant and littermate controls in an experiment were always processed and stained at the same conditions.

Brains were dissected and post-fixed for 3 hours in 4% PFA or 10% TCA at RT for immunostainings or overnight in 4% PFA at 4°C for FISH. For immunostainings, brains were sectioned coronally at 80 μm using a vibratome. For FISH, retinas and brains were cryoprotected by consecutive immersions in 15% and 30% sucrose in PBS at 4°C. Samples were then embedded in Tissue-Tek O.C.T. compound and frozen on dry ice. Coronal sections with a thickness of 16 μm were generated using a cryostat microtome. For retinal whole-mount staining, eyes were collected from mutant mice and their control littermates and fixed in 4% PFA solution overnight at 4°C or for 2 hours at RT.

Embryonic, postnatal, adult brain, and retina sections were incubated with primary antibodies after 30 min or 1 hour of blocking and permeabilization with 5 to 10% normal donkey serum (NDS), 0.5% Triton X-100 in PBS, respectively. For BrdU immunodetection, 80 μm vibratome sections were pretreated with 2 N HCl for 30 min and subsequently neutralized with sodium tetraborate (Na2B4O7, 0.1 M).

For immunostaining of whole retinas, retinas were isolated, blocked, and permeabilized in 5% NDS and 0.5% Triton X-100 in PBS at RT for 1 hour. Primary antibodies were diluted in 5% NDS in PBS. Incubations were performed overnight at 4°C. For retinal vessel visualization, retinas were incubated with IB4 (1:200) in 1% Triton X-100 in PBS.

Primary brain endothelial and bEND.3 cells were fixed in cold 10% TCA for 10 min and incubated with primary antibodies in 2% NDS in PBS for 2 hours at RT after 15 min of blocking and permeabilizing with 5% NDS and 0.1% Triton X-100 in PBS.

HUVEC and MLEC cultures were fixed with 4% PFA for 20 min, incubated with NH4Cl for 10 min at RT, blocked and permeabilized with 4% NDS, 0.2% Triton X-100 and 2% bovine serum albumin (BSA) in PBS for 30 min. Cells were incubated with the primary antibody in blocking solution for 1 hour at RT.

The following primary antibodies were used: rabbit anti-Cux1 1:100 (Santa Cruz; SC-13024), rabbit anti-aquaporin4 1:100 (Millipore; AB2218), mouse anti-BrdU 1:200 (Millipore; MAB3424), rabbit anti-ERG 1:200 (Abcam; ab92513), rabbit anti-phospho-Histone H3 1:200 (Millipore; 06-570), rat anti-mouse CD29 (activated integrin-β1) 1:100 (BD Pharmingen; 550531), rabbit anti–brain lipid binding protein (BLBP) 1:100 (Millipore; ABN14), rabbit anti-Glut1 1:200 (07-1401; Millipore), rat anti-laminin α-2 1:100 (abcam; ab11576), goat anti-laminin α-4 1:100 (R&D; AF3837), rabbit anti-laminin α-5 1:100 (Novus Biologicals; NBP1-18714), mouse anti-NeuN 1:200 (Millipore; MAB377B), rabbit anti-NeuN 1:200 (Millipore; ABN78), rabbit anti-Pax6 1:200 (Covance; PRB-278P), rabbit anti-Tbr1 1:200 (Abcam; ab31940), rabbit anti-Tbr2 1:200 (Abcam; ab183991), goat anti-PDGFRβ 1:200 (Neuromics; GT15065-100), rabbit anti-phospho-Dab1 (Y232) 1:200 (Cell Signaling Technology; 3325S), rat anti-VE-cadherin 1:100 (BD Pharmingen; 555289) and goat anti-reelin 1:100 (R&D Systems; AF3820). After primary antibody incubation, samples were washed with PBS (cells and retinas) or TBS-T (150 mM NaCl, 25 mM Tris base, 0.1% Tween 20; pH 7.6; brain sections) for three times and incubated with the appropriate fluorophore-coupled secondary antibodies for 1 hour at RT (cells and postnatal retinas) or overnight (brain sections and adult retinas). The secondary antibodies were Alexa Fluor 488-, 555-, 568- and 647-conjugated donkey anti-rabbit/mouse/goat 1:200 (Life Technologies) or anti-rabbit-Cy3 secondary antibody 1:200 (Jackson Immunoresearch). Nuclei were counterstained with DAPI. Sections were washed with PBS or TBS-T before mounting them using fluorescence mounting medium. Proliferating endothelial cells were quantified as described (71). The tip cell/stalk cell ratio was assessed as suggested by others (72, 73). Retinal tip cell filopodia quantification was performed as described (27). The number of branch points in the retina vasculature was assessed by counting the number of branch points between artery and vein and normalizing against the selected area. Vascular density in adult retinas was assessed by thresholding and measuring the area covered by IB4 signal in all three layers of the adult retina. Brain vascular parameters (vessel density, total vessel length, vessel orientation, and number of branch points) were analyzed in the cortical area using AngioTool (74) and ImageJ software (75). Aqp4 coverage was quantified by measuring the area of the vessel covered with staining. The area of Aqp4 staining was normalized to the vessel area (podocalyxin staining). Lama-4 and integrin-β1 signals were quantified by measuring the immunofluorescence intensity at the vascular walls normalized to the intracellular compartment. Metamorph software was used for these morphometric analyses.

For FISH, whole C57BL/6 adult mice brains were dissected, and RNA was extracted using TRIzol reagent. RNA was reverse-transcribed into cDNA using High Capacity cDNA Reverse Transcription Kit and the resulting cDNA used to obtain the PCR products. The primer sequences used for each probe are: Dab1-probe1-Fw AACCTGTTATCCTGGACTTGA, ISH: Dab1-probe1-Rv TGAACAAGGGGCTGCTGGCC, ISH: Dab1-probe2-Fw, GTCCATAAATCATGGGACTGGT, ISH: Dab1-probe2-Rv, TGGAGAGACTCAGATAGCCACA, ISH: ApoER2-Fw, TCTACTGGACAGACTCAGGCAA and ISH: ApoER2-Rv, CGGTAGCATCTCTTCATGTCTG. Probes for Dab1-probe2 and ApoER2 were obtained from Allen Brain Atlas:; PCR conditions, PCR product purification, cloning, transformation, and plasmid amplification were done as described (76). Plasmids with the right sequence were linearized using restriction enzymes and purified with Wizard SV Gel and PCR Clean-Up System. Finally, linear plasmids were transcribed into RNA probes labeled with digoxigenin (DIG). FISH was performed as described (76) with minor modifications: (i) slices of P7-8 brains and retinas were incubated in proteinase K solution (12 μg/ml) at 37°C for 12 min, (ii) Dab1 detection was performed by mixing two different riboprobes mixed together, (iii) anti-DIG-alkaline phosphatase was incubated together with goat anti-podocalyxin 1:200 (R&D Systems; AF1556), (iv) signal of DIG was detected using HNPP/Fast Red and podocalyxin with Alexa Fluor 647-conjugated donkey anti-goat, 1:250 (Life Technologies).

Images were taken using a laser scanning confocal spectral microscope. Brightness and contrast of the images were adjusted using the software Adobe Photoshop CS6 or ImageJ. Figures were prepared using Adobe Illustrator CS5.1.

Transmission electron microscopy (TEM)

For assessing the rate of transcytosis, HRP-diaminobenzidine histochemistry was performed on brain vibratome sections from animals injected with HRP (see above). Sections were incubated with 3,3′-diaminobenzidine tetrahydrochloride (DAB) for 30 min at RT and washed with PBS. Subsequently, samples were processed for TEM imaging.

Tissue was post-fixed with 6% glutaraldehyde/0.4 M PBS for 24 hours at RT and subsequently washed 5 times in 0.1 M Epon-PBS. Small tissue samples were cut out from the parasagittal area and processed with a tissue processor with 1% osmium tetroxide. Dehydration steps were followed by using increasing ethanol concentrations (25%, 35%, 50%, 70%, 75%, 85%, 100%). Prior to embedding, tissue combined with resin (Agar 100 Resin Kit) was dehydrated in an exsiccator for 24 hours. After embedding, the samples were kept in a steaming cabinet at 60°C for minimum of 4 days.

From the resin-embedded tissue, ultrathin sections (0.23 μm) were cut with a microtome and placed at 200 mesh copper grids (3.05 mm). Ultrathin sections were then contrasted with EM AC20 (0.5% uranyl acetate/Ultrostain I and 3% lead citrate/Ultrostain II). Samples were examined with a transmission electron microscope equipped with a Slowscan-2K-CCD-digital camera (2K-wide-angle). Morphometric analyses were performed with ImageSp software. For quantifications, numbers of HRP-filled vesicles were normalized per length of vascular lumen, and the angle of tight junction in relation to the lumen surface was measured with ImageJ software.

Quantitative real-time PCR

Whole C57BL/6 adult mice brains were dissected, and RNA was extracted using TRIzol reagent. RNA was reverse-transcribed into cDNA using High Capacity cDNA Reverse Transcription Kit. Quantitative PCR assays were performed using an ABI 7500 Fast Real-Time PCR System using TaqMan Fast Universal PCR master mix and TaqMan Gene Expression probes for mouse Dab1 (Mm01256039_m1), mouse Reelin (Mm00465200_m1), mouse Vegf (Mm00437306_m1), mouse Laminin4 (Mm01193660_m1), mouse Laminin-α5 (Mm01222029_m1), and mouse β2m (Mm00437762_m1), which served as an endogenous control.

PCR for Dab1 gene excision

Genomic DNA was extracted from the tail of the animals and PCR was performed using 5 μl of the DNA extract. The primers used to amplify the excised Dab1 fragment were: 5′GGTTCAGTGCCTATCATGTATC3′(Fwd) 5′CCTATACTTTCTAGAGAATAGGAAC3′ (Rv). PCR was performed with Promega GoTaq Green Master Mix using Tm = 54°C and 38 cycles. PCR product was loaded in a 3% agarose gel and DNA was stained with 0.01% ethidium bromide. After electrophoresis, gel was imaged in a transilluminator using a UV light source.

Attachment assay

bEND.3 cells were seeded in quintuplicates on a 96-well plate. bEND.3 cells were grown overnight at 37°C and were then starved with MEM containing penicillin/streptomycin (100 μg/ml) for 2 hours at 37°C. After starvation, bEND.3 cells were stimulated with recombinant reelin at 100 ng/ml overnight at 37°C in starving medium. Primary astrocytes were incubated with 6 μM Texas Red Hydrazide in order to label them fluorescently and with rat anti-mouse integrin-β1 blocking antibody (BD Biosciences; 553715) at the indicated concentrations (53) for 30 min at 37°C. After the stimulation of the bEND.3 cells, astrocytes were pelleted, resuspended in medium and seeded (40,000/well) onto the bEND.3 cell monolayer for 3 hours at 37°C. After the attachment incubation, non-attached astrocytes were washed gently 3 times with PBS. Finally, 100 μl of 1% SDS solution was added per well and fluorescent intensity was read (λex 590 nm; λem 620 nm).

Permeability assay

We used an in vitro model of BBB based in a coculture of endothelial cells and astrocytes seeded in the opposite sides of a prehydrated 24-well membrane insert. 40,000 bEND.3 cells were seeded onto the luminal side of the insert and cultured in bEND.3 medium for 24 hours in a 37°C, 5% CO2 cell culture incubator until monolayer was formed. Prior to reelin stimulation, bEND.3 cells were washed with PBS and starved in MEM for 2 hours in an incubator. bEND.3 cells were then stimulated with recombinant reelin (100 ng/ml) for 24 hours. Shortly prior to adding astrocytes, reelin was removed and bEND.3 cells were kept in bEND.3 medium. Primary cortical astrocytes were pre-incubated with blocking integrin-β1 antibody for 30 min. Pellet of blocked astrocytes was resuspended in bEND.3 medium. 40,000 astrocyte cells were seeded on the abluminal side of the insert to let them attach for 3 hours in the incubator. Prior to permeability assay, both luminal and abluminal sides of the membrane insert were washed with PBS to remove unattached astrocytes and the excess medium. Insert was transferred to a new 24-well plate filled with 500 μl of PBS. In vitro BBB permeability was assessed by adding 200 μl of sodium fluorescein (100 μg/ml) to the luminal side of the insert. After permeation time of 15 min, the insert was transferred to a new well and another permeation was repeated (in total 4 times). The liquid in the well plate (now containing sodium fluorescein that crossed the cell layers) was thoroughly mixed and 100 μl of this was transferred to 96-well plate. The plate was read in a fluorescence plate reader (λex 460 nm; λem 515 nm).

Western blot

Tissue and cells were lysed in lysis buffer (50 mM Tris-HCl, pH 7.5; 150 mM NaCl; 1% Triton X-100; 1 mM sodium orthovanadate; 10 mM NaPPi; 20 mM NaF) or RIPA buffer (150 mM sodium chloride; 1% Triton X-100; 0.5% sodium deoxycholate; 0.1% SDS; 50 mM Tris, pH 8.0) and 1% complete protease inhibitor cocktail (Complete EDTA-free Proteinase inhibitor cocktail tablets). Protein content was determined using Pierce BCA Protein Assay Reagent according to manufacturer’s instructions and samples were separated by SDS-PAGE.

For immunoprecipitation of VEGFR2, endothelial cells were lysed with NET lysis buffer (50 mM Tris HCl buffer, pH 7.4, 15 mM EDTA pH 7.4, 1% NP-40, 150 mM NaCl, 10 mM sodium pyrophosphate, 20 mM NaF, 1 mM sodium orthovanadate, and 1% complete protease inhibitor cocktail) for 30 min at 4°C, centrifuged at 21,000g for 15 min and supernatants were collected. Samples were pre-incubated with protein G–Sepharose beads for 1 hour at 4°C. Beads were removed by centrifugation at 400g. Supernatants were incubated with protein G–Sepharose beads, rabbit anti-VEGFR2 antibody (Cell Signaling; 2479) and TBS/0.1% NP40 for 2 hours at 4°C. Samples were washed with NENT 300 washing buffer (20 mM Tris pH 7.4, 300 mM NaCl, 1 mM EDTA pH 7.4, 0.1% NP40, 25% glycerol) and TBS/0.1% NP40. For immunoprecipitation of Dab1, a similar protocol was applied with minor changes. Lysis buffer was used for lysis of MLEC, incubation of lysates with the antibody (goat anti-Dab1 antibody, Abcam, Ab16674) coupled sepharose beads and washing of the beads prior to Western blot analysis.

Protein samples from total lysates or IP were boiled with sample buffer (8% SDS, 200 mM Tris-HCl pH 6.8, 400 mM DTT, 0.4% Bromophenol blue, 40% Glycerol) prior to separation by SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked in TBS-T with skimmed milk powder (3% or 5%) or BSA (5%), depending on the antibody manufacturer’s recommendation. The following antibodies were used: rabbit anti-phospho-Dab1 Y232 1:500 (Cell Signaling Technology; 3325), goat anti-Dab1 1:1000 (Abcam; Ab16674), rabbit anti-VEGFR2 1:1000 (Cell Signaling Technology; 2479), rabbit anti-phospho-VEGFR2 Y1175 1:1000 (Cell Signaling Technology; 2478), rabbit anti-ApoER2 1:1000 (Sigma-Aldrich; A3481), goat anti-laminin-α4 1:1000 (R&D; AF3837), rabbit anti-laminin-α5 1:1000 (Novus Biologicals; NBP1-18714), rabbit anti-aquaporin4 1:1000 (Millipore; AB2218). Goat anti-actin, 1:1000 (Santa Cruz; sc-1615) and mouse anti-pan-cadherin 1:1000 (Sigma; C1821) were used as a loading controls. Primary antibodies were incubated overnight at 4°C, and membranes were subsequently incubated with HRP-conjugated secondary antibodies goat anti-rabbit HRP, donkey anti-goat HRP and goat anti-mouse HRP 1:1000 (Jackson Immuno Research Laboratories) in blocking solution 2 hours at RT. HRP activity was detected using enhanced chemiluminescence detection reagent (ECL) and the ImageQuant LAS 4000 system.

Statistical analysis

Quantifications were normalized to control and represented as percentage of control (%), unless otherwise indicated. Statistical significance was determined using 2-tailed unpaired Student’s t-test when comparing 2 variables, unless otherwise indicated. Statistical analysis was performed with Prism version 5. One-way analysis of variance (ANOVA) was performed in Prism Version 5 to assess statistical significance of the differences between multiple measurements. All animal experiments included animals from at least two litters. Statistical significance was defined as P < 0.05 (*), P < 0.01 (**) and P < 0.001 (***). All values indicate mean ± SEM.

Supplementary Materials

References and Notes

Acknowledgments: We thank U. Müller, J. Herz, E. Wagner, and R. Adams for providing mouse strains; S. Sawamiphak for initial help in the project; U. Bauer, K. Hammer, S. Seidel, D. Schmelzer, and T. Belefkih for technical support; and A. Frangakis and H. Zimmermann for helpful discussions. Funding: Supported by ERC_AdG_Neurovessel (project 669742); Deutsche Forschungsgemeinschaft grants SFB 834, SFB1080, FOR2325, EXC 115, and EXC 147; the Max Planck Fellow Program and Gutenberg Research College (GRC) at Johannes Gutenberg University Mainz (A.A.-P.); and the Marie Curie–CIG 293902 (M.S.). Author contributions: M.S. performed initial experiments and designed and supervised the project; M.R.A. designed, performed, and supervised experiments in the vascular-glia and BBB parts; F.C. designed, performed, and supervised experiments in the vascular part; C.L.-C. performed experiments in the vascular-glia and BBB parts; R.H. and I.B. performed experiments in the vascular part; M.D. performed PLA assays and VEGFR2 cortex analysis; M.P. performed in situ hybridizations; D.H. performed astrocyte and BBB permeability experiments and designed the model cartoons; A.S., H.S., and T.A. designed, performed, and supervised the electron microscopy; L.M. and M.R. performed initial experiments in the vascular part; L.T.-M. performed astrocyte binding assays; A.A.-P. analyzed data and designed and supervised all stages of the project; M.S. and A.A.-P. wrote the manuscript with contributions from M.R.A. and F.C.; and all authors discussed and interpreted the data and gave input to the written manuscript. Competing interests: The authors declare no competing interests. Data and materials availability: All data needed to evaluate the conclusions in this paper are present either in the main text or the supplementary materials. The Dab1 floxed mice are available from U. Müller under a material agreement with The Scripps Research Institute.
View Abstract

Stay Connected to Science

Navigate This Article