Canonical Wnt Signaling Regulates Organ-Specific Assembly and Differentiation of CNS Vasculature

See allHide authors and affiliations

Science  21 Nov 2008:
Vol. 322, Issue 5905, pp. 1247-1250
DOI: 10.1126/science.1164594


Every organ depends on blood vessels for oxygen and nutrients, but the vasculature associated with individual organs can be structurally and molecularly diverse. The central nervous system (CNS) vasculature consists of a tightly sealed endothelium that forms the blood-brain barrier, whereas blood vessels of other organs are more porous. Wnt7a and Wnt7b encode two Wnt ligands produced by the neuroepithelium of the developing CNS coincident with vascular invasion. Using genetic mouse models, we found that these ligands directly target the vascular endothelium and that the CNS uses the canonical Wnt signaling pathway to promote formation and CNS-specific differentiation of the organ's vasculature.

Several signaling pathways have been defined that generally regulate vascular development [e.g., (1)]. However, much less is known about the control of organ-specific vascularization and endothelial cell differentiation. The endothelium of the central nervous system (CNS) differs from most other organ systems in that the vascular cells establish a blood-brain barrier (BBB) that serves a critical neuroprotective role in preventing the free flow of substances between the blood and CNS [e.g., (2)]. Consequently, the BBB also presents a physical block to the passage of potentially therapeutic agents. Diseases of the brain's vasculature are the third leading cause of death in the United States (3). Thus, defining the signaling pathways that promote the formation and differentiation of the CNS vasculature has important clinical ramifications.

At embryonic day 8.5 (E8.5) in the mouse, migrating paraxial mesoderm–derived angio-blasts form a perineural vascular plexus (PNVP) surrounding the neural tube, the CNS anlagen. Endothelial cell sprouting from the PNVP initiates intraneural vascular plexus (INVP) formation at E9.5. Several Wnt family members, including Wnt7a and Wnt7b, are expressed in the neural tube coincident with neural tube angiogenesis (4). By E10.5, Wnt7a and Wnt7b are expressed in broad overlapping domains along the dorsal-ventral axis in the presumptive spinal cord and complementary patterns in the future forebrain (fig. S1). Because single null mutants for either Wnt7a or Wnt7b (5, 6) do not exhibit an early neural tube phenotype, we derived Wnt7a/b double mutant embryonic stem cell lines and used tetraploid aggregations to demonstrate a neural tube–specific hemorrhaging phenotype in embryos lacking both signaling activities.

To examine the role of these Wnts in detail, we generated a more tractable genetic system in which the embryo-specific removal of Wnt7a and Wnt7b function avoided an early lethality due to Wnt7b function in placental development (7). In this model, all Wnt7a/b double mutants die around E12.5, displaying a severe CNS-specific hemorrhaging phenotype and disorganization of neural tissue (Fig. 1, A and E, and figs. S2 and S3). These data indicated that Wnt signaling is essential for CNS vascular development. In most instances, when a single active Wnt7a or Wnt7b allele remained, no phenotype was observed (figs. S2 and S3).

Fig. 1.

(A to L) Neuroepithelial Wnt7a and Wnt7b expression is critical for CNS vascularization. Wnt7a+/+;Wnt7bc3/+ [(A) to (D)] mutants appeared phenotypically normal at E12.5. In contrast, Sox2Cre+/–;Wnt7a–/–;Wnt7bc3/d3 [(E) to (H)] and NestinCre+/–;Wnt7a–/–;Wnt7bc3/d3 [(I) to (L)] mutants exhibited severe hemorrhaging restricted to the CNS [arrowheads in (E) and (I) point to hemorrhaging in the CNS]. Furthermore, FLK1 [(B), (D), (F), (H), (J), and (L)] and PDGFRβ [(C), (G), and (K)] immunodetection of endothelial cells and pericytes, respectively, demonstrated that CNS vascularization was normal in Wnt7a+/+;Wnt7bc3/+ [(B) to (D)] mutants but was perturbed ventrally in Sox2Cre+/–;Wnt7a–/–;Wnt7bc3/d3 [brackets, (F) and (G)] and to a lesser extent in NestinCre+/–;Wnt7a–/–;Wnt7bc3/d3 mutants [(J) to (L)]. In the dorsal neural tube of Sox2Cre+/–;Wnt7a–/–;Wnt7bc3/d3 embryos, endothelial cells and pericytes formed abnormal clusters and enlarged lumens in vascular structures [e.g., arrowhead in (F)]. Boxes in (B), (F), and (J) show the approximate region of (D), (H), and (L), respectively. tel, telencephalon; sc, spinal cord; d, dorsal; v, ventral; NT, neural tube.

A detailed molecular analysis of CNS vascular development was performed using antibodies to several general endothelial markers, including fetal liver kinase 1 (FLK1) and the pericyte marker platelet-derived growth factor receptor–β (PDGFRβ) (Fig. 1, B to D and F to H, and figs. S3 and S4). Consistent with normal vascular endothelial growth factor (VEGF) signaling (8), a PNVP surrounded the neural tube in Wnt7a/b double mutants at E12.5. However, endothelial cells and pericytes were absent from all ventral neural regions of the presumptive spinal cord except the floor plate (Fig. 1, F to H). In the dorsal neural tube, endothelial cells and pericytes were present but clustered abnormally and formed vessels with expanded lumens, most noticeably where the expression of dorsal Wnts intersects with that of Wnt7a and Wnt7b (Fig. 1F and fig. S5). Thus, whereas Wnt7a and Wnt7b are critical for ventral CNS vascularization, other Wnts may also participate in this process in more dorsal domains of the presumptive spinal cord. When embryos were examined at E10.5, ventral vasculature was clearly already defective (fig. S6). In contrast, no obvious differences could be observed in neural tube patterning, neural proliferation, cell death, neural tube organization, or localization of the general angiogenic factor VEGF when control embryos were compared with Wnt7a/b double mutants (figs. S7 to S9).

To confirm that the Wnt7a/b double mutant phenotype was specific to Wnt7a/b action in the neuroepithelium, we used a NestinCre allele to conditionally remove Wnt7b activity from neuroepithelial cells on a Wnt7a mutant background (figs. S10 and S11) (9, 10). Although only partial recombination occurred by E10.5 in this model (fig. S10) (11), we observed CNS-specific hemorrhaging in the NestinCre;Wnt7a/b double mutants at E12.5 (Fig. 1I and fig. S2). A molecular analysis revealed the presence of a PNVP and an INVP (Fig. 1, J to L, and fig. S4); however, the numbers of endothelial cells and pericytes in the INVP were greatly reduced and the morphology of blood vessels was abnormal.

We next examined the cellular target and signaling mechanisms underlying Wnt7a/b action. Canonical Wnt signaling produces a transcriptional activation complex between β-catenin (βCat) and its DNA binding partners, LEF/TCF factors. Analysis of a mouse carrying a β-catenin–activated nuclear β-galactosidase transgene (a BAT-Gal reporter mouse) revealed sites of canonical Wnt signaling (12). β-Galactosidase was observed in FLK1/friend leukemia integration 1 (FLI1)–positive endothelial cells (13) in the PNVP and INVP at E10.5 (Fig. 2, A to D) (12). Several members of the Frizzled Wnt receptor family (fig. S12) and lymphoid enhancer binding factor 1 (LEF1), a critical component and direct feedforward target of canonical Wnt signaling, showed similar endothelial distribution (Fig. 2, E to H). Together these data suggest that Wnt7a and Wnt7b produced by the pseudostratified neuroepithelium may directly target endothelial cells in the basally positioned PNVP, thereby triggering angiogenesis through a canonical Wnt signaling pathway.

Fig. 2.

(A to H) Canonical Wnt signaling in endothelial cells is critical for CNS vascularization. In BAT-Gal mice, β-galactosidase was observed in FLI1/FLK1-positive endothelial cells in the PNVP and INVP of the E10.5 spinal cord [arrowheads, (A) to (D)]. The Wnt target LEF1 showed a similar endothelial cell distribution [arrowheads, (E) to (H)]. (I to P) Endothelial-specific removal of βCat function in Flk1Cre+/–;βCatc/n embryos resulted in a severe CNS-specific hemorrhaging phenotype in mice that survived to E12 [(I) and (M); arrowhead in (M) points to hemorrhaging in the CNS]. FLK1 [(J), (L), (N), and (P)] and PDGFRβ [(K) and (O)] immunodetection of endothelial cells and pericytes, respectively, demonstrated that a PNVP formed in Flk1Cre+/–;βCatc/n embryos [arrowheads in (N) to (P)], but there was an almost complete failure of angiogenesis into the adjacent neural tube [(N) to (P)]. Boxes in (J) and (N) show the approximate region of (L) and (P), respectively.

To examine this possibility, we removed βCat from vascular precursors combining a Flk1Cre driver line (14) with a conditional βCat allele (figs. S13 and S14) (15). In those embryos that survived until E12, the vascular phenotype was enhanced (Fig. 2, I and M, and fig. S15); the PNVP was present, but endothelial cells and pericytes were entirely absent from the neuroepithelium (Fig. 2, J to L and N to P). In contrast, when βCat activity was removed from neural progenitors, no vascular phenotype was observed (fig. S16). Thus, βCat plays a neural tube–specific role in regulating vascular development. The more severe phenotype in the Flk1Cre;βCat model is consistent with a broad role extending to other Wnt expression domains in the developing CNS.

Another report of βCat removal using a different vascular Cre driver (Tie2Cre) suggested an enhanced vascular fragility through destabilization of adherens junctions (16). However, our observations, including normal VE-cadherin in the Flk1Cre;βCat model (fig. S17), argue against a broad structural role for βCat in regulating vascular integrity and are more consistent with a localized signaling function. An up-regulation in γ-catenin in endothelial cells deficient in βCat (fig. S14) may functionally substitute for βCat's structural role, as in other systems (17, 18).

The development of a mature BBB results from activation of a CNS-specific endothelial cell differentiation program, the trigger for which is unclear. The first molecular changes are apparent within the PNVP coincident with vascular invasion of the neural tube; formation of a full BBB occurs over a prolonged period of development (2). The glucose transporter GLUT-1 is present in neuroepithelial cells before vascular invasion but is down-regulated in the neuroepithelium and up-regulated in the developing vascular endothelium as angiogenesis occurs; high levels of GLUT-1 are a hallmark of the adult BBB (fig. S18) (2). At E12.5, Wnt7a/b double mutants showed reduced levels of GLUT-1 in the PNVP and INVP (Fig. 3, A and E), and persistent GLUT-1 expression was observed in the ventral neuroepithelium in the absence of angiogenesis (Fig. 3, A to H). In Flk1Cre;βCat mutant embryos, GLUT-1 was also markedly down-regulated in the PNVP and persisted throughout the neuroepithelium (Fig. 3, I to P). Unlike the developing CNS vasculature, GLUT-1 levels were unaltered in other regions of Flk1Cre;βCat mutants where GLUT-1 levels were high, such as the limb vasculature. Taken together, these findings support a direct role for canonical Wnt signaling in organ-specific endothelial cell differentiation.

Fig. 3.

Canonical Wnt signaling is necessary for differentiation of the CNS vasculature. (A to H) The BBB marker GLUT-1 was severely down-regulated at E12 to E12.5 in the PNVP and INVP in Sox2Cre+/–;Wnt7a–/–;Wnt7bc3/d3 double mutants [(E) to (H)] relative to control embryos [(A) to (D); arrowhead in (D) points to a GLUT-1–positive vessel]. Some vascular segments expressed low levels of GLUT-1 in the mutants [arrowhead in (E)]. Weak GLUT-1 activity remained in the ventral neuroepithelium when vascular invasion failed [bracket in (E), arrowhead in (F)]. (I to P) In Flk1Cre+/–;βCatc/n embryos, at E12 a reduction of GLUT-1 in the PNVP relative to control embryos was evident [(I) to (L), (M) to (P); solid arrowhead in (P)]. However, strong GLUT-1 was detected in the neuroepithelium [(M) to (P); open arrowhead in (P)]. Boxes in (A), (E), (I), and (M) show the approximate region of (B) to (D), (F) to (H), (J) to (L), and (N) to (P) in separate sections, respectively.

To determine whether Wnt7 subfamily signaling is sufficient to induce GLUT-1 in endothelial cells, we targeted the Rosa26 locus with a construct enabling Cre-mediated control of the expression of a bicistronic message encoding Wnt7a and lacZ (fig. S19). In addition to driving Cre expression in the developing neuroepithelium, NestinCre showed low-level sporadic recombination in scattered cells outside the CNS (Fig. 4, A and E). NestinCre-driven ectopic Wnt7a expression in these regions enhanced GLUT-1 synthesis in endothelial cells (Fig. 4, B to D and F to H). Thus, Wnt7 signaling regulates an early feature of a CNS-specific vascular program.

Fig. 4.

Ectopic Wnt7a expression is sufficient to induce GLUT-1 in endothelial cells outside the CNS. (A to H) NestinCre-mediated ectopic activation of a bicistronic message encoding Wnt7a and E. coli nuclear β-galactosidase in scattered cells outside the neural tube [(A), arrowheads in (E)] resulted in up-regulation of GLUT-1 in vascular endothelium outside the CNS [compare (F) to (H) with control embryos in (B) to (D)].

Previous in vivo studies have suggested roles for various Wnts in specific aspects of vascular development. Wnt7b action has been linked to regression of the hyaloid vasculature by triggering apoptosis, as well as development of the lung vasculature via an indirect mechanism (19, 20). In addition, Wnt2 is thought to play a role in placental vasculature (21) and the putative Wnt receptor Fzd5 in vascular development of the yolk sac (22). In vitro studies have also suggested roles for Wnt signaling in the proliferation, survival, and formation of capillary-like networks (2327). Our study provides direct in vivo evidence for Wnt7a- and Wnt7b-mediated regulation of organ-specific vascular development. Their cellular target is the nascent endothelial network of the neural tube. These, and likely other Wnt family members, act through a canonical Wnt signaling pathway to promote formation and differentiation of the CNS vasculature. Whether this pathway also plays a later role in vascularization of other organ systems remains to be determined.

Our findings may have important clinical ramifications. For example, local reductions in Wnt signaling levels could potentially lead to malformation of CNS vasculature. In addition, if BBB properties in the adult are regulated by Wnt, altering Wnt activity may be a fruitful strategy for delivery of pharmacological agents to the CNS. Interestingly, there is a correlation between neo-angiogenesis and βCat accumulation in the endothelium of brain tumors such as gliomas and human glioblastoma multiforme (28, 29). This raises the possibility that canonical Wnt signaling may not only support vascular development but also promote tumor pathogenesis in the CNS.

Supporting Online Material

Materials and Methods

Figs. S1 to S19


References and Notes

Stay Connected to Science

Navigate This Article