Defective Angiogenesis in Mice Lacking Endoglin

See allHide authors and affiliations

Science  28 May 1999:
Vol. 284, Issue 5419, pp. 1534-1537
DOI: 10.1126/science.284.5419.1534


Endoglin is a transforming growth factor–β (TGF-β) binding protein expressed on the surface of endothelial cells. Loss-of-function mutations in the human endoglin gene ENGcause hereditary hemorrhagic telangiectasia (HHT1), a disease characterized by vascular malformations. Here it is shown that by gestational day 11.5, mice lacking endoglin die from defective vascular development. However, in contrast to mice lacking TGF-β, vasculogenesis was unaffected. Loss of endoglin caused poor vascular smooth muscle development and arrested endothelial remodeling. These results demonstrate that endoglin is essential for angiogenesis and suggest a pathogenic mechanism for HHT1.

HHT is an autosomal dominant vascular dysplasia characterized by recurrent epistaxis, telangiectasia, gastrointestinal hemorrhage, and pulmonary, cerebral, and hepatic arteriovenous malformations (1). ENG, the gene responsible for HHT1, encodes an endothelial transmembrane protein that binds to members of the TGF-β superfamily and their receptor complexes (2, 3). TGF-β signaling is required for the first stage of vascular development, vasculogenesis, when the primary capillary network, composed of interconnected and homogeneously sized endothelial tubes, is formed (4). The second stage of vascular development, angiogenesis, involves remodeling the primary endothelial network into a mature circulatory system (5,6). To understand the role of endoglin in vascular development, we used gene targeting to generate mice lacking endoglin.

The targeting vector was designed to replace the first two exons with a gene that conferred neomycin resistance (7) (Fig. 1A). Three targeted embryonic stem (ES) cell clones were identified and used to generate chimeric mice by morula aggregation. Southern blot analysis confirmed germ line transmission of the targeted allele (Fig. 1B). Immunohistochemistry was used to detect endoglin in the endothelium of Eng +/+ and Eng +/− mice by embryonic day 8.5 (E8.5), but neither endoglin protein nor mRNA was detected in Eng −/− mice (8) (Fig. 1, C to E). The life expectancy, fertility, and gross appearance of Eng +/− F1and F2 mice were normal; however, no homozygotes were found among 150 newborn animals from heterozygous intercrosses. By examining embryos from heterozygous intercrosses at different developmental stages, we determined that no Eng −/− mice survive after E11.5.

Figure 1

Targeted inactivation of murineEng results in defective vascular development. (A) Restriction maps of Eng genomic fragment, targeting construct, and predicted structure of targeted Engallele. Hatched boxes represent regions of homology shared by the targeting vector and genomic Eng. The probe for Southern blot analysis detects an 11.3-kb Hind III fragment from the disrupted allele and a 4.5-kb Hind III fragment from the wild-type allele. (B) Southern blot analysis of yolk sac DNA from E10.5 embryos probed for homologous recombination. (C) Absence of Eng transcript in E9.5 Eng −/− mice. Primers amplifying a 282–base pair region of Eng were used to amplify cDNA from total RNA. (D and E) Endoglin immunostain of E8.5 embryos demonstrates the presence and absence, respectively, of endoglin in Eng +/+ and Eng −/− mice. (F and G) Photomicrographs of E10.5 yolk sacs. The vasculature of the Eng +/+ yolk sac is well defined. Pockets of red blood cells are observed in the Eng −/− yolk sac with no discernible vessels. (Hand I) PECAM immunostain of yolk sacs at E10.5. Eng −/− endothelium fails to organize into vitelline vessels. (J and K) PECAM immunostain of head vessels at E10.5. The perineural capillary plexus fails to organize and the carotid artery (CA) is atretic in Eng −/− embryos. (D to G, J, and K) Bar = 1.0 mm; (H and I) bar = 0.1 mm.

At E10.5, Eng −/− mice were three times smaller than Eng +/+ mice and had fewer somites (18 to 22 in Eng −/− mice and 32 to 35 in Eng +/+ mice). These Eng −/− embryos exhibited an absence of vascular organization and the presence of multiple pockets of red blood cells on the surface of the yolk sac (Fig. 1, F and G). Expression of endothelial markers such as Flk-1, Flt-1, Tie-1, and Tie-2 and hematopoietic markers Gata-1 and Il-3r were not disrupted in Eng −/− mice (9). Thus, in contrast to TGF-β1 or its signaling receptor, there is no evidence that endoglin is required for endothelial differentiation or primitive hematopoiesis (4). The absence of organized vessels in the Eng −/− yolk sacs was confirmed by immunohistochemical staining for the endothelial marker platelet-endothelial cell adhesion molecule (PECAM) (Fig. 1, H and I) (10). The persistence of an immature perineural vascular plexus indicated a failure of endothelial remodeling in Eng −/− embryos (Fig. 1, J and K). At E10.5, the cardiac tube did not complete rotation in Eng −/− mice and was associated with a serosanguinous pericardial effusion (11). Although the cardiac tube continued to circulate blood at E10.5, by E11.5 there was evidence of resorption and necrosis in Eng −/− embryos.

PECAM immunostains demonstrated that the first organ system affected in Eng −/− embryos was the vascular system. At E8.5 and E9.5, the endothelial organization of Eng +/+ and Eng −/− embryos was similar (Fig. 2, A, B, E, and F). However, between E9.5 and E10.5, vascular development was disrupted in Eng −/− mice. Although there was extensive endothelial remodeling of the vasculature with expansion of existing vessels and sprouting and branching of new ones in Eng +/+ embryos at E10.5, the major vessels including the dorsal aortae, intersomitic vessels, branchial arches, and carotid arteries were atretic and disorganized in Eng −/− embryos (Fig. 1, J and K, and Fig. 2, I and J).

Figure 2

Poor vascular smooth muscle development in Eng −/− embryos precedes disruption in endothelial remodeling. Immunohistochemistry with antisera to PECAM (A, B, E, F, I, and J) and α-smc actin (C, D, G, H, K, and L). (A and B) At E8.5 the organization of Eng +/+ and Eng −/− endothelial tubes is indistinguishable. (C and D) At E8.5, initiation of vsmc differentiation occurs at the cranial-most aspect of the dorsal aortae (arrowheads). (E and F) At E9.5 the endothelial organization of Eng +/+ and Eng −/− embryos remains similar. The dorsal aorta (DA), branchial arches (BA), and intersomitic vessels (ISV) are identified. (G and H) vsmc formation in the Eng +/+ embryos extends caudally in the dorsal aortae and rostrally to the carotid arteries (arrowheads). vsmc development in Eng −/− embryos fails to progress from E8.5 to E9.5. (I and J) There is a marked maturation of endothelial organization in Eng +/+ embryos that is lacking in Eng −/− embryos. Large arteries like the carotid arteries, dorsal aortae, and intersomitic vessels are atretic in Eng −/− embryos. (K and L) vsmc surround the carotid arteries and the dorsal aortae in Eng +/+ embryos (arrowheads). In comparison, vsmc formation in Eng −/− embryos remains incomplete and sparse. (A, B, G, H, K, and L) Bar = 1.0 mm; (C, D, E, F, I, and J) bar = 0.2 mm.

Because TGF-β signaling has been shown to regulate vascular smooth muscle cell (vsmc) differentiation in vitro, we hypothesized that the disrupted angiogenesis in Eng −/− embryos was due to poor vsmc development (6). We stained embryos with an antibody to alpha smooth muscle cell actin (α-smc actin) to assess vsmc development. At E8.5 (10- to 12-somite stage), Eng +/+ and Eng −/− embryos were indistinguishable, with a foci of vsmc forming at the cranial-most aspect of the dorsal aortae (Fig. 2, C and D). By E9.5 (18- to 20-somite stage), Eng +/+ vsmc had extended rostrally to the carotid arteries and caudally through the dorsal aortae (Fig. 2G). At E10.5, vsmc of the Eng +/+ embryos surrounded the dorsal aortae, branchial arches, and carotid arteries (Fig. 2K). In contrast, there was poor vsmc formation of Eng −/− embryos at both E9.5 and E10.5 (Fig. 2, H and L). Thus, significant differences in development of Eng +/+ and Eng −/− vsmc were apparent by E9.5 and preceded the differences in endothelial organization observed between E9.5 and E10.5.

The failure in endothelial remodeling was not restricted to embryonic tissue. Vascular organization of E8.5 Eng +/+ and Eng −/− yolk sacs was similar and consisted of a primary endothelial network (Fig. 3, A and B). At E9.5, distinct vessels were forming in Eng +/+ yolk sac (Fig. 3E). In contrast, the vasculature of E9.5 Eng −/− yolk sacs failed to organize (Fig. 3F). By E10.5, distinct vessels were prominent in Eng +/+ mice but absent in Eng −/− mice (Fig. 1, H and I). Although disruption of endothelial organization occurs between E9.5 and E10.5, poor vsmc development is evident by E8.5 in Eng −/− yolk sacs. At E8.5, vsmc coated selective endothelial tubes in the Eng +/+ yolk sac but were scarce in the Eng −/− yolk sac (Fig. 3, C and D). By E9.5, vsmc in Eng +/+ yolk sacs outlined distinct vessels, whereas no progression was seen in Eng −/− yolk sacs (Fig. 3, G and H). Thus, the vsmc defect in Eng −/− extraembryonic tissue was evident by E8.5 and preceded the defect in endothelial remodeling.

Figure 3

Poor vascular smooth muscle development in Eng −/− yolk sacs precedes disruption in endothelial remodeling. Immunohistochemistry using antisera to endothelial markers Flk-1 (A and B) and PECAM (E and F) and the vsmc marker α-smc actin (C, D, G, and H). (A and B) At E8.5, a primary endothelial network is present in both Eng +/+ and Eng −/− yolk sacs. (Cand D) At E8.5 vsmc (arrowhead) develop around selective endothelial tubes from Eng +/+ yolk sacs. vsmc formation is scarce and unorganized in Eng −/− yolk sacs. (E andF) At E9.5, the primary endothelial network remodels into distinct vessels in Eng +/+ yolk sacs (arrowheads). There is no evidence of endothelial remodeling in E9.5 Eng −/− yolk sacs. (G and H) At E9.5, vsmc define distinct vessels in Eng +/+ yolk sac but not in Eng −/− yolk sac (arrowhead). Bar = 0.1 mm.

We used histologic analysis, in situ hybridization, and ultrastructural analysis to confirm that vascular development is disrupted in Eng −/− mice. Cross sections of α-smc actin immunostains identified vsmc between the endoderm and endothelium of the yolk sac. Few vsmc formed between these layers in Eng −/− compared with Eng +/+ yolk sacs (Fig. 4, A and B). Transverse sections of dorsal aortae showed vsmc developing around the endothelium of an Eng +/+ embryo at E9.5 (Fig. 4C). No vsmc are observed in a comparable section of an Eng −/− embryo (Fig. 4D). In situ hybridization for an early molecular marker of vsmc development, SM22α, showed a failure of vsmc to develop in E9.5 Eng −/− yolk sac and embryos (12) (Fig. 4, E, F, G, and H). Electron micrographs of E9.5 Eng −/− yolk sacs illustrated the absence of supporting cells, presumably pericytes or vsmc precursors, around the endothelium of the capillary network (13) (Fig. 4, I and J). Because vascular defects in Eng −/− mice are observed before embryonic circulation is established and before defects in cardiogenesis are documented, it is unlikely that failed vsmc development and arrested angiogenesis are secondary to impaired blood flow. These data support our conclusion that endoglin is required for normal vsmc development.

Figure 4

Poor vsmc formation in Eng −/− mice. (A to D) Transverse sections of anti-smc actin-stained embryos and yolk sacs at E9.5. (A and B) vsmc can be identified between the endoderm and endothelium in Eng +/+ and are scarce in Eng −/− yolk sacs (arrowheads). Cross reactivity of α-smc actin antisera with the mesothelium is observed in both Eng +/+ and Eng −/− yolk sacs. (C and D) vsmc form around the dorsal aortae (DA) of Eng +/+ embryos. No vsmc are identified in the dorsal aortae of Eng −/− embryos. (E to H) In situ hybridization of yolk sacs and embryos at E9.5 using an RNA probe for the vsmc marker SM22α. (Eand F) Expression of SM22α outlines Eng +/+ vessels and is absent in Eng −/− yolk sacs. (G and H) Expression of SM22α is present throughout the dorsal aortae of Eng +/+ embryos (arrowheads) but is absent from Eng −/− embryos. (I and J) Electron micrographs of Eng +/+ and Eng −/− yolk sac at E9.5. Supporting cells (SC), presumably vsmc or pericytes, are seen between the endoderm (N) and endothelium (E) of Eng +/+ yolk sacs but are absent in the Eng −/− yolk sac. M indicates mesothelium. (A, B, C, and D) Bar = 0.1 mm; (E, F, G, and H) bar = 1.0 mm.

Angiogenesis involves the differential growth and sprouting of endothelial tubes and recruitment and differentiation of mesenchymal cells into vsmc and pericytes (5). Our experiments demonstrate that endoglin is required for both processes. Because endoglin binds members of the TGF-β superfamily and interacts with their receptors, it is likely that endoglin regulates TGF-β signaling. This conclusion is supported by in vitro heterotypic coculture experiments in which endothelial cells induced vascular smooth muscle differentiation through a TGF-β pathway (14). Thus, our experiments indicate that TGF-β signaling is essential for angiogenesis.

Communication between the endothelium and mesenchyme is important for angiogenesis (5). Mesenchymal cells signal endothelial cells via the angiopoietin/Tie-2 signaling pathway, whereas endothelial cells induce differentiation of pericytes through the platelet-derived growth factor (PDGF) signaling pathway (15, 16). Although PDGF signaling is important for microvascular pericyte formation in the brain, we demonstrate that endothelial expression of endoglin is essential for vsmc development throughout the circulatory system. The subsequent failure of the endothelium to remodel in Eng −/− mice after arrested vsmc development suggests that vsmc may also play a role in regulating endothelial organization. Thus, we conclude that endoglin mediates a third pathway of endothelial-mesenchymal communication that is essential for angiogenesis and important to the pathogenesis of vascular disease.

  • * To whom correspondence should be addressed. E-mail:{at}


Stay Connected to Science

Navigate This Article