Endoglin in angiogenesis and vascular diseases
- First Online:
- Cite this article as:
- ten Dijke, P., Goumans, M. & Pardali, E. Angiogenesis (2008) 11: 79. doi:10.1007/s10456-008-9101-9
- 915 Views
Endoglin is a transmembrane auxillary receptor for transforming growth factor-β (TGF-β) that is predominantly expressed on proliferating endothelial cells. Endoglin deficient mice die during midgestation due to cardiovascular defects. Mutations in endoglin and activin receptor-like kinase 1 (ALK1), an endothelial specific TGF-β type I receptor, have been linked to hereditary hemorrhagic telangiectasia (HHT), an autosomal dominant vascular dysplasia characterized by telangiectases and arteriovenous malformations. Endoglin heterozygote mice develop HHT-like vascular abnormalities, have impaired tumor and post-ischemic angiogenesis and demonstrate an endothelial nitric oxide synthase-dependent deterioration in the regulation of vascular tone. In pre-eclampsia, placenta-derived endoglin has been shown to be strongly upregulated and high levels of soluble endoglin are released into the circulation. Soluble endoglin was found to cooperate with a soluble form of vascular endothelial growth factor receptor 1 in the pathogenesis of pre-eclampsia by inducing endothelial cell dysfunction. Endoglin is highly expressed in tumor-associated endothelium, and endoglin antibodies have been successfully used to target activated endothelial cells and elicit anti-angiogenic effects in tumor mouse models. These exciting advances provide opportunities for the development of new therapies for diseases with vascular abnormalities.
KeywordsAngiogenesisBMPEndothelial cellsHereditary hemorrhagic telangiectasiaSignal transductionSmadTGF-β
Endoglin is a TGF-β type III auxiliary receptor (TβRIII)  that is not directly involved in signaling, but modulates signaling responses of multiple members of the TGF-β family . In addition, endoglin is involved in TGF-β independent signaling . Moreover, two endoglin splice variants exist which can exert opposite effects [37, 38]. Furthermore, a soluble form of endoglin (sEnd) has been found, most likely generated by proteolytic shedding, which antagonizes the membrane bound form . Taken together, multiple layers of complexity exist by which the function of endoglin is regulated. In this review, we discuss the latest advances in our understanding of its mechanism of action and function in angiogenesis and vascular diseases.
Human endoglin is a homo-dimeric protein of 658 amino acid residues that contains an extracellular domain, a single transmembrane domain and a short intracellular domain (Fig. 1) . Endoglin is structurally related to betaglycan [39, 40], another TβRIII, with high similarity in the transmembrane and intracellular regions. Both endoglin and betaglycan are glycoproteins with N-linked and O-linked glycans; however, endoglin lacks the glycosamino-glycan (GAG) chains that are characteristic for betaglycan. Both type III receptors can form heteromeric complexes with TβRII, but whereas endoglin requires TβRII for binding to TGF-β1 and -β3, betaglycan alone can bind all three TGF-β isoforms [41, 42]. Endoglin and betaglycan are usually expressed as homodimers on the cell surface, which in the case of endoglin are linked by disulfide bridges. Heteromeric complexes between endoglin and betaglycan have been observed in microvascular endothelial cells . Besides the membrane-bound form, both type III receptors can occur in soluble form. High circulating levels of sEnd are detected in patients with cancer [44, 45] or pre-eclampsia [13, 46]. Betaglycan can be shed by membrane-type metalloprotease 1 (MT1-MMP  and sEnd is likely also generated by proteolytic cleavage of the membrane-bound form.
The extracellular domain of human endoglin contains an Arg-Gly-Asp (RGD) tripeptide sequence that is an interaction motif for integrins. However, this sequence is not conserved in mouse pig and rat [2, 48]. In its extracellular region endoglin harbors a zona pellucida (ZP) domain of approximately 260 amino acid residues, a feature shared with betaglycan . The ZP domain is involved in endoglin oligomerization and in ligand (in)dependent heteromeric interactions with the TGF-β receptors TβRII and ALK5 . Using single-particle electron microscopy, the three-dimensional structure of the extracellular domain of endoglin was determined at 25 Å resolution. The two monomers are positioned in an anti-parallel fashion with each monomer consisting of three well-defined domains , i.e., one domain with unknown structural homology and a ZP domain with bipartite structure (Fig. 1).
The intracellular domain of endoglin lacks an enzymatic motif but contains many serine and threonine residues of which certain residues are phosphorylated by TGF-β receptor kinases (see below) . At its C-terminus endoglin contains a PDZ interaction motif (SerSerMetAla). There are two splice isoforms, termed Long (L)- and Short (S)-endoglin, which differ in their intracellular part; L and S have cytoplasmic tails of 47 and 14 amino acids, respectively, and have only 7 juxtamembrane amino acids in common . The L form is most abundantly expressed and is also the predominant form in endothelial cells while the S-form is expressed in liver and lung at significant levels . Both forms can bind TGF-β , and differently regulate TGF-β-induced responses. L-endoglin enhanced the TGF-β/ALK1 pathway, while S-endoglin promoted the TGF-β/ALK5 route upon ectopic expression in myoblasts . Forced expression of S-endoglin in vascular endothelium in mice results in reduced tumor growth and neovascularization and suggest that S-endoglin (in contrast to L-endoglin) has anti-angiogenic activity. S-endoglin may elicit this effect by inhibiting the formation of L–L homodimers as L-and S-forms have been shown to form heterodimers .
Cellular and tissue distribution
Whereas endoglin is expressed at low to non-detectable levels in resting endothelial cells within normal tissues, it is highly expressed in vascular endothelial cells in sites of active angiogenesis during embryogenesis, in inflamed tissues, and within and surrounding tumors [3–5]. After ischemia-reperfusion injury and myocardial infarction, endoglin expression is upregulated in the ischemic area and border zone . Besides hypoxia, TGF-β, BMP9, and constitutively active (ca)ALK1 also potently stimulate endoglin expression [55–57], whereas TNFα inhibits endoglin expression in endothelial cells . The endoglin promoter contains multiple Sp1 binding sites that are shown to play a critical role for basal transcription . Hypoxia and TGF-β cooperate in endoglin promoter activation . This appears to be mediated via a multi-protein complex consisting of Smad3/4, Sp1 and HIF1α/β bound to their cognate DNA binding elements.
Despite the fact that endoglin is considered to be an endothelial specific marker, several other cell types have been shown to express endoglin. For example, endoglin is present on monocytes and upregulated during the monocyte-macrophage transition . Endoglin was also found to have a crucial role in monocyte-mediated vascular repair . In addition, endoglin is expressed in syncytiotrophoblasts of term placenta  and in pre-eclampsia its expression is highly elevated [13, 46]. While endoglin in normal smooth muscle cells is low, its expression is strongly upregulated in vascular smooth muscle cells in atherosclerosis . During development endoglin is expressed in a subset of neural crest stem cells and is required for myogenic differentiation . Moreover, endoglin is expressed in adult bone marrow hematopoietic stem cells (HSCs)  and is a functional marker that defines long-term repopulating hematopoietic stem cells . Using the differentiation of ES cells into embryoid bodies as an assay system, endoglin was shown to be required for efficient myelopoiesis and definitive erythropoiesis . In addition, overexpression or knockdown of endoglin in hematopoietic stem cells revealed that, while endoglin is not required for engraftment and reconstituting capacities, it regulates adult erythroid development .
Furthermore, endoglin expression is also found in certain tumor cells, including primary and metastatic lesions of melanoma  and ovary  and prostate cancer cells . In prostate cancer cells, endoglin suppresses cell adhesion, motility and invasion by activating the TGF-β/ALK2/Smad1 pathway [72, 73], in a manner that is reminiscent of TGF-β/ALK1/Smad1 signaling in endothelial cells . Consistent with the notion that endoglin expression is attenuated during prostate cancer progression, it has been postulated that endoglin has a tumor suppressor role in prostate cancer. Endoglin is also expressed in epidermal keratinocytes . In a multistage model for mouse skin carcinogenesis using wild type and endoglin heterozygous mice, endoglin was shown to have a dual role by inhibiting benign tumor formation, but accelerating the malignant conversion in skin carcinogenesis. Endoglin haplo-insufficieny may elicit these effects by enhancing the tumor suppressing and inhibiting the tumor promoting effects of TGF-β/ALK5 signaling in tumor cells .
Interplay with TGF-β and other signal transduction pathways
Besides the TGF-β dependent interaction of endoglin with TβRII , endoglin can also interact with TβRII and type I receptors ALK1 and ALK5 in a ligand independent manner and both extra- and intracellular domains contribute to this interaction [50, 75]. Endoglin can bind several other ligands besides TGF-β, including activins and BMPs and can interact with activin type II receptors . Interestingly, BMP9 can bind endoglin with high affinity independently from type I or type II receptors (Fig. 2) .
In endothelial cells TGF-β can signal via two distinct type I receptor pathways, i.e., ALK1 and ALK5 . Whereas TGF-β/ALK1 signaling stimulates cell proliferation and migration of endothelial cells, TGF-β/ALK5 signaling inhibits these responses (Fig. 2) . Recent studies have revealed an intricate interplay between the two signaling pathways, and with endoglin. ALK5/Smad2/3 inhibits ALK1-Smad1/5 signaling and vice versa, and ALK5 is required for efficient TGF-β/ALK1 signaling . Forced overexpression of endoglin inhibits TGF-β/ALK5 signaling [78, 79] and TGF-β-induced growth inhibition [42, 61, 80]. Conversely, inhibiting endoglin function by specific knockdown or treatment with neutralizing antibodies inhibits TGF-β/ALK1 signaling, and (in) directly potentiates TGF-β/ALK5 signaling [80–83]. Establishment of endothelial cell lines from endoglin deficient mouse embryos (mouse embryonic endothelial cells, MEECs) demonstrated that endothelial cells deficient in endoglin do not proliferate efficiently, possibly because TGF-β/ALK1 signaling is attenuated and TGF-β/ALK5 signaling is enhanced. Surviving cells were found to have downregulated ALK5 expression, possibly caused by an adaptive response to overcome the increased TGF-β/ALK5 induced growth arrest . Similar results were obtained with primary cultures of outgrown endothelial cells from the blood of HHT1 patients . However, in another study using MEECs derived from endoglin null embryos, it was demonstrated that endoglin is not required for TGF-β-induced Smad1/5 activation and that endoglin may regulate TGF-β receptor affinity .
The intracellular region of endoglin can be phosphorylated by TβRII and type I receptors . TβRII and ALK5 preferentially phosphorylate Ser634 and Ser635, which greatly facilitate subsequent phosphorylation by caALK1 on threonine residues (Fig. 1). The PDZ-binding motif appears to negatively regulate phosphorylation as its removal greatly increases the serine phosphorylation level by all three receptor kinases. The endoglin phosphorylation status influences its subcellular distribution and may modulate endoglin-mediated effects on endothelial cell adhesion and proliferation.
Endoglin can localize to caveolae . Interestingly, caveolin-1, the major protein component in caveolae, was recently shown to interact and functionally cooperate with TGF-β/ALK1 signaling in endothelial cells . Specific knockdown of caveolin-1 or caveolae disruption by cholesterol depletion attenuated ALK1 signaling. This in contrast to TGF-β/ALK5 signaling for which caveolin-1 has an inhibitory effect . Interestingly, cavaolin-1 knock-out mice develop vascular abnormalities with increased endothelial permeability .
Several studies have suggested that endoglin has functions that are independent from TGF-β family signaling (Fig. 1). The endoglin-induced inhibition of apoptosis in endothelial cells subjected to hypoxic stress has been suggested to occur independently of TGF-β . Endoglin has also been shown to induce activation of MAP kinases . Endoglin interacts with zyxin and ZRP-1, LIM domain containing proteins which are concentrated in focal adhesions [90, 91]. Via these interactions endoglin may control cell migration independently of TGF-β. The endoglin cytoplasmic tail interacts with β-arrestin2 in endothelial cells . This interaction results in endoglin internalization in endosomal vesicles and was shown to inhibit TGF-β-induced ERK activation and migration in endothelial cells. Using the cytoplasmic tail as bait in yeast two hybrid interaction screen, Tctex2β, a dynein light chain (DLC) family member, was identified . Tctex2β was also found to interact with TβRII. Ectopic expression of Tctex2β had an inhibitory effect on TGF-β-induced transcriptional reporter activity. Endoglin, TβRII, and Tctex2β colocalize in the plasmamembrane. DLC protein family members are involved in retrograde protein transport, but whether the function of endoglin is regulated and directed by this remains to be investigated.
Endoglin expression is elevated during alterations in vascular structure as they occur during embryogenesis, inflammation, and wound healing [94, 95]. The link between mutations in the endoglin gene and the vascular disorder HHT  also indicates an important role for endoglin in maintaining normal vascular architecture (see discussion below). Mice deficient in endoglin die 10.5 days p.c. and fail to form mature blood vessels in the yolk sac [8–10]. Vessels do form, but are dilated and fragile, and easily rupture. The endoglin deficient embryos are much more fragile and smaller than their wild type litter mates. Most embryos lack vitelline vessels, which connect the yolk sac with the embryo proper. This may also explain the pericardial effusion, indicative of a circulation defect that is observed in most mutant embryos. This phenotype of endoglin deficient embryos is reminiscent of that of TGF-β1, TβRII, ALK5, and ALK1 , and suggests a functional link of endoglin with these other TGF-β receptors during extra-embryonic vascular development. In contrast to the lack of vascular networks in the yolk sac of endoglin deficient mice, the vasculature in the embryo proper was normal with the exception of the heart [8–10]. Endoglin null mice were found to have several cardiac defects; almost all endoglin deficient embryos demonstrated enlarged cardiac ventricles and dilated outflow tracts and the atrioventricular canal endocardium failed to undergo mesenchymal transformation and to form cushions. These defects in heart valve formation in the endoglin knockouts might also be related to perturbation of TGF-β signaling. TGF-β ligands and signaling receptors are also expressed in the developing heart  and defects in heart looping and morphogenesis have been reported in transgenic mice with misexpressed TGF-β receptor .
TGF-β has an important role in vascular morphogenesis, and is involved in endothelial cell function and differentiation of pericytes and smooth muscle cells . Endoglin mutant mice also demonstrate defects in both endothelial and smooth muscle cell function [8–10]. Its predominant expression in endothelial cells suggests that the primary defect is endothelial and that defects in smooth muscle cell differentiation are secondary. Consistent with this notion, endoglin deficient mice have decreased paracrine TGF-β signaling from endothelial cells to neighboring mesothelial cells in the yolk sac . This may lead to defective mesothelial cell differentiation into pericytes or vascular smooth muscle cells. As a result the vessels are weak and susceptible to damage and hemorrhage. However, recent studies also demonstrate that endoglin may have a cell autonomous role in vascular smooth muscle cells. Ectopic expression of endoglin in neural crest cells resulted in pronounced hemorrhage in cranial vessels and in the pericardial cavity due to aberrant smooth muscle cells in the vascular wall .
Mice heterozygotes for endoglin serve as a good model for HHT . They have dilated and fragile blood vessels, telangiectases, and nosebleeds, which are the clinical manifestations that are also observed in HHT patients. Some strains of mice are more affected than others, suggesting the existence of modifier genes. In addition, shear, blood pressure, and inflammation products may contribute to disease heterogeneity. This has also been suggested to be the case for HHT patients . Endoglin heterozygous mice have reduced eNOS expression and NO synthesis-dependent vasodilation is impaired [86, 101]. Endothelial cells derived from endoglin heterozygous mice showed a significantly attenuated proliferation, migration, increased collagen production, and mitigated NO synthase expression and VEGF secretion. Compared to wild type animals, mice heterozygous for endoglin showed impaired angiogenesis as observed by a delayed reperfusion following hindlimb ischemia . In addition, when matrigel plugs were implanted in endoglin heterozygote mice to measure endothelial cell outgrowth and invasion into the extracellular matrix in vivo, it was found that they had significantly less vascular structures compared to wild type mice .
Hereditary hemorrhagic telangiectasia (HHT)
HHT, also termed Osler–Weber–Rendu disease, is an autosomal dominant vascular disorder with an incidence of about 1 in 10,000 [100, 102]. Characteristic clinical features include small dilated bloodvessels (telangiectases) and arteriovenous malformations (AVMs) in the vasculature of lung, liver, and brain. The disease manifests itself often during puberty with spontaneous recurrent nose bleeds. Later in life the intensity of the nosebleeds increase and also AVMs become larger and problematic. However, onset and clinical manifestations of HHT are heterogeneous between different individuals and even within families [100, 102]. Two HHT variants, i.e., HHT1 and HHT2, have been described that have been linked to mutations in endoglin and ALK1, respectively. HHT1 differs from HHT2 in an earlier onset, stronger prevalence and higher frequency of pulmonary AVMs. In HHT2 the gastrointestinal bleeding and liver involvement appears to occur frequently than in HHT1 families [100, 102]. Recently, another gene in the TGF-β signaling pathway has been implicated in HHT; Smad4 is mutated in a subset of HHT patients with juvenile polyposis, lacking mutations in endoglin and ALK1 . This syndrome is now termed JP-HHT.
Mutations in endoglin include deletions, insertions and missense mutations, and splice site changes, and in about 80% of the cases lead to premature stop codons and truncated endoglin proteins [100, 102]. It has been suggested that these mutant proteins may achieve a dominant negative effect by disrupting normal endoglin function. However, mutated endoglin proteins were found to be expressed at low levels, mediated by nonsense mediated mRNA decay. In addition, mutated proteins may be misfolded and not stable and/or do not reach the cell surface and are thereby unable to form heterodimers with normal endoglin. Indeed, patients with endoglin mutations have significantly lower endoglin levels in peripheral blood monocytes compared to control group . Haplo-insufficiency, therefore, likely provides the explanation for the mechanism underlying HHT1. The lower levels of endoglin expression leads to a dysfunction in TGF-β signal transduction. Whether other non-TGF-β signaling pathways are affected as well by lowered endoglin expression is not clear.
As already mentioned before, studies in mouse models suggest that defective paracrine TGF-β signaling of endothelial cells to adjacent pericytes/smooth muscle cells lead to an inefficient maturation of these cells . This makes the vessels highly susceptible to rupture, a main characteristic of HHT. Interestingly, HHT patients with endoglin mutations have been found to have low circulating levels of TGF-β . Several mechanisms have been suggested through which the AVMs develop. During development there may be a loss of arterial and venous identity whereby the normal separation between veins and arteries are disrupted. In addition, defective vascular remodeling and dilatation may occur following local inflammation. Moreover, cells that form the capillary endothelial bed that separates the arteries and veins may be removed by apoptosis in response to hypoxic stress .
Pre-eclampsia is characterized by the onset of high blood pressure and significant amounts of protein in the urine in the third trimester of pregnancy . It affects both the fetus and the mother and occurs in about 5% of the pregnancies. Severe pre-eclampsia leads to appearance of Hemolysis Elevated Liver enzymes and Low Platelets (HELLP) syndrome, seizures and/or fetal growth restriction, and can result in death [105, 106]. The endothelial dysfunction in pre-eclampsia is thought to be caused by circulating factors that are released from the placenta [107, 108]. Circulating levels of placenta-derived soluble form of vascular endothelial growth factor receptor (VEGFR)1 (also known as sFlt), which sequesters the pro-angiogenic proteins placenta growth factor (PlGF) and VEGF, are increased before onset and correlate with the severity of pre-eclampsia . Interestingly, a 65 kDa sEnd is increased in sera of pregnant women, and strongly elevated in pre-eclamptic patients. SEnd levels also correlate with disease severity . Whereas forced expression of soluble endoglin increased vascular permeability and induced a modest increase in blood pressure without significant proteinuria, ectopic expression of both sFlt and sEnd in pregnant rats induced the hallmarks of severe pre-eclampsia .
In vitro studies on endothelial cell lines showed that sEnd interferes with TGF-β signaling and eNOS activation and thereby causes endothelial dysfunction . The soluble form of endoglin functions likely as a scavenger of specific circulating TGF-β family ligands, such as TGF-β or BMP9 [13, 14]. sEnd and sFlt act differently and may therefore cooperate to induce the endothelial cell dysfunction . As sEnd levels increase 2–3 months before the onset of pre-eclampsia, sEnd (and sFlt and PlGF) levels may be used as diagnostic marker to prioritize patients and thereby prevent pre-eclampsia-induced death . Inhibition of the putative protease involved in endoglin shedding may be of therapeutic benefit in the treatment of pre-eclampsia.
Whereas in pre-eclampsia there is an increase in sEnd, patients suffering from an acute myocardial infarction were found to have a significantly lower level of sEnd in their serum . Interestingly, patients that died had significantly lower levels of sEnd than those that survived. Low levels of sEnd may therefore be used as a prognostic marker after acute myocardial infarction.
Anti-angiogenic activity of endoglin antibodies
Tumors depend on angiogenesis to grow beyond a few cubic millimeters; the exchange of oxygen and nutrients with carbon dioxide and waste products cannot rely any longer on diffusion, but require blood vessels. Also metastasis, the spreading and colonization of the primary tumor to distant organs, requires tumor angiogenesis . Anti-angiogenic therapy, in which proliferating endothelial cells within and surrounding tumor associated vessels are selectively targeted, has been shown (in combination with chemotherapy) to induce tumor regression and inhibit metastasis .
Endoglin is expressed in tumor associated endothelium in many solid cancers, including breast, prostate, and cervical cancer [113–116]. Intra-tumor microvessel density as determined by anti-endoglin staining and circulating levels of soluble endoglin have prognostic significance in cancer. Antibodies that specifically detect endoglin have successfully been used for tumor imaging and endoglin antibodies coupled with toxins  or radioactivity  have been used with favorable outcome for vascular targeting [119, 120]. Monoclonal antibodies to endoglin can inhibit proliferation of endothelial cells  and have been shown to have anti-angiogenic effects . Moreover, immunotherapy with the extracellular domain of porcine endoglin was found to induce cytotoxic T lymphocyte (CTL)-mediated cytotoxicity and inhibits both endothelial cell proliferation and tumor growth in mouse cancer models [123, 124]. However, the suitability of endoglin as a target for anti-angiogenic therapy for cancer has been questioned; endoglin expression has also been observed in the endothelial cells of normal tissues [125, 126]. Differences may prevail between different antibodies, and testing of different endoglin antibodies to select the best one for antibody-based therapeutic approaches has been proposed [119, 126].
The anti-angiogenic effects of sEnd may be used to interfere with angiogenically active tumors. Besides endoglin, also the signaling TGF-β receptors can be targeted to achieve anti-angiogenic effects. In this respect, it is of interest to note that treatment of mice with a small molecule ALK5 kinase inhibitor decreased the pericyte coverage of the endothelial vessels, and in particular those ones present in the tumor neovasculature. This activity was used to stimulate a greater accumulation of anticancer nanocarriers in tumors .
Endoglin plays a critical role in angiogenesis and dysregulation of its expression and/or activity has been implicated in multiple vascular diseases, most notably HHT, pre-eclampsia and tumor angiogenesis. While much needs to be investigated on the molecular mechanisms that underlie its role in vascular development and disease, the recent advances provide ample opportunities for better diagnosis and development of new therapies for diseases with vascular abnormalities.
Our studies on endoglin are supported by Dutch Cancer Society (UL-2005-3371) and EU grants Angiotargeting and Tumor-Host-Genomics.