Biochemical Society Transactions

Advances in the Cellular and Molecular Biology of Angiogenesis

Angiogenesis regulation by TGFβ signalling: clues from an inherited vascular disease

Marwa Mahmoud , Paul D. Upton , Helen M. Arthur


Studies of rare genetic diseases frequently reveal genes that are fundamental to life, and the familial vascular disorder HHT (hereditary haemorrhagic telangiectasia) is no exception. The majority of HHT patients are heterozygous for mutations in either the ENG (endoglin) or the ACVRL1 (activin receptor-like kinase 1) gene. Both genes are essential for angiogenesis during development and mice that are homozygous for mutations in Eng or Acvrl1 die in mid-gestation from vascular defects. Recent development of conditional mouse models in which the Eng or Acvrl1 gene can be depleted in later life have confirmed the importance of both genes in angiogenesis and in the maintenance of a normal vasculature. Endoglin protein is a co-receptor and ACVRL1 is a signalling receptor, both of which are expressed primarily in endothelial cells to regulate TGFβ (transforming growth factor β) signalling in the cardiovasculature. The role of ACVRL1 and endoglin in TGFβ signalling during angiogenesis is now becoming clearer as interactions between these receptors and additional ligands of the TGFβ superfamily, as well as synergistic relationships with other signalling pathways, are being uncovered. The present review aims to place these recent findings into the context of a better understanding of HHT and to summarize recent evidence that confirms the importance of endoglin and ACVRL1 in maintaining normal cardiovascular health.

  • activin receptor-like kinase 1 (ACVRL1)
  • angiogenesis
  • endoglin
  • hereditary haemorrhagic telangiectasia (HHT)
  • transforming growth factor β (TGFβ)
  • vascular disease

Hereditary haemorrhagic telangiectasia

HHT (hereditary haemorrhagic telangiectasia) is an inherited vascular disorder associated with altered TGFβ (transforming growth factor β) signalling in endothelial cells. It is inherited in an autosomal dominant fashion and affects approximately 1 in 5000 individuals. Clinically, HHT is characterized by multiple red spots known as telangiectases that develop on the skin, in the mucocutaneous lining of the oronasal and GI (gastrointestinal) tract, and are prone to severe haemorrhage [1]. The telangiectases form due to local vascular abnormalities: the emergence of direct connections between a small artery and vein and associated loss of a capillary bed [2,3]. Larger AVMs (arteriovenous malformations), forming as a result of abnormal connections between more substantial arteries and veins, may occur in the lung, brain and liver of HHT patients, causing major clinical problems [1].

The majority of HHT patients carry mutations in one of two genes: ENG (endoglin), a co-receptor for ligands of the TGFβ superfamily [4], or ACVRL1 (activin receptor-like kinase 1, also known as ALK1), a type I TGFβ superfamily receptor [5]. Mutations in other (as yet unidentified) genes have also been mapped in other HHT families [1], and a rare combined syndrome of juvenile polyposis and HHT is due to mutations in SMAD4 [6].

Endoglin and ACVRL1 protein structures

Endoglin and ACVRL1 are transmembrane proteins, found primarily on the surface of endothelial cells, that play an important role in TGFβ superfamily signalling [7]. The endoglin protein has a large highly glycosylated EC (extracellular) domain, containing a ZP (zona pellucida) domain that, together with the orphan domain, forms a dome-like structure on the cell surface [8] (Figure 1). The ZP domain is also a feature of the related accessory protein betaglycan, suggesting an important role for this region in TGFβ signalling [9]. The IC (intracellular) domain of endoglin terminates with a PDZ domain-binding motif that interacts with the scaffold protein GIPC [GAIP (Gα-interacting protein)-interacting protein C-terminus, also known as synectin] to modulate downstream signalling and regulate endothelial cell migration [10]. The IC domain of endoglin is constitutively phosphorylated on particular serine residues and, to a minor extent, on threonine residues [11]. The type II TGFβ receptor, TGFBR2, preferentially phosphorylates endoglin on Ser634 and Ser635, which subsequently permits ACVRL1 to phosphorylate endoglin on three threonine residues, Thr640, Thr647 and Thr654 [12]. In addition, ALK5 (activin receptor-like kinase 5), the type I TGFβ receptor (also known as TGFBR1) phosphorylates endoglin on Ser646 and Ser649 [13] (Figure 1). The precise role of phosphorylation of the IC domain of endoglin is not understood, but it is known to affect TGFβ family signalling and endothelial cell migration, and may also affect interactions between endoglin and cytoskeletal or signalling proteins. The IC domain of endoglin interacts with the focal adhesion proteins zyxin and zyxin-related protein to remodel the actin cytoskeleton [14,15]. There may also be effects on the microtubule machinery as a result of endoglin interacting with the dynein-related protein Tctex2β [16]. Furthermore, the IC domain of endoglin interacts with β-arrestin2 in a manner that depends on Thr650 of endoglin and leads to increased endoglin internalization and altered ERK (extracellular-signal-regulated kinase) signalling [17].

Figure 1 ACVRL1 and endoglin proteins

ACVRL1 (A) and endoglin (B and C) are transmembrane receptors composed of 503 and 658 amino acids respectively. Their protein structures (A and B) can be divided into four distinct regions: the signal peptide (SP), which is removed during processing to generate the mature protein; the EC domain, important for ligand binding; and the TM and IC domains. The IC domain of ACVRL1 contains the GS domain at the juxtamembrane position and the kinase responsible for phosphorylation of SMAD1/5/8 (as well as selected threonine residues in endoglin). Endoglin has a large glycosylated EC domain and a short non-signalling IC domain with a PDZ-binding motif (SSMA) at its C-terminus. Alternative splicing generates two different IC domains for endoglin, L-Eng and S-Eng, and various serine and threonine residues are phosphorylated in L-Eng by receptor kinases (see the text). (C) The three-dimensional structure of the EC domain of endoglin comprises the N- and C-terminal regions of the ZP domain (ZP-N and ZP-C), and the third is an orphan domain of no known homology [8]. Endoglin normally forms a disulfide-linked homodimer at the cell surface and the dome consists of two antiparallel oriented monomers, but only one monomer is depicted for simplicity.

Two protein isoforms of human endoglin have been characterized. L (large)-endoglin is the predominant isoform with a cytoplasmic domain of 47 residues. In contrast, S (short)-endoglin, the minor isoform, has a cytoplasmic domain of only 14 residues and arises by alternative splicing [18]. These isoforms have opposing roles in angiogenesis, with S-endoglin having anti-angiogenic effects [19], whereas L-endoglin is pro-angiogenic [20]. The increased S/L ratio of endoglin isoforms in senescent human endothelial cells is regulated by the ASF (alternative splicing factor)/SF2 (splicing factor-2) and is consistent with an anti-angiogenic role for S-endoglin [21,22].

Endoglin protein is also expressed in the placental syncytiotrophoblasts and can be shed from the cell surface to generate Sol-Eng (soluble endoglin). Increased circulating levels of Sol-Eng are associated with pre-eclampsia [23], and Sol-Eng also inhibits angiogenesis [24]. The regulatory mechanisms involved in endoglin shedding are not yet fully understood, but exposure of endothelial cells to the inflammatory cytokine TNFα (tumour necrosis factor α) leads to increased shedding of endoglin protein in vitro and this may be through increased expression of MMP14 (matrix metalloproteinase 14) which promotes cleavage of endoglin protein at residues 586–587 [24].

The ACVRL1 receptor, similar to other type I receptors of the TGFβ family, is composed of a cysteine-rich EC domain, a single TM (transmembrane) region and an IC domain containing a serine/threonine kinase as well as a highly conserved GS (glycine/serine-rich) domain that plays an important regulatory role in kinase activity (Figure 1). Essentially, phosphorylation of the GS domain of a type I receptor such as ACVRL1 by a TGFβ family type II receptor is required to activate its kinase signalling activity [25]. There are various splice variants for human ACVRL1 in the Ensembl database, but little is known about the function of the different isoforms.

TGFβ family signalling: the role of endoglin and ACVRL1

TGFβ superfamily ligands mediate their effects through heteromeric complexes of type I and type II receptors (Figure 2A). Following ligand binding and receptor activation, the TGFβ type I receptor phosphorylates specific R (receptor-regulated)-SMADs that can then interact with the Co (common-mediator)-SMAD4 and translocate to the nucleus to regulate transcription of specific target genes, usually in complex with other transcription factors [26,27] (Figure 2). Another class of SMAD proteins, the I (inhibitory)-SMADs (SMAD6 and SMAD7), inhibit activated R-SMADs by competing with them for receptor interaction, promoting the proteosomal degradation of the activated type I receptor or by recruiting phosphatases to dephosphorylate the activated type I receptor [28].

Figure 2 TGFβ family signalling: basic canonical pathway

(A) TGFβ superfamily ligands mediate their effects by binding to specific, constitutively phosphorylated, type II receptors. This induces the recruitment and phosphorylation of specific type I receptors which leads to phosphorylation of specific R-SMAD signal transduction proteins. Activated R-SMADs form a complex with the Co-SMAD SMAD4, and this SMAD complex translocates to the nucleus where it binds DNA and regulates the transcription of specific target genes. Endoglin and β-glycan co-receptors (also called type III receptors) can be recruited to the TGFβ receptor complex and modulate signalling. (B) Translocation of SMAD2 protein to the nucleus of endothelial cells over the 60 min following treatment with 5 ng/ml TGFβ1 ligand. Images courtesy of Leon Jonker (Newcastle University).

In endothelial cells, TGFβ1 can propagate signalling via two distinct type I receptors with opposing effects on angiogenesis [29] (Figure 3). TGFβ1 predominantly signals via the type I receptor ALK5, leading to activation of SMAD2/3 and transcription of target genes such as PAI-1 (plasminogen-activator inhibitor 1) associated with endothelial cell quiescence. TGFβ1 can also signal via ACVRL1, which results in the activation of R-SMAD1/5/8 and the transcription of target genes such as ID1 (inhibitor of DNA binding 1) associated with endothelial cell proliferation and angiogenesis. ALK5 mediates the TGFβ-dependent recruitment of ACVRL1 to the receptor complex and is required for optimal ACVRL1 activation [30]. Endoglin recruitment to the TGFBR2–ACVRL1–ALK5 receptor complex promotes signalling via the SMAD1/5/8 pathway and this indirectly inhibits SMAD2/3 signalling via the ALK5/SMAD2/3 pathway [30,31].

Figure 3 Role of endoglin and ACVRL1 in modulating TGFβ family signalling in endothelial cells

TGFβ superfamily signalling depends on ligand context and the composition of the receptor complex. (A) TGFβ1 can activate two distinct signalling pathways in endothelial cells. Low levels of circulating TGFβ1 ligand promote signalling through the ACVRL1–ALK5–TGFBR2 receptor complex and activation of SMAD1/5/8 which leads to the transcription of pro-angiogenic target genes. On the other hand, high levels of TGFβ1 ligand induce signalling through the ALK5–TGFBR2 receptor complex and activation of SMAD2/3, which leads to the transcription of anti-angiogenic target genes. Endoglin promotes activation of SMAD1/5/8 and indirectly inhibits activation of SMAD2/3. Note that TGFβ1 may also exert anti-angiogenic effects on the underlying mural cells via ALK5–TGFBR2 and R-SMAD2/3 signalling. (B) BMP9 ligand can also activate two distinct signalling pathways in endothelial cells via two distinct type II receptors. BMP9 signalling via the ACVRL1–BMPR2 and ACVRL1–ACTRII receptor complexes both activate R-SMAD1/5/8, with loss of one type II receptor compensated for by the other. This response promotes the transcription of target genes required for maintaining endothelial cell quiescence. BMP9 signalling via the ACVRL1–ACTRII receptor complex also activates SMAD2, with BMPR2 contributing to this response. The opposing roles of ID1 as pro- or anti-angiogenic (compare A and B) is likely to depend on the longevity of the ID1 signal and how it is integrated with other transcriptional mediators. ECM, extracellular matrix; IL8, interleukin 8; PAI-1, plasminogen-activator inhibitor 1.

ACVRL1 can also bind to BMP9 (bone morphogenetic protein 9) with high affinity in association with BMPR2 (type II BMP receptor) or ACTRII (activin type II receptor) to activate SMAD1/5/8 [32,33] (Figure 3). Unlike TGFβ1 signalling via ACVRL1, which promotes endothelial cell proliferation and angiogenesis [29,31], BMP9 signalling via the BMPR2–ACVRL1 receptor complex inhibits FGF (fibroblast growth factor)-induced cell proliferation, VEGF (vascular endothelial growth factor)-induced angiogenesis and maintains endothelial cell quiescence [32,34]. Conversely, BMP9 can also act in a pro-angiogenic capacity in combination with TGFβ1, as both ligands act synergistically to improve the endothelial cell response to VEGF [35]. BMP9 can also activate signalling via ACVRL1 in vascular endothelial cells independently of ALK5 and endoglin [36] (Figure 3), leading to the phosphorylation of SMAD1/5/8, induction of E-selectin and IL-8 (interleukin 8) expression and the promotion of endothelial cell quiescence. These findings point to a dual role for ACVRL1 in promoting or inhibiting endothelial cell activation dependent on cytokine context and the composition of the signalling receptor complex.

Although not a core receptor for TGFβ ligands, endoglin can indirectly modulate signalling by refining the selectivity of ligand binding [37]. Endoglin binds to TGFβ isoforms 1 and 3 in the presence of TGFBR2, and to BMP9 and BMP10 in a receptor-independent manner, and, in both cases, promotes signalling through ACVRL1 [31,32,38]. Circulating Sol-Eng, on the other hand, can sequester ligands of the TGFβ superfamily and thereby has the potential to reduce ligand availability at the endothelial cell surface [23,39]. Thus endoglin can modulate TGFβ family signalling at a number of different levels.

Non-canonical TGFβ signalling

In addition to canonical TGFβ signalling via the SMAD proteins, TGFβ can activate several non-canonical signalling pathways such as the ERK, MAPK (mitogen-activated protein kinase), Rho-like GTPase and PI3K (phosphoinositide 3-kinase)/Akt pathways [25]. The extent to which endoglin and ACVRL1 play a role in these non-canonical pathways is not clear. However, the highly similar defects reported for the ACVRL1, endoglin and TAK1 (TGFβ-activated kinase 1) constitutive knockout mice indicates possible cross-talk between these signalling proteins and involvement of the MAPK pathway [4042]. Constitutively active Acvrl1 inhibited activation of JNK (c-Jun N-terminal kinase) and ERK in microvascular endothelial cells [43], whereas the interaction of endoglin with β-arrestin2 discussed above [17] is enhanced by ACVRL1 and results in the internalization and accumulation of endoglin and β-arrestin2 in endocytic vesicles. This appears to have no effect on SMAD activation, but down-regulates ERK activation and alters the subcellular distribution of activated ERK in endothelial cells.

Expression of endoglin and ACVRL1

ACVRL1 and endoglin are both expressed in the developing vasculature of the mouse embryo [44,45]. Endoglin appears to have a broader expression pattern in the developing endocardium and cardiac cushions of the heart, as well as all blood vessel types, whereas ACVRL1 expression appears to be more restricted to the arteries. Both are expressed at lower levels in the lymphatic endothelial cells, and recent evidence suggests ACVRL1 is important in lymphatic development and remodelling [46]. In the adult, ACVRL1 exhibits preferential expression in the blood vessels of the lung, where it is predominantly expressed in the pulmonary capillaries and pre-capillary arterioles [47]. Endoglin is also highly expressed in pulmonary blood vessels in the adult mouse and has been shown to co-localize with ACVRL1 in the distal regions of the pulmonary vasculature, where associated activation of SMAD1/5/8 is observed [47]. The expression of both the Eng and Acvrl1 genes is up-regulated during periods of active angiogenesis during dermal wound healing, heart repair and tumour angiogenesis [4851], where hypoxia, which leads to up-regulation of endoglin, is likely to be an important trigger [52].

In addition to their expression in vascular endothelium, endoglin and ACVRL1 are also found in other cell types. Endoglin is expressed in bone marrow with a reported role in erythropoiesis; in myofibroblasts with effects on fibrosis; in macrophages with a likely role in inflammation; and in mesenchymal cells as a marker of mesenchymal stem cells [53]. ACVRL1 has been reported in hepatic stellate cells [54] and in chondrocytes [55] where it is thought to regulate the balance of ALK5 and ACVRL1 signalling.

Role of endoglin and ACVRL1 in vascular development

Mice that are homozygous for loss-of-function mutations in Acvrl1 or Eng develop severe vascular abnormalities and die at mid-gestation, consistent with a critical role for both genes during angiogenesis [40,41,56,57]. In addition, Eng-null embryos exhibit heart defects that may be a consequence of reduced endothelial–mesenchymal transition of the endocardial cells overlying the cardiac cushions [41,58]. Examination of the vascular defects in the yolk sac of Eng-null embryos reveals reduced TGFβ signalling ‘cross-talk’ between endothelial cells and adjacent peri-endothelial support cells, leading to a failure in muscle maturation [59]. This reduced muscularization is also seen in the vessels of Eng-heterozygous mice [60], and can be rescued by thalidomide treatment, which increases the expression of PDGFβ (platelet-derived growth factor β) by endothelial cells, potentially compensating for the reduced TGFβ signalling [61]. Importantly, thalidomide treatment has been used successfully in a small group of HHT patients to reduce bleeding, confirming the validity of these findings [61].

HHT exhibits an age-related penetrance, suggesting that haploinsufficiency for the ACVRL1 or ENG alleles is not sufficient to drive disease, and that a further somatic insult affecting the intact allele is required. In support of this, AVMs that are typical of HHT have been successfully modelled in mice, but a conditional knockout approach is required to remove both copies of the target gene. For example, loss of ACVRL1 in endothelial cells of the brain and lung leads to multiple arteriovenous fistulas and haemorrhage in pups, causing lethality at 5 days of age. Also, removing ACVRL1 from adult mice leads to a rapid development of multiple AVMs and associated micro-haemorrhage in lung and GI tract, whereas AVM formation in the skin requires a wounding stimulus, suggesting that an angiogenic trigger is required [62]. Furthermore, local depletion of Acvrl1 in the adult mouse brain also leads to severe AVMs in the presence of an angiogenic stimulus [63]. In zebrafish, loss of ACVRL1 lead to AVMs that are dependent on blood flow [64].

Similarly, loss of endothelial endoglin in the adult mouse results in the formation of AVMs. Using the neonatal retinal vascular plexus as a model of angiogenesis, AVMs appear to develop by gradual vessel enlargement over several days; they retain a venous identity and are associated with increased endothelial cell proliferation [65]. These studies all suggest that pathological or developmental angiogenesis is required to trigger the formation of AVMs in the absence of endoglin or ACVRL1, and that blood flow may be required to maintain an AVM. Thus the heterogeneity seen in HHT patients may be due to somatic loss of the second allele in combination with an angiogenic insult such as may occur during inflammation [65].

Endoglin, ACVRL1 and disease therapies

Some small-scale drug trials have been completed in HHT patients with positive preliminary outcomes. Thalidomide treatment promotes vascular stability and reduced bleeding in a small group of HHT patients [61]. Preliminary studies also suggest that anti-VEGF targeting is successful in treating HHT patients, probably due to its anti-angiogenic effects. On a broader scale, ACVRL1 and endoglin are important for angiogenesis in a wide range of disorders that depend on new blood vessel growth. For example, drugs that inhibit tumour angiogenesis are currently being used to treat cancer patients, and endoglin and ACVRL1 both represent new and potentially valuable anti-angiogenic cancer targets. Even haploinsufficient levels of endoglin or ACVRL1 lead to reduced angiogenesis in preclinical mouse models [35,66]. Anti-endoglin therapy is being tested in clinical trials [67], whereas anti-ACVRL1 therapy also appears to be a promising approach in anti-angiogenic and anti-lymphatic tumour therapy [35,46].

Conclusions and future perspectives

Studies of HHT have revealed that endoglin and ACVRL1 are essential for cardiovascular development and play key roles in angiogenesis. Work using mouse models has shown that loss of either gene in adult life leads to major vascular pathologies in the context of an angiogenic insult, and these mice are proving to be valuable models of HHT. In addition, spontaneous AVMs, the most frequent cause of cerebral haemorrhage in the young, may relate to abnormal function of ENG or ACVRL1 genes. In the context of general cardiovascular health, circulating Sol-Eng is not only an important biomarker of pre-eclampsia, but also a predictor of cardiovascular events in chronic coronary artery disease patients [68]. It is almost certain that further fundamental roles of endoglin and ACVRL1 in the cardiovasculature will be revealed in the near future. For example, a recent genome-wide association study has revealed an association between an endoglin haplotype and bicuspid aortic valve, one of the most common cardiac malformations [69]. For HHT patients, the challenge for the next decade is to better understand the molecular basis of the disease and to develop more effective therapies. Fortunately, we now have excellent mouse models with which to make rapid progress in these important fields of research.


M.M. is supported by the Wellcome Trust; P.D.U. and H.M.A. are supported by the British Heart Foundation.


  • Advances in the Cellular and Molecular Biology of Angiogenesis: A Biochemical Society/Wellcome Trust Focused Meeting held at the University of Birmingham, U.K., 27–29 June 2011. Organized and Edited by Stuart Egginton and Roy Bicknell (Birmingham, U.K.).

Abbreviations: ACTRII, activin type II receptor; ACVRL1, activin receptor-like kinase 1; ALK5, activin receptor-like kinase 5; AVM, arteriovenous malformation; BMP, bone morphogenetic protein; BMPR2, type II BMP receptor; Co-SMAD, common-mediator SMAD; EC, extracellular; ENG, endoglin; ERK, extracellular-signal-regulated kinase; GI, gastrointestinal; GS, glycine/serine-rich; HHT, hereditary haemorrhagic telangiectasia; IC, intracellular; L-endoglin, large endoglin; ID1, inhibitor of DNA binding 1; MAPK, mitogen-activated protein kinase; R-SMAD, receptor-regulated SMAD; S-endoglin, short endoglin; Sol-Eng, soluble endoglin; TGFβ, transforming growth factor β; TGFBR, TGFβ receptor; TM, transmembrane; VEGF, vascular endothelial growth factor; ZP, zona pellucida


View Abstract