RESEARCH ARTICLE

Soluble endoglin regulates expression of angiogenesis-relatedproteins and induction of arteriovenous malformations in a mousemodel of hereditary hemorrhagic telangiectasiaEunate Gallardo-Vara1, Simon Tual-Chalot2, Luisa M. Botella1, Helen M. Arthur2,* and Carmelo Bernabeu1,*,‡

ABSTRACTEndoglin is a transmembrane glycoprotein expressed in vascularendothelium that plays a key role in angiogenesis. Mutations in theendoglin gene (ENG) cause hereditary hemorrhagic telangiectasiatype 1 (HHT1), characterized by arteriovenous malformations(AVMs) in different organs. These vascular lesions derive fromabnormal processes of angiogenesis, whereby aberrant vascularremodeling leads to focal loss of capillaries. Current treatments forHHT1 include antiangiogenic therapies. Interestingly, a circulatingform of endoglin (also known as soluble endoglin, sEng),proteolytically released from the membrane-bound protein anddisplaying antiangiogenic activity, has been described in severalendothelial-related pathological conditions. Using human andmouse endothelial cells, we find that sEng downregulates severalpro-angiogenic and pro-migratory proteins involved inangiogenesis. However, this effect is much reduced in endothelialcells that lack endogenous transmembrane endoglin, suggestingthat the antiangiogenic activity of sEng is dependent on thepresence of endogenous transmembrane endoglin protein. In fact,sEng partially restores the phenotype of endoglin-silencedendothelial cells to that of normal endothelial cells. Moreover,using an established neonatal retinal model of HHT1 with depletedendoglin in the vascular endothelium, sEng treatment decreases thenumber of AVMs and has a normalizing effect on the vascularphenotype with respect to vessel branching, vascular density andmigration of the vascular plexus towards the retinal periphery. Takentogether, these data show that circulating sEng can influencevascular development and AVMs by modulating angiogenesis, andthat its effect on endothelial cells depends on the expression ofendogenous endoglin.

This article has an associated First Person interview with the firstauthor of the paper.

KEY WORDS: Angiogenesis, Endoglin, HHT, AVM, TGF-β,Endothelial cells

INTRODUCTIONEndoglin is a homodimeric transmembrane glycoprotein that actsas an auxiliary receptor for members of the transforming growthfactor-β (TGF-β) family of cytokines. It is expressed primarily invascular endothelium, but also in mesenchymal cells, and plays akey role in vascular physiology, including angiogenesis andvascular remodeling (ten Dijke and Arthur, 2007; López-Novoaand Bernabeu, 2010; Núñez-Gómez et al., 2017; Redgrave et al.,2017; Ruiz-Llorente et al., 2017). In humans, there are two differentprotein isoforms, the predominantly expressed L-endoglin, andthe minor isoform S-endoglin, generated by alternative splicing(Gougos and Letarte, 1990; Bellón et al., 1993; Blanco andBernabeu, 2011). Both endoglin isoforms are identical in theirextracellular and transmembrane regions, but they differ from eachother in their cytoplasmic domain (Bellón et al., 1993). In additionto these two membrane-bound proteins, a circulating form ofendoglin, originally named as soluble endoglin (sEng) (Venkateshaet al., 2006; Gregory et al., 2014), containing the extracellularregion has been described. Circulating sEng is shed frommembrane-bound endoglin by the proteolytic activity of thematrix metalloprotease 14 (MMP14 or MT1) (Hawinkels et al.,2010; Valbuena-Díez et al., 2012) and can be released from theplacenta in exosomes enriched with certain sphingomyelin species,upon clustering with MMP14 (Ermini et al., 2017). However, thespecific nature of the circulating sEng, whether as an individualsoluble protein or complexed within exosomes, is still not fullyunderstood. Shedding of sEng can be triggered by inflammation,tumor necrosis factor-α (TNF-α), endothelial injury or anti-endoglin antibodies (Li et al., 2003; Sunderland et al., 2011;Kumar et al., 2014; Gallardo-Vara et al., 2016). High levels of sEnghave been reported in several endothelium-related pathologicalconditions (Bernabeu et al., 2009; Blázquez-Medela et al., 2010;Oujo et al., 2013; Gregory et al., 2014). Among these, markedlyelevated levels of sEng are found in pre-eclampsia, a disease of highincidence in pregnant women which, if left untreated, can lead to thedeath of mother and baby (Venkatesha et al., 2006). Pre-eclampsiais characterized by hypertension and proteinuria associated withendothelial dysfunction. Several lines of evidence supporta pathogenic role of sEng in pre-eclampsia, includingantiangiogenic activity, increased vascular permeability andhypertension (Venkatesha et al., 2006; Hawinkels et al., 2010;Valbuena-Díez et al., 2012). Also, sEng has pro-inflammatoryactivity via nuclear factor kappa-light-chain-enhancer of activated Bcells (NFκB) and interleukin-6 (IL6) (Varejckova et al., 2017), andcan modulate inflammation-associated leukocyte adhesion andtransmigration (Rossi et al., 2013). Studies in transgenic animalsoverexpressing human sEng suggest that sEng also contributes toendothelial dysfunction (Jezkova et al., 2016; Rathouska et al.,2017). Despite its critical importance in vascular pathology, theReceived 28 February 2018; Accepted 29 July 2018

1Centro de Investigaciones Biológicas, Consejo Superior de InvestigacionesCientıf́icas (CSIC), and Centro de Investigación Biomédica en Red deEnfermedades Raras (CIBERER), 28040 Madrid, Spain. 2Institute of GeneticMedicine, Centre for Life, Newcastle University, Newcastle NE1 3BZ, UK.*These authors contributed equally to this work

‡Author for correspondence (bernabeu.c@cib.csic.es)

C.B., 0000-0002-1563-6162

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1

© 2018. Published by The Company of Biologists Ltd | Disease Models & Mechanisms (2018) 11, dmm034397. doi:10.1242/dmm.034397

Disea

seModels&Mechan

isms

http://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalmailto:bernabeu.c@cib.csic.eshttp://orcid.org/0000-0002-1563-6162http://creativecommons.org/licenses/by/3.0http://creativecommons.org/licenses/by/3.0

molecular mechanism of action of sEng remains poorly understood.It has been postulated that sEng activity is based on its capacityto antagonize the function of membrane-bound endoglin. Forexample, sEng binds to bone morphogenetic protein 9 (BMP9 orGDF2), a member of the TGF-β family, with high affinity(Castonguay et al., 2011; Alt et al., 2012; Saito et al., 2017) andthis can lead to sequestration of BMP9, preventing its binding tosurface endoglin and the subsequent downstream intracellularsignaling of the TGF-β receptor complex (Venkatesha et al., 2006;Hawinkels et al., 2010; Gregory et al., 2014). Also, both membrane-bound endoglin and sEng contain an accessible arginine-glycine-aspartic acid (RGD) sequence, which is a consensus binding motiffor integrin recognition (Gougos and Letarte, 1990; Saito et al.,2017), and it has been shown that sEng can inhibit integrin-mediated cell adhesion to endothelial endoglin, likely by competingwith its binding to integrins (Rossi et al., 2013, 2016). Nevertheless,the exact molecular mechanisms of action of sEng remain to beelucidated.Endoglin plays a key role in endothelial cells (ECs), as shown by

numerous in vitro and in vivo studies. Roles include regulating cellproliferation and migration, actin cytoskeleton, endothelial nitricoxide synthase (eNOS) expression and activity, endothelial cellpermeability, leukocyte extravasation, vessel maturation, arterialand venous specification, and vessel diameter in response to flow(López-Novoa and Bernabeu, 2010; Núñez-Gómez et al., 2017;Rossi et al., 2013, 2016; Mahmoud et al., 2010; Jerkic and Letarte,2015; Baeyens et al., 2016; Sugden et al., 2017; Jin et al., 2017). Inaddition, mutations in the human endoglin gene (ENG) are theunderlying cause of hereditary hemorrhagic telangiectasia (HHT)type 1 (HHT1), an autosomal-dominant disorder characterized bythe presence of arteriovenous malformations (AVMs) in differentorgans (McAllister et al., 1994; Abdalla and Letarte, 2006; Shovlin,2010). It has been postulated that the vascular lesions derive fromabnormal processes of angiogenesis and vascular remodeling,leading to focal loss of capillaries and, as a consequence, a directconnection between venules and arterioles (Wetzel-Strong et al.,2017; Cunha et al., 2017). The current treatments for HHT1 includeseveral antiangiogenic therapies (Ruiz-Llorente et al., 2017;Ardelean and Letarte, 2015). Interestingly, sEng displaysantiangiogenic activity (Hawinkels et al., 2010; Venkatesha et al.,2006), but its putative role in AVM formation and treatment has notyet been explored.In this study, we find that sEng downregulates several pro-

angiogenic and pro-migratory proteins involved in angiogenesis,and this effect is dependent on the presence of endogenoustransmembrane endoglin. Furthermore, using an inducible endoglin(Eng) knockout (KO) animal model, we show that sEng treatmentdecreases the number of AVMs in the neovascularized retina. Takentogether, this work reveals, for the first time, the context-dependentrole of sEng in regulating angiogenesis and vascular pathogenesis.

RESULTSsEng inhibits endothelial tubulogenesis and cell migrationThe effect of sEng on tubulogenesis and in wound healing assayswas analyzed in cultured human umbilical vein-derived endothelialcells (HUVECs) using a physiological range of sEngconcentrations: low (1-10 ng/ml) and mid to high concentrations(40-100 ng/ml). For 3D tubulogenesis assays, HUVECs werecultured on vascular endothelial growth factor (VEGF)-containingMatrigel, and treated with increasing concentrations of sEng(Fig. 1A). After 6 h of culture with sEng, a dose-dependentinhibition of tubular network formation was observed, compared

with control HUVECs without sEng treatment. This inhibition wasmost evident at doses of 40 ng/ml and 100 ng/ml sEng. Next, cellmigration experiments were performed by measuring theendothelial ‘scratch wound’ closure of HUVECs monolayers overtime (Fig. 1B). Under normal conditions, without sEng treatment,the wound was almost closed between 6 h and 8 h (75% and 90%,respectively). However, at 100 ng/ml sEng, closure was only 35% at6 h and 60% at 8 h (Fig. 1B), suggesting that sEng induced aninhibitory effect. Because this type of wound healing assaymeasures a combination of cell migration and proliferation, wealso analyzed the individual effect of sEng on HUVECproliferation. We found no effect of sEng on cell proliferation(data not shown), suggesting that the overall changes observed in thescratch wound assay are mainly due to inhibition of cell migrationby sEng. Our observations that sEng inhibits both endothelial celltubulogenesis and migration are in agreement with the reportedantiangiogenic activity in vivo of sEng, including the inhibition ofVEGF-induced blood vessel formation and sprouting (Venkateshaet al., 2006; Hawinkels et al., 2010; Castonguay et al., 2011).

sEng affects the expression of soluble proteins related toangiogenesis in HUVECs and MLECsTo investigate possible molecular mechanisms underlying theantiangiogenic activity of sEng, conditioned media from HUVECsandmouse lung endothelial cells (MLECs) were analyzed followingtreatment with sEng. Changes in the secreted angiogenic proteinprofile were assessed using angiogenesis-specific antibody arrays.Upon treatment with sEng, HUVECs and MLECs reduced theexpression of 15 of 55 (HUVECs), and 19 of 53 (MLECs),angiogenic proteins present in both human and mouse arrays(Table S1A). Most of the downregulated proteins have pro-angiogenic properties, consistent with the antiangiogenic andantimigratory effects of sEng observed above. Ascertaining whichangiogenesis-related proteins were downregulated in both cell typeswas partly limited by the fact that the arrays had only 72% proteinscommon to both human and mouse panels. Nevertheless, sEngtreatment led to significant downregulation of three pro-angiogenicproteins in both human and mouse ECs [insulin-like growth factor-binding protein 1 (IGFBP-1) and 2 (IGFBP-2), and platelet-derivedendothelial cell growth factor (PD-ECGF)] (Fig. 2A; Table S1A).An additional six proteins [endothelin, CXCL-4 (PF4), dipeptidylpeptidase 4 (DPP4), heparin binding-like epidermal growth factor(HB-EGF), platelet-derived growth factor (PDGF-AA or PDGFA)and thrombospondin-2 (TSP-2 or THBS2)] appeared to be reducedin both human andmouse ECs, but reached statistical significance inonly one cell type (Fig. 2A; Table S1A). In contrast to the strikingdecrease in pro-angiogenic proteins following sEng treatment, onlytwo proteins in HUVECs and none in MLECs were significantlyincreased in response to sEng (Table S1B). Interestingly, maspin,one of the upregulated proteins, displayed angioinhibitory activity,in line with an antiangiogenic effect of sEng treatment.

sEng regulates the secretion of angiogenic proteins and thiseffect is dependent on endoglin gene expressionTo determine whether sEng treatment had a similar antiangiogeniceffect on endoglin-deficient cells, as a model of HHT1, we used aMLEC line isolated from Engfl/fl;RosaCreERT mice (Anderberget al., 2013). The efficiency of 4-OH-tamoxifen-induced endoglingene silencing in these MLECs was confirmed by RT-PCR andimmunostaining (Fig. S1). Differentially expressed angiogenicproteins from Eng-KO and control MLECs, with and withouttreatment with sEng, were determined using the mouse angiogenic

2

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm034397. doi:10.1242/dmm.034397

Disea

seModels&Mechan

isms

http://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplemental

protein array (Table S2). Eng-KO MLECs show altered expressionof many angiogenic proteins compared with controls, in line withthe known importance of endoglin in angiogenesis. However, sEngtreatment normalizes several of these differences back to controllevels. Of eight angiogenic proteins significantly downregulated inEng-KO MLECs (including endogenous Eng), one was restored tonormal values upon treatment with sEng, whereas there was asimilar trend for several other proteins (Table S2). These includeheparin-binding EGF-like growth factor (HB-EGF), involved invascular remodeling; the chemokine (C-X-C motif ) ligand 1(CXCL1), involved in angiogenesis, arteriogenesis, inflammation,and wound healing; pentraxin-3 (PTX3), an acute-phase-responseprotein that regulates angiogenesis after ischemia; and proliferin, a

prolactin/growth hormone-like peptide with angiogenic properties.In addition, expression of angiogenin, a ribonuclease with potentproangiogenic activity, and placental growth factor (PlGF or PGF),a VEGF family member involved in angiogenesis andvasculogenesis, was increased towards normal levels followingsEng treatment. Similarly, of three proteins significantlyupregulated in Eng-KO MLECs, two were reduced to normallevels by sEng. These were coagulation factor III, also known astissue factor (TF), involved in the initial steps of blood coagulation,and PD-ECGF, a thymidine phosphorylase which promotesangiogenesis in vivo and stimulates the growth of ECs in vitro.Several other upregulated proteins, including the pro-inflammatorychemokine MCP-1 (CCL2), showed a similar trend of restoration

Fig. 1. Effect of sEng on tubulogenesis and woundhealing. (A) HUVECs were incubated on Matrigel platesat 37°C in VEGF-enriched EBM2/EGM2 mediumcontaining increasing doses of sEng (0-100 ng/ml). Thecord network formation was visualized by taking picturesat different times up to 6 h after cell plating. Theappearance of a complete network is achieved by 6 h inuntreated cells or cells treated with 1 ng/ml sEng, while athigher concentrations of sEng, cells remain in opentubules with some patches of disorganized and sparsecells. A representative assay of more than three differentexperiments per condition is shown. Scale bars: 100 µm.The tube density (closed tubes) in the network wasquantified, normalized and represented in the histogram.*P

towards normal by sEng treatment (Table S2). To confirm this trend,enzyme-linked immunosorbent assay (ELISA) was used toascertain the levels of two upregulated (coagulation factor III andMCP-1) and two downregulated (HB-EGF and CXCL1) proteins inEng-KO MLECs with and without sEng treatment (Fig. 2B). Thisanalysis confirmed the array data showing that for all four proteins,sEng significantly altered their levels towards that of controls. Takentogether, these results show that sEng can partially rescue theangiogenic protein expression imbalance observed in Eng-KO

MLECs, suggesting that sEng can help to compensate for the lack ofendogenous transmembrane endoglin.

sEng inhibits the development of retinal arteriovenousmalformations in an HHT1 murine modelWe noticed that many of the genes encoding those proteinsidentified in our study (Fig. 2B) are expressed in endothelial cellsduring development of the mouse retina (Jeong et al., 2017), whichis awidely used animal model of angiogenesis (Selvam et al., 2018).In order to assess the effect of sEng on AVM formation duringvascular development, we turned to endothelial-specific tamoxifen-inducible endoglin KO (Eng-iKOe) mice, which develop AVMs inthe neonatal retina (Mahmoud et al., 2010). Wild-type (WT) andendothelial-specific Eng-iKOemicewere each intraocularly injectedwith sEng in the left eye and vehicle (PBS) in the right eye. Retinalvasculature was visualized 2 days later by immunofluorescentstaining. At the retinal periphery, where there is active proliferationof ECs involved in angiogenesis, endoglin expression was increasedcompared with the remodeled vessels of the central zone (Fig. 3A;Fig. S2A). Whereas isolectin staining did not show any significantdifference among the four conditions, endoglin staining wasdecreased in Eng-iKOe retinas, compared with WT samples, asexpected (Fig. 3B). We confirmed that sEng treatment did not affectthe level of endoglin staining, when comparing treated versusuntreated mice, thus ruling out a potential artifact owing to sEngprotein accumulation in the vasculature. As previously described(Mahmoud et al., 2010), loss of endoglin expression in ECs led tothe formation of AVMs within the retinal plexus, delayedprogression of the vascular plexus towards the periphery, andhyperbranching of the peripheral vessels (Fig. 3). The effect of sEngtreatment on these vascular parameters in WT and Eng-iKOe retinaswas assessed. We show here, for the first time, that treatment withsEng inhibits vascular plexus migration inWT retinas (Fig. 3C,D), afinding compatible with the antimigratory and antiangiogenicactivity of sEng observed in vitro (Fig. 1). In contrast, sEngtreatment of Eng-iKOe retinas favors vascular migration, suggestinga possible mechanism of phenotype ‘normalization’ or ‘rescue’ bysEng. Normalized vascular density and area covered by alphasmooth muscle actin (αSMA) staining were higher in Eng-iKOe

(>30%) compared with WT retinas (Fig. 4A,B), and this phenotypewas also partially rescued by sEng treatment, which decreasedvascular density by ∼15%, but had no significant effect on vasculardensity ofWT retinas (Fig. 4A). In addition, a significant increase incaliber or width of veins, but not of arteries, in Eng-iKOe retinas,compared with WT retinas was observed, although this was notaffected by sEng treatment (Fig. 4C,D).

Compared with WT retinas, Eng-iKOe retinas display morecapillary junctions or branching (Fig. 4E1,E2) in the peripheralareas, where angiogenesis is more active, and a higher number offilopodia at the leading edge of migrating tip cells (Fig. 4F1,F2).However, intraocular treatment with sEng significantly decreasedboth branching and filopodia in Eng-iKOe retinas (Fig. 4E1,F1) by19% (branching) and 17% (filopodia number), thus partiallyrestoring the phenotype of Eng-iKOe retinas to that of WT.

Based on the observed role of sEng on angiogenesis and vascularremodeling, we next examined the effect of sEng on AVM formationin Eng-iKOe retinas (Fig. S2A,B). Treatment with sEng showed atendency to decrease the size of AVMs (Fig. S3A,B). Moreover, sEngsignificantly decreased the incidence of AVMs in Eng-iKOe retinaswhen compared with vehicle treatment (Fig. 5A; Fig. S2A,B).Overall, the mean number of AVMs was reduced from 4.0 to 2.5AVMs/retina in Eng-iKOe mice upon treatment with sEng (Fig. 5B).

Fig. 2. Analysis of deregulated proteins upon treatment with sEng.Differentially expressed secreted angiogenic proteins in HUVECs, MLECs andEng-KO MLECs, previously identified in protein arrays, using a cutoff of 0.9 or1.1 (Tables S1 and S2) were analyzed. (A) Venn diagram representing thedifferential protein expression between HUVECs and MLECs upon treatmentwith sEng. Among the downregulated proteins, from human and mousepanels, 12 were found only in HUVECs, 16 were found only in MLECs andthree were found in both HUVECs and MLECs. Within this last set of proteins,statistical significance in both cell types was found for IGFBP-1, IGFBP-2 andPD-ECGF. (B) ELISA analysis of differentially secreted angiogenic proteins inEng-KO MLECs in the absence (KO) or presence (KO+sEng) of solubleendoglin compared with control MLECs. Results were normalized to the totalprotein concentration in supernatants, and compared with untreated MLECs,which was given an arbitrary value of 1. For each protein, a minimum of fourdifferent experiments, each in triplicate, were carried out. Fold change (FC)measurements±s.e.m. and P-values between sEng-treated and untreated KOMLECs are represented.

4

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm034397. doi:10.1242/dmm.034397

Disea

seModels&Mechan

isms

http://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplemental

In the neonatal retina, vascular smooth muscle cell (vSMC)coverage is associated with muscularized arteries, whereas it isnormally absent from veins at this stage of development. We find

Fig. 3. Analysis of vascular markers andmigration in retinas. (A) P7 retinasfrom Eng-iKOe and WT mice were stained with isolectin (green) and anti-endoglin (red), visualized by fluorescence microscopy and analyzed using theFiji-ImageJ program. Decreased endothelial endoglin expression and increasedabnormalities in the neonatal vascular plexus from retinas of Eng-iKOe micecompared with WT animals were observed. The presence of veins (V) andarteries (A) is indicated. Endoglin staining (red fluorescence) predominates inveins. Loss of endoglin protein expression in endothelial cells leads to AVMs(arrows) and hyperbranching in the periphery of retinas (asterisk). White dottedline boxes in upper panels represent regions enlarged in lower panels. Imagesare taken at different magnifications (5×, 10× or 20×). Scale bars: 500 µm (1,2),400 µm (3,4) and 100 µm (5,6). (B) Quantification of vascular staining insEng-intraocularly treated and untreated mice. WT and Eng-iKOe mice weretreatedwith or without sEng, as indicated. Fluorescence intensity of eachmarkerwas quantified using the Fiji-ImageJ2 software, and normalized with respect toWT controls. Endoglin expression is markedly decreased in Eng-iKOe versusWT retinas, whereas it is not affected by sEng treatment. (***P20mice per condition. (C,D) Analysis of vascular migration. (C) Examples ofstained retinas to illustrate the measurement of vascular and retina radius areshown. Red arrows indicate the vascular radius (measurement from the centerto the edge of the vascular front), whereas purple arrows indicate the retinaradius (measurement from the center to the edge of retina). The black areasindicate areas in which there has been some breakage in the fragile retina.Magnification, 5×. Scale bars: 500 µM (D) Quantification of vascular migration.The vascular radius/retina radius ratio, represented as a percentage, withrespect to the control, was used to quantify vascular migration. Progression ofthe vascular plexus to retinal periphery is delayed in Eng-iKOe compared withWT control retinas, and this delay is partly reversed when Eng-iKOe animals aretreated with sEng. In WT retinas, sEng treatment decreases vascular migration.***P

that the development of AVMs is associated with an ‘arteriolization’effect owing to increased blood flow through these vessels, leadingto an increased number of αSMA-positive vSMCs at the site ofAVMs (Fig. S2B), as reported (Mahmoud et al., 2010). Of note, wefound a significantly increased expression of αSMA in Eng-iKOe

retinas (compared with WT retinas) that was associated with AVMs(Fig. 4B; Figs S2B and S3C). Importantly, the area occupied byvSMCs in Eng-iKOe retinas is significantly decreased after sEngtreatment (Fig. 4B; Fig. S3C), consistent with the reduced caliber,area (Fig. S3A,B) and number (Fig. 5) of AVMs.Taken together, the above results (summarized in Table 1) show

that sEng can partially compensate for membrane-bound endoglininsufficiency to promote rescue of vascular defects in Eng-iKOe

mice.

DISCUSSIONAbnormal levels of sEng are found in several endothelium-relatedpathological conditions, including pre-eclampsia (Venkatesha et al.,

2006; Oujo et al., 2013; Gregory et al., 2014), atherosclerosis,hypercholesterolemia (Blann et al., 1996; Blaha et al., 2008;Rathouska et al., 2015), diabetes mellitus (Blázquez-Medela et al.,2010; Ceriello et al., 2015; Emeksiz et al., 2016), hypertension(Blázquez-Medela et al., 2010), diabetic retinopathy (Malik et al.,2005), coronary artery disease (Li et al., 2000a; Ikemoto et al., 2012;Saita et al., 2017), HHT1 (Letarte et al., 2005; Botella et al., 2015),acute myocardial infarction (Cruz-Gonzalez et al., 2008) and cancer(Li et al., 2000b; Bernabeu et al., 2009). In many of these diseases,the deregulated levels of sEng in plasma, serum or urine frompatients have been postulated to be a reliable biomarker for severitycorrelation and prognosis. Furthermore, sEng appears to play anactive role in disease pathogenesis. For example, it has beenreported that, in pre-eclampsia, sEng contributes to hypertensionand renal involvement (Venkatesha et al., 2006; Valbuena-Díezet al., 2012); in atherosclerosis, sEng induces a pro-inflammatoryresponse, leading to endothelial dysfunction (Varejckova et al.,2017; Jezkova et al., 2016); and, in cancer, sEng acts as anantiangiogenic protein by inhibiting the ongoing neoangiogenesisassociated with the growth of solid tumors (Hawinkels et al., 2010;Castonguay et al., 2011). In spite of the wide range ofpathophysiological effects of sEng reported in the cardiovascularsystem, its underlying mechanism of action on ECs is poorlyunderstood. Of note, the balance between pro- and antiangiogenicfactors in angiogenesis/vascular remodeling is crucial (Potente andCarmeliet, 2017). One of the aims of this work was to analyzethe levels of angiogenesis-related proteins released from ECs inthe presence of sEng, to provide some insights into the alteredangiogenesis/vascular remodeling processes associated withincreased circulating levels of endoglin. Using human and mouseECs, both expressing high levels of membrane-bound endoglin, wefound that sEng induced a protein expression pattern with apredominant antiangiogenic profile. After sEng treatment, the levelsof secreted pro-angiogenic proteins, such as VEGF, fibroblastgrowth factor (FGF), PDGF, IGFBP proteins and thrombospondin,were all significantly decreased, consistent with the reportedantiangiogenic effect of sEng. In this regard, sEng stimulated theexpression of only one pro-angiogenic protein (leptin) andthe antiangiogenic protein (maspin) (Table S1B). Surprisingly, the

Fig. 5. Effect of sEng treatment on the number of AVMs in Eng-iKOe retinas. AVMs from at least 30 retinas from each group were analyzed. (A) Histogramrepresentation of the number of AVMs found relative to the number of counted retinas. Retinas treated with sEng show a lower incidence of AVMs thanuntreated retinas. (B) Mean number of AVMs corrected for the number of retinas analyzed in Eng-iKOe mice, treated with or without sEng. Values show 95%confidence intervals. **P

cellular response to sEng in the absence of membrane-boundendoglin yielded different response to ECs expressing normal levelsof endoglin (Tables S1 and S2). Indeed, altered levels of secretedangiogenic proteins in Eng-iKOe ECs, including coagulation factorIII, PD-ECGF, HB-EGF, CXCL1 and MCP-1, tend to benormalized towards that of control ECs after treatment with sEng(Fig. 2B; Table S2B). These results clearly indicate that membrane-bound endoglin is key in the regulation of the angiogenic signalingand its absence produces an imbalance that can be compensated bysEng.Because endoglin is a component of the TGF-β receptor complex

(López-Novoa and Bernabeu, 2010; Mahmoud et al., 2011), it istempting to speculate that the effects of sEng on protein expressionare mediated by this signaling pathway. In this regard, it is worthmentioning that SERPINE1 (also known as PAI-1), a characteristicdownstream target of the TGF-β route in ECs. Thus, basal levels ofPAI-1 in Eng-iKOe cells tend to be higher than those in control ECs(Table S2B), a finding compatible with the upregulated expression ofthe PAI-1 gene found in gene arrays of ECs derived from HHTpatients (Fernandez-Lopez et al., 2007) and with increased ALK5(TGFBR1) signaling in the absence of endoglin (Lebrin et al., 2004).However, sEng treatment tends to restore PAI-1 levels in Eng-iKOe

cells to similar levels of control ECs (Table S2B), potentially byrestoring the normal balance of the TGF-β signaling pathway. Ofnote, the extracellular part of endoglin specifically interacts with theTGF-β type I receptors ALK1 (ACVRL1) and ALK5 and with theTGF-β type II receptor (Guerrero-Esteo et al., 2002; Blanco et al.,2005). Also, endoglin is released from the placenta into the maternalcirculation via sphingomyelin (18:0)-enriched exosomes, togetherwith ALK5 and the TGF-β type II, where it can associate with thesereceptors, forming a functional receptor complex and modulating thevascular effects of TGF-β in the circulation (Ermini et al., 2017).Although soluble endoglin does not bind on its own to TGF-β1(Gregory et al., 2014), it is possible that sEng, which contains most ofthe extracellular part of endoglin, could bind to the TGF-β receptorcomplex, mimicking membrane-bound endoglin and thusmodulating its downstream signaling. This is an interestingpossibility that could explain why in endoglin-silenced ECs, sEngrestores the protein expression pattern of normal endothelial cells, andin an HHT1 animal model, sEng decreases the number of AVMs.The extracellular region of endoglin binds with high affinity to

BMP9 (Castonguay et al., 2011; Alt et al., 2012; Saito et al., 2017),a TGF-β family member involved in the development of blood andlymphatic vessels (Levet et al., 2013; Chen et al., 2013; Li et al.,2016) and in endoglin-dependent chemokine responses of ECs(Young et al., 2012). In addition to HHT1 patients carryingmutations in endoglin, an HHT-like disorder has also been reportedin patients heterozygous for mutations in BMP9 (Wooderchak-Donahue et al., 2013), whereas single-allele mutations in the ALK1gene give rise to HHT2 (Johnson et al., 1996). Accordingly, BMP9,ALK1 and endoglin proteins participate in a common signalingpathway also involving the type II receptors BMPRII and ActRII(Mahmoud et al., 2011; Tillet and Bailly, 2015; Roman and Hinck,2017; Ruiz-Llorente et al., 2017) (Fig. 6A,B). Thus, in normal ECsit is possible that upon binding to BMP9, sEng (either alone or incomplex with exosomes), ‘hijacks’ the ligand, prevents its bindingto the receptor complex, and inhibits the downstream intracellularsignaling (Hawinkels et al., 2010; Gregory et al., 2014; Venkateshaet al., 2006; Wang et al., 2017) (Fig. 6C). However, in endoglin-silenced ECs, BMP9 cannot bind to membrane endoglin, but canstill associate with sEng (Castonguay et al., 2011; Saito et al., 2017).Furthermore, because sEng-bound BMP9 can directly interact with

ALK1 (Blanco et al., 2005; Castonguay et al., 2011; Saito et al.,2017), it can be hypothesized that the effects of BMP9/sEng involveits interaction with ALK1 to promote proangiogenic ALK1signaling and decrease the incidence of AVMs (Fig. 6D). As sEngand the type II receptors BMPRII or ActRII bind to BMP9 in amutually exclusive fashion, sEng would be released, potentiallyenabling repeated enhancement of signaling. However, furtherdetailed studies are needed to elucidate the exact mechanism ofaction of sEng in this pathway.

As discussed above, sEng displays antiangiogenic activity andthe protein array data suggest that sEng regulates the expression ofdifferent proteins involved in angiogenesis and vascularremodeling, which are key processes involved in the generation ofAVMs. In addition, using an HHT1 model of ECs, sEng was able to

Fig. 6. Hypothetical model of sEng action on the vasculature. (A,B) Innormal endothelial cells, membrane endoglin is a component of a receptorcomplex that contains type I (RI; ALK1), and type II (RII; BMPR2/ActRII) TGF-βreceptors, which can be activated by BMP9, leading to an equilibrium betweenpro- and antiangiogenic factors (A). In endoglin-silenced endothelial cells(Eng-iKOe), BMP9 cannot bind to endoglin and BMP9-dependent signaling isdisturbed, leading to a dysregulated expression of angiogenic factors,decreased migration during angiogenesis and the presence of AVMs (B).(C,D) In normal endothelial cells, the circulating extracellular region of endoglin(sEng) can interact with BMP9, sequestering the ligand, interfering with theintracellular signaling of the receptor complex and changing the angiogenesisbalance towards a dysregulated state (C). In Eng-iKOe, BMP9 cannot bind tomembrane endoglin, but interacts with sEng, and the resulting BMP9/sEngcomplex interacts with the ALK1 receptor on the cell membrane, enhancing theproangiogenic ALK1 signaling and decreasing the incidence of AVMs (D). Theinvolvement of endoglin in the TGF-β1/ALK5 signaling pathway of endothelialcells has been omitted for simplification.

7

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm034397. doi:10.1242/dmm.034397

Disea

seModels&Mechan

isms

http://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplemental

partly restore the protein expression pattern of normal ECs. Becauseseveral antiangiogenic therapies have been used in HHT patients(Lebrin et al., 2010; Dupuis-Girod et al., 2012, 2016; Albiñanaet al., 2012) to inhibit vascular bleeding produced by the rupture ofAVMs, we analyzed whether sEng could also modulate theoccurrence of AVMs formed in an established mouse model ofHHT1 (Mahmoud et al., 2010; Tual-Chalot et al., 2015). In theabsence of endoglin, retinal vascular development shows delayedvascular plexus remodeling, increased endothelial cell proliferation,formation of AVMs, and increased SMA expression in AVMs as aresult of increased flow (Mahmoud et al., 2010). We find thatintraocularly injected sEng promotes rescue of abnormalvasculature, and decreases the incidence and size of AVMs in thismodel (Table 1). Upon sEng treatment, the number of AVMs wasdecreased, the number of capillary unions and sprouts of tip cells inthe periphery were normalized, and the vascular density and thelength of veins were restored, with respect to the WT phenotype.It has been proposed that sEng might play a role in promoting

AVMs (Chen et al., 2009). Patients with sporadic brain AVMs hadhigher levels of sEng compared with controls, but membrane-boundendoglin levels were normal and there was no history of HHT. Ourfindings would suggest that sEng acts differently in the presenceof membrane-bound endoglin, compared with cases in whichmembrane-bound endoglin is reduced. Interestingly, one of thehallmarks of HHT1 is the deficient expression of endoglin (Abdallaand Letarte, 2006; Ruiz-Llorente et al., 2017) and, accordingly, itcould be speculated that addition of exogenous sEng in this contextcould have a beneficial effect in counteracting the formation of AVMs.However, the underlying mechanism of action of sEng in AVMformation in both sporadic cases and in HHT remains to be elucidated.In addition to the putative role of sEng in the TGF-β/BMP9

pathway, its function as a modulator of cell adhesion cannot beexcluded. In this regard, the extracellular region of endoglin displays,within its zona pellucida domain, an RGDmotif, which is a consensussequence implicated in integrin-based interactions with other proteins(Gougos and Letarte, 1990; Saito et al., 2017; Llorca et al., 2007).Accordingly, it has been shown that sEng can modulate integrin-mediated cell adhesion involving membrane-bound endothelialendoglin (Rossi et al., 2013, 2016; Ruiz-Remolina et al., 2017), andthis function might have an impact on the active angiogenesis andvascular remodeling processes in the neovascularized retina of Eng-iKOe mice. In this regard, upon an inflammatory stimulus, leukocyterecruitment to the vasculature involves endothelial endoglin, vialeukocyte integrins (Rossi et al., 2013), as well as BMP9 (Mitrofanet al., 2017). Because the inflammatory infiltrate of leukocytes appearsto be involved in the vascular remodeling leading to AVMs in HHTpatients (van Laake et al., 2006; Zhang et al., 2016), a role for sEng inthis cell adhesion-dependent process can be postulated (Dingenoutset al., 2015; Rossi et al., 2015).Endoglin and ALK1 play a key role in angiogenesis (López-

Novoa and Bernabeu, 2010; Núñez-Gómez et al., 2017; Mahmoudet al., 2011; Roman and Hinck, 2017), and targeting these proteinshas been used as a therapeutic approach to treat tumor angiogenesis.For example, TRC105 is a humanized monoclonal antibody againstendoglin that not only binds to membrane-bound endoglin, but isalso able to stimulate shedding of sEng (Kumar et al., 2014; Rosenet al., 2014). TRC105 is currently used in antiangiogenic therapiesof several types of cancer (Liu et al., 2014; Paauwe et al., 2016).Moreover, ongoing clinical trials are testing two ALK1-relatedpharmacological inhibitors: dalantercept/ACE-041, an ALK1extracellular domain fusion protein, and PF-03446962, anantibody against the extracellular domain of ALK1 (de Vinuesa

et al., 2016). Whether sEng can also be used as an antiangiogenicdrug remains to be determined.

Taken together, our results suggest that sEng has context-dependent effects. In the presence of endogenous transmembraneendoglin, it has antiangiogenic properties, but, in the absence ofendogenous endoglin, sEng can rescue angiogenic defects inHHT1. This would suggest that under these latter conditions, sEngis pro-angiogenic. However, this hypothesis needs to be confirmedby further investigations on the exact mechanistic role of sEng in theabsence of endogenous endoglin. Although we have shown thatsEng can reduce the generation of AVMs, additional studies areneeded to explore whether sEng is able to regress AVMs once theyhave been formed, and whether sEng could be used in the future as atherapeutic drug for the resolution of AVMs in HHT1 patients.These studies could also be helpful to better understand the functionof sEng in conditions such as pre-eclampsia, cancer or inflammatorydisease, where sEng is present at high levels.

MATERIALS AND METHODSCell culture and treatmentsAll cells were incubated routinely at 37°C in a humidified atmosphere with5% CO2. HUVECs were purchased from Lonza and used at early passages(3-5). HUVECs were grown on 0.2% gelatin (Sigma-Aldrich) pre-coatedplates, in endothelial growth medium (EGM-2) supplemented with 10%heat-inactivated fetal bovine serum (FBS, Gibco), 2 mM L-glutamine,100 U/ml penicillin and 100 µg/ml streptomycin (Gibco), unless otherwisenoted. MLECs were derived from Engfl/fl;RosaCreERT mice, as previouslydescribed (Anderberg et al., 2013). Briefly, endothelial cells wereisolated with anti-CD31 (PECAM1)-coated Dynabeads sheep anti-rat IgG(Invitrogen) after dissociation of the lung tissue with 1 mg/ml collagenase,and cultured in MV2 endothelial cell basal medium (PromoCell) with 10%FBS. Treatment with 1 µM 4-OH-tamoxifen for 48 h was used to activateCreERT, leading to endoglin deletion. Treatments with human recombinantsEng (R&D Systems), reconstituted in PBS with 0.1% bovine serumalbumin (BSA), were carried out either in serum-free medium or in 2% FBSmedium, as indicated. For angiogenic protein assays, HUVECs or MLECs(with or without endoglin KO) were incubated with or without 100 ng/mlsEng for 24 h in EBM-2 or MV2 serum-free medium, respectively.The resulting culture supernatants were used to evaluate secretedangiogenic proteins using protein antibody arrays (described below).Culture supernatants were also analyzed using ELISA kits specific formouse CXCL1 (DY453-05), coagulation factor III (DY3178-05), HB-EGF(DY8239-05) and MCP-1 (DY479-05) from R&D Systems. Theimmunoassays were performed according to the manufacturer’s protocoland measured in a Glomax® Multi Detection System (Promega).

Wound healing and tube formation assaysIn vitro scratch wounds were created by scraping confluent HUVECmonolayers in 24-well plates with a sterile pipette tip. Fresh EBM2medium supplemented with 2% FBS and EGM2 (Lonza) with differentconcentrations of sEng was added, and samples were incubated for up to10 h. Endothelial cell migration into the denuded area was monitored at 0, 4,6, 8 and 10 h postwound. The ImageJ program (Schneider et al., 2012) wasused to quantify the wound healing process. For tube formation assays,HUVECs were seeded on 24-well plates, previously covered with 100 μlstandardMatrigel (BDBioscience) diluted 1:2 in serum-free EBM2mediumsupplemented with VEGF-containing EGM2 (Lonza). Samples wereincubated at 37°C in EBM2 medium supplemented with 2% FBS andEGM2 in the presence of sEng (0-100 ng/ml) for 6 h, as indicated. Imageswere taken with an Olympus digital camera, and quantification of closedtubules was performed using Adobe Photoshop CS3 software.

Angiogenesis protein arraysConditioned medium from HUVECs and MLECs cells, previouslyincubated with 100 ng/ml sEng in serum-free medium for 24 h, wereused. The relative expression profile of angiogenesis-related proteins were

8

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm034397. doi:10.1242/dmm.034397

Disea

seModels&Mechan

isms

quantified using a proteome profiler mouse angiogenesis array kit (ARY015-mouse; ARY 007-human; R&D Systems) according to themanufacturer’s instructions. The relative intensities of the spots werenormalized per total protein concentration of each supernatant, measuredwith the BCA protein assay kit (Pierce), and by subtracting the backgroundof the membranes. Each protein array experiment was performed intriplicate.

Mice and treatmentsThe Engfl/fl mouse and Eng-iKOe mouse lines have been previouslydescribed (Mahmoud et al., 2010; Allinson et al., 2007). Neonates wereinjected subcutaneously with 0.5 mg tamoxifen (Sigma-Aldrich) atpostnatal day (P) 2 and P4 to activate Cdh5-CreERT2, as reported(Mahmoud et al., 2010). Controls were tamoxifen-treated Engfl/fl

littermates. For intraocular treatments, P5 neonates were anesthetized by3% isofluorane and intraocularly injected with 0.3 µl of 100 ng/ml sEng inPBS/0.1% BSA or control solution (PBS/0.1% BSA) in the left and righteye, respectively, using a Hamilton syringe (80001 10 μl SYR) attached to afine needle (Needle PRE-33013 TSK Laboratory).

Whole-mount, immunofluorescence staining and analysis ofdifferent parameters in mouse retinasMouse eyes were enucleated immediately postmortem at P7. Retinas wereprepared and stained following the whole-mount immunofluorescenceprotocol from Tual-Chalot et al. (2013). Briefly, retinas were fixed in 4%(wt/vol) paraformaldehyde in PBS and then stained with Alexa Fluor488-conjugated G. simplicifolia isolectin B4 (Life Technologies).Immunostaining using antibody against mouse endoglin (MJ7/18,eBioscience) was detected using an Alexa Fluor 568-conjugatedsecondary anti-rat antibody (Thermo Fisher Scientific). Vascular smoothmuscle was detected using anti-αSMA-Cy3 antibody (Sigma-Aldrich).Stained retinas were flat mounted using ProLong Gold Antifade Mountmedium (Thermo Fisher Scientific) and examined under a Zeiss Axioimagermicroscope, and images were analyzed using Zeiss-ZEN software. Analysisof vascular parameters was performed using Fiji-ImageJ2 software(Schindelin et al., 2012). Vascular migration was calculated by dividingthe length of the vessel radius by the total length of the retina radius, bothmeasured from the center of the retina in at least three different directions.Analysis of the branch points was calculated by counting the closedcapillaries in at least three different areas near the retinal periphery.Measurements of the number of filopodia, at the leading edge of migratingtip cells, were normalized to the selected field of view, in at least fourdifferent sections of the retina.

The percentage area covered by isolectin-positive ECs, and αSMA- orendoglin-positive cells was calculated as the area occupied by the intensityof each marker normalized by the total retinal vascular area. Nonspecificstaining at the edges of the retina or remnants of the hyaloid vessel werediscarded. The number of AVMs was measured throughout the retina. Atleast 15 retinas were analyzed per parameter and per group.

Statistical analysisWound healing, tubulogenesis and array data were analyzed by Student’st-test. ANOVA test was used for the comparison of more than one group.Nonparametric data were analyzed using the Kruskal–Wallis test. Variancesbetween the different groups were determined by the Levene’s test. Posthoc Tamhane or Bonferroni tests were performed for homogeneous ornonhomogeneous variances, respectively. Analysis was performed usingSPSS software. All results shown in graphical representations are the meanof a minimum of three replicates. Data are presented as mean±s.e.m. withthe corresponding P-values.

EthicsAll animal experiments were performed under UK Home Office license,with approval from the Newcastle University Ethical Review Committeeand following the guidelines of EU Directive 2010/63/UE for animalexperiments. All applicable international, national and/or institutionalguidelines for the care and use of animals were followed.

AcknowledgementsWe thank Dr Marcus Fruttiger (UCL Institute of Opthalmology, London, UK) fortraining advice in intraocular injections; Darroch Hall, Esha Singh and Elena de Blasfor technical support; Rafael Nun ̃ez for bioinformatics support; and Laura Barrios forhelp with statistical analysis.

Competing interestsThe authors declare no competing or financial interests.

Author contributionsConceptualization: E.G.-V., L.M.B., H.M.A., C.B.; Methodology: E.G.-V., S.T.-C.;Validation: E.G.-V.; Formal analysis: E.G.-V.; Investigation: E.G.-V.; Resources:H.M.A., C.B.; Writing - original draft: E.G.-V., L.M.B., H.M.A., C.B.; Writing - review &editing: E.G.-V., H.M.A., C.B.; Visualization: E.G.-V., H.M.A., C.B.; Supervision:H.M.A., C.B.; Project administration: H.M.A., C.B.; Funding acquisition: H.M.A., C.B.

FundingThis work was supported by Ministerio de Economıá, Industria y Competitividad,Gobierno de Espan ̃a (SAF2010-19222 and SAF2013-43421-R to C.B.), Centro deInvestigacion Biomedica en Red de Enfermedades Raras (CIBERER; ISCIII-CB06/07/0038) and the British Heart Foundation (PG/14/86/31177 to H.M.A.). E.G.-V. wassupported by the International Mobility Program of Ministerio de Economıá, Industriay Competitividad, Gobierno de Espan ̃a (EEBB-I-14-09020 and EEBB-I-15-10398).

Supplementary informationSupplementary information available online athttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplemental

ReferencesAbdalla, S. A. and Letarte, M. (2006). Hereditary haemorrhagic telangiectasia:

current views on genetics andmechanisms of disease. J. Med. Genet. 43, 97-110.Albin ̃ana, V., Recio-Poveda, L., Zarrabeitia, R., Bernabéu, C. and Botella, L. M.

(2012). Propranolol as antiangiogenic candidate for the therapy of hereditaryhaemorrhagic telangiectasia. Thromb. Haemost. 108, 41-53.

Allinson, K. R., Carvalho, R. L. C., van den Brink, S., Mummery, C. L. andArthur, H. M. (2007). Generation of a floxed allele of the mouse Endoglin gene.Genesis 45, 391-395.

Alt, A., Miguel-Romero, L., Donderis, J., Aristorena,M., Blanco, F. J., Round, A.,Rubio, V., Bernabeu, C. andMarina, A. (2012). Structural and functional insightsinto endoglin ligand recognition and binding. PLoS ONE 7, e29948.

Anderberg, C., Cunha, S. I., Zhai, Z., Cortez, E., Pardali, E., Johnson, J. R.,Franco, M., Páez-Ribes, M., Cordiner, R., Fuxe, J. et al. (2013). Deficiency forendoglin in tumor vasculature weakens the endothelial barrier to metastaticdissemination. J. Exp. Med. 210, 563-579.

Ardelean, D. S. and Letarte, M. (2015). Anti-angiogenic therapeutic strategies inhereditary hemorrhagic telangiectasia. Front. Genet. 6, 35.

Baeyens, N., Larrivée, B., Ola, R., Hayward-Piatkowskyi, B., Dubrac, A., Huang,B., Ross, T. D., Coon, B. G., Min, E., Tsarfati, M. et al. (2016). Defective fluidshear stress mechanotransduction mediates hereditary hemorrhagictelangiectasia. J. Cell Biol. 214, 807-816.

Bellón, T., Corbı,́ A., Lastres, P., Calés, C., Cebrián, M., Vera, S., Cheifetz, S.,Massague, J., Letarte, M. and Bernabéu, C. (1993). Identification andexpression of two forms of the human transforming growth factor-beta-bindingprotein endoglin with distinct cytoplasmic regions. Eur. J. Immunol. 23,2340-2345.

Bernabeu, C., Lopez-Novoa, J. M. and Quintanilla, M. (2009). The emerging roleof TGF-beta superfamily coreceptors in cancer. Biochim. Biophys. Acta 1792,954-973.

Blaha, M., Cermanova, M., Blaha, V., Jarolim, P., Andrys, C., Blazek, M., Maly,J., Smolej, L., Zajic, J., Masin, V. et al. (2008). Elevated serum soluble endoglin(sCD105) decreased during extracorporeal elimination therapy for familialhypercholesterolemia. Atherosclerosis 197, 264-270.

Blanco, F. J. and Bernabeu, C. (2011). Alternative splicing factor or splicing factor-2 plays a key role in intron retention of the endoglin gene during endothelialsenescence. Aging Cell 10, 896-907.

Blanco, F. J., Santibanez, J. F., Guerrero-Esteo, M., Langa, C., Vary, C. P. H. andBernabeu, C. (2005). Interaction and functional interplay between endoglin andALK-1, two components of the endothelial transforming growth factor-betareceptor complex. J. Cell. Physiol. 204, 574-584.

Blann, A. D., Wang, J. M., Wilson, P. B. and Kumar, S. (1996). Serum levels of theTGF-beta receptor are increased in atherosclerosis. Atherosclerosis 120,221-226.

Blázquez-Medela, A. M., Garcıá-Ortiz, L., Gómez-Marcos, M. A., Recio-Rodrıǵuez, J. I., Sánchez-Rodrıǵuez, A., López-Novoa, J. M. and Martıńez-Salgado, C. (2010). Increased plasma soluble endoglin levels as an indicator ofcardiovascular alterations in hypertensive and diabetic patients. BMCMed. 8, 86.

9

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm034397. doi:10.1242/dmm.034397

Disea

seModels&Mechan

isms

http://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dmm.biologists.org/lookup/doi/10.1242/dmm.034397.supplementalhttp://dx.doi.org/10.1136/jmg.2005.030833http://dx.doi.org/10.1136/jmg.2005.030833http://dx.doi.org/10.1160/TH11-11-0809http://dx.doi.org/10.1160/TH11-11-0809http://dx.doi.org/10.1160/TH11-11-0809http://dx.doi.org/10.1002/dvg.20284http://dx.doi.org/10.1002/dvg.20284http://dx.doi.org/10.1002/dvg.20284http://dx.doi.org/10.1371/journal.pone.0029948http://dx.doi.org/10.1371/journal.pone.0029948http://dx.doi.org/10.1371/journal.pone.0029948http://dx.doi.org/10.1084/jem.20120662http://dx.doi.org/10.1084/jem.20120662http://dx.doi.org/10.1084/jem.20120662http://dx.doi.org/10.1084/jem.20120662http://dx.doi.org/10.3389/fgene.2015.00035http://dx.doi.org/10.3389/fgene.2015.00035http://dx.doi.org/10.1083/jcb.201603106http://dx.doi.org/10.1083/jcb.201603106http://dx.doi.org/10.1083/jcb.201603106http://dx.doi.org/10.1083/jcb.201603106http://dx.doi.org/10.1002/eji.1830230943http://dx.doi.org/10.1002/eji.1830230943http://dx.doi.org/10.1002/eji.1830230943http://dx.doi.org/10.1002/eji.1830230943http://dx.doi.org/10.1002/eji.1830230943http://dx.doi.org/10.1016/j.bbadis.2009.07.003http://dx.doi.org/10.1016/j.bbadis.2009.07.003http://dx.doi.org/10.1016/j.bbadis.2009.07.003http://dx.doi.org/10.1016/j.atherosclerosis.2007.04.022http://dx.doi.org/10.1016/j.atherosclerosis.2007.04.022http://dx.doi.org/10.1016/j.atherosclerosis.2007.04.022http://dx.doi.org/10.1016/j.atherosclerosis.2007.04.022http://dx.doi.org/10.1111/j.1474-9726.2011.00727.xhttp://dx.doi.org/10.1111/j.1474-9726.2011.00727.xhttp://dx.doi.org/10.1111/j.1474-9726.2011.00727.xhttp://dx.doi.org/10.1002/jcp.20311http://dx.doi.org/10.1002/jcp.20311http://dx.doi.org/10.1002/jcp.20311http://dx.doi.org/10.1002/jcp.20311http://dx.doi.org/10.1016/0021-9150(95)05713-7http://dx.doi.org/10.1016/0021-9150(95)05713-7http://dx.doi.org/10.1016/0021-9150(95)05713-7http://dx.doi.org/10.1186/1741-7015-8-86http://dx.doi.org/10.1186/1741-7015-8-86http://dx.doi.org/10.1186/1741-7015-8-86http://dx.doi.org/10.1186/1741-7015-8-86

Botella, L.-M., Albin ̃ana, V., Ojeda-Fernandez, L., Recio-Poveda, L. andBernabéu, C. (2015). Research on potential biomarkers in hereditaryhemorrhagic telangiectasia. Front. Genet. 6, 115.

Castonguay, R., Werner, E. D., Matthews, R. G., Presman, E., Mulivor, A. W.,Solban, N., Sako, D., Pearsall, R. S., Underwood, K. W., Seehra, J. et al.(2011). Soluble endoglin specifically binds bonemorphogenetic proteins 9 and 10via its orphan domain, inhibits blood vessel formation, and suppresses tumorgrowth. J. Biol. Chem. 286, 30034-30046.

Ceriello, A., La Sala, L., De Nigris, V., Pujadas, G., Testa, R., Uccellatore, A. andGenovese, S. (2015). GLP-1 reduces metalloproteinase-14 and soluble endoglininduced by both hyperglycemia and hypoglycemia in type 1 diabetes. Endocrine50, 508-511.

Chen, Y., Hao, Q., Kim, H., Su, H., Letarte, M., Karumanchi, S. A., Lawton, M. T.,Barbaro, N. M., Yang, G.-Y. and Young, W. L. (2009). Soluble endoglinmodulates aberrant cerebral vascular remodeling. Ann. Neurol. 66, 19-27.

Chen, H., Brady Ridgway, J., Sai, T., Lai, J., Warming, S., Chen, H., Roose-Girma, M., Zhang, G., Shou, W. and Yan, M. (2013). Context-dependentsignaling defines roles of BMP9 and BMP10 in embryonic and postnataldevelopment. Proc. Natl. Acad. Sci. USA 110, 11887-11892.

Cruz-Gonzalez, I., Pabón, P., Rodrıǵuez-Barbero, A., Martıń-Moreiras, J.,Pericacho, M., Sánchez, P. L., Ramirez, V., Sánchez-Ledesma, M., Martıń-Herrero, F., Jiménez-Candil, J. et al. (2008). Identification of serum endoglin asa novel prognostic marker after acute myocardial infarction. J. Cell. Mol. Med. 12,955-961.

Cunha, S. I., Magnusson, P. U., Dejana, E. and Lampugnani, M. G. (2017).Deregulated TGF-β/BMP signaling in vascular malformations. Circ. Res. 121,981-999.

de Vinuesa, A. G., Bocci, M., Pietras, K. and Ten Dijke, P. (2016). Targetingtumour vasculature by inhibiting activin receptor-like kinase (ALK)1 function.Biochem. Soc. Trans. 44, 1142-1149.

Dingenouts, C. K. E., Goumans, M.-J. and Bakker, W. (2015). Mononuclear cellsand vascular repair in HHT. Front. Genet. 6, 114.

Dupuis-Girod, S., Ginon, I., Saurin, J.-C., Marion, D., Guillot, E., Decullier, E.,Roux, A., Carette, M.-F., Gilbert-Dussardier, B., Hatron, P.-Y. et al. (2012).Bevacizumab in patients with hereditary hemorrhagic telangiectasia and severehepatic vascular malformations and high cardiac output. JAMA 307, 948-955.

Dupuis-Girod, S., Ambrun, A., Decullier, E., Fargeton, A.-E., Roux, A., Bréant,V., Colombet, B., Rivier̀e, S., Cartier, C., Lacombe, P. et al. (2016). Effect ofbevacizumab nasal spray on epistaxis duration in hereditary hemorrhagictelangectasia: a randomized clinical trial. JAMA 316, 934-942.

Emeksiz, H. C., Bideci, A., Damar, Ç., Derinkuyu, B., Çelik, N., Döğer, E., Yüce,Ö., Özmen, M. C., Çamurdan, M. O. and Cinaz, P. (2016). Soluble endoglin levelincrease occurs prior to development of subclinical structural vascular alterationsin diabetic adolescents. J. Clin. Res. Pediatr. Endocrinol. 8, 313-320.

Ermini, L., Ausman, J., Melland-Smith, M., Yeganeh, B., Rolfo, A., Litvack, M. L.,Todros, T., Letarte, M., Post, M. and Caniggia, I. (2017). A single sphingomyelinspecies promotes exosomal release of endoglin into the maternal circulation inpreeclampsia. Sci. Rep. 7, 12172.

Fernandez-Lopez, A., Garrido-Martin, E. M., Sanz-Rodriguez, F., Pericacho, M.,Rodriguez-Barbero, A., Eleno, N., Lopez-Novoa, J. M., Düwell, A., Vega,M. A., Bernabeu, C. et al. (2007). Gene expression fingerprinting for humanhereditary hemorrhagic telangiectasia. Hum. Mol. Genet. 16, 1515-1533.

Gallardo-Vara, E., Blanco, F. J., Roqué, M., Friedman, S. L., Suzuki, T., Botella,L. M. andBernabeu, C. (2016). Transcription factor KLF6 upregulates expressionof metalloprotease MMP14 and subsequent release of soluble endoglin duringvascular injury. Angiogenesis 19, 155-171.

Gougos, A. and Letarte, M. (1990). Primary structure of endoglin, an RGD-containing glycoprotein of human endothelial cells. J. Biol. Chem. 265,8361-8364.

Gregory, A. L., Xu, G., Sotov, V. and Letarte, M. (2014). Review: the enigmatic roleof endoglin in the placenta. Placenta 35 Suppl., S93-S99.

Guerrero-Esteo, M., Sánchez-Elsner, T., Letamendia, A. and Bernabéu, C.(2002). Extracellular and cytoplasmic domains of endoglin interact with thetransforming growth factor-beta receptors I and II. J. Biol. Chem. 277,29197-29209.

Hawinkels, L. J. A. C., Kuiper, P., Wiercinska, E., Verspaget, H. W., Liu, Z.,Pardali, E., Sier, C. F. M. and ten Dijke, P. (2010). Matrix metalloproteinase-14(MT1-MMP)-mediated endoglin shedding inhibits tumor angiogenesis. CancerRes. 70, 4141-4150.

Ikemoto, T., Hojo, Y., Kondo, H., Takahashi, N., Hirose, M., Nishimura, Y.,Katsuki, T., Shimada, K. and Kario, K. (2012). Plasma endoglin as a marker topredict cardiovascular events in patients with chronic coronary artery diseases.Heart Vessels 27, 344-351.

Jeong, H.-W., Hernández-Rodrıǵuez, B., Kim, J. M., Kim, K.-P., Enriquez-Gasca, R., Yoon, J., Adams, S., Schöler, H. R., Vaquerizas, J. M. and Adams,R. H. (2017). Transcriptional regulation of endothelial cell behavior duringsprouting angiogenesis. Nat. Commun. 8, 726.

Jerkic, M. and Letarte, M. (2015). Increased endothelial cell permeability inendoglin-deficient cells. FASEB J. 29, 3678-3688.

Jezkova, K., Rathouska, J., Nemeckova, I., Fikrova, P., Dolezelova, E.,Varejckova, M., Vitverova, B., Tysonova, K., Serwadczak, A., Buczek, E.et al. (2016). High levels of soluble endoglin induce a proinflammatory andoxidative-stress phenotype associated with preserved NO-dependentvasodilatation in aortas from mice fed a high-fat diet. J. Vasc. Res. 53, 149-162.

Jin, Y., Muhl, L., Burmakin, M., Wang, Y., Duchez, A.-C., Betsholtz, C., Arthur,H. M. and Jakobsson, L. (2017). Endoglin prevents vascular malformation byregulating flow-induced cell migration and specification through VEGFR2signalling. Nat. Cell. Biol. 19, 639-652.

Johnson, D. W., Berg, J. N., Baldwin, M. A., Gallione, C. J., Marondel, I., Yoon,S.-J., Stenzel, T. T., Speer, M., Pericak-Vance, M. A., Diamond, A. et al. (1996).Mutations in the activin receptor-like kinase 1 gene in hereditary haemorrhagictelangiectasia type 2. Nat. Genet. 13, 189-195.

Kumar, S., Pan, C. C., Bloodworth, J. C., Nixon, A. B., Theuer, C., Hoyt, D. G.and Lee, N. Y. (2014). Antibody-directed coupling of endoglin and MMP-14 is akey mechanism for endoglin shedding and deregulation of TGF-β signaling.Oncogene 33, 3970-3079.

Lebrin, F., Goumans, M.-J., Jonker, L., Carvalho, R. L. C., Valdimarsdottir, G.,Thorikay, M., Mummery, C., Arthur, H. M. and ten Dijke, P. (2004). Endoglinpromotes endothelial cell proliferation and TGF-beta/ALK1 signal transduction.EMBO J. 23, 4018-4028.

Lebrin, F., Srun, S., Raymond, K., Martin, S., van den Brink, S., Freitas, C.,Bréant, C., Mathivet, T., Larrivée, B., Thomas, J.-L. et al. (2010). Thalidomidestimulates vessel maturation and reduces epistaxis in individuals with hereditaryhemorrhagic telangiectasia. Nat. Med. 16, 420-428.

Letarte, M., McDonald, M.-L., Li, C., Kathirkamathamby, K., Vera, S., Pece-Barbara, N. and Kumar, S. (2005). Reduced endothelial secretion and plasmalevels of transforming growth factor-beta1 in patients with hereditary hemorrhagictelangiectasia type 1. Cardiovasc. Res. 68, 155-164.

Levet, S., Ciais, D., Merdzhanova, G., Mallet, C., Zimmers, T. A., Lee, S.-J.,Navarro, F. P., Texier, I., Feige, J.-J., Bailly, S. et al. (2013). Bonemorphogenetic protein 9 (BMP9) controls lymphatic vessel maturation andvalve formation. Blood 122, 598-607.

Li, C. G., Bethell, H., Wilson, P. B., Bhatnagar, D., Walker, M. G. and Kumar, S.(2000a). The significance of CD105, TGFbeta and CD105/TGFbeta complexes incoronary artery disease. Atherosclerosis 152, 249-256.

Li, C., Guo, B., Wilson, P. B., Stewart, A., Byrne, G., Bundred, N. and Kumar, S.(2000b). Plasma levels of soluble CD105 correlatewith metastasis in patients withbreast cancer. Int. J. Cancer 89, 122-126.

Li, C., Guo, B., Ding, S., Rius, C., Langa, C., Kumar, P., Bernabeu, C. andKumar, S. (2003). TNF alpha down-regulates CD105 expression in vascularendothelial cells: a comparative study with TGF beta 1. Anticancer Res. 23,1189-1196.

Li, W., Salmon, R. M., Jiang, H. and Morrell, N. W. (2016). Regulation of the ALK1ligands, BMP9 and BMP10. Biochem. Soc. Trans. 44, 1135-1141.

Liu, Y., Tian, H., Blobe, G. C., Theuer, C. P., Hurwitz, H. I. andNixon, A. B. (2014).Effects of the combination of TRC105 and bevacizumab on endothelial cellbiology. Invest. New Drugs 32, 851-859.

Llorca, O., Trujillo, A., Blanco, F. J. and Bernabeu, C. (2007). Structural model ofhuman endoglin, a transmembrane receptor responsible for hereditaryhemorrhagic telangiectasia. J. Mol. Biol. 365, 694-705.

López-Novoa, J. M. and Bernabeu, C. (2010). The physiological role of endoglin inthe cardiovascular system. Am. J. Physiol. Heart Circ. Physiol. 299, H959-H974.

Mahmoud, M., Allinson, K. R., Zhai, Z., Oakenfull, R., Ghandi, P., Adams, R. H.,Fruttiger, M. and Arthur, H. M. (2010). Pathogenesis of arteriovenousmalformations in the absence of endoglin. Circ. Res. 106, 1425-1433.

Mahmoud, M., Upton, P. D. and Arthur, H. M. (2011). Angiogenesis regulation byTGFβ signalling: clues from an inherited vascular disease. Biochem. Soc. Trans.39, 1659-1666.

Malik, R. A., Li, C., Aziz, W., Olson, J. A., Vohra, A., McHardy, K. C., Forrester,J. V., Boulton, A. J. M., Wilson, P. B., Liu, D. et al. (2005). Elevated plasmaCD105 and vitreous VEGF levels in diabetic retinopathy. J. Cell. Mol. Med. 9,692-697.

McAllister, K. A., Grogg, K. M., Johnson, D. W., Gallione, C. J., Baldwin, M. A.,Jackson, C. E., Helmbold, E. A., Markel, D. S., McKinnon, W. C., Murrell, J.et al. (1994). Endoglin, a TGF-beta binding protein of endothelial cells, is the genefor hereditary haemorrhagic telangiectasia type 1. Nat. Genet. 8, 345-351.

Mitrofan, C.-G., Appleby, S. L., Nash, G. B., Mallat, Z., Chilvers, E. R., Upton,P. D. and Morrell, N. W. (2017). Bone morphogenetic protein 9 (BMP9) andBMP10 enhance tumor necrosis factor-α-induced monocyte recruitment to thevascular endothelium mainly via activin receptor-like kinase 2. J. Biol. Chem. 292,13714-13726.

Nún ̃ez-Gómez, E., Pericacho, M., Ollauri-Ibán ̃ez, C., Bernabéu, C. and López-Novoa, J. M. (2017). The role of endoglin in post-ischemic revascularization.Angiogenesis 20, 1-24.

Oujo, B., Perez-Barriocanal, F., Bernabeu, C. and Lopez-Novoa, J. M. (2013).Membrane and soluble forms of endoglin in preeclampsia. Curr. Mol. Med. 13,1345-1357.

Paauwe, M., Heijkants, R. C., Oudt, C. H., van Pelt, G. W., Cui, C., Theuer, C. P.,Hardwick, J. C. H., Sier, C. F. M. and Hawinkels, L. J. A. C. (2016). Endoglin

10

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm034397. doi:10.1242/dmm.034397

Disea

seModels&Mechan

isms

http://dx.doi.org/10.3389/fgene.2015.00115http://dx.doi.org/10.3389/fgene.2015.00115http://dx.doi.org/10.3389/fgene.2015.00115http://dx.doi.org/10.1074/jbc.M111.260133http://dx.doi.org/10.1074/jbc.M111.260133http://dx.doi.org/10.1074/jbc.M111.260133http://dx.doi.org/10.1074/jbc.M111.260133http://dx.doi.org/10.1074/jbc.M111.260133http://dx.doi.org/10.1007/s12020-015-0565-2http://dx.doi.org/10.1007/s12020-015-0565-2http://dx.doi.org/10.1007/s12020-015-0565-2http://dx.doi.org/10.1007/s12020-015-0565-2http://dx.doi.org/10.1002/ana.21710http://dx.doi.org/10.1002/ana.21710http://dx.doi.org/10.1002/ana.21710http://dx.doi.org/10.1073/pnas.1306074110http://dx.doi.org/10.1073/pnas.1306074110http://dx.doi.org/10.1073/pnas.1306074110http://dx.doi.org/10.1073/pnas.1306074110http://dx.doi.org/10.1111/j.1582-4934.2008.00156.xhttp://dx.doi.org/10.1111/j.1582-4934.2008.00156.xhttp://dx.doi.org/10.1111/j.1582-4934.2008.00156.xhttp://dx.doi.org/10.1111/j.1582-4934.2008.00156.xhttp://dx.doi.org/10.1111/j.1582-4934.2008.00156.xhttp://dx.doi.org/10.1161/CIRCRESAHA.117.309930http://dx.doi.org/10.1161/CIRCRESAHA.117.309930http://dx.doi.org/10.1161/CIRCRESAHA.117.309930http://dx.doi.org/10.1042/BST20160093http://dx.doi.org/10.1042/BST20160093http://dx.doi.org/10.1042/BST20160093http://dx.doi.org/10.3389/fgene.2015.00114http://dx.doi.org/10.3389/fgene.2015.00114http://dx.doi.org/10.1001/jama.2012.250http://dx.doi.org/10.1001/jama.2012.250http://dx.doi.org/10.1001/jama.2012.250http://dx.doi.org/10.1001/jama.2012.250http://dx.doi.org/10.1001/jama.2016.11387http://dx.doi.org/10.1001/jama.2016.11387http://dx.doi.org/10.1001/jama.2016.11387http://dx.doi.org/10.1001/jama.2016.11387http://dx.doi.org/10.4274/jcrpe.2906http://dx.doi.org/10.4274/jcrpe.2906http://dx.doi.org/10.4274/jcrpe.2906http://dx.doi.org/10.4274/jcrpe.2906http://dx.doi.org/10.1038/s41598-017-12491-4http://dx.doi.org/10.1038/s41598-017-12491-4http://dx.doi.org/10.1038/s41598-017-12491-4http://dx.doi.org/10.1038/s41598-017-12491-4http://dx.doi.org/10.1093/hmg/ddm069http://dx.doi.org/10.1093/hmg/ddm069http://dx.doi.org/10.1093/hmg/ddm069http://dx.doi.org/10.1093/hmg/ddm069http://dx.doi.org/10.1007/s10456-016-9495-8http://dx.doi.org/10.1007/s10456-016-9495-8http://dx.doi.org/10.1007/s10456-016-9495-8http://dx.doi.org/10.1007/s10456-016-9495-8http://dx.doi.org/10.1016/j.placenta.2013.10.020http://dx.doi.org/10.1016/j.placenta.2013.10.020http://dx.doi.org/10.1074/jbc.M111991200http://dx.doi.org/10.1074/jbc.M111991200http://dx.doi.org/10.1074/jbc.M111991200http://dx.doi.org/10.1074/jbc.M111991200http://dx.doi.org/10.1158/0008-5472.CAN-09-4466http://dx.doi.org/10.1158/0008-5472.CAN-09-4466http://dx.doi.org/10.1158/0008-5472.CAN-09-4466http://dx.doi.org/10.1158/0008-5472.CAN-09-4466http://dx.doi.org/10.1007/s00380-011-0163-zhttp://dx.doi.org/10.1007/s00380-011-0163-zhttp://dx.doi.org/10.1007/s00380-011-0163-zhttp://dx.doi.org/10.1007/s00380-011-0163-zhttp://dx.doi.org/10.1038/s41467-017-00738-7http://dx.doi.org/10.1038/s41467-017-00738-7http://dx.doi.org/10.1038/s41467-017-00738-7http://dx.doi.org/10.1038/s41467-017-00738-7http://dx.doi.org/10.1096/fj.14-269258http://dx.doi.org/10.1096/fj.14-269258http://dx.doi.org/10.1159/000448996http://dx.doi.org/10.1159/000448996http://dx.doi.org/10.1159/000448996http://dx.doi.org/10.1159/000448996http://dx.doi.org/10.1159/000448996http://dx.doi.org/10.1038/ncb3534http://dx.doi.org/10.1038/ncb3534http://dx.doi.org/10.1038/ncb3534http://dx.doi.org/10.1038/ncb3534http://dx.doi.org/10.1038/ng0696-189http://dx.doi.org/10.1038/ng0696-189http://dx.doi.org/10.1038/ng0696-189http://dx.doi.org/10.1038/ng0696-189http://dx.doi.org/10.1038/onc.2013.386http://dx.doi.org/10.1038/onc.2013.386http://dx.doi.org/10.1038/onc.2013.386http://dx.doi.org/10.1038/onc.2013.386http://dx.doi.org/10.1038/sj.emboj.7600386http://dx.doi.org/10.1038/sj.emboj.7600386http://dx.doi.org/10.1038/sj.emboj.7600386http://dx.doi.org/10.1038/sj.emboj.7600386http://dx.doi.org/10.1038/nm.2131http://dx.doi.org/10.1038/nm.2131http://dx.doi.org/10.1038/nm.2131http://dx.doi.org/10.1038/nm.2131http://dx.doi.org/10.1016/j.cardiores.2005.04.028http://dx.doi.org/10.1016/j.cardiores.2005.04.028http://dx.doi.org/10.1016/j.cardiores.2005.04.028http://dx.doi.org/10.1016/j.cardiores.2005.04.028http://dx.doi.org/10.1182/blood-2012-12-472142http://dx.doi.org/10.1182/blood-2012-12-472142http://dx.doi.org/10.1182/blood-2012-12-472142http://dx.doi.org/10.1182/blood-2012-12-472142http://dx.doi.org/10.1016/S0021-9150(99)00476-1http://dx.doi.org/10.1016/S0021-9150(99)00476-1http://dx.doi.org/10.1016/S0021-9150(99)00476-1http://dx.doi.org/10.1002/(SICI)1097-0215(20000320)89:2%3C122::AID-IJC4%3E3.0.CO;2-Mhttp://dx.doi.org/10.1002/(SICI)1097-0215(20000320)89:2%3C122::AID-IJC4%3E3.0.CO;2-Mhttp://dx.doi.org/10.1002/(SICI)1097-0215(20000320)89:2%3C122::AID-IJC4%3E3.0.CO;2-Mhttp://dx.doi.org/10.1042/BST20160083http://dx.doi.org/10.1042/BST20160083http://dx.doi.org/10.1007/s10637-014-0129-yhttp://dx.doi.org/10.1007/s10637-014-0129-yhttp://dx.doi.org/10.1007/s10637-014-0129-yhttp://dx.doi.org/10.1016/j.jmb.2006.10.015http://dx.doi.org/10.1016/j.jmb.2006.10.015http://dx.doi.org/10.1016/j.jmb.2006.10.015http://dx.doi.org/10.1152/ajpheart.01251.2009http://dx.doi.org/10.1152/ajpheart.01251.2009http://dx.doi.org/10.1161/CIRCRESAHA.109.211037http://dx.doi.org/10.1161/CIRCRESAHA.109.211037http://dx.doi.org/10.1161/CIRCRESAHA.109.211037http://dx.doi.org/10.1042/BST20110664http://dx.doi.org/10.1042/BST20110664http://dx.doi.org/10.1042/BST20110664http://dx.doi.org/10.1111/j.1582-4934.2005.tb00499.xhttp://dx.doi.org/10.1111/j.1582-4934.2005.tb00499.xhttp://dx.doi.org/10.1111/j.1582-4934.2005.tb00499.xhttp://dx.doi.org/10.1111/j.1582-4934.2005.tb00499.xhttp://dx.doi.org/10.1038/ng1294-345http://dx.doi.org/10.1038/ng1294-345http://dx.doi.org/10.1038/ng1294-345http://dx.doi.org/10.1038/ng1294-345http://dx.doi.org/10.1074/jbc.M117.778506http://dx.doi.org/10.1074/jbc.M117.778506http://dx.doi.org/10.1074/jbc.M117.778506http://dx.doi.org/10.1074/jbc.M117.778506http://dx.doi.org/10.1074/jbc.M117.778506http://dx.doi.org/10.1007/s10456-016-9535-4http://dx.doi.org/10.1007/s10456-016-9535-4http://dx.doi.org/10.1007/s10456-016-9535-4http://dx.doi.org/10.2174/15665240113139990058http://dx.doi.org/10.2174/15665240113139990058http://dx.doi.org/10.2174/15665240113139990058http://dx.doi.org/10.1038/onc.2015.509http://dx.doi.org/10.1038/onc.2015.509

targeting inhibits tumor angiogenesis and metastatic spread in breast cancer.Oncogene 35, 4069-4079.

Potente, M. and Carmeliet, P. (2017). The link between angiogenesis andendothelial metabolism. Annu. Rev. Physiol. 79, 43-66.

Rathouska, J., Jezkova, K., Nemeckova, I. and Nachtigal, P. (2015). Solubleendoglin, hypercholesterolemia and endothelial dysfunction.Atherosclerosis 243,383-388.

Rathouska, J., Fikrova, P., Mrkvicova, A., Blazickova, K., Varejckova, M.,Dolezelova, E., Nemeckova, I., Vitverova, B., Peslova, L., Gallardo-Vara, E.et al. (2017). High soluble endoglin levels do not induce changes in structuralparameters of mouse heart. Heart Vessels 32, 1013-1024.

Redgrave, R. E., Tual-Chalot, S., Davison, B. J., Singh, E., Hall, D., Amirrasouli,M. M., Gilchrist, D., Medvinsky, A. and Arthur, H. M. (2017). Cardiosphere-derived cells require endoglin for paracrine-mediated angiogenesis. Stem CellRep. 8, 1287-1298.

Roman, B. L. and Hinck, A. P. (2017). ALK1 signaling in development and disease:new paradigms. Cell. Mol. Life Sci. 74, 4539-4560.

Rosen, L. S., Gordon, M. S., Robert, F. and Matei, D. E. (2014). Endoglin fortargeted cancer treatment. Curr. Oncol. Rep. 16, 365.

Rossi, E., Sanz-Rodriguez, F., Eleno, N., Düwell, A., Blanco, F. J., Langa, C.,Botella, L. M., Caban ̃as, C., Lopez-Novoa, J. M. and Bernabeu, C. (2013).Endothelial endoglin is involved in inflammation: role in leukocyte adhesion andtransmigration. Blood 121, 403-415.

Rossi, E., Lopez-Novoa, J. M. and Bernabeu, C. (2015). Endoglin involvement inintegrin-mediated cell adhesion as a putative pathogenic mechanism in hereditaryhemorrhagic telangiectasia type 1 (HHT1). Front. Genet. 5, 457.

Rossi, E., Smadja, D. M., Boscolo, E., Langa, C., Arevalo, M. A., Pericacho, M.,Gamella-Pozuelo, L., Kauskot, A., Botella, L. M., Gaussem, P. et al. (2016).Endoglin regulates mural cell adhesion in the circulatory system. Cell. Mol. LifeSci. 73, 1715-1739.

Ruiz-Llorente, L., Gallardo-Vara, E., Rossi, E., Smadja, D. M., Botella, L. M. andBernabeu, C. (2017). Endoglin and alk1 as therapeutic targets for hereditaryhemorrhagic telangiectasia. Expert Opin. Ther. Targets 21, 933-947.

Ruiz-Remolina, L., Ollauri-Ibán ̃ez, C., Pérez-Roque, L., Nún ̃ez-Gómez, E.,Pérez-Barriocanal, F., López-Novoa, J. M., Pericacho, M. and Rodrıǵuez-Barbero, A. (2017). Circulating soluble endoglin modifies the inflammatoryresponse in mice. PLoS ONE 12, e0188204.

Saita, E., Miura, K., Suzuki-Sugihara, N., Miyata, K., Ikemura, N., Ohmori, R.,Ikegami, Y., Kishimoto, Y., Kondo, K. and Momiyama, Y. (2017). Plasmasoluble endoglin levels are inversely associated with the severity of coronaryatherosclerosis-brief report. Arterioscler. Thromb. Vasc. Biol. 37, 49-52.

Saito, T., Bokhove, M., Croci, R., Zamora-Caballero, S., Han, L., Letarte, M., deSanctis, D. and Jovine, L. (2017). Structural basis of the human endoglin-BMP9interaction: insights into BMP signaling and HHT1. Cell Rep. 19, 1917-1928.

Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch,T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B. et al. (2012). Fiji: anopen-source platform for biological-image analysis. Nat. Methods 9, 676-682.

Schneider, C. A., Rasband,W. S. and Eliceiri, K. W. (2012). NIH Image to ImageJ:25 years of image analysis. Nat. Methods 9, 671-675.

Selvam, S., Kumar, T. and Fruttiger, M. (2018). Retinal vasculature developmentin health and disease. Prog. Retin. Eye Res. 63, 1-19.

Shovlin, C. L. (2010). Hereditary haemorrhagic telangiectasia: pathophysiology,diagnosis and treatment. Blood Rev. 24, 203-219.

Sugden, W. W., Meissner, R., Aegerter-Wilmsen, T., Tsaryk, R., Leonard, E. V.,Bussmann, J., Hamm, M. J., Herzog, W., Jin, Y., Jakobsson, L. et al. (2017).Endoglin controls blood vessel diameter through endothelial cell shape changesin response to haemodynamic cues. Nat. Cell Biol. 19, 653-665.

Sunderland, N. S., Thomson, S. E., Heffernan, S. J., Lim, S., Thompson, J.,Ogle, R., McKenzie, P., Kirwan, P. J., Makris, A. and Hennessy, A. (2011).Tumor necrosis factor α induces a model of preeclampsia in pregnant baboons(Papio hamadryas). Cytokine 56, 192-199.

ten Dijke, P. and Arthur, H. M. (2007). Extracellular control of TGFbeta signalling invascular development and disease. Nat. Rev. Mol. Cell. Biol. 8, 857-869.

Tillet, E. and Bailly, S. (2015). Emerging roles of BMP9 and BMP10 in hereditaryhemorrhagic telangiectasia. Front. Genet. 5, 456.

Tual-Chalot, S., Allinson, K. R., Fruttiger, M., Arthur, H. M. (2013). Whole mountimmunofluorescent staining of the neonatal mouse retina to investigateangiogenesis in vivo. J. Vis. Exp. e50546.

Tual-Chalot, S., Oh, S. P. and Arthur, H. M. (2015). Mouse models of hereditaryhemorrhagic telangiectasia: recent advances and future challenges. Front.Genet. 6, 25.

Valbuena-Dıéz, A. C., Blanco, F. J., Oujo, B., Langa, C., Gonzalez-Nun ̃ez, M.,Llano, E., Pendas, A. M., Dıáz, M., Castrillo, A., Lopez-Novoa, J. M. et al.(2012). Oxysterol-induced soluble endoglin release and its involvement inhypertension. Circulation 126, 2612-2624.

van Laake, L. W., van den Driesche, S., Post, S., Feijen, A., Jansen, M. A.,Driessens, M. H., Mager, J. J., Snijder, R. J., Westermann, C. J. J.,Doevendans, P. A. et al. (2006). Endoglin has a crucial role in blood cell-mediated vascular repair. Circulation 114, 2288-2297.

Varejckova, M., Gallardo-Vara, E., Vicen, M., Vitverova, B., Fikrova, P.,Dolezelova, E., Rathouska, J., Prasnicka, A., Blazickova, K., Micuda, S.et al. (2017). Soluble endoglin modulates the pro-inflammatory mediators NF-κBand IL-6 in cultured human endothelial cells. Life Sci. 175, 52-60.

Venkatesha, S., Toporsian, M., Lam, C., Hanai, J., Mammoto, T., Kim, Y. M.,Bdolah, Y., Lim, K.-H., Yuan, H.-T., Libermann, T. A. et al. (2006). Solubleendoglin contributes to the pathogenesis of preeclampsia.Nat. Med. 12, 642-649.

Wang, Y., Chen, Q., Zhao,M.,Walton, K., Harrison, C. andNie, G. (2017). Multiplesoluble TGF-β receptors in addition to soluble endoglin are elevated inpreeclamptic serum and they synergistically inhibit TGF-β signaling. J. Clin.Endocrinol. Metab. 102, 3065-3074.

Wetzel-Strong, S. E., Detter, M. R. andMarchuk, D. A. (2017). The pathobiology ofvascular malformations: insights from human and model organism genetics.J. Pathol. 241, 281-293.

Wooderchak-Donahue, W. L., McDonald, J., O’Fallon, B., Upton, P. D., Li, W.,Roman, B. L., Young, S., Plant, P., Fülöp, G. T., Langa, C. et al. (2013). BMP9mutations cause a vascular-anomaly syndrome with phenotypic overlap withhereditary hemorrhagic telangiectasia. Am. J. Hum. Genet. 93, 530-537.

Young, K., Conley, B., Romero, D., Tweedie, E., O’Neill, C., Pinz, I., Brogan, L.,Lindner, V., Liaw, L. and Vary, C. P. H. (2012). BMP9 regulates endoglin-dependent chemokine responses in endothelial cells. Blood 120, 4263-4273.

Zhang, R., Han, Z., Degos, V., Shen, F., Choi, E.-J., Sun, Z., Kang, S., Wong, M.,Zhu, W., Zhan, L. et al. (2016). Persistent infiltration and pro-inflammatorydifferentiation of monocytes cause unresolved inflammation in brain arteriovenousmalformation. Angiogenesis 19, 451-461.

11

RESEARCH ARTICLE Disease Models & Mechanisms (2018) 11, dmm034397. doi:10.1242/dmm.034397

Disea

seModels&Mechan

isms

http://dx.doi.org/10.1038/onc.2015.509http://dx.doi.org/10.1038/onc.2015.509http://dx.doi.org/10.1146/annurev-physiol-021115-105134http://dx.doi.org/10.1146/annurev-physiol-021115-105134http://dx.doi.org/10.1016/j.atherosclerosis.2015.10.003http://dx.doi.org/10.1016/j.atherosclerosis.2015.10.003http://dx.doi.org/10.1016/j.atherosclerosis.2015.10.003http://dx.doi.org/10.1007/s00380-017-0976-5http://dx.doi.org/10.1007/s00380-017-0976-5http://dx.doi.org/10.1007/s00380-017-0976-5http://dx.doi.org/10.1007/s00380-017-0976-5http://dx.doi.org/10.1016/j.stemcr.2017.04.015http://dx.doi.org/10.1016/j.stemcr.2017.04.015http://dx.doi.org/10.1016/j.stemcr.2017.04.015http://dx.doi.org/10.1016/j.stemcr.2017.04.015http://dx.doi.org/10.1007/s00018-017-2636-4http://dx.doi.org/10.1007/s00018-017-2636-4http://dx.doi.org/10.1007/s11912-013-0365-xhttp://dx.doi.org/10.1007/s11912-013-0365-xhttp://dx.doi.org/10.1182/blood-2012-06-435347http://dx.doi.org/10.1182/blood-2012-06-435347http://dx.doi.org/10.1182/blood-2012-06-435347http://dx.doi.org/10.1182/blood-2012-06-435347http://dx.doi.org/10.3389/fgene.2014.00457http://dx.doi.org/10.3389/fgene.2014.00457http://dx.doi.org/10.3389/fgene.2014.00457http://dx.doi.org/10.1007/s00018-015-2099-4http://dx.doi.org/10.1007/s00018-015-2099-4http://dx.doi.org/10.1007/s00018-015-2099-4http://dx.doi.org/10.1007/s00018-015-2099-4http://dx.doi.org/10.1080/14728222.2017.1365839http://dx.doi.org/10.1080/14728222.2017.1365839http://dx.doi.org/10.1080/14728222.2017.1365839http://dx.doi.org/10.1371/journal.pone.0188204http://dx.doi.org/10.1371/journal.pone.0188204http://dx.doi.org/10.1371/journal.pone.0188204http://dx.doi.org/10.1371/journal.pone.0188204http://dx.doi.org/10.1161/ATVBAHA.116.308494http://dx.doi.org/10.1161/ATVBAHA.116.308494http://dx.doi.org/10.1161/ATVBAHA.116.308494http://dx.doi.org/10.1161/ATVBAHA.116.308494http://dx.doi.org/10.1016/j.celrep.2017.05.011http://dx.doi.org/10.1016/j.celrep.2017.05.011http://dx.doi.org/10.1016/j.celrep.2017.05.011http://dx.doi.org/10.1038/nmeth.2019http://dx.doi.org/10.1038/nmeth.2019http://dx.doi.org/10.1038/nmeth.2019http://dx.doi.org/10.1038/nmeth.2089http://dx.doi.org/10.1038/nmeth.2089http://dx.doi.org/10.1016/j.preteyeres.2017.11.001http://dx.doi.org/10.1016/j.preteyeres.2017.11.001http://dx.doi.org/10.1016/j.blre.2010.07.001http://dx.doi.org/10.1016/j.blre.2010.07.001http://dx.doi.org/10.1038/ncb3528http://dx.doi.org/10.1038/ncb3528http://dx.doi.org/10.1038/ncb3528http://dx.doi.org/10.1038/ncb3528http://dx.doi.org/10.1016/j.cyto.2011.06.003http://dx.doi.org/10.1016/j.cyto.2011.06.003http://dx.doi.org/10.1016/j.cyto.2011.06.003http://dx.doi.org/10.1016/j.cyto.2011.06.003http://dx.doi.org/10.1038/nrm2262http://dx.doi.org/10.1038/nrm2262http://dx.doi.org/10.3389/fgene.2014.00456http://dx.doi.org/10.3389/fgene.2014.00456http://dx.doi.org/10.3791/50546http://dx.doi.org/10.3791/50546http://dx.doi.org/10.3791/50546http://dx.doi.org/10.3389/fgene.2015.00025http://dx.doi.org/10.3389/fgene.2015.00025http://dx.doi.org/10.3389/fgene.2015.00025http://dx.doi.org/10.1161/CIRCULATIONAHA.112.101261http://dx.doi.org/10.1161/CIRCULATIONAHA.112.101261http://dx.doi.org/10.1161/CIRCULATIONAHA.112.101261http://dx.doi.org/10.1161/CIRCULATIONAHA.112.101261http://dx.doi.org/10.1161/CIRCULATIONAHA.106.639161http://dx.doi.org/10.1161/CIRCULATIONAHA.106.639161http://dx.doi.org/10.1161/CIRCULATIONAHA.106.639161http://dx.doi.org/10.1161/CIRCULATIONAHA.106.639161http://dx.doi.org/10.1016/j.lfs.2017.03.014http://dx.doi.org/10.1016/j.lfs.2017.03.014http://dx.doi.org/10.1016/j.lfs.2017.03.014http://dx.doi.org/10.1016/j.lfs.2017.03.014http://dx.doi.org/10.1038/nm1429http://dx.doi.org/10.1038/nm1429http://dx.doi.org/10.1038/nm1429http://dx.doi.org/10.1210/jc.2017-01150http://dx.doi.org/10.1210/jc.2017-01150http://dx.doi.org/10.1210/jc.2017-01150http://dx.doi.org/10.1210/jc.2017-01150http://dx.doi.org/10.1002/path.4844http://dx.doi.org/10.1002/path.4844http://dx.doi.org/10.1002/path.4844http://dx.doi.org/10.1016/j.ajhg.2013.07.004http://dx.doi.org/10.1016/j.ajhg.2013.07.004http://dx.doi.org/10.1016/j.ajhg.2013.07.004http://dx.doi.org/10.1016/j.ajhg.2013.07.004http://dx.doi.org/10.1182/blood-2012-07-440784http://dx.doi.org/10.1182/blood-2012-07-440784http://dx.doi.org/10.1182/blood-2012-07-440784http://dx.doi.org/10.1007/s10456-016-9519-4http://dx.doi.org/10.1007/s10456-016-9519-4http://dx.doi.org/10.1007/s10456-016-9519-4http://dx.doi.org/10.1007/s10456-016-9519-4