Sulindac metabolites decrease cerebrovascularmalformations in CCM3-knockout miceLuca Bravia, Noemi Rudinia, Roberto Cuttanoa, Costanza Giampietroa,b, Luigi Maddalunoa, Luca Ferrarinia,Ralf H. Adamsc, Monica Coradaa, Gwenola Bouldayd, Elizabeth Tournier-Lasserved, Elisabetta Dejanaa,e,1,2,and Maria Grazia Lampugnania,f,1,2

aFondazione Italiana per la Ricerca sul Cancro (FIRC) Institute of Molecular Oncology Fondazione, 20139 Milan, Italy; bDepartment of Biosciences, Universityof Milan, 20136 Milan, Italy; cFaculty of Medicine, University of Münster, D-48149 Münster, Germany; dInstitut National de la Santé et de la RechercheMédicale, UMR-S 740, 75010 Paris, France; eDepartment of Immunology, Genetics and Pathology, Uppsala University, 75185 Uppsala, Sweden; and fMarioNegri Institute for Pharmacological Research, 20156 Milan, Italy

Edited by Harry C. Dietz, The Johns Hopkins University School of Medicine, Baltimore, MD, and approved May 28, 2015 (received for review January 26, 2015)

Cerebral cavernous malformation (CCM) is a disease of the centralnervous system causing hemorrhage-prone multiple lumen vascularmalformations and very severe neurological consequences. At pre-sent, the only recommended treatment of CCM is surgical. Becausesurgery is often not applicable, pharmacological treatment would behighly desirable. We describe here a murine model of the diseasethat develops after endothelial-cell–selective ablation of the CCM3gene. We report an early, cell-autonomous, Wnt-receptor–indepen-dent stimulation of β-catenin transcription activity in CCM3-deficientendothelial cells both in vitro and in vivo and a triggering of a β-catenin–driven transcription program that leads to endothelial-to-mesenchymal transition. TGF-β/BMP signaling is then required for theprogression of the disease. We also found that the anti-inflammatorydrugs sulindac sulfide and sulindac sulfone, which attenuate β-catenintranscription activity, reduce vascular malformations in endothelialCCM3-deficient mice. This study opens previously unidentified per-spectives for an effective pharmacological therapy of intracranialvascular cavernomas.

cerebral cavernous malformation | endothelial cells | β-catenin |sulindac metabolites | vascular pathology

The vascular malformations that characterize the disease knownas cerebral cavernous malformation (CCM) are concentrated

in the central nervous system, and they typically show multiplelumens and vascular leakage (1). These abnormalities can result insevere neurological symptoms, including hemorrhagic stroke (2),and, to date, the only possible therapy is surgery (3). In humans,loss-of-function mutations in any one of three independent genesknown as cerebral cavernous malformation 1, 2, and 3 (CCM1,CCM2, and CCM3) are the cause of the genetic form of CCM (4).Similarly, in murine models, the vascular phenotype can bereproduced by endothelium-specific loss-of-function mutations ofany one of these three CCM-linked genes (5–7).We have recently reported (7) that TGF-β/BMP signaling is ac-

tivated after ablation of CCM1, CCM2, or CCM3 and induces en-dothelial-to-mesenchymal transition (EndMT) that plays a crucialrole in the development of vascular malformations. Nevertheless, thesequence of signaling responses elicited by ablation of CCM genesstill remains to be defined. Inhibitors of the TGF-β/BMP-signalingpathways reduce the number and size of the malformations, butnot completely (7), suggesting that other signaling pathways maybe implicated.TheWnt/β-catenin pathway, in synergy with TGF-β signaling (8),

is responsible for the EndMT switch of endothelial cells giving riseto the heart cushion in the embryo. In addition, the knockdown ofCCM1 and CCM3 expression in cultured aortic and artery endo-thelial cells promotes β-catenin signaling (9, 10), although no directlink with the in vivo model of the disease has been made. Activa-tion of canonical Wnt/β-catenin signaling is critical for brain vas-cularization and acquisition, by the microvasculature, of blood–brainbarrier properties (11–13). Endothelial signaling by β-catenin must

be tightly regulated: it is high during blood–brain barrier devel-opment, but it is rapidly abrogated postnatally (13). Long-lastinghigh levels of β-catenin signaling in the vasculature may cause strongalterations in vascular stability and lumen malformations (14).In the present study, we bring evidence that sustained β-catenin

signaling plays an important role in the EndMT and in the devel-opment of brain vascular malformations in CCM3-deficient mice.Activation of the β-catenin pathway is followed by TGF-β/BMPsignaling that supports further evolution of the vascular lesions.Importantly, pharmacological inhibitors of β-catenin signaling sig-nificantly reduce the development of the vascular malformationsin CCM3-deficient mice. This work introduces a previously un-identified therapeutic strategy to counteract the formation of brainvascular cavernomas in genetic patients.

ResultsTranscription Activity of β-Catenin Precedes TGF-β/BMP SignalingDuring the Development of CCM Vascular Malformations. The invivo mouse system used was generated through a cross of CCM3-flox/flox mice with Cdh5(PAC)-CreERT2 mice to obtain tamoxifen-inducible endothelial-cell–specific expression of Cre recombinaseand CCM3 gene recombination (CCM3-ECKO mice). These micewere then crossed with BAT-gal reporter mice, which show β-catenin–activated expression of nuclear β-galactosidase (β-gal). For details,see SI Appendix, Methods, and Fig. S1. CCM3 mutations occur with

Significance

Cerebral cavernous malformation (CCM) disease can lead tobrain hemorrhages, seizures, and paralysis. No pharmacologicaltherapy is currently available. Here we define, to our knowledgefor the first time in vivo, the sequence of molecular events thatlead to CCM vascular cavernomas. We found that β-catenin ac-tivation is the first trigger followed by TGF-β signaling, which, inturn, mediates the progression of the disease. We also show thatβ-catenin signaling is cell-autonomous and independent of Wnt-receptor activation. Most importantly, these studies promptedus to identify pharmacological agents that, by targeting the al-tered β-catenin signaling, limit the formation of brain vascularcavernomas in mice with CCM3 ablation in endothelial cells.These drugs are currently used in clinics for different pathologiesand may be repurposed for CCM therapy.

Author contributions: E.D. and M.G.L. designed research; L.B., N.R., R.C., C.G., L.M., L.F.,and M.C. performed research; R.H.A., G.B., and E.T.-L. contributed new reagents/analytictools; L.B. analyzed data; and E.D. and M.G.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1E.D. and M.G.L. contributed equally to this work.2To whom correspondence may be addressed. Email: mariagrazia.lampugnani@ifom.euor elisabetta.dejana@ifom.eu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1501352112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1501352112 PNAS | July 7, 2015 | vol. 112 | no. 27 | 8421–8426

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

May

31,

202

0

low frequency (about 10%) in genetic patients (15 and referencestherein) and determines the most severe CCM (16).As reported in Fig. 1A, Upper, we could observe a significantly

higher β-catenin transcription signal in the nuclei of endothelialcells in CCM3-ECKO mice in comparison with matched controls.This difference was detectable at early stages [3 day postnatal (dpn)]after induction of CCM3 recombination (at 1 dpn). In brain sectionsof CCM3-ECKO mice, endothelial cells with β-gal–positive nucleicould be found both in pseudonormal vessels and in cavernae of anysize (SI Appendix, Fig. S2). In contrast, phospho-Smad1 (p-Smad1)staining in separate sections (Fig. 1A, Lower) and in costaining forβ-gal (Fig. 1B) was not enhanced after CCM3 ablation in 3-dpnpups, but it was increased in 9-dpn pups (Fig. 1C). p-Smad1 wassignificantly high in middle-to-large lesions only (maximal di-ameter ≥50 μm in 9-dpn pups) (SI Appendix, Fig. S2). Ex-pression of stem-cell/EndMT markers (7) (Klf4, S100a4, andId1) was high in 3-dpn CCM3-ECKO pups (Fig. 2 A–D, singlepositive) and was concentrated in endothelial cells with β-gal–positive nuclei (Fig. 2D, colocalization). At 9 dpn, β-gal ex-pression in endothelial cells of CCM3-ECKO pups decreasedwhereas stem-cell/EndMT markers remained high (Fig. 2D).In cultured endothelial cells, acute abrogation of CCM3 by

siRNA induced early up-regulation of β-catenin target genes andstem-cell/EndMT markers at times when phosphorylation of Smad1was not yet enhanced (SI Appendix, Fig. S3 A–C). In contrast, last-ing CCM3 abrogation induced by CCM3-flox/flox recombination by

Adeno-Cre (SI Appendix, Methods) induced higher levels of bothp-Smad1 and p-Smad3 (SI Appendix, Fig. S3D). Furthermore, sustainedactivation of β-catenin signaling by Lef-ΔβCTA induced phos-phorylation of both Smad1 and Smad3 (SI Appendix, Fig. S3E),whereas acute stimulation with Wnt3a did not (SI Appendix, Fig.S3G). Taken together, these data suggest a temporal link be-tween β-catenin and TGF-β/BMP signaling.

β-Catenin Signaling in CCM3-Deficient Endothelial Cells Is ActivatedThrough Cell-Autonomous, Wnt-Receptor–Independent Mechanism.Recombination of the CCM3 gene in vitro in freshly isolatedbrain endothelial cells or in a lung endothelial cell line fromCCM3-flox/flox mice (SI Appendix) showed active β-catenin[dephosphorylated on Ser37 and Thr41 (17)] in the nucleus (byimmunofluorescence microscopy in SI Appendix, Fig. S4 A and B,KO, arrowheads; confirmed by cell fractionation in SI Appendix,Fig. S4 C and D). This effect paralleled a strong alteration ofadherens junctions with delocalization of both β-catenin andVE-cadherin (SI Appendix, Fig. S4 A and B).Enhanced β-catenin transcription activity in CCM3-deficient

endothelial cells was further confirmed by (i) increased expres-sion of β-catenin target genes abrogated by a dominant-negativemutant of TCF4 (dnTCF4) (18) (Fig. 3A) and (ii) activation ofTcf/Lef-dependent transcription of the exogenous luciferase geneas measured in Top/Fop Flash reporter assays (see controls inSI Appendix, Fig. S7B).

Fig. 1. Brain endothelial cells in CCM3-ECKO mice show enhanced β-catenin transcription activity earlier than activation of TGF-β/BMP signaling. (A) Brainsections from wild-type (WT) and CCM3-ECKO mice stained for β-gal (red, Upper), and p-Smad1 (red, Lower) in endothelial cells (Podocalyxin-positive, green)at early (3 dpn) and late (9 dpn) time points after CCM3 recombination (1 dpn). Blue, DAPI-stained nuclei. (B) Costaining for β-gal (red), p-Smad1 (green), andPodocalyxin (blue). (A and B) Arrows, β-gal– and p-Smad1–positive nuclei; empty arrows, p-Smad1–negative nuclei in endothelial cells of a vascular mal-formation (caverna in A) and telangiectasia in B) in 3-dpn pups. (Insets) Magnification of boxed areas. (Scale bars, 50 μm; Insets in A, 10 μm.) (C) Quantificationof β-gal–positive and p-Smad–positive endothelial cells in brain sections as in A and B. At least 450 nuclei were counted in 40 random fields at 63× mag-nification for each condition in samples from matched littermate pups in three independent experiments. *P < 0.01 versus indicated controls (t test).

8422 | www.pnas.org/cgi/doi/10.1073/pnas.1501352112 Bravi et al.

Dow

nloa

ded

by g

uest

on

May

31,

202

0

Furthermore, transcription of genes related to acquisition andmaintenance of the stem-cell/EndMT phenotype (19) [Klf4, Ly6a,S100a4, Id1, Cdh2, and Acta2 (20)] was significantly enhanced underbasal conditions in the CCM3-knockout endothelial cells (Fig. 3B)and inhibited by dnTCF4 (18) (Fig. 3B and SI Appendix, Fig. S8B).Activation of β-catenin–mediated transcription in CCM3-

knockout endothelial cells was cell-autonomous because (i) itwas observed in absence of exogenous Wnt; (ii) the porcupineinhibitors IWP2 and IWP12 (21, 22), as well as Dkk1, a compet-itor of Wnt coreceptor Lrp5/6 (21, 22), did not inhibit tran-scription of typical β-catenin targets (SI Appendix, Fig. S5 A–D);(iii) Lrp6 phosphorylation was not increased (SI Appendix, Fig.S5 E and F); and (iv) stimulation by exogenous Wnt3a did notinduce expression of stem-cell/EndMT markers, whereas the con-stitutively active form of β-catenin, Lef-ΔβCTA (23), did (SI Appendix,Fig. S5 G and H).Taken together, these data indicate that in CCM3-knockout

endothelial cells enhanced nuclear localization and transcrip-tion activity of β-catenin do not depend on a classical ligand–receptor interaction.We reported previously that silencing or dismantling of

VE-cadherin from endothelial junctions can up-regulate β-cateninsignaling (18). Consistently, we observed here that silencing VE-cadherin by siRNA activated the expression of EndMT markers(S100a4 and Id1) in addition to typical β-catenin targets (Axin2,Ccnd1, and Nkd1, SI Appendix, Fig. S5I) and promoted nuclearlocalization of active β-catenin (SI Appendix, Fig. S6A). More-over, VE-cadherin knockdown did not enhance the phosphor-ylation of Smad1 (SI Appendix, Fig. S6B).These data suggest that in CCM the first trigger of β-catenin

signaling is the dismantling of VE-cadherin junctions that, inturn, causes the release of β-catenin in the cytoplasm and nuclear

translocation. This process precedes and possibly contributes to theactivation of TGF-β/BMP signaling for lesion progression (7).

Sulindac Sulfide Reduces β-Catenin Transcription Activity and Expressionof Stem-Cell/EndMT Markers in Endothelial Cells of CCM3-ECKO Mice. Inthe attempt to translate the results described above into therapeuticopportunities, we investigated the effects of inhibitors of β-cateninsignaling on expression of target genes in CCM3-knockout endo-thelial cells. We tested a range of pharmacological agents alreadyused in humans and able to affect β-catenin signaling: sulindac

Fig. 2. Brain endothelial cells in CCM3-ECKO mice express stem-cell/EndMT markers in association with enhanced β-catenin transcription activity. (A–C) Brainsections from wild-type (WT) and CCM3-ECKO mice stained for β-gal (red) in combination with Podocalyxin (blue) and different stem-cell/EndMT markers(Klf4, S100a4, Id1, all green), at 3 and 9 dpn after CCM3 recombination at 1 dpn. Arrows point to endothelial cells (Podocalyxin-positive) expressing both β-galand stem-cell/EndMT markers (see Merge, yellow). (Scale bar, 40 μm.) (D) Quantification of endothelial nuclei positive for β-gal, Klf4, S100a4, and Id1 (singlepositive) and of their colocalization in brain sections as in A–C. Colocalization was calculated in two populations of endothelial cells: the β-gal–positive onewith EndMT-positive nuclei and the EndMT-positive one with β-gal–positive nuclei. At least 600 nuclei were counted in 50 random fields at 63× magnificationfor each condition in samples from matched littermate pups in three independent experiments. *P < 0.05 versus respective WT values (t test); ^P < 0.05 versusvalue in 3-dpn CCM3-ECKO pups.

Fig. 3. β-Catenin controls the expression of stem-cell/EndMT markers inCCM3-knockout endothelial cells in culture. (A and B) Quantification oftypical β-catenin transcription targets (A) and of stem-cell/EndMT markers(B) without (−) and with (+) expression of a dominant-negative Tcf4 in lungwild-type (WT) and CCM3-knockout (KO) endothelial cells. Data are means(±SD) of triplicate RT-PCR assays from three independent experiments. Tu-bulin transcripts (α, β), which are not targets of the CCM3 knockout, werenot modified by dominant-negative Tcf4. *P < 0.05 for CCM3-knockoutversus control (WT). ^P < 0.05 for CCM3 knockout plus dominant-negativeTcf4 versus CCM3 knockout plus GFP (t test).

Bravi et al. PNAS | July 7, 2015 | vol. 112 | no. 27 | 8423

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

May

31,

202

0

sulfide, sulindac sulfone (24, 25), silibinin, curcumin, and resveratrol(26) (SI Appendix, Fig. S7A).Sulindac sulfide and sulfone were the most effective inhibitors of

the β-catenin target genes (SI Appendix, Fig. S7A). We concen-trated our efforts on sulindac sulfide. This drug inhibited β-cateninsignaling by Top/Fop Flash reporter assay (SI Appendix, Fig. S7B)and active β-catenin nuclear localization in CCM3-knockout en-dothelial cells (SI Appendix, Figs. S7C and S8C). Furthermore,protein expression and staining of β-catenin–dependent stem-cell/EndMT markers were also reduced in cultured CCM3-knockoutendothelial cells by treatment with sulindac sulfide (SI Appendix,Figs. S8C and S9).The inhibitory effects of sulindac sulfide on β-catenin signaling

paralleled the effect of the drug in restoring a correct organi-zation of β-catenin and VE-cadherin at adherens junctions byimmunofluorescence costaining (SI Appendix, Fig. S7C magnifi-cation in i and ii; quantification in SI Appendix, Fig. S7D) and bycoimmunoprecipitation (SI Appendix, Fig. S7 E and F).Furthermore, the small GTPase Rap1, which controls the or-

ganization of endothelial adherens junctions and is upstream ofCCM1 activity (27), was inhibited in CCM3-knockout endothelialcells, but was restored by sulindac sulfide treatment (SI Appendix,Fig. S7 G and H).Treatment of CCM3-ECKO mice with sulindac sulfide reduced

the expression of the β-catenin reporter gene in brain endothelialcells at different stages after CCM3 ablation (Fig. 4 A and B) and,in parallel, inhibited junction dismantling (Fig. 4C). Sulindacsulfide also decreased the expression of the stem-cell/EndMTmarkers (SI Appendix, Figs. S10–S13) and limited increasedendothelial cell proliferation in the brain lesions (SI Appendix,Fig. S14).

Sulindac Sulfide Reduces Development of Vascular Lesions in the Brainand Retina of CCM3-ECKOMice. Sulindac sulfide significantly reducedthe number and size of the superficial CCM vascular malforma-tions in the cerebellum and in the deeper layers of the brain (Fig. 5A–C). The vessels of the retina of the CCM3-ECKO mice showedvascular malformations with multiple lumens concentrated at the

periphery of the vascular network. Such lesions develop from veins,which present an enlarged lumen (compare Vehicle in WT andCCM3-ECKO in Fig. 5D). Sulindac sulfide partially normalizedthis aberrant vascular network of CCM3-ECKO mice (Fig. 5 Dand E) and reduced the enlargement of the most central tract ofthe veins (diameter 89.5 ± 7.1 μm in Vehicle-CCM3-ECKO versus35 ± 7.8 μm in sulindac sulfide CCM3-ECKO, mean ± SD) of 30measurements in seven retinas each) (Fig. 5F). In addition, sulin-dac sulfide significantly prolonged the survival of CCM3-ECKOpups in both acute and subacute gene-recombination schedules(details in the figure legend) (P = 0.0046 and P = 0.0032, log-ranktest, respectively, Fig. 5 G and H).Sulindac sulfide has been reported to inhibit cyclooxygenase,

potentially leading to inhibition of platelet aggregation (24) and,possibly, increasing the hemorrhagic tendency of these patients.However, sulindac sulfone, which does not inhibit cyclooxygenase(24), showed a comparable activity to sulindac sulfide in inhibitingβ-catenin signaling and expression of stem-cell/EndMT markers incultured CCM3-knockout endothelial cells (SI Appendix, Figs. S7Aand S16). As with sulindac sulfide, the reduction of active β-cat-enin in the nucleus corresponded to the increased localization ofthis mediator at cell junctions together with VE-cadherin (SI Ap-pendix, Fig. S16). Sulindac sulfone also limited the expression ofstem-cell/EndMT markers and strongly reduced the number oflesions in the brain of the CCM3-ECKO mice (SI Appendix, Figs.S17 and S18).In conclusion, both sulindac metabolites inhibit β-catenin sig-

naling and, in parallel, exert therapeutic effects in hindering thedevelopment of brain vascular cavernomas in a preclinical model.

DiscussionHere, we report that endothelial-cell–selective deletion of theCCM3 gene activates β-catenin transcription signaling in vivo. Thisresponse develops early after CCM3 deletion, before activation ofTGF-β/BMP signaling (7), and contributes to the pathogenesis ofCCM3 cavernomas in this model. Furthermore, pharmacologicalinhibition of β-catenin transcription activity by sulindac sulfide and

Fig. 4. Brain endothelial cells in CCM3-ECKO mice show sulindac sulfide reduction of β-catenin transcription activity and induction of relocalization ofVE-cadherin from diffused distribution to adherens junctions. (A) Brain sections without (Vehicle) and with sulindac sulfide treatment of the CCM3-ECKO miceat different time points after CCM3 recombination. Arrowheads, β-gal reactivity (red) in the nucleus of endothelial cells (Podocalyxin-positive, green). (Scalebar, 50 μm.) (B) Quantification of immunofluorescence microscopy data as in A. At least 500 nuclei were counted in 40 random fields at 63× magnification foreach condition in samples from matched littermate pups in three independent experiments. *P < 0.05 versus respective Vehicle (t test). (C) Brain sections(9-dpn pups) stained for VE-cadherin (green), diffused (Vehicle), and junctional (sulindac sulfide, arrowheads) in blood-vessel endothelial cells of CCM3-ECKOand of wild-type (WT) mice (junctional, arrowheads). (Scale bar, 25 μm.)

8424 | www.pnas.org/cgi/doi/10.1073/pnas.1501352112 Bravi et al.

Dow

nloa

ded

by g

uest

on

May

31,

202

0

sulindac sulfone reduced the number and size of the cerebral vas-cular malformations in this murine model.CCM malformations develop largely, although not exclusively,

in the central nervous systems in patients and in mouse models(5, 6). The Wnt pathway is a well-established determinant forthe specification of the blood–brain barrier, and it is under tightcontrol during physiological angiogenesis of the central nervoussystem (11–13).Our data support the hypothesis of a cell-autonomous deregulated

activation of the β-catenin pathway in response to CCM3 deletionand delineate an unconventional mechanism of a β-catenin–driventranscription program instructing endothelial dedifferentiation.We report that in CCM3-deficient endothelial cells activation

of β-catenin–mediated transcription is independent of both en-dogenous Wnt ligand and Wnt-receptor stimulation.We show that, in the absence of CCM3, endothelial adherens

junctions are dismantled, as observed after ablation of both CCM1(27, 28) and CCM2 (5). In sparse endothelial cells and in VE-cadherin–null endothelial cells, when junctions are disorganized,β-catenin dissociates from cell–cell junctions and accumulates tothe nucleus (18). We observe here that in CCM3-deficient cells theamount of β-catenin associated with VE-cadherin is indeed re-duced by 35–50% concomitantly with the increase of this activemediator in the nucleus.Furthermore, by treating CCM3-knockout endothelial cells in

vitro and in vivo with sulindac metabolites, we were able to restorejunction organization and β-catenin association to VE-cadherinwhile reducing β-catenin nuclear signaling. Overall, these datastrongly suggest that a significant aspect of the function of the CCMcomplex is that of stabilizing endothelial cell junctions maintainingβ-catenin at the membrane, preventing, in this way, an uncontrolledβ-catenin transcriptional signaling.We have recently demonstrated that after endothelial-cell–

selective ablation of CCM1 in mice the TGF-β/BMP pathway isactivated and sustains the progression of the pathology (7). Weconfirm here these observations in CCM3-ECKO mice. In addi-tion, we observed that activation of β-catenin–driven transcriptionand nuclear localization precedes the initiation of TGF-β/BMPsignaling in CCM3-ablated endothelial cells both in vivo and invitro and that β-catenin contributes to such stimulation. Thesefindings delineate a sequence of signaling steps in responseto ablation of CCM3 in which, as an early event and concurrentlywith disorganization of adherens junctions, β-catenin concentratesin the nucleus to drive the expression of a dedifferentiation pro-gram, which comprises expression of stem-cell/EndMT markersand activation of TGF-β/BMP signaling for the progression of thevascular lesions. TGF-β/BMP signaling can then follow β-cateninsignaling and, possibly, contribute to the decrease in β-catenin sig-naling that we observed in vivo in endothelial cells of late stagevascular malformations (29). Similar sequential activation of dif-ferent signaling pathways after early β-catenin signaling has beenreported in epithelial transformation in colon cancer (30).Considering future pharmacological interventions, we have

identified two sulindac metabolites, sulindac sulfide and sulindacsulfone, that significantly inhibit β-catenin–stimulated transcriptionof stem-cell/EndMT markers and the development of vascular

Fig. 5. CCM3-ECKO mice show sulindac-sulfide–induced constraint of brainand retinal vascular lesions and prolonged survival. (A) Macroscopic ap-pearance of CCM3-ECKO mice brains following dissection without (Vehicle)and with sulindac sulfide treatment in pups at different time points afterCCM3 recombination at 1 dpn. (B) Endothelial cells (Pecam-positive, red) ofdifferent types of vascular lesions [mulberry (multiple cavernae), singlecaverna, or telangiectases (Telang.: tortuous small vessels with abnormallydilated lumen) (33)] in brain sections without (Vehicle) and with sulindacsulfide treatment of the CCM3-ECKO mice (9 dpn). (C, Left and Middle)Quantification of mean number of brain lesions as illustrated in B. Matchedlittermates from five independent litters were Vehicle-treated (n = 8) orsulindac sulfide-treated (n = 7). *P < 0.005, Wilcoxon signed-rank test.(Right) Quantification of mean size of brain lesions (μm; see SI Appendix,Methods for details). *P < 0.05, t test. (D) Endothelial cells (Pecam-positive,red) of vessels in the retina of wild-type (WT) and CCM3-ECKO mice without(Vehicle) and with sulindac sulfide treatment. Multiple-lumen vascular le-sions (arrowheads) develop from veins (arrow). (E) Percentages of the retinalperimeter affected by vascular lesions illustrated in D (n = 7 for both Vehicleand sulindac sulfide treatments). *P < 0.05, t test. (F) Detail of the retinavascular lesions illustrated in D. As well as the peripheral vascular malfor-mations, sulindac sulfide induced reductions in the diameters of the veins(green, isolectin B4-labeled endothelial cells). Arteries of these CCM3-ECKOmice do not show this aberrant phenotype. [Scale bars, 0.3 cm (A); 100 μm(B); 700 μm (D); 60 μm (F).] (G) CCM3-flox/flox–Cdh5(PAC)-CreERT2–BAT-galpups were treated with tamoxifen at 1 dpn to induce endothelial-specificrecombination of CCM3. Treatment with sulindac sulfide was started the

following day. Kaplan–Meier curves of Vehicle- and sulindac- sulfide-treatedpups were significantly different (P = 0.0046, log-rank test). Matched lit-termate pups were the following: Vehicle-treated, n = 13; sulindac sulfide-treated, n = 15 (in three independent experiments). (H) Pups wererecombined for CCM3 at 6 dpn to retard the development of CCM lesionsand prolong life span. Sulindac sulfide was started the following day.Kaplan–Meier curves of Vehicle- and sulindac sulfide-treated pups weresignificantly different (P = 0.0032, log-rank test). Matched littermate pupswere the following: Vehicle-treated, n = 8; sulindac sulfide-treated, n = 9(in two independent experiments).

Bravi et al. PNAS | July 7, 2015 | vol. 112 | no. 27 | 8425

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

May

31,

202

0

lesions. These agents are nonsteroidal anti-inflammatory drugs thathave shown significant inhibition of cancer progression in differentpreclinical models and have been used in the treatment of coloncancer in human patients (24). The mechanism of action of thesedrugs is complex (31) and includes, among various targets, inhi-bition of proteasome-dependent degradation of β-catenin, down-regulation of β-catenin transcription, and inhibition of β-cateninnuclear localization through inhibition of cGMP-PDE5 and acti-vation of PKG (25, 31). We report here a comparable activity ofthe two compounds in constraining CCM cavernomas in mice, butthe sulfone metabolite is more promising for further pharmaco-logical development because it is devoid of cyclooxygenase andplatelet inhibitory activity (32, 24).In conclusion, the targeting of β-catenin signaling with specific

pharmacological tools represents a promising strategy for the re-duction or prevention of cavernoma development in genetic CCMpatients, and it deserves consideration for clinical trials.

Materials and MethodsEndothelial-Cell–Specific Recombination in CCM3-flox/flox Mice. The CCM3-flox/flox mice were generated at TaconicArtemis. Two P-lox sequences wereinserted that flank exons 4 and 5 of the murine CCM3 gene to produce a loss-of-function mutation after excision by Cre recombinase. Detailed description ofthe breedings of these mice to generate endothelial-specific recombinationof CCM3 and β-catenin–driven expression of β-galactosidase as well as Cre-recombinase-reporter expression of enhanced yellow green fluorescentprotein is presented in SI Appendix.

Treatment with Sulindac Sulfide and Sulindac Sulfone. Both sulindac sulfide(Sigma) and sulindac sulfone (Sigma) were dissolved in DMSO and furtherdiluted 1:50 in corn oil. They were administered intragastrically daily (30 mg/kg

body weight), starting 1 d after the induction of recombination unless oth-erwise described. The control mice were treated in parallel with Vehicle only[corn oil plus 2% (vol/vol) DMSO]. Matched littermate pups were treated inparallel with either the drug or the Vehicle. Animal experimentation has beenapproved by the FIRC Institute of Molecular Oncology Institutional AnimalCare and Use Committee and was performed according to the guidelines ofthe Italian Ministry of Health regulating animal experimentation.

In Vitro Isolation, Culture, and Recombination of Endothelial Cells from theCCM3-flox/flox Mice. Endothelial cells from the CCM3-flox/flox mice (8–10 wkold) were isolated from the brain as previously described (13). CCM3 ablationwas with AdenoCre viral vector as described in SI Appendix.

Antibodies. The full list of the antibodies used is reported in SI Appendix.

Western Blotting and Immunoprecipitation. Standard procedures were used toextract and analyze the protein content by Western blotting and immuno-precipitation (28). Nuclear fractionation was as described in SI Appendix.

Assessment of Lesion Burden. For the classification and counting of lesions, entirebrains were sectioned, stained, and analyzed as described in the SI Appendixaccording to McDonald et al. (33).

ACKNOWLEDGMENTS. This study was supported by grants (to E.D.) fromTELETHON–GGP14149, Associazione Italiana per la Ricerca sul Cancro (AIRC)(AIRC IG 14471), “Special Program Molecular Clinical Oncology 5x1000” toAIRC-Gruppo Italiano Malattie Mieloproliferative, Fondazione Cassa di Ris-parmio delle Provincie Lombarde (CARIPLO) Contract 2012-0678, the Euro-pean Community (Wnt for Brain Contract 268870; Innovative TrainingNetworks Vessel 317250, Endostem-Health-2009-241440), and FondazioneCARIPLO Contract 2014-1038 (to N.R.). R.C. was supported by the FIRC fel-lowship 16617. Part of this work was funded by the European consortiumEuropean Research Area Network NEURON (to E.T.-L.).

1. Clatterbuck RE, Eberhart CG, Crain BJ, Rigamonti D (2001) Ultrastructural and im-munocytochemical evidence that an incompetent blood-brain barrier is related to thepathophysiology of cavernous malformations. J Neurol Neurosurg Psychiatry 71(2):188–192.

2. Rigamonti D, et al. (1988) Cerebral cavernous malformations. Incidence and familialoccurrence. N Engl J Med 319(6):343–347.

3. Li DY, Whitehead KJ (2010) Evaluating strategies for the treatment of cerebral cav-ernous malformations. Stroke 41(Suppl 10):S92–S94.

4. Plummer NW, Zawistowski JS, Marchuk DA (2005) Genetics of cerebral cavernousmalformations. Curr Neurol Neurosci Rep 5(5):391–396.

5. Boulday G, et al. (2011) Developmental timing of CCM2 loss influences cerebral cav-ernous malformations in mice. J Exp Med 208(9):1835–1847.

6. He Y, et al. (2010) Stabilization of VEGFR2 signaling by cerebral cavernous malfor-mation 3 is critical for vascular development. Sci Signal 3(116):ra26.

7. Maddaluno L, et al. (2013) EndMT contributes to the onset and progression of cere-bral cavernous malformations. Nature 498(7455):492–496.

8. Liebner S, et al. (2004) Beta-catenin is required for endothelial-mesenchymal trans-formation during heart cushion development in the mouse. J Cell Biol 166(3):359–367.

9. Glading AJ, Ginsberg MH (2010) Rap1 and its effector KRIT1/CCM1 regulate beta-catenin signaling. Dis Model Mech 3(1-2):73–83.

10. DiStefano PV, Kuebel JM, Sarelius IH, Glading AJ (2014) KRIT1 protein depletionmodifies endothelial cell behavior via increased vascular endothelial growth factor(VEGF) signaling. J Biol Chem 289(47):33054–33065.

11. Stenman JM, et al. (2008) Canonical Wnt signaling regulates organ-specific assemblyand differentiation of CNS vasculature. Science 322(5905):1247–1250.

12. Daneman R, et al. (2009) Wnt/beta-catenin signaling is required for CNS, but not non-CNS, angiogenesis. Proc Natl Acad Sci USA 106(2):641–646.

13. Liebner S, et al. (2008) Wnt/beta-catenin signaling controls development of theblood-brain barrier. J Cell Biol 183(3):409–417.

14. Corada M, et al. (2010) The Wnt/beta-catenin pathway modulates vascular remod-eling and specification by upregulating Dll4/Notch signaling. Dev Cell 18(6):938–949.

15. Liquori CL, et al. (2006) Low frequency of PDCD10 mutations in a panel of CCM3probands: Potential for a fourth CCM locus. Hum Mutat 27(1):118.

16. Shenkar R, et al. (2015) Exceptional aggressiveness of cerebral cavernous malforma-tion disease associated with PDCD10 mutations. Genet Med 17(3):188–196.

17. van Noort M, Meeldijk J, van der Zee R, Destree O, Clevers H (2002)Wnt signaling controlsthe phosphorylation status of beta-catenin. J Biol Chem 277(20):17901–17905.

18. Taddei A, et al. (2008) Endothelial adherens junctions control tight junctions by VE-cadherin-mediated upregulation of claudin-5. Nat Cell Biol 10(8):923–934.

19. Fadini GP, Losordo D, Dimmeler S (2012) Critical reevaluation of endothelial pro-genitor cell phenotypes for therapeutic and diagnostic use. Circ Res 110(4):624–637.

20. Medici D, Kalluri R (2012) Endothelial-mesenchymal transition and its contribution tothe emergence of stem cell phenotype. Semin Cancer Biol 22(5-6):379–384.

21. Clevers H, Nusse R (2012) Wnt/β-catenin signaling and disease. Cell 149(6):1192–1205.22. Anastas JN, Moon RT (2013) WNT signalling pathways as therapeutic targets in can-

cer. Nat Rev Cancer 13(1):11–26.23. Vleminckx K, Kemler R, Hecht A (1999) The C-terminal transactivation domain of

beta-catenin is necessary and sufficient for signaling by the LEF-1/beta-catenin com-plex in Xenopus laevis. Mech Dev 81(1-2):65–74.

24. Gurpinar E, Grizzle WE, Piazza GA (2013) COX-independent mechanisms of cancerchemoprevention by anti-inflammatory drugs. Front Oncol 3:181.

25. Li N, et al. (2013) Sulindac selectively inhibits colon tumor cell growth by activatingthe cGMP/PKG pathway to suppress Wnt/β-catenin signaling. Mol Cancer Ther 12(9):1848–1859.

26. Teiten MH, Gaascht F, Dicato M, Diederich M (2012) Targeting the wingless signalingpathway with natural compounds as chemopreventive or chemotherapeutic agents.Curr Pharm Biotechnol 13(1):245–254.

27. Glading A, Han J, Stockton RA, Ginsberg MH (2007) KRIT-1/CCM1 is a Rap1 effectorthat regulates endothelial cell cell junctions. J Cell Biol 179(2):247–254.

28. Lampugnani MG, et al. (2010) CCM1 regulates vascular-lumen organization by in-ducing endothelial polarity. J Cell Sci 123(Pt 7):1073–1080.

29. He XC, et al. (2004) BMP signaling inhibits intestinal stem cell self-renewal throughsuppression of Wnt-beta-catenin signaling. Nat Genet 36(10):1117–1121.

30. Polakis P (2012) Wnt signaling in cancer. Cold Spring Harb Perspect Biol 4(5):4.31. Gurpinar E, Grizzle WE, Piazza GA (2014) NSAIDs inhibit tumorigenesis, but how? Clin

Cancer Res 20(5):1104–1113.32. Piazza GA, et al. (1997) Sulindac sulfone inhibits azoxymethane-induced colon carci-

nogenesis in rats without reducing prostaglandin levels. Cancer Res 57(14):2909–2915.33. McDonald DA, et al. (2012) Fasudil decreases lesion burden in a murine model of

cerebral cavernous malformation disease. Stroke 43(2):571–574.

8426 | www.pnas.org/cgi/doi/10.1073/pnas.1501352112 Bravi et al.

Dow

nloa

ded

by g

uest

on

May

31,

202

0