FOXC2 controls Ang-2 expression and modulatesangiogenesis, vascular patterning, remodeling,and functions in adipose tissueYuan Xue*, Renhai Cao*, Daniel Nilsson†, Shaohua Chen*, Rickard Westergren†, Eva-Maria Hedlund*, Cecile Martijn‡,Lena Rondahl‡, Per Krauli‡, Erik Walum‡, Sven Enerback†, and Yihai Cao*§

*Laboratory of Angiogenesis Research, Department of Microbiology, Tumor and Cell Biology, Karolinska Institute, 171 77 Stockholm, Sweden; †Departmentof Medical and Clinical Genetics, Institute of Biomedicine, the Sahlgrenska Academy, Goteborg University, Box 440, SE 405 30 Goteborg, Sweden;and ‡Biovitrum AB, Strandbergsgatan 49, P12-4, SE-112 76 Stockholm, Sweden

Edited by Tak Wah Mak, University of Toronto, Toronto, ON, Canada, and approved April 29, 2008 (received for review March 17, 2008)

Adipogenesis is spatiotemporally coupled to angiogenesis through-out adult life, and the interplay between these two processes iscommunicated by multiple factors. Here we show that in a transgenicmouse model, increased expression of forkhead box C2 (FOXC2) in theadipose tissue affects angiogenesis, vascular patterning, and func-tions. White and brown adipose tissues contain a considerably highdensity of microvessels appearing as vascular plexuses, which showredistribution of vascular smooth muscle cells and pericytes. Dysfunc-tion of these primitive vessels is reflected by impairment of skinwound healing. We further provide a mechanistic insight of thevascular phenotype by showing that FOXC2 controls Ang-2 expres-sion by direct activation of its promoter in adipocytes. Remarkably, anAng-2-specific antagonist almost completely reverses this vascularphenotype. Thus, the FOXC2–Ang-2 signaling system is crucial forcontrolling adipose vascular function, which is part of an adaptationto increased adipose tissue metabolism.

adipogenesis � neovascularization � wound healing � obesity � metabolism

Obesity is a risk factor for diabetes, dyslipidemia, cardiovasculardisease, cancer, and sleep breathing disorders (1–3). Although

genetic defects contribute to obesity, high-caloric diet and lowphysical activity are critical etiological factors responsible for mostof today’s obesity (4). Unlike most other tissues, adipose tissueconstantly experiences expansion and regression throughout adultlife. The plasticity of adipose tissue requires continuous remodelingof the vasculature that controls energy expenditure, metaboliteexchange, transport of adipokines or hormones, and adipocytehypertrophy and hyperplasia (5–7). Vascular remodeling and func-tions are regulated by a number of growth factors and inhibitors.For example, leptin produced by the adipose tissue possessesangiogenic activity, and its expression level is correlated withadipogenesis (8–10). Adiponectin, an adipose tissue-derived hor-mone, has a reverse correlation with adipose tissue growth andinhibits angiogenesis (11). In addition, known angiogenic factorssuch as VEGF, angiopoietin (Ang), insulin-like growth factor,hepatocyte growth factor, and FGF produced by adipocytes andnonadipocytes are involved in regulation of adipose angiogenesis (5,12–21). Indeed, angiogenesis inhibitors could prevent obesity bynormalizing metabolisms and without significantly affecting foodintake in both leptin-deficient genetic and in high-caloric diet-fedmouse models (22, 23). Interestingly, reduction of fat mass in theangiogenesis inhibitor-treated animals is accompanied by improvedinsulin sensitivity (22).

Ang is an important group of vascular remodeling factors thatcontrol vessel maturation, patterning, and stabilization (24).Although Ang-1 and Ang-2 could bind and activate the sameTie-2 tyrosine kinase receptor on endothelial cells, they seem todisplay opposing vascular functions (25). Ang-1 and VEGF exertcomplementary effects during early embryonic development(26). Although VEGF initiates vascular formation, Ang-1 pro-motes subsequent remodeling, maturation, and stabilization

(26). Ang-2 plays a complex role in regulation of vascularremodeling that leads to either vessel sprouting or regression,depending on its expression relation with other angiogenicstimuli. For example, in the presence of VEGF, Ang-2 wouldpotentiate angiogenic sprouting (27). However, in the absence ofVEGF, Ang-2 might lead to vascular regression. Despite theseknown angiogenic factors expressed in adipose tissue, underlyingmechanisms of regulation of adipose vascular patterning, re-modeling, and functions remain uncharacterized.

FOXC2 is a member of the forkhead/winged helix transcriptionfactor family, playing an important role in regulation of metabolism,arterial specification, and vascular sprouting (28). Based on patientstudies involving measurements of FOXC2 mRNA levels in whiteadipose tissue (WAT), identification of a single nucleotide poly-morphism in the FOXC2 promoter region and detection of aframe-shift mutation in FOXC2 has led us and others to suggestthat FOXC2 plays a role in human conditions such as obesity,dyslipidemia, and type 2 diabetes (29–32). Human FOXC2 muta-tions are associated not only with obesity and type 2 diabetes butalso with defects in lymphangiogenesis (lymphoedema-distichiasissyndrome) (33). Previous work showed that specific expression ofFOXC2 in adipocytes led to conversion of WAT to a ‘‘brown’’-likeadipose tissue phenotype in transgenic mice (FOXC2-TM) (34).Adipocytes from such mice exhibit a 4-fold increase in oxygenconsumption (34). Thus, this is an optimal model to study howvascularization is regulated in response not only to tissue remod-eling (more brown-like adipocytes) but also to an increased met-abolic rate. Our present results show that FOXC2 directly controlsthe promoter activity of Ang-2, which alters vascular patterning,remodeling, maturation, and functions.

ResultsVascular Phenotypes of WAT and Brown Adipose Tissue (BAT). Nec-ropsy analysis of FOXC2-TM mice revealed that both axillary andinguinal WAT appeared as a reddish or brownish tissue, which wassimilar to the color of BAT (34) (Fig. 1A). Immunohistochemicalanalysis of the epididymal and inguinal WAT with an anti-CD31antibody showed that a high density of ‘‘plexus-like’’ vascularityexisted in FOXC2-TM. These microvessels appeared as ‘‘honey-comb-like’’ vascular networks, which encapsulated adipocytes (Fig.1 B and C). In contrast, microvascular networks of WATs derived

Author contributions: Y.X., D.N., S.E., and Y.C. designed research; Y.X., R.C., D.N., S.C., R.W.,C.M., L.R., and P.K. performed research; Y.X., R.C., D.N., R.W., E.-M.H., E.W., S.E., and Y.C.analyzed data; and Y.X., S.E., and Y.C. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

§To whom correspondence should be addressed. E-mail: yihai.cao@ki.se.

This article contains supporting information online at www.pnas.org/cgi/content/full/0802486105/DCSupplemental.

© 2008 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0802486105 PNAS � July 22, 2008 � vol. 105 � no. 29 � 10167–10172

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

June

11,

202

0

from WT mice were highly organized and appeared in a relativelylow density. Similar to WAT, interscapular BAT of FOXC2-TMappeared as an exceptionally high density of vascular networks,which virtually consisted of only vascular plexuses (Fig. 1D). Sig-nificantly fewer vascular plexuses were detected in WATs or BATof WT mice (Fig. 1F). It should be emphasized that the vascularphenotype of WAT in FOXC2-TM resembled the phenotypesfound in BAT in WT mice, suggesting a possible transition fromWAT toward BAT (Fig. 1 C and D). Quantification demonstrated2- to 3-fold increases of vascular density of both WAT and BAT inFOXC2-TMcomparedwith thoseofcontrols (Fig.1E).These findingsdemonstrate that overexpression of FOXC2 in the adipose tissueleads to alterations of both vessel numbers and vascular structures.

Vascular Phenotype of Dermal and Subcutaneous Tissues of FOXC2-TM. We examined blood vessels in the dermal and subcutaneoustissues by immunohistological analysis using an anti-CD31-specific antibody. A high microvessel density was detected in thedermal and subcutaneous tissues of FOXC2-TM compared withthose of the WT mice [supporting information (SI) Fig. S1 A andB]. The structure of these vessels appeared to be a dense networkwith irregular and tangled vascularity. In some areas, bloodvessels tended to form ragged vascular plexuses due to fusion ofmultiple capillaries/microvessels. Strikingly, the average inter-capillary distance in the subcutaneous tissue of FOXC2-TM wassignificantly shorter than that of the adipose tissue of WT mice.

Quantification showed that the total area of the CD31-positivestructures and the number of vascular plexuses were significantlyhigher in both dermal and subcutaneous tissues of FOXC2-TMthan those of WT mice (Fig. S1 C and D).

Alteration of Vascular Remodeling and Maturation. Altered pattern-ing of adipose vasculatures prompted us to study vascular remod-eling and maturation in both WAT and BAT of FOXC2-TM. InWT mice, vascular smooth muscle cells (VSMCs) were all associ-ated with large arterial blood vessels, and no �-smooth muscle actin(�-SMA)-positive signals were found in microvessels in either WATor BAT (Fig. 2A). Interestingly, the total number of �-SMA-positive arterial vessels was significantly decreased in both WATand BAT of FOXC2-TM as compared with WT mice (Fig. 2B).Surprisingly, a considerable number of �-SMA-positive VSMCswere associated with microvessels, including the dense vascularplexuses, suggesting redistribution of VSMCs (Fig. 2A). There wasa similar trend for redistribution of NG2-positive pericytes inmicrovasculatures of WAT and BAT of FOXC2-TM. In WT mice,NG2-positive pericytes were mainly associated with arterial vesselsand broadly distributed in the intervascular spaces but remained

Fig. 1. Patterning, remodeling, and maturation of adipose vasculature inFOXC2-TM. (A) Necropsy analysis of subcutaneous adipose tissue of FOXC2-TMand WT mice. Arrows point to reddish WAT of FOXC2-TM. (B–D) Epididymal (B)and inguinal (C) WATs and interscapular BAT (D) of both FOXC2-TM and WTmice were stained with an anti-CD31 antibody. Vascular patterns and struc-tures were revealed by 3D projection by using confocal laser scanning micro-scope analysis of the whole-mount tissues. Arrows point to disorganizedvascular plexuses. (Bar: 50 �m.) (E and F) Total vessel areas (E) and numbers ofdisorganized vascular plexuses (F) were quantified, and the data are pre-sented as mean (�SD). epiWAT, epididymal WAT; ingWAT, inguinal WAT;intBAT, interscapular BAT.

Fig. 2. Association of smooth muscle cells and pericytes. (A) Inguinal WATand interscapular BAT were triple-stained with an anti-CD31, an anti-�-SMA,and an anti-NG2 antibody. Blood vessels (red), VSMCs (green), and pericytes(blue) were revealed by confocal laser scanning microscopy of the frozentissues. Arrows point to triple-positive signals. (Bar: 50 �m.) (B–E) The totalnumber of �-SMA-positive large arterial vessels (B), the total �-SMA-positivearea (C), the total CD31-positive area (D), and the total NG2-positive area werequantified as per optical field (�20), and the data of 9 or 10 random fields arepresented as mean (�SD). intBAT, interscapular BAT; ingWAT, inguinal WAT.

10168 � www.pnas.org�cgi�doi�10.1073�pnas.0802486105 Xue et al.

Dow

nloa

ded

by g

uest

on

June

11,

202

0

less associated with microvessels (Fig. 2A). However, virtually allNG2-positive pericytes were found to be associated with primitivevascular networks in both WAT and BAT of FOXC2-TM. Inaddition to microvessels, large arterial vessels are also partiallycoated with NG2-positive pericytes (Fig. 2A). Quantification dem-onstrated that the total area of �-SMA-positive structures in bothWAT and BAT was significantly increased in FOXC2-TM (Fig.2C). In contrast, total numbers of NG2-positive pericytes weresignificantly decreased in WAT and BAT of FOXC2-TM, butnearly all pericytes remained associated with microvessels. Takentogether, these findings demonstrate that FOXC2 affects adiposevessel maturation by regulating distribution of both VSMCs andpericytes. Vascular remodeling is probably a prerequisite for bloodvessels to coordinate with the high metabolic rate in FOXC2-expressing adipose tissue.

Up-Regulation of Angiogenic Factors and Their Signaling Components.To study the underlying mechanisms of the angiogenic pheno-type switched by FOXC2, we compared gene expression profilesof the abdominal WAT of FOXC2-TM and WT mice by usingan Affymetrix microarray gene chip and PCR analyses. Se-quences of primers used for quantitative real-time PCR assayscan be found in Table S1. Interestingly, among up-regulated geneproducts, the expression levels of several potent angiogenicfactors were significantly increased (Table S2). These angiogenicgene products include members of the VEGF, PDGF, Ang,TGF-�, TNF, ephrin, insulin-like growth factor, and endothelinfamilies. Interestingly, several of the significantly up-regulatedangiogenic factors such as VEGF-C, Ang-2, and PDGF-AA havebeen reported to induce angiogenesis and are involved in regulationof vascular maturation, remodeling, and stabilization (35–37). Itshould be emphasized that Ang-2 was one of the most up-regulatedgene products in the adipose tissue of FOXC2-TM. Thus, these datasupport the active angiogenic phenotype in FOXC2-TM.

Additionally, levels of several receptor signaling molecules ex-pressed in endothelial cells or VSMCs/pericytes, which are directlyor indirectly involved in the angiogenesis and vascular remodelingsignaling systems, were also increased (Table S3). Importantly,tyrosine kinase receptors, including the PDGF receptor-�(PDGFR-�) and FGF receptor type 2 (FGFR-2), were significantlyup-regulated. In addition to tyrosine kinase receptors, the levels ofa couple of GTP-coupled signaling components were also signifi-cantly increased. These data provide a possible molecular basis forthe observed angiogenic phenotypes.

FOXC2 Transcriptionally Regulates Ang-2 Expression. To furthervalidate gene expression profiles regulated by FOXC2, quanti-tative real-time PCR analysis was performed with RNAs ex-tracted from WAT of FOXC2-TM and WT mice. Among allanalyzed gene transcripts, Ang-2 was the most up-regulated genein the adipose tissue of FOXC2-TM, which was consistent withthe gene array findings. A nearly 6-fold increase of Ang-2 wasdetected in WAT of FOXC2-TM as compared with that of WTmice (Fig. 3A). In addition, the level of placental growth factorwas also significantly increased. There was a trend for elevatedexpression levels of EfnB-2, Notch-3, and PDGFR-�. Thus,these results obtained from real-time PCR generally validatedthe data from Affymetrix array analysis.

Sequence analysis of the ligated sequences by using pDRAW(Acaclone Software) revealed five putative forkhead-bindingsites by using the forkhead-binding consensus sequence (RY-MAAYA; R � A/G; Y � C/T; M � A/C), suggesting that Ang-2might be a direct target gene for FOXC2. To investigate thispossibility, Ang-2 promoter fused with the luciferase reporterwas cloned and transfected into 3T3 diffentiated preadipocytesin the presence and absence of FOXC2. Interestingly, Angpromoter activity was increased in a dose-dependent fashionafter the addition of FOXC2 (Fig. 3C Left). Nearly an 8-fold

increase of luciferase activity was observed at 200 ng of FOXC-2.These findings show that FOXC-2 directly activates Ang-2promoter activity and controls its expression.

To further delineate the FOXC2-responsive sequences in theAng-2 promoter, various deletion mutant promoter constructswere used to drive luciferase gene expression. Primers used tomake these constructs are listed in Table S4. Among all five Fkhregions in the Ang-2 promoter, deletion of Fkh4 nearly abro-gated the promoter activity induced by FOXC2, suggesting thatthe Fkh4 region was essential for FOXC2-induced transcriptionactivity (Fig. 3C Right). In contrast, deletion of the other fourFkh regions in the Ang-2 promoter did not affect the reportergene expression in any significant way.

We next isolated preadipocytes from FOXC2�/�, FOXC2�/�

heterozygous, and FOXC2�/� knockout mice to quantitativelycorrelate expression levels of FOXC2 with those of Ang-2. Expect-edly, an ideal correlation of expression levels between FOXC2 andAng-2 existed in FOXC2�/� preadipocytes (Fig. 3B). Approxi-mately 50% reduction of expression levels of FOXC2 and Ang-2was detected in FOXC2�/� preadipocytes. Interestingly, completeknockout of FOXC2 gene in preadipocytes did not further decrease

Fig. 3. Detection of gene expression by real-time PCR and analysis of Ang-2promoter activity. (A) mRNAs extracted from epididymal WAT and mouseembryo fibroblasts (MEFs) were used for quantitative amplification of Ang-2,EfnB2, placental growth factor (PlGF), Notch3, and PDGFR-� by real-time PCRanalysis. (B) FOXC2�/�, FOXC2�/�, and FOXC2�/� MEFs were used for correlat-ing Ang-2 and FOXC2 expression levels. (C) The Ang-2 promoter-luciferasereporter construct was used to transfect 3T3 fibroblast-differentiated prea-dipocytes, and induction of luciferase activity was determined. (Left) The foldincrease of luciferase activity after various amounts of FOXC2 DNA wasquantified, and the average data of each sample are presented as mean (�SD).(Right) A series of deletion constructs consisting of various regions of theAng-2 promoter and luciferase reporter were generated and transfected intopreadipocytes (see Table S4). delFkh4, deletion of Fkh4.

Xue et al. PNAS � July 22, 2008 � vol. 105 � no. 29 � 10169

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

June

11,

202

0

Ang-2 expression as compared with that of FOXC2�/� preadipo-cytes, suggesting that other mechanisms might exist to control thebasal expression level of Ang-2 in adipocytes. Taken together, thesedata demonstrate that FOXC-2 transcriptionally up-regulatesAng-2 expression in adipocytes.

Ang-2 Is the Essential Mediator for the FOXC2-Induced AngiogenicPhenotype. Transcriptional regulation of Ang-2 expression byFOXC2 in adipocytes suggested that Ang-2 might be responsiblefor switching on the angiogenic phenotype. To explore this possi-bility, a known Ang-2-specific inhibitor, L1-10 (38), was used toreverse the FOXC-2-induced angiogenic phenotype in FOXC2-TM. Interestingly, administration of L1-10 to FOXC2-TM at a doseof 4 mg/kg, a dose known to block Ang-2 function in vivo, virtuallycompletely reversed the FOXC-2-induced angiogenic phenotype.Morphologically, the reddish appearance of axillary and inguinaladipose tissue in FOXC2-TM was converted into a relatively palecolor in the L1-10-treated group as compared with that of thebuffer-treated group (Fig. 4A). Immunohistological analysisshowed that the primitive vascular plexus-shaped vessels in inguinalWAT were normalized to well structured vascular networks, whichwere indistinguishable from those in WT adipose tissues (Fig. 4 Cand E). In addition to structural changes, the unusually high densityof microvessels in adipose tissues in FOXC2-TM was also normal-ized to the level of WT adipose tissues (Fig. 4 B and C). Similarly,redistribution of �-SMA-positive VSMCs was reversed to therelatively large arterial vessel, as seen in those in both WAT andBAT of WT mice (Fig. 4D). Quantification showed that blockageof Ang-2 led to decreased total number of VSMC-coated vasculararea and increased large arterial vessel association (Fig. 4 F and G).These data provide compelling evidence that Ang-2 is responsiblefor the FOXC-2-induced vascular maturation and patterning in theadipose tissue.

Impairment of Vascular Function. The vascular phenotype in dermaland subcutaneous tissues of FOXC2-TM suggested possible alter-ations of vascular functions. To explore this possibility, we per-formed full skin wound-healing experiments. At day 11 after thecreation of the wound, all WT mice exhibited complete healing ofthe wound beds (n � 10). In contrast, a significantly delayed woundhealing was observed in FOXC2-TM. Notably, significantly largerdiameter wounds already became obvious at day 4 after the creationof the wound and significant differences remained throughout theentire experiments (Fig. 5 B–D). Approximately 3 days of delayedwound healing was observed in FOXC2-TM, which showed com-plete healing at day 14 (Fig. 5 C and D). Immunohistochemicalanalysis showed that a significantly higher number of CD31-positivevessels were present in wound tissue of FOXC2-TM than in that ofWT mice, suggesting that impairment of wound healing was not dueto defects of neovascularization but abnormality of vascular func-tion (Fig. 5 A and E). These results demonstrate that the vascularadaptation seen in FOXC2-TM results in remodeling of existingvessels and formation of premature new vessels that lead to delayedwound healing.

DiscussionThe plasticity of adipose tissue throughout adult life requiresconstant vessel growth, regression, and remodeling. In addition toadipose tissue growth, conversion of the WAT into a BAT-likephenotype demands a high metabolic rate by activation of theadrenergic/cAMP/protein kinase A signaling pathway and increas-ing oxygen consumption, which requires increased blood supply(39). Accumulating evidence shows that adipocytes cross-communicate with neighboring endothelial cells via paracrinesignaling pathways, extracellular components, and direct cell–cellinteractions (5, 40–42). For instance, the adipocyte-derived hor-mone leptin induces Ang-2 expression in adipocytes (43). In brownadipocytes, it is well established that an increased oxygen consump-

tion/metabolic rate, induced by cold exposure, is associated withincreased synthesis of angiogenic factors such as VEGF, which inturn stimulates vessel growth and remodeling in response to met-abolic needs (44–46). However, the molecular identity and geneticcontrol of adipose-derived paracrine factors in regulation of vesselgrowth, maturation, remodeling, patterning, and function remainpoorly characterized. Here we show that in a transgenic mousemodel, with elevated metabolism in WAT, FOXC2 transcription-ally switches on an angiogenic phenotype in the adipose tissue byincreasing expression of angiogenic factors, including Ang-2.

Fig. 4. Reversal of vascular phenotype by blocking Ang-2. Ang-2 inhibitor L1-10was administrated into 3-week-old FOXC2-TM and WT mice for 4 weeks. (A)Axillary, inguinal, and other subcutaneous adipose tissues were exposed aftereuthanizing mice. Relatively pale WATs were found in the L1-10-treatedFOXC2-TM animals. (B) Quantification of CD31-positive blood vessels. (C) Whole-mount tissues of inguinal WAT (ingWAT) and interscapular BAT (intBAT) werestained with an anti-CD31 antibody. (D) Frozen sections of inguinal WAT andinterscapular BAT were used for double staining of endothelial cells (red) andVSMCs (green, revealed by presence of �-SMA). White arrows point to double-positive large arterioles, and arrowheads point to microvessel double-positivesignals (yellow). (Bar: 50 �m.) (E–G) Quantification of vascular plexuses (E),�-SMA� large arterioles (F), and the total area of �-SMA� area (G). Data arepresented as mean (�SD) of 7–10 randomized fields.

10170 � www.pnas.org�cgi�doi�10.1073�pnas.0802486105 Xue et al.

Dow

nloa

ded

by g

uest

on

June

11,

202

0

To reveal the identity of FOXC2-regulated soluble paracrine andendocrine factors produced by adipocytes, Affymetrix gene arrayanalysis showed that Ang-2 is one of a few angiogenic gene productsthat are up-regulated at high levels. Up-regulation of Ang-2 hasbeen confirmed by quantitative real-time PCR, and its promoteractivity could be directly induced by FOXC-2. Among five Fkhregions, Fkh4 is the only essential element in the Ang-2 promoterresponsible for FOXC2 regulation. Ang-2 is one of the few vascularfactors known for regulation of vascular patterning, remodeling,and maturation (24). Although Ang-2 and Ang-1 bind to the sametyrosine kinase receptor Tie-2, they display opposing activity onblood vessel remodeling and maturation. Although Ang-1 promotesvessel maturation by recruiting VSMCs onto the nascent vascula-ture, Ang-2 repels mural cell association with blood vessels (25, 27).Without mural cells, the newly formed vasculature remains unsta-ble and experiences either growth or regression depending on thepresence of other angiogenic stimuli. The primitive plexus-shapedvascular network observed in the adipose tissue of FOXC2-TManimals is consistent with vascular functions of Ang-2. ThisFOXC2-Ang-2-induced phenotype resembles the phenotypes intumors induced by Notch antagonists, which induce primitive,disorganized, and nonproductive vascular networks (47).

Although several other angiogenic factors and receptor signalingmolecules are also up-regulated in FOXC-2-expressing adiposetissue, their vascular roles in relation to Ang-2 are unclear. It isknown, however, that in the presence of Ang-2, VEGF couldfurther accelerate neovascularization by broadly acting on non-mural-cell-coated endothelium (48). Other angiogenic factorscould also exert a similar synergistic effect with Ang-2 on vascu-larization and remodeling. Although Ang-2 repels pericytes andVSMCs from large vessels, it is unclear why these mural cells areredistributed from large vessels to coat microvessels. This effectprobably requires an intimate interplay between Ang and PDGFsystems. For example, growing cones of tip endothelial cells areknown to produce PDGF-BB, which recruits pericytes and VSMCsonto the newly formed vasculature (49, 50). It is possible that Ang-2repels mural cells from large vessels and that PDGF-BB redirectsthem onto microvessels. Because little is known about angiogenesisand vascular remodeling in response to increased metabolism in

adipose tissue, our model system might provide a unique oppor-tunity to study the underlying mechanisms by which Ang-2 andother vascular factors cooperatively control vascular maturation,remodeling, and function.

One of the most intriguing findings in our study is that a specificAng-2 inhibitor could reverse the FOXC-2-induced vascular phe-notype in adipose tissue. These functional data provide convincingevidence that Ang-2 is the target gene product responsible for theobserved vascular phenotype. Several independent studies showthat Ang-2 is constitutively up-regulated in expanding adiposetissues (5, 21, 51). In addition to having its direct angiogenicfunction, Ang-2 is required for vascular patterning, remodeling, andmaturation in growing adipose tissue. Ang-2 is a soluble vascularfactor that targets both proximal and distal vasculatures (52, 53).Malformation of vascular networks in nonadipose tissues as aresponse to an ‘‘overshoot’’ in adipose tissue-derived angiogenicfactors leads to functional defects. For example, skin wound healingis significantly delayed in FOXC-2-TM.

Taken together, our work provides compelling evidence andmolecular mechanisms that FOXC-2 regulates the cross-talkbetween adipocytes and vascular cells, including endothelial cellsand mural cells. Ang-2 as a direct target for FOXC2 is involvedin vascular remodeling, patterning, and maturation in adiposetissues. We would like to speculate, based on the findingspresented, that via regulation of FOXC2 adipose tissue canadapt its degree of vascularization to meet present metabolicdemand. Thus, functional interference with FOXC2 or Ang-2might provide a therapeutic approach for prevention and treat-ment of obesity or its related disorders.

Experimental ProceduresAnimals. Animals were anesthetized by an injection of a mixture of Dormicumand Hypnorm (1:1, Roche and VetaPharma) before all procedures and killed bya lethal dose of CO2 followed by cervical dislocation. All animal studies werereviewed and approved by the Animal Care and Use Committee of the NorthStockholm Animal Board. See SI Methods for information about reagents.

Plasmid Construction, Mutagenesis, Cell Culture, and Transfection. Plasmidconstruction, mutagenesis, cell culture procedure, and DNA transfection aredescribed in SI Methods.

Fig. 5. Wound healing. Full skin wounds werecreated on the backs of WT and FOXC2 mice. (A)The wound beds were stained with H&E (Left) ordouble-stained with propidium iodide (PI, red) andan anti-CD31 antibody (green) (Right). Arrowspoint to CD31-positive vessels. (Bar: 50 �m.) Diam-eters of wounds were measured every other day (Band D), and the percentages of animals with com-pletely healed wounds were recorded (C). (E) CD31-positive blood vessels were quantified as vesselarea per optical field (�10), and data representmean of 9 or 10 samples. ***, P � 0.001.

Xue et al. PNAS � July 22, 2008 � vol. 105 � no. 29 � 10171

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

June

11,

202

0

Whole-Mount Staining and Confocal Analysis. The whole thicknesses of dorsalmiddle region skin tissue, inguinal and epididymal WAT, and interscapular BATwere freshly resected from FOXC2-TM and WT mice and were fixed with 4%paraformaldehyde overnight. Tissue samples were transferred into PBS and weredigested with proteinase K (20 �g/ml) for 5 min followed by staining overnight at4°C with a rat anti-mouse CD31 antibody (1:100). After rigorous rinsing, bloodvessels were detected with labeled secondary antibodies. After washing, slideswere mounted in Vectashield mounting medium (Vector Laboratories) andstored at �20°C in the dark before examination under a confocal microscope(Zeiss confocal LSM510 laser scanning microscope, CLSM). The images werefurther analyzed with the Adobe PhotoShop CS software program.

Real-time Quantitative RT-PCR. See SI Methods for details.

Immunohistochemistry. See SI Methods for details.

Vascular Quantification Analysis. Vascularization areas were quantified by usingan Adobe PhotoShop program. Briefly, 10 random fields (200 � 200 �m persquare) of CD31-, NG2-, or �-SMA-positive structures from three to five animalswere activated and calculated with a computerized mathematic method.

Ang-2 Blocking Experiments. Three-week-old female FOXC2-TM mice wererandomly divided into two groups (five animals per group), and the samenumber of WT mice were used as controls in each group. The detailed protocolfor Ang-2 blockage experiments is described in the SI Methods.

Statistical Analyses. Statistical analysis of the in vitro and in vivo results wasmade by a standard two-tailed Student’s t test by using Microsoft Excel 2003.P � 0.05, P � 0.01, and P � 0.001 were deemed as significant, highly significant,and extremely significant, respectively.

ACKNOWLEDGMENTS. We thank Drs. Yu Li, Sharon Lim, and Anne Hennig fortechnical support. We thank Amgen for providing L1-10 and Dr. NaoyukiMiura at Hamamatsu University School of Medicine for providing Foxc2�/�

mice for our studies. Y.C.’s laboratory is supported by research grants from theSwedish Research Council, the Swedish Heart and Lung Foundation, theSwedish Cancer Foundation, the Karolinska Institute Fund, the SoderbergFoundation, the European Union integrated projects of Angiotargeting, andVascuPlug. S.E. is supported by the Swedish Research Council, European Uniongrants, the Arne and IngaBritt Foundation, the Swedish Foundation forStrategic Research, and the Soderberg Foundation.

1. Friedman JM (2000) Obesity in the new millennium. Nature 404:632–634.2. Kopelman PG (2000) Obesity as a medical problem. Nature 404:635–643.3. Roth J, Qiang X, Marban SL, Redelt H, Lowell BC (2004) The obesity pandemic: Where

have we been and where are we going? Obes Res 12(Suppl 2):88S–101S.4. Flegal KM (2000) Obesity, overweight, hypertension, and high blood cholesterol: The

importance of age. Obes Res 8:676–677.5. Cao Y (2007) Angiogenesis modulates adipogenesis and obesity. J Clin Invest 117:2362–

2368.6. Dallabrida SM, et al. (2003) Adipose tissue growth and regression are regulated by

angiopoietin-1. Biochem Biophys Res Commun 311:563–571.7. Planat-Benard V, et al. (2004) Plasticity of human adipose lineage cells toward endo-

thelial cells: Physiological and therapeutic perspectives. Circulation 109:656–663.8. Bouloumie A, Drexler HC, Lafontan M, Busse R (1998) Leptin, the product of Ob gene,

promotes angiogenesis. Circ Res 83:1059–1066.9. Cao R, Brakenhielm E, Wahlestedt C, Thyberg J, Cao Y (2001) Leptin induces vascular

permeability and synergistically stimulates angiogenesis with FGF-2 and VEGF. ProcNatl Acad Sci USA 98:6390–6395.

10. Sierra-Honigmann MR, et al. (1998) Biological action of leptin as an angiogenic factor.Science 281:1683–1686.

11. Brakenhielm E, et al. (2004) Adiponectin-induced antiangiogenesis and antitumoractivity involve caspase-mediated endothelial cell apoptosis. Proc Natl Acad Sci USA101:2476–2481.

12. Baillargeon J, et al. (2006) Obesity, adipokines, and prostate cancer in a prospectivepopulation-based study. Cancer Epidemiol Biomarkers Prev 15:1331–1335.

13. Friedman JM, Halaas JL (1998) Leptin and the regulation of body weight in mammals.Nature 395:763–770.

14. Hiraoka Y, et al. (2006) In situ regeneration of adipose tissue in rat fat pad bycombining a collagen scaffold with gelatin microspheres containing basic fibroblastgrowth factor. Tissue Eng 12:1475–1487.

15. Kuo LE, Abe K, Zukowska Z (2007) Stress, NPY and vascular remodeling: Implications forstress-related diseases. Peptides 28:435–440.

16. Li J, Yu X, Pan W, Unger RH (2002) Gene expression profile of rat adipose tissue at theonset of high-fat-diet obesity. Am J Physiol 282:E1334–E1341.

17. Lijnen HR, et al. (2006) Impaired adipose tissue development in mice with inactivationof placental growth factor function. Diabetes 55:2698–2704.

18. Mu H, et al. (2006) Adipokine resistin promotes in vitro angiogenesis of humanendothelial cells. Cardiovasc Res 70:146–157.

19. Saiki A, Watanabe F, Murano T, Miyashita Y, Shirai K (2006) Hepatocyte growth factorsecreted by cultured adipocytes promotes tube formation of vascular endothelial cellsin vitro. Int J Obes (London) 30:1676–1684.

20. Samad F, Pandey M, Loskutoff DJ (1998) Tissue factor gene expression in the adiposetissues of obese mice. Proc Natl Acad Sci USA 95:7591–7596.

21. Voros G, et al. (2005) Modulation of angiogenesis during adipose tissue developmentin murine models of obesity. Endocrinology 146:4545–4554.

22. Brakenhielm E, et al. (2004) Angiogenesis inhibitor, TNP-470, prevents diet-inducedand genetic obesity in mice. Circ Res 94:1579–1588.

23. Rupnick MA, et al. (2002) Adipose tissue mass can be regulated through the vascula-ture. Proc Natl Acad Sci USA 99:10730–10735.

24. Yancopoulos GD, et al. (2000) Vascular-specific growth factors and blood vesselformation. Nature 407:242–248.

25. Maisonpierre PC, et al. (1997) Angiopoietin-2, a natural antagonist for Tie2 thatdisrupts in vivo angiogenesis. Science 277:55–60.

26. Gale NW, Yancopoulos GD (1999) Growth factors acting via endothelial cell-specificreceptor tyrosine kinases: VEGFs, angiopoietins, and ephrins in vascular development.Genes Dev 13:1055–1066.

27. Gale NW, et al. (2002) Complementary and coordinated roles of the VEGFs andangiopoietins during normal and pathologic vascular formation. Cold Spring HarborSymp Quant Biol 67:267–273.

28. Seo S, et al. (2006) The forkhead transcription factors, Foxc1 and Foxc2, are required forarterial specification and lymphatic sprouting during vascular development. Dev Biol294:458–470.

29. Carlsson E, Almgren P, Hoffstedt J, Groop L, Ridderstrale M (2004) The FOXC2 C-512Tpolymorphism is associated with obesity and dyslipidemia. Obes Res 12:1738–1743.

30. Pajukanta P, et al. (2003) Combined analysis of genome scans of dutch and finnishfamilies reveals a susceptibility locus for high-density lipoprotein cholesterol on chro-mosome 16q. Am J Hum Genet 72:903–917.

31. Ridderstrale M, et al. (2002) FOXC2 mRNA expression and a 5� untranslated regionpolymorphism of the gene are associated with insulin resistance. Diabetes 51:3554–3560.

32. Yildirim-Toruner C, et al. (2004) A novel frameshift mutation of FOXC2 gene in a familywith hereditary lymphedema-distichiasis syndrome associated with renal disease anddiabetes mellitus. Am J Med Genet A 131:281–286.

33. Fang J, et al. (2000) Mutations in FOXC2 (MFH-1), a forkhead family transcriptionfactor, are responsible for the hereditary lymphedema-distichiasis syndrome. Am JHum Genet 67:1382–1388.

34. Cederberg A, et al. (2001) FOXC2 is a winged helix gene that counteracts obesity,hypertriglyceridemia, and diet-induced insulin resistance. Cell 106:563–573.

35. Cao R, et al. (2002) Angiogenesis stimulated by PDGF-CC, a novel member in the PDGFfamily, involves activation of PDGFR-alphaalpha and -alphabeta receptors. FASEB J16:1575–1583.

36. Cao Y, et al. (1998) Vascular endothelial growth factor C induces angiogenesis in vivo.Proc Natl Acad Sci USA 95:14389–14394.

37. Vajkoczy P, et al. (2002) Microtumor growth initiates angiogenic sprouting withsimultaneous expression of VEGF, VEGF receptor-2, and angiopoietin-2. J Clin Invest109:777–785.

38. Oliner JD, Min H (2006) US Patent 7,138,370.39. Chernogubova E, Cannon B, Bengtsson T (2004) Norepinephrine increases glucose

transport in brown adipocytes via beta3-adrenoceptors through a cAMP, PKA, andPI3-kinase-dependent pathway stimulating conventional and novel PKCs. Endocrinol-ogy 145:269–280.

40. Bouloumie A, Lolmede K, Sengenes C, Galitzky J, Lafontan M (2002) Angiogenesis inadipose tissue. Ann Endocrinol (Paris) 63:91–95.

41. Hutley LJ, et al. (2001) Human adipose tissue endothelial cells promote preadipocyteproliferation. Am J Physiol 281:E1037–E1044.

42. Varzaneh FE, Shillabeer G, Wong KL, Lau DC (1994) Extracellular matrix componentssecreted by microvascular endothelial cells stimulate preadipocyte differentiation invitro. Metabolism 43:906–912.

43. Cohen B, et al. (2001) Leptin induces angiopoietin-2 expression in adipose tissues. J BiolChem 276:7697–7700.

44. Asano A, Morimatsu M, Nikami H, Yoshida T, Saito M (1997) Adrenergic activation ofvascular endothelial growth factor mRNA expression in rat brown adipose tissue:Implication in cold-induced angiogenesis. Biochem J 328:179–183.

45. Fredriksson JM, Lindquist JM, Bronnikov GE, Nedergaard J (2000) Norepinephrineinduces vascular endothelial growth factor gene expression in brown adipocytesthrough a beta-adrenoreceptor/cAMP/protein kinase A pathway involving Src butindependently of Erk1/2. J Biol Chem 275:13802–13811.

46. Tonello C, et al. (1999) Role of sympathetic activity in controlling the expression ofvascular endothelial growth factor in brown fat cells of lean and genetically obese rats.FEBS Lett 442:167–172.

47. Ridgway J, et al. (2006) Inhibition of Dll4 signalling inhibits tumour growth byderegulating angiogenesis. Nature 444:1083–1087.

48. Holash J, et al. (1999) Vessel cooption, regression, and growth in tumors mediated byangiopoietins and VEGF. Science 284:1994–1998.

49. Gerhardt H, et al. (2003) VEGF guides angiogenic sprouting utilizing endothelial tip cellfilopodia. J Cell Biol 161:1163–1177.

50. Lindahl P, Johansson BR, Leveen P, Betsholtz C (1997) Pericyte loss and microaneurysmformation in PDGF-B-deficient mice. Science 277:242–245.

51. Silha JV, Krsek M, Sucharda P, Murphy LJ (2005) Angiogenic factors are elevated inoverweight and obese individuals. Int J Obes (London) 29:1308–1314.

52. Geva E, et al. (2002) Human placental vascular development: Vasculogenic and angio-genic (branching and nonbranching) transformation is regulated by vascular endo-thelial growth factor-A, angiopoietin-1, and angiopoietin-2. J Clin Endocrinol Metab87:4213–4224.

53. Hu B, et al. (2003) Angiopoietin-2 induces human glioma invasion through the acti-vation of matrix metalloprotease-2. Proc Natl Acad Sci USA 100:8904–8909.

10172 � www.pnas.org�cgi�doi�10.1073�pnas.0802486105 Xue et al.

Dow

nloa

ded

by g

uest

on

June

11,

202

0