Alternatively spliced tissue factor inducesangiogenesis through integrin ligationY. W. van den Berga, L. G. van den Hengela, H. R. Myersa, O. Ayachia, E. Jordanovab, W. Rufc, C. A. Spekd, P. H. Reitsmaa,V. Y. Bogdanove, and H. H. Versteega,1

aThe Einthoven Laboratory for Experimental Vascular Medicine and bDepartment of Pathology, Leiden University Medical Center, Albinusdreef 2, 2333 ZA,Leiden, The Netherlands; cDepartment of Immunology and Microbial Science, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA92037; dCenter for Experimental and Molecular Medicine, Academic Medical Center, Meibergdreef 9, 1105 AZ, Amsterdam, The Netherlands; and eDivisionof Hematology/Oncology, Department of Internal Medicine, University of Cincinnati College of Medicine, 3125 Eden Avenue, Cincinnati, OH 45267

Edited by Charles T. Esmon, Oklahoma Medical Research Foundation, Oklahoma City, OK, and approved September 25, 2009 (received for reviewMay 15, 2009)

The initiator of coagulation, full-length tissue factor (flTF), in complexwith factor VIIa, influences angiogenesis through PAR-2. Recently, analternatively spliced variant of TF (asTF) was discovered, in which partof the TF extracellular domain, the transmembrane, and cytoplasmicdomains are replaced by a unique C terminus. Subcutaneous tumorsproduced by asTF-secreting cells revealed increased angiogenesis, butit remained unclear if and how angiogenesis is regulated by asTF.Here, we show that asTF enhances angiogenesis in matrigel plugs inmice, whereas a soluble form of flTF only modestly enhances angio-genesis. asTF dose-dependently upregulates angiogenesis ex vivoindependent of either PAR-2 or VIIa. Rather, asTF was found toligate integrins, resulting in downstream signaling. asTF-�V�3 inte-grin interaction induces endothelial cell migration, whereas asTF-dependent formation of capillaries in vitro is dependent on �6�1integrin. Finally, asTF-dependent aortic sprouting is sensitive to �1and �3 integrin blockade and a TF-antibody that disrupts asTF-integrin interaction. We conclude that asTF, unlike flTF, does notaffect angiogenesis via PAR-dependent pathways but relies on inte-grin ligation. These findings indicate that asTF may serve as a targetto prevent pathological angiogenesis.

cancer � coagulation � endothelial cells � integrins

The development of blood vessels out of existing vessels,termed angiogenesis, occurs in embryonic development,

wound healing, and cancer (1). Angiogenesis depends on initialtip cell migration from the existing vessel, followed by migrationof stalk cells, which divide and form a capillary (2), and therecruitment of pericytes aligning the capillary. Angiogenesis isregulated by such proteins as vascular endothelial growth factor(VEGF), metalloproteinases, and interleukin-8 (1, 3, 4), but alsodepends on integrins. In particular, �1- and �3-type integrinsplay a role in endothelial and pericyte migration as well ascapillary formation (5, 6).

Upon vessel damage, tissue factor (TF) together with factorVII(a), generates factor Xa, thrombin, and fibrin, yielding a hemo-static plug (7). TF is indispensable for life since TF-deficientembryos die in utero (8). Moreover, TF affects tumor angiogenesisin colorectal cancer and breast cancer models, through TF:VIIa-dependent protease-activated receptor-2 (PAR-2) activation (9–11), resulting in expression of VEGF, IL-8, MMP-7, and CXCL-1(10–13). In addition, TF binds �3�1 and �6�1 integrins, anddisruption of this complex downregulates pro-angiogenic signalingand suppresses tumor growth in vivo (11, 14).

Recently, a TF isoform, asTF, which results from alternativesplicing, was discovered, in which the transmembrane and cyto-plasmic domains are replaced by a unique 40 amino acid, C-terminal domain, rendering asTF soluble (15). asTF incorporates inthrombi and asTF secreted by endothelium exhibits pro-coagulantactivity (16). While some groups failed to detect pro-coagulantactivity of asTF in transfected cells (15–17), native asTF can bepro-coagulant (16) and is expressed in malignancies such as pan-creatic and lung cancer (18, 19). Hobbs et al. demonstrated that

pancreatic cancer cells transfected to express asTF produce moreblood vessels (20), but it remained mechanistically unclear if andhow angiogenesis is regulated by asTF. One possibility is that asTFstimulates cancer cells to produce angiogenic factors, but it is alsoplausible that asTF enhances angiogenesis via paracrine stimulationof endothelial cells. Moreover, the role of VIIa, PAR-2 activationand downstream coagulation activation remains obscure.

In this study, we report that asTF enhances angiogenesis in vivo,ex vivo, and in vitro independent of downstream coagulation factorsor activation of PAR-2, but dependent on �v�3 and �6�1 integrinfunction.

ResultsasTF Induces Angiogenesis In Vivo and Ex Vivo. To study a potentialeffect of asTF on angiogenesis, we evaluated the properties ofrecombinant asTF. Purified asTF migrated as a single band ofapproximately 26 kDa, corresponding to its predicted molecularweight, on SDS/PAGE (Fig. S1A). The preparation reacted with amonoclonal antibody against the N-terminal part (9C3), but notwith antibodies against the C-terminal part of full-length TF (flTF)(10H10, 5G9) (Fig. S1B), confirming that the expressed protein wasindeed asTF. Analogous immunoreactivity of these antibodies wasobserved with asTF expressed in eukaryotic cells (Fig. S1C).

Whether asTF induced angiogenesis was tested in a matrigel plugassay. Matrigel supplemented with asTF, VEGF, or buffer controlwas injected into C57BL/6 mice, and angiogenesis was analyzed 7days later. Both asTF and VEGF enhanced angiogenesis comparedto the buffer control (Fig. 1A), showing that asTF can influenceangiogenesis in the absence of tumor cells. Similar results wereobtained in an ex vivo aortic sprouting model (21) in whichsegmented aortas isolated from C57BL/6 mice were implanted inmatrigel containing asTF, VEGF, or control buffer (Fig. 1B).Similar ex vivo experiments revealed that asTF concentration-dependently enhanced the formation of sprouts at concentrationsas low as 1 nM (Fig. 1C). Active site-blocked VIIa (VIIai) orhirudin, inhibitors of VIIa and thrombin, respectively, did notinhibit asTF-dependent sprouting, indicating that generation ofcoagulation proteases was not required (Fig. 1C). Moreover, addingasTF in combination with VIIa did not enhance aortic sproutingcompared to asTF alone (Fig. 1D), supporting our finding that asTFinduced angiogenesis independent of coagulation activation. Pres-ence of asTF in the matrigel, but not in the media placed on top ofthe matrigel, enhanced angiogenesis, suggesting that asTF’s pres-

Author contributions: Y.W.v.d.B., P.H.R., and H.H.V. designed research; Y.W.v.d.B.,L.G.v.d.H., H.R.M., O.A., and E.J. performed research; W.R., C.A.S., and V.Y.B. contributednew reagents/analytic tools; Y.W.v.d.B., L.G.v.d.H., H.R.M., O.A., E.J., and H.H.V. analyzeddata; and Y.W.v.d.B., P.H.R., V.Y.B., and H.H.V. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

1To whom correspondence should be addressed. E-mail: h.h.versteeg@lumc.nl.

This article contains supporting information online at www.pnas.org/cgi/content/full/0905325106/DCSupplemental.

www.pnas.org�cgi�doi�10.1073�pnas.0905325106 PNAS � November 17, 2009 � vol. 106 � no. 46 � 19497–19502

MED

ICA

LSC

IEN

CES

ence in the extracellular matrix is required (Fig. 1E). RecombinantasTF was produced in E. coli, thus LPS contamination couldpotentially influence sprout formation. However, inclusion of theLPS inhibitor polymyxin B did not affect aortic sprouting inducedby asTF (Fig. S2A), and similar sprouting was observed in aserum-free system (Fig. S2B), a condition that does not allow forefficient LPS signaling (22).

asTF-Dependent Angiogenesis Is Not Dependent on PAR-2. Next, weexamined the mechanism behind asTF-induced angiogenesis. Un-der basal conditions, endothelial cells express low levels of PAR-2,but these levels go up during angiogenesis (10). flTF in complexwith VIIa was previously shown to promote PAR-2-dependentangiogenesis, presumably through activation of, among others, theMAP kinase signal transduction pathway. To study the influence ofasTF on PAR-2 activation in endothelial cells, we used an endo-thelial cell line (ECRF) adenovirally transduced to express PAR-2,after which these cells were stimulated with asTF, VIIa, or thecombination of asTF and VIIa. A truncated soluble form of flTF(sTF) in complex with VIIa and a PAR-2 agonist, SLIGRL, wereused as positive controls. VIIa, asTF, or the combination of asTF

and VIIa were not able to induce phosphorylation of MAP kinase,whereas VIIa in complex with sTF or SLIGRL elicited robustphosphorylation within 10 min (Fig. 2A). Similar results wereobtained using primary endothelial cells (HUVECs, Fig. S3).

Since angiogenesis is dependent on endothelial cell prolifer-ation, we also determined whether asTF induced proliferation inPAR-2 transduced cells. FCS induced significant cell prolifera-tion, but incubation with asTF, VIIa or the combination of asTFand VIIa were without effect (Fig. 2B and Fig. S3). In agree-ment, we found that aortic segments isolated from PAR-2-/-mice displayed the same increase in asTF-dependent sproutformation as wild-type segments, although basal sprouting inPAR-2-/- segments was lower (Fig. 2C). Thus, the effect of asTFon angiogenesis is not dependent on VIIa-mediated activation ofPAR-2. Importantly, the effect of asTF on sprouting was muchmore profound than that observed using sTF, either in wild-typeaortas or PAR-2-/- aortas (Fig. 2C), suggesting that asTF is themain TF isoform that mediates angiogenesis.

asTF Binds Endothelial Integrins. Since flTF was previously shown tobind integrins, the effect of asTF on angiogenesis may be dependenton integrin ligation. To study this, tissue culture plates were coatedwith asTF and blocked with BSA, after which ECRF cells wereseeded. ECRF cells bound time-dependently to asTF-coated plates,whereas cells bound very poorly to BSA-only treated plates (Fig.3A). Cells bound to asTF displayed enhanced phosphorylation ofFocal Adhesion Kinase (FAK), p42/p44 MAP kinase, p38 MAPkinase, and Akt—events commonly elicited by integrin ligation(Fig. 3B). Inclusion of polymyxin B did not influence cell adhesionto asTF (Fig. S4A). Depletion of asTF from our preparation usingnickel-NTA abolished cell adhesion, whereas incubation of asTFpreparations with non-nickel bound control beads did not preventcell adhesion (Fig. S4B). Moreover, cell adhesion to asTF was fully

Fig. 1. asTF induces angiogenesis in vivo and ex vivo. (A) Matrigel containing100 nM asTF, 50 ng/mL VEGF, or buffer control was injected s.c. into mice.FITC-dextran was tail vein-injected before sacrifice. Extracted plugs wereexamined using MZ 16FA stereo microscope and DFC 420C camera (Leica), andthe number of invading vessels was counted. The graph shows quantificationof this experiment (n � 10 � SEM) (B) Aortic segments from C57BL/6 mice wereimplanted into matrigel supplemented with solvent control, asTF, or VEGF.Outgrowing sprouts were visualized and counted on day 5. The graph showsquantification of this experiment (n � 8 � SEM). (C) Aortic segments inmatrigel supplemented with solvent control or various concentrations of asTF.The coagulation inhibitors VIIai (100 nM) or hirudin (500 nM) were included insome of the conditions. (D) Aortic segments in matrigel supplemented with100 nM VIIa, 100 nM asTF, or the combination of asTF and VIIa. (E) Sproutingoccurs in asTF-containing matrigel, but not when media overlying the matri-gel contains asTF.

Fig. 2. asTF does not activate PAR-2. (A) PAR-2-expressing ECRF cells werestimulated with 100 nM asTF � 100 nM VIIa, VIIa alone, 100 nM truncated flTF(sTF) � VIIa, or SLIGRL (100 �M). MAPK phosphorylation was determined byWestern blotting. (B) Cells were serum starved and stimulated with asTF �VIIa, VIIa alone, or 15% FCS. Cell proliferation was assessed using MTT assays.(C) Aortic segments from wild-type or PAR-2-/- C57BL/6 mice were implantedin matrigel supplemented with solvent control, asTF, or sTF, and the numberof sprouts was determined as described.

19498 � www.pnas.org�cgi�doi�10.1073�pnas.0905325106 van den Berg et al.

inhibited in the presence of the anti-TF antibody 6B4, whichinterferes with TF-integrin binding (14) (Fig. 3C). Thus, enhance-ment of cell adhesion was entirely dependent on the presence ofasTF. The use of sTF yielded similar results in this assay.

To identify the responsible integrins, ECRF cells were preincu-bated with specific blocking antibodies. �2 or �4 blockade did notinhibit cell adhesion to asTF, �3 blockade modestly inhibited celladhesion, and �1 blockade resulted in a 50% reduction of celladhesion (Fig. 3D). The combination of �1 and �3 blockadereduced cell adhesion to basal levels. Similar results were obtainedusing a murine endothelial cell line (Fig. S4C), validating the use ofhuman asTF in murine models such as the matrigel plug assay andthe aortic ring assays. In conclusion, both �1 and �3 integrins bindto asTF to mediate cell adhesion.

asTF Induces Endothelial Cell Migration. During angiogenesis, en-dothelial tip cells initially migrate out of the intima and subse-quently, endothelial cell proliferation and differentiation resultin capillary formation. Using a transwell assay, we assessedwhether asTF induces cell migration. asTF induced a 4-foldupregulation in ECRF migration when present in the lowercompartment, and cells appeared rounded (Fig. 4A). When thelower sides of transwells were coated with asTF, a 10-foldupregulation of cell migration was observed while cells displayeda flattened morphology.

Pericytes also contribute to angiogenesis. We found that asTF didnot facilitate pericyte migration, whereas a mix of extracellularmatrix proteins optimized for pericyte binding supported potentmigration (Fig. 4B). Thus, asTF does not induce pericyte migration,but selectively leads to migration of endothelial cells. Interestingly,endothelial migration was reduced on sTF when compared to asTF,which may explain the lower angiogenic potential of sTF (Fig. 4C).

The identity of integrins involved in asTF-enhanced migrationwas determined using integrin-blocking antibodies. �1, �2, and �4blocking antibodies, as well as a control antibody, were unsuccessfulin blocking migration, but �3 blockade fully inhibited migration(Fig. 4D). An anti-�v�3 antibody (LM609) also blocked asTF-dependent migration, demonstrating that �v�3 is required forasTF-dependent migration (Fig. 4E). Similarly, �v�3 blockadeinhibited migration of primary HUVECs (Fig. S5A). It is importantto note that �v�3 integrin blockade by LM609 induces apoptosis inendothelial cells (23), which may be responsible for the inhibitingeffect on asTF-induced migration, but apoptosis was not observed(Fig. S5B). Polymyxin B did not affect migration, confirming thatpossible traces of LPS did not contribute to this effect (Fig. 4E).

To investigate the signal transduction pathways involved inasTF-dependent migration, we preincubated cells with specifickinase inhibitors. The p38 MAP kinase inhibitor SB203580 and thePI3 kinase inhibitor LY294002, but not the p42/p44 MAP kinaseinhibitor, completely abolished migration (Fig. 4F). An Akt inhib-itor only partially inhibited asTF-dependent migration (Fig. 4G),suggesting that PI3-kinase acts through Akt-dependent and inde-pendent pathways.

asTF Enhances Endothelial Capillary Formation. Next, we determinedasTF’s potential to support capillary formation by seeding ECRFcells on matrigel supplemented with asTF or control buffer. asTFenhanced the formation of capillaries 2.5-fold (Fig. 5 A and B).When asTF was added to the medium, capillary formation wasinhibited, most likely by binding and blocking the integrins requiredfor capillary formation on matrigel. Surprizingly, �1 but not �3blockade inhibited capillary formation to subbasal levels in ECRFand primary HUVECs (Fig. 5B and Fig. S6). Apparently, endo-thelial cell migration and capillary formation induced by asTF weredependent on different integrin heterodimers. Blockade of �3 and�6 integrin subunits that were previously shown to interact withflTF (10) was also tested. Integrin �6 blockade inhibited asTF-induced capillary formation to subbasal levels (Fig. 5C). In contrast,�3 blockade modestly inhibited capillary formation, suggesting that�3 integrins do not play a major role in this process.

Matrigel primarily consists of collagen and laminin, thus �3 and�6 may also participate in the attachment of cells to matrigel. Toverify whether the inhibition of capillary formation by �3 and �6blockade was due to diminished adhesion to asTF, we tested �3 and�6 blockade on cell adhesion to asTF. �6, but not �3, blockadepartially inhibited cell binding to asTF (Fig. 5D). Cell adhesion wasreduced to basal levels upon inclusion of a �3 blocking antibody,supporting the conclusion that endothelial cells employ differentintegrin heterodimers, namely �6�1 and �v�3, but only require�6�1 integrin for asTF-induced capillary formation. Preincubationof ECRF cells with SB203580, LY294002 and PD98059 showed thatasTF-induced capillary formation was dependent on p42/p44 MAPkinase and PI3 kinase action, but not the promigratory p38 MAPkinase cascade (Fig. 5E), demonstrating that distinct integrin-dependent pathways are involved in asTF-dependent migration andcapillary formation. In support of that, cells allowed to formcapillaries on matrigel supplemented with asTF showed higherlevels of p42/p44 and Akt, but not p38 phosphorylation.

Similar to what was observed in aortic sprouting experiments,VIIai, hirudin, and polymyxin B did not influence capillary forma-tion (Fig. 5 C and E), demonstrating that asTF-dependent capillaryformation was not dependent on coagulation activation and/or LPScontamination.

asTF-Enhanced Aortic Sprouting Is Dependent on Integrins. We nextassessed the relative contributions of �1 and �3 integrins toasTF-dependent sprout formation in matrigel, using integrin block-ing antibodies. Blockade of �1 inhibited basal sprouting andasTF-induced sprouting to similar levels (Fig. 6A), suggesting thatboth matrigel and asTF ligation of integrins was inhibited. Blockade

Fig. 3. asTF binds integrins on endothelial cells. (A) Cells were seeded onBSA- or asTF-coated culture wells and adhered cells were counted at theindicated times. (B) Cells were left to adhere to BSA- or asTF-coated wells forthe indicated times, lysed, and the lysates were checked for FAK, p42/22MAPK, p38 MAPK, and c-Akt phosphorylation on Western blot. Total levels ofthese proteins were assessed to verify equal loading. (C) Culture wells werecoated with 50 �g/mL asTF or sTF and cell adhesion was assayed in theabsence/presence of 100 �g/mL 6B4. (D) Endothelial cells were preincubatedwith integrin-blocking antibodies and seeded onto BSA- or asTF-coated wells.Adhered cells were counted.

van den Berg et al. PNAS � November 17, 2009 � vol. 106 � no. 46 � 19499

MED

ICA

LSC

IEN

CES

of �3 did not affect basal sprouting, but reduced asTF-dependentsprouting to basal levels. This is expected because matrigel containsrelatively little �3-ligating matrix proteins. 6B4, which inhibitsasTF-integrin interaction, similarly reduced asTF-dependent aorticsprouting (Fig. 6B).

asTF is found in thrombi and could therefore potentiate angio-genesis in a fibrin matrix. asTF dose-dependently upregulatedsprouting in a fibrin milieu, similarly to that observed in matrigel(Fig. 6C); sTF also enhanced sprouting, but at much higherconcentrations. Inclusion of 6B4 and blockade of �1 and �3integrins led to inhibition of asTF-dependent sprouting (Fig. 6D).In contrast to what was observed in matrigel, �3 but not �1blockade inhibited sprouting to subbasal levels, reflecting therequirement of �3 integrin activation by fibrin. The asTF/integrin-blocking antibody 6B4 reduced vessel formation in vivo, andalthough sTF modestly induced vessel formation, 6B4 did not havean effect (Fig. 6E). Thus, integrins are instrumental in asTF-dependent sprouting in matrigel and fibrin and in vivo angiogenesis.

Intratumoral asTF Concentrations. To test whether asTF concentra-tions that induce angiogenesis are pathologically relevant, weassessed intratumoral asTF concentrations in a set of cervicaltumors. We chose cervical cancer since high levels of asTF mRNAwere observed in tumors from seven patients (Fig. 7A). Ten tumorspecimens were homogenized and analyzed for asTF expression byWestern blot using a specific anti-asTF antibody. Bands werequantified using an asTF concentration curve. All but one specimenrevealed asTF concentrations that fell within the range that inducesangiogenesis (Fig. 7B), indicating that the observed asTF-inducedangiogenesis is pathobiologically relevant.

DiscussionHere, we show that asTF induces angiogenesis in a matrigel plugassay and aortic sprouting in matrigel and fibrin. asTF inducedangiogenesis in a VIIa/PAR-2-independent manner but actedthrough ligation of �1 and �3 integrins, at concentrations as low as1 nM. asTF-induced endothelial cell migration and capillary for-mation was dependent on �v�3 and �6�1, respectively. Ourconclusion that asTF binds integrins is based on the followingobservations: (i) endothelial cell adhesion to asTF as well asasTF-induced capillary formation and aortic sprouting were po-tently inhibited by integrin blocking antibodies, (ii) inclusion ofasTF in the medium blocked formation of capillaries, most likely byblocking integrins, and (iii) an antibody that inhibits TF-integrininteraction similarly inhibited sprouting and in vivo angiogenesis.Only the combination of �1 and �3 integrin blocking antibodiescompletely abrogated asTF-dependent cell adhesion, thus both �1and �3 integrin appear to physically bind to asTF. This distinguishesasTF from matrigel and fibrin which predominantly activate �1 and�3 integrins, respectively. Although the importance of other inte-grin ligands, for example fibronectin and vitronectin, cannot for-mally be excluded, they are not the constituents of the tumorextracellular matrix (matrigel) or the provisional angiogenic fibrinmatrix, and therefore play no role in our model. Endothelial cellmigration was dependent on �v�3 integrin, p38 MAP kinase, andPI3-kinase, whereas capillary formation was dependent on �6�1,p42/p44 MAP kinase, and PI3-kinase. It is noteworthy that tip cellmigration and capillary formation represent different stages ofangiogenesis and involves different integrins (24). Concordantly,�v�3 ligation results in directional migration, whereas �1 ligationinduces non-directional migration that is likely to be more impor-tant for capillary formation (25). In wound healing, �v�3 is focally

Fig. 4. asTF induces �v�3-dependent endothelial cell migration. (A) Transwell inserts were coated with BSA (negative control), 50 �g/mL asTF, or 1% gelatin(positive control). Endothelial cells were left to migrate for 5 h. Cells were also seeded in uncoated inserts and migration was induced by placing 100 nM asTFinto the lower well. Note the rounded-up morphology of cells migrating toward a gradient of asTF versus the flattened morphology of cells migrating throughasTF-coated inserts. The graph on the right shows a quantification of these results (n � 9). (B) Pericytes were seeded into asTF-coated inserts or inserts coatedwith a pericyte extracellular matrix protein mix (Cell Systems) and left to migrate for 5 h. (C) Effect of asTF versus sTF coating on migration of ECRF cells. (D)Endothelial cells were preincubated with � integrin-blocking antibodies and seeded into asTF-coated transwell inserts. Migration was assessed as describedabove. (E) Endothelial cells were preincubated with �3 or �v�3-blocking antibodies. asTF-induced cell migration was assessed as described. Inclusion of PM inthe assay did not inhibit migration. (F) Effects of MAPK pathway inhibitor PD98059 (20 �M), p38 MAPK inhibitor SB203580 (10 �M), or the PI-3 kinase inhibitorLY294002 (10 �M) on asTF-induced migration. (G) Effect of c-Akt inhibitor (10 �M) on asTF-induced migration.

19500 � www.pnas.org�cgi�doi�10.1073�pnas.0905325106 van den Berg et al.

expressed at the tips of invading capillary sprouts whereas �1integrins are broadly expressed on the entire capillary (26). Itappears that asTF-driven migration through �v�3 integrin mimicsendothelial tip cell migration, whereas �1-dependent capillaryformation through �6�1 recapitulates the formation of vessels fromnon-tip cell endothelial cells through migration and differentiation.In aortic sprouting assays, which combine these features, asTF-induced angiogenesis was indeed sensitive to both �1 and �3blockade.

flTF on tumor cells has also been shown to induce angiogenesis,but in a mechanistically different manner. Like asTF, flTF interactswith integrins, but rather than serving as receptors to TF, integrin�3�1, and �6�1 control flTF-VIIa signaling via PAR-2 (10). Thefact that flTF is membrane bound and binds integrins expressed onthe same cell surface, whereas asTF is a soluble protein capable ofbeing deposited into the extracellular matrix; that is, at sites notaccessible to flTF, suggests flTF and asTF may bind integrins indistinct orientations, thus determining specificity of integrins forthese two TF forms. Our results, showing a lack of asTF-�3�1integrin interaction, support this concept.

Another important finding is that asTF is a much more potentinducer of angiogenesis that a truncated form of flTF (sTF),probably because asTF is more pro-migratory than sTF, renderingextracellular asTF the principal candidate to engage in integrin-dependent angiogenesis, especially because sTF is not known tooccur naturally in vivo. The underlying basis may be the differential

Fig. 5. asTF enhances �6�1-dependent endothelial capillary formation. (A)Endothelial cells were seeded on matrigel supplemented with solvent controlor 100 nM asTF. Photographs were taken after O/N incubation. (B) Cells wereseeded on matrigel containing solvent control or asTF. Alternatively, asTF wasadded tot the medium containing the cells and the mix was added onto amatrigel layer without asTF. Cells were also preincubated with � integrinblocking antibodies and seeded onto asTF-containing matrigel. Total capillarylength was quantified. (C) Cells were preincubated with �3 or �6-blockingantibodies and seeded onto asTF-containing matrigel. Non-treated cells werealso seeded onto asTF-cotaining matrigel in the presence of VIIai or hirudin.(D) Cells were preincubated with �3 or �6-blocking antibodies in the presenceor absence of �3-blocking antibody before seeding of cells onto immobilizedasTF. Adhered cells were counted after 4 h. (E) Effect of PD98059, SB203580 orLY294002 on asTF-induced capillary formation. (F) asTF-dependent phosphor-ylation of MAPK, p38, and Akt in capillary-forming endothelial cells. Cellswere left to form capillaries for the times indicated on matrigel in thepresence/absence of 100 nM asTF.

Fig. 6. asTF induces integrin-dependent angiogenesis in vivo and ex vivo. (A)Wild-type aortic segments were implanted in solvent control-containing ma-trigel or asTF-containing matrigel. Murine �1 or �3-blocking antibodies wereadded to the matrigel. The number of sprouts was determined as described.(B) Wild-type aortic segments were implanted in solvent control-containingmatrigel or matrigel containing 6B4-preincubated (100 �g/mL) asTF. (C) Wild-type aortic segments in fibrin supplemented with various concentrations ofasTF or sTF. (D) Wild-type aortic segments in fibrin � 100 nM asTF, in thepresence/absence of 50 �g/mL integrin-blocking antibodies or 100 �g/mL 6B4.(E) In vivo matrigel plug assay; matrigel was supplemented with control bufferVEGF, asTF or sTF. asTF and sTF were also preincubated with 100 �g/mL 6B4before addition to the matrigel. Vessels were counted as described above.

Fig. 7. Cervical tumors express high levels of asTF. (A) mRNA obtained fromseven cervival tumors was subjected to RT-PCR using asTF- and �-actin-specificprimers. (B) Ten cervical tumor specimens were homogenized and lysed insample buffer. Lysates were analyzed for asTF expression on Western blot, andasTF bands were quantified using an asTF concentration curve.

van den Berg et al. PNAS � November 17, 2009 � vol. 106 � no. 46 � 19501

MED

ICA

LSC

IEN

CES

integrin activation by secreted asTF and membrane bound flTF.Indeed, asTF differs from flTF such that asTF displays a decreasedaffinity for VII.

In our experiments, asTF induced angiogenesis at concentrationsas low as 1 nM. Interestingly, in a panel of 10 cervical tumors, wefound that nine out of 10 tumors contained asTF at levels in therange within which asTF is pro-angiogenic, rendering asTF apotential key player in tumor angiogenesis. In agreement, asTF isexpressed in pancreatic cell lines (18), squamous cell carcinoma ofthe lung (27), and advanced stages of non-small cell lung cancer(19); asTF appears to drive tumor growth in xenograft models (20).It remains controversial whether asTF always displays coagulantactivity and can be effectively secreted (16). Regardless of whetherasTF is procoagulant in vivo, we did not observe extensive thrombusformation in our in vivo experiments using either asTF or sTF,ruling out that these proteins induce thrombus formation in newlyformed vessels. Szotowski et al. (17) showed that asTF can besecreted after appropriate stimulation and asTF is found in normalhuman plasma, making up approximately 30% of the total blood-borne TF pool (15). Although the cue to asTF secretion in tumorsremains to be identified, hypoxia and/or tumor-expressed cytokinesmay contribute to this phenomenon.

Materials and MethodsMaterials, Cell Lines. Protein Expression, and Survival Assay. Materials, cell lines,protein expression and survival assays are described in detail in the SI Text.

Endothelial Cell Adhesion Assays. Ninety-six-well plates were coated with 50�g/mL asTF or 10% BSA as negative control. Uncoated areas were blocked with10% BSA. Cells (20,000) were seeded per well and plates were incubated at 37 °Cfor 4 h. In some experiments, cells were preincubated with 50 �g/mL integrinblocking antibodies. Photographs were taken and cells adopting a flattenedmorphology were counted. Data are shown as mean � SD, n � 4.

Cell Migration Assays. Cell migration was assessed using a transwell assay.Polycarbonatemembranetranswell inserts (8.0 �m)(CorningCostar,CorningLifeSciences, Corning) were coated with 1% gelatin or asTF (50 �g/mL). Cells (25,000)wereseededper insertaftera20-minpretreatmentwith integrinblockingmAb’s,when appropriate. Cells were allowed to migrate for 5 h at 37 °C; migrated cellswere fixed, stained with crystal violet for 10 min, and counted per field of 40�magnification.

In Vitro Capillary Formation Assay. Ninety-six-well plates were coated with 50 �Lmatrigel, supplemented with asTF according to the experimental conditions.Endothelial cells (20,000) were seeded after a 20-min pretreatment with blockingantibodies when appropriate, allowed to form capillaries for 18 h (ECRF) or 6 h(HUVECs), and the lengths of tubular networks was measured.

Mouse Aortic Ring Assay. Mouse thoracic aortas were isolated and cleaned of thesurroundingtissue inserum-freeRPMI(Invitrogen)containing50�g/mLpenicillinand 50 �g/mL streptomycin. Dissected aortas were flushed, cut into equal seg-ments, embedded in matrigel, and covered with EBM containing 2% serum andpenicillin/streptomycin. Sprouts were counted on day 5. Aortas were also em-bedded in fibrin that was prepared by mixing 2 mg/mL fibrinogen with 0.1 U/mLthrombin. Aortas were then overlayed with medium containing 5 U/mL hirudin.

Matrigel Plug Assay. Eight-week-old C57Bl6 mice (n � 10 per group) wereanesthetized with isoflurane and injected s.c. into the flank with 0.5 mL ice-coldmatrigel. Matrigel was either supplemented with 100 nM asTF or sTF in thepresence/absence of 100 �g/mL 6B4 and 50 ng/mL mouse recombinant VEGF orPBS. After 7 days, 150 �L FITC-Dextran (30 mg/mL) was injected into the tail vein.After 15 min the animals were killed, implants were extracted, fixed in 10%formalin, and analyzed on a Leica MZ16 FA stereomicroscope.

Statistics. Data are mean �/- SD unless otherwise stated. Statistical analysiswas performed using Student’s t-test. *, P � 0.05; **, P � 0.01; ***, P � 0.001.

ACKNOWLEDGMENTS. We acknowledge Lars C. Petersen (Novo Nordisk) for thegift of VIIai. H.H.V. is supported by The Netherlands Scientific Organisation (grantno.916.76.012).W.R. is supportedbytheNational InstitutesofHealth(NIH) (grantnos. HL060742 and HL016411). V.Y.B. is supported by NIH (grant no. HL094891).

1. Carmeliet P (2005) Angiogenesis in life, disease and medicine. Nature 438:932–936.2. Holderfield MT, Hughes CC (2008) Crosstalk between vascular endothelial growth

factor, notch, and transforming growth factor-beta in vascular morphogenesis. CircRes 102:637–652.

3. Rundhaug JE (2005) Matrix metalloproteinases and angiogenesis. J Cell Mol Med9:267–285.

4. Li A, Dubey S, Varney ML, Dave BJ, Singh RK (2003) IL-8 directly enhanced endothelialcell survival, proliferation, and matrix metalloproteinases production and regulatedangiogenesis. J Immunol 170:3369–3376.

5. Brooks PC, Clark RA, Cheresh DA (1994) Requirement of vascular integrin alpha v beta3 for angiogenesis. Science 264:569–571.

6. Bloch W, et al. (1997) Beta 1 integrin is essential for teratoma growth and angiogenesis.J Cell Biol 139:265–278.

7. Versteeg HH, Ruf W (2006) Emerging insights in tissue factor-dependent signalingevents. Semin Thromb Hemost 32:24–32.

8. Carmeliet P, et al. (1996) Role of tissue factor in embryonic blood vessel development.Nature 383:73–75.

9. Yu JL, et al. (2005) Oncogenic events regulate tissue factor expression in colorectalcancer cells: Implications for tumor progression and angiogenesis. Blood 105:1734–1741.

10. Versteeg HH, et al. (2008) Inhibition of tissue factor signaling suppresses tumor growth.Blood 111:190–199.

11. Versteeg HH, et al. (2008) Protease-activated receptor (PAR) 2, but not PAR1, signalingpromotes the development of mammary adenocarcinoma in polyoma middle T mice.Cancer Res 68:7219–7227.

12. Milsom CC, et al. (2008) Tissue factor regulation by epidermal growth factor receptorand epithelial-to-mesenchymal transitions: Effect on tumor initiation and angiogen-esis. Cancer Res 68:10068–10076.

13. Zhang JQ, et al. (2008) The FVIIa-tissue factor complex induces the expression of MMP7in LOVO cells in vitro. Int J Colorectal Dis 23:971–978.

14. Dorfleutner A, Hintermann E, Tarui T, Takada Y, Ruf W (2004) Cross-talk of integrinalpha3beta1 and tissue factor in cell migration. Mol Biol Cell 15:4416–4425.

15. Bogdanov VY, et al. (2003) Alternatively spliced human tissue factor: A circulating,soluble, thrombogenic protein. Nat Med 9:458–462.

16. Censarek P, Bobbe A, Grandoch M, Schror K, Weber AA (2007) Alternatively splicedhuman tissue factor (asHTF) is not pro-coagulant. Thromb Haemost 97:11–14.

17. Szotowski B, Antoniak S, Rauch U (2006) Alternatively spliced tissue factor: A previouslyunknown piece in the puzzle of hemostasis. Trends Cardiovasc Med 16:177–182.

18. Haas SL, et al. (2006) Expression of tissue factor in pancreatic adenocarcinoma isassociated with activation of coagulation. World J Gastroenterol 12:4843–4849.

19. Goldin-Lang P, et al. (2008) Tissue factor expression pattern in human non-small celllung cancer tissues indicate increased blood thrombogenicity and tumor metastasis.Oncol Rep 20:123–128.

20. Hobbs JE, et al. (2007) Alternatively spliced human tissue factor promotes tumorgrowth and angiogenesis in a pancreatic cancer tumor model. Thromb Res 120 Suppl2:S13–S21.

21. Masson V, et al. (2002) Mouse aortic ring assay: A new approach of the moleculargenetics of angiogenesis. Biol Proced Online 4:24–31.

22. Wright SD, Ramos RA, Tobias PS, Ulevitch RJ, Mathison JC (1990) CD14, a receptor forcomplexes of lipopolysaccharide (LPS) and LPS binding protein. Science 249:1431–1433.

23. Kuzuya M, et al. (1999) Induction of apoptotic cell death in vascular endothelial cellscultured in three-dimensional collagen lattice. Exp Cell Res 248:498–508.

24. Von Offenberg SN, et al. (2005) Cyclic strain-mediated regulation of vascular endo-thelial cell migration and tube formation. Biochem Biophys Res Commun 329:573–582.

25. Danen EH, Sonneveld P, Brakebusch C, Fassler R, Sonnenberg A (2002) The fibronectin-binding integrins alpha5beta1 and alphavbeta3 differentially modulate RhoA-GTPloading, organization of cell matrix adhesions, and fibronectin fibrillogenesis. J CellBiol 159:1071–1086.

26. Clark RA, Tonnesen MG, Gailit J, Cheresh DA (1996) Transient functional expression ofalphaVbeta 3 on vascular cells during wound repair. Am J Pathol 148:1407–1421.

27. Rauch U, et al. (2005) Association of tissue-factor upregulation in squamous-cellcarcinoma of the lung with increased tissue factor in circulating blood. Lancet Oncol6:254.

19502 � www.pnas.org�cgi�doi�10.1073�pnas.0905325106 van den Berg et al.