Canstatin Inhibits Akt Activation and Induces Fas-dependent Apoptosis in Endothelial Cells*

Canstatin, a 24-kDa peptide derived from the C-terminal globular non-collagenous (NC1) domain of the α2 chain of type IV collagen, was previously shown to induce apoptosis in cultured endothelial cells and to inhibit angiogenesis in vitro and in vivo. In this report, we demonstrate that canstatin inhibits the phosphorylation of Akt, focal adhesion kinase, mammalian target of rapamycin, eukaryotic initiation factor-4E-binding protein-1, and ribosomal S6 kinase in cultured human umbilical vein endothelial cells. It also induces Fas ligand expression, activates procaspases 8 and 9 cleavage, reduces mitochondrial membrane potential, and increases cell death (as determined by propidium iodide staining). Canstatin-induced activation of procaspases 8 and 9 as well as the induced reduction in mitochondrial membrane potential and cell viability were attenuated by the forced expression of FLICE-inhibitory protein. Canstatin-induced procaspase 8 activation and cell death were also inhibited by a neutralizing anti-Fas antibody. Collectively, these data indicate that canstatin-induced apoptosis is associated with phosphatidylinositol 3-kinase/Akt inhibition and is dependent upon signaling events transduced through membrane death receptors.

Type IV collagen, a complex vascular basement membrane protein consisting of six distinct gene products (␣1-␣6), is thought to play a crucial role in endothelial cell (EC) 1 adhesion, migration, and differentiation (1)(2)(3). The isolated C-terminal globular non-collagenous (NC1) domains of several type IV collagen ␣ chains, however, function as potent angiogenesis inhibitors. For example, canstatin, an endogenously produced fragment of the NC1 domain of the ␣2 chain (4), was recently shown to inhibit EC proliferation and tube formation in vitro and to suppress the growth of implanted PC-3 human prostate carcinoma and 786 -0 renal cell carcinoma cells in severe com-bined immunodeficiency and athymic nude mice, respectively (4). In these xenograft models, tumors excised from canstatintreated mice were less extensively infiltrated with CD31-staining cells than were control tumors. This reduced tumor vascularity and the fact that canstatin had no effect on tumor cell growth in vitro indicate that angiogenesis suppression is the dominant mechanism by which canstatin retards tumor growth.
The 26-kDa NC1 domain of the type IV collagen ␣1 chain (known as arresten) also inhibits endothelial tube formation and migration in vitro (5) and, like canstatin, blocks tumor growth in the prostate and renal cell carcinoma xenograft models cited above. The NC1 domain of the ␣3 chain (known as tumstatin) is similarly angiostatic (6) but, in addition, possesses intrinsic antitumor activity against melanoma cells, an effect attributable to a peptide sequence (amino acids 185-203) distinct from that which mediates its angiostatic properties (amino acids 54 -132) (7,8). Although this angiostatic peptide binds to both melanoma and endothelial cells, its antiproliferative effects are restricted to EC. As with canstatin and arresten, tumstatin also has potent antitumor activity in vivo. Thus, the NC1 domains of at least three distinct type IV collagen ␣ chains have documented angiostatic activity and, in this respect, seem to resemble the previously characterized angiogenesis inhibitor endostatin, a peptide derived from the NC1 domain of the ␣1 chain of type XVIII collagen (9 -11).
In addition to its suppressive effects on EC proliferation and tube formation, canstatin induces apoptosis in EC (4). We showed previously that this process was associated with the down-modulation of the anti-apoptotic protein FLIP (4). FLIP is a structural homologue of procaspases 8 and 10, but unlike these proteases, FLIP lacks a critical cysteine at what would otherwise be the active site of a functional protease (12)(13)(14). FLIP is recruited in competition with procaspase 8 to the plasma membrane through an interaction between its death effector domain and that of the adaptor protein FADD. This recruitment is triggered by the multimerization of FADD, an event usually initiated by an interaction between the death domain of FADD and one of the TNF receptor family members (e.g. Fas, TRAIL death receptors 4 and 5) induced by the binding of Fas ligand or TRAIL to its respective receptor (15)(16)(17)(18). Once complexed with FADD, FLIP interacts with TRAFs 1 and 2, receptor-interacting protein, and Raf to facilitate the activation of the NFB and mitogen-activated protein kinase pathways (19), both of which are associated with apoptosis resistance. It also interferes with autoprocessing of procaspase 8 in the multimolecular complex generated in response to FADD clustering (12)(13)(14)(15)(16)(17)(18). Therefore, one would expect that the down-modulation of FLIP induced by exposure to canstatin would sensitize endothelial cells to apoptosis mediated through procaspase 8 activation.
In our previous study, we provided no information on the mechanism by which canstatin down-modulates FLIP or data about whether its disappearance from endothelial cells contributed to the anti-angiogenic effects of canstatin. In this report, we demonstrate that canstatin inhibits the activation of Akt, a kinase previously shown to regulate FLIP levels in several cell types including EC (20,21). We also establish an unambiguous causal link between the loss of FLIP and the induction of apoptosis by canstatin.

EXPERIMENTAL PROCEDURES
Cell Lines and Reagents-Human umbilical vein endothelial cells (HUVEC) were purchased from American Type Culture Collection and maintained in McCoy's 5A media supplemented with 20% fetal bovine serum, 2 mM L-glutamine, 50 g/ml gentamycin, 5 g/ml polymyxin B (to inhibit the effects of residual lipopolysaccharide in the canstatin preparations), and 100 g/ml endothelial cell growth factor (ECGF, Biomedical Technologies, Stoughton, MA). The vinculin antibody used in our studies was purchased from Sigma. The pFAK-Tyr 397 antibody and the anti-Fas antibody used in Western blots were purchased from Calbiochem. The rabbit anti-FasL, pp70 s6k , caspase-8, and FLIP antibodies were obtained from Santa Cruz Biotechnology. The anti-Fas antibody (ZB4) used in the neutralization assays was obtained from Upstate Biotechnology (Lake Placid, NY). The pAkt (Ser 473 ) antibody was purchased from BD Biosciences, and the procaspase 9, phospho-mTOR, and phospho-4E-BP1 antibodies were obtained from Cell Signaling Technology (Beverly, MA). The canstatin cDNA was provided by Raghu Kalluri (4). Recombinant canstatin was produced and purified as reported previously (4).
Western Analyses-HUVEC were lysed in 62.5 mM Tris-HCl buffer, pH 6.8, containing 1% SDS, protease (phenylmethylsulfonyl fluoride and leupeptin), and phosphatase (NaF, Na 3 VO 4 , glycerophosphate, and Na 4 P 2 O 7 ) inhibitors. Lysates were separated on 12% SDS-PAGE gels, and the fractionated proteins were transferred to nitrocellulose. The blots were probed first with rabbit or murine antibodies specific for the protein of interest and then with either a goat anti-rabbit or anti-mouse antibody conjugated to horseradish peroxidase. The blots were then treated with SuperSignal chemiluminescent substrate (Pierce) and then exposed to Kodak X-Omat Blue XB-1 film. The films were analyzed by densitometry with a Bio-Rad densitometer.
Cytotoxicity and Mitochondrial Membrane Potential Assays-In these assays, adherent cells were detached by gentle trypsinization and combined with detached, floating cells. Propidium iodide (5 ng/ml), annexin V-FITC (5 l), or 5 g/ml BD Mitosensor reagent (BD Biosciences) were added to the cell pool, and the cells were then analyzed by flow cytometry with a BD Biosciences FACScan.
Infection of HUVEC with FLIP-Tet Adenovirus-Cells were co-infected with the rTA adenovirus and an adenovirus into which the cDNA of the long isoform human FLIP had been inserted (obtained from Ken Walsh, Boston University Medical Center) (21), each at a multiplicity of infection of 10. After overnight incubation, the medium was replaced. The FLIP gene is regulated by doxycycline and readily induced by exposing infected cells to an antibiotic concentration of 300 ng/ml.

Canstatin Inhibits the Phosphorylation of Akt and FAK and Induces Fas Ligand Expression in HUVEC-
We demonstrated previously that cultured bovine pulmonary arterial endothelial cells undergo apoptosis when exposed to canstatin (4). Although the biochemical mechanism underlying this effect was not explored in our earlier report, we did show that canstatin exposure was associated with the down-modulation of the antiapoptotic protein FLIP. We and others have since shown that FLIP expression is determined primarily by the activity of the phosphatidylinositol 3-kinase (PI3K)/Akt signaling pathway in a variety of cell types including tumor and endothelial cells (20,21). These observations suggested that canstatin-induced FLIP down-modulation might be caused by PI3K/Akt inhibition. To test this hypothesis, HUVEC were first incubated for 4 h in culture medium containing no exogenous growth factor and only 5% serum to reduce background signaling. Canstatin (20 g/ml) was then added to some of the cells, and 1 h later, the medium was replaced by fresh serum-supplemented (20%) medium containing ECGF (100 g/ml) with or without canstatin (20 g/ml). The cells were lysed at various time points, and the lysates were analyzed by Western blot for the phosphorylation of Akt and focal adhesion kinase (FAK), an integrin-associated kinase upstream of PI3K/Akt (22). As shown in Fig. 1, the addition of serum/ECGF resulted in a prompt increase in Akt phosphorylation. This effect was totally blocked by canstatin. The initial exposure to canstatin slightly increased FAK phosphorylation over background but completely blunted the subsequent response to ECGF. As reported previously, FLIP expression was suppressed in the canstatin-treated cells (4). Exposure to canstatin modestly increased Fas expression in HUVEC and markedly enhanced that of Fas ligand, especially at the later time points tested. These data confirm that canstatin does indeed suppress Akt phosphorylation in HUVEC as proposed, and they suggest that the disruption of this signaling pathway may be responsible for the disappearance of FLIP, the induction of Fas ligand, and other events triggered by canstatin.
Canstatin Inhibits the Phosphorylation of mTOR, 4E-BP1, and p70 s6k -Many of the effects of Akt on cell growth, protein synthesis, and cell cycle progression are mediated by the kinase mTOR and its effectors 4E-BP1 and p70 s6k (23,24). To determine whether canstatin-induced inhibition of Akt affected the activity of these downstream targets, HUVEC were exposed first to canstatin as described above, and 1 h later, ECGF was added to the culture. The cells were then lysed and analyzed by Western blot. As shown in Fig. 2A, even the 1 h of exposure to canstatin that preceded the addition of ECGF (the zero time point) was sufficient to completely inhibit mTOR phosphorylation. The phosphorylation of 4E-BP1 was also markedly attenuated with similar kinetics. Canstatin also completely blocked the phosphorylation of p70 s6k (Fig. 2B), although this effect was not evident until 24 h.
Role of FLIP, Fas, and Fas Ligand in Canstatin-induced Apoptosis-The observation that exposure to canstatin reduces FLIP levels and induces Fas ligand expression in HUVEC suggests that canstatin-induced apoptosis might be mediated through a death receptor (i.e. Fas)-dependent pathway (12)(13)(14). To test this hypothesis, we sought to determine whether exposure to canstatin induced the cleavage of procaspase 8 and if this effect was blocked by forced expression of FLIP. In this experiment, HUVEC were first co-infected with an rTA adenovirus and one containing a doxycycline-inducible FLIP construct. Doxycycline (300 ng/ml) was then added to some of the cell cultures to induce FLIP expression (21). Twenty-four hours later, the cells were placed into culture in ECGF-containing medium in the presence or absence of canstatin (20 g/ml), and after 48 h of incubation the cells were lysed and analyzed by Western blot for procaspase 8 activation. As shown in Fig. 3A, the antibiotic augmented FLIP expression in the infected HU-VEC. Exposure to canstatin resulted in procaspase 8 cleavage, suggesting that canstatin-induced apoptosis in HUVEC might indeed be mediated through Fas. Furthermore, procaspase 8 cleavage was evident in the absence of doxycycline but not in its presence, indicating that the process could be inhibited by FLIP.
To determine whether the forced expression of FLIP was able to maintain the viability of canstatin-treated HUVEC, the adenovirus-infected HUVEC were cultured with or without doxycycline (300 ng/ml) for 24 h and then placed in medium containing ECGF (100 g/ml) with or without canstatin (20 g/ml) for 48 h, as described above. Spontaneously detached cells were combined with those removed by trypsinization and analyzed by flow cytometry after staining with annexin V and propidium iodide (PI). As shown in Fig. 3B, exposure of the antibioticuntreated HUVEC to canstatin doubled the number of PIstaining cells, whereas no increase in the number of PI-staining cells was observed in the antibiotic-pretreated HUVEC.
As shown in Fig. 1, canstatin reduces FLIP expression and induces that of Fas ligand. To determine whether Fas-Fas ligand interactions might play a role in canstatin-induced activation of procaspase 8, HUVEC were placed in serum-supplemented medium containing canstatin with or without ECGF or a blocking anti-Fas antibody (ZB4, 1 g/ml) for 24 h. Procaspase 8 activation was then assessed by Western blot. As shown in Fig. 4A, canstatin induced procaspase 8 cleavage independently of the presence of ECGF, and this cleavage was substantially blocked by the anti-Fas antibody. This anti-Fas antibody also reduced the number of PI-staining cells in both canstatin-treated and untreated cultures (Fig. 4B), implicating membrane Fas-Fas ligand interactions in both the spontaneous (background) apoptosis observed in HUVEC cultures and the augmented apoptosis induced by canstatin.
Role of the Mitochondria in Canstatin-induced Apoptosis in HUVEC-To determine whether the mitochondria play a role in canstatin-induced apoptosis, HUVEC were first infected with the rTA and FLIP-tet adenoviruses as described above and cultured for 24 h in the presence or absence of doxycycline (300 ng/ml). The cells were then placed in medium containing ECGF (100 g/ml) with or without canstatin (20 g/ml) for 48 h. Spontaneously detached cells were combined with adher-ent cells removed by trypsinization, stained with the MitoSensor reagent, and analyzed by dual color flow cytometry for the presence of intramitochondrial dye aggregates (FL2-H) and intracytoplasmic monomers (FL1-H). Healthy cells with mitochondria that maintained a normal transmembrane potential (i.e. those with a high aggregate/monomer ratio) are depicted in the upper left quadrant of each panel in Fig. 5A, whereas those with lower ratios fall along a diagonal to the right. Cells with a high ratio appear more numerous in the lower two panels, which were generated with antibiotic-treated (FLIP-expressing) HUVEC than in the upper panels, which were generated with untreated HUVEC. In the absence of FLIP, canstatin exposure resulted in a thinning of this population, with an accumulation of the cells along a diagonal (cells with a lower aggregate/monomer ratio), whereas no such shift was observed in the doxycycline-treated cells. Exposure to canstatin reduced the mean aggregate/monomer ratio from 43 to 36 in HUVEC not treated with doxycycline but only from 43 to 42 in cells pretreated with the antibiotic. These data indicate that exposure of HUVEC to canstatin does reduce mitochondrial membrane potential and that this effect is blocked by forced expression of FLIP.
A reduction in mitochondrial membrane potential is associated with the release of several mitochondrial proteins including cytochrome c, second mitochondria-derived activator, and apoptosis-inducing factor, all of which participate in apoptosis (25)(26)(27). Cytochrome c released from the mitochondria binds to

FIG. 3. Canstatin induces procaspase 8 cleavage and cell death in HUVEC, and these effects are blocked by the forced expression of FLIP. A, pretreatment of FLIP-tet adenovirus-infected cells with doxycycline induces FLIP expression (top)
. No procaspase 8 cleavage was apparent in HUVEC not exposed to canstatin or in the canstatin-treated cells previously exposed to doxycycline. The p44 and p20 fragments of procaspase 8 were, however, readily detectable in the lysates of cells exposed to canstatin without doxycycline pretreatment. B, without prior doxycycline treatment (to induce FLIP expression), exposure to canstatin increased the number of PI/annexin V-staining HUVEC from 12.5 to 25.9%. Canstatin, however, had no effect on the viability of antibiotic-treated cells. the adaptor protein Apaf-1, which activates procaspase 9. To determine whether the reduction in mitochondrial membrane potential induced by canstatin results in procaspase 9 activation, adenovirus-infected HUVEC, some of which had been pretreated with doxycycline to induce FLIP expression, were exposed to canstatin, and procaspase 9 cleavage was then assessed by Western blot. As shown in Fig. 5B, exposure to canstatin results in increased procaspase 9 cleavage, an effect nearly completely inhibited by forced FLIP expression. These data show that the Fas-dependent apoptotic signaling events triggered by canstatin are amplified in the mitochondria.

DISCUSSION
Canstatin is a recently discovered, endogenously produced angiostatic peptide derived from the NC1 domain of the ␣2 chain of type IV collagen (4). In our initial description of this novel inhibitor, we demonstrated that recombinant canstatin inhibited endothelial cell proliferation, migration, and tube formation in vitro and suppressed the growth and vascularization of human tumor xenografts (4). We also showed that canstatin induced apoptosis in endothelial cells and reduced the levels of the anti-apoptotic protein FLIP but provided no additional information on the mechanism of FLIP down-modulation or its biological significance. In this report, we show that the disappearance of FLIP induced by canstatin is associated with reduced activity of the phosphatidylinositol 3-kinase/Akt signaling pathway. We also demonstrate that FLIP down-modulation plays an essential permissive role in canstatin-induced apoptosis.
The mechanism by which canstatin inhibits Akt activation in EC is unclear, but it may involve the disruption of signaling downstream of FAK, as has been proposed for endostatin (28) and tumstatin (29). As shown in Fig. 1, canstatin modestly increases basal FAK phosphorylation but completely blunts the inductive effect of ECGF. Similar results have been reported with endostatin, which increases basal FAK phosphorylation but inhibits the increase that would otherwise be induced by fibroblast growth factor (28). When recruited to the cytoplasmic domains of integrins clustered at focal adhesions, FAK directly interacts with the SH2 and SH3 domains of the p85 regulatory subunit of phosphatidylinositol 3-kinase and activates Akt (22). It is possible that canstatin interferes with some aspect of either integrin-dependent adhesion or downstream signaling. Tumstatin inhibits FAK activation by directly binding to the endothelial integrin ␣ v ␤ 3 (9,29,30). It is possible that canstatin functions similarly, although the binding of canstatin to an integrin has not yet been demonstrated.
In addition to its transcriptional regulation by Akt, FLIP is also subject to post-transcriptional regulation. FLIP is rapidly degraded in the proteasome in response to such diverse stimuli as p53 activation (31) or exposure to peroxisome proliferatoractivated receptor-␥ ligands (32). Because of its rapid turnover, protein synthesis inhibitors such as cycloheximide markedly reduce FLIP levels (33). The rapamycin analogue CCI-779, which inhibits protein synthesis by specifically blocking the Akt target mTOR (24), has similar effects. 2 5. Effects of canstatin on HUVEC mitochondrial membrane potential and procaspase 9 activity. A, HUVEC were infected with the rTA and FLIP-tet adenoviruses as described above. The infected cells were then incubated in medium with or without doxycycline (to induce FLIP expression), and subsequently, some were exposed to canstatin. Dye oligomers were detected in FL2-H and monomers in FL1-H. Data were reported as the ratio of the mean fluorescence intensity in FL2-H to that in FL1-H. Without prior doxycycline treatment, exposure to canstatin reduced the number of cells in the upper left region of the panel (high aggregate/monomer ratio) and increased the fraction of cells with data points along an apparent diagonal (lower ratio). The mean fluorescence intensity ratio decreased from 43 to 36 in response to canstatin. Cells pretreated with doxycycline, however, were essentially unaffected by canstatin (mean fluorescence intensity decreased from 43 to 42). B, canstatin activates procaspase 9. This effect, however, was markedly attenuated by the induction of FLIP.
shown to inhibit Akt and mTOR, resulting in a decline in overall protein synthesis (29). In fact, this phenomenon has been invoked as an explanation for its angiostatic effects. Our data demonstrating that canstatin completely inhibits mTOR and its downstream targets 4E-BP1 and p70 s6k suggest that canstatin may function in a similar manner. Therefore, it is likely that the nearly complete disappearance of FLIP from endothelial cells exposed to canstatin is caused by both reduced gene expression as a result of Akt inhibition and reduced FLIP protein synthesis due to mTOR inhibition.
The activation of procaspase 8 induced in HUVEC by canstatin is associated with changes in mitochondrial membrane potential and the activation of procaspase 9. This amplification is usually mediated through the cleavage of the proapoptotic Bcl-2 homology 3-domain-only Bcl-2 family member Bid, which, once cleaved, triggers the oligomerization of Bak or Bax and the release of mitochondrial proteins (34). Although we have been unable to detect Bid in HUVEC (data not shown), recent studies have indicated that this protein is not absolutely essential for apoptosis in response to procaspase 8 activation. Kandasamy et al. (35) recently showed that TRAIL induces apoptosis and the release from the mitochondria of cytochrome c (but not second mitochondria-derived activator) in Bid Ϫ/Ϫ murine embryonic fibroblasts as long as either Bak or Bax were present. Thus, canstatin seems to be able to reduce mitochondrial membrane potential and activate procaspase 9 in HUVEC despite limited Bid expression.
The mechanism by which canstatin activates procaspase 8 and initiates an apoptotic signaling cascade is unclear. Stupack et al. (36) demonstrated that the proenzyme is directly recruited to the cytoplasmic domains of unoccupied integrins and activated there in a manner similar to that induced by the multimerization of FADD. It is conceivable that canstatin functions by uncoupling endothelial integrins from the underlying matrix, as has been proposed for tumstatin and endostatin (9), and that procaspase 8 is subsequently recruited to these unoccupied integrins and activated. Such a model, however, would not explain the ability of FLIP to block canstatin-induced procaspase 8 activation without invoking the prospect that FLIP might be similarly recruited to integrins. This model would also fail to explain the Fas dependence of canstatin-or disadhesioninduced apoptosis (i.e. anoikis) (37) unless one proposes that the initial procaspase 8 activation triggered by unoccupied integrins is later amplified by membrane Fas-Fas ligand interactions. Although our data do not exclude a contribution of integrin-mediated procaspase 8 activation, they indicate that Akt inhibition with the induction of Fas ligand, the downmodulation of FLIP, and the resulting interactions between Fas ligand and Fas at the cell membrane are the primary determinants of canstatin-induced apoptosis.