The PTEN/Akt Pathway Dictates the Direct α_Vβ₃-Dependent Growth-Inhibitory Action of an Active Fragment of Tumstatin in Glioma Cells In vitro and In vivo

Tumors, such as glioblastoma multiforme, the most aggressive form of human glioma, are characterized by marked endothelial cell proliferation and extensive angiogenesis, which are in turn believed to be driven by the high metabolic demands of the tumor, hypoxia, and the release of factors that can lead to uncontrolled neovascularization (1, 2). Because tumors, such as glioblastoma multiforme, depend on vessel formation to supply oxygen necessary for growth, a significant effort has been made to better understand the angiogenic process in these tumors (3, 4). Results from such studies suggest that the angiogenic process is controlled by the balance between proangiogenic molecules that promote vessel sprouting (5–8) and negative regulators of the angiogenic process.

Among the most interesting and least understood of the antiangiogenic molecules are the endogenous inhibitors of angiogenesis. These proteins, including canstatin, endostatin, and tumstatin, are bioactive fragments of larger proteins commonly found in the vascular basement membrane (VBM; refs. 9–14). As an example, tumstatin is an NC1 domain fragment of the α3 chain of type IV collagen, which is located in a variety of basement membranes including those in the brain (13). Release of endogenous inhibitors of angiogenesis from the VBM occurs by the same matrix metalloprotease–mediated mechanism that degrades the VBM and drives the early stages of angiogenesis. Endogenous inhibitors of angiogenesis, however, rather than stimulating angiogenesis, bind to integrin complexes and initiate a cascade of events that suppress the angiogenic process. In the case of tumstatin, the peptide can bind α_Vβ₃ through amino acids 54 to 132 and 185 to 203 (13, 15). Although tumstatin does not alter tumor cell growth directly, a NH₂-terminal truncated protein or subfragments containing amino acids 185 to 203 can block the proliferation of melanoma cells by suppression of the focal adhesion kinase (FAK)/phophoinositol-3 kinase (PI3K) pathway (15). Intact tumstatin, however, has been shown to have only antiangiogenic activity, and this activity has been localized to fragments containing amino acids 54 to 132 (also known as Tum-5), which bind α_Vβ₃ on a variety of cells but only inhibit the PI3K/Akt/mammalian target of rapamycin (mTOR) pathway and trigger subsequent apoptosis in endothelial cells (13). This latter activity has been localized to amino acids 69 to 98, and a peptide fragment containing these amino acids (also known as T3) is considered a promising α_Vβ₃-dependent inhibitor of angiogenesis (16).

Although the ability of the T3 peptide to bind α_Vβ₃, block the Akt/mTOR pathway, and trigger apoptosis in endothelial cells has been well documented, the basis for its endothelial-specific action remains unclear, particularly in light of the fact that many tumor cells, including those at the invading edges of glioblastoma multiforme, express α_Vβ₃ and rely on Akt activation for growth (17, 18). Although a variety of α_Vβ₃-expressing tumor cells have been shown to be insensitive to T3-mediated suppression of cell growth and /or apoptosis, it is worth noting that most of these tumors have also been reported to be PTEN deficient and/or to exhibit constitutive activation of the Akt/mTOR pathway. If these cells have alterations that activate the Akt pathway downstream of the point at which T3 blocks Akt activation, these cells might seem insensitive to T3. Conversely, tumors that do not have PTEN alterations or high levels of phospho-Akt (pAkt; a group that includes the majority of glioblastoma multiforme) might, along with endothelial cells, be directly targeted by T3. In such tumors, T3 might have a broader range of action than previously reported.

To address this possibility, we collected a variety of transformed glial and glioblastoma multiforme cells, genetically modulated the PTEN/Akt/mTOR pathway, and examined the effects of T3 in vitro and in vivo in response to these modulations. The results of the studies show that at concentrations that suppress the growth of endothelial cells, T3 also suppresses the growth of glioblastoma multiforme cells, but only if the cells express α_Vβ₃ and have a functional PTEN/low levels of pAkt. These results suggest that in a subset of glioblastoma multiforme and potentially genetically similar tumors, T3 might have a dual mechanism of action, both directly and indirectly suppressing tumor growth.

Reagents. T3 peptide (amino acids 69-88 of human tumstatin) was obtained from Phoenix Pharmaceuticals, Inc. (Belmont, CA). For distribution studies, mono-5-(and-6)-carboxyfluorescein (FAM)-labeled T3 was used. The maximum excitation/emission of FAM is 495/519 nm, respectively.

Cell lines. Genetically modified human astrocytes containing constructs encoding HPV16 E6, HPV16 E7, hTERT, V12 H-Ras, and either a blank vector or constructs encoding human integrin β₃ [wild-type (WT) or constitutively active (CA) D723R; Ras + blank vector, Ras + β3WT, and Ras + β3CA cells, respectively] were created as described (19, 20). Cells were further infected with a blank vector, with pWZL-hygro encoding CA myristilated Akt (myrAkt Δ4-129), or with pWZL-hygro encoding eukaryotic translation initiation factor 4E (eIF4E). All established human glioblastoma multiforme cell lines were obtained from the University of California San Francisco (UCSF) Brain Tumor Research Center Tissue Bank, and select lines were also infected with blank vector, pMXI-egfp-human integrin β₃ WT, and/or pWZL-hygro-myrAktΔ4-129. An immortalized human dermal microvascular endothelial cell line (HDMEC) was created from cryopreserved cells isolated from human dermal tissue (PromoCell, Heidelberg, Germany) and immortalized with SV40 Tag (21) and was provided by Dr. Gabriele Bergers (UCSF Department of Neurological Surgery). These cells were further infected with blank vector or pMXI-egfp-human integrin β3WT. Mouse astrocytes derived from β3WT or β₃ knockout (β3KO) animals were transformed by retroviral infection with V12 H-Ras and HPV16 E6/HPV16 E7 as described (2, 19, 20, 22). The cells were further infected with blank vector, pWZL-hygro-myrAktΔ4-129, or pMXI-egfp-human integrin β3WT, after which expression of β3WT or myrAkt was verified by PCR and Western blot. Mouse astrocytes derived from WT or PTEN conditional KO mice [P2 animals with genotypes Pten^+/+;GFAP-cre (WT) and PtenloxP/loxP;GFAP-cre (Pten cKO; ref. 23)] and transfected with SV40 large T antigen and mutant V12 H-Ras were also provided by Dr. Gabriele Bergers.

For enhanced green fluorescent protein (egfp) selection, retrovirally infected egfp-positive cells (>10,000 per population) were sorted 5 days after infection on a FACSVantage (Becton Dickinson, San Jose, CA) and pooled. For Hygromycin B selection, cells were grown for 5 days in Hygromycin B (300 μg/mL) beginning 72 hours after infection.

Cells were maintained as monolayers in a complete medium consisting of DMEM supplemented with 10% fetal bovine serum (FBS). HDMEC cells were cultured in MCDB131 (Life Technologies, Carlsbad, CA) with 10% FBS and 1:100 l-Glutamine. All cells were cultured at 37°C in a humidified 5% CO₂ atmosphere.

Small interfering RNA studies. Cells were seeded at 2.0 × 10⁵ per six-well plate 24 hours before transfection and transfected with 50 or 100 nmol/L small interfering RNA (siRNA) targeting PTEN (5′-AACAGTAGAGGAGCCGTCAAA-3′; B-Bridge International, Inc., Sunnyvale, CA) or control non-targeting siRNA (Dharmacon, Lafayette, CO), using Oligofectamine (Invitrogen, Carlsbad, CA). After 4 hours of transfection, the media were replaced with complete growing media containing 10% FBS. Cells were harvested 48 or 96 hours after transfection and analyzed by immunoblotting or plated for further assays.

Cell viability/growth assays. For cell viability/growth assays, pre-plated cells (1,000 cells per 96-well plate) were serum starved for 24 hours, after which medium was changed to normal growing medium containing 10% FBS with or without T3 peptide (0-50 μmol/L). Seventy-two hours after incubation, 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt (MTS) reagent (Promega, Madison, WI) was added and used to analyze the ability of the cells to chemically reduce the MTS reagent to formazan, which was subsequently detected spectrophotometrically (490 nm) according to manufacturer's protocol. Absorbance of control cultures (which was not significantly influenced by 24 hours of growth in serum-free conditions) was defined as 1.0 and compared with that of T3-treated cultures.

Western blot analysis. Cells were serum starved for 24 hours, after which medium was changed to normal growing medium containing 10% FBS with or without T3 peptide (10 μmol/L). After 4 hours of incubation, cells were harvested and lysed as described previously (24). Whole-cell lysate (10 μg) was subjected to gel electrophoresis and electroblotted onto an Immobilon-P membrane (Millipore, Bedford, MA). The membrane was blocked in 5% skim milk and incubated with antibodies against pAkt, total Akt, PTEN, integrin β₃, phospho-S6 kinase (pS6K), total S6K, eIF4E-binding protein 1 (4E-BP1), eIF4E (1:1,000; all from Cell Signaling Technology, Danvers, MA), or α-tubulin (1:2,000; Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C. Bound antibody was detected with horseradish peroxidase–conjugated secondary antibody using Enhanced Chemiluminescence Western blotting detection reagents (Pierce, Milwaukee, WI).

Bromodeoxyuridine cell cycle analysis. Cells were serum starved for 24 hours, after which medium was changed to normal growing medium that contained 10% FBS with or without 10 μmol/L T3 peptide. After 24 hours of incubation, bromodeoxyuridine (BrdUrd; 10 μmol/L) was added to medium, and the cells were incubated for 1 hour at 37°C. The cells were then trypsinized, fixed, and stained according to the manufacturer's instruction (BD Biosciences PharMingen, San Diego, CA).

Analysis of cell cycle distribution/Annexin V-FITC assay. Cells were trypsinized after T3 incubation, washed in PBS, and fixed in 70% ethanol at −20°C. Following incubation in PBS containing 40 μg/mL propidium iodine and 200 μg/mL RNase A (Sigma, St. Louis, MO) for 1 hour at room temperature in the dark, stained nuclei were analyzed on a FACScan machine (Becton Dickinson) with 10,000 events per determination as described previously (25). ModFit LT software (Verity Software House, Inc., Topsham, ME) was used to assess cell cycle distribution. The Annexin V-FITC binding assay was done using an ApoAlert Annexin V kit (Clontech, Palo Alto, CA) as described previously (26), with 20,000 events per determination.

Analysis of cell surface expression of integrin α_Vβ₃. To confirm integrin α_Vβ₃ expression on the cell surface, cells were trypsinized, washed with PBS, and incubated with anti-human integrin α_Vβ₃ antibody (1:200 in PBS with 1% bovine serum albumin; Chemicon, Temecula, CA) for 1 hour at room temperature. After washing with PBS, cells were incubated with phycoerythrin-conjugated anti-mouse IgG secondary antibody (1:200; Vector Lab, Burlingame, CA) for 1 hour at room temperature in the dark. Stained cells were then analyzed on FACScan machine (Becton Dickinson).

Mice genotyping. To confirm the integrin β₃ genotype, three-primer PCR was carried out as described previously (27) using the following conditions: 35 cycles of 94°C (1 minute)/60°C (1 minute)/72°C (1 minute) followed by 72°C (10 minutes). The primer sets were 5′-CTTAGACACCTGCTACGGGC-3′ (primer 1; common forward primer), 5′-CTGAGGCTGAGTGTGATGG-3′ (primer 2; WT specific), and 5′-CACGAGACTAGTGAGACGTG-3′ (primer 3; mutant specific). PCR products were electrophoresed in 2% agarose gel and observed under UV light. PCR products were 538 bp (mutant) and 446 bp (WT).

Mice and tumorigenicity assay and T3 convection-enhanced delivery. β₃ Integrin KO (22) mice were anesthetized with ketamine, and tumor cells were i.c. injected as previously described (19, 20). Briefly, 3 × 10⁶ cells were resuspended in 2.5 μL of PBS and stereotactically injected (>5 minutes) into the striatum of age- and sex-matched mice (20, 28), after which the needle was left in place for another 5 minutes to stabilize the injected cells. Fourteen and 18 days after implantation, tumors were exposed to T3 peptide using convection-enhanced delivery (CED) as described previously (29–31). The infusion cannula was mounted onto a stereotactic device and inserted into the brain using the same coordinate used for tumor cell implantation. Ascending infusion rates (0.2 μL/min × 15 minutes, 0.5 μL/min × 10 minutes, and 0.8 μL/min × 15 min) were applied to achieve the 20 μL (40 μmol/L) infusion. Mice were euthanized 21 days after tumor implantation (7 days after initial CED), and tumor-bearing brains were sectioned coronally at the point of cellular implantation. After H&E staining, sections were photographed, and the length (a) and width (b) of the largest tumor cross-sectional areas were determined. To obtain another variable of width (c), samples were cut into 10-μm serial coronal sections, and every 10th section was stained for H&E. Tumor volume was calculated using the standard formula: V = length (a) × width (b) × width (c) × 0.52.

Distribution of T3 after CED/evaluation of toxicity. Two control mice and two tumor-bearing mice were used to evaluate the distribution of T3 peptide (FAM labeled) after CED infusion into their striatum or brain tumor, respectively. Fourteen days after tumor implantation, T3 peptide was infused locally by CED as described above. Four hours after infusion, mice were euthanized, and their brains were harvested. After embedding in ornithine carbamyl transferase compound, samples were cut into 12-mm serial sections and stained with H&E and observed under the fluorescent microscope. To evaluate the local and systemic toxicity, T3 peptide was infused twice into intact mice brain by CED followed the same protocol as for tumor-implanted mice. Seven days after initial CED, mice were euthanized, and harvested brains were histologically analyzed.

Immunohistochemistry. Five-micrometer paraformaldehyde-fixed, paraffin-embedded sections or 10-μm frozen sections were used for immunohistochemical analysis as described previously (20). Primary antibodies used for immunohistochemistry were rabbit anti-von Willebrand factor (vWF; 1:800; Abcam, Cambridge, MA), rabbit anti-Ki67 (1:200; Abcam), and rat anti-mouse CD31 (1:50; BD PharMingen, San Jose, CA). These primary antibodies were incubated overnight at 4°C. After incubation with a biotinylated secondary antibody (goat anti-rabbit, 1:200; Santa Cruz Biotechnology) for 60 minutes at room temperature, antigens were revealed with streptavidin-conjugated horseradish peroxidase and diaminobenzamide. Nuclei were counterstained with methyl green. Vessel (pixels) area was measured using Openlab software (Improvison, Lexington, MA). CD31-positive cell vessels were selected with the wand tool using a threshold of 50 pixels, and the area of these vessels was recorded. Vessel density was assessed by counting the number of vWF-positive vessels in 20 independent tumor fields at ×400 with the aid of an ocular grid (20). Vessel area and vessel density were reported as the mean ± SD of results from six animals per group.

Statistical analysis. All statistical analyses were done using the Student's t test, with significance defined as P < 0.05.

T3 induces apoptosis in endothelial cells but not glioma cells. The active T3 peptide fragment of tumstatin has been previously shown to suppress cell growth and induce apoptosis in a variety of α_Vβ₃-expressing endothelial cells but not in tumor cells. To verify these results, we first examined α_Vβ₃ expression and T3-induced effects on cell growth and apoptosis in HDMEC and in α_Vβ₃-expressing glioblastoma multiforme cell lines. Incubation of HDMEC (which under our culture conditions express very low levels of total cellular and cell surface α_Vβ₃; Fig. 1A and D, respectively) with T3 peptide (0-10 μmol/L, 72 hours) neither altered the number of viable cells (MTS assay; Fig. 1B,, right) relative to control cultures nor induced apoptosis (Annexin V/propidium iodide staining; Fig. 1C). Introduction of a construct encoding integrin β₃ (the limiting component of α_Vβ₃ assembly in these cells), however, significantly increased both β₃ expression (Fig. 1A and D) and the ability of T3 to decrease viable cell number (Fig. 1B,, right) and induce apoptosis (Fig. 1C). In contrast, the initial set of β₃-expressing glioma cell lines examined (Fig. 1A and D) were immune to T3-induced effects on cell growth (Fig. 1B) and apoptosis (data not shown). These results are consistent with previously reported data suggesting that T3 functions as an inducer of apoptosis in α_Vβ₃-expressing endothelial cells but not tumor cells.

T3 has direct integrin β₃– and Akt-dependent antiproliferative effects on transformed human astrocytes. Although T3 has been shown to bind integrin α_Vβ₃; to suppress FAK, PI3K, Akt, and mTOR activation; and to induce apoptosis in an endothelial cell–specific manner, the majority of tumor cells described as T3 insensitive (including those in Fig. 1A) are unlike normal endothelial cells in that they contain PTEN mutations and/or have high endogenous levels of pAkt (Fig. 1A). To examine the possibility that T3 can suppress cell viability in tumors in which the PTEN/Akt pathway is not dysregulated, we first examined T3 sensitivity in human astrocytes transformed by introduction of hTERT, E6/E7, and mutant V12 H-Ras. These PTEN-proficient transformed cells exhibited low levels of pAkt under normal growth conditions (Fig. 2A,, left), did not express integrin β₃ (Fig. 2A,, left), and as expected by their β₃ status did not exhibit growth suppression (MTS assay; Fig. 2A,, right), increased apoptosis (lack of cells with sub-G₁ DNA content by fluorescence-activated cell sorting analysis), or changes in the percentage of cells incorporating BrdUrd in response to exposure to T3 (Fig. 2B). Introduction of a WT or CA β₃-encoding construct (D723R) increased α_Vβ₃ levels in these transformed glial cells relative to blank vector controls without altering pAkt levels (Fig. 2A,, left and middle). Although the cells expressing the CA β₃ (Ras + β3CA; whose activation cannot be blocked by T3) were insensitive to T3-induced effects on cell growth, apoptosis, and entry into S phase (Fig. 2A,, right and Fig. 2B); the number of viable Ras + β3WT cells; and the percentage of Ras + β3WT cells in S phase were both significantly reduced following T3 exposure (Fig. 2B). To more clearly define how T3 suppressed the growth of β3WT-expressing transformed human astrocytes, the pathway shown to link β₃ to pAkt and mTOR in endothelial cells was examined in these transformed astrocytes. Serum-starved T3-insensitive transformed astrocytes (Ras + blank vector cells or Ras + β3CA cells) exhibited an increase in levels of pAkt and pS6K (an indicator of mTOR activity; refs. 32–35) in response to serum stimulation (Fig. 2C). These effects were not altered by incubation of the cells with T3. Serum-starved T3-sensitive Ras cells expressing β3WT also exhibited an increase in pAkt and pS6K levels in response to serum stimulation, although in these cells, exposure to T3 significantly suppressed both pAkt and pS6K levels and also led to hypophosphorylation (activation) of the translation inhibitor 4EBP1 (Fig. 2C,, bottom arrow). These results suggest that the growth-suppressive actions of T3 in transformed astrocytes are, as in endothelial cells, linked to suppression of the Akt/mTOR pathway. To more fully address this possibility, T3-sensitive β₃-expressing transformed astrocytes were retrovirally infected with blank vector constructs or constructs encoding either a constitutively activated Akt (myrAktΔ4-129), or eIF4E, the downstream target of mTOR involved in stimulation of protein synthesis, after which effects of pathway alterations on T3-induced decreases in viable cell number were monitored. As shown in Fig. 2D, the ability of T3 to suppress the growth of β₃-expressing Ras cells (relative to control cultures) could be overcome by overexpression of either myrAkt or the translational activator and 4EBP1 target eIF4E (Fig. 2D). These results suggest that T3 can suppress the viability of transformed glial cells as well endothelial cells in a non-apoptotic, pAkt-dependent manner.

T3 has direct integrin β₃– and Akt-dependent effects on viable human glioblastoma multiforme cell number. Because studies in transformed human astrocytes suggested that low pAkt levels were a prerequisite for T3 sensitivity, we further expanded our analysis to include three glioblastoma multiforme cells lines which, as part of a subset of glioblastoma multiforme subset with an intact PTEN gene, have low endogenous levels of pAkt (Fig. 2A). Although none of these cell lines expressed β₃ under the culture conditions used (Fig. 3A), introduction of β3WT into each of these cells increased α_Vβ₃ expression (relative to vector controls) without altering pAkt levels and significantly increased the ability of T3 to suppress viable cell number (Fig. 3A). In each case, the actions of T3 could be reversed by increasing pAkt levels, either by introduction of a myrAkt (Fig. 3A) or by exposure to a PTEN-targeted siRNA (Fig. 3B). These results show that the β₃- and Akt-dependent effects of T3 on viable cell number are not limited to human astrocytes transformed in vitro but can be readily seen in human glioblastoma multiforme cells as well.

T3 has direct integrin β₃– and Akt-dependent antiproliferative effects in genetically defined transformed mouse astrocytes. Because our studies indicate that the growth-suppressive effects of T3 are β₃ and PTEN dependent, we took advantage of mice genetically engineered to be deficient in β₃ or PTEN expression, as well as our ability to transform astrocytes in vitro, to examine the effects of β₃ and PTEN expression on T3 sensitivity in a defined genetic model. For these studies, mouse astrocytes from WT, β3KO, or PTEN KO animals were isolated and transformed by expression of mutant V12 H-Ras in combination with either E6/E7 or large T antigen. The resultant tumorigenic cells were subsequently examined for sensitivity to T3 in vitro. As shown in Fig. 4A, T3 exposure decreased the growth of transformed astrocytes derived from β3WT animals as well as the percentage of these cells incorporating BrdUrd (Fig. 4B), whereas the effects of T3 exposure in similarly transformed astrocytes from β3KO animals were negligible. As with transformed human cells, T3-mediated inhibition of viable cell number of transformed mouse astrocytes was associated not with induction of apoptosis but with suppression of pAkt and pS6K levels, and with hypophosphorylation/activation of 4EBP1 (Fig. 4B and C), and could be reversed by introduction of a CA myristilated Akt -encoding construct (Fig. 4D). Conversely, the T3-insensitive transformed mouse astrocytes derived from β3KO animals could be sensitized by introduction of a construct encoding β3WT (Fig. 4D; Supplementary Fig. S1). Similarly, the number of viable transformed astrocytes from PTEN WT mice (which had minimal levels of pAkt; Fig. 5A) was suppressed by exposure to T3 (Fig. 5B), whereas that of similarly transformed astrocytes from PTEN KO embryos (which had high levels of pAkt) was not (Fig. 5A and B). In combination with the previously described work in human transformed cells and glioblastoma multiforme cells, these results show that T3 has direct growth suppressive actions on transformed glial cells, but that these effects are highly dependent on the status of the PTEN/Akt/mTOR pathway.

T3 directly effects glioblastoma multiforme growth in vivo. To assess the function of T3 in vivo, we generated a model in which T3 actions on the tumor could clearly be separated from those on the tumor vasculature. In this system, transformed mouse astrocytes derived from β3WT or β3KO mice were i.c. implanted into β3KO mice. After allowing the cells to establish a tumor, T3 peptide or vehicle was given twice (14 and 18 days after tumor implantation) to the tumor by CED. Tumor volume, vessel area, and vessel density were then monitored 7 days after the initial drug treatment. Because the host animal provides the vasculature necessary for tumor growth, and because the host endothelial cells and vasculature in this model are deficient in β₃ expression and therefore insensitive to the effects of T3, the model allows any effects of T3 administration on tumor growth to be ascribed to direct effects of T3 on the tumor.

To first show that T3 peptide had no adverse effects on normal tissue, the T3 peptide solution to be used in tumor studies was infused twice by CED into the brains of two normal mice. The animals were monitored for general health and body weight for 7 days after the initial infusion, after which the animals were sacrificed and harvested brains were examined histologically. As shown in Supplementary Fig. S2, no toxicity was noted around the area of T3 infusion. To show that T3 peptide could be successfully given to i.c. tumors by CED, a fluorescently labeled T3 preparation (FAM-T3) was delivered by CED to two tumors 14 days following implantation of cells. The animals were sacrificed 4 hours after infusion, after which the brains were harvested, fixed, sectioned, stained with H&E, and examined under a light or fluorescent microscope. As shown in the serial sections in Fig. 6A, CED of T3 resulted in uniform distribution of the labeled peptide throughout the tumor. Having shown that the T3 peptide could be delivered safely and effectively by CED, we tested the effect of CED-delivered T3 on the i.c. growth of β3WT or β3KO transformed mouse astrocytes in β3KO mice. As shown in Fig. 6B, both transformed β3WT and β3KO mouse astrocytes formed tumors upon i.c. injection (although the tumors derived from β3WT cells were slightly larger than those formed by β3KO cells; data not shown). Although CED infusion of T3 peptide did not alter the volume of β3KO tumors formed relative to vehicle control, it significantly reduced the volume of β3WT tumors formed (Fig. 6B). The effects of T3 peptide on tumor volume were not a result of alteration in vessel area (Fig. 6C; Supplementary Fig. S3) or vessel density (data not shown), which was comparable in β3WT and β3KO tumors infused with either vehicle or T3. These results are consistent with the observation that β₃-deficient mice form normal tumor vasculatures, but that lack of β₃ expression renders cells of the tumor vasculature insensitive to the proapoptotic effects of T3. The Ki67 labeling index, a measure of proliferation, of β3WT tumor cells exposed in vivo to T3, however, was significantly less than that of the same tumor cells exposed to vehicle, or of the Ki67 labeling index of β3KO tumor cells exposed in vivo to either T3 or vehicle (Fig. 6D; Supplementary Fig. S3), whereas terminal deoxynucleotidyl transferase–mediated nick-end labeling analysis showed no significant increase in apoptotic cells in T3-treated sections (data not shown). These results show that T3 has direct, β₃- and PTEN/Akt–dependent growth-suppressing actions on tumor cells in vivo as well as in vitro.

Angiogenesis is a prerequisite for tumor growth and plays an especially important role in rapidly growing tumors, such as glioblastoma multiforme (3, 36, 37). A variety of factors controlling the angiogenic balance have been described, and among these, the endogenous inhibitor of angiogenesis tumstatin and one of its active subfragments T3 have drawn considerable attention. Although tumstatin and specifically T3 have been shown to block the proliferation of endothelial cells, studies of the direct effects of T3 on tumor growth have been limited to a relatively small number of tumor cell lines and transformed fibroblasts (38). Because the pathway inhibited by T3 in endothelial cells is also critical, and in many cases constitutively activated, in tumors, we considered the possibility that tumor-directed actions of T3 might be only be apparent in tumors in which the Akt pathway activation was not present. The results of the present study show that this is the case, and that T3 can suppress the growth of a variety of cell lines providing that the cells express α_Vβ₃ and have not constitutively activated the Akt/mTOR pathway. These results show that T3 and perhaps tumstatin itself have direct as well as indirect inhibitory actions on tumor growth.

Several previous studies have investigated the role of tumstatin and T3 in tumor growth suppression. T3, as well as intact tumstatin, have been shown to have direct proapoptotic effects on endothelial cells. These effects, coupled with a lack of direct effects on tumor cell growth, led to the conclusion that antiangiogenic peptides indirectly suppress tumor growth (16). Our in vitro and in vivo studies clearly showed that in addition to having indirect tumor-suppressive actions, T3 also has direct growth-suppressive activity in tumor cells with an appropriate genotype. The action of T3 in the tumor seemed to be cytostatic and not tumoricidal, and in in vitro studies, removal of T3 from the media allowed resumed cell growth (data not shown). Because glioma cells are highly resistant to apoptosis, particularly when compared with endothelial cells or lymphoid cells, the direct antitumor effects of T3 might be less dramatic in gliomas than other tumor types. The direct antitumor effects of T3 have not been previously noted perhaps because most studies have focused on PC-3 prostate adenocarcinoma and WM-164 melanoma cells (13, 16, 18, 38, 39), both of which have mutant PTEN (40–42) and high levels of pAkt relative to endothelial cells (38). It seems likely, therefore, that PC-3 and WM-164 cells, along with the PTEN-deficient cells examined in this study, are insensitive to the growth-suppressive actions of T3 not because of their tumorigenic status, but because of PTEN loss and subsequent Akt activation. The reversal of T3 sensitivity upon suppression or loss of PTEN or introduction of Akt/eIF4E as noted in the present study further supports this idea. The concept that Akt activation can suppress T3-mediated cell death is also consistent with suggestions that PTEN loss/Akt activation may be a means by which tumor cells escape integrin-mediated death triggered by unligated (or in this case, antagonized) α_Vβ₃ functioning as a “dependence receptor” (43). Previous studies, however, have also suggested that T3-induced suppression of phospho-FAK (pFAK) levels, which is linked to decreased Akt and mTOR activation, does not occur in T3-insensitive tumor cells, and that as such, the insensitivity of these tumors to T3-induced growth suppression is a consequence of alterations upstream, rather than downstream, of FAK. Introduction of functional PTEN, however, has been shown to suppress levels of pFAK, and in hepatocarcinoma tissues and cell lines, an inverse association exists between PTEN function and pFAK levels (44). The reported inability of T3 to suppress pFAK levels in T3-resistant WM-164 cells may, therefore, simply be a result of enforced FAK activation resulting from loss of PTEN function in these cells.

The present work not only adds to our understanding of tumstatin function but may also have clinical implications. The present work defines PTEN-proficient tumors and/or tumors with low endogenous levels of pAkt as a subset sensitive to the direct growth-suppressive actions of T3. Clinical trials with tumstatin, therefore, might benefit from stratification of patients based on PTEN status. This type of stratification could easily be done in glioblastoma multiforme, only 30% of which exhibit PTEN alterations (45–47). Alternatively, the present work suggests that inhibition of the Akt/mTOR axis by agents, such as epidermal growth factor receptor inhibitors or rapamycin, might further sensitize tumors to the direct antitumor actions of T3. We have also shown in the present study that T3 can be safely and effectively delivered in vivo across the entire i.c. tumor volume by CED at higher concentrations than could likely be achieved systemically, thereby bypassing possible concerns about blood-brain barrier impermeability (30, 31). As the CED technique becomes more refined and applied in clinical settings, prolonged localized delivery of T3 might allow long-term control of glioma growth and progression, whereas avoiding the autoimmune responses possible following long-term systemic administration (13, 48). The identification of a dual mechanism of action of T3 in the present study, therefore, not only provides insight into the function of endogenous inhibitors of angiogenesis but also may help influence how patients are selected for such therapy and ultimately may suggest alternative ways to use such agents clinically.

Grant support: NIH grant CA94989 (R.O. Pieper), NIH (S.J. Baker), and American Lebanese and Syrian Associated Charities (S.J. Baker).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.