SD-208, a Novel Transforming Growth Factor β Receptor I Kinase Inhibitor, Inhibits Growth and Invasiveness and Enhances Immunogenicity of Murine and Human Glioma Cells In vitro and In vivo

Human glioblastoma patients experience a median survival of little more than 1 year with the standard treatment of surgery, radiotherapy, and nitrosourea-based chemotherapy (1). For many years, immunotherapy has been explored as an alternative approach for these tumors because human glioma patients exhibit specific deficits in their cellular immune response ex vivo (2) and because human glioma cells are paradigmatic for the property of cancer cells to express immunosuppressive molecules. These include soluble factors such as transforming growth factor (TGF)-β (3), prostaglandins (4), or interleukin (IL)-10 (5), as well as cell surface molecules such as CD70 (6) or HLA-G (7). Among these, TGF-β has attracted the most interest (8), resulting in experimental therapeutic approaches using antisense strategies (9), gene transfer of TGF-β antagonists such as decorin (10), inhibition of TGF-β–processing proteases of the furin family (11), or drugs such as tranilast (12). The undesirable effects of TGF-β in malignant glioma are not restricted to the induction of immunosuppression in the host but include a critical role of TGF-β in migration and invasion (13). Here we examine a novel therapeutic principle of TGF-β antagonism in malignant glioma, defined by SD-208, a pharmacological agent that blocks TGF-β receptor (TGF-βR) I signaling, similar to SB-505124, an inhibitor of activin-like kinase receptors 4 and 5 (TGF-βRI) and 7 (14).

Phytohemagglutinin (PHA) was from Biochrom (Berlin, Germany). [Methyl-³H]Thymidine was obtained from Amersham (Braunschweig, Germany). ⁵¹Cr was purchased from New England Nuclear (Boston, MA). Human recombinant TGF-β1, TGF-β2, and mouse IL-2 were obtained from Peprotech (London, United Kingdom). Neutralizing pan–anti-TGF-β antibody was purchased from R&D (Wiesbaden, Germany). Specific enzyme-linked immunosorbent assay (ELISA) kits (R&D) were used for the detection of human and murine TGF-β. The human malignant glioma cell line LN-308 was kindly provided by N. de Tribolet (Lausanne, Switzerland). The murine glioma line SMA-560 was a kind gift of D. D. Bigner (Durham, NC). CCL64 mink lung epithelial cells were obtained from American Type Culture Collection (Manassas, VA).

The glioma cells and CCL64 cells were maintained in Dulbecco’s modified Eagle’s medium supplemented with 2 mmol/L l-glutamine (Gibco Life Technologies, Inc., Paisley, United Kingdom), 10% fetal calf serum (Biochrom), and penicillin (100 IU/mL)/streptomycin (100 μg/mL; Gibco Life Technologies, Inc.). Growth and viability of the glioma cells were examined by crystal violet staining, lactate dehydrogenase release (Roche, Mannheim, Germany), and trypan blue dye exclusion assays. For crystal violet staining, the cell culture medium was removed, and surviving cells were stained with 0.5% crystal violet in 20% methanol for 10 minutes. The plates were washed extensively under running tap water and air dried, and absorbance values were read in an ELISA reader at 550 nm wavelength. Human peripheral blood mononuclear cells (PBMCs) were isolated from healthy donors by density gradient centrifugation (Biocoll; Biochrom). Monocytes were depleted by adhesion and differential centrifugation to obtain peripheral blood lymphocytes (PBLs). To obtain purified T cells, PBMCs were depleted of B cells and monocytes using LymphoKwik T reagent (One Lambda, Canoga Park, CA). The purity of this population was >97% verified by flow cytometry using antihuman CD3-phycoerythrin antibody (Becton Dickinson, Heidelberg, Germany). Human polyclonal natural killer (NK) cell populations were obtained by culturing PBLs on irradiated RPMI 8866 feeder cells for 10 days (15). Murine NK cells were prepared from splenocytes from VM/Dk mice by positive selection using DX5 monoclonal antibody-coupled magnetic beads with the corresponding column system (Miltenyi Biotech, Bergisch Gladbach, Germany) and cultured with mouse IL-2 (5,000 units/mL) for at least 10 days before use. The human polyclonal NK cell cultures, PBLs, T cells, and mouse NK cells were grown in RPMI 1640 supplemented with 10% fetal calf serum, 2 mmol/L l-glutamine, 1 mmol/L sodium pyruvate, 50 μmol/L β-mercaptoethanol, and penicillin (100 IU/mL)/streptomycin (100 μg/mL).

SD-208 is a TGF-βRI kinase inhibitor developed by Scios Inc. (Fremont, CA). To assess the specificity of SD-208 for TGF-βRI, various kinase activities were assayed by measuring the incorporation of radiolabeled ATP into a peptide or protein substrate. The reactions were performed in 96-well plates and included the relevant kinase, substrate, ATP, and appropriate cofactors. The reactions were incubated and then stopped by the addition of phosphoric acid. Substrate was captured onto a phosphocellulose filter, which was washed free of unreacted ATP. The counts incorporated were determined by counting on a microplate scintillation counter (TopCount; Perkin-Elmer Corp., Boston, MA). The ability of SD-208 to inhibit the respective kinase was determined by comparing counts incorporated in the presence of compound with those incorporated in the absence of compound.

The levels of bioactive TGF-β were determined using the CCL64 bioassay. Briefly, 10⁴ CCL64 cells were adhered to 96-well plates for 24 hours and then exposed to recombinant TGF-β1, TGF-β2, or glioma cell culture supernatant (SN) diluted in complete medium for 72 hours. Growth was assessed by crystal violet staining at 72 hours. Glioma cell SNs were harvested from subconfluent cultures maintained for 48 hours in serum-free medium and heat-treated (5 minutes, 85°C) to activate latent TGF-β (11).

Glioma cells were cultured in the absence or presence of SD-208 (1 μmol/L) for 48 hours. The cells were pulsed for the last 24 hours with [methyl-³H]thymidine (0.5 μCi) and harvested (Tomtec, Hamden, CT), and incorporated radioactivity was determined in a liquid scintillation counter (Wallac, Turku, Finland).

The adherent glioma cells were detached nonenzymatically using cell dissociation solution (Sigma, Taufkirchen, Germany). Cell cycle analysis of glioma or immune effector cells was performed on fixed and 70% EtOH-permeabilized glioma or immune effector cells. RNA was digested with RNase A (GIBCO Life Technologies, Inc.). DNA was stained with propidium iodide (50 μg/mL). Fluorescence was measured in a Becton Dickinson FACSCalibur (Heidelberg, Germany).

Intracellular TGF-β signaling was assessed by reporter assays using pGL2 3TP-Luc (16) or pGL3 SBE-2 Luc (17) reporter gene plasmids kindly provided by J. Massagué (New York, NY) and B. Vogelstein (Baltimore, MD). The pGL2 3TP-Luc construct contains a synthetic promoter composed of a TGF-β–responsive plasminogen activator inhibitor 1 promoter fragment inserted downstream of three phorbol ester-responsive elements. The pGL3 SBE-2-Luc reporter contains two copies of the Smad-binding element GTCTAGAC. LN-308 and SMA-560 cells were transfected using FuGene (Roche). At 24 hours after transfection, the cells were pretreated in serum-containing medium with SD-208 for 12 hours (1 μmol/L). TGF-β1 (5 ng/mL) was then added for another 16 hours. The cells were lysed and transferred to a LumiNunc plate (Nunc, Roskilde, Denmark), and luminescence was measured in a LumimatPlus (EG&G Berthold, Pforzheim, Germany), using a luciferase assay substrate (Promega, Mannheim, Germany). For T cell assays, 5 × 10⁶ freshly isolated PBLs were cotransfected with 4.5 μg of pGL2–3TP-Luc or pGL3-SBE-2 Luc reporter gene plasmid and 0.5 μg of pRL-CMV (Promega), using the Nucleofector device and the cell type-specific human T-cell Nucleofector kit (Amaxa, Cologne, Germany). IL-2 (50 units/mL) was added 4 hours after nucleofection, and the cells were pretreated with SD-208 for 1 hour before TGF-β1 (5 ng/mL) was added for another 16 hours. The respective activities of firefly and Renilla reniformis luciferase were determined sequentially using the Firelite dual luminescence reporter gene assay (Perkin-Elmer, Rodgau-Jügesheim, Germany). Counts obtained from the measurement of firefly luciferase were normalized with respect to pRL-CMV.

Phosphorylated Smad (p-Smad) 2 levels in glioma cells were analyzed by immunoblot using 20 μg of protein per lane on 12% sodium dodecyl sulfate-polyacrylamide gels. PBMCs were analyzed using 100 μg per lane and 10% gels. After transfer to a polyvinylidene difluoride membrane (Amersham), the blots were blocked in PBS containing 5% skim milk and 0.05% Tween 20 and incubated overnight at 4°C with p-Smad2 antibody (2 μg/mL; Cell Signaling Technology, Beverly, MA). Visualization of protein bands was accomplished using horseradish peroxidase-coupled secondary antibody (Sigma) and enhanced chemiluminescence (Amersham). Total Smad2/3 levels were assessed using a specific Smad2/3 antibody (1 μg/mL; Becton Dickinson).

Invasion of glioma cells was measured by the invasion of 10,000 cells through Matrigel-coated Transwell inserts (Becton Dickinson). Briefly, Transwell inserts with 8-μm pore size were coated with Matrigel, and preincubated SMA-560 cells were applied to the upper wells and allowed to transmigrate through the membrane toward conditioned medium derived from NIH-3T3 fibroblasts that was added to the lower wells. Migrated cells on the lower side of the membrane were fixed, stained in toluidine blue solution (Sigma), and counted in five microscopic high-power fields using a microgrid.

Multicellular SMA-560 glioma cell spheroids were cultured in 25-cm² culture flasks base-coated with 1% Noble Agar (Difco Laboratories, Detroit, MI). Briefly, 4 × 10⁵ cells were suspended in 10 mL of medium, seeded onto 1% agar plates, and cultured until spheroids had formed. Spheroids of about 200 μm in diameter were selected for the experiments. Preincubated spheroids were seeded into collagen I and fibronectin-containing wells. Spheroid radius, which is determined by the invasion of single cells into the matrix (18), was analyzed by morphometry using the MCID digitalization system (Imaging Research, Ontario, Canada) at 24, 48, and 72 hours.

HLA-A2–mismatched human PBLs (10⁵ PBLs per well) were cocultured with 10⁴ irradiated (30 Gy) LN-308 glioma cells in 96-well plates in triplicates for 5 days. Some cocultures received PHA (5 μg/mL). The cells were pulsed for the last 24 hours with [methyl-³H]thymidine (0.5 μCi) and harvested (Tomtec), and incorporated radioactivity was determined in a liquid scintillation counter (Wallac).

HLA-A2–mismatched PBLs or T cells (10⁷ cells per 25-cm² flask) were cocultured with 10⁶ irradiated (30 Gy) LN-308 glioma cells for 5 days. Glioma cell targets were labeled using ⁵¹Cr (50 μCi, 90 minutes) and incubated (10⁴ cells per well) with effector PBLs harvested from the cocultures at effector to target ratios of 100:1 to 3:1. The maximum ⁵¹Cr release was determined by addition of 1% Nonidet P-40 (Sigma). After 4 hours, the SNs were transferred to a Luma-Plate TM-96 (Packard, Dreieich, Germany) and measured. The percentage of ⁵¹Cr release was calculated as follows: 100 × [experimental release − spontaneous release]/[maximum release − spontaneous release].

Interferon (IFN)-γ, tumor necrosis factor (TNF)-α, and IL-10 release by immune effector cells was assessed by Elispot assay in multiscreen-HA 96-well plates (Millipore, Eschborn, Germany) coated with corresponding antihuman capture antibodies (Becton Dickinson). Briefly, 5 × 10⁴ glioma cells were cocultured for 24 hours with 10⁵, 2.5 × 10⁵, or 5 × 10⁵ HLA-A2–mismatched, prestimulated (5 days) PBLs. The cells were removed using double-distilled water, and captured cytokines were visualized using biotinylated antibodies and streptavidin-alkaline phosphatase (Becton Dickinson). Spots were counted on an Elispot reader system (AID Diagnostic GmbH, Strassberg, Germany).

Male mice (BALB/c; Jackson Laboratories, Bar Harbor, ME) were studied in six groups of eight animals each. For each drug group, a single volume of SD-208 was administered by oral gavage 1 hour before dosing with TGF-β1 (R&D Systems) diluted in 100 μL of 0.1% bovine serum albumin (BSA)/4 mmol/L HCl/PBS by intravenous injection. The mice were sacrificed by cervical dislocation 1 hour later. Tissues were removed and lysed in 20 mmol/L Tris (pH 7.5) containing 1 mmol/L EDTA, 0.5% Triton X-100, 0.5% Nonidet P-40, 150 mmol/L NaCl, 1× protease inhibitor mixture (Roche), and 1× phosphatase inhibitor mixture set II (Calbiochem, San Diego, CA). Tissue was homogenized using an Ultra-turrax T8 (Reyom Instruments, Brabcova, Czech Republic). Tissue homogenates were clarified by centrifugation, and the SN fraction was collected. Protein concentrations were determined with a bicinchoninic acid protein assay (Pierce, Rockford, IL). The levels of p-Smad were determined by sandwich ELISA. Briefly, 96-well ELISA plates were coated with an anti-Smad2/3 monoclonal antibody (100 ng/well; Becton Dickinson) for 18 hours at 4°C. Excess antibody was removed, and the wells were treated with blocking buffer (0.3% BSA/PBS) for 2 hours at room temperature. Tissue lysates (125–150 μg of total protein) were added to each well and incubated overnight at 4°C. Wells were rinsed before adding a polyclonal anti–p-Smad 2/3 antiserum diluted in 2% BSA/0.5% Tween 20/PBS. After a 2-hour incubation at room temperature, the wells were washed, and secondary antibody was applied (horseradish peroxidase-conjugated goat antirabbit IgG; Southern Biotech, Birmingham, AL). After 1 hour, the wells were developed with tetramethylbenzidine (Sigma). The plate was incubated for 5 to 30 minutes before the reaction was stopped with 0.5 N H₂SO₄ and read at 450 nm in a SpectraMax 250 plate reader (Molecular Devices, Sunnyvale, CA).

VM/Dk mice were purchased from the TSE Resource Center (Berkshire, United Kingdom). Mice of 6 to 12 weeks of age were used for the survival experiments. The experiments were performed according to the German animal protection law. Groups of eight mice were anesthesized before all intracranial procedures and placed in a stereotaxic fixation device (Stoelting, Wood Dale, IL). A burr hole was drilled in the skull 2 mm lateral to the bregma. The needle of a Hamilton syringe (Hamilton, Darmstadt, Germany) was introduced to a depth of 3 mm. SMA-560 cells [5 × 10³ cells (19)] resuspended in a volume of 2 μL of PBS were injected into the right striatum. Three days later, the mice were allowed to drink SD-208 at 1 mg/mL in deionized water. The mice were observed daily and, in the survival experiments, sacrificed on development of neurologic symptoms.

Glioma-bearing mice were sacrificed 10 days after tumor implantation by cardiac puncture, perfused, and postfixed in 4% paraformaldehyde (Sigma) overnight. Five-micrometer paraffin sections were cut at 150-μm intervals from each brain. The sections were deparaffinized, rehydrated, and stained with hematoxylin and eosin [H&E (Harri’s; American Master Tech, Lodi, CA)], rat antimouse monoclonal CD34 IgG2a (1:100; CL8927AP; Cedarlane, Hornby, Canada), rabbit polyclonal anti-Ki67 (1:100; ab833-500; Novus Biologicals, Littleton, CO), rabbit antimouse active caspase 3 (1:400; AF835; R&D), antimouse CD8 (1:50; 53-6.7; BD Biosciences, Heidelberg, Germany), antimouse CD11b (1:50; M1/70; BD Biosciences), or anti-Ly-49G2 (1:50; 4D11; BD Biosciences). Biotinylated secondary antibodies (1:150; Zymed, San Francisco, CA) were used for detection. Streptavidin-alkaline phosphatase (1:100) was added, and the staining was developed with naphtol as substrate and levamisole as inhibitor of endogeneous alkaline phosphatase (Fast Red Tablets; Roche). The negative control for CD34 was normal rat IgG2a (CBL605; Chemicon International, Temecula, CA). The negative control used for Ki67 and caspase 3 was normal rabbit IgG (SC-2027; Santa Cruz Biotechnology, Santa Cruz, CA). The negative control for CD8, Ly-49G, and CD11b was rat IgG2a or rat IgG2b. In these assays, murine spleen served as a positive control, and normal murine brain served as a negative control. The total number of CD34+ microvessels was counted in an area of 0.63 mm² corresponding to two high-power fields in two nonconsecutive sections in the tumor center. To assess the percentage of proliferating cells, the number of Ki67-positive nuclei was counted. At least 600 nuclei were counted in four high-power fields in two nonconsecutive sections in the tumor center. To assess the degree of apoptosis, caspase 3-positive cells were counted in the tumor center in two nonconsecutive sections.

Glioma-bearing mice were sacrificed 10 days after tumor cell injection. Splenocytes were isolated and used in 24-hour IFN-γ Elispot assays as described above. Furthermore, these cells were stimulated with IL-2 (5,000 units/mL) for 10 days to generate lymphokine-activated killer (LAK) cells, which were used in ⁵¹Cr release assays against SMA-560 glioma cells as targets.

The experiments were usually performed at least three times with similar results. Significance was tested by Student′s t test. P values are derived from two-tailed t tests.

The initial characterization of SD-208 in cell-free assays led to its identification as a potent inhibitor of TGF-βRI. SD-208 exhibited an in vitro specificity for TGF-βRI kinase of >100-fold compared with TGF-βRII kinase and at least 20-fold over members of a panel of related protein kinases (Table 1). The differential patterns of release of TGF-β1/2 by human glioma cell lines have been reported previously (11). The TGF-β release by the cell lines examined here was ascertained by ELISA. The levels of TGF-β released into the SN were unaffected by SD-208 within the concentration and time frame of the ensuing experiments (Fig. 1,A). CCL64 mink lung epithelial cells are sensitive to the growth-inhibitory effects of human TGF-β1 and TGF-β2 at EC₅₀ concentrations of 0.5 ng/mL. The inhibitory effects of recombinant TGF-β as well as those of TGF-β–containing glioma cell SN were abrogated by specific TGF-β antibodies (11). The CCL64 bioassay was used here to verify the TGF-β–antagonistic properties of SD-208 (Fig. 1,B). SD-208 rescued the inhibition of growth mediated by TGF-β1 or TGF-β2 (10 ng/mL) or diluted SMA-560 (data not shown) or LN-308 glioma cell SN in a concentration-dependent manner, with an EC₅₀ concentration of 0.1 μmol/L (Fig. 1,C). When these bioassays were performed in the absence of serum in the CCL64 medium, the EC₅₀ for SD-208 was 0.03 μmol/L (data not shown), corresponding to the data in Table 1.

We next examined the biological effects of SD-208 on murine and human glioma cells in vitro. The concentrations required to block the growth-inhibitory effects of TGF-β in the CCL64 bioassay had no effect on the proliferation of either glioma cell line. SD-208 did not reduce proliferation assessed by [methyl-³H]thymidine incorporation or viability assessed by lactate dehydrogenase release or trypan blue dye exclusion assays at concentrations up to 1 μmol/L for 48 hours in SMA-560 (Fig. 2,A) or LN-308 cells (data not shown). Furthermore, neither exogeneous TGF-β nor neutralizing TGF-β antibodies had such an effect. Accordingly, cell cycle analysis showed no difference in both cell lines after exposure to SD-208 for 48 hours (Fig. 2,B and C). The inhibition of TGF-β signaling transduced by endogenous or exogenous TGF-β was ascertained by demonstrating that SD-208 interfered with Smad2 phosphorylation without altering total cellular Smad2/3 levels (Fig. 2,D). Similarly, two different reporter assays revealed a strong inhibition of TGF-β signaling when the glioma cells were treated with SD-208 (Fig. 2 E).

We next examined the biological effects of SD-208 on SMA-560 cells in two independent migration and invasion paradigms in vitro. SD-208 reduced the invasion of glioma cells from a spheroid into a three-dimensional collagen I gel and also the transmigration of glioma cells in a Boyden chamber assay. Moreover, the proinvasive effect of exogenous TGF-β was neutralized by SD-208 in both assays (Fig. 3 A and B).

Consistent with the effects of SD-208 on TGF-β–mediated signaling in glioma cells (Fig. 2,D and E), immunoblot analysis and reporter assays also revealed a strong resistance to TGF-β signaling in the presence of SD-208 in immune cells (Fig. 4,C and D). Again, SD-208 did not affect cell cycle distribution, apoptosis, or proliferation (Fig. 4,A and B). The next series of experiments was designed to examine whether SD-208 restores allogeneic immune cell responses to cultured human glioma cells. When HLA-A2–mismatched PBLs or purified T cells were cocultured with irradiated glioma cells in the presence of SD-208, their lytic activity in a subsequent 4-hour ⁵¹Cr release assay was significantly enhanced (Fig. 5,A). Similar effects were obtained using neutralizing TGF-β antibodies (10 μg/mL, added every 2 days; data not shown). The release of IFN-γ by HLA-mismatched PBLs was strongly inhibited when the priming had taken place in the presence of glioma cells. SD-208 restored the IFN-γ release to levels comparable with PBLs precultured in the absence of LN-308 cells (Fig. 5,B). Similar results were obtained for TNF-α (Fig. 5,C). In contrast, IL-10 release was stimulated after coculturing with LN-308 cells, and SD-208 reduced the release of IL-10 by immune effector cells generated both from unstimulated and glioma cell-primed cultures (Fig. 5,D). The lytic activity of polyclonal NK cells against LN-308 targets was inhibited by exogenous TGF-β, and TGF-β–mediated inhibition was relieved by SD-208 (Fig. 5,E). Similarly, LN-308 SN inhibited NK cell activity, and this inhibition was also blocked by SD-208 (Fig. 5 F) or neutralizing TGF-β antibodies (data not shown). Note that whereas the absolute lytical activity was donor specific, the effect of SD-208 was consistent through all experiments.

To verify the bioavailability of orally administered SD-208, we ascertained that SD-208 inhibited the TGF-β–dependent in vivo phosphorylation of Smad2/3 in spleen and brain (Fig. 6,A and B). Note that exogenous TGF-β was a more potent inducer of Smad2/3 phosphorylation in spleen than in brain, but that SD-208 was an equally effective antagonist of TGF-β in both tissues. The therapeutic effects of SD-208 administered via the drinking water (1 mg/mL) were assessed in the syngeneic SMA-560 mouse glioma paradigm. The development of neurologic symptoms was delayed in SD-208–treated mice, and mean survival was prolonged to 25.1 ± 6.5 days (median, 23 days) compared with 18.6 ± 2.1 days (median, 18 days) in vehicle-treated animals (Fig. 6,C; P = 0.004, t test). The survival rate at 30 days was 29% in SD-208–treated animals but 0% in control animals. Preliminary evidence also indicated that SD-208 modulated the immune response of the glioma-bearing animals. Elispot assays for IFN-γ release by splenocytes harvested at day 7 after the initiation of SD-208 treatment revealed an increase over background in three of five SD-208–treated animals but in only one of five control animals (data not shown). Furthermore, LAK cells generated from the splenocytes of SD-208–treated animals showed an enhanced lytic activity against SMA-560 as targets (Fig. 5 D).

The tumor volumes at day 10 showed a trend toward a lower tumor burden in SD-208–treated animals, but these changes were not significant (data not shown), consistent with the clinical course at day 10 shown in Fig. 6,C. Immunohistochemistry revealed no significant difference in blood vessel formation (CD34), tumor cell proliferation (Ki67), or apoptosis (active caspase 3) between the gliomas of untreated and SD-208–treated animals (Table 2). There was little tumor infiltration by immune or inflammatory cells in the vehicle group as detected by H&E staining. However, H&E staining revealed major differences in the sizes and the histology of the gliomas of SD-208–treated animals: as indicated by the variable clinical course (Fig. 6,C), there appeared to be nonresponder and responder animals. Animals with small tumors had stronger immune cell infiltration of their tumors than mice bearing larger tumors. Typical staining patterns for CD8 T cells, NK cells (Ly49G), and macrophages/neutrophils (CD11b) in a vehicle-treated animal, an SD-208–treated nonresponder animal, and an SD-208–treated responder animal are presented in Fig. 7.

Antagonizing the biological effects of TGF-β has become a very attractive experimental strategy to combat various types of cancer including malignant gliomas. Current rationales for anti–TGF-β strategies include the putative role of TGF-β in migration and invasion (13), metastasis (20), and tumor-associated immunosuppression (8, 21). All of the TGF-β–based therapeutic approaches evaluated in experimental gliomas thus far appear to have limitations with regard to their transfer into the clinic. Antisense oligonucleotides pose severe problems in terms of delivery to the desired site of action. The same applies to gene therapy strategies based on the transfer of the decorin gene (10). Inhibition of furin-like proteases aimed at limiting TGF-β bioactivity at the level of TGF-β processing (11) may not be achieved with acceptable specificity at present because a whole variety of molecules require processing by such enzymes (22). More specificity may result from the use of soluble TGF-βR fragments that act to scavenge bioactive TGF-β before it reaches the target cell population (20, 21). However, this approach may find its limits in the complex pathways of storage and activation of TGF-β. These considerations suggest that specific small molecules designed to protect cells from the actions of TGF-β at the level of receptor-dependent intracellular signal transduction are a particularly promising alternative for antagonizing TGF-β (14).

Here we characterize the activity of one such candidate agent, SD-208, against murine and human glioma cells in vitro and in vivo. Human LN-308 cells were chosen because they are paradigmatic for their prominent TGF-β synthesis (refs. 3 and 11; Fig. 1,A). SMA-560 cells transplanted in syngeneic VM/Dk mice probably represent the best model for the immunotherapy of rodent gliomas (19). We show that SD-208 is a potent TGF-βRI kinase inhibitor (Table 1) that blocks the biological effects of TGF-β1 and TGF-β2 as well as glioma cell SN in the CCL64 mink lung epithelial assay (Fig. 1,B and C). Because SD-208 did not modulate glioma cell proliferation at concentrations of up to 1 μmol/L (Fig. 2,A), we did not confirm a negative growth-regulatory effect of TGF-β on SMA-560 cells (23). Smad2 phosphorylation is induced by TGF-β in a SD-208–sensitive manner (Fig. 2,C), indicating that TGF-β signaling is not abrogated constitutively in glioma cells but may not play a role in the modulation of glioma cell proliferation. Moreover, as expected (13), the antagonism of autocrine and paracrine signaling by TGF-β in SD-208–treated glioma cells, as confirmed by reporter assay (Fig. 2,E), resulted in a potent inhibition of migration and invasion (Fig. 3).

We then focused on the desired immune modulatory effect of SD-208, which should result in an enhanced immunogenicity of glioma cells as a consequence of reduced TGF-β bioactivity. As predicted, human PBLs and purified T cells developed enhanced lytic activity against LN-308 glioma cell targets when prestimulated with glioma cells in the presence of SD-208 (Fig. 5,A). This was paralleled by an enhanced release of proinflammatory cytokines such as IFN-γ and TNF-α and a reduced release of the immunosuppressive cytokine IL-10 in SD-208–treated cells (Fig. 5,B–D). Similarly, SD-208 restored the lytic activity of polyclonal NK cell cultures cocultured with TGF-β or LN-308 SN (Fig. 5 E and F).

The strong reduction of Smad phosphorylation in the unlesioned mouse brain indicates that SD-208 may reach sufficient levels beyond the intact blood–brain barrier to counteract the biological effects of tumor-derived TGF-β (Fig. 6,B). Accordingly, SD-208 prolonged the median survival of SMA-560 glioma-bearing mice significantly (Fig. 6,C). No dose-limiting toxicity was reached in these experiments, but higher doses could not be administered via drinking water because of the poor solubility of SD-208, suggesting that the therapeutic effect of SD-208 or related agents might even be improved in that glioma model. The therapeutic effect of SD-208 might be mediated by inhibition of glioma cell migration and invasion (13), promotion of antiglioma immune responses (8), or both. An immune contribution is suggested by the histologic analyses (Fig. 7), which delineated an interrelation between tumor shrinkage and the degree of immune cell infiltration.

The present data strongly suggest a role for SD-208 or related molecules in the treatment of gliomas. Such a systemic treatment with TGF-βRI kinase inhibitors might well be combined with local approaches to limit the bioavailability of TGF-β, e.g., TGF-β antisense oligonucleotides that are already evaluated clinically.