The peroxisome proliferator-activated receptor-γ agonist troglitazone inhibits transforming growth factor-β–mediated glioma cell migration and brain invasion

Gliomas are the most common primary tumors in the central nervous system, with glioblastomas as the most malignant entity (1). Despite multimodal therapy regimens, including neurosurgical resection, radiochemotherapy, and polychemotherapy, the prognosis of glioma patients remains poor. Less than 3% of affected patients survive >5 years after diagnosis (2). Rapid proliferation and diffuse brain invasion are histopathologic hallmarks of these tumors and are likely to determine unfavorable prognosis. Among the plethora of molecular-genetic abnormalities identified within cell cycle control mechanisms of malignant glioma cells (1), the transforming growth factor-β (TGF-β) signaling pathway plays a decisive role for tumor growth and cell motility (3) and remarkable efforts were initiated to target the TGF-β pathway for adjuvant glioma therapy (4).

The peroxisome proliferator-activated receptor-γ (PPAR-γ) is a ligand-activated transcription factor that belongs to the superfamily of hormone receptors (5). Major physiologic mechanisms of PPAR-γ include the regulation of adipocyte differentiation and glucose homeostasis. The synthetic thiazolidinediones rosiglitazone, pioglitazone, ciglitazone, and troglitazone were developed as specific PPAR-γ activators. Beyond the established clinical use of thiazolidinediones for oral therapy of type 2 diabetes, recent in vitro and in vivo studies have shown that thiazolidinediones also play an important role as regulators of inflammation, cell proliferation, and resistance to apoptosis in cancer (6). The antitumor efficacy of thiazolidinediones was explored in several in vitro and in vivo studies as well as in clinical trials addressing angiosarcomas, breast cancers, prostate cancers, advanced melanomas, and soft tissue sarcomas (7). Analysis of the four thiazolidinediones rosiglitazone, pioglitazone, ciglitazone, and troglitazone shown in this study emphasized their potency to reduce glioma cell growth, whereas troglitazone seems to be more efficient than the remaining thiazolidinediones tested. By applying an ex vivo tumor implantation model, micromolar doses of troglitazone diminished glioma progression in an organotypic brain environment. We show that this property is based on the ability of troglitazone to induce cell cycle arrest and to reduce glioma cell migration and brain invasion. Troglitazone-induced inhibition of glioma cell motility occurred in a PPAR-γ–independent manner and is in concordance with its ability to modulate the expression of genes involved in TGF-β signaling. Based on the identification of troglitazone as a potent inhibitor of TGF-β release in glioma cells already at low micromolar doses and the increasing evidence for a prominent role of TGF-β in glioma cell motility, troglitazone and troglitazone derivatives may be considered for adjuvant glioma therapy to counteract TGF-β–mediated brain invasion.

Thiazolidinediones were purchased from Alexis Biochemicals. GW9662 and SB431542 were obtained from Sigma-Aldrich. All compounds were dissolved in DMSO and diluted in cell culture medium before experimental use.

SMA-560 mouse glioma cells (8) were kindly provided by D.D. Bigner (Duke University, Durham, NC). The rat glioma cell line F98 and the human glioma cell line U-87 MG were obtained from the American Type Culture Collection. Glioma cells were cultivated as described (9).

Cell proliferation was measured using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium-bromide (MTT) assay (10) as described (9) and using the 5-bromo-2′-deoxyuridine labeling and detection kit III (Roche). For the 5-bromo-2′-deoxyuridine assay, 50,000 cells/mL were seeded into 96-well plates (100 μL/well), treated as described for MTT assay (9), and analyzed after 96 h according to the manufacturer's protocol. Lactate dehydrogenase, a cytosolic enzyme that is released from the cell upon loss of plasma membrane integrity, was measured using the CytoTox 96 Assay (Promega) according to the manufacturer's protocol (10,000 cells per well using 96-well plates, 48 h). For all assays, each experiment was repeated eight times (n = 8).

Forty thousand cells per milliliter were seeded in 25 cm² culture flasks. After incubation with the test compound or solvent only, cells were washed in PBS, fixed in methanol, washed twice in PBS, and incubated with PBS containing 2% FCS and 40 μg/mL RNase A (Roche). Cells were suspended in PBS containing 20 μg/mL propidium iodide (Sigma-Aldrich). The rate of apoptosis and the cell cycle distribution were analyzed by flow cytometry (10,000 counts per sample) using FACSCalibur equipped with Cell Quest software (BD Biosciences). The data were evaluated using Win MDI version 2.8 (TSRI) and Cylchred version 1.0.2 (Cardiff University, Wales, United Kingdom) as described (11, 12). For each experiment, analyses were repeated thrice (n = 3).

RNA isolation was done using the RNeasy Mini Kit (Qiagen). For nonquantitative analyses, reverse transcription was done using oligodeoxythimidylate primers and 1 μg of total RNA by applying the SuperScript First-Strand Synthesis System (Invitrogen). Two microliters of each 20 μL reverse transcription reaction were used for each reverse transcription-PCR (RT-PCR). Mouse, rat, and human β-actin and PPAR-γ cDNA were amplified using primers that bind to all three orthologues of each gene (β-actin-fwd AACGGCTCCGGCATGTGCAA, β-actin-rev CTCAAACATGATCTGGGTCATCTT; PPARγ-fwd CATGCTTGTGAAGGATGCAAG, PPARγ-rev CCCATCATTAAGGAATTCATGTC). PCR conditions were as follows: 5 min initial denaturation (94°C), followed by 36 cycles (94°C for 45 s, 60°C for 45 s, 72°C for 45 s) and final extension (72°C for 7 min). PCR products were visualized on agarose gels by ethidium bromide staining. For quantitative RT-PCR analyses, one-step RT-PCR was preformed using the QuantiTect SYBR Green RT-PCR kit and validated QuantiTect Primer Assays (Qiagen) on an Applied Biosystems 7500 Real-time PCR System. Each 20 μL RT-PCR reaction contained 10 ng total RNA. To ensure that equal amounts of total RNA have been used for each analysis, the exact RNA concentration of each sample was determined by applying the RiboGreen RNA Quantitation Kit (Invitrogen) using a Tecan Safire² microplate reader (Tecan). The following QuantiTect primer assays were used: Tgf-β₁, Tgf-β₂, Tgf-β₃, Tgf-βrI, Tgf-βrII, and Fkbp1a. The comparative method of relative quantification (2^−ΔCt) was used to calculate the expression levels of each target gene relative to time- and solvent-matched controls. The RT-PCR specificity was verified by melting curve analysis. Real-time RT-PCR was conducted thrice for each gene and each RNA sample with four technical repetitions (n = 12).

Organotypic entorhino-hippocampal slice cultures (OHSC) were prepared and maintained according to the interface technique (13) as described in detail (9). In brief, 7-day-old Wistar rats were decapitated; the brains were rapidly removed and placed into ice-cold preparation medium. After dissection of the frontal pole of the hemispheres and the cerebellum, the brains were cut in 350-μm-thick horizontal slices using a vibratome (Leica VT 1000S) in preparation medium. Brain slices were transferred onto culture plate insert membranes (Becton Dickinson; 0.4 μm pore diameter) and subsequently into six-well culture dishes containing 1.2 mL OHSC medium (9). The medium was changed 1 day after preparation and every 2nd day thereafter. For the assessment of potential neurotoxicity to rat brain parenchyma, OHSCs were treated either with the test compound or solvent only. The medium, including the test compound or solvent, was changed every 2nd day. Six days after initial treatment, irreversibly damaged cell bodies were visualized by propidium iodide staining as described (14). For each experiment, eight brain slices were used (n = 8).

Transfection of the pEGFP-N1 expression vector (Clontech) into F98 glioma cells was done as described (9, 15). Transfected cells were cultured in selection medium containing 500 μg G418/mL (Sigma-Aldrich). Green fluorescent protein (GFP)–positive F98 glioma cells (5,000) were implanted within a total volume of 0.1 μL medium into the entorhinal cortex 1 day after slice preparation (9, 14). One day after implantation and every following day, glioma growth and invasion were evaluated using a fluorescence microscope as described below. Seven brain slices were used for each experiment (n = 7).

Cell migration was analyzed using the monolayer wound-healing assay as described (16). Glioma cells were cultured until confluency was achieved. Scratch wounds were added using standard 1,000 μL pipette tips followed by treatment with the test compound or solvent only. The wounds were monitored by phase contrast microscopy. Images were taken at 2-h intervals over the time course of 12 to 24 h and analyzed as described below. Eight images were used for each experiment (n = 8). The in vitro migration assay was done using a Boyden chamber (QCM-FN Migration Assay, Chemicon). Briefly, 2.5 × 10⁵ F98 glioma cells, pretreated with the test compound/compounds or solvent for 24 h, were transferred into each Boyden chamber. At 24 h after incubation, cells that migrated through the fibronectin-coated chamber membranes (8 μm pore diameter) were quantified according to the manufacturer's protocol. Experiments were repeated thrice (n = 3).

Morphometric analyses were done using high-power optical fields digitized with a charge coupled device camera (Color View II, Soft Imaging System) equipped with a BX51 microscope (Olympus) using analySIS documentation software (Soft Imaging System). Fluorescence (enhanced GFP)–labeled tumor cells transplanted into OHSCs as well as propidium iodide staining intensities were analyzed by an Olympus IX70 microscope equipped with a tetramethyl rhodamine isothiocyanate (excitation filter 520–550 nm, barrier filter 580 nm) and FITC (excitation filter 450–490 nm, band filter 520–550) narrow band filter and a charge coupled device camera (F-View II; Soft Imaging System).

For the analysis of TGF-β_1-3 expression in cell culture supernatants, F98 glioma cells were cultivated as described. Forty percent to 50% confluent F98 cell cultures were washed with PBS, cultivated in serum-free DMEM medium, and treated with the test compound or solvent only for 48 h. Cell culture supernatants were collected and centrifuged to remove persistent cells, followed by protein concentration using Amicon Ultra-15 centrifugal filter devices (Millipore). After centrifugation (45 min, 4000 × g, room temperature), the concentrated retentates (30 μg total protein per lane) were subjected to Western blot analyses using standard protocols (12% SDS-PAGE). Equal loading of each well was confirmed by Ponceau staining. The rabbit polyclonal antibody TGF-β1/2/3 (H-112; Santa Cruz Biotechnology) was added in a dilution of 1:250 to detect all three TGF-β variants. PPAR-γ and β-actin expression levels in rat brain and F98 whole-cell lysates were analyzed using rabbit polyclonal PPAR-γ (1:400, Abcam) and mouse monoclonal β-actin (1:20,000, Sigma-Aldrich) antibodies. Peroxidase-conjugated goat anti-rabbit IgG (1:10,000, Pierce) and goat anti-mouse IgG (1:10,000, Jackson ImmoResearch) were used as secondary antibodies. Detection of signals was carried out using the Super Signal West Pico chemiluminescence reagent (Pierce).

To quantify TGF-β₁ protein levels in glioma cell culture supernatants, ELISA assays were done according to the manufacturer's protocols (Mouse/Rat/Porcine/Canine Quantikine TGF-β1, Human Quantikine TGF-β1 ELISA Kit, R&D Systems). Briefly, a monoclonal antibody specific for the mature TGF-β₁ ligand has been precoated onto the bottom of 96-well plates. Standards, controls, and samples are transferred into the wells and any TGF-β₁ present is bound by the immobilized antibody. Subsequently, an enzyme-linked polyclonal antibody against TGF-β₁ was added to the wells to sandwich the immobilized TGF-β₁. A substrate solution is added to the wells and color develops in proportion to the amount of TGF-β₁ bound in the initial step. The absorbance was measured using a Tecan Safire² microplate reader at an absorbance of 450 nm (reference wavelength 540 nm). Each experiment was repeated thrice (n = 3).

Concentration-effect curves were analyzed by nonlinear regression analysis (GraphPad Software 4.03) as described (9), by applying a four-variable logistic equation containing the variables top plateau, bottom plateau, inflection point (IC₅₀), and curve slope. Drug concentrations inducing a 50% and 90% reduction of cell growth/TGF-β₁ release, respectively, were read from the best-fit curves. For statistical analysis, single comparisons were analyzed by unpaired t tests, whereas multiple comparisons among the data sets were done by two-way ANOVA (GraphPad). Three levels of statistical significance were discriminated (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Concentration-dependent inhibition of glioma cell growth by PPAR-γ agonists was investigated using cell lines derived from mouse (SMA-560), rat (F98), and human (U-87 MG). As shown by MTT assay, all compounds inhibited glioma cell growth in a concentration-dependent manner. The reduction of cell growth occurred with steep curve slopes and consequently there was a narrow range of effective concentrations producing a 10% to 90% inhibition (Fig. 1A). Irrespective of the drug and cell line tested, the IC₅₀ was found at doses ranging from 20 to 85 μmol/L (variables characterizing the cell growth inhibition are given in Fig. 1 legend). With IC₅₀ doses of 20 μmol/L (SMA-560), 30 μmol/L (F98), and 34 μmol/L (U-87 MG), troglitazone seemed to be more potent than the remaining thiazolidinediones tested. The reduction of glioma cell proliferation was confirmed by measuring 5-bromo-2′-deoxyuridine uptake. Using 40 μmol/L doses, all compounds significantly reduced glioma cell proliferation, whereas troglitazone again had a higher efficacy compared with the remaining drugs (Fig. 1B). The cell lines did not essentially differ in their dose-dependent response to each drug (P > 0.05, two-way ANOVA). This finding is particularly interesting because the PPAR-γ receptor is expressed in SMA-560 and U-87 MG, but only barely detectable in F98 glioma cells (Fig. 2A and B). Treatment of F98 cells with IC₅₀ (30 μmol/L) or IC₉₀ (76 μmol/L) doses of troglitazone did not induce PPAR-γ expression (Fig. 2C). In all, these data suggest that the antiproliferative effects of troglitazone do not correlate with PPAR-γ expression.

Subsequent analyses focused on troglitazone due to its superior efficacy compared with the remaining test drugs. Consistent with previous findings using other cancer cell types (17–21), repeated flow cytometric analyses of F98 glioma cells after exposure to IC₉₀ concentrations of troglitazone for 48 h revealed a significant accumulation of glioma cells in the G₁ phase (control, 34.5 ± 2.2%; troglitazone IC₉₀, 46.6 ± 3.5%; P < 0.05, t test) and reduction of cells in the S phase (control, 44.2 ± 7.2%; troglitazone IC₉₀, 32.0 ± 4.0%; P < 0.05, t test), confirming that troglitazone blocks events critical for G₁-S progression. The share of cells in the G₂ phase was not significantly altered compared with time- and solvent-matched controls (control, 17.2 ± 6.0%; troglitazone IC₉₀, 16.3 ± 2.2%; P > 0.05, t test). The proportion of apoptotic cells in the sub-G₁ phase was also not significantly elevated (control, 4.2 ± 1.7%; troglitazone IC₉₀, 5.1 ± 2.3%, P > 0.05, t test). Consistently, exposure of F98 cells to troglitazone IC₉₀ concentrations did not increase lactate dehydrogenase release (P > 0.05, t test, data not shown), indicating that troglitazone acts antiproliferative rather than proapoptotic in F98 cells at the doses used.

We used rat organotypic hippocampal brain slice cultures (OHSC) to monitor glioma cell proliferation and brain invasion in the organotypic brain environment (9). Initially, potential neurotoxic effects of troglitazone were evaluated ex vivo using the rat OHSC model. Increased propidium iodide uptake, which indicates irreversible cell damage, was not observed after treatment with IC₉₀ (76 μmol/L) or 2× IC₉₀ (152 μmol/L) concentrations, indicating that troglitazone is not neurotoxic to highly vulnerable rat brain parenchyma at concentrations shown to effectively diminish glioma cell proliferation in vitro (data not shown). To evaluate the propensity of troglitazone to reduce glioma progression in an organotypic brain environment, enhanced GFP–labeled F98 glioma cells were implanted into the entorhinal cortex of OHSCs (Fig. 3A). The tumor infiltration area was quantified each day after implantation up to 6 days by fluorescence microscopy. A continuous increase of the bulk tumor mass was observed in solvent-matched control experiments at all time periods. Six days after implantation, the tumor infiltration area increased ∼11-fold compared with the initial tumor size at day 1 after implantation (Fig. 3B and C). In contrast, the tumor infiltration size remained stable over the period of 6 days after treatment with IC₉₀ doses, indicating that troglitazone effectively impeded tumor progression ex vivo (Fig. 3B and C). Increased, nontoxic doses of troglitazone (2× IC₉₀), diminished tumor progression in a similar manner as observed after treatment with IC₉₀ concentrations (Fig. 3B and C). Interestingly, even 2× IC₉₀ doses of troglitazone were not capable to significantly reduce initial tumor infiltration area (P > 0.05, t test), which is consistent with the absence of apoptotic tumor cell death after troglitazone treatment (see above). This finding indicates that troglitazone is not able to reduce existing tumor masses, but effectively inhibits tumor progression in an organotypic environment.

Using the rat organotypic tumor invasion model, we have shown that enhanced GFP–labeled F98 glioma cells migrate into the adjacent brain parenchyma, visible as a diffuse corona surrounding the bulk tumor mass (9). Diffuse infiltration of F98 cells into the brain parenchyma was observed in solvent-matched controls, but not after treatment with IC₉₀ and 2× IC₉₀ doses of troglitazone (Fig. 3B). Thus, we hypothesized that the lack of tumor progression and brain invasion observed after troglitazone treatment was accompanied by a reduction of glioma cell migration. By applying a monolayer wound-healing assay, we showed that the F98 glioma cell locomotion was substantially altered by troglitazone (Fig. 4A and B). After mechanical injury, bordering glioma cells show a lamellar protusion and migration toward the center of the wound, resulting in a progressive reduction of the wound area (Fig. 4A). In control experiments, the wound area was reduced by 30 ± 5% at 12 h and by 53 ± 6% at 22 h after injury, respectively, whereas the reduction of the wound area was diminished in a dose-dependent manner after troglitazone treatment (Fig. 4A and B). To confirm this finding, glioma cell migration was further analyzed by using a Boyden chamber assay. Compared with the migratory activity of solvent-treated controls, troglitazone dose-dependently decreased F98 glioma cell migration by 83.3 ± 2.3% (IC₅₀) and 90.8 ± 2.1% (IC₉₀), respectively. The migratory activity of solvent-treated F98 cells and the antimigratory effects of troglitazone were not impaired by the irreversible and selective PPAR-γ inhibitor GW9662 (22), indicating that the antimigratory properties of troglitazone were not PPAR-γ mediated (Fig. 4C).

TGF-β signaling has shown to be crucially implicated in glioma cell migration and brain invasion. Exogenously added TGF-β₁ and TGF-β₂ were found to elicit a strong stimulation of migration and invasiveness in a variety of glioma cells in vitro (23–26), whereas small interfering RNA–mediated Tgf-β₁ and Tgf-β₂ gene silencing has shown to counteract glioma cell motility and invasiveness (27). Consistently, selective inhibition of TGF-β signaling by SB431542, a small-molecule inhibitor of the type I TGF-β receptor kinase activity (28), reduced F98 migration by 73.9 ± 1.9% similar to troglitazone (Fig. 4C). Based on this finding, we hypothesized that troglitazone is able to counteract TGF-β–mediated glioma cell migration and brain invasion. Three variants of TGF-β (TGF-β_1-3) are encoded by three homologous genes. Real-time PCR analysis revealed a dose-dependent down-regulation of Tgf-β_1-3 and their receptors (Tgf-βrI, Tgf-βrII) after troglitazone treatment, whereas the FK506-binding protein 1A (Fkbp1a) gene expression, which encoded the protein that inhibits TGF-β receptor I activity (29), remained unchanged (Fig. 5). To analyze the levels of released TGF-β_1-3, supernatants of F98 glioma cell cultures were subjected to immunoblotting. Using a polyclonal antibody directed against all three TGF-β isoforms, TGF-β–specific signals were abundant in solvent-matched controls, but not after troglitazone treatment (IC₉₀), giving evidence that troglitazone decreased the TGF-β release to very low levels (Fig. 6A). A comparative real-time PCR analysis of basal Tgf-β_1-3 expression levels in nontreated F98 cells identified >2⁶-fold higher expression levels of Tgf-β₁ compared with Tgf-β₃ and Tgf-β₂, thus identifying Tgf-β₁ as the predominant Tgf-β isoform (data not shown).

Extracellular TGF-β₁ and TGF-β₂ concentrations of 2 to 5 ng/mL cell culture medium have shown to elevate the migratory activity of glioma cells (24, 25). We did an absolute TGF-β₁ quantification to assess whether F98 glioma cells release TGF-β₁ at relevant doses. Promigratory concentrations of 5.92 ng/mL (±0.10 ng/mL) cell culture medium were observed in solvent-matched controls, suggesting that F98 cells promote migration by an autocrine signaling pathway. In contrast, troglitazone decreased TGF-β₁ protein levels by 94.93 ± 0.22% using IC₅₀ and almost completely using IC₉₀ doses, respectively, to 0.30 ng (±0.01 ng) and 0.01 ng/mL (±0.00 ng/mL). An extended, dose-dependent analysis revealed that already 7.41 μmol/L doses of troglitazone reduced TGF-β₁ release by 50%. A very similar reduction was observed in SMA-560 and U87-MG glioma cells. Both cell lines released TGF-β₁ at relevant doses (SMA-560, 2.37 ± 0.11 ng/mL; U87-MG, 2.30 ± 0.02 ng/mL), whereas 8.32 and 8.51 μmol/L doses of troglitazone, respectively, reduced TGF-β₁ levels by 50%, thus identifying troglitazone as a potent inhibitor of TGF-β₁ release at low micromolar concentrations (Fig. 6B). Using 10 μmol/L concentrations, the remaining test compounds also reduced TGF-β₁ release in all cell lines tested, but significantly less efficient than troglitazone (Fig. 6C).

The thiazolidinediones rosiglitazone, pioglitazone, ciglitazone, and troglitazone decreased glioma cell proliferation in a dose-dependent manner, whereas troglitazone showed a superior potency in all cell lines tested (Fig. 1A and B). The effects of troglitazone on the cell cycle distribution have intensively been studied in various cancer cell lines (17–21). We confirmed that troglitazone causes a G₁ arrest, which has shown to be associated with altered expression levels of genes critically involved in the G₁-S cell cycle progression (17–21, 30). For example, troglitazone has shown to increase the expression of the DNA damage inducible gene 45 (Gadd45) as well as the cyclin-dependent kinase inhibitors p21^CIP1/WAF1 and p27^KIP and down-regulates the expression of the cyclin-dependent kinases Cdk2, Cdk4, and Cdk6 and the DNA polymerase δ auxiliary protein (Pcna), a nuclear protein whose appearance correlates with the proliferative state of the cell. By quantitative real-time PCR analysis, we confirmed that troglitazone (IC₅₀, IC₉₀) significantly reduced the expression of Pcna, Cdk2, Cdk4, and Cdk6, and increased the Gadd45 expression in F98 glioma cells in a dose-dependent manner (data not shown), suggesting that the troglitazone-induced G₁ arrest identified in F98 glioma and other neoplastic cells share common mechanisms. However, the effects of troglitazone on the F98 glioma cell transcriptome are by far not restricted to the genes described. A preliminary microarray analysis (Affymetrix REA230A) revealed that the expression of 4.0% (643 of 15,923) of all transcripts was at least 2-fold increased, whereas the expression of 7.9% (1,258 of 15,923) of all transcripts was decreased by 50% or more after troglitazone treatment (IC₉₀, 48 h), indicating that troglitazone has extensive pleitropic effects on the glioma cell transcriptome.

The primary mechanism(s) by which troglitazone modulates the expression of genes involved in glioma cell proliferation and cell cycle distribution remain elusive. The PPAR family of ligand-activated transcription factors consists of three subtypes (PPAR-γ, PPAR-α, and PPAR-δ), which form heterodimers with the nuclear retinoid X receptor and then bind to PPAR response elements in the regulatory regions of target genes. Although pioglitazone and rosiglitazone activate both PPAR-γ and PPAR-α, respectively, troglitazone has shown to be PPAR-γ selective (31). Several lines of evidence support that the inhibitory effects of thiazolidinediones on tumor cell proliferation are not PPAR-γ mediated. As confirmed in our study, the sensitivity of cancer cells to thiazolidinedione-mediated growth inhibition does not correlate with the level of PPAR-γ expression (Figs. 1 and 2) and there exists a three orders of magnitude discrepancy between the concentration required to produce antitumor effects and that to improve insulin resistance via PPAR-γ activation (32). Using PPAR-γ–positive and PPAR-γ–deficient mouse embryonic stem cells, it has been shown that troglitazone inhibited the growth of tumors formed by injection of PPAR-γ +/+ and PPAR-γ −/− embryonic stem cells to the same extent, indicating that PPAR-γ is not essential for the antiproliferative effects of troglitazone (33). Moreover, troglitazone derivatives that are unable to activate PPAR-γ suppress cancer cell proliferation similar to troglitazone, giving further evidence that the antiproliferative effects of troglitazone are PPAR-γ independent (32).

Using the rat organotypic glioma invasion model, we show that micromolar doses of troglitazone are capable to block glioma progression (Fig. 3). This intriguing antiglioma property of troglitazone seems to be only partially based on its antiproliferative effects, which have been suggested to be moderate (32). In subsequent analyses, we identified troglitazone as a potent inhibitor of glioma cell migration and diffuse brain invasion, which occurs in a PPAR-γ–independent manner (Figs. 3 and 4). The antimigratory effects of troglitazone could be mimicked by inhibition of TGF-β signaling, which has shown to be intimately involved in glioma cell migration, suggesting both mechanisms to be interlinked (Fig. 4C). Several studies using surgically resected glioma tissues revealed an intriguing correlation between tumor grade and the expression of TGF-β_1-3 and their receptors I and II. High-grade gliomas express high levels of TGF-βRI, TGF-βRII, and TGF-β_1-3, whereas the expression levels of these molecules have shown to be weak in low-grade gliomas and normal brain tissue (34–36). The inverse correlation between TGF-β_1-3, TGF-βRI, and TGF-βRII expression and patient's survival support its pivotal role in malignant glioma progression. The role of TGF-β as molecular target for glioma therapy has been facilitated by numerous studies demonstrating that TGF-β levels determine glioma cell motility. Exogenously added TGF-β₁ and TGF-β₂ elicit a strong stimulation of migration in a variety of glioma cells (23–26), whereas Tgf-β₁ and Tgf-β₂ gene silencing has shown to reduce glioma cell motility and invasiveness (27). We identified troglitazone to significantly down-regulate the expression of Tgf-β_1-3 and their receptors (Tgf-βrI, Tgf-βrII), whereas expression of Fkbp1a, a protein that inhibits TGF-βRI activity (29), remained unchanged (Fig. 5). Using a polyclonal antibody directed against all three TGF-β isoforms, we show that troglitazone substantially decreased TGF-β release (Fig. 6A). An absolute quantification of TGF-β₁ protein levels revealed that rat F98, mouse SMA-560, and human U87-MG glioma cells release TGF-β₁ at relevant doses, suggesting that glioma cells promote migration by an autocrine signaling pathway. A dose-dependent analysis indicated that troglitazone significantly inhibits TGF-β₁ release already at low micromolar doses in all three glioma cell lines applied, with a narrow range of concentrations producing a 10% to 90% inhibition (Fig. 6B). In all, these results are in line with the propensity of troglitazone to counteract TGF-β–mediated glioma cell motility and invasiveness into brain parenchyma (Figs. 3 and 4).

Several studies are consistent with our finding that troglitazone counteracts TGF-β signaling. For example, TGF-β has shown to increase the expression of the matrix metalloproteinases MMP-2 and MMP-9, thus promoting glioma cell migration (3, 37), whereas troglitazone has shown to decrease MMP-2 and MMP-9 protein levels (38). TGF-β has shown to induce the expression of the caspase-8 inhibitor FLICE inhibitory protein, which plays a critical role as an endogenous modulator of apoptosis (39, 40). In contrast, troglitazone has shown to cause a pronounced reduction of FLICE inhibitory protein levels in glioma cells, thus sentisizing glioma cells to tumor necrosis factor–related apoptosis-inducing ligand–induced apoptosis (41). Moreover, troglitazone significantly counteracts hepatocellular carcinoma, thyroid cancer, and prostate cancer progression in xenograft mouse models (18, 42, 43). Especially the present data concerning prostate cancer and glioma treatment seem to share similarities. Very similar to glioma cells, TGF-β₁ promotes the motility of prostate cancer cells (44), whereas elevated TGF-β expression is associated with poor clinical outcome (45). A phase II clinical study in prostate cancer patients revealed a prolonged stabilization of prostate-specific antigen and long periods of stable disease in patients treated with troglitazone (46). In reference to these findings, one might speculate that numerous effects of troglitazone described thus far are due to its ability to counteract TGF-β signaling.

Because TGF-β antagonism is considered as a therapeutic strategy, including the development of antisense regimens, inhibition of pro–TGF-β processing, scavenging of TGF-β by the TGF-β–binding protein decorin, and blocking of TGF-β receptor I kinase activity (4), our results suggest troglitazone as a candidate drug for adjuvant glioma therapy. Pharmacokinetic studies in healthy individuals revealed that, after p.o. application of 400 mg of troglitazone, serum levels of 5.42 mg/L (12.3 μmol/L) can be achieved (47). At these doses, troglitazone displayed only marginal effects on glioma cell proliferation (Fig. 1A), but effectively inhibited TGF-β₁ release in vitro (Fig. 6B and C). With regard to the pharmacokinetic data, one might speculate that only antimigratory but not antiproliferative doses of troglitazone are clinically achievable. It has not been established whether troglitazone penetrates the blood-brain barrier in humans; however, rat studies provide evidence that troglitazone is actively incorporated by the bidirectional transporter Oatp14 (Slco1c1) expressed in brain capillary endothelial cells, which is likely to provide troglitazone homeostasis (48). We show that the antimigratory effects of troglitazone are PPAR-γ independent (Fig. 4C). Thus, the yet unknown protein/proteins mediating the troglitazone-induced inhibition of brain invasion and/or glioma proliferation remain to be identified and may represent future targets for structure-relationship studies. Moreover, already established antiproliferative and PPAR-γ–inactive troglitazone derivatives (32) may retain their propensity to counteract TGF-β–dependent glioma cell migration and might be further developed to minimize potential PPAR-γ–mediated side effects in glioma patients. Because local invasion of neoplastic cells into the surrounding brain is perhaps the most important aspect on the biology of gliomas that precludes successful treatment, troglitazone and its derivatives may be considered for the adjuvant therapy of glioma and other highly migratory tumor entities.