Nuclear Factor of Activated T Cells-dependent Down-regulation of the Transcription Factor Glioma-associated Protein 1 (GLI1) Underlies the Growth Inhibitory Properties of Arachidonic Acid*

Numerous reports have demonstrated a tumor inhibitory effect of polyunsaturated fatty acids (PUFAs). However, the molecular mechanisms modulating this phenomenon are in part poorly understood. Here, we provide evidence of a novel antitumoral mechanism of the PUFA arachidonic acid (AA). In vivo and in vitro experiments showed that AA treatment decreased tumor growth and metastasis and increased apoptosis. Molecular analysis of this effect showed significantly reduced expression of a subset of antiapoptotic proteins, including BCL2, BFL1/A1, and 4-1BB, in AA-treated cells. We demonstrated that down-regulation of the transcription factor glioma-associated protein 1 (GLI1) in AA-treated cells is the underlying mechanism controlling BCL2, BFL1/A1, and 4-1BB expression. Using luciferase reporters, chromatin immunoprecipitation, and expression studies, we found that GLI1 binds to the promoter of these antiapoptotic molecules and regulates their expression and promoter activity. We provide evidence that AA-induced apoptosis and down-regulation of antiapoptotic genes can be inhibited by overexpressing GLI1 in AA-sensitive cells. Conversely, inhibition of GLI1 mimics AA treatments, leading to decreased tumor growth, cell viability, and expression of antiapoptotic molecules. Further characterization showed that AA represses GLI1 expression by stimulating nuclear translocation of NFATc1, which then binds the GLI1 promoter and represses its transcription. AA was shown to increase reactive oxygen species. Treatment with antioxidants impaired the AA-induced apoptosis and down-regulation of GLI1 and NFATc1 activation, indicating that NFATc1 activation and GLI1 repression require the generation of reactive oxygen species. Collectively, these results define a novel mechanism underlying AA antitumoral functions that may serve as a foundation for future PUFA-based therapeutic approaches.

oxygen species. Treatment with antioxidants impaired the AAinduced apoptosis and down-regulation of GLI1 and NFATc1 activation, indicating that NFATc1 activation and GLI1 repression require the generation of reactive oxygen species. Collectively, these results define a novel mechanism underlying AA antitumoral functions that may serve as a foundation for future PUFA-based therapeutic approaches.
Epidemiological and experimental studies have demonstrated a role for polyunsaturated fatty acids (PUFAs) 3 in different aspects of tumor development (1)(2)(3). PUFAs have been reported to have both pro-and antitumoral properties (4). In particular, the roles of dietary -3 and -6 PUFAs in cancer prevention have been explored (5,6). Both in vitro and in vivo studies as well as clinical studies have implicated different fatty acids in tumor development and progression. Increased cellular levels of PUFAs have been shown to inhibit tumor growth (7). In vivo intratumoral injection of PUFAs induces tumor regression (8 -10) and improves survival (11). Moreover, clinical studies have shown that intratumoral injection of PUFAs in patients with intractable gliomas improves survival and induces partial tumor regression without causing side effects (12,13).
PUFAs can also act as cytotoxic molecules, activating different cell signaling pathways that modulate proliferation, cell death, and migration of tumor cells (14,15). The cytotoxic effects of PUFAs have been suggested to occur in part due to alterations in reactive oxygen species, changes in cell membrane fluidity or conversion of PUFAs to highly bioactive metabolites such as prostaglandins and leukotrienes, and/or altered expression of genes that regulate apoptotic cell death (6, 16 -22). However, the molecular pathways by which PUFAs regulate cell death in cancer cells are poorly understood.
Here, using the PUFA arachidonic acid (AA) in vitro and in vivo, we define a novel mechanism underlying the tumor inhibitory function of this fatty acid. AA is a member of the -6 family, derived by successive elongation and desaturation from the essential fatty acid linoleic acid (LA), the major PUFA of cell membranes present at the sn-2 position of glycerophospholipids. Direct cell supplementation with AA has cytotoxic activity and decreases cell viability in different tumor cell types, including cervical, hepatic, and pancreatic (23)(24)(25). In this study, we identify GLI1, a known oncogenic transcription factor (26), as a target of AA signaling in cancer cells. GLI1 is a member of the GLI family of C 2 H 2 zinc finger transcription factors. It plays a crucial role in carcinogenesis, contributing to the development of tumors in the prostate, lung, pancreas, and breast (27,28). We show that AA down-regulates GLI1 expression through the modulation of its promoter in an NFATc1-dependent manner. Finally, the blockade of GLI1 results in decreased expression of prosurvival target genes (BCL2, BFL1/A1, and 4-1BB), leading to increased cell death both in vivo and in vitro. Together, these findings define a novel signaling cascade controlling cancer cell survival and help to delineate the mechanisms underlying AA antitumoral properties.

Experimental Procedures
Cell Lines and Culture Conditions-LM3 cells were kindly provided by Dr. Bal de Kier Joffé from the Institute of Oncology Angel H. Roffo (University of Buenos Aires, Buenos Aires, Argentina). PANC1 and MCF7 cell lines were obtained from ATCC and were cultured as reported previously (29 -31). L3.6 cells were kindly provided by Dr. Fidler (M. D. Anderson Cancer Center, Houston, TX). For fatty acid treatments, AA, palmitoleic acid (PA), and LA (Nu-Chek Prep, Inc., Elysian, MN) were dissolved in absolute ethanol to an initial dilution of 100 mg/ml. The dilution was prepared under a nitrogen atmosphere and then stored at Ϫ80°C. For all fatty acid treatments, we used 5% fetal bovine serum (FBS) in the cell culture media as a fatty acid carrier. To avoid any issues with the use of FBS, we ran parallel experiments using 0.5 and 1% BSA as a carrier. Both protocols show similar biological effects (data not shown). For some of the experiments, a general caspase (CASP) inhibitor, Q-VD-OPH (SM Biochemicals LLC, Yorba Linda, CA), was used at a final concentration of 20 M and added at the beginning of the cell culture process. N-Acetylcysteine (NAC) and L-glutathione (both from Sigma-Aldrich) were used at final concentration of 500 M. Cyclosporin A (CsA) and ionomycin (both from Sigma-Aldrich) were used for some of the expression and reporter assays at a concentration of 1 and 0.5 M, respectively.
Cell Viability Assay-Cell viability was determined by a colorimetric and fluorometric technique using resazurin (Sigma-Aldrich). In each assay, 5 ϫ 10 4 PANC1 cells/well or 7.5 ϫ 10 4 MCF7 cells/well were plated in 96-well plates and cultured in their respective media. The following day cells were treated with different concentrations of AA (5, 20, 40,  In Vivo and in Vitro Apoptosis Assays-In the in vivo studies, apoptosis was determined by the terminal deoxyribonucleotidyltransferase-mediated dUTP nick-end labeling (TUNEL) method in which paraffin-embedded tumor tissues were used for the specific detection and quantification of apoptotic cells within a cell population using the DeadEnd TM Fluorometric TUNEL System (Promega, Madison, WI) in accordance with the manufacturer's protocol. Images were obtained using an Axio Observer epifluorescence microscope (Carl Zeiss AG, Oberkochen, Germany). For the in vitro studies, apoptosis was determined by DNA fluorescent labeling with Hoechst 33258 (32). Cells (2.5 ϫ 10 5 /well) were plated in 6-well plates and then incubated for 24 h in their respective media. After 24 h of AA treatment or 48 h with GANT61 or Glabrescione B, the medium was removed, and the cells were washed once with PBS and fixed with 500 l of PBS solution and 500 l of 3:1 methanol:glacial acetic acid for 5 min. Five hundred microliters of 3:1 methanol:glacial acetic acid was then added for 10 min. The medium was removed, and 1 ml of PBS plus Hoechst 33258 dye was added at a final concentration of 5 g/ml and then incubated for 10 min at 37°C before being examined under a fluorescence microscope (Axioskop epifluorescence microscope, Carl Zeiss AG). In addition, the Apo-ONE homogeneous CASP3/7 assay (Promega) was performed. Cells were plated at 5 ϫ 10 4 cells/well in 96-well plates. After treatment, 100 l of Apo-ONE CASP3/7 reagent was added to each well containing 100 l of cells in culture and incubated at room temperature for 6 h. The fluorescent signal was measured at 485 nm excitation/ 527 nm emission wavelength.
Expression and shRNA Constructs, siRNAs, and Transfection-Expression vectors for NFATc1 and the shRNA constructs were described previously in Elsawa et al. (33) and Köenig et al. (34). GLI1 expression construct and siRNA targeting GLI1 were described previously (35). The GLI-luciferase reporter containing eight consecutive GLI binding sites upstream of the luciferase gene (GLI-LUC) was kindly provided by Dr. Chi-chung Hui (University of Toronto, Toronto, Ontario, Canada). Human BFL1/A1, 4-1BB, and GLI1 promoter reporter constructs were a gift from Drs. Gelinas (University of Medicine and Dentistry of New Jersey, Piscataway, NJ), Kang (College of Pharmacy, Seoul National University, Seoul, South Korea), and Aberger (University of Salzburg, Salzburg, Austria), respectively. Human BCL2 promoter containing the Ϫ1428 to Ϫ799 bp upstream of exon1 was cloned using standard DNA recombinant protocols. Mutations of GLI binding sites in wild type promoters were performed as follows. For BCL2 promoter, the GLI1 canonical binding sequence from Ϫ1125 to Ϫ1123 bp was changed from CAC to AAA using the QuikChange II XL site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA) and the following primers: GGCA-AAGGTGGAGACCTTTAGGAGAAAAAAACCCCAGCG-TTAGGACGGTGGGCC (sense) and GGCCCACCGTCCTA-ACGCTGGGGTTTTTTTTCTCCTAAAGGTCTCCACCT-TTGCC (antisense). For BFL1/A1 promoter mutants, two core sequences of the GLI binding sites spanning from Ϫ5 to Ϫ3 bp and ϩ27 to ϩ29 bp were changed from CAC to AAA using the Q5 site-directed kit (New England Biolabs, Ipswich, MA) and the following set of primers: TTCACATTCCAAACAGCAAT-CTACAAG (sense) and ATGGTACAACCCCTGTGAA-AATG (antisense) and CTAATTTCTCAAAATCCTGCATT-TAAGAC (sense) and AATCCTTGTAGATTGCTGTTTG (antisense), respectively. For the 4-1BB promoter, the core sequence of the GLI binding site that goes from Ϫ12 to Ϫ10 bp was changed from CAC to AAA using the primer CCCGCCC-CTGGCTGAGTaaaCGCACTCCTG (sense), the site spanning 510 -512 bp was changed from CCC to AAA using the primer TGGTACTGCAAAGTACCTATGAAAGTCAAAAAAAGT-TCTTGCTATTGTTATTTCTGC, and the site from 806 to 808 bp was changed from CAG to AAA using the QuikChange multisite-directed mutagenesis kit (Agilent) and the primer CTTTTGCAGAGTGACATTTGTGAGACAAACTAATTT-GATTAAAATTCTCTTGGAATC (sense). In all cases, mutagenesis was performed according to the manufacturer's protocol, and changes were verified by sequencing.
PANC1 and MCF7 cells were transfected with X-treme-GENE HP (Roche Applied Sciences) according to the manufacturer's instructions. Twenty-four hours prior to transfection, 5 ϫ 10 4 and 7 ϫ 10 5 cells were seeded per well in 12-well plates and 100-mm dishes, respectively. For luciferase assays, 2 g of GLI1, GLI-LUC, BCL2, BFL1/A1, and 4-1BB promoter reporter vectors were used. For transfection of multiple constructs, 10 g of DNA were used, consisting of 1 g of promoter reporters and 9 g of the expression vectors GLI1 and NFATc1. In GLI1 knockdown experiments, 10 g of shGLI1 or control vector were transfected for 72 h. In the knockdown experiments where GLI1 and NFATc1 siRNAs (FlexiTube siRNA, Qiagen, Valencia, CA) were used along with a non-targeting siRNA negative control (Allstars Negative Control siRNA, Qiagen), cells were transfected using Oligofectamine (Invitrogen) to a final concentration of 100 nM siRNA and harvested 48 h after transfection. In rescue experiments, PANC1 cells were transfected with 10 g of GLI1 expression construct or empty vector. Twentyfour hours post-transfection, cells were treated with AA at 60 g/ml for 6 h. Then RNA was extracted to determine the expression of BCL2, BFL1/A1, and 4-1BB.
Luciferase Reporter Assay-Cells were grown and transfected as indicated above. For the assays, cells were plated in triplicate at 5 ϫ 10 4 cells/well in 12-well plates in the corresponding medium. After 24 h of transfection, cells were treated with AA, and then samples were harvested and prepared for luciferase assays in accordance with the manufacturer's protocol (Promega). The total protein of samples in each well was quantitated using the Bio-Rad protein assay, and luciferase readouts were normalized to protein content. Relative luciferase activity represents luciferase/protein concentration readouts normalized to control group within each experiment (36).
Amplification was performed using the C1000 Thermal Cycler (Bio-Rad). Each RNA level was normalized by comparison with the corresponding housekeeping RNA level in the same sample. The results were calculated following the 2⌬Ct method.
Chromatin Immunoprecipitation (ChIP) Assay-ChIP was conducted following a modified version of the Magna ChIP kit protocol (EMD Millipore, Bedford, MA). Briefly, PANC1 cells (7 ϫ 10 6 ) were cross-linked with 1% formaldehyde directly into the medium for 10 min at room temperature. The cells were then washed and scraped with PBS, collected by centrifugation at 800 ϫ g for 5 min at 4°C, resuspended in cell lysis buffer, and incubated on ice for 15 min. The pellet was then resuspended in nuclear lysis buffer and sheared to fragment DNA to ϳ700 bp. Samples were then immunoprecipitated using a GLI1 antibody (Novus Biologicals, Littleton, CO), NFATc1 antibody (Santa Cruz Biotechnology, Santa Cruz, CA), normal rabbit IgG (EMD Millipore), or normal mouse IgG (EMD Millipore) overnight at 4°C on a rotating wheel. Following immunoprecipitation, cross-links were removed, and immunoprecipitated DNA was purified using spin columns and subsequently amplified by PCR. PCR was performed using a primer set to amplify a 350-bp region of the GLI1 promoter containing a potential NFATc1 binding site (sense, AGCAGTATAGGGTCCCTCAAG; antisense, CGCGAGAAGCGCAAACT) and primer sets to amplify BCL2, BFL1/A1, and 4-1BB promoters containing potential GLI1 binding sites as follows: BCL2: sense, ACACACGTCTG-CGAGTGTGAATGT; antisense, TCCCTCTGTCCCTAAC-ACCTTT; BFL1/A1: sense, GAGTATCTGTTGGGTGTATA-CCTATG; antisense, GATGATACATGGAGGCTGGTG-GAA; 4-1BB: sense, ATCTGTGACACATCCTGACAGTAGA; antisense, ACATGTGGAAAGACCTCCCTGTTC; sense; GAACAGGGAGGTCTTTCCACATGT; antisense, TGATG-AAATCTGGCACAGGTATGAT. PCR products were visualized using a 2% agarose gel.
For the transient ChIP assay, PANC1 cells (1 ϫ 10 6 ) were grown overnight in 100-mm dishes and then transfected with 3 g of the BCL2, 4-1BB, or BFL1/A1 promoter reporter vectors with WT or mutant GLI1 biding sites using X-tremeGENE HP. Twenty-four hours after transfection, cells were cross-linked with formaldehyde and harvested, and chromatin immunoprecipitations were performed as described above. Antibodies used in the procedure include anti-GLI1 (Novus) and anti-rabbit IgG (EMD Millipore). The resulting DNA was analyzed by real time PCRs with the same set of primers described previously. Quantitative SYBR PCR was performed in triplicate for each sample using the C1000 Thermal Cycler. Results are represented as -fold of enrichment where each antibody signal was relative to its respective input and then normalized to the nonimmune IgG control signal.
Immunoblotting-Western blotting was performed as described previously (28,36). 1 ϫ 10 6 PANC1 and MCF7 cells were plated in standard medium with 10% FBS and used 24 h later for treatment or transfection experiments. Cells were lysed 48 h after transfection. GLI1 and NFATc1 antibodies were obtained from Cell Signaling Technology (Danvers, MA). BCL2 antibody was purchased from Bethyl Laboratories (Montgomery, TX), and BFL1/A1 and 41BB antibodies were obtained from Abcam (Cambridge, MA). Anti-␣-tubulin was obtained from Sigma-Aldrich. Peroxidase-conjugated secondary antibodies were used, and immunoreactive proteins were detected by chemiluminescence (GE Healthcare).
Fluorescence Microscopy-PANC1 cells were transfected for 48 h with an NFATc1-GFP expression construct (Addgene, Cambridge, MA). Cells were treated with NAC, ionomycin, CsA, or AA as described above and in figure legends. PANC1 cells were then fixed in 4% formaldehyde and mounted in Prolong with DAPI (Invitrogen). Microscopy was performed using an Olympus AX70 equipped with a Hamamatsu C4742-95 camera. Images were captured using MetaMorph (version 7.3.2; Universal Imaging Corp.) and analyzed with ImageJ.
In Vivo Studies-BALB/c mice were housed in groups of two to four in polycarbonate cages. Food and water were provided ad libitum. Animals were maintained at a constant temperature of 20°C and subjected to a 12-h light-12-h darkness cycle. Animal studies were conducted in accordance with the guidelines of the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and all the procedures were approved by the Institutional Committee for the Care and Use of Laboratory Animals of the School of Medicine, University of Cordoba, Argentina.
Three-month-old mice were inoculated subcutaneously in the left flank with 1 ϫ 10 6 viable LM3 murine mammary adenocarcinoma cells with lung moderate metastatic capability (31) in 0.02 ml serum-free minimum Eagle's medium. When tumors became palpable, mice were randomized in treatment or control groups. The treatment group (n ϭ 10) received pure AA at a dose of 0.70 mg AA/g of body weight injected in the tumor bed every 7 days over 21 days. The control group (n ϭ 7) received injections of normal saline buffer in the tumor bed. The experimental period of 21 days was determined according to the time required for the generation of metastasis and before the tumor reaches the maximum standard size allowed. Tumor volume was measured with digital slide calipers at the beginning of treatment and every 7 days for 3 weeks. The tumor volume was calculated from the transverse (width) and longitudinal (length) diameters in millimeters of the tumor using the formula: Volume ϭ Length ϫ Width 2 /2 (37). Twenty-one days after treatment, the animals were sacrificed. The tumors and lungs were rapidly removed, and samples preserved in formalin for subsequent H&E histopathological analysis and immunohistochemical studies.
We also evaluated the potential acute toxicity after AA administration through examination of general conditions, body weight, food intake, and mortality. Major organs were examined macroscopically, and lung, liver, and spleen were fixed and stained with H&E for histopathological examination (38). We determined that the dose used in this study was well tolerated by the experimental mice and had no toxicity. All mice survived until the end of the experimental study, and no adverse effects were observed. Mice displayed normal behavior, normal feeding, normal energy, and shining hair. Also, AAtreated animals as well as controls showed no weight loss, no salivation or vomiting, no mouth or nose dryness or edema, no eye secretion or running nose, and no flare or ulceration in the skin. At necropsy, no alteration or significant difference was recorded in internal organ evaluation. Moreover, histopathological analysis of the different tissue showed no significant changes compared with the control group.
For the GANT61 experiment, tumors were inoculated following the above protocol. When tumors were palpable, mice were randomized and treated daily with GANT61 at a dose of 50 mg/kg of body weight. The control group received injections of normal saline buffer in the tumor bed. The experimental period of 21 days was determined according to the time required for the generation of metastasis and before the tumor reaches the maximum standard size allowed. Tumor volume was measured with digital slide calipers at the beginning of treatment and every 7 days for 3 weeks.
Ferric Xylenol Orange Peroxide Assay-To determine the levels of hydroperoxides, we measured the dye xylenol orange and Fe 3ϩ generated as an iron-xylenol orange complex in the visible absorbance range. For the assay, 5 ϫ 10 4 cells/well were plated in a 96-well plate, and 24 h later cells were treated with AA for 24 h as we detailed above. Cells were lysed with 10 l of 1% (v/v) SDS and mixed with the appropriate volume of reagents to give final concentrations of 25 mM H 2 SO 4 , 100 mM xylenol orange, and 250 mM ferrous ammonium sulfate. After 30 min in the dark, the absorbance was read at 560 nm with xylenol orange as a blank with a GloMax multimode microplate reader (Promega) (39).
␥-Glutamyl Transpeptidase Activity-␥-Glutamyl transpeptidase activity was measured based on the principle that ␥-glutamyl-p-nitroanilide acts as a donor and glycylglycine acts as an acceptor substrate (40). Five thousands cells/well were plated in a 96-well plate. After attachment, the cells were treated with AA for 24 h as detailed above. A kinetic modified Szasz method was performed according to the kit instructions of ␥-G-test cinética AA (Wiener Laboratorios, Rosario, Santa Fe, Argen-tina). The kinetics of the reaction was determined measuring the absorbance at 450 nm at time 0 and after 10 and 20 min with a GloMax multimode microplate reader. Total protein was determined by Bradford assay. The ␥-glutamyl transpeptidase specific activity was presented in ␥-glutamyl transpeptidase international units (IU) ϭ ⌬A/min ϫ 1.158 IU of GGPT/mg of proteins (41).
Statistical Analysis-Analysis of variance models were used to evaluate differences among the experimental groups fol-

AA Treatment Induces Apoptosis in Vitro-Using in vitro
models, we demonstrated that AA decreased cell viability in both LM3 and MCF7 mammary cancer cell lines and in pancreatic cancer cell lines PANC1 and L3.6. Cancer cell viability was reduced in a dose-dependent manner in all cell lines tested (Fig.  1A). Further analysis shows that AA induced apoptosis after 24 h of treatment as measured by nuclear fragmentation (Fig. 1, B and C) with a significant increase (p Ͻ 0.05) at a concentration of 60 g/ml in PANC1 and MFC7 cell lines and 40 g/ml in LM3 and L3.6 cell lines relative to controls, concentrations found to have significant effects in the previous cell viability experiments. In addition, in PANC1 and MCF7 cell lines, AA treatment elicited a significant increase in CASP3/7 activity after 6 (left panel) and 24 h (right panel) of treatment (Fig. 1D). Taken together, these data support a role for AA as an anticancer agent by affecting cell growth and inducing apoptosis.
AA Treatment Decreases the Expression of Antiapoptotic Molecules BCL2, BFL1/A1, and 4-1BB-To define the mechanisms behind AA-induced apoptosis, we examined by q-PCR the expression of a set of genes with a known role in regulation of cell growth and apoptosis. We identified the antiapoptotic molecules BCL2, BFL1/A1, and 4-1BB (42,43) as down-regulated in AA-treated cells. As shown in Fig. 2A, AA treatment of MCF7 and PANC1 cells decreased BCL2, BFL1/A1, and 4-1BB mRNA levels compared with controls. A significant difference (p Ͻ 0.05) was observed in all genes with the exception of the 4-1BB antiapoptotic gene in MCF7 cells, the levels of which showed a tendency to decrease. Immunoblot analysis shows a similar effect of AA on the protein expression of these antiapoptotic molecules (Fig. 2B). This effect was specific for these molecules as the expression of other antiapoptotic proteins, MCL1 and BCL-xL, was not affected by the AA treatment (Fig.  2C). Next we determined whether these effects of AA were also shared by other fatty acids with known antitumoral activity, PA and LA. Similar to AA, these fatty acids decrease viability in PANC1 cells (Fig. 2D) but do not significantly change the expression of the antiapoptotic molecules except BFL1/A1, which is down-regulated by the treatment with PA and LA (Fig. 2E).
BCL2, BFL1/A1, and 4-1BB Are Direct Targets of GLI1-Bioinformatics analysis of the promoters of BCL2, BFL1/A1, and 4-1BB identified the transcription factor GLI1 as a candidate regulator of the expression of these molecules (Fig. 3A). To define the role of GLI1 in antiapoptotic BCL2, BFL1/A1, and 4-1BB promoter genes, we utilize a combination of reporter, expression, and chromatin assays. Using two independent shRNA constructs targeting GLI1, we showed that knockdown of GLI1 diminished BCL2, BFL1/A1, and 4-1BB mRNA expression in PANC1 cells (Fig. 3B). Similar results were obtained using an siRNA targeting GLI1 (data not shown). A ChIP assay confirmed the binding of endogenous GLI1 to a region located Ϫ1234 to Ϫ1025 bp upstream of the first exon for BCL2, Ϫ121 to ϩ226 bp from the first exon for BFL1/A1, and two regions for 4-1BB located Ϫ192 to ϩ341 bp (a) and ϩ317 to ϩ873 bp (b) from the first exon (Fig. 3C). In addition, overexpression of GLI1 led to an increase in the promoter activity of BCL2, BFL1/ A1, and 4-1BB in reporter assays (Fig. 3D). Finally, to demonstrate the specificity of the GLI1 binding sites identified for each gene promoter, we utilized a transient ChIP (44) using WT promoter construct and mutant versions of these vectors lacking the GLI1 binding sites. In each case, we found a decreased binding of GLI1 to the mutant promoters compared with their WT sequence (Fig. 3E). These results confirm a previously known GLI1 target gene (BCL2) (45) and demonstrate that BFL1/A1 and 4-1BB are novel direct targets of GLI1 transcription factor.
AA Diminishes GLI1 Expression by Inducing the Repressor Activity of NFATc1-To determine whether GLI1 plays a role in the AA-induced down-regulation of antiapoptotic genes, we investigated the effect of AA treatment on GLI1 expression and activity. We observed that AA treatment decreased the mRNA expression of GLI1 in PANC1 and MCF7 cells (Fig. 4A). This effect was specific to AA as LA and PA did not repress GLI1 mRNA expression (data not shown). GLI1 protein expression was also reduced by AA as shown by Western blot analysis in PANC1 cells at 6 and 24 h post-treatment with AA (Fig. 4B). Altered GLI1 expression was accompanied by a decrease in GLI1-associated transcriptional activity as demonstrated by lower activity of the GLI luciferase reporter (GLI-LUC) in AAtreated cells (Fig. 4C). To rule out the possibility that the effect of AA on GLI1 activity was a consequence of cells undergoing apoptosis, we performed the AA treatment in the presence of a CASP inhibitor (Q-VD-OPH). AA was able to decrease GLI1 activity even in the presence of Q-VD-OPH as shown in Fig. 4C. As a control for these treatments, we measured the levels of activated CASP3/7 and cell viability, which were altered by AA treatment but restored to basal levels by the CASP inhibitor (Fig. 4D). Next, we sought to determine whether GLI1 overexpression was able to block AA induction of cell death. As shown in Fig. 4E, GLI1 overexpression in PANC1 cells was able to reverse AA-induced CASP3/7 activation. Similarly, overexpression of GLI1 restored the expression of BCL2, BFL1/A1, and 4-1BB in cells treated with AA (Fig. 4F). Finally, we demonstrated that pharmacological or genetic inhibition of GLI1 mimics the effect of AA on cell growth. Knockdown of GLI1 led to an increase in the number of apoptotic cells in both PANC1 and MCF7 lines (Fig. 4G). Treatment of these cells with GLI1 inhibitors GANT61 (Fig. 4H) and Glabrescione B (Fig. 4I) led to decreased cell viability. Collectively, these results suggest that down-regulation of GLI1 is required for AA induction of cell death. The above results demonstrated that AA modulates the expression and activity of GLI1 in cancer cells. Further analysis showed that AA decreased the transcriptional activity of the GLI1 promoter in PANC1 (left) and MCF7 (center) starting after 30 min and maintained up to 6 h (Fig. 5A). Even upon CASP inhibition, AA treatment decreased GLI1 promoter activity (Fig. 5B). Reporter studies using deletion constructs of this promoter defined an area located Ϫ200/Ϫ100 bp upstream of exon 1 as the AA-responsive element (Fig. 5C). In silico examination of transcription factor binding sites in this region  (EtOH). B, Western blot analysis shows that AA decreases GLI1 protein expression at 6 and 24 h post-treatment. C, relative changes in luciferase activity in PANC1 cancer cells transfected with a GLI-LUC reporter and subsequently treated with the CASP inhibitor Q-VD-OPH at 20 M or the control vehicle DMSO and co-treated with AA (60 g/ml) or control vehicle (EtOH) for 6 h. Bars and error bars represent means and S.E., respectively; n ϭ 3. *, p Ͻ 0.05 versus control group. D, apoptosis was measured by homogeneous CASP3/7 assay and viability was measured by resazurin assay in PANC1 cells treated with the CASP inhibitor Q-VD-OPH at 20 M. Cells were subsequently co-treated with AA (60 g/ml) or control vehicle for 6 h. Bars and error bars represent means and S.E., respectively; n ϭ 3. *, p Ͻ 0.05 versus control group. E, PANC1 cells were transfected with the GLI1 expression construct or the control vector. After 24 h, transfected cells were treated with AA at 60 g/ml or control vehicle (EtOH) for 6 h. A homogeneous CASP3/7 assay was performed to measure apoptosis. Bars and error bars represent means and S.E., respectively; n ϭ 3. *, p Ͻ 0.05 versus control vector. F, mRNA expression of BCL2, BFL1/A1, and 4-1BB in PANC1 and MCF7 cells transfected with GLI1 expression construct and control vector and treated with AA at a concentration of 60 g/ml. G, apoptosis was evaluated by fluorescence staining with Hoechst 33258 in PANC1 and MCF7 transfected with siRNA non-targeting (siNT) control and siGLI1. (TESS and Tfsitescan database programs) identified NFATc1 as a candidate mediator of AA regulation of the GLI1 promoter in cancer cells (Fig. 5D). Overexpression of NFATc1 indeed caused a dramatic decrease in GLI1 promoter activity in both PANC1 and MCF7 cells (Fig. 5E). Conversely, siRNA knockdown of NFATc1 increased the expression of GLI1 (Fig. 5F). We also found that AA requires NFATc1 to decrease GLI1 expression by showing that knockdown of NFATc1 impairs AA-induced reduction of GLI1 (Fig. 5G). Next, we showed that NFATc1 nuclear translocation was increased upon administration of AA (Fig. 6A), and NFATc1 binding to a region located Ϫ200/Ϫ100 bp upstream of the first exon of GLI1 was increased by treatment with this fatty acid (Fig. 6B). These studies indicate that AA suppresses GLI1 expression via NFATc1 binding to the GLI1 promoter.
NFATc1 activation is often mediated by the phosphatase calcineurin (46,47). Thus, we next investigated the involvement of calcineurin in the AA-initiated repression of GLI1 expression. We found that treatment with the calcineurin inhibitor CsA inhibited the nuclear localization of NFATc1 and rescued the AA-mediated down-regulation of GLI1 expression in PANC1 cells (Fig. 6C, right panel). Conversely, treatment with the calcium ionophore ionomycin, which is known to activate calcineurin (46,48), stimulated NFATc1 nuclear translocation FIGURE 5. AA represses GLI1 promoter activity in cancer cells. A, relative changes in luciferase activity in PANC1 (left) and MCF7 (right) cancer cells transfected with GLI1 promoter reporter vector and subsequently treated with AA (60 g/ml) or control vehicle (EtOH) for 30 min (Ј) and 6 h. Bars and error bars represent means and S.E., respectively; n ϭ 3. *, p Ͻ 0.05 versus the control group. B, relative changes in luciferase activity in PANC1 cancer cells transfected with a GLI1 promoter reporter and subsequently treated with the CASP inhibitor Q-VD-OPH at 20 M or the control vehicle DMSO and co-treated with AA (60 g/ml) or the control vehicle (EtOH) for 6 h. Bars and error bars represent means and S.E., respectively; n ϭ 3. *, p Ͻ 0.05 versus the control group. C, relative luciferase activity in PANC1 cells transfected with a series of truncated GLI1 promoter reporter constructs and then treated with AA or control (EtOH). *, p Ͻ 0.05 versus the control group. D, diagram showing AA-responsive region and NFATc1 binding site within the GLI1 promoter. E, relative changes in luciferase activity in PANC1 and MCF7 cancer cells co-transfected with the GLI1 promoter reporter and with the NFATc1 expression constructs or PCMV control vector. Bars and error bars represent means and S.E., respectively; n ϭ 3. *, p Ͻ 0.05 versus the control group. F, PANC1 cells transfected with an siRNA targeting NFATc1 or a non-targeting control (siNT) were examined for GLI1 and NFATc1 mRNA expression levels by real time RT-PCR. Bars and error bars represent means and S.E., respectively; n ϭ 3. *, p Ͻ 0.05 versus non-targeting control. G, GLI1 mRNA and NFATc1 expression levels were determined by q-PCR in PANC1 transfected with a non-targeting (siNT) control and siRNA targeting NFATc1 and then treated with AA (60 g/ml) or the control vehicle for 3 h. Bars and error bars represent means and S.E., respectively; n ϭ 3 expressed as relative percentage to control. *, p Ͻ 0.05 versus control group (EtOH). and lowered GLI1 expression (Fig. 6C). Finally, CsA treatment prevented the repression of GLI1 transcription by AA (Fig. 6D). These data suggest that a Ca 2ϩ -calcineurin axis mediates the suppression of GLI1 expression by AA.
Because intracellular Ca 2ϩ can be increased by reactive oxygen species, a common feature of cells treated with fatty acids, including AA (49 -51), we examined the role of oxidative stress in the regulation of GLI1. As shown in Fig. 7A, the levels of hydroperoxides and activity of ␥-glutamyl transpeptidase, two markers of oxidative stress, were elevated in PANC1 and MCF7 cells treated with AA. Treatment with antioxidants NAC and L-glutathione reversed the effect of AA on activation of CASP3 (Fig. 7B), GLI1 expression and promoter activity (Fig. 7, C and  D), and nuclear translocation of NFATc1 (Fig. 7D, right panel). Taken together, these results indicated that GLI1 is a novel target of the NFATc1 transcription factor, which acts as a mediator of the AA cytotoxic effect in cancer cells downstream of a reactive oxygen species-calcineurin axis.

AA Treatment Impairs Tumor Growth and Metastasis and
Induces Apoptosis in Vivo-Next we defined the biological significance of the above findings by testing the ability of AA to inhibit tumor growth in vivo using a syngeneic mouse mammary tumor model (31). A significant reduction in tumor volume was observed in animals injected with AA every 7 days in the tumor bed compared with the control group (p Ͻ 0.05) (212.13 Ϯ 13.4 versus 507.83 Ϯ 139.06 mm 3 ) at 21 days posttreatment (Fig. 8A). Further analysis showed that AA reduced the number of lung metastases. Sections stained with H&E showed that mice treated with AA have a significantly lower incidence of lung metastases. Although metastases were observed in 57.14% of saline-treated mice, AA treatment resulted in a significant reduction in the incidence of lung metastasis (11.11%) (p Ͻ 0.05) (Fig. 8B).
Characterization of this phenotype demonstrated an increase in apoptosis in AA-treated tumors. TUNEL staining of tumors treated with AA showed an increased number of apo- B, ChIP assay was performed in PANC1 lyses cells pretreated with AA (60 g/ml) or control vehicle (EtOH) for 6 h; NFATc1 antibody or IgG control antibody was used. PCR was performed using a specific set of primers indicated under "Experimental Procedures" for the Ϫ350 bp upstream GLI1 promoter region. IP, immunoprecipitation. C, PANC1 cells were treated for 3 h with 1 M CsA or 0.5 M ionomycin (Ion) for 15 min (Ј) or vehicle control (Con) and examined for GLI1 mRNA expression by q-PCR (upper panel). PANC1 cells transfected with NFATc1-GFP for 48 h were similarly treated with CsA, ionomycin, or vehicle and viewed by fluorescence microscopy (lower panel) to validate the effects of these treatments on NFATc1 localization. Green, NFATc1-GFP; blue, DAPI. D, PANC1 cells were treated for 1 h with CsA and then for 24 h with AA or control vehicle. RNA was isolated, and GLI1 mRNA expression was quantified by q-PCR. Bars and error bars represent means and S.E., respectively.
ptotic cells compared with control-treated mice (Fig. 8C). To confirm the effect of AA on the gene expression in vivo, we determined the expression of GLI1 and the antiapoptotic genes in LM3 tumors treated with or without AA. AA significantly decreased the mRNA expression of Gli1 (Fig. 8D) and Bcl2, Bfl1/a1, and 4-1bb genes (Fig. 8E) in LM3 tumors (n ϭ 5).
Next, we examined whether pharmacologic inhibition of GLI1 using GANT61 (52) would mimic the effect of AA in LM3 cells in vitro and in vivo. Indeed, treatment of these cells with this inhibitor decreased cell viability and expression of Bcl2, Bfl1/a1, and 4-1bb genes (Fig. 9A) as well as tumor growth (Fig.  9B). Taken together, these data support a role for AA as an anticancer agent by affecting cell growth and inducing apoptosis in vivo and in vitro.

Discussion
PUFAs can regulate cellular behavior in all stages of carcinogenesis by acting as cytotoxic compounds, although in some cases these lipids promote tumor growth (20, 53-60). Our data show that the PUFA AA treatment impairs tumor volume in vivo and induces cell death in vivo and in vitro. Moreover, AA treatment decreases tumor progression as demonstrated by the lower number of lung metastasis. Consistent with the results presented here, local injection of the -6 PUFA ␥-linolenic acid into a glioma tumor in a rat model has been shown to lead to FIGURE 7. Involvement of reactive oxygen species in AA effects on GLI1 expression. A, hydroperoxides and ␥-glutamyl transpeptidase (GGT) activity were measured in PANC1 and MCF7 cells treated ϮAA. Activated CASP3/7 (B) and GLI1 mRNA expression (C) were measured in PANC1 and MCF7 cells pretreated with the antioxidants NAC and L-glutathione (L-GLUT) before treatment ϮAA. C, control. D, transfected PANC1 cells were pretreated with NAC before treatment ϮAA. Left panel, luciferase activity measured in cells transfected with the GLI1 promoter reporter shows that NAC abolishes the downregulation of GLI1 promoter activity caused by AA. Right panel, measurement of the nuclear/cytoplasmic ratio of NFATc1 in cells transfected with NFATc1-GFP demonstrates that NAC prevents the AA-promoted increase in nuclear translocation of NFATc1. n Ն 140 cells per group. *, p Ͻ 0.05 versus the control group. Bars and error bars represent means and S.E., respectively. FIGURE 8. AA treatment inhibits tumor growth and metastasis. A, BALB/c mice were inoculated subcutaneously with 1 ϫ 10 6 LM3 murine mammary adenocarcinoma cells. Tumor volume was measured in the mice treated ϮAA every 7 days for 21 days. Bars and error bars represent means and S.E., respectively, calculated as follows: Tumor volume ϭ Length ϫ Width 2 /2. AA group, n ϭ 10; control group, n ϭ 7. *, p Ͻ 0.05 versus control group. B, lung tissues sections stained with H&E were analyzed for detection of metastasis. C, apoptosis evaluation in LM3 tumor section tissues by TUNEL assay (FITC labeling). TUNEL-positive cells in the nucleus are stained in green, and total nuclei are stained with DAPI in blue (at left). Data at right show bars and error bars representing means and S.E., respectively, for quantification of the apoptosis rates (average number of TUNEL-positive cells (stained in green) per visual field). Experiments were repeated in five tumor tissue sections for each condition. Ten fields of each section were selected at random for counting in a blinded manner; n ϭ 5. *, p Ͻ 0.05 versus control group. D and E, relative mRNA levels determined by quantitative PCR for Gli1 (D) and Bcl2, Bfl1/a1, and 4-1bb (E) in tumors from control and AA-treated animals. Bars and error bars represent means and S.E., respectively; n ϭ 5. *, p Ͻ 0.05 versus control group.
tumor regression (8,9). In addition, clinical studies show that intracerebral injection of ␥-linolenic acid in patients with gliomas induces decreased tumor size (12,13). Moreover, earlier studies had also shown that injection of LA and ␣-linolenic acid increases survival of rats with Mat 1376 ascites tumors (11). Subsequently, other studies showed that systemic treatment with the -6 PUFA LA decreased tumor growth and the incidence of metastases in a mouse model of lymphoma EL4 (10). Other reports have shown that AA has cytotoxic action, decreasing viability and inducing apoptosis in drug-resistant cancer cells (23)(24)(25)56). Moreover, similar to our data, AA-enriched human colon cancer cells exhibited diminished cell viability and increased CASP-dependent apoptosis (20,59). Although no studies have previously demonstrated a direct cytotoxic effect of AA on tumor development in vivo, it has been shown that, when intravenously infused, different radiolabeled PUFAs such as AA and docosahexaenoic acid induce rapid changes in the serum oxidant and antioxidant status, suggesting a possible therapeutic application in cancer (61,62).
Our study elucidates a novel mechanism for AA as a cytotoxic lipid in cancer cells. We define a new signaling pathway induced by AA leading to a decrease in cell viability and increase in cell death by reducing the expression of a subset of antiapoptotic genes, BCL2, BFL1/A1, and 4-1BB. We showed that AA reduces the expression of these genes by regulating their promoter activity in a transcriptional manner. Several studies have demonstrated that AA reduces the expression of apoptosis-related proteins, including the known antiapoptotic protein BCL2 (20, 63, 64). However, no studies have linked AA or other PUFAs to the antiapoptotic genes BFL1/A1 and 4-1BB. In addition, our results establish for the first time a transcriptional mechanism that explains the cytotoxic antitumoral effectsofAAinwhichthereducedtranscriptionoftheseantiapoptotic genes is related to the down-regulation of the transcription factor GLI1.
We propose a new GLI1 regulatory mechanism that is independent of known cascades regulating the activity and expression of this transcription factor (65)(66)(67). Our data show that NFATc1 acts as a repressor by binding to the promoter of the GLI1 gene in cells treated with AA. We further demonstrate that repression of GLI1 expression by NFATc1 in AA-treated cells is inhibited by antioxidants, suggesting that GLI1 repression by NFATc1 requires reactive oxygen species (Fig. 10). This conclusion is supported by the increase in reactive oxygen species detected in cells treated with AA ( Fig. 6A and Refs. 17, 49, and 51). We also show that GLI1 repression by AA is blocked by CsA, an inhibitor of calcineurin. This result suggests that calcineurin plays a role in the signal cascade initiated by AA treatment in our studies, which is consistent with the known role of calcineurin in the activation of NFATc1 (46,47). However, further studies are needed to determine the precise mechanisms of NFATc1 activation by AA. NFATc1 was first described as a regulator of T lymphocyte activation and differentiation and was subsequently implicated in the control of critical processes in many vertebrate systems (68,69). NFATc1 has been reported to form nuclear complexes with different transcription factors (Stat3, NFB, and CCAAT/enhancer-binding protein), thus activating the expression of genes that promote cell transformation and tumorigenesis (70 -73). Nevertheless, recent studies indicate that NFAT complexes also can induce gene silencing via cooperation of transcription repressor proteins and members of the histone deacetylase family (74 -76). NFATc1 complexes might similarly repress GLI1 expression by altering histone acetylation associated with the GLI1 promoter.
Taken together, the results of this work show that natural nutrients such as PUFAs control the expression of target genes involved in carcinogenesis. Knowledge of the mechanisms by