Inactivation of the Transcription Factor GLI1 Accelerates Pancreatic Cancer Progression*

Background: GLI1 is required for pancreatic tumor initiation; however, its role at later stages of carcinogenesis remains elusive. Results: Genetic inactivation of GLI1 accelerates pancreatic cancer progression. Conclusion: GLI1 can act as both a promoter and suppressor of pancreatic carcinogenesis depending on the tumor stage. Significance: This knowledge increases our understanding of pancreatic carcinogenesis and may help the design of therapies targeting GLI1-related pathways. The role of GLI1 in pancreatic tumor initiation promoting the progression of preneoplastic lesions into tumors is well established. However, its function at later stages of pancreatic carcinogenesis remains poorly understood. To address this issue, we crossed the gli1 knock-out (GKO) animal with cre-dependent pancreatic activation of oncogenic kras concomitant with loss of the tumor suppressor tp53 (KPC). Interestingly, in this model, GLI1 played a tumor-protective function, where survival of GKO/KPC mice was reduced compared with KPC littermates. Both cohorts developed pancreatic cancer without significant histopathological differences in survival studies. However, analysis of mice using ultrasound-based imaging at earlier time points showed increased tumor burden in GKO/KPC mice. These animals have larger tumors, decreased body weight, increased lactate dehydrogenase production, and severe leukopenia. In vivo and in vitro expression studies identified FAS and FAS ligand (FASL) as potential mediators of this phenomenon. The FAS/FASL axis, an apoptotic inducer, plays a role in the progression of pancreatic cancer, where its expression is usually lost or significantly reduced in advanced stages of the disease. Chromatin immunoprecipitation and reporter assays identified FAS and FASL as direct targets of GLI1, whereas GKO/KPC mice showed lower levels of this ligand compared with KPC animals. Finally, decreased levels of apoptosis were detected in tumor tissue in the absence of GLI1 by TUNEL staining. Together, these findings define a novel pathway regulated by GLI1 controlling pancreatic tumor progression and provide a new theoretical framework to help with the design and analysis of trials targeting GLI1-related pathways.

Pancreatic cancer is a devastating disease with a poor prognosis due in part to the complex network of pathways driving aberrant gene expression and tumor progression (1)(2)(3)(4)(5)(6)(7)(8)(9). Thus, an understanding of the mechanisms controlling pancreatic cancer growth and maintenance are essential for the development of future therapeutic approaches for this disease.
Our group is particularly interested in understanding the role of transcription factor GLI1, 2 a member of the GLI family of transcription factors, in pancreatic carcinogenesis. GLI1 has been shown to exert an oncogenic function working as an effector of the Hedgehog pathway in multiple malignancies (8, 10 -13). This transcription factor is expressed at low levels in normal adult tissue, but it is often re-expressed and de-regulated in many human cancers carrying signature mutations in the Kras gene (10 -13). Recently, we demonstrated that the loss of GLI1 in a genetically engineered mouse model (GEMM) blocks the progression of pancreatic intraepithelial neoplasms (PanIN) induced by oncogenic KRAS (10). In agreement with these findings, using a transgenic model overexpressing a repressor form of the transcription factor GLI3 to antagonize GLI1 activity, Rajurkar et al. (11) showed a requirement for this transcription factor in pancreatic cancer initiation. Furthermore, several studies demonstrate that blocking Sonic Hedgehog (SHH), the most well established upstream regulator of GLI1, down-regulates GLI1 expression and inhibits tumor growth in established xenograft models, suggesting a role for GLI1 in pancreatic tumor progression (13).
To define the biological significance and molecular mechanisms associated with expression of GLI1 in the advanced stages of tumor progression, and the mechanisms driving this phenomenon, we generated a quadruple cross GEMM resem-bling these late stages of the disease. This mouse model contained p48 cre-dependent activation of oncogenic Kras and loss of tumor suppressor tp53 (KPC) and gli1 resulting in the GKO/ KPC mouse. We hypothesized that the loss of Gli1 in this model would result in prolonged survival and lower frequency of pancreas tumors. Interestingly, both cohorts retained the common biological and molecular features of advanced pancreatic cancer, but GKO/KPC mice developed early signs of tumor burden and perished within the 1st month of life. Loss of GLI1 accelerated disease progression as shown by decreased survival, body weight, fatigue, increased tumor volume, increased serum lactate dehydrogenase (LDH), and leukopenia. Analysis of the mechanism showed that both FAS and FAS ligand (FASL) expression were significantly decreased in the GKO/KPC cohort. Down-regulation of these pro-apoptotic proteins is often associated with the progression of many cancers in conjunction with loss of tumor suppressor p53 (14,15), but more recent evidence suggests other factors may contribute to this phenomenon (16). Using ChIP assays, we demonstrate endogenous GLI1 binding to the FAS and FASL promoters. Furthermore, overexpression and knockdown studies confirm a role for GLI1 in the regulation of the expression of these genes. Finally, TUNEL staining of pancreas sections resulted in decreased levels of apoptosis in the GLI1-null group demonstrating the existence of a novel GLI1-FASL/FAS axis in pancreatic cancer. We hope this knowledge will provide a new theoretical framework to help with the analysis of existing studies targeting GLI1-related pathways as well as the design of future clinical trials.
Genotyping-Allele-specific PCR verified the presence of mutant gene arrangements in the DNA as described previously (16) with the addition of tp53 F primers as follows: tp53 tm1Brn , forward 5Ј-aaggggtatgagggacaagg-3Ј and reverse 5Ј-gaagacagaaaaggggaggg-3Ј.
Ultrasound in Vivo Imaging-Mice pancreata were imaged via ultrasound using the VEVO2100 small imaging system. Briefly, mice were anesthetized using 1.5 liters/min oxygen/ isofluorane mixture and then gently taped to the imaging stage. One-two ml of normal saline was injected into the right flank via intraperitoneal injection. Visual signs of distress as well as vital signs were monitored on the VEVO2100 (VisualSonics Inc, Ontario, Canada). Mice were imaged in both supine and oblique positions for better imaging. Still images were captured using the abdominal package in M-mode, and three images per pancreas (10 fields/analysis) were captured via three-dimensional Doppler for volume analysis following manufacturer's (VisualSonics) and published literature procedures (17,18). KPC tumors have a highly fibrotic nature and appeared extremely bright in the ultrasound field; thus, we used WT or KC tumors as controls. These tumors appeared seedy/fluid and dark/dense as described for WT animals (Fig. 3A) (18).
Lactate Dehydrogenase (LDH) Assay-Serum LDH was measured using a colorimetric kit (Abcam) following the manufacturer's recommended protocol. Briefly, 2.5 l of serum was mixed with assay buffer, at time 0 and 30 min after incubation at 37°C, and activity was measured by spectrophotometry at 450 . LDH activity was analyzed utilizing a linear standard curve equation (nmol/min/ml (in milliunits/ml)) by using the following equation: B ϫ sample dilution/(t 1 Ϫ t 0 ) ϫ V; where B is NADH (M r ϭ 763 g/mol) generated between t 0 and t 1 in nanomoles, and t 0 and t 1 indicate the absorbance at time 0 and after 30 min, respectively. V is the pretreated sample volume added into the reaction well (in milliliters).
Lymphocyte Isolation and Analysis-Mice were autopsied at 1 month of age. CD4 ϩ and CD8 ϩ T cells were isolated from KPC and GKO/KPC spleens, and lymphocytes from spleen, mesenteric lymph node, thymus, and blood (WT and GKO not expressing oncogenic mutant kras). T-cells were isolated using the standard MACs isolation kit (Miltenyi, Cambridge, MA) and biotin/antibody mixture/microbeads. The FACS analysis was performed using a FACSCanto system (BD Biosciences) and FACSDiva software. Data were analyzed using FloJo software.
Tissue Culture and Transfections-Immortalized mouse embryonic fibroblasts were transfected with GLI1 overexpression and shRNA plasmids using XtremeGENE transfection reagent (Roche Applied Science). Overexpression studies were analyzed at 48 h and knockdown studies at 72 h.
Chromatin Immunoprecipitation Assay (ChIP)-ChIP was performed in mouse fibroblasts with the EZ-Magna ChIP TM G kit (Millipore, Billerica, MA) using 1 g of rabbit IgG or GLI1 antibody (Novus Biologicals, Littleton, CO) as described previously (10) using the following primer sets for fas and fasl upstream of the first exon in each of the promoters as follows: fas forward 5Ј-gagttgtgtgggtctcagttt-3Ј and reverse 5Ј-catgaaggtgatctatagccaaataatg-3Ј ENSMUST00000025691; and fasl forward 5Ј-tggttaaggcacacttgctat-3Ј and reverse 5Ј-gaagtcccttccctgaacatc-3Ј ENSMUST00000000834. Alternative primers sets to confirm the specificity of the GLI1 antibody were designed to amplify regions upstream of the first exon in both fas and fasl at least 200 bp from the canonical GLI1-binding sites. The primers used were fas forward 5Ј-gcctattatctctacctctgctg-3Ј and reverse 5Ј-ggcctctgaatggtaattaccac-3Ј; fasl forward 5Ј-ggctgattccgtggaatcatgag-3Ј and reverse 5Ј-gtagaacttagctctgtagctgc-3Ј. Results were represented as percentage of input, where each antibody signal was relative to its respective input.
Luciferase Reporter Assay-For the assays, mouse embryonic fibroblasts were plated in triplicate into six-well plates in medium containing 10% FBS. Twenty four hours later, cells were transfected as indicated above. Samples were collected and processed in accordance with the manufacturer's protocol (Promega, Madison, WI). Intersample variations were controlled by measuring the total protein of samples in each well (Bio-Rad), and luciferase readouts were normalized to protein content. Relative luciferase activity represents luciferase readouts/protein concentrations normalized to control cells within each experiment.
Patient Expression Study-Patient samples analyzed for desmoplasia and GLI1 expression were cases from the Mayo Clinic that were recruited from October 1, 2000 to July 1, 2010. Recruitment, consent, and specimen collection were IRB-approved (IRB protocol number 06-002459) by our institution. Patient records were reviewed by a gastrointestinal specialist and confirmed as pancreatic adenocarcinoma. This study was done without the bias of underlying pathologies, prior treatment or demographics (gender, age, etc). Staging was performed using American Joint Committee on Cancer (AJCC) 6th Edition criteria. Patients were identified as resected, locally advanced, and/or metastatic. Resected patients were further categorized by stage IA, IIB, IIA, and IIB. Blinded histopathology analysis was performed by two expert pathologists at the Mayo Clinic, and desmoplasia was scored using the following criteria: 1) none; 2) mild; 3) moderate; and 4) severe. Median desmoplasia expression level was categorized into three groups as follows: (a) none/mild versus moderate versus severe; (b) severe versus none/mild/moderate; (c) moderate/severe versus none/mild. Analyses were performed using SAS software, version 9.3 (SAS Institute, Cary, NC). RNA was reverse-transcribed into cDNA (as described above). qPCR was performed using TaqMan primer/probe sets for human GLI1 and 18 S using ABI7900 technology and software (Applied Biosystems, Foster City, CA). Relative target gene expression was normalized to the reference gene using the 2⌬Cp method, and all reactions were performed in triplicate. Wilcoxon/Kruskal Wallis test (p value Ͻ 0.05) was used to test for statistically significant differences in expression across desmoplasia groups in both histology scoring and expression studies.
Terminal Deoxynucleotidyltransferase dUTP Nick End Labeling (TUNEL)-The DeadEnd fluorometric TUNEL system (Promega) was used to detect fragmented DNA of apoptotic cells in mouse pancreas tissues. PCNA (Abcam, Cambridge, MA) double staining was used to detect areas of proliferation. Seven mouse pancreas tissue sections per group, and four fields at ϫ10 were visualized, analyzed, and captured using the LSM 510 confocal microscopy (Zeiss, Thornwood, NY) and ImageJ software (National Institutes of Health).
Statistical Analysis-All data except the results from the human samples expression studies are expressed as the mean Ϯ S.E. Differences between groups were compared using a twotailed Student's t test. One-way analysis of variance was used to compare samples from the different time points in the same group.

Loss of GLI1 Accelerates Pancreatic Tumorigenesis-To
define the biological significance of GLI1 at later stages of pancreatic carcinogenesis, we established cohorts of mice containing common features of advanced pancreatic cancer. This model includes the well established cre-dependent, pancreasspecific activation of oncogenic mutant kras and loss of the tumor suppressor tp53 (2), coupled with the loss of gli1 (GKO/ KPC) (10). Mice from the GKO/KPC cohort succumbed to death at 1 month of age on average compared with 2.5 months in the KPC cohort containing normal expression levels of GLI1 (p ϭ 0.0001) (Fig. 1A). Furthermore, gli1 heterozygous mouse (GHet)/KPC survival was similar to KPC (p ϭ 0.2210) mice suggesting that one wild-type gli1 allele is sufficient to maintain functional GLI1 activity (Fig. 1B). All three cohorts (KPC, GKO/KPC, and GHet/KPC) developed fibrotic pancreas tumors that were similar in size (Fig. 1C) and showed similar oncogenic KRAS activity, as determined by the activation of the downstream effectors ERK (Fig. 1D and data not shown for GHet/KPC). Two independent expert pathologists completed blinded histopathological analysis of the three cohorts of experimental mice. Surprisingly, pancreas tumors were phenotypically indistinguishable between the varying degrees of GLI1 expression. Both KPC and GKO/KPC pancreata contained high grade PanIN II and III lesions, and more than 50% of pancreata in the cohorts presented with moderate fibrosis, inflammation, and pancreatic cancer ( Fig. 2A). Furthermore, no significant or very minor differences were observed between the cohorts with respect to epithelial cell content (CK19), inflammatory macrophages (F4/80), microvessel density (CD31), and fibrosis/collagen content or deposition (Trichrome staining) in areas of pancreas tumors in this end point survival study (Fig. 2B). Finally, we engaged in a detailed study of the mice lacking oncogenic kras re-arranged alleles to assess the transgenic colony for anomalies associated with cre-p48 or cre-tp53-null mediated effects. Histopathology of the pancreas, liver, spleen, stomach, heart, and lungs was completed over an extended period of time, and it was determined that no developmental defects or cre-mediated effects were present at 2, 4, 6 and 8 months of age, and the tissues appeared normal at all time points considered. No differences in survival were noted between cohorts over an 18-month period (data not shown). Together, these results suggest that the loss of GLI1 hastens mortality in the KPC GEMM with similar pancreas pathology in this end point study.
GLI1 Loss Accelerates Cancer Progression-To determine the cause of accelerated morbidity in the GLI1-null cohorts, we evaluated mice at earlier time points by ultrasound imaging at 25 days of age, and biological specimens were collected on autopsy at day 30. Both supine and oblique images were cap- tured and analyzed by ultrasound enabling a detailed capture of the pancreas and tumors (Fig. 3A). Using three-dimensional mode imaging, 10 fields per tumor region were captured and analyzed. The average tumor volume was smaller in KPC pancreas (56.14 Ϯ 14.27 mm 3 , n ϭ 14) compared with GKO/KPC pancreas (162.1 Ϯ 15.8 mm 3 , n ϭ 12) (p Ͻ 0.0001) (Fig. 3B). On autopsy, GKO/KPC average body weight was decreased (15.734 Ϯ 0.3725 g, n ϭ 15) compared with KPC (21.46 Ϯ 0.6697 g, n ϭ 15) (p ϭ 0.0001) (Fig. 3C), and average pancreas weight was augmented in GKO/KPC (0.590 Ϯ 0.0208 g, n ϭ 15) compared with the KPC cohort (0.290 Ϯ 0.0195 g, n ϭ 15) (p Ͻ 0.0001) (Fig. 3D). At 30 days, the GKO/KPC mice present with the physical attributes of advanced disease or cancer progression compared with KPC littermates. Increased LDH activity is associated with tissue damage and is often used as a prognostic tool to measure progression of several diseases, including cancer (20 -24). Next, we analyzed the serum of mice at 30 days post-natal. A direct correlation between increased tumor burden and elevated LDH activity was noted in the GKO/KPC cohort (0.397 Ϯ 0.005 nmol/min/ml, n ϭ 9) compared with KPC (0.363 Ϯ 0.009 nmol/min/ml, n ϭ 9) (p ϭ 0.0015) (Fig. 3E). Together, these results reflect animals that are experiencing accelerated cachexia due to accelerated tumor growth that is reminiscent of human pancreatic ductal adenocarcinoma progression. Because advanced stage cancer is often associated with progressive lymphopenia, particularly within the CD4 ϩ compartment (25-27), we evaluated CD4 ϩ /CD8 ϩ ratios from splenocyte populations as a clinically relevant marker of advanced disease. Lymphocytes isolated from the spleen were analyzed by FACS and showed a reduction in CD4 ϩ /CD8 ϩ T-cell ratio in the GKO/KPC cohort (1.745 Ϯ 0.1179) compared with KPC (2.440 Ϯ 0.2367) (p ϭ 0.0545) (Fig. 3F). Overall, these results resemble the dismal features of clinical disease that many patients present at the time of diagnosis. T-cell ratios in wild-type and GKO mice in the absence of kras or tp53 mutations were analyzed to determine whether the loss of GLI1 alone caused lymphopenia in the experimental mice. No significant alterations within the lymphocyte CD4 ϩ /CD8 ϩ subpopulations were noted in the spleen (WT 2.379 Ϯ 0.3629 and GKO 3.038 Ϯ 0.197, p ϭ 0.4145), mesenteric lymph (WT 2.41 Ϯ 0.2631 and GKO 2.822 Ϯ 0.553, p ϭ 0.5782), or thymus (WT 86.17 Ϯ 0.4333 and GKO 86.30 Ϯ 0.7506, p ϭ 0.8875) (Fig. 3G).
Not surprisingly, at day 30 GKO/KPC histopathology showed a predominance of late stage PanIN III lesions and carcinoma. Alternatively, PanIN II and few PanIN III lesions were noted in the majority of the pancreata in the KPC cohort (Fig.  4A). Immunohistochemical analysis of pancreas markers in time point studies failed to reveal statistically significant changes in other pathology between the cohorts, but the analysis was limited due to the rapid development of tumor and extensive fibrosis in GKO/KPC and a more normal architecture in the KPC pancreas. Hence, analysis and comparisons were confined to similar areas between the cohorts (Fig. 4B).
To ascertain whether this study recapitulates the clinical features of this disease, samples from patients with varying stages of the disease were analyzed by qPCR and histopathology. GLI1 expression levels were grouped as containing either high or low levels of GLI1. Matched H&E slides were reviewed and scored by a trained pancreas pathologist. Interestingly, patients with severe desmoplasia, a histopathological feature associated with poor prognosis, appeared to have a significantly lower median level of GLI1 expression when compared with those patients with none, mild, or moderate desmoplasia (0.48 versus 2.08 median expression, Kruskal Wallis (p ϭ 0.0238)). These results, coupled with recent reports (28,29), suggest that lower levels of GLI1 at later stages of pancreas tumorigenesis promote tumor progression. Together, these findings demonstrate that the loss of GLI1 results in increased tumor burden ultimately leading to accelerated morbidity and poor prognosis.
Regulation of FASL Underlies the Role of GLI1 during Tumor Progression-Cancer progression is often marked by aberrant expression of cytokines (30 -32). These molecules modulate growth, proliferative, migration, and apoptotic pathways, often resulting in metastasis and advanced stage disease (33)(34)(35)(36)(37)(38)(39)(40). Next, we evaluated expression of a panel of cytokines associated with disease progression using pancreas tumor tissues from the same cohort in the previous section by qPCR. Several cytokines showed trends in differential expression patterns, but only FASL (18-fold, p ϭ 0.010) was significantly decreased in GKO/ KPC pancreata ( Table 1). The FASL receptor, FAS, was also decreased (14-fold, p ϭ 0.008) in GKO/KPC samples (Table 1). We confirmed FASL and FAS expression results in a separate larger cohort of animals (n ϭ 7/group) with the same results (Fig. 5A). Interestingly, the changes in cytokine expression were accompanied by changes in the SHH pathway activation, the most well characterized regulator of GLI1 function. Expression studies demonstrate that the levels of receptor Patched, a marker of the activity of SHH, were down-regulated in GKO/ KPC mice compared with the KPC animals. Similarly, Smoothened, GLI2, and GLI3 were significantly diminished in the GKO/KPC, but the levels of SHH remained unchanged (Fig. 5,  B-G).
Bioinformatics analysis identified six candidate GLI1-binding sites in the promoter sequences of both FASL and FAS proteins (data not shown). ChIP assays confirmed the binding of endogenous GLI1 to the FAS promoter at Ϫ1147 to Ϫ943 bp upstream of the first exon. Alternative FAS primers amplifying a region without GLI1 consensus sequences were used to control for antibody specificity at Ϫ613 to Ϫ486 bp on the FAS promoter (Fig. 6A, left panel). GLI1 also bound to the FASL promoter at a region from Ϫ539 to Ϫ293 bp upstream of the first exon. Similar to FAS promoter, alternative FASL primers without GLI1 consensus sites in the promoter region Ϫ1646 to Ϫ1517 bp were also negative (Fig. 6A, right panel).
Overexpression of GLI1 in fibroblasts increased expression of FAS expression by 2-fold (p ϭ 0.0198) and FASL by 4.5-fold   (p ϭ 0.0211) (Fig. 6B, left panel). Conversely, knockdown of GLI1 resulted in a decrease in FAS expression by 1.8-fold (p ϭ 0.003), and a decrease in FASL expression by Ͼ8-fold (p ϭ 0.0285) (Fig. 6B, right panel) (Fig. 6C, right panel). Together, these studies identify FAS and FASL as direct targets of GLI1 and define a novel transcriptional mechanism for these proapoptotic molecules. FASL and FAS are members of the tumor necrosis ligand and receptor superfamily, and they play a critical role in apoptosis in multiple cell types (41)(42)(43). Thus, we analyzed proliferation (PCNA) and apoptosis (TUNEL) using triple fluorescence staining of pancreatic tumor tissues from both cohorts. GKO/  KPC showed a decrease in the number of apoptotic cells (0.78 Ϯ 0.466, n ϭ 7) compared with KPC (4.75 Ϯ 1.48, n ϭ 7) (p ϭ 0.002) (Fig. 6D). Cell proliferation (PCNA stain) in both cohorts confirmed that the cellular tumor bed of the tissue analyzed is not necrotic. DAPI staining provided reference for cell nuclei. Together, these results provide evidence of a novel GLI1-FAS/ FASL interaction that may play a role in the progression of pancreatic tumorigenesis.

DISCUSSION
GLI1, like other molecules (e.g. TGF␤), can act as both a positive and negative regulator of tumor development (8, 10, 11, 29, 44 -47). Examination of human and mouse tumor samples by immunohistochemistry or global arrays (49,(52)(53)(54)(55) demonstrates that the expression of GLI1 in many cancers often correlates with poor prognosis. Therefore, GLI1 expression signatures are often associated with cancer progression and metastasis (50,51). However, several in vitro and in vivo studies either inhibiting effectors upstream of GLI1 or GLI1 directly with genetic or pharmacological tools show that loss of this transcription factor decreases processes like epithelial-to-mesenchymal transition and cellular de-differentiation due to a less aggressive tumorigenic phenotype (56). Although GLI1 expression in pancreatic cancer clinical samples and pancreatic cancer GEMMs is well documented, the biological relevance of this result in pancreatic cancer progression is not completely understood. This is important because several clinical trials inhibiting pathways known to induce GLI1 activity (e.g. Hedgehog) are underway (12,48,49,(57)(58)(59). The success of these trials has been mixed, possibly due to the complexity of the signaling processes involved in pancreatic cancer progression. Recently, our group and others have identified GLI1 as a mediator in the process of cellular transformation in vivo where loss of GLI1 in the presence of oncogenic KRAS inhibited late stage PanIN formation and pancreas cancer initiation (10,11). However, the significance of GLI1 expression and the molecular mechanisms regulating expression in pancreatic cancer progression are conflicting. Thus, we generated a GLI1-null GEMM retaining the molecular and biological features of advanced pancreatic cancer. We hypothesized that the reduction of GLI1 in this model of tumor progression would slow tumorigenesis and cancer progression thereby extending survival. Surprisingly, the GKO/KPC died at a rate that was 50% faster than the KPC littermates. Histologically, the pancreas was similar in both cohorts. There appeared to be minor increases in fibrosis in the GKO/KPC cohort, but these alterations could not be scored by the pathologist as significantly different between the two cohorts. Marker analysis failed to provide insight to the relevant pathways that might be involved in progression in part due to a fairly normal phenotype in the KPC mice at 30 days. Thus, accurate comparisons could not be made and were primarily focused on comparing fibrotic areas of the pancreas that were present on both cohorts. Although the vasculature was diagnosed as normal in appearance, we saw an increase in microvessel density in GKO/KPC in the survival cohort, but the result was not significant. Only one section was stained and analyzed, and a more detailed study, including blood flow analysis, is warranted to conclude whether or not the findings are significant. Finally, to define the mechanism underlying these biological findings, we explored the role of cytokines known to modulate tumor progression, especially in the tumor microenvironment that might participate in escalating this phenotype. The panels of cytokines evaluated showed trends, but significant results were not achieved for most cytokines. However, expression of both FASL and FAS was significantly reduced in the GKO/KPC mice with a tight fit compared with KPC littermates. Both proteins are critical modulators in tumorigenesis and intimately involved in homeostasis inducing apoptosis on target cells (60,61). We demonstrated endogenous GLI1 binding to both FASL and FAS promoters and define these pro-apoptotic molecules as novel direct transcriptional targets of GLI1 in fibroblasts. Similar results were seen in other cell types that are known to express GLI1 in pancreatic cancer, such as cancer cells and lymphocytes (data not shown). Furthermore, we define a novel regulatory role for GLI1 in the FAS/FASL axis suggesting a possible explanation for the transformed cellular phenotype to propagate. To confirm this hypothesis, we examined the functional output of FAS/FASL axis and noted lower levels of apoptosis in the GKO/KPC cohort. Thus, we defined a GLI1-FAS/FASL axis as a novel regulator of cell survival in the advanced stages of pancreatic tumorigenesis.
Together, these results define GLI1 as a novel regulator of tumor progression and expand the list of pathways playing a dual role at different stages of tumor development. We provide evidence that the loss of GLI1 in a cancer progression model hastens morbidity over animals with wild-type levels of GLI1 suggesting that inhibition of GLI1 may be detrimental in some contexts in patients diagnosed with pancreatic cancer. Future studies will include the identification of genetic and epigenetic modulators regulating GLI1, FAS, and FASL transcription and expression with the hope that novel therapies to treat this dismal disease and other types of cancer will finally come to fruition.