Genome-Wide Screening Reveals an EMT Molecular Network Mediated by Sonic Hedgehog-Gli1 Signaling in Pancreatic Cancer Cells

Aims The role of sonic hedgehog (SHH) in epithelial mesenchymal transition (EMT) of pancreatic cancer (PC) is known, however, its mechanism is unclear. Because SHH promotes tumor development predominantly through Gli1, we sought to understand its mechanism by identifying Gli1 targets in pancreatic cancer cells. Methods First, we investigated invasion, migration, and EMT in PC cells transfected with lentiviral Gli1 interference vectors or SHH over-expression vectors in vitro and in vivo. Next, we determined the target gene profiles of Gli1 in PC cells using cDNA microarray assays. Finally, the primary regulatory networks downstream of SHH-Gli1 signaling in PC cells were studied through functional analyses of these targets. Results Our results indicate there is decreased E-cadherin expression upon increased expression of SHH/Gli1. Migration of PC cells increased significantly in a dose-dependent manner within 24 hours of Gli1 expression (P<0.05). The ratio of liver metastasis and intrasplenic miniature metastasis increased markedly upon activation of SHH-Gli1 signals in nude mice. Using cDNA microarray, we identified 278 upregulated and 59 downregulated genes upon Gli1 expression in AsPC-1 cells. The data indicate that SHH-Gli1 signals promote EMT by mediating a complex signaling network including TGFβ, Ras, Wnt, growth factors, PI3K/AKT, integrins, transmembrane 4 superfamily (TM4SF), and S100A4. Conclusion Our results suggest that targeting the molecular connections established between SHH-Gli1 signaling and EMT could provide effective therapies for PC.


Introduction
Sonic hedgehog (SHH) is involved in embryonic organogenesis as a morphogen. Inappropriate activation of SHH signals during pancreas formation results in agenesis and several pancreatic diseases [1]. SHH is excluded from the developing pancreas as well as the mature organ, but is upregulated in chronic pancreatitis, early pancreatic intraepithelial neoplasia (PanIN) lesions, and invasive pancreatic cancer (PC) [2]. Aberrant SHH upregulation was reported in subtotal human PC cells and might be a primary critical mediator of PC development [3].
The Hedgehog (HH) signaling pathway is closely related to tumor metastasis and prognosis in clinical studies and is required for PC tumor metastasis in orthotopic mouse models [3,4]. Recently, this pathway was thought to orchestrate the reprogramming of cancer cells via epithelial mesenchymal transition (EMT).
Interestingly, recent evidence found that SHH was significantly upregulated in gemcitabine-resistant PC cells that simultaneously express cancer stem cell (CSCs) markers [5]. Because the SHHinduced target gene products could contribute to the self-renewal, survival, and migration of cancer progenitor cells and Gli1 may play a crucial role in the malignant behavior of PC cells [6,7], identifying Gli1 targets is a logical step to understand its mechanism in PC cells.
The goal of this study was to provide a framework for the primary regulatory networks downstream of SHH-Gli1 signaling in PC cells. We also sought to determine if specific Gli1 target genes connect SHH-Gli1 signaling and EMT, thus providing a therapeutic strategy for PC.

Cell culture
The PC cell lines (BxPC3, AsPC-1, and Panc-1 were all saved by the Chinese Academy of Sciences.) were cultured in RPMI-1640 supplemented with 10% fetal calf serum (FCS). All cells were incubated at 37uC in a humidified atmosphere of 5% CO 2 in air.

Vector construction and cell infection
Lentiviral transfer vectors for human Gli1 shRNA or SHH cDNA were constructed by Genechem Co., Ltd, Shanghai, China. This system includes the lentiviral vector pLVTHM, the envelope plasmid pMD2G, and the packaging plasmids pRsv-REV and pMDlg-pRRE. The lentivirus-SHH (L-SHH) contains a 3.3-kb SHH coding sequence and the lentivirus-Gli1i (L-Gli1i) contains small hairpin Gli1 RNA to the targeting sequence of the shRNA, as previously described (59-CTCCACAGGCATACAGGAT-39) [8]. The lentivirus-control (L-C) did not include Gli1 interference sequences or SHH cDNA sequences and served as control. Lentiviral constructs were verified by DNA sequencing. Recombinant lentivirus was produced by transiently transfecting 293T cells following a standard protocol. When BxPC3, AsPC-1, and Panc-1 cells were approximately 50% confluent (in RPMI-1640 containing 2% FCS), they were infected with the lentiviral constructs at MOI of 5. Cells were harvested after 72 hours for further experiments. To identify functional L-SHH and L-Gli1i constructs, we routinely analyzed SHH and Gli1 expression by qRT-PCR.

RNA extraction and real time RT-PCR assays
Total RNA was extracted with Trizol reagent (Invitrogen Corporation, Carlsbad, CA, USA) according to the manufacturer's protocol. Total RNA (100 ng) was reverse transcribed in 20 ml volume and 2 ml cDNA was used for PCR, according to the manufacturer's instructions. (TaKaRa Biotechnology, Dalian, China). The primer sequences are shown in Table 1. CT (cycle threshold) values were standardized to CT values of GAPDH.

Protein extraction and western blotting assays
Total protein was extracted with RIPA buffer according to standard methods and samples were normalized for protein content using a commercially available kit (Bio-Rad Laboratories Inc Philadelphia, PA USA). Protein samples were separated by 6% SDS-PAGE (for Gli1 protein) and 12% SDS-PAGE (for SHH, Ecadherin, and GAPDH). Proteins were transferred to PVDF membranes and membranes were incubated for 2 h in TBST buffer, followed by incubation overnight at 4uC with the primary antibodies [1:1000 (v/v) for SHH, E-cadherin, or GAPDH and 1:500 (v/v) for Gli1] in blocking solution and visualization using the ECL detection system (GE Healthcare Biosciences, Piscataway, NJ, USA).

Transwell assays
Cell invasion assays (24-well sample kits; Chemicon, Bedford, MA, USA) were used to study PC cell line invasion and migration. Briefly, PC cells (1610 5 ) were separately seeded in serum-free media in Matrigel pre-coated transwell chambers (upper chamber), which contained polycarbonate membranes with 8-mm pores. Media containing 2% FCS was added into the bottom chamber. The transwell chambers were then placed on the 24-well plates. After incubation for 24 h, migration of PC cells was determined by photographing the membrane through the microscope. Counts were recorded from the 5 areas with the highest cell concentrations at high power magnification (6200). The mean value of the fields was considered the migration count of PC cells.

Liver metastases induction by splenic injection
Three groups of AsPC-1 cells (lentivirus-Gli1i, lentiviruscontrol, and lentivirus-SHH) were used to detect metastasis after intrasplenic inoculation into nude mice as previously described [9].
Briefly, mice were anesthetized with methoxyflurane, a minor abdominal left flank incision was made, and the spleen was exposed. AsPC-1 cells were injected into the spleen with a 30gauge needle. The spleen was returned to the abdomen, and the wound was closed in one layer with wound clips. After 8 weeks, we harvested the liver and spleen and produced continuous frozen sections. We stained the sections with hematoxylin and eosin and counted spleen tumors, intrasplenic miniature metastases, and liver metastases under a fluorescence microscope and optical microscope. All animal experiment protocols used in this study were approved by the Animal Research Committee of Tongji University.

cDNA microarray analyses
AsPC-1 cells transduced with L-Gli1i and L-C were used in cDNA microarray assays with the Affymetrix Human Genome U133 Plus2.0 Array GeneChip. Three experiments were performed on a single total RNA preparation from the cells. Signal values are presented as the mean value of 3 replicate experiments. cDNA microarray assays and statistical analyses of the gene expression results were performed as described previously [10].

Statistical Analyses
For all statistical analyses, we used SPSS17.0 software (SPSS, Inc, Chicago, IL, USA). Continuous variables are expressed as the mean 6 SE. Non-paired Student's t-tests were used for statistical evaluation. P,0.05 was considered statistically significant.

Lentiviral-Gli1i and -SHH transduction efficiency and PC cell EMT is regulated by SHH-Gli1 signaling
We transfected three PC cell lines with the lentiviral Gli1 interference vector (L-Gli1i), SHH over-expression vector (L-SHH), and control vector (L-C). We verified alterations in activation of SHH-Gli1 signaling by evaluating the expression of SHH, Gli1, and Patched1 using real-time RT-PCR. The real-time RT-PCR data revealed that the Gli1 and Patched1 genes were significantly downregulated by L-Gli1i transduction, whereas Gli1 and Patched1 were upregulated by L-SHH transduction compared with L-C (P,0.01; Figure 1A). Gli1 and Patched1 were target genes in most cell types with SHH signaling activated, therefore, the results suggest the lentiviral vectors efficiently changed the activation of SHH-Gli1 signals. The E-cadherin mRNA levels were drastically reduced by increased SHH/Gli1expression in PC cells. A similar trend was observed with the Ecadherin protein.

PC cell invasion and migration is regulated by SHH-Gli1 signaling
Data from the transwell assays showed that an increased number of cells from the PC cell lines invaded in a Gli1 dosedependent manner through the Matrigel-coated filter within 24 hours (P,0.05; Figure 2A1-A9, B).       The SHH-Gli1 signaling pathway regulates PC cell proliferation Our MTT data showed that the L-Gli1i/SHH transduction did not significantly influence cell proliferation within 24 hours. However, after 48 hours, PC cell proliferation increased with viral transduction (Figure 2C1-C3).

Liver metastases after injection of AsPC-1 cells into nude mice is regulated by SHH-Gli1 signaling
Our data from the nude mice model showed that 8 weeks after intrasplenic injection of AsPC-1 cells, there were spleen tumors in 8 of 10 mice in the L-Gli1i group, 8 of 9 mice in L-C group, and 9 of 9 animals in the L-SHH group. The average numbers of splenic miniature tumors were 2.6, 4.9, and 8.9, respectively. The incidence of liver metastases was 3 of 8 mice in the L-Gli1i group, 5 of 8 mice in the L-C group, and 8 of 9 animals in the L-SHH group. The average numbers of liver metastases were 2.7, 4.2, and 6.7, respectively ( Figure 3, Table 2).

cDNA microarray analyses of Gli1 target genes in AsPC-1 cells
The Patched1 gene, a direct target of Gli1, was upregulated 1.71341-fold in this study. Therefore, we set 1.7-fold regulation as the target gene standard. Using this threshold, the target gene profile data showed that 278 genes were upregulated and 59 genes were downregulated upon Gli1 in AsPC-1 cells. (Table 3). The regulated genes were classified into different categories based on well-documented and established biological or pathological function. Genes regulated by Gli1 belong principally belong to the following categories: cell invasion/migration, angiogenesis, cell survival, transport, metabolism, signal transduction, and immune system defense (Figure 4). We then compared these target genes with previous data by searching the Medline database to screen for differentially expressed PC genes and SHH signaling pathway Table 4. PC-related genes and Hedgehog-related genes reported previously in the target genes profile data.  target genes. Utilizing this approach, we identified 58 upregulated genes (Table 4) and 1 downregulated gene upon Gli1 inhibition in our screen that were previously been found to be similarly regulated in PC. Using the same method, we found 22 upregulated genes upon Gli1 inhibition that were previously found to be correlated with SHH signaling (Table 4). Moreover, 15 of 22 genes that were reported to be overexpressed in PC were involved in cell metastasis, including ITGB4, ANG, VEGFA, S100A4, WNT5A, and TGFB2 as well as cell survival, such as BCL2, BIRC3, IGFBP6, KLF4, and PLAU. At least 8 genes (WNT5A, BCL2, IGFBP6, PTCH1, MSX2, TGFB2, HOXC6, and SOX13) were previously demonstrated to be direct targets of SHH signaling [10,11].

Discussion
In this study, cell survival target genes could be divided into several types: (1) proliferation-related genes, such as IGFBP6, IGF1R, IRS1, EGFR, and ALOX5, (2) apoptosis-related genes, such as BIRC3 and Bcl-2, (3) Cell cycle-related genes, such as CCNG2, CDC2L6, and CDK6, and (4) CSC orCSCs maintenance-related genes. The stem cell phenotype predominantly included EMT, anti-treatment, and stem cell markers. The IGF signaling pathway was a key proliferation-related pathway and the Bcl-2 family was an important classic apoptotic signaling pathway. It was reported that Gli1 directly regulates CCND transcription and our data suggests it may regulate CCNG2 in the same manner [7]. The ABCC3 gene encodes multidrug resistance-associated protein 3 (MRP3), which is involved in chemotherapy resistance of cancer cells [12]. Moreover, MTS upregulation and CTPS downregulation has also been reported to lead to chemotherapy resistance [13]. In addition, KLF4 is a stem cell marker that promotes cancer stem cell population maintenance and CD59 upregulation may be associated with tumor cell immune escape [14,15]. Interestingly, HUS1 downregulation likely weakens the DNA damage repair mechanisms [16].
Angiogenesis is necessary for cancer metastasis as well as for CSCs microenvironment maintenance. Substantial evidence suggests that activated SHH signaling may be one angiogenesisinitiating signaling pathway during pancreas carcinogenesis, though its exact mechanism is not known [17]. In this study, we found that Gli1 significantly upregulated pro-angiogenic factors, including ANG, VEGFA, PDGFA, TNFRSF12A, and IL-8, suggesting it has an important regulatory role in PC angiogenesis. Moreover, in this study, VEGF and PDGF were upregulated at the same time, suggesting that the proangiogenesis mechanisms of the SHH pathway are not just involved in endothelial cells (ECs) tuberformation, but also vessel wall maturation.
It was reported that SHH signaling pathway activation accompanied EMT, and EMT is required for migration of SHH-responsive cells during tissue morphogenesis. However, there was no evidence that Gli1 directly regulated Snail or Slug transcription. In the present study, target profile data showed that SHH signaling in EMT involved a complex crosstalk network ( Figure 5A). The EMT-related target genes are summarized as follows: (1) TGF-b signaling pathway: TGFb2 and TGFbR3. Previous studies showed that TGFb signaling is significantly elevated in PC with Smad4 mutation, resulting in the loss of Smad4-dependent cell growth inhibition and increased Smad4 independent EMT [18]. (2) Ras signaling pathway: RAB27B, RAB8B, RASA4, RHEBL1, RHOU, RRAS, and RIN2. Data indicate the Ras/ERK1/2 pathways are involved in the mesenchymal transformation of PC cells [19]. (3) Wnt signaling pathway: wnt5a. Previous study indicate that wnt5a promotes EMT through a non-classical pathway [20]. (4) PI3K/AKT signaling pathway: ITPKB. PI3K was found to strengthen Snail nuclear colonization through PAK1 activation of the AKT signaling pathway in EMT [21]. AKT functions as a central point to transduce extracellular (growth factors including insulin, IGF-1, and EGF) and intracellular (such as mutated or activated receptor tyrosine kinases, PTEN, Ras, and Src) signals [22]. (5) Growth factor and receptor signaling pathways: IGF1R, IGFBP6, IGFL2, EGFR, PDGFA, and VEGFA. Previous studies have demonstrated that abnormal activation of these pathways promotes epithelial-derived tumor expansion and progression through promotion of EMT-like transitions. Regarding mechanisms, IGFR signaling induces expression of the transcription factors Snail and Zeb [23]. PDGF may induce EMT via activation of the Wnt signaling pathway [24]. VEGF and EGF can increase of Snail and Twist protein expression [25]. (6) Integrins: ITGA3, ITGB4, and ITGB6. It has been reported that the a3 and b4 subunits can make up lamininbinding integrins with other subunits, such as a3b1 or a6b4, and these subunits can be palmitoylated that may contribute to integrin-tetraspanin interactions [26]. The potential prometastatic functions of these integrin subunits, particularly b4, were reported previously and tyrosine phosphorylation of the b4 Shc-binding site results in disassembly of hemidesmosomes and mobilization of signaling-activated a6b4 integrin. Mobilized a6b4 switches from keratin to actin filament association and may mediate migration and invasion of laminin isoforms [26]. (7) TM4SF: TSPAN1. TSPAN1 gene over-expression was detected in liver cancer, prostate cancer, gastric cancer, cervical cancer, and colorectal cancer [27]. It has been proposed that TSPAN1 gene expression correlates with cell proliferation and cancer prognosis. Our data suggests that TSPAN1, as a member of TM4SF, may participate in the EMT process of PC cells. However, it remains to be determined how it interacts with integrins, growth factors, or other TM4S proteins. (8) MicroRNAs: miR-21. Studies have shown that miR-21 is associated with PC metastasis and prognosis and may play a role in TGF-b -induced EMT [28,29]. (9) S100A gene family: S100A4. It has been reported that S100A4 and E-cadherin are inversely regulated in several cell systems and that S100A4 promotes the expression of the essential transcription factors, Twist and Snail, in the EMT process, as well as mesenchymal markers, including vimentin and MMPs [30,31].
Interesting, our data suggests that the EMT molecular network mediated by SHH signaling may contain at least two important positive feedback loops in PC cells. The first is the positive feedback between SHH and TGFb signaling. In vitro and in vivo evidence suggests the crosstalk between TGFb and SHH results in reciprocal induction. TGFb upregulated SHH and activated Gli1 during EMT induction; however, SHH signaling upregulated TGFb2 and TGFBR3 as demonstrated in this and a previous study [32,33]. The second positive feedback loop is between SHH and Ras signaling. Previously, studies showed that k-ras mutation was an essential mechanism of SHH and Gli1 upregulation in PC cells and in this study, we found that Gli1 upregulated several Rasrelated genes to activate Ras signaling [34].
Based on previous studies and our data, we speculate that this molecular network might start with k-ras mutations, followed by SHH signaling activation, and finally, the TGFb signal joins and a positive feedback loop forms between the three pathways. The SHH signal was continuously enhanced through this positive feedback and directly promotes EMT via regulation of EMTrelated Gli1 target genes, such as IGFR1, VEGF, EGF, and S100A4. ( Figure 5B). However, this molecular network model may be more complex with the participation of additional signaling proteins, such as integrins, PI3K/AKT, and WNT.