Abstract
Pancreatic cancer aggressiveness is characterized by its high capacity for local invasion, ability to promote angiogenesis, and potential to metastasize. Hypoxia is known to represent a crucial step in the development of aggressive malignant features of many human cancers. However, micrometastatic tumors are not typically subjected to hypoxic events during early stages of dissemination; therefore, it is unclear how these tumors are able to maintain their aggressive phenotype. Thus, the identification of regulators of hypoxia-related genes in aggressive/metastatic tumors represents a fundamental step for the design of future therapies to treat pancreatic cancer. To this end, transcriptomic profiles were compared between the nonmetastatic pancreatic cancer cell line FG (LMET) and its angiogenic/metastatic derivate L3.6pl (HMET) under normoxic or hypoxic conditions. Cluster analysis revealed a number of transcripts that were induced by hypoxia in nonmetastatic cancer cells. Strikingly, this cluster was determined to be constitutively activated under normoxia in the metastatic cancer cells and could not be further induced by hypoxia. A subset of these transcripts were regulated by the transcription factor SOX9 in the aggressive-metastatic cells, but driven by hypoxia-inducible factor-1α (HIF-1α) in the parental nonmetastatic cell line. Moreover, these transcripts were enriched in cancer-related networks including: WNT, CXCR4, retinoic acid, and (FAK) focal adhesion kinase, gene PTK2 signaling pathways. In functional assays, inhibition of SOX9 expression in HMET cells led to increased apoptosis and reduced migration in vitro and a significant reduction in primary tumor growth, angiogenesis, and metastasis following orthotopic tumor cell injection. At the molecular level, the control of SOX9 expression was associated with changes in the methylation status of the SOX9 promoter. Finally, SOX9 upregulation was verified in a series of tumor specimens of patients with pancreatic carcinoma.
Implications: SOX9 represents a novel target for pancreatic cancer therapy. Mol Cancer Res; 12(3); 421–32. ©2013 AACR.
Introduction
Pancreatic cancer is a leading cause of cancer-related deaths in western countries (1). The biologic aggressiveness of pancreatic cancer is defined by local invasion, tumor angiogenesis, and its potential to metastasize. Hypoxia plays a significant role in these processes by readjustment of gene expression to support tumor cell survival. Here, we demonstrate that the transcription factor SOX9 also plays an important role in the oxygen independent regulation of hypoxia-associated gene expression responsible for tumor progression and metastases in pancreatic cancer.
SOX9 [SRY (sex determining region Y)-box 9] was originally identified as a factor necessary for development of the testis (2). SOX9 acts as a transcription factor, binding to a defined consensus sequence (3, 4). Phosphorylation of Sox9 on 2 S64 and S181 protein kinase A sites is cAMP-dependent and associated with activation of the transcription factor (5).
During cartilage development, hypoxia helps promote the differentiation of mesenchymal cells into chondrocytes in part through the activation of the transcription factor Sox9 via a hypoxia-inducible factor-1α (HIF-1α)–dependent mechanism. This expression of SOX9 is abolished by deletion of hypoxia responsive element sites in the SOX9 promoter (6).
SOX9 has also been described as a specific marker and maintenance factor for multipotential progenitors during pancreas organogenesis. In the developing pancreas, SOX9 expression is restricted to a mitotically active, Notch-responsive subset of PDX1(+) pluripotent progenitors, and is absent from committed endocrine precursors or differentiated cells. Seymour and colleagues demonstrated that SOX9 maintains pancreatic progenitors by stimulating their proliferation, survival, and persistence in an undifferentiated state (7).
The pancreatic carcinoma cell line L3.6pl (ref. 8; HMET) contains typical genetic alterations that are associated with an early developmental phase of human pancreatic cancer including somatic mutation of the KRAS gene (KRASmut) as well as an amplification of the ERB2 gene (9). L3.6pl (HMET) contains KRASG12D, the clinically most relevant activating point mutation found in Codon 12 of the KRAS gene (10–12). This mutation is not present in the parental cell line and thought to have arising during in vitro selection of the tumor line (10, 12).
Typical hypoxic gene regulation would bring advantage of more flexible rearrangement of metabolism and regulations toward hypoxic functions already for micrometastasis. Nevertheless, small cellular aggregates typical for micrometastasis do not yet suffer on hypoxia and therefore HIF-1α cannot be the main regulator causing the regulatory and metabolic rearrangement. In this study, we demonstrate that overexpression of SOX9 is associated with the unusual regulation of hypoxia-inducible genes such as VEGF with consecutive activation of primary tumor growth and metastasis in clinically relevant human pancreatic cancer.
Materials and Methods
Cell culture
The following human pancreatic carcinoma cell lines were used in this study: FG (low-metastatic, e.g. LMET) and L3.6pl (high-metastatic, e.g. HMET; ref. 8). Both cell lines were maintained in Dulbecco's Modified Eagle Medium under culture conditions and with supplements as described previously (8). The last 24 hours cells were incubated either in hypoxic (1% O2, 5% CO2, 94% N2) or remained at atmospheric conditions.
VEGF ELISA
To analyze the protein concentration of VEGF in the conditioned media, the human VEGF ELISA Kit (BenderMedsystems) was used according to the manufacturer's protocol.
Microarray analysis
Total RNA was isolated from the cultured cells by the TRizol method assayed for integrity and purity by agarose gel electrophoresis. Synthesis of cDNA, transcription labeling, and purification was performed with Affymetrix Kits according to the manufacturer's protocol. Affymetrix GeneChips HGU 133 plus 2.0 were hybridized, stained, and scanned. The full dataset is available at the GEO database (http://www.ncbi.nlm.nih.gov/geo/) under the accession number GSE9350. Differentially expressed genes were identified by fitting a linear model for a 2 × 2 factorial design with the factors culture conditions (hypoxia vs. atmospheric oxygen pressure) and cell line with the R-package LIMMA (13). Three independent biologic replicates were subjected to analysis. The threshold of significance in the moderated t test was set to a P value < 0.01 after false discovery rate adjustment and to a fold change >2.0. The expression values for the genes significant for a specific pair-wise comparison extracted from the linear model were converted to log fold change relative to the mean of all samples and clustered using the SOTA algorithm with a Pearson uncentered metric within the program suite MeV 4.0 (http://www.tm4.org/mev.html). The genes constituting a cluster were analyzed for enrichment of transcription factor binding sites with the program oPOSSUM 2.0 using the recommended significance threshold of Z-score > 10.0 and Fisher P-value < 0.01 (14).
Methylation assay
Methylation assay was performed using the MetylCollector Kit (Active Motif) with MseI digestion according to the manufacturer's protocol. The presence and quantity of purified methylated-Sox9 fragments were detected via PCR using the primers (Supplementary Materials). Methylated BRCA-DNA delivered with the kit was used as a positive control. Products of purification in absence of methyl-CpG-binding domain protein were used as a negative control.
Construction of plasmids for regulated inhibition of Sox9 expression
pENTRSox9shRNA and control pENTRLacZshRNA based on the plasmid vector pENTR/H1/TO were constructed using BLOCK-iT Inducible H1 RNAi Entry Vector Kit (Invitrogen) with oligonucleotides Sox9shRNAforw, Sox9shRNArev (see Supplementary Materials) and verified by DNA sequencing. The negative control construct expresses shRNA with no target in human cells (Escherichia coli LacZ gene) was constructed using oligonucleotides LacZshRNAforw and LacZshRNArev (see Supplementary Materials).
Inhibition of Sox9 expression in cell culture
HMET cells were stably transfected with the plasmid pcDNA6TR (Invitrogen) and evaluated for overproduction of E. coli tetracycline repressor via Western blotting with anti-tetracycline repressor (MoBiTech). The overproducers were then stably transfected with Sox9shRNA or LacZshRNA constructs and selected. Expression of the respective shRNAs was induced by treatment with 10 μg/mL tetracycline for 48 hours.
Determination of apoptosis using FACS
To evaluate the fraction of apoptotic cells, cell-cycle analyses were performed according to methods described elsewhere (15). Tumor cells transfected with shRNA against Sox9 were cultivated in 6-well plates with or without tetracycline treatment.
Orthotopic tumor injection
For analysis of orthotopic tumor growth, HMET Sox9sh and HMET LacZsh cells, respectively, were injected into the pancreas of 20 male athymic nu/nu BALB/c mice (Charles River WIGA), as described previously (8). Expression of Sox9 was inhibited by induction of transcription of Sox9-specific shRNA with tetracycline. Therefore, the animals were divided into 2 experimental groups (10 animals each): one group received 1 g/L tetracycline with glucose in the drinking water, the other group received isotonic solution of glucose (Sigma) in tetracycline-free MilliQ-water. Orthotopic tumor growth was monitored every second day by measuring the tumor volume using a caliper. On day 37 after tumor cell injection, animals were sacrificed and examined for primary tumor growth, lymph node, and liver metastases.
RT-PCR
Reverse transcription (RT)-PCR or quantitative reverse transcription PCR (qRT-PCR) was performed using One-Step RT-PCR Kits with Platinum Taq High Fidelity (Invitrogen) and Light Cycler with total RNA isolated from cell culture or tissue samples using RNeasy Kit (Qiagen) according appropriate manufacturer's protocol. The following primers were designed to span 2 exons if it sequence allowed. Primer sequences: see Supplementary Materials.
Chromatin immunoprecipitation
Chromosomal immunoprecipitation (ChIP) of the promoters binding Sox9 was performed using ChIP-IT Express Magnetic Chromatin Immunoprecipitation Kit (Active Motif). Sox9 complexes were immunoprecipitated using anti-Sox9 antibody (Tebu-bio). Immunoprecipitated promoter sequences were identified using PCR with the following primers (see Supplementary Materials). The primers recognize the sequence either within published promoter sequence or 1 kb upstream of transcription start. One twentieth of the immunoprecipitated sample and one twentieth of the sample precipitated using the negative control antibody, or sample containing no antibody or beads were subjected to PCR using promoter-specific primers and analyzed by agarose gel electrophoresis.
Western blotting
Proteins from cell lysates or from immunoprecipitation were subjected to Western blotting using the appropriate antibody according to method described elsewhere (15). Bands were visualized using WesternDot 625 Western Blot Kit (Invitrogen) and documented using gel documentation system.
Investigation of HIF-1α stability
HMET Sox9shRNA was cultivated under hypoxia or normoxia in presence or absence of tetracycline, respectively. Nuclear and cytoplasmatic extracts were prepared using Nuclear Extract Kit (Active Motif) according manufacturer's manual. Both extracts were subjected to Western blotting using anti-HIF-1α antibody (Abcam), the membrane was stripped and reprobed with anti-β-actin (Sigma) or anti-H4 histone antibody (Abcam), respectively. Data show that reduced amount of Sox9 does not influence hypoxic stabilization of HIF-1α.
Examination of SOX9, VEGF expression, and SOX9 methylation in patient samples
The collection of human tumor tissue was approved by the local ethics committee. Operative specimens of tumor and corresponding normal tissues were obtained from 15 patients with histologically confirmed pancreatic ductal adenocarcinoma. The tissues were snap frozen and stored in liquid nitrogen. RNA or chromosomal DNA extraction were done using RNeasy Mini Kit or DNeasy Blood & Tissue Kit, respectively (both Qiagen). Quantitative results were normalized for glyceraldehyde-3-phosphate dehydrogenase.
Boyden chamber assay
Cell migration activity was determined using a colorimetric QCM Haptotaxis Cell Migration Assay—Fibronectin (Millipore). Complete growth medium [supplemented with 10% fetal calf serum (FCS)] was used as chemo-attractant. Untreated cells as well as membrane coated with arbitrary protein (bovine serum albumin, BSA) were used as a negative control.
Statistical analysis
Data are given as the mean ± SE in quantitative experiments. For statistical analysis of differences between the groups, an unpaired Student t test was done. Resulting P values were utilized for testing of the significance of certain event. As the significant are qualified only events that difference shows P value < 0.05.
Results
VEGF expression is increased in metastatic HMET cells independent of hypoxia
The LMET and HMET pancreatic cell lines derive from the same parental cell line, but HMET shows a more “aggressive” metastatic phenotype. The biology of these cell lines was examined with regards to their response to hypoxic stress and related processes associated with the development of a metastatic phenotype.
Increased expression of VEGF is an important indicator of hypoxic stress. VEGF protein and RNA expression in LMET and HMET cells was evaluated under normoxic and hypoxic conditions. Interleukin (IL)-6 mRNA expression was used as an additional control of hypoxic inducibility (16, 17). Results presented in Fig. 1A and B show inducible upregulation of VEGF RNA and protein expression in LMET cells (P < 0.0001), whereas metastatic HMET cells constitutively exhibited increased VEGF expression under normoxic conditions (P < 0.0001), which was not further inducible by hypoxia (Fig. 1A and B). In these cells, constitutive oxygen-independent upregulation of VEGF was detected on transcriptional (Fig. 1A) and protein levels (Fig. 1B).
Microarray analysis revealed an enrichment of genes with binding sites for SOX9 under hypoxia in LMET cells
Microarray analysis was used to profile the transcriptome of the cell lines cultivated under normoxic or hypoxic conditions. RNA was isolated and hybridized to Affymetrix human U133 Plus 2.0 GeneChip microarrays. Of the 24,734 probe sets, analysis yielded 2,986 differentially expressed genes at a significance threshold of P < 0.01 and a fold change higher than 2. A total of 742 genes were found to respond (either up or down) to hypoxic changes in LMET cells (Supplementary Fig. S1A; patterns 1 and 2), whereas these genes showed only slight expression changes (Supplementary Fig. S1A; pattern 1), or were not regulated (Supplementary Fig. S1A; pattern 2) in the HMET cells where 280 genes were found to be oxygen-responsive (Supplementary Fig. S1A; pattern 3).
We hypothesized that the regulatory activity and gene expression seen may result from the actions of a small set of specific transcription factors linked to the constitutive upregulation of “hypoxia” associated genes in the HMET line. In this regard, transcriptomic profiling identified hypoxia-induced expression of SOX9 in the LMET cells, whereas it was found to be constitutively expressed in HMET cells (3.5-fold; P = 0.014; normoxia vs. hypoxia; GEO database: accession number GSE9350).
Sox9 is constitutively phosphorylated in HMET cells independent of hypoxia
The mRNA and protein expression of SOX9 was investigated in more detail in the LMET and HMET cells under hypoxic and normoxic conditions. The results showed increased SOX9 expression in HMET at the protein level (Fig. 2A). SOX9 and VEGF transcription was constitutively upregulated in HMET cells and was not further inducible under hypoxic culturing conditions (Fig. 2B). The oxygen-dependent regulation of VEGF exhibited a similar transcriptional pattern to that seen for SOX9 (Fig. 2B). The results showed no significant difference in protein activation (phosphorylation of Ser-181; ref. 5) when normalized to total amount of SOX9 in both cell lines (Fig. 2C).
Pathway analysis and expression of SOX9 target genes
The synchronization of gene expression linked to regulatory pathways can be associated with the activity of transcription factors such as SOX9. To better understand the functional consequences of the gene regulation seen in HMET, the genes were grouped according their differential reactivity to hypoxia in LMET versus HMET cells. The genes were also analyzed for presence of Sox9-binding site in their promoter region using the GeneQuest program from Lasergene package from DNASTAR. Genes from this subgroup were ordered into known pathways and analyzed using Pathvisio software (18). These data show a high prevalence of potential SOX9 target genes in cancer relevant pathways including the CXCR4, focal adhesion, retinoic acid, and WNT pathways. The expression of key genes in these pathways was then verified by qRT-PCR. The results as depicted in Fig. 3A shows the expression of Sox9 target genes and Fig. 3B describes the fold change of transcription of the investigated gene under hypoxia compared with normoxia (values greater than 1 show upregulation whereas less than 1 define downregulation under hypoxia). A significant difference between hypoxia-induced regulation of expression in LMET versus HMET cells of the gene FYN belonging to CXCR4 pathway, PAK3 linked to the focal adhesion pathway, Retinoic acid receptor (RXRA) associated with the retinoic acid pathway, and the WNT11 gene belonging to WNT pathway was observed in LMET versus HMET cells.
ChIP verified SOX9 binding to target gene promoters
ChIP was used to verify binding of SOX9 protein to promoter sequences of the investigated genes. Cross-linked chromosomal fragments were immunoprecipitated using an anti-SOX9 antibody. The results showed SOX9 binding to promoters of the genes from investigated pathways as follows (Fig. 4): CXCR4, FYN, and RHOB from CXCR4 pathway, FN1 and PIK3R1 from focal adhesion pathway, RDH10, RETSAT, and RXRA from the retinoic acid pathway, as well as WNT11 and FZD7 from the canonical WNT pathway. No binding of Sox9 to GRB7, PAK3, and SMAD3 promoters could be demonstrated (data not shown). Sox9 binding was further verified on the promoters of other cancer relevant genes, which contain conserved Sox9 binding sites including; VEGF, mTOR, β-catenin, and IFIT3 (Supplementary Fig. S1B).
Epigenetic changes are linked to Sox9 induction
To investigate whether differential methylation of the SOX9-promoter could underlie the changes in SOX9 expression seen in LMET versus HMET cells, the methylation analysis was performed. Chromosomal DNA was isolated from both cell lines grown under hypoxia or normoxia and digested with the methylation-sensitive MseI restriction endonuclease. Under normoxic conditions, the Sox9-promoter in LMET cells was hypermethylated as compared with HMET cells (Supplementary Fig. S2A). However, under hypoxic conditions methylation of the SOX9-promoter was substantially reduced in the LMET cells, in contrast, methylation of the SOX9-promoter in HMET cells was constitutively low suggesting reduced potential responsiveness to hypoxia. Methylation of the SOX9-promoter in LMET and HMET cells seems to be inversely related to the expression of VEGF and SOX9 at the protein and RNA level under normoxia and hypoxia.
Inhibition of SOX9 expression impairs the aggressive behavior of HMET cells
We then asked whether inhibition of SOX9 expression could revert the biologic aggressive phenotype of HMET cells toward a less aggressive phenotype as in LMET cells. To this end, HMET cells were stably transfected with an inducible vector expressing SOX9 shRNA. Treatment with 10 μg/mL tetracycline for 48 hours induced shRNA expression. A reduction of SOX9 expression was confirmed by Western blotting (Supplementary Fig. S1D).
Induced expression of Sox9shRNA lead to changes in the expression of all investigated downstream genes. CXCR4 (CXCR4 pathway), FN1, GRB7, PAK3, and PIK3R1 (focal adhesion pathway), RDH10, RETSAT, and RXRA (retinoic acid pathway), SMAD3 and FZD7 (WNT pathway) also showed reduced expression after inhibition of Sox9 expression. Expression of FYN and RHOB from CXCR4 pathway as well as WNT11 from WNT pathway showed a significant increase in expression following knockdown of Sox9 expression (Fig. 5A). The results also showed (Supplementary Fig. S4) that reduction of SOX9 expression did not lead to alterations in the steady-state concentration of HIF-1α in the nucleus or in cytoplasm suggesting that SOX9-dependent regulation of hypoxic transcription represents an alternative to HIF-1α–mediated gene transcription.
Resistance to apoptosis is associated with a malignant tumor phenotype. An increased fraction of apoptotic cells under both serum depletion as well as normal culturing conditions was also detected in the Sox9shRNA transfectants [Fig. 5B: typical fluorescence-activated cell sorting (FACS) scans]. The induced inhibition of Sox9 expression also led to a significant reduction in VEGF production (Fig. 5C, P < 0.0001, Sox9 shRNA transfectants: normoxia: + Tet vs. −Tet, hypoxia: + Tet vs. −Tet), and importantly, restored the hypoxia-associated inducibility of VEGF expression (Fig. 5C, P < 0.0001, Sox9 shRNA transfectants: + Tet: hypoxia vs. normoxia) in this cell line. Migration and invasion is an important component in tumor aggressiveness. A significant reduction in cell migration was observed after inhibiting SOX9 expression (Fig. 5D, P < 0.0005).
These phenomenon were then evaluated in an in vivo setting. The cell lines were implanted orthotopically in nude mice. Tetracycline-induced inhibition of SOX9 expression in HMET cells in vivo led to a significant reduction in primary tumor growth and metastasis as compared with control tumors (Fig. 5E and Table 1). Following implantation of HMET, LacZshRNA cells developed large tumors [average volumes 2.1 (±0.3) cm3 −tet, 2.2 (±0.4) cm3 +tet] and high metastatic activity (table 1). Animals with implanted HMET Sox9sh cells exhibited substantially [0.8 (±0.4) cm3] and statistically significant (P < 0.01) reduction of primary tumor volume and incidence of liver metastasis (Table 1).
. | . | . | Metastasis . | . | |
---|---|---|---|---|---|
. | Tetracycline . | n . | Liver . | Lymph nodes . | Death . |
LacZ shRNA | − | 10 | 9 | 10 | 2 |
+ | 10 | 9 | 10 | 1 | |
Sox9 shRNA | − | 10 | 7 | 9 | 2 |
+ | 10 | 0 | 2 | 0 |
. | . | . | Metastasis . | . | |
---|---|---|---|---|---|
. | Tetracycline . | n . | Liver . | Lymph nodes . | Death . |
LacZ shRNA | − | 10 | 9 | 10 | 2 |
+ | 10 | 9 | 10 | 1 | |
Sox9 shRNA | − | 10 | 7 | 9 | 2 |
+ | 10 | 0 | 2 | 0 |
NOTE: HMET stable transfectants expressing Sox9 shRNA or E. coli LacZ shRNA in the presence of tetracycline. Tetracycline treatment, total number of mice, and the incidence of liver or lymph node metastases are indicated.
Histologic sections prepared from the HMET tumors were stained for antigen Ki-67 specific for proliferating cells and counterstained using hematoxylin and eosin. A significant reduction in proliferation activity of the HMET cells in vivo was seen after inhibition of SOX9 expression (proliferative index: 0.73 vs. 0.27). Expression of unspecific shRNA showed no significant effect on proliferation of the tumor cells in vivo (proliferative index: 0.82 vs. 0.85; Fig. 5F). The tissue sections were then fluorescent stained using Terminal deoxynucleotidyl transferase dUTP nick end labeling. An increased apoptosis was seen in tumors expressing Sox9shRNA construct after induction by tetracycline. Negative control cells expressing non-relevant shRNA showed no significant effect on apoptosis (Supplementary Fig. S1F).
Clinical significance of SOX9 expression
To examine the potential role of Sox9 in clinical settings, paired tumor and normal tissue samples obtained from the same human pancreas were analyzed for expression of VEGF and SOX9 genes. In 22 of 28 analyzed tissue samples, significant upregulation of SOX9 expression was observed in pancreatic cancer in comparison to normal tissue. Analysis of VEGF transcription in the same samples revealed the significant upregulation in 23 of 28 tumor compared with normal samples (Fig. 6A).
To further validate the clinical role of SOX9 at the level of promoter methylation, chromosomal DNA was isolated from the same tissue samples used for investigation of Sox9 and VEGF transcription described earlier, and analyzed for SOX9-promoter methylation. SOX9-promoter methylation was increased in normal tissue as compared that seen in corresponding tumor tissues in 9 of the 15 investigated paired samples (Supplementary Fig. S5A and S5B). Three samples exhibited increased SOX9-promoter methylation in cancer tissues in comparison to normal tissue. Interestingly, SOX9 or VEGF expression is in these samples stronger in normal tissue or they show no significant difference in both paired samples (Supplementary Fig. S5B). Elevation of VEGF expression (tumor vs. normal tissue) in patient samples strongly correlates with the elevation of SOX9 expression in tumor samples (Fig. 6B). We have observed less significant (correlation coefficient: 0.7) negative correlation between SOX9 promoter methylation and SOX9 expression (Fig. 6C).
Discussion
SOX9 expression was linked to aggressive pancreatic tumor phenotypes. SOX9 expression was found to cluster with constitutively upregulated genes in the HMET cells under normoxic conditions (but not hypoxia). Genes within the clusters were demonstrated to be differentially responsive to hypoxia and to be targets of SOX9 regulation. Four regulatory pathways associated with tumorigenesis: the CXCR4 pathway, focal adhesion pathway, retinoic acid pathway, and WNT pathways were found to be overrepresented among SOX9 target genes. Sox9shRNA inducible transfectants in the HMET cell line showed reduced expression of genes identified in the HMET cluster suggesting a central role for SOX9 in pancreatic tumor progression (Fig. 5 and Supplementary Fig. S1E; refs. 19–21).
Tumor progression is associated with hypoxia. HIF helps to mediate responses to hypoxia (22). VEGF family members are commonly overexpressed in human pancreatic cancer and contribute to tumor growth and metastasis (23). We recently demonstrated that the metastatic potential of a human pancreatic cancer cell line HMET as compared with its nonmetastatic derivative LMET that was associated with increased expression of proangiogenic molecules, in particular, VEGF. As expected, VEGF RNA expression and protein production was inducible by hypoxia in LMET cells (reviewed in ref. 24). Interestingly, in HMET cells VEGF expression was constitutively expressed under normoxia and not inducible by hypoxia. An increase in the biological aggressiveness of HMET cells in vivo was previously reported as compared with LMET cells based on the development of an angiogenic phenotype and an increase of their metastatic potential (8).
Transcriptomic profiling defined genes that coclustered in HMET cells with VEGF. These genes were found to be inducible by hypoxia in LMET cells, whereas in HMET cells the same genes were constitutively upregulated under normoxia and not inducible by hypoxia (Supplementary Fig. S1A). Presumably, the aggressive HMET pancreatic tumor cells gained via in vivo selection the capability to overexpress hypoxia inducible genes under normoxic conditions.
The promoter sequence analysis of the gene cluster revealed a significant enrichment for the binding site of the transcription factor Sox9 (14). We then hypothesized that SOX9-dependent transcriptional regulation of hypoxia-associated genes under normoxic conditions could be responsible for the aggressive biology of HMET-induced pancreatic cancer, and may therefore contribute to aggressive tumor stages of pancreatic cancer. A series of survival and migration-related genes are found in the gene clusters, exhibiting the differential hypoxic response in LMET versus HMET cells, suggesting a potential role for these gene products in the observed phenotype (Table 2). Although HIF-1α is a major factor driving hypoxia-inducible genes in pancreatic tumors under oxygen deprivation (25), SOX9-based transcription of hypoxia-associated genes may represent an alternative pathway for the induction of these tumor-critical genes in the absence of hypoxia. Knockdown of SOX9 led to restoration of HIF-1α–driven hypoxic inducibility of the downstream target genes (Fig. 5C for VEGF), suggesting overlap of activity between these transcription factors.
Gene . | Migration . | Survival . | HMET/LMET (Hpx) . | HMET/LMET (Nmx) . |
---|---|---|---|---|
MAP2 | − | − | 0.395 | 0.082 |
SYNJ2 | + | 2.6 | 3.18 | |
ETV1 | + | + | 4.29 | 5.7 |
SEMA3A | + | 2.2 | 7.16 | |
SEMA3C | + | 3.6 | 6.87 | |
WNT11 | − | + | 4.85 | 5.43 |
FOXC1 | + | + | 2.04 | 8.39 |
ITGA6 | + | + | 6.36 | 10.63 |
HAS3 | + | 4.66 | 3,691 | |
ERF | + | − | 0.28 | 0.17 |
SMAD7 | − | − | 0.19 | 0.0011 |
ITGB8 | + | 2.97 | 9.78 | |
FOXP1 | − | 0.44 | 0.46 | |
LAMA3 | + | 2.3 | 2.38 | |
ARHGAP21 | − | 0.3 | 0.22 | |
HBEGF | + | + | 16.22 | 95.01 |
RASSF8 | + | 10.34 | 16.34 | |
CPEB4 | − | 0.22 | 0.24 | |
PCDH17 | − | 0.017 | 0.087 | |
TBCCD1 | + | 2.71 | 6.62 | |
CALD1 | − | 0.31 | 0.22 | |
SGPP1 | − | 0.29 | 0.28 | |
EREG | + | + | 26.72 | 26.17 |
NOV | + | 19.56 | 6.49 | |
SSH2 | + | 2.3 | 2.53 | |
THBS1 | − | − | 0.37 | 0.31 |
GRB7 | + | + | 2.38 | 3.46 |
EHF | + | 3.68 | 9.38 | |
TPM1 | − | 0.12 | 0.0063 | |
DUSP6 | + | + | 2179.23 | 51,418 |
CAST | + | 2.53 | 4.08 | |
TIAM1 | + | + | 4.79 | 4.82 |
PIK3R1 | + | + | 3.43 | 5.66 |
PTPRR | − | 0.000006 | 0.0013 | |
LATS2 | − | 0.4 | 0.36 | |
PPP2R2B | − | 0.28 | 0.13 | |
SSBP3 | − | 0.45 | 0.37 | |
HHEX | + | 5.62 | 9.32 | |
DEDD2 | − | 0.23 | 0.21 | |
DOCK1 | + | 2.23 | 2.75 | |
TNIK | + | 3.43 | 3.14 | |
ATF3 | − | 0.44 | 0.0093 | |
SH3RF1 | − | 0.36 | 0.19 | |
DLC1 | − | − | 0.34 | 0.4 |
SMAD3 | + | + | 4.23 | 5.94 |
IGFBP6 | + | + | 14.22 | 30.48 |
HIF1AN | + | 2.57 | 3.74 | |
ETV5 | + | 5.43 | 8.57 |
Gene . | Migration . | Survival . | HMET/LMET (Hpx) . | HMET/LMET (Nmx) . |
---|---|---|---|---|
MAP2 | − | − | 0.395 | 0.082 |
SYNJ2 | + | 2.6 | 3.18 | |
ETV1 | + | + | 4.29 | 5.7 |
SEMA3A | + | 2.2 | 7.16 | |
SEMA3C | + | 3.6 | 6.87 | |
WNT11 | − | + | 4.85 | 5.43 |
FOXC1 | + | + | 2.04 | 8.39 |
ITGA6 | + | + | 6.36 | 10.63 |
HAS3 | + | 4.66 | 3,691 | |
ERF | + | − | 0.28 | 0.17 |
SMAD7 | − | − | 0.19 | 0.0011 |
ITGB8 | + | 2.97 | 9.78 | |
FOXP1 | − | 0.44 | 0.46 | |
LAMA3 | + | 2.3 | 2.38 | |
ARHGAP21 | − | 0.3 | 0.22 | |
HBEGF | + | + | 16.22 | 95.01 |
RASSF8 | + | 10.34 | 16.34 | |
CPEB4 | − | 0.22 | 0.24 | |
PCDH17 | − | 0.017 | 0.087 | |
TBCCD1 | + | 2.71 | 6.62 | |
CALD1 | − | 0.31 | 0.22 | |
SGPP1 | − | 0.29 | 0.28 | |
EREG | + | + | 26.72 | 26.17 |
NOV | + | 19.56 | 6.49 | |
SSH2 | + | 2.3 | 2.53 | |
THBS1 | − | − | 0.37 | 0.31 |
GRB7 | + | + | 2.38 | 3.46 |
EHF | + | 3.68 | 9.38 | |
TPM1 | − | 0.12 | 0.0063 | |
DUSP6 | + | + | 2179.23 | 51,418 |
CAST | + | 2.53 | 4.08 | |
TIAM1 | + | + | 4.79 | 4.82 |
PIK3R1 | + | + | 3.43 | 5.66 |
PTPRR | − | 0.000006 | 0.0013 | |
LATS2 | − | 0.4 | 0.36 | |
PPP2R2B | − | 0.28 | 0.13 | |
SSBP3 | − | 0.45 | 0.37 | |
HHEX | + | 5.62 | 9.32 | |
DEDD2 | − | 0.23 | 0.21 | |
DOCK1 | + | 2.23 | 2.75 | |
TNIK | + | 3.43 | 3.14 | |
ATF3 | − | 0.44 | 0.0093 | |
SH3RF1 | − | 0.36 | 0.19 | |
DLC1 | − | − | 0.34 | 0.4 |
SMAD3 | + | + | 4.23 | 5.94 |
IGFBP6 | + | + | 14.22 | 30.48 |
HIF1AN | + | 2.57 | 3.74 | |
ETV5 | + | 5.43 | 8.57 |
NOTE: Relevance to cell migration and survival based on published data are indicated (+, increase; −, decrease) as well as differential expression shown as fold change between average value in HMET/LMET under normoxia or hypoxia, respectively.
The concept of “stemness” has been linked to the regulation of tumor aggressiveness, including metastasis, chemoresistance, and adaptation to hypoxia. To date, the transcription factor SOX9 has not been associated with pancreatic cancer development and progression. Galmiche and colleagues described cytoplasmatic expression of SOX9 and PDX1 associated with β-catenin mutations in pediatric solid pseudopapillary tumors (26). The authors suggested that these tumors may originate from the transformation of normally quiescent pancreatic stem cells (26). Results from the ChIP analysis presented here showed in vivo binding of SOX9 to the β-catenin gene promoter (Supplementary Fig. S1B). Seymour and colleagues described SOX9 as a specific marker and maintenance factor for multipotential progenitors during pancreas organogenesis (7). Genes identified in their study represent downstream targets of SOX9 regulation and may therefore be active in the establishment of a tumor progenitor phenotype. In this study, we identified SOX9 target genes exhibiting differential hypoxic responsiveness to hypoxia in HMET and specifically belonging to the WNT and CXCR4 pathways. The constitutive normoxic upregulation of SOX9 may thus have a direct effect on tumor growth, invasiveness, and metastasis.
The knockdown of SOX9 led to restoration of oxygen sensitivity presumably because of a reactivated HIF-1α pathway. Our data also show that SOX9 expression does not seem to interfere with HIF-1α gene regulating machinery (Supplementary Fig. S4). The stability of HIF-1α protein is regulated the same regardless of the intracellular concentration of SOX9. SOX9 activation seems to have a stronger effect on the transcriptome and therefore can overcome hypoxic HIF-1α–based hypoxic regulation.
Four pathways associated with tumorigenesis were linked to SOX9 expression: the CXCR4 pathway, focal adhesion pathway, retinoic acid, and WNT pathways (Supplementary Fig. S6). Sox9 is a positive regulator of expression of CXCR4 from CXCR4 pathway, FN1, GRB7, PAK3, PIK3R1 from the focal adhesion pathway, RHD10, RETSAT, and RXRA from the retinoic pathway, and SMAD3 and FZD7 from the canonical WNT pathway. SOX9 was also shown to be a negative regulator of FYN and RHOB expression in CXCR4 pathway, and WNT11 from the WNT pathway. FYN interacts with FAK (27), a kinase that plays a central role in directed cell movement. The Akt pathway focuses survival signals from other pathways regulated by SOX9: focal adhesion pathway via at least GRB7, PAK3, and PIK3R1gene. All-trans-retinoic acid is an important regulator of epithelial differentiation, embryogenesis, and cancer (28). RXRA as well as at least 2 genes (RDH10 and RETSAT) linked to the biosynthesis of retinoic acid were identified as targets of SOX9 regulation in this setting (Fig. 5A). Potential points of cross-talk between the regulatory pathways are depicted in Supplementary Fig. S6.
Oxygen presence affected methylation of the SOX9 promoter in an inverse manner in relation to gene expression: hypoxia-reduced methylation induced downregulation of SOX9 expression in LMET cells. Interestingly, methylation of the SOX9 promoter in HMET cells was only weakly sensitive to the oxygenation status of the cells analogous to the marginal alterations in gene expression of SOX9 itself and other genes of the gene cluster (e.g., VEGF) (Supplementary Fig. S2B compare with Supplementary Fig. S2B). These data show that differential methylation of the SOX9 promoter is at least in part responsible for the differential expression of SOX9 in LMET versus HMET cells. Finally, increased expression of SOX9 and VEGF mRNA was found in human pancreatic adenocarcinoma specimens as compared with normal pancreatic tissue (Fig. 6), suggesting that SOX9 may help drive expression of VEGF in vivo. Differential methylation of the SOX9 promoter may be responsible in part for the regulation of SOX9 seen. A reduction in SOX9-promoter methylation correlated with increased expression of SOX9 and VEGF expression in patient tumor samples.
We are aware that our experimental model is dealing with one couple of isogenic cell lines supported with the results of animal experiments. Because we assume that the mechanism of overcoming of HIF-1α regulation via SOX9 has more general significance, we tried to verify our data obtained on cell lines LMET and HMET in paired samples of primary tumor versus metastasis. We searched for gene array data of analyzed patient's specimen comparing primary pancreatic tumors and metastasis for SOX9-based regulation patterns. In our Supplementary Material, we added the respective results of the GEO database from the manuscript of Mozon and colleagues (29). Unfortunately, the majority of such data were generated on gene array platforms incompatible to the one used in this manuscript. However, a small amount of gene array data generated on a HG-U133A platform revealed an upregulation of CXCR4 and LAMC genes as well as a downregulation of FYN, PIK3R1, ARHGAP, COL3A1, and Sox2 in metastases in comparison to primary tumors. We obtained similar data using LMET and HMET cells. Yu and colleagues (30) demonstrated by RNA sequencing of pancreatic circulating tumor cells that the WNT and CXCR4 signaling are important pathways for the establishment of metastases comparable to own results about the genes WNT11, CXCR4, and CTNNB1. A recently published report demonstrated the upregulation of SOX9 in colorectal carcinoma samples in comparison to normal tissue specimens. This regulation correlated with upregulation of β-catenin and PPAR-γ expression in cell cultures (31). Both those genes are members of the SOX9 target gene cluster identified in this study. In case of β-catenin, a direct binding of SOX9 to the promoter was demonstrated via ChIP (Supplementary Fig. S1B). The data presented here support the general relevance of SOX9 regulation for overcoming the lack of HIF-1α–based hypoxic induction of gene expression in micrometastasis followed persistent regulation of SOX9 target genes in metastases. SOX9 may represent an important new therapeutic target for the treatment of pancreatic carcinoma.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: P. Camaj, P.J. Nelson, C.J. Bruns
Development of methodology: P. Camaj, E.N. DeToni, C.J. Bruns
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Camaj, S. Krebs, H. Blum, K.-W. Jauch, C.J. Bruns
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Camaj, C. Jäckel, S. Krebs, H. Blum, K.-W. Jauch, P.J. Nelson, C.J. Bruns
Writing, review, and/or revision of the manuscript: P. Camaj, E.N. DeToni, P.J. Nelson, C.J. Bruns
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.J. Bruns
Study supervision: P.J. Nelson, C.J. Bruns
Acknowledgments
The authors thank M. Hofstetter (Clinical Biochemistry Group, Medizinische Klinik und Poliklinik IV, University of Munich, Germany) and A. Tischmacher (Department of Surgery, Department of Gastroenterology, Munich University Medical Center, Munich, Germany) for the excellent technical assistance.
Grant Support
This work was supported by grants of Deutsche Forschungsgemeinschaft NE 648/2-4 to P.J. Nelson and by grants from the Deutsche Forschungsgemeinschaft (DFG BR 1614/7-1 to C.J. Bruns; DFG SPP1190 “Tumor Vessel Interface” to C.J. Bruns and P.J. Nelson).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.