The first 2 authors contributed equally to this article.
Prolyl hydroxylase-2 (PHD2) exerts tumor-suppressive activity in pancreatic cancer†
Version of Record online: 26 JUL 2011
Copyright © 2011 American Cancer Society
Volume 118, Issue 4, pages 960–972, 15 February 2012
How to Cite
Su, Y., Loos, M., Giese, N., Metzen, E., Büchler, M. W., Friess, H., Kornberg, A. and Büchler, P. (2012), Prolyl hydroxylase-2 (PHD2) exerts tumor-suppressive activity in pancreatic cancer. Cancer, 118: 960–972. doi: 10.1002/cncr.26344
- Issue online: 3 FEB 2012
- Version of Record online: 26 JUL 2011
- Manuscript Accepted: 16 MAY 2011
- Manuscript Revised: 10 MAY 2011
- Manuscript Received: 13 FEB 2011
- pancreatic cancer;
- prolyl hydroxylase-2;
- hypoxia-inducible factor
Pancreatic cancer is 1 of the most common and poorly treated tumors. In search of new therapeutic approaches, the oxygen sensors prolyl hydroxylases (PHD) are potential targets. PHD2 is considered the key oxygen sensor-regulating hypoxia-inducible factor (HIF). Currently, there is conflicting evidence regarding the exact role of PHD2 in tumorigenesis. The objective of this study was to investigate the role of PHD2 in pancreatic cancer growth and progression.
PHD2 expression was analyzed by quantitative real-time polymerase chain reaction analysis and immunohistochemistry in human tissue specimens and cell lines. Knockdown of PHD2 was done by using short-interfering RNAs (siRNAs) specific against PHD2, and PHD2 overexpression was achieved by stable combinational DNA transfection. In vivo, an orthotopic murine model was used. Angiogenic cytokines were assessed with enzyme-linked immunosorbent assays, and invasion was studied with Matrigel assays.
PHD2 expression was not altered substantially in cancer tissues and their metastases. Lymph node-negative tissues had higher levels of PHD2 than lymph node-positive tissues. PHD2 was hypoxia-inducible in pancreatic cancer cell lines and regulated cell growth through cyclin D1 down-regulation samples with PHD2 suppression and through p21 up-regulation in samples with of PHD2 overexpression. In vivo, PHD2 caused tumor growth retardation and reduced tumor invasion by inhibiting angiogenesis. This observation was caused by the suppression of angiogenic cytokines and tumor invasion.
The current results indicated that PHD2 plays an important role in pancreatic tumorigenesis. In summary, the authors concluded that PHD2 may function as a tumor suppressor gene in pancreatic cancer and, thus, may define a potential target for the treatment of pancreatic cancer. Cancer 2012;. © 2011 American Cancer Society.
Accurate regulation of oxygen homeostasis is essential for survival in health and disease.1 In the context of tumor expansion, excessive tumor cell proliferation that distances cells from oxygen-rich blood vessels rapidly outstrips the supply of nutrients.2 Oxygen sensing, therefore, is a central mechanism of tumor growth and vasculogenesis, because pancreatic cancer grows in extensive areas of tumor hypoxia.3 At the heart of this regulatory system is hypoxia-inducible factor (HIF), which is activated in pancreatic cancer.4 This finding has placed the hypoxia-signaling pathway at the forefront of nutritional control and ongoing tumor growth. HIF itself does not directly sense variations in oxygen tension (pO2).5 The prolyl hydroxylase domain (PHD) proteins represent the true oxygen-sensing molecules, and their activity depends strictly on cellular pO2.6-8
PHDs hydroxylate HIF-1/HIF-2α at 2 specific prolyl residues in the oxygen-dependent degradation domain (P402 and/or P564), allowing capture by the von Hippel-Lindau tumor suppressor protein (pVHL), ubiquitination, and subsequent proteasomal degradation.9 Of the 3 PHD isoforms in humans, PHD2 is the key limiting enzyme that targets HIF-1α for degradation under normoxic conditions, whereas the physiologic roles of PHD1 and PHD3, which are active under chronic hypoxia, remain to be investigated.10
It is believed commonly that HIF promotes tumor growth, but it also is known that the HIF pathway exerts proapoptotic activities, suggesting that HIF function in tumors may be more complex.11 Indeed, several studies have reported conflicting data with regard to the effect of HIF manipulation on tumor growth. Some have reported impaired tumor growth and angiogenesis after HIF-1α loss of function.12, 13 In contrast, others have provided evidence for the tumor suppressive function of HIF.14, 15 Even less is known about the function of PHD2 in human cancer growth. In vitro, PHD2 biphasically modulates tumor formation. In MSU-1 (fibroblast) cells, an inverse correlation was reported between PHD2 expression and tumor-forming potential.16 A small decrease of PHD2 levels in cell lines caused malignant transformation, whereas severe decreases of PHD2 did not. Other authors postulated that PHD2 acts as a tumor suppressor gene.17, 18 The role of PHD2 in tumor angiogenesis also is controversial: Some reported that tumor vasculature is regulated by PHD2,19 whereas others observed that reduced PHD2 expression led to the normalization of blood vessels with reduced tumor metastasis.20 In the current study, we analyzed PHD2 expression in large number of pancreatic cancer tissues and cell lines and studied the possible regulatory function of PHD2 in tumor growth and progression.
MATERIALS AND METHODS
All tissue specimens that were analyzed in this study were obtained according to institutional review board-approved procedures for consent. Human pancreatic cancer tissue samples were obtained from 62 patients (30 women and 32 men; median age, 64.3 years; age range, 39-82 years) who underwent pancreatic resection for pancreatic cancer at the University Hospital of Berne (Berne, Switzerland) or Heidelberg (Heidelberg, Germany). Three patients presented with stage I disease, 35 patients presented with stage II disease, 23 patients presented with stage III disease, and 1 patient presented with stage IV disease according to the sixth edition (2002) of the International Union Against Cancer classification. Normal human pancreatic tissue samples were obtained from 20 previously healthy individuals (median age, 45 years; age range, 18-74 range) in an organ donor program who had other organs taken for transplantation when no recipients for the pancreas were present. Immediately after surgical removal, tissue samples were either snap-frozen in liquid nitrogen (for RNA and protein extraction) or fixed in 5% formalin and embedded in paraffin 24 hours later (for histologic analysis).
For our experiments, we used the established human pancreatic cancer cell lines. AsPc-1, HPAF-II, MIA PaCa-2, and PANC-1 cells were purchased from the American Type Culture Collection (Rockville, Md). The parental mouse hepatoma cell line Hepa-1c1c7 and the derived mutant c4 subclone deficient for an obligatory component of the HIF-1 heterodimer, HIF-1β, were described previously.21, 22 The c4 cell line carries a mutated period circadian protein-aryl hydrocarbon nuclear translocator protein-single-minded protein (PAS) region of the Ah receptor nuclear translocator (ARNT) gene, causing impaired hypoxic induction of HIF binding to DNA.
RNA Preparation and Real-Time Quantitative Polymerase Chain Reaction
Messenger RNA was prepared by automated isolation using MagNA Pure LC instrument and isolation kits as reported previously (Roche Applied Science, Mannheim, Germany).23 Complementary DNA (cDNA) was prepared using the first-strand cDNA Synthesis Kit for reverse transcriptase-polymerase chain reaction (RT-PCR) according to the manufacturer's instructions. Real-time PCR was performed with the LightCycler FastStart DNA SYBR Green kit as described previously. The number of specific transcripts was calculated from the standard curve of 2 housekeeping genes, cyclophilin B and hypoxanthine phosphoribosyl transferase. Data from 2 independent analyses for each sample and parameter were averaged and are presented as adjusted transcripts per μL cDNA.
Western Blot Analysis
The following other primary antibodies were used: mouse monoclonal antibodies against cyclin D1 (Zymed Laboratories, Carlsbad, Calif), retinoblastoma (Rb) (Pharmingen-BD Biosciences, San Jose, Calif), rabbit polyclonal antibodies against cyclin D1 (Millipore Corporation, Billerica, Mass) and against cyclin D3 (C-16), cyclin E (M-20), p21, and p27 (Santa Cruz Biotechnology, Heidelberg, Germany). Cells were lysed in suspension buffer (50 mM Tris/HCl, 150 mM NaCl, 2 mM ethylene diamine tetra acidic acid, and 1% sodium dodecyl sulfate) that contained complete protease inhibitor cocktail tablets from Roche Applied Science. Protein concentrations were measured with a bicinchoninic acid protein assay (Pierce Chemical Company, Rockford, Ill). The cell lysates (30 μg aliquots) were separated on sodium dodecyl sulfate-polyacrylamide gels and electroblotted onto nitrocellulose membranes. Membranes were then incubated in blocking solution (5% nonfat milk in 20 mM Tris/Cl, 150 mM NaCl, and 0.1% Tween-20 [TBS-T]), followed by incubation with primary antibodies. PHD2 and HIF-1α antibodies (1:1000 dilution; Novus Biologicals, Cambridge, United Kingdom) were incubated at 4°C overnight. The membranes then were washed in TBS-T and incubated with horseradish peroxidase-conjugated secondary antibodies (Amersham Life Science, Buchinghamshire, United Kingdom) for 1 hour at room temperature. Antibody detection was performed with an enhanced chemiluminescence reaction (Amersham Life Science).
Immunohistochemistry was performed using the Dako Envision Systems (Dako Cytomation GmbH, Hamburg, Germany). Consecutive sections (3-5 μm thick) were dewaxed and rehydrated. Antigen retrieval was done in citrate buffer, pH 6.0, in a microwave oven for 10 minutes. Thereafter, slides were cooled and washed in washing buffer (Tris-buffered saline with 0.1% bovine serum albumin) for 10 minutes. Endogenous peroxidase activity was quenched; and the slides were incubated first for 30 minutes at room temperature with normal goat serum and then at 4°C overnight with the primary antibodies. Thereafter, the sections were rinsed with washing buffer and incubated with horseradish peroxidase-linked goat-antirabbit antibodies, followed by reaction with diaminobenzidine and counterstaining with Mayer hematoxylin. In addition, to confirm the specificity of the primary antibodies and the technique used, tissue sections were incubated in the absence of the primary antibodies and with negative control rabbit immunoglobulin G. Under these conditions, no specific immunostaining was detected.
Silencing of PHD2 by Short-Interfering RNA
Cells were 50% confluent and transfected with 5.0 nM PHD2 short-interfering RNA (siRNA) at 24 hours and 48 hours using Oligofectamine (Invitrogen GmbH, Darmstadt, Germany) according to the manufacturer's instructions. The negative control siRNA (Qiagen GmbH, Hilden, Germany) was used as negative control for all experiments. Sequences of PHD2 siRNA (Genbank accession number NM_022051) corresponded to nucleotides 3901 through 3921 (aaggacatccgaggcgataag) and nucleotides 4077 through 4097 (aacgggttatgtacgtcatgt).
Generation of PHD2-Overexpressing MIA PaCa-2 and Panc-1 Cells
MIA PaCa-2 and PANC-1 cells were converted into PHD2-expressing cells by stable transfection. The Malian expression plasmid enhanced green fluorescent protein-N1 (pEGFP-N1) (Clontech, Palo Alto, Calif) carrying full-length PHD2 cDNA was transfected into MIA PaCa-2 and PANC-1 cells using TransFectin Lipid Reagent (Bio-Rad Laboratories, Munich, Germany) and selected in G418-containing culture medium. Empty pEGFP/N1 plasmids were transfected as a negative control.
Cell Proliferation Assays
To analyze DNA synthesis as an index of cellular proliferation, MIA PaCa-2 and PANC-1 cells were plated onto 48-well plates (5000/cm2) in growth medium, incubated overnight, and serum-deprived (1% fetal calf serum) for 24 hours. Replicate wells were then stored at −70°C for baseline (day 0) cell counts, and fresh medium with or without growth factors was added to the remaining wells, which were incubated for 72 to 96 hours in 20% or 5% O2. Day 0 and day 3 or 4 cell counts were determined by lysing cells in a buffer that contained a fluorescent dye, which, alone, has minimal fluorescence but fluoresces when bound to DNA or RNA (Invitrogen GmbH). Absolute cell numbers were calculated by comparing the fluorescence of specimens with that of a standard curve similarly prepared using a known number of cells.
Determination of vascular endothelial growth factor (VEGF), interleukin 8 (IL-80, and angiopoietin-1 (Ang1) protein levels in cell supernatants was done according to the manufacturer's instructions using an enzyme-linked immunosorbent assay kit (R&D Systems GmbH, Wiesbaden, Germany; Thermo Fisher Scientific, Bonn, Germany). Protein levels were quantified (pg/mL) and compared with the expression level of control experiments, which was set at 1 pg/mL.
Cell invasion was studied using the BioCoat Matrigel Invasion Chambers (Becton Dickinson, Bedford, Mass) according to the manufacturer's instructions. 2.5 × 104 cells were seeded into the upper chamber of the invasion chambers and incubated for 24 hours. For hypoxia studies, the BioCoat Matrigel invasion chamber was placed in the hypoxic chamber, which was flushed for 20 minutes with a gas mixture consisting of 0.75% O2, 10% CO2, and 89.25% N2. The noninvasive cells were removed from the upper surface of the membrane by wiping with a cotton-tipped swab. Cells that adhered to the lower surface were fixed in 75% methanol mixed with 25% acetone and then stained with 1% Toluidine blue. Cells were counted under a microscope at ×200 magnification. The invasion index was expressed as the ratio between the number of invaded test cells to the number of invaded control cells.
Five-week-old male nude mice (BALB/cA) were used for subcutaneous and orthotopic tumor implantation. The experimental protocol was approved by the Chancellor's Animal Research Committee of the University of Heidelberg, Germany, in accordance with the national guidelines for animal care and use of laboratory animals. After a 4-week period of subcutaneous tumor formation, 1 small tumor fragment from the external rim of the growing subcutaneous tumor was used for transplantation. Orthotopic transplantation was performed according to the method described by Fu with minor modifications, as previously described.24, 25 A laparotomy was performed, the spleen with tail of the pancreas was exteriorized, and a pancreatic parenchyma pocket was created using microscissors. One tumor piece was implanted into the pancreatic parenchyma. After transplantation, mice were inspected daily and were killed 6 weeks after orthotopic tumor induction. All animals underwent autopsies. The tumor size was measured with a caliper in 3 perpendicular dimensions. The tumor volume was calculated as described previously using the following formula: tumor volume = 0.5 × (length × width × depth).25 Then, the mice were killed, and xenograft tumors were harvested and immediately snap-frozen or fixed in formalin for further analysis.
Assessment of Microvessel Density
To assess vascularity, frozen sections, each 5 μm thick, were cut, and endothelial cells were immunostained with antimouse CD31 antibody (Pharmingen-BD Biosciences) as described previously.24 Stained tissue specimens were analyzed by 2 independent observers who were unaware of the animal's status. Microvessel density was determined as described by Vermeulen et al.26 The highest area of vascularization was determined by scanning each slide at low-power magnification (×100). Each slide was scanned first at low magnification (×100) to identify the 5 areas with the highest density of microvessels; then, each field was evaluated at high-power magnification (×200), and the number of stained microvessels per high-power field was determined. Any brown-stained endothelial cell was considered a single, countable microvessel.
Experiments were done in triplicate and were repeated at least twice. The results are expressed as means ± standard errors. Statistical significance was determined with the Student t test (P < .05).
PHD2 Expression in Human Pancreatic Cancer
PHD2 messenger RNA (mRNA) expression was determined by quantitative real-time RT-PCR in 20 normal pancreatic tissue samples, 62 pancreatic cancer tissue samples, 16 specimens of liver metastases, and 10 lymph node metastases that originated in pancreatic cancer (Fig. 1). The mean PHD2 mRNA expression in normal pancreatic tissue specimens was 996 ± 95 mRNA copies per μL and was slightly decreased in pancreatic cancer tissues to 854 ± 57 mRNA copies per μL. Similarly, there was no change in PHD2 mRNA expression in liver metastases (931 ± 69 mRNA copies per μL) or in lymph node metastases (794 ± 95 mRNA copies per μL) from pancreatic cancer. However, a gradual loss of PHD2 mRNA expression was observed in the less differentiated tumor samples (780 ± 46 copies per μL) compared with the well differentiated tumor specimens, in which PHD2 mRNA expression was maintained at higher levels (1538 ± 289 mRNA copies per μL) (Fig. 1B). Similarly, the mean PHD2 mRNA expression was higher in lymph node-negative tumor specimens (1067 ± 175 mRNA copies per μL) PHD2 mRNA than in lymph node-positive specimens, which had a mean of 791 ± 51 mRNA copies per μL PHD2 mRNA moieties (Fig. 1C).
Localization of PHD2 expression in human tissue specimens revealed that pancreatic cancer cells expressed PHD2 (Fig. 1D,E), but stromal cells and the endothelial linings of blood vessels (Fig. 1F) also were positive for PHD2 in immunohistochemical analyses. Normal pancreatic tissues and normal-appearing pancreas adjacent to cancer tissues also were stained positive for PHD2 (Fig. 1F).
PHD2 Is a Hypoxia-Inducible Gene in Pancreatic Cancer Cells
We used quantitative real-time RT-PCR and Western blot analysis to investigate the expression of PHD2 in pancreatic cancer cell lines. PHD2 protein (approximately 48 kD) was expressed constitutively in normoxic pancreatic cancer cells and was readily up-regulated in whole-cell extracts from cells that had been exposed to hypoxia for 16 hours (Fig. 2). The level of protein expression differed markedly between pancreatic cancer cell lines, with higher expression levels observed in AsPC-1 and HPAF-II cells and lower expression levels observed in MIA PaCa-2 and PANC-1 cells. PHD2 mRNA levels were inducible by severe hypoxia for 16 hours with a greater than 2-fold induction of PHD2 mRNA expression under low oxygen conditions (Fig. 2)
PHD2 Knockdown Inhibits Cell Growth Through the Suppression of Cyclin D1 Expression
Knockdown of PHD2 was done using previously validated, PHD2-specific siRNAs. The cellular effects of PHD2 knockdown were studied in 2 cell lines with constitutive PHD2 expression (AsPC-1 and HPAF-II). Down-regulation of PHD2 reduced cell growth in AsPC-1 and HPAF-II cells compared with control experiments using scrambled siRNAs and untreated cells, which were set at 1.0 pg/mL (Fig. 3A). The growth-inhibitory effect also was observed in 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assays, in which PHD2 knockdown reduced cell growth under both normoxic and hypoxic conditions (Fig. 3B). Overall growth inhibition was more pronounced under low oxygen conditions. In flow cytometric analyses, a moderate accumulation of cells in G1-phase was observed (data not shown). Because the G1/S transition is closely linked to cyclin D1 levels, we tested weather cyclin D1 expression was affected by PHD2. In Western blot analyses, a decrease in cyclin D1 protein expression was detectable under normoxic conditions (Fig. 3C). No changes were observed for cyclin D2 or cyclin D3. Hypoxia itself also down-regulated the expression of cyclin D1 (data not shown). No changes in expression of cyclin-dependent protein kinase inhibitors p16, p21, p27 or pRb were observed (not shown). To test whether the modulation of cyclin D1 expression is mediated by HIF-1 activation, we used an HIF-1-deficient mouse hepatoma cell line that carries a mutated PAS region of the ARNT gene, causing impaired hypoxic induction of HIF binding to DNA. We transfected Hepa-1c1c7 parental cells and their derived HIF-1β-deficient subclone c4 with siRNA directed against PHD2. Down-regulation of cyclin D1 after knockdown of PHD2 expression was not affected by concurrent inactivation of HIF-1β suggesting that expression of cyclin D1 is modulated directly by PHD2 inactivation and is not mediated by HIF-1 activation (Fig. 3D).
PHD2 Overexpression Maintains Cell Homeostasis Under Hypoxic Conditions
To study the effect of PHD2 overexpression in cell lines with constitutively low PHD2 expression, we transfected PHD2 nonexpressing/low expressing MIA PaCa-2 and PANC-1 cells. Similar to the knockdown experiments, a reduction in cell growth was detectable in normoxic cell cultures (Fig. 4B). Under conditions of low oxygen, cell death was prevented by PHD2 expression. This hypoxic rescue activity of PHD2 overexpression also was confirmed by MTT assays, in which PHD2 expression maintained cell viability under hypoxic conditions (Fig. 4A,B). Similar to PHD2 knockdown, flow cytometry resulted in a G1-phase arrest under normoxic conditions (data not shown). In contrast to the knockdown of PHD2, no changes in cyclin D1, cyclin D2, or cyclin D3 expression levels were observed upon PHD2 overexpression. Because replicative senescence is a response to various stimuli, we determined variation in other cyclin-dependent kinase inhibitors, such as p16, p21, p27, and pRb levels. The p16 level was undetectable in both normoxic and hypoxic cell clones. No significant alterations to p27 or pRb protein levels or to the phosphorylation status of pRb were observed. Only the p21 cyclin-dependent kinase inhibitor was up-regulated in response to PHD2 transfection under conditions of low oxygen. This up-regulation was observed in both cell lines. The up-regulation of p21 probably was caused by HIF-1 activation, because p21 up-regulation was not present in ARNT-deficient cells (Fig. 4D).
In Vivo, PHD2 Overexpression Inhibited Tumor Growth and Angiogenesis in an Orthotopic Mouse Model
To determine the effect of PHD2 overexpression on tumor growth and angiogenesis in pancreatic cancer, we used an orthotopic murine model. Stably transfected MIA PaCa-2 and PANC-1 cells were injected into donor mice for subcutaneous tumor formation. After a 4-week period of subcutaneous tumor formation, 1 small tumor fragment was used for orthotopic tumor transplantation into the tail of the mouse pancreas. Six weeks after orthotopic tumor induction, the mice were killed. PHD2 overexpression resulted in reduced tumor growth compared with control xenograft tumors (P < .05). Because PHD2 serves as an oxygen sensor to mediate the proteolytic degradation of HIF-1α, it may affect the transcription of proangiogenic cytokines, including VEGF, Ang1, and IL-8. Therefore, the microvessel density of the harvested tumors was examined by staining tumors with anti-CD31. Tumors that resulted from the implantation of PHD2-overexpressing MIA PaCa-2 and PANC-1 cancer cells had reduced microvessel density compared with the microvessel density in control tumors. No difference was observed in metastatic scoring between PHD2-overexpressing tumors and control tumors. The local invasion index, which reflects local tumor aggressiveness and invasion into adjacent organs, was increased in control tumors (Fig. 5D).
PHD2 Regulates Invasiveness but Not VEGF-Mediated Mitogenesis
Previous studies have indicated that knockdown of prolyl hydroxylase activity with dimethyloxalylglycine (DMOG) inhibits fibroblast growth factor 2 and platelet-derived growth factor-induced cell proliferation of vascular smooth muscle cells under normoxic and hypoxic conditions. Therefore, we tested whether VEGF-induced cell proliferation is influenced by PHD2 knockdown as well DMOG. It is noteworthy that VEGF directly stimulates the growth of pancreatic cancer cells and, thus, is recognized as an autocrine growth factor for pancreatic cancer cells.27 A strong mitogenic effect of VEGF on pancreatic cancer cells was detectable (Fig. 6A), but DMOG was not able to inhibit VEGF-induced mitogenesis in pancreatic cancer, suggesting that PHD2 does not participate in VEGF signaling (Fig. 6A). There was a strong increase in tumor cell invasion determined in a Matrigel-based invasion assay in the case of PHD2 knockdown under both normoxic and hypoxic culture conditions (Fig. 6B). The most pronounced effect was observed under hypoxic culture conditions in AsPC-1 cells, which almost doubled the invasive phenotype upon PHD2 suppression (Fig. 6B). Conversely, PHD2 overexpression caused a slight decrease in tumor cell invasion, although the effect was less pronounced compared with PHD2 knockdown.
PHD2 Decreased Angiogenic Cytokines
PHD2-overexpressing xenograft tumors exhibited fewer blood vessels than control tumors. Nevertheless, the role of PHD2 in tumor angiogenesis remains controversial, because some have demonstrated its ability to suppress angiogenesis, whereas others have observed that PHD2 promotes angiogenesis. Therefore, we examined whether PHD2 modulation and/or hypoxia could influence expression levels of VEGF, Ang1, and IL-8. All of these have been reported as hypoxia-inducible genes, a finding that also was reproducible in pancreatic cancer cells (Fig. 7), and the involvement of PHD2 in the regulation of these genes was tested by cell transfection and knockdown experiments. Enzyme-linked immunosorbent assays were used to determine the levels of VEGF, Ang1 and IL-8. Hypoxia increased the expression of all 3 angiogenic cytokines to various extents (Fig. 7). PHD2 overexpression resulted in the suppression of VEGF and Ang1 (Fig. 7A,B). and IL-8 expression also was compromised under hypoxic conditions (Fig. 7E). However, the suppression of PHD2 expression led to the up-regulation of VEGF, Ang1, and IL-8 under both normoxic and hypoxic culture conditions (Fig. 7B,D,F). To test whether the regulation of angiogenic cytokines by PHD2 is mediated by HIF-1 activation, we transfected Hepa-1c1c7 and its HIF-1β-deficient subclone c4 with PHD2 expression plasmids or with siRNAs. Like what we observed in pancreatic cancer cells, PHD2 knockdown increased angiogenic cytokines in all cell lines tested (data not shown). These findings suggest that the observed regulation of angiogenic cytokines was independent of HIF-1.
In the current study, we have provided evidence that, at first glance, PHD2 influences tumor growth bidirectionally. PHD2 was expressed in normal pancreatic tissues and pancreatic cancer cells. In contrast to normal pancreatic tissues, PHD2 mRNA levels were decreased markedly in pancreatic cancer tissues. Primary tumor specimens from patients without lymph node metastases had higher PHD2 levels than those from patients with lymph node metastases. To date, the functional difference and biologic relevance of this observation remain unclear, because we could not identify PDH2 target genes involved in the process of lymph node metastasis. In the metastatic tumor specimens (liver, lymph nodes), PHD2 expression was equal to that in normal pancreatic tissues and cancer tissues. Low PHD2 mRNA levels have been correlated with poorer tumor differentiation. PHD2 immunohistochemistry produced a positive signal in ductal cancer cells, in normal acinar tissues, and in adjacent tumor blood vessels (Fig. 1D-F) without any obvious spatial restriction.
Among the pancreatic cancer cell lines, AsPC-1 and HPAF-II cells exhibited high basal PHD2 levels, whereas MIA PaCa-2 and PANC-1 cells expressed low PHD2 levels. Similar to a previous observations in HIF-1α, all pancreatic cancer cells had increased levels of PHD2 under hypoxic culture conditions, confirming that PHD2 is a hypoxia-inducible gene in pancreatic cancer.4 PHD2 knockdown inhibited pancreatic cancer cell growth under conditions of normoxia and even more under hypoxia. We have demonstrated that this effect probably was caused by the regulation of cyclin D1 leading to G1 cell cycle arrest. It has been demonstrated previously that cyclin D1 is down-regulated by hypoxia itself, without PHD2 manipulation. Therefore, PHD2 knockdown may mimic a “hypoxic cellular state” in an otherwise normoxic environment. In contrast to previous reports,28 the effect on the cell cycle was not caused by HIF-1 activation, because HIF-1β-deficient cells displayed the same mode of regulation as HIF-1-competent cells, suggesting that cyclin D1 is not regulated by HIF-1 as recently suggested.29
The overexpression of PHD2 slowed tumor cell growth. However, the mechanism was different from that observed in PHD2 knockdown, because the expression of cyclin D1, cyclin D2, and cyclin D3 was unaltered. PHD2 up-regulation, however, increased the cyclin-dependent kinase inhibitor p21. This is not surprising, because the p21 promoter can be transactivated by HIF-1, which indicates that p21 is an HIF-1 target gene.30 Furthermore, hypoxia-induced p21 expression was abrogated in cells that lacked HIF-1α but not in parental cells.14
In vivo, PHD2-overexpressing xenograft tumors had significant growth retardation in nude mice accompanied by decreased blood vessels density compared with control tumors. There was no difference with regard to metastasis, but local invasion of PHD2-overexpressing tumors was reduced. Therefore, we further investigated potential mechanisms by which PHD2 exerts its tumor suppressive effects. One possible mechanism is that PHD2 inhibition may abrogate growth factor-driven mitogenic or angiogenic signals. In contrast to platelet-derived growth factor and fibroblast growth factor 2, PHD2 did not interfere with mitogenic VEGF signaling.31 However, the silencing of PHD2 increased angiogenic cytokines, whereas overexpression suppressed VEGF, IL-8, and Ang1 secretion. The observation that PHD2 expression exerts anti-invasive properties also was demonstrated in a Matrigel invasion assay, in which PHD2 suppression increased the invasive phenotype particularly under hypoxic conditions, whereas PHD2 overexpression was associated with a less invasive phenotype.
In summary, the function of PHD2 in pancreatic cancer is not yet clear, because there are conflicting data regarding the function and significance of PHD2 for cancers in general. PHD2 has been associated with tumor suppressive functions,20 whereas others have reported the tumor promoting activity of PHD2.19 In pancreatic cancer, PHD2 is not overexpressed compared with its expression in normal pancreatic tissue. Nevertheless, well differentiated tumors expressed the highest levels of PHD2. In mice, the overexpression of PHD2 decreased pancreatic cancer growth probably because of halted angiogenesis, as supported by the results from angiogenic cytokine profiling in vitro. However, we must keep in mind that PHD2 also may act independent of HIF, causing entirely different effects on tumor growth. Further analysis of the function of PHD2 will be crucial for understanding the role of PHD2 in tumor progression.
No specific funding was disclosed.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.