Article Text

Download PDFPDF

Protein kinase D2 is a crucial regulator of tumour cell–endothelial cell communication in gastrointestinal tumours
  1. Ninel Azoitei1,
  2. Ganesh Varma Pusapati1,
  3. Alexander Kleger1,2,
  4. Peter Möller3,
  5. Rainer Küfer4,
  6. Felicitas Genze4,
  7. Martin Wagner1,
  8. Johan van Lint5,
  9. Peter Carmeliet6,
  10. Guido Adler1,
  11. Thomas Seufferlein1,7
  1. 1Department of Internal Medicine l, University of Ulm, Germany
  2. 2Institute of Molecular Medicine and Max-Planck-Research Group on Stem Cell Aging, Ulm, Germany
  3. 3Department of Pathology, University of Ulm, Germany
  4. 4Department of Urology, University of Ulm, Germany
  5. 5Department of Molecular Cell Biology, K.U. Leuven, Belgium
  6. 6Vesalius Research Center (VRC), Belgium
  7. 7Department of Internal Medicine 1, Martin-Luther-University Halle-Wittenberg, Germany
  1. Correspondence to Professor Thomas Seufferlein, Department of Internal Medicine 1, Division of Medicine, Martin-Luther-University Halle-Wittenberg, Ernst-Grube-Straße 40, 06120 Halle (Saale), Germany; thomas.seufferlein{at}medizin.uni-halle.de

Abstract

Background Tumour angiogenesis is crucially dependent on the communication between the tumour and the associated endothelium. Protein kinase D (PKD) isoenzymes mediate vascular endothelial growth factor-A (VEGF-A) induced endothelial cell proliferation and migration and are also highly expressed in various tumours.

Aim To examine the role of PKDs for tumour proliferation and angiogenesis selectively in pancreatic and gastric tumours and in tumour-associated endothelium in vitro and in vivo.

Methods PKD2 expression in human tumours was determined by immunohistochemistry. The effect of PKD2 depletion in endothelial cells by siRNAs was examined in sprouting assays, the chorioallantois model (CAM) and tumour xenografts. In murine endothelium in vivo PKD2 was knocked-down by splice switching oligonucleotides. Human PKD2 was depleted in xenografts by siRNAs and PKD2-miRs. PKD2 activation by hypoxia and its role for hypoxia-induced NR4/TR3- and VEGF-A promoter activity, expression and secretion was investigated in cell lines.

Results PKD2 is expressed in gastrointestinal tumours and in the tumour-associated endothelium. Tumour growth and angiogenesis in the CAM and in tumour xenografts require PKD expression in endothelial cells. Conversely, hypoxia activates PKD2 in pancreatic cancer cells and PKD2 was identified as the major mediator of hypoxia-stimulated VEGF-A promoter activity, expression and secretion in tumour cells. PKD2 depletion in pancreatic tumours inhibited tumour-driven blood vessel formation and tumour growth in the CAM and in orthotopic pancreatic cancer xenografts.

Conclusion PKD2 regulates hypoxia-induced VEGF-A expression/secretion by tumour cells and VEGF-A stimulated blood vessel formation. PKD2 is a novel, essential mediator of tumour cell–endothelial cell communication and a promising therapeutic target to inhibit angiogenesis in gastrointestinal cancers.

  • Angiogenesis
  • protein kinase D2
  • PKD2
  • tumour
  • hypoxia
  • VEGF-A
  • pancreatic cancer
  • cell proliferation
  • cell signalling
  • gastric cancer
  • pancreatic cancer
View Full Text

Statistics from Altmetric.com

Significance of this study

What is already known about this subject?

  • Protein kinase D (PKD) isoforms are expressed in various tumours including prostate and gastric cancer.

  • PKDs play a role in cell motility and migration in various cell models.

  • PKD isoforms mediate vascular endothelial growth factor (VEGF-A) induced endothelial proliferation and migration.

What are the new findings?

  • PKD2 is expressed in gastrointestinal tumours, including pancreatic and colorectal cancers, both in the tumour cells and in the tumour-associated endothelial cells.

  • Selective PKD depletion in the tumour-associated endothelium impairs vessel formation and blocks tumour growth both in vivo and in vitro.

  • PKD2 is activated by hypoxia in pancreatic cancer cells and mediates hypoxia-induced VEGF promoter activity, VEGF expression and VEGF secretion by these cells via induction of the nuclear transcription factor NR4/TR3. Consequently, selective knock-down of PKD2 in tumour cells also prevents tumour angiogenesis and as a result tumour growth in vivo.

How might it impact on clinical practice in the foreseeable future?

  • PKD2 is a crucial mediator of induction of angiogenic factors in pancreatic cancer cells and the angiogenic response of the host vasculature. Due to this two-pronged mechanism of action in the endothelial and the tumour cells, targeting PKD2 is a promising therapeutic approach to switch off tumour angiogenesis.

Introduction

Most solid tumours require an adequate oxygen delivery. Rapid tumour proliferation often results in hypoxia that in turn induces angiogenesis, the generation of new blood vessels that supply the tumour with the necessary metabolites. Because of the intricate dependence of tumours on an adequate blood supply, tumour angiogenesis has become a very attractive target for tumour therapy.

The hypoxic response is triggered to a large extent by the transcription factor hypoxia-inducible factor-1 (HIF-1) that controls the expression of multiple target genes whose products are implicated in angiogenesis, metabolism and cell survival.1 Among the most prominent targets of HIF-1 is vascular endothelial growth factor (VEGF-A). Induction of VEGF-A expression is an ubiquitous response to a hypoxic atmosphere in a wide range of cultured cells.2 3 VEGF-A induces tumour angiogenesis by stimulating proliferation, survival and migration of endothelial cells.4 In addition to hypoxia, VEGF-A expression and secretion is stimulated by oncogenes in a variety of tumour cells.5 The precise signalling components regulating VEGF-A induced tumour angiogenesis in response to hypoxia, thereby controlling tumour progression, are the subject of intense research.

Recently, members of the protein kinase D (PKD) family of serine threonine kinases have been implicated in the signalling cascade induced by VEGF-A in endothelial cells. VEGF-A rapidly induces the activation of PKDs via a VEGF receptor 2 (VEGFR2)→PLCg→PKC pathway.6 Furthermore, VEGF-A-stimulated endothelial cell proliferation and migration is mediated by a pathway involving VEGFR2→PLCg→PKC→PKDs→HDAC5→NR4A1/TR3.6–9 In addition, recent data indicate that in endothelial cells PKD2 is the predominant PKD isoform that is required for proliferation, migration, in vitro angiogenesis and expression of VREGFR2 and fibroblast growth factor receptor-1.10

The PKD family belongs to the calcium/calmodulin-dependent protein kinase superfamily11 and comprises PKD1, PKD2 and PKD3.12 PKDs are activated by various stimuli including phorbol esters, reactive oxygen species, G protein coupled receptors and receptor tyrosine kinases such as the VEGFR2.6 12 13 PKCs directly activate PKDs via phosphorylation at two critical serine residues within the activation loop of the catalytic domain.14 15 PKDs play a critical role in cell motility, migration and invasion of cancer cells.16–20 Expression of PKD isoforms is high in various tumours21 (figure 1) and PKDs have been implicated in the regulation of tumour cell proliferation, migration and apoptosis.22–24

Figure 1

Protein kinase D2 (PKD2) is expressed in human epithelial tumours. (A) PKD2 was stained in 53 human pancreatic cancers using standardised protocols as described in Materials and methods. All cancer samples exhibited PKD2 immunoreactivity. Therefore normal ducts were taken as an internal reference and PKD2 staining intensity in normal ducts was considered to be low (+). Intermediate and high PKD2 staining intensity are labelled ++ and +++, respectively. A representative image for each staining intensity is shown. Scale bar: 50 μm. (B) The 53 human pancreatic cancer samples were grouped according to the criteria outlined above. Bar graphs represent the tumours in each category of PKD2 immunoreactivity (+, ++, +++). (C) A pancreatic cancer tissue array (n=32) was stained for PKD2 (green), CD34 (red) and DAPI (blue). In addition, a modulation contrast image is shown (phase contrast). Shown are images representative for a tumour sample with >90% PKD2 positive vessels. The yellow arrows mark the coexpression of cytoplasmic PKD2 and surface localised CD34 in the vessels surrounding the tumour cells. Scale bar: 25 μM. (D) The same tissue array was stained for PKD2 (red) and CD34 (black). Scale bar: 25 μM. A representative image of tumour cells with double positive vessels (black arrows) is shown. (E) The percentage of PKD2 positive vessels in the respective tumours in the tissue array was calculated as described in Materials and methods; tumours were grouped according to the percentage of PKD2 positive vessels surrounding the tumour cells (<10%, 10–50%, 50–90%, >90%). N indicates the number of tumour samples in each group. (F) Lysates of HUVEC or HEK293 cells were subjected to western blotting using specific antibodies directed against PKD1, PKD2 and PKD3.

The objective of this study was to investigate the role of protein kinase D2 in tumour angiogenesis, focusing not only on the surrounding tumour endothelium, but also on the tumour itself. Indeed, because of the multitude of tumour hallmarks that are affected by PKDs (proliferation, apoptosis, migration), we wanted to investigate whether PKD2 is also involved in the angiogenic programme (HIF-1 signalling, VEGF production) of the tumour cell itself. We hypothesised that if PKD2 would indeed play such a dual role in angiogenesis, then targeting PKD2 could provide a very powerful approach for anti-angiogenic therapy.

We investigated a potential role of PKDs in tumour angiogenesis in the tumour cells as well as in the endothelium using distinct in vitro and in vivo model systems that allowed a selective targeting of PKD2 either in the endothelium or in the tumour. Our study shows that PKD expression in endothelial cells is required for pancreatic cancer progression and angiogenesis in vivo in a chorioallantois model and in athymic mice. Next, we found that PKD2 expression in pancreatic cancer cells is required for hypoxia-induced regulation of VEGF-A promoter activity, VEGF-A protein expression and VEGF-A secretion by the tumour cells. We established a novel signalling axis induced by hypoxia in human cancer cells that includes PKD2, TR3 and VEGF-A. Consequently, interrupting the crosstalk between the tumour and the vessels by depletion of PKD2 in pancreatic cancer cells substantially inhibited tumour-driven blood vessel formation and subsequently tumour growth.

In conclusion, we demonstrate a novel essential role of PKD2 in two key aspects of tumour angiogenesis: VEGF-A expression and secretion by the tumour cells; and VEGF-A stimulated blood vessel formation by the tumour-associated endothelial cells. Furthermore, these results show that PKD2 inhibition provides a powerful two-pronged approach to inhibit tumour angiogenesis.

Materials and methods

Cell culture

Human PaTu2 and PancTu1 pancreatic cancer cells (gift of Simone Fulda, University of Ulm) and AGS gastric cancer cells were cultured in DMEM (Invitrogen, Karlsruhe, Germany) supplemented with 10% fetal calf serum (FCS; PAA, Germany) and 1% penicillin/streptomycin as described previously. Human umbilical vein endothelial cells (HUVECs) and human microvascular endothelial cells (HUMECs) were purchased from PromoCell (Heidelberg, Germany) and cultivated in endothelial cell basal medium (ECBM) supplemented with 2% FCS, 0.1 ng/ml endothelial growth factor (EGF), 1 μg/ml hydrocortisone and 1 ng/ml basic fibroblast growth factor (FGF).

Knockdown by RNA interference

Pancreatic cancer cells were seeded in six-well tissue culture dishes and allowed to grow for 24 h. Cells were transfected two consecutive days with 800 ng/ml siRNA: siPKD1 (GCAACUAUUGGAAGAUUGGAUAGCAAA) (#45-1812), siPKD2 (AAUGACCUUAACUGCCACGUCCCGG), siPKD3 (45-1814) (GCUGCUUCUCCGUGUUCAAGUCCUA) and siTR3 (CCAAGUACAUCUGCCUGGCUAACAA) (all Invitrogen) using TransMessenger reagent from Qiagen (Germany; #301525) according to the manufacturer's instructions. Scrambled control siRNA (UUCUCCGAACGUGUCACGU) was from Qiagen (Germany; #1022076).

Lentiviral mediated generation of PKD2 deficient PancTu1 cells

The BLOCK-iT Pol II miR RNAi sequences targeting human PKD2 (NM_016457) were cloned into the pcDNA 6.2-GW/EmGFP-miR vector followed by the transfer along with the CMV promoter to the HiPerform pLenti6.4/R4R2/V5-DEST vector via a Gateway recombination reaction (Invitrogen). High-titre virus-containing supernatants of 293FT after transient co-transfection of pLenti6.4-CMV/EmGFP-PKD2 with packaging plasmids were used for the lentiviral mediated transduction of pancreas cancer cells PancTu1. Stable cell lines were generated by selection with blasticidin for the next 3 weeks. PKD2-deficient PancTu1 cell lines were finally tested for PKD2 expression by western blotting

Promoter assays

Pancreatic cancer cells or HUVECs plated in six-well dishes were transfected with 1 μg of the respective promoter plasmid as indicated in the figure legends using TransMessenger reagent according to the manufacturer's instructions. Luciferase activity was determined using the Dual Luciferase Assay Kit (Promega, Mannheim, Germany). Firefly luciferase relative light units were normalised with renilla luciferase activity after co-transfection of a pRL-TK plasmid (Promega, Germany).

Hypoxia experiments

A hypoxic environment was generated using a modular incubator chamber from Billups-Rothenberg (Del Mar, California, USA). The chamber was flushed three times at 0.5 bar for 5 min with a gas mixture consisting of 1% oxygen, 5% carbon dioxide and 94% nitrogen, and incubated for 24–40 h at 37°C.

Western blotting

For western blotting, whole cell extracts (50–100 μg protein) were subjected to SDS–PAGE and transferred to PVDF membranes (Millipore Corporation, Bedford, Massachusetts, USA). Membranes were blocked with 5% dry milk in phosphate buffered saline (PBS) containing 0.2% Tween20. For subsequent washes, 0.2% Tween 20 in PBS was used. Membranes were incubated overnight with specific antibodies at 4°C under shaking conditions. To detect PKD isoforms and phosphorylated PKD antibodies against PKD1 (Santa Cruz Biotechnology, California, USA; #sc-935), PKD2 (Upstate; #07-488), PKD3 (Bethyl, TX, USA; #A300-319A) and pPKD Ser744/748 (Cell Signalling; #2054) were used. VEGF-A was from Abcam (#ab9544), and HIF-1a from BD Transduction Laboratories (#610959). b-actin antibody was used as a loading control (Sigma, Germany; #A2228). Goat anti-rabbit secondary antibody was from BioRad (Hercules, California, USA; #172-1019). Chemiluminescence was detected using the ChemiSmart Chemiluminescence 5000 documentation system (PeqLab Biotechnologie, Germany).

ELISA assay

Experiments were performed using the Quantikine human VEGF-A165 Immunoassay Kit (R&D Systems, Minnesota, USA, #DVE00) according to the manufacturer's instructions.

3D-spheroid collagen gel assay

Monoculture spheroids containing 750 cells were generated overnight by cultivating either HUVEC or PaTu2 cells in growing media containing 20% methyl cellulose (viscosity 4000 cP; Sigma, Steinheim, Germany) into U-shaped 96-well dishes for non-adherent cells. Co-culture spheroids were generated by co-cultivating 375 endothelial HUVEC together with 375 pancreas cancer cells (PaTu2) as described above. The following day mono- and/or co-culture spheroids were embedded into collagen gels. A collagen stock solution was prepared before use by mixing 8 volumes of acidic collagen of rat tails (Collagen Type I, BD Biosciences, Massachusetts, USA; equilibrated to 2 mg/ml) with 1 volume cold 10× EBSS (PromoCell, Heidelberg, Germany; #C-40196) and 1 volume of cold 0.2 N NaOH to adjust the pH to 7.4. This stock solution was mixed 1:1 with growing media containing 20% methylcellulose and the spheroids previously collected by centrifugation at 900 rpm. The spheroids containing gel was rapidly transferred into a pre-warmed 24-well plate and allowed to polymerise for 15 min after which 200 μl ECBM was applied onto the gel. The gels were incubated at 37°C, in 5% carbon dioxide for 24 h. In vitro invasion was digitally quantitated by measuring the number and the length of the sprouts that had grown out of each spheroid. Knock-down of PKD2 was achieved by transfecting specific siRNA twice into endothelial HUVEC cells 32 h and 8 h, respectively, prior to multicellular spheroid overnight generation. Scrambled siRNA was applied accordingly.

Chorioallantois membrane assay

Shells of fertilised chicken eggs were opened on day 4 and silicon rings (5 mm in diameter) were applied onto the chorioallantois membrane (CAM). After two consecutive siRNA deliveries to the cells as indicated, 1×106 pancreatic cancer cells were transplanted at day 8 onto CAM to generate tumours on the CAM. After 48 h tumours were harvested and photographed. For chicken PKD1 and PKD3 silencing, siRNA (in a transfection cocktail with TransMessenger reagent) was applied directly to the CAM at days 6 and 7.

Immunohistochemistry of human specimens

Human pancreatic, gastric and colon cancer samples were obtained from the tissue bank of the Department of Pathology, University of Ulm (Ulm, Germany); they were fixed in 10% buffered formalin and paraffin-embedded. Human pancreatic cancers were classified according to the World Health Organization grading classification and then analysed for expression of PKD2. Immunohistochemistry was done on 3 μm thick paraffin-embedded sections using the biotin–streptavidin peroxidase LSAB kit (Vector Labs via Biozol, Eching, Germany). Following pressure cooker (20 min) antigen retrieval (Antigen-retrieval solution pH 6.0, DAKO, Copenhagen, Denmark) an antibody against PKD2 (Epitomics; rabbit monoclonal; 1:200, Biomol, Germany) was applied at room temperature, followed by application of the biotin–streptavidin reagents according to the manufacturer's instructions and labelled using the alkaline phosphatase substrate kit 1 for PKD2 (Vector Labs). Each section was then counterstained with hemalaun. Colorectal cancer and gastric cancer samples were processed accordingly and five cases of each tumour type were analysed. A section with skipped primary antibody for each tumour served as a negative control. Double staining of human pancreatic cancer samples with PKD2 and CD34 was done according to standard immunofluorescence protocols. Briefly, after antigen retrieval sections were blocked with goat serum and incubated with primary antibody mixture (PKD2, CD34) for 1 h. Secondary antibodies were labelled with either Alexa 488 (PKD2) or Alexa 568 (CD34) and incubated for 1 h (1:1000 dilution). Nuclei were stained with DAPI. Immunohistochemistry based double staining was performed with the Vectastain ABC Kit according to the manufacturer's procedures. PKD2 expression in each tumour sample was quantified by an experienced pathologist. Tissues were analysed and photographed using a Leica microscope (Leica Camera, Solms, Germany).

Quantification of PKD2 staining intensity in human pancreatic cancer samples

To quantify the intensity of PKD2 staining within different human pancreatic cancers, we stained a collective of 53 samples for PKD2 via immunohistochemistry. To guarantee equal exposure time, substrate (same lot for all tumours) was incubated for exactly 10 min. The reaction was stopped after washing with distilled water. Evaluation was done by an experienced pathologist. As all tumours stained positive for PKD2, staining intensities were defined as low (+), intermediate (++) and high (+++) using normal ducts as an internal reference. Clinical parameters (age, sex, grading, tumour localisation, size etc.) were obtained from the CCCU (Comprehensive Cancer Center Ulm; Professor Möller, University Hospital of Ulm).

Quantification of PKD2 positive tumour vessels in human pancreatic cancer samples

To quantify the percentage of PKD2 positive vessels within different human pancreatic cancers, commercially available tissue arrays (AccuMax) were chosen as templates. Thirty-two tumours were stained in duplicate with antibodies against PKD2 and CD34 via immunofluorescence. An average of 6.8 high power field images were chosen randomly per tumour sample within the red channel (CD34), followed by automatic photographing of all three channels (green, PKD2; red, CD34; blue, nuclei) with a Leica microscope. On average, 21.5 (6–47) vessels were counted per high power field. Evaluation was done by two independent investigators. Only vessels with typical vessel morphology and clear CD34 staining signal were taken into the analysis for PKD2 positivity. The percentage of PKD2 positive vessels was calculated after dividing the number of PKD2 positive vessels by the number of counted vessels per high power field. The percentages of PKD2 positive tumours in the following groups were determined: <10% positive vessels, 10–50% positive vessels, 50–90% positive vessels or >90% positive vessels.

Immunohistochemistry of CAM and mouse tumours

Formalin fixed tumours were embedded in paraffin using standard histological procedures. The 5 μm sections were processed and stained with antibodies directed against Ki67 (1:100; Dako), pan-cytokeratin (1:100), PKD2 (1:250; Epitomics), desmin (1:100); von Willebrand factor (vWF) (1:100), CD31 (1:100; Dako); CD34 (1:100; Dako); HIF-1a (1:7; BD Transduction Laboratories).

Xenograft mouse models

All animal experiments were conducted according to German animal institutional regulations and research protocols were approved by the relevant authorities. Orthotopic xenograft experiments were performed under narcosis. Briefly, 700 000 PancTu1 cells modified as described in the figure legends in 50 μl serum and antibiotic free DMEM media were orthotopically delivered into the pancreas. NMRI mice were monitored for the next 3 weeks. After tumour retrieval, tumour volume was calculated according to the formula 0.5×L×W×T (L, length; W, width; T, thickness). Tumours were further processed for immunohistological analysis. For tumour inoculation, 1×106 PancTu1 cells were inoculated subcutaneously at the left and right dorsal sides of 6-week-old female NMRI nu/nu mice (Janvier Laboratories, Le Genest-St Isle, France). Control Vivo-Morpholino or PKD2 Vivo-Morpholino, respectively, were delivered intravenously at 12.5 mg/kg starting at day 2 after tumour inoculation and continued every third day for two consecutive weeks (five Vivo-Morpholino intravenous injections in total). Tumour size was assessed twice weekly as described above. Three weeks after tumour inoculation, tumours were retrieved for immunohistological analysis. Another two mice per group were sacrificed 24 h after the fifth intravenous injection, liver tissues were retrieved and protein extracts were prepared.

Statistical analysis

Data are presented as means ± SEM of three independent experiments if not stated differently. The paired t test was applied using Sigma Plot software (version 11.0) to establish statistically significant differences. Differences were considered statistically significant at p<0.05.

Results

PKD2 is expressed in pancreatic, colorectal and gastric cancers

PKD2 is expressed in multiple tissues.25 We examined the expression of PKD2 by immunohistochemistry in a set of 53 human pancreatic cancers. Virtually all tumours stained positive for PKD2. PKD2 immunoreactivity was largely detectable in the cytoplasm of the tumour cells but also in the nucleus and the perinuclear area. Staining intensities were defined as low (+), intermediate (++) and high (+++) using normal ducts as an internal reference (figure 1A). The predominant staining intensity was intermediate, followed by strong and low (47%, n=25; 37%, n=20; 15%, n=8, respectively; figure 1B). There was a non-significant trend towards lower PKD2 immunoreactivity in tumours with higher grading (data not shown). To examine PKD2 expression in the endothelia surrounding the tumour we examined a second set of pancreatic cancer samples (n=32) by double staining of PKD2 and CD34 as a marker for small vessels. We detected coexpression of PKD2 and CD34 in most of the vessels surrounding the tumour cells (figure 1C,D). To define PKD2 expression in these vessels in greater detail we defined four groups that exhibited <10%, 10–50%, 50–90% and >90% PKD2-poitive vessels per tumour. Most of the tumours could be allocated to the group that contained 50–90% PKD2 positive vessels (41%, n=13), followed by tumours with >90% of PKD2-positive vessels (31%, n=10) and with 10–50% positive vessels (25%, n=8). There was only one tumour that exhibited <10% PKD2 positive vessels. To determine whether this expression pattern was exclusive for pancreatic cancers we also examined a set of gastric and colorectal cancers (n=5 each). Similarly to pancreatic cancer, virtually all gastric and colorectal cancers exhibited PKD2 immunoreactivity (supplementary figure S1). PKD2 was also expressed in pancreatic and gastric cancer cell lines in agreement with previous data (data not shown).26 PKD2 was also a prominent PKD isoform to be expressed in HUVECs (figure 1F) in agreement with previous data.10

Silencing of chicken PKD1 or PKD3 in the CAM results in impaired vessel formation and decreased tumour growth

To directly examine the contribution of PKDs to tumour blood vessel formation in an in vivo approach, we specifically targeted chicken PKDs in the CAM, a widely used model to study pro- and anti-angiogenesis.27 The CAM allows the selective targeting of PKDs in chicken endothelial cells or human tumour cells applied to the CAM due to species differences in expression and DNA sequence of the members of this kinase family.28 29 There are only two PKD isoforms in chicken tissues corresponding to human PKD1 and PKD3, but no isoform corresponding to human PKD2.30 We established a method to successfully deliver siRNA to chicken CAM that allowed us to achieve a specific and almost complete knockdown of chicken PKD1 (cPKD1) or PKD3 (cPKD3), respectively (figure 2A). These siRNA constructs did not affect the expression level of human PKD1 and PKD3 (supplementary figure S2). Targeting cPKD1 or cPKD3 in the CAM resulted in a significant reduction in size of either pancreatic tumours (figure 2B,C, left panel) or gastric tumours (figure 2C, right panel and data not shown) growing on the CAM compared to scrambled siRNA oligonucleotides (figure 2B,C).

Figure 2

Selective knockdown of chicken protein kinase D1 (cPKD1) and cPKD3 in the chorioallantois membrane (CAM) results in reduced growth of cancer cells. (A) Scrambled or chicken-specific siRNA oligonucleotides targeting cPKD1 or cPKD3 were delivered to the CAM as described in Materials and methods. forty-eight hours after transfection, membranes were explanted and RNA was extracted followed by RT-PCR analysis using chicken-specific cPKD1 and cPKD3 primers. (B) Tumours arising from pancreatic cancer cells on CAMs transfected with scrambled siRNA, cPKD1 or cPKD3 specific siRNA oligonucleotides. (C) Quantification of the tumour area (mm2) of Panc-Tu1 or AGS cell induced tumours. Results represent the means of at least six tumours from three independent experiments (*p<0.05; **p<0.01).

Pancreatic and gastric tumours growing on CAM treated with cPKD1 or cPKD3 siRNA were less invasive as judged by the amount of cytokeratin-positive tumour cells invading the CAM (figure 3A). There was also a substantial reduction in vWF and desmin positive cells surrounding tumour cells in the CAM figure 3A–C. Thus, depletion of cPKD1 and cPKD3 in the CAM results in a marked decrease in peritumoural endothelial cells and pericytes, respectively. Consequently, there was a substantial reduction in Ki67 positive, proliferating tumour cells in both pancreatic and gastric tumours growing on cPKD1- or cPKD3-depleted CAM as compared to tumours growing on CAM treated with scrambled siRNA (figure 3D).

Figure 3

Selective knockdown of chicken protein kinase D1 (cPKD1) and cPKD3 in the chorioallantois membrane (CAM) results in impaired angiogenesis and reduced growth of cancer cells. (A) Immunohistochemical analysis of pancreatic and gastric tumours growing on the CAM was performed using specific antibodies directed against cytokeratin, desmin, and von Willebrand factor (vWF). Arrows indicate peritumoural endothelial cells and pericytes. (B,C) Quantification of desmin (B) or vWF (C) staining intensity in human pancreatic (left; PancTu1) and gastric tumours (right, AGS) growing on the CAM. Desmin and vWF immunoreactivities were quantified by subtracting background staining from the specific desmin or vWF staining using Optima software (*p<0.05; ***p<0.001). (D) Immunohistochemical analysis of pancreatic (PancTu1) and gastric tumours (AGS) growing on CAM using specific antibodies directed against Ki67. Representative images of at least 12 tumours are depicted. Scale bars indicate 500 μm (upper panels; cytokeratin) and 125 μm (other panels).

PKD2 depletion in HUVECs and HUMECs decreases basal and VEGF-A induced endothelial cell sprouting in tumour cell–HUVEC or tumour cell–HUMEC co-culture spheroids

We further characterised the role of PKD2 in tumour endothelium using an in vitro sprouting assay of HUVECs and HUMECs as an alternative model for tumour angiogenesis.31 32 We confirmed that VEGF-A165 treatment of HUVECs induced phosphorylation of PKDs at the two critical serine residues within the activation loop as well as autophosphorylation of PKD2 at Ser876 in a concentration- (figure 4A) and time-dependent manner (data not shown).6 10 PaTu2 pancreatic cancer cells and either HUVECs or HUMECs were used to generate 3D-spheroids, and endothelial cell sprouting in collagen gels was assayed in the presence or absence of VEGF-A165. Both HUVEC and HUMEC monoculture spheroids displayed a basal sprouting which was further enhanced on addition of VEGF-A165 (figure 4C). As expected, monoculture spheroids consisting of pancreatic cancer cells did not sprout in the absence or presence of exogenous VEGF-A165 (figure 4C). PaTu2-HUVEC and PaTu2-HUMEC coculture spheroids exhibited an up to 30% higher basal level of sprouting compared to HUVEC or HUMEC monoculture spheroids, indicating a stimulatory effect induced by the pancreatic cancer cells (figure 4C–E). Depletion of PKD2 by specific siRNA oligonucleotides in both endothelial cell lines (figure 4B) resulted in a substantial inhibition of basal as well as VEGF-A165 stimulated sprouting in PaTu2-HUVEC or PaTu2-HUMEC co-culture spheroids growing in 3D collagen gels. Both sprout number and sprout length were significantly reduced on depletion of PKD2 in the absence as well as the presence of VEGF-A165 (figure 4C–E). These data indicate that PKD2 is a critical mediator of tumour angiogenesis in human endothelial cells.

Figure 4

(A) Vascular endothelial growth factor-A (VEGF-A) activates protein kinase D (PKD) in HUVECs. Lysates from HUVECs incubated for 15 min with various concentrations of VEGF-A165 as indicated were subjected to western blotting using antibodies against PKD2, PKD2 phosphorylated at Ser876 (pPKD2 876) or PKDs phosphorylated at the two serine residues within the activation loop (pPKD S744/748). PMA at 400 nM was used as a control. (B) Lysates of HUVECs or HUMECs transfected with siPKD2 were subjected to anti-PKD2 or anti-β-actin western blotting. (C) PKD2 knockdown in human endothelial cells results in decreased sprouting of spheroid culture in 3D-collagen gels. HUVECs or HUMECs transfected with siPKD2 and PaTu2 cells were used to generate single spheroids or HUVEC/PaTu2 or HUMEC/PaTu2 co-spheroids as described in Materials and methods that were then incubated with 20 ng/ml VEGF-A165 as indicated. Invasion of HUVECs or HUMECs into the collagen gel was monitored for 24 h. Phase-contrast images of spheroids are shown. The scale bar indicates 50 μm. (D,E) Number and length of the sprouts of the various spheroids were determined by digitised image analysis. The bar graphs represent the means ± SEM of three independent experiments. *p<0.05; **p<0.01.

Targeting murine PKD2 reduces growth of human pancreatic cancer xenografts in nude mice

To further prove that targeting of PKD2 in endothelial cells reduces tumour growth in vivo, we used the systemic delivery of a splice-switching oligonucleotide (SSO) to mice as gene targeting approach. SSOs are morpholino oligomers conjugated to an arginine-rich peptide.33 We designed an SSO that specifically targeted the intron16–exon17 junction of murine PKD2 (mPKD2). Exon 17 harbours the key serine residues that are important for the catalytic activity of PKD2 (figure 5A). To determine the efficacy of this in vivo approach, we examined livers, lungs and kidneys of mice treated systemically with the morpholino that specifically targeted murine PKD2 (mPKD2-MO) or a control morpholino (Co-MO), respectively. In mice receiving the mPKD2-specific morpholino, there was a substantial reduction of mPKD2 expression in all organs analysed compared to the organs of mice receiving the control morpholino (figure 5B). Three weeks after inoculation, tumours developing in mice that received a systemic application of the mPKD2-MO were significantly smaller compared to those developing in mice injected with the Co-MO (figure 5C,D). Immunohistochemistry of tumours retrieved from mPKD2-MO injected mice showed a substantial decrease in desmin and CD31 immunoreactivity in vessels surrounding the tumour and within the tumour xenografts compared to tumours of mice treated with the Co-MO (figure 5E and data not shown). Furthermore, we observed a marked decrease in the number of Ki67 positive tumour cells in mice receiving the mPKD2-MO compared to mice that were treated with the control morpholino (figure 5E, right panels). Thus, these data confirm a major role of endothelial PKD2 in tumour angiogenesis in vivo.

Figure 5

Knock-down of murine protein kinase D2 (mPKD2) in vivo results in impaired human tumour growth and angiogenesis. (A) Schematic representation of the Vivo-Morpholino splice-blocking oligonucleotide targeting the intron 16–exon 17 junction of murine mPKD2. Key serines in the activation loop of PKD2 are highlighted in red. (B) A specific mPKD2-specific Vivo-Morpholino (mPKD2-MO) or a control Vivo-Morpholino oligonucleotide (Co-MO) was delivered to mice as described in Materials and methods and pancreatic cancer cells were implanted subcutaneously. Extracts of mouse livers, kidneys and lungs were subjected to western blotting using anti-PKD2 or anti-β-actin antibodies. (C) Photographs of mice treated with Co-MO (top panels) or mPKD2-MO (bottom panels) three weeks after inoculation of the pancreatic cancer cells (left panels) and of the respective tumours (right panels). (D) Tumour volume was quantified as described in Materials and methods. Bar graphs represent the means ± SEM of six tumours. *p<0.05. (E) Immunohistochemical analysis of subcutaneously growing pancreatic tumours three weeks after subcutaneous inoculation in athymic NMRI mice using antibodies directed against desmin, CD31 and Ki67. Arrows indicate peritumoural and intratumoural pericytes and endothelial cells, respectively. The scale bar indicates 125 μm.

PKD2 is activated by hypoxia

Since epithelial tumour cells express PKD2 (figure 1) and the tumour cells stimulated angiogenesis in the sprouting assay (figure 4), we wanted to determine whether PKD2 in the cancer cells could also contribute to tumour angiogenesis. Many epithelial tumours including pancreatic cancers are characterised by growing in a hypoxic environment34 35 characterised by an up-regulation of HIF-1α, a subunit of the HIF-1 transcription factor that induces transcription of several genes involved in the cellular response to hypoxia. Hypoxia is likely to trigger the VEGF-A mediated tumour angiogenesis in vitro and in vivo in various epithelial tumours, including pancreatic cancer.35 36 We detected a marked expression of HIF-1α in the pancreatic tumours growing on the CAM, suggesting that this experimental model provides a hypoxic environment to the tumour cells (figure 6A). Hypoxia-induced VEGF-A expression can be mediated by PKC-dependent mechanisms37 38 and PKCs are upstream activators of PKD2.12 15 Therefore we first examined the effect of hypoxia on PKD activity in human pancreatic cancer cells. In a hypoxic atmosphere, we observed a marked increase in PKD activation as determined by an activation-specific antibody that detects phosphorylation of PKDs at two critical serine residues within the activation loop (figure 6B). Hypoxia-induced PKD activation was comparable to the effect of a maximal effective concentration of phorbol myristate acetate (PMA). Thus, hypoxia triggers activation of PKDs. We could exclude the possibility that activation of PKD2 by hypoxia was due to an autocrine VEGF loop in the tumour cells, because both cell lines used lack VEGFR1 and VEGFR2 and do not respond to exogenously added VEGF-A165 (data not shown).

Figure 6

Hypoxia-induced expression and secretion of vascular endothelial growth factor-A (VEGF-A) by the tumour cells requires TR3 expression. (A) PancTu1 cells were allowed to form tumours on chicken CAM for 48 h followed by immunohistochemistry using an anti-HIF-1α antibody. A representative example of six tumours is shown. The scale bar indicates 1000 μm (left panel) and 125 μm (right panel). (B) PaTu2 cells were incubated under normoxic or hypoxic conditions for 40 h. Cell lysates were then subjected to anti-pPKD-Ser706/710 and anti-PKD2 western blotting, respectively. Quantification of pPKD S744/748 band intensity is presented. (C) PaTu2 cells were cultured under normoxic or hypoxic conditions for 24 h and further analysed by RT-PCR using TR3-specific primers. Expression of GAPDH was used as control. (D) PaTu2 cells were transfected with scrambled siRNA or TR3-specific siRNA oligonucleotides as indicated; 6 h after transfection, cells were cultured under normoxic or hypoxic conditions for 24 h followed by western blotting using anti-VEGF-A165 (upper panels) or anti-β-actin antibodies (lower panel). Parallel samples were subjected to RT-PCR analysis using specific primers for TR3 (lower panels). GAPDH was used as control. (E) PaTu2 cells transfected with TR3-specific or scrambled siRNA oligonucleotides as indicated were incubated under normoxic or hypoxic conditions for 24 h. Secreted VEGF-A165 in the supernatant was determined by ELISA. Bars represent the means ± SEM of two independent experiments each performed in duplicate. **p<0.01.

Hypoxia-induced VEGF-A expression and secretion requires TR3 in human pancreatic cancer cells

HIF-1α induces TR3, an orphan member of the steroid/thyroid receptor superfamily of transcriptional factors that positively or negatively regulate gene expression.39 TR3 has been shown to regulate VEGF-A induced angiogenesis through its transcriptional activity.40 In pancreatic cancer cells, hypoxia induced TR3 expression (figure 6C). Since hypoxia stimulates both VEGF-A expression and TR3, we examined whether TR3 could mediate hypoxia-induced VEGF-A165 expression in pancreatic cancer cells. Depletion of TR3 in pancreatic cancer cells prevented hypoxia-induced VEGF-A protein expression (figure 6D). Furthermore, depletion of TR3 significantly reduced hypoxia-induced secretion of VEGF-A165 into the tumour cell supernatant (figure 6E). These data show a link between hypoxia, TR3 activation and VEGF165 expression/secretion in pancreatic cancer.

PKD2 regulates hypoxia-induced VEGF-A165 secretion through induction of TR3 expression in pancreatic cancer cells

We have previously shown that TR3 is a target of active PKDs including active PKD2. PKDs induce the expression of TR3 by phosphorylating class 2 HDACs.41–43 Therefore, we reasoned that PKD2 could also be an important mediator in hypoxia-induced TR3 expression. Indeed, depletion of PKD2 in PaTu2 and PancTu1 cells, respectively, virtually abolished hypoxia-induced TR3 promoter activity (figure 7A) and markedly reduced hypoxia-induced TR3 mRNA expression in the pancreatic cancer cells (figure 7B). VEGF-A induced TR3 promoter activity in endothelial cells was also critically dependent on PKD2, as shown by depleting PKD2 in HUVEC cells (Figure 7C). In the absence of PKD2, VEGF-A165 failed to induce TR3 promoter activity. Furthermore, active PKD2-S244/706/710E induced TR3 mRNA expression to a similar degree as incubation of cells with VEGF-A165 in HUVEC cells (figure 7D).

Figure 7

(A) Protein kinase D2 (PKD2) mediates hypoxia-induced TR3 promoter activity in epithelial tumour cells. Pancreatic cancer cells transfected with PKD2-specific (siPKD2) or scrambled siRNA oligonucleotides were subsequently transfected with pTR3-luc and pTK-Renilla for 4 h. Cells were then incubated for another 24 h under normoxic or hypoxic conditions and then cell lysates were subjected to luciferase assays. Bars are the means ± SEM of three independent experiments. *p<0.05. (B) PaTu2 cells transfected with PKD2-specific (siPKD2) or scrambled siRNA oligonucleotides were cultured under normoxic or hypoxic conditions for 24 h followed by RT-PCR analysis using TR3-specific primers. GAPDH was used as control. (C) Vascular endothelial growth factor-A (VEGF-A165) induced TR3 promoter activity requires PKD2. HUVECs transfected with PKD2-specific (siPKD2) or scrambled siRNA oligonucleotides and subsequently pTR3-luc and pTK-Renilla were incubated for with 20 ng/ml VEGF-A165 for 8 h as indicated and then luciferase activity was determined. Bars are the means ± SEM of three independent experiments.*p<0.05. (D) HUVECs transfected with GFP-tagged wild-type PKD2 (PKD2-wt) or a constitutively active PKD2 mutant (PKD2-S244/706/710E, PKD2-3SE) were incubated with 20 ng/ml VEGF-A for 8 h as indicated and subjected to RT-PCR analysis using a TR3- specific primer. GAPDH was used as control. (E) Hypoxia-induced VEGF-A promoter activity requires PKD2 in epithelial tumour cells. PaTu2 cells co-transfected with pGL2-VEGF4-luc, pTK-Renilla and GFP-tagged PKD2-wt, constitutively active PKD2-S244/706/710E (PKD2-3SE) or kinase-dead PKD2 (PKD2-D695A; PKD2-DA) were incubated under normoxic or hypoxic conditions for 24 h followed by luciferase activity assays. Bars are the means ± SEM of three independent experiments. ***p<0.001. (F) PaTu2 cells transfected with PKD2-specific (siPKD2) or scrambled siRNA oligonucleotides as indicated and subsequently with pGL2-VEGF4-luc and pTK-Renilla were incubated for 24 h under normoxic or hypoxic conditions followed by determination of luciferase activity. Bars are the means ± SEM of three independent experiments. **p<0.01. (G) PaTu2 cells transfected with PKD2-specific (siPKD2) or scrambled siRNA oligonucleotides were cultured under normoxic or hypoxic conditions for 24 h. Cell lysates were subjected to western blotting using anti-VEGF165 and anti-b-actin antibodies. (H) PKD2 is required for hypoxia-induced VEGF-A165 secretion by epithelial tumour cells. Pancreatic cancer cells transfected with PKD2-specific (siPKD2) or scrambled siRNA oligonucleotides were incubated under normoxic or hypoxic conditions for 24 h. VEGF-A was determined in supernatant of the tumour cells using a specific ELISA. Bars are the means ± SEM of three independent experiments.*p<0.05; **p<0.01.

Having established that TR3 expression is critical for hypoxia-induced VEGF expression and secretion and that PKDs are activated under hypoxic conditions and required for hypoxia or VEGF-A165 induced TR3 expression, we wondered whether PKD2 could mediate hypoxia-induced VEGF-A165 expression in epithelial tumour cells. Pancreatic cancer cells were transfected with a VEGF-A-luc promoter plasmid together with wild type PKD2, constitutively active PKD2-S244/706/710E or catalytically inactive PKD2-D695A and were exposed to a normoxic or hypoxic atmosphere, respectively. We observed a twofold increase in VEGF-A promoter activity under hypoxic conditions compared to a normoxic atmosphere in cells expressing wild type PKD2. Expression of predominantly nuclear, constitutively active PKD2-S244/706/710E43 resulted in maximum activation of VEGF-A promoter activity under normoxic conditions; this was not further enhanced when cells were exposed to hypoxia. Conversely, expression of catalytically inactive PKD2 completely prevented hypoxia-induced VEGF-A promoter activity (figure 7E). In addition, depletion of PKD2 in the tumour cells also completely prevented hypoxia-induced VEGF-A promoter activity. Combined knockdown of PKD1, PKD2 and PKD3 by specific siRNA oligonucleotides did not result in a further reduction in VEGF-A promoter activity compared to depletion of PKD2 alone (figure 7F). Consequently, depletion of PKD2 also resulted in a substantial reduction of hypoxia-induced VEGF-A165 protein expression in the cancer cells (figure 7G). Finally, we determined the amount of VEGF-A165 protein in the supernatant of PancTu1 and PaTu2 cells transfected with scrambled siRNA or siRNA oligonucleotides that specifically target PKD2. On silencing of PKD2, hypoxia-induced VEGF-A165 secretion into the supernatant of both cell lines was significantly reduced (figure 7H). Thus, PKD2 regulates hypoxia-induced VEGF-A165 expression and secretion via TR3 in pancreatic cancer cells.

PKD2 depletion in tumour cells results in impaired tumour angiogenesis and reduced tumour growth on the CAM

If the signalling cascade described above is functional, targeting of PKD2 in the tumour should affect vessel formation and consequently tumour growth on the CAM. PancTu1 pancreatic cancer cells were transfected either with a scrambled siRNA or siRNA oligonucleotides that specifically and efficiently targeted PKD2 or each of the three PKD isoforms, respectively (figure 8A). The cells were then subjected to CAM assays and allowed to form tumours. Depletion of PKD2 in PancTu1 cells resulted in a substantial, more than 80% decrease in tumour growth on the CAM (figure 8B). The depletion of PKD1, PKD2 and PKD3 using specific siRNAs did not lead to a further reduction in tumour growth compared to the PKD2 siRNA oligonucleotide alone (figure 8B).

Figure 8

Protein kinase D2 (PKD2) depletion in pancreatic cancer cells impairs both tumour angiogenesis and tumour growth in the chicken chorioallantois membrane (CAM). (A) PancTu1 cells transfected with siRNA directed against PKD1, PKD2 or PKD3 (siPKD1, siPKD2, siPKD3) or scrambled siRNA oligonucleotides were lysed and lysates subjected to western blotting using PKD1, PKD2 or PKD3 specific antibodies. Membranes were reprobed with b-actin antibody. (B) Left panels: photographs of pancreatic tumours arising from cells transfected with either siRNA targeting PKD2, a combination of siPKD1, siPKD2 and siPKD3, or scrambled siRNA oligonucleotides as indicated growing in a silicon ring on the CAM. Right panel: quantification of the tumour areas shown in the left panel. Results represent the means ± SEM of at least six tumours from three independent experiments. **p<0.01. (C) Left panels: Expression of pan-cytokeratin and desmin (arrows) in tumours of cells transfected with scrambled siRNA or PKD2-specific siRNA oligonucleotides (siPKD2) 48 h after plating of the cancer cells. Representative images of at least 12 tumours are presented. Right panel: quantification of desmin staining in the tumours was performed by subtracting the background staining from the specific staining using dedicated image analysis software (Optima). **p<0.01; ***p<0.001. (D) Left panels: Ki67 expression in tumours of cells transfected with scrambled siRNA (left) or PKD2-specific siRNA oligonucleotides (right). Representative images of either scrambled or siPKD2-transfected tumours are presented. Right panel: quantification of Ki-67 positive cells in the tumours. Results represent the percentage of Ki67 positive cells from 300 tumour cells each in at least four tumours. **p<0.01. Scale bar: 500 μm (upper panels) and 125 μm lower panels.

Depletion of PKD2 in the human tumour cells resulted in a marked reduction of chicken blood vessel formation as determined by a reduction in desmin-positive pericytes in the CAM which is in line with the prediction of our model (figure 8C). Ki67 immunostaining revealed a significantly reduced proliferation rate in those tumours generated by PKD2-depleted pancreatic cancer cells compared to the tumours generated by cells transfected with scrambled siRNA (figure 8D) which corresponds well to the markedly reduced angiogenesis due to PKD2-knockdown in the tumours. These data suggest that PKD2 mediates tumour progression by promoting tumour angiogenesis in vivo.

PKD2 silencing in pancreatic cancer cells impairs tumour angiogenesis and growth in an orthotopic mouse model of pancreatic cancer

To further substantiate the data obtained in the CAM model we examined the effect of PKD2 depletion in the tumour cells on tumour angiogenesis and growth in a second in vivo model, an orthotopic pancreatic cancer model in athymic mice. PancTu1 human pancreatic cancer cells stably expressing a green fluorescence protein-microRNA (GFP-miR) that specifically targets PKD2 (GFP-miR-PKD2) or a scrambled GFP-miR were generated as described in the Materials and methods section and orthotopically transplanted to the pancreas of athymic mice. The expression of GFP-miR-PKD2 resulted in a markedly reduced PKD2 expression in clones #2 and #4 (figure 9A). Orthotopic implantation of clones #2 and #4 in the pancreas of athymic mice resulted in a significantly decreased tumour growth compared to tumours expressing the scrambled GFP-miR (figure 9B,C). Immunohistochemistry revealed that GFP was highly expressed in scrambled-GFP-miR and GFP-miR-PKD2 transduced tumours. However, PKD2 expression was decreased by more than 90% in tumours originating from GFP-miR-PKD2 transduced pancreatic cancer cells (figure 9D,E). Examination of peritumoural blood vessel formation in tumours expressing GFP-miR-PKD2 revealed a marked reduction in desmin, vWF, CD31 and CD34 immunoreactivity within and around the tumours compared to tumours expressing the scrambled GFP-miR (figure 9D). The marked reduction in angiogenesis in the tumours expressing GFP-miR-PKD2 was accompanied by a significant decrease in the number of Ki67-positive tumour cells (figure 9F).

Figure 9

Knock-down of protein kinase D2 (PKD2) in pancreatic cancer cells impairs tumour angiogenesis and tumour growth in an orthotopic mouse model of pancreatic cancer. (A) Cell extracts of PancTu1 cells stably transduced with scrambled GFP-miR or various PKD2-GFP-miR constructs as indicated were subjected to western blotting using a PKD2-specific antibody. For further experiments GFP-miR-PKD2 #2 and #4 were used. (B) PancTu1 human pancreatic cancer cells stably transduced either with GFP-miR scrambled or GFP-miR-PKD2 #2 or #4 were orthotopically delivered to the mouse pancreas as described in Materials and methods. (C) Quantification of tumour volumes three weeks after inoculation of cells. The bar graph represents the mean ± SEM of 3–4 tumours. *p<0.05. (D) Expression of GFP, PKD2, desmin, von Willebrand factor (vWF), CD31 and CD34 in human pancreatic tumours transduced with GFP-miR scrambled or GFP-miR-PKD2 #2 and orthotopically growing on NMRI nu/nu mouse pancreas. Representative images are shown. (E) Graphic representation of PKD2-positive cells in GFP-miR scrambled or GFP-miR-PKD2 #2-transduced human pancreatic cancer cells growing orthotopically in mouse pancreas. Results represent the percentage of PKD2-positive cells from around 700 randomly counted tumour cells. ***p<0.001. (F) Ki67 staining of orthotopically growing human pancreatic cancer cells transduced with GFP-miR scrambled or GFP-miR-PKD2 #2. Lower panel: graphic representation of Ki67 positive cells. Results represent the percentage of Ki67 positive cells of 700 counted tumour cells. ***p<0.001. Scale bar: 125 μm.

These data clearly show a markedly reduced tumour angiogenesis in this model and further substantiate a major role of PKD2 in tumour angiogenesis and consequently tumour progression in vivo by a two-pronged mechanism: (1) PKD2 is a major regulator of VEGF-A expression/secretion in the tumour; and (2) the kinase is a crucial mediator of VEGFR2-induced signalling in the endothelial cells surrounding the tumour (figure 10).

Figure 10

Model of protein kinase D2 (PKD2) mediated tumour angiogenesis. (A) Hypoxia activates PKD2 that induces transcription of TR3 and in turn leads to activation of the vascular endothelial growth factor-A (VEGF-A) promoter and VEGF-A secretion into the supernatant of the tumour cells. Tumour-derived VEGF-A binds to the VEGFR2 on the endothelial cells leading to activation of PKD2. Active PKD2 induces activation of TR3 which results in proliferation of endothelial cells, increased angiogenesis and as a consequence enhanced tumour growth. Blocking of PKD2 in either the tumour cells (B) or the endothelial cells (C) interrupts this feedback loop and results in both cases in markedly decreased angiogenesis and substantially reduced tumour growth in response to hypoxia. Thus, PKD2 is a major mediator of hypoxia induced angiogenesis acting in both epithelial tumour cells and endothelial cells.

Discussion

Angiogenesis is essential for cancer development and tumour progression.44 VEGF-A is a key mediator of angiogenesis in cancer, in which it is up-regulated by oncogene expression, a variety of growth factors and hypoxia. The production of VEGF and other growth factors by the tumour results in the 'angiogenic switch', where new vasculature is formed in and around the tumour, allowing it to grow exponentially. The central role of VEGF in the production of tumour vasculature made it a rational target for anticancer therapy and most efforts to stimulate or inhibit angiogenesis in the past were focused on VEGF. However, there is mounting evidence that inhibition of VEGF causes resistance and class-specific side effects, causing the need for alternative strategies to block tumour angiogenesis. In this study, we have investigated the role of protein kinase D2 in tumour angiogenesis, focusing not only on the endothelium surrounding tumour cells, but also on the tumour itself. We found that, besides its role in endothelial cells, PKD2 is also involved in the angiogenic programme (hypoxia signalling, VEGF production) of the tumour cells. Thus, PKD2 plays a dual role in angiogenesis, and targeting PKD2 could provide a very powerful two-pronged approach for anti-angiogenic therapy.

We found that PKD2 is expressed in human epithelial tumours, including pancreatic, colorectal and gastric cancers. The expression of PKD2 was not limited to the tumour cells, but was also evident in the surrounding endothelial cells. A marked expression of PKD2 in gastrointestinal epithelial tumour cells is consistent with earlier findings of high PKD expression in other epithelial tumours such as prostate cancer21 and skin basal cell carcinoma.45

We also found that PKD depletion in the ‘host’ (ie, depletion of PKD outside the tumour cells, eg in endothelial cells) impairs vessel formation in two experimental in vivo systems, the chicken CAM and human xenograft tumours in nude mice and in one in vitro system, tumour cell–HUVEC co-culture spheroids. The in vivo results demonstrate the feasibility of selective targeting of vessel PKD2 as a means to inhibit tumour angiogenesis. Moreover, the in vivo data confirm and extend earlier in vitro data pointing to an essential role for PKD in endothelial cell proliferation, migration and tubulogenesis.6 8–10 46

When we investigated a possible pro-angiogenic role for PKDs in tumour cells, we observed that PKD2 is activated by hypoxia. This identifies PKD2, for the first time, as a new component of hypoxia signalling. We next wondered whether there could be a connection between hypoxia-induced PKD2 activation and hypoxia-induced VEGF-A expression, and how this could possibly be mediated. In this context we were reminded of our earlier work demonstrating that PKD2 activates the expression of TR3/Nur77,42 43 a protein that has recently been shown to be a stabiliser of HIF1 and mediator of VEGF expression.40 47 Our experiments demonstrate that hypoxia-induced VEGF-A expression/secretion requires TR3 expression in epithelial tumour cells, and that TR3 expression, in turn, crucially requires PKD2. These findings establish a novel hypoxia-induced signalling pathway in tumour cells involving PKD2→TR3→VEGF expression.

Based on these findings, we next wondered whether PKD2 depletion in tumours would impair angiogenesis. Depletion of PKD2 only in tumour cells prevented tumour angiogenesis on the CAM and in turn tumour cell proliferation. Similarly, tumour angiogenesis and tumour growth were markedly reduced in an orthotopic pancreatic cancer mouse model when the tumour cells lacked PKD2. The finding that the mere depletion of PKD2 in the tumour is capable of impairing tumour angiogenesis indicates that PKD2 is very likely a mediator in the induction of expression/secretion of a broad spectrum of angiogenic factors.

Our findings that PKD2 is a crucial mediator of both induction of angiogenic factors in tumours and of the angiogenic response of the host vasculature, opens some very interesting perspectives. First, PKD2 inhibition offers the possibility to interfere with hypoxia signalling in the tumour, which is an aspect of tumour angiogenesis that hitherto has not been easy to target.48 49 Second, PKD2 inhibitors may offer a two-pronged approach to block angiogenesis, which has the potential to be more powerful than an approach focusing on the vasculature only. Third, PKD2 inhibition may help to avoid metastasis problems associated with therapeutic strategies targeting VEGF. Indeed, it has been reported that VEGF inhibition causes even more hypoxia and thereby increases hypoxic signalling, which in turn contributes to the expression of invasion and metastasis promoting factors.50 51 Inhibition of PKD2 may help to block this hypoxia induced expression of invasion and metastasis promoting factors by also acting on the tumour cells. Fourth, PKD2 inhibition may help to alleviate resistance problems associated with the use of VEGF inhibitors. Indeed, since the mere inhibition of PKD2 in the tumour is capable of inhibiting angiogenesis, it is very likely that the expression/secretion of a broad spectrum of angiogenic factors is affected by PKD inhibition, thereby reducing the chance of a compensatory switch to the secretion of other pro-angiogenic factors.52 53 Such a ‘broad spectrum effect’ of PKD2 inhibition could also help to overcome anti-angiogenic therapy resistance caused by pre-existing multiplicity of pro-angiogenic signals.52 54 This mechanism is also likely to be relevant for human pancreatic cancer, in particular in the metastatic setting, since metastases do not exhibit the marked desmoplastic reaction observed in the primary tumour.

In conclusion, our data suggest a central role for PKD2 in particular in the regulation of tumour angiogenesis and define a novel model in which hypoxia activates PKD2 in pancreatic cancer cells leading to activation of TR3, expression and secretion of VEGF and activation of the VEGFR2 on endothelial cells. The VEGFR2 on endothelial cells, in turn, activates endothelial PKD2 which again results in the activation of TR3 and in angiogenesis and consequently tumour proliferation (figure 10). Inhibiting PKD2 in either compartment blocks tumour growth and angiogenesis. Thus we propose that PKD2 is a novel interesting target for antiangiogenic strategies.

Acknowledgments

We thank Ralf Köhntop and Claudia Ruhland for expert technical assistance.

References

View Abstract

Supplementary materials

Footnotes

  • GVP and AK contributed equally.

  • Funding This project was funded in part by the German Federal Ministry of Education and Research (BMBF) grant no. PKB-01GS08209-4 to TS and by the DFG (SFB518 B3, to TS). The authors are responsible for the content of this paper. Work in the laboratory of JVL is supported by the Fonds voor Wetenschappelijk Onderzoek–Vlaanderen (G.0612.07) and by the IAP programme of the Belgian Federal Government (IAP 6/18).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.