Neuropilin-2–Mediated Tumor Growth and Angiogenesis in Pancreatic Adenocarcinoma

In 2008, there will be an estimated 37,680 new cases of pancreatic ductal adenocarcinoma (PDAC) in the United States (1). PDAC is associated with a dismal prognosis, with a 5-year survival rate of only 3% to 4% independent of stage (2). For more than a decade, the deoxycytidine analogue gemcitabine was the only U.S. Food and Drug Administration-approved chemotherapeutic agent used for PDAC based on its alleviation of disease-related symptoms and its modest survival advantage over 5-fluorouracil (3). More recently, the addition of targeted therapeutics to gemcitabine regimens has been thoroughly evaluated in phase III clinical trials. In one trial that compared gemcitabine and gemcitabine plus erlotinib, the epidermal growth factor receptor tyrosine kinase inhibitor, the combination therapy provided a small but statistically significant improvement in survival, which led to eventual Food and Drug Administration approval (4). In a similarly designed trial, the addition of cetuximab, the monoclonal antibody to epidermal growth factor receptor, to a gemcitabine regimen did not improve the efficacy of gemcitabine. Despite promising results in a phase II trial, the addition of bevacizumab, the anti-vascular endothelial growth factor (VEGF) monoclonal antibody, to a gemcitabine regimen did not provide any benefit over gemcitabine alone (5). Given the dismal prognosis for patients with PDAC and the lack of significant improvement in overall survival with the current mainstays of targeted therapy, a better understanding of tumor growth, with the goal of identification of novel targets for therapy, is imperative.

Although bevacizumab, when combined with gemcitabine, did not improve the effects of gemcitabine in a phase III clinical trial, recent preclinical evidence suggests that other VEGF-associated molecules may be valid targets for therapy in patients with PDAC (6–8). The VEGF family of ligands and receptors is involved in many physiologic and pathologic processes (9). In humans, there are five known VEGF family ligands (VEGF-A, VEGF-B, VEGF-C, VEGF-D, and placental growth factor) and three structurally related tyrosine kinase receptors [VEGF receptor (VEGFR)-1, VEGFR-2, and VEGFR-3; reviewed in ref. 10]. Coreceptors for the VEGFRs include neuropilin-1 and neuropilin-2 (NRP-2), which were first identified based on their role in axonal guidance and neuronal development (11–14).

Neuropilin-1 and NRP-2 are 130- to 140-kDa nontyrosine kinase transmembrane glycoproteins with unique protein structures that share 44% sequence homology (15). Although expression of the neuropilins was originally thought to be limited to neurons, these receptors have since been identified on inflammatory cells (16), vascular smooth muscle cells (17), endothelial cells (13, 18), and, most recently, tumor cells (reviewed in ref. 10). Given their lack of an intracellular kinase domain, it was originally thought that the downstream effects of neuropilins were dependent on their association with other transmembrane receptors and their ability to increase the binding affinity of the ligands to these receptors, thereby increasing downstream signaling. However, recent data suggest that neuropilins can signal via their short intracellular domain directly by recruiting neuropilin-interacting protein or synectin to the cell membrane (19–21).

NRP expression on tumor cells is correlated with a more malignant phenotype in melanoma (22) and in breast (23, 24), prostate (25), colorectal (26–28), and pancreatic (8, 28) cancers. Recently, our laboratory showed that NRP-2 in colorectal cancer regulates tumor growth, and exploration of small interfering RNA therapies identified this molecule as a potential therapeutic target in this disease (26). Using an antibody targeting murine NRP-2, others showed that inhibition of NRP-2 in a preclinical lung metastasis model prevented tumor metastases by blocking the formation of tumor-associated lymphatics (29). We and others have shown that PDAC specimens exhibit greater expression of NRP-2 than nonmalignant ductal epithelium does (6). Given these findings, we sought to identify the role of NRP-2 in PDAC. Through a series of experiments, we studied the effects of reducing NRP-2 on PDAC cells both in vitro and in xenograft models.

Human tissue specimens and cell lines. Formalin-fixed, paraffin-embedded primary PDAC specimens and adjacent nonmalignant pancreatic tissue were obtained from an established tumor bank at The University of Texas M. D. Anderson Cancer Center following protocols approved by the institutional review board.

The human PDAC cell lines AsPC-1, BxPC3, MiaPaCa2, MPanc96, and Panc-1 were obtained from the American Type Culture Collection. L3.6pl cells were kindly provided by I.J. Fidler D.V.M., PhD. (The University of Texas M. D. Anderson Cancer Center). All cells were cultured and maintained at 37°C with 5% CO₂ in MEM supplemented with 10% fetal bovine serum (FBS), 2 units/mL penicillin-streptomycin mixture (Flow Laboratories), vitamins (Life Technologies), 1 mmol/L sodium pyruvate, 2 mmol/L l-glutamine, and nonessential amino acids. In vitro experiments were done at 60% to 80% cell confluence and at early passages after receipt from their supplier.

Generation of NRP-2 short-hairpin RNA cell lines. Short-hairpin RNA (shRNA) expression vectors targeting NRP-2 were created as described previously (26). Briefly, two NRP-2 targeting sequences were identified, and oligonucleotides encoding each sequence followed by a 9-bp hairpin sequence were generated. The targeting sequences were as follows: 5′-CCCAACCAGAAGATTGTCC-3′ for NP2#1 and 5′-GTCAGCACTAATGGAGAGG-3′ for NP2#2. The oligonucleotides were then ligated into the pSilencer 4.0 shRNA expression system (Ambion) at compatible sites, generating the shRNA expression plasmids, shNP2Vec#1 and shNP2Vec#2. Negative-control shRNA expression plasmids, shConVec#1 and shConVec#2, were generated in a similar fashion by scrambling the NRP-2 targeting sequences and verifying through NIH Basic Local Alignment Search Tool analysis that the scrambled sequences were not substantially homologous with any vertebrate genes.

The stable NRP-2 knockdown cell lines, shNRP2-C21 and shNRP2-C23, were created by transfecting BxPC3 cells with 0.5 ng of each shNP2Vec plasmid using FuGENE 6 transfection reagent (Roche Diagnostics) according to the manufacturer's protocols. The stable control-transfected cell line, shCon, was generated similarly using both scramble sequence-encoding shNRP expression plasmids. Selected clones were isolated and maintained in medium containing 50 μg/mL hygromycin B (Roche Diagnostics). NRP-2 expression levels in all resulting cell lines were determined by Western blot analysis.

Western blotting. Cells were plated and grown to 70% to 80% confluence before protein extraction. Whole-cell lysates were obtained using radioimmunoprecipitation assay B protein lysis buffer as described previously (30). To isolate secreted proteins, cells were plated in 1% FBS-supplemented medium for 48 h, and conditioned medium was harvested and concentrated using Amicon Ultra Centrifugal Filter Devices (Millipore). Isolated proteins were quantitated using a modified Bradford assay (Bio-Rad Laboratories). Protein samples for Western blotting were prepared, boiled, and separated using SDS-PAGE on an 8% or 15% gel and transferred to a polyvinylidene difluoride membrane (Millipore) by electroblotting. Antibodies were diluted in TBS and 0.1% (v/v) Tween with 5% nonfat dry milk after 1 h protein blocking in the absence of antibody. Membranes were incubated at 4°C overnight and then washed and incubated with the appropriate horseradish peroxidase-conjugated secondary antibody (Amersham Biosciences) for 1 h at room temperature. Protein bands were visualized using a commercially available enhanced chemiluminescence kit (Amersham Biosciences). Antibodies used included neuropilin-1, NRP-2, and Jagged-1 (Santa Cruz Biotechnology); phospho-Akt^Ser473, Akt, phospho-extracellular signal-regulated kinase 1/2^{Thr202/Tyr204}, extracellular signal-regulated kinase 1/2, phospho-Src^Tyr416, and Src (Cell Signaling Technology); phospho-VEGFR-1^Tyr1213 (Millipore); VEGFR-1 (Oncogene Research Products); and VEGF-A (R&D Systems). For verification of equal protein loading, all membranes were stripped and reprobed for either β-actin or vinculin (Sigma-Aldrich). When indicated, densitometric analysis was used to quantitate differences in protein levels from blots using NIH ImageJ v1.34 software.⁴

Cell proliferation and chemosensitivity.In vitro cellular proliferation was determined using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay as described previously (26). Briefly, 2,500 cells of each cell line (parental, shCon, shNRP2-C21, and shNRP2-C23) were plated in each well. At each time point (0, 24, 48, and 72 h), 40 μL 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide solution was added to each well, and the plates were incubated for 1 h at 37°C. Colorimetric analysis after the addition of DMSO was done using a standard microplate reader.

Migration and invasion assays. Migration assays were done as described previously (7). Briefly, 75,000 control- or shNRP2-transfected cells were suspended in 500 μL of 1% FBS-supplemented medium and placed in an insert with 8 μm pores, which was lowered into 750 μL of 10% FBS-supplemented medium in a standard Boyden chamber assay. After 24 h incubation, migrated cells on the underside of the membrane were fixed and stained using the Protocol HEMA 3 stain set (Fisher Scientific). Membranes were excised, mounted, and examined under light microscopy at ×20 magnification. Migrated cells were counted in five random fields.

Invasion assays were done using a similar protocol with minor modifications. The inserts used in the invasion assays were coated with Matrigel (BD Biosciences) and prehydrated with 1% FBS-supplemented medium for 30 min before the addition of the cell suspension. Invasion chambers were incubated for 48 h, and the numbers of invading cells were again quantified.

Anchorage-independent growth assays. Soft-agar assays were used to determine the effect of reduced NRP-2 expression on the ability to grow in anchorage-independent conditions. Each well of a 6-well plate was coated with 1 mL of 10% FBS-supplemented medium with 1% agarose. After 20 min, cell suspensions containing control- and shNRP2-transfected cells (500 cells each) were added in 1 mL medium with 0.5% agarose. Cells were incubated for 14 days under standard conditions (37°C, 5% CO₂) with the addition of 300 μL medium every 3 days to hydrate the exposed agarose. At the end of the incubation period, wells were examined under a light microscope at ×20 magnification, and the number of colonies larger than 50 μm was counted per well.

Xenograft models. Male athymic nude mice, 6 to 8 weeks old, were obtained from the National Cancer Institute-Frederick Cancer Research Facility and acclimated for 2 weeks. All animal studies were conducted under approved guidelines of the Animal Care and Use Committee of The University of Texas M. D. Anderson Cancer Center. Equal numbers of cells (10⁶) from the shCon, shNRP2-C21, and shNRP2-C23 cell lines were suspended in 100 μL PBS and injected subcutaneously into the right rear flank of each mouse (10 mice per group). Tumor growth was observed and recorded over 10 weeks. When tumors in the control group exceeded 1.5 cm in longest diameter, mice were killed by CO₂ asphyxiation according to protocol, and tumors were excised. Tumors were weighed and measured, and a portion of each was placed in either 10% formalin (for paraffin embedding) or OCT compound or was snap-frozen in liquid nitrogen. Tumor volume was calculated as 0.5 × (width²) × (length).

The results of the subcutaneous xenograft study were validated with an orthotopic pancreatic tumor model. BxPC3-shCon and shNRP2 cells were infected with a luciferase-reporter gene using a recombinant lentivirus as described previously (31). In each of 10 mice, a suspension of 2 × 10⁶ luciferase-labeled shCon or shNRP2 cells in 50 μL PBS was injected into the tail of the pancreas through a left-flank incision under ketamine/xylazine (Sigma-Aldrich) anesthesia. Mice were killed at 50 days when 2 to 3 mice in any group showed signs of lethargy. Tumors were weighed, measured, and processed as in the subcutaneous model.

Immunohistochemical and immunofluorescence analysis. Tumors preserved in formalin were placed in paraffin blocks and sectioned onto positively charged microscope slides. They were deparaffinized in xylene, hydrated in graded alcohol, and pretreated for antigen retrieval in citrate buffer for 20 min in a 98°C steamer. Tumor sections embedded in OCT compound were sectioned onto positively charged microscopy slides and serially immersed in acetone, a 1:1 (v/v) acetone/chloroform mixture, and then acetone. Slides were then stained with H&E to assess morphology, with proliferating cell nuclear antigen to visualize proliferative nuclei, with terminal deoxynucleotidyl transferase-mediated biotin-dUTP nick end labeling to visualize apoptotic cells, or with CD31 to visualize blood vessels. All immunohistochemical sections were counterstained with Gill no. 3 hematoxylin (Sigma-Aldrich).

Antibodies and materials used included rat anti-mouse CD31 (Pharmingen), mouse anti-human proliferating cell nuclear antigen PC-10 (DAKO), and the DeadEnd Fluorometric Terminal Deoxynucleotidyl Transferase-Mediated Biotin-dUTP Nick End Labeling System (Promega). Immunofluorescence slides were examined using a Nikon Microphot FXA fluorescence microscope, and representative images were obtained.

To determine the digitized microvascular area (D-MVA), CD31 fluorescently stained slides were analyzed using NIH ImageJ v1.34 software. The red channel, corresponding to CD31 staining, was isolated and digitized into a binary image, with black indicating stained vessels and white indicating no staining. Vessels with lumens were digitally filled, and a composite D-MVA was quantitated.

In vivo bioluminescence imaging. Bioluminescence imaging of luciferase-expressing cells in the orthotopic tumor model was done using the IVIS 100 imaging system coupled to a data acquisition personal computer equipped with Living Image software (Xenogen). Tumor cell-inoculated mice were anesthetized with a 1.5% isoflurane-oxygen mixture and injected intraperitoneally with luciferase potassium salt solution (Sigma-Aldrich) at a dose of 150 mg/kg body weight immediately before imaging. In vivo images were obtained on days 3, 18, 39, and 50, and photon emission representative of luciferase activity was used assess relative tumor burden in the mice.

Expression of NRP-2 in human pancreatic adenocarcinoma. Western blot analysis of 6 commonly used pancreatic cancer cell lines showed that 5 of 6 expressed NRP-2 to varying degrees (Fig. 1A). When tissue from PDAC surgical specimens was analyzed for NRP-2 expression, NRP-2 was detected in 7 of 11 (64%) adenocarcinomas but not in any of 4 specimens of adjacent nonmalignant tissue; representative images are shown in Fig. 1B. In specimens designated positive, all visualized tumor cells stained positively for NRP-2 expression.

Reduced NRP-2 expression alters constitutive signaling in pancreatic cancer cells. shRNA-NRP-2 decreased NRP-2 without any effect on neuropilin-1. Because the neuropilins are coreceptors for VEGFRs in both tumor cells and endothelial cells, we determined the effect of reduced NRP-2 expression on constitutive activation of VEGFR-1 in BxPC3 cells. This cell line expresses VEGFR-1 but not VEGFR-2 or VEGFR-3 (by reverse transcription-PCR analysis; data not shown). The reduction of NRP-2 expression was associated with a decrease in phosphorylation of VEGFR-1 in these cells relative to control-transfected cells (Fig. 2A). Furthermore, survival signaling, as measured by Western blotting for phospho-Akt, was reduced in cells transfected with shRNA to NRP-2. This finding was confirmed in transfected MPanc96 cells (data not shown). Other intracellular signaling molecules were evaluated to investigate the role of NRP-2 in cellular signaling. We determined that Src phosphorylation was also moderately reduced in shNRP-2 cells, but extracellular signal-regulated kinase 1/2 levels were unchanged relative to those in control-transfected cells (Fig. 2B).

Reduced NRP-2 expression decreases migration, invasion, and anchorage-independent growth. A 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay was used to determine the effect of reduced NRP-2 expression on in vitro proliferation rates. The shCon and shNRP2 clones showed similar in vitro proliferation rates up to 72 h after plating (Fig. 3A).

To evaluate the effect of reduced NRP-2 expression on in vitro migration and invasion, a standard Boyden chamber assay was used. Pancreatic cancer cells with shRNA to NRP-2 showed a 40% to 70% decrease in ability to migrate (Fig. 3B; P < 0.05) and a 50% to 70% decrease in ability to invade through a Matrigel-coated membrane (Fig. 3C; P < 0.05).

The ability of cells to survive and replicate under anchorage-independent conditions was evaluated using a soft-agar growth assay. Cells with reduced NRP-2 produced significantly fewer clones in soft agar than did control-transfected cells (P < 0.05).

Reduced NRP-2 expression leads to decreased in vivo tumor growth and alters the tumor vasculature. Given the decrease in migratory and invasive behavior and the reduction in anchorage-independent growth of cells deficient in NRP-2, we next investigated the role of NRP-2 in in vivo growth. Growth characteristics were first evaluated using a murine subcutaneous xenograft model. After 10 weeks, mice inoculated with shNRP2-transfected cells grew tumors that were 63% to 95% smaller than those of mice inoculated with shCon-transfected cells (Figs. 4A and 5A-C; P < 0.05). Furthermore, despite similar in vitro proliferation rates, analysis of tumors by proliferating cell nuclear antigen staining showed a significant decrease in in vivo proliferation in the shNRP2 groups. Specifically, tumors derived from shNRP2-transfected cells had ∼65% fewer proliferative nuclei per field relative to those in tumors from shCon-transfected cells (P < 0.05). No differences were seen in the numbers of apoptotic cells in the tumors by terminal deoxynucleotidyl transferase-mediated biotin-dUTP nick end labeling staining (data not shown).

A follow-up study was designed to validate our findings in an orthotopic model using shCon- and shNRP2-transfected cells labeled with firefly luciferase. Before mouse implantation, 2 × 10⁵ cells were plated in 12-well plates, and equal luciferase activity in each cell line was verified in vitro (data not shown). Cells were then prepared and injected orthotopically into the tails of the mouse pancreases. As seen on bioluminescence imaging, tumors derived from shNRP2 cells grew more slowly than did the shCon-derived tumors (Fig. 6A and B; P < 0.05). At the time of tumor harvest, shNRP2 tumors were significantly smaller than shCon tumors; average tumor masses were 0.50 g for shCon tumors and 0.08 g for shNRP2 tumors (Fig. 6C; P < 0.05), corroborating the data from the subcutaneous model.

Reduced NRP-2 expression on tumor cells alters the tumor vasculature. Tumor morphologic characteristics were evaluated by standard H&E staining, which showed qualitative differences between groups (Fig. 5D-F). Although tumors derived from shCon-transfected cells formed contiguous tumor cell clusters with stromal elements, those from shNRP2-transfected cells contained patchy acellular areas suggestive of regions of restrained growth within the tumors. Given the decreased number of proliferative cells in the shNPR2-derived tumor sections (Fig. 5G-I), we hypothesized that the inhibition of growth may be due to a secondary effect via limited tumor angiogenesis. Immunohistochemical staining for CD31 was then used to evaluate the number and morphologic characteristics of blood vessels within each tumor. Vessels in shCon tumors were subjectively larger with more visible patent lumens (Fig. 5J-L). Vessels were enumerated by counting the number of discrete stained structures within each field without regard to vessel size or patency. There was no difference in absolute vessel number between groups. D-MVA was analyzed to incorporate vessel size and patency into our analysis of the tumor vasculature by providing an estimate of integrated lumen area and presumably blood flow orthogonal to the tumor section. shNRP2 tumors had a 61% to 66% decrease in D-MVA than shCon tumors (Figs. 4B and 5M-O; P < 0.05).

Reduced NRP-2 expression is associated with decreased Jagged-1 levels. The reduction of D-MVA led us to hypothesize that the effects on development of the tumor vasculature may be due to altered angiogenic mediator expression in the tumor cells themselves. To investigate this hypothesis, we used Western blotting to identify differences in protein levels of several known angiogenic mediators. There were no differences in VEGF-A, VEGF-C, or delta-like ligand-4 levels between shCon- and shNRP2-transfected cells (data not shown); however, there was a significant reduction (46% in shNRP2-C21 and 53% in shNRP2-C23) in Jagged-1 levels in cells deficient in NRP-2 relative to that in control cells (Fig. 7A and B).

PDAC is associated with poor outcome and a high mortality rate. Attempts to treat this malignancy with standard chemotherapeutics or, more recently, targeted agents have led to minimal improvements in disease-related measures or the overall mortality rate. New therapeutic strategies are essential if we are to improve the lives of patients with this disease.

Our results showed that NRP-2 is overexpressed in pancreatic cancer cell lines and human PDAC specimens relative to nonmalignant ductal epithelium. Reduced NRP-2 expression in cell lines led to decreased tumor growth, identifying NRP-2 as a potential therapeutic target in this malignancy. Although a decrease in NRP-2 expression had no effect on in vitro proliferation, tumors derived from NRP-2 knockdown pancreatic cancer cells had significantly fewer proliferative nuclei and a smaller functional blood vessel area in cross-sectional analysis. An orthotopic xenograft tumor model confirmed the findings of the subcutaneous model: NRP-2-deficient cells had significantly less in vivo tumor growth than control-transfected cells did.

Although the expression of NRP-2 in vascular and lymphatic endothelial cells has been described, few studies have investigated its role in tumor cells. Our group recently showed that NRP-2 in colorectal carcinoma plays a role in several critical aspects of the malignant phenotype (26). The results of this study confirm prior work done in our laboratory on PDAC and extend it by highlighting the importance of tumor-derived NRP-2 and its indirect effect on the tumor vasculature. We found that, similar to the case in colorectal cancer, NPR-2 in PDAC is involved in survival signaling, migration, invasion, and anchorage-independent growth in vitro. Furthermore, in vivo, we showed in two independent experiments at two different sites that cells deficient in NRP-2 had decreased tumor growth. We showed that NRP-2-deficient cells in vitro had similar proliferation rates to those of control-transfected cells, suggesting that the reduction of tumor growth in vivo was not due to inherent proliferative capacity.

These data suggest that the reduction of tumor growth may have been due to the secondary effect on angiogenesis as shown by D-MVA quantitation and the decrease in functional vasculature within the tumor. The reduction in tumor vasculature with the loss of NRP-2 expression was associated with a decrease in Jagged-1 expression in the tumor cells. Jagged-1 is a member of the Notch family of ligands and receptors known to play a role in developmental angiogenesis and, more recently, tumor angiogenesis (reviewed in ref. 32). Others have shown that increased Jagged-1 expression in tumor cells leads to increased endothelial capillary network sprouting in vitro and angiogenesis in vivo (33). Our data showed that a decrease in Jagged-1 expression after experimental reduction of NRP-2 levels in tumor cells is associated with a lower D-MVA, although a causal relationship has not been established. In conclusion, our data provide evidence that NRP-2 on tumor cells may serve as a therapeutic target in PDAC.

L.M. Ellis: commercial research grants, Sanofi-Aventis, honoraria, Genentech.