Cell Culture and Reagents
Human carcinoma cell lines included pancreatic adenocarcinoma (MiaPaCa2 and Panc1), head and neck squamous cell carcinoma (QLL2 and SCC25), salivary and lung mucoepidermoid carcinoma (H3118 and H292), and human cervical cancer (HeLa) cell lines. MiaPaCa2, Panc1, and SCC25 were purchased from the American Type Culture Collection (Manassas, VA). H3118 and H292 were gifts from Dr Frederic Kaye (National Cancer Institute). HeLa was provided by Dr Christopher Moskaluk (University of Virginia). QLL2 was derived from a patient at Memorial Sloan-Kettering Cancer Center (MSKCC). Cells were grown in vitro in Dulbecco's modified Eagle medium (DMEM), Eagle modified essential medium, or RPMI-1640 containing 10% fetal calf serum (FCS), penicillin and streptomycin, and incubated in a 5% CO2-humidified incubator at 37°C.
Dorsal root ganglia (DRG) from newborn male Balb/c mice were isolated as previously described (11
). DRG were removed rapidly, placed in phosphate-buffered saline (PBS contains [in grams/liter]: NaCl ; KCl [0.2]; Na2
[0.24], pH 7.4), and digested in 0.125% collagenase for 45 minutes followed by 20 minutes in 0.25% trypsin at 37°C. The DRG were then washed in media and gently dissociated with a pipette. The cell suspension was centrifuged through a 15% FCS solution (Sigma, St Louis, MO). The cell pellet was resuspended in cultured media containing Neurobasal, B-27 serum-free (Invitrogen Corporation, Carlsbad, CA), 2 mmol/L glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin. Neurons were then plated on 24-well culture plates precoated with poly-ornithine, fibronectin, and laminin (Falcon Plastics, Franklin Lakes, NJ). Neurons were incubated for 10 days in the bottom well of a Boyden chamber with Neurobasal, B-27 media, or with DMEM supplemented with 10% FCS. Dissociated neurons survived well without adding neurotrophic factors. Before the migration assay, neurons were incubated overnight with serum-free DMEM. We detected no morphological changes or neural cell death when neurons were incubated up to 3 days with serum-free DMEM.
Neural Invasion In Vitro Model for Assessment of Nerve–Cancer Cell Interactions
The in vitro PNMC model is based on a technique originally described for prostate cancer cells (12
). Mice (Balb/c, 2–4 weeks old, n
60) were killed using CO2
euthanasia according to MSKCC institutional guidelines, and their excised DRG were implanted approximately 500 μm adjacent to a colony of carcinoma cells in growth factor–depleted Matrigel matrix (BD Biosciences, Bedford, MA). Cultures were grown in DMEM, RPMI-1640, or MEM containing 10% FCS in 37°C and 5% CO2
incubation conditions. The FCS was removed as soon as cancer cells made contact with the DRG neurites (~5 days after implantation). In experiments with HeLa cells, 3% FCS was added to support cell survival. The neural invasion index was calculated by dividing the distance traversed by the invading cancer cells at day 10 along the DRG neurites (α) by the total distance measured between the tumor colony and the DRG (β).
For expression of phospho-ERK1/2, total ERK1/2, phospho-MEK-1/2, and total MEK-1/2, cells were grown in Matrigel with and without contact with the DRG for 10 days. Following 24 hours of incubation with serum-free media, cells were released without enzymatic digest using cell recovery solution (BD Biosciences) at 4°C. Cancer cells were mechanically dissociated gently under the microscope and isolated from the DRG without disrupting its integrity. Cell pellets were sonicated for 10 seconds and clarified by centrifugation. Proteins (20–50 μg) were subjected to electrophoresis in 7.5% or 10% Tris–HCl gels (Bio-Rad, Hercules, CA) and transferred to polyvinylidene difluoride membranes, blocked with TBS-T (20 mM Tris–HCl, pH 7.6; 154 mM NaCl; and 0.1% Tween 20), containing 5% nonfat dry milk (Bio-Rad), and exposed to primary antibody followed by a secondary antibody conjugated to horseradish peroxidase. Protein–antibody complexes were visualized using an ECL Plus detection system (Amersham, Piscataway, NJ). Density was quantified using a computer-controlled carge coupled device camera (AlphaImager Imaging Systems; Alpha Innotech, San Leandro, CA).
Glass dishes containing DRG and cancer cells grown in Matrigel were washed once with PBS and fixed with −20°C methanol for 10 minutes followed by brief exposure to −20°C acetone. Cells were then washed with PBS for 5 minutes at room temperature. Polyclonal rabbit anti-cytokeratin 8/18 or mouse anti-neurofilament primary antibodies were used for staining of cancer cells and nerves. Secondary antibodies included goat anti-mouse Cy5 and Alexa Fluor 488 and goat anti-rabbit: Cy5, Alexa Fluor 488, Alexa Fluor 633, and Alexa Fluor 568 (Molecular Probes, Eugene, OR). Preparations were incubated overnight with primary antibodies at 1:250 dilution or saline as control at 4°C. Cells were washed with PBS for 15 minutes and incubated with a secondary antibody at room temperature for 25 minutes at 1:200 and 1:1000 dilutions. Slides were mounted and examined by confocal microscopy (LSM510; Carl Zeiss, Oberkochen, Germany).
Immunohistochemistry of Murine Sciatic Nerves to Differentiate Between Nerve and Cancer Cells
Sciatic nerve specimens were excised 50 days after tumor injection, frozen in OCT, and cut into 8-μm-thick sections on glass slides. Some of the slides were fixed and stained with hematoxylin and eosin. To differentiate between nerve tissue and cancer cells, the sciatic nerve preparation was immunostained with polyclonal rabbit anti-cytokeratin 8/18 antibody (Santa Cruz Biotechnology, Santa Cruz, CA). Slides were fixed with 2% formaldehyde and 0.2% glutaraldehyde, quenched, blocked, and incubated with primary antibodies (1:100 and 1:200 dilutions) overnight at 4°C. Slides were incubated with a biotinylated secondary antibody and visualized with an avidin–biotin complex kit (Santa Cruz Biotechnology). Slides were counterstained with hematoxylin and reviewed in a blinded fashion by two of the authors (Z. Gil and A. Rein).
Immunohistochemistry of Human Pancreatic Cancers to Assess Expression of RET and GFRα1
The study population included randomly selected pancreatic adenocarcinomas case patients who had been identified through a search of the MSKCC Departments of Surgery and Pathology Tissue Bank after approval by the Institutional Review Board. Archival samples of pancreatic adenocarcinoma from tumors that had been surgically resected between January 1, 1980, and December 31, 2006, at MSKCC were selected for analysis. Only samples from previously untreated patients with available paraffin-embedded primary tumor tissue were obtained. All selected tumors demonstrated histological evidence of neural invasion. The demographic and pathological details of the patients are shown in Supplementary Table 1
(available online). A detailed histopathologic review of all cancers was performed by an experienced pathologist (D. L. Carlson). Tumor was identified and marked on representative sections stained with hematoxylin and eosin for each patient. Sections of 5 μm thickness were cut from paraffin blocks and placed on charged polylysine-coated slides. Representative sections were stained with hematoxylin and eosin and analyzed to confirm the presence of nerve invasion. Immunostaining was performed with a Ventana automated staining system (Ventana Medical Systems, Tucson, AZ). Monoclonal mouse anti-human RET and GFRα1 antibodies (Santa Cruz Biotechnology) were used at a dilution of 1:100 to 1:500. Brain tissue known to express RET and GFRα1 was used as the positive control, whereas normal connective tissue (fibroblasts and fat tissue) was used as the negative control. Relative protein expression was estimated using dedicated software (MetaMorph; Molecular Devices, Union City, CA), which analyzed the intensity and area of fluorescence.
Migration Assays for Studying the Effects of GDNF on Cancer Cells
Polyethylene terephthalate 8.0-μm pore transparent inserts (BD Biosciences, Bedford, MA) were used in 24-well plates. Cancer cells were incubated in FCS-free media overnight, and 1 × 105 cells were added to each of the inserts in 0.5 mL of media without FCS. Below the inserts, 0.7 mL of FCS-free media was added with GDNF (1–100 ng/mL) or DRG into each of the wells. FCS-free media alone was used as a control. In additional experiments, dissociated neurons were treated with anti-GDNF antibodies (0.5 or 4 μg/mL), or MiaPaCa2 cells were incubated with pyrazolopyrimidine-1 (PYP1, 2 μmol/L). The polyethylene terephthalate inserts were removed after 24 hours. The nonmigrating cells were wiped off from the superior aspect of the membranes using a cotton swab. The migrating cells present on the undersurface of the membrane were fixed in 100% alcohol and stained with 1% methylene blue in 1% borax. The circular membranes were then excised using a scalpel and mounted on glass slides. Migrating cells were quantified by counting stained cells in five high-power fields at predetermined areas on the membrane.
Cytotoxicity Assays for Assessing Cancer Cell Viability With Small-Molecule Inhibitors
Pancreatic cancer cells (MiaPaCa2) were plated at 2 × 104 cells per well in 12-well plates in 2 mL medium. After incubation for 24 hours, PD98059 (25 μmol/L) or PYP1 (2 μmol/L) was added to each well. Cell viability was tested at daily intervals by lactate dehydrogenase assays. On day 4, fresh medium with the same drug concentration was added to the wells to feed remaining viable cells. Cells were washed with PBS and lysed with Triton X (1.35%) to release intracellular lactate dehydrogenase, which was quantified from cell lysates with a Cytotox 96 kit (Promega, Madison, WI) at 450 nm on a spectrophotometer (EL321e; Bio-Tek Instruments, Winooski, VT). Results are expressed as the ratio of surviving cells. This ratio was determined by comparing the measured lactate dehydrogenase of each infected sample relative to control untreated cell samples, which were considered 100% viable. All samples were analyzed in triplicate.
Lentiviral Gene Transduction and RNA Interference for Inhibition of GDNF Expression by DRG
Lentiviral pLKO.1 vectors expressing short hairpin RNA against murine GDNF (NM_010275) were purchased from Open Biosystems (Huntsville, AL) and from OriGene Technologies (Rockville, MD). pLKO.1 empty vector was used as a negative control. High-titer lentiviral vector stock was produced in 293T cells by using calcium phosphate–mediated transient transfection and cotransfection with pCMV and pUC19 packaging vectors as described previously (13
). The efficiency of the transfection was evaluated using fluorescence microscopy to monitor enhanced green fluorescent protein expression that was typically greater than 90%. Supernatants were harvested after 24 hours. DRG neurons (~2 × 105
in 24-well dishes) were transduced with virus plus 4 μg/mL polybrene. A second infection was performed 24 hours after the first infection and the medium changed 24 hours later. Protein expression was determined 48 hours later by immunoblotting. DRG treated with anti-GDNF short hairpin RNA was used for migration experiments but not for the Matrigel DRG experiments because infection of the DRG through the Matrigel was inefficient.
Small interfering RNA (siRNA) oligonucleotides directed against human RET (Stealth Select RNAi) were custom-designed and validated by Invitrogen Corporation (Carlsbad, CA). Cells were transfected with three different transcripts of RET siRNA (UniGeneID: Hs.350321, Cat. No. HSS109180, HSS109181, HSS109182) or with a nontargeting siRNA used as control (Stealth RNAi Negative control Low GC, No. 12935-200). MiaPaCa2 cells were cotransfected with 600 pmol siRET or control siRNA using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. Green fluorescent protein cDNA was used to optimize transfection efficiency before siRNA transfection. After 72 hours, cells were harvested, and RET expression was assessed with immunoblotting.
The siRNA transcripts that induced maximal block of RET expression (HSS109180, 5′-GGCGGAAGAUGAAGAUUUCGGAUUU-3′) were also used for the nerve invasion DRG assays. siRET-transfected MiaPaCa2 cells were harvested 24 hours after transfection and plated in Matrigel DRG assays for subsequent experiments.
Video Microscopy and Computer-Assisted Time-Lapse Analysis for In Vitro Migration Assays
Cancer cells and DRG were maintained in Delta T 0.17 mm culture dishes (Bioptechs, Inc, Butler, PA) containing a buffered medium. During live video microscopy imaging, the dish was placed in a closed environment at a mean ± SD temperature of 37 ± 0.1°C and 5% CO2 (Tokai Hit, Fujinomiya, Japan). Cells were recorded by using a carge coupled device camera (CoolSNAP; Photometrics, Tucson, AZ) placed on a microscope (Axiovert 200M; Zeiss, Göttingen, Germany). Cells were imaged at days 7–10. Digitized images were recorded every 10 minutes for up to 24 hours to follow cell locomotion. In each experiment, three to six different cells were imaged simultaneously and up to seven focus planes were acquired for each cell. Data acquisition and analysis were performed using specialized software (MetaMorph 7; MDC, Sunnyvale, CA) for off-line quantitative measures of migrating cells. This software determines the xy coordinates of a cell at any given time, along with the cell trajectory, distance from origin, and speed of motion. The maximal distance from origin was calculated as the greatest linear distance between a cell's origin and its final position. The mean velocity was calculated by measuring the travel distance between subsequent positions in 10-minute intervals. The accuracy of the measurements was reviewed and confirmed for each cell independently.
In Vivo Model of Sciatic Nerve Function
Nude athymic mice (n
14) were anesthetized using isoflurane (1%–3% for maintenance and up to 5% for induction), and their left sciatic nerve was exposed (Supplementary Figure 2
, available online). The nerve could be easily identified deep to the femoro coccygeous and biceps femoris muscles. Carcinoma cell lines (MiaPaCa2 or QLL2) were microscopically injected into the sciatic nerve, distal to the bifurcation of the tibial and common peroneal nerves. Microinjection of 3 μL of cell suspension at a concentration of 1 × 105
cells per microliter was performed using a 10-μL Hamilton syringe for a 2-minute period. Sciatic nerve function was measured weekly as described previously (14
). The sciatic nerve innervates the hind limb paw muscles. Sciatic nerve function was monitored by using the following measurements: 1) gross behavior––signs of motor weakness or repetitive biting of the hind limb were monitored for 10 minutes; 2) sciatic neurological score––nerve function was graded according to hind limb paw response to manual extension of the body, from 4 (normal) to 1 (total paw paralysis) (15
); 3) sciatic nerve function index––this was calculated as the spread length (in millimeters) between the first and fifth toes of the mouse hind limbs. Sciatic nerve function was normal in all mice after the operation (n
Imaging of Neural Invasion In Vivo
A dedicated 55-MHz high-resolution small animal ultrasound system (Vevo 770; VisualSonic, Toronto, Canada) was used to measure nerve and tumor diameters once a week (n
14 mice). The sciatic nerve diameter was measured 5 mm proximal to the cancer cell injection site, just before it enters the spinal column. Using this method, we were able to identify neural invasion at a spatial resolution of 50–100 μm (16
). Eight mice were subjected to magnetic resonance imaging, performed on a Bruker USR 4.7-T 40-cm bore scanner (Bruker Biospin MRI, Inc, Billerica, MA) using a custom-designed active decoupled radiofrequency surface coil (Stark MRI Contrast Research, Erlangen, Germany) to improve signal to noise ratio and spatial resolution. The total imaging time was 90 minutes for each mouse. For each mouse, three scan sequences were acquired in the following order: a coronal T2
-weighted fast spin-echo image and a coronal T1
-weighted gradient-echo image pre- and postcontrast (gadolinium-diethylenetriamine penta-acetic acid) injection. T2
-imaging parameters included TE = 40.2 milliseconds, TR = 2000 milliseconds, field of view 3.0 × 3.0 cm, matrix 256 × 192, number of acquisitions
12, slice thickness of 800 μm, interslice spacing of 800 μm, and in-plane resolution of 117 × 156 μm. T1
-imaging parameters included TE
2.3 milliseconds, TR
1180 milliseconds, flip angle 30°, field of view 3.0 × 3.0 cm, matrix 126 × 126, number of acquisitions
8, slice thickness of 800 μm, and interslice spacing of 800 μm. Dynamic contrast-enhanced scans were obtained after injection of gadolinium-diethylenetriamine penta-acetic acid, 0.1 mmol/kg, every 2 minutes after injection with an in-plane resolution of 234 × 234 μm. Mouse respiration was monitored during the imaging session (model 1025; Small Animal Instruments, Inc, Stony Brook, NY).
Measurement of Toxicity of PYP1 on Mice
The toxicity of PYP1 was estimated using the following parameters: body weight, change in behavior pattern (grooming and general activity), skin condition (wounds and infections), and mortality. All protocols were approved by the MSKCC institutional animal care committee.
Student t test and analysis of variance test with multiple range tests adjusted for contrast with control were used for statistical analysis as appropriate. All P values were calculated using two-sided tests. Mantel–Haenszel and Fisher exact tests were used for evaluating differences in drug toxicity between groups. SPSS (SPSS, Inc, Chicago, IL) and Origin (OriginLab Corporation, Northampton, MA) were used for graphs and statistical analysis. Differences were considered statistically significant if P was less than .05. Error bars in the graphs represent 95% confidence intervals (CIs). All experiments were repeated at least three times. Data from representative experiments are shown.