By serial transplantation of human glioblastoma biopsies into the brain of immunodeficient nude rats, two different tumor phenotypes were obtained. Initially, the transplanted xenografts displayed a highly invasive phenotype that showed no signs of angiogenesis. By serial transplantation in animals, the tumors changed to a less invasive, predominantly angiogenic phenotype. To identify novel proteins related to the invasive phenotype, the xenografts were analyzed using a global proteomics approach. One of the identified proteins was protein disulfide isomerase (PDI) A6 precursor. PDI is a chaperone protein that mediates integrin-dependent cell adhesion. It is both present in the cytosol and at the cell surface. We show that PDI is strongly expressed on invasive glioma cells, in both xenografts and at the invasive front of human glioblastomas. Using an in vitro migration assay, we also show that PDI is expressed on migrating glioma cells. To determine the functional significance of PDI in cell migration, we tested the effect of a PDI inhibitor, bacitracin, and a PDI monoclonal antibody on glioma cell migration and invasion in vitro. Both tumor spheroids derived from human glioblastoma xenografts in nude rat brain and cell line spheroids were used. The PDI antibody, as well as bacitracin, inhibited tumor cell migration and invasion. The anti-invasive effect of bacitracin was reversible after withdrawal of the inhibitor, indicating a specific, nontoxic effect. In conclusion, using a global proteomics approach, PDI was identified to play an important role in glioma cell invasion, and its action was effectively inhibited by bacitracin. (Cancer Res 2006; 66(20): 9895-902)
- protein disulfide isomerase
- tumor invasion
Glioblastoma multiforme is the most common primary brain tumor in adults. Despite aggressive treatment, including surgery, radiotherapy, and chemotherapy, the majority of patients die within 1 year from diagnosis. Tumor cell invasion is a major problem in the management of glioblastoma multiforme, where infiltrative tumor cells prevent curative resection leading to relapse and a fatal outcome. The mechanisms of glioma cell invasion have been addressed in different studies and experimental settings, yet there is a need for novel markers characterizing the invasive phenotype ( 1, 2). Several high-throughput strategies have been used to identify molecular events regulating the invasive process ( 3, 4). For instance, cDNA microarray experiments have identified many differentially expressed genes that may promote invasion ( 5– 7). One major problem underlying this approach is that the functional cellular phenotype is ultimately defined by the protein expression level. This level shows only a limited correlation with the gene expression level, as it is also regulated by post-translational events.
Another methodologic problem is that xenotransplanting human glioma cell lines into the central nervous system (CNS) produces circumscribed tumors with little or no single cell infiltration into the brain parenchyma ( 1). To avoid this problem, we have developed a xenograft model where human brain tumor biopsy spheroids are transplanted into the nude rat brain. The tumors derived from such spheroids show a highly infiltrative behavior in the CNS, which reflects the invasive characteristics of the human tumors in situ (low-generation tumors; refs. 8, 9). On serial transplantation in vivo, the tumors will develop a less invasive and more angiogenic phenotype (high-generation tumors). Thus, in vivo tumor xenografts obtained from the same patient provide two different phenotypes: one that is predominantly invasive, showing little or no signs of angiogenesis, and one that is less invasive and predominantly angiogenic. By comparing the two different phenotypes using a proteomics approach, including the use of two-dimensional protein electrophoresis (2DE), mass spectrometry (MS) and bioinformatics, we identified novel proteins that were differentially expressed by the invasive phenotype. Using both in vitro and in vivo assay systems, we present one of the identified proteins, protein disulfide isomerase (PDI) that was overexpressed in the invasive phenotype. Moreover, its involvement in glioma cell invasion was confirmed in functional assays, where PDI inhibition caused a reduced cell migration and invasion. Finally, we show that PDI is strongly expressed in human glioblastoma multiforme biopsies taken from the invasive rim of infiltrative brain tumor.
Materials and Methods
Tissue culture. Tumor fragments were obtained at surgery from two glioblastoma patients. The collection of tumor tissue was approved by the regional ethical committee at Haukeland University Hospital (Bergen, Norway). The specimens were collected from tumor areas appearing macroscopically viable, corresponding to regions with contrast enhancement on preoperative magnetic resonance imaging (MRI) scans and immediately transferred under aseptical conditions to test tubes containing a complete growth medium, which consisted of DMEM, supplemented with 10% fetal bovine serum (FBS), 2% l-glutamine, and 4× the prescribed concentration of nonessential amino acids, penicillin, and streptomycin.
Spheroids were prepared as described previously ( 10). Briefly, tissue samples were minced into 0.5-mm fragments, placed into 80-cm2 tissue culture flasks (Nunc, Roskilde, Denmark), and base coated with 10 mL of 0.75% agar (Difco, Detroit, MI) in culture medium (DMEM). The spheroids were maintained in a standard tissue culture incubator with 5% CO2 in air and 100% relative humidity at 37°C and the medium was changed once weekly. After 1 to 2 weeks in culture, spheroids with diameters between 200 and 300 μm were selected for i.c. implantation ( Fig. 1 ).
The permanent glioblastoma cell line U373 was cultured in 80-cm2 culture flasks in the complete medium and passaged at confluence. Monolayers of cells were treated with trypsin-EDTA (BioWhittaker, Verviers, Belgium) as described before ( 11). The cell line spheroids were initiated by seeding cells into 80-cm2 cell culture flasks base coated with agar. Cell aggregates with size 100 to 200 μm were selected for the experiments.
Animal experiments. Nude rats (Han:rnu/rnu Rowett) were fed a standard pellet diet and provided water ad libitum. All procedures were approved by The National Animal Research Authority. Biopsy spheroids were stereotactically implanted into the right brain hemisphere as described elsewhere ( 8). The animals were sacrificed when symptoms developed, and the brains were then removed. All surgical procedures were done on animals anesthetized by s.c. injection of Dormicum (Roche, Basel, Switzerland), 0.2 g/100 g of body weight, and fentanyl/fluanisone (Hypnorm, Janssen, Oxford, United Kingdom; fentanyl citrate, 0.15 mg/mL; fluanisone, 10 mg/mL), 0.0126 g/100 g of body weight. The rats were immobilized in a stereotactic frame (model 900, David Kopf Instruments, Tujunga, CA). After local anesthesia with Xylocain (AstraZeneca, Zug, Switzerland), a burr hole was made 3.0 mm to the right of the sagittal suture and 1 mm posterior to the bregma. PBS (25 μL) containing 10 spheroids was injected into the forebrain to a depth of 2.5 mm under the brain surface. Closure was effected with 3.0 ethilon suture. Tumor growth was monitored at regular intervals during 4 to 20 weeks after implantation using a MRI Magnetom Vision Plus 1.5 T scanner (Siemens, Erlangen, Germany) and a small loop finger coil as described previously ( 12, 13). The animals were sacrificed by CO2 inhalation when neurologic signs were evident. The tumors were removed and new spheroids were generated, which again were transplanted into new animals. A schematic description of the experimental procedure is provided in Fig. 1. The brains were removed and fixed in 4% formaldehyde or snap frozen in liquid N2 for further studies.
Two-dimensional protein electrophoresis. Novel proteins were identified by 2DE followed by gel comparisons. Differentially expressed proteins were selected manually, excised, and trypsinized.
The tryptic digest was applied on a MALDI-TOF plate and peptide mass spectra were generated. The achieved mass spectra were compared with in silico–generated theoretical ones and the score, indicating the probability of correct protein identification was obtained. Only proteins identified with significant scores were further verified.
The brain tumor specimens were stored in liquid N2. After thawing, they were washed in TRIS/sucrose solution [0.25 mol/L sucrose in 10 mmol/L Tris (pH 7.4); Tris, Merck, Darmstadt, Germany; sucrose, Sigma-Aldrich, St. Louis, MO] and placed in sample buffer containing 7 mol/L urea, 2 mol/L thiourea (Merck), 4% CHAPS (Sigma-Aldrich), 100 mmol/L DTT (Merck), and 1% pharmalyte (Amersham Biosciences, Uppsala, Sweden) pI 3-10: pI 5-6 = 3:1. Afterwards, the tissue samples were homogenized 3 × 10 seconds, sonicated 3 × 1 seconds, and centrifuged at 1,600 × g for 20 minutes at 4°C using an Eppendorf centrifuge (Eppendorf, Hamburg, Germany). The supernatant was transferred to a new tube and the protein concentration was estimated using the Bradford reagent (Bio-Rad, Hercules, CA). For the analytic gels, a protein load of 100 μg per gel was applied. The protein load for micropreparative gels was 400 μg/gel.
The first dimension of separation was carried out on IPGphore (Amersham Biosciences) at 20°C. The desired protein amount, 100 μg for analytic gels and 400 μg for the micropreparative ones, was diluted in 350 μL sample buffer and applied on an 18-cm (pH 3-10) linear immobilized pH gradient (IPG) Immobiline DryStrip IPG (Amersham Biosciences) by in-sample rehydration. After 1 hour of passive rehydration, the strips were covered with mineral oil (Bio-Rad) to prevent drying, and 50 V voltage was applied overnight (16 hours). The day after, the electrodes were covered with Milli-Q water moistened paper pads and the voltage increased to 250 V for 15 minutes followed by a gradient to 8,000 V and step-in-hold 8,000 V to final 60,000 Vhrs. After 1 hour of separation, the electrode pads were changed to remove the electrolytes. After the first dimension of separation, the IPG strips were stored at −80°C until they were subjected to the second dimension of separation.
Before the second dimension of electrophoresis, the strips were equilibrated for 15 minutes in solution containing 100 mmol/L DTT, 6.5 mol/L urea (Merck), 26% (w/v) glycerol (Merck), 2% (w/v) SDS solution (Bio-Rad), and 50 mmol/L Tris-HCl (pH 6.8). Thereafter, they were subjected to a new equilibration solution for 15 minutes, where DTT was replaced by 243 mmol/L iodoacetamide (Sigma-Aldrich). After equilibration, the strips were applied on 13.75% acrylamide/N,N′-methylenebisacrylamide (Bio-Rad) gel, and the electrophoresis was carried out overnight. Separation in the second dimension of electrophoresis was carried out on a Protean II Cell (Bio-Rad) using 13.75% SDS-polyacrylamide gel with 5% stacking gel at a temperature of 15°C and an initial voltage of 35 V for 7.5 hours, 65 V for 8 hours, and 95 V until the electrophoresis was finished.
After the second dimension electrophoresis, the analytic gels were silver stained, dried, and analyzed manually. Spots overexpressed in the invasive phenotype in both patients were selected. Thereafter, micropreparative gels with higher protein load were prepared and stained with SYPRO Ruby (Bio-Rad) immediately after the SDS-PAGE electrophoresis according to the manufacturer's protocol. The gels were incubated in the SYPRO Ruby overnight. The images of SYPRO Ruby–stained gels were obtained on image analyzer LAS-100 (Fuji, Tokyo, Japan).
Mass spectrometry. After manual excision, the differentially expressed spots were stored in -80°C until further analysis. During the preparation of protein samples for MS, the gel pieces were washed in NH4HCO3/acetonitrile (Sigma-Aldrich) solution and dried in speed-vac for 20 minutes and in-gel trypsin digestion (Promega, Madison,WI) was done overnight at 37°C. Thereafter, the peptides were extracted, lyophilized, reconstituted, and mixed with α-cyano-4-hydroxycinnamic acid (Promega) matrix solution directly on the MALDI target plate. Peptide mass spectra were generated on an Ultraflex MALDI-TOF (Bruker Daltonics, Bremen, Germany). The experimental peptide mass spectrum was matched to the theoretical spectra using a peptide mass fingerprinting technique and a MASCOT search engine ( 14). A probability-based scoring, showing a match between the experimental data and mass values calculated from candidate peptide sequences, was then obtained.
Western blot. The tissue pieces were homogenized in homogenization buffer, containing CHAPS, K2HPO4, KH2PO4, EDTA (Merck), and protease inhibitor cocktail (Complete, Roche Diagnostics, Manheim, Germany) and centrifuged and the supernatant was transferred into a new tube before the protein concentration was estimated. The samples were kept frozen at -80°C until use.
A 12.5% gel was prepared and 30 μg proteins were subjected to SDS-PAGE electrophoresis at 100 V for 2 hours. The bands were transferred onto a polyvinylidene difluoride transfer membrane Hybond-P (Amersham Biosciences). Expression of PDI was detected using an anti-PDI mouse monoclonal antibody (Alexis Biochemicals, Lausen, Switzerland) at a concentration of 1:1,000. The images were obtained on image analyzer LAS-100.
Immunohistochemistry. Monoclonal mouse antibodies to PDI were applied on formalin-fixed, paraffin-embedded tissue slices at a concentration of 1:100 and 1:200. The application of a secondary antibody and visualization was done with the DakoCytomation EnVision+ System-HRP kit according to the protocol of the manufacturer (DakoCytomation, Carpinteria, CA).
For immunofluorescence, the cells were fixed with paraformaldehyde in PBS, permeabilized with Triton X-100 (Sigma) in PBS, and incubated with primary PDI antibody at a concentration of 1:100. The secondary antibody used was isotope-specific goat anti-mouse conjugated to FITC (SouthernBiotech, Birmingham, AL).
Flow cytometry. U373 cells were grown as monolayers to 90% confluence in 80-cm2 cell culture flasks. Thereafter, the cells were dissociated with 2 mmol/L EDTA solution without trypsin. The cell suspension was then transferred to 15 mL centrifuge tubes and the cell numbers were determined using a Bürker chamber. Thereafter, the cell suspensions were centrifuged at 800 rpm for 5 minutes, the supernatant was discarded, and the pelleted cells were resuspended in fresh 4% paraformaldehyde in PBS. After fixation, 4 × 105 cells were transferred into 96-well multidishes (V-shaped bottom). The dishes were centrifuged at 1,000 rpm for 5 minutes. The supernatant was discarded and the cells were washed twice in PBS. The cells were not permeabilized. The cells were then incubated with a mouse monoclonal antibody to PDI at a dilution of 1:50 for 45 minutes on ice. After washing twice in PBS, the cells were incubated for 45 minutes with a FITC-conjugated goat anti-mouse F(ab′)2 secondary antibody (DAKO Denmark A/S, Glostrup, Denmark). All the steps were carried out on ice. The control cells were incubated with the secondary but not with the primary antibody. After incubation, the cells were washed again in PBS and stored at 4°C until analysis. The flow cytometry was done using a FACSCalibur flow cytometer (Becton Dickinson, San Diego, CA). To verify PDI expression at the cell surface, nonpermeabilized cells were also examined by confocal microscopy. The same cells, as used for the flow cytometric analysis, were mounted on slides using 4′,6-diamidino-2-phenylindole (DAPI) containing mounting medium (Vectashield, Vector Laboratories, Inc., Burlingame, CA) whereupon they were analyzed using a Zeiss confocal microscope (Zeiss LSM 510 Meta, Jena, Germany).
Adhesion assay. The cells were grown in 80-cm2 culture flasks and nonenzymatically detached from the bottom as described before. Thereafter, they were centrifuged at 800 rpm for 5 minutes and the pellet was resuspended in 5 mL medium without serum. The cells were then counted and incubated for 1 hour on ice with a monoclonal PDI antibody diluted in complete medium without serum. The control cells were incubated in complete medium without serum. After 1 hour, the cells were counted and equal number of control and treated cells was seeded in 5 parallel wells of 24- and 48-well multidishes. The number of attached cells was determined after 2 and 24 hours. The experiment was repeated four times with different number of seeded cells.
Migration assay. To assess functional variables related to PDI expression, six spheroids derived from the highly invasive rat xenograft (low-generation tumor) and six spheroids derived from the less invasive, angiogenic xenograft (high-generation tumor) were placed in 24-well plates, one in each well and overlaid with 0.5 mL medium. After 4 hours, 0.5 mL medium containing 10 mmol/L bacitracin (Sigma), which is a specific inhibitor of PDI ( 15, 16), was added to three wells containing low-generation and three containing high-generation spheroids to a final concentration 5 mmol/L bacitracin. The remaining six wells (three with low-generation spheroids and three with high-generation spheroids, respectively) represented the control group, which received normal growth medium. The experiments were done twice.
In one set of experiments, bacitracin was withdrawn from the cell cultures (treatment day 0-2, treatment day 0-4, and treatment day 0-7). There were 6 spheroids in each treatment group (total of 24 spheroids).
To verify that the migration inhibition seen by bacitracin represented a PDI-specific effect, we also did the same experiment using a PDI-specific antibody that inhibited the expression of PDI on U373 cells (which express high levels of PDI). There were three U373 spheroids in each treatment group [i.e., treatment with 2.5 and 5 mmol/L bacitracin, treatment with PDI antibody at the dilution of 1:10, 1:50, 1:100, and 1:500, and 6 spheroids in the control group (total of 24 spheroids)]. The experiments were done twice and terminated after 4 days.
The diameter of the spheroids was measured at the time of plating, and the cell migration area was measured daily using an inverted phase-contrast light microscope with a micrometer in the ocular. Cell migration was calculated as the distance from the center of the spheroid to the population of migrating cells most distant from the spheroid minus the original spheroid radius. The medium was changed every second day. The longest treatment with bacitracin was 7 days. Measurements were recorded for up to 13 days. The monoclonal PDI antibody was diluted in medium and added to the culture in the same way as for bacitracin. In the dose response studies, the following bacitracin concentrations were used: 2.5, 5, and 10 mmol/L. The experiments with monoclonal antibody to PDI were carried out in 96-well plates containing 160 μL medium.
In the experiments where possible additive effects of bacitracin and β1 and β3 integrin blocking antibodies on glioma cell migration were tested, the U373 glioma cell line spheroids were placed in 48-well multidishes, one in each, and covered with 200 μL medium. After 4 hours, when the spheroids were attached to the bottom, the normal medium was replaced with medium containing bacitracin at concentrations of 2.5 and 5 mmol/L with or without β1 integrin blocking antibody (dilution 1:10, 1:50, and 1:100; SouthernBiotech). The same experiment was also done using a β3 blocking antibody (Chemicon International, Temecula, CA). Cell migration was then measured at regular intervals as decribed above.
Invasion assay. The collagen invasion assay was carried out as described previously ( 17). Briefly, the collagen solution was prepared by mixing 3.2 mg/mL collagen type I in 0.012 mol/L HCl (Vitrogen, Cohesion, Palo Alto, CA) and 10-fold concentrated DMEM without FBS and antibiotics. The pH of the solution was adjusted using 0.2 mol/L NaOH and 500 μL of this solution was added to 24-well plates. Twelve spheroids (six treated and six untreated in the control group) were implanted into the gel so that the spheroids were completely surrounded by the collagen matrix. After gelation at 37°C, the gel was overlaid with 500 μL supplemented DMEM and the inhibitor. When bacitracin was added for the first time, a concentration of 10 mmol/L of bacitracin was applied to saturate the collagen phase. During medium replacement, 5 mmol/L bacitracin in DMEM was used. Cell invasion distances were measured using an inverted phase-contrast light microscope with a micrometer in the ocular. Invasion was measured daily, and medium was renewed every 3rd day. To measure the invasion distance in the collagen gel, the same criteria as used in the migration assays were applied. Measurements were taken for 10 days. The longest treatment with bacitracin was 9 days. The experiment was done twice.
Statistical analyses. A migration distance of treated and untreated spheroid cells was compared with Student's t test (independent samples, two tailed). Type I error level was set to α = 0.05.
Human tumor xenografts in nude rats. The tumor biopsies were taken from an area of contrast enhancement on MRI and characterized by an extensive cellular pleomorphism with extensive angiogenesis and areas of necrosis ( Fig. 2A ). The low-generation tumors developed after 4 to 5 months as verified by MRI imaging ( Fig. 2B, top). These tumors were highly cellular and showed little evidence of angiogenesis and necrosis ( Fig. 2B, bottom). On serial transplantation in vivo, a new MRI contrast-enhancing tumor phenotype emerged, which was less invasive and displayed extensive angiogenesis and necrosis ( Fig. 2C).
Proteomic experiments. Comparative proteome analysis of the xenografts, displayed in Fig. 2B and C, revealed several differentially expressed spots in the low-generation compared with the high-generation tumors ( Fig. 3 ). Only the spots found in both patient samples showing the same tendency of differential expression were subjected to further analysis. To identify differentially expressed proteins, we did 2DE of micropreparative gels with higher protein load. These gels were stained with SYPRO Ruby; selected spots were manually excised and subjected to trypsinization before further analysis using MS. All of the identified proteins were human. Among the identified proteins were structural proteins, stress-related proteins, brain-specific proteins, extracellular matrix (ECM) proteins, metabolic proteins, and others. One of the identified proteins with a significant score was PDI A6 precursor (PDA6_HUMAN, Swiss Prot No: Q15084, Fig. 3). PDI has been reported previously to be expressed on the cancer cell surface ( 18, 19) and because we identified a member of the PDI protein family in the invasive glioma cells, we decided to further characterize this protein both in the xenografts as well as in human brain tumors.
PDI is expressed in the human xenografts. PDI was shown by immunohistochemistry and Western blots to be overexpressed in the invasive low-generation tumors ( Fig. 4A and B ). Although some PDI staining was also observed in the less invasive, high-generation tumors, the expression was mainly confined to the tumor periphery ( Fig. 4A, bottom). The immunoblots verified the immunohistochemical observations ( Fig. 4B).
PDI is expressed in human glioblastomas. Immunohistochemistry of glioblastoma biopsy specimens collected from the invasive edge (see Fig. 2A) revealed strong PDI expression in the invasive tumor rim ( Fig. 4C). Next, to verify that PDI was expressed on migrating tumor cells, we immunostained the migrating cells using a spheroid migration assay ( Fig. 4D). As seen in Fig. 4D, PDI was strongly expressed by the migrating cells.
PDI is expressed on the surface of glioblastoma cells. The surface localization of PDI on U373 glioblastoma cells was verified by flow cytometry. Compared with the controls, the flow cytometric analysis revealed a significant increase in fluorescence intensity in the nonpermeabilized U373 cells [mean fluorescence intensity (MFI), 86.37] compared with the control cells (MFI, 3.39; Fig. 5A ). The same cells were also studied by confocal microscopy, showing PDI to be expressed at the cell surface ( Fig. 5A).
Cell adhesion assay. To determine if cell surface PDI expression influences glioma cell adhesion, we incubated U373 cells with a PDI antibody before seeding them on a plastic surface. Cell attachment was determined 2 and 24 hours after seeding. Already after 2 hours, some cells were observed attaching to the plastic surface and the number of attached cells increased with time. However, the number of attached cells was clearly lower in the dishes, where cells were pretreated with the PDI antibody. This difference was seen both after 2 and 24 hours ( Fig. 5B).
PDI inhibition reduces tumor cell migration. Because PDI was expressed by the invasive tumor cells in vivo and by migrating glioma cells in vitro, we assessed the effect of a well-known PDI inhibitor, bacitracin ( 15, 16), on glioma cell migration and invasion.
Our previous experiments showed that PDI was stronger expressed in the low-generation, invasive tumors. We therefore tested whether the low-generation glioma spheroids were more sensitive to PDI inhibition by bacitracin than those derived from the high-generation tumors. We exposed both low- and high-generation spheroids to 5 mmol/L bacitracin, 4 hours after plating them into 24-well plates. The results show that cell migration from both low- and high-generation spheroids was inhibited by bacitracin and that the inhibition was stronger in cells derived from low-generation spheroids, as an almost complete inhibition of migration was obtained ( Fig. 6A ). In contrast, the treated high-generation spheroids attached to the bottom and some cells migrated, but the migration distance was strongly reduced compared with the control group. The inhibitory effect of bacitracin on the high-generation spheroids was more pronounced after 3 days, and the difference in migration distance compared with the control group increased with time ( Fig. 6A).
Figure 6B shows the difference in sensitivity to bacitracin between low- and high-generation spheroids. The almost complete inhibition of low-generation spheroid migration was defined as 100%. The relative migration inhibition was calculated according to the following equation: Relative migration inhibition = (average migration distance of untreated spheroids − average migration distance of treated spheroids) × 100% / average migration distance of untreated spheroids.
To eliminate possible toxic effects of 5 mmol/L bacitracin exposure, we investigated whether its inhibitory effect on cell migration was reversible. We divided spheroids into four groups, six spheroids in each group. Control spheroids did not receive any treatment, whereas other groups were treated for 2, 4, and 7 days before bacitracin removal. The results show that removal of the inhibitor caused an immediate migration of glioma cells similar to that seen in the untreated group. Thus, the effect of 5 mmol/L bacitracin was reversible ( Fig. 6C).
A concentration of 10 mmol/L bacitracin had a toxic effect on the spheroids, characterized by no attachment of spheroids to a plastic surface and spheroid dissociation (data not shown). Bacitracin (5 mmol/L) had a clear effect on both high-and low-generation spheroids, as well as on glioma cell line spheroids without causing toxicity ( Fig. 6C). Bacitracin at the concentration of 2.5 mmol/L also reduced the glioma cell migration, but the effect was weaker than that seen by 5 mmol/L ( Fig. 6D).
Because bacitracin is a chemical inhibitor that may unspecifically interfere with other intracellular processes not related to PDI, we also wanted to see if a PDI-specific biological inhibitor showed similar results to those obtained for bacitracin. We therefore tested the inhibitory effect of a monoclonal antibody against PDI on a highly invasive and tumorigenic glioblastoma cells U373 ( 11), which express high levels of PDI (data not shown). As shown in Fig. 6D, the monoclonal antibody inhibited cell migration in a dose-dependent manner. The inhibitory effect of the PDI antibody was significant (P ≤ 0.05) but less than that observed using 5 mmol/L bacitracin.
Bacitracin and β3 integrin antibody show additive effects on cell migration. The effects of β1 and β3 integrin blocking antibodies on glioma U373 cell migration were determined and the inhibitory effects were compared with those obtained using 2.5 mmol/L bacitracin. Blocking the β1 integrin had no effect on cell migration (data not shown). In contrast, the blocking β3 integrin antibody reduced migration similar to that observed using 2.5 mmol/L bacitracin. Moreover, when both bacitracin at the concentration of 2.5 mmol/L and β3 integrin antibody at the dilution 1:100 were used together, the inhibitory effect was stronger than that observed if the agents were used separately ( Fig. 6F).
Bacitracin inhibits glioma invasion in vitro. In addition to the inhibitory effect of bacitracin on glioma cell migration in vitro, we confirmed that bacitracin reduces glioma cell penetration into collagen gels, which indicates an inhibitory effect on glioma invasion in vitro. We applied bacitracin at a concentration of 5 mmol/L on six low-generation spheroids, placed in the collagen gel. There was clear inhibition of cell penetration into the collagen gel ( Fig. 6E).
At present, several experimental models have been developed to study glioma cell invasion in vitro and in vivo ( 20, 21). In our study, we used a unique animal model, where the glioma phenotype changed, in the nude rat brain, by serial xenotransplantation of human glioblastoma biopsy spheroids. The first tumors that developed showed no obvious signs of angiogenesis and displayed a highly infiltrative phenotype (low-generation tumors). By serial transplantation in rats, the tumor became highly angiogenic but showed less invasion (high-generation tumors). Thus, from one tumor sample, two distinct phenotypes were obtained, in which each mimicked different aspects of human brain tumor biology in situ. This allowed a direct comparison of differentially expressed proteins characterized by the phenotypic shift from low-generation to high-generation tumors. Moreover, we compared such phenotypic shifts in two patients and only proteins that were up-regulated in both patients were analyzed by MS. We then used proteomics to identify protein fingerprints that characterize the two phenotypes. The 2DE method has its limitations because only the most abundant proteins are identified ( 22, 23). To optimize the sensitivity, we used silver staining for the analytic gels and MS compatible SYPRO Ruby for micropreparative gels. Among the proteins that were identified, we found structural proteins, stress-related proteins, brain-specific proteins, metabolic proteins, and ECM proteins (data not shown). According to the MASCOT search engine, all of the identified proteins were human.
One of the proteins identified was PDI A6 precursor (synonyms: PDI P5, thioredoxin domain-containing protein 7, human P5). PDI A6 is a 440-amino acid protein, belonging to the PDI family. Human P5 contains two thioredoxin-like domains, like human PDI, and these domains share 47% amino acid sequence identity in PDI and P5. P5 belongs to the PDI protein family ( 24), and PDI is expressed on the cancer cell surface. We verified the presence of PDI on the cell surface using flow cytometry and confocal microscopy. We also verified that PDI was differentially expressed in the two phenotypes. It has been proposed that P5 is functionally related to PDI, which is a chaperone protein, involved in the protein quality control machinery ( 25). Its major role is in rearrangement of both intrachain and interchain disulfide bonds, where it acts catalytically to both initiate and reduce disulfide bonds. It has been postulated that PDI, on the cell surface, acts as a reductase cleaving the disulfide bonds of proteins attached to the cell surface and inside the cell it forms disulfide bonds ( 26). It may also control the function of certain extracellular proteins by regulating their red-ox state ( 27).
P5 cDNA was first isolated from a hydroxyurea-resistant hamster cell line ( 28). It has been proposed that P5 plays an important role in acquiring hydroxyurea resistance ( 29). It has also been shown that PDI is up-regulated in glial cells during hypoxia/brain ischemia, and a role of PDI in apoptosis resistance has been postulated ( 25, 30). Here, we identified this protein to be strongly expressed by the most invasive glioma phenotype.
Immunohistochemistry and immunoblots revealed that PDI was expressed in both phenotypes. However, there was a quantitative difference in its expression between the two phenotypes. In the low-generation tumors, PDI was extensively expressed throughout the tumor mass, whereas, in the high-generation tumors, PDI was confined to the tumor periphery and was not present in the angiogenic core. The same observations were made in the biopsy specimens obtained from patients, where PDI was expressed in the tumor periphery. PDI was also expressed by migrating tumor cells in vitro, suggesting that PDI has a functional role in glioma cell migration. To address this issue, we exposed the cells derived from the two phenotypes to a PDI-specific inhibitor, bacitracin ( 15). Interestingly, bacitracin had an inhibitory effect on migration and invasion of tumor cells derived both from the low- and high-generation tumors. However, the effect was strongest on the invasive phenotype ( Fig. 6A and B).
It has been suggested that PDI functions at the cell surface, where it may be involved in disulfide exchange required for cell-mediated adhesion by integrins ( 31). Several integrins have been linked to glioma cell invasion, and a particular focus has been on the β1 and β3 integrins ( 1). Interestingly, it has been shown that PDI can mediate conformational changes in both β1 and β3 integrins, which may lead to cell adhesion to a particular substrate ( 32). Here, we show that a blocking β3 integrin antibody has an inhibitory effect on U373 cell migration in vitro and that this effect does not abrogate further migration inhibition by bacitracin.
In addition, a direct intervention of PDI in cellular adhesion has been shown in retina cells from chicken embryo, where a cell surface adhesion protein was identified to be PDI ( 16). In our experiments, we have shown that incubation of glioblastoma cells with PDI antibody inhibits adhesion of U373 cells to the plastic surface, indicating a role of PDI in glioma cell adhesion.
The fact that a PDI-specific antibody also inhibited cell migration suggests a functional role of PDI in glioma cell migration and invasion. The inhibitory effect using the PDI antibody was less than that observed using bacitracin. This observation is consistent with earlier reports showing a 49% inhibition of PDI activity by using the PDI antibody ( 15).
In conclusion, using a novel human glioblastoma xenograft model and proteomics screening technology, we were able to identify a novel protein, PDI, which is strongly expressed by invasive glioma cells. Moreover, inhibition of PDI led to reduced glioma cell migration and invasion. PDI may therefore represent a potential therapeutic target in malignant brain tumors.
Grant support: Norwegian Cancer Society, The Norwegian Research Council, Centre Recherche Public de la Santé, Luxembourg, Helse Vest Haukeland University Hospital, and the sixth European Union Framework Programme grants (integrated project “Angiotargeting,” contract no 504743 in the areas of “Life sciences, genomics, and biotechnology for health”).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Erna Finsås for technical assistance with protein electrophoresis and Western blot, Tove Johansen for the help with cell culture, Tore Jacob Raa and Linda Vabø for taking care of the animals, and colleagues at PROBE for all the help and expertise in doing MS. 4
- Received December 27, 2005.
- Revision received June 21, 2006.
- Accepted August 4, 2006.
- ©2006 American Association for Cancer Research.