Proteomics-based analysis of invasion-related proteins in malignant gliomas

Cell Culture

The J3T canine glioma cell was a generous gift from Michael E. Berens MD, PhD (Translation Genomics Research Institute, Phoenix, AZ, USA).³ Two cell lines—J3T-1 and J3T-2—were developed from the parental J3T cell line, as previously described.^1,4 In brief, J3T cells (5 × 10⁶ cells) were subcutaneously implanted into the flanks of two athymic mice (NCr/Sed, nu/nu, 20 g). After 6 weeks, two tumors were established in each animal. These tumors were harvested in a sterile fashion, minced with a scalpel into 1-mm³ cubes, treated for 1 h with 1 mg/ml Collagenase/Dispase (Roche, Basel, Switzerland), and subsequently cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 units penicillin, and 0.1 mg streptomycin per milliliter. Both cell lines were derived from single subcutaneous tumors.

For enhanced visualization of J3T-1 and J3T-2 cells, cell lines were established that stably expressed green fluorescent protein (GFP), as previously described.¹ In brief, J3T-1 and J3T-2 cells were transfected with the pAcGFP1-C1 plasmid (Clontech Laboratories Inc, Mountain View, CA, USA), which encodes for GFP using the TransIT-LT1 reagent (Takara Bio Inc, Otsu, Japan) to obtain J3T-1G and J3T-2G, respectively. Cells were cultured in DMEM supplemented with 10% FBS, 100 units penicillin, and 0.1 mg streptomycin per milliliter in a standard tissue incubator at 37°C with a 5% CO₂ atmosphere.

2DGE

The cells were washed with a solution of phosphate-buffered saline (PBS) that contained 0.27% CaCl₂ • H₂O, 0.02% MgCl₂ • 6H₂O, 1% FBS, and 5 mM sodium pyruvate. Cells were harvested by mechanical scraping during the exponential growth phase. Cells were then centrifuged, and the cell pellets were dissolved in a lysis buffer that consisted of 5 M urea, 2 M thiourea, 2% 3-(3-cholamidopropyl)dimethylammonio-1-propane sulfonate (CHAPS), 2% sulfobetaine 3-10 (SB3-10), 1% dithiothreitol (DTT), and a protease inhibitor cocktail (Sigma-Aldrich, St. Louis, MO, USA). After three freeze-thaw cycles, the cell pellets were sonicated for 30 s and ultracentrifuged at 75,000 ×g for 30 min at 10°C using the Optima™ TLX ultracentrifuge (Beckman Coulter, Brea, CA, USA). The supernatant was transferred to a new tube and treated with ReadyPrep 2D Cleanup Kit (Bio-Rad, Hercules, CA, USA) to remove any ions, DNA, RNA, etc. The protein concentration was estimated using the RC-DC Protein Assay kit (Bio-Rad) according to manufacturer with two-washed standard protocol. For isoelectric focusing (IEF), Carrier ampholyte 5/8 (Bio-Rad) was formulated to increase the resolution at the basic end of the flatbed isoelectric focusing gel.

The first-dimensional IEF was performed using a 17-cm immobilized nonlinear pH gradient (5–8) strip (DryStrip; Bio-Rad). After rehydration for 15 h in a 300-μl buffer that consisted of 5M urea, 2M thiourea, 2% CHAPS, 3% SB3-10, 1% DTT, and 0.2% Bio-Lyte^® 5/8 ampholyte (Bio-Rad), 60-μg samples of each protein were loaded onto the strips. Focusing was performed in following step; (I) 250 V with linear increase for 40 min, (II)10000 V with linear increase for 4 h and (III) 10000 V with rapid increase for 7 h (total 70000V-hr); the samples were then maintained at 500 V, as needed.

The focused strips were then equilibrated in buffer I (6 M urea, 2% sodium dodecyl sulfate (SDS), 0.375 M Tris-HCl, 20% glycerol, and 2% DTT; pH 8.8) for 30 min and then buffer II (6 M urea, 2% SDS, 0.375 M Tris-HCl, 20% glycerol, and 2.5% iodoacetamide; pH 8.8) for 15 min with gentle shaking. The second-dimensional separation was carried out on 12% SDS-polyacrylamide gels using PROTEAN II Cell (Bio-Rad) at 20°C and 40 mA/gel for 4 h.

After 2DGE, the gels were stained with SYPRO Ruby (Invitrogen, Carlsbad, CA, USA) and Dodeca Silver Stain Kits (Bio-Rad), according to the manufacturer’s protocol. Each experiment was performed in triplicate.

Image Analysis of 2DGE Gels

Images of the SYPRO Ruby-stained gels were obtained using a FLA-3000 image analyzer (Fujifilm, Tokyo, Japan). The images were analyzed using PDQuest Advanced, version 8.0 (Bio-Rad); analysis included background subtraction, spot detection, and volume normalization. The intensity of each spot was quantified by calculating the spot volume after normalization to the local regression model. The intensities of three sets of matched spots of each J3T-1 and J3T-2 gel were compared, and a threshold ratio of 1.5 was used to screen for differences. Visual inspection confirmed the differences indicated by PDQuest analysis.

Protein Identification

The selected protein spots were manually excised from the gels after visualizing the spots with silver staining. After destaining and drying the excised spots, in-gel trypsin digestion using the Protein In-Gel Tryptic Digestion kit (Agilent Technologies, Santa Clara, CA, USA) was performed overnight at 30°C. The obtained peptide digests were analyzed using a fully automated nanoflow liquid chromatography-ion trap-tandem mass spectrometer (nLC-IT-MS/MS; Agilent 1100 LC/MSD Trap XCT Ultra, Agilent Technologies).

Protein identification was performed using the Spectrum Mill MS Proteomics Workbench platform (version A.03.02; Agilent Technologies), according to the workflow parameters specified by the manufacturer. The identification parameters were set as follows: database, NCBInr; species, Canis familialis; enzyme, trypsin; monoisotopic masses used; precursor mass tolerance (peptide tolerance), ±2.5 Da; product mass tolerance (MS/MS tolerance), ±0.8; fixed modification, carbamidomethylation (cysteine); and variable modification, oxidation (methionine). Two missed cleavages with trypsin were allowed and the instrument’s setting was specified as ‘‘ESI ion trap’’. The probability scores calculated by the software were used as the criteria for correct identification.

Quantitative reverse-transcription PCR (QRT-PCR)

Total RNA was isolated from cultured J3T-1 and J3T-2 cells using RNeasy^® Mini Kit (QIAGEN, Valencia, CA, USA) and reverse-transcribed with oligo dT primers using SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen), according to manufacturer’s instructions. Primers specific for each target gene were designed using Primer Express (Applied Biosystems, Foster City, CA, USA) and synthesized by Invitrogen. The resulting cDNA was amplified by PCR using gene-specific primers and the 7300 Real Time PCR system (Applied Biosystems) and QuantiTect™ SYBR^® Green PCR kit (QIAGEN). A log-linear relationship between the amplification curve and quantity of cDNA in the range of 1 to 1000 copies was observed.

The cycle number at the threshold was used as the threshold cycle (Ct). The different expression of mRNA was deducted from 2^-ΔΔCt using the 7300 Real Time PCR System with the Sequence Detection software (version 1.4; Applied Biosystems). The amount of cDNA in each sample was normalized to the crossing point of the housekeeping gene GAPDH. The following thermal cycling parameters were used: denaturation at 95°C for 15 min, followed by 40 cycles of 30 s at 94°C, 30 s at 55°C, and 1 min at 72°C. The relative mRNA upregulation for each gene in the J3T-2 cells was calculated using their respective crossing points in the following formula:

• where, F = fold difference, T = J3T-2 cells, O = J3T-1 cells,

• H = housekeeping gene (GAPGH), and G = gene of interest.

Immunohistochemistry

All experimental animals were housed and handled in accordance with the Okayama University Animal Research Committee guidelines. Before implantation, 85–90% of the confluent J3T-1, J3T-2, J3T-1G, and J3T-2G cells were trypsinized, rinsed with a solution of DMEM and 10% FBS, and centrifuged at 800 rpm for 5 min. The resulting pellets were resuspended in PBS and the concentration was adjusted to 1 × 10⁵ cells/μl of PBS. To establish the brain tumor models, athymic rats (F344/N-nu/nu; CLEA Japan, Inc., Tokyo, Japan) were anesthetized with an intraperitoneal injection of nembutal (30 mg/kg) and placed in a stereotactic apparatus (Narishige, Tokyo, Japan). Tumor cells (5 × 10⁵ cells/5 μl) were slowly injected into the basal ganglia of the right cerebral hemisphere (4 mm lateral and 1 mm anterior to the bregma at a depth of 4 mm) using a Hamilton syringe (Hamilton, Reno, NV, USA), according to previously published procedures.^1,5 For histological examination, the athymic rats (J3T-1, n = 5; J3T-2, n = 5) were sacrificed 4–5 weeks after tumor inoculation. Before sacrifice, the animals were anesthetized. Death was induced via cardiac puncture, then the animals were perfused with 100 ml PBS and fixated with 200 ml of 4% paraformaldehyde. The brains were removed and stored in 4% paraformaldehyde for at least 24 h. Snap-frozen tissue samples were embedded in an optimal cutting temperature compound for cryosectioning. Sixteen-micrometer cryostat sections were processed for indirect immunofluorescence analysis. The slides were incubated with 10% horse serum in PBS at room temperature and then with the anti-annexin A2 antibody (1:20 mouse IgG; Zymed Laboratories, San Francisco, CA, USA) diluted with 1% horse serum in PBS, and stored overnight at 4°C. After three washes with PBS for 5 min, the slides were incubated with Cy3-conjugated antibodies against mouse IgG (1:300) (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA) and 4′, 6-diamino-2-phenylindole in PBS (1:500; Invitrogen) for 60 min. The slides were then washed with PBS and mounted.

Human glioma samples were intraoperatively obtained from patients undergoing surgical removal of tumors for diagnostic and therapeutic purposes (Okayama University Hospital, Okayama, Japan). Written informed consent was taken in all the cases before surgery. Lobectomies were performed to remove the entire tumor and any surrounding brain parenchyma beyond the boundaries of the tumors observed on MRI. All glioma specimens were diagnosed and graded according to the World Health Organization classification of tumors of the central nervous system, and all five specimens were diagnosed as glioblastomas. No patient had received radiation or chemotherapy before surgery.

After deparaffinization in xylene and rehydration in decreasing concentrations of ethanol, 4-μm-thick sections were incubated in 0.3% hydrogen peroxide (30 min) and autoclaved with distilled water for 10 min at 121°C. After three washes in PBS, the sections were incubated with anti-annexin A2 monoclonal antibodies (1:20) (mouse IgG; Zymed Laboratories) that had been diluted with PBS and 5% skim milk (60 min) at room temperature. Secondary antibody was applied using the DakoCytomation Envision⁺ System-HRP kit, according to the manufacturer’s protocol (DakoCytomation, Carpinteria, CA, USA). The sections were visualized using 3,3'-Diaminobenzidine (DAB).

Annexin A2 expression was qualitatively scored as follows: (−) for no staining (nearly 0% of cells were labeled), (+) for a trace of positive cells (less than 30% of cells were labeled), (++) for moderately diffuse staining or sparsely intensive staining (30%–60% of cells were labeled, and less than 30% of cells were labeled with the strong intensity), and (+++) for strongly diffuse staining (60%–100% of cells were labeled). T.M. and S.I. classified the tumors according to 3 representative fields without prior knowledge of the patients’ clinical or radiological data.

2DGE patterns and differential analysis of J3T-1 and J3T-2

The 2DGE gel images of J3T-1 and J3T-2 cells showed similar protein expression patterns, probably reflecting the genetic similarity of J3T-1 and J3T-2 (Fig. 1).

Using the PDQuest 2DGE gel analysis software, approximately 2800 well-stained and clearly delineated protein spots were detected. Twenty-two spots showing more than 1.5-fold change in spot intensity; five of these protein spots were upregulated in J3T-1 and 17 were upregulated in J3T-2. These differentially expressed protein spots are illustrated with circles and spot numbers in Fig. 1A and 1B.

Proteins identified by LC-MS/MS

Differentially expressed protein spots were excised from the 2DGE gels and analyzed for protein identification by LC-MS/MS and database searching. As a result, 18 proteins were identified in those spots; these proteins are shown in Table 1. The Spectrum Mill software calculates the MS/MS Search score as described below. Each peak assigned to an allowed fragment ion type for a candidate peptide sequence is given ''bonus points''. In contrast, each unassigned peak is given ''penalty points''. Penalty value is based on peak height, - (peak height / height of tallest peak). MS/MS score is summation of these bonus points and penalty points, indicating probability of the protein. Thus, the higher score indicates the better probability. The identified proteins that were upregulated in J3T-1 included one metabolic enzyme, one storage protein, one structural protein, and one phospholipid-binding protein. Proteins that were upregulated in J3T-2 included eight metabolic enzymes, two translational regulatory proteins, two RNA-binding proteins, one GTP-binding protein, and one signal transduction protein. Metabolic enzymes constituted more than half of the proteins that were upregulated in J3T-2.

Fig. 2A and 2B shows the distribution of all the 18 identified proteins that were upregulated in J3T-1 and J3T-2 according to their general functions as described in NCBI Entrez Gene Database (http://www.ncbi.nlm.nih.gov/sites/entrez?db=gene) and Human Protein Reference Database (http://www.hprd.org/index_html). Fig. 2C shows the representative MS/MS spectrum for the peptide sequence matched to annexin A2 identified from spot No.4.

Candidate protein validation by QRT-PCR

A summary of the QRT-PCR data is shown in Fig. 3. Similar patterns were observed for the gene expression profiles obtained with the proteomics and QRT-PCR data. The directionality of expression of 17 of the 18 genes coincided with the expression of the corresponding proteins. Fourteen of 18 genes demonstrated more than 1.5-fold change in expression (as denoted by the asterisks [*] in Fig. 3), while the rest of the proteins did not show marked difference in expression between J3T-1 and J3T-2 cells.

Immunohistochemical corroboration of candidate proteins observed in the invasive animal models

Representative photomicrographs of the invasive animal models are presented in Fig. 4A, B. J3T-1 cells formed well-demarcated and highly angiogenic tumors. In contrast, J3T-2 cells formed tumors with obscure margins via single-cell infiltration into the normal brain parenchyma.

The results of this proteomics-based analysis of the proteins involved in invasion were corroborated by immunohistochemical analysis of the J3T-1G and J3T-2G xenograft models. One of the identified proteins was annexin A2, which has been reported to be involved in tumor invasion and metastasis.^6–8 Therefore, this protein was further studied using both animal models and human glioma samples. The annexin A2 antibody was readily available for immunohistochemical analysis of the frozen sections. Intense staining of the annexin A2 protein in the cytoplasm of the J3T-1 tumor cells was observed, whereas almost no staining was observed in the J3T-2 tumor cells (Fig. 4 C-H), as expected based on the results of the proteomic analysis.

Immunohistochemical analysis of annexin A2 in human glioma samples

To analyze expression pattern of annexin A2 in relation to vasculature, immunohistochemical staining with anti-annexin A2 antibody was performed in 5 human glioblastoma samples. Patient’s age, sex, tumor locations, and degree of annexin A2 expression are summarized in Table 2. The center of the tumor comprised an area of high-density tumor cells which were negative or weakly positive for annexin A2. Marked angiogenesis, which was characterized by thick endothelial proliferation, was seen in the center and at the tumor borders. At the borders, clusters of annexin A2-positive tumor cells were observed around dilated vessels in all cases (Fig. 5A). Furthermore, diffuse single cell infiltration from the tumor core to the surrounding normal brain parenchyma, which was independent of vascular structure, was observed. Those cells were mostly negative (four cases) or weakly positive (one case; Patient 3) for annexin A2 (Fig. 5B).

Advantages of the proteomic method for identifying invasion-related proteins

This study describes how four proteins were observed in J3T-1 cells and 14 proteins were observed in J3T-2 cells, out of 2800 protein spots that were reproducibly identified on 2DGE gels at pH 5–8.

The similarities in the protein expression patterns on the 2DGE gel images of J3T-1 and J3T-2 cells reflect their genetic proximity. At the same time, significant differences in protein expression between these two cell lines do exist, which could be directly related to their invasive characteristics. Taken together, the proteomic method could be used to sensitively capture the features of the cell lines used in this study.

The results of these proteomic analyses were corroborated by the results of the QRT-PCR and immunohistochemical analyses. Nearly all proteins that were shown to be upregulated by the 2DGE method were also upregulated at the mRNA level. This method is simple, reproducible, and reliable and may prove useful for gaining a broader understanding of invasion in general as well as some of the advantages and pitfalls of other in vitro modeling systems.

As a shortcoming, 2DGE analysis usually cannot detect hydrophobic membrane proteins, proteins larger than 100 kDa, low-abundance proteins,⁹ or proteins with pI values outside the pH range.¹⁰ These limitations of standard 2DGE probably account for the less chance of this study to identify these types of proteins in J3T-1 and J3T-2 cells.

Advantages of the two subclones for proteomic analysis

There are several advantages in using invasive animal models and proteomic analysis to study the molecular basis of glioma invasion.

First, the animal models used in this study are one of the few models available for studying invasive glioma in experimental settings.^1,2 Traditional animal glioma models do not exhibit invasive growth into the surrounding normal brain areas and have been criticized for not recapitulating the main pathological features of human glioma.¹¹ Recently, several experimental glioma models showing varying degrees of invasion have been developed to study glioma cell invasion both in vitro and in vivo.^12,13

Second, the two subclones that were established from the same parental cell line showed very different invasive phenotypes when inoculated in rat brains; however, the subclones have inherently similar genetic backgrounds.⁴ These differences in phenotypes are supposed to arise from differences in genotype. This allowed for direct comparison of the differentially expressed proteins that characterize the phenotypic differences in their invasive capabilities.

Third, the animal models used in this study are reproducible and suitable for in vivo experiments. Therefore, candidate proteins that might affect the invasive capability of the tumor can be directly verified in vivo using these animal models.

Annexin A2 as a candidate invasion-related protein

One of the proteins identified, annexin A2, was expressed at higher levels in J3T-1 cells than in J3T-2 cells. Annexin A2 has been previously reported to be expressed on the tumor cell surface and to be involved in tumor invasion and metastasis.^6–8 Therefore, this protein’s role in both animal models and human glioma samples was further examined.

Annexins are a superfamily of closely related calcium- and membrane-binding proteins that are expressed in specific cell types. Annexins comprise numerous gene products, including 12 that are formed in vertebrates (group A annexins), and have been proposed to be involved in a variety of functions, including control of membrane structures, vesicle trafficking, cell division, apoptosis, calcium signaling, and growth regulation.^14–20 All annexins share a characteristic molecular structure that consists of a variable N-terminal domain and a highly conserved C-terminal domain, which is also called the annexin core. The core domain of annexin A2 harbors several Ca²⁺-binding sites and one actin-binding domain, whereas the N-terminal domain contains the binding site for the cellular ligand S100A10 and a Ca²⁺-independent membrane-binding domain. In cells, annexin A2 exists either as a monomer or as a heterotetrameric complex with S100A10.²¹ Cell-surface annexin A2 has been shown to be a receptor/binding protein for both proteases (cathepsin B, plasminogen, and tissue-type plasminogen activator [tPA]) and extracellular matrix proteins (type I collagen and tenascin C) that are involved in tumor invasion and metastasis.^6–8 tPA has been shown to bind specifically to annexin A2 on the extracellular membrane of pancreatic cancer cells where it activates plasmin production and promotes tumor cell invasion. Annexin A2 is also expressed on the surface of vascular endothelial cells and has been shown to play a significant role in the regulation of tPA activities, plasminogen activation, and angiogenic functions.^22,23 Increased expression of annexin A2 has been described in several types of tumor, including gastric carcinoma,²⁴ colorectal cancer,²⁵ pancreatic cancer,²⁶ breast cancer,²⁷ high-grade gliomas,²⁸ kidney cancer,²⁹ and vascular tumors.³⁰ In gliomas, annexin A2 has been identified more frequently in high-grade gliomas than in low grade gliomas¹⁰ and is known to be an independent prognostic factor for poor clinical outcomes in patients with gliomas.^28,31 Because of the ability of annexin A2 to interact with the actin cytoskeleton^14,15,32 and tumor-released proteases,¹⁵ annexin A2 has become a promising candidate for influencing the invasion processes of glioma cells. Recent proteomic investigations of the pseudopodia extensions of highly migratory U87MG glioma cells have reported increased levels of annexin A2.³³ Tatenhorst et al. demonstrated that the migration of human glioma cells is significantly inhibited following annexin A2 depletion by RNA interference techniques in vitro.³⁴ Moreover, cell-surface annexin A2 is a receptor for angiostatin, which is an internal fragment of plasminogen. Annexin A2-dependent localized plasmin generation could contribute to angiogenesis and metastasis. Therefore, by blocking annexin A2-dependent plasmin production, e.g., by using angiostatin, it may be possible to develop new anti-angiogenic therapies.³⁵

A previous study reported that there are at least two invasive and angiogenic phenotypes—angiogenesis-dependent invasion and angiogenesis-independent invasion—and the novel animal glioma models used in this study were able to exhibit these phenotypes.^1,2 The histopathological features of angiogenesis-dependent invasion are characterized by robust neovascular formation in the normal parenchyma adjacent to the main tumor mass and the formation of clusters of glioma cells. In contrast, angiogenesis-independent invasion is characterized by the formation of poorly demarcated tumors that show single-cell infiltrations into the surrounding normal brain tissue without angiogenic activities.

In the present study, annexin A2 was upregulated in J3T-1 cells, which typically show angiogenesis-dependent invasion in rat brain. In addition, immunohistochemical analysis of human glioblastoma samples revealed that annexin A2 is expressed in clusters of tumor cells around dilated vessels located at the border of the tumor, suggesting that annexin A2 has a functional role in angiogenesis-dependent invasion. The authors consider that neo-angiogenesis is an important factor that contributes to the regulation of invasion patterns. Recent studies have shown that glioma cell lines derived from transgenic mice that lack VEGF expression form tumors with irregular borders and are highly invasive along blood vessels, as compared with the sharp borders seen in tumors derived from the glioma cells that have been transfected to overexpress VEGF.³⁶ This observation is consistent with the result of the present study, suggesting that annexin A2 plays a key role in angiogenesis-dependent invasion by interacting with VEGF and other proangiogenic factors. With regard to the relationship between annexin A2 and VEGF, Zhao et al. demonstrated that the increased expression of annexin A2 in ischemic retinas is induced by increased levels of VEGF mRNA. Zhao et al. concluded that annexin A2 is a functional protein in VEGF expression in retinal neovascularization and is induced by hypoxia.³⁷

Additional findings

Other noticeable proteins expressed higher in J3T-2 are heterogeneous nuclear ribonucleoproteins (hnRNPs) and Septin 11.

hnRNPs comprise a large family of proteins, including approximately 30 members that share some structural domains. The results obtained in our study show HNRNPK and HNRNPA1 are up-regulated in J3T-2. hnRNPs play important roles in telomere biogenesis, DNA repair, cellular signaling, and the regulation of expression at both the transcriptional and translational levels.³⁸ In addition, the nuclear shift of HNRNPK in dividing cells suggests a role in proliferating cells.³⁹ So far, there have been no detailed reports about functions of hnRNPs in gliomas.

Septins are a highly conserved subfamily of GTPases that play an important role in maintaining cytoskeletal structures necessary for the control of cell division, and are required for cytokinesis.⁴⁰ Kim et al. showed that septins are variably expressed in human brain tumor cell lines and specimens using antibodies against septins 2, 3, 4, 5, 6, 7, 9, and 11 in immunofluorescence and Western blot analysis.⁴¹ In their study, septin 11 was immunolocalized along with the actin microfilamentous system, suggesting possible involvement in cell motility function.

These proteins are not fully investigated in terms of their function in glioma invasion and await further study.