Abstract
Purpose: There are a limited number of noninvasive methods available for the monitoring of neoplastic disease in the central nervous system. The goal of our study was to find reliable markers that could be used for disease monitoring as well as to identify new targets for the therapeutic intervention for malignant astrocytoma (WHO grades 3 and 4).
Experimental Design: We employed proteomic techniques to identify secreted proteins in the cerebrospinal fluid that were specific to patients with malignant astrocytoma.
Results: Among 60 cerebrospinal fluid samples of patients with various central nervous system diseases, attractin was consistently found to be elevated in the samples of patients with malignant astrocytoma. To independently validate these results, we examined attractin expression in a new set of 108 normal and tumoral brain tissue specimens and found elevated expression in 97% of malignant astrocytomas, with the highest levels in grade 4 tumors. Using immunohistochemistry, we further showed that attractin is produced and secreted by the tumor cells. Finally, we showed that cerebrospinal fluid from brain tumor patients induces glioma cell migration and that attractin is largely responsible for this promigratory activity.
Conclusions: Our results find attractin to be a reliable secreted marker for high-grade gliomas. Additionally, our migration studies suggest that it may be an important mediator of tumor invasiveness, and thus, a potential target in future therapies.
- Astrocytoma
- Proteomics
- Attractin
- migration
- Brain/central nervous system cancers
- CARCINOGENESIS
Diffusely infiltrating astrocytomas are among the most common forms of intracranial tumors, with malignant astrocytoma (WHO grades III and IV) being the most aggressive subtypes. Patients with high-grade (WHO grade IV) astrocytomas have a life expectancy of <1 year even after surgery, chemotherapy, and radiation therapy (1). Postoperative recurrence from residual tumor cells deeply infiltrated in the normal surrounding brain is one of the major reasons these tumors are incurable. Additionally, there are limited noninvasive methods available for the monitoring of neoplastic disease in the central nervous system and there is an urgent need to find reliable markers that can be used for disease monitoring as well as to identify new targets for therapeutic intervention (2).
The cerebrospinal fluid (CSF) is well-suited for sampling, and its accessibility through a lumbar puncture is less invasive than neurosurgical procedures (3). Because the CSF is continuous with the extracellular compartment of the central nervous system, it is a useful and currently underutilized source for searching proteins released during astrocytoma growth and progression. Moreover, the normal CSF protein concentration is 100- to 400-fold lower than in serum, allowing a much greater signal-to-noise ratio for detecting biomarker proteins (4–7).
Attractin is a member of the CUB family of cell adhesion and guidance proteins (8) existing in secreted (175 kDa) and transmembrane (∼200 kDa) forms due to alternative splicing of the ATRN gene (9). Secreted attractin is an abundant serum glycoprotein, but is normally absent in the central nervous system and is undetectable in the CSF (10, 11). Using proteomic analyses of the CSF of patients with brain tumors, we have demonstrated for the first time that attractin levels are elevated in the CSF of grades III to IV astrocytoma, with the highest levels in grade IV tumors. Furthermore, we show that attractin is a promotility factor for glioma cells, which is an important finding because invasiveness is a critical feature of glioma malignancy and resistance to therapy.
Materials and Methods
Source of CSF and sample preparation. Sixty CSF and cyst fluid (CF) samples (32 astrocytoma, 7 primary brain tumor, 8 metastatic brain tumor, 3 infectious, and 10 nontumoral controls) were individually centrifuged at 750 × g at 4°C to remove cells and debris, and stored in aliquots at −70°C until use (Supplemental Table S1). The samples (1 mL) were concentrated 5-fold using Centricon 3 kDa filtration (Millipore, Billerica, MA) and partially depleted of IgG and albumin using Proteoprep Blue albumin depletion kit (Sigma, St. Louis, MO). The proteins from each sample were precipitated using 15% trichloroacetic acid, resuspended in denaturing buffer (8 mol/L urea, 4% CHAPS, and 100 mmol/L protease inhibitor cocktail; Roche, Mannheim, Germany), and stored at −20°C.
Two-dimensional PAGE. Each CSF sample was individually analyzed in duplicate using two-dimensional PAGE as described (12). The protein content of each sample was measured and normalized so that 200 μg of protein was analyzed per sample. The first dimension was done on an IPGphor system using Immobiline Dry strips (pH 3-10 NL and 4-7; 13 cm; Amersham Biosciences, Piscataway, NJ) rehydrated overnight with 200 μg of protein (total run, 130,000 Vh). The second dimension was run on 12.5% polyacrylamide gels with 2% SDS using the Protean II xi system (Bio-Rad, Hercules, CA) and protein spots were visualized using Silver Stain Plus kit (Bio-Rad). The gels were analyzed using Melanie (EXPASY SwissPro 2-D database; http://au.expasy.org/melanie/) and ImageMaster (Amersham Biosciences) softwares. Attractin was identified following the in-gel digestion of proteins with trypsin followed by MALDI-TOF/TOF-MS analysis.
Cleavable isotope-coded affinity tag analysis. Cleavable isotope-coded affinity tag cICAT technology uses stable isotope tags (12C or 13C), in combination with two-dimensional chromatography, to compare the levels of complex peptide mixtures present in two samples (13). Equal volumes of pooled normal and pooled astrocytoma (AII or AIV) samples (three different samples/group) were analyzed. Two pools were made for each category (six total samples/category) and analyzed in two independent cleavable ICAT analyses (Applied Biosystems, Foster City, CA). The control and astrocytoma samples were labeled with light and heavy reagents, respectively, mixed, digested overnight in a 1:100 w/w trypsin/protein ratio, purified using a monomeric avidin column (Applied Biosystems), and analyzed using Ultimate nano-HPLC LC-MS/MS (Dionex/LC Packings, Sunnyvale, CA) with a Vydac C18 silica (5 μm, 300 Å, 5 μm × 150 mm) column interfaced to a QSTAR XL Qtof type mass spectrometer (Applied Biosystems). The MS/MS data was processed using ProICAT software for protein identification and quantification. Two ICAT comparisons using two separate sets of pooled samples (three/group) were done in triplicate to identify AII- and AIV-specific proteins. Attractin was found with a ProtScore >1.5 (>95% confidence) in all analyses.
Scratch-wound assay for glioma cell migration. A wound was generated in a monolayer of confluent glioma cells (14) by removing a row of cells using a 20-μL pipette tip and cells were switched to serum-free media. Purified recombinant secreted attractin (rAtrn; ref. 15) or bovine serum albumin (BSA) as a control was added in increasing amounts (50, 150, 300, and 500 ng/mL). The width of the scratch was measured at 0 and 24 hours posttreatment to measure the distance traveled by the cells. Experiments were repeated thrice in duplicate with comparable results.
Modified Boyden chamber migration assay for glioma cell migration. Cell culture inserts (8.0 μm pore) in a 24-well format (Becton Dickinson, Franklin Lakes, NJ) were plated with 2.5 × 104 cells/well. Cell migration was measured in response to increasing amounts (50, 150, 250, 500, and 1,000 ng/mL) of rAtrn or BSA as a control in a 12-hour assay. For CSF/CF, we used a 10% dilution in medium (v/v) of pooled samples (three samples/pool) as a chemoattractant in the bottom chamber. Pooled samples from nontumoral controls, AII, AIII, or AIV patients either crude or after preincubation for 2 hours at 4°C with 50 μg/mL antiattractin antibody or an irrelevant IgG as a control were used. Medium with 50 μg/mL of antiattractin antibody alone was also used as a separate control. After overnight migration, the migrated cells were stained with Diff-Quik (Dade Behring, Deerfield, IL), photographed, and counted at 40× from three separate fields. The entire experiment was repeated thrice.
3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide proliferation and cytotoxicity assay. Human glioblastoma cells (LN-Z308, LN229, and U87MGD; ref. 14) were plated in triplicate at 5 × 103 cells/well in 96-well plates and treated with serial dilutions of rAtrn from 4,800 to 0 ng/mL in media. The relative cell number was quantified at 48, 96, 120, and 144 hours post-treatment by using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide kit (Roche, Basel, Switzerland).
Attractin immunohistochemistry on paraffin-embedded tissue samples. Attractin expression in paraffin-embedded tumor sections was detected by immunohistochemical analysis using rabbit polyclonal antiattractin antibody (1:10,000) and DakoCytomation kit following the manufacturer's protocol (Dako Behring, Glostrup, Denmark). An irrelevant IgG was used as the primary antibody for controls.
Results
Attractin is secreted in the CSF of patients with malignant astrocytoma. To discover new secreted proteins involved in brain tumor development, we established protein profiles in the CSF or CF of 60 patients (47 with brain tumors and 13 nontumoral controls; Supplemental Table S1) using two independent proteomic techniques: two-dimensional PAGE and cleavable ICAT analysis. Both techniques revealed secreted attractin to be almost exclusively present in the samples of patients with grades III and IV astrocytoma (Fig. 1A and B ). High levels of attractin were confirmed by Western blot in 100% (19 of 19) of the AIV samples analyzed (Fig. 1C). Furthermore, lower levels of attractin were detected in the samples of 57% (4 of 7) of the AIII, and 33% (2 of 6) of the AII patients, whereas only 10% (1 of 10) of controls and 39% (7 of 18) of non–astrocytic tumors show low but detectable levels of attractin (Fig. 1C; Supplemental Table S1).
Attractin is secreted in the CSF of patients with high-grade astrocytoma. A, differential expression of attractin on enlarged sections of representative two-dimensional PAGE gels of control, AII, AIII, and AIV CSF. Encircled areas, proteins detected by silver staining. B, representative spectrum from one of the three peptides found from silver-stained two-dimensional PAGE gel spots for attractin identified by MS/MS. The spectrum identifies a tryptic peptide (LTGSSGFVTDGPGNYK) with a Mowse score (ms) of 98 that was also found in the initial characterization of attractin (15). The other two peptides were “LTLTPWVGLR” (ms = 37) and “SCALDQNCQWEPR” (ms = 98). m/z, mass to charge ratio; amu, atomic mass unit. C, representative Western blot verifying the secretion of attractin (175 kDa; polyclonal antiattractin; 1:1,000; ref. 11) in CSF and CF samples from patients with high-grade disease. Transthyretin (TTR) is included as a loading control.
Attractin in the CSF originates from tumors. To examine whether the elevated attractin found in the CSF of patients was derived directly from the tumor, we analyzed 46 astrocytoma (WHO grades II-IV) and 16 control brain tissue samples by Western blot analysis (Supplemental Table S2; Fig. 2A ). Attractin was absent in 14 of the 16 controls, whereas the remaining 2 samples had trace amounts. As additional controls, we also examined attractin levels in 26 nonglioma primary brain tumors and 12 metastatic brain tumors (Supplemental Table S2). Low to moderate levels of attractin were observed in only 23% (7 of 31) of primary brain tumors and 36% (5 of 14) of the metastatic brain tumor samples tested (Fig. 2A). In contrast, moderate levels of attractin were seen in 64% (7 of 11) of AII, and 100% of AIII (9 of 9) samples, whereas significantly higher levels of attractin were observed in 96% (25 of 26) of AIV tissue lysates (Fig. 2A). These results suggest that high levels of soluble attractin distinguish grade IV astrocytoma from lower grade astrocytoma and other brain tumors.
Attractin in CSF originates from tumor cell secretions. A, Western analysis of tissue lysates from normal brain (C), primary brain tumors (PBT), and metastatic brain tumors (MBT) compared with AII to AIV tumor samples shows strong up-regulation of attractin in patients with high-grade astrocytoma. Antiactin (sc-1615; 1:1,000; Santa Cruz Biotechnology) was used as a loading control. Protein lysates (25 μg) were loaded per sample. B, immunohistochemical analyses show positive staining for attractin exclusively in neurons (arrows) in normal brain (a and e). The majority of tumor cells stained moderately for attractin in AII (b and f); whereas strong positive staining in astrocytoma cell cytoplasm and extracellular compartments was seen in AIV (c and g). No staining of vascular structures (arrows) was evident (c and g). Staining was negative in schwannomas (d and h). a-d, H&E stain (H&E); e-h, antiattractin immunohistochemistry. An irrelevant IgG was used as the primary antibody for control and showed no staining (data not shown).
Attractin is secreted by astrocytoma cells. Next, we sought to identify the source of attractin production through the immunohistochemical analysis of sections from varying grades of astrocytoma and compared them to normal brain and/or other forms of brain tumor malignancies (Fig. 2B). The antibody used recognizes both the transmembrane as well as the secreted form of attractin (9). Therefore, the secreted form was differentiated by its presence in the extracellular spaces along with the cytoplasm. Attractin staining was seen only in the neurons but not in glial cells in normal brain samples, as expected (n = 3; ref. 10). In contrast, weak cytoplasmic staining for attractin was observed in 100% of AII (n = 4) and AIII (n = 4) sections, whereas much stronger cytoplasmic and extracellular staining was observed in all AIV (n = 7) tumors. Finally, consistent with two-dimensional PAGE and Western blotting results, the staining was either extremely weak or absent in 100% of schwannoma (n = 4) and meningioma (n = 3) brain tumor samples (Fig. 2B). The high levels of cytoplasmic staining found, specifically in grade IV astrocytoma samples, confirm that the glioma cells are the most likely source of secreted attractin in the CSF or CF of these patients (Fig. 2B). This data directly supports the proteomic analysis–based proposal that elevated levels of attractin in the CSF or CF represent a predominantly AIV-specific biological characteristic, and show that it is produced by astrocytoma cells within the tumor.
Secreted attractin enhances the migration potential of glioma cells in vitro. Subsequently, we examined the possible roles for secreted attractin in the process of gliomagenesis. Attractin is rapidly expressed and secreted by activated T cells and mediates the spreading of monocytes and other immune-related cells (11). Given attractin's ability to regulate cellular protrusions in neurons and monocytes, we hypothesized a role for attractin in increasing the migratory potential of glioblastoma cells. To examine this hypothesis, we stimulated three genetically different glioblastoma cell lines (14) with increasing amounts of purified recombinant attractin in two different cell migration assays: the scratch-wound and modified Boyden chamber assays. The cells showed a dose-dependent 2- to 4-fold increase in their migration in both assays with maximal effects at ∼500 ng/mL. BSA was used as a control protein and did not elicit any response (Fig. 3A-D ). The observed changes were not due to alterations in cell viability or cell proliferation (Fig. 3E).
rAtrn increases glioma cell migration in vitro. A and B, scratch-wound assay measuring increased cell migration of three human glioblastoma cell lines (LN-Z308, LN229, and U87MGD) in response to rAtrn in the medium. Width of scratch (A) was measured and photographed at 0 and 24 hours (results shown for U87MGD). Quantification of results (B) showed a dose-dependent increased cell migration in response to increasing concentrations (0, 50, 150, 300, and 500 ng/mL) of rAtrn in the medium, whereas BSA at 500 ng/mL had no effect (B). C and D, modified Boyden chamber migration assay measuring increased cell migration in response to increasing concentrations (0, 125, 250, 500, or 1,000 ng/mL) of rAtrn or BSA (as control) present in the lower well. Quantification of results (D) indicated a dose-dependent increase in cell migration in response to rAtrn, whereas BSA (1,000 ng/mL) had no effect. E, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide proliferation assay showed no change in relative cell number (#) in glioma cells with treatment up to 4,000 ng/mL of rAtrn in the medium for 48 hours.
Secreted attractin in the CSF elicits the migration of glioma cells in vitro. To determine whether attractin present in the complex protein mixture of the CSF could also elicit migration, we tested CSF diluted 1:10 in medium for its promotility properties. The CSF from all three astrocytoma grades augmented the baseline migration of three genetically diverse human glioblastoma cell lines in overnight assays with the highest levels obtained with AIV samples (Fig. 4A-C ). This effect on migration was completely inhibited after 2 hours of preincubation of CSF with antiattractin but not control antibodies. Taken together, these results suggest that the increase in the migratory ability of glioma cells, elicited in the presence of CSF from patients with astrocytoma, is largely mediated by attractin.
Secreted attractin in the CSF from patients with astrocytoma increases glioma cell migration. A, B, and C, modified Boyden chamber assay showing increased migration of three glioblastoma cell lines (LN-Z308, LN229, and U87MGD) in response to CSF from control (C) or glioma patients (AII-AIV). Quantification of results for LN-Z308 (A) and U87MGD (B) glioma cells lines and representative pictures of migrated LN229 cells across the membrane (C). Reduction in cell migration upon incubation with antiattractin antibody (A, B, and C). Irrelevant IgG used at similar concentrations had no effect (data not shown).
Discussion
In this report, we examine attractin as a biologically relevant protein marker secreted into the CSF and CF of patients with malignant astrocytoma (WHO grades III and IV). Using two independent proteomic screens (two-dimensional PAGE and cleavable ICAT analyses), we found significantly higher levels of secreted attractin in the majority of the CSF and CF from patients with malignant astrocytoma versus low-grade astrocytoma.
We further discovered that the CSF of glioma patients stimulated the migration of glioma cells, and that attractin was largely responsible for this effect. Attractin was able to elicit glioma cell migration in vitro and antiattractin antibodies neutralized the promotility activity of the CSF. The mechanism by which attractin may modulate glioma cell migration is unknown. Attractin derives its name from its ability to promote T cell migration towards adherent monocytes following immune activation (11). Additionally, it mediates the extension of dendrites from maturing neurons and may play a role in axonal regeneration following traumatic neural injury (10, 16). Attractin is a multidomain protein containing CUB and C-lectin domains along with several epidermal growth factor–like domains. The expression of both CUB and the epidermal growth factor–like domains is associated with proteins known to mediate cell adhesion, growth, migration, and differentiation (17). Additionally, attractin also has an associated dipeptidyl peptidase IV (DPPIV) activity (8). The DPPIV activity has the potential to regulate the chemotactic activity of chemokines through targeted proteolysis leading to the activation and inactivation of target substrates (19). Several members of the DPPIV-like activity and/or structural homologues family have known roles in cancer-related processes including the interaction and degradation of ECM, regulation of cell survival, intracellular signaling capacity, modification of immune response, and even drug resistance (19). Recently, DPPIV-like activity was found to be up-regulated in WHO grades III and IV–derived human glioma cell lines, and it was suggested that this activity might be due to attractin (19, 20).
In general, quiescent glial and microglial cells do not express attractin (21); thus, the regulated expression may prove to be a sensitive marker of glial cell transformation. Even more significantly, secreted attractin transcription is essentially shut down during neuron differentiation, with messages for the secreted transcript isoform almost undetectable in all regions of the normal adult human brain (10). As a consequence of this tight regulation of alternative splicing, the secreted attractin protein cannot be detected in non–pathologic CSF. This raises the possibility that following blood-brain barrier disruption in high-grade astrocytoma, secreted attractin present at high levels in the plasma could be leaking to the CSF from the periphery (10). Although we cannot formally exclude this possibility, we show that a major source of secreted attractin is the tumor itself, as the transformed glioma cells were found to produce it following immunohistochemistry and Western blot analyses.
Functionally, it has been suggested that attractin may bind positively charged growth factors and help maintain growth factor gradients, or modulate local concentrations (8, 11). Such growth factors include most of the neurotrophins and include glial cell–derived neurotrophic factor, which has a positive charge at physiologic pH. Future studies are needed to elucidate the mechanism by which secreted attractin results in the increased migratory ability of glioma cells, including the possibility that its DPPIV activity might be involved.
In summary, using proteomic analyses, we provide strong evidence for the abnormal presence of elevated levels of secreted attractin in CSF and CF from patients with grade IV astrocytoma. Our studies suggest that the monitoring of attractin levels in the CSF of patients with grades II to IV astrocytoma should be explored further as a potential biomarker to monitor disease progression or therapeutic response. Attractin's role in glioma migration suggests that it may also hold value as a novel therapeutic target.
Acknowledgments
We greatly appreciate the patients who donated their samples for our study. We thank Dr. Jan Pohl, Matthew Reed, and Pavel Svoboda (Microchemical Core Facility) for their continuous helpful advice and technical support for this project.
Footnotes
-
Grant support: CA 86335 from the NIH (E.G. Van Meir), the Genetics and Molecular Biology Program of Emory University's Graduate Division of Biological and Biomedical Sciences, and the National Science Foundation (PRISM, DGE0231900).
-
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.
-
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
-
Project concept, experimental designs, result interpretation, and writing of manuscripts were done by F.W. Khwaja and E.G. Van Meir. All experiments were done by F.W. Khwaja, D.J. Brat did the neuropathology, and J.S. Duke-Cohan provided attractin reagents and advice. All authors read the manuscript.
- Accepted August 10, 2006.
- Received May 30, 2006.
- Revision received July 17, 2006.