SEARCH

SEARCH BY CITATION

Keywords:

  • Cancer stem cells;
  • Adult glioma;
  • Two-dimensional differential gel electrophoresis;
  • Secreted factor;
  • Neoangiogenesis;
  • Tumor progression

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Presence in glioblastomas of cancer cells with normal neural stem cell (NSC) properties, tumor initiating capacity, and resistance to current therapies suggests that glioblastoma stem-like cells (GSCs) play central roles in glioblastoma development. We cultured human GSCs endowed with all features of tumor stem cells, including tumor initiation after xenograft and radio-chemoresistance. We established proteomes from four GSC cultures and their corresponding whole tumor tissues (TTs) and from human NSCs. Two-dimensional difference gel electrophoresis and tandem mass spectrometry revealed a twofold increase of hepatoma-derived growth factor (HDGF) in GSCs as compared to TTs and NSCs. Western blot analysis confirmed HDGF overexpression in GSCs as well as its presence in GSC-conditioned medium, while, in contrast, no HDGF was detected in NSC secretome. At the functional level, GSC-conditioned medium induced migration of human cerebral endothelial cells that can be blocked by anti-HDGF antibodies. In vivo, GSC-conditioned medium induced neoangiogenesis, whereas HDGF-targeting siRNAs abrogated this effect. Altogether, our results identify a novel candidate, by which GSCs can support neoangiogenesis, a high-grade glioma hallmark. Our strategy illustrates the usefulness of comparative proteomic analysis to decipher molecular pathways, which underlie GSC properties. STEM CELLS 2012;30:845–853


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Human glioblastomas are the most common and most malignant primary tumors of the human adult central nervous system. Despite aggressive association of treatments after surgical resection, these tumors remain of poor prognosis. The median survival time following standard therapeutic strategy combining temozolomide, chemotherapy, and radiation therapy after neurosurgery is of 14.6 months [1].

Initiation, progression, and recurrence of glioblastomas might be supported by cancer stem cells. Glioblastoma stem-like cells (GSCs) differ from the other cells within the tumor mass because of their ability to grow as neurospheres in defined medium, to maintain a cellular hierarchy, to self-renew, and to share common molecular markers with neural stem cells (NSCs) [2]. Similar to NSCs, maintenance of GSC properties is thought to depend upon their interactions with perivascular niches, the formation of which NSCs and GSCs can stimulate through production of the angiogenic factor, vascular endothelial growth factor (VEGF) [3–6]. Conversely, endothelial cells may stimulate GSC stemness in the vascular niche, particularly through a mTOR-dependent signaling [7]. GSC injection into the brain of immunocompromised mice results in the development of tumors, which phenocopy the original patient cancerous lesions. This unique preservation of both genotype and phenotype, unachieved with traditional glioma cell lines grown in serum, makes GSCs the best glioblastoma model to date [8]. The pertinence of GSCs for the study of glioblastomas is reinforced by their resistance to current chemo- and radio-therapies [6, 9–13]. Since the discovery of an adult neurogenesis arising from NSCs [14], and the recent establishment of the antitumorigenic role of NSCs against GSCs [15], the therapeutic challenge is to target GSCs without affecting NSCs. Despite the variety and extent of the genomic anomalies sustained by GSCs, they exhibit a transcriptomic pattern similar to NSCs [8], and both cell types have in common a number of molecular pathways necessary for the maintenance of their self-renewal properties [16–22]. In order to pinpoint molecular pathways specific to GSCs, we compared at the protein level four human primary cultures of GSCs, the four corresponding tumor tissues (TTs) from which they derived, and four primary cultures of human NSCs. Among the proteins over-represented in GSCs, we identified hepatoma-derived growth factor (HDGF), a multifunctional growth factor that has been related to tumor development and angiogenesis [23–30]. Interestingly, although both GSCs and NSCs express HDGF, only GSCs were able to secrete HDGF. Moreover, GSC-produced HDGF could mediate human endothelial cell migration in vitro and neoangiogenesis in vivo.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Human Sample Collection

The tumors from four adult patients (Tumor 1–4) were surgically removed at our institution. In accordance with French law, all patients signed an informed consent form that allowed their tissues to be used in this research project. The institutional review board of Sainte Anne hospital, Paris, France approved this study. Tumor diagnosis was performed along the Sainte Anne classification and grading system, and the World Health Organization (WHO) guidelines of brain tumors. Tumors were diagnosed as glioblastomas according to the WHO classification system [31] and as malignant glio-neuronal tumors according to the Sainte Anne classification and grading system [32]. After surgical removal of tumor samples, fragments free of gross hemorrhage or coagulation artifacts were divided into fragments: one was immediately frozen in liquid nitrogen and stored at −80°C until protein extraction; the second was fixed in formaldehyde 5% (vol/vol), NaCl 8 g/l, and ZnSO4 3 g/l and paraffin embedded for histological diagnosis; the third was immediately processed for cell culture studies. The four patients had not received radiotherapy or chemotherapy prior neurosurgery. All studies with human fetal normal tissue were performed under the Ethical Approval from the University Paris-Descartes Internal Review Board using tissue donated with informed consent after elective termination of pregnancy.

Cell Cultures

Isolation and characterization of four fetal NSCs primary cultures (NSC A to D) were performed as previously described [33]. Isolation of GSCs from TTs (GSC 1–4) was made as previously described [33] and are named according to the number attributed to TTs from which they derived. Their characterization was previously reported [7, 34, 35]. Of note, two of them bear a TP53 mutation preventing DNA binding of p53 (R175Q or R175H), whereas mutations in PTEN or IDH1 coding regions were not found. GSCs were cultured in defined medium containing basic fibroblast growth factor (bFGF) and epidermal growth factor (EGF) [34]. Immortalized human cerebral microvascular endothelial cells (hCMEC/D3) were cultured as previously described [36].

Protein Extraction for Two-Dimensional Differential Gel Electrophoresis Analyses

Frozen fragments of tumors 1–4 were cut with a cryostat at −20°C (10 μm-thick sections). TT was microdissected after routine hemalun phloxin coloration and analysis by the pathologist. Proteins were extracted from 10 to 15 pooled microdissected sections. A minimum of five 106 cells of each GSC and NSC culture were collected by centrifugation and rinsed twice with phosphate buffer saline (PBS) (pH 7.4). Lysis was performed with 8 M urea, 2 M thiourea, 4% (vol/vol) CHAPS, 60 mM Dithiothréitol (DTT) (GE Healthcare, Paris, France) for 30 minutes in ice. The protein extracts were clarified by ultracentrifugation at 100,000g for 1 hour at 12°C. The supernatants were treated with the reagents of a two-dimensional (2D) Clean-Up kit (GE Healthcare) according to the manufacturer's instructions. The dried pellets were suspended in lysis buffer devoid of DTT at a final concentration of 7 mg/ml (pH 8.5).

2D Differential Gel Electrophoresis

Each protein sample (50 μg) was labeled with CyDyes Fluor minimal dyes (GE Healthcare) according to the manufacturer's instructions. Two samples of each experimental group (namely TT 1-4, GSC 1-4, and NSC A-D) were labeled with Cy3 and two with Cy5 (CyDyes Differential In-Gel Electrophoresis (DIGE) Fluor minimal dyes, GE Healthcare). Equal amounts of all samples were pooled, Cy2-labeled, and used as the internal standard.

Six analytical gels were run, each containing two samples from different groups and the internal standard. Isoelectric focusing, gel scans, and analysis were performed as previously described [37]. Briefly, isoelectric focusing was performed with 150 μg of labeled proteins (50 μg each of the Cy3-, Cy5-, and Cy2-labeled proteins) and Immobiline Drystrips pH 4–7 (GE Healthcare) with an IPGphor system (GE Healthcare). The second dimension was performed using in-house made 8%–18% gradient polyacrylamide gels. The 2D differential gel electrophoresis (2D-DIGE) gels were scanned using a Typhoon 9400 scanner (GE Healthcare). Differentially expressed proteins were analyzed using DeCyder 2D 7.0 Software (GE Healthcare). The DeCyder Extended Data Analysis module was used to perform principal component analysis and hierarchical clustering.

Three preparative gels (one for each set, TT, GSC, and NSC) were prepared as described above using immobilized pH gradient (IPG) strips rehydrated with 350 μg of proteins. After electrophoresis, gels were fixed overnight in 30% (vol/vol) ethanol, 2% (vol/vol) phosphoric acid (two changes, 30 minutes each), washed three times for 10 minutes each with 2% (vol/vol) phosphoric acid, and then stained for 72 hours in 0.01% (wt/vol) Coomassie Brilliant Blue G-250, 12% (wt/vol) ammonium sulfate, 18% (vol/vol) ethanol, and 2% (vol/vol) phosphoric acid.

Mass Spectrometry

Tryptic in-gel digestion was performed as described [37] with minor modifications and automated with a Freedom EVO 100 digester/spotter robot (Tecan, Männedorf, Switzerland). After digestion and desalting, pooled eluates were dried at ambient temperature. The samples were analyzed by matrix-assisted laser desorption/ionization-time of flight-time of flight (MALDI-TOF-TOF) 4800 mass spectrometer (ABSciex, Foster City, CA). Mascot version 2.2 (MatrixScience, London, U.K.) in the GPS Explorer Software package version 3.6 (ABSCiex) was used to search the SwissProt databank (www.expasy.org) for human protein sequences that matched those of tandem mass spectrometry (MS)/MS data. The search parameters allowed variable modifications for carbamidomethylated cysteines and oxidized methionine residues. One uncleaved tryptic site was permitted and the mass accuracy tolerances were 30 ppm for the parent peptides and 0.3 Da for the MS/MS fragments. Positive identification was based on a Mascot score above the significance level (i.e., p < .05). The identified proteins are those that had the greatest number of peptide matches. Given the stringency of our identification criteria, no result matched multiple members of a protein family.

Immunoblotting and Immunocytochemistry

Lysis was performed in the following lysis buffer: 50 mM Tris-HCl pH 6.8 buffer containing 1% Triton X-100, 150 mM NaCl, 0.5 mM EGTA, 0.5 mM EDTA, 1 mM orthovanadate, 50 mM β-glycerophosphate, 5 mg/ml leupeptine, 5 mg/ml pepstatine, 5 mg/ml aprotinin, and 1 mM phenylmethanesulfonylfluoride (PMSF). Culture media were precipitated with trichloroacetic acid (1:100, vol/vol, overnight incubation at −20°C). The precipitates were rinsed with acetone, prior to be resuspended into lysis buffer. Protein extracts (40 μg) were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to Hybond-C Extra nitrocellulose membranes (Amersham Biosciences, London, U.K.) as described [37]. Immunoblotting was achieved with anti-HDGF (C14)-R (Santa Cruz Biotechnology, Paris, France, 1/800) and peroxidase-conjugated secondary anti-rabbit IgG (Amersham Biosciences, Paris, 1/10,000). Signal detection was performed with the enhanced chemoluminescence (ECL) + chemiluminescence detection system (PerkinElmer, Paris, France). Immunocytochemistry and image acquisition were performed as previously described [33, 38], using the anti-HDGF antibody at 1/500 or anti-von Willebrand Factor (Abcam, Paris, France, 1/300) following proteinase K (10 μg/ml for 10 minutes, 25°C).

Migration Assays of Human Cerebral Endothelial Cells

Chemotactic migration assays were performed using a polyvinyl pyrrolidone-free polycarbonate filter with 12 μm pores (Corning Costar, Dutcher, Paris, France) precoated with collagen (10 μg/ml, BD, Paris, France), as previously described [39]. Fifty microliters of serum-starved hCMEC/D3 (106 cells per microliter) were added to the upper chamber, whereas the conditioned media (CM) or corresponding fresh media were added to the lower chamber. Alternatively, CM were preincubated 1 hour at 37°C with 10 μg/μl anti-ERK1 or anti-HDGF antibodies (Santa Cruz). After 16 hours of incubation, filters were scrubbed to remove the unmigrated cells, fixed with methanol, stained with 4′,6′-diamidino-2-phenylindole (DAPI) (Vectashield), and analyzed by confocal microscopy. The number of cells per field of view was manually counted. Results are presented as mean ± SEM of three independent experiments.

siRNA Transfection

GSCs were transfected by electroporation with 50 μM of control (Ambion AM4611), A (Ambion s229694) or B (Ambion s6500) anti-HDGF siRNAs. The transfection was performed using the L transfection solution (AMAXA) with the X005 program. The CM of GSCs were collected 1 week post-transfection for use in in vivo assays. The corresponding cells were lysed and HDGF protein levels were measured by immunoblotting.

In Vivo Angiogenesis Assays

GSC-CM were mixed with Matrigel (BD Biosciences, Paris, France) and injected into the flanks of 7-week-old female C57BL/6 mice (Charles River, Paris, France) (450 microliters per graft; 24 volume Matrigel:1 volume CM). The plugs were retrieved 1 week later, fixed in 4% paraformaldehyde, and paraffin embedded. Each plug was cut into serial 8 μm-thick sections, which were colored with hemalun and phloxin. Images were acquired on a digital still camera (DXM 1200, Nikon Instruments, Melville, NY, http://www.nikoninstruments.com) using the Lucia software (Laboratory Imaging, Ltd, http://www.laboratory-imaging.com). Determination of endothelial cell and neovessel numbers was performed on one every other section.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Glioblastoma TTs, GSCs, and NSCs Exhibit Distinct Proteomic Profiles

2D-DIGE was used to obtain a quantitative global overview of the proteins differentially expressed between the three conditions, namely GSC, NSC, and TT. Principal component analysis revealed patterns that segregated into three distinct groups. This segregation separation persisted when each detected spot was aligned (2,083) as well as after a quantitative selection refinement. This latter selection, based on a significant variation of intensity in one group as compared to each of the two others (>1.4-fold; Student's t test, p < .04), unmasked 353 spots of interest (Fig. 1). Unsupervised clustering also resulted into three distinct patterns, one for each sample group (i.e., NSC, GSC, and TT; Fig. 1). Of note, GSCs positioned between NSCs, with whom they share stem-like properties, and TTs with which they share the tumor features.

thumbnail image

Figure 1. Specific and distinct proteome patterns for GSCs, glioblastoma TTs, and NSCs, as shown by unsupervised principal component and hierarchical clustering analysis of 353 protein spots. The proteins selected for analysis displayed statistical significant differential expressions between GSCs, NSCs, and TTs (at least 1.4-fold increase or decrease in spot intensities and Student's t test p value ≤.04). (Left panel) Principal component analysis. Each dot represents the proteome map of one sample. (Right panel) Hierarchical clustering displayed as a heat map. Spots of interest with similar overall variation of intensities are grouped together (lines), and proteomes with similar variation of expression profiles are grouped together (columns (A–D) and 1–4). The color scale changes from green (decreased protein expression) to black (no change in expression) to red (increased protein expression). A spot colored white was not detected or not matched. Note in both representations the segregation of each sample proteome into its respective groups of origin (GSC, TT, or NSC). Abbreviations: GSCs, glioblastoma stem-like cells; NSCs, neural stem cells; TTs, tumor tissues.

Download figure to PowerPoint

Following this, 220 spots out of the 353 of interest were isolated, trypsin-digested, and analyzed using tandem mass spectrometry in conjunction with a search of the SwissProt databank by Mascot. One hundred and eight proteins were then identified (Table 1 and Supporting Information Tables S1–S5). Eighteen proteins were overexpressed selectively in the GSC group (Supporting Information Table S1), 23 in the NSC group (Supporting Information Table S2), and 26 in the TT group (Supporting Information Table S3). In addition, 22 proteins were common to the NSC and GSC groups (Supporting Information Table S4), and 19 were common to the GSC and TT groups (Supporting Information Table S5).

Table 1. Functional classification of the proteins identified
inline image

Most proteins commonly overexpressed by NSCs and GSCs, as compared to TTs, were related to chromatin, mRNA, and DNA processing (Supporting Information Table S4). This list included also vimentin, an intermediary filament protein, routinely used as a marker of NSCs and GSCs [40]. In contrast, two other additional NSC/progenitor markers, the microtubule regulating phosphoprotein Stathmin [41] and FABP7 involved in lipid metabolism [42], were over-represented in NSCs, as compared to GSCs (Supporting Information Table S2). A vascularization-related protein group was found in TTs as compared to both NSCs and GSCs that might account for tumor-induced angiogenesis associated with glioblastoma development (Supporting Information Table S3). Metabolic enzymes dominated among the proteins commonly overexpressed in both TTs and GSCs, as compared to NSCs (Supporting Information Table S5). Among the enzymes involved in energy metabolism, nicotinamide-N-methyltransferase (NNMT) signed for GSCs (fivefold and 3.9-fold vs. NSC and TT, respectively). Of note, 15% of the proteins identified corresponded to activation targets of the transcription factors Nanog, Oct4, or Sox2 (Supporting Information Tables S1–S5).

Among the list of 18 proteins specifically overexpressed in GSCs (Supporting Information Table S1), we selected for further study the HDGF, because this candidate could directly relay paracrine communication between GSCs and their vascular environment. Thus, our nonbiased proteomic approach allows the identification of a specific hallmark for GSCs, where only HDGF could be viewed as a signaling factor operating in niches.

HDGF Is Selectively Secreted by GSCs

GSCs and NSCs expressed the HDGF isoform a (Swiss-Prot accession number: P51858), as evidenced by the five peptides identified through MS/MS (Fig. 2A). The twofold overexpression of HDGF in GSCs as determined by 2D-DIGE was confirmed using immunoblotting assays (Fig. 2B).

thumbnail image

Figure 2. HDGF is overexpressed and secreted specifically by GSCs. (A): Sequence of human HDGF protein. Peptides identified during the tandem mass spectrometry (MS)/MS analysis are colored in red. (B): Left panel. Western blot assays on total protein extracts from GSC and NSCs. Actin was used as a loading control. Similar results were obtained with three others cultures (two experiments each). Right panel. Densitometry analysis of HDGF immunoreactive signals assessed by Western blot. Graphs present the mean ± SEM of four different cultures. (C): Western blot assays on proteins precipitated from the conditioned media derived from GSC primary cultures (GSC) and from NSC primary cultures (NSC). Proteins precipitated from the naive culture media were loaded as negative controls. Representative of three independent experiments. Abbreviations: CM, conditioned media; GSCs, glioblastoma stem-like cells; HDGF, hepatoma-derived growth factor; NM, naïve medium; NSCs, neural stem cells. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Although HDGF lacks the secretion signal sequence found in most secreted proteins [43], its initial description as a secreted factor by the hepatoma-derived cell line HuH7 [44] prompted us to seek for its presence in GSC-CM. HDGF was indeed detected by immunoblotting in GSC-CM, but not in NSC-CM, in 1-week-old culture (Fig. 2C).

GSC-Secreted HDGF Induces Cerebral Endothelial Cell Migration

As GSCs and endothelial cells are found in close interaction in human glioblastomas, we then aimed at determining the role of HDGF on cultured brain endothelial cells, using a well-characterized human blood-brain barrier model [36]. CM prepared from GSCs enhanced migration of hCMEC/D3 by a twofold factor, as shown using Boyden chambers (Fig. 3). Anti-HDGF antibodies severely impaired GSC-CM-induced hCMEC/D3 migration, suggesting a key role for HDGF in endothelial migration (Fig. 3). These results demonstrate that GSC-produced HDGF might trigger in vitro migration of human cerebral endothelial cells.

thumbnail image

Figure 3. HDGF promotes human cerebral endothelial cell migration in vitro. C-: endothelial brain medium (EBM)-serum-free medium (negative control). bFGF: EBM-serum-free medium containing 50 ng/ml bFGF (positive control). NM: EBM-serum-free medium with NM used to culture GSC and diluted at 1/20. CM: EBM-serum-free medium with GSC-conditioned medium diluted at 1/20. NM and CM were preincubated with either a rabbit irrelevant antibody (−) or a rabbit anti-HDGF antibody (+) prior to be added to the cerebral endothelial cells. Graphs present the mean ± SEM of three independent experiments. Abbreviations: bFGF, basic fibroblast growth factor; CM, conditioned media; HDGF, hepatoma-derived growth factor; NM, naïve medium.

Download figure to PowerPoint

GSC-Secreted HDGF Promotes Angiogenesis In Vivo

The ability of GSC-secreted HDGF to stimulate neoangiogenesis was further evaluated using an in vivo mouse model. We used subcutaneous grafts since cerebral and extracerebral endothelial cells respond in a similar manner to angiogenic factors. This experimental setting allows testing neoangiogenesis in a well-circumscribed volume while avoiding lesion-induced inflammation, which takes place following any nervous tissue lesion [45]. CM were collected from GSCs transfected with either anti-HDGF siRNAs or scrambled siRNAs. First, two independent siRNA sequences targeting HDGF significantly reduced HDGF expression in GSCs, as compared to scrambled siRNA and mock-transfected cells (Fig. 4A). Second, such GSC-CM were mixed with Matrigel and injected subcutaneously in C57BL/6 mice. One week later, scrambled siRNA-transfected GSC-CM plugs exhibited a reddish color indicative of blood perfusion resulting from neoangiogenesis (Fig. 4B). In contrast, plugs containing CM derived from anti-HDGF siRNA-transfected GSCs or naive control exhibited little coloration, indicating poor perfusion (Fig. 4C). Density of host endothelial cells within the plugs was decreased to 40% in plugs containing HDGF siRNA-CM as compared to scrambled siRNA-CM (Fig. 4D–G). This reduced migration of endothelial cells was accompanied with a dramatic decrease in amount of functional blood vessels identified by their content of red blood cells (Fig. 4H). Altogether, our results suggest that GSC-derived HDGF acts as an angiogenic factor in vivo.

thumbnail image

Figure 4. GSC-secreted HDGF promotes in vivo angiogenesis. (A): Decreased expression of HDGF in GSCs transfected with anti-HDGF siRNAs, as compared to mock-transfected GSCs or to GSCs transfected with scrambled siRNAs. (B, C): Representative Matrigel plugs containing conditioned medium derived either from scrambled siRNA-transfected GSCs (B) or from HDGF siRNA-transfected GSCs (C). Microphotographs taken at 1 week after subcutaneous injection. Note the reddish color indicating the presence of blood vessels in (B). Bar = 0.5 cm. (D): Numerous mature blood vessel structures (black arrow) identified by the presence of red blood cells in their lumen (arrowhead) were observed in plugs containing conditioned media of scrambled transfected GSCs. Bar = 25 μm. (E): In contrast, plugs containing medium derived from HDGF siRNA-transfected GSCs showed poorly perfused vessel-like structures without detectable red blood cells within their lumen (white arrow). (F): Example of blood vessel structure visualized with either hemalun phloxin coloration (top panel) or anti-Von Willebrand Factor immunocytochemistry (bottom panel). (G, H): Plugs containing medium derived from HDGF siRNA-transfected GSCs exhibited decreased number of endothelial cells (*p = .003, t test) (G) and a strong reduction in the number of neovessels containing red blood cells (H) (**p value = .0002). Abbreviation: HDGF, hepatoma-derived growth factor.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

This study was designed to unravel proteins differentially expressed by GSCs, which might rationalize GSC properties. This comparative approach allowed the identification of distinct sets of proteins specifically over-represented in each group. Complementary in vitro and in vivo assays unveil a novel molecular component for glioma perivascular niches. Previous proteomic studies devoted to NSCs explored the role of proteins involved in their aggregation as neurospheres [46] or in their neuronal differentiation [47]. Other studies focused on proteins accompanying the differentiation of embryonic stem cells in NSCs [48–50]. With respect to GSCs, a single proteome analysis has been reported by the group of C.A. Conrad, who highlighted the IL-6/STAT3/HIF1α pathway in one single GSC line [51, 52]. This study provides a direct comparison of proteomes from four GSCs, their TT of origin (TT), and NSCs. A number of proteins identified here are associated with tumor behavior in gene ontology (GO) annotation (CSDE1, ERP29, HDGF, LMNA, PNP, RPSA, prostaglandin E synthase [PTGES3], CCT3, ENO1, SERPINA3, annexin 2 [ANXA2], YWHAG, and NNMT). Among them, PTGES3/P23 and ANXA2 are overexpressed in gliomas [53, 54]. Of note, altered interaction of the PTGES3 with HSP90 chaperone protein could limit the nuclear localization, and hence the activity of human telomerase reverse transcriptase (hTERT) [55, 56], a specialized reverse transcriptase preserving telomere ends in cancer cells [55]. Its enrichment in GSCs as compared to TTs and NSCs is thus consistent with the expected integrity of telomeres in these cells. Increased expression of HSP90B1 in GSCs raises the possibility that it might be the HSP90 family member interacting with PTGES3. Interestingly, this secreted chaperone, also named gp96 or endoplasmin, has been associated with aggressive esophageal squamous cell carcinoma and chemoresistance to paclitaxel [57]. ANXA2 expression has been correlated with increased glioma malignancy and shown to promote glioma cell motility [54, 58, 59]. NNMT, responsible for the methylation of nicotinamide, is a key enzyme of NAD+ recycling. It plays an essential role in energy metabolism and DNA repair [60]. A role for NNMT in cancer cell growth, survival, radiation resistance, and invasiveness has been demonstrated in various cancers [61–64]. NNMT overexpression correlated with radioresistance of mesenchymal cancer stem cells [65]. Most interestingly, our unpublished results indicate that downregulation of its activity accompanies induced loss of stem properties by GSCs. Several proteins overexpressed in GSCs and TTs are encoded by target genes of Nanog, Oct4, and Sox2 in embryonic stem cells [66]. These transcription factors are core components of the molecular networks that control embryonic stem and/or somatic stem cell behaviors. Our observation is in accordance with the known enrichment of their activation targets in glioblastomas, their participation in an embryonic stem-like signature [66–68], and their reported involvement—at least for Nanog and Sox2—in the governance of the stem properties of GSCs [16–19].

Our results show that only GSCs secrete HDGF in a significant manner and act via this growth factor on endothelial cell migration. In addition to HDGF described here, GSCs are known to produce several growth factors with angiogenic properties, such as VEGF, bFGF, transforming growth factor alpha (TGFα), or stromal derived factor 1 (SDF1) [4, 69–71]. Whether one of them behaves as a dominant driver or they cooperate remains unknown. Either anti-HDGF or anti-VEGF antibodies have been shown to inhibit to the same extent neoangiogenesis, without additive effects in a lung cancer model [72]. In addition, the fact that HDGF and VEGF may mobilize independent transduction pathways in endothelial cells [24] favors the possibility of concerted effects. Cooperation may also take place through cell-cell interactions and cancer and endothelial cells sharing the same array of angiogenic factors, which may act in both a paracrine [73] and autocrine manner [24]. An intracrine effect of HDGF on GSCs as well as on NSCs cannot however be excluded. Although initially identified as a secreted factor, HDGF has subsequently been localized in the nucleus as well as in the cytoplasm [74] and found endowed with DNA-binding activity through its PWWP domain [75]. In addition, siRNA viral-mediated HDGF downregulation in a glioblastoma cell line has been shown to decrease cell proliferation [76]. The recent establishment of the HDGF interactome supports its involvement in RNA processing, DNA damage repair, and transcriptional regulation [77]. In this context, it is interesting to note that heterogeneous nuclear ribonucleoprotein K (HNRNPK), a ribonucleoprotein reported to interact with HDGF [77], was found overexpressed in GSCs. Interestingly, increased nuclear expression of HDGF in glioblastoma has been linked with poorer postsurgery survival of the patients [76].

Glioblastomas are among the most vascularized human tumors. Their persistence and growth depend on the pathological formation of new capillary blood vessels [78]. Development of neovessels in glioblastomas is characterized by several mechanisms including GSCs production of VEGF and genesis of tumor endothelial cells from these stem-like cells [79–81]. Bevacizumab, a humanized monoclonal antibody directed against VEGF, improves survival and quality of life. Its effect is however transient, tumors eventually resuming their progression, and half of the patients failing to respond to the treatment [82, 83]. Moreover, interfering with VEGF implies interfering with normal neurogenesis, as this cytokine also participates to adult neurogenesis and mediates crosstalk between NSCs and endothelial cells [84, 85]. The discovery of HDGF, as a proangiogenic molecule specific to GSCs opens the possibility to target GSCs while sparing NSCs.

SUMMARY

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

In summary, our results founded on the comparison of proteomes from human GSCs, their TT of origin and normal NSCs, identify HDGF secretion as a specific candidate underlying GSC-promoting effects on angiogenesis, an event characteristic of high-grade gliomas, and essential to support tumor growth. They illustrate the functional interest of comparative proteomic analysis to decipher GSC molecular pathways.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

We thank Nicolas Cagnard (Plateforme Bioinformatique Necker, Paris, France) for bioinformatic assistance and Amélia Morais-Dias and Silvina Dos Reis Tavares for excellent technical assistance in cell culture and immunocytochemistry. We are highly grateful to Dr. René Frydman (Hôpital Antoine Béclère, Clamart, France) and Dr. Hervé Coffigny for providing fetal brain tissues and the Sainte-Anne neurosurgery team for providing glioblastoma tissues. We thank Guilhem Clary, Morgane Le Gall, and Joanna Lipecka for support in proteomic analysis. This work was supported by ARC and by Région Ile de France-Canceropôle (Ph.D. fellowship).

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
STEM_1062_sm_SuppTab1.tif544KSuppl. Table 1. Proteins identified as overexpressed in GSCs as compared to the corresponding tumor tissues (TTs) and NSCs. GSC: glioblastoma stem-like cell group; TT: glioblastoma tumor tissue group; NSC: neural stem cell group. (a): Involved in cancer; (b): Sox2 activation target in embryonic stem cells; (c): Oct4 activation target in embryonic stem cells 66.
STEM_1062_sm_SuppTab2.tif584KSuppl. Table 2. Proteins overexpressed by NSCs. (b): Sox2 activation target in embryonic stem cells; (c): Oct4 activation target in embryonic stem cells; (d): Nanog activation target in embryonic stem cells 66.
STEM_1062_sm_SuppTab3.tif620KSuppl. Table 3. Proteins overexpressed by TTs. (a): involved in cancer; (b): Sox2 activation target in embryonic stem cells; (c): Oct4 activation target in embryonic stem cells; (d): Nanog activation target in embryonic stem cells 66.
STEM_1062_sm_SuppTab4.tif626KSuppl. Table 4. “Stem-like”-proteins commonly overexpressed by GSCs and NSCs as compared to TTs. (a): involved in cancer; (b): Sox2 activation target in embryonic stem cells; (d): Nanog activation target in embryonic stem cells 66.
STEM_1062_sm_SuppTab5.tif517KSuppl. Table 5. “Tumor-related” proteins commonly overexpressed by GSCs and TTs, as compared to NSCs. (a): involved in cancer; (b): Sox2 activation target in embryonic stem cells; (c): Oct4 activation target in embryonic stem cells; (d): Nanog activation target in embryonic stem cells 66.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.