Human glioblastoma-initiating cells invade specifically the subventricular zones and olfactory bulbs of mice after striatal injection

Animals

We used adult P40 female immunodeficient athymic nude mice, Crl:NU-Foxn1^nu, obtained from Charles River® animal facilities (Charles River Laboratories®, Wilmington, UK). All animals were taken care of in accordance with the declaration of Helsinki and following the guidelines of the Belgium ministry of agriculture in agreement with EC laboratory animal care and use regulation (86/609/CEE, CE of J n°L358, 18 December 1986). The athymic nude mice were housed in sterilized filter-topped cages.

Cell cultures

One primary GIC primary culture (GB138) was established from acutely resected human GBM originated from a woman and validated as previously described.22, 23 Briefly, tumor tissue was washed twice in phosphate buffer saline (PBS, Gibco®), minced and placed in T75 flask (Corning®, Corning, NY, USA) in DMEM containing 10% of foetal bovine serum (FBS, Invitrogen®, Merelbeek, Belgium). After 2–3 weeks, cells were collected and resuspended at the density of 25,000 cells/mL in serum-free defined medium consisting of DMEM/F12 containing B27 (Invitrogen®) and supplemented daily with recombinant human epidermal growth factor and recombinant human fibroblast growth factor 2 (EGF, 20 ng/mL and FGF-2, 10 ng/mL, respectively, Preprotech®, Rocky Hill, NJ, USA). Under these culture conditions, cells were maintained as floating spheres. When spheres became visible under microscope, we tested their ability to self-renew using secondary spheres culture assays. Briefly, spheres were incubated 15 min at 37°C in papaïn (Wothington®, Lakewood, NJ, USA), the reaction was stopped with ovomucoïde (Wothington®) and then, cells were mechanically dissociated. Finally, the cell suspension was cultured as the primary culture sphere assays.

Glioma cell line U87MG was obtained from the American Type Culture Collection (ATCC) and cultivated in serum-containing medium. U87MG/DsRed cells expressing red fluorescent protein were also used. To stimulate the formation of floating spheres, cells were cultivated in B27-supplemented DMEM/F12 medium as described above. Cultures were maintained at 37°C under a humidified atmosphere containing 5% carbon dioxide.

Intracranial transplantation of tumor cells into nude mice

Mice were anaesthetized with an intraperitoneal injection of a Rompun® (Sedativum 2%, Bayer®, Bruxelles, Belgium) and Ketalar® (Ketamin 50 mg/mL, Pfizer®, Bruxelles, Belgium) solution (V/V) prepared just before injection. A suspension of 50,000 or 100 GBM cells in 2 μL of PBS solution was injected into the right striatum of the athymic nude mice. The mouse head was maintained within a stereotactic frame, allowing a precise and reproducible injection site, using the intersection of the coronal and sagittal sutures (bregma) as reference. The coordinates used were 0.5 mm anterior and 2.5 mm lateral from the bregma and at a depth of 3 mm. This injection site is 1.2 mm distant from the SVZ. Mice were sacrificed at 24 hr, 1, 2, 3 or 4 weeks after injection for U87MG cells and after 4, 6, 8, 10 weeks for GB138 GIC. For secondary injections, the OB, the SVZ and the tumor mass (TM) were isolated from 400 μm thickness sections obtained with a Leica vibratome (Leica VT1000S, Groot Bijgaarden, Belgium). Thereafter, tissues were incubated in papaïn (Wothington®, Lakewood, NJ, USA) for 30 min at 37°C, and ovomucoïde (Wothington®) was added to stop the dissociation. Then, U87MG cells were plated in six-well dishes in DMEM containing 10% of FBS and GB138 cells in serum-free defined medium at the density of 25,000 cells/mL. Human cells were selected with one passage allowing the selection of more than 99% of human cells. GBM cells were then passed and grafted as described above in a secondary mouse brain or cultured as floating spheres in specific medium.

Processing of tissue sections and cell cultures for immunolabelling

Mice were anaesthetised with a Nembutal® injection (Pentobarbital 60 mg/mL, Ceva Sante Animal®, Bruxelles, Belgium) before an intracardiac perfusion with a NaCl 0.9% solution (VWR International®, Prolabo, USA), followed by paraformaldehyde (P.F.A.) 4% at 4°C (4,3g/L NaOH, 40g/L paraformaldehyde, 18.8 g/L NaH2PO4). Brains were removed and postfixed in P.F.A. 4% for 2 hr, then cryoprotected for 48 hr in PBS containing 20% sucrose before being frozen at −20°C in a 2-methylbutane solution (Sigma®, Bornem, Belgium). Fourteen micrometer thick coronal and sagittal sections were cut on a cryostat and stored at −20°C.

For immunocytofluorescence, cells were plated on coverslips coated with polyornithine for 1 hr at room temperature (0.1 mg/mL, Sigma®). After a short washing in PBS, cells were fixed by incubation for 15 min at room temperature (RT) in P.F.A. 4% solution, then washed three times for 5 min in PBS.

Immunolabelling and histology

Permeabilization and blocking of unspecific binding sites were obtained by a 30 min incubation at RT in the blocking solution (5% donkey serum and 0.3% Triton X-100 in PBS). Primary antibodies were diluted in a carrier solution containing 0.2% donkey serum and 0.1% Triton X-100 in PBS. We used antibodies directed against Nestin (chicken polyclonal IgG, 1:250, Novus Biologicals®, Littleton, CO, USA), GFAP (mouse monoclonal IgG, 1:500, Sigma®, Bornem, Belgium), human mitochondria (mouse monoclonal IgG1, 1:250, Millipore®, Brussels, Belgium), human nuclei (mouse monoclonal IgG1, 1:250, Millipore®), Sox2 (rabbit polyclonal IgG, 1:250, Abcam®, Cambridge, UK and goat polyclonal IgG, 1:200, Santa-Cruz®), Ki67 (rabbit polyclonal IgG, 1:200, Abcam®), CD133 (rabbit polyclonal IgG, 1:250, Abcam®), CD31 (rabbit polyclonal IgG, 1:500, Abcam®), Doublecortin (goat polyclonal IgG, 1:250, Santa-Cruz®), β-III-tubulin (rabbit polyclonal IgG, 1:1000, Covance®, Belgium), NeuN (rabbit polyclonal IgG, 1:500, Chemicon®) and GABA (guinea pig polyclonal IgG, 1:750, Chemicon®). Brain sections or fixed cells were incubated with primary antibodies at RT for 2 hr or at 4°C for the night. Three 15 min washes were performed in PBS at RT. We used biotinylated (Vector®, Brussels, Belgium) or RRX-, FITC- and Cy5-conjugated (Jackson Immunoresearch Laboratories®, West Grove, PA, USA) secondary antibodies. All secondary antibodies were diluted at 1:500 in the carrier solution. Finally, tissue sections were washed three times with PBS and coverslipped in an assembly DAPI-containing Vectashield® solution (Hard Set Mounting Medium®, Vector laboratory, Burlingame, CA, USA) when appropriate. The slides and cultures were stored in the dark at 4°C. Omission of primary antibodies resulted in a complete loss of detectable signal. Moreover, some slides were stained with Hematoxylin/Eosin to observe the tumor mass.

Image acquisition and data analysis

Immunostained sections were imaged and examined using a laser-scanning confocal microscope equipped with a krypton/argon gas layer (Olympus® Fluoview 1000, Aartselaar, Belgium). We also used Zeiss Axiovert 10® microscope (Car Zeiss®, Zaventem, Belgium) coupled with Mercator® software (Explora Nova®, La Rochelle, France) for automatically cell counting within anatomical regions. Figures were composed using ImageJ® (public domain Java image processing program, author: Wayne Rasband, National Institute of Mental Health, Bethesda, Maryland, USA) and Gimp software® (GNOME foundation, Cambridge, MA, USA).

Statistical analysis

Brain xenografts and tumor sphere experiments were repeated at least three times and representative results are illustrated. Where applicable, chi-square test was used to compare tumorigenicity between groups. Quantitative data were presented as means ± SD and a p value of <0.05 was considered statistically significant.

Validation of the intrastriatal xenograft of GBM cells as a model of cancer cell migration

To test whether human glioblastoma (GBM) cells are attracted to neurogenic zones, human U87MG cells were injected into the right striatum of immunodeficient mice using a stereotactic frame. This site is 1.2 mm distant from the ipsilateral subventricular zone (SVZ) but also 1.1 mm from the dentate gyrus in the hippocampus (Fig. 1a). We first checked that the injection is not responsible of expelling cells into the SVZ by sacrificing animals 24 hr after the surgery. We located human cells in the adult mouse brain using antibodies specifically directed against human mitochondria (Fig. 1b) or nuclei. The species-specificity of these antibodies was confirmed by the selective labeling of human astrocytes but not mouse astrocytes (data not shown) and by the injection of cells of the U87MG cell line previously stably transfected with a plasmid expressing a red fluorescent protein (Fig. 1d, DsRed). The small red tumor formed after 2 weeks was specifically labeled with the antibodies directed against human mitochondria (Fig. 1d, green) or nuclei (data not shown). In contrast, the mouse brain was completely clear of any signal in these conditions.

GBM cells invade the adult mouse subventricular zones and migrate to the OB after their injection

Four weeks after the injection in the right striatum, as mice developed TM, we identified U87MG cells in the right SVZ and in the left (contralateral) SVZ (Fig. 1f). These cells appeared to have migrated and reached these regions during the 4th week postinjection, as they are not seen during the first 3 weeks postimplantation (Fig. 1e). Migrating human cells were observed in one the main fiber tract structure connecting the right and left SVZ (i.e., the corpus callosum, Fig. 1f′) but did not follow the blood vessels (Fig. 1f, CD31-staining of endothelial cells). No cells could be observed in the right or left dentate gyrus (data not shown). In the SVZ, U87MG cells were intermingled with normal NSC (Fig. 2a). As adult NSC in the SVZ can divide asymmetrically and give rise to neuroblasts that differentiate while migrating toward the OB through the rostral migratory stream (RMS),24 we looked for the presence of human cells both in the RMS and in the OB. Human U87MG cells were indeed found migrating to the granular layer of OB (Fig. 2b). To better analyze this migration, we performed sagittal sections of the left striatum, opposite to the tumor injection site, and demonstrated the presence of glioma cells in the more anterior part of the left SVZ, throughout the left RMS and into the glomerular layer of the left OB (Fig. 2c).

GB138 cells first established a GIC long term expanding culture (GB138 cells, see under “Cell Cultures”) and conclusively demonstrated its tumorigenic properties following serial orthotopic transplantations in immunodeficient mice. We next investigate if GB138 cells invade the SVZ niches in a similar fashion as their U87MG counterpart. Tumors formed by GB138 cells were infiltrative, do not form tumor mass and some of these cells were found in the ipsilateral SVZ 6 weeks after the injection (Supporting Information Table 1b). Two weeks later, GB138 cells migrate across the corpus callosum until the controlateral neurogenic niche (Supporting Information Table 1c). Moreover, some GB138 cells migrate through the RMS to the OB of both hemispheres within the next 2 weeks (Figs. 3a and 3b). Interestingly, we never observed GB138 cells in other anatomical structures such as the controlateral cortex or the dentate gyrus of the hippocampus, even 10 weeks after the implantation (Fig. 3a). These results suggest a tropism of GB138 invading cells for the SVZ and from there, for the OB.

U87MG and GB138 human GBM cells express stem cells markers in vivo

To better characterize the phenotype of tumor cells found in the neurogenic zones, we studied the expression of Sox2, Nestin and CD133. Sox2 is a highly conserved HMG box transcription factor that plays an important role in the maintenance of NSC and is expressed in SVZ in the adult brain.25 Moreover, the inhibition of Sox2 in GIC leads these particular cells to stop their proliferation and to lose their tumorigenicity in immunodeficient mice.26 In our experiments, 2 weeks after stereotactic right striatal injection, Sox2-positive U87MG cells were present heterogeneously in the tumor mass (Fig. 4a). This observation is not surprising as it was recently shown that Sox2 is widely expressed in GBM.27 However, 4 weeks after surgery, Sox2-expressing U87MG cells were mainly observed at the edge of the tumor mass facing the neurogenic niche, along and into the SVZ (Figs. 4b and 4c). Unlike U87MG cells, GB138 cells form a more invasive tumor and are Sox2-positive wherever they are located in the brain, in the neurogenic zone, in the OB or during their migration through the corpus callosum and the RMS (Fig. 5a). We next examined the expression of Nestin. This intermediate filament-associated protein is expressed mostly in dividing cells during the early stages of development and the adult neurogenesis and now generally regarded as a classical marker for brain NSC.28 We demonstrated that Nestin is expressed by Sox2-positive U87MG invading cells (Fig. 4c) and by most of the GB138 cells (Fig. 5b). Finally, we used an antibody directed against CD133, a cell surface glycoprotein expressed in GBM.29 We did not observe any staining of U87MG cells both in vivo (and in vitro, data not shown). In contrast, GB138 cells largely expressed CD133 (Fig. 5c) and are labeled with an antibody directed against GFAP (Fig. 5d).

At the same time, we studied the expression of neuronal markers (NeuN, GABA and B-III-tubulin) to highlight the differentiation of U87MG cells migrating to the OB but did not observe any signals (data not shown). Indeed, true neuroblasts generated from NSC in SVZ differentiate into neurons while integrating the OB.30 However, some U87MG cells located in SVZ expressed Doublecortin (Dcx), a microtubule associated-protein (Supporting Information Fig. 1). Dcx plays a role in the neuroblasts migration to the OB both during development and in adulthood and is considered as an early marker of neuronal specification and differentiation.31 Interestingly, U87MG cells expressing Dcx are nestin-negative as it is observed within host neuroblast.

To conclude, we were unable to distinguish a specific profile of stem cells markers for GB138 cells invading the SVZ from those located in the bulk of cells in the striatum as we did for U87MG cell line. These results correlate with recent studies proving that self-renewing capacities of GIC seem to be not limited to a uniform population of cancer stem cells but can be linked with cell lineages expressing a range of markers.32

U87MG cells isolated from the TM, SVZ and the OB formed self-renewable spheres under specific in vitro conditions

To further investigate the biological properties of GBM cells found in the neurogenic zones, we decided to carry out their ability to initiate self-renewable “spheres” when cultured in specific conditions. Notably, GIC has to meet this criterion. Because of the very slow growth rate of GB138 cells, we decided to focus on U87MG cells. Thus, we isolated U87MG cells specifically from the TM, the ipsi- and contro- lateral SVZ and the OB of immunodeficient mice 4 weeks after their striatal injection (Fig. 6a). Cells obtained from these mouse brain regions were cultivated in a specific cell culture medium used for NSC and GIC. In these conditions, we could easily select U87MG cells that were able to form Nestin-positive spheres (Fig. 6b). Moreover, single cells obtained from the dissociation of these primary spheres have the ability to form secondary spheres that are also Nestin-positive (Fig. 6b). These experiments suggest the self-renewability of the U87MG cell populations derived from the four anatomical regions (TM, ipsi- or controlateral SVZ and OB).

Injection of distant invaders to new immunodeficient mice rapidly induced the formation of a tumor

The operational definition of GIC includes the ability to initiate tumor in xenograft models. To evaluate the experimental tumorigenicity of the migrating GBM cells, we performed injection of 50,000 human U87MG cells harvested from the primary tumor mass, the ipsilateral SVZ, the controlateral SVZ and the OB into immunosuppressed mice. Accordingly with former studies, mice grafted with these previously injected and then repurified U87MG cells had to be sacrificed 2 weeks after the grafting procedure because they were dying. Hematoxylin/Eosin (H/E) coloration showed the formation of a very large tumor in the right hemisphere (Fig. 6c). The large size of these secondary tumors observed 2 weeks after injection were never observed in the case of primary tumors 4 weeks after injection and prevents us to observe a new specific migration of human U87MG cells into neurogenic zones. It is now known that GIC responsible to experimental tumor represent a minor subset of the total cells within GBM, like in U87MG cell line.33 It is possible to estimate the amount of GIC using dilution assays through xenograft in immunodeficient mice brain.5 To do so, we then injected only 100 U87MG cells harvested from the primary tumor mass, the ipsilateral SVZ, the controlateral SVZ and the OB into the striatum of adult mice. As injection of 100 native U87MG cells never induce tumors, we observed the formation of three tumors out of 12 mice injected with 100 cells from the primary tumor mass, five tumors out of 11 mice injected with 100 cells from the controlateral SVZ, one tumor out of six mice injected with 100 cells from the ipsilateral SVZ and finally two tumors out of nine mice injected with 100 cells from the OB (Fig. 6c). Those tumors harbor the same infiltrative behavior as the naïve U87MG cells. Chi-square test analysis between the four conditions was not significant showing the same enrichment in GIC in the different populations of U87MG cells. In conclusion, we have demonstrated that a few GBM cells refugee in the neurogenic niches could initiate tumor.