Central neurocytoma (CN) is a peculiar type of brain tumor that occurs most frequently in young adults, developing in the subventricular zone (SVZ) area of the lateral ventricle. They were initially classified as Organization (WHO) grade I lesions, but in 1993 CNs were upgraded to WHO grade II lesions when it was recognized they could exhibit more aggressive behavior [1] and recurrences [2]. Malignant neoplasms such as glioblastoma (GBM), in rare case, have been reported to arise from dedifferentiation of CN in an area distal to the SVZ [3] in the previous study, in which histological diagnosis of the tumor located in the right lateral ventricle was CN whereas two tumors of the left temporal lobe were GBM, and the specimen of GBM showed CN histology. This clinical observation indicates that CN itself has inter-tumor heterogeneity observed between patients with the same tumor type due to genetic variability, heterogeneity in microenvironment [4]. Similarly, previous findings that despite being a neuronal tumor, CN cells have the capacity to differentiate into both neurons and glia in an in vitro culture environment [5, 6] corresponds well with the some case of relatively aggressive form of CN.

However, it has been unclear which cells within the CN tumor mass are responsible for tumor initiation and maintenance. Hassoun et al. [7] suggested that CN may originate from small gray nuclei of the septum pellucidum, which appear to be the site of tumor attachment, whereas von Deimling et al. [8] hypothesized that CN originates postnatally from the remnants of the subependymal germinal plate of the lateral ventricles. Similarly, Sim et al. hypothesized that CN derives from a neoplastic GFAP+ cell that undergoes selective expansion when removed to culture, and which is able to regenerate itself and form new neuroblasts [5]. However, the latter study was limited by the use of serum in the culture conditions: to maintain neural stem cells (NSCs) in an undifferentiated state, they must be grown in serum-free media [9]. The persistence of germinal regions and the presence of NSCs and transit amplifying progenitors (TAP) in the adult mammalian brain has suggested that tumor cells may arise from cells other than mature neural cells [10]. Indeed, brain tumor stem cells have been shown to be involved in the initiation and propagation of several types of brain tumors, namely: GBM (CD133+) [11], which is the most frequent primary intraparenchymal neoplasm in the elderly; medulloblastomas (MB) (CD133+) [12], which has the highest incidence in children; and ependymomas (EP) (CD133+, Nestin+, and BLBP+) [13]. The SVZ has been shown to be a source for tumor stem cells that initiate gliomagenesis [14]. SVZ astrocytes (Type B cells) have been identified as NSCs in the mouse [15, 16] and adult mammalian brain [17]. Type B cells/NSCs have ultrastructural characteristics of brain astrocytes and express GFAP [15]. Importantly, in human, Middeldorp et al., showed that the adult SVZ is indeed a remnant of the fetal SVZ, which develops from radial glial cell (RGC). Furthermore, the study provides evidence that GFAP-delta can distinguish actively proliferating Type B cells/NSCs in SVZ [18]. Dividing GFAP-delta+ Type B cells/NSCs give rise to type C cells/TAPs [16, 18], a population of rapidly dividing, immature-appearing cells. The Type C cells, which are likely TAPs, express OLIG2 [19], MASH (ASCL1) [19], EGFR [20], SOX2 [21]. Type C cells/TAPs in turn produce type A cells, the neuroblasts which express TUJ1 and the PSA-NCAM [22-24].

In light of the anatomical origin of CN origin and its phenotype, we hypothesized that CN has neuroblast characteristics and originates from NSCs or TAPs in the SVZ. We first isolated CN-spheroids from CN-differentiated cells, and showed that tumor spheroids have TAP and RGC characteristics, and generate both neuron and astrocyte based on immune-histochemical and electrophysiological examinations. We then performed a comparative transcriptomic analysis of CN-spheroids compared to CN-differentiated cells and human NSCs, and found the tumor spheroids showed increased expression of genes for TAP, RGC, pluripotency, and tumor stem cell compared to these other cell types. Gene set enrichment analysis (GSEA) algorithm also showed enrichment of tumor spheroids-specific gene signatures in CN-derived tumor spheroids compared to CN-differentiated cells and SVZ-NSCs. In addition, intuitive neurogenesis signaling was detected in CN-differentiated cells and SVZ-NSCs.

Experimental design

We obtained tissues of CNs from surgeries of three patients (N=3), and then we performed primary cultures of CN tissues. To define cellular composition of CN tissues and primary cells, three biological replicates (N=3) from different patients were analyzed, and tumor spheroids were isolated from differentiated cells for characterization (RT-PCR and immunostaining) and electrophysiological study. Single tumor sphere RT-PCR was done to confirm that all individual spheres express tumor stem cell markers and to rule out any artifact which can arise from mixed population-derived RNA analysis. One of the three samples yielded a notable number of tumor spheroids for microarray. Other samples also made tumor spheroids, but the amount of RNA was not optimal for the microarray. This reflects that rare case report of malignancy of CN. As shown in Fig. 1b, a CN tissue that has a bigger size of tumor mass on MRI imaging made a robust number of tumor spheroids. For gene expression analysis, we compared CN-differentiated cells (N=1) and CN-spheroids (N=1) to defined specific genes, expressed on CN-spheroids. Gene expression of CN-spheroids were also compared with fetal SVZ-derived NSCs (N=1) to detect overlapped and differential expression. For a comparative study between samples, expression data was standardized using z-score transformation.

Human samples

Use of human CN tumor samples (N=3) for primary cultures in this study was approved by the Institutional Review Board (IRB) of Seoul National University Hospital (IRB approval C-0710-040-223). The human CN, normal ependymal tissues were obtained from the Brain Bank of Seoul National University Hospital, as approved by IRB of Seoul National University Hospital (IRB approval H-0B05-036-243). Fetal SVZ tissues (N=1) was obtained from inevitable abortion because of fatal diseases, and those tissues were primary cultured, as approved by IRB of Seoul National University Hospital (IRB approval H-0408-130-002). All methods were performed in accordance with the relevant guidelines and regulations of IRB board of Seoul National University Hospital. All the patient samples were obtained after obtaining written informed consent.

Primary culture of CN tissue

Methods for establishment of primary cultures for this study were approved by the Institutional Review Board (IRB) at Seoul National University Hospital (IRB No; C-0710-040-223). CN tissues from three patients (N=3) were mechanically dissected and enzymatically dissociated to single-cell suspensions using papain (Sigma-aldrich, St. Louis, MO) and DNase I (Sigma-aldrich) as described previously [25]. The cells were then suspended in DMEM/F12 (Invitrogen, Carlsbad, CA) supplemented with 5 μg/ml insulin (Sigma-aldrich), 50 μg/ml transferrin (Sigma-aldrich), 30 nM selenium chloride (Sigma-aldrich), and 5 μg/ml fibronectin (Sigma-aldrich) (ITSFn medium), and DMEM-FBS20%.

Immunostaining and fluorescent in situ hybridisztion (FISH)

Immunohistochemistry and immunocytochemistry [25, 26]were performed as described previously. Briefly, the cells were rinsed with PBS, fixed with 4% paraformaldehyde for 30 minutes at room temperature, and washed three times with PBS. Tissues were post-fixed by ice-cold 4% paraformaldehyde in PBS (pH 7.4) for one day at 4°, and then cryoprotected overnight in the same fixative supplemented with 25% sucrose. The tissues were embedded in OCT compound (Sakura Finetek, Inc., Torrance, CA), and frozen at -70° with dry ice. Sections (14 lm thick) were cut with a cryotome (Leica Microsystems, Wetzlar, Germany).

CN were analyzed by FISH for 1p and 19q on paraffin-embedded (FFPE) tissue sections to verify its diagnosis as describe in the previous study [27]. Dual, fluorescent-labelled DNA probes are used to detect 1p and 19q loci within the interphase nuclei of individual glioma cells from FFPE tissue sections transcribed on to unstained slides. Changes in the 1p and 19q probe signals compared with controls are used to determine the presence of 1p/19q-co-deletion.

List of antibodies

Primary antibodies used in this study are Nestin rabbit IgG (AB5922, 1:500, Millipore Sigma, Carlsbad, CA, USA); GFAP mouse IgG (IF03L, 1:1000, Millipore Sigma) and rabbit IgG (04-1062, 1:1000, Millipore Sigma); Human nuclei Millipore mouse IgG (MAB1281, 1:500, Millipore Sigma); BLBP rabbit IgG (ABN14, 1:500, Millipore Sigma); GFAP-delta rabbit IgG (AB9598, 1:500, Millipore Sigma); MAP2 rabbit IgG (AB5622, 1:500, Millipore Sigma) and mouse IgG (05-346, 1:500, Millipore Sigma); Tuj1 mouse IgG (MAB1637, 1:500, Millipore Sigma); PSA-NCAM mouse IgG (MAB5324, 1:500, Millipore Sigma); Sox2 rabbit IgG (ab75627, 1:500, Abcam, Cambridge, UK); NeuN mouse IgG (MAB377, 1:500, Millipore Sigma); EGFR Rabbit IgG (ab32562, 1:500, Abcam, Cambridge, UK). All secondary antibodies are Goat anti-Mouse IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488 (A-11001, 1:600, Invitrogen, Carlsbad, CA, USA); Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488 (A-11008, 1:600, Invitrogen); Goat anti-Mouse IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 568 (A-11004, 1:600, Invitrogen); Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 568 (A-11011, 1:600, Invitrogen).

Confocal microscopy

Fluorescently immunolabeled sections were analyzed on a Meta confocal microscope (model LSM 510; Carl Zeiss MicroImaging, Inc., Jena, Germany) equipped with four lasers (Diode 405, Argon 488, HeNe 543, and HeNe 633). Each channel was separately scanned using a multitrack PMT configuration to avoid crosstalk between fluorescent labels. Cells were mounted with antifading solution containing 4’-6-diamidino-2-phenylindole (DAPI; H-1200, Vector Laboratories Burlingame, CA, USA), and observed under a confocal microscope (Carl Zeiss MicroImaging).

RT-PCR

Total RNA was isolated from cells that had been cultured with either ITSFn medium or DMEM-FBS), using an RNeasy mini kit (Qiagen, GmbH, Germany). First-strand cDNA synthesis was carried out by random priming of the total RNA using a random primer mixture (Invitrogen) and reverse transcriptase with superscript 3 (Invitrogen). All designed primers (Bioneer, Republic of Korea) were screened using BLAST (Basic Local Alignment Search Tool) to ensure specificity of binding. Primers were used at a concentration of 250 nM. The PCR program entailed 10 min of the 9494T Basic Local Alignment Search Tool to ensure specificity of binding. The PCR program was as follows: 10 min 94° pre-run, 30 s at 94°, 30 s at 55°, 2 min at 72° for 35 cycles, and 10 min 72° post-run. No products were amplified in water. All the primer sequences are listed in Table 1.

Electrophysiology

CN cells were cultured as monolayers on 12 mm glass coverslips coated with poly-L-lysine (Sigma). External solution contained (in mM): 150 NaCl, 10 HEPES, 3 KCl, 2 CaCl², 2 MgCl², 22 sucrose, 10 glucose; pH was adjusted to 7.4 and osmolarity to 325 mOsm. The solution to fill the electrodes was composed of (mM) 140 K-gluconate, 0.5 CaCl², 1 MgCl², 5 EGTA, 10 HEPES, 4 Mg-ATP and 0.3 Na³-GTP. Whole cell patch recordings were obtained from CN-differentiated cells in voltage clamp or current clamp configuration using a multi-clamp 700B (Axon instruments, Union City, CA, USA) and patch pipette of 4~7 MΩ resistance. The solution to fill the electrodes was composed of (mM) 140 K-gluconate, 0.5 CaCl2, 1 MgCl2, 5 EGTA, 10 HEPES, 4 Mg-ATP and 0.3 Na3-GTP with 0.5% biocytin for morphological analysis.

Calcium imaging

Cells were loaded in 5 mM Fura 2-AM (Molecular Probes, Eugene, OR) and 0.01% pluronic acid in HEPES-buffered salt solution (HBSS) for 45 min at 37uC, and then unloaded in HBSS for another 15 min. Cells were plated onto poly-L-lysine coated 12 mm coverslips. Coverslips with Fura 2-AM loaded cells were then transferred to a perfusion chamber on the stage of an upright microscope (Olympus BX50, Japan). Cells were illuminated by a Xenonlamp and observed with a 40X UV water-immersion objective lens (Olympus, Tokyo, Japan). For Fura 2-AM excitation, the shutter and filter wheel (polychrome-IV; TILL-Photonics, Martinsried, Germany) were controlled by Axon Imaging Workbench (AIW) software 2.1 (Axon Instruments, Foster City, CA) to provide sequential illumination at two alternating wavelengths, 340 and 380 nm. Fluorescence of Fura 2-AM was detected at an emission wavelength of 510 nm. Video images were acquired using an intensified CCD camera (LUCA; Andor, UK). Fluorescence emission ratios following excitation at 340 and 380 nm were calculated by dividing averaged pixel values in circumscribed regions of individual responding cells in the field of view. The values were exported from AIW to Origin 8.0 for additional analysis and plotting. The composition of the HBSS was (in mM); NaCl, 137; KCl, 5; MgSO4, 0.9; CaCl2, 1.4; NaHCO3, 3; Na2HPO4, 3;Na2HPO4, 0.6; KH2PO4, 0.4; glucose, 5.6; and HEPES, 20; pH 7.4. For the depolarization conditioning, Ca2+ imaging were performed in HBSS containing high K+ (133 mM K+ with substitution of Na+).

Microarray analysis

After hybridization, the chips were stained and washed in a GeneChip Fluidics Station 450. After the final wash and staining step, the microarray arrays were scanned using Affymetrix Model 3000 7G scanner and the image data was extracted through Affymetrix GeneChip Command Console software 1.1. For the normalization, Robust Multi-Average algorithm implemented in Affymetrix Expression Console software 1.1 was used. Use of human CN tumor samples (Biological replicates, N=3) for primary cultures in this study was approved by the Institutional Review Board (IRB) of Seoul National University Hospital (IRB approval C-0710-040-223). Attributing to its benign characteristics of CN, only one sample among those three was able to make robust number of tumor spheroids. Other sample also made tumor spheroids, but the amount of RNA was not optimal for the microarray. We compared CN-differentiated cells (N=1) and CN-spheroids (N=1) to defined specific genes, expressed on CN-spheroids. Gene expression of CN-spheroids were also compared with Fetal SVZ-derived NSCs to detect overlapped and differential expression. For a comparative study between samples, expression data was standardized using z-score transformation. The GEO accession numbers for the gene expression and comparative genomic data used in this article are GSE42670 ‘Comparative transcriptomic analysis of central neurocytoma-derived tumor spheres as compared to its differentiated cells and fetal SVZ-derived neural stem cells’. For a normal control we used “GSM469410” from the public GEO database (http://www.ncbi.nlm.nih.gov/geo/).

GSEA analysis

We used GSEA [28] utilizing Molecular Signatures Database (MSigDB) [29] to determines whether an a priori defined set of genes (The False Discovery Rate (FDR)<0.25, p<0.01) shows statistically notable differences between CN-spheroids in comparison with CN-differentiated genes and SVZ-NSCs. For the analysis, we utilized category C2 among the MSigDB gene sets. Category C2 is curated gene sets from online pathway databases, publications in PubMed, and knowledge of domain experts. The FDR is calculated by comparing the actual data with 1000 Monte-Carlo simulations. The density of modified genes in the dataset is computed by the NES (Normalized Enrichment Score) with the random expectancies, and then it is normalized by the number of genes found in a given gene cluster.

Engraftment of cells

All procedures for animal experiments were approved by the IACUC (#07212) at Clinical research institute, Seoul National University Hospital. Experiments were performed in accordance with NIH guidelines. We assessed whether ITSFn-cultured cells are bipotential in vivo by evaluating their fate after engraftment to three immune-deficient NOD-SCID mice (male; N=3). CN cells (1×10⁵) and SVZ-derived NSCs (1×10⁵) were transplanted four weeks after primary culture to assess their lineage potential after transplantation (1×10⁵ cells per ml) along the anterior-posterior axis into the target brain at these coordinates: from the bregma, 2 mm anteriorly, 3 mm laterally and to 5 mm depth. The cells were live-stained with DiI (Invitrogen) delivered by an infusion pump at 0.5 ml/min. The needle was left in situ for 2 min post-injection before being slowly removed. The recipients were sacrificed and fixed four weeks after transplantation.

Quantification and statistical analysis section

For the quantitative analysis, a similar threshold was set for all images (three slices per sample) of in vitro CN-derived cells and mice tissues cells on which CN-derived cells are transplanted. Five regions of interest (ROIs) were randomly selected in the different regions of the images, and the area of specific immunoreactivity was measured using an image analyzer (Image-Pro Plus, Media Cybernetics Co., Silver Spring, MD). Statistical analysis was performed using GraphPad Prism 5. All data are presented as the mean+SD. Statistical significance was determined using unpaired two tailed t test, One-Way and Two-Way ANOVA followed by Bonferroni correction and t-test. Statistical details (p value, F, and degree of freedom (Df) are provided in Figure legends. A p value of <0.05 (*) or 0.01 (**), and 0.001 (***) was used to denote statistical significance.

Data availability statement

All the microarray data is available on GEO database. The GEO accession numbers for the gene expression and comparative genomic data used in this article are GSE42670, which includes four sample, GSM2977251: CN-differentiated cells, GSM2977252: CN-tumor spheres, and GSM2977253: Fetal-SVZ-derived NSCs.

The majority of CN cells display neuroblast characteristics

CN tumor masses are located in the SVZ of the patients as shown on representative MRI (magnetic resonance imaging) images (Fig. 1a, b). These tumors are diagnosed as CN based on its location and immunohistochemistry, in which H&E shows sheet of monotonous rounded cells with rounded nuclei, and Synaptophysin and NeuN are robustly positive in tumor cells of CN (Fig. 1a, b). In addition, FISH results show that two red and two green signals are evident in CN tissues, which implies no deletion of chromosome 1p and 19q (Fig. 2s). No complete co-deletion of short arm of chromosome 1 (1p) and the long arm of chromosome 19 (19q) (1p/19q co-deletion) shows typical characteristics of CN.

We detected expression of PSA-NCAM and Tuj1 (Fig. 1c) and light expression of NeuN and MAP2 (Fig. 1d), which suggests that most cells within CN tumors are not fully matured neurons. BMP2, PSA-NCAM, and Tuj1 were expressed in CN tissue sections (Fig. 1c) and CN-primary cells in ITSFn media, which promoted survival of NSCs at 1 week after culture (Fig. 1e). The majority of CN cells are round in shape in original tumor tissues (Fig. 1c, d) and primary cell population (Fig. 1f, e) without typical mature neuronal morphology. Semi-quantitative RT-PCR revealed that BMP2, BMPR1B, NCAM, and TUJ1, but not BMP4, were expressed in CN tissues and its differentiated cells (Fig. 1g). This phenotype is consistent with neuroblasts [22] supporting the hypothesis that CN cells have neuroblast characteristics. Interestingly, a few BLBP+/Nestin+ RGC like cells (Fig. 1h), numerous BLBP+/PSA-NCAM+ intermediate cells or PSA-NCAM+/Tuj1 neuroblasts (Fig. 1i) were observed in the CN tissue, suggesting that PSA-NCAM+/Tuj1+ cells might originate from BLBP+ cells in CN. These results are supported by gene expression of BLBP, Nestin, GFAP (Fig. 1j) and Tuj1 (Fig. 1g) in CN tissue and its primary cells at 1 week after dissociation.

CN has SVZ-hierarchal neurogenic cell composition

We next performed immunocytochemistry to examine the identity and distribution of neurogenic cell types present in CN. GFAP-delta, a marker for the SVZ-NSCs [18], and Tuj1 were immunostained on CN-primary cells (Fig. 2a~l) and CN tissue (Fig. 2m~p). In particular, the heterogeneity of the neurogenic cell types results in a non-uniform distribution of subpopulations across and within disease sites of CN tissue (Fig. 2m~p). We observed that, in the ITSFn-cultured CN primary cells at 1 week (Fig. 2a), CN tissue (Fig. 2m), and normal subependymal tissue (Fig. 3a), three kinds of cells were present: NSC-like GFAP-delta+, TAP-like GFAP-delta+/Tuj1+; GFAP-delta+/PSA-NCAM+; and neuroblast-like Tuj1+; PSA-NCAM+ cells. In contrast, we detected no TAP-like GFAP-delta+/Tuj1+ cells in the SVZ of fetal brain tissue (Fig. 3d). Importantly, GFAP-delta+ cells were predominantly located in the vicinity of the SVZ area and did not co-label with Tuj1 in the SVZ of fetal brain tissue, consistent with the immature status of fetal cells, compared to adult cells in the SVZ. This observation is supported by the previous studies as following. It has been reported that the transitional state of sibling TAPs express low levels of GFAP and DCX [30], and that GFAP+ cells differentiate to form intermediate Hu+(neuronal marker)/GFAP+ progenitors [31]. Similarly, an earlier study reported the presence of GFAP+/MASH1+ adult NSCs and MASH+/DCX+ adult TAPs in adult SVZ, which indicates a more mature state for adult NSC compared to fetal NSCs [32].

We also observed that in CN primary cells and tissue, NSC-like GFAP-delta+ cells (Fig. 2b, i, n) grew considerably in size exhibiting large bipolar-oriented processes. TAP-like GFAP-delta+/Tuj1+ cells (Fig. 2c, f, g) have an enlarged cell body with several processes. In contrast, the neuroblast-like Tuj1+ cells are round in shape with one side tail with the smallest size of cell body. (Fig. 2j). This observation is correlated with the previous report which shows a highly significant difference between the size of astroglial NSCs (largest) and either TAPs or neuroblasts (smallest) [30]. Interestingly, some of NSC-like GFAP-delta+ cells and TAP-like GFAP-delta+/Tuj1+ cells were localized in close opposition with a “twin-like” morphology (Fig. 2b, e), suggesting that NSC-like GFAP-delta+ cells divide asymmetrically to self-renew (GFAP-delta+), and to generate TAP-like GFAP-delta+/Tuj1+ cells. Ki67 expression shows that GFAP-delta+ cells are mitotic (Fig. 2k). Similarly, TAP-like GFAP-delta+/Tuj1+ cells and postmitotic neuroblast-like Tuj1+ cells were also located adjacent to each other, implicating these cells may also undergo asymmetric cell divisions reminiscent of RGC (Fig. 2c, d, f). Consistent with that notion, the small nuclei attached to TAP-like GFAP-delta+/Tuj1+ cells in Fig. 2c suggest asymmetric divisions have occurred. Moreover, these GFAP-delta+/Tuj1+ cells are resembled with their daughter Tuj1+ cells (Fig. 2c, d, f). Interestingly, Tuj1+ cell and GFAP-delta+/Tuj1+ cell in Fig. 2g recapitulate a neuroblast migrating along a long RGC fiber. Majority of GFAP-delta+/Tuj1+ cells display bipolar morphology, extending long process that was interestingly reminiscent of RGCs in the brain, and Tuj1+ cells have a round shape with tails attached on the fiber (Fig. 2g, h). In the prolonged culture (3 weeks after culture), the basal processes of GFAP-delta+/Tuj1+ cells were wrapped around the bundles of Tuj1+ neuroblasts (Fig. 2l). This observation suggests that TAP-like GFAP-delta+/Tuj1+ cells may constitute the tumor stem cells in CN. In particular, mitotic –and its lineage differentiation resulted in the heterogeneity of the neurogenic cell types as a non-uniform distribution of subpopulations across and within disease sites of CN tissue (Fig. 2m~p). Clustered TAP-like GFAP-delta+/Tuj1+ cells (Fig. 2p) were also observed in CN tissue, and their morphology resembled with daughter neuroblasts (Tuj1+). Fig. 2q shows summarized observation of the neurogenic subpopulation of the CN and characteristic of each cell type including distinctive cellular morphology and phenotypic expression. NSC-like cells (GFAP-delta+) comprised significantly a small portion of the CN primary cells (9.167%±1.878 N=3, p<0.05 and p<0.001 by t test compared to TAP-like cell and daughter neuroblasts, whereas TAP-like cells (GFAP-delta+/Tuj1+) comprised 22.67%±2.333 N=3, and daughter neuroblasts (Tuj1+) comprised 67.47%±2.784 N=3 (Fig. 2r). Overall, our results support the concept that the CN cell population has hierarchical neurogenic components similar to the SVZ.

CN-derived tumor spheroids show TAP and RGC-like characteristics

We initially maintained CN-derived cells in two culture conditions: ITSFn, which is optimal for proliferation and maintaining non-differentiation; and DMEM-FBS, which promotes differentiation. During the first week, the cells were predominantly flattened and attached. At that point, tumor spheroids began to form, not only in ITSFn media but also, surprisingly, in the DMEM-FBS media (Fig. 4a). Among three patient CN tissues, one sample made a significant number of tumor spheroids than other two samples. CN tissue, which has a bigger size of tumor mass on MRI imaging (Fig. 1b), made a robust number of tumor spheroids. To characterize properties of the CN-spheroids, we performed RT-PCR and IF on single tumor spheroids. We detected RGC makers, GFAP, NESTIN, and BLBP, in both fractions (Fig. 4b up), implicating an RGC origin for the CN-spheroids. Statistically, ITSFn-cultured CN-spheroids showed higher expression of NESTIN, SOX2, EGFR, and hTERT as compared to DMEN-FBS-cultured CN-spheroids (Fig. 4b bottom). TAP markers, MASH1, EGFR, and SOX were also detected in both spheroids (Fig. 4b). Neither fraction expressed NF 200, MAP2, and little TUJ1 was expressed (Fig. 4b), indicating the cells were in a TAP-like immature transient state. The expression of human Telomerase (hTERT) was rarely detected in DMEN-FBS-cultured CN-spheroids but pronounced in ITSFn-cultured CN-spheroids (Fig. 3b). Both ITSFn- and DMEM-FBS-cultured CN-spheroids displayed mixed population of NSCs (Nestin+/BLBP+/GFAP-delta+), TAPs (BLBP+/PSA-NCAM+; GFAP-delta+/Tuj1+), as well as a few neuroblasts (PSA-NCAM+ or Tuj1+) in the periphery of spheroids (Fig. 4c~f). In particular, CN-spheroids shows immature transient state with merged expression of BLBP+/PSA-NCAM+ and GFAP-delta+/Tuj1+ (Fig. 4d~f). This observation is consistent with clustered cell population of GFAP+/Tuj1+ in CN tissue (Fig. 2p). Accordingly, CN-spheroids expressed representative markers of TAP, namely ALCL1 (MASH1), EGFR (Fig. 4g), and SOX2 (Fig. 4h) in consistent with the RT-PCR result (Fig. 4b). Taken together, these observations suggest that CN tissue contains RGC- and TAP-like tumor stem cells in vivo.

The bipotentiality of RGC-like CN-spheroids

The observation that tumor spheroids can be generated in cultures of CN-derived cells prompted us to examine whether these cells can generate both neurons and glia. Accordingly, to evaluate differentiation potency, CN-spheroids were attached on the plate, and cultured up to 4 weeks (Fig. 4i, m). We observed that differentiation began in the cells localized on the periphery of the spheroids, while cells localized to the interior of the sphere remained SOX2+ immature cells (Fig. 4l, p). Then individual cells migrated out from spheroids (Fig. 4i, m) and begun to differentiate to form MAP2+/Tuj1+ neurons or GFAP+ astrocytes (Fig. 4j, n). Some of the cell were still immature Nestin+ cells (Fig. 4k, o). Majority of ITSFn-cultured cells showed neuronal differentiation (Fig. 4j), and DMEM-FBS-cultured cells generated mostly astrocytes (Fig. 4n), both spheroids are potential though. In parallel, elevations in intracellular free Ca2+ concentration in response to high KCl was two folds higher in the cells migrating from the ITSFn-cultured colony when compared to DMEM-cultured colony (Fig. 4q~s). We performed single spheroid RT-PCR on each ITSFn-cultured colony and DMEM-FBS-cultured colony. GFAP, MAP2, and TUJ1 expression was detected in both fractions, suggesting that both CN-spheroids are bipotential (Fig. 4t). Meanwhile, the detection of NESTIN and BLBP by RT-PCR, as shown in Fig. 4t, is attributed to the presence of Nestin+ cells (Fig. 4k, o). In that case, the persistence of mitotic neurogenesis and gliogenesis by single spheroids is consistent with the presence of cycling multipotential cells. Taken together, these observations suggest that CN tissue contains RGC- and TAP-like tumor stem cells in vivo.

In the prolonged culture up to 4 weeks, expression of MAP2 and NeuN decreased notably in DMEM-FBS-cultured cells, which resulted in widespread GFAP+ glia cells (Fig. 5a, b). Accordingly, DMEM-cultured cells had no changes in intracellular free Ca2+ concentration by membrane depolarization with high KCl (Fig. 5c). Meanwhile, in ITSFn-culture neural cells, expressions of MAP2 and NeuN (Fig. 5d, e) were maintained up to 8 weeks, and high KCl bathing solution for 60 sec increased intracellular free Ca2+ concentration (Fig. 5f). Neuron-like cells displayed rapid, reversible, >100% elevations in intracellular free Ca2+ concentration in response to high KCl, consistent with the activity of neuronal voltage-gated calcium channels (Fig. 5f), and it was blocked by 2 M nimodipine, a voltage-dependent calcium channel blocker (Fig. 5g). In addition, ITSFn-cultured cells was able to develop fast sodium currents and action potentials characteristic of eletrophysiologically competent neurons. Whole-cell patch-clamp recording was performed during current stimulation. CN-derived neurons in ITSFn showed action potentials by current steps of 20 pA (Fig. 5h). The action potentials were blocked by 0.5 M TTX (Fig. 5i). A total of 14 cells were recorded. Of these, 3 cells showed voltage-activated sodium ion currents. In addition, none of cells cultured in DMEM-FBS showed substantial current-induced sodium currents (Data not shown). Together, these results indicated that neurons arising from ITSFn-cultured CN cells developed mature electrophysiology functions, including both fast sodium currents and action potential. The biocytin solution was filled into the electrodes for morphological analysis, and these cells were confirmed as MAP2+ by immunostaining (Fig. 5j). Based on these observations, CN-derived cells have bipotential in response to different environmental niche.

GSEA indicated tumor stem cell features of CN-spheroids

GSEA [28] utilizing Molecular Signatures Database (MSigDB) [29] is a computational method that determines whether an a priori defined set of genes (FDR<0.25, p<0.01) shows statistically significant, concordant differences between two biological states. First, in comparison with CN-differentiated cells, CN-spheroids showed upregulation of gene sets in GBM cell lines displaying spherical growth (cluster-1) compared to those displaying semi adherent or adherent growth phenotype (cluster-2) (Fig. 6a) [33]. Gene sets of packaging of telomere ends (reactome) (Fig. 6b) and EGFR signaling (Fig. 6c) [34] were upregulated in CN-spheroids as well. These results are well correlated with the gene expression of hTERT and EGFR (Fig. 4b) and EGFR+ cells (Fig. 4g) in ITSFn-cultured CN-spheroids. In addition, CXCR chemokine receptor binding (Gene Ontology (GO)) (Fig. 6d) [35], KEGG glycine and serine and threonine metabolism (Fig. 6e) and KEGG steroid hormone biosynthesis (Fig. 6f) were enriched in CN-spheroids as compared to their differentiated cells. In particular, the serine, glycine, one-carbon (SGOC) metabolic network is implicated in cancer pathogenesis metabolism [36].

Secondly, in comparison with SVZ-NSCs, CN-spheroids showed upregulation of gene sets of brain cancer progenitors (Fig. 6g) [37], specific to esophageal adenocarcinoma (EAC) relative to normal tissue (Fig. 6h) [38], stromal stem cells (Fig. 6i), mesenchymal type of GBM tumors (Fig. 6j) [39], CDH1 signaling via CTNNB1 (Fig. 6k) [40], invasiveness signature resulting from cancer cell/microenvironment interaction (Fig. 6l) [41], Transforming growth factor beta binding (Gene Ontology (GO) (Fig. 6m), and LIF signaling (Fig. 6n) [42]. In addition, when compared to SVZ-NSCs, genes consistently overexpressed with age [43] were upregulated in CN-spheroids in consistent with its older age in comparison with fetal cells (Fig. 6o).

Analysis of differentially expressed genes (DEGs) shows RGC & TAP characteristics of CN-spheroids

We next performed differential expression gene (DEG) analysis (Fig. 7 and 8) on CN-spheroids, CN-differentiated cells, and SVZ-NSCs, to compare expression of genes associated with cancer stem cells, pluripotency, tumor suppression, and cancer-related signaling pathways. Importantly, we found that CN-spheroids were upregulated in expression of the cancer stem cell-associated genes CXCR12 [44], CXCR4 [44], CDK4 [45], PCDHB7 [46], AQP4 [47], TPD52L1 [48], ID1 [49], ID3 [49], CD44 [50], and PDPN [51], and the pluripotency-related genes OCT4 [52], NANOG, SOX2, KLF4, and CBX7 (Fig. 7b) [53], when compared to both CN-differentiated cells and SVZ-NSCs. Interestingly, Cbx7 has been shown to maintain pluripotency by repressing lineage-specific genes such as Cbx2 [53]. CN-spheroids showed a decrease, whereas, in the expression of the tumor suppressor-associated genes MCM2 [54], EPHB3 [55-57], and EPHB4 [55-57], and the neuronal differentiation-related genes FOXO3 [58], CCND1 [34], WNT7A [59], CDH2 [60], CNTN1 [61], SIRT1 [62], CBX2 [53]. ASPM [63], and INA (Fig. 8) [64].

We also found that in comparison with CN-differentiated cells, both CN-spheroids and SVZ-NSCs showed upregulation of genes for RGC (NOTCH3 [65], FABP7 [66], SLC1A3 [66], DAB1 [67], ERBB4 [68], and VCAM [69]), TAP (OLIG1 [19], OLIG2 [19], ASCL1 [19], SOX2 [21], and EMX2 [70]), SVZ-stem cell (CSPG2-VCAN, CSPG3-NCAN, CSPG5 [71], HAS3 [72], CDH1 (E-cadherin) [40], FZD3 [73], and AQP6) (Fig. 7c). RGC gene upregulation is well corresponded with BLBP (FABP7)+cells on CN- spheroids (Fig. 4c, d) and CN tissues (Fig. 1h, i).

TAP gene upregulation is also supported by protein expression of EGFR, MASH (Fig. 4g), and SOX2 (Fig. 4h) on CN-spheroids. On the other hand, CN-differentiated cells showed upregulation of neuroblast genes, ASPM [63], INA [64] and NRG1 [74] in comparison with CN-spheroids (Fig. 8a, b). This observation is corresponded well with the strong expression of PSA-NCAM and Tuj1 in CN tissues (Fig. 1c) and dissociated cells (Fig. 1e) at 1 week. Conversely, both CN-tumor spheres and SVZ-NSCs, were downregulated in expression of neuronal development related genes such as FABP5 [75], DIRAS [76], NEUROD1 [77], NTF3 [78], SLIT2 [79], CD24 [80], EMX1 [81], STMN2 [82], NRG1 [74], RELN [83], HES6 [84], L1CAM [85], and IGF2 [86], and the choroid plexus genes AQP1 [87] and EPHB2 (Fig. 8b) [88]. In particular, CSPG2 [71], CHL1 [46, 89], and OLIG1 [19] were significantly upregulated in both CN-spheroids and SVZ-NSCs as compared to CN-differentiated cells (Fig. 7e). CHL1 is expressed and functions as a malignancy promoter in glioma cells [89]. Among the CSPG family genes, CSPG2 was most strikingly upregulated in CN-spheroids as compared to their differentiated cells. Increased versican (CSPG2) expression is often observed in brain tumor growth [90].

In regard to tumorigenesis, both CN-tumor spheres and CN-differentiated cells showed an increase in expression of the embryonic oncogenesis-related genes: NET1 [91], MET [92], CD274 [93], and WNT2 [94], and the proliferation related genes FGF2 [78], EGF [20], EGFL6 [95], and TGBI (Fig. 7d). In contrast, they showed a decrease in expression of genes related to tumor suppression and negative control of proliferation, namely DIRAS1 [96], CXXC4 [97], EFNB2 [98], and EFNB3 [98] (Fig. 8).

Differentiation potential of the CN-derived cells in vivo

We next assessed the bipotentiality of ITSFn-cultured cells in vivo by evaluating their fate after engraftment to immune-deficient NOD-SCID mouse. CN cells (1×10⁵) were transplanted adjacent to the SVZ four weeks after primary culture to assess their lineage potential after transplantation. The recipients were sacrificed and fixed four weeks after transplantation. Human donor cells were identified by immunofluorescence of brain sections to detect human nuclear antigen (hNA) and DiI. Immunofluorescence analysis showed the presence of hNA+ cells in the grafts, demonstrating survival of grafted cells in the brain after transplantation (Fig. 9a, g). Most of the DiI+ CN-derived cells grafted in the striatum near SVZ were PSA-NCAM+ (64.00±3.786%, N=3) or Tuj1+ (75.67±2.333%, N=3) whereas fewer cells (17.00±2.082%, N=3%) were hNestin+ (Fig. 9b~f, k). Majority of Tuj1+ cells showed weak expression of GFAP (73.67±2.333% N=3), and this implicates these cells are differentiated from GFAP+/Tuj1+ TAPs. The PSA-NCAM+ cells displayed morphology similar to that of migrating neuroblasts (Fig. 9c). In contrast, most of the fetal brain-derived SNSCs grafted in the striatum near SVZ were Nestin+ (89.67±3.528% N=3), and very few cells were Tuj1+ (12.67±2.333 N=3) (Fig. 9h~j, l). Taken together, the data suggest that, in this experimental context, cell fate is determined more by intrinsic factors than by the environmental niche up to 4 weeks.

In this study, we demonstrated SVZ-neurogenic cell composition in CN tissue and primary cells. Majority of CN cells have neuroblast characteristics, and its spheroids possess TAP and RGC-like characteristics based on their gene and protein marker expressions. First we showed that CN tissue and CN-derived cells are comprised of three kinds of cells: GFAP-delta+, GFAP-delta+/Tuj1+, and Tuj1+ cells. The presence of GFAP-delta+/Tuj1+ cells is supported by an earlier study showing that the major population of GFAP+ proliferating cells were Tuj1-negative in the early SVZ, but Tuj1-positive in the adult SVZ, suggesting a decrease in postnatal neurogenesis occurs during development [99]. We hypothesize that GFAP-delta+, GFAP-delta+/Tuj1+, and Tuj1+ cells in CN correspond with NSCs, TAPs, and neuroblasts respectively, and CN-spheroids are a mixture of NSCs and TAPs, and that the majority of CN-differentiated cells are neuroblasts. In support of that hypothesis, we observed an upregulation of EPHB1 expression in CN-spheroids, which supports an NSC identity, whereas upregulation of EPHB2 in the CN-differentiated cells is consistent with neuroblast characteristics. This finding is supported by a previous report showing that EphB1 is expressed in SVZ astrocytes and that EphB2 is expressed Tuj1+ neuroblasts [88]. In consistent with the previous study which revealed that cell size decreases significantly from astroglial NSCs to TAPs or neuroblasts in vivo [22] , we observed that NSC-like GFAP-delta+ cells had elongated cell bodies with long process, and that TAP-like GFAP-delta+/Tuj1+ had an enlarged cell body with processes. Conversely, Tuj1+/PSA-NCAM neuroblasts have a round shaped morphology without mature neuronal structures, and they have the smallest cell body size. Although CN has been known as neuronal tumor, we observed that majority of CN cells express a neuroblast-associated PSA-NCAM, and DEG analysis showed upregulation of neuroblast genes, ASPM [63], INA [64] and NRG1 [74] in CN-differentiated cells when compared with CN-spheroids.

On the other hand, CN-spheroids showed more pronounced TAP-characteristics compared to SVZ-NSC, suggesting that these TAP-like cells may function as tumor stem cells of CN. First DEG analysis showed a notable expression of MASH1 [19], EGFR [20] and SOX2 [21], which are key markers for TAPs, on CN-spheroids in comparison with SVZ-NSC, and their expression on CN-spheroids is also confirmed by immunostaining and RT-PCR. The TAP characteristics of CN-spheroids is further supported by our observation that TAP-specific genes, OLIG1 [19], OLIG2 [19], and EMX2 [70] are upregulated in CN-spheroids, as compared to their differentiated cells. In particular GSEA analysis also showed that gene sets of EGFR signal are enriched in CN-spheroids in comparison with CN-differentiated cells. Our results are supported by the previous report Doetsch et al. [22], 2000, which revealed that the majority of EGF-responsive cells producing neurospheres in vivo are not derived from relatively quiescent NSC, but from the highly mitotic TAPs, and that TAPs maintain high levels of EGFR expression, perhaps related to their rapid division, and neuroblasts downregulate expression of the EGFR [22]. CN-spheroids also upregulated the gene sets of CSPG signaling such as CSPG2 (VCAN), CSPG3 (NCAN), and CSPG5 [71] which stimulate NSC survival through enhanced EGFR signaling [100]. We speculate that CN tumor stem cells have TAP characteristics, and EGFR signaling has a critical role in tumorigenesis or spheroid formation of CN (Fig. 10), and that their ephemeral nature of TAPs, destined to neuroblast, might reflect benign characteristics of CN.

Meanwhile, a downregulation of putative tumor suppressor genes, EPHB2, EPHB3 and EPHB4 of eph-ephrin signaling [55-57] in CN-spheroids could also induce tumorigenesis of CN. This hypothesis is supported by the early study which shows that breakdown of EPHB2 signaling results in invasion of SVZ astrocytes into the ependymal layer, which is accompanied by bulbous hyperplasia’s along the lateral ventricle [98]. Similarly, the EphrinB3−/− mouse showed an increase in NPC proliferation and the presence of more neurosphere-forming NPCs in the developing SVZ [101]. Further supporting the RGC characteristics of CN-spheroids, we found that in CN-spheroids and SVZ-derived NSCs, the NOTCH3 [65] coupled genes FABP7 [66] (BLBP) and DAB1 [67], SLC1A3 (GLAST) [66], VCAM [69], and ERBB4 [68] were upregulated. Interestingly, the ERBB4 oncogene, which is expressed at high levels in EP and is associated with increased tumor-proliferative capacity, is an important target of NOTCH signaling in RGC [102]. Most strikingly, IGF2 was downregulated 20- and 116-fold in CN-spheroids and SVZ-derived NSCs, respectively, as compared to CN-differentiated cells. Although IGF2 has been reported to also be important for the proliferation of MB cells [103], and the findings of Sim et al. [5], suggested that the dysregulated autocrine signaling by IGF2 contributes significantly to the genesis and expansion of neurocytoma, our findings suggest that IGF2 is more likely to be involved in neurogenesis than in NSC proliferation [5]. Our result is supported by a previous study reporting that knockdown of IGF2 in vitro or within the adult SVZ does not affect NSC proliferation, and IGF2 is strongly expressed in subsets of immature neuroblastoma cells [86], which correspond to the neuroblast-like CN-derived cells in our study.

CN is itself a rare tumor, and the aggressive form of CN is even rarer. Due to its rarity, this study might be more meaningful for the understanding of signaling and microenvironment which make aggressiveness of the tumor. In this study, all three samples made tumor spheres and all expressed cancer stem cell-specific markers, and one of the three samples, particularly, made notable number of tumor spheres. This observation reflects of variance among same diseases or heterogeneity of the individuals. And our observation of aggressive form of CN corresponds with one case report which addressed that some of aggressive form of CN progressed to glioma [3]. Intriguingly, MRI image of this case showed the size of Central neurocyoma is big which is full of the ventricular areas, and it is the same with the Patient 2 in our study. Signaling and gene expression of the rarely aggressive form of CN to our knowledge, have been never touched due to its scarcity. Here our study elucidates some enhanced genes in tumor spheroids of aggressive form CN which causes aggressiveness of this benign tumors. And these findings would give insightful information in searching for therapeutics of brain tumors.

In future study, it will be interesting to understand what intrinsic or micro-environmental factors contribute to their benign status compared with more aggressive tumors with similar gene expression profiles. Searching for genes that are differentially expressed in the aggressive tumor stem cells compared to CN tumor stem cells will be helpful for developing cancer targeted therapies.

In conclusion, here we show that tumor stem cells of CN possess TAP- and RGC-like characteristics with enhanced EGFR signaling which implicates its origin. In addition, by comparative DEG and GSEA analysis, we have identified CN spheroid genes that are consistent with a tumor stem cell function, as well as critical genes for NSC maintenance as well as neurogenesis. This research also provides innate neurogenesis signal, which is synchronized inartificially versus directed differentiation of human embryonic stem cells (hESCs) or induced pluripotent stem cells (iPSCs) into neurons.

This research was supported by the Korea Healthcare Technology R&D Project (grant no. HI11C21100200) of the Ministry of Health & Welfare, Republic of Korea; the Technology Innovation Program (grant no. 10050154, Business Model Development for Personalized Medicine Based on Integrated Genome and Clinical Information) of the Ministry of Trade, Industry & Energy (MI, Korea); the Bio & Medical Technology Development Program of the National Research Foundation (grant no. 2015M3C7A1028926) of the Ministry of Science and ICT, Republic of Korea; and the National Research Foundation of Korea Grant (grant no. NRF-2017M3C7A1047392) of the Ministry of Science and ICT, Republic of Korea to Sun Ha Paek.

Oligonucleotide primers and PCR conditions used for RT-PCR