FMRP modulates the Wnt signalling pathway in glioblastoma.

Converging evidence indicates that the Fragile X Messenger Ribonucleoprotein (FMRP), which absent or mutated in Fragile X Syndrome (FXS), plays a role in many types of cancers. However, while FMRP roles in brain development and function have been extensively studied, its involvement in the biology of brain tumors remains largely unexplored. Here we show, in human glioblastoma (GBM) biopsies, that increased expression of FMRP directly correlates with a worse patient outcome. In contrast, reductions in FMRP correlate with a diminished tumor growth and proliferation of human GBM stem-like cells (GSCs) in vitro in a cell culture model and in vivo in mouse brain GSC xenografts. Consistently, increased FMRP levels promote GSC proliferation. To characterize the mechanism(s) by which FMRP regulates GSC proliferation, we performed GSC transcriptome analyses in GSCs expressing high levels of FMRP, and in these GSCs after knockdown of FMRP. We show that the WNT signalling is the most significantly enriched among the published FMRP target genes and genes involved in ASD. Consistently, we find that reductions in FMRP downregulate both the canonical WNT/β-Catenin and the non-canonical WNT-ERK1/2 signalling pathways, reducing the stability of several key transcription factors (i.e. β-Catenin, CREB and ETS1) previously implicated in the modulation of malignant features of glioma cells. Our findings support a key role for FMRP in GBM cancer progression, acting via regulation of WNT signalling.

Introduction

Glioblastoma (GBM), the most frequent and malignant primary brain tumor in adults [1], is characterized by uncontrolled cellular proliferation, robust angiogenesis, propensity for necrosis, diffuse infiltration, resistance to apoptosis and genomic instability [2]. Even with current multimodal therapies, including maximal safe resection and radiotherapy supported by treatment with the alkylating agent temozolomide (TMZ), the mean survival rate for GBM patients is only 15 months [3]. Over the last decades, comprehensive studies have revealed a profound cellular and molecular heterogeneity in GBM [4-8]. The identification of a small population of tumor-initiating cells with stem properties, termed GBM stem-like cells (GSCs), in perivascular and hypoxic niches introduced a new and revolutionary target for therapy. GSCs, uniquely endowed with self-renewal capacity, multi-potency and induction of tumorigenesis [9, 10], are considered the primary cause of GBM chemo/radio-resistance and relapse, representing a relevant target for medical interventions [11-13].

The intricate molecular network underlying tumor progression makes highly complex the identification of potential pharmacological treatments. RNA-binding proteins (RBPs) are emerging as key regulators in cancer development, playing a key role in post-transcriptional control and gene expression homeostasis [14]. In this scenario, growing evidence shows the overexpression of the RBP Fragile X Mental Retardation Protein (FMRP) in several kinds of tumors, regulating cancer progression and invasiveness [15-18]. The FMR1 gene, encoding FMRP, is expressed in different tissues and cancer cell types (https://www.genevestigator.com/gv/).

FMRP is an RBP involved in multiple aspects of the mRNA metabolism, including negative regulation of translation [19-23], modulation of mRNA stability [24-26], transport [27-29], pre-mRNA splicing [30, 31] or RNA editing [32-34].

In addition to its well-established functions in the brain, independent clinical and molecular indications suggest that FMRP may also play a role in tumorigenesis. The absence of FMRP is known to cause Fragile X Syndrome (FXS), the most frequent form of inherited intellectual disability and syndromic autism [35]. Interestingly, a decreased risk of tumor incidence has been reported in a Danish cohort of patients with FXS [36]. Molecularly, a large subset of mRNAs, which play major roles in brain function and cancer, are targets of FMRP [37], and several FMRP protein interactors are mutated in various cancers [38]. Clinically, FMRP has been shown to be overexpressed in highly aggressive triple-negative breast cancers (TNBC). High FMRP levels also lead to increased number of metastasis through the regulation of mRNAs encoding proteins involved in epithelial to mesenchymal transition (EMT) [39]. Furthermore, recent data have shown that dysregulation of the N-methyl-D-aspartate (NMDA) receptor signalling pathway results in aberrant FMRP levels in pancreatic neuroendocrine tumors [40]. Finally, enhanced FMRP expression promotes proliferation through activation of MEK/ERK signalling in astrocytomas [41]. Notably, a case study of a boy with FXS diagnosed with GBM at 10 years old reported no specific neurological abnormalities throughout the clinical assessment, until the age of 18 [42]. Combined, these findings suggest that the absence of FMRP might have a “protective” effect against tumor growth, including in GBMs.

Here we investigated the contribution of FMRP to the biology of the most aggressive brain cancer, namely GBM. Given the importance of GSCs in GBM progression and their emerging role as therapeutic target, we investigated the molecular mechanisms driving GBM aggressiveness by analyzing GSCs derived from 28 GBM patients. Our data show that FMRP downregulation affects GSC proliferation ability in vivo and in vitro. In addition, transcriptome analysis highlights the WNT mediated signalling as a key dysregulated pathway upon FMRP reduction, with this inhibition of canonical and non-canonical WNT pathway leading to decreased expression and activity of different transcription factors, known to affect GMB proliferation. These findings demonstrate the key role for FMRP in the aggressive proliferation of GSCs and support the hypothesis that FMRP may constitute a potential therapeutic target for GBM.

Materials and Methods

Patients, diagnosis, and tumor characterization

Tumor tissue samples were derived from adult patients with GBM tumors (WHO grade IV) undergoing surgical resection at the Institute of Neurosurgery, Catholic University School of Medicine in Rome. Informed consent was obtained from the patients before surgery. All experiments involving human specimens were conformed to the principles described in the NMA Declaration of Helsinki and the NIH Belmont report. The expression of the proliferation marker Ki-67 and phosphatase and tension homolog (PTEN) were characterized on tumor specimen by immunohistochemistry. O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation patterns were assessed on genomic DNA extracted from paraffin-embedded tissue by methylation-specific PCR as previously described [43]. Levels of VEGF and EGFRvIII were assessed as previously described [44, 45].

Establishing GSC cultures

GSCs from 28 patients were isolated and characterized. Surgical specimens were subjected to mechanical dissociation and the resulting cell suspension was cultured in a serum-free medium supplemented with EGF and FGF as previously described [43]. Cell lines actively proliferating require 3–4 weeks to be established. In these conditions, cells grew as clusters (neurospheres) of undifferentiated cells, as indicated by morphology and expression of stem cell markers such as CD133, SOX2, Musashi-1, and nestin. The in vivo tumorigenic potential of GBM neurospheres was assayed by intracranial or subcutaneous cell injection in immunocompromised mice. GBM neurospheres were able to generate tumors with histological features mirroring the human parent tumor. GSC lines were validated by Short Tandem Repeat (STR) DNA fingerprinting. Nine highly polymorphic STR loci plus amelogenin (Cell IDTM System, Promega Inc., Madison, WI, USA) were used. Detection of amplified fragments was obtained by ABI PRISM 3100 Genetic Analyzer (Applied Biosystems, Carlsbad, CA, USA). Data analysis was performed by GeneMapper® software, version 4.0 (Biological Bank and Cell Factory, National Institute for Cancer Research, IST, Genoa, Italy). All GSC line profiles were challenged against public databases to confirm authenticity.

Mice and animal care

Immunosuppressed SCID mice used in this study are male, 4–6 weeks old, 20–25 g of body weight (Charles River, Milan, Italy). Mice were kept under pathogen-free conditions in positive-pressure cabinets (Tecniplast Gazzada, Varese, Italy) and observed daily for neurological behavior. A 12-h light/dark cycle was used, and food and water were available ad libitum. Experiments involving animals were approved by the Ethical Committee of the Università Cattolica del Sacro Cuore (UCSC) in Rome (Pr. No. CESA/P/51/2012). Animal experiments described in this study were conducted according to the guidelines set by the European Directive 2010/63/EU and the European Recommendations 526/2007 and the Italian D.Lgs 26/2014. Sample size for animal studies was chosen in order to have the smallest number of cases that allowed a statistical analysis.

Glioblastoma cell Line

The human GBM cell line T98G was purchased from the American Type Culture Collection (Manassas, VA, USA). The GBM cell line was cultured in Dulbecco’s Modified Eagle’s Medium (DMEM; Thermo Fisher Scientific, MA, USA), supplemented with 10% fetal bovine serum (FBS; Thermo Fisher Scientific, MA, USA) and 1% penicillin/streptomycin (Thermo Fisher Scientific, MA, USA) at 37 °C with 5% CO2.

Protein extract preparation

GSCs were lysed in ice-cold buffer (250 mM NaCl, 10 mM Tris/HCl pH 7.4, 10 mM MgCl2, 1% Triton X-100, 10 µl/ml Protease inhibitor cocktail (Roche, Mannheim, Germany), 10 µl/ml Phosphatase inhibitor cocktails II and III (Merck KGaA, Darmstadt, Germany), 40 U/ml RNaseOUT (Invitrogen, Thermo Fisher Scientific, MA, USA).

Western blotting

Standard methodologies were used. Protein extracts were separated by SDS–PAGE electrophoresis, and transferred to a PVDF membrane (GE Healthcare, Milan, Italy). FMRP levels were detected using a specific polyclonal antibody, namely PZ1. A synthetic peptide corresponding to aminoacid 548 to 564 of the mouse FMRP (gene ID 14265) was used for the production of the rabbit polyclonal antibodies employed in this study (21st Century Biochemicals). As shown in the Fig. S1A, the antibody recognizes at least three FMRP isoforms in mouse neuronal and brain protein extracts and at least two FMRP isoforms in human fibroblasts and iPSCs protein extracts).

Membranes were incubated using the following specific antibodies including human anti-Vinculin (1:2000, Merck, Cod. V9131), anti-GSK3β (1:1000, Cell Signaling Technology Inc., Danvers, MA, USA, Cod. 4337), anti-phospho-GSK3β (Ser9, 1:1000, Cell Signaling Technology Inc., Cod. 5558), anti-β-Catenin (1:2000, Invitrogen, Cod. 13-8400), anti-WNT5B (1:500, Abcam, Cambridge, MA, USA, ab124818), anti-p44/42 MAPK (Erk1/2, 1:1000, Cell Signaling Technology Inc., Cod. 137F5), anti-phospho-p44/42 MAPK (Erk1/2 Thr202/Tyr204, 1:1000, Cell Signaling Technology Inc., Cod. 9101), anti-ETS1 (1:1000, Cell Signaling Technology Inc., Cod. 14069), anti-phospho-ETS1 (T38, 1:500, Abcam, Cod. ab59179), anti-CREB (1:1000, Cell Signaling Technology Inc., Cod. 48H2), anti-phospho-CREB (Ser133, 1:1000, Cell Signaling Technology Inc., Cod. 87G3), anti-FXR2P (1:1000, Merck, Cod. F1554). Secondary HRP-conjugated anti-rabbit, anti-mouse antibodies (1:10000) were purchased from Promega (Milan, Italy, Cod. W4011 and W402B, respectively). Proteins were revealed using an enhanced chemiluminescence kit (Bio-Rad) and the imaging system LAS-4000 mini (GE Healthcare). Quantification was performed using the IQ ImageQuant TL software (GE Healthcare). Coomassie staining of the membranes and Vinculin signals were used as normalizers.

The full and uncropped figures are provided as Supplemental Material.

Quantitative real-time-PCR (RT-qPCR)

Total RNA was extracted from GSCs with TRIzol according to the manufacturer’s protocol (Life Technologies, Thermo Fisher Scientific, Cod. 15596026). For the cDNA synthesis 500 ng of total RNA was used as input into a 20 μl reaction using p(dN)6 and 200U/ μl M-MLV RTase (Invitrogen, Thermo Fisher Scientific, 28025013). mRNAs were quantified by real-time PCR using SYBR® Green Master Mix (Bio-Rad, Segrate (MI), Italy) on StepOnePlus™ Real-Time PCR machine (Life Technologies, Thermo Fisher Scientific) according to the manufacturer’s instructions using specific primers. The mRNAs levels were expressed in relative abundance compared to HPRT1 or H3 gene (2ˆ(-delta delta CT) method). Specific primers were used for the amplification of the selected genes:

hFMR1 5′-TGTCAGATTCCCACCTCCTG-3′; 5′-TAACCACCAACAGCAAGCCT-3′

hHPRT1 5′-TGCTGAGGATTTGGAAAGGGT-3′; 5′-TCGAGCAAGACGTTCAGTCC-3′

hACTB 5′-ACCGAGCGCGGCTACAG-3′; 5′-CTTAATGTCACGCACGATTTCC-3′

hWNT5B 5′-GCAGAAGGTTGACAGCTTCAGT-3′; 5′-ACAGTTTCCAGAGTAGGGTTCC-3′

hCTNNB1 5′-TTGAAGGTTGTACCGGAGCC-3′; 5′-GCAGCTGCACAAACAATGGA-3′

hH3 5′-GTGTCATCCATGCCAAACGG-3′; 5′-GTGGCGAGATAGCCCTCCTA-3′

Polysome-mRNP analysis

GSCs were lysed as previously described [46, 47] with slight modifications described below. The supernatant was loaded onto a 15–50% (w/v) sucrose gradient and sedimented by centrifugation at 4 °C for 150 min at 37,000 rpm in a Beckman SW41 rotor (Fullerton, CA). Each gradient was collected into 10 fractions (1–6 = polysomes; 7–10 = mRNPs) while reading the absorbance at 254 nm, followed by the addition of 50 pg of spike-in control (luciferase control RNA, Promega L456A). Total RNA was extracted from each fraction, precipitated and the RNA quality was assessed by gel electrophoresis and spectrophotometry (ND-1000 spectrophotometer, Nanodrop Technology). The mRNAs of interest (FMR1, and luciferase) were quantified by RT-qPCR.

Immunohistochemistry for FMRP and staining evaluation score from GBM tissues

Formalin-fixed, paraffin-embedded sections (40 µm thick) were mounted on positive charged glass slides. The series of GBM samples used for IHC (n = 60) differed from that used for the generation of GSCs (n = 28). The procedure to obtain these samples is different. While the paraffin-embedded specimens for IHC can be available just one week after surgery, bona fide GSCs require 12.8 ± 7.7 weeks to obtain sizable tumorspheres [48] and additional 4–6 months for cell grafting and in vivo studies. Only bona fide GSCs were used. For antigen retrieval to detect FMRP protein, deparaffinized and rehydrated sections were boiled in TRIS-EDTA buffer solution (pH 9) for 20 min. The slides were cooled, and endogenous peroxidase were blocked with peroxidase block buffer (citric acid 0.04 M, Na2HPO4 × 2H2O 0.12 M, NaN3 0.03 M, and H2O2 at 1.5% v/v) for 15 min at room temperature. Then, the sections were incubated for 2 h with rabbit polyclonal antibodies against FMRP (1:200, PZ1).

The primary antibodies were visualized using the avidin-biotin-peroxidase complex method (UltraTek HRP Anti-polyvalent, ScyTek, Logan, UT) according to the instruction manual. 3,3′ diaminobenzidine was used as substrate to observe the specific antibody localization, and Mayer hematoxylin was used as a nuclear counterstain.

Staining intensity of tissue slides was evaluated independently by 2 observers (L.M.L. and M.M.), who were blinded toward the patients’ characteristics and survival. Cases with disagreement were discussed using a multi-headed microscope until agreement was achieved. To assess differences in staining intensity, an immunoreactivity scoring system was applied. FMRP expression in each specimen was scored according to the percentage of stained cells and intensity of nuclear staining. Immunohistochemistry (IHC) was scaled as 0 for no IHC signal at all, 1 for 1–30%, 2 for 31–70%, and 3 for 71–100% of tumor cells stained. The score for IHC intensity was also scaled as 0 for no IHC signal, 1 for weak, 2 for moderate, and 3 for strong IHC signals. The final score used in the analysis was calculated by multiplying the extent score and intensity score, with a maximum score equal to 9. Samples with score 0–3 were arbitrary identified as low FMRP expression, while samples with score 4–9 were identified as high FMRP expression. The specificity of FMRP detection was tested on GBM sections stained in the absence of the primary antibody and on human breast cancer sections expressing high or no FMRP levels as in [39]. At least two sections were stained for each sample, FMRP expression was highly reproducible.

RNA extraction from GBM tissues

RNA was extracted from three, independent, 10-µm section from paraffin-embedded tissues of each patient using RNeasy FFPE Kit (QIAGEN, Milan, Italy), following the manufacturer’s protocol. Real-time PCR was performed using the KAPA SYBR FAST One-Step qRT-PCR Kit (KAPA-Biosystems, Boston, Massachusetts, USA), following the manufacturer’s protocol using the CFX96™ Real-Time PCR Detection System (Bio-Rad). Each analysis was performed in duplicate and the expression level of FMR1 mRNA was normalized by ACTB mRNA.

Lentiviral production and stable cell lines

The lentiviral expression vector to silence FMR1 (shFMR1) or non-target control (shNTC) were purchased from Sigma (Merck, pLKO.1-puro-CMV-tGFP shRNA control, and pLKO.1-puro-CMV-tGFP shFMR1 clone ID: TRCN0000059761 and shFMR1 clone ID: TRCN0000298271). The overexpression of FMR1 mRNA was obtained using the lentiviral vector (207.pRRLsyn.PPTs.hCMV.GFP.Wpre) kindly provided by Prof. Suzanne Zukin (Albert Einstein College of Medicine, USA). These vectors also contain the GFP as reporter gene that allowed the selection of transduced cells by Fluorescent Activated Cell Sorting (FACS).

Lentiviral particles were produced by the calcium-phosphate transfection protocol in the packaging human embryonic kidney cell line 293T. Briefly, the lentiviral construct was cotransfected with pMDL, pRSV-REV and pVSV-G. The calcium-phosphate DNA precipitate was removed after 8 h by replacing the medium. Viral supernatants were collected 48 h post transfection, filtered through a 0.45 μm pore size filter, and added to GSCs in the presence of 8 μg/ml polybrene. Cells were centrifuged for 30 min at 1800 rpm. After infection, the fluorescence of transduced cells was evaluated by FACSCanto (Becton Dickinson).

In vivo xenografts

Immunosuppressed SCID mice were anesthetized with intraperitoneal injection of diazepam (2 mg/100 g) followed by intramuscular injection of ketamine (4 mg/100 g). Animal skulls were immobilized in a stereotactic head frame and a burr hole was made 2 mm right of the midline and 1 mm anterior to the bregma. The tip of a 10 µl-Hamilton microsyringe was placed at a depth of 3 mm from the dura and 2 × 104 of either shFMR1 GSC#148 or shNTC GSC#148 or shFMR1 GSC#163 or shNTC GSC#163 were slowly injected. After grafting, the animals were kept under pathogen-free conditions in positive-pressure cabinets (Tecniplast Gazzada, Varese, Italy) and observed daily for neurological signs. For survival curves, the mice were sacrificed when the body weight dropped to 80% of initial weight or at appearance of neurological signs. The mice were deeply anesthetized and transcardially perfused with 0.1 M PBS (pH 7.4) then treated with 4% paraformaldehyde in 0.1 M PBS. The head was fixed in the stereotactic head frame, the skull was removed, and the brain was blocked 2 mm posteriorly to the grafting site. The anterior block of the brain was stored in 30% sucrose buffer overnight at 4 °C and serially cryotomed at 20 µm on the coronal plane. Sections were collected in distilled water and mounted on slides with Vectashield mounting medium (Bio-Optica, Milan, Italy). Images were acquired with a laser scanning confocal microscope (LSM 500 META, Zeiss, Milan, Italy). The tumor volume was determined according to the equation: V = (a2 × b)/2, where a is the mean transverse diameter calculated through the tumor epicenter and b is the cranio-caudal extension of the tumor. The brain posterior to the grafting site was post-fixed in 10% formalin for 48 h, embedded in paraffin, and cut at 4 µm thick sections.

Fluorescence Microscopy and Immunohistochemistry on brain xenografts

For immunofluorescence, coronal sections of the brain were blocked in PB with 10% BSA, 0.3% Triton X-100 for 45 min. Sections were incubated overnight at 4 °C with primary antibodies in PB with 0.3% Triton X-100 and 0.1% normal donkey serum (NDS). Polyclonal antibodies used was Ki67 (1:500, Thermo Fisher Scientific, Cod. RM9106S1). For detecting brain microvessels, sections were incubated overnight at 4 °C in PB with 0.3% Triton X-100 and 0.1% NDS with Lectin from Lycopersicon esculentum (tomato) biotin conjugate (1:500, Sigma-Aldrich, St. Louis, MO) together with primary antibodies. Slices were rinsed and incubated in PB containing 0.3% Triton X-100 with secondary antibodies for 2 hours at RT. For lectin staining, sections were incubated for 2 h at RT in PB containing 0.3% Triton X-100 with streptavidin protein, DyLight 405 conjugate or streptavidin Alexa Fluor® 647 conjugate (1:200, Thermo Fisher Scientific, Waltham, MA, USA). Before mounting, slices were incubated with DAPI (1:4000; Sigma-Aldrich) for 10 min. Immunofluorescence was observed with a laser confocal microscope (SP5; Leica) and images were acquired. Image analysis was performed with Leica Application Suite X software. The MIB-1 staining index was determined as the percentage of Ki67-positive cells relative to the total number of cells in high power fields (400×). In each tumor specimen, at least 1500 tumor cells were counted. The number of neoformed vascular structures was assessed in lection stained coronal sections of xenografted brains through the tumor epicenter.

Proliferation and cell viability assays

Cell viability: shNTC or shFMR1 transduced GSCs were plated at density of 2 × 103/ml in 96 well plates in triplicate. Cell viability was monitored by counting the cells and confirmed by using the CellTiter-Blue™ Viability Assay (Promega). Cell proliferation: was evaluated by Bromo-2′-deoxyuridine (BrdU) incorporation using BrdU Cell proliferation ELISA kit (Abcam, Cod. ab126556) following the manufacturer's instructions. The motility of transduced GSCs was evaluated by plating in Corning FluoroBlokTM Multiwell Inserts System (Corning Life Sciences, Tewksbury, MA), according to the manufacturer's instruction. Briefly, 3x103 cells were added to the upper chambers in stem cell medium without growth factors (GFs). GF completed medium was used as chemoattractant in the lower chambers. The plates were incubated for 48 h at 37 °C, after which the fluorescent dye calcein acetoxymethylester (calcein AM, Life Technologies) was added to the lower chamber for 30 min. The cell viability indicator calcein AM is a non-fluorescent, cell-permeant compound that is hydrolyzed by intracellular esterases into the fluorescent anion calcein and can be used to fluorescently label viable cells before microscope observation. The number of migrated cells was evaluated by counting the cells after imaging acquisition using a fluorescence microscope (Nikon Eclipse TS100).

RNA-seq and statistical analysis

To study the FMRP-regulated pathways in GBM, we compared existing datasets derived from human glioblastoma and mouse brain. For the human dataset we interrogated the genes mutated in GBM listed in the cBioPortal for cancer genomics (http://cbioportal.org) selecting six studies of glioblastoma stemming from The Cancer Genome Atlas (TCGA, four studies), Mayo Clinic (one study in 2019, Columbia University (one study in 2019) for a total of 1184 patients. Genes that were mutated in at least 1% of all patients were considered for further analysis (1077 genes are mutated). For the mouse datasets, FMRP target mRNAs in mouse hippocampus derived from two CLIP-seq studies were selected. For each of the two relevant studies [49, 50], the log2 of the count of sequence tags (indicating the FMRP binding strength) was z-score normalised. Genes that had, on average, a z-score of at least +1 (top ~37% of FMRP crosslinked mRNA targets) were considered (550 genes), and their human homologs were identified using the vertebrate homology table of the Mouse Genome Informatics (MGI) database (http://www.informatics.jax.org/downloads/reports/index.html#homology). The overlap (52 genes) of both gene lists (1077 and 550) was screened for enrichment in the PANTHER pathway database using the tool provided (pantherdb.org); the reference list was set to all genes that were reported either in cBioPortal/GBM or in both of the CLIP studies. The overlap (73 genes) of the genes frequently mutated in GBM (see above) and the list of genes associated with autism (1023 genes, Simons Foundation Autism Research Initiative, SFARI, https://gene.sfari.org) was analyzed in the same way.

RNA-seq was performed on control and FMR1-silenced cells. GSCs stably transduced with shNTC or shFMR1 vectors were expanded and GFP-positive cells selected by FACS. Total RNA was extracted from control or FMR1 silenced cells using TRIzol reagent (Life Technologies). Three independent cell expansion and FACS selection rounds were performed. The quality and quantity of the RNA were measured by spectrophotometry (Nanodrop, Thermo Fisher Scientific), and RNA integrity was verified on a Bioanalyser (Agilent). If quantity (minim 100 ng) and quality (RIN factor of 5) of the RNA was deemed sufficient mRNA expression profile (RNA-seq) was analyzed by Illumina HiSeq at the Genomic Technologies Facility (GTF) of the University of Lausanne, Switzerland (www.unil.ch/gtf). Reads were aligned to the human genome (GRCh38.86) and assigned to the respective mRNAs using the Cufflinks algorithm; differential expression between samples has been analyzed by the DESeq module of the Bioconductor suite. To gauge the effect of FMR1 silencing, the log2(FPKM) of all samples for a given cell line was compared by principle component analysis (PCA, function prcomp() in R), plotting the first and second dimension (package ggplot). Human genes linked to WNT-related pathways (excluding “cell polarity”) were downloaded from geneontology.org (441 genes, 370 of which were detected in the RNA-seq experiment in the three cell lines). Gene Set Enrichment Analysis (GSEA) was performed as follows: the Wnt-associated genesets (N=70) was downloaded from the Molecular Signature Database (https://www.gsea-msigdb.org/gsea/msigdb/). The log fold change (logFC) of gene expression was calculated comparing shFMR1 and the shNTC GSCs transcriptome for a total of 16572 genes, which were then used to run pre-ranked GSEA (GSEA, https://www.gsea-msigdb.org/gsea/index.jsp) using weighted enrichment statistics and 1000 random sample sets permutation.

RNA stability assay

T98G cells were silenced for FMRP using FMR1-specific siRNAs from Life Technologies (Carlsbad, CA, USA) (AM 16708, ID 10824, 10919, and 11010). As a nonspecific control, a scrambled siRNA was used (Cod. 4390843; Life Technologies). siRNA was transfected into T98G cells using Lipofectamine RNAiMAX (Life Technologies, Cod. 13778100), according to the manufacturer’s instructions. Transfections were carried out with 90 pmol of siRNA, and cells were used for the experiment after 48h. T98G control cells and T98G FMR1-silenced cells were treated at t = 0 with Actinomycin D (1 mg/ml) for 0, 2, 4, 6, 8 h. RNA was extracted and RT‐qPCR performed as previously described.

Quantification and statistical analysis

Data quantification has been described in the figure legends and in some of the methods in this section. Statistical analysis was performed using GraphPad-Prism 5 software (Graph Pad Software, San Diego, CA) and MedCalc version 10.2.0.0 (MedCalc Software, Mariakerke, Belgium). Comparisons between the two conditions, shNTC and shFMR1, were performed using two-sample two-tailed Student’s t-tests or One-sample t-test. Growth curves were analyzed using two-way ANOVA. Correlation was assessed by Pearson regression analysis. Comparison of categorical variables was performed by chi-square statistic, using the Fisher’s exact test. Kaplan-Meier survival curves were plotted and differences in survival between groups of patients were compared using the log-rank test. Statistical comparison of continuous variables was performed using Mann-Whitney U-test. Significance was denoted as *P < 0.05, **P < 0.01, ***P < 0.001. Error bars represent the standard error of the mean (SEM).