Aberrant coupling of glutamate and tyrosine kinase receptors enables neuronal control of brain tumor growth

Summary

Direct and paracrine neuron-cancer interactions govern tumor development and progression. While neuron-elaborated neurotransmitters, like glutamate, support neoplastic growth, the mechanism underlying tumor intracellular mitogenic signaling and proliferation remains an unresolved question in cancer neuroscience. Herein, we discover that glutamate receptor (GluR) stimulation phosphorylates Src to activate platelet-derived growth factor (PDGF) receptor-α-dependent ERK signaling and drive glioma growth. Using single-cell transcriptomic datasets and unique laboratory-generated humanized models of the most common brain tumor in children (pilocytic astrocytoma, PA), we find tumor cell glutamatergic pathway enrichment, such that glutamate increases PA proliferation without changing membrane depolarization. Aberrant GRID2 and GRIK3 GluR expression increases RAS/ERK signaling by selective Src-mediated PDGFRα activation. Moreover, genetic or pharmacologic GRID2/GRIK3 and PDGFRA inhibition reduce PDGFRα/RAS/ERK activation, PA cell proliferation, and PA xenograft growth. Taken together, these observations establish a conceptual framework for understanding similar neurotransmitter dependencies in other cancers.

eTOC blurb

Anastasaki and colleagues establish a previously unknown glutamate growth dependency in pediatric low-grade brain tumors (gliomas). Glioma cells hijack normal neuron-glial molecular circuits, present during neurodevelopment, using aberrant glutamate receptor stimulation to activate PDGFRα/ERK via Src. This uncovers a molecular link between neurotransmitter signaling and tumor growth in children.

Graphical Abstract

Introduction

While most human cancers arise in close association with nerves, the idea that neurons can directly influence tumor growth has only gained traction over the past decade, ushering in the new field of cancer neuroscience¹. Numerous studies have revealed that neuronal activity can drive tumor initiation and progression through the establishment of direct synaptic connections^2–5 or the elaboration of activity-regulated paracrine factors (e.g., NLGN3^6,7, BDNF⁸, IGF-1⁹). For example, neurons originating in the olfactory bulb (olfactory receptor neurons) and retina (retinal ganglion cells) release paracrine factors critical for brain tumor (glioma) initiation (e.g., IGF-1, NLGN3^9,10), while other neuron bioactive molecules, such as neurotrophic factors (e.g., BDNF, GDNF^8,11,12), collagen subunits¹³, and midkine¹³, support tumor progression in the brain and peripheral organs. In addition to these growth regulatory molecules, neurons release neurotransmitters, which can similarly function as critical regulators of cancer growth and spread.

While numerous neurotransmitters have been implicated in cancer growth, including acetylcholine^14,15 and beta-adrenergic receptor ligands^16,17, glutamate is the most widely described driver of tumor biology throughout the body (reviewed in^1,18,19). Within the brain, glutamate receptor activation is important for pediatric and adult high-grade (malignant) glioma growth^2,4,20,21, invasion³ and metastatic colonization by breast cancer²². Similarly, non-central nervous system cancers, such as pancreatic neuroendocrine tumors²³, melanoma²⁴, breast cancer^21,25, ovarian cancer²¹, and lung cancer²⁶ cells are controlled by glutamate receptor expression and signaling²¹.

However, despite years of investigation, the mechanism underlying glutamate-induced intracellular mitogenic signaling and tumor growth remains unclear. Herein, we sought to answer this unresolved question using the most common brain tumor encountered in children (World Health Organization grade 1 pilocytic astrocytomas; PAs). We specifically chose this typically non-fatal cancer histotype for proof-of-concept studies using versatile human ex vivo and in vivo models, since PAs are caused by well-defined genetic drivers (either bi-allelic loss of the neurofibromatosis type 1 [NF1] gene or a genomic KIAA1549: BRAF rearrangement)^27,28 that result in increased oncogenic pathway activation. Additionally, similar to high-grade gliomas ^29–35, PAs likely arise from oligodendrocyte precursor cells (OPCs)^35,36, which naturally interact with excitatory neurons in the developing and mature brain in an activity-dependent manner through the elaboration of glutamate^37–40. Leveraging a combination of human PA single-cell RNA sequencing (scRNAseq) datasets, numerous newly established patient-derived PA cell lines, and unique humanized PA-patient derived xenograft (PDX) mouse models, we discovered a previously unknown tumor growth dependency created by coupling glutamate receptor (GluR) stimulation to activation of platelet-derived growth factor (PDGF) receptor (PDGFRα)-mediated mitogenic signaling. These findings illustrate how tumor cells usurp physiologic neuron-OPC interactions to establish new stromal dependencies that favor cancer growth, as well as provide a conceptual framework for studying neurotransmitter receptor growth regulation in other cancers both within and outside of the brain.

Results

Glutamate increases PA cell growth

To determine whether the most common pediatric brain tumors (PAs) establish neurotransmitter growth dependencies, we first leveraged three independently-generated pediatric PA single-cell RNA sequencing datasets: GSE244433, recently generated by our group⁴¹, and two publicly available datasets (GSE222850⁴², phs001854.v1.p1⁴³), containing a total of 15 sporadic and 1 NF1-associated PA tumors (Figure 1). Whether analyzed separately as individual tumors or as an aggregate of 16 total tumors, the PAs segregated into 8 distinct cell clusters. In the case of GSE244433, two smaller clusters that did not express any discernable cell type markers were removed from subsequent analyses (Figure S1A–S1B). The remaining 6 clusters included a tumor cell cluster (ASCL1, ETV5, FABP7, GFAP, PDPN, GPR17), as well as TAMs (TMEM119, P2RY12, CX3CR1), T cells (CD3E, CD8A, CD8B, CD27), dendritic cells (CLEC10A), B cells (MS4A1) and neurons (MAP2, NEUROD1 neurons) (Figure S1C). Additionally, the tumor cell cluster had enriched expression of MAPK signaling (e.g., SPRY4, SPRY2; Figure S1D⁴²), oligodendrocyte progenitor cell (e.g., OLIG1, OLIG2) and glioma (e.g., FABP7, GFAP, PDPN^44,45) genes (Figure S1E). The genes used to classify the PA cell clusters are included in Figure S1.

Figure 1. Glutamate receptor expression is increased in PA tumor cell clusters.

A. UMAP plot from scRNAseq analyses and summary of demographic data of 5 pediatric pilocytic astrocytomas (PAs; GSE244433). Each cluster is represented by a different color and indicated within each panel. B-C. (B) Graphic representation and (C) table summarizing the top 10 upregulated gene ontology (GO) pathways in the tumor cells relative to all other cells in the five GSE244433 PAs. (D) Bubble plots illustrating relative glutamate receptor expression enrichment within each cell cluster of the GSE244433 PAs. E-F. UMAP plots from scRNAseq analyses of (E) five (GSE222850), and (F) six (phs001854.v1.p1) independent pediatric PAs. Each cluster is represented by a different color and indicated within each panel. G-H. Bubble plots illustrating relative glutamate receptor expression enrichment within each cell cluster in (G) the GSE222850, and (H) the phs001854.v1.p1 sets. See also, Figure S1, and Tables S1–S2.

Focusing on the comparison of the tumor cluster to all other clusters, we performed differential pathway and gene expression analyses and found that the top upregulated gene ontology (GO) terms in the tumor cells from the 5 PAs in the Washington University dataset (GSE244433) included glutamatergic signaling, with all top 10 upregulated pathways involving synaptic and neuronal processes (Figures 1B–C, Tables S1–S2). Mirroring these results, following differential gene expression analyses, we found that PA tumor cells predominantly expressed glutamate receptors (GluRs; Figure 1D), rather than other neurotransmitter receptors. In addition, parallel analyses on 11 additional PAs from the other two available scRNAseq datasets (GSE222850, phs001854.v1.p1) revealed similar tumor cell glutamatergic pathway (not shown) and glutamate receptor (Figures 1E–H, Figures S1C–E) enrichment, raising the possibility that glutamate receptor signaling might regulate low-grade glioma biology.

To determine whether this GluR enrichment contributed to PA cell growth in vitro, primary pediatric PA cell lines, representing both sporadic (Sp-PA, n=6) and NF1-associated (NF1-PAs, n=4; Figure 2A) were analyzed. For these studies, we used two previously published PA cell lines (Res186⁴⁶, JHH-NF1-PA1⁴⁷) and 8 new PA cell lines independently generated from St Louis Children’s Hospital operative specimens (Sp-PAs harboring KIAA1549:BRAF rearrangements, n=5 [WUPA1, WUPA6, WUPA8, WUPA11, WUPA12]; NF1-PAs with biallelic NF1 loss, n=3 [WUPA3, WUPA4, WUPA10]). Glutamate increased the proliferation (%Ki67⁺ cells and BrdU incorporation) of all 10 PA lines in a dose- (1.5–1.8-fold increase in %Ki67⁺ cells and 1.2–2.2-fold increase in BrdU incorporation relative to PBS vehicle controls; Figures 2B–C, Figure S2A) and time– (1.1–1.7-fold increase %Ki67⁺ cells and 1.9–3.1-fold increase in BrdU relative to vehicle controls; Figures 2D–E, Figure S2A) dependent manner. Moreover, compounds that selectively bind to and activate NMDA, AMPA and kainate GluRs increased PA cell proliferation to comparable levels as glutamate (1.8-fold increase in %Ki67⁺ cells and 2-fold increase in BrdU incorporation relative to vehicle-treated controls, Figures 2F–G, Figure S2B). It should be noted that glutamate, NMDA, AMPA, and kainate also had no effect on programmed cell death (apoptosis, cleaved caspase-3⁺ cells; Figure 2H, Figure S2B). Similarly, glutamate had no effect on cellular senescence (Figure 2I, Figure S2C) or metabolic conversion to α-ketoglutarate (Figure S2D). Since glutamate increased the proliferation all ten PA cell lines in a similar manner (1.8-fold increase in %Ki67⁺ cells and 2.1-fold increase in BrdU incorporation, Figure 2B–C, 1.9-fold increase in resorufin fluorescence, Figure S2E), we selected representative sporadic and NF1-associated PA lines for all subsequent experiments.

Figure 2. Glutamate increases patient-derived pediatric PA cell proliferation in vitro.

A. Summary of demographic, brain location, and molecular and pathologic data from the pediatric PA specimens acquired from others or at WUSM used to generate six sporadic (Res186⁴⁷, WUPA1, WUPA6, WUPA8, WUPA11, WUPA12) and four NF1-associated (JHH-NF1-PA1⁴⁶, WUPA3, WUPA4, WUPA10) PA tumor cell lines. B-G. Quantification of PA cell proliferation (%Ki67⁺ cells and BrdU incorporation) following (B-C) increasing doses of glutamate and (D-E) increasing incubation times (50μM glutamate). Color-coding of PA lines is indicated within the graph (B, D. n=3 biological replicates for all cell lines for all conditions; C, 0μM glutamate, n=3 biological replicates for all cell lines; 10–50μM glutamate, n=4 biological replicates for all cell lines; E, 0h, Res186, WUPA6, WUPA8, WUPA11, WUPA12, WUPA3, WUPA4, n=4; WUPA1, JHH-NF1-PA1, WUPA10 n=3; 1h, Res186, WUPA10, n=3; remaining cell lines, n=4; 6h, WUPA8, n=3; remaining cell lines, n=4; 12–24h, n=4 for all cell lines and conditions). (F-G) Exposure to 50μM glutamate, 10μM NMDA, 10μM AMPA, and 10μM kainate similarly increase WUPA cell proliferation (n=3 biological replicates for each cell line for each condition). H. Glutamate, NMDA, AMPA, and kainate do not induce apoptosis (%cleaved caspase-3⁺ cells; n=3 biological replicates for each cell line for each condition). I. 50μM glutamate does not induce senescence (%β-galactosidase⁺ cells; n=3 biological replicates for each cell line for each condition). Color-coding of PA lines is indicated within the graph in panel C. Data are shown as the mean ±SEM. B, D, F-H one-way ANOVA with Dunnett’s post-test correction, C, E two-way ANOVA with Tukey’s multiple comparison test, I two-tailed student’s t-test. p values are indicated within each graph. ns, not significant. See also Figure S2.

Next, we performed several experiments to determine whether the mitogenic glutamate derives from neurons in proximity to the PA cells. First, we isolated primary cerebellar neurons derived from Rag1−/− neonatal mice, the host strain of PA-PDXs previously established by our group⁴⁵. These neurons were selected as they project to and innervate the two most common regions where PAs arise in children (cerebellum and brainstem). We found that cerebellar neurons produce glutamate (15μM/mg total protein; Figure S3A), which is similar to the doses administered to PA cells our in vitro experiments (Figure 2B–C). Moreover, Rag1−/− cerebellar neurons increased PA cell proliferation when maintained as co-cultures (31% increase relative to controls, Figure S3B), as well as when Rag1−/− cerebellar neuron conditioned medium (CM) was added (2.2-fold increase in resorufin fluorescence relative to controls, Figure S3C). Importantly, in these neuron-PA cell co-cultures, Rag1−/− cerebellar neuronal processes encircled the PA cells, with presynaptic proteins (e.g., synapsin, bassoon) localizing to PA cell borders (Figure S3D). Together, these data support the hypothesis that neurons associate with PA cells and that neuron-derived glutamate is sufficient to induce PA cell proliferation in vitro.

Glutamate does not alter PA cell electrical activity in vitro

Glutamatergic neuronal stimulation of malignant gliomas evokes action potentials in tumor cells to regulate their proliferation^2,4. To determine whether PA cells similarly respond to glutamate by increasing membrane excitability, we employed three complementary methods: (1) direct whole cell patch-clamp recordings (Figure 3A, B), (2) multi-electrode arrays (Figure 3C, D), and (3) calcium imaging (Figure 3E, F). In striking contrast to primary murine hippocampal neurons, which responded to glutamate exposure (50μM) during a step current injection, a method used to determine activity response to prolonged depolarization (Figure 3A), PA cells did not exhibit action potentials and showed no evidence of any glutamate-gated conductance, following glutamate application (Figure 3B), or evidence of excitatory postsynaptic currents. Moreover, glutamate did not alter PA cell spiking rates (Figures 3C, D) or calcium influx (Figures 3E, F), suggesting that glutamate drives PA cell growth through a mechanism unrelated to PA cell excitability.

Figure 3. Glutamate does not induce ionotropic receptor activity or alter network activity in PA cell electrical activity in vitro.

A-B. Representative traces (left panels) and quantification (right panel) of whole cell patch clamped (A) primary mouse hippocampal neurons (positive controls; n=3 biological replicates for each condition) and (B) PA cells (WUPA4; n=4 biological replicates for each condition) are shown. Application of 50μM glutamate induces an inward current in neurons, but not in PA cells, at −70 and −30mV. C-F. Representative traces (C, E) and quantification (D, F) of spikes in sporadic (left panels, Res186, black dots; WUPA1, grey dots) and NF1-associated (right panels, JHH-NF1-PA1, blue dots; WUPA4, green dots) PA cell lines reveal no changes in cell electrical activity following 50μM glutamate treatment as measured by MEA (C, D; Veh, n=4; Glu, n=3 biological replicates for all cell lines) and calcium imaging (E, F; Res186, WUPA1, WUPA4, n=3; JHH-NF1-PA1, n=6 biological replicates for each condition). Data are shown as the mean ±SEM. Two-tailed student’s t-test. ns, not significant.

Pharmacologic inhibition or genetic silencing of GluR reduces PA cell proliferation in vitro and in vivo

To determine whether glutamate receptors could be targeted to inhibit PA cell growth, we employed an array of inhibitors designed to block AMPA (perampanel, NBQX, topiramate), kainate (topiramate, CNQX) and NMDA (memantine, felbamate) glutamate receptors (Figure 4, Figure S3E–G). Of these agents, only memantine reversed glutamate-induced PA cell proliferation in vitro (47–56% reduction in %Ki67⁺ cells relative to glutamate exposure; Figure 4A).

Figure 4. Pharmacologic glutamate receptor inhibition reduces PA cell proliferation in vitro and in vivo.

A. Memantine (2μM) inhibits glutamate-induced proliferation (%Ki67⁺ cells; Res186, Veh n=5, Glu, n=5, Glu+Mem, n=4 biological replicates; WUPA1, n=5 biological replicates for all conditions; JHH-NF1-PA1, Veh, n=5, Glu, n=5, Glu+Mem, n=4 biological replicates; WUPA4, n=5 biological replicates for all conditions). B. Representative H&E and Ki67 staining of the parental tumors and respective WUPA-PDX tumors growing in Rag1^−/− mice at 1-month post-injection (mpi). The dotted lines demonstrate the tumor boundary. The percentage of Ki67⁺ cells is indicated within each respective panel. Scale bars, 100μm. C. Immunofluorescent imaging demonstrates the proximity of presynaptic neuronal terminals (synapsin⁺ and bassoon⁺ cells) to PA-PDX tumor cells in 1mpi Rag1−/− mice in vivo. The dotted lines highlight the brain-tumor interface. Scale bar, 50μm. D. (Left) Schematic illustration of the experimental design used for these proof-of-concept treatment studies. (Right) Representative immunohistochemistry images and quantitation demonstrate that in vivo memantine treatment (20mg/kg/day) reduces tumor proliferation (%Ki67⁺ cells) in Sporadic- (n=7 mice in both treatment groups), and NF1-PA-PDX tumors (Veh, n=4 mice; Mem, n=5 mice). The dotted lines indicate the tumor boundary. Data are shown as the mean ±SEM. A. One-way ANOVA with Dunnett’s post-test correction, C, two-tailed student’s t-test. P values are indicated within each graph. Veh, vehicle; Glu, glutamate; Mem, memantine; LGG, low-grade glioma. See also Figure S3.

Since memantine is being currently evaluated in pediatric clinical trials for its ability to reduce neurocognitive adverse effects in children undergoing radiotherapy for brain tumors (NCT03194906), we employed a previously established patient-derived PA xenograft (PA-PDX) platform in which neonatal Rag1^−/− mice are injected with 5×10⁵ tumor cells in the brain location from which the original patient tumor arose⁴⁵. All PA lines formed tumors at 1-month post-injection (mpi) and were scored as low-grade gliomas by an experienced neuropathologist (F.J.P.) based on their histologic appearance (H&E), low proliferative index (%Ki67⁺ cells), GFAP/OLIG2 immunopositivity, and lack of differentiated neuronal synaptophysin staining (Figure 4B, Figure S3H–I). Similar to children with these low-grade neoplasms, PA-PDX implantation did not affect the overall survival of injected mice and did not result in any observed clinical symptomatology.

To assess the proximity of neurons to PA-PDX tumors in vivo, using thunder microscopy we observed synapsin⁺ and bassoon⁺ presynaptic structures extending into the PA-PDXs (Figure 4C). Using this PA-PDX platform in which neurons are in proximity to the tumor cells, following LGG establishment, 4 weeks of oral memantine (20mg/kg) treatment was sufficient to reduce PA-PDX proliferation in vivo (79–84% reduction in %Ki67⁺ cells; Figure 4D, Figure S3J). Together, these data establish that PA tumor growth is dependent upon glutamate in vitro and in vivo.

GRID2 and GRIK3 GluR subunit expression is enriched in PAs

To identify specific glutamate receptor subunit expression in pediatric PAs, we first leveraged an additional bulk RNA transcriptomic dataset with 33 unique PAs (GSE163071¹⁰). When we examined the expression of all glutamate receptor genes in pediatric PAs relative to non-neoplastic pediatric brain tissue, we identified GRID2 and GRIK3 as two GluRs consistently increased in a total of 48 tumors, including this new dataset (GSE163071; Figure 5A), and the 16 previously analyzed scRNAseq PAs (GSE244433, GSE222850, phs001854.v1.p1, Figures 5B, C). Second, we leveraged several tissue microarrays containing a total of 80 distinct PA tumor specimens (Table S3) to demonstrate that 72.5% and 73.8% of pediatric PAs, respectively, exhibited GRID2 and GRIK3 expression relative to control non-neoplastic cortex tissue, as measured by immunohistochemical staining (Figure 5D). In contrast, the non-neoplastic cortex tissue lacked GRID2 and GRIK3 expression. Third, to ascertain whether the observed increase in GRID2 and GRIK3 expression was a consequence of NF1 loss or specific to tumor development, we compared human iPSC-derived control and NF1-deficient OPCs, the putative PA cells of origin to PA tumor cells. We found that GRID2 and GRIK2 expression was increased in PA cells relative to both NF1-null and control OPCs (2.4- to 4.9-fold increase in GRID2, 12.5- to 39.7-fold increase in GRIK3 expression relative to NF1-OPCs; Figures 5E–F), whereas there was no difference between control and NF1-null OPCs. Fourth, to establish the necessity of GRID2 and GRIK3 in mediating glutamate-driven PA cell proliferation, we employed lentiviral shRNA constructs to genetically silence GRID2 or GRIK3 expression (Figure 5G–H, Figure 4A–C). Knockdown of either gene reduced glutamate-induced PA cell proliferation (37–45% reduction in %Ki67⁺ cells, Figure 5G–H; 49–51% reduction in resorufin fluorescence, Figure S4D–E), NMDA, AMPA and kainate-induced PA cell proliferation (47–50% reduction in resorufin fluorescence, Figure S4F), or Rag1−/− cerebellar neuron-induced proliferation (49–42% reduction in %Ki67⁺ cells and resorufin fluorescence, respectively; Figure S4G–H) relative to control medium-treated cells. In addition, genetic GRID2 or GRIK3 silencing had no effect on apoptosis or senescence in vitro (Figure 5G–H, Figure S4D–I). Importantly, as the effect of simultaneous GRID2 and GRIK3 silencing did not confer an additional growth disadvantage to PA cells relative to single GRID2 or GRIK3 silencing (Figure S4G–H), we used only single knockout lines for all subsequent experiments. Furthermore, tumor proliferation at 1-mpi was reduced in shGRID2- or shGRIK3-knockdown PA-PDXs in vivo relative to control PA-PDXs (59–63.9% reduction in %Ki67⁺ PA cells relative to control PA-PDXs) (Figure 5I, Figure 4J). Fifth, to assay the direct effect of GRID2 or GRIK3 in increasing PA proliferation, we employed two separate lentiviral constructs to ectopically express GRID2 (1.2×10³-fold increase in GRID2 expression, Figure S5A) or GRIK3 (3.1×10⁵-fold increase in GRIK3 expression; Figure S5B) in representative PA cell lines. We demonstrated that ectopic expression of GRID2 or GRIK3 increased PA cell proliferation (resorufin fluorescence) by 2.2- or 2.3-fold, respectively, relative to GFP control-infected PA cells (Figure S5C). Taken together, these findings demonstrate that both GRID2 and GRIK3 are necessary and sufficient for glutamate-dependent PA cell proliferation in vitro and in vivo.

Figure 5. GRID2 and GRIK3 expression is enriched in pediatric PAs and their genetic silencing reduces PA cell proliferation in vitro and in vivo.

A. Glutamate receptor RNA expression in pediatric PAs from the GSE244433 PA scRNAseq dataset and GSE163071 bulk RNAseq samples of non-neoplastic brain, sporadic PAs, and NF1-PAs. The increase in GRID2 and GRIK3 RNA expression is highlighted. B, C. Feature plots illustrating GRID2 and GRIK3 expression in the tumor cell clusters from three independent PA scRNAseq datasets. Relative expression is indicated by increasing color saturation. D. GRID2 and GRIK3 protein expression is increased in 58/80 and 59/80, respectively, of pediatric PA samples relative to non-neoplastic human cortex. E, F. Quantitative RT-PCR reveals increased (E) GRID2, and (F) GRIK3 expression in sporadic and NF1-PA cells relative to CTL and NF1-null OPCs. n=3 biological replicates of each cell line in each condition. Data are shown as the mean ± SEM. One-way ANOVA with Dunnett post-test correction. P values are indicated within each graph. G-H. Quantitation of Ki67⁺ and cleaved caspase-3⁺ PA cells demonstrate a reduction in proliferation (%Ki67⁺ cells) following (G) GRID2 (n=3 biological replicates of each cell line in each condition) or (H) GRIK3 knockdown (n=3 biological replicates of each cell line in each condition) using three independent shRNA constructs, but no effect on PA cell apoptosis (%cleaved caspase-3⁺ cells; G, n=3 biological replicates of each cell line in each condition; H, n=4 biological replicates of each cell line in each condition). I. Representative immunohistochemistry images and quantitation of Ki67⁺ cells demonstrate that either shGRID2 or shGRIK3 knockdown reduces PA tumor proliferation in Rag1^−/− mice (n=6 mice for each PA-PDX cohort). The dotted lines indicate the tumor boundary. Data are shown as the mean ±SEM. One-way ANOVA with Dunnett’s post-test correction. P values are indicated within each graph. ns, not significant. Scale bars, 100μm. See also Table S3, and Figures S4–S5.

Glutamate receptor-driven PA proliferation is mediated by PDGFRα-MEK/ERK signaling

As PAs are considered tumors of deregulated MEK/ERK signaling as a result of KIAA1549: BRAF-mediated MEK or biallelic NF1 loss-induced RAS/MEK activation^27,28, we assayed RAS and ERK1/2 activity in control and shGRID2 and shGRIK3 PA cells. We found that GRID2 and GRIK3 silencing reduced both RAS activity (RAS-GTP) (shGRID2, 48–51%; shGRIK3, 50% RAS-GTP reduction; Figure 6A) and ERK1/2^Thr202/204 phosphorylation in vitro (shGRID2, 55–64%; shGRIK3, 49–53% phospho-ERK1/2 reduction; Figure 6B, Figure SD-BE) relative to control PA cells and in vivo relative to shCTL PA-PDXs (Figure S5F). Conversely, ectopic GRID2 or GRIK3 expression increased ERK activation (shGRID2, 1.9-fold increase, shGRIK3, 1.75-fold increase; Figure S5G) relative to control-infected PA cells. Moreover, memantine treatment following glutamate exposure reduced RAS-GTP (45% reduction relative to glutamate exposure alone; Figure 6C), similar to MEK (PD0325901, 51% reduction) and RAS (IN-1, 54% reduction) inhibition (Figure 6C), as well as reduced ERK1/2 phosphorylation in PA cells (44% reduction relative to glutamate exposure alone; Figure 6D, Figure S5H) and in PA-PDXs in vivo (Figure S5I). It is worth noting that these glutamate-mediated effects operate independently of PI3K/AKT/mTOR signaling (phospho-Akt^{Thr308/ Ser473}; phospho-S6^Ser235/236) (Figure S5J–K).

Figure 6. Glutamate increases PA cell proliferation in a PDGFRα/RAS/ERK-dependent manner.

A-D. 50μM glutamate (Glu) increases RAS/MAPK activation in PA cells, which is blocked by (A) GRID2 or GRIK3 shRNA knockdown, as well as (C) glutamate receptor (2μM memantine, Mem), MEK (1μM PD0325901; PD) and pan-RAS (0.5μM IN-1) inhibition (n=3 biological replicates in all cell lines and conditions). B, D. Western blotting quantitation reveals that glutamate induces ERK1/2^Thr202/204 phosphorylation (activation), which is blocked by (B) GRID2 or GRIK3 shRNA knockdown (n=3 biological replicates in all cell lines and conditions), as well as (D) 2μM memantine treatment (sporadic PA, WUPA1, black dots; Veh, n=5, Glu, n=5, Glu+Mem, n=3 biological replicates; NF1-associated PA, WUPA4, blue dots; Veh, n=3, Glu, n=4, Glu+Mem, n=3 biological replicates). E. Feature plots indicate expression of the PDGFRA RTK in the tumor cell clusters from each of the three PA scRNAseq datasets. F. Quantitative RT-PCR reveals increased PDGFRA expression in CTL and NF1-null OPCs, as well as in sporadic (grey) and NF1-associated (green) PA cells relative to CTL and NF1-null NPCs and GRPs (n=3 biological replicates for all cell types and conditions). Data are shown as the mean ± SEM. One-way ANOVA with Dunnett post-test correction. G. Glutamate (50μM) induces PDGFRα phosphorylation (activation). Sporadic PA, WUPA1, black dots; NF1-associated PA, WUPA4, blue dots (n=3 biological replicates for all cell lines and conditions). H. Glutamate increases PA cell proliferation (%Ki67⁺ PA cells), which is inhibited by PDGFRα inhibition (1μM avapritinib, Ava). Apoptosis (%cleaved caspase-3⁺ PA cells) is not affected by any treatment. (sporadic PA cells: Res186, black dots; WUPA1, grey dots; NF1-PA cells: JHH-NF1-PA1, blue dots; WUPA4, green dots; n=3 biological replicates for all cell lines and conditions). I. Representative immunohistochemistry images and quantitation of Ki67⁺ cells demonstrate that avapritinib treatment (50mg/kg/day) reduces PA tumor proliferation in Rag1^−/− mice. The dotted lines indicate the tumor boundary. Data are shown as the mean ±SEM (Sp.PA-PDX, black dots; n=3 mice for all conditions; NF1-PA-PDX, green dots; Veh, n=3 mice, Ava, n=4 mice). A, C, F, H, one-way ANOVA with Dunnett’s post-test correction; D, one-way ANOVA with Tukey’s post-test correction; B, G, I, two-tailed student’s t-test. P values are indicated within each graph. ns, not significant. Veh, vehicle; Glu, glutamate; Mem, memantine; Ava, avapritinib. See also Figure S6.

Since RAS/ERK activation is most commonly induced by receptor tyrosine kinase (RTK) signaling, we used all 16 unique PAs analyzed by scRNAseq (GSE244433, GSE222850, phs001854.v1.p1) to identify potential RTKs expressed in the tumor cells. While some RTKs commonly implicated in cancers were not expressed in PA tumor cells (e.g., EGFR, HGFR, KIT, RET, FGFR2; Figure S6A), and NTRK2 was only expressed in the tumor cells of the GSE244433 dataset (Figure S6A), PDGFRA, and FGFR1 were detected in the tumor cells (Figure 6E, Figure S6B). TRKB, the NTRK2 protein, and FGFR1 were not activated (phosphorylated) in in PA lines following glutamate exposure (Figure S6C), suggesting they are not responsible for GluR-mediated ERK activation in PAs. However, expression of PDGFRA, which is the major mitogenic RTK in OPCs, was increased in PA cells (23-fold increase relative to control and NF1-deficient NPCs and GRPs). In contrast, PDGFRA was expressed at comparable levels to CTL and NF1-null hiPSC-derived OPCs (19- to 27-fold increase relative to control and NF1-deficient NPCs and GRPs; Figure 6F), demonstrating that its increased expression was not a consequence of NF1 loss. Importantly, PDGFRα activation (phosphorylated; PDGFRα^{Tyr754/Tyr849}) was increased following glutamate exposure (2.4-fold increase relative to vehicle treatment; Figure 6G, Figure S6D). PDGFRα pharmacologic inhibition using avapritinib, a selective and well-tolerated PDGFRα inhibitor that can cross the blood-brain barrier⁴⁸, blocked glutamate-mediated PA cell proliferation in vitro (71% reduction in %Ki67⁺ cells, Figure 6H, Figure S6E) and in PA-PDX tumors in vivo (69% reduction in %Ki67⁺ cells, Figure 6I) as well as reduced ERK activation in vivo (Figure S6F).

To determine PDGFRA necessity for mediating glutamate-driven PA cell growth, we genetically silenced PDGFRA using lentiviral shRNA constructs (88% reduction, Figure S6G) and demonstrated that, following glutamate incubation in vitro, PDGFRA-silenced PA cells have reduced pERK activation (52% reduction, Figure S6H) and proliferation (54% resorufin fluorescence reduction, Figure S6I) relative to shCTL PA cells. It should be noted that genetic PDGFRA knockdown following ectopic GRID2 or GRIK3 expression restored GRID2/GRIK3-mediated ERK activation and PA cell proliferation to control shRNA-infected PA cell levels (Figure S6J–K).

To examine a potential epistatic relationship between PDGFRA and the GRID2/GRIK3 glutamate receptors, we leveraged a lentiviral PDGFRA construct to ectopically express PDGFRA in control and shGRID2 or shGRIK3 PA cells. We found that ectopic PDGFRA expression (21.6-fold increase in PDGFRA mRNA, 9.8-fold increase in PDGFRA protein; Figure S6L) increased control PA cell proliferation by 2.24-fold and pERK activation by 2.3-fold (Figure S6M–N). However, ectopic PDGFRA expression following either GRID2 or GRIK3 knockdown had no effect (Figure S6M–N).

Taken together, these data establish that PDGFRα functions downstream of glutamate receptor activation to control PA tumor growth. While PDGFRα is required for glutamate receptor-mediated and ERK-dependent PA cell proliferation and glutamate receptor activation regulates PDGFRα-MEK signaling, the exact mechanism underlying this coupling requires further investigation.

PA cells couple GluR and PDGFRα-ERK signaling to drive glioma growth

Since during brain development PDGF-AA is produced by neurons and functions as the physiologic ligand for PDGFRα in OPCs⁴⁹, we assessed the effect of glutamate and PDGF-AA on normal control and NF1-deficient (NF1^−/−) OPC ERK activation and proliferation (Figures 7A–B). While in OPCs, there was no increase in ERK activation (phosphorylation, phospho-ERK1/2^Thr202/204) or proliferation following glutamate exposure, PDGF-AA increased both ERK phosphorylation (2.1–2.5-fold increase, Figure 7A, Figure S7A) and proliferation (CTL,3.9-fold; NF1^−/− 1.7-fold increase, Figure 7B, Figure 7B) relative to untreated OPCs. In contrast, PDGF-AA and glutamate both increased ERK1/2 phosphorylation in PA cells, and combined PDGF-AA/glutamate exposure increased ERK activation (2.9-fold increase, Figure 7C, Figure S7C) and PA cell proliferation (4.1-fold increase, Figure 7D, Figure S7D) to levels greater than either mitogen alone.

Figure 7. Glutamate receptor-PDGFRα convergence underlies RTK-ERK-dependent PA growth regulation.

A-B. 20mg/mL PDGF-AA (PDGF-AA), but not 50μM glutamate (Glu), induces (A) ERK1/2 phosphorylation (activation; each dot represents the average of two pooled biological replicates) and (B) increased proliferation (%Ki67⁺ cells) in CTL (black dots; A, Veh, n=4, Glu, n=2, PDGF-AA, n=2 replicates; B, n=4 biological replicates in all conditions) and NF1-null OPCs (blue dots; A, Veh, n=4, Glu, n=2, PDGF-AA, n=2 replicates; B, n=3 biological replicates in all conditions). C-D. Concomitant exposure to 50μM glutamate (Glu) and 20mg/mL PDGF-AA increases PA cell (C) ERK1/2 phosphorylation (activation; Vehicle, n=4 both cell lines; PDGF-AA, n=3 both cell lines; Glu, n=3 both cell lines; PDGF-AA+Glu, WUPA1, n=4, WUPA4, n=3 biological replicates) and (D) proliferation (%Ki67⁺ cells; n=3 biological replicates for both cell lines for all treatment groups) relative to either treatment alone. E-F. PDGFRα stimulation by 20mg/mL PDGF-AA increases PA cell (E) RAS activity (RAS-GTP; n=3 biological replicates for both cell lines and treatment groups) and (F) ERK1/2 phosphorylation (activation; WUPA1, Veh, n=4, all other treatments, n=3 biological replicates; WUPA4, n=3 biological replicates for all treatment groups), which are inhibited by glutamate receptor inhibition (2μM Memantine, Mem). 50μM glutamate increases (E) RAS activation and (F) ERK1/2 phosphorylation (activation), which are attenuated by PDGFRα inhibition (1μM Avapritinib, Ava). G. Quantification of Ki67⁺ and cleaved caspase-3⁺ PA cells demonstrate that PDGF-AA increases PA cell proliferation, which is inhibited by memantine (2μM) exposure. Apoptosis is not affected by any treatment (WUPA1, Veh, n=5, PDGF-AA, n=3, PDGF-AA+Mem, n=3; WUPA4, Veh, n=4, PDGF-AA, n=3, PDGF-AA+Mem, n=3 biological replicates). C-G. Sporadic PA, WUPA1, black dots; NF1-associated PA, WUPA4, blue dots. Data are shown as the mean ±SEM. A, B, E, G, One-way ANOVA with Dunnett’s post-test correction. C, D, F, One-way ANOVA with Tukey’s post-test correction. P values are indicated within each graph. ns, not significant. See also Figure S7.

To determine whether GluRs could increase PDGFRα-mediated RAS/ERK signaling, we performed several experiments. First, we showed that GluR inhibition (memantine) reduced PDGF-AA-induced PA cell RAS activity (51% reduction; Figure 7E), ERK1/2 phosphorylation (44% reduction; Figure 7F, Figure S7E), and proliferation in vitro (45% reduction in %Ki67⁺ cells; Figure 7G, Figure S7F) relative to PDGF-AA exposure. Second, PA cells with GRID2 or GRIK3 knockdown did not increase ERK1/2 activation to the same levels as control PA cells (Figure S7G) in response to PDGF-AA. Third, pharmacological PDGFRα inhibition (avapritinib) blocked glutamate-mediated RAS and ERK activation (64% and 46% reduction relative to glutamate exposure, respectively; Figures 7E–F, Figure S7E), PDGF-AA-mediated ERK activation (Figure S7H), PA cell proliferation in vitro (Figure 6H), and PA-PDX tumor growth in vivo (Figure 6I).

GluR- and PDGFRα-ERK signaling are coupled in a Src-dependent manner to drive glioma growth

As glutamate receptors can transactivate RTKs through Src kinase activity (Figure 8A)^50,51, we measured Src activity (Src ^Tyr416 phosphorylation) in PA cells, and found a 2.7-fold increase in Src activation following glutamate treatment (Figure 8B, Figure S8A). Glutamate-driven Src phosphorylation resulted in increased PDGFRα and ERK1/2 activation, as well as increased PA cell proliferation (1.7-fold increase in resorufin fluorescence), which was blocked by glutamate receptor (memantine) or Src (dasatinib) inhibitors (Figure 8C–F, Figure S8B). Consistent with glutamate receptor stimulation acting upstream of PDGFRα activation, ectopic expression or genetic silencing of PDGFRA had no effect on Src phosphorylation (Figure S8C). Conversely, GRID2 and GRIK3 ectopic expression induced Src activation, while their genetic silencing reduced Src phosphorylation (Figure S8D).

Figure 8. Src mediates glutamate-induced PDGFRα activation to increase low-grade glioma proliferation.

A. Glutamate receptor activation by glutamate can transactivate receptor tyrosine kinases, such as PDGFRα, through Src phosphorylation. The residues undergoing tyrosine (Tyr) phosphorylation are included. B. 50μM glutamate increases Src activation (Tyr⁴¹⁶ phosphorylation) in PA cells in vitro (Veh, n=3 biological replicates; Glu, n=4 biological replicates). C-F. Src inhibition with 1μM dasatinib (Das), reduces (C) PDGFRα (Tyr⁷⁵⁴ and Tyr⁸⁴⁹; n=3 biological replicates for both treatments), (D) Src (Tyr⁴¹⁶; n=3 biological replicates for both treatments), and (E) ERK1/2 (Tyr²⁰²/Tyr²⁰⁴; n=3 biological replicates for both treatments) phosphorylation, as well as (F) proliferation (resorufin fluorescence; n=4 biological replicates for all cell lines and conditions) in PA cells in vitro (Sp-PA cells: WUPA8, black dots; WUPA11, grey dots; NF1-PA cells: WUPA3, blue dots; WUPA4, green dots). Data are shown as the mean ±SEM. One-way ANOVA with Dunnett’s post-test correction. P values are indicated within each graph. ns, not significant. G. Representative immunohistochemistry images and quantitation of Ki67⁺ cells demonstrate that dasatinib treatment (20mg/kg/day) of 1mpi PA-PDX-bearing Rag1−/− mice in vivo reduces tumor proliferation at 2mpi (Sp-PA-PDXs, Veh, n=6 mice, Das, n=7 mice; NF1-PA-PDXs, Veh, n=5 mice, Das, n=7 mice). The dotted lines indicate the tumor boundary. Veh, vehicle; Glu, glutamate; Mem, memantine; Das, dasatinib. H. Illustration of proposed mechanism underlying GluR/RTK convergence-mediated PA growth regulation. In normal OPCs, neuronal PDGF-AA drives PDGFRα-regulated ERK-mediated proliferation, while neuron-derived glutamate increases OPC electrical activation, differentiation, and axonal myelination. In contrast, PA cells couple glutamatergic receptor and PDGFRα activation through Src-phosphorylation to support ERK-mediated glioma cell proliferation. See also Figure S8.

To extend our observations to high-grade gliomas (HGGs), we established a series of HGG cell lines from operative specimens at St Louis Children’s Hospital (Figure S8E). Similar to PAs, glutamate increased PDGFRα-ERK activation and HGG cell proliferation, which was attenuated by glutamate receptor (memantine), Src (Dasatinib), or PDGFRα (avapritinib) inhibition in vitro (Figure S8F–G). Importantly, pharmacologic Src inhibition (20mg/kg/day dasatinib) blocked PA-PDX proliferation (72–81% reduction in %Ki67⁺ cells, Figure 8G), as well as reduced ERK activation in vivo (Figure S8H). These data demonstrate that GluR activation is coupled to PDGFRα-ERK signaling through Src.

Collectively, these findings provide a mechanistic explanation for glutamate/GluR brain tumor growth regulation and reveal how tumor cells can usurp normal physiologic relationships between neurons and oligodendrocytes to create new clinically targetable glioma stromal dependencies.

Discussion

While several reports have described the effect of the glutamate on cancer biology^{2,8,18,21–23}, exactly how this excitatory neurotransmitter functions to increase mitogenic signaling within the cancer cell remains unresolved. Herein, we provide evidence for a previously unknown molecular mechanism linking neurotransmitter-driven neuron-cancer cell crosstalk through the non-canonical activation of a key oncogenic pathway (RAS/ERK). In this manner, normal neuron-OPC communication is hijacked in the setting of cancer by joining glutamate and PDGFRα signaling through the aberrant expression of GluR subunits and Src activation (Figure 8H). These findings additionally raise several points relevant to cancer neuroscience and the intersection between brain organogenesis and tumorigenesis.

First, both low-grade^35,36 and malignant gliomas^{29–35,52,53} are hypothesized to arise from OPCs. This cellular origin was also demonstrated using hiPSC-derived KIAA1549:BRAF-expressing (sporadic) and NF1-null OPCs, which generate low-grade gliomas in Rag1−/− mice in vivo⁴⁵. In the healthy brain, OPCs express AMPA, NMDA and kainate glutamate receptors and form bona fide glutamatergic synapses with neurons^38,40 to directly regulate their differentiation⁵⁴, myelination^55,56, and proliferation⁵⁴. Given this ontological relationship, it is logical that glioma cells should respond similarly to neuronal glutamate. However, unlike malignant gliomas, which generate action potentials following glutamate synaptic transmission, PA cells exhibited a mitogenic response to glutamate exposure with no change in membrane excitability. Similar mitogenic effects of deregulated glutamatergic signaling have been reported in malignant melanoma⁵⁷, osteosarcoma⁵⁸, colorectal^59,60, lung⁶¹, and breast cancer⁶². Investigation of this dissociation between tumor membrane excitability and cell proliferation in pediatric low-grade gliomas versus pediatric and adult high-grade gliomas, as well as those from other organs, will require comparative studies of tumors with different cellular origins and growth regulatory mechanisms.

Second, the finding that glutamate increases PA tumor growth in a GluR-dependent manner adds to a growing appreciation that trophic factors released by neurons act on RTKs in the normal brain and in the context of neoplasia. In this respect, neuronal activity-dependent release of the same neurotransmitters produced during development enhance tumor cell growth. For example, during normal development, neuregulin-1 released by neurons acts on the OPC ErbB3 receptor to control prefrontal cortex myelination and normal cognitive function in mice⁶³, while it enhances glioma cell survival, motility, and migration^64,65. Similarly, neuronal BDNF-TrkB signaling to OPCs underlies adaptive myelination, both in health⁶⁶ and in response to chemotherapy⁶⁷, while in the setting of cancer, neuronal BDNF promotes AMPA receptor trafficking to increase glutamate-induced currents in murine models of malignant glioma⁸. Additionally, neuron-secreted nerve growth factor (NGF) activation of TrkA increases OPC differentiation in response to stress during development⁶⁸ and mediates p75NTR proteolytic-dependent proliferation in malignant glioma tumor-initiating cells⁶⁹. Our finding that the GluR-mediated growth advantage in PAs does not operate through mitogenic pathways previously reported in normal OPCs, such as JNK1⁷⁰ or AKT^71,72 activation, but rather through MEK/ERK signaling as described in zebrafish⁷³ and mice⁷⁴, may reflect the MAPK dependency conferred by PA driver genomic alterations (KIAA1549; BRAF rearrangement, NF1 loss).

Third, our discovery that aberrant expression of glutamate receptor subunits (GRID2, GRIK3) in PA tumors creates a convergence of two physiologically important, but distinct, OPC dependencies (neuronal glutamate and PDGF-AA production) to control tumor growth suggests that similar mechanisms may operate in other cancers. While little is known about the specific functions of GRID2 and GRIK3 in healthy OPCs, their aberrant expression has been implicated in other solid tumors. As such, the GRIK3 kainate receptor has been shown to regulate WNT/β-catenin signaling and proliferation in breast and small cell lung cancer ^75,76. Overexpression of GRIK3 or other kainate receptor family members correlates with tumor growth and reduced patient survival in breast cancer (GRIK3⁷⁵), gastric cancer (GRIK3⁷⁷; GRIK5⁷⁸), colorectal cancer (GRIK3⁷⁹) glioblastoma (GRIK1⁸⁰), and glioma (GRIK2⁸¹, GRIK4–5⁸²). GRIK2 has been hypothesized to represent a targetable marker for bladder cancer⁸³ and low-grade glioma⁸¹. Ionotropic GRID2 expression is increased in endometrial carcinoma⁸⁴, prostate cancer⁸⁵ and GRID2 interacting protein has been suggested as a biomarker for colorectal cancer⁸⁶. As glutamate receptors comprise heterogeneous subunits, further investigation will be required to understand the basis for the observed differences between specific GluR agonist and antagonist effects on PA cell growth. Similarly, a more comprehensive analysis of the specific functions of GRID2 and GRIK3, as well as other kainate and ionotropic receptor family members, may reveal additional insights into the roles of glutamate or other neurotransmitter receptor subunit overexpression in cancer biology.

Fourth, our demonstration that glutamate/GluRs regulate PDGFRα function extends prior observations in which other G protein-coupled receptors (GPCRs) interact with RTKs to amplify mitogenic signaling. Prior studies have shown that GluR interacts with the epidermal growth factor receptor (EGFR) in lung cancer brain metastases⁸⁷ and in malignant glioma⁸⁸. In addition, D4 dopamine receptor activation increases PDGFRβ-ERK1/2 signaling independent of PDGF-BB ligand binding⁸⁹, while D2 receptor activation inhibits NMDA-mediated PDGFRβ signaling⁹⁰. Moreover, serotonin (5-HT) receptors can transactivate PDGFRβ to increase neuroblastoma cell line growth⁹¹ and GABA receptor-induced EGFR transactivation promotes prostate cancer cell migration⁹². Similarly, muscarinic acetylcholine receptors can activate EGFR, PDGFR ⁹³ and VEGFR2 RTKs in neuroblastoma cell lines^94,95, while adrenergic receptor stimulation by norepinephrine increases BDNF receptor phosphorylation and ERK/ PI3K signaling⁹⁶. Based on our findings, it will be exciting to determine whether aberrant neurotransmitter receptor coupling to receptor tyrosine kinase-mediated mitogenic signaling pathways also holds true for additional cancers growing inside and outside the brain.

Taken together, our findings that PA tumor cells create a new stromal dependency in which neuronal glutamate confers a growth advantage via activation of PDGFRα signaling combines two normal physiological inputs that regulate OPC physiology to favor PA tumor growth. Similar aberrant use of normal neurodevelopmental transcriptional and epigenetic circuits has been reported in malignant brain tumors, where peritumoral GBM cells resemble uncommitted OPCs to drive invasion and metastasis⁵³. While future studies will be required to determine whether additional receptor dependencies and convergencies exist in high-grade brain tumors or other cancers reliant on neurotransmitter trophic support, this hijacking of neurotransmitter receptor function suggests that selective targeting of these and other receptors using agents in clinical practice for treating neurological and psychiatric disorders might emerge as tractable therapeutic options for brain tumors (NCT05664464; NCT05775458). Ongoing screening efforts are already aimed leveraging neuroactive drugs that can identify potential cooperative therapeutic targets for high-grade gliomas⁹⁷. Particularly for tumors characterized by glutamatergic dysregulation, such as high- and low-grade gliomas, memantine, a NMDA receptor antagonist approved for dementia and Alzheimer’s disease, represents a compelling chemotherapeutic candidate. In addition, memantine can act directly on tumor cells by disrupting glutamate-mediated mitogenic pathways, as well as confer a neuroprotective effect by reducing neuronal excitotoxicity^98,99. Collectively, our findings in both low- and high-grade pediatric gliomas establish a previously unknown mechanism for linking G-protein coupled receptor to receptor tyrosine kinase signaling and provide a roadmap for studying other cancers similarly dependent on stromal neurotransmitter availability.

Resource Availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, David H. Gutmann, MD, PhD (gutmannd@wustl.edu).

Materials availability

The unique PA cell lines generated in this study will be made available on request, but we may require a payment and will require a completed Materials Transfer Agreement. Further information and requests should be directed to the lead contact, David H. Gutmann, MD, PhD (gutmannd@wustl.edu).

Data and code availability

The raw sequencing data utilized in this study has been deposited to Gene Expression Omnibus (GSE244433). Additional publicly available RNA sequencing datasets utilized in this study are deposited at GEO (GSE222850, GSE163071) and dbGAP (phs001854.v1.p1). This study did not generate any unique codes. Any additional information required to reanalyze the data reported in this work paper is available from the lead contact, David H. Gutmann, MD, PhD (gutmannd@wustl.edu), upon request.

STAR★Methods

Experimental model and study participant details

Mouse husbandry, injections, and treatments

Mice were maintained on a 12-hour light/ dark cycle in a barrier facility, at 21°C and 55% humidity with ad libitum access to food and water under an approved and active Animal Studies Committee at the Washington University School of Medicine (Institutional Animal Care and Use Committee). Rag1−/− (The Jackson Laboratory, B6.129S7-Rag1tm1Mom/J; JAX:002216) mice were intercrossed and neonatal pups (P0–5) intracranially injected as previously described⁴⁵. Briefly, anesthetized mice were injected with 5 × 10⁵ cells resuspended in 2 μL ice-cold PBS were injected 0.7 mm to the right of the midline 0.5 mm posterior to Lambda: 2 mm deep with a Hamilton syringe. Injected pups were allowed to recover and were subsequently immediately returned to their maternal cage until weaning. Mice of both sexes were randomly assigned to all experimental groups, and no sex-related differences were observed in any findings. Animals were aged to 1 month for tissue collection or underwent a 4-week treatment (5 days/week) with memantine hydrochloride, avapritinib, or dasatinib by oral gavage at a dose of 20 mg/kg body weight and were harvested at 2 months post-injection (mpi).

Human PA and HGG collection, dissociation, in vitro maintenance, and treatments

Human operative specimens and dissociated cell suspensions were acquired from the St Louis Children’s Hospital Tumor Bank under an active and approved Human Studies (IRB) protocol at the Washington University School of Medicine. Basic demographic information for these donors is reported in Figure 2A. For cell isolation and cell line generation, fresh operative specimens (pediatric pilocytic astrocytomas, PAs, or high-grade gliomas, HGG) were dissociated in 1U papain for 12hours, prior to red blood cell lysis, clearing and resuspension in ice-cold PBS for intracranial injections of neonatal mice, or in culture medium for cell line propagation (DMEM supplemented with 10% FBS, Pen/Strep and Glutamax). Cells were propagated in uncoated vented flasks and passaged with Accutase upon confluency. For all small molecule treatments, control medium did not contain Glutamax (glutamine). Established PA cells underwent treatment with agonists (0–100μM Glutamate, 10 μM NMDA, 10 μM AMPA, 10 μM Kainate and 20ng/mL PDGF-AA), glutamate receptor inhibitors (2 μM memantine, 1 μM perampanel, 1 μM NBQX,10 μM topiramate, 1 μM CBQX, and 10 μM felbamate), PDGFRα inhibitor (1 μM Avapritinib), Src inhibitor (1μM Dasatinib), MEK inhibitor (1μM PD0325901) or a pan-RAS inhibitor (0.5 μM IN-1).

Primary mouse neuron cultures

Primary neuron cultures were established from hippocampus and cerebellum dissected from postnatal day 5 (PN5) Rag1−/− mice. Tissues were dissected in Hibernate-A, dissociated in papain following manufacturer’s instructions (Papain dissociation kit, Worthington), and 0.5–1×10⁶ purified primary neurons were plated on a well of a poly-D-lysine/laminin pre-coated 6-well plate and were incubated at 37°C 5% CO₂. Neuron maintenance medium (Neurobasal medium supplemented with B27, 2mM L-Glutamine, Pen/Strep) was refreshed every 2 days. Neuronal cultures were established for 7 days in vitro (div). Conditioned medium was collected from 7div neurons and filtered through 0.45μm PVDF syringe filters prior to further use. For PA cell- co-cultures, PA cells were seeded directly onto established 7div primary cerebellar neurons and were incubated in conventional neuronal maintenance medium for 24h prior to further immunocytochemical analysis.

Method details

Colorimetric Glutamate Assay

Whole cortex and cerebellum derived from neonatal (PN5) and adult (PN30) Rag1−/− mice, as well as 7div primary cortical or cerebellar neurons derived from neonatal Rag1−/− mice were lysed in 0.1M HCl pH7.0 by sonication and underwent glutamate production quantification following manufacturer’s instructions.

scRNA sequencing and analysis

Five samples from pediatric PAs, each obtained in a pre-dissociated frozen single-cell suspension, were obtained from the St. Louis Children’s Hospital Tumor Repository, were counted for viability and 10,000 cells underwent 10xGenomics scRNA sequencing at Genome Technology Access Center (GTAC, Washington University). For all samples, raw sequencing data was processed using the 10X Genomics Cell Ranger pipeline (version 6.0.1) to generate gene count matrices and then aligned to the GRCh38 human reference genome. All scRNA sequencing data were analyzed using the Seurat (version 5.0) R package. For each sample, the cell quality was evaluated individually by assessing the number of reads, detected gene counts, and confirming that percentages of mitochondrial reads were fewer than 10%. Doublet removal was, then, performed using “doubletFinder” function in DoubletFinder (version 2.0.6) R package. Following the quality control steps, the “SCTransform” (version 2) function was used to normalize the data, followed by principal component analysis (PCA) and Seurat-CCA integration by “IntegrateLayers” to correct any batch effects between samples. The “FindNeighbors” and “FindClusters” functions were used for subsequent cell clustering and the ensuing groups were visualized by Uniform Manifold Approximation and Projection (UMAP), based on the first 30 PCs. Gene differential expression analyses were performed with the Seurat “FindMarkers’” function based on Wilcox test scores and the significant results were select with p_val_adj <0.05, and |log2FC| > 1. Gene ontology analysis was performed using DAVID Bioinformatics. Significant pathways were selected by FDR < 0.1. Raw data from these analyses are deposited to GEO (GSE244433). Additional publicly available databases utilized for validation in this study include GSE222850 and dbGap: phs001854.v1.p1.

FFPE Immunohistochemistry and Immunofluorescence

For paraffin-embedded tissue immunohistochemistry, previously prepared paraffin sections or Ringer’s and 4% PFA-transcardially perfused PA-PDX-bearing mice were used. Brain tissues were post-fixed in 10% buffered formalin prior to paraffin embedding. Paraffin-embedded tissues were serially sectioned (5μm) and immunostained with Ki67, GFAP, Olig2, phospho-ERK1/2, Synaptophysin, GRIK3, and GRID2 antibodies (Key Resources Table). Immunohistochemical staining was performed using the Vectastain ABC kit and appropriate biotinylated secondary antibodies. For immunofluorescence assays, sections were immunostained with synapsin I, or bassoon antibodies (Key Resources Table) and appropriate AlexaFluor 488 and AlexaFluor568 secondary antibodies. Hematoxylin and eosin (H&E) staining was performed following the manufacturer’s instructions (StatLab). Brightfield images were acquired using Image Studio Lite Version 5.2 software, and LAS AF Lite 3.2.0 software, while fluorescent images were acquired with a Leica DMi1 fluorescent microscope using the Leica LAS X software, or a Leica Thunder Imager 3D Assay system following the manufacturer’s instructions. All images were analyzed using ImageJ 1.53a software utilizing the cell counter plugin, as well as Adobe Photoshop version 21.1.1 software.

Key resources table.

REAGENT or	SOURCE	IDENTIFIER
Antibodies
AKT	Cell Signaling Technologies	Cat# 9272S
Alexa Fluor 488 anti-mouse	Invitrogen	Cat# A11029
Alexa Fluor 568 anti-rabbit	Invitrogen	Cat# A11011
Alpha Tubulin	Fisher	Cat# T9026
Bassoon	Abcam	Cat# ab110426
Biotinylated anti-mouse	Vector laboratories	Cat# BA-9200
Biotinylated anti-rabbit	Vector laboratories	Cat# BA-1000
Cleaved caspase-3	Cell Signaling Technologies	Cat# 9661S
ERK1/2	Cell Signaling Technologies	Cat# 9102S
FGFR1	Cell Signaling Technologies	Cat# 9740S
GAPDH	Abcam	Cat# ab8245
GRID2	Abcam	Cat# ab251953
GRIK3	Abcam	Cat# ab183035
GFAP	Thermo Fisher Scientific	Cat# GF28R
Ki67	BD Pharmingen	Cat# BDB556003
Ku80	Cell Signaling Technologies	Cat# 2180S
Mouse Secondary NIR	Bio-Techne Sales Corp.	Cat# 042-206
Olig2	GeneTex	Cat# GTX132732
PDGFRa	Cell Signaling Technologies	Cat# 3174S
Phospho-AKT Thr308	Cell Signaling Technologies	Cat# 9275S
Phospho-AKT Ser473	Cell Signaling Technologies	Cat# 4060S
Phospho-ERK1/2 (for Western Blotting)	Cell Signaling Technologies	Cat# 9101S
Phospho-ERK1/2 (for Immunohistochemistry)	Novus Biologicals	Cat# AF1018
Phospho-FGFR1 (Tyr653/654)	Cell Signaling Technologies	Cat# 52928S
Phospho-PDGFRa	Cell Signaling Technologies	Cat# 2992S
Phospho-S6	Cell Signaling Technologies	Cat# 2215S
Phospho-Src	Cell Signaling Technologies	Cat# 6943S
Phospho-TRKB (Tyr 516)	Cell Signaling Technologies	Cat# 4619S
Rabbit Secondary - HRP	Bio-Techne Sales Corp.	Cat# 042-206
S6	Cell Signaling Technologies	Cat# 2217S
Src	Cell Signaling Technologies	Cat# 2109S
Synaptophysin	Abcam	Cat# ab32127
Synapsin I	Abcam	Cat# ab64581
TRKB	Cell Signaling Technologies	Cat# 4603S
Biological samples
JHH-NF1-PA1 cells	Gift from Fausto J. Rodriguez46	N/A
Res-186 PA cells	Gift from Charles G. Eberhart47	N/A
WUPA1	St Louis Children’s Hospital Tumor Bank, this paper	N/A
WUPA10	St Louis Children’s Hospital Tumor Bank, this paper	N/A
WUPA11	St Louis Children’s Hospital Tumor Bank, this paper	N/A
WUPA12	St Louis Children’s Hospital Tumor Bank, this paper	N/A
WUPA3	St Louis Children’s Hospital Tumor Bank, this paper	N/A
WUPA4	St Louis Children’s Hospital Tumor Bank, this paper	N/A
WUPA6	St Louis Children’s Hospital Tumor Bank, this paper	N/A
WUPA8	St Louis Children’s Hospital Tumor Bank, this paper	N/A
WUHG1	St Louis Children’s Hospital Tumor Bank, this paper	N/A
WUHG2	St Louis Children’s Hospital Tumor Bank, this paper	N/A
WUHG3	St Louis Children’s Hospital Tumor Bank, this paper	N/A
Human CTL iPSCs	Washington University Genome Engineering and iPSC Center45	N/A
Human NF1−/− iPSCs	Washington University Genome Engineering and iPSC Center45	N/A
Chemicals, peptides, and recombinant proteins
10% buffered formalin	Fisher Scientific	Cat# SF100-4
Accutase	Sigma-Aldrich	Cat# A6964-500ML
AMPA	Tocris	Cat# 0254
Antigen Retrieval Buffer (100X Citrate Buffer, pH 6.0)	Abcam	Cat# ab93678
Antigen Retrieval Buffer (100X Tris-EDTA Buffer, pH 9.0)	Abcam	Cat# ab93684
Avapritinib	MedChem Express	Cat# HY-101561
B27 Supplement – serum-free	Gibco	Cat# 17504044
CaCl2	Sigma-Aldrich	Cat# C4901
CNQX	Tocris	Cat# 0190
Cs-methanesulfonate	Sigma-Aldrich	Cat# C1426
Dasatinib	Med-Chem Express	Cat# HY-10181-500mg
DMEM/F12	Gibco	Cat# 11-320-033
EGTA	Sigma-Aldrich	Cat# E4378
Felbamate	MedChem Express	Cat# HY-B0184
Fetal bovine serum	Thermo Scientific	Cat# A5256701
Fluo-8/AM	AAT Bioquest	Cat# 1345980-40-6
Fugene HD	Promega	Cat# E2311
Glutamax	Thermo Scientific	Cat# 35050061
Glycine	Sigma-Aldrich	Cat# G8790-100G
Goat serum	Thermo Scientific	Cat# 16210072
HEPES	Gibco	Cat# 15630080
Hibernate-A	Gibco	Cat# A1247501
Hoechst 33258, pentahydrate	Thermo Scientific	Cat# H3569
IN-1	MedChem Express	Cat# HY-101295
Kainic acid	MedChem Express	Cat# HY-N2309
K-gluconate	Sigma-Aldrich	Cat# P1847
L-Glutamate	Sigma-Aldrich	Cat# G1251-100G
L-Glutamine	ThermoFisher	Cat# 25030081
Memantine hydrochloride	Tocris	Cat# 0773
NaCl	Sigma-Aldrich	Cat# S5886
NBQX	MedChem Express	Cat# HY-15068
Neuro-background suppressor	ThermoFisher	Cat# F10489
NMDA	Tocris	Cat# 0114
Papain	Worthington Biochemicals	Cat# LS003126
Paraformaldehyde	Sigma-Aldrich	Cat# 158127
PD0325901	STEMCELL Technologies	Cat# 72184
PDGF-AA	Peprotech	Cat# 100-13A
Penicillin / Streptomycin	Sigma-Aldrich	Cat# 11074440001
Perampanel	Cayman Chemical Company	Cat# 23003
Phosphatase inhibitor cocktail	Cell Signaling Technologies	Cat# 5870S
Phosphate buffered saline	Gibco	Cat# 10010023
PowerLoad	ThermoFisher	Cat# P10020
Probenecid	ThermoFisher	Cat# P36400
Protease/phosphatase inhibitor cocktail	Cell Signaling Technologies	Cat# 5872S
Red blood cell lysis buffer	Sigma	Cat# R7757-100ML
Resazurin	R&D Systems	Cat# AR002
RIPA buffer	Abcam	Cat# ab156034
Topiramate	MedChem Express	Cat# HY-B0122
Triton-X	Millipore Sigma	Cat# X100-500ML
Critical commercial assays
96-Well Ras Activation ELISA Kit	ThermoFisher Scientific	Cat# STA-440
Apo-ONE Homogeneous Caspase-3/7 Assay	Promega	Cat# G7790
Applied Biosystems TaqMan Fast Advanced Master Mix	Applied Biosystems	Cat# 4444557
cDNA Reverse transcription kit	Applied Biosystems	Cat# 4368814
Cell Proliferation ELISA, BrdU (colorimetric)	Sigma-Aldrich	Cat# 11647229001
Cellular Senescence Plate Assay Kit - SPiDER-βGal	Dojindo	Cat# SG05-05
Glutamate Assay Kit, Colorimetric	LS Bio	Cat# LS-K264-100
Nucleospin RNA purification kit	Macherey-Nagel	Cat# 740955.50
Papain dissociation system	Worthington Biochemicals	Cat# LK003153
Senescence-associated galactosidase assay	Cell Signaling Technologies	Cat# 9860S
Vectastain ABC HRP Detection kit	Vector Laboratories	Cat# PK-4000
Deposited data
PA single-cell RNA sequencing data	41	GEO: GSE244433
PA single-cell RNA sequencing data	42	GEO: GSE222850
PA single-cell RNA sequencing data	Publicly available at dbGAP	phs001854.v1.p1
Raw PA bulk RNA sequencing data	10	GEO: GSE163071
Experimental models: Cell lines
HEK-293T cells (passage 10–20)	ATCC	Cat# CRL-1573
Experimental models: Organisms/strains
B6.129S7-Rag1tm1Mom/J mice	Jackson laboratories	Strain JAX:002216
Oligonucleotides
GAPDH Taqman Probe	Fisher Scientific	Cat# Hs02786624_g1
GRID2 Taqman Probe	Fisher Scientific	Cat# Hs00910015_m1
GRIK3 Taqman Probe	Fisher Scientific	Cat# Hs01012793_m1
PDGFRA Taqman Probe	Fisher Scientific	Cat# Hs00998018_m1
Recombinant DNA
pLV100-Puro-EF1A>hGRID2	Vector-Builder	Cat# VB250320-1316qzj
pLV100-Puro-CMV>hGRIK3	Vector-Builder	Cat# VB250314-1464rvk
pLV100-Puro-CMV>hPDGFR	Vector-Builder	Cat# VB250320-1318za
Lenti-GFP	OriGene Technologies	Cat# TR30021V
shGRID2	MISSION Sigma-Aldrich	Cat# TRCN0000063068
shGRID2	MISSION Sigma-Aldrich	Cat# TRCN0000063069
shGRID2	MISSION Sigma-Aldrich	Cat# T RCN0000063070
shGRIK3	MISSION Sigma-Aldrich	Cat# TRCN0000063283
shGRIK3	MISSION Sigma-Aldrich	Cat# TRCN0000063284
shGRIK3	MISSION Sigma-Aldrich	Cat# TRCN0000063285
shPDGFRA	MISSION Sigma-Aldrich	Cat# TRCN0000195132
shPDGFRA	MISSION Sigma-Aldrich	Cat# TRCN0000194855
Software and algorithms
Adobe Illustrator v21.1.1	Adobe Creative Cloud	RRID:SCR_010279
Adobe Photoshop v21.1.1	Adobe Creative Cloud	RRID:SCR_014199
AXION Biosystems integrated studio (AxIS) version 2.5.1	AXION Biosystems	RRID:SCR_016308
Bio-Render	www.BioRender.com	www.biorender.com
GraphPad Prism 5 (10.3.1)	GraphPad Software	RRID: SCR_002798
Image Studio Lite Version 5.2	Li-Cor	RRID: SCR_013715
ImageJ 1.53a	ImageJ	https://imagej.nih.gov/ij/
LAS AF Lite 3.2.0	Leica	https://leica-las-af-lite.software.informer.com/3.2/
Leica LAS X	Leica	RRID: SCR_013673
MPM6 v6.3	Bio-Rad Laboratories	https://www.bio-rad.com/en-us/product/microplate-manager-sofware-6?ID=613864bd-2fb8-4b3c-b88b-2f5ba823ba06
pClamp 10.4	Molecular Devices	https://www.moleculardevices.com/products/axon-patch-clamp-system/acquisition-and-analysis-software/pclamp-software-suite
Seurat (version 5.0) R package	Satija Lab	https://satijalab.org/seurat/articles/install_v5.html
SkanIt Microplate Reader Software	ThermoFisher	https://www.thermofisher.com/us/en/home/life-science/lab-equipment/microplate-instruments/plate-readers/software.html

Immunocytochemistry and BrdU Proliferation Assays

Immunocytochemistry was performed on PA cells using cleaved caspase-3, bassoon, Ki67, Ku80 and synapsin I primary antibodies (Key Resources Table). Briefly, adherent cells were washed, fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 in PBS, before blocking in 10% goat serum and incubation overnight at 4 °C in primary antibodies diluted in 2% goat serum at the manufacturer’s suggested concentrations. Appropriate secondary Alexa-Fluor-conjugated secondary antibodies were employed and bis-benzamide (Hoechst 33258) was used as a nuclear counterstain. Cells were imaged on a Leica DMi1 fluorescent microscope using the Leica LAS X software, following the manufacturer’s instructions, and were analyzed using ImageJ 1.53a software using the cell counter plugin, as well as Adobe Photoshop version 21.1.1 software. For BrdU-based proliferation assays (Key Resources Table), 10μL BrdU was added directly to each 100μL cell medium on live cells, 12 hours prior to fixation of the cells and anti-BrdU colorimetric detection, following manufacturer’s instructions. Data was collected on a Bio-Rad iMark microplate reader and analyzed using MPM6 v6.3 (Bio-Rad Laboratories) software.

Resazurin Proliferation Assays

Multi-well plate-based proliferation assays were performed on low- and high-grade glioma cells following various in vitro treatments. For these assays, 50,000 cells were seeded per well of a 96-well plate 24h prior to treatment initiation. Unless otherwise stated, cells were treated for a total of 24h. During the last 4 h of treatment, resazurin was added to each well, following manufacturer’s instructions. Metabolized resazurin-to-resorufin fluorescence was recorded with a Varioskan^™ LUX Multimode Microplate Reader (ThermoFisher) plate reader and data were analyzed using SkanIt (ThermoFisher) software.

RAS Activity Assays

RAS activity assays were performed on sonicated cell pellets following the manufacturer’s instructions. Each assay was performed using a minimum of three independently generated biological replicates. Data was collected on a Bio-Rad iMark microplate reader and analyzed using MPM6 v6.3 (Bio-Rad Laboratories) software.

Apoptosis and senescence assays

Apoptosis and senescence were measured in a 96-well plate using the Apo-ONE^® Homogeneous Caspase-3/7 Assay format, or SPIDER β-galactosidase assays, respectively, following the manufacturer’s instructions. Fluorescence was recorded with a Varioskan^™ LUX Multimode Microplate Reader (ThermoFisher) plate reader and analyzed using SkanIt software. Imaging-based senescence-associated β galactosidase assays were performed following manufacturer’s instructions (Cell Signaling Technologies). Brightfield images were captured with a Leica stereomicroscope and analyzed using ImageJ 1.53a software. The total number of cells per field of view was recorded. A minimum of 3 fields of view were captured per cell line and the total number of cells was averaged.

Multi-electrode arrays

150,000 PA cells from each of the cell lines assayed were plated on AXION Biosystems 48-well MEA plates and grown for 3 days. A minimum of four individual wells were analyzed per cell line. All cells were recorded for 3 min at a 5 standard deviation threshold level and 5000 Hz as a digital filter using AXION Biosystems integrated studio (AxIS) version 2.5.1 software. Spike firing rates were calculated from the total number of spikes/3 min and are represented as spikes/min, only accounting for active electrodes. Representative traces of spikes were extracted using the AXION Biosystems neural metric tool.

Calcium imaging

50,000 PA cells were plated onto uncoated 96-well plates for 1 day before being treated with Fluo-8/AM, PowerLoad, and Probenecid for 30 min at 37 °C and for another 30 min at room temperature. Cells were subsequently washed with HBSS and incubated for a minimum of 10 min in fresh culture medium supplemented with 5% neuro-background suppressor and for a subset of them with 50uM glutamate. Cells were imaged on a Nikon spinning disk upright epi-fluorescence confocal microscope equipped with a 10× dry objective, and a 488 nm wavelength laser was used for wide-field imaging. Cells were stimulated by a Ti LAPP DMD (Deformable Mirror Device) LED source for ultrafast photo-stimulation, with 0.1 mW applied during each recording for Fluo-8 excitation. Fluo-8 images were collected at 15 Hz (2048 × 2048 pixels, 1 × 1 mm) and the duration of each region of interest (ROI) was limited to 10 min. The fluorescence intensity and optical response to depolarizing membrane potential transients (ΔF/F) were calculated in Matlab programming environment to generate single-neuron activity traces. The ΔF/F threshold was set at 4 standard deviations beyond baseline fluorescence. A minimum of five wells were seeded per cell line assayed and at least three cells were recorded per well.

Whole-Cell Patch-Clamp Recording

Primary mouse hippocampal neurons and PA cells were recorded using standard whole-cell techniques at room temperature. Pipettes were made from borosilicate glass pipettes (World Precision Instruments, Inc) with resistance of 3–6 MΩ. The bath solution contained (in mM): 138 NaCl, 4 KCl, 2 CaCl₂, 1 MaCl₂, 10 glucose, and 10 HEPES; pH 7.25. For glutamate-induced current recordings, cells were patched at voltage-clamp mode at −70 mV and −30 mV (to relieve Mg²⁺ block of NMDA receptors) as indicated. All bath and perfusion solutions included 10 μM glycine as an NMDA receptor co-agonist. Patch pipettes were filled with solution containing (in mM): 130 Cs-methanesulfonate, 4 NaCl, 0.5 CaCl₂, 5 EGTA, and 10 HEPES; pH 7.25. Glutamate (50μM) was applied with a multi-barrel, gravity-driven local perfusion system. For action potential recordings, cells were recorded in current-clamp mode with pipettes filled with solution containing (in mM): 130 K-gluconate, 4 NaCl, 0.5 CaCl₂, 5 EGTA, and 10 HEPES; pH 7.30. Step currents from −50 pA with increments of 50 pA were injected to probe for action potentials and active conductances. Recordings were conducted using a MultiClamp 700B amplifier (Molecular Devices), Digidata 1550 16-bit A/D converter, and pClamp 10.4 software (Molecular Devices). Currents were recorded at 5 kHz and filtered at 2 kHz using an eight-pole Bessel filter.

Western blotting

Western blotting was performed on snap-frozen cells, lysed in RIPA buffer supplemented with a protease and phosphatase inhibitor cocktail and were blotted using appropriate primary and NIR-conjugated secondary antibodies (Key Resources Table). Images were captured and analyzed using the Li-Cor Image Studio Lite Version 5.2 software and are representative of more than three independently generated biological replicates. A subset of lysates underwent simple western blotting (Bio-Techne) following the manufacturer’s instructions. Briefly, snap-frozen cells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors, as above, 1.8μg of protein were loaded per capillary and incubated with appropriate primary antibodies diluted at 1:50, as well as secondary HRP- or NIR-conjugated antibodies (Key Resources Table). Compass for SimpleWestern (6.3.0) software directly generated all digital images used for this publication.

Lentiviral production and infection

Three independent pre-designed lentiviral short hairpins for GRID2 and GRIK3, two independent short hairpins for PDGFRA, and lentiviral vectors expressing human GRID2, GRIK3, and PDGFRA open reading frames were purified. These along with packaging vectors were transfected into HEK293T cells using Fugene HD following the manufacturer’s instructions. Viral supernatants were collected 48 h post-transfection, filtered, and used to directly infect PA cells for 24 h. GRID2, GRIK3, and PDGFRA knockdown and ectopic expression was confirmed by quantitative PCR and/or Western blotting. Infected cells were used for in vitro analyses or intracranial injections of Rag1^−/− neonatal mice.

RNA extraction and quantitative RT-PCR

Total RNA was extracted from snap-frozen PA cell pellets following the manufacturer’s instructions and reverse-transcribed using a high-capacity cDNA reverse transcription kit. Quantitative RT-PCR was performed using TaqMan gene expression assays (GRID2, GRIK3, PDGFRA; Key Resources Table) and a TaqMan Fast Advanced Master Mix according to the manufacturer’s instructions. All reactions were performed using the Bio-Rad CFX96 Real-Time PCR system equipped with Bio-Rad CFX Manager 3.1 software. Gene expression levels of technical replicates were estimated by ΔΔCt method using GAPDH (Key Resources Table) as a reference gene.

Quantification and statistical analyses

All statistical tests were performed using GraphPad Prism 5 software. Two-tailed Student’s t-tests, or one-way analysis of variance (ANOVA) with Dunnett’s or Tukey’s post-test correction using GraphPad Prism 5 software. Statistical significance was set at p < 0.05, and individual p values are indicated within each graphical figure. A minimum of 3 independently generated biological replicates was employed for each of the analyses. All experiments were designed in consultation with an experienced bioinformatician (OC).

Supplementary Material

2

Table S1. Pathway enrichment analyses of the tumor cluster in the GSE244433 dataset. Related to Figure 1.

3

Table S2. Pathway enrichment analyses of the tumor cluster in the GSE244433 dataset. Related to Figure 1.

4

Table S3. GRID2 and GRIK3 expression is increased in pediatric PAs. Related to Figure 5.

Highlights.

• scRNA-seq reveals enriched glutamatergic signaling in human pediatric gliomas

• Human pediatric low-grade gliomas exhibit a glutamate growth dependency

• Glutamate stimulation enhances tumor cell ERK signaling, not membrane excitability

• Glutamate receptor activation of PDGFRα is mediated by Src to drive glioma growth