Gliomas phenocopy an inborn error of metabolism to drive neuronal activity and tumor growth

SUMMARY

The metabolic hallmarks of high-grade glioma (HGG) are not fully understood. Human brain tissue metabolomics revealed that the creatine synthesis pathway intermediate guanidinoacetate (GAA) accumulated ~100-fold in HGGs relative to controls, which was caused by imbalanced activities of enzymes in this pathway. Glioma cells secreted GAA rather than using it to produce creatine, implicating an alternative function. GAA accumulates in GAMT Deficiency, an inborn error of metabolism, and elevates neuronal excitability. Neuronal excitability is also increased in glioma and drives tumor growth through neuron-glioma interactions. We hypothesized that glioma-generated GAA excites surrounding neurons. Indeed, GAA induced neuronal hyperactivity by activating GABA_A receptors and causing depolarizing currents in glioma-associated neurons with dysregulated chloride homeostasis. Depleting tumoral GAA decreased electrochemical activity, neuron-glioma interactions, and tumor aggressiveness. Our findings unveil a mechanism linking cancer metabolism with cancer neuroscience and leverage human genetics to nominate GAA synthesis as a target in gliomas.

In Brief

Glioma-derived guanidinoacetate depolarizes neighboring neurons through γ-aminobutyric acid (GABA) A receptors, and depleting this metabolite reduces both neuronal activity and glioma growth.

Graphical Abstract

INTRODUCTION

Gliomas are the most common central nervous system malignancy in adults. High-grade gliomas (HGG, Grade 4) carry a median life expectancy of only 1–2 years. There is an urgent need for insights into molecular mechanisms underlying aggressive glioma behavior and therapeutic strategies that disrupt these processes. Adult HGGs encompass two diagnostic categories defined by isocitrate dehydrogenase (IDH) mutational status¹: 1) Glioblastoma (GBM), IDH-wildtype, Grade 4, and 2) Astrocytoma, IDH-mutant, Grade 4. IDH mutant enzymes display neomorphic synthesis of the oncometabolite (R)-2-hydroxyglutarate [(R)-2HG], which reprograms the epigenome^2–9 to promote glioma development and maintenance^10–12. This metabolic alteration underpins the antitumor activity of the mutant IDH1/2 inhibitor vorasidenib, which was recently approved by multiple regulatory agencies for the treatment of Grade 2 IDH-mutant gliomas¹³. These advances provide proof-of-principle that altered metabolism can serve as both an oncogenic driver and a therapeutic target in glioma. However, knowledge of metabolic mechanisms that sustain HGGs across genetic and transcriptomic subtypes is limited.

Prior research has provided vital insights into the biochemical landscape of adult-type gliomas^14–17. HGGs display reduced creatine content^15,18,19 and harbor elevated levels of nucleotides and their precursors^14,20,21, tryptophan¹⁵, lysine catabolites²², lipid species^14,17, and redox metabolites¹⁴ including glutathione¹⁵. Nevertheless, comprehensive metabolomic analyses of HGG, lower-grade glioma (LGG, Grades 2 and 3), and non-malignant brain specimens have not kept pace with comparable genomics and transcriptomics analyses of these tissues^23–25. This limited momentum has impeded insights into metabolic alterations that are causally linked to aggressive glioma behavior.

Growth-promoting interactions of gliomas with the neural circuits they invade have emerged as a major axis of pathophysiology. The activity of neurons, especially glutamatergic neurons, drives the growth^26–31 and invasion^32,33 of multiple glioma subtypes through paracrine signaling factors²⁶ and bona fide neuron-to-cancer synapses^28,31. In turn, gliomas increase the excitability of neurons^34–37, contributing to glioma-associated seizures and further augmenting glioma growth-promoting, activity-regulated neuronal interactions. The role that metabolic mechanisms may play in this pathophysiology, however, is largely unexplored.

To identify biochemical patterns specific to the HGG metabolome and define their contributions to HGG maintenance and progression, we collected 91 primary human brain tissue samples and conducted multi-omics analyses. We discovered a metabolic phenotype present in HGG that spans genetic and transcriptomic disease subtypes and recapitulates the biology of a rare inborn error of creatine metabolism, GAMT Deficiency. This metabolic phenotype drove hyperexcitability of the HGG microenvironment and tumor growth, thereby revealing a tractable target for therapeutic intervention in patients with HGG.

RESULTS

Metabolomic and lipidomic profiling of human brain tissues

We prospectively collected 91 adult human brain tissue samples from tumor or seizure foci resections. Samples included LGGs (Grades 2 and 3 IDH-mutant astrocytomas and oligodendrogliomas), HGGs (Grade 4 IDH-mutant astrocytomas and IDH-WT GBMs), brain metastases, and non-malignant brain specimens (NMBS) (Figure S1A). Tissues were divided and processed for metabolomic, lipidomic, and transcriptomic analyses. We used defined gene expression programs²⁵ to classify glioma samples as Proneural, Classical, or Mesenchymal (Figure S1B). We integrated data for 563 polar metabolites and lipid species with matched clinical and molecular information and performed unsupervised clustering (Figure 1A and Table S1). Seven dominant clusters emerged, including subgroups enriched in Mesenchymal GBMs, recurrent gliomas, index LGGs, brain metastases, NMBS, and two mixed populations without defined correlates with molecular or clinical features. Supervised dimensionality reduction analysis demonstrated that HGGs, LGGs, brain metastases, and NMBS were distinguishable by biochemical profiles (Figure 1B). Moreover, glioma samples of different histological subtypes, grades, and transcriptional subtypes could be discriminated based on metabolite content (Figures S1C-E).

Figure 1. Metabolomic stratification of human brain tissues.

(A) Polar metabolites and lipids (n=563) in specimens from adult HGG (n=39), LGG (n=25), brain metastases (n=18), and non-malignant brain (Brain; n=9). (B) Partial least-squares discriminant analysis (PLS-DA) of tissue samples. (C-E) Metabolite set enrichment analyses of (C) HGGs, (D) LGGs, or (E) metastases relative to non-malignant brain specimens. See also Figure S1.

To validate datasets, we assessed expected metabolic differences among gliomas and NMBS. 2HG levels were markedly elevated in IDH-mutant gliomas relative to IDH-wildtype gliomas and NMBS (Figure S1F). L-DOPA was enriched in NMBS compared with HGGs (Figure S1G), consistent with prior work³⁸. Additionally, intermediates in lysine degradation²² (aminoadipate) and nucleotide metabolism^20,21 (orotate and deoxyguanosine) pathways were elevated while the ratio of N-acetylaspartate to creatine (an imaging biomarker of HGG^18,19) decreased in HGG versus NMBS (Figures S1H-K).

Pathway-level changes in metabolism of amino acids, including arginine, proline, glutamate, histidine, alanine, and aspartate, were observed in tumors independent of subtype (Figures 1C–1E). HGGs showed specific changes in the pentose-glucuronate interconversion pathway (Figure 1C), driven by enrichment of levels of hexose monophosphates and glucuronic acid and depletion of pentitols relative to NMBS. Consistent with prior studies^39–42, glutathione metabolism scored highly in LGG compared to NMBS (Figure 1D). Butanoate metabolism differed between metastases and NMBS (Figure 1E) (principally reflecting lower GABA levels in metastases), likely due to lower neuronal content of metastases relative to diffuse gliomas and NMBS.

Guanidinoacetate accumulates in HGG and is associated with a bottleneck in the creatine synthesis pathway

To identify metabolic hallmarks associated with aggressive glioma behavior independent of IDH mutational status, we rank-ordered metabolites according to degree of enrichment or depletion in HGG versus NMBS (Figure 2A). An intermediate in the creatine synthesis pathway, guanidinoacetate (GAA), accumulated ~100-fold in HGG versus NMBS (Figures 2B and 2C). In contrast, creatine and creatinine were decreased while phosphocreatine was unchanged (Figures 2D-F). Therefore, GAA upregulation in HGG did not appear to be associated with a global increase in creatine synthesis. GAA accumulation was specific because other metabolites containing guanidino functional groups, guanidinosuccinate and γ-guanidinobutyrate, were modestly enriched and unchanged, respectively, in HGGs (Figure S2A-B).

Figure 2. Identification of guanidinoacetate accumulation as a metabolic hallmark of high-grade glioma.

(A) Metabolite abundance in HGG vs non-malignant brain specimens. (B) Schema depicting the creatine synthesis pathway. (C-F) Peak intensities of (C) GAA (FDR=0.007), (D) creatine (FDR<0.001), (E) phosphocreatine (FDR=0.5), and (F) creatinine (FDR=0.01) in index samples of HGG (n=26), LGG (n=15), and non-malignant brain specimens (Brain, n=9). Welch’s ANOVA with Benjamini-Hochberg FDR correction. (G-H) ROC analysis of GAA content in index HGG tissues (n=26) compared to (G) non-malignant brain specimens (n=9) or (H) index LGG tissues (n=15). (I-J) Sankey diagrams depicting interactions between ¹⁵N-labeled tracers (top) and labeled metabolites (bottom) in (I) NHA or (J) TS516 cells (n=1 per line). (K) Differential labeling scores calculated from (I-J). (L-S) Labeling of (L-O) GAA or (P-S) creatine by indicated tracer (n=3 per cell line). (T-U) (T) Intracellular and (U) secreted GAA (n=3 per cell line; Welch’s t-tests). Data are means ± SEM. n.s.=not significant, n.d.=not detected, *p<0.05, **p<0.01, ***p<0.001. See also Figure S2.

We evaluated patterns of GAA accumulation across glioma subgroups. GAA, but not other creatine synthesis pathway intermediates, was increased in a subset of recurrent LGG samples relative to index LGG samples (Figures S2C-F). Stratifying HGGs by IDH mutational status and gene expression programs, we found that GAA selectively accumulated in Grade 4 IDH-mutant astrocytoma and all GBM subgroups (Figures S2G-J). In contrast, we found small increases in both GAA and creatine in lung squamous tumors⁴³ relative to benign adjacent lung tissues and did not observe GAA enrichment in kidney^44,45 or prostate cancers (Figures S2K-R). Thus, isolated GAA enrichment is a phenotype specific to glioma.

Receiver operating characteristic (ROC) analyses indicated that GAA content served as a robust tissue biomarker discriminating HGG from LGG or NMBS (Figures 2G, 2H, S2S, and S2T, Table S2). In contrast, GAA levels weakly distinguished LGG and NMBS (Figures S2U and S2V, Table S2). These data establish that GAA accumulation is a recurrent, specific metabolic hallmark of GBMs and likely of Grade 4 IDH-mutant astrocytomas.

Our findings suggested that the creatine synthesis pathway may be reprogrammed in HGG. To probe substrate-product interactions in this pathway, we performed global nitrogen metabolism profiling⁴² of non-transformed, immortalized human astrocytes (NHA) and TS516 GBM glioma stem-like cells (GSCs) (Figures 2I and 2J, Table S3). In the creatine synthesis pathway, the enzyme arginine:glycine amidinotransferase (AGAT) catalyzes the conversion of arginine and glycine to GAA and ornithine, after which guanidinoacetate N-methyltransferase (GAMT) methylates GAA in an S-adenosylmethionine (SAM)-dependent manner (Figures 2B and S2W). We ranked interactions between ¹⁵N-labeled tracers and metabolites based on differences between NHA and TS516 lines and found that ¹⁵N₄-arginine-labeled GAA was strongly enriched in GBM cells, but ¹⁵N₄-arginine-labeled creatine was not (Figure 2K).

We validated these findings using low-throughput tracing assays (Figure S2W). TS516 cells displayed strong GAA labeling by ¹⁵N₄-arginine, ¹⁵N-glycine, and ¹⁵N-serine (a precursor to glycine), but not ¹⁵N₃-creatine, while GAA was below the limit of detection in NHA cells (Figures 2L–2O and S2X-S2Y). Neither NHA nor TS516 cells used ¹⁵N₄-arginine, ¹⁵N-glycine, or ¹⁵N-serine tracers to produce creatine (Figures 2P–2R and S2Z-S2AA). Both cell types predominantly consumed extracellular ¹⁵N₃-creatine to sustain creatine pools (Figure 2S). The creatine pool was larger in TS516 cells than NHA cells, which may explain less intracellular creatine labeling by extracellular ¹⁵N₃-creatine in TS516 cells. TS516 cells, but not NHA cells, readily accumulated and secreted GAA (Figures 2T and 2U). These data suggest that HGG cells synthesize GAA but preferentially export this metabolite rather than using it to produce creatine.

Creatine synthesis pathway reprogramming and GAA secretion are recurrent features of HGG

To address the reproducibility of GAA accumulation in HGG, we evaluated two independent metabolomics datasets generated from human brain tissue specimens. The first included HGG, LGG, and NMBS (Figures 3A–3D) while the second comprised only HGG and LGG samples (Figures 3E–3G). Indeed, GAA, but not other creatine synthesis pathway metabolites, preferentially accumulated in HGG versus LGG or NMBS.

Figure 3. HGG cells display a bottleneck in the creatine synthesis pathway and GAA secretion.

(A-D) (A) GAA (p=0.0002), (B) creatine (p=0.0475), (C) phosphocreatine (p=0.202), and (D) creatinine (p=0.0342) levels in HGG (n=28), LGG (n=9), and non-malignant brain (n=6) samples. (E-G) (E) GAA, (F) creatine, and (G) creatinine in HGG (n=9) and LGG (n=27) samples. (H-I) (H) Intracellular or (I) secreted GAA in non-malignant cells (n=4), IDH-mutant GSCs (n=6), and IDH WT GSCs (n=7) (n=3 per cell line). (J-K) Labeling of (J) GAA and (K) creatine by ¹⁵N₄-arginine (n=3 per cell line). (L) GAA secretion of normal-appearing brain (Brain, n=7) or GBM (n=5) explants. (M-Q) Spatially resolved (M) GAA secretion in explants from a human GBM resection from (N) normal-appearing brain (Brain), (O) FLAIR-enhancing tumor-adjacent tissue (FLAIR), (P) tumor margin (Margin), and (Q) tumor core (Core). MRIs and H&E stains are shown. Scale bar=20 μm. Data are means (A-D and H-L) or medians (E-G) ± SEM, or individual data points (M). n.d.=not detected. n.s.= not significant, *p<0.05, **p<0.01, ***p<0.001 (Welch’s ANOVA in A-D, Mann-Whitney test in E-G, Kolmogorov-Smirnov test in L). See also Figure S3.

We next quantified intracellular GAA content, rates of GAA secretion, and ¹⁵N₄-arginine-dependent labeling of GAA and creatine in a panel of IDH-mutant and IDH-wildtype GSC lines, NHA lines derived from two donors, and non-transformed neural stem cell (NSC) lines (ENSA and NSC11). All non-malignant NHA and NSC lines displayed low or undetectable GAA cellular content, negligible GAA secretion, and minimal de novo creatine synthesis pathway activity (Figures 3H–3K). Most GSC lines displayed creatine synthesis pathway reprogramming, including GAA accumulation and secretion as well as robust labeling of GAA, but not creatine, by ¹⁵N₄-arginine. Differences in ¹⁵N₄-arginine metabolism were not explained by differences in tracer uptake (Figure S3A). Thus, GSCs display tumor-specific, cell-intrinsic GAA production and export.

Using TCGA and GTEx RNAseq datasets, we found AGAT was upregulated while GAMT and distal enzymes in the creatine synthesis pathway were downregulated in HGG tissues relative to NMBS (Figures S3B-C and Table S4). In cultured cell lines, transcript and protein levels of AGAT were upregulated in all GSC lines except those (MGG152 and BT260) that did not display GAA accumulation or secretion (Figure S3D-G). AGAT^high cells nearly exclusively expressed one transcript (ENST00000396659) (Figures S3H-S3I), which is associated with promoter and enhancer elements surrounding exon 1 and an intragenic enhancer located between exons 2 and 3 of the GATM gene (encoding the AGAT protein). ATAC-sequencing revealed AGAT^high GSCs displayed more open chromatin at these promoter and enhancer sequences (Figure S3J) and footprinting analysis demonstrated increased transcription factor occupancy at these loci. Therefore, overexpression of the AGAT enzyme in HGGs is linked to chromatin remodeling within the GATM gene and enhanced transcription factor recruitment to constituent cis-regulatory elements.

To assess GAA secretion from human tissues, we leveraged recent advances in organoid modeling^46–49 to generate tissue explants from non-malignant brain and GBM surgical samples (Table S5). GBM but not NMBS explants displayed high GAA secretion (Figure 3L). Next, we established explants from spatially defined tissues obtained from a GBM resection, including normal-appearing brain tissue, tumor-adjacent edematous tissue marked by Fluid-Attenuated Inversion Recovery (FLAIR) hyperintensity, and tissues from the tumor margin and tumor core (Figures 3M-Q). Explants with high tumor cell content from the margin and core secreted more GAA relative to FLAIR-positive and normal-appearing brain tissues with low tumor cell content. These findings indicate that HGG cells synthesize GAA and release it into tumor interstitial fluid, leading to 10- to 1000-fold increases in GAA content in the HGG microenvironment.

Imbalanced AGAT and GAMT expression causes GAA accumulation in glioma

We hypothesized that GAA accumulation in HGG is driven by an imbalance in the expression and activity of AGAT and GAMT enzymes. We evaluated AGAT and GAMT expression in human GSCs and non-malignant NSCs (Figures 4A-B) and found AGAT was broadly upregulated in GSCs while expression of GAMT did not differ between GSCs and NSCs. Using a published single-cell RNA sequencing dataset of human GBM, we found that tumor cells and oligodendrocytes displayed the highest levels of AGAT expression and AGAT:GAMT ratios (Figures S4A-C). Myeloid cells in the GBM microenvironment synthesize creatine and export it to neighboring tumor cells⁵⁰. While AGAT and GAMT transcripts were expressed by macrophage and microglia populations, we did not observe GAA secretion from cultured microglia (Figure S4D). Thus, unlike tumor cells, myeloid cells in the HGG microenvironment display canonical creatine synthesis pathway activity. Using matched proteomics and metabolomics data from human gliomas, we compared protein-level AGAT:GAMT expression with tumoral GAA content (Figures 4C). We observed a positive correlation between these features, driven by differences in AGAT (Figure S4E) rather than GAMT (Figure S4F) expression.

Figure 4. Imbalanced AGAT and GAMT expression is necessary and sufficient for GAA accumulation.

(A-B) (A) AGAT and (B) GAMT transcripts in GSCs (n=44) and non-malignant NSCs (n=9) from Mack et al., 2019. (C) Correlation of tumor GAA content with AGAT:GAMT protein expression in human gliomas (n=50). (D) Representative immunoblot of NHA cells expressing indicated cDNAs and sgRNAs with Cas9 (n=3). sgAAVS1, sgRNA targeting the AAVS1 safe harbor locus. (E) GAA levels in cells from (D) (n=9 for all lines except AGAT WT, sgGAMT #2 cells, for which n=8). (F-G) Labeling of (F) GAA and (G) creatine by ¹⁵N₄-arginine in cells from (D) (n=3). (H) Representative immunoblot of NHA cells expressing indicated cDNAs and sgRNAs with Cas9 (n=3). s.e., short exposure, l.e., long exposure. (I) GAA levels in cells from (H) (n=3). (J-K) Labeling of (J) GAA and (K) creatine by ¹⁵N₄-arginine in cells from (H) (n=3). Data are means ± SEM. n.s.=not significant, *p<0.05, **p<0.01, ***p<0.001 (Welch’s or unpaired t-tests). See also Figure S4.

To test whether increased AGAT activity without a commensurate increase in GAMT activity causes GAA accumulation, we transduced AGAT^low NHA cells with wildtype AGAT, a catalytically dead AGAT mutant (C407A⁵¹), or an empty vector control in the presence or absence of GAMT (Figure 4D). Overexpressing AGAT increased GAA levels (Figure 4E) and triggered strong labeling of GAA, but not creatine, by ¹⁵N₄-arginine (Figures 4F-G and S4G) in a catalytic activity-dependent manner. Increasing the ratio of AGAT to GAMT activity further by knocking out GAMT evoked higher GAA accumulation. Therefore, imbalanced expression and activity of AGAT and GAMT enzymes is sufficient to explain GAA enrichment in HGG.

To determine if imbalanced AGAT and GAMT activities are necessary to sustain high cellular GAA levels, we overexpressed wildtype GAMT, a catalytically dead GAMT mutant (E45S⁵²), or an empty vector control in NHA lines stably expressing AGAT wildtype or C407A mutant enzymes or an empty vector control (Figure 4H). Overexpressing wildtype, but not catalytically dead, GAMT in AGAT-overexpressing cells depleted GAA (Figure 4I) and increased creatine labeling by ¹⁵N₄-arginine (Figures 4J-K, and S4H). Our findings indicate that GAMT occupies a critical node in the creatine synthesis pathway, ultimately dictating whether AGAT activity leads to creatine production or GAA accumulation and secretion.

GAA induces excitatory GABA_A receptor-dependent signaling in the HGG microenvironment

GAA is known to accumulate broadly in tissues of patients with GAMT Deficiency, an inborn error of creatine metabolism caused by loss-of-function mutations in the GAMT gene that lead to intractable seizures⁵³ (Figure 5A). This clinical phenotype suggests that GAA elicits neuronal excitability in the brain. Patients with HGG, like those with GAMT Deficiency, also experience seizures, cognitive dysfunction, and neurological deficits. Glioma-infiltrated tissues are more electrically active than non-malignant brain tissues³¹, and microenvironmental neuronal activity in turn drives glioma growth through neuron-to-glioma synapses and activity-regulated signaling^{26–28,33,36,54,55}. Nevertheless, we do not have a comprehensive understanding of the molecular mechanisms driving electrical activity in the HGG microenvironment. Thus, we hypothesized that GAA secretion by HGG cells may represent a key mechanism of glioma-neuron crosstalk.

Figure 5. GAA activates GABA_A receptors and neuronal signaling in the glioma microenvironment.

(A) Schema of pathogenic mechanisms in GAMT Deficiency and HGG. (B-C) GAA levels in brains of (B) Gamt-deficient (n=10) and control (n=6) mice or (C) human non-malignant brain (n=7) and HGG (n=9) tissues. In (B), data are from Schmidt et al., 2004⁷⁵. (D-G) Spatially resolved GAA content in human GBM. (D) GAA content from n=4 cases, including samples from case #5852: (E) normal-appearing brain (Brain), (F) tumor margin, and (G) tumor core. MRIs, H&E stains, and immunohistochemistry for AGAT and GAMT are shown. Scale bars=50 μm. (H) GAA and γ-aminobutyric acid (GABA) structures. (I) Patch-clamp recording traces of neurons in brain slices from 5–7 week old mice treated with 100 μM GAA or creatine with or without 10 μM gabazine (GBZ). (J) Area under the curve (AUC) values of traces represented in (I): GAA (n=5), GAA + gabazine (n=6), creatine (n=5). (K-L) Normalized input resistance of neurons treated with GAA wash-on and untreated (control, n=3), or pre-treated with 10 μM of the NMDA receptor inhibitor CPP and 20 μM of the AMPA receptor inhibitor DNQX (E blockers, n=5), 100 μM of the GABA_A receptor inhibitor picrotoxin and 2 μM of the GABA_B receptor inhibitor CGP52432 (I blockers, n=5), all inhibitors (E+I blockers, n=8), or picrotoxin (n=8). In (L), data reflect post-wash-on normalized input resistance. (M-N) Schemas depicting maintenance of [Cl⁻]_i by NKCC1 and KCC2 in (M) the adult brain and (N) HGGs. (O-P) Multielectrode array (MEA) assay of neuronal activity in response to GAA in acute brain slices from adult mice. Normalized spike rates are shown prior to and after treatment with 20 μM GAA for (O) tumor-infiltrated right hemisphere (n=3 slices) or (P) contralateral brain (n=3 slices). (Q-T) Perforated patch clamp analysis of E_GAA. (Q-R) Representative traces of chloride flux in tumor-infiltrated and contralateral cortex. (S) E_GAA and (T) [Cl⁻]_i in neurons from tumor-adjacent or contralateral cortex upon treatment with GAA (n=4 neurons). Data are means ± SEM. n.s.=not significant, *p<0.05, **p<0.01, ***p<0.001 (Welch’s t-test in B-C, Friedman test in D, unpaired t-test for GAA vs. GAA+gabazine in J, Mann-Whitney test for GAA vs. creatine in J, unpaired t-test in L, paired t-test in O-P, Mann-Whitney test in S-T). See also Figure S5.

We compared GAA accumulation in brain tissues from mice deficient for Gamt and from patients with HGG (Figures 5B and 5C). Although control samples showed species-specific numerical differences in GAA content, we observed ~100-fold increases in GAA levels in brain tissues from both Gamt-knockout mice and HGG patients relative to controls. We also measured GAA concentrations in spatially defined tissues from GBM resections. GAA levels sharply rose with increasing proximity to the tumor core (Figures 5D, S5B-N). Immunohistochemistry analysis confirmed association with an isolated increase in AGAT expression (Figures 5E-G, and S5A).

We investigated mechanisms through which GAA may contribute to increased electrochemical signaling in the glioma microenvironment. GAA resembles the neurotransmitter γ-aminobutyric acid (GABA) (Figure 5H) and activates GABA_A receptor-dependent chloride fluxes^56–58, the directionality of which are regulated by intracellular chloride ion concentrations ([Cl⁻]_i). Whole-cell voltage-clamp recordings from CA1 pyramidal neurons in acute hippocampal slices from adult mice revealed that exogenous GAA (but not creatine) application induced a large outward current that was completely blocked by the GABA_A receptor inhibitor gabazine (GBZ) (Figures 5I–5J). GABA_A receptor-dependent currents were evoked in a dose-dependent fashion by both GAA and GABA, with EC₅₀ values of 133.7 μM and 446.1 μM, respectively (Figures S5O-S5R). These data suggest that GAA levels in human HGG tumors (Figures 5C-D) are sufficient to activate GABA_A receptors. However, additional work is required to understand how cell type-specific differences in GABA_A receptor subunit composition influence GAA binding. Current-clamp recordings from cortical pyramidal neurons in acute brain slices from juvenile mice revealed that GAA reduced neuron input resistance, consistent with ion channel opening (Figures 5K–5L). Blockade of GABAergic, but not glutamatergic, receptors fully prevented the decrease in input resistance, an effect recapitulated by inhibiting GABA_A receptors alone with the GABA_A-selective inhibitor picrotoxin. Although GAA has been reported to inhibit synaptic glutamate uptake⁵⁹, our data suggest that this effect makes a negligible or minor contribution to GAA-dependent neuromodulation and instead highlight a prominent role for GAA as a GABA_A receptor agonist^56–58.

Our findings presented an apparent paradox: GAA engaged GABA_A receptors and hyperpolarized mature neurons from naïve mice but also accumulates in the hyperexcitable microenvironment of HGGs. NKCC1 and KCC2 transporters import and export neuronal Cl⁻, respectively, thereby influencing whether GABA_A receptor agonism is depolarizing or hyperpolarizing. Healthy adult neurons express KCC2 and NKCC1 at levels that maintain relatively low [Cl⁻]_i, causing chloride influx and neuronal hyperpolarization upon GABA_A receptor activation (Figure 5M). In contrast, immature neurons display low KCC2 expression, high [Cl⁻]_i and are depolarized by GABA_A receptor agonism⁶⁰. GBM-infiltrated brain displays hyperexcitability partly due to KCC2 downregulation, dysregulation of [Cl⁻]_i homeostasis, and depolarizing GABA signaling^61,62, recapitulating immature neuronal phenotypes (Figure 5N). Multi-electrode array (MEA) analysis of cultured neonatal rat cortical neurons demonstrated acute, durable increases in activity following GAA treatment (Figure S5S-S5U). In patch clamp studies of pyramidal neurons in acute brain slices from juvenile mice, GAA wash-on caused rapid and reversible decreases in input resistance and action potential firing, consistent with an inhibitory effect of GABA_A receptor activation at this later developmental stage (Figure S5V-S5X). We asked if GAA-dependent GABA_A receptor agonism in HGGs is associated with excitatory electrochemical signaling in glioma-adjacent neurons. We established orthotopic TS516 xenografts in mice displaying tumor infiltration in one hemisphere but not the other. In acute cortical slices, high-density MEA analysis revealed that GAA wash-on led to an increase in neuronal firing in tumor-infiltrated brain tissue but not in the contralateral hemisphere (Figures 5O–5P). To determine whether tumor-adjacent neurons display increased GAA reversal potential that contributes to depolarizing GABAergic chloride transients, we conducted perforated patch voltage clamp recordings with local application of GAA (GAA “puffing”). The reversal potential of GAA-induced currents (E_GAA) and the calculated [Cl⁻]_i were significantly increased in tumor-adjacent versus contralateral cortical neurons (Figures 5Q–5T). These findings align with prior reports of depolarizing GABAergic currents evoked by GABA in the GBM microenvironment^61,62 and support a model in which GAA stimulates neuronal signaling through a similar mechanism.

To validate increased [Cl⁻]_i in peritumoral neurons, we imaged murine cortical neuron (mCN) monocultures and cocultures of mCN and TS516 cells using the MQAE dye, fluorescence of which negatively correlates with [Cl⁻]_i. Co-culture of mCNs with GBM cells decreased neuronal MQAE fluorescence relative to mCN cultured alone, indicating that tumor cells cause chloride accumulation in neighboring neurons (Figure S5Y-S5Z). In agreement with prior work^61,62, peritumoral neurons exhibited lower expression of KCC2 compared to neurons in the contralateral hemisphere (Figures S5AA-S5BB). Together, our findings indicate that GAA accumulation cooperates with tumor-specific neuronal [Cl⁻]_i dysregulation to promote GABA_A receptor-dependent electrochemical signaling in gliomas.

AGAT inhibition represses electrochemical signaling and glioma growth

To determine whether inhibiting AGAT-dependent GAA synthesis and consequent neuronal activation attenuates GBM growth, we established isogenic AGAT-expressing (AGAT WT) and AGAT-knockout (AGAT KO) TS516 and UTSW63 GBM cell cultures (Figures 6A and S6C). AGAT knockout inhibited intracellular GAA accumulation and GAA secretion (Figures 6B-C and S6D-E). AGAT inhibition did not affect the growth of GBM cell monocultures in vitro (Figures 6D and S6F), but prolonged survival of mice bearing orthotopic GBM xenografts in both models (Figures 6E, and S6G), underscoring the central role of the brain microenvironment in dictating the oncogenicity of GAA accumulation. AGAT KO tumors displayed reduced GAA but not creatine content (Figures 6F-G, S6A-B and S6H-I), consistent with minimal de novo creatine synthesis in HGG cells (Figure 3K).

Figure 6. GAA accumulation drives neuronal hyperexcitability and glioma aggressiveness.

(A) Representative immunoblots in AGAT WT and AGAT KO TS516 GSCs. (B-C) (B) Intracellular and (C) secreted GAA from cells in (A) (n=3 per cell line). (D) In vitro growth of cells from (A) analyzed by nonlinear regression (n=3). (E) Overall survival of mice bearing orthotopic xenografts from cells in (A) (n=10 mice per arm). (F-G) Absolute (F) GAA and (G) creatine in tumors from (E) (n=7 AGAT WT, n=8 AGAT KO). (H-I) Representative H&E and cFOS immunofluorescence for (H) AGAT WT and (I) AGAT KO TS516 tumors. (J) cFOS⁺ nuclei in peritumoral cortex in AGAT WT and AGAT KO TS516 tumors (n=6 each). (K) Immunofluorescent imaging of vGLUT1 in AGAT WT and AGAT KO TS516 tumors. (L) vGLUT1 foci in tumor beds of AGAT WT and AGAT KO TS516 tumors (n=7 each). (M) Schema of optogenetics study. (N) Representative immunofluorescent imaging for eYFP⁺ neurons and HNA⁺ tumor cells in xenograft-bearing Thy1-ChR2-YFP mice. Scale bar=150 μm. (O-P) Representative imaging of EdU incorporation by HNA⁺ (O) TS516 AGAT WT and (P) AGAT KO xenografts in Thy1-ChR2⁻ and Thy1-ChR2⁺ mice 24 hours after optogenetic stimulation. Scale bar=50 μm. (Q-R) EdU incorporation by (Q) HNA⁺ TS516 AGAT WT (n=4 Thy1-Chr2⁻ mice, n=6 Thy1-Chr2⁺ mice) and (R) AGAT KO (n=4 Thy1-Chr2⁻ mice, n=7 Thy1-Chr2⁺ mice) xenografts. Data are means ± SEM. n.s.=not significant, *p<0.05, **p<0.01 (Welch’s t-test in B, C, J, Q, R; nonlinear regression in D; log-rank test in E; unpaired t-test in F, G, L). See also Figure S6.

We next evaluated whether AGAT knockout attenuates neuronal hyperactivity in the HGG microenvironment. We quantified nuclei expressing the electrochemical signaling marker cFOS in cortical sections of tumor-infiltrated brain tissues harvested from mice bearing AGAT WT or AGAT KO GBM xenografts. cFOS expression was nearly exclusively observed in neurons (Figures S6N-O) and declined in brain tissue adjacent to AGAT KO tumors from both models (Figures 6H-J and S6J-K). Decreased cFOS levels were associated with reduced tumor cell proliferation and interaction with vGLUT1+ neurons (a marker of excitatory neurons) in the GBM microenvironment (Figures 6K-L, S6L-M and S6P-Q). To ask if heightened tumor-neuron interaction in GAA-enriched xenografts drives tumor cell proliferation, we established AGAT WT and AGAT KO TS516 GBM xenografts in mice harboring or lacking expression of Channelrhodopsin-2 (ChR2) in Thy1+ neurons and performed optogenetic stimulation and EdU labeling 14–15 days after tumor cell implantation (Figures 6M-N). Acute optogenetic stimulation of glutamatergic neurons in Thy1::ChR2 mice increased tumor cell proliferation in AGAT WT, but not AGAT KO, tumors (Figures 6O-R). We did not observe an apparent increase in proliferation in AGAT WT versus AGAT KO cells in the unstimulated condition, which is likely explained by lower GAA accumulation at this early time point compared to later time points (Figure 6F). Collectively, these data establish causal links between GAA accumulation, tumor-neuron interactions, and aggressive glioma growth.

GAA accumulation drives HGG growth in a neuron-dependent manner

To ask if neurons are required for GAA signaling to stimulate tumor cell proliferation in HGGs, we established co-cultures of mCNs and TS516 cells and treated them with or without GAA. GAA treatment increased the proliferative index of TS516 cells cultured with mCN by approximately 50% relative to untreated co-cultures (Figures 7A-B). GAA treatment had no effect on TS516 cells grown as monocultures (Figure 7C) or with HMC3 microglia (Figure S7A). In these assays, the effects of endogenous GAA synthesis and secretion by tumor cells were negligible because secreted GAA fell below the limit of detection due to low tumor cell densities (Figure S7B). We also evaluated the effects of GAA on glioma cell migration by establishing cerebral organoids and co-culturing them with GBM neurospheres in the presence or absence of GAA. Treatment of GBM neurosphere-organoid co-cultures with GAA increased tumor cell migration (Figure S7C-D). This effect was specific to the cerebral microenvironment because GAA did not increase migration or invasion in GBM cell monocultures (Figure S7E-H).

Figure 7. GAA accumulation drives HGG growth in a neuron-dependent manner.

(A) Representative immunofluorescence of HNA, MAP2, and Ki67 in tumor-neuron co-cultures in the presence or absence of 5 μM GAA. (B) Ki67⁺HNA⁺ cells as a fraction of total HNA⁺ cells in cells treated as in (A) for 72 hours (n=4 co-cultures). (C) EdU incorporation by TS516 monocultures grown in the presence or absence of 5 μM GAA for 72 hours (n=4). (D) Growth of TS516 AGAT WT and AGAT KO subcutaneous (SubQ) xenografts (n=10 mice per arm). (E-F) (E) GAA and (F) creatine in TS516 AGAT WT and AGAT KO subcutaneous xenografts (n=10). (G) Schema of DREADD expression study. (H) Survival of AAV9-hSyn-mCherry-injected mice bearing TS516 AGAT WT or AGAT KO xenografts (n=8 mice per arm) and AAV9-hSyn-HM3Dq-mCherry-injected mice bearing TS516 AGAT KO xenografts (n=10 mice). (I) Representative immunofluorescent stains for KU80, Ki67, and mCherry in mice bearing TS516 AGAT KO xenografts from (H). (J) Tumor cell proliferation at the tumor margin in the indicated mice (n=3 mice). (K) GAA labeling by ¹⁵N₄-arginine with or without 150 μM ornithine (n=3). (L) Survival of TS516 xenograft-bearing mice receiving a standard (n=10 mice) or GAMT Deficiency diet (n=10 mice). (M-N) (M) GAA and (N) creatine in standard diet- or GAMT Deficiency diet-treated tumors at 30 days (n=9). (O) Representative immunofluorescence imaging for KU80, Ki67, and NeuN at the margin of TS516 xenografts from mice treated with standard or GAMT Deficiency diet. NeuN staining was performed on an adjacent tissue section. (P) Ki67⁺KU80⁺ cells as a fraction of total KU80⁺ cells in tumors from (O) (n=7 mice). Data are means ± SEM; n.s.=not significant, *p<0.05, **p<0.01, ***p<0.001 (unpaired t-test in B-F, N, P; log-rank test in H, L; Welch’s t-test in J, M.). See also Figure S7.

To ask if neuronal activity in the brain microenvironment is required for GAA signaling to promote glioma aggressiveness, we grew AGAT WT and AGAT KO TS516 cells as subcutaneous xenografts, which are virtually devoid of neurons (Figure S7I). In contrast to tumor promotion by AGAT activity in orthotopic xenografts (Figures 6E and S6G), AGAT activity minimally affected subcutaneous GBM growth (Figure 7D). AGAT WT subcutaneous xenografts displayed higher levels of both GAA and creatine compared to AGAT KO tumors (Figures 7E-F). We hypothesized that neuronal activity accounts for differences in AGAT WT and AGAT KO GBM xenograft growth in the brain and flank. We sought to determine whether chronic stimulation of neurons³² in AGAT KO orthotopic GBM xenografts rescues aggressive tumor behavior by expressing the excitatory hM3Dq designer receptor exclusively activated by designer drugs (DREADD) in neurons, which is activated by the agonist deschloroclozapine (DCZ)⁶³. AGAT WT and KO orthotopic GBM xenografts expressing mCherry alone served as controls (Figure 7G). Among DCZ-treated control mice, mice bearing AGAT KO GBM xenografts survived longer than those bearing AGAT WT tumors (Figure 7H), as expected (Figure 6E). Chronic, GAA-independent neuronal excitation in DCZ-treated mice expressing the hM3Dq DREADD in cerebral neurons restored aggressiveness of AGAT KO tumors and abolished the survival benefit conferred by GAA depletion (Figure 7H). These effects were associated with enhanced proliferation of AGAT KO tumor cells neighboring DREADD-expressing neurons relative to those neighboring mCherry-expressing neurons (Figures 7I-J). Thus, neuronal stimulation is sufficient to rescue decreased tumor aggressiveness caused by loss of GAA signaling in the GBM microenvironment.

GAA did not alter immune cell frequencies or functional markers in any lymphoid or myeloid population isolated from syngeneic CT-2A glioma allograft mice, except for a modest but statistically significant decrease in granzyme B expression in CD8⁺ T cells (Figure S7J-S7O). We observed countervailing effects of GAA on cultured myeloid cells, including minor but statistically significant decreases in both inflammatory and suppressive markers, particularly in bone marrow-derived macrophages (BMDMs) (Figure S7P). In primary T cells, GAA caused small, hypoxia-specific increases in proliferation of stimulated CD4+ and CD8+ cells and did not affect unstimulated cells (Figure S7Q). Single-cell RNA sequencing of AGAT WT and AGAT KO TS516 GBM xenografts identified diverse murine stromal cell and human tumor cell populations (Figures S7R-S). Tumor cell frequencies increased in AGAT WT versus AGAT KO tumors, consistent with tumor cell proliferation driven by GAA signaling (Figure S7T). A transcriptional signature of GABAergic signaling was enriched in peritumoral neurons from AGAT WT tumors compared to AGAT KO controls (Figure S7U). Astrocyte reactivity increased in AGAT WT tumors (possibly as a consequence of neuronal hyperexcitability⁶⁴), while a myelination gene program in oligodendrocytes was unaffected by AGAT status (Figures S7V-W). These findings support a model in which GAA signaling in the GBM microenvironment principally affects neurons through GABA_A receptor activation.

To translate these mechanistic insights, we tested a dietary therapy featuring arginine restriction and ornithine supplementation that is used clinically to treat GAMT Deficiency (hereafter the “GAMT Deficiency diet”). This diet attenuates AGAT activity, depletes GAA, and reduces neuronal hyperactivity and seizure incidence in patients^65,66. Ornithine supplementation reduced AGAT activity in cultured TS516 cells (Figure 7K). Mice administered the GAMT Deficiency diet did not display weight loss or reduced chow consumption relative to a standard diet (Figures S7X-S7Y). Placing mice bearing TS516 orthotopic xenografts on the GAMT Deficiency diet prolonged their survival relative to mice given a standard diet (Figure 7L) and reduced tumoral GAA without affecting creatine (Figures 7M-7N). The GAMT Deficiency diet recapitulated the survival benefit conferred by genetic AGAT knockout (Figure 6E) and diminished tumor cell proliferation near cortical neurons (Figures 7O-7P). This dietary therapy may constitute a clinically validated approach to block aberrant AGAT activity and oncogenic GAA signaling in patients with HGG.

DISCUSSION

We generated a deep metabolite profiling dataset of human brain tissues to discover GAA accumulation as a recurrent metabolic feature of HGG that manifests independently of canonical genetic and transcriptomic disease subtypes. GAA accumulation was caused by a bottleneck in the de novo creatine synthesis pathway in GSCs that functionally disconnects GAA production from creatine synthesis. Increased GAA content served as a surrogate marker of tumor grade, suggesting that GAA may have utility as a glioma tissue biomarker in diagnostic and therapeutic applications^67–71.

The creatine synthesis pathway is regulated in a cell type-specific manner in HGGs. Tumor-associated macrophages and microglia supply creatine to GBM cells to promote tumor tolerance to microenvironmental stressors, including hypoxia and acidification⁵⁰. This interaction may allow HGG cells to repurpose the creatine synthesis pathway to produce and secrete GAA into interstitial fluid⁷². GAA accumulation in HGG is caused by AGAT upregulation in tumor cells in the absence of a commensurate increase in GAMT. AGAT upregulation was associated with chromatin remodeling at key cis-regulatory elements in the GATM gene. In human tumors, the ratio of AGAT:GAMT protein expression positively correlated with tissue GAA content and tumor grade.

Human genetics and pathophysiology underscore the clinical relevance of GAA accumulation. Patients with GAMT Deficiency develop neurological symptoms that mirror the clinical phenotype commonly seen in glioma. GAA has been shown to regulate neuronal activity by serving as an agonist for GABA_A receptors^56,58 and repressing Na⁺/K⁺ ATPase activity⁷³ and synaptic glutamate uptake⁵⁹. The standard therapy for GAMT Deficiency involves restricting arginine (an AGAT substrate) and supplementing creatine and ornithine (an end-product AGAT inhibitor). This treatment causes GAA depletion, electroencephalogram normalization, and decreased seizure frequency in GAMT Deficiency patients^65,66. These results are not achieved by creatine supplementation alone. Intractable seizures are uncommon in other inborn errors of creatine metabolism that do not feature GAA accumulation⁵³, reinforcing the link between GAA signaling and neuronal hyperactivity. Genetic or dietary inhibition of GAA synthesis reduced activity of glioma-associated neurons and constrained HGG growth in vivo. Restoring neuronal activity in AGAT KO tumors via excitatory DREADD expression was sufficient to rescue GBM aggressiveness, supporting a neuron-dependent mechanism of tumor promotion by GAA signaling. Inspired by clinical management of GAMT Deficiency, we modeled an arginine-restricted, ornithine-supplemented dietary therapy and found that it was well tolerated, decreased tumoral GAA, and reproduced effects of genetic AGAT inhibition on survival of glioma-bearing mice. Thus, this dietary therapy used in the clinical management of GAMT Deficiency may display therapeutic activity in patients with HGG.

Advances in cancer neuroscience demonstrate that gliomas elevate neuronal activity of glioma-infiltrated circuits to further augment neuron-glioma interactions and accelerate tumor growth^{26–28,31,33,54,55}. Concordantly, tumor-infiltrated brain tissues are more electrically active than healthy tissue³¹. Release of glutamate^28,34 and synaptogenic factors^36,37,54 are implicated in this process, but the full complement of molecules involved in glioma-to-neuron communication is still emerging. Our data implicate GAA accumulation as a mechanism of neuronal activation in HGG. GAA resembles GABA, and our electrophysiology data confirm previous studies showing that GAA is a GABA_A receptor agonist^56,58. Recent work has shown that GABAergic signaling promotes progression of diffuse midline gliomas⁷⁴ and that [Cl⁻]_i dysregulation converts GABAergic signaling from inhibitory to excitatory within the GBM microenvironment^61,62. Consistent with these studies, GAA selectively stimulated neuronal firing in tumor-infiltrated brain tissues and elevated electrochemical signaling markers in xenograft HGG tissues. Together, our results identify GAA accumulation as a metabolic hallmark of HGG, establish the mechanism of GAA production in GSCs, connect GAA accumulation to neuronal hyperexcitability, and nominate AGAT inhibition as a therapeutic strategy to disrupt neuronal activity-dependent glioma growth.

Limitations of the Study

Neuronal GAA reversal potentials in GAMT Deficiency are uncharacterized, underscoring the need for further electrophysiological studies in this disease context. Chloride fluxes are determined both by intracellular and extracellular [Cl⁻], although changes in the latter remain unclear in HGG. We demonstrate that tumor-adjacent neurons exhibit dysregulated Cl⁻ homeostasis, mediated at least partly by reduced KCC2 expression. Additional studies assessing contribution of Cl⁻ transporters to peritumoral neuronal [Cl⁻]_i dysregulation are warranted. While tumor cells readily secrete GAA, the mechanism of GAA efflux in GSCs is undefined. Our data indicate that GAA enrichment occurs broadly in adult HGGs; however, the current study is underpowered with respect to Grade 4 IDH-mutant astrocytoma due to sample size. Future studies should evaluate the prevalence of GAA accumulation and signaling across the spectrum of glioma subtypes, particularly adult progressive LGGs and pediatric gliomas.

STAR METHODS

RESOURCE AVAILABILITY

Lead Contact

• Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Samuel K. McBrayer (Samuel.McBrayer@utsouthwestern.edu).

Materials Availability

• Plasmids generated in this study are available from Addgene.

Data and Code Availability

• Metabolomics data have been deposited by our group or others at National Metabolomics Data Repository (NMDR) Metabolomics Workbench (accession numbers ST005093, ST005089, ST000784, ST003143, ST005087). Metabolomics data deposited by our group are publicly available as of the date of publication. Kidney cancer metabolomics data from Hakimi et al. were accessed via the National Institutes of Health Manuscript System (NIHMS) (accession number NIHMS746037). RNA and ATAC sequencing data have been deposited by our group or others at National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) (accession numbers GSE265788, GSE312748, GSE343207, GSE343542, GSE119834, GSE274546). RNA and ATAC sequencing data deposited by our group are publicly available as of the date of publication. The Cancer Genome Atlas (TCGA) GBM and Genotype-Tissue Expression (GTEx) brain RNA sequencing data were accessed via the TCGA portal (https://portal.gdc.cancer.gov/projects/TCGA-GBM) and the GTEx portal (file version: 2017-06-05_v8_brain_cortex), respectively.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the Lead Contact upon request.

Experimental model and study participant details

Human Subjects

The study was conducted according to the principles of the Declaration of Helsinki. Patient tissue and blood were collected following ethical and technical guidelines on the use of human samples for biomedical research after informed patient consent under an Institutional Review Board (IRB)-approved protocol. All patient samples were de-identified before processing. All patient samples and organoid explants were diagnosed and graded according to the 2021 WHO Classification of Tumors of the Central Nervous System (CNS), 5th edition¹. Patient samples for the Discovery Cohort were collected under a protocol approved by the IRB of the University of Texas Southwestern Medical Center (STU 022011-70). Patient samples for Validation Cohort #1 were collected under a protocol approved by the IRB of Northwestern University (17-048-CDH). Patient samples for the Validation Cohort #2 are part of a collection of specimens associated with a study supported by the NCI CPTAC group and are described in a manuscript in preparation. Patient samples used to generate organoid explants were collected under a protocol approved by the IRB of the University of Pittsburgh (STUDY19080321). Information related to primary samples can be found in Tables S1 and S5.

Cell lines

NHA Donor #1 cells (non-malignant human astrocytes, immortalized with human papillomavirus E6, E7, and hTERT; sex unknown)⁷⁸ were obtained from R. Pieper at the University of California in San Francisco. NHA Donor #2 cells were generated from commercially available astrocytes (Lonza CC-3187) as described previously¹¹. HEK293T cells (female, ATCC CRL-3216) and 293AD cells (female, CellBioLabs AD-100) were obtained commercially. BT054⁷⁹ (female) were provided by S. Weiss at the University of Calgary. BT260⁵ cells (sex unknown) were provided by K. Ligon at Dana Farber Cancer Institute. HK157 (female), HK211 (female), HK213 (male), HK252 (male), and HK308 (female) cells were provided by H. Kornblum at the University of California in Los Angeles⁸⁰. TS516 (sex unknown, RRID: CVCL_A5HY) and TS603 (sex unknown, RRID: CVCL_A5HW) were provided by I. Mellinghoff at Memorial Sloan-Kettering Cancer Center⁸¹. ENSA (sex unknown) and NSC11 (sex unknown) cells were provided by J. Rich at University of North Carolina-Chapel Hill⁸³. MGG152 (male) cells were provided by D. Cahill at MGH⁸⁴. UTSW63 (male) and UTSW71 (male) were generated as previously described⁸². HMC3 cells (Sex unknown, ATCC CRL-3304, RRID: CVCL_II76) were acquired commercially.

UTSW5 GBM cells (female) were generated as described previously⁸². Briefly, fresh tumor tissue from the operating room was suspended in ice cold Hibernate A (BrainBits HA) with 100 U/mL and 100 μg/mL, respectively, of penicillin/streptomycin (Thermo Fisher 15–140-148) for transfer on ice to the laboratory, then transferred to RBC lysis buffer (ThermoFisher 00433357) for 10 minute incubation at room temperature before washing in Hibernate A supplemented with 2mM Glutamax (ThermoFisher 35050061), penicillin/streptomycin (100 U/mL and 100 μg/mL, respectively; Thermo Fisher 15–140-148), and Amphotericin B (0.25 μg/mL, Gemini Bio-Products 400104). Following washing, tissue was dissociated using the Brain Tumor Dissociation Kit (Miltenyi Biotec 130-095-942) per manufacturer instructions, and single-cell suspensions were cultured in NeuroCult NS-A Basal Medium (Human) with 1x Proliferation Supplement (STEMCELL Technologies 05751), supplemented with 20 ng/mL EGF (GoldBio 1150-04); 20 ng/mL bFGF (GoldBio 1140-02); 2 μg/mL heparin (STEMCELL Technologies 07980); 100 U/mL and 100 μg/mL, respectively, of penicillin/streptomycin (Thermo Fisher 15–140-148); 250 ng/mL amphotericin B (GeminiBio 400104); and 0.25 μg/mL Plasmocin (InvivoGen ant-mpp) on ultra-low adherence plates (6-well plates: Corning 3471, 10 cm dishes: Corning 4615) in 5% CO₂ and at ambient oxygen at 37°C.

All cell lines were routinely evaluated for mycoplasma contamination with the e-Myco Mycoplasma PCR Detection Kit (Bulldog Bio 2523348), e-Myco PLUS Mycoplasma PCR Detection Kit (Bulldog Bio 25233), or MycoAlert Mycoplasma Detection Kit (Lonza LT07–318), per manufacturer’s instructions, and routinely tested negative throughout the course of this study. As reference short term tandem repeat profiles have not been established for these lines, no cell line authentication was performed. Sex and source of each line is stated above or listed as unknown if unreported in the original publication describing its derivation.

Cell Culture

BT054, BT260, HK157, HK211, HK213, HK252, HK308, TS516, TS603, UTSW5, UTSW63, and UTSW71 human GSCs were cultured in NeuroCult NS-A Basal Medium (Human) with 1x Proliferation Supplement (STEMCELL Technologies 05751), supplemented with 20 ng/mL EGF (GoldBio 1150-04); 20 ng/mL bFGF (GoldBio 1140-02); 2 μg/mL heparin (STEMCELL Technologies 07980); 100 U/mL and 100 μg/mL, respectively, of penicillin/streptomycin (Thermo Fisher 15–140-148); 250 ng/mL amphotericin B (GeminiBio 400104); and 0.25 μg/mL Plasmocin (InvivoGen ant-mpp) on ultra-low adherence plates (6-well plates: Corning 3471, 10 cm dishes: Corning 4615) in 5% CO₂ and at ambient oxygen at 37°C.

ENSA, NSC11, and MGG152 cells were cultured in Neurobasal Medium (Thermo Fisher 21103049) supplemented with 3mM glutamine (Thermo Fisher 25030081); 1× B27 supplement (Thermo Fisher 17504044); 0.25× N2 supplement (Thermo Fisher 17502048); 20 ng/mL EGF (GoldBio 1150-04); 20 ng/mL bFGF (GoldBio 1140-02); 2 μg/mL heparin (STEMCELL Technologies 07980); 50 U/mL and 50 μg/mL, respectively, of penicillin/streptomycin (Thermo Fisher 15–140-148); 125 ng/mL amphotericin B (GeminiBio 400104); and 0.25 μg/mL Plasmocin (InvivoGen ant-mpp) on ultra-low adherence plates (6-well plates: Corning 3471, 10 cm dishes: Corning 4615) in 5% CO₂ and at ambient oxygen at 37°C.

NHA Donor #1, NHA Donor #2, HMC3, HEK293T, and 293AD were cultured in DMEM (Thermo Fisher 11995-065) supplemented with 10% fetal bovine serum (FBS; GeminiBio 100–106) and 100 U/mL and 100 μg/mL, respectively, of penicillin/streptomycin (Thermo Fisher 15–140-148) on tissue culture-treated plates in 5% CO₂ and at ambient oxygen at 37°C. HMC3 microglia were cultured in EMEM (ATCC 30–2003) supplemented with 10% fetal bovine serum (FBS; GeminiBio 100–106) and 100 U/mL and 100 μg/mL, respectively, of penicillin/streptomycin (Thermo Fisher 15–140-148) on tissue culture-treated plates in 5% CO₂ and at ambient oxygen at 37°C.

UTSW63 AGAT KO human GSCs were cultured in NeuroCult NS-A Basal Medium (Human) with 1x Proliferation Supplement (STEMCELL Technologies 05751) as described above, with the addition of anhydrous creatine (Millipore Sigma C0780) to a final concentration of 40 μM on ultra-low adherence plates (6-well plates: Corning 3471, 10 cm dishes: Corning 4615) in 5% CO₂ and at ambient oxygen at 37°C.

Animals

All care and treatment of experimental animals were carried out in strict accordance with Good Animal Practice as defined by the US Office of Laboratory Animal Welfare and approved by the University of Pittsburgh Medical Center (protocols 25066832 and 25087128), UT Southwestern Medical Center (protocols 2019–102795 and 2015–101252) Institutional Animal Care and Use Committee, and Northwestern University (protocol IS00023467). Animal welfare assessments were carried out daily during treatment periods. Animals were housed in a pathogen-free environment between 20–26°C and at 30–70% humidity, with a 12 hour:12 hour light:dark cycle. ICR-SCID (Taconic) mice were obtained from Taconic at 6–7 weeks of age. C57BL/6J mice were obtained from Jackson Laboratories. NSG mice were bred by the UT Southwestern Immunocompromised Mouse Breeding Facility from NSG founders (Jackson Laboratories). Mice were housed together (2–5 mice of the same sex per cage) and provided free access to water and chow diet (Purina ISO Pro Rodent 3000 “5P76” or Teklad 2916).

Generation of DREADD-expressing NSG mice via intracranial AAV injection

Mice in studies incorporating chemogenetic manipulation of neuronal activity via DREADD induction were bred via timed breedings of NSG mice. Pregnant dams were monitored twice-daily for litter delivery. AAV for expression of DREADD receptors was obtained commercially from Addgene. At P2, NSG pups were anesthetized via induced hypothermia, then injected intracranially with 1 μL (all virus preps ≥ 1 × 10¹³ GC/mL) of either AAV9-hSyn-HM3D-Gq-mCherry (Addgene 50474-AAV9) or AAV9-hSyn-mCherry (Addgene 114472-AAV9) into the subventricular zone at 1 mm posterior and .83 mm lateral to the bregma at a depth of 3 mm. The whole injection procedure, including hypothermia-induced anesthesia, was kept under 10 minutes. Pups were allowed to recover on a pre-warmed heating pad. Pups were returned to their maternal cage following injection once normal body temperature and skin color were restored, weaned at P25–28, and maintained on standard chow until xenografting.

Generation of patient-derived xenograft (PDX) mouse models

TS516 orthotopic xenografts for experiments assessing neuronal network activity in tumor-bearing and non-tumor-bearing brain hemispheres were established by intracranial injection of 5,000 cells into 11-week-old female ICR SCID mice (IcrTac:ICR-Prkdc^scid, Taconic ICRSC). TS516 orthotopic xenografts for xenograft studies of AGAT expression were established by intracranial injection of 1,000 TS516 cells into 8-week-old female ICR SCID mice (Taconic ICRSC). TS516 orthotopic xenografts for xenograft studies of dietary ornithine supplementation and arginine restriction were established by intracranial injection of 1,000 TS516 cells into 7-week-old female ICR SCID mice (Taconic ICRSC). Mice were anesthetized with isoflurane and immobilized using a stereotactic frame. An incision was made to expose the skull surface, and a hole was drilled into the skull. Cells suspended in 5 μL cell culture medium were injected into the brain through the hole using a 5 μL syringe (Hamilton) at 0.5–1 mm anterior and 2 mm lateral to the bregma and a depth of 3 mm from the brain surface. The skin was closed with surgical clips and analgesia was administered. Mice receiving experimental diets received either Teklad amino acid diet (Inotiv TD.01084) or diet supplemented with 0.665% ornithine and arginine-free (Inotiv TD.240691). Survival analyses were performed by researchers who were not blinded to the treatment arms or genotypes of the mice. Mice were euthanized when they displayed neurological symptoms or became moribund.

UTSW63 orthotopic xenografts for experiments for xenograft studies of AGAT expression were established by intracranial injection of 10,000 cells into 6–8 week-old NSG mice. Males and females were both used for these experiments in equal proportions for each arm. Mice were anesthetized with isoflurane and immobilized using a stereotactic frame. An incision was made to expose the skull surface, and a hole was drilled into the skull. Cells suspended in 5 μL cell culture medium were injected into the brain through the hole using a 10 μL syringe (Hamilton) at 0.5–1.0 mm anterior and 2 mm lateral to the bregma at a depth of 3 mm from the brain surface. The skin was closed with surgical sutures, and buprenorphine SR and carprofen were given for analgesia. Survival analyses were performed by researchers who were not blinded to the treatment arms or genotypes of the mice. Mice were euthanized when they displayed neurological symptoms or became moribund.

TS516 orthotopic xenografts for experiments assessing chemogenetic manipulation of neuronal network activity in tumor-bearing mice were established by intracranial injection of 1,000 cells into P33-P36 DREADD-expressing NSG mice. Males and females were both used for these experiments in equal proportions for each cohort. Mice were anesthetized with isoflurane and immobilized using a stereotactic frame. An incision was made to expose the skull surface, and a hole was drilled into the skull. Cells suspended in 2.5 μL cell culture medium were injected into the brain through the hole using a 5 μL syringe (Hamilton) at 0.5 mm posterior and 2 mm lateral to the bregma at a depth of 2 mm from the brain surface. The skin was closed with surgical sutures, and buprenorphine SR and carprofen were given for analgesia. After three days’ postsurgical recovery, all xenografted mice were transitioned to a diet containing the DREADD ligand deschloroclozapine (DCZ; Sigma-Aldrich SML3651) at a concentration of 5.042 mg DCZ/kg chow (Research Diets D26011201), a concentration estimated to yield a daily PO dose per animal of 0.5 mg DCZ/kg¹⁰⁰, on the OpenStandard Diet background (Research Diets D11112201) and were maintained on this diet for the remainder of the study.

TS516 flank xenografts were established by subcutaneous injection of 1 × 10⁶ TS516 cells in 100 μL of a 1:1 mixture of complete NeuroCult NS-A Medium (Human) (STEMCELL Technologies 05751) and Matrigel Matrix (Corning 354234) into 6–8-week-old female nude mice (Taconic NCRNU). Animals were monitored daily following injection for development of a measurable tumor. Tumors were then measured every 3–4 days using calipers (Fisher Scientific 06–664-16), and tumor volume calculated as $V=\frac{1}{2}\mathrm{*}L\mathrm{*}{W}^2$, where $L$ was the longest axis of measurement of the tumor and $W$ was the axis 90° perpendicular to $L$. Animals were sacrificed upon attaining a total tumor volume of 1000 mm³, a tumor length ≥ 20 mm, or development of ulceration of the tumor.

Method Details

Chemicals

Where indicated, cells or brain slices were treated with gabazine (EMD Millipore SR95531), GAA (Millipore Sigma G11608, or TCI G0167), creatine monohydrate (Thermo Fisher B25009.22), 10 μM 3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic acid (CPP) (Hello Bio HB0036), 20 μM (6,7-dinitroquinoxaline-2,3-dione (DNQX) (Tocris 2312), 100 μM picrotoxin (Hello Bio HB0506), and/or 2 μM CGP 52432 (Tocris 1246).

Vectors

lentiCRISPRv2-GFP (Addgene 82416) and lentiCRISPRv2 mCherry (Addgene 99154) were digested with FastDigest Esp3I (Thermo Fisher FD0454) or BsmBIv2 (New England Biolabs R0739). sgRNAs targeting the AAVS1 safe harbor locus⁸⁵ (sg #1: GTCACCAATCCTGTCCCTAG), AGAT (sg #1: TGTGGGCAATGAGATTATCG), or GAMT (sg #1: GGCCAGCGCGTGCATATAGG; sg #2: GTACGACACGTACCCACTCT) were ligated into lentiCRISPRv2-GFP to generate a construct expressing Cas9 and the desired sgRNA. sgRNAs targeting the AAVS1 safe harbor locus (sg #2: TGTTAGGCAGATTCCTTATC) or AGAT (sg #2: ACTTCAATGACCAGTCAATG) were ligated into lentiCRISPRv2-mCherry to generate a construct expressing Cas9 and the desired sgRNA. Reaction mixtures were transformed into XL10-Gold Ultracompetent Cells (Agilent 200315) and validated by whole plasmid sequencing (Plasmidsaurus).

CRISPR/Cas9 and fluorophore expression cassettes for lentiCRISPRv2-GFP-sgAAVS1 #1, lentiCRISPRv2-GFP-sgAGAT #1, lentiCRISPRv2-mCherry-sgAAVS1 #2, and lentiCRISPRv2-mCherry-sgAGAT #2 were amplified from these constructs and appended with 5’ attB1 and 3’ attB2 sites using the following primers:

attB1_U6_Fwd_v2:

GGGGACAAGTTTGTACAAAAAAGCAGGCTCTGAGGGCCTATTTCCCATG

bGHpolyA_attB2_Rev_v2:

GGGGACCACTTTGTACAAGAAAGCTGGGTTCCATAGAGCCCACCGCAT

PCR product was gel-purified with the QIAquick Gel Extraction Kit (Qiagen 28706), purified with the Monarch PCR & DNA Cleanup Kit (New England Biolabs T1030) and cloned into the Gateway vector pDONR223 by BP reaction (Thermo Fisher 11789100) overnight. The resultant products were transformed into HB101 Competent Cells (Promega L2015), and identity of the final products were validated by whole-plasmid sequencing (Plasmidsaurus).

pENTR223 plasmids containing CRISPR/Cas9 expression cassettes for GFP-sgAAVS1 #1, GFP-sgAGAT #1, mCherry-sgAAVS1 #2, and mCherry-sgAGAT #2 were cloned into the Gateway vector pAd/PL-DEST (Thermo Fisher V49420) by LR reaction (Thermo Fisher 11791020) overnight. Resultant pAdCRISPR-GFP-sgAAVS1–1, pAdCRISPR-GFP-sgAGAT-1, pAdCRISPR-mCherry-sgAAVS1–2 and pAdCRISPR-mCherry-sgAGAT-2 vectors for adenoviral transduction of mammalian cells were transformed into HB101 Competent Cells (Promega L2015) and validated by whole plasmid sequencing (Plasmidsaurus).

AGAT (RefSeq NM_001482.2) WT cDNA was generated from the Ultimate ORF Human Clone Library (Life Technologies HORF01) in a pENTR221 Gateway vector. The C407A point mutation was generated by inverse PCR of pENTR221-AGAT using CloneAmp HiFi PCR Master Mix (Takara 639298) and the following primers:

GATM_C407A_InvPCR_F: TTCCATGCCTGGACCTGCGATGTC

GATM_C407A_InvPCR_R: GGTCCAGGCATGGAAGCCTCCT

The amplified inverse PCR product was then gel-purified with the QIAquick Gel Extraction Kit (Qiagen 28706), PCR purified with the Monarch PCR & DNA Cleanup Kit (New England Biolabs T1030) and assembled using the In-Fusion HD Cloning Kit (Takara Bio 102518). The assembled product was then transformed into XL10-Gold Ultracompetent Cells (Agilent 200315). Final pENTR221-AGAT_C407A product was validated by whole plasmid sequencing (Plasmidsaurus).

5’ attB1 and 3’ attB2 sites were appended to AGAT WT and AGAT C407A cDNA using Q5 Hot Start High-Fidelity 2X Master Mix polymerase (New England Biolabs M0494) and the following primers:

attB_AGAT_F: GGGGACAAGTTTGTACAAAAAAGCAGGCTCTGCCACCATGCTGCGGGT

attB_AGAT_R: GGGGACCACTTTGTACAAGAAAGCTGGGTTTCAGTCCAAGTAGGACTGTAAGGTGCC

PCR product was gel-purified with the QIAquick Gel Extraction Kit (Qiagen 28706), purified with the Monarch PCR & DNA Cleanup Kit (New England Biolabs T1030) and cloned into the Gateway vector pDONR223 by BP reaction (Thermo Fisher 11789100) overnight. The resultant products were transformed into HB101 Competent Cells (Promega L2015). pENTR223-AGAT_WT and pENTR223-AGAT_C407A were validated by whole plasmid sequencing (Plasmidsaurus).

GAMT WT and GAMT E45S cDNA were synthesized as gBlocks by Twist Biosciences. attB sites were added and sequences amplified using Q5 Hot Start High-Fidelity 2X Master Mix polymerase (New England Biolabs M0494) and the following primers:

attB_GAMT_F: GGGGACAAGTTTGTACAAAAAAGCAGGCTCTGCCACCATGAGCGCCCC

attB_GAMT_R: GGGGACCACTTTGTACAAGAAAGCTGGGTTTCAGCCTTTGGTCACCAGGGG

PCR product was gel-purified with the QIAquick Gel Extraction Kit (Qiagen 28706), PCR purified with the Monarch PCR & DNA Cleanup Kit (New England Biolabs T1030) and cloned into the Gateway vector pDONR223 by BP reaction (Thermo Fisher 11789100) overnight to generate pENTR223-GAMT_WT and pENTR223-GAMT_E45S plasmids. The resultant products were transformed into HB101 Competent Cells (Promega L2015). pENTR223-GAMT_WT and pENTR223-GAMT _E45S were validated by whole plasmid sequencing (Plasmidsaurus).

pENTR223-EV, pENTR223-AGAT_WT, and pENTR223-AGAT_C407A were cloned into pLenti-EF1α-DEST-IRES-Neo⁴² (Addgene cat# 253766) by Gateway LR reaction (Thermo Fisher 11791020) overnight. The lentiviral vectors pLenti-EF1α-EV-IRES-Neo, pLenti-EF1α-AGAT_WT-IRES-Neo, and pLenti-EF1α-AGAT_C407A-IRES-Neo were transformed into HB101 Competent Cells (Promega L2015) and validated by whole plasmid sequencing (Plasmidsaurus).

pENTR223-EV, pENTR223-GAMT_WT, and pENTR223-GAMT_E45S were cloned into pLenti-EF1α-DEST-PGK-Hygro by Gateway LR reaction (Thermo Fisher 11791020) overnight. The lentiviral vectors pLenti-EF1α-EV-PGK-Hygro, pLenti-EF1α-GAMT_WT-PGK-Hygro, and pLenti-EF1α-GAMT_E45S-PGK-Hygro were transformed into HB101 Competent Cells (Promega L2015) and validated by whole plasmid sequencing (Plasmidsaurus).

Cassettes containing 4 CRISPR-Cas12a crRNA guides targeting either the AAVS safe harbor locus (TCTTCTGTGCACTACTTGTC; ATGGCTTCTAGGTTGCGGAC; ATTTCCCAGCATGCCCTGCGAGG; TGGACAACCCCAAAGTACCCCG) or AGAT (GTCTGACAGGTTTATACAGTGCA, GGAGCTGTTGTCCACTTGGCGCC, ATAATTGACCTGTACGCTCGGTA, TGCTCTGCCCACTATCACTTCC) were designed using the following template, with individual guides substituting for poly-N sequences in the order presented:

AATGATACGGCGACCACCGAcgtctcgAGATnnnnnnnnnnnnnnnnnnnnTAATTTCTACTATTGTAGATnnnnnnnnnnnnnnnnnnnnAAATTTCTACTCTAGTAGATnnnnnnnnnnnnnnnnnnnnTAATTTCTACTGTCGTAGATnnnnnnnnnnnnnnnnnnnnTTTTTTGAATggagacgATCTCGTATGCCGTCTTCTGCTTG

Guide-containing cassettes were ordered as g-blocks from IDT. Guide-containing g-blocks and pRDB-054 (Addgene 216096) were each digested with BsmBI-v2 (New England Biolabs R0739). Digested pRDB-054 was gel-purified with the QIAquick Gel Extraction Kit (Qiagen 28706), dephosphorylated using Quick CIP (New England Biolabs M0525), and purified with the Monarch PCR & DNA Cleanup Kit (New England Biolabs T1030). Guide cassette digests were heat-inactivated at 80° C, then ligated into pRDB-054 using Salt T4 DNA Ligase (New England Biolabs M0467). Reaction mixtures were transformed into XL10-Gold Ultracompetent Cells (Agilent 200315) and validated by whole plasmid sequencing (Plasmidsaurus).

The mCherry fluorophore was amplified from lentiCRISPRv2 mCherry and overhangs added for hybridization using Q5 Hotstart High-Fidelity 2X Master Mix polymerase (New England Biolabs M0494) and the following primers:

LGmCherry_F: TCGACGAATTCGGATCCGCAATGGTGAGCAAGGGCGAG

LGmCherry_R: ATTAAGTTTAAACGCGGCCATTACTTGTACAGCTCGTCCATG

The ccdB-containing Gateway destination cassette was removed from pLV-EF1α-IRES-puro (Addgene cat# 253769) and the resulting product amplified using Q5 Hotstart High-Fidelity 2X Master Mix polymerase (New England Biolabs M0494) and the following primers:

pLV-EF1a-IRES-Puro DEST_fwd: TGGCCGCGTTTAAACTTAATTATC

pLV-EF1a-IRES-Puro DEST_rev: TGCGGATCCGAATTCGTC

Each of the above PCR products was gel-purified with the QIAquick Gel Extraction Kit (Qiagen 28706) and PCR purified with the Monarch PCR & DNA Cleanup Kit (New England Biolabs T1030). mCherry was then cloned into the linearized pLV-EF1α-IRES-puro cassette using NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs E2621) to generate pLV-mCherry-EF1α-IRES-puro. The resultant products were transformed into HB101 Competent Cells (Promega L2015).

The GFP fluorophore was amplified from pLenti-EF1α-GFP-IRES hygro and appended with attB flanking sites using Q5 Hotstart High-Fidelity 2X Master Mix polymerase (New England Biolabs M0494) and the following primers:

attB_GFP_F: GGGGACAAGTTTGTACAAAAAAGCAGGCTTAggatccgcatggtgagcaag

attB_GFP_R: GGGGACCACTTTGTACAAGAAAGCTGGGTTttaattaagtttaaacgcggccttacttgtac

PCR product was gel-purified with the QIAquick Gel Extraction Kit (Qiagen 28706), PCR purified with the Monarch PCR & DNA Cleanup Kit (New England Biolabs T1030) and cloned into the Gateway vector pDONR223 by BP reaction (Thermo Fisher 11789100) overnight to generate pENTR223-GFP, which was then transformed into HB101 Competent Cells (Promega L2015) and validated by whole plasmid sequencing (Plasmidsaurus).

pDONR223-GFP was cloned into pLenti-CMV-DEST-PGK-Puro by Gateway LR reaction (Thermo Fisher 11791020) overnight. The pLenti-CMV-GFP-PGK-Puro lentiviral vector was then transformed into HB101 Competent Cells (Promega L2015) and validated by whole plasmid sequencing (Plasmidsaurus).

Transfection and viral transduction

Lentiviral particles were produced by transfection of HEK293T cells with expression vectors and packaging plasmids psPAX2 (Addgene 12260, gift of Didier Trono) and pMD2.G (Addgene 12259, gift of Didier Trono) in a ratio of 4:3:1 using TransIT-LT1 transfection reagent (Mirus Bio MIR2 B-304). Media was discarded and replaced at 24 hours post-transfection, then virus-containing media was collected 48 and 72 hours post-transfection and passed through a 0.45 μm filter (Corning 431220). 1 volume LentiX Concentrator (Takara 631232) was added per 3 volumes clarified supernatant. The mixture of viral supernatant and LentiX Concentrator was incubated at 4°C overnight, then centrifuged at 1500 × g for 45 minutes. After centrifugation, supernatant was discarded, and the remaining pellet was resuspended in ⅙ original viral supernatant volume of DMEM supplemented with 10% fetal bovine serum (FBS; GeminiBio 100–106) and 100 U/mL and 100 μg/mL, respectively, of penicillin/streptomycin (Thermo Fisher 15–140-148). For lentiviral transduction of UTSW63 and TS516 GSCs, lentivirus was produced and concentrated as above, then resuspended in 1/100 original viral supernatant volume of DPBS (Fisher Scientific 14–190-250) to prevent exposure of GSCs to FBS, and the resulting viral titer was quantified using LentiX GoStix Plus (Takara 631281) to permit normalization of viral dose for transductions.

To generate adenovirus, pAdCRISPR constructs were linearized by PacI (New England Biolabs R0547S) digestion at 37°C overnight. 293Ad cells were subsequently transfected with 2.5 μg linear pAdCRISPR constructs using TransIT-LT1 transfection reagent (Mirus Bio MIR2 B-304). Media was changed at 24 hours after transfection, and cells were expanded to a 10 cm dish and incubated for 10–14 days with regular media changes. 10–14 days following transfection, 293Ad cells and media were harvested and centrifuged at 3739 × g for 5 minutes. Supernatant was discarded and the pellet was resuspended in 1 mL media. Resuspended pellet was subjected to 3 freeze-thaw cycles with 30 minutes at −80°C followed by 15 minutes at 37°C. After final thaw, suspensions were centrifuged at 2093 × g for 15 minutes to generate crude adenovirus. Presence of adenovirus capsid in crude adenovirus prep was confirmed using Adeno-X GoStix (Takara 632270). Crude adenovirus was then amplified 2–3 times by inoculating 293Ad cells with crude virus and harvesting as above after 2–4 days. Amplified virus was purified using Virabind Adenovirus Miniprep Kit (Cell Biolabs VPK-099) according to manufacturer’s instructions.

NHA Donor #1 cells were plated at a density of 1.5 × 10⁵ cells per well in 2 mL DMEM in a six well plate and allowed to adhere overnight. The next day, 3 uL polybrene (8 μg/mL, MedChemExpress HY-112735) was added to each well along with 1 mL viral pLenti-EF1a-EV-IRES-Neo, pLenti-EF1a-AGAT_WT-IRES-Neo, or pLenti-EF1a-AGAT_C407A-IRES-Neo supernatant. Plates were centrifuged at 4,000 × g for 30 minutes at room temperature, then incubated overnight. The following day, cells were expanded and replated in a 10 cm dish. After initial selection at 1500 μg/mL G418 (GoldBio G-418), stable cell lines were maintained in 600 μg/mL G418.

NHA Donor #1 EV, AGAT WT, and AGAT C407A cells were plated at a density of 1.5 × 10⁵ cells per well in 2 mL DMEM without G418 in a six well plate and allowed to adhere overnight. The next day, 3 uL polybrene (8 μg/mL, MedChemExpress HY-112735) was added to each well along with 1 mL virus for pLenti-EF1a-EV-PGK-hygro, pLenti-EF1a-GAMT_WT-PGK-Hygro, or pLenti-EF1a-GAMT_E45S-PGK-Hygro; or 1 mL lentiCRISPRv2-GFP-sgAAVS1–1, lentiCRISPRv2-GFP-sgGAMT-1, or lentiCRISPRv2-GFP-sgGAMT-2 lentivirus. Plates were centrifuged at 4,000 × g for 30 minutes at room temperature, then incubated overnight. The following day, cells were expanded and replated in a 10 cm dish. For cells transduced with pLenti-EF1a-EV-PGK-hygro, pLenti-EF1a-GAMT_WT-PGK-Hygro, or pLenti-EF1a-GAMT_E45S-PGK-Hygro, following initial selection at 250 μg/mL hygromycin (GoldBio H-270-EZ50), stable cell lines were maintained in 100 μg/mL hygromycin. Cells transduced with lentiCRISPRv2-GFP constructs were sorted for GFP positivity 14 days post-transduction using a FACS Aria II SORP Four-Laser instrument (BD Biosciences) in the Children’s Research Institute Flow Cytometry Core.

For lentiviral transduction, UTSW63 and TS516 AGAT WT GSCs were plated at 3 × 10⁵ cells in 1.6 mL NeuroCult NS-A Basal Medium (Human) per well in a 6-well format and incubated with lentivirus (pRDB_054_4x_crAAVS1 or pRDB_054_4x_crAGAT for UTSW63; pLV-EF1α-mCherry-IRES-puro for TS516 AGAT WT) and 1.6 μL polybrene overnight. The following day, virus-containing media was removed, and like-transduced wells were combined into 10 cm ultra-low adherence dishes in 10 mL NeuroCult NS-A Basal Medium. TS516 cells were selected for mCherry positivity 3 days following lentiviral transduction using a FACS Aria II SORP Four-Laser instrument (BD Biosciences) in the Children’s Research Institute Flow Cytometry Core. UTSW63 cells were selected in puromycin at 1 μg/mL beginning 3 days after lentiviral transduction.

For adenoviral transduction, TS516 cells were plated at 1 × 10⁷ cells per dish in 10 mL NeuroCult NS-A Basal Medium (Human) and treated with adenovirus produced from pAdCRISPR-GFP-sgAAVS1–1 or pAdCRISPR-GFP-sgAGAT-1 at a concentration empirically determined to yield 30–50% transduction efficiency for each viral preparation. At 16–18 hours, media was changed to fresh NeuroCult NS-A Basal Medium (Human) and cells were split out to 2 dishes each. At 7 days post-transduction, cells were sorted for GFP positivity using a FACS Aria II SORP Four-Laser instrument (BD Biosciences) in the Children’s Research Institute Flow Cytometry Core. Sorted GFP⁺ cells were transduced with adenovirus produced from pAdCRISPR-mCherry-sgAAVS1–2 or pAdCRISPR-mCherry-sgAGAT-2 adenovirus 14 days following initial sort as previously done with GFP-containing virus, then sorted 7 days following secondary transduction using a FACS Aria II SORP Four-Laser instrument (BD Biosciences) in the Children’s Research Institute Flow Cytometry Core. After two weeks in culture, sorted mCherry⁺ cells were sorted for GFP and mCherry dual-negative populations to establish stably-transduced AGAT WT and AGAT KO lines.

Preparation of human plasma-like medium (HPLM) and tracer HPLM

HPLM was prepared as previously described¹⁰¹. Tracer-containing HPLM library was as previously described⁴², with the addition of a medium containing ¹⁵N₃-creatine (Cambridge Isotope Laboratories NLM-9218). Tracer-containing HPLM used in nitrogen metabolism profiling, profiling validation, and ¹⁵N₄-arginine tracing experiments were generated identically to the unlabeled version, with labeled metabolites replacing their unlabeled counterparts at equimolar concentrations. HPLM for ¹⁵N₄-arginine tracing in TS516 under ornithine-supplemented conditions was generated without arginine or ornithine, then 400 μM ¹⁵N_4-arginine (Cambridge Isotope Laboratories NLM-396) and either 0 μM or 150 μM ornithine (Millipore Sigma O2375) were added. Tracer HPLM was assembled by mixing thawed frozen or freshly-prepared stocks of unlabeled HPLM pools, replacing one pool per tracing condition with a version containing an individual ¹⁵N tracer. Media was adjusted to pH 7.4 using NaOH or HCl, then diluted to the final volume with ultrapure water, sterile filtered with a 0.22 μm PES filter (Millipore Sigma SCGP00525, Corning 431153, Corning 431097, or Corning 431098), and supplemented as described below. To prevent glutamate toxicity in culture, HPLM was prepared without glutamate for experiments in GSCs, NSCs, and immortalized astrocyte cells, except for conditions in which glutamate was traced.

For experiments in GSCs and NSCs, HPLM was supplemented with 1 × B27 supplement (Thermo Fisher 17504044); 0.25 × N2 supplement (Thermo Fisher 17502048); 20 ng/mL EGF (GoldBio 1150-04); 20 ng/mL bFGF (GoldBio 1140-02); 2 μg/mL heparin (STEMCELL Technologies 07980); 50 U/mL and 50 μg/mL, respectively, penicillin/streptomycin (Thermo Fisher 15–140-148); 125 ng/mL amphotericin B (GeminiBio 400104); and 0.25 μg/mL Plasmocin (InvivoGen ant-mpp). For experiments in differentiated cells, HPLM was supplemented with 10% dialyzed FBS (GeminiBio 100–108) and 100 U/mL and 100 μg/mL, respectively, of penicillin/streptomycin (Thermo Fisher 15140148).

Nitrogen metabolism profiling platform

Nitrogen metabolism profiling was conducted as described previously⁴². Briefly, NHA Donor #1 cells were plated in 6-well plates (2.5 × 10⁴ cells per well) in 2 mL DMEM prepared as described above. TS516 cells were plated in 6-well ultra-low adherence plates (2 × 10⁵ cells per well) in 2 mL NeuroCult NS-A Basal Medium (Human) prepared as described above. After 24 hours, 2 mL unlabeled HPLM prepared as described above was added to produce a mixture of 50% native medium and 50% HPLM. 24 hours later, media was changed to 100% unlabeled HPLM (10 mL for NHA Donor #1 cells and 4.5 mL for TS516 cells). 24 hours later, media was changed to fresh 100% HPLM with one nitrogen-containing metabolite per sample exchanged for an equimolar amount of its ¹⁵N-labeled counterpart (10 mL for NHA Donor #1 cells, 4.5 mL for TS516 cells). After 18 hours, samples were harvested and prepared for LC-MS analysis as described below.

Creatine synthesis pathway tracing in cells

For tracing of creatine synthesis pathway substrates, differentiated cells (NHA Donor #1 and #2) were plated at 2.5 × 10⁴ cells per well for NHA Donor #1 and 1 × 10⁵ cells per well for NHA Donor #2 in 2 mL DMEM prepared as described above. GSCs with exception of MGG152 were plated at 2 × 10⁵ cells per well in 2 mL NeuroCult NS-A Basal Medium (Human) prepared as described above in 6-well ultra-low adherence plates. NSCs and MGG152 were plated at 2 × 10⁵ cells per well in 2 mL NeuroBasal prepared as described above in 6-well ultra-low adherence plates. 24 hours after plating, 2 mL unlabeled HPLM prepared as described above was added to produce a 1:1 mixture of HPLM and native medium. 24 hours later, media was changed to 10 mL 100% HPLM for NHA Donor #1 and #2 or 4.5 mL HPLM for GSCs or NSCs. 24 hours later, media was changed to 10 or 4.5 mL fresh HPLM in which the unlabeled metabolite was exchanged for an equimolar concentration of ¹⁵N₄-arginine (Cambridge Isotope Laboratories NLM-396), ¹⁵N₃-creatine (Cambridge Isotope Laboratories NLM-9218), ¹⁵N-glycine (Cambridge Isotope Laboratories NLM-202), or ¹⁵N-serine (Cambridge Isotope Laboratories NLM-2036). Cells were incubated with tracer for 18 hours, then harvested and prepared for LC-MS as described below. For each cell line, counts and diameter were quantified from parallel cell cultures using a Beckman Coulter Vi-CELL XR cell viability analyzer (RRID: SCR_019664) at the time of sample harvest.

Intracellular GAA content in cultured cells

Differentiated cells (NHA Donor #1 and #2) were plated at 5 × 10⁴ cells per well for NHA Donor #1 and 2 × 10⁵ cells per well for NHA Donor #2 in 2 mL DMEM. GSCs with exception of MGG152 were plated at 4 × 10⁵ cells per well in 2 mL NeuroCult NS-A Basal Medium (Human) prepared as described above in 6-well ultra-low adherence plates. NSCs and MGG152 were plated at 4 × 10⁵ cells per well in 2 mL NeuroBasal medium prepared as described above in 6-well ultra-low adherence plates. 24 hours after plating, 2 mL unlabeled HPLM prepared as described above was added to produce a 1:1 mixture of HPLM and native medium. 24 hours later, media was changed to 10 mL 100% HPLM for NHA Donor #1 and #2 or 4.5 mL HPLM for GSCs or NSCs. 24 hours later, media was changed to 2 mL fresh HPLM. 24 hours later, media was changed to 2 mL fresh HPLM per well. 24 hours later, cells were collected, washed in Optima saline, and snap frozen. For each cell line, cell counts and cell diameter were quantified from parallel cell cultures using a Beckman Coulter Vi-CELL XR cell viability analyzer (RRID: SCR_019664) at the time of sample harvest.

Secretion of GAA in cell culture

Differentiated cells (NHA Donor #1 and #2) were plated at 5 × 10⁴ cells per well for NHA Donor #1 and 2 × 10⁵ cells per well for NHA Donor #2 in 2 mL DMEM in 6-well plates. HMC3 microglia were plated at 5 × 10⁴ cells per well in 2 mL EMEM in 6-well plates. GSCs with exception of MGG152 were plated at 4 × 10⁵ cells per well in 2 mL NeuroCult NS-A Basal Medium (Human) prepared as described above in 6-well ultra-low adherence plates. NSCs and MGG152 were plated at 4 × 10⁵ cells per well in 2 mL NeuroBasal prepared as described above in 6-well ultra-low adherence plates. 24 hours after plating, 2 mL unlabeled HPLM prepared as described above was added to produce a 1:1 mixture of HPLM and native medium. 24 hours later, media was changed to 10 mL 100% HPLM for NHA Donor #1 and #2 or 4.5 mL HPLM for GSCs or NSCs. 24 hours later, media for all cells was changed to 2 mL fresh HPLM. 48 hours later, media samples were collected from wells, spun at 4,000 × g for 3 minutes, 100 μL transferred to a fresh microcentrifuge tube, and snap frozen in liquid nitrogen. For each cell line, cell counts and cell diameter were quantified from parallel cell cultures using a Beckman Coulter Vi-CELL XR cell viability analyzer (RRID: SCR_019664) at the time of sample harvest.

Preparation of cultured cells for LC-MS

Prior to LC-MS analysis, cell samples were suspended in 1 μL ice-cold 80% acetonitrile (Thermo Fisher A9554) per 1,000 cells. Adherent cells (NHA Donor #1 and NHA Donor #2) were lifted from plates by scraping to facilitate resuspension. Cell samples resuspended in acetonitrile were platform-vortexed for 20 minutes at 4°C and centrifuged for 10 minutes at 17,000–21,100 × g at 4°C. The supernatant was transferred to a fresh tube and again centrifuged for 10 minutes at 17,000–21,100 × g at 4°C. The resultant supernatant was then analyzed by LC-MS. For experiments in which absolute quantification of GAA in cell samples was performed, internal standards 1,2-¹³C₂, 3-¹⁵N-guanidinoacetic acid (Cambridge Isotope Laboratories CNLM-8300) or 2,2-D₂-guanidinoacetic acid (Cambridge Isotope Laboratories DLM-9998) at concentrations of 20, 100, or 200 nM were added. Standard curves for GAA (Millipore Sigma G11608) were created fresh for each submission in neat 80% acetonitrile over a range of 1–10,000 nM. All reagents used were Optima-grade.

Preparation of media for LC-MS

Prior to LC-MS analysis, media was diluted 1:10 in ice-cold 80% acetonitrile (Thermo Fisher A9554) and vortexed for 20 minutes at 4°C, then centrifuged for 10 minutes at 17,000–21,100 × g at 4°C. The supernatant was transferred to a fresh tube and again centrifuged for 10 minutes at 17,000–21,100 × g at 4°C. The resultant supernatant was then analyzed by LC-MS. For experiments in which absolute quantification of GAA was performed in media, an internal standard of 1,2-¹³C₂, 3-¹⁵N-guanidinoacetic acid (Cambridge Isotope Laboratories CNLM-8300) or 2,2-D₂-guanidinoacetic acid (Cambridge Isotope Laboratories DLM-9998) was added to all samples and standards at concentrations of 20, 100, or 200 nM. Standard curves for GAA (Millipore Sigma G11608) were created fresh for each submission in neat 80% acetonitrile over a range of 1–10,000 nM. All reagents used were Optima-grade.

Preparation of tissue for LC-MS

Tissues collected for metabolomic profiling of human brain tumors were obtained in the operating room by neurosurgeons during stereotactically guided surgical resections to allow for targeting of solid tumor tissue. Intraoperative areas with hemorrhage, necrosis, or non-viable tissue were avoided. The brain specimens were immediately stored dry on ice, transferred to the laboratory and snap frozen in less than 30 minutes. Samples were prepared as previously described¹⁰², with some modifications. Briefly, samples were cut with pre-cold forceps on dry ice and approximately 10 mg tissue per sample was weighed, thawed on ice, washed in 1 mL ice-cold normal saline, and resuspended in 50 μL ice-cold 80% methanol (Thermo Fisher A456) per mg of tissue. Tissue samples were homogenized in a TissueLyser II (Qiagen 85300, RRID: SCR_018623) at 4°C at an oscillation rate of 25 Hz. 520 μL tert-Butyl methyl ether (Millipore Sigma 650560) and 380 μL water were added to 500 μL of tissue homogenate in a glass vial, then vortexed at room temperature for 1 hour at 1,000 rpm in a thermal mixer (Benchmark Scientific H5000-HC). Samples were centrifuged for 10 minutes at 1,000 × g at 4°C.

For polar metabolomics analysis, 800 μL of the lower aqueous phase was combined with 300 μL of the upper organic phase, avoiding interlayer debris. A SpeedVac (Thermo Fisher SPD2030) was used to dry down polar metabolite extracts at room temperature, and dried metabolites were then resuspended in 100 μL ice-cold 80% acetonitrile (Thermo Fisher A9554). A mixture of internal standards, composed of 3 μM ¹³C₅-glutamine (Cambridge Isotope Laboratories CLM-1822), 3 μM ¹⁵N-valine (Cambridge Isotope Laboratories NLM-316), 3 μM ¹⁵N-methionine (Cambridge Isotope Laboratories NLM-752), and 3 μM chloramphenicol (Millipore Sigma C0378) was added. Samples were sonicated in a room temperature water bath for 10 minutes, then centrifuged for 10 minutes at 21,100 × g at 4°C. The resultant supernatant was then analyzed by LC-MS.

For lipidomics analysis, 100 μL of upper organic phase was transferred to a new glass vial. The lipid extract was dried using nitrogen to avoid oxidation, and dried lipids were resuspended in 500 μL isopropanol and methanol mixture (65:35, V/V) with 15 μL of Avanti Sph/Cer mix standard (Avanti Research, 330707). Samples were then vortexed at 4°C for 10 minutes, then centrifuged for 10 minutes at 1000 × g at 4°C. The resultant supernatant was then analyzed by LC-MS.

Both polar metabolite and lipid samples were run in a single batch to avoid batch-related variation. In addition, pooled quality control (QC) samples were injected after every 10 study samples throughout LC-MS acquisition to monitor instrument stability and analytical reproducibility.

For experiments in which absolute quantification of GAA in human tissue samples or xenograft tissue samples was performed, all samples were snap-frozen immediately in liquid nitrogen in the operating room or animal procedure room within minutes of surgical collection and were not stored in ice-cold Hibernate A prior to freezing. After snap-freezing, tissues were stored at or below −80°C until analysis. Tissues were homogenized in 50 μL ice-cold 80% Optima methanol (Thermo Fisher A456) per mg of tissue using a TissueLyser II (Qiagen 85300, RRID: SCR_018623) with two carbide beads per tube (Qiagen 69997) at an oscillation rate of 25 Hz. Samples were vortexed for 20 minutes at 4°C, then centrifuged for 10 minutes at 17,000–21,100 × g at 4°C. The supernatant was transferred to a fresh tube then centrifuged for 10 minutes at 17,000–21,100 × g at 4°C. 200 or 250 μL of supernatant was transferred to a fresh tube, then dried using a SpeedVac (Thermo Fisher SPD2030). Dried metabolites were resuspended in 7.5 μL ice-cold 80% acetonitrile (Thermo Fisher A9554) per mg tissue, vortexed for 20 minutes at 4°C, and centrifuged for 10 minutes at 17,000–21,100 × g at 4°C. The resultant supernatant was then diluted 1:10 and analyzed by LC-MS. For experiments in which absolute quantitation of GAA was performed in tissues, 1,2-¹³C₂, 3-¹⁵N-guanidinoacetic acid (Cambridge Isotope Laboratories CNLM-8300) or 2,2-D₂-guanidinoacetic acid (Cambridge Isotope Laboratories DLM-9998) was added as an internal standard to all samples and standards. Standard curves for GAA (Millipore Sigma G11608) were created fresh for each submission in neat 80% acetonitrile over a range of 1–10,000 nM. All reagents used were Optima LC/MS grade.

Extraction of TS516 AGAT WT and AGAT KO xenograft tissues collected for quantification of GAA and creatine was performed using pre-chilled 80% methanol (Thermo Fisher A456). A volume of 80% MeOH equivalent to 50 μL per mg of tissue was added to each sample tube. Two tungsten carbide beads (Qiagen 69997) were added to each tube for homogenization. Tissue homogenization was carried out using a TissueLyser set to 25 Hz for 1-minute cycles. Following homogenization, samples were vortexed for 20 minutes at 4 °C. Samples were centrifuged twice at maximum speed (~13,000–15,000 × g) for 10 minutes at 4 °C. The resulting supernatant was transferred and analyzed by LC-MS/MS and internal standards for both GAA (Guanidineacetic-2,2-d2 Acid, CDN Isotopes D-6320) and creatine [Creatine-d5 H2O (N-methyl-d3; glycine-2,2-d2), LGC Standards TRC-C781496] were added to final supernatant. A standard curve of creatine (LGC Standards TRC-C781483) and GAA (TRC-G821250) was created for each run with duplicate QCs (low, medium, and high) to ensure the results for each day were accurate. The linear range for GAA and creatine in brain was 1–25,000 ng/mL.

Metabolite quantification by LC-MS

10–20 μL of metabolite extract in acetonitrile was injected and analyzed with a Q-Exactive HF-X (Thermo, RRID: SCR_020425), Orbitrap LUMOS (Thermo, RRID: SCR_020562), or Orbitrap Exploris 480 (Thermo, RRID: SCR_027000) hybrid quadrupole-orbitrap mass spectrometer coupled to a Vanquish Flex UHPLC system (Thermo Fisher), as described previously^103,104; or with an AB QTRAP 5500 (Applied Biosystems Sciex) or AB QTRAP 6500+ (Applied Biosystems Sciex) with ESI source coupled to a Nexera X2 LC-30AD HPLC (Shimadzu Corporation). Chromatographic separation of samples acquired on the Q-Exactive or Exploris Orbitrap instruments was accomplished using a Millipore-Sigma ZIC-pHILIC column using a binary gradient of 10 mM ammonium formate pH 9.8 and acetonitrile. Chromatographic separation of samples acquired on the Orbitrap LUMOS instrument was accomplished using a binary gradient of 50% water / 50% acetonitrile, 0.1% formic acid, 5 mM ammonium formate and 88% isopropanol / 10% acetonitrile / 2% water, 0.1% formic acid, 5 mM ammonium formate. Spectra acquired on Orbitrap instruments were acquired with resolving power of 60,000, 120,000, or 240,000 full width at half maximum (FWHM), and a scan range set to 60–900, 70–1050, 80–1,200 m/z, or 250–1500 m/z, collecting data in both positive and negative polarities. To improve signal resolution of GAA and its isotopologues, a targeted selected ion monitoring (tSIM) scan event using a 7 Da window around GAA was used in tandem with full scan data acquisition to capture all relevant stable isotope labeling events with an AGC targeted of 1 × 10⁵ ions.

Chromatographic separation of samples acquired on the AB QTRAP 5500 or AB QTRAP 6500+ was accomplished with a SeQuant^® ZIC^®-pHILIC HPLC column (150 × 2.1 mm, 5 μm, polymeric) or a BEH Z-HILIC VanGuardTM Fit HPLC column (150 × 2.1 mm, 5 μm), using an electrospray ionization (ESI) source in multiple reaction monitoring (MRM) mode. Samples acquired on the AB QTRAP 5500 were resolved using a binary gradient of 10 mM ammonium acetate in water (pH 9.8 adjusted with ammonium water; mobile phase A) and 100% ACN (mobile phase B), with gradient elution as follows: 0–15 minutes, linear gradient 90–30% B, followed by 3 minute wash with 30% B before reconditioning for 6 minutes using 90% B. Samples acquired on the AB QTRAP 6500 were resolved using a binary gradient of 10 mM ammonium acetate in water (pH 9.8 adjusted with ammonium water; mobile phase A) and 90% ACN:10% mobile phase A (mobile phase B), with gradient elution as follows: 0–15 minutes, linear gradient 100–33% B, followed by 3 minute wash with 33% B before reconditioning for 6 minutes using 90% B. A flow rate of 0.25 mL/minute was used with injection volume 10–20 μL. MRM data was acquired with Analyst 1.6.3 software (SCIEX).

For samples acquired on the AB QTRAP 5500 in which isotopologue species of GAA and creatine pathway-associated metabolites were quantified, the MRMs used for each metabolite were as follows: For GAA, M+0 (Q1/Q3: 118/76, CE: 16), M+1 (Q1/Q3: 119/76 and 119/77, CE: 16), M+2 (Q1/Q3: 120/76, 120/77, 120/78; CE: 16), M+3 (Q1/Q3: 121/76, 121/77, 121/78, CE: 16), internal standard M+3 (Q1/Q3: 121/79, CE: 16); for arginine, M+0 (Q1/Q3: 175/70, CE: 25), M+1 (Q1/Q3: 176/70, 176/71; CE: 25), M+2 (Q1/Q3: 177/70, 177/71, 177/72; CE: 25), M+3 (Q1/Q3: 178/70, 178/71, 178/72, 178/73; CE: 25), M+4 (Q1/Q3: 179/70, 179/71, 179/72, 179/73, 179/74; CE: 25); for ornithine, M+0 (Q1/Q3: 133/70, CE: 25), M+1 (Q1/Q3: 134/70, 134/71; CE: 25), M+2 (Q1/Q3: 135/70, 135/71, 135/72; CE: 25); for creatine, M+0 (Q1/Q3: 132/90, CE: 34), M+1 (Q1/Q3: 133/90, 133/91; CE: 34), M+2 (Q1/Q3: 134/90, 134/91, 134/92; CE: 34), M+3 (Q1/Q3: 135/90, 135/91, 135/92, 135/93; CE: 34); for creatinine, M+0 (Q1/Q3: 114/86, CE: 15), M+1 (Q1/Q3: 115/86, 115/87; CE: 15), M+2 (Q1/Q3: 116/86, 116/87, 116/88; CE: 15), M+3 (Q1/Q3: 117/86, 117/87, 117/88, 117/89; CE: 15); and for phosphocreatine, M+0 (Q1/Q3: 210/79, 210/97; CE: −26), M+1 (Q1/Q3: 211/79, 211/97; CE: −26), M+2 (Q1/Q3: 212/79, 212/97; CE: −26), M+3 (Q1/Q3: 213/79, 213/97; CE: −26).

Peaks for full scan data were integrated using El-Maven 0.12.0 software (Elucidata) using a targeted in-house spectral library detecting metabolites using precursor and product ion spectra¹⁰⁵, and peaks for GAA tSIM data were integrated using a targeted method in TraceFinder 5.1 SP2 (Thermo Fisher OPTON-31001). Peaks from data obtained on the AB QTRAP 5500 and 6500+ were integrated using MultiQuant v2.1.1 (Sciex) software. For tracing experiments acquired on the AB QTRAP 5500, peak areas for each MRM corresponding to a given isotopologue species, as outlined above, were summed to yield the total abundance of each isotopologue. Total ion counts for full scan data were quantified using TraceFinder 5.1 SP2 software (Thermo Fisher OPTON-31001, RRID: SCR_023045). Peaks were normalized using probabilistic quotient normalization¹⁰⁶ and total ion counts using R⁴² (RRID: SCR_001905) or by normalization to an internal standard. For stable isotope tracing studies, correction for natural abundance of metabolite labeling was performed using AccuCor 0.3.0⁸⁶ (RRID: SCR_023046) in R (RRID: SCR_001905) or by manual correction.

Tissue collection, processing, and extraction – glioma metabolomics Validation Cohort 1

De-identified tumor and matched normal brain tissue specimens were collected intraoperatively, immediately snap-frozen in liquid nitrogen, and stored at −80°C until shipment. Approximately 10 mg of frozen tissue per sample was submitted to Metabolon, Inc. (Morrisville, NC) for global metabolomic profiling. Recovery standards were added to each specimen prior to extraction to enable quality control monitoring of instrument performance and data normalization. Protein was precipitated by vigorous shaking with methanol, followed by centrifugation to recover chemically diverse metabolites. The resulting supernatant was split into multiple fractions optimized for four parallel UPLC-MS/MS methods: (1) reverse-phase positive ion electrospray ionization (ESI) optimized for hydrophilic species, (2) reverse-phase positive ion ESI optimized for hydrophobic species, (3) reverse-phase negative ion ESI, and (4) hydrophilic interaction chromatography (HILIC) negative ion ESI. A fifth aliquot was reserved as backup. Following drying under nitrogen, extracts were concentrated with a TurboVap^® and maintained under inert gas overnight until analysis by MS.

Each fraction was reconstituted in a solvent system compatible with the intended UPLC-MS/MS method, each containing internal standards at fixed concentrations to verify injection and chromatographic consistency. Analyses were performed on a Waters ACQUITY UPLC system coupled to a Thermo Fisher Q-Exactive high-resolution mass spectrometer (Orbitrap, 35,000 FWHM resolution) equipped with a heated electrospray ionization (HESI-II) source. For reverse-phase methods, extracts were eluted from a Waters BEH C18 column using gradients of water, methanol, and acetonitrile containing 0.05% perfluoropentanoic acid and 0.1% formic acid. HILIC separations employed a BEH Amide column with water/acetonitrile and 10 mM ammonium formate at pH 10.8. MS analysis alternated between full MS and data-dependent MSⁿ scans across 70–1000 m/z, with dynamic exclusion applied to enhance coverage. Metabolites were identified by comparison with Metabolon’s spectral library, using retention time, accurate mass, and MS/MS data. Relative quantitation was performed by peak area integration, with values normalized to tissue mass. Missing data were imputed with the minimum observed value. Quality control included pooled technical replicates, blanks, and internal standards in every run. Instrument performance and reproducibility were monitored by calculating relative standard deviations, with experimental samples randomized and QC injections interspersed.

Absolute quantification of GAA in cells, conditioned media, and tissues

Metabolite preparation for absolute quantification of GAA in cells, media, and tissue was performed as described above. Standard curves for absolute quantification were generated using MultiQuant v2.1.1 or GraphPad Prism 10.4.1 by regression with 1/x² weighting. Parallel cultures for each cell line evaluated were used to obtain cell counts and average cellular diameter values using a Vi-CELL XR cell viability analyzer (Beckman Coulter). Total cellular volume was calculated using the formula $V=\frac{4}{3}\mathrm{*}\pi \mathrm{*}n\mathrm{*}{\left(\frac{d}{2}\right)}^3$, where $V$ represents cell volume per μL acetonitrile extract, $d$ = average cell diameter, and $n$ = number of cells per μL extract. Media GAA content was calculated as $GA{A}_{\mathit{secreted}}=\frac{[GAA{]}_{CMExtract}\mathrm{*}D\mathrm{*}V-[GAA{]}_{UCMExtract}\mathrm{*}D\mathrm{*}V}{n}$, where [GAA] is the calculated concentration of GAA in conditioned media (CM) and unconditioned media (UCM) extract, $D$ is the dilution factor of media in acetonitrile, $V$ is the volume of media in which cells were incubated for the assay, and $n$ represents the cell count in thousands. For quantification of media GAA secretion, a signal:noise threshold of conditioned > 2.5 × unconditioned media was applied to all samples, values below which were considered undetectable. For tissue samples, volumes were calculated using tissue weights and a previously published value of brain tissue density¹⁰³.

LC-MS/MS assays for absolute quantification of GAA and creatine in tumor xenograft tissue

LC-MS/MS assays for the quantification of GAA and creatine in AGAT KO tumor studies were developed and validated according to the FDA guidelines for bioanalytical method validation¹⁰⁷. Briefly, a LC-MS/MS system consisting of a Thermo Fisher Vanquish UPLC and Thermo Fisher TSQ Quantis Plus that was equipped with a heated ESI (HESI) source was used. The SRM transitions used for quantitation are as follows: for GAA, m+0 (Q1/Q3: 118.1/76.0), m+2 ISTD (Q1/Q3: 120.1/78.0); for creatine, m+0 (Q1/Q3: 132.1/90.0), m+5 ISTD (Q1/Q3: 137.1/95.2). 50 μL of plasma or brain tissue was protein precipitated using 450 μL of methanol containing the internal standards. The sample was vortexed and then centrifuged at 10,000 × g for 10 minutes before the supernatant was transferred and analyzed by LC-MS/MS. Chromatographic separation of the samples was accomplished with a Waters Acquity BEH c18 (2.1 × 100 mm, 1.7 μm) column with an isocratic elution using 80:20 (A:B, v:v) of water with 0.1% formic acid (A) and methanol (B) at a flow rate of 0.25 mL/minute. The total runtime was 3 minutes.

A standard curve was created for each run with duplicate QCs (low, medium, and high) to ensure the results for each day were accurate. The linear range of the standard curves for GAA and creatine in plasma was 50–25,000 ng/mL and 1–25,000 ng/mL in brain tissue. For both matrices, each analyte showed excellent linearity (r² ≥ 0.9955). All FDA guideline criteria were met for each analyte in brain tissue. High analyte recovery was demonstrated (94% to 112%), and the intra- and inter- day accuracy (±11.2%) and precision (±8.7%) for quality controls were excellent. Other criteria, including dilution (±9.6%), stability after three 24 hour freeze thaw cycles (±4.7%), autosampler stability for 72 hours (±10.2%), bench top stability at RT for 4 hours (±5.3%), and matrix effect (±11.0%) for GAA and creatine also passed.

Immunoblot analysis of protein expression

Cells were lysed in EBC lysis buffer containing a protease inhibitor cocktail (Millipore Sigma 11836153001). Lysates were resolved by SDS-PAGE and transferred to nitrocellulose membranes (Bio-Rad 1620112 or Bio-Rad 1704158). Primary antibodies used included: anti-GAPDH (D16H11, Cell Signaling Technology 5174S, 1:5000, rabbit monoclonal, RRID: AB_10622025), anti-vinculin (Sigma V9131, 1:100,000, mouse monoclonal, RRID:AB_477629), anti-AGAT (Abcam 119269, 1:1000, mouse monoclonal, RRID: AB_10902241), anti-GAMT (PTG 10880–1-AP, 1:1000, rabbit polyclonal, RRID: AB_2109304). HRP-conjugated secondary antibodies used included: anti-Mouse IgG (Thermo Fisher 31430, 1:2,000, goat polyclonal, RRID: AB_228307) and anti-Rabbit IgG (Thermo Fisher 31460, 1:2,000, goat polyclonal, RRID: AB_228341). Densitometry analysis was performed in ImageJ.

ATAC-sequencing and transcription factor footprinting analysis of GATM

Omni-ATAC-seq was performed as described by Corces et al¹⁰⁸. 50,000 cells were collected and DNase treated for 30 minutes at 37°C prior to nuclei extraction and transposition. After transposition, cleanup was performed using Zymo Clean and Concentrator-5 Kit (Zymo Research D4013) and DNA was eluted in 21 uL of elution buffer. To add adaptor sequences, 5 cycles of amplification were performed using NEBNext 2× Master Mix and qPCR (New England Biolabs M0541S) was used to determine the addition of 0–3 cycles based on sample amplification profiles. Samples were purified using Zymo Clean and Concentrator-5 Kit. Libraries were sequenced on the NextSeq 500 (Illumina, RRID:SCR_014983) using paired-end sequencing of 75 base pair reads by the Molecular Biology Core Facilities at Dana Farber Cancer Institute.

Fastq files associated with NHA, AGAT-low, and AGAT-high GSCs were analyzed using the Nextflow-based BICF Astrocyte 2.2.0 ATAC-seq analysis workflow from the UT Southwestern Bioinformatics Core. Briefly, adaptors were trimmed with TrimGalore⁸⁸, and reads mapped with BWA⁸⁹, filtered with SamTools⁹¹, and sorted with Sambamba⁹⁰. Duplicates were then marked with Sambamba, reads filtered with SamTools, and percentage of reads in mitochondria calculated and library complexity calculated with SamTools and bedtools⁹². Cross-correlation was calculated using PhantomPeakQualTools^93,109, peaks were called with MACS2⁹⁴ from overlaps of pooled replicates, and consensus peaks were called and annotated from ChipSeeker output files.

Footprinting analysis was performed using TOBIAS through UT Southwestern BioHPC. Peaks were filtered for only Chr peaks and to remove blacklisted peaks. .narrowPeak output from the above ATACseq workflow was then merged from all samples and replicate bam files were merged with SamTools. Bam files were then merged according to AGAT expression and malignancy status.

Splice isoform analysis of GATM

Isoform-specific GATM gene expression data were generated by reanalysis of RNAseq fastq files^6,42 using the hg38 build through Astrocyte 2.2.0 to derive relative expression of transcript variants. Differential gene expression analysis was then performed through Astrocyte 2.2.0.

Single cell RNA sequencing analysis of human GBM samples

Human GBM single cell RNA sequencing analysis was conducted using a previously published dataset (GSE274546)⁷⁶. Single cell UMI count files from all primary tumors were downloaded from the NCBI Sequence Read Archive (SRA) and processed using the R package Seurat. Cells with less than 1000 or more than 7000 genes, or with greater than 3% mitochondrial fraction were excluded from further analysis. Subsequent steps including normalization, finding highly variable genes, scaling, cell clustering and UMAP construction were performed. Marker genes for each Seurat cluster were identified by using the Seurat “FindAllMarkers” function, cell clusters were annotated by marker gene expression as in the original publication⁷⁶. Gene expression was visualized by using the Seurat “VlnPlot” function.

Single cell RNA sequencing analysis of mouse xenograft samples

TS516 AGAT WT and AGAT KO intracranial xenografts were established via intracranial injection of 1,000 cells in 13-week old NSG mice at the following coordinates from bregma: 2.0 mm lateral, 0.5 mm anterior, and 3 mm deep from the surface of the brain. 31 days after intracranial injections of either TS516 AGAT WT or TS516 AGAT KO cells, mice were euthanized, cardiac perfusion with phosphate-buffered saline (PBS) was performed and whole brain was extracted. Xenograft brain tissues were dissected and immediately dissociated to single cell suspension using the neural tissue dissociation kit (Miltenyi Biotec 130-092-628). Cells were stained with calcein violet (Invitrogen, C34858) and ethidium homodimer-1 (Invitrogen, E1169) for 15 mins and sorted on a FACS Aria II SORP (4 lasers) from BD Biosciences. Live cells (calcein violet positive, ethidium homodimer-1 negative) were isolated for downstream sequencing analyses. Single cell RNAseq libraries were generated using the using the 10X Genomics Chromium Controller following the manufacturer’s protocol for the Chromium Next GEM Single Cell 3ʹ Kit v3.1 (10X Genomics, PN-1000269). Generated libraries were quantified by Qubit DNA high sensitivity assay (Thermo Fisher Scientific, Q33231) and fragment size was determined by Agilent Tapestation high sensitivity D5000 tapes. Library sequencing was performed at the Children’s Research Institute’s Sequencing Facility at UT Southwestern Medical Center.

Raw sequencing reads were aligned and quantified using Cell Ranger (v10.0.0) against a combined human-mouse reference (refdata-gex-GRCh38_and_GRCm39–2024-A). For each sample, the filtered feature-barcode matrix was analyzed with the Read10X function with Seurat (V5) in R (v4.4.3). Two single-cell objects were generated per sample: a human (tumor) object restricted to barcodes classified as GRCh38 and a mouse (stromal microenvironment) object restricted to barcodes classified as GRCm39. Per-sample Seurat objects were created with the CreateSeuratObject function. Human cells with fewer than 400 or more than 7,500 genes, or with greater than 5% mitochondrial fraction were excluded from further analysis. Mouse cells with fewer than 200 or more than 4,000 genes, or with greater than 10% mitochondrial fraction were excluded from further analysis. For both human and mouse datasets, genes detected in fewer than three cells were removed. All post-QC objects were merged and processed with the standard Seurat workflow (NormalizeData, FindVariableFeatures, ScaleData, RunPCA, RunUMAP). For final integration, post-QC objects were merged and normalized with SCTransform, and PCA was computed on the SCT-normalized expression matrix. Sample-level technical batch effects were corrected with Harmony, applied through Seurat V5’s IntegrateLayers function. Human and mouse compartments were integrated independently. The mouse cell clusters were annotated by marker gene expression as in our previous publication⁷.

Single-cell scoring was computed using the AddModuleScore_UCell function from the UCell R package (v2.10.1). The Molecular Signatures Database (MSigDB) gene signatures were used to calculate GABAergic signaling scores in neurons (“GOBP_SYNAPTIC_TRANSMISSION_GABAERGIC”), astrocyte activation scores in astrocytes (“GOBP_ASTROCYTE_ACTIVATION”), and oligodendrocyte myelination scores in oligodendrocytes (“GOBP_REGULATION_OF_MYELINATION”).

RNA sequencing and transcriptional subtype classification of human primary glioma samples

Total RNA from primary human glioma samples was extracted using RNeasy Plus Universal kit (Qiagen 73404) according to the manufacturer’s instructions. RNA libraries were generated using Kapa RNA HyperPrep Kits with RiboErase (HMR) (Roche KK8561) following manufacturer instructions. Briefly, all RNA samples underwent rRNA depletion, RNAse H and DNase treatment and fragmentation. The RNA fragments were used as template for cDNA synthesis. The cDNA fragments went through the ligation process with molecular identifier adapters synthesized by IDT. The products were purified and enriched with PCR amplification to create the final cDNA library. The library quality was verified on Agilent 2100 Bioanalyzer and sequenced on Illumina NovaSeq 6000 sequencer platform using 150 bp paired-end protocol at the UT Southwestern Genomics & Microarray core.

Sequencing adapters from raw reads were trimmed using Trim Galore (v0.6.4). Trimmed reads were aligned to the human genome (hg38) using STAR (v2.7.3a). Mapped reads were quantified using featureCounts from the Subread package (v1.6.3). Transcriptional subtype classification was done by scoring each glioma samples with gene signatures^25,110 of each GBM subtype by the ssGSEA method from the R package GSVA.

Immunohistochemistry of human brain samples

H&E and immunohistochemistry procedures for the spatially-resolved human glioma case (patient ID #5852) were performed at HistoWiz, Inc, using the Leica Bond RX automated stainer (Leica Microsystems), a Standard Operating Procedure, and a fully automated workflow. Samples were fixed in 10% formalin for 24 hours, washed in 70% ethanol, and embedded in paraffin. Samples were shipped as paraffin embedded blocks and sectioned at 4 μm. One section per sample was stained in hematoxylin and eosin (H&E). Immunohistochemistry (IHC) was conducted using a Leica Bond RX automated stainer (Leica Microsystems). HistoWiz in-house murine skeletal muscle and kidney tissue were used as negative and positive controls, respectively, for optimization of IHC staining for both AGAT and GAMT. Tissue slides were dewaxed with xylene and alcohol-based solutions. Heat-induced epitope retrieval (HIER) was performed for 20 minutes using BOND Epitope Retrieval Solution 1 (H1–20; citrate-based buffer, pH 6.0; Leica Biosystems AR9961) for the GATM antibody, and Epitope Retrieval Solution 2 (H2–20; EDTA-based buffer, pH 9.0 AR9640) for the GAMT antibody. Slides were then incubated with anti-GATM (HPA026077, Atlas Antibodies, RRID: AB_1849528) at a 1:100 dilution and anti-GAMT (PA5–119778, Thermo Fisher, RRID:AB_2913350) at a 1:5,000 dilution for 30 minutes at room temperature. Signal detection was carried out using the Leica Bond Polymer Refine Detection Kit (Leica Biosystems DS9800, RRID:AB_2891238), with 3,3’-diaminobenzidine (DAB) as the chromogen and counterstained with hematoxylin following the manufacturer’s protocol. After staining, slides were dehydrated and coverslipped using the Tissue-Tek Film^® Automated Coverslipper (Sakura Finetek). Whole-slide images were acquired at 40× magnification using a Leica Aperio AT2 scanner (Leica Microsystems, RRID:SCR_021256).

Histology and immunofluorescence of murine PDX samples

Tissues were collected from DREADD-expressing mice or animals treated with the GAMT deficiency diet 30 days after intracranial xenograft injection. DREADD-expressing mice were anesthetized with isoflurane, then perfused with 25 mL of cold 1xPBS (Fisher Scientific BP399–20) followed by 25 mL of cold 4% paraformaldehyde in PBS (ThermoScientific J19943.K2) for fixation. Brains were then dissected and postfixed overnight in 4% paraformaldehyde (ThermoScientific J19943.K2) at 4°C. For all other experiments, mouse brain tissue was dissected and fixed in 10% neutral buffered formalin (VWR 95042–908) for 24 hours. Samples were processed for embedding and sectioning by the Histology Core at the University of Pittsburgh or the Tissue Management and Shared Resource at the UT Southwestern Medical Center (UTSW). H&E was performed by the Tissue Management and Shared Resource at the UT Southwestern Medical Center (UTSW), and the slides were imaged using a Hamamatsu Nanozoomer S60 (RRID:SCR_023762) at the UT Southwestern Whole Brain Microscopy Facility.

For immunofluorescence labeling, paraffin-embedded brain was sectioned at 4 μm and deparaffinized in xylene and re-hydrated using graded ethanol series and distilled water. Antigen retrieval was performed by boiling slides in 1 x EDTA buffer, pH 8.5 (Sigma E1161), 1x EDTA buffer, pH 8.0 (Abcam ab93680), or citric acid buffer, pH 6.0 (Fisher Scientific NC9401067) in a microwave oven for 25 minutes, followed by blocking (10% normal donkey serum, 0.4% Triton X-100 in PBS) at room temperature for 1 hour. Slides were incubated at 4 °C overnight in the following primary antibodies: anti-KCC2 (1:500, Millipore Sigma MABN88, RRID: AB_10807704), anti-cFOS (1:250, Cell Signaling Technology 2250, RRID:AB_2247211), anti-cFOS (1:1000, Synaptic System 226009, RRID: AB_2943525), anti-KU80 (1:250, Cell Signaling Technology 2180, RRID: AB_2218736), anti-mCherry (1:1000, ThermoFisher PA5–143590, RRID: AB_2942819), anti-Ki67 (1:100, ThermoFisher 14–5698-82, RRID: AB_10854564), anti-Ki67 (1:100, Abcam ab156956, RRID:AB_2732028), or anti-vGLUT1 (1:500, Millipore Sigma AB5905, RRID: AB_2301751). The following day, slides were incubated at room temperature for 1 hour with secondary antibodies: 1:300 Alexa Fluor 488 AffiniPure donkey anti-rabbit IgG (H+L) (Jackson ImmunoResearch, 711–546-152, RRID:AB_2340619); 1:500 Alexa Fluor 488 donkey anti-chicken IgΥ (H+L) (Invitrogen, A78948); 1:300 Alexa Fluor 555 AffiniPure donkey anti-goat IgG (H+L) (Jackson ImmunoResearch, 705565147, RRID: AB_3095465), 1:500 Alexa Fluor 594 goat anti-rabbit IgG (H+L) (Invitrogen, A11012, RRID:AB_2534079); 1:300 Alexa Fluor 647 AffiniPure donkey anti-guinea pig IgG (H+L) (Jackson ImmunoResearch, 706–606-148, RRID:AB_2340477), 1:300 Alexa Fluor 647 AffiniPure donkey anti-rat IgG (H+L) (Jackson ImmunoResearch, 712606153, RRID: AB_2340696) and 1:500 Alexa Fluor 647 donkey anti-rabbit IgG (H+L) (Invitrogen A32795).

Both primary and secondary antibodies were diluted in PBS containing 10% normal donkey serum (Jackson ImmunoResearch 017000121 or Sigma-Aldrich S30-M) and 0.2% Triton X-100 (Millipore Sigma T8787). After the secondary antibody incubation, the nuclei and slides were stained 5ug/mL DAPI (Cayman Chemical 14285) for 10min at room temperature followed by mounting with ProLong Diamond Antifade Mountant (ThermoFisher P36970) or ProLong Gold Antifade Mountant (ThermoFisher P36934). The images were scanned by Zeiss Axioscan 7 Digital Slide Scanner (RRID:SCR_027284) at the Whole Brain Microscopy Facility or taken by the Zeiss LSM880 with Airyscan (RRID:SCR_020925) Confocal Scanning Microscope in the Quantitative Light Microscopy Core (QLMC) at the UTSW. Quantification of immunofluorescent images was performed using ImageJ 1.43p (Wayne Rasband and contributors, National Institutes of Health, USA) and ZEISS ZEN 3.11 (Carl Zeiss Microscopy GmbH).

Acute brain slice preparation

Acute hippocampal slices used in single-cell electrophysiological recordings in experiments in which slices were treated with GAA, creatine, or GABA in the presence or absence of gabazine were prepared as previously described¹¹¹ from C57BL/6J mice aged 5–7 weeks old. Briefly, mice were decapitated and the brain was rapidly extracted in oxygenated (95% O₂ and 5% CO₂) and ice-cold sucrose-based artificial cerebrospinal fluid (sucrose-aCSF) at pH 7.4 and 300 mOsm containing: 185 mM sucrose (Sigma Aldrich S9378), 25 mM NaHCO₃ (Thermo Fisher, BP328), 2.5 mM KCl (Millipore Sigma, P9333), 25 mM glucose (Millipore Sigma, G8270), 1.25 mM NaH₂PO₄ (Sigma Aldrich S8282), 10 mM MgCl₂ (Millipore Sigma, M9272), and 0.5 mM CaCl₂ (Millipore Sigma, 223506). The brain was dissected, glued to the specimen holder, and acute slices (300 μm) were prepared on a vibratome (Leica VT1200 S). Slices were transferred to oxygenated and heated (32°C) sucrose-aCSF for 30 minutes. After 30 minutes, the water bath temperature was turned off and slices were transferred to oxygenated recording aCSF at pH 7.4 and 300 mOsm containing: 125 mM NaCl (Millipore Sigma S9888), 25 mM NaHCO₃ (Thermo Fisher BP328), 2.5 mM KCl (Millipore Sigma P9333), 10 mM glucose (Millipore Sigma G8270), 2 mM CaCl₂ (Millipore Sigma 223506), and 2 mM MgCl₂ (Millipore Sigma M9272), and allowed to recover for an additional 30 minutes before recordings. Slices were maintained in oxygenated aCSF at room temperature for the rest of the day.

Brains for experiments measuring changes in input resistance in response to GAA in the presence or absence of inhibitors of glutamatergic and GABAergic signaling were prepared as described previously¹¹². Briefly, wild-type male mice on postnatal days 18–23, were anesthetized by intraperitoneal injection of ketamine/xylazine. After toe pinch to confirm sufficient depth of anesthesia, mice were euthanized by rapid decapitation. Acute coronal slices (300 μm thickness) containing somatosensory barrel cortex were generated using a Leica VT1200S Vibratome in a semi-frozen 300 mOsm dissection buffer containing: 110 mM choline chloride (Millipore Sigma, C1879), 25 mM NaHCO₃ (Millipore Sigma, S6014) 25 mM D-glucose (Millipore Sigma, G8270), 11.6 mM ascorbic acid (Millipore Sigma, A92902), 2.5 mM KCl (Millipore Sigma, P5405), 1.25 mM Na₂H₂PO₄ (Millipore Sigma, S9638), 3.1 mM Na-pyruvate (Millipore Sigma, P5280), 7 mM MgCl₂ (Millipore Sigma, M8266), 0.5 mM CaCl₂ (Millipore Sigma, C7902), and 1 mM kynurenic acid (Millipore Sigma, K3375). Brains and slices were continually aerated with 95% O₂ and CO₂ prior to and during the slicing procedure. After slicing, slices were then transferred to 300mOsm normal artificial cerebrospinal fluid (ACSF) solution containing: 125 mM NaCl (Millipore Sigma, S7653), 25 mM NaHCO₃ (Millipore Sigma, S6014), 10 mM D-glucose (Millipore Sigma, G8270), 2.5 mM KCl (Millipore Sigma, P5405), 1.25 mM Na₂H₂PO₄ (Millipore Sigma, S9638), 2 mM MgCl₂ (Millipore Sigma, M8266), and 2 mM CaCl₂ (Millipore Sigma, C7902) to recover at 37°C for 30 minutes, then at 21°C for 30 minutes prior to recording.

Acute brain slices for experiments assessing neuronal network activity in tumor-bearing and contralateral brain hemispheres were obtained from 14-week-old adult mice, 3 weeks following stereotactic xenograft injection. Slices were prepared at a thickness of 300 μm using a VT1000P vibratome (Leica) in oxygenated, ice-cold cutting buffer containing 205 mM sucrose (Thermo Fisher S5), 2.5 mM KCl (Thermo Fisher P330), 1.25 mM NaH₂PO₄ (Thermo Fisher S397), 26 mM NaHCO₃ (Thermo Fisher P233), 10 mM glucose (Gibco 15023-021), 0.5 mM CaCl₂ (Thermo Fisher 012316.A1), and 5 mM MgSO₄ (Thermo Fisher 033337.36). Following dissection, slices were incubated in recording buffer at 34°C for 40 minutes.

Single cell patch-clamp electrophysiological recordings

Slices generated for electrophysiology experiments in which slices were treated with GAA, creatine, or GABA in the presence or absence of gabazine were transferred under an upright microscope (Scientifica) and held under a harp in a chamber. Slices were perfused at 2 mL/minute with oxygenated aCSF heated to 32°C. The CA1 pyramidal cell layer was visually identified under a 4× objective (NA = 0.1, Olympus) and visually targeted for whole-cell patch clamp recordings using a 40× objective (NA = 0.8, Olympus). Electrodes were made from borosilicate glass capillaries (BF150–86-10HP, Sutter Instrument) on a P-1000 micropipette puller (Sutter Instrument). Recording electrodes were filled using a pH 7.3 and 290 mOsm solution containing: 130 mM Cs-methanesulfonate (Millipore Sigma C1426), 10 mM HEPES (Millipore Sigma H3375), 2 mM MgCl₂ (Millipore Sigma M9272), 4 mM ATP (Millipore Sigma A9187), 0.3 mM GTP (Millipore Sigma G8877), 7 mM Phosphocreatine di(tris) (Millipore Sigma P7936), 0.6 mM EGTA (Millipore Sigma E3889), and 5 mM KCl (Millipore Sigma P9333). The electrophysiological signal was amplified using a Multiclamp 700B (Molecular Devices) and digitized at 10 kHz with a Digidata 1550B (Molecular Devices) before being displayed and recorded on a personal computer using Clampex v11.3 (Molecular Devices). After establishing the whole cell configuration, neurons were assessed for electrophysiological parameters and voltage-clamped at 0 mV to begin experiments. Only neurons with a resting membrane potential more hyperpolarized than −55 mV and access resistance < 30 MΩ were included for analysis. The liquid junction potential was not corrected. Recording epochs consisted of a 3 minute baseline with aCSF, followed by application of GAA, creatine, or GABA. GAA, creatine, and GABA were dissolved in aCSF and prepared fresh every day. 10 μM gabazine (EMD Millipore SR95531) was pre-applied and continuously applied in experiments where indicated. aCSF and all compounds were delivered using a gravity-driven perfusion system. Data from hippocampal patch-clamp slice recordings was analyzed in Clampfit v11.4 and Igor Pro v9.05. The area under curve was measured in Clampfit between recording time 4–10 minutes.

Recordings for experiments measuring changes in input resistance in response to GAA in the presence or absence of inhibitors of glutamatergic and GABAergic signaling were conducted at 21°C in a submerged patch-clamp setting and perfused with oxygenated (95% O₂/5% CO₂) ACSF (1mL/minute) as described previously¹¹². Depending on experiment, the ACSF contained the following inhibitors of neurotransmitter-gated ion channels: 10 μM CPP (Hello Bio HB0036), 20 μM DNQX (Tocris 2312), 100 μM PTX (Hello Bio HB0506), 2 μM CGP 52432 (Tocris 1246), and/or 1 μM tetrodotoxin (Hello Bio HB1034). Layer V pyramidal neurons were visualized with differential interference contrast microscopy. Patch pipettes with 5–7 MΩ were pulled from borosilicate glass. Whole-cell recordings were performed on layer 5a pyramidal neurons using a Multiclamp 700A amplifier (Molecular Devices). Internal solution contained: 130 mM K-gluconate (Millipore Sigma P1847), 0.2 mM EGTA (Millipore Sigma E3889), 6 mM KCl (Millipore Sigma P5405), 3 mM NaCl (Millipore Sigma S7653), 10 mM HEPES (Millipore Sigma H4034), 14 mM phosphocreatine-tris (Millipore Sigma P1937), 4 mM Mg-ATP (Millipore Sigma A9187) and 0.4 mM Na-GTP (Millipore Sigma G3776). Recordings were conducted in voltage clamp, holding the membrane potential at −65 mV. Neurons with resting membrane potential >−60mV and series resistance <40 MΩ were included in analysis. Input resistance in voltage clamp was measured with a −10 mV voltage step. For each recording, baseline electrophysiological measurements were obtained, followed by a 100 μM GAA wash-in for 5 minutes. Following 5-minute GAA wash-in, post wash-in measurements were made. Data acquisition and analysis was performed using custom Labview 8.6 software (National Instruments). Input resistance following GAA treatment was normalized to pre-treatment input resistance averages for each neuron.

Multielectrode array recording in tumor-bearing brain slices

Spike counts were recorded from acute tumor-bearing brain slices using a 4096-electrode Complementary Metal–Oxide Semiconductor (CMOS)-based HD-MEA (BioCam DupleX with CorePlate^™ 3D, 3Brain). Prior to use, chips were rinsed three times with deionized water, treated with 70% ethanol, re-rinsed, and preconditioned with PBS.

After incubation at 34°C, slices were incubated an additional 40-minutes at room temperature prior to recording. The recording buffer consisted of 126 mM NaCl (Thermo Fisher P271), 3.5 mM KCl (Thermo Fisher P330), 1.25 mM NaH₂PO₄ (Thermo Fisher S397), 26 mM NaHCO₃ (Thermo Fisher P233), 10 mM glucose (Gibco 15023-021), 1.6 mM CaCl₂ (Thermo Fisher 012316.A1), and 1.25 mM MgSO₄ (Thermo Fisher 033337.36). Spontaneous activity was sampled continuously for 5 minutes at 20 Hz across sequential pharmacological conditions: baseline aCSF and 20 μM GAA. Recordings were acquired in either contralateral left hemisphere or tumor-infiltrated right hemisphere (n=3 brains each).

Spike detection was performed using the Precise Timing Spike Detection (PTSD)¹¹³ algorithm in BrainWave 5 software (3Brain), with a threshold set at 8 standard deviations. Spike counts were analyzed in R (v4.5.0); data were reshaped into long format using tidyverse⁹⁹ tools and log-transformed as log₁₀(spike count + 1).

Brain slice preparation for perforated patch clamp recordings

For perforated patch-clamp experiments, TS516 AGAT WT mCherry xenografts were established via intracranial injection of 1000 cells in 2.5 μL culture media at 2 mm lateral, 0.5 mm posterior, and 2 mm deep from the bregma landmark in NSG mice aged 42–50 days. 26–33 days following xenografting, xenograft-bearing mice were used for acute-slice electrophysiology. Mice were anesthetized by intraperitoneal injection of ketamine/xylazine, then decapitated and submerged in semi-frozen dissection buffer for 30 seconds. Acute coronal slices containing the tumor and its contralateral hemisphere were obtained in a semi-frozen dissection buffer. After slicing, slices were then transferred to oxygenated sucrose-substituted ACSF containing in mM 100 NaCl, 2.5 KCl, 25 glucose, 25 NaHCO₃, 1.25 NaH₂PO₄, 30 sucrose, 2 MgCl₂ and 1 CaCl₂ to recover at 37°C for 30 minutes. After recovery at 37° C, slices were kept at 21°C for another 30 minutes before recording.

Perforated patch recordings of tumor-bearing brain slices

For perforated patch-clamp recording, gramicidin (50μg/ml) (Sigma-Aldrich G5002) was added to the internal recording solution to induce chloride-impermeable pore formation. Also, internal recording solution was filled with Alexa Fluor 488 hydrazide sodium salt (50μM) (Invitrogen A10436) to visualize potential pore rupture under blue field. An access resistance of <100 MΩ was obtained about 20~30 minutes after initial patch formation. To obtain reversal potential, neurons were clamped at −65 mV and voltage steps to −85 to −40 mV with 5mV increments were applied while GAA (10mM) was transiently (5–10 msec) and locally applied through a patch pipette using a Picospritzer III (Parker Instrumentation). The GAA pipette was 10–20 μm from the recorded neuron and aimed at the soma. Tumor cells were visualized by mCherry fluorescence under green field with fluorescence and differential interference contrast microscopy and pyramidal neurons were using visualized differential interference contrast microscopy. GraphPad Prism was used to perform analysis of GAA reversal potentials. Intracellular chloride concentration was derived from reversal potentials using the Nernst equation: $E=\frac{R\mathrm{*}T}{n\mathrm{*}F}\mathrm{*}\mathrm{l}\mathrm{n}(\frac{{\left[C{l}^{-}\right]}_{\mathit{out}}}{{\left[C{l}^{-}\right]}_{\mathit{in}}})$, where $\mathrm{n}=-1$ for ${\mathrm{C}\mathrm{l}}^{-}$.

Culture of neonatal rat cortical neurons

Primary neonatal rat cortical neuron cultures were generated as previously described¹¹⁴. Briefly, Sprague-Dawley (Charles River Laboratories) rat neonates aged P0-P3 were anesthetized using isoflurane and rapidly decapitated. Their brains were then dissected and cortices isolated from other neural tissue in a modified Puck’s dissociation medium D1 containing 5 mM HEPES, 16.5 mM glucose, 22 mM Sucrose, 137 mM NaCl, 0.32 mM Na₂HPO₄, 0.22 mM KH₂PO₄ in deionized water, pH 7.4. Whole rat cortices were then dissociated into single cells in a Puck’s/papain solution [1.5 mM CaCl₂, 0.5 mM EDTA, 0.75% papain (Worthington Biochemical LS003126) and added L-cysteine in D1 medium]. Following dissociation, cells were plated in Rat Neuron Plating Medium consisting of 1x B-27 (Thermo-Fisher 17504044), 5% fetal bovine serum, 2x antibiotic-antimycotic mix (Thermo-Fisher 15240062), 5 mM HEPES, and 1.2 mM L-glutamine in Neurobasal Medium (Thermo-Fisher 21103049), pH 7.4, at 1.25 × 10⁵ cells per well of a 96-well Axion CytoView microelectrode array (MEA) plate (Axion Biosystems M768-tMEA-96). Culture wells were precoated with 1 mg/mL of poly-D-lysine (PDL) in borate buffer, pH 8.4. Dissociated neurons were maintained in a cell culture incubator at 37° C in 5% CO₂ and ambient oxygen. 24 hours after dissociation, media was exchanged for Rat Cell Culture Medium consisting of 1x B-27 (Thermo-Fisher 17504044), 1x antibiotic-antimycotic mix (Thermo-Fisher 15240062), 5 mM HEPES, and 1.2 mM L-glutamine in Neurobasal Medium (Thermo-Fisher 21103049), pH 7.4, after which point half-media changes occurred every 5–6 days until assays were performed roughly 17 days post-dissection.

MEA assay of neonatal rat cortical neuronal network activity

Rat cortical neuronal firing was recorded at a sampling rate of 12.5 kHz using a Maestro Pro MEA system (Axion BioSystems) and the Neural Metrics Tool (Axion BioSystems) to detect neuronal action potentials (spikes), defined as when the recorded trace exceeded a threshold of ±6 standard deviations from the baseline signal. Cortical rat neuron culture media was changed 24 hours prior to experimental treatment and recording. On the day of recording, rat neuron culture medium was exchanged for ACSF consisting of 145 mM NaCl, 2.5 mM KCl, 10 mM HEPES, 2 mM CaCl₂, 10 mM glucose, 200 nM glycine, and 1 mM MgCl₂. A 15 minute recording was then performed to establish a baseline firing rate for each well; electrodes demonstrating a spike frequency <5 Hz were excluded from downstream analysis. Wells were then treated with either PBS, 100 μM 4-aminopyridine (4-AP), or 0.1 μM GAA, and 5 minute recordings were performed at 5 minutes, 2 hours, and 5 hours following treatment. Average neuronal spike rates were calculated for each well using only the active electrodes in each well. All measures of neuronal activity in each well in each experimental condition and at each time point (${\mathrm{x}}_1$) were normalized to the burst rate and firing rate observed during the baseline period (${\mathrm{x}}_0$) using a measure of normalization bounded between −1 and 1: $\left({\mathrm{x}}_1-{\mathrm{x}}_0\right)/\left({\mathrm{x}}_1+{\mathrm{x}}_0\right)$. The average spiking rate across all active electrodes in a given well was then taken as a biological replicate for a given condition at a given timepoint.

Cell growth assay

TS516 and UTSW63 AGAT WT and AGAT KO cells were plated in 6-well ultra-low adherence plates at 2 × 10⁵ cells per well in NeuroCult NS-A Basal Medium (Human) prepared as described above. Cell counts were obtained at 0, 24, 48, 72, 96, and 120 hours following dissociation in Accutase (StemCell 07922) using a Vi-CELL XR cell viability analyzer. Media on remaining wells was changed at 48 and 96 hours. Growth rates were compared using nonlinear regression to fit exponential growth curves in GraphPad Prism 10.4.1.

Xenograft studies of AGAT expression

Mice were randomly assigned to receive intracranial injections of either TS516 AdCRISPR AGAT WT or TS516 AdCRISPR AGAT KO cells, or UTSW63 AGAT WT or UTSW63 AGAT KO cells. Cells were injected in 5 μL NeuroCult NS-A Basal Medium (Human) 0.5–1 mm anterior and 2 mm lateral to the bregma, at a depth of 3 mm from the brain surface. Animals were monitored daily until the appearance of neurological symptoms, at which point they were euthanized.

Stereotactic glioma cell implantation and optical fiber placement for optogenetics studies

Male and female mice were used in cohorts equally. For optogenetic experiments performed in immunodeficient Thy1-ChR2-YFP mice, patient-derived glioblastoma cultures (TS516 AGAT-WT and TS516 AGAT-KO; 1,000 cells per mouse) were injected into the M2 cortex. A single-cell suspension of cell cultures was prepared in sterile culture medium immediately prior to surgery. Animals at P28–P35 were anesthetized with 1–4% isoflurane and positioned in a stereotaxic apparatus. Under sterile conditions, the cranium was exposed via a midline incision, and a 31-gauge burr hole was made at the specified coordinates. For cortical M2 injections, coordinates were: AP = +1.0 mm (from bregma), ML = +0.5 mm, DV = −1.0 mm. Cells were injected using a 31-gauge Hamilton syringe at an infusion rate of 0.5 μl min⁻¹ with a digital pump. At completion of infusion, the syringe needle was allowed to remain in place for a minimum of 5 minutes, then manually withdrawn. For optogenetic stimulation, optical fibers with a 200 μm core diameter (Doric Lenses Inc. MFC_200/240-0.22_4.5mm_MF2.5_FLT) were implanted 0.3mm above the tumor cell injection site. All optical fibers were secured using stainless steel screws (thread size 00–90 × 1/16, Antrin Miniature Specialties AMS90/1P) placed contralateral to the fiber, along with C&B Metabond (Parkell S380) and light-cured dental adhesive cement (Geristore A&B paste, DenMat 34523101).

Optogenetic stimulation of Thy1-ChR2-YFP mice

Optogenetic stimulation was performed two weeks after tumor cell implantation and optic ferrule placement. A 473 nm diode-pumped solid-state laser system was used to stimulate ChR2-expressing cortical neurons. Light pulses (10 mW at the fiber tip) were delivered at 20 Hz for 30 s, followed by a 90 s recovery period, in repeated cycles for a total duration of 30 minutes. Illumination was delivered through a 200 μm core optic fiber (NA = 0.22; Doric Lenses). Potential light-induced heating damage was excluded by histological examination of stimulated tissue after each session. Animals received an intraperitoneal injection of EdU (5-ethynyl-2′-deoxyuridine, 40 mg/kg; Invitrogen E10187) prior to stimulation.

Mouse cortical neuron culture

Embryonic (E19) mouse cortical neuron cultures from C57BL6 mice (Jackson Laboratory) were prepared as described previously¹¹⁵. The cerebral cortex of embryos was dissected under a microscope using aseptic techniques, with tissues from 3–6 embryos pooled and processed together. To isolate cortical neurons, the cortices were minced with scalpels on an ice-cold culture dish and digested using a papain dissociation kit (Worthington Biochemical, LK003150), per the manufacturer’s instructions. Briefly, papain digestion and subsequent protease inhibition were performed at 37 °C for 7 and 3 min, respectively. Cell pellets were washed with Neurobasal medium (Thermo Fisher, 21103049) supplemented with Deoxyribonuclease I (1 mg/mL; Millipore Sigma, DN25). The cell suspension was filtered through a 40 μm filter, counted, and centrifuged. Neuron culture media (NCM) was prepared by mixing Neurobasal medium with 1× B-27 (Thermo Fisher, 17504044), 1% Glutamax (Thermo Fisher, 35050061), 1% sodium pyruvate (Thermo Fisher, 11360070), and 1% Antibiotic-Antimycotic (Thermo Fisher, 15240062). Cells were resuspended in neuron plating media (NPM), prepared by adding 4.5% fetal bovine serum (FBS) to the NCM, then plated onto poly-D-lysine and laminin-coated coverslips (NeuVitro, GG-12–1.5-laminin) at a density of 2.5 × 10⁵ cells per well in 24-well plates. After 24 h, the culture media was replaced with serum-free NCM. Cultures were maintained by replacing half the culture media with fresh NCM every 3–4 days. For neuron-glioma co-culture experiments, AGAT WT cells were added to neurons at a density of 1.5 × 10⁴ cells per coverslip. After 4 days of neuron-glioma co-culture, GAA at 5 μM was added to the treatment wells and was replenished for next 3 days. Following 3 days of GAA supplementation, neuron-glioma co-cultures with and without GAA were fixed for immunocytochemistry.

Immunofluorescence staining of tumor-neuron co-cultures

Following removal of the culture medium and a PBS wash, cells were fixed with ice-cold 4% paraformaldehyde at 4 °C for 30 min. Coverslips were then washed twice with ice-cold PBS and stored at 4 °C until immunofluorescence staining was performed. Staining was carried out in 24-well tissue culture plates with coverslips. Cells were blocked with a buffer containing 0.25% Triton X-100 and 5% normal goat serum in PBS for one hour at RT, then incubated overnight at 4 °C with primary antibodies diluted in fresh blocking buffer. The primary antibodies and their dilutions were as follows: anti-MAP2 (chicken, 1:1000, EnCor, CPCA-MAP2, RRID: AB_2138173), HNA (mouse, 1:500, Millipore, MAB1281, RRID: AB_94090) and rabbit anti-Ki-67 (1:500, Abcam, ab15580, RRID: AB_443209). The next day, following two washes with RT PBS, cells were stained with secondary antibodies prepared in 0.25% Triton X and 1% normal goat serum in PBS. The secondary antibodies used were Alexa 488 goat anti-chicken IgG (Thermo Fisher A-11039, RRID: AB_142924), Alexa 568 goat anti-mouse IgG (Thermo Fisher A-11004, RRID: AB_2534072) and Alexa 647 goat anti-rabbit IgG (Thermo Fisher A-21245, RRID: 2535813) all used at 1:500 in antibody diluent solution for 1 h at room temperature. After 1 hour incubation in secondary antibody, cells were washed twice with PBS and mounted cell-side down on glass slides using mounting media with DAPI (H-1200).

In vitro chloride imaging

Imaging of intracellular Cl⁻ in embryonic (E19) mouse cortical neurons co-cultured with AGAT WT GBM cells grown on poly-D-lysine and laminin-coated coverslips was performed with the fluorescent chloride-sensitive indicator MQAE [N-(Ethoxycarbonylmethyl)-6-Methoxyquinolinium-Bromide; Thermo Fisher E3101], as previously described^116,117. The MQAE dye detects Cl⁻ ions via diffusion-limited collisional quenching, resulting in a concentration-dependent decrease of fluorescence emission following an increase in Cl⁻ concentration¹¹⁷. Therefore, a decrease in MQAE fluorescence is indicative of a higher [Cl⁻]_i and vice versa. On the day of imaging, neuron-glioma co-cultures were incubated in Tyrode solution (136 mM NaCl, 5 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂, 11 mM glucose, 10 mM HEPES, pH 7.4) containing 5 mM MQAE for 90 minutes at 37°C in a humidified incubator with 5% CO₂. After incubation, cells were washed once with Tyrode solution to remove excess dye and immediately imaged. Images were acquired with a Zeiss spinning disk microscope equipped with a 10× air-objective (NA 0.75). MQAE was excited with a 405 nm diode laser and fluorescence collected with a 525/50 nm band-pass emission filter. All excitation and acquisition parameters (laser intensity, PMT offset and gain) were kept constant throughout experiments. Image analysis was performed with Zen 3.11 software by measuring the mean fluorescent intensity of ROIs placed on the cell body of individual neurons from 6 randomly selected fields for each coverslip.

Flow cytometry EdU incorporation assay

TS516 AGAT WT mCherry cells were plated at a density of 3 × 10⁴ cells per well on Neuvitro laminin-coated coverslips (NeuVitro, GG-12–1.5-laminin) in 1 mL Neuron Culture Media in a 24-well plate. In wells in which TS516 cells were co-cultured with HMC3 microglia cells, HMC3 microglia were plated simultaneously with TS516 cells at a density of 1.5 × 10⁴ cells per well. GAA was added to GAA treated wells from a 5 mM stock in DPBS for a final concentration of 5 μM. Media was changed every 24 hours. Cells were incubated in the presence or absence of GAA for 72 hours. 2 hours prior to collection, EdU was added to a final concentration of 10 μM. Following cell collection, cells were fixed, permeabilized, and EdU detected using the Invitrogen Click-iT Plus EdU Alexa Fluor 488 Flow Cytometry Assay Kit (Thermo-Fisher C10632). TS516 cells were gated for mCherry positivity and % EdU incorporation analyzed as a fraction of the mCherry population using a FACS LSRFortessa (BD Biosciences) in the UT Southwestern Flow Cytometry Core. Data were processed using FCS Express software (Dotmatics).

Xenograft studies of dietary GAA modulation

Mice were randomly assigned to receive one of the following diets: Amino Acid Diet (Inotiv TD.01084) or Arginine-Deficient, 0.665% Ornithine Diet (GAMT Deficiency Diet, Inotiv TD.240691). Cells were injected in 5 μL NeuroCult NS-A Basal Medium (Human) 0.5–1 mm anterior and 2 mm lateral to the bregma, at a depth of 3 mm from the brain surface. Mice were monitored daily for the development of neurological symptoms, at which point they were euthanized.

Human induced pluripotent stem cell culture

Reprogrammed feeder-independent iPS cells (Wisconsin International Stem Cell Bank, Wicell Research Institute, iPS (IMR-90)-4 cells) (RRID: CVCL_C437) were grown on Matrigel (Corning 354230) and maintained in culture in mTeSR Plus media (STEMCELL Technologies 100-0276). hiPSCs were passaged with Gentle Cell Dissociation Reagent (GCDR) (STEMCELL Technologies 100-0485) every 2–3 days when they reached a confluence of 60–70%.

Generation of cerebral organoids from hIPSCs

The STEMdiff Cerebral Organoid Kit (containing Induction Medium, Expansion Medium, and Maturation Medium) (STEMCELL Technologies 08570) was used for cerebral organoid differentiation. hiPS cultures at 70–80% confluence were dissociated into single cells using Accutase (STEMCELL Technologies 07922). iPS cells were seeded in the AggreWell 800 microwell culture plates (STEMCELL Technologies 34811) with mTeSR Plus medium in the presence of 10μM Rho-associated protein kinase (ROCK) inhibitor (STEMCELL Technologies 72304) for embryoid body (EB) formation. Half of the media was changed in the AggreWells for the first 5 days. On day 5, the medium was switched to induction medium to induce neuroectoderm formation. On day 7, EBs were transferred from the AggreWells into Matrigel (Corning 356234) droplets by pipetting into cold ECM gel on a sheet of Parafilm with small 3 mm dimples. The Matrigel droplets containing the EBs were allowed to polymerize at 37°C for 10 minutes, then transferred into a six-well ultra-low attachment plate containing expansion medium for neural tube formation up to day 10. On day 10, the medium was switched from expansion medium to maturation medium for cortical layer formation, and the plate was placed on an orbital shaker. From day 10 to day 40, organoids were cultured in maturation media with media changes every other day.

Co-culture of iPSC-derived cerebral organoids and TS516 neurospheres

Three-month-old mature cerebral organoids were placed individually into a 96-well round-bottom plate. A single TS516 AGAT WT GFP neurosphere of approximately 300 μm was added to each well. Plates were centrifuged at 100 × g for 2 minutes to induce contact between the neurosphere and the organoid. The co-cultures were maintained in the conditions for 48 hours to ensure stable attachment. The attached co-cultures were transferred to an ultra-low attachment 24-well plate with media changes daily. The pairs were treated with either 5 μM GAA or vehicle control for the next 14 days.

Organoid-neurosphere co-culture imaging and migration analysis

Before imaging, NucBlue Live ReadyProbes Reagent (1:50, Thermo Fisher R37605) was added to the media to stain the organoid-neurosphere co-cultures. Imaging was performed using a confocal microscope every 7 days. Migration was calculated by measuring the distance from the point of attachment of the neurosphere to the farthest point reached by the GFP-positive tumor cells along the surface of the organoid. Three measurements were obtained for each co-culture using the point-to-point length measurement function in the NIS Elements AR software, and the mean was used as the migration distance. After imaging on day 14, co-cultures were fixed and paraffin-embedded for histological analysis at the UPMC Histology Core facility. Hematoxylin and eosin staining (H&E) was performed, and slides were imaged via confocal microscopy.

3D invasion assay of TS516 neurospheres

3D invasion assays were performed as previously described with minor modifications¹¹⁸. Briefly, neurospheres were generated by seeding 2,000 cells in 100 μL of Neuro-cult medium in ULA 96-well round-bottom plates (Corning) three days prior to the assay. On day −1, spheres were collected and their diameters were measured; spheres of approximately 200 μm were selected for subsequent experiments and transferred to a well of a ULA 96-well round-bottom plate containing 50 μL of culture medium. An equal volume (50 μL) of Matrigel (Corning) was then added to each well, and plates were incubated at 37°C, 5% CO₂ for 1 hour to allow gel polymerization. After polymerization, 50 μL of culture medium was added to each well. Images were captured at 0 h and 72 h using a Nikon microscope. Image analysis was performed by measuring the diameter of the invasive area, and invasion rate was quantified as the difference in total area from day 0 to day 3 as a fraction of the original area of the cultured neurosphere.

3D migration assay of TS516 neurospheres

3D migration assays were performed as previously reported with modifications¹¹⁹. Flat-bottom 96-well plates (Falcon) were coated with Matrigel (Corning) and incubated to allow polymerization. Following coating, 100 μL of culture medium was added to each well. TS516 neurospheres (200 μm), prepared as described above, were transferred into the coated wells by adding 100 μL of spheroid-containing medium to each well (final volume: 200 μL). Images were captured at 0 h and 72 h using a Nikon microscope. Migration rate was quantified as the difference in total area from day 0 to day 3 as a fraction of the original area of the cultured neurosphere.

Secretion of GAA by human tissue explants

Tumor tissue was collected and tissue explants generated as described previously⁴². Briefly, tissue collected from the operating room was suspended in Hibernate A (BrainBits HA) and transferred to RBC Lysis Buffer (Thermo Fisher 00433357) within 30 minutes of excision for 10 minutes with rocking. Tissue was washed three times with Hibernate A supplemented with 2 mM GlutaMAX (Thermo Fisher 35050061), 100 U/mL and 100 μg/mL (respectively) penicillin/streptomycin (Thermo Fisher 15140148), and 250 ng/mL amphotericin B (GeminiBio 400104). Dissection scissors or scalpels were used to parcellate tissues into 1–2 mm³ pieces which were then plated in separate wells of a 24-well ultra-low adherence plate (Corning 3473) in 1 mL HPLM supplemented with 1× B27 supplement minus vitamin A (Thermo Fisher 12587010), 1× N2 supplement (Thermo Fisher 17502048), 100 U/mL and 100 μg/mL (respectively) penicillin/streptomycin (Thermo Fisher 15140148), 250 ng/mL plasmocin (InvivoGen ant-mpp), 55μM 2-mercaptoethanol (Thermo Fisher BP176100), and 2.375–2.875 μg/mL insulin (Millipore Sigma I9278). Tissues were randomized to groups of 1 to 4 technical replicates per tracer from each biologic sample, depending on available tissue; remaining tissue was used for histopathology. After 30 minutes, media was exchanged for 1 mL HPLM as described previously⁴². Tissues were incubated for 18 hours in a 37°C incubator at ambient oxygen and 5% CO2 with shaking. After 18 hours, conditioned media was harvested, snap-frozen in liquid nitrogen, and stored at −80°C until analysis. Tissues were then prepared for LC-MS analysis and quantification of GAA secretion in media as described above. As in quantification of media GAA secretion by GSCs, a signal:noise threshold of GAA concentration in conditioned > 2.5 × unconditioned media was applied to all samples to discriminate GAA-secreting samples. As no cell count information could be obtained for these explants, secreted GAA was rendered as the unconditioned-subtracted GAA content of conditioned media for those samples whose GAA content exceeded this threshold. Secretion values were then normalized to organoid mass.

Histopathology of human explants

Residual human tumor and brain tissues from explant preparation were fixed for 1 hour in neutral buffered 10% formalin solution (Millipore Sigma HT501128). Following fixation, tissues were washed and stored in 70% ethanol. Tissues were embedded in paraffin, sectioned, and stained with hematoxylin and eosin (H&E) by the University of Texas Southwestern Histo Pathology Core. H&E sections were reviewed by a board-certified neuropathologist (T.E.R).

Ex-vivo immune phenotyping assay

C57BL/6 mice were implanted with CT-2A (RRID:CVCL_ZJ44) tumors as done previously¹²⁰. When mice were moribund, TILs were isolated using a Percoll gradient (Cytiva 17089101) and plated in RPMI media +/− 1 mM GAA (Sigma-Aldrich). After 72 hours flow cytometry was performed as previously described¹²⁰. The following two flow cytometry antibody panels were used, with clones in parentheses: T-cell Panel: PD1-FITC (J43) (eBioscience 11–9985-82; RRID:AB_465472), CD62L-PercpCy5.5 (MEL-14) (BioLegend 104431; RRID:AB_2187123), Foxp3-PacBlue (FJK-16s) (Fisher Scientific 48–5773-82; RRID:AB_465935), CD45-BV510 (30-F11) (BioLegend 103137; RRID:AB_2561392), CD8a-BV605 (53–6.7) (BioLegend 100743; RRID: AB_2561352), CD4-BV711 (GK1.5) (BioLegend 100447; RRID:AB_2564586), Granzyme B-AF647 (GB11) (BioLegend 515405; RRID:AB_2566333), CD25-AF700 (PC61) (BioLegend 102024; RRID:AB_493709), L/D-APC-Eflour 780 (Thermo Fisher Scientific 65-0865-14), CD44-PeCy7 (IM7) (BioLegend 103029; RRID:AB_830786), CD11b-BUV395 (M1/70) (BD Biosciences 565976; RRID:AB_2738276); Myeloid Panel: Ly6G-FITC (1A8) (BioLegend 127605; RRID:AB_1236488), MHCII-PCP (M5/114.15.2) (BioLegend 107625; RRID:AB_2191072), TGFβ/LAP-BV421 (TW7–20B9) (BioLegend 141407; RRID:AB_2561580), CD19-BV711 (6D5) (BioLegend 115555; RRID:AB_2565970), P2RY12-APC (S16007D) (BioLegend 848005; RRID:AB_2721468), Ly6C-AF700 (HK1.4) (BioLegend 128023; RRID:AB_10640119), L/D-APC-Efluor780 (live/dead, no clone) (Thermo Fisher Scientific 65-0865-14), PDL1-PeCy7 (10F.9G2) (BioLegend 124313; RRID:AB_10639934). Intracellular staining was performed using the eBioscience^™ Intracellular Fixation & Permeabilization Buffer Set (Thermo Fisher 88–8824-00). Samples were run on a BD FACS Symphony Analyzer, and data were analyzed using FlowJo software (Becton Dickinson).

In-vitro myeloid assay

Bone marrow cells were isolated from naïve mice and differentiated into: BMDMs using recombinant 40ng/ml M-CSF (PeproTech 315-02) for 6 days¹²¹, TAMs using 40ng/ml M-CSF in 50:50 complete RPMI and conditioned media from CT-2A cells cultured for 72 hours following plating at 2E6 cells¹²² starting 3 days after differentiation, or MDSC using 40ng/ml GM-CSF (PeproTech 315-03) for 6 days while being cultured with 50% CT-2A conditioned media for 6 days¹²³. All cells were then treated with 1 mM GAA for 48 hours before flow cytometry analysis. The following panel was used, clones in parentheses. All antibodies were obtained from BioLegend. In-vitro myeloid panel: Ly6G-FITC (1A8) (BioLegend 127605), MHCII-PercpCy5.5 (M5/114.15.2) (BioLegend 107135), TGFβ/LAP-BV421 (TW7– 20B9) (BioLegend 141407), CD45-BV510 (30-F11) (BioLegend 103137), CSF1R-BV605 (AFS98) (BioLegend 135513), CD80-BV711 (16–10A1) (BioLegend 104739), Ly6C-AF700 (HK1.4) (BioLegend 128023), live/dead-APC-Cy7 (BioLegend 423105), CD86-PE (GL1) (BioLegend 105007), PDL1-PeCy7 (10F.9G2) (BioLegend 124313), and CD11b-BuV396 (M1/70) (BD Biosciences 563553). Samples were run on a BD FACS Symphony Analyzer, and data were analyzed using FlowJo software (Becton Dickinson).

In-vitro T cell assay

T cells were isolated from the spleens of naïve mice using the EasySep Mouse T Cell Isolation Kit (Stemcell 19851). The T cells were then labeled with the CellTrace Violet Cell Proliferation Kit (Thermo Fisher C34557), and activation was done using ImmunoCult Mouse T Cell Activator Kit (Stemcell 100–1572), supplemented with 20 IU murine IL-2 (PeproTech 212–12). GAA (1 mM or 10mM) was added at the time of stimulation. The cells were incubated in either 21% O₂ or 1% O₂ in a Heracell Tri-Gas CO₂ incubator (Fisher). For flow cytometry, the cells were stained with CD4-PE-Cy7 (BioLegend), CD8-BV605 (53–6.7) (BioLegend 100743; RRID: AB_2561352), and CD44-PerCP-Cy5.5 (BioLegend). Samples were run on a BD FACS Symphony Analyzer, and data were analyzed using FlowJo software (Becton Dickinson).

QUANTIFICATION AND STATISTICAL ANALYSIS

Analysis of nitrogen metabolism profiling platform data

Analysis of nitrogen metabolism profiling platform data was performed as previously described⁴². Briefly, integrated peaks from LC-MS analysis of nitrogen metabolism profiling platform data were analyzed using R (RRID: SCR_001905). Metabolites which were not quantified in more than one sample, had a mean total pool size below a threshold of 1 × 10⁶, had calculated total labeling across all conditions of less than 1%, had quantified labeling of >4% in any unlabeled sample, had calculated total labeling across all conditions of greater than 500% were filtered and removed from subsequent analysis. Differential labeling scores were calculated using the formula: $\left|{\text{ln}}^{fractionalenrichmen{t}_{condition1}}/fractionalenrichmen{t}_{condition2}\right|\times \left(fractionalenrichmen{t}_{condition1}-fractionalenrichmen{t}_{condition2}\right)$. For visualization, metabolites with less than 5% total labeling were filtered and removed from Sankey diagrams.

TCGA and GTEX RNA sequencing analysis

Results are based in part upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga¹²⁴. RNA sequencing data for patient glioblastoma samples were downloaded from the TCGA portal (https://portal.gdc.cancer.gov/projects/TCGA-GBM), comprising 173 samples including 154 index tumor samples, 14 recurrent tumor samples, and 5 normal controls sequenced by Illumina HiSeq 2000 RNA Sequencing. Of these, 151 primary GBM tissue samples with accompanying survival and prognostic information were selected for analysis. RNA sequencing data from non-malignant brain tissue samples were downloaded from the GTEx portal (https://gtexportal.org, file version:2017-06-05_v8_brain_cortex, accession date August 26, 2019)^125,126. Gene expression values for GATM, GAMT, CKB, CKMT1A, CKMT2A, and CKMT2B were first converted to Transcripts Per Million (TPM) to enable comparison across samples. A log₂(TPM + 1) transformation was then applied to stabilize variance and normalize the distribution. For each gene, Z-score transformation was performed across all samples to standardize expression levels and highlight relative up- or down-regulation. Samples were annotated according to associated histological subtype, IDH mutation status, CIMP status, and MGMT promoter methylation.

RNAseq analysis of GATM and GAMT expression in NSCs and GSCs

RNAseq data were reanalyzed from a publicly available dataset generated by Mack et al.⁷⁷ (GSE119834). Gene expression levels were downloaded as FPKM values as provided in the original dataset, composed of RNAseq values for 9 NSC and 44 GSC cell lines. Transcript abundances in fragments per kilobase million (FPKM) for GATM (AGAT) and GAMT were extracted for each sample; a +0.1 transformation was applied to FPKM values for all samples to facilitate log-scaled data visualization. To evaluate the relative expression between these two genes, ratio of GATM to GAMT expression was calculated for each sample.

Other statistical analyses

Information related to data presentation and statistical analysis for individual experiments can be found in the corresponding figure legends. Pathway and PLS-DA analyses of metabolite profiles of human brain tissue samples were performed using MetaboAnalyst 5.0¹²⁷. Other statistical analyses were carried out using GraphPad Prism software (version 10.4.1). Significance of all comparisons involving two groups was calculated by unpaired two-tailed t-test. For comparisons of two groups with significantly different variances, Welch’s t-test was used. For comparisons of two groups without significant differences in variances, Student’s t-test was used. Mann-Whitney tests or Kolmogorov-Smirnov tests were used for comparisons between groups in which assumptions of normality were not appropriate. Significance of comparisons involving three or more groups was calculated by one-way ANOVA, Welch’s ANOVA, or Friedman test. Outlier analyses were conducted via ROUT method or Grubbs’ test. Survival data from intracranial xenograft studies were analyzed and statistical significance was determined using log-rank tests. For all tests, two-tailed p-values < 0.05 were considered statistically significant.

Key Resources Table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit monoclonal anti-GAPDH (D16H11)	Cell Signaling Technology	Cat# 5174S, RRID: AB_10622025
Mouse monoclonal anti-VINCULIN	Sigma	Cat# V9131, RRID: AB_477629
Mouse monoclonal anti-AGAT	Abcam	Cat# 119269, RRID: AB_10902241
Rabbit polyclonal anti-AGAT	ATLAS	Cat# HPA026077, RRID: AB_1849528
Rabbit polyclonal anti-GAMT	Proteintech	Cat# 10880-1-AP, RRID: AB_2109304
Rabbit polyclonal anti-GAMT	Thermo Fisher Scientific	Cat# PA5-119778, RRID: AB_2913350
Rabbit monoclonal anti-cFOS	Cell Signaling Technology	Cat# 2250, RRID: AB_2247211
Chicken monoclonal anti-cFOS	Synaptic Systems	Cat# 226 009, RRID: AB_2943525
Rabbit monoclonal anti-KU80	Cell Signaling Technology	Cat# 2180, RRID: AB_2218736
Guinea pig polyclonal anti-VGLUT1	Millipore	Cat# AB5905, RRID: AB_2301751
Mouse monoclonal anti-KCC2	Millipore	Cat# MABN88, RRID: AB_10807704
Goat polyclonal anti-mCherry	ThermoFisher	Cat# PA5-143590, RRID: AB_2942819
Rat monoclonal anti-Ki67 (SolA15)	ThermoFisher	Cat# 14-5698-82, RRID: AB_10854564
Rat monoclonal anti-Ki67	Abcam	Cat# ab156956, RRID:AB_2732028
Goat polyclonal anti-Mouse IgG (H+L), HRP	Thermo Fisher Scientific	Cat# 31430, RRID: AB_228307
Goat polyclonal anti-Rabbit IgG (H+L), HRP	Thermo Fisher Scientific	Cat# 31460, RRID: AB_228341
Alexa Fluor 488 AffiniPure donkey anti-rabbit IgG (H+L)	Jackson ImmunoResearch	Cat# 711-546-152, RRID:AB_2340619
Alexa Fluor 488 donkey anti-chicken IgΥ	Invitrogen	Cat# A78948
Alexa Fluor 594 goat anti-rabbit IgG (H+L)	Invitrogen	Cat# A11012, RRID:AB_2534079
Alexa Fluor 647 AffiniPure donkey anti-guinea pig IgG (H+L)	Jackson ImmunoResearch	Cat# 706-606-148, RRID:AB_2340477
Alexa Fluor 647 donkey anti-rabbit IgG (H+L)	Invitrogen	Cat# A32795
Alexa Fluor 555 AffiniPure donkey anti-goat IgG (H+L)	Jackson ImmunoResearch	Cat# 705565147, RRID: AB_3095465
Alexa Fluor 647 AffiniPure donkey anti-rat IgG (H+L)	Jackson ImmunoResearch	Cat# 712606153, RRID: AB_2340696
Chicken polyclonal anti-MAP2	EnCor	Cat# CPCA-MAP2, RRID:AB_2138173
Mouse monoclonal anti-HNA	Millipore Sigma	Cat# MAB1281, RRID:AB_94090
Rabbit polyclonal anti-Ki-67	Abcam	Cat# ab15580, RRID:AB_443209
Alexa 488 goat anti-chicken IgG	Thermo Fisher	Cat# A-11039, RRID:AB_142924
Alexa 568 goat anti-mouse IgG	Thermo Fisher	Cat# A-11004, RRID:AB_2534072
Alexa 647 goat anti-rabbit IgG	Thermo Fisher	Cat# A-21245, RRID:AB_2535813
Armenian hamster monoclonal anti-PD1-FITC (J43)	eBioscience / Thermo Fisher	Cat# 11-9985-82, RRID:AB_465472
Rat monoclonal anti-CD62L-PerCP-Cy5.5 (MEL-14)	BioLegend	Cat# 104431, RRID:AB_2187123
Rat monoclonal anti-Foxp3-Pacific Blue (FJK-16s)	eBioscience / Fisher	Cat# 48-5773-82, RRID:AB_465935
Rat monoclonal anti-CD45-BV510 (30-F11)	BioLegend	Cat# 103137, RRID:AB_2561392
Rat monoclonal anti-CD8a-BV605 (53-6.7)	BioLegend	Cat# 100743, RRID:AB_2561352
Rat monoclonal anti-CD4-BV711 (GK1.5)	BioLegend	Cat# 100447, RRID:AB_2564586
Mouse monoclonal anti-Granzyme B-AF647 (GB11)	BioLegend	Cat# 515406, RRID:AB_2566333
Rat monoclonal anti-CD25-AF700 (PC61)	BioLegend	Cat# 102024, RRID:AB_493709
Rat monoclonal anti-CD44-PE-Cy7 (IM7)	BioLegend	Cat# 103029, RRID:AB_830786
Rat monoclonal anti-CD11b-BUV395 (M1/70)	BD Biosciences	Cat# 565976, RRID:AB_2738276
Rat monoclonal anti-Ly6G-FITC (1A8)	BioLegend	Cat# 127605, RRID:AB_1236488
Rat monoclonal anti-MHCII-PerCPcy5.5 (M5/114.15.2)	BioLegend	Cat# 107625, RRID:AB_2191072
Mouse monoclonal anti-TGFβ/LAP-BV421 (TW7-20B9)	BioLegend	Cat# 141407, RRID:AB_2561580
Rat monoclonal anti-CD19-BV711 (6D5)	BioLegend	Cat# 115555, RRID:AB_2565970
Rat monoclonal anti-P2RY12-APC (S16007D)	BioLegend	Cat# 848005, RRID:AB_2721468
Rat monoclonal anti-Ly6C-AF700 (HK1.4)	BioLegend	Cat# 128023, RRID:AB_10640119
Rat monoclonal anti-PDL1-PE-Cy7 (10F.9G2)	BioLegend	Cat# 124313, RRID:AB_10639934
Rat monoclonal anti-CSF1R-BV605 (AFS98)	BioLegend	Cat# 135513, RRID:AB_2562667
Armenian hamster monoclonal anti-CD80-BV711 (16-10A1)	BioLegend	Cat# 104739, RRID:AB_2687094
Rat monoclonal anti-CD86-PE (GL1)	BioLegend	Cat# 105007, RRID:AB_313150
Bacterial and virus strains
AAV9-hSyn-HM3D-Gq-mCherry	Addgene	50474-AAV9
AAV9-hSyn-mCherry	Addgene	114472-AAV9
HB101 Competent Cells	Promega	Cat# L2015
XL10-Gold Ultracompetent Cells	Agilent	Cat# 200315
Biological samples
8656 nonmalignant human brain tissue sample	This study	N/A
6377 nonmalignant human brain tissue sample	This study	N/A
3218 nonmalignant human brain tissue sample	This study	N/A
3242 nonmalignant human brain tissue sample	This study	N/A
7346 nonmalignant human brain tissue sample	This study	N/A
6518 nonmalignant human brain tissue sample	This study	N/A
7768N nonmalignant human brain tissue sample	This study	N/A
2739 human glioblastoma tissue sample	This study	N/A
1207 human glioblastoma tissue sample	This study	N/A
4564 human glioblastoma tissue sample	This study	N/A
8025 human glioblastoma tissue sample	This study	N/A
3280 human glioblastoma tissue sample	This study	N/A
4617 human glioblastoma tissue sample	This study	N/A
7757 human glioblastoma tissue sample	This study	N/A
8547 human glioblastoma tissue sample	This study	N/A
7768T human glioblastoma tissue sample	This study	N/A
5852 spatially-resolved human glioblastoma tissue sample	This study	N/A
3768 spatially-resolved human glioblastoma tissue sample	This study	N/A
5646 spatially-resolved human glioblastoma tissue sample	This study	N/A
8814 spatially-resolved human glioblastoma tissue sample	This study	N/A
1068 nonmalignant brain explant	Savani et al42	N/A
3649 nonmalignant brain explant	Savani et al42	N/A
4875 nonmalignant brain explant	This study	N/A
6777 nonmalignant brain explant	Savani et al42	N/A
8995 nonmalignant brain explant	This study	N/A
9230 nonmalignant brain explant	Savani et al42	N/A
9413 nonmalignant brain explant	Savani et al42	N/A
1232 glioblastoma brain explant	Savani et al42	N/A
3313 glioblastoma brain explant	Savani et al42	N/A
5565 glioblastoma brain explant	Savani et al42	N/A
7602 glioblastoma brain explant	Savani et al42	N/A
9041 glioblastoma brain explant	Savani et al42	N/A
Chemicals, peptides, and recombinant proteins
Recombinant Human Epidermal Growth Factor	GoldBio	Cat# 1150-04
Recombinant Human Fibroblast Growth Factor 2	GoldBio	Cat# 1140-02
Heparin	STEMCELL Technologies	Cat# 07980
Penicillin/Streptomycin	Thermo Fisher	Cat# 15140148
Amphotericin B	GeminiBio	Cat# 400104
Plasmocin	InvivoGen	Cat# ant-mpp
L-Glutamine	Thermo Fisher	Cat# 25030081
B27 Supplement	Thermo Fisher	Cat# 17504044
N2 supplement	Thermo Fisher	Cat# 17502048
BenchMark Fetal Bovine Serum	GeminiBio	Cat# 100-106
Accutase	STEMCELL Technologies	Cat# 07922
Deschloroclozapine (DCZ)	Sigma-Aldrich	Cat# SML3651
Matrigel Basement Membrane Matrix, LDEV-free	Corning	Cat# 354234
Gabazine	EMD Millipore	Cat# 5059860001
Guanidinoacetic acid	Sigma-Aldrich	Cat# G11608
Guanidinoacetic acid	TCI	Cat# G0167
Creatine monohydrate	Thermo Scientific	Cat# B25009.22
3-(2-Carboxypiperazin-4-yl)propyl-1-phosphonic acid (CPP)	Hello Bio	Cat# HB0036
CGP 52432	Tocris	Cat# 1246
6,7-dinitroquinoxaline-2,3-dione (DNQX)	Tocris	Cat# 2312
Picrotoxin	Hello Bio	Cat# HB0506
Esp3I FastDigest	Thermo Fisher	Cat# FD0454
BsmBI-v2	New England Biolabs	Cat# R0739
Salt T4 DNA Ligase	New England Biolabs	Cat# M0467
TransIT Transfection Reagent	Mirus Bio	Cat# MIR2 B-304
LentiX Concentrator	Takara	Cat# 631232
PacI	New England Biolabs	Cat# R0547S
Polybrene	MedChemExpress	Cat# HY-112735
G418 Sulfate	GoldBio	Cat# G-418
Hygromycin	GoldBio	Cat# H-270-EZ50
Glucose	Thermo Fisher	Cat# 15023021
(R)-(–)-3-hydroxybutyric acid sodium salt	Millipore Sigma	Cat# 298360
2-hydroxybutyric acid sodium salt	Millipore Sigma	Cat# 220116
Acetone	Millipore Sigma	Cat# AX0120
Ammonium chloride	Millipore Sigma	Cat# A9434
Betaine	Millipore Sigma	Cat# 61962
Calcium chloride	Millipore Sigma	Cat# C5670
Calcium nitrate tetrahydrate	Millipore Sigma	Cat# C1396
Citric acid	Millipore Sigma	Cat# 251275
Creatine	Millipore Sigma	Cat# C0780
Creatinine	Millipore Sigma	Cat# C4255
D-(-)-fructose	Millipore Sigma	Cat# F3510
D-(+)-galactose	Millipore Sigma	Cat# G5388
Formic acid	Millipore Sigma	Cat# 1002640100
Glycerol	Millipore Sigma	Cat# G2025
Glycine	Millipore Sigma	Cat# G7126
Hypoxanthine	Millipore Sigma	Cat# H9377
L-(-)-malic acid	Millipore Sigma	Cat# M7397
L-2-aminobutyric acid	Millipore Sigma	Cat# A2536
L-alanine	Millipore Sigma	Cat# A7627
L-arginine monohydrochloride	Millipore Sigma	Cat# A5131
L-asparagine	Millipore Sigma	Cat# A0884
L-aspartic acid	Millipore Sigma	Cat# A9256
L-carnitine hydrochloride	Millipore Sigma	Cat# C0283
L-citrulline	Millipore Sigma	Cat# C7629
L-cysteine hydrochloride	Millipore Sigma	Cat# C1276
L-cystine	Millipore Sigma	Cat# C8755
L-glutamic acid	Millipore Sigma	Cat# G1251
L-glutamine	Millipore Sigma	Cat# G3126
L-glutathione reduced	Millipore Sigma	Cat# G6013
L-histidine monohydrochloride monohydrate	Millipore Sigma	Cat# H5659
L-isoleucine	Millipore Sigma	Cat# I2752
L-leucine	Millipore Sigma	Cat# L8000
L-lysine monohydrochloride	Millipore Sigma	Cat# L5626
L-methionine	Millipore Sigma	Cat# M9625
L-ornithine monohydrochloride	Millipore Sigma	Cat# O2375
L-phenylalanine	Millipore Sigma	Cat# P2126
L-proline	Millipore Sigma	Cat# P0380
L-serine	Millipore Sigma	Cat# S4500
L-threonine	Millipore Sigma	Cat# T8625
L-tryptophan	Millipore Sigma	Cat# T0254
L-tyrosine	Millipore Sigma	Cat# 93829
L-valine	Millipore Sigma	Cat# V0500
Magnesium chloride	Millipore Sigma	Cat# M8266
Magnesium sulfate	Millipore Sigma	Cat# M2643
Malonic acid	Millipore Sigma	Cat# M1296
N-acetylglycine	Millipore Sigma	Cat# A16300
O-acetyl-L-carnitine hydrochloride	Millipore Sigma	Cat# A6706
Phenol red sodium salt	Millipore Sigma	Cat# P5530
Potassium chloride	Millipore Sigma	Cat# P5405
RPMI 1640 vitamins solution	Millipore Sigma	Cat# R7256
Sodium acetate	Millipore Sigma	Cat# S5636
Sodium bicarbonate	Millipore Sigma	Cat# S5761
Sodium chloride	Millipore Sigma	Cat# S7653
Sodium L-lactate	Millipore Sigma	Cat# L7022
Sodium phosphate dibasic heptahydrate	Millipore Sigma	Cat# S9390
Sodium pyruvate	Millipore Sigma	Cat# P2256
Succinic acid	Millipore Sigma	Cat# S3674
Taurine	Millipore Sigma	Cat# T0625
trans-4-hydroxy-L-proline	Millipore Sigma	Cat# H5534
Urea	Millipore Sigma	Cat# U5378
Uric acid	Millipore Sigma	Cat# U2625
α-ketoglutaric acid	Millipore Sigma	Cat# K1128
Ammonium chloride (15N)	Cambridge Isotope Laboratories	Cat# NLM-467
Hypoxanthine (15N4)	Cambridge Isotope Laboratories	Cat# NLM-8500
L-alanine (15N)	Cambridge Isotope Laboratories	Cat# NLM-454
L-arginine hydrochloride (15N4)	Cambridge Isotope Laboratories	Cat# NLM-396
L-asparagine hydrochloride (15N2)	Cambridge Isotope Laboratories	Cat# NLM-3286
L-aspartic acid (15N)	Cambridge Isotope Laboratories	Cat# NLM-718
L-citrulline (ureido-15N1)	Cambridge Isotope Laboratories	Cat# NLM-6850
L-creatine (15N3)	Cambridge Isotope Labs	Cat# NLM-9218
L-cysteine (15N)	Cambridge Isotope Laboratories	Cat# NLM-2295
L-cystine (15N2)	Cambridge Isotope Laboratories	Cat# NLM-3818
L-glutamic acid (15N)	Cambridge Isotope Laboratories	Cat# NLM-135
L-glutamine (15N2)	Cambridge Isotope Laboratories	Cat# NLM-1328
L-glycine (15N)	Cambridge Isotope Laboratories	Cat# NLM-202
L-histidine hydrochloride monohydrate (15N3)	Cambridge Isotope Laboratories	Cat# NLM-1513
L-isoleucine (15N)	Cambridge Isotope Laboratories	Cat# NLM-292
L-leucine (15N)	Cambridge Isotope Laboratories	Cat# NLM-142
L-lysine dihydrochloride (15N2)	Cambridge Isotope Laboratories	Cat# NLM-1554
L-methionine (15N)	Cambridge Isotope Laboratories	Cat# NLM-752
L-ornithine hydrochloride (15N2)	Cambridge Isotope Laboratories	Cat# NLM-3610
L-phenylalanine (15N)	Cambridge Isotope Laboratories	Cat# NLM-108
L-proline (15N)	Cambridge Isotope Laboratories	Cat# NLM-835
L-serine (15N)	Cambridge Isotope Laboratories	Cat# NLM-2036
L-threonine (15N)	Cambridge Isotope Laboratories	Cat# NLM-742
L-tryptophan (15N2)	Cambridge Isotope Laboratories	Cat# NLM-800
L-tyrosine (15N)	Cambridge Isotope Laboratories	Cat# NLM-590
Sodium nitrate (15N)	Cambridge Isotope Laboratories	Cat# NLM-157
Taurine (15N)	Millipore Sigma	Cat# 491330
Urea (15N2)	Cambridge Isotope Laboratories	Cat# NLM-233
Uric acid (1,3-15N2)	Cambridge Isotope Laboratories	Cat# NLM-1697
Uridine (15N2)	Cambridge Isotope Laboratories	Cat# NLM-812
L-valine (15N)	Cambridge Isotope Laboratories	Cat# NLM-316
L-glutamine (amide-15N)	Cambridge Isotope Laboratories	Cat# NLM-557
Dialyzed Fetal Bovine Serum	GeminiBio	Cat# 100-108
tert-Butyl methyl ether	Millipore Sigma	Cat# 650560
L-glutamine (13C5)	Cambridge Isotope Laboratories	Cat# CLM-1822
Chloramphenicol	Millipore Sigma	Cat# C0378
Isopropanol	Thermo Fisher	Cat# A461
SPLASH™ LIPIDOMIX™ Sph/Cer mix standard	Avanti Research	Cat# 330707
Acetonitrile (Optima LC/MS Grade)	Thermo Fisher	Cat# A9554
1,2-13C2, 3-15N-guanidinoacetic acid	Cambridge Isotope Labs	Cat# CNLM-8300
2,2-D2-guanidinoacetic acid	Cambridge Isotope Labs	Cat# DLM-9998
Guanidineacetic acid	Sigma-Aldrich	Cat# G11608
Methanol	Thermo Fisher	Cat# A456
Creatine-d5 H2O (N-methyl-d3; glycine-2,2-d2)	LGC Standards	Cat# TRC-C781496-50MG
Guanidineacetic-2,2-d2 Acid	CDN Isotopes	Cat# D-6320
Creatine	LGC Standards	Cat# TRC-C781483
Guanidinoacetic Acid	LGC Standards	Cat# TRC-G821250-25G
Hibernate A	BrainBits	Cat# HA
RBC Lysis Buffer	Thermo Fisher	Cat# 00433357
cOmplete Mini Protease Inhibitor Cocktail	Millipore Sigma	Cat# 11836153001
CellTrace™ Calcein Violet, AM, for 405 nm excitation	Thermo Fisher	Cat# C34858
Ethidium homodimer-1	Thermo Fisher	Cat# E1169
4% Paraformaldehyde in PBS	Thermo Fisher	Cat# J19943.K2
DAPI	Cayman Chemical	Cat# 14285
Sucrose	Sigma Aldrich	Cat# S9378
Sodium bicarbonate	Thermo Fisher	Cat# BP328
Potassium chloride	Millipore Sigma	Cat# P9333
D-glucose	Millipore Sigma	Cat# G8270
Sodium phosphate monobasic	Sigma Aldrich	Cat# S8282
Magnesium chloride	Millipore Sigma	Cat# M9272
Calcium chloride	Millipore Sigma	Cat# 223506
Sodium chloride	Millipore Sigma	Cat# S9888
Choline chloride	Millipore Sigma	Cat# C1879
Sodium bicarbonate	Millipore Sigma	Cat# S6014
D-glucose	Millipore Sigma	Cat# G8270
Ascorbic acid	Millipore Sigma	Cat# A92902
Potassium chloride	Millipore Sigma	Cat# P5405
Sodium phosphate dibasic	Millipore Sigma	Cat# S9638
Sodium pyruvate	Millipore Sigma	Cat# P5280
Magnesium chloride	Millipore Sigma	Cat# M8266
Calcium chloride	Millipore Sigma	Cat# C7902
Kynurenic acid	Millipore Sigma	Cat# K3375
Sodium chloride	Millipore Sigma	Cat# S7653
Sucrose	Thermo Fisher	Cat# S5
Potassium chloride	Thermo Fisher	Cat# P330
Sodium phosphate monobasic	Thermo Fisher	Cat# S397
Sodium bicarbonate	Thermo Fisher	Cat# P233
D-glucose	Gibco	Cat# 15023-021
Calcium chloride	Thermo Fisher	Cat# 012316.A1
Magnesium sulfate	Thermo Fisher	Cat# 033337.36
Cesium methanesulfonate	Millipore Sigma	Cat# C1426
HEPES	Millipore Sigma	Cat# H3375
Adenosine 5′-triphosphate magnesium salt	Millipore Sigma	Cat# A9187
Guanosine 5′-triphosphate sodium salt hydrate	Millipore Sigma	Cat# G8877
Phosphocreatine di(tris)	Millipore Sigma	Cat# P7936
EGTA	Millipore Sigma	Cat# E3889
Tetrodotoxin	Hello Bio	Cat# HB1034
Potassium gluconate	Millipore Sigma	Cat# P1847
Phosphocreatine di(tris) salt	Millipore Sigma	Cat# P1937
HEPES	Millipore Sigma	Cat# H4034
Guanosine 5′-triphosphate sodium salt solution	Millipore Sigma	Cat# G3776
Sodium chloride	Thermo Fisher	Cat# P271
Alexa Fluor 488 hydrazide sodium salt	Thermo Fisher	Cat# A10436
Gramicidin	Sigma-Aldrich	Cat# G5002
Papain	Worthington Biochemical	Cat# LS003126
Antibiotic-Antimycotic (100X)	Thermo Fisher	Cat# 15240062
5-ethynyl-2′-deoxyuridine	Thermo Fisher	Cat# E10187
Deoxyribonuclease I	Millipore Sigma	Cat# DN25
Sodium pyruvate	Thermo Fisher	Cat# 11360070
N-(Ethoxycarbonylmethyl)-6-Methoxyquinolinium-Bromide	Thermo Fisher	Cat# E3101
GlutaMAX	Thermo Fisher	Cat# 35050061
B27 Supplement minus Vitamin A	Thermo Fisher	Cat# 12587010
2-mercaptoethanol	Thermo Fisher	Cat# BP176100
Insulin	Millipore Sigma	Cat# I9278
Zombie NIR Fixable Viability Live/Dead Kit	BioLegend	Cat# 423105
Percoll	Cytiva	Cat# 17089101
Recombinant M-CSF	Peprotech	Cat# 315-02
Recombinant GM-CSF	Peprotech	Cat# 315-03
Murine IL-2	Peprotech	Cat# 212-12
RPMI 1640	Thermo Fisher	Cat# 11875093
Critical commercial assays
e-Myco PCR Detection Kit	Bulldog Bio	Cat# 2523348
e-Myco PLUS PCR Detection Kit	Bulldog Bio	Cat# 25233
MycoAlert Mycoplasma Detection Kit	Lonza	Cat# LT07-318
Brain Tumor Dissociation Kit	Miltenyi Biotec	Cat# 130-095-942
Q5 Hot Start High-Fidelity 2X Master Mix	New England Biolabs	Cat# M0494
QIAquick Gel Extraction Kit	QIAGEN	Cat# 28706
Monarch PCR & DNA Cleanup Kit	New England Biolabs	Cat# T1030
Gateway BP Clonase II	Thermo Fisher	Cat# 11789100
Gateway LR Clonase II	Thermo Fisher	Cat# 11791020
In-Fusion HD Cloning Kit	Takara Bio	Cat# 102518
Virabind Adenovirus Miniprep Kit	Cell Biolabs	Cat# VPK-099
Adeno-X GoStix	Takara	Cat# 632270
Lenti-X GoStix Plus	Takara	Cat# 631281
Zymo Clean and Concentrator-5 Kit	Zymo Research	Cat# D4013
NEBNext 2x MasterMix	New England Biolabs	Cat# M0541S
RNeasy Plus Universal Kit	Qiagen	Cat# 73404
Neural Tissue Dissociation Kit (P)	Miltenyi Biotec	Cat# 130-092-628
Chromium Next GEM Single Cell 3’ Kit v3.1	10X Genomics	Cat# PN-1000269
Qubit DNA high sensitivity assay	Thermo Fisher	Cat# Q33231
Kapa RNA HyperPrep Kit with RiboErase	Roche	Cat# KK8561
Leica Bond Polymer Refine Detection Kit	Leica	Cat# DS9800,RRID:AB_2891238
Papain dissociation kit	Worthington Biochemical	Cat# LK003150
Invitrogen Click-iT Plus EdU Alexa Fluor 488 Flow Cytometry Assay Kit	Thermo Fisher	Cat# C10632
EasySep Mouse T Cell Isolation Kit	Stemcell Technologies	Cat# 19851
CellTrace Violet Cell Proliferation Kit	Thermo Fisher	Cat# C34557
ImmunoCult Mouse T Cell Activator Kit	Stemcell Technologies	Cat# 100-1572
eBioscience Fixation-Permeabilization Kit	Thermo Fisher / Fisher	Cat# 88-8824-00
Deposited data
Metabolomics analysis of human brain tissue specimens	This study	NMDR Metabolomics Workbench ST005093
Lipidomics analysis of human brain tissue specimens	This study	NMDR Metabolomics Workbench ST005089
Metabolome in a group of AA and EA matched pairs of prostate cancer (PCa) and benign adjacent tissues	Sreekumar et al.	NMDR Metabolomics Workbench ST000784
Metabolomics analysis of matched ccRCC and benign adjacent tissues, Dataset 1	Bezwada et al.44	NMDR Metabolomics Workbench ST003143
Metabolomics analysis of matched ccRCC and benign adjacent tissues, Dataset 2	Hakimi et al.45	NIHMSID: NIHMS746037
Metabolomics analysis of TS516 and NHA Donor #1 cells in nitrogen metabolism profiling platform	This study	NMDR Metabolomics Workbench ST005087
TCGA-GBM RNA sequencing analysis of human GBM tissue	TCGA	https://www.cancer.gov/tcga, https://portal.gdc.cancer.gov/projects/TCGA-GBM
GTEx RNA sequencing analysis of human brain tissue	GTEx	GTEx portal, file version: 2017-0605_v8_brain_cortex
RNA sequencing of glioma stem-like cells and NHA Donor #1 cells	Wu et al6, Savani et al48	NCBI GEO GSE265788, GSE312748
The Multi-Layered Transcriptional Architecture of Glioblastoma Ecosystems; single-cell RNA sequencing of human glioblastoma	Nomura et al76	NCBI GEO GSE274546
ATAC sequencing of NHA cells and AGAThigh and AGATlow GSC lines	This study	NCBI GEO GSE343207
RNA sequencing analysis of human brain tissue specimens	This study	NCBI GEO GSE343542
RNA sequencing of GSC and NSC lines	Mack et al.77	NCBI GEO GSE119834
Experimental models: Cell lines
NHA Donor #1 (immortalized with HPV E6 and E7 and hTERT)	Sonoda et al.78; Provided by R. Pieper	N/A
NHA Donor #2 (immortalized with HPV E6 and E7 and hTERT)	Shi et al11.	N/A
HEK293T	ATCC	Cat# CRL-3216, RRID: CVCL_0063
293AD	CellBioLabs	Cat# AD-100, RRID: CVCL_KA63
BT054	Kelly et al.79; Provided by S. Weiss	N/A
BT260	Koivunen et al5; provided by K. Ligon	N/A
HK157	Laks et al.80; Provided by H. Kornblum	N/A
HK211	Laks et al.80; Provided by H. Kornblum	N/A
HK213	Laks et al.80; Provided by H. Kornblum	N/A
HK252	Laks et al.80; Provided by H. Kornblum	N/A
HK308	Laks et al.80; Provided by H. Kornblum	N/A
TS516	Rohle et al.81; Provided by I. Mellinghoff	RRID: CVCL_A5HY
TS603	Rohle et al.81; Provided by I. Mellinghoff	RRID: CVCL_A5HW
UTSW5	This study	N/A
UTSW63	Nguyen et al82	N/A
UTSW71	Nguyen et al82	N/A
ENSA	Wang et al83	N/A
NSC11	Wang et al83	N/A
MGG152	Wakimoto et al84	N/A
HMC3	ATCC	Cat# CRL-3304, RRID: CVCL_II76
CT-2A tumor cells	Millipore Sigma	Cat# SCC194, RRID: CVCL_ZJ44
iPS(IMR-90)-4	WiCell	RRID:CVCL_C437
Experimental models: Organisms/strains
C57BL/6J mice; 5-7 weeks old	Jackson Labs	Cat# 000664, RRID:IMSR_JAX:000664
C57BL/6J mice; P18-23	Jackson Labs	Cat# 000664, RRID:IMSR_JAX:000664
ICR-SCID mice; 11 weeks old, female	Taconic	Cat# ICRSC-F
ICR-SCID mice; 7 weeks old, female	Taconic	Cat# ICRSC-F
ICR-SCID mice; 8 weeks old, female	Taconic	Cat# ICRSC-F
NSG mice; P2	Jackson Labs	Cat# 005557, RRID:IMSR_JAX:005557
NSG mice; 6-8 weeks old	Jackson Labs	Cat# 005557, RRID:IMSR_JAX:005557
Nude mice; 6-8 weeks old, female	Taconic	Cat# NCRNU-F
Sprague-Dawley rat pups; P0-P3	Charles River Laboratories	Cat# Crl:CD(SD)
Thy1::ChR2+/-;NSG mice; P28-P35	Venkatesh et al26	N/A
Oligonucleotides
sgAAVS1 #1: GTCACCAATCCTGTCCCTAG	Chen et al85	N/A
sgAGAT #1: TGTGGGCAATGAGATTATCG	This study	N/A
sgGAMT #1: GGCCAGCGCGTGCATATAGG	This study	N/A
sgGAMT #2: GTACGACACGTACCCACTCT	This study	N/A
sgAAVS1 #2: TGTTAGGCAGATTCCTTATC	Chen et al85	N/A
sgAGAT #2: ACTTCAATGACCAGTCAATG	This study	N/A
attB1_U6_fwd: GGGGACAAGTTTGTACAAAAAAGCAGGCTCTGAGG GCCTATTTCCCATG	This study	N/A
bGHpolyA_attB2_Rev_v2: GGGGACCACTTTGTACAAGAAAGCTGGGTTCCATA GAGCCCACCGCAT	This study	N/A
GATM_C407A_InvPCR_F: TTCCATGCCTGGACCTGCGATGTC	This study	N/A
GATM_C407A_InvPCR_R: GGTCCAGGCATGGAAGCCTCCT	This study	N/A
attB_AGAT_F: GGGGACAAGTTTGTACAAAAAAGCAGGCTCTGCCA CCATGCTGCGGGT	This study	N/A
attB_AGAT_R: GGGGACCACTTTGTACAAGAAAGCTGGGTTTCAGTC CAAGTAGGACTGTAAGGTGCC	This study	N/A
attB_GAMT_F: GGGGACAAGTTTGTACAAAAAAGCAGGCTCTGCCA CCATGAGCGCCCC	This study	N/A
attB_GAMT_R: GGGGACCACTTTGTACAAGAAAGCTGGGTTTCAGC CTTTGGTCACCAGGGG	This study	N/A
EnAsCas12a crAAVS1 #1: TCTTCTGTGCACTACTTGTC	This study	N/A
EnAsCas12a crAAVS1 #2: ATGGCTTCTAGGTTGCGGAC	This study	N/A
EnAsCas12a crAAVS1 #3: ATTTCCCAGCATGCCCTGCGAGG	This study	N/A
EnAsCas12a crAAVS1 #4: TGGACAACCCCAAAGTACCCCG	This study	N/A
EnAsCas12a crAGAT #1: GTCTGACAGGTTTATACAGTGCA	This study	N/A
EnAsCas12a crAGAT #2: GGAGCTGTTGTCCACTTGGCGCC	This study	N/A
EnAsCas12a crAGAT #3: ATAATTGACCTGTACGCTCGGTA	This study	N/A
EnAsCas12a crAGAT #4: TGCTCTGCCCACTATCACTTCC	This study	N/A
LGmCherry_F: TCGACGAATTCGGATCCGCAATGGTGAGCAAGGGC GAG	This study	N/A
LGmCherry _R: ATTAAGTTTAAACGCGGCCATTACTTGTACAGCTCG TCCATG	This study	N/A
pLV-EF1a-IRES-Puro DEST_fwd: TGGCCGCGTTTAAACTTAATTATC	This study	N/A
pLV-EF1a-IRES-Puro DEST_rev: TGCGGATCCGAATTCGTC	This study	N/A
attB_GFP_F: GGGGACAAGTTTGTACAAAAAAGCAGGCTTAggatccg catggtgagcaag	This study	N/A
attB_GFP_R: GGGGACCACTTTGTACAAGAAAGCTGGGTTttaattaag tttaaacgcggccttacttgtac	This study	N/A
Recombinant DNA
lentiCRISPRv2-GFP	Addgene	Addgene, cat# 82416
lentiCRISPRv2-GFP-sgAAVS1-1	Addgene	Addgene, cat# 261346
lentiCRISPRv2-GFP-sgGAMT-1	Addgene	Addgene, cat# 261347
lentiCRISPRv2-GFP-sgGAMT-2	Addgene	Addgene, cat# 261348
lentiCRISPRv2-mCherry	Addgene	Addgene, cat# 99154
pDONR223	Thermo Fisher	N/A
pAd/PL-DEST	Thermo Fisher	V49420
pAdCRISPRv2-GFP-sgAAVS1-1	Addgene	Addgene, cat# 261349
pAdCRISPRv2-GFP-sgAGAT-1	Addgene	Addgene, cat# 261351
pAdCRISPR-mCherry-sgAAVS1-2	Addgene	Addgene, cat# 261350
pAdCRISPR-mCherry-sgAGAT-2	Addgene	Addgene, cat# 261352
pENTR221-AGAT WT from Ultimate ORF Human Clone Library	Life Technologies	Cat# HORF01; RRID:SCR_005371
pLenti-EF1α-DEST-IRES-Neo	Savani et al42	Addgene, cat# 253766
pLenti-EF1α-EV-IRES-Neo	Savani et al42	Gift of G. Lu
pLenti-EF1α-AGAT_WT-IRES-Neo	This study	Addgene, cat# 261068
pLenti-EF1α-AGAT_C407A-IRES-Neo	This study	Addgene, cat# 261069
pLenti-EF1α-EV-PGK-Hygro	This study	Addgene, cat# 261070
pLenti-EF1α-GAMT_WT-PGK-Hygro	This study	Addgene, cat# 261071
pLenti-EF1α-GAMT_E45S-PGK-Hygro	This study	Addgene, cat# 261072
pRDB-054	McGee et al.	Addgene, cat# 216096
pRDB-054_4x_crAAVS1	This study	Addgene, cat# 261344
pRDB-054_4x_crAGAT	This study	Addgene, cat# 261345
pLV-EF1α-DEST-IRES-Puro	Savani et al	Addgene, cat# 253769
pLV-EF1α-mCherry-IRES-Puro	This study	Addgene, cat# 261073
pLenti-CMV-GFP-PGK-Puro	This study	Addgene, cat# 261074
psPAX2	Addgene	Addgene, cat# 12260
pMD2.G	Addgene	Addgene, cat# 12259
Software and algorithms
El-MAVEN (0.12.0)	Elucidata	https://elucidata.io/el-maven/, RRID: SCR_022159
TraceFinder (5.1 SP2)	Thermo Fisher	Cat# OPTON-31001, RRID: SCR_023045
MetaboAnalyst 5.0	MetaboAnalyst	https://www.metaboanalyst.ca/home.xhtml, RRID: SCR_015539
Multiquant v2.1.1	Sciex	https://sciex.com/products/software/multiquant-software
R (4.4.2)	The R Project for Statistical Computing	https://www.r-project.org/, RRID: SCR_001905
RStudio (2024.09)	Posit Software	https://posit.co/products/open-source/rstudio/, RRID: SCR_000432
AccuCor (0.3.0)	Su et al.86	https://github.com/lparsons/accucor, RRID: SCR_023046
Metabolomics and Stable Isotope Tracing Postprocessing Script	Savani et al42	https://zenodo.org/records/18736712
GraphPad Prism (10.4.1)	GraphPad	https://www.graphpad.com/scientific-software/prism/, RRID: SCR_002798
Astrocyte 2.2.0	UT Southwestern Bioinformatics Core	N/A
ImageJ	Schneider et al87	RRID:SCR_003070
Trimgalore v 0.6.4	Krueger et al88	https://github.com/FelixKrueger/TrimGalore, RRID:SCR_011847
BWA	Li et al89	https://bio-bwa.sourceforge.net/, RRID:SCR_010910
Sambamba	Tarasov et al90	http://www.open-bio.org/wiki/Sambamba.Sambamba v0.5.0, RRID:SCR_024328
SamTools	Danecek et al91	https://www.htslib.org/, RRID:SCR_002105
bedtools	Quinlan et al92	https://github.com/arq5x/bedtools2, RRID:SCR_006646
PhantomPeakQualTools	Kharchenko et al93, Landt et al	https://code.google.com/p/phantompeakqualtools/, RRID:SCR_005331
MACS2	Zhang et al94	https://github.com/macs3-project/MACS, RRID:SCR_013291
Seurat (V5)	Satija et al95	https://satijalab.org/seurat/, RRID:SCR_016341
STAR v2.7.3a	Dobin et al96	https://github.com/alexdobin/STAR, RRID:SCR_004463
Subread v1.6.3	Liao et al97	https://subread.sourceforge.net/, RRID:SCR_009803
GSVA	Hӓnzelmann et al98	https://www.bioconductor.org/packages/release/bioc/html/GSVA.html, RRID:SCR_021058
Clampex v11.3	Molecular Devices	https://www.moleculardevices.com/products/axon-patch-clamp-system/acquisition-and-analysis-software/pclamp-software-suite, RRID: SCR_011323
Clampfit v11.4	Molecular Devices	https://www.moleculardevices.com/products/axon-patch-clamp-system/acquisition-and-analysis-software/pclamp-software-suite, RRID: SCR_011323
Igor Pro v9.05	WaveMetrics	https://www.wavemetrics.com/, RRID:SCR_000325
Labview 8.6	National Instruments	https://www.ni.com/en/support/downloads/software-products/download.labview.html#570679, RRID: SCR_014325
BrainWave 5	3Brain	https://www.3brain.com/products/software/brainwave5
Tidyverse	Wickham et al99	https://www.tidyverse.org/, RRID:SCR_019186
Cell Ranger (v10.0.0)	10x Genomics	https://github.com/10XGenomics/cellranger, RRID:SCR_017344
UCell R package	Andreatta & Carmona	https://github.com/carmonalab/UCell
Human MSigDB v2026.1.Hs	Subramanian, Tamayo, et al	https://www.gsea-msigdb.org/gsea/msigdb
Mouse MSigDB v2026.1.Mm	Castanza et al	https://www.gsea-msigdb.org/gsea/msigdb
Zeiss Zen 3.11	ZEISS Carl Zeiss Microscopy GmbH	https://knowledge.zeiss.com/rms/en/zen/release-notes/zen-3-11; RRID:SCR_013672
NDP.view 2	Hamamatsu Photonics	https://www.hamamatsu.com/jp/en/product/life-science-and-medical-systems/digital-slide-scanner/U12388-01.html,
Neural Metric Tools	Axion Biosystems	https://www.axionbiosystems.com/products/mea/mea-software/neural-module
FCS Express	Dotmatics	https://www.dotmatics.com/solutions/fcs-express
FlowJo Software	BD Biosciences	https://www.bdbiosciences.com/en-us/products/software/flowjo-software?tab=flowJo-v11-software
Other
NeuroCult NS-A Basal Medium (Human) with Proliferation Supplement	STEMCELL Technologies	Cat# 05751
Neurobasal Medium	Thermo Fisher	Cat# 21103049
DMEM	Thermo Fisher	Cat# 11995-065
EMEM	ATCC	Cat# 30-2003
EDTA buffer pH 8.5	Sigma	Cat# E1161
EDTA buffer pH 8.0	Abcam	Cat# 93680
Citric acid buffer pH 6.0	Fisher Scientific	Cat# NC9401067
Normal donkey serum	Jackson ImmunoResearch	Cat# 17000121
Normal donkey serum	Sigma-Aldrich	Cat# S30-M
Triton X-100	Millipore Sigma	Cat# T8787
ProLong Diamond Antifade Mountant	Thermo Fisher	Cat# P36970
ProLong Gold Antifade Mountant	Thermo Fisher	Cat# P36934
Clear Flat Bottom Ultra-Low Attachment 6-well Plates	Corning	Cat# 3471
Clear Flat Bottom Ultra-Low Attachment 24-well Plates	Corning	Cat# 3473
100 mm Ultra-Low Attachment Culture Dishes	Corning	Cat# 4615
FACS Aria II SORP 4-Laser	BD Biosciences	Cat# N/A
SpeedVac	Thermo Fisher	Cat# SPD2030
TissueLyser II	QIAGEN	RRID: SCR_018623
ViCELL XR Cell Viability Analyzer	Beckman Coulter	RRID: SCR_019664
Q-Exactive HF-X	Thermo Fisher	RRID: SCR_020425
Orbitrap Exploris 480	Thermo Fisher	RRID: SCR_027000
Vanquish Flex UHPLC	Thermo Fisher	N/A
AB QTRAP 5500	Applied Biosystems Sciex	RRID: SCR_020517
AB QTRAP 6500+	Applied Biosystems Sciex	RRID:SCR_021831
Nexera X2 LC-30AD HPLC	Shimadzu	N/A
NextSeq 500 System	Illumina	RRID: SCR_014983
Leica Bond RX Automated Stainer	Leica	RRID:SCR_025548
BOND Epitope Retrieval Solution 1	Leica	Cat# AR9961
BOND Epitope Retrieval Solution 2	Leica	Cat# AR9640
Tissue-Tek Film® Automated Coverslipper	Sakura Finetek	Cat# 4740, 4743
Leica Aperio AT2 Scanner	Leica Microsystems	RRID:SCR_021256
NanoZoomerS60 Digital Slide Scanner	Hamamatsu	RRID:SCR_023762
Zeiss Axioscan 7	Zeiss	RRID:SCR_027284
Zeiss LSM880 with Airyscan Confocal Laser Scanning Microscope	Zeiss	RRID:SCR_020925
VT1200S Vibratome	Leica	Cat# N/A; RRID:SCR_018453
VT1000P Vibratome	Leica	N/A
Borosilicate glass capillary	Sutter Instrument	Cat# BF150-86-10HP
P-1000 micropipette puller	Sutter Instrument	RRID:SCR_021042
Multiclamp 700B Microelectrode Amplifier	Molecular Devices	RRID:SCR_018455
Digidata 1550B	Molecular Devices	N/A
Multiclamp 700A Microelectrode Amplifier	Molecular Devices	RRID:SCR_021040
96-well Axion CytoView microelectrode array (MEA) plate	Axion Biosystems	M768-tMEA-96
Maestro Pro MEA system	Axion BioSystems	N/A
BioCam DupleX with CorePlate™ 3D	3Brain	N/A
Dual Fiber-Optic Cannula	Doric Lenses Inc.	Cat# MFC_200/240-0.22_4.5mm_MF2.5_FLT
Stainless steel screws, thread size 00-90 × 1/16	Antrin Miniature Specialties	Cat# AMS90/1P
C&B Metabond	Parkell	Cat# S380
Geristore A&B paste	DenMat	Cat# 34523101
Poly-D-lysine and laminin-coated coverslips	NeuVitro	Cat# GG-12-1.5-laminin
ISO Pro Rodent 3000 Diet	Purina Prolab	Cat# 5P76
Teklad Irradiated Global 16% Protein Rodent Diet	Inotiv	Cat# 2916
Teklad Amino Acid Diet	Inotiv	Cat# TD.01084
Teklad 0.665% Ornithine, Arginine-restricted diet	Inotiv	Cat# TD.240691
24-Well Ultra-Low Adherence Plate	Corning	Cat# 3473
Zeiss spinning disk microscope equipped with a 10× air-objective (NA 0.75)	Zeiss	N/A
FACS LSRFortessa	BD Biosciences	N/A
BD FACSymphony A5 Analyzer	BD Biosciences	N/A
STEMdiff Cerebral Organoid Kit	Stem Cell Technologies	Cat# 08570

Highlights.

• Guanidinoacetate (GAA) accumulation is a metabolic hallmark of high-grade glioma

• Imbalanced creatine synthesis pathway enzyme activity causes GAA secretion

• GAA excites local neurons by activating γ-aminobutyric acid (GABA) A receptors

• Genetic or dietary GAA depletion reduces neuronal activity and glioma growth