The genomic alterations in glioblastoma influence the levels of CSF metabolites

Abstract

Cerebrospinal fluid (CSF) analysis is underutilized in patients with glioblastoma (GBM), partly due to a lack of studies demonstrating the clinical utility of CSF biomarkers. While some studies show the utility of CSF cell-free DNA analysis, studies analyzing CSF metabolites in patients with glioblastoma are limited. Diffuse gliomas have altered cellular metabolism. For example, mutations in isocitrate dehydrogenase enzymes (e.g., IDH1 and IDH2) are common in diffuse gliomas and lead to increased levels of D-2-hydroxyglutarate in CSF. However, there is a poor understanding of changes CSF metabolites in GBM patients. In this study, we performed targeted metabolomic analysis of CSF from n = 31 patients with GBM and n = 13 individuals with non-neoplastic conditions (controls), by mass spectrometry. Hierarchical clustering and sparse partial least square-discriminant analysis (sPLS-DA) revealed differences in CSF metabolites between GBM and control CSF, including metabolites associated with fatty acid oxidation and the gut microbiome (i.e., carnitine, 2-methylbutyrylcarnitine, shikimate, aminobutanal, uridine, N-acetylputrescine, and farnesyl diphosphate). In addition, we identified differences in CSF metabolites in GBM patients based on the presence/absence of TP53 or PTEN mutations, consistent with the idea that different mutations have different effects on tumor metabolism. In summary, our results increase the understanding of CSF metabolites in patients with diffuse gliomas and highlight several metabolites that could be informative biomarkers in patients with GBM.

Keywords: Biomarker; CSF; Carnitine; Cerebrospinal fluid; Choline; GABA; Glioblastoma; Lactate; Metabolomics; PTEN; Shikimate; TP53.

Fig. 1

Patient characteristics. CSF samples were collected from patients with GBM (n = 31) and individuals without a history of cancer as controls (n = 13). (A) Characteristics of patients with CSF collected pre-treatment (n = 13) and controls (n = 13). Four patients (patients 5, 6, 7, and 10) received no additional treatment (i.e., no chemotherapy or radiation). (B) Characteristics of patients for which tumor mutation information was available (n = 18). All patients with mutation information had CSF collected post-treatment. (C) Legend for panels A and B. Overall survival (OS) spans from 2 to 73 months. Treatments included chemotherapy, radiation therapy (RT), tumor-treating field (TTF), or was not applicable for control patients. All GBM patients underwent surgery. CSF was collected via lumbar puncture (LP) (n = 18), sulcus sampling (n = 17), ventricular sampling (n = 7), reservoir/shunt (n = 1), or cisterna magna (n = 1)

Fig. 2

CSF metabolites differ between patients with GBM and controls. (A) sPLS-DA plot with the 31 GBM samples shows that GBM samples collected pre-treatment and post-treatment cluster in separate groups. (B) Volcano plot comparing CSF metabolites in CSF from GBM pre-treatment vs. GBM post-treatment. Colored points represent metabolites that are present at significantly different levels and fold change (-log10(p-value) > 1.3, log2(Post/Pre) > + 1 or < -1) in CSF samples from GBM pre-treatment vs. GBM post-treatment. Several CSF metabolites are present at significantly different levels in GBM pre-treatment compared to GBM post-treatment. P-value and fold change for differentially abundant metabolites are listed in Supplementary Material 5. (C) sPLS-DA plot with the GBM pre-treatment (n = 13) and control (n = 13) CSF samples shows that GBM and control samples cluster in separate groups. (D) Supervised heat map of metabolite levels (n = 125) in CSF samples from patients with GBM (n = 31) and controls (n = 13). GBM patients either had CSF collected pre-treatment (n = 13) or post-treatment (n = 18). (E) Volcano plot comparing CSF metabolites in CSF from GBM vs. controls. Colored points represent metabolites that are present at significantly different levels and fold change (-log10(p-value) > 1.3, log2(GBM/Control) > + 1 or < -1) between GBM and control CSF samples. Seven metabolites are present at significantly higher levels in the CSF of GBM patients

Fig. 3

CSF from GBM patients exhibits significantly higher levels of (A) carnitine, (B) 2-methylbutyrylcarnitine, (C) shikimate, (D) aminobutanal, € uridine, (F) N-acetylputrescine, and (G) farnesyl diphosphate than control CSF. Each colored dot represents a patient, small black dot indicates samples that are 1.5 times the interquartile range above the upper quartile or below the lower quartile

Fig. 4

CSF metabolite levels differ with TP53 mutation status. (A) Supervised heat map of metabolites (n = 125) with CSF samples grouped based on TP53 status (mutant vs. wildtype). (B) Volcano plot of metabolites comparing GBM-TP53-mutant vs. GBM-TP53-wildtype. Colored points represent metabolites that are present at significantly different levels (-log10(p-value) > 1.3, log2(Mut/WT) > + 1.5 or < -1.5) between GBM-TP53-wildtype and GBM-TP53-mutant samples. Five carnitine compounds, choline, and γ-aminobutyric acid (GABA) are highly abundant in CSF from TP53-wildtype patients. (C) sPLS-DA plot of CSF metabolites status shows clear separation between TP53-mutant vs. TP53-wildtype samples

Fig. 5

CSF metabolite levels differ with PTEN mutation status. (A) Supervised heat map based on PTEN status (mutant vs. wildtype). (B) Volcano plot of metabolites comparing CSF samples from patients with GBM-PTEN-mutant vs. GBM-PTEN-wildtype. (C) sPLS-DA plot shows clear separation between GBM-PTEN-mutant and GBM-PTEN-wildtype CSF samples

Fig. 6

Pairwise comparisons of carnitine compound levels in CSF of control, GBM-TP53-mutant, and GBM-TP53-wildtype patients. (A) carnitine, (B) propionylcarnitine, (C) 2-methylbutyrylcarnitine, (D) isobutyryl-L-carnitine, and (E) deoxycarnitine. The abundance of carnitine compounds was not significantly different between control and GBM-TP53-mutant groups except for 2-methylbutyrylcarnitine (C). Metabolite abundance was significantly different between control and GBM-TP53-wildtype samples for all 5 carnitine compounds

Fig. 7

Levels of metabolites measured by MRS (Lactate, γ-aminobutyric acid (GABA), and choline) are influenced by tumor mutations. CSF samples from GBM TP53-wildtype patients show significantly elevated levels of (A) lactate, (B) GABA, and (C) choline. CSF samples from GBM PTEN-mutant patients show elevated levels of (D) GABA, (E) lactate, and (F) choline. Each colored dot represents a patient, small black dot indicates samples that are 1.5 times the interquartile range above the upper quartile or below the lower quartile

Fig. 8

Kaplan-Meier plots showing survival probability for 9 GBM, IDH-wildtype patients divided into “high” (in red) and “low” (in blue) metabolite level groups. A patient is classified as “high” if the individual metabolite level is above the cutoff value calculated by CutoffFinder and classified as “low” if below the cutoff. CSF levels of 2-methylbutyrylcarnitine, aminobutanal, and acetylcholine are all inversely associated with survival. (A) 2-methylbutyrylcarnitine (p = 0.046); median survival in days (High = 681, Low = N/A). (B) Aminobutanal (p = 0.022); median survival in days (High = 382, Low = N/A). (C) Acetylcholine (p = 0.022); median survival in days (High = 382, Low = N/A)